Transport of Nano Sized Zero Valent Iron Colloids During Injection Into the Subsurface

Heft 215 Cjestmir Volkert de Boer

Transport of Nano Sized Zero Valent Iron Colloids during Injection into the Subsurface

Von der Fakultät Bau- und Umweltingenieurwissenschaften der Universität Stuttgart zur Erlangung der Würde eines Doktor-Ingenieurs (Dr.-Ing.) genehmigte Abhandlung

Vorgelegt von Cjestmir Volkert de Boer aus Langedijke, Königreich der Niederlande

Hauptberichter: Prof. Dr.-Ing. Rainer Helmig Mitberichter: Prof. Dr. Ruud Schotting Prof. Dr. Rajandrea Sethi

Tag der mündlichen Prüfung: 19. Juli 2012

Institut für Wasser- und Umweltsystemmodellierung der Universität Stuttgart 2012

Heft 215 Transport of Nano Sized Zero Valent Iron Colloids during Injection into the Subsurface

von Dr.-Ing. Cjestmir Volkert de Boer

Eigenverlag des Instituts für Wasser- und Umweltsystemmodellierung der Universität Stuttgart D93 Transport of Nano Sized Zero Valent Iron Colloids during Injection into the Subsurface

Bibliografische Information der Deutschen Nationalbibliothek Die Deutsche Nationalbibliothek verzeichnet diese Publikation in der Deutschen Nationalbibliografie; detaillierte bibliografische Daten sind im Internet über http://www.d-nb.de abrufbar

de Boer, Cjestmir Volkert: Transport of Nano Sized Zero Valent Iron Colloids during Injection into the Subsurface von Cjestmir Volkert de Boer. Institut für Wasser- und Umweltsystemmodellierung, Universität Stuttgart. - Stuttgart: Institut für Wasser- und Umweltsystemmodellierung, 2012

(Mitteilungen Institut für Wasser- und Umweltsystemmodellierung, Universität Stuttgart: H. 215) Zugl.: Stuttgart, Univ., Diss., 2012 ISBN 978-3-942036-19-1 NE: Institut für Wasser- und Umweltsystemmodellierung : Mitteilungen

Gegen Vervielfältigung und Übersetzung bestehen keine Einwände, es wird lediglich um Quellenangabe gebeten.

Herausgegeben 2012 vom Eigenverlag des Instituts für Wasser- und Umwelt- systemmodellierung

Druck: Document Center S. Kästl, Ostfildern

“Muad’Dib could indeed see the Future, but you must understand the limits of this power. Think of sight. You have eyes, yet cannot see without light. If you are on the ﬂoor of a valley, you cannot see beyond your valley. Just so, Muad’Dib could not always choose to look across the mysterious terrain. He tells us that a single obscure decision of prophecy, perhaps the choice of one word over another, could change the entire aspect of the future. He tells us, ‘The vision of time is broad, but when you pass through it, time becomes a narrow door.’ And always, he fought the temptation to choose a clear, safe course, warning, ‘That path leads ever down into stagnation.’ -from ‘Arrakis Awakening’ by the Princess Irulan” Frank Herbert - Dune - 1965 Acknowledgments

I sincerely thank professors Rainer Helmig, Ruud Schotting and Rajandrea Sethi for their continuous support, constructive feedback and above all their patience. I enjoyed the nu- merous discussions I had in the beginning with Ruud and Rainer about how to extract a dissertation from the results of my work at VEGAS and later the in-detail discussions about my work with Rajandrea. Beside that, during several visits to Utrecht there was often time to join the great group activities and while staying in Torino Rajandrea even introduced me to the local cuisine and wine tasting traditions. More than once I had serious doubts about finishing this dissertation, but each time I found support and motivation from the talks I had with each one of you, which motivated me to continue and finish this work. I thank all the colleagues from VEGAS for these wonderful years and great working experience. Special thanks go out to my supervisors Jurgen¨ Braun and Norbert Klaas, and also Oliver, Steffen, Ralf, Bojan, Hubert, Henning and Tanja for their great help in working out experimental ideas. I learned a lot from you all and I am confident that this knowledge will continue to help me in my future work. During the two projects after finishing my Masters, I had the chance to supervise several Diplom, Masters and Bachelors students, which helped to keep the fundamental research going while I was working on the more applied side to reach the project goals. The dynamic interaction with Stefan, Weining, Willem-Bart, Rein and Dave has helped me keeping hold of my motivation for being a researcher at the university. It is to a large extent due to them that I decided to choose for a career in science. I would like to thank my parents Joke and Gerrit for their unconditional confidence in me and their support in keeping me on the job of finishing this work. Throughout my research and especially in the final phase of writing, I have loved the mental support from Flora, who at the same time was writing her dissertation in Geneva. Thank you for being such a great friend throughout all these years! I am very grateful to Tobias for his help with the German summary and being such a supportive and good friend. Special picture credits go out to AndréBuchau (2.4, 2.5, 2.8), Bojan Skodic (4.6), David Estrella (4.23), Flora Boekhout (4.1a), Hua Li (2.12-2.15, 2.19, 2.23) and Stefan Steiert (2.24, 3.9) The presented research was performed within three projects at the Research Facility for Subsurface Remediation (VEGAS), University of Stuttgart, Germany. I gratefully

I thank Jurgen¨ Braun for letting me work on these projects and all the effort he has put into getting these grants. Two projects were funded by the State of Baden-Wurttemberg,¨ Germany, the first through BW-PLUS under the project number BWR25001, the second under the project number 111-047588.6 / 973.049977.9, BUT 013. And one by the European Commission through the 7th Research Framework Programme: FP7 ENV 2008.3.1.1.1., AQUAREHAB. Finally, I know I would not have finished my dissertation without the mental support from my friends and family, “Thank You All!” Cjestmir de Boer October 2012, London (ON), Canada

II Contents

Acknowledgments I

List of Figures VI

List of Tables IX

Notation IX

Abstract XVI

Kurzfassung XVII

1 Introduction 1 1.1 Background ...... 1 1.2 Remediation Technologies ...... 2 1.2.1 Ex-Situ ...... 2 1.2.2 In-Situ...... 3 1.2.3 Nanotechnology for Groundwater Remediation ...... 3 1.2.4 Chemical Reactions and the Chemical Composition of nZVI . . . . 6 1.2.5 Field Application of nZVI ...... 8 1.2.6 Detection and Concentration Measurement of nZVI in the Subsurface 10 1.3 Research Questions ...... 11 1.4 Structure of the Dissertation ...... 11

2 Detection and Concentration Measurement of nZVI in the Subsurface 13 2.1 Motivation...... 13 2.2 Susceptibility ...... 14 2.3 Measurement of nZVI Concentration Proﬁles in a Column ...... 15 2.3.1 Experimental Set Up ...... 16 2.3.2 Concentration Proﬁles and Calibration ...... 16 2.3.3 About the Metal Detector ...... 19 2.4 Measuring Iron Break Through Curves in the Container Experiment . . . . 20 2.4.1 Data Analysis and Post Processing Algorithms ...... 22 2.4.2 Sensor Design and Data Acquisition ...... 24

III 2.4.3 Experimental Veriﬁcation ...... 28 2.5 Measuring Iron Break Through Curves in the Field ...... 30 2.5.1 Sensor Design and Data Acquisition ...... 31 2.5.2 Measuring Device ...... 35 2.5.3 Post-Processing Algorithms ...... 35 2.5.4 Sensor Calibration ...... 39 2.5.5 Sensor Test ...... 40 2.6 Chemical Measurement of nZVI in Soil & Suspension ...... 43 2.6.1 Materials & Methods ...... 44

3 Fundamentals of Transport of Nano Sized Zero Valent Iron Colloids 47 3.1 Motivation...... 47 3.2 Colloid Transport & Filtration in Porous Media ...... 47 3.2.1 Colloid Filtration Theory ...... 48 3.2.2 Attachment Eﬃciency ...... 49 3.2.3 Sensitivity of Colloid Filtration Theory Parameters ...... 50 3.3 Characteristics of the nZVI Suspension ...... 54 3.3.1 nZVI Suspension ...... 54 3.3.2 Aging of nZVI During Storage ...... 55 3.3.3 Aging of nZVI after Dilution ...... 57 3.3.4 Size of nZVI in Suspension using Stokes’ Law ...... 57 3.3.5 Size of nZVI in Suspension using a Laser Detector ...... 60 3.4 Conceptual Model for Transport of nZVI ...... 61 3.5 Mathematical Model for Transport of nZVI ...... 64 3.5.1 Groundwater Flow and Transport ...... 64 3.5.2 Transport and Kinetic Removal ...... 65 3.5.3 nZVI Removal in Porous Media ...... 65 3.6 Transport Experiments (1D) ...... 67 3.6.1 Materials and Methods ...... 68 3.6.2 Conditions ...... 72 3.6.3 Results & Discussion ...... 73 3.7 Numerical Model ...... 80 3.7.1 Numerical Simulation with Fitting on Transient Column Data . . . 80 3.8 Summarizing the Findings ...... 83

4 Colloid Transport in a Radial Flow Field 85 4.1 Motivation...... 85 4.2 Flow and Transport in Radial Systems ...... 86 4.2.1 Concept ...... 86 4.2.2 Mathematical Description ...... 86

IV 4.3 Container Experiments with a Radial Flow Field ...... 88 4.3.1 Motivation ...... 88 4.3.2 Material & Methods ...... 89 4.3.3 Conditions ...... 95 4.3.4 Results and Discussion ...... 97 4.4 Discretized Radial Flow Field Reproduced with Columns ...... 106 4.4.1 Motivation ...... 106 4.4.2 Concept ...... 106 4.4.3 Methods ...... 112 4.4.4 Conditions ...... 113 4.4.5 Results & Discussion ...... 115 4.5 Numerical Simulations ...... 119 4.5.1 Motivation ...... 119 4.5.2 Solver ...... 120 4.5.3 FEniCS vs. Matlab ...... 124 4.5.4 Results ...... 125 4.5.5 Discussion ...... 130 4.6 Comparison of Radial Flow Simulations Methods ...... 131

5 Feasibility of Injecting nZVI into the Subsurface 133 5.1 Motivation...... 133 5.2 Prerequisites for Success ...... 133 5.3 Determination of Necessary Amount of nZVI ...... 137 5.4 Application of Feasibility Test Methods ...... 138 5.5 Application of Methods in the Field ...... 141

Conclusion 143

Outlook 145

Bibliography 147

V List of Figures

1.1 NAPL Distributions in the Subsurface ...... 2 1.2 From Meter to Nanometer ...... 4 1.3 Conceptual Injection Field ...... 9 1.4 Dissertation Structure ...... 12

2.1 Photo of the Metal Detector used for Column Experiments ...... 18 2.2 Calibration Set Up and Data from the Metal Detector ...... 19 2.3 Calibration Curve of the Metal Detector ...... 20 2.4 Magnetic Field Strength of a Cylindrical Coil ...... 21 2.5 Equivalent Network of the Container Sensor ...... 23 2.6 Coilwindingmachine...... 25 2.7 Container Sensor Development ...... 26 2.8 Iso Surface of the Magnetic Field Strength of the Primary Coil ...... 27 2.9 Container nZVI Detection Electronics ...... 28 2.10 Inﬂuence of Fluid Chemistry on Dual Coil Sensor ...... 29 2.11 Schematic Overview of an In-Situ Application of nZVI ...... 30 2.12 nZVI Field Sensor ...... 31 2.13 Equivalent Network of the Field Sensor ...... 32 2.14 Cross Section of the Field Sensor ...... 33 2.15 Response of the Field Sensor on nZVI at Increasing Distance ...... 33 2.16 Field Sensor Development ...... 34 2.17 Field nZVI Measurement System Overview ...... 36 2.18 Calibration Set Up for the Field Sensor ...... 40 2.19 Susceptibility of nZVI at Diﬀerent Concentrations ...... 40 2.20 Field Sensor Test Set Up ...... 41 2.21 Field Measuring Electronics ...... 42 2.22 Propagation of nZVI Inside a Small Container ...... 42 2.23 Concentration and Temperature During the Injection ...... 43 2.24 Measuring Set Up for Analytical ZVI Measurement ...... 45

3.1 Transport Mechanisms of Particle Filtration ...... 48 3.2 Contact Eﬃciencies as a Function of Diﬀerent Parameters ...... 52 3.3 F e0 Reduction of Stored nZVI ...... 56

VI 3.4 Oxidation of nZVI Suspensions after Dilution ...... 57 3.5 Sedimentation Curves of nZVI Suspensions ...... 59 3.6 Size of Nano-Fraction of nZVI in Suspension ...... 61 3.7 Conceptual Model for nZVI Transport ...... 63 3.8 Set Up of the Column Experiment ...... 67 3.9 Column Tracer Experiment ...... 70 3.10 Mixer and Disperser ...... 71 3.11 Schematic Overview of the 1-D Experiments ...... 72 3.12 Photos of the Propagation of nZVI in a Column Experiment ...... 74 3.13 Transient Concentration Profiles of Column Experiments ...... 75 3.14 Pressure Gradient Development ...... 76 3.15 Water Flushing After Injection ...... 77 3.16 nZVI Breakthrough at 50 cm from Column Inlet ...... 78 3.17 Normalized Transport Distance ...... 79 3.18 Numerically Fitted Model on Column Data ...... 81 0 3.19 Fitting Parameter Smax in Relation to C0 ...... 83 4.1 Hyperbolically Decreasing Seepage Velocity Around a Well ...... 86 4.2 Photo of the Container Experiment ...... 89 4.3 Schematic Overview of the Container Experiment Set Up ...... 90 4.4 Filter Screens of the Container Experiment ...... 91 4.5 Two Different Types of Pumps Used ...... 92 4.6 3-D Graphical Overview of the Container Experiment ...... 93 4.7 View into the Container During Filling ...... 94 4.8 Samples Taken for Chemical Analysis ...... 95 4.9 Reynolds Number in the Container Experiments ...... 98 4.10 nZVI Propagation in the Container ...... 99 4.11 Magnetic Susceptibility Measured During Container Exp. 1 ...... 100 4.12 Photos of Container Experiment Results at 1000 l/h ...... 102 4.13 Chemically Analyzed nZVI from Container Exp. 1 & 4 ...... 102 4.14 Photos of Container Experiment Results at 500 & 1000 l/h ...... 104 4.15 Chemically Analyzed nZVI from Container Exp. 3 & 4 ...... 104 4.16 Photos of Container Experiment Results with Different Pumps ...... 105 4.17 Chemically Analyzed nZVI from Container Exp. 2 & 3 ...... 105 4.18 Graphical Overview of all Notation for Radial Flow Discretization . . . . . 108 4.19 Discretization of a Radial Flow Field into Three Sections ...... 112 4.20 Results of Column Set 1 for Container Exp. 1 ...... 116 4.21 Results of Column Set 3 for Container Exp. 3 ...... 116 4.22 Manual Correction of Recorded Data ...... 119 4.23 Example Mesh for Radial Flow Simulation ...... 123

VII 4.24 Comparison of the Matlab and FEniCS models ...... 124 4.25 Hydraulic Head Distribution for Container Exp. 1 & 4 Simulation . . . . . 125 4.26 nZVI Distribution from Simulation of Container Exp. 1 ...... 126 4.27 Simulation and Container Exp. 1 Data ...... 126 4.28 Hydraulic Head Distribution for Container Exp. 3 Simulation ...... 127 4.29 nZVI Distribution from Simulation of Container Exp. 3 ...... 128 4.30 Simulation and Container Exp. 3 Data ...... 128 4.31 nZVI Distribution from Simulation of Container Exp. 4 ...... 129 4.32 Simulation and Container Exp. 4 Data ...... 129

5.1 Injection Wells and Sensor Locations ...... 136 5.2 Diﬀerently Possible Distributions of nZVI ...... 138 5.3 Decision Chart for Using nZVI ...... 140 5.4 Drawing of Funnel Shaped nZVI Reservoir ...... 141

VIII List of Tables

1.1 Pollutants that can be Remediated by nZVI ...... 6

2.1 Susceptibilities Measured and Calculated ...... 39

3.1 Filtration Sensitivity Analysis Constants ...... 50 3.2 Filtration Sensitivity Analysis Variables ...... 51 3.3 nZVI Aging of Diﬀerent Lots ...... 55 3.4 Conditions Used for the Column Experiments ...... 72 3.5 Normalized Transport Distances of the Column Experiments ...... 79 3.6 Parameters of the Column Experiment Simulations ...... 82

4.1 Conditions of the Container Experiments ...... 92 4.2 Parameters of the Container Experiments ...... 96 4.3 Parameters for Column Set 1 ...... 114 4.4 Parameters for Column Set 3 ...... 114 4.5 Measured Input and Eﬄuent Concentrations ...... 115 4.6 Mass Balance Calculations ...... 118 4.7 Parameters Used for Model Comparison ...... 124

IX Notation

Abbreviation Deﬁnition

1-D one dimensional 2-D two dimensional 3-D three dimensional A/D analog digital AC alternating current ADC analog to digital converter CHC chlorinated hydrocarbons DNAPL dense non aqueous phase liquid GSM global system for mobile communications ID inner diameter L, W, H length, width, height L, T , M dimensions: Length, Time, Mass LNAPL light non aqueous phase liquid MTBE Methyl-tert-butyl-ether PCE Tetrachloroethylene PDE differential equation P e Peclet number PEN project on emerging nanotechnologies PP Polypropylene PRB permeable reactive barrier PTFE Polytetrafluoroethylene (Teflon) PV pore volume Re Reynolds number RMS root mean square RMSE root mean square error RNIP reactive nano iron particle TCE Trichloroethylene TTL transistor-transistor logic UV ultra violet

X Abbreviation Deﬁnition

ZVI zero valent iron modem modulator-demodulator nZVI nano sized zero valent iron µZVI micro sized zero valent iron nr. number

Symbol Deﬁnition Dimension or SI Unit

α empirical attachment efficiency [ ] − αagg empirical attachment efficiency of aggregates [ ] − αk fitting coefficient for the geometry factor [ ] − αl dispersion length [L] χD average susceptibility measured [ ] − χ observed susceptibility [ ] − χm susceptibility in accessible domain [ ] − χnZV I susceptibility of nZVI [ ] − η effective single-collector contact efficiency [ ] − ηD single-collector contact efficiency for Brown- [ ] − ian diffusion

ηD single-collector contact eﬃciency for gravita- [ ] − tional eﬀects

ηI single-collector contact eﬃciency for inter- [ ] − ception ηagg summed single-collector contact eﬃciency of [ ] 0 − aggregates

η0 summed single-collector contact efficiency of [ ] − colloids µ absolute fluid viscosity [ML−1T −1] −1 µ0 permeability of vacuum [ ](T mA ) − Ω domain [ ] − Φ magnetic flux [ ](T m2) − −3 ρsand sand density [ML ] −3 ρbulk = ρb sand bulk density (porosity included) [ML ]

XI Symbol Deﬁnition Dimension or SI Unit

−3 ρf ﬂuid density [ML ] θ cylindrical coordinate (angle) [ ](Rad) − εr electromotive force [ ](V Volt) − ξ velocity ratio between two discretization sec- [ ] − tions

Hamaker constant [ ](J = kg m2 s−2) A −2 · · Acolumn cross-sectional area of the column [L ] 2 Ai cross-sectional area of the boundary at ri in [L ] a cylinder

AM sum of the measured data points [ ] − B magnetic flux density [ ](T Tesla) − b base height of confined aquifer or container [ ] − bcolumn number of data points [ ] − C generic constant [ ] − C concentration [ML−3] −3 Cmax maximum concentration [ML ] −3 CnZV I concentration of nZVI [ML ] Ct concentration of nZVI in suspension at time [ ] − step t −3 Ceffluent output (effluent) concentration [ML ] CF e0 weight percentage of zero-valent iron in total [w%] iron −3 Cinput = C0 input concentration [ML ] (in,out) −3 Ci input, effluent concentration of the boundary [ML ] at radial distance ri C1 capacitance of the primary coil [ ](F Farad) − C11 capacitance of the primary reference coil [ ](F Farad) − C21 capacitance of the primary detection coil [ ](F Farad) − Ck capacitance of a cable [ ](F Farad) − cM constant [ ] − c1,2,3,4 integration constants [ ] − D hydrodynamic dispersion (tensor) [L2T −1] 2 −1 Dl longitudinal hydrodynamic dispersion [L T ] Dinf diffusion in an infinite medium [ ] −

XII Symbol Deﬁnition Dimension or SI Unit

dp particle (colloid) diameter [L] agg dp aggregate of colloids diameter [L] dc collector (sand grain) diameter [L] d10 grain size that exceeds the size (diameter) of [L] 10 % of the material by weight d50 grain size that exceeds the size (diameter) of [L] 50 % of the material by weight 0 2+ 3+ F etotal total iron (summed F e , F e & F e )[M] F mass fraction [ ] −2 −1 fn flux for the Neumann boundary condition [L T ] f frequency [ ](Hz) − f (0) frequency at χ = 0 [ ](Hz) − f 0 frequency at two calibration measurements a [ ](Hz) a,b − & b at χ = 0 g gravitational acceleration [LT −2] H magnetic field strength [ ](Am−1 Ampère- − turn per meter) h piezometric head [L]

IC1 current through the primary (generating) [ ](A Amp´ere) − reference coil

IC2 current through the primary (generating) de- [ ](A Amp´ere) − tection coil

I1 electric current towards the primary (gener- [ ](A Ampére) − ating) coil −1 −2 J(i) mass flux density (at boundary ri) [MT L ] k Boltzmann constant [ ](J K−1 = kg − · · m2 s−2 K−1) · · K hydraulic conductivity [LT −1] K˜ average hydraulic conductivity [LT −1] kaging nZVI aging coefficient [ ] col − katt attachment coefficient of colloids [ ] agg − k attachment coefficient of aggregates [ ] att − kd detachment coefficient [ ] − kexp experimental geometry factor [ ] −

XIII Symbol Deﬁnition Dimension or SI Unit

kF e conversion ratio from measured data point [ ] − into concentration of nZVI

ksim simulation geometry factor [ ] − L characteristic ﬁltration length [ ] − L1,2 inductance of coil 1,2 [ ](H Henry) − MF e total injected nZVI [M] M F e molecular mass of iron [M]

Mmi1,2 mutual inductance of the dual (primary and [ ](H Henry) − secondary) coil system, 1 for reference and 2 for detection

Mt value of measured data point [ ] − M0 value of measured background data point [ ] ˙ − −1 M(i) mass ﬂux of injected nZVI (in section i)[MT ] ∆M total injected nZVI in time ∆t [M]

msand total mass of sand [M] mtracer mass of tracer [M] N number of discretization sections [ ] − nH amount of H2 molecules [ ] 2 − n porosity [ ] − P pressure [ML−1T −2] Q discharge, ﬂow rate or injection rate [L3T −1] q Darcy velocity (vector) [LT −1] q Darcy velocity [LT −1] −1 qi Darcy velocity at the radial distance ri [LT ] −1 q˜i average Darcy velocity over section i [LT ] R1 resistance [ ] (Ω Ohm) − r cylindrical coordinate (radial distance from [L] injection well)

ri radial distance of the boundary of section i [L] −1 Smax maximum mass of nZVI per unit mass of dry [MM ] soil 0 −1 Smax maximum mass of colloids per unit mass of [MM ] dry soil −1 Sr mass of aggregates per unit mass of dry soil [MM ] −1 Sc mass of colloids per unit mass of dry soil [MM ]

XIV Symbol Deﬁnition Dimension or SI Unit

Sb surface [ ] − −1 Stotal = S mass of attached and suspended nZVI per [MM ] mass of dry soil −1 Stotal wet mass of nZVI per mass of wet soil [MM ] | T temperature [ ](K Kelvin) − t time [T ] ∆t period of time or time step [T ] (0) U electric voltage over the resistor R1 at the [ ](V Volt) 1 − primary coil, (0) for voltage at χ = 0 U (0) induced voltage at the secondary coil, (0) for [ ](V Volt) 2 − voltage at χ = 0 U 0 voltages at two calibration measurements a [ ](V Volt) 1,2|a,b − & b at χ = 0 U (0) induced voltage at the secondary reference [ ](V Volt) 12 − coil, (0) for voltage at χ = 0 U (0) induced voltage at the secondary detection [ ](V Volt) 22 − coil, (0) for voltage at χ = 0 U (0) induced voltage at the secondary coil (com- [ ](V Volt) 3 − bined), (0) for voltage at χ = 0 V volume [L3] 3 VnZV I volume occupied by nZVI [L ] −1 vp particle settling velocity [LT ] v seepage velocity [LT −1] ˙ 3 −1 Vi volume ﬂux in section i [L T ] w number of windings on a coil [ ] − ω test function [ ] − XF normalized travel distance of mass fraction [ ] − F xF travel distance of a mass fraction F [L] ∆x element size [L] −1 ZC1 impedance of a coil [ ](VA − Volt Amp`ere−1)

XV Abstract

One of the recent In-Situ groundwater remediation techniques under development uses reactive zero valent iron (ZVI) to turn highly toxic chlorinated hydrocarbons (CHCs) into harmless compounds. CHCs are non miscible and characterized by a low solubility which determines their slow dissolution (over decades or centuries) into groundwater, forming plumes that can target drinking water wells, rivers and lakes. Injection of nano sized zero valent iron (nZVI) suspension into the subsurface could target the contaminants directly in the source zone. The high reactivity of nZVI together with the injection into the source fastens the depletion of the contaminant and interrupts the plume generation. The presented work focused on the transport of nZVI during the injection. To make quantitative descriptions of transport possible, an effective detection technique was developed. Exact concentrations of nZVI inside the porous medium was measured through changes in susceptibility detected with electromagnetic induction sensors. Mobility tests with different suspension concentrations were performed in a 1-D horizontally orientated two meter long column. Continuous concentration measurements were performed over the whole length of the column. In a near field scale container experiment a confined aquifer with a radial flow field over a radius of almost two meters was simulated. Differ- ent injection rates and pumping techniques were tested inside this experimental set up. A discretization method to represent all effects of a radial flow field using sets of columns was developed. The method could be verified successfully by comparing the concentration profiles to the results obtained from the container experiments. A mathematical model, developed by starting from the classic colloid filtration theory and by considering the transport of primary colloids and aggregates separately, was able to be fitted on the 1-D results. After implementation in a numerical solver, the model was furthermore capable of providing a very good fit on the results of the radial geometry tests while using exclusively the fitted parameters obtained from the 1-D tests. Throughout the work a better understanding of the transport of nZVI during the injection was developed. It was demonstrated that transport of nZVI without modification was possible over a distance of two meters in both 1-D and radial geometry flow fields. An extrapolation of the work for field application was furthermore described. By applying the methods developed in this work the necessary suspension concentration, the volume of suspension and the injection rate could be determined in advance. The presented work showed promising results and could be a sound scientific basis for further investigations and case studies on nZVI based remediation.

XVI Kurzfassung

Einleitung

Hintergrund Die Wahrnahme der Gefahren, die von verunreinigten Standorten ausgehen, hat in den letzten paar Jahrzehnten stark zugenommen. Viele dieser Standorte wurden kontaminiert mit chlorierten Kohlenwasserstoffen (CKW’s), welche bereits in niedrigsten Konzentratio- nen äußerst toxisch und umweltschädlich sind und von denen einige sogar krebserregend fur¨ den Menschen sind. Flussige¨ CKW’s haben eine höhere Dichte als Wasser, wodurch Sie bis auf große Tiefen im Untergrund versickern können. Ihre Löslichkeit im vorbeiströ- menden Grundwasser ist sehr gering, weshalb die CKW’s den Aquifer im unterstromigen Bereich fur¨ mehrere Jahrhunderte verunreinigen können. In den letzten Jahrzehnten wurden die unterschiedlichsten Methoden entwickelt, um diese Verunreinigungen zu beseitigen. Eine von den neuesten und vielversprechendsten Methoden ist die Sanierung mittels nullwertigem Eisen in nanoskaliger, kolloidaler Form. Diese sogenannten Nano-Eisen-Kolloide haben durch ihre geringe Größe eine sehr große Oberfläche und damit eine sehr hohe Reaktivität. Neben CKW’s kann ein ganzes Spek- trum von Schadstoffen mit ihrer Hilfe abgebaut und somit unschädlich gemacht werden. Bis vor Kurzem wurde hauptsächlich die Reaktivität von Nano-Eisen erforscht. Durch die geringe Partikelgröße wurde angenommen, dass sich die Kolloide einwandfrei in den Untergrund einbringen ließen. Leider musste jedoch bei mehreren Pilotanwendungen [Muller¨ et al., 2006a,b] und Vorversuchen im Labor [de Boer, 2007; Koch, 2007] festgestellt werden, dass sich die Kolloide deutlich schlechter im Untergrund verteilen als erwartet.

Forschungsfragen Um ein besseres Verständnis vom Transport nanoskaliger Eisenkolloide im Untergrund zu bekommen und deren Anwendbarkeit fur¨ die Sanierung von Schadensfällen besser abschätzen zu können, wurden im Rahmen dieser Arbeit die folgenden Forschungsfragen aufgestellt. (a) Ist es möglich, Nano-Eisen direkt im Untergrund zu messen? (b) Welche Faktoren bedingen die niedrige Transportfähigkeit im porösen Medium und wie kann der Transport optimiert werden?

XVII (c) Welche Reichweite kann in einem radialen Str¨omungsfeld erreicht werden? (d) Wie l¨asst sich die Machbarkeit der Injektion an einem Standort feststellen?

Detektion von Nano-Eisen und Konzentrationsmessung im Untergrund

Die Messung von Nano-Eisen im Untergrund ist mit den gängigen Messtechniken nicht ohne weiteres möglich, weshalb im Rahmen dieser Arbeit ein neues Messverfahren speziell fur¨ die Messung von Nano-Eisen entwickelt wurde. Das Messverfahren wurde sukzessive fur¨ drei verschiedene Skalen angepasst. Zunächst wurde die Messung von Konzentra- tionsprofilen entlang einer Säule realisiert. Danach folgte die Aufnahme von Durchbruch- skurven in einem großskaligen Containerversuch mittels stationärer, durchströmter Son- den. Schlussendlich wurde die Technik dann bis zur Feldtauglichkeit weiterentwickelt.

Messprinzip Das neue Messverfahren basiert auf den ferromagnetischen Eigenschaften von nullwertigem Eisen. Durch den Ferromagnetismus haben die Kolloide eine messbare Suszeptibil- ität. Die Suszeptibilität ist das Verhältnis aus der Magnetisierung in einer Substanz und der daraus resultierenden magnetischen Kraft. Die Auswahl eines Messverfahrens, das die Anderung¨ der magnetischen Suszeptibilität im Boden erfassen kann, liegt also nahe. Die Anderung¨ der Suszeptibilität im Boden ist direkt proportional zur Suszeptibilität von Nano-Eisen und damit dessen Konzentration am Ort der Messung.

Messtechnik fur¨ Säulenversuchen Zur Bestimmung der Transporteigenschaften von Nano-Eisen wurde ein Säulenversuchs- stand aufgebaut. Um Konzentrationsprofile entlang der Säule messen zu können, wurde zunächst eine modifizierte Version eines kommerzielles verfugbaren¨ Minensuchdetek- tor (FX2FD, Minex, Reutlingen) genutzt. Der Detektor wurde an einen Rechner angeschlossen, sodass die Rohdaten ausgelesen werden konnten. Zur Aufnahme von Konzentrationsprofilen wurde das Minensuchgerät auf einem fahrbaren Wagen montiert, der mittels eines Schrittmotors entlang der Säule verfahren werden konnte. Vor der Injektion wurde eine Hintergrundmessung durchgefuhrt.¨ Dann wurde während der Injektion alle zehn Minuten ein Profil gemessen und die Hintergrunddaten abgezogen. Um die Rohdaten in reale Konzentrationen umzusetzen, wurde die Gesammtmasse an in- jiziertem Nano-Eisen zum Zeitpunkt der Messung bestimmt und mit der Integralsumme der gemessenen Masse uber¨ alle Messpunkte gleichgesetzt. Somit wurde ein Faktor fur¨

XVIII jeden Messpunkt ermittelt , der zur Umrechung aller Rohdaten eines Versuchs in Konzen- trationsproﬁle herangezogen wurde.

Messtechnik fur¨ die großskaligen Containerversuche Das bei den Säulenversuchen verwendete Messgerät war nicht fur¨ den Einsatz im großskali- gen Containerversuch geeignet. Die Messtechnik musste direkt im Boden eingegraben werden. Die dafur¨ nötigen Messsensoren sowie die dazu gehörige Messelektronik wurden eigens entwickelt. Es wurde eine lange, zylinderförmige Spule gewickelt, um ein elektromagnetisches Feld aufzubauen (Erregerspule). Um diese Spule wurde eine kurze Spule gewickelt, in der durch Induktion eine Spannung erzeugt wurde (Messspule). Die Messsensoren wurde während des Einbaus in den Container mit Bodenmaterial gefullt¨ und so angeordnet, dass die Grundwasser-Strömungslinien genau mittig durch die Spule verliefen. Wenn nun die Nano-Eisen Suspension durch die Spulen strömt, ändert sich die Suszeptibilität im Inneren der Spule, die Spannung der Messspule steigt in der Folge an und der Strom in der Erregerspule sinkt ab. Die Anderung¨ der induzierten Spannung und der durch die Erregerspule fließende Strom wurden mittels einer elektronischen Schaltung erfasst und die Daten auf einem Rechner gespeichert. Zur Umrechnung der Daten in reale Konzentrationen, wurden nach Ablauf des Ver- suches Bodenproben aus den Spulen entnommen und mittels chemischer Analyse die Konzentration an Nano-Eisen bestimmt. Diese Konzentration wurde dann mit dem letzten gemessenen Wert des Sensors gleichgesetzt und die während der gesamten Injek- tionsdauer gemessenen Konzentrationen entsprechend korrigiert.

Messtechnik furs¨ Feld Im nächsten Schritt wurde die Messtechnik fur¨ den Feldeinsatz angepasst. Im Feld sollen die Sensoren in einen Brunnen oder mittels anderer Sondierungsverfahren in der Tiefe angebracht werden. Dabei ist die parallele Ausrichtung der Sensoren zu den Strömungslin- ien nicht immer gewährleistet und die Spulen werden unter Umständen nicht immer gleich durchströmt. Aus diesem Grund wurde ein Spulenpaar entwickelt, bei dem das elektromagnetische Feld nach außen gerichtet ist. Die Sensitivität der Messung wurde aber hierdurch deutlich reduziert, was eine Neuentwicklung der Elektronik erforderlich machte. Die Elektronik wurde außerdem um die Möglichkeit erweitert, die Daten direkt im Messgerät zu speichern und uber¨ ein Master-Slave-Verfahren abzurufen. Fur¨ jede Messlanze mit mehreren uber¨ die Tiefe verteilten Sensoren wurde ein Slave eingesetzt und zentral uber¨ einen Master abgerufen.

XIX Da uber¨ chemische Verfahren keine genaue Konzentration im Untergrund zu bestimmen ist, wurde außerdem eine Methode entwickelt, die Sensoren bereits im Vorfeld zu kalibrieren. Diese Methode wurde vor dem Einsatz im Feld mittels eines kleinen Containerversuchs uberpr¨ uft.¨

Messung von nullwertigem Eisen in Boden- und Flussigproben¨ Um die Bodenproben vom großskaligen Containerversuch aber auch den Anteil an nullwertigem Eisen in den Nano-Eisen-Partikeln bestimmen zu können, wurde ein Messstand aufgebaut wie z.B. in Elion and Elion [1933] beschrieben. Die Messung beruht darauf, dass bei einer Reaktion von nullwertigem Eisen mit Salzsäure ein genau definiertes Volu- men an Wasserstoff freigesetzt wird. Der gebildete Wasserstoff kann somit sehr genau in die Masse an nullwertigem Eisen umgerechnet werden, die zum Zeitpunkt der Messung in der Probe enthalten war.

Grundlagen zum Transport von Nano-Eisen-Kolloiden

Hintergrund Das Hauptziel der Arbeit ist ein besseres Verständis des Transports kolloidaler Nano- Eisen-Partikel während der Injektion in den Untergrund. Hierzu wurden zuerst beste- hende Theorien zur Filtration von Kolloiden im Detail betrachtet. Anschließend wurde die zu verwendende Nano-Eisen Suspension charakteriziert und ein konzeptionelles Modell entwickelt. Auf Basis dieses Modells konnte dann ein mathematisches Modell hergeleitet werden welche dann mittels eines numerischen Modells und eindimensionaler Säulenver- suche uberpr¨ uft¨ wurde.

Kolloidfiltrationstheorie In porösen Medien wird der Transport von Kolloiden gehemmt durch Filtrationseffekte. Die Filtration fuhrt¨ dazu, dass die Kolloide aus der Suspension entfernt werden. Die am häufigste eingesetzte Filtrationstheorie ist die Theorie von Tufenkji and Elimelech [2004]. Der Theorie beschreibt die Filtration auf Basis von drei verschiedenen Filtrationsmech- anismen: Festsetzung am porösen Medium durch Interzeption, Gravitationseffekte und Brownsche Diffusion. Das Zusammenspiel der drei Mechanismen ergibt die Kontaktef- fizienz. Da mit der Theorie nicht alle Effekte vollständig beschrieben werden können, wird ein Kontaktseffizienz Koeffizient eingesetzt, welche experimentell zu bestimmen ist. Uber¨

XX eine Sensitivitätsanalyse wurde festgestellt, dass die Kolloidgröße den größten Einfluss auf die Kontakteffizienz hat.

Charakterisierung der Nano-Eisen Suspension Die ausgewählte Suspension war RNIP 10-E von Toda Kogyo, Japan. Die Suspension wird vom Hersteller als sehr hoch konzentrierter Schlamm geliefert, der vor der Verwen- dung verdunnt¨ werden muss. Uber¨ Vorversuche [de Boer, 2007] wurde festgestellt, dass die Kolloide im Schlamm stark aggregiert sind, jedoch mittels hoher Scherkräfte von einen Dispergiergerät wieder aufgebrochen werden können. Ein weiteres Ergebnis der Vorun- tersuchungen war, dass sich eine Suspensionskonzentration von etwa 10 g/l am besten im porösen Medium transportieren lässt. Es stellte sich heraus, dass der Schlamm während der Lagerung altert und der Anteil an nullwertigem Eisen mit der Zeit abnimmt. Hierzu wurden Langzeitmessungen uber¨ mehr als funf¨ Jahre durchgefuhrt,¨ woraus Alterungsgleichungen hergeleitet werden konnten. Aus der in Sedimentationsversuchen bestimmten Absatzkurve konnte festgestellt werden, dass auch nach der Dispergierung noch Aggregate mit einem mittleren Durchmesser von 4, 4 µm vorlagen. Mit Hilfe von Messungen mit einem Partikelgrößenmessgerät, das auf der Streuung von einem Helium-Laser basiert, konnten jedoch auch die primären Nano-Eisen-Kolloide mit einer Größe zwischen 60nm und 120nm nachgewiesen werden.

Konzeptionelles Transportmodell Auf Basis der Charakterisierung und der Filtrationstheorie konnte ein konzeptionelles Modell aufgesetzt werden. Das Modell lässt sich zusammenfassen in vier Punkte: Die Suspension besteht aus nanoskaligen Kolloiden und mikroskaligen, kolloidalen Ag- ◦ gregaten. Die Kolloide binding an der Kornoberfläche, wobei die verfugbaren¨ Bindungsplätze ◦ begrenzt sind. Aggregate werden durch Gravitationseffekte und auf Basis ihrer Größe aus der Suspen- ◦ sion entfernt. Hierzu gibt es kein Maximum außer dem verfugbaren¨ Porenvolumen, was aber hier vernachlässigt wird. Die Filtrationstheorie von Tufenkji and Elimelech [2004] kann sowohl die Entfernung ◦ der Kolloide als Aggregate beschreiben, gegeben Sie werden getrennt betrachtet.

1-D Transport Modell Fur¨ das konzeptionelle Modell wurde eine mathematische Beziehung hergeleitet. Diese Transportgleichung wurde dann in ein numerisches Modell (in Matlab) integriert, mit Hilfe dessen der Transport berechnet werden konnte. Anschließend wurden S¨aulenversuche durchgefuhrt.¨ Da in der Filtrationstheorie die Suspensionskonzentration nicht

XXI berucksichtigt¨ wird, musste zunächst festgestellt werden, ob die Transportgleichung fur¨ unterschiedliche Eingangskonzentrationen gleichermaßen gilt oder ob dazu zusätzliche Konstitutivbeziehungen nötig sind. Die Säulenversuche wurden sodann mit verschiedenen Eingangskonzentrationen durchgefuhrt¨ und dabei jeweils Konzentrationsprofile zu verschiedenen Zeitpunkten aufgezeichnet. Das numerische Modell wurde dann mit drei Parametern iterativ an die Daten angepasst. Die Parameter waren die Kontaktseffizienz Koeffizient fur¨ die Kolloide und die Aggregate, und die Maximalkonzentration der gebundene Kolloide. Mit Hilfe der Experimente konnte festgestellt werden, dass die Filtrationstheorie fur¨ die Kolloide eine sehr gute Vorhersage lieferte. Einzig der Kontaktseffizienz Koeffizient fur¨ die Aggregate musste recht niedrig eingestellt werden, um eine gute Ubereinstimmung¨ mit den experimentellen Daten zu erreichen. Beide Koeffizienten wiesen keine erkennbare Abhängigkeit von der Kolloidkonzentration auf. Die Maximalkonzentration der Kolloide war jedoch konzentrationsabhängig, wozu eine lineare Beziehung aufgestellt werden konnte.

Transport von Nano-Eisen-Kolloiden im radialen Str¨omungsfeld

Hintergrund Säulenversuche können nur bedingt die Transportfähigkeit von Nano-Eisen darstellen, da Sie nur ein sehr einfaches, eindimensionales Strömungsfeld besitzen. Um das Trans- portverhalten von Nano-Eisen während der Injektion im Untergrund richtig beschreiben zu können, muß das Strömungsfeld mindest radialsymmetrisch gewählt werden. In einem radialsymmetrischen Strömungsfeld weist unter anderem die Fließgeschwindigkeit eine hy- perbolische Abnahme mit zunehmender Entfernung von der Injektionsstelle auf. Dieser Effekt lässt sich mittels standard Säulenversuchen nicht darstellen. Im Rahmen dieser Arbeit wurden drei verschiedene Methoden entwickelt, um den Transport in einem radialsymmetrischen Strömungsfeld nachbilden bzw. vorhersagen zu können.

Großskalige Containerversuche Zur Erstellung eines hyperbolisch abnehmenden Fließgeschwindigkeitsprofils wurde ein Container in Form eines gleichseitigen Dreiecks zur Annäherung an ein Zylindersegment aufgebaut (Länge = 2 m,Höhe = 60 cm, Winkel = 60◦). Ein Dreieck wurde ausgewählt, da dies annähernd die gleichen Strömungseffekte wie ein kompletter Zylinder nachweist aber nur ein Sechstel an Bodenmaterial und Suspension benötigt. Ein Dreiecksschenkel

XXII wurde mit einer Glasscheibe versehen, um die Ausbreitung der Kolloide auch visuell beobachten zu können. In der Spitze wurde ein Injektionsbrunnen eingebracht und entlang des gegenuberliegenden¨ Dreiecksschenkels wurden Drainagerohre installiert, die an einen Auslaufbehälter mit Festpotential angeschlossen wurden. Der Behälter wurde ho- mogen mit Sand befullt¨ und mit einem festen Deckel verschlossen, sodass ein gespannter Aquifer gebildet war. Im Container wurden verschiedene Versuche durchgefuhrt.¨ So konnte gezeigt werden, dass eine kontinuierliche Pumprate bei der Injektion besser ist als eine pulsierende. Auch wurde festgestellt, dass eine Halbierung der Injektionsrate nur einen geringen Effekt auf den Ausbreitungsradius hatte. Bei den Versuchen war es möglich, das Nano-Eisen uber¨ eine Distanz von etwa zwei Metern zu transportieren, wenn etwa drei Porenvolumina uber¨ eine Injektionsdauer von einer Stunde injiziert wurden.

Abbildung eines radialen Strömungsfeldes mittels Säulenversuchen Da die großskaligen Containerversuche sehr arbeitsintensiv waren, viel Vorbereitungszeit brauchten und sehr große Mengen an Suspension benötigten, wurde nach einer einfacheren und schnelleren Methode gesucht, um den Transport in einem radialen Strömungsfeld darzustellen. Hierzu wurde zunächst mathematisch das Strömungsfeld aufgeteilt in kleinere Kreis- segmente. Dann wurde die Fließrate in jedem Segment auf eine Säule gleicher Länge ubertragen.¨ Bei der Injektion der Suspension in eine der jeweils ein Segment represen- tierenden Säulen wurde der Auslauf gesammelt, sodass er als Eingangssuspension fur¨ die nächste Säule genutzt werden konnte. Somit wurde eine Versuchsmethode entwickelt, um das radiale Strömungsfeld diskretisiert darzustellen. Es wurden zwei Versuche mit jeweils drei Segmenten durchgefuhrt.¨ Die Ergebnisse wurden sodann mit denen von zwei großskaligen Containerversuchen verglichen. Hierzu wurde von jeder einzelnen Säule das Konzentrationsprofil aufgezeichnet. In der graphis- chen Darstellung wurden sodann die einzelnen Profile entsprechend der Position der Säule im Gesamtablauf aneinandergehängt. Es konnte somit nachgewiesen werden, dass die Methode in der Lage ist, die Realität akzeptabel bis gut zu repräsentieren. Ein Hauptproblem bei der Durchfuhrung¨ war die genaue Einstellung der Ein- gangskonzentration durch Verdunnung¨ aus dem Nano-Eisen-Schlamm. Da die Trans- porteigenschaften der Suspension stark von deren Einganskonzentration abhängig sind, ist es schwierig, verschiedene Fälle zu vergleichen oder Vorhersagen bezuglich¨ des Trans- portverhaltens in einem bestimmten System zu treffen, wenn auch nur geringfugig¨ verän- derte Eingangskonzentrationen vorliegen.

XXIII Numerische Simulation Als dritte Methode wurde die hergeleitete Transportgleichung in ein numerisches Model mit mehrere Dimensionen umgesetzt. Um die Robustheit vom Modell zu steigern und die Möglichkeiten der Anwendbarkeit zu vergrößern, wurde der Finite-Elemente-Solver FEniCS [Logg et al., 2011] gewählt. Auch hier wurde das Modell getestet gegen die Resul- tate der großskaligen Containerversuche, wobei die Eingangsparameter entsprechend dem großskaligen Containerversuche gewählt wurden. Dazu wurden die Kontaktseffizienz Ko- effizienten fur¨ die Kolloide und die Aggregate und die Gleichung fur¨ die Maximalkonzen- tration der Kolloide eingesetzt, die aus den Säulenversuchen bestimmt wurden. Das Re- sultat wurde dann vergleichend mit den Endresultaten vom großskaligen Containerversuch graphisch dargestellt. Auffällig war, dass die Ergebnisse der numerischen Simulation fast genau durch die Mittelwerte der gemessenen Konzentrationen gingen. Durch den Einsatz der Transportgleichung in einer numerischen Simulation konnte also eine sehr gute Vorhersage des Transports im radialen Strömungsfeld getroffen werden.

Machbarkeitsanalyse zum Einsatz von Nano-Eisen im Feld

Hintergrund Die bisher in dieser Arbeit präsentierten Erkenntnisse sind weitgehend als Grundlagen- forschung anzusehen und stellen die Basis fur¨ ein vertieftes Prozessverständnis zum Transport von injizierten Nano-Eisen-Partikeln im Untergrund. Uber¨ die rein wis- senschaftlichen Aspekte hinaus können die entwickelten Methoden und hergeleiteten Beziehungen auch zur Einschätzung der Machbarkeit einer Injektion im Feld herangezogen werden. Einschränkend lässt sich nochmals darauf hinweisen, dass alle hier beschriebenen Meth- oden auf der Annahme basieren, dass Strömung und Transport im Untergrund nur mittels Permeation im porösen Medium stattfinden und nicht durch hydraulisch erzeugte Klufte.¨ Außerdem wurden fur¨ die Beweisfuhrung¨ nur dispergierte, verdunnte¨ Nano-Eisen Suspen- sionen eingesetzt. Die Ubertragbarkeit¨ auf Suspensionen mit anderen Eigenschaften ist daher vorher im Einzelfall zu uberpr¨ ufen.¨

Voraussetzungen fur¨ einen erfolgreichen Einsatz Bevor Nano-Eisen zur Grundwassersanierung eingesetzt werden kann, muss bekannt sein, ob ein Einsatz mit einer hinreichend hohen Wahrscheinlichkeit zum Erfolg fuhrt.¨ Hierzu sollten bereits in einem fruhen¨ Stadium der Planung einige grundlegende Fragen beantwortet werden, um die grundsätzliche Machbarkeit einer Injektion einschätzen zu können.

XXIV Befindet sich zum Beispiel der Schadensherd hauptsächlich in der ungesättigten Zone, ist der Einsatz von Nano-Eisen nicht sinnvoll, da es sofort durch den hohen Sauerstoffge- halt oxidieren wurde. Ein weiteres Ausschlusskriterium ist die hydraulische Leitfähigkeit des Untergrundes. Ist diese zu niedrig, kann höchstwahrscheinlich kein Transport uber¨ Permeation stattfinden. Falls alle sonstigen Kriterien aber fur¨ den Einsatz von Nano- Eisen sprechen, sollten Laborversuche durchgefuhrt¨ werden, um die Machbarkeit näher zu untersuchen. Zunächst sollte festgestellt werden, wie im Einzellfall ein Erfolg definiert ist und wie dieser nachzuweisen ist. Hauptsächlich sollte bewiesen werden, dass die erforderliche Aus- breitung der Partikel während der Injektion erreicht wurde. Dazu kann beispielsweise die im Rahmen dieser Arbeit entwickelte Feldmesstechnik eingesetzt werden. Die Sensoren werden dazu idealerweise direkt zwischen zwei Injektionspunkten installiert. Des Weit- eren sollte die Reduktion der unterstromigen Schadstofffracht nachgewiesen werden. Zu beachten ist dabei, dass bei der Injektion von Nano-Eisen-Suspension zumeist auch viel sauberes Wasser injiziert wird. Bevor aus der Abnahme der Schadstofffracht Schlussfol- gerungen gezogen werden können, muss gewährleistet werden, dass diese Injektionsflus-¨ sigkeit vor der Beprobung komplett ausgespult¨ wurde. Des Weiteren können vor und nach der Injektion durchgefuhrte¨ Tracerversuche dabei helfen, die in die Frachtberech- nung eingehende Grundwasserfließgeschwindigkeit möglichst genau zu bestimmen

Anwendung von Entwickelte Machbarkeitstests Die in dieser Arbeit beschriebenen Methoden zur Vorhersage der Ausbreitung von Nano- Eisen-Suspension im radialen Strömungsfeld können dazu eingesetzt werden, sowohl die Eingangskonzentration als auch die Injektionsrate fur¨ einen Feldeinsatz zu bemessen. Gegebenenfalls kann zusätzlich noch die Distanz zwischen den Injektionsbrunnen optimiert werden, wobei fur¨ deren Lage jedoch häufig andere Faktoren (z.B. Bebauung) ausschlaggebend sind. Ziel einer Auslegung sollte es sein, eine Kombination von Eingangskonzentration und Injektionsrate zu finden, mit der eine zuvor bestimmte Mindestkonzentration an Nano- Eisen in möglichst großen Bereichen des zu sanierenden Untergrundes erreicht werden kann. Die benötigte Menge an Nano-Eisen kann somit minimiert werden.

Schlußfolgerungen Der Einsatz von Nano-Eisen ist ein vielversprechendes Verfahren zur Sanierung von verunreinigten Aquiferen. Das Ziel eines solchen Einsatzes ist die Entfernung des Schaden- sherdes in-situ, also direkt im Untergrund. Das Transportverhalten von Nano-Eisen w¨ahrend der Injektion wurde in dieser Arbeit im Detail untersucht und es konnte ein deutlicher Fortschritt im Verst¨andnis des Transportverhaltens verzeichnet werden. Die

XXV chemische Machbarkeit einer Sanierung mit Nano-Eisen wurde in dieser Arbeit nicht betrachtet. Abschließend soll nun auf die eingangs gestellten vier Forschungsfragen einge- gangen werden, welche alle vier eindeutig beantwortet werden konnten: (a) Ist es möglich Nano-Eisen direkt im Untergrund zu messen? Es gab bis zum Anfang der Arbeit noch keinen verfugbaren¨ Messtechnik. Deshalb wurde im Rahmen der Arbeit eine neue Messtechnik entwickelt mit Hilfe derer die Nano-Eisen-Konzentration zerstörungsfrei gemessen werden kann. (b) Welche Faktoren bedingen die niedrige Transportfähigkeit im porösen Medium und wie kann der Transport optimiert werden? Die Größe der Kolloide wurde als Hauptfaktor fur¨ den Transport bestimmt. Beim be- trachteten Nano-Eisen-Material (RNIP 10-E, Toda Kogyo, Japan) wurde festgestellt, dass die Kolloide zur Bildung von Aggregaten neigen und der angelieferte Schlamm fur¨ eine direkte Einbringung in den Untergrund zu hoch konzentriert ist. Ein vorrangiges Ziel dieser Arbeit war es, die minimal nötige Vorbehandlung der Suspension fur¨ eine erfolgreiche Injektion zu ermitteln. Die kolloidalen Aggregate können unter Einsatz hoher Scherkräfte mittels eines Dispergiergerätes in kleinere Aggregate aufgebrochen werden. Die Primärkolloide können aber selbst damit nur bedingt wiederhergestellt werden. Zusätzlich muss der angelieferte Schlamm runterverdunnt¨ werden, um den Transport im Untergrund möglich zu machen. (c) Welche Reichweite kann in einem radialen Strömungsfeld erreicht werden? In dieser Arbeit wurde erstmalig ein Transport von fast zwei Metern in einem radialen Strömungsfeld nachgewiesen. (d) Wie lässt sich die Machbarkeit der Injektion an einem Standort feststellen? Es wurde ein Fließschema erstellt, um bei der Einschätzung der Machbarkeit einer Injektion zu helfen. Es wurden zwei alternative Methoden entwickelt, um den Trans- port im radialen Strömungsfeld vorhersagen zu können. Der Einsatz einer dieser Methoden kann in der Planungsphase helfen, die richtigen Entscheidungen zu treffen und die benötigten Ablaufbedingungen festzulegen. Die Ergebnisse dieser Arbeit sind vielversprechend und stellen eine gute Basis fur¨ weitere Forschung bezuglich¨ des Einsatzes von Nano-Eisen bei der Sanierung von kontaminierten Aquiferen dar. Des Weiteren können die hier vorgestellten Ergebnisse bei der Beurteilung von entsprechend gelagerten Fallbeispielen helfen.

XXVI 1 Introduction

1.1 Background

The awareness of the dangers of contaminated sites and the resulting contamination in groundwater has been increasing over the last decades. The number of sites that have been recognized for the need to be treated is large, in the European Union there are over three million sites contaminated [Swartjes, 2011]. In Germany 350 000 sites are potentially contaminated and are a danger for the environment and the groundwater [Hahn, 2006]. A major separation of contaminant types is made based on their density (Figure 1.1). Contaminants that have a higher density than water are called dense nonaqueous phase liquids (DNAPL), if their density is lower than water they are light nonaqueous phase liquids (LNAPL). Among the sites in Germany, 35% are contaminated by chlorinated hydrocarbons (CHCs), these are very toxic and many of them are carcinogenic. CHCs like Tetrachloroethylene (PCE) and Trichloroethylene (TCE) are non ﬂammable solvents for organic materials and were thus used in many industrial areas, mainly to degrease metal parts. Furthermore, PCE was also applied in many dry-cleaning facilities, which were often situated in urban areas. Liquid CHCs like PCE and TCE are DNAPLs. If spilled in large volumes, due to their high density, they can penetrate deep into the subsurface, moving through the unsaturated zone and pass the groundwater table into the saturated zone, leaving a trail of residual NAPL (ganglia) behind, until a less permeable layer like a clay layer or an aquitard (typically the bedrock formation), holds them from moving further downward, where a highly saturated pool forms. DNAPLs like PCE and TCE dissolve (albeit, very slowly) into the groundwater passing through the zone of ganglias and pool (source zone), the formed contaminated groundwater plume in turn often ends up in drinking water wells, rivers and lakes. Due to the high toxicity of these CHCs, even very low concentrations in the groundwater can form a risk for human health and the environment. Furthermore, because CHC phase is so persistent, downstream groundwater can end up to be polluted for several centuries.

1 1. Introduction

Figure 1.1: LNAPL and DNAPL spill distributions in the subsurface, after G¨odeke [2011]

1.2 Remediation Technologies

At present, many different types of remediation techniques exist. They can be classified in: Ex-Situ methods that remove the contaminant from the subsurface to treat it on the surface, both on-site and off-site and In-Situ techniques to remediate a contamination in the subsurface without bringing the contaminant to the surface.

1.2.1 Ex-Situ

For sites with near surface pollution (often these are LNAPL spills), excavation is the most common applied method. Though effective in achieving a full clean up, the main disadvantages of this technique are the ecologically ineffectiveness due to the high energy consumption for excavation, transport and off-site treatment as well as the rather unsuitability for residential areas and saturated aquifers [Fetter, 1999]. In case of a saturated aquifer or for deeper contaminations the pump-and-treat remediation is currently the most applied method for treatment. With pump-and-treat the groundwater with the

2 1.2. Remediation Technologies dissolved contaminant is pumped out of the ground and treated on the surface, for example by using activated-carbon ﬁlters. With this technique it often takes many years or even decades to remediate a site if NAPL phase or sorbed contaminants is present [Fetter, 1999].

1.2.2 In-Situ

Different In-Situ techniques have been developed to remove CHCs from the groundwater and source zone. Some methods involve only a single action after which chemical reactions or biology removes the contaminants or when the natural conditions are suitable to remove the contaminants it might even involve no action beside close observation and monitoring of the contaminants in the downstream groundwater. Contaminant plumes downstream of the source zone can be cut off by placing permeable reactive barriers (PRBs) or they can be treated by enhanced natural attenuation (ENA). ENA is mainly the controlled and enhanced biodegradation by bacteria or plants (e.g. Ferguson and Pietari [2000]; Yang et al. [2009]). The PRB method was proposed by Gillham and O’Hannesin [1992], these barriers are in principle trenches where the aquifer material has been replaced by reactive material, the exactly chosen chemical substance can differ for each pollution origin. Once the polluted groundwater flows through the PRB, the reactive components of the PRB transform the pollutants into harmless or immobile end products. One of the materials often used (e.g. Vogan et al. [1999] or VanStone et al. [2005]) in PRBs is zero valent iron (also denoted as “elementary iron” or “F e0”). Zero valent iron is capable of remediating several types of pollutants, the most common target pollutions of PRBs are chlorinated hydrocarbons. Most of them can be dechlorinated by zero-valent iron.

1.2.3 Nanotechnology for Groundwater Remediation

In an interview in The Progressive [Higgs, 2009], David Rejiski of the Project on Emerging Nanotechnologies (PEN) states that products that include nano-sized particles are boom- ing with 212 diﬀerent products in March 2006 up to 1015 products in August 2009. This organization also notes a 50% increase in the number of organizations that are involved in nanotechnology between 2006 and 2008. Nanotechnology takes place at a scale of one to several billionths of a meter. For comparison, the diameter of a human hair varies between 17 000 and 181 000 nanometer

3 1. Introduction

Human Hair, diameter 17 181 µm Micro Fe-Colloids, diameter− 1 15 µm Carbon 60 Credit: BASF, Germany − Diameter 0.7 nm Soccer Ball ≈ Diameter 22 cm ≈

1 2 3 4 5 6 7 8 9 10 10m 1m 10− m 10− m 10− m 10− m 10− m 10− m 10− m 10− m 10− m 10− m

(1 cm) (1 mm) (1 µm) (100 nm) (1 nm)

nZVI Particles Micro Iron Particles Diameter: 60 120 nm Diameter 30 100 µm Credit: Li, Elliot≈ and− Zhang [2006] Credit: Höganäs,Sweden≈ − Figure 1.2: Overview of different sizes

[Ley, 2010]. A meter scale is given in Figure 1.2 to give an overview of the different size scales. Since science and industry are now able to work on these incredibly small scales, all sorts of intriguing possibilities come within reach. Also for the groundwater remediation field it is nano-technology that appears to provide very promising possibilities. Different types of nanoparticles are being created to treat specific contaminants. Nanotitaniumoxide particles (T iO2) were found capable of immo- bilizing pertechnetate anions in situ [Mattigod et al., 2005]. T iO2 particles ( 20 nm) ∼ doped with gold nanoparticles ( 2 nm) were found to be highly effective to perform ∼ a photocatalytic clean-up of methyl-tert-butyl-ether (MTBE) and 4-chlorophenol contaminated groundwater [Orlov et al., 2007]. A nanosensor was developed based on a semiconductor film of ZnO ( 27 nm) which was found to quench emitting visible light ∼ when chlorinated phenols were present in the water on top of the film. The quenching was quantitative to the concentration and could be used to detect concentration as low as

1 ppm. Furthermore these semiconductors1 were able to completely degrade the organic contaminants when radiated with UV-light [Kamat et al., 2002].

4 1.2. Remediation Technologies

Based on the success of zero valent iron used in PRBs, an In-Situ method was proposed by Cantrell and Kaplan [1997]. Instead of an excavated trench filled with ZVI filings, they proposed the injection of a zero valent iron colloid suspension into the subsurface using wells. This method forms a reactive permeable zone which is no longer bound to limited depths and plume treatment, but can also be used directly in the contaminant source zone. By treating the source directly, the contaminant removal time is significantly reduced. The ZVI removes the contaminant from the water in the direct vicinity of the contaminant phase, thus increasing the dissolution rate. Furthermore, the contaminant plume is cut off and no new plume can develop. In the first tests, Cantrell and Kaplan [1997] used iron of micrometer scale. These particles were rather heavy and gravitational settling occurred during injection, limiting the transport and feasibility of this method. Due to the recent developments in the nano technology branche, it has become possible to create zero- valent iron particles at the nanometer scale. Some of the currently commercially available nano-sized zero-valent iron (nZVI) colloids are between 10 and 100 nm in diameter (e.g. RNIP, Toda Kogyo, Japan or NANOFER 25, NanoIron, Czech Republic). These are provided suspended in water (with additives like surfactants and/or polymers) as a highly concentrated slurry. Common aquifer pore diameters range from 200 nm for a silty aquifer to 100 000 nm in a gravel aquifer. Thus nZVI colloids are small enough to be transported through the finer pore spaces of the porous media where micro sized particles would be filtered out [Elliott and Zhang, 2001]. Considering the small size of nZVI, filtration effects and the chance of clogging the aquifer are reduced and also they will have a lower settling velocity, which should make it possible to transport them farther away from the injection well. Still though, Tratnyek and Johnson [2006] describe that the transport distance in an aquifer is expected to be limited based on the deep-bed filtration theory. Transport of larger colloids is reduced mainly due to gravitational forces and straining effects (trapping in pore throats that are too small to allow passage), whereas smaller colloids will filter out of suspension mainly due to Van der Waals forces and Brownian diffusion [Tufenkji and Elimelech, 2004]. While several field tests have been performed with nZVI, the delivery of the particles to the desired location often was unsuccessful or at least disputable [Schrick et al., 2004] or the transported distance was very small resulting in a very narrow grid of boreholes needed for the injection (e.g. Muller¨ et al. [2006b]).

5 1. Introduction

1.2.4 Chemical Reactions and the Chemical Composition of nZVI

The reactivity of nZVI colloids is mainly controlled by their specific surface, a higher specific surface resulting in a higher reactivity [Bradford and Torkzaban, 2008]. The specific surface of nZVI colloids is approximately 33.5 m2/g [Elliott and Zhang, 2001], in comparison, for micro iron particles 0.1 1 m2/g [Nurmi et al., 2005] and for granular − iron filings approximately 0.004 m2/g [Huang et al., 2003], although the latter can due to surface roughness and angularity reach a specific surface of approximately 0.5 m2/g. The high reactivity makes the nZVI colloids reactive towards many contaminants (Table 1.1) and facilitate to deplete a source zone with high concentrations of the contaminant and stop the generation of a pollution plume. The chemical reaction that takes place between nZVI and contaminants (e.g. CHCs) is well understood and has been described by several authors in detail [Bartzas et al., 2006; Lien and Zhang, 2001; Liu et al., 2005; Steiert, 2008; Zhang, 2003].

Table 1.1: Pollutants that can be remediated by nZVI. After Muller¨ et al. [2006a]; Zhang [2003]

Chlorinated methanes Chlorinated benzenes Pesticides

Carbon tetrachloride (CCl4) Hexachlorobenzene (C6Cl6) DDT (C14H9Cl5)

Chloroform (CHCl3) Pentachlorobenzene (C6HCl5) Lindane (C6H6Cl6)

Dichloromethane (CH2Cl2) Tetrachlorobenzenes (C6H2Cl4) Organic dyes

Chloromethane (CH3Cl) Trichlorobenzenes (C6H3Cl3) Orange II (C16H11N2NaO4S)

Dichlorobenzenes (C6H4Cl2) Chrysoidine (C12H13ClN4)

Chlorobenzene (C6H5Cl) Tropaeolin O (C12H9N2NaO5S) Heavy metal ions Trihalomethanes Chlorinated ethenes 2+ Mercury (Hg ) Bromoform (CHBr3) Tetrachloroethene (C2Cl4) 2+ Nickel (Ni ) Dibromochloromethane (CHBr2Cl) Trichloroethene (C2HCl3) + Silver (Ag ) Dichlorobromomethane (CHBrCl2) cis-Dichloroethene (C2H2Cl2) 2+ Cadmium (Cd ) trans-Dichloroethene (C2H2Cl2) 2+ Cobalt (Co ) 1,1-Dichloroethene (C2H2Cl2) 2+ Tin (Sn ) Vinyl chloride (C2H3Cl) Lead (P b2+) Copper (Cu2+) Polychlorinated hydrocarbons Other organic contaminants Inorganic anions 2− PCB’s N-nitrosodimethylamine (NDMA) Dichromate (Cr2O7 ) 3− Dioxins (C4H10N2O) Arsenic (AsO4 ) − Pentachlorophenol (C6HCl5O) TNT (C7H5N3O6) Perchlorate (ClO4 ) − Nitrate (NO3 )

6 1.2. Remediation Technologies

Though, the colloids are not only more reactive towards the contaminants, they are also strongly aﬀected by undesired side reactions. The main side reaction is the tendency to quickly corrode. Dry nZVI colloids would even self ignite due to the reaction with oxygen in the air. Aerobic corrosion also occurs in suspension with the dissolved oxygen (Equation 1.1). Hence, the colloids are unsuitable for the remediation of pollutants in the unsaturated zone, whereas fully saturated aquifers are in general completely anaerobe and thus provide a suitable environment.

0 + 2+ 2F e + 4H + O2(aq) 2F e + 2H2O(l) (1.1) (s) (aq) → (aq) An other side reaction is anaerobic corrosion (Equation 1.2), this reaction though is much slower and strongly pH dependent [Bartzas et al., 2006]. During storage the colloids are in an anaerobic environment, in which, even tough no free oxygen is present, they can thus still corrode. The anaerobe corrosion is in a closed environment self inhibiting because the corrosion sets hydrogen and hydroxide free which increases the pH and consequently reduces the corrosion rate.

0 2+ − F e + 2H2O(aq) F e + H2(g) + 2OH (1.2) (s) → (aq) (aq) The amount of zero valent iron inside a colloid therefore reduces with residence time in water. Anaerobe corrosion sets hydrogen gas free (Equation 1.2) which could clog the porous media. To minimize the corrosion problem, nZVI colloids are often covered with a thin shell that shields the zero valent iron from direct contact with water. The nZVI colloids as presented by Wang and Zhang [1997] were produced with a palladium acetate

([P d](C2H3O2)2]3) that reacts with a small outer part of the zero-valent iron colloid, creating a small shield against corrosion and will function as a catalyst. Towards several contaminants additional metals like the palladium acetate can function as a catalyst as well, other metals that can be used with the same purposes are for example platinum (F e/P t), silver (F e/Ag), nickel (F e/Ni), cobalt (F e/Co) and copper (F e/Cu). The particles of Toda Kogyo (Japan) and Nano Iron (Czech republic) are shielded with crys- talline magnetite (F e3O4)(F e/F e) [Wang and Zhang, 1997]. So far only the last type has shown to be producible in larger amounts for a fairly acceptable price. The chemical feasibility of the remediation technique has been worked out in detail with Steiert [2008]. Within his research it was furthermore discovered that the production of hydrogen gas and with that the clogging of the porous media can be signiﬁcantly reduced

7 1. Introduction

by adding burned chalk (Ca(OH)2) in granular form to the suspension before injection. This technique has been filed for patenting [Klaas et al., 2010b]. Though very small, nZVI colloids are still ferro magnetic and contain a positive and negative pole (i.e. they are bipolar). The colloids thus tend to get attracted to each other and form aggregates. The aggregation and gelation (the building of a network of aggregates) can easily result in pore plugging and gravity settling, strongly reducing transportability [Saleh et al., 2006]. To avoid aggregation, a nZVI suspension can be stabilized through the following repelling mechanisms: Electrostatic repulsion, which is the mutual repulsion of like electrical charges, influenced by the pH value of suspension [Hong et al., 2009]; Steric repulsion, adsorbed long polymers (e.g. guar gum) on the particle surface, which prevents particles to get close [Tiraferri et al., 2008]; Electrosteric repulsion, which is the combination of both using long steric repelling polymers [Phenrat et al., 2008]. Unfortunately, it was also found that the addition of electrosteric polymers significantly reduced the reactivity because the obstructed contact between the contaminant and the colloid [Phenrat et al., 2009, 2007].

1.2.5 Field Application of nZVI

The injection of a nZVI suspension in the field could be performed using a simple screened well, with packers to inject over a small height, by using a short filter screen at the tip of a direct push rod, or hydro fracturing (either by using a concrete lined well, or direct push). The injections using filter screens result generally in a near radially symmetrical flow field around the injection well. Spherical flow fields will be less profound since most aquifers show a horizontal layered structure with a large anisotropy. A radial flow field results in a hyperbolically reducing seepage velocity with increasing distance from the well screen. When hydro fracturing is applied, the main transport takes place in large fractures and there is negligible porous media flow (permeation) thus this technique falls beyond the scope of this research. In a general sense, there are two different possible application concepts for injectable nZVI, source zone- and plume treatment. F To prevent the migration of a plume, or to secure an area (e.g. residential- or water- protection area) from an approaching plume a barrier can be formed with injectable nZVI (similar to a PRB). The concentrations in a downstream plume are much lower than the concentrations in a source zone. The removal of the contaminant in the plume will not

8 1.2. Remediation Technologies

Figure 1.3: Conceptual concentration distribution of an injection field affect the dissolution rate of the contaminant in the source zone. This application type is thus mainly useful in case of an unknown, unreachable or already removed source. To make sure that the whole contaminant is in contact with the injected barrier, it should be guaranteed that the nZVI is spread between injection wells without voids where the groundwater could pass through. This can be achieved by placing the wells in triangular positions. The concentration of nZVI is expected to decrease with increasing distance from the injection well. Hence, a possible concentration profile could look like the one presented in Figure 1.3. The injected barrier has to be reactive for a long period and a full consumption of the reactive iron is likely to occur before the source zone stops emitting. Injectable nZVI can then make it possible to re-inject fresh reactive iron at the moment that the previous dose has been consumed. F To apply source zone treatment, the location of the source inside the subsurface needs to be well known. The nZVI has then to be injected, such that it gets very close to the contaminant phase. Once in place, the dissolved contaminant concentration in the direct vicinity of the contaminant phase is being reduced, this in response will increase the dissolution rate of the contaminant phase. The downstream plume will be cut off but will not be removed. By natural attenuation or by applying a pump and treat method the plume could be removed if necessary.

9 1. Introduction

1.2.6 Detection and Concentration Measurement of nZVI in the Subsurface

To do a source zone treatment, as for most methods targeting the source zone, it is essential that the exact location of the source is well known. This is far from trivial and puts a high demand on the measuring techniques used for the field characterization. Once the site has been characterized and the contaminant source localized, the nZVI has to be transported to these locations. Hence, an accurate detection technique is necessary to determine the location of nZVI during and after the injection. Since the reaction in this active zone is expected to be fast, nZVI will be consumed rapidly as the reaction proceeds. An accurate monitoring technique would thus be preferable to determine when the nZVI has been consumed and a renewal is needed. So far several measuring techniques have been applied during pilot tests to proof the effectiveness of injected nZVI. Muller¨ et al. [2006b] used geophysical measurements, like ground penetrating radar, geomagnetic and geoelectric mapping. Glazier et al. [2003] and Zhang [2003] used in their field test measurements of flow rate, water level, oxidation reduction potential (ORP), dissolved oxygen (DO), pH, specific conductance, temperature and the contaminant concentration reduction to show the radius of influence of the injection and the location of nZVI during and after the injection. Elliott and Zhang [2001] monitored the total iron and dissolved iron concentrations in combination with the pH and ORP values at monitoring wells throughout the injection to determine the migration of nZVI. Quinn et al. [2005] used soil samples from cores taken at different times after the injection of emulsified zero valent iron and compared the TCE content in the samples to samples from cores taken prior to the injection. They also measured the TCE concentrations in ground water samples at different times after the injection and compared them with samples taken prior to the injection. The aforementioned measuring techniques all share some common problems. The methods based on ground water sampling are not conclusive since during the injection a large amount of clean water (used as carrier fluid for nZVI) was injected, pushing away the dissolved contaminants. It takes time before the contaminants again dissolve into the fresh water up to the old concentration and also the groundwater flow velocity through the contaminated zone might be reduced due to the presence of the nZVI and possibly hydrogen gas. Using cores from different locations to compare samples after the injection with samples taken prior to the injection will likely be inconclusive even if the cores are taken close to each other since the subsurface is heterogeneous. Therefore, geogenic iron

10 1.3. Research Questions background values are likely to diﬀer more than the additional iron added due to injection of nZVI from one location to another. To quantitatively and conclusively measure the nZVI presence in the subsurface during and after the injection, a new technique able to monitor directly iron concentrations without background interference is desirable.

1.3 Research Questions

In the previous sections it has been emphasized that the treatment of contaminations in the groundwater with injectable nZVI is promising. Nevertheless, there are still questions to be answered and technical problems to be overcome before the method can be applied. The goal of this doctoral research is to get a better description and understanding of the transport behavior of nZVI during the injection into the subsurface which led to the following research questions: (a) Is it possible to detect In-Situ the concentration of iron and determine the transport distance? (b) Which conditions have the largest effect on mobility and how can they be optimized? (c) What is the transport distance in a radial flow field? (d) Which method can test the feasibility of injection into the subsurface?

1.4 Structure of the Dissertation

To obtain a better description and understanding of the transport behavior of nZVI in porous media the research started at the fundamental of colloid filtration and worked this out through 1-D and 2-D experiments and numerical simulations to recommendations for the field, covering a wide range of scales. For this need, it was found that no non-destructive and direct measuring technique was available and thus first had to be developed to describe and understand the transport behavior (Figure 1.4). This work distinguishes four main chapters after this introduction. Chapter 2 presents the developed quantitative non-destructive measuring and detection techniques based on the ferro magnetic property of nZVI to measure the nZVI concentration in the porous medium at the column, container and field scale. A chemical (destructive) analysis method was furthermore developed to provide a chemical verification method of the non- destructive measurements as well as to characterize nZVI suspensions. Chapter ??

11 1. Introduction

Figure 1.4: Dissertation research structure provides a description of colloid transport and constitutive relationships derived from the colloid filtration theory, column experiments and numerical simulations of the column experiments. Chapter 4 focuses on the transport in a radial flow field. A large scale container experiment is presented. The results of this experiment were used to verify a quick screening column experiment and a numerical predictive model. Chapter 5 shows how the experiments and the numerical simulations provide a method to test the injection feasibility and can be used to design a field application. Recommendations and the preferred approach for the optimized design are presented. At the end of the work the answers on the research questions are summarized in the conclusion and an outlook provides further research topics that could be performed.

12 2 Detection and Concentration Measurement of nZVI in the Subsurface

2.1 Motivation

Research on the transport of nZVI is only possible with accurate methods to determine the concentration inside porous media. So far most research was based on break through curves of columns (e.g. Baumann and Werth [2005]; Bradford and Bettahar [2006]), some- times supported by a destructive chemical analysis of the column contents after the experiments were finished (e.g. Tufenkji and Elimelech [2005b]). Break through curves can be useful and through a combination with numerical simulations [Tosco et al., 2009], many transport parameters can be determined by fitting. The exact distribution inside the porous medium though is far from sure to be realistic, many different scenarios can lead to the same break through curve. Visual observation can not provide quantitative concentrations, since the iron already colors the sand black at very low concentrations, only the front of the iron can be determined and further no separation between low and high concentrations is possible. In a chemical determination the natural presence (geogenic) of iron in the sand will create a large unknown factor in the determination. Because even the geogenic iron content of sand packed in a column or container is not homogeneously spread, it is not possible to make a fully reliable difference measurement based on chemical iron detection. Furthermore, a chemical analysis can first be performed after the experiment is finished because sand samples have to be taken. To facilitate a quantitative measurement of the nano sized zero valent iron (nZVI) concentration during and after the injection into a column, a container or an aquifer, a new measuring technique has been developed. The method has been produced in different stages for different uses. The first stage was the adaptation of existing hardware to measure concentrations inside a column. The second phase was the development of a

13 2. Detection and Concentration Measurement of nZVI in the Subsurface sensor and the accompanying electronics to measure break through curves within a large container experiment. The final stage was the development of sensors and electronics to measure concentrations in an aquifer for a field application. All three different types of sensors and accompanying electronics have been tested in the lab. The following sections separately describe the measuring techniques for all three uses.

Collaborations

The container and field sensor development was performed in collaboration with André Buchau and Hua Li of the Institute of Theoretic Electrical Engineering (ITE), Univer- sity of Stuttgart. They performed numerical simulations of the electromagnetic fields of different types of sensors to optimize the sensors for the container and the field and developed post processing algorithms. The coordination, design of the sensor production techniques, production and experimental testing of all the sensors and implementation of the post processing algorithms into data analysis software were performed by the author. The electronics for the container and field sensors was developed in collaboration with Hubert Hermes from Hermes Messtechnik, Stuttgart, Germany. The building and testing of all the electronics based on the electrical layouts were performed by the author.

2.2 Susceptibility

Susceptibility is the ratio of the magnetization in a substance to the corresponding mag- netizing force, iron has a significant susceptibility [Getzlaff, 2008]. Simple experiments on nZVI showed that this material property is still observable [de Boer, 2007]. Hence, a measurement system that detects changes in susceptibility of the monitored subsurface is possible. The system should be designed in a way that the magnetic field strength is small enough to neglect the non-linear behavior of susceptibility of iron. Then, the magnetic flux density B is simply connected to the magnetic field strength H by

B = µ0(1 + χnZV I )H (2.1) where χnZV I is the susceptibility of nZVI and µ0 is the permeability of vacuum [Getzlaﬀ,

2008]. The volume of nZVI (VnZV I ), which is injected into an aquifer, is very small. Furthermore, it is assumed that the nZVI near the sensor is approximately homogeneously distributed in the aquifer. Then, the observed susceptibility χ is expected to be in the

14 2.3. Measurement of nZVI Concentration Proﬁles in a Column range of:

χ VnZV I χnZV I (2.2) ' The measured susceptibility in Equation 2.2 depends both on the concentration of nZVI and on its susceptibility [Bregar and Pavlin, 2004]. If the susceptibility of nZVI is constant, the resulting susceptibility depends only on the concentration of nZVI. Non-linear behaviour of nZVI is excluded by small magnetic ﬁelds to ensure the recommended constant susceptibility of nZVI. By using an alternating magnetic ﬁeld (e.g. between 1 kHz and 20 kHz), there is no chance of polarizing the nZVI nor is there any magnetic attraction or repelling. Starting from the statement in Equation 2.2, the concentration of nZVI in the porous medium (CnZV I ) can be related to the total susceptibility

CnZV I = s (χ) (2.3)

The function s (χ) can be obtained from posteriori chemical analysis (e.g. using the method described in Elion and Elion [1933] or Liu and Lowry [2006]), or a priori calibration of the sensor. The calibration of the column experiments was performed by calculating the total injected mass of iron present inside the column at the moment the susceptibility was measured along the column. The posteriori chemical analysis method has been used for the container experiments and is described in more detail in Section 2.6, for the ﬁeld measurement a method to calibrate the sensors before installation has been developed and is presented in Section 2.5.4. Since the total susceptibility χ is very small, a high sensitivity and precision of the electronics was necessary for a reliable measurement. In the following three sections the measurement of the susceptibility at three diﬀerent scales is presented.

2.3 Measurement of nZVI Concentration Proﬁles in a Column

A column experiment was set up to investigate the transportability of nZVI. To measure the particle concentration and determine transport characteristics in porous media, a commercially available metal detector was adopted. Institute Dr. Foerster (Reutlingen, Germany) provided one of their research and development devices (Minex FX2FD) which could be connected to a computer to store all the raw data internally produced by the

15 2. Detection and Concentration Measurement of nZVI in the Subsurface detector.

2.3.1 Experimental Set Up

To prevent interference of metal on the detection, the framework to hold the column was made completely from wood and contained no metal parts. Due to the framework, the column was situated 1.5 m above the steel reinforced concrete floor and the metal track which was used to move a wooden metal detector carrier at a constant velocity along the column. Within one of the sensor openings of the metal detector the column was placed (Figure 2.1). In basics, the detector consists of an oscillator producing an alternating current at 19.2 kHz that passes through a coil producing an alternating electromagnetic field. Another coil is used to measure the electromagnetic field through an alternating voltage induced in that coil. This voltage changes when the susceptibility within the electromagnetic field changes. The coils of the Minex FX2FD are 8-shaped and placed on top of each other, creating roughly one electromagnetic field within each of the loops which both point in opposing directions [Auslander et al., 1991]. This is useful when searching small metal parts in the soil to determine if they are located closer to the one or other opening of the sensor. Though, for the column experiment this functionality is not necessary and is not used by placing the column inside only one of the loops. The digital output of the detector is linearly proportional to the magnetic susceptibility (detailed testing provided in the following section). The susceptibility in turn is directly proportional to the iron concentration [Bregar and Pavlin, 2004; Buchau et al., 2010].

2.3.2 Concentration Proﬁles and Calibration

During the injection the concentration profile inside the column was measured approximately every ten minutes. A high accuracy of the nZVI content change inside the column due to the injection was reached because geogenic iron occurrences in the sand and background metal presences were removed by subtracting a background measurement. The carrier moved with 10 mm/s and the metal detector registered 25 readings per second, resulting in a resolution of 0.4 mm. To convert the measured signal profile into a real concentration profile, one measured profile of each experiment was used with a clear distribution of iron inside the column. The total injected nZVI MF e at the moment of the recording was calculated based on the injection rate Q, duration t and the suspension input concentration Cinput (i.e. MF e = QCinputt). The concentration Cinput was determined using the phenanthroline colorimetric method and a photometer (Lambda 12,

16 2.3. Measurement of nZVI Concentration Proﬁles in a Column

Perkin Elmer, Germany), this method provides a measurement of the total iron (F etotal) in suspension. All the iron inside the suspension was dissolved into solution by using nitric acid (H2SO4). Preferably no iron break through at the outlet was reached. When such reading was not available, the diﬀerence between the injected concentration and the eﬄuent concentration (Ceffluent) was used to determine the total iron inside the column:

MF e(t) = CinputQt Ceffluent(Qt PV ) (2.4) − − where PV denotes one pore volume of the column. The measured value Mt minus the background value M0 were then integrated over the length of the column (Equation 2.6), from this sum AM and the total injected iron MF e a conversion ratio kF e was calculated (Equation 2.6), which was used to convert the rest of the measurements into concentration proﬁles. X AM = (Mt M0) (2.5) − 0→bcolumn MF e bcolumn kF e = (2.6) AM msand where bcolumn is the number of data points along the length of the column and msand the total mass of sand inside the column. Since the metal detector can not distinguish between colloids in suspension and colloids attached to the porous medium, the concentration proﬁles always present the summed concentration of the attached mass of nZVI on the porous medium and the mass of nZVI in suspension, this summed concentration is presented in the rest of this work as Stotal to present the total mass of nZVI per mass of dry soil (Equation 2.7) in each data point.

Stotal = kF e (Mt M0) (2.7) − when the concentration as a function of wet soil would be needed, Stotal wet could be | calculated following: ρsand(1 n) Stotal wet = kF e (Mt M0) − (2.8) | − ρsand(1 n) + ρf (n) − −3 −3 where ρsand is the density of sand [ML ], ρf the density of water [ML ] and n the porosity [ ]. −

17 2. Detection and Concentration Measurement of nZVI in the Subsurface

Figure 2.1: Setup of the metal detector to non-destructively measure concentration proﬁles along the column. The metal detector was placed on a wooden frame which could move over a rails. The column was placed inside the lower opening of the metal detector ring

2.3.2.1 Linearity of the Detector

To show the linear response of the metal detector to the concentration of iron, different amounts of dry iron powder were mixed with sand. The iron powder was each time completely mixed with 140 g sand, this filled a section of 10 cm in the column. In the column the iron sand mixtures were separated by a section with 10 cm clean sand for the low concentrations and a section with 15 cm clean sand for the higher concentrations (see top of Figure 2.2). The whole column was measured by moving the detector at a rate of 10 mm/s along the column. The signal of the metal detector responded to the iron content in the vicinity of the location of the measuring coil (Figure 2.2). Thus the signal was already changing in the clean sand area when it got close to the section containing iron. In the middle of each iron filled section, the signal was at a maximum because on both sides of the measuring coil an equal amount of metal was present. To determine the maximum value for each metal containing section, over a length of 10 mm around the maximum value (12 data points in each direction plus the center point, giving 25 data points in total), the sum of the response frequency was taken. This integral value was divided by the amount of data point in 10 mm to get an average response for 1 mm of column length. Figure 2.3 shows the determined values for each integrated section.

18 2.3. Measurement of nZVI Concentration Proﬁles in a Column

Figure 2.2: Calibration setup and the recorded RAW data of the detector. The upper values above the column present the iron mass (g) per mass of sand (kg), the lower values present the iron mass (mg) per mm of column length. Each iron containing section was separated by a section containing no iron

2.3.3 About the Metal Detector

A small noise was observed in the recorded signal, present when the detector moved and when it stood still. The movement of the metal detector on the rails was nearly constant, but small vibrations in the direction of travel at the measure coil was observed. The difference between a recording with the detector standing still and moving though showed negligible difference in the noise, indicating that the vibrations could be neglected, especially when higher concentration changes are to be measured. Because the metal detector mainly responds to zero valent iron, and the calibration is performed with analytically determined values of the total iron in suspension (F etotal), the concentration of zero valent iron inside the colloids has an influence on the calibration

19 2. Detection and Concentration Measurement of nZVI in the Subsurface

1e+08

1e+07

1e+06

100000

10000

1000 Calibration values RAW data integrated over 1mm [-] f(S )=9.18⋅107S (+/- 1.137 %) 100 total total 1e-05 0.0001 0.001 0.01 0.1 1

Stotal [g/mm]

Figure 2.3: Calibration curve of the metal detector. The measured RAW data was integrated over 1 mm and set out against the average iron content in 1 mm column length. The linear 7 2 ﬁtted relation is: f(Stotal) = 9.18 10 Stotal with R = 0.9986 · coeﬃcient. Therefore the calibration has to be performed for each experiment when the concentration of zero valent iron inside the colloids has changed. This happens with time for all nZVI colloids used for this research (Section 2.6).

2.4 Measuring Iron Break Through Curves in the Container Experiment

A large container experiment was developed to determine transport behavior of nZVI in a near field scale radial flow regime. Since the container simulates a confined aquifer of 60 cm thickness, no metal detector could be conveniently moved along the radial distance of the container to obtain transient concentration profiles, like in the column experiment set up. Therefore, a fixed sensor system was thought of, installing several sensors at different locations within the container during filling it with sand. The magnetic field of a cylindrical coil (Figure 2.4) is large near the coil. Hence, changes of susceptibility are mainly detected in the near-field of the coil. Then, an optimal configuration for the measurement of changes in susceptibility is obtained if the

20 2.4. Measuring Iron Break Through Curves in the Container Experiment

Figure 2.4: Streamlines of the magnetic field strength of a cylindrical coil; energy density of the magnetic field in a cutting plane axis of the coil is parallel to the flow direction of nZVI and if this flow goes through the coil. A fairly standard technique in the context of iron detection is to use two coils for the measurement of the magnetic flux density. One coil measures the magnetic flux density of the examined material and the second coil is used for a reference measurement without magnetic material. Both coils are oppositely connected and only a difference signal is obtained. An advantage of such a configuration is that unbalances are eliminated at low-level in hardware. Though, here, a reference measurement is impossible. The measurement coils are buried and a domain without nZVI is not available while the experiment runs, and cannot be created at justifiable costs. For that reason, a post- processing of the measured signals was used instead of the reference coil [Klaas et al., 2010a]. Since the measured susceptibility of nZVI in the subsurface was unknown but was expected to be very small, the measurement system had to be as sensitive as possible. Therefore a configuration of the coils where the flow of nZVI goes through the coils was chosen. Furthermore, a coil design that concentrates the magnetic energy very close to the coil was used, because in such a measurement system the influence between neighboring coils is negligible. This fact is important, since several coils have to be used at the same time to monitor the spatial distribution of nZVI.

21 2. Detection and Concentration Measurement of nZVI in the Subsurface

2.4.1 Data Analysis and Post Processing Algorithms

Sensitivity of the measurement method depends mainly on the accuracy of the measured input current and the output voltage of the measuring coil sensor. A post-processing, which is described in the following in detail, reduces errors of the measuring components and eliminates inherent deviations of the measurement system. The magnetic ﬁeld strength H is proportional to the electric current I in the ﬁrst coil

H = CI (2.9)

C is a constant that depends on the location of the point, in which the magnetic ﬁeld is calculated, but is independent of susceptibility and electric current. This approximation is valid, if susceptibility is constant near the coil.

Since it is very diﬃcult to keep I1 constant during an experiment, I1 is determined and logged measuring the voltage U1 at the resistor R1 , which is connected in series to the coil (Figure 2.5):

U1 I1 = (2.10) R1

The electric current in the ﬁrst coil is time-harmonic (with a frequency of 10 kHz).

Hence, a voltage U2 is induced in the second coil. The induced voltage depends on the magnetic ﬂux density B, described by the frequency, susceptibility and current in the ﬁrst coil:

U2 = c2f (χ + 1) U1 (2.11)

The constant c2 is the result of the ﬂux computation through the second coil.

In practice, the voltages U1 and U2 are measured with a circuit, which converts the voltage into a digital signal for post processing on a computer (Figure 2.5). Tolerances in the circuits are eliminated by calibration of each measuring channel. For instance, voltage is precisely measured and the corresponding digital signal is stored in a look-up table. Hence, even small non-linearities of the circuit are eliminated.

The easiest way to determine the constant c2 is to perform a reference measurement with vanishing susceptibility χ = 0, all parameters measured without nZVI (χ = 0) are indicated with a prime (0).

0 0 0 U2 = c2f U1 (2.12)

22 2.4. Measuring Iron Break Through Curves in the Container Experiment

Figure 2.5: Equivalent network of the ﬁrst coil, which is used to determine the electric current in the coil IC1 from the measured current I1; measured signals are sent to a PC

where the prime denotes the measurement at χ = 0. Note, the voltage U2 in Equation 2.11 0 0 and U2 in Equation 2.12 depend directly on the frequencies f and f . Since the frequency produced by a signal generator is generally not stable enough for the desired accuracy, f is recorded during the measurement by performing a TTL-conversion of the sine wave. Finally, susceptibility is obtained from:

0 0 U2f U1 χ = 0 1 (2.13) U2fU1 −

Experimental data showed, however, that an analysis purely based on Equation 2.13 is not accurate enough in practice. The voltage U2 changed when water was injected into the container, even though water with a zero nZVI concentration has no susceptibility. An important quantity is the capacitance of the coil, which is caused by the electric field between the windings of the coil. The capacitance of a coil is proportional to the permittivity, which can be influenced by salts and surfactants in the water. Surfactants are necessary for the transport of nZVI and the injected water was a degassed tap water, without further treatment to remove salts. The flow of water could thus cause a change in the permittivity and with that influence U2 due to salts and surfactants in the water.

The equivalent network of the ﬁrst coil is depicted in Figure 2.5, the electric current I1, which is measured with help of the voltage U1, it is not equal to the current IC1 through the ﬁrst coil, though preferred, it is impossible to measure IC1 directly. A solution to avoid this problem, would be to decrease the frequency of the system. Though, that

23 2. Detection and Concentration Measurement of nZVI in the Subsurface would also signiﬁcantly decrease the measured voltage at the second coil and with that the sensitivity of the system. Hence, a solution, which avoids the inﬂuence of permittivity change, was preferred. By covering the coils completely in an epoxy resin the water can not get in between the wires and the capacitance of the coils stays unchanged. A transfer function between the voltage at the primary coil and the voltage at the secondary coil was obtained by an analysis of the equivalent network (Figure 2.5) and the basic ideas of Equation 2.10 and Equation 2.11. Small terms in the transfer function are neglected, to simplify the equation:

c3f (1 + χ) U1 U2 = 2 (2.14) 1 c4(εr)f (1 + χ) − where c3 and c4 are two integration constants, εr is the electromotive force, the voltage generated around a closed loop in an electromagnetic ﬁeld. Two calibration measurements (subscripts a and b) with non-magnetic matter (χ = 0) but slightly diﬀerent frequency are to be performed. Then, the constants c3 and c4 are given by:

0 0 0 0 0 0 2 0 0 U2b fafbU1aU2b (fb) U1bU2a c3 = 0 1 0 0 0 0 − 0 2 0 0 (2.15) fbU − f f U U (f ) U U 1b b a 1a 2b − a 1b 2a and

0 0 0 0 0 0 faU1aU2b fbU1bU2a c4 = − (2.16) (f 0)2f 0 U 0 U 0 (f 0 )2f 0U 0 U 0 b a 1a 2b − a b 1b 2a The susceptibility χ is ﬁnally obtained from U2 2 (1 c4f ) 1 c3fU1 − − χ = (2.17) U2 2 1 + c2f c4fU1

2.4.2 Sensor Design and Data Acquisition

Numerical simulations of the magnetic fields of the dual coil sensor were applied to find a suitable design of the coils. Further prerequisites were that the coils should be small enough to measure the local concentration of nZVI, but should still be large enough not to disturb the flow of nZVI in the aquifer. A cylindrical coil with a length of 108 mm and an inner diameter of 35 mm was found most suitable for generating the magnetic field in the first coil. A copper wire with a

24 2.4. Measuring Iron Break Through Curves in the Container Experiment

Figure 2.6: Coil winding machine diameter of 0.6 m (0.674 mm with a varnish) was wound in four layers of 160 windings each. The thickness of these four layers was 2.4 mm in total. For the experiment, a self-supporting coil was obtained by a varnish coated copper wire, which melts because of heating during winding and which consolidates afterwards (Figure 2.6). The second coil surrounds this first coil. The length of the second coil is 14 mm. It consists of four layers of each 40 windings. The diameter of the copper wire is 0.3 mm (0.354 mm including varnish) and the thickness of these four layers is 1.4 mm (Figure 2.7). An iso-surface was computed for the value of the magnetic field strength, which is 0.1% of the maximum value of the magnetic field strength (Figure 2.8). The color indicates the distance to the center of the coil and ranges between 200 mm in the radial direction and 300 mm in the axial direction. The streamlines of the magnetic field are cut at this iso surface. The minimum distance between two neighboring coils in an experiment can be derived by analyzing the magnetic field. If a neighboring coil lies outside the iso surface, the error, which is caused in the neighboring coil, is kept lower than 0.1%. The first coil is excited by a time harmonic electric current. Frequency and amplitude of this current are given. Here, a frequency of 10 kHz was found to be most suitable. The voltage,

25 2. Detection and Concentration Measurement of nZVI in the Subsurface

(a) Tools to wind cylindrical coils on (b) Uncoated dual coil sensors

(c) Tool to coat the dual coil sensor with (d) Coated with epoxy resin dual coil sensor epoxy resin Figure 2.7: Dual coil sensor development which is measured at the second coil, is proportional to the magnetic flux density and the unknown susceptibility, respectively. The first experiments were performed using the bare coils, later experiments were performed with coils completely covered by an epoxy resin. A special tool was developed to cover the coils with the resin (Epoxy Resin L, R&G Faserverbundwerkstoffe GmbH, Germany) (Figure 2.7c) During the first two experiments in the container the signal of the metal detection sensors was strongly influenced by the water chemistry (mainly due to surfactants in the water used to inhibit the aggregation of iron colloids). For the third transport experiments in the large container (2-D radial symmetric flow container) the sensors were coated with an epoxy resin to avoid the influence of water chemistry on the measured signal. The

26 2.4. Measuring Iron Break Through Curves in the Container Experiment

Figure 2.8: Iso surface of 0.1% of the maximum of the magnetic field strength of the first coil of the dual coil sensor coated sensors only show a change in the measured signal when ferromagnetic material is nearby. Unfortunately, however, the resin layer also reduced the magnitude of the signal and, hence, significantly decreased the sensitivity of the measuring system. The electronics developed for these sensors had thus to be optimized after the sensors were coated. Furthermore the above presented method of performing measurements at different frequencies had to be made possible. A new controller unit was designed to make it possible to measure quickly after each other at two different frequencies. This made the external sine wave generator (Figure 2.9c) obsolete. The electronics was found to be highly sensitive to sudden temperature changes in the lab (e.g. direct sunlight on the electronics or opening and closing of the building door) and daily temperature changes of day and night, the latter made it almost impossible to calibrate the system in the morning and perform measurements in the afternoon, therefore the electronics were all placed inside an isolated box in which the temperature was maintained constant (Figure 2.9d). Also, the measuring device showed fluctuations due to a 50 Hz background noise from the

27 2. Detection and Concentration Measurement of nZVI in the Subsurface

(a) Ampliﬁer card of the measuring (b) Sallen key low pass ﬁlter extension device

(c) First set up of the electronics (d) Final set up of the electronics inside an isolated box. with an external sine wave The front lid is opened generator Figure 2.9: Container measuring electronics electricity net. Sallen key low pass ﬁlters [Texas Instruments, 2002] between the peak detectors and the A/D converter were added to remove this artifact (Figure 2.9b).

2.4.3 Experimental Veriﬁcation

Two small column experiments were performed to verify the previously mentioned influence of surfactants on the measured magnetic susceptibility. Two dual coil sensors, one uncoated and one coated with epoxy resin (Figure 2.7), were each placed inside a glass column (ID: 10 cm, L: 20 cm) and filled with sand (Dorsilit, nr. 8, 0.3 0.8 mm). − First the columns were filled with degassed tap water, next two pore volumes of nZVI

28 2.4. Measuring Iron Break Through Curves in the Container Experiment

1.004 1.12 uncoated coil 1.0035 coated coil 1.1 1.003 1.08 1.0025 (-) (-) 1.002 1.06 coated uncoated χ

1.0015 χ 1.04 1.001 1.02 1.0005

1 1 0 1 2 3 4 5 Pore Volumes (a) magnetic susceptibility curves for the coated and (b) Two test columns filled with uncoated sensor; both have their own y-axis sand, the coated and uncoated sensors are inside. The left column is being injected with nZVI suspension Figure 2.10: Experiment to verify that a coated sensor is not influenced by the injection fluid chemistry suspension (RNIP 10 E, Toda Kogyo, Japan), diluted to 10 g/l, dispersed with a T25 Ultra Turrax (IKA, Germany) without further treatment were injected, the effluent was circulated for another pore volume (Figure 2.10b). Afterwards the columns were flushed 0 0 with degassed tap water for three pore volumes. At the start of each experiment U1, U2, and the frequency were measured, during the experiments U1, U2, and the frequency were continuously measured. Next, the data was processed using the post processing algorithms previously described. The nZVI suspension was injected and spread through the whole column, after flushing with fresh water most of the nZVI particles stayed inside the column. The resulting magnetic susceptibility curves of both sensors (Figure 2.10a) show that the uncoated sensor increased strongly during the presence of surfactants whereas the coated sensor shows no increased apparent susceptibility because of the surfactants in the fluid. The uncoated shows an increased magnetic susceptibility (right y-axis) in the column after the removal of the surfactants, which is because of the presence of nZVI in and around the sensor. The coated sensor shows a much smaller increase in susceptibility (left y-axis), but the change is still clearly visible. The larger apparent magnetic susceptibility of the uncoated sensor after flushing with fresh water indicates that there are still some remnants of the surfactants present.

29 2. Detection and Concentration Measurement of nZVI in the Subsurface

2.5 Measuring Iron Break Through Curves in the Field

The next stage was the development of sensors, electronics and accompanying installation techniques to measure nZVI concentrations in an aquifer during a field application (Figure 2.11). The aim of the sensor is to measure a low concentration of nZVI over a relatively large measuring range around the sensor. The sensor developed for the container experiment, was designed such that most of the magnetic field strength was focused in the center of the coil, because the nZVI could flow through the center of the sensor. By installing a sensor in the subsurface it can no longer be guaranteed that nZVI will flow exactly through the sensor and thus the sensor has to measure the change in susceptibility in the area outside of the sensor. At the ITE a numerical simulation was performed to design and optimize the shape and dimensions of the coils by computing the magnetic field with the finite element method in COMSOL Multiphysics [Li et al., 2012].

Figure 2.11: Schematic overview of an in-situ application of nZVI including monitoring

30 2.5. Measuring Iron Break Through Curves in the Field

2.5.1 Sensor Design and Data Acquisition

Compared to other shapes the sharply edged so-called racetrack coil has the widest out- wards oriented magnetic field. Moreover, there is no need to consider the orientation of the coil, since the magnetic field penetrates a circular domain around the well when installed such that the axial orientation is vertically aligned with the well (Figure 2.12). The inductive sensor presented here is composed of two identical sets of two racetrack coils. The second set of coils is connected in series to the first coil as reference. The potential difference between a region with nZVI and without nZVI is measured instead of an absolute value. The susceptibility of the considered domain in the aquifer, which is related to the concentration of nZVI, is calculated according to this voltage. The advantage of a series circuit in the primary coils is that two primary coils have the same current and generate the same magnetic field. Both of the secondary coils are oppositely connected so that the potential difference can be measured directly. This offers the possibility to measure very small changes in susceptibility and thus the electronics can be kept fairly simple. The influence of a long cable between the coils can be significant, because the resistance and capacitance of the cable increase with the length of the cable. However, the inductance of a long cable is small enough to be neglected (Figure 2.13). It is important to consider the accuracy of the measuring device, so that the device can detect the smallest change of potential difference successfully. Taking all factors into account the dimensions of the racetrack coil were obtained by a parameter sweep. An optimized design of the set of two racetrack coils was found to be able to measure 1 g/kg of nZVI at a measuring range of 10 cm, considering that a change in voltage of

Figure 2.12: Dual racetrack coil sensor. The primary racetrack coil generates the magnetic ﬁeld and the secondary racetrack coil measures the magnetic ﬂux density

31 2. Detection and Concentration Measurement of nZVI in the Subsurface

3 22

Figure 2.13: Equivalent network of the inductive dual coil measurement system with a reference sensor

10 µV can be detected by the measuring electronics. The optimized dimensions of the primary racetrack coil are 52 10 100 mm. The × × secondary coil, which fits within the primary coil, is 36 37 80 mm. The primary coil × × uses a copper wire with 0.3 mm diameter and has 513 windings. The secondary coil wire has a diameter of 0.2 mm and 4228 windings. Due to the higher amount of windings in the secondary coil, the induced voltage due to nZVI presence is amplified. A display version of the designed sensor with a transparent plastic casing is shown in Figure 2.16d. The measuring range of the sensor was determined by simulating rings of 1 cm with an inner diameter varying from 8.5 cm to 18.5 cm (Figure 2.14). The inner diameter of the well is 7.5 cm and the induced voltage by each 1 cm ring was calculated from the simulation results (Figure 2.15). A suitable magnetic field to measure nZVI is generated with an input current density of 0.125 106 A/m2. The optimal frequency was determined to be 2 kHz. At this frequency · the skin effect in the coil can be neglected and the presence of nZVI still provides a measurable effect in the secondary coil. The coils were again wound on the winding machine using two special tools (Fig- ure 2.16a). To be able to install the sensor at a field site a special casing was designed.

32 2.5. Measuring Iron Break Through Curves in the Field

Figure 2.14: Horizontal plane cross section view of the sensor in the well and its measuring domain with nZVI

Figure 2.15: Induced voltage of a ring-shaped volume of 1 cm width containing nZVI at a concentration of 1 g/kg positioned at diﬀerent measuring distances

33 2. Detection and Concentration Measurement of nZVI in the Subsurface

The plastic casing was made out of two halves, each with a space removed where the two coils could be placed in and positioned (Figure 2.16b). On both halves, half of a channel was cut out through which, when the halves were screwed together, an epoxy resin could be injected with a syringe. In total 45 ml of resin was needed to fill up the complete interior of the casing, the air and excess resin could come out on the top through the cable channel. The epoxy resin protects and fixates the coils inside, but is mainly meant to coat the coils against water chemistry. On the bottom and top a male and female thread was placed, this way the sensor could be connected to standard one inch pipes which are commonly used to build observation wells. Several sensors could be connected together with different lengths of pipe between them, to reach the desired distance between the sensors and depth of installation (Figure 2.16c). The installation would preferably be performed by using a hollow pipe direct push method (e.g. using a 3.2500 outer diameter Geoprobe) or a hollow auger drilling technique. The hollow pipe would then be pushed down with a lost tip at the bottom, closing off the pipe. The sensors will have to be a

(a) Tools to wind angular coils on (b) Coils for the dual coil ﬁeld sensor

(c) Field sensors connected on a one inch pipe which can be connected to the (d) Look- threads on the top and bottom to provide an easy installation of the sensor at through larger depths version Figure 2.16: Field sensor development

34 2.5. Measuring Iron Break Through Curves in the Field certain distance away from this metal tip, preferably a blind pipe of one meter would be connected to the bottom of the deepest sensors. Once down, the sensors connected to the pipe with all cables bound to the pipe, could be inserted. Afterwards the hollow pipe would be removed with the tip left below and the aquifer material falls back towards the sensors and pipe. By placing bentonite plugs or a socking filled with bentonite pellets, cross connections and preferential flow paths along the installation could be avoided. Be- side the sensor other measuring techniques could be installed as well together on the same pipe, creating a multi functional observation well. These could be temperature sensors, fluorescence sensors (glass fiber cables) and small liquid sampling ports.

2.5.2 Measuring Device

Since the detection of nZVI is needed at different wells and depths, the measurement system is set up modular, and can thus consist of several amplifier cards. Each amplifier card holds a sine generator with an audio power amplifier. This generates the excitation AC voltage of approximately 14 V RMS at a frequency of 2 kHz. This voltage is sent through both primary coils of the detection and reference system, which are interconnected as described above. The current through the excitation coils is measured by a shunt resistor with an adjacent RMS detector and 24 bit analog digital converter (ADC). An operational amplifier is used to amplify the few µV RMS (root mean square) difference of the voltages induced in the secondary coils. The signal then is also led to an RMS voltage amplitude detector and an ADC. In this way a small change can be measured. Due to geometrical and electrical differences of the coils, there is a signal offset to be expected, even without the presence of iron. The more similar the two sets of coils are, the lower this offset will be. With the produced coils used here, this difference was around 100 mV . The data should be collected at a central point to access all the data at once and make remote control through a GSM modem possible. Therefore, the whole system is set up as a master-slave system (Figure 2.17). The slaves, on request, will start measuring and send their data to the central master. Each slave is positioned at a well and can have up to 16 amplifier cards for sensors at 16 different depths.

2.5.3 Post-Processing Algorithms

The equivalent circuit diagram of two windings of primary coil is composed of the distributed capacitance between these windings, which is parallel to resistance and induc-

35 2. Detection and Concentration Measurement of nZVI in the Subsurface tance connected in series. However, it is difficult to measure the distributed capacitance of the coil directly by using a simple measurement. Instead, an equivalent capacitance C1 is used for the equivalent circuit diagram of the coil (Figure 2.13). The voltages and currents of the primary coil at different frequencies are measured by the measuring device. Though it is difficult to measure the capacitance of the coil, the impedance ZC1 of the coil, which is also related to the frequency, can be calculated according to Ohm’s law:

U11 ZC1 = (2.18) IC1 where ZC1 is the impedance of the coil, U11 the measured voltage at the coil, IC1 is the measured current in the coil. From the equivalent circuit diagram of the coil (Figure 2.13), the impedance ZC1 of the coil is:

R1 + j2πfL1 ZC1 = 2 (2.19) (1 (2πf) )L1C1 + j2πfR1C1 −

Figure 2.17: Measuring device and the measurement system overview for application in the ﬁeld

36 2.5. Measuring Iron Break Through Curves in the Field

where j indicates the imaginary part, R1 the resistance, L1 the inductance and C1 the capacitance of the ﬁrst reference coil.

The values of R1, L1 and C1, were determined by performing a measurement where the frequency f was varied from 100 Hz to 3 kHz. The lower and upper bounds of

R1, L1 were estimated by a measuring device, the bounds of C1 were computed by the simulation. The measured impedance ZC1 was imported and plotted in the curve ﬁtting tool of MATLAB, to ﬁt Equation 2.19 to the measured data by varying C1. The current in the primary coil decreases because of the increased inductance, when 0 0 0 0 0 nZVI is in the measuring domain of the sensor. The symbols of I1, IC1, IC2 and U12, U22, 0 U3 in Figure 2.13 are the currents and voltages without nZVI (χ = 0). The following symbols of I1, IC1, IC2 and U11, U12, U22 are the currents and voltages with nZVI χ = 0. 0 0 0 6 The currents I1, I1 and the voltages U12, U12, U22, U22 can be measured by the measuring device. The currents in the primary coils are

0 2 −1 0 IC1, (I ) = (1 (2πf) L1C1) + j2πfR1C1 I1, (I ) (2.20) C1 − 1 0 2 −1 0 IC2, (I ) = (1 (2πf) L2(C2 + Ck)) + j2πfR2(C2 + Ck) I1, (I ) (2.21) C2 − 1 (0) where C1 is the capacitance and L1 the inductance and I1 the current of the primary reference coil, and Ck is the capacitance of the cable, C2 the capacitance and L2 the inductance of the primary detection coil. Hence, if there is no nZVI around the coil, the 0 0 induced voltages U12 und U22 in each secondary coil are

0 0 0 0 U12, (U22) = j2πfMmi1IC1, (j2πfMmi2IC2) (2.22) where Mmi1 and Mmi2 are the mutual inductances of the dual coil system (reference and detection respectively). Which can be calculated according to the induced voltage U12,22 and the current IC1,C2, which are measured in the circuit by the measuring device. If there is nZVI around the sensor, the induced voltages U12 and U22 are

U12 = j2πfMmi1IC1 (2.23) χm U22 = j2πf 1 + Mmi2IC2 (2.24) kexp where kexp is the measured geometry factor as deﬁned in Equation 2.34), and χm the measured susceptibility. Since the two dual coil systems are connected in series, the

37 2. Detection and Concentration Measurement of nZVI in the Subsurface potential diﬀerence of the secondary coils from the reference and detection sensor is:

U 0 = U 0 U 0 (2.25) 3 22 − 12 0 0 0 U = j2πf Mmi2I Mmi1I (2.26) 3 C2 − C1 U3 = U22 U12 (2.27) − χm 0 0 U3 = j2πf 1 + Mmi2IC2 Mmi1IC1 (2.28) kexp − Finally, the susceptibility is obtained from

 0  U3 − U3 M (I0 I ) 0 2πf mi1 C1 C1 IC2 χm = kexp  − − + 1 (2.29)  Mmi2IC2 IC2 − 

where U3(0) is the potential diﬀerence of the reference and detection sensors’ secondary coils. The plastic casing of the sensor obstructs the nZVI from getting too close or within the coil center. The average susceptibility measured χD thus consists of two domains, one where nZVI can enter χm and one where it cannot (Figure 2.14). To calculate only the susceptibility in the area around the sensor where nZVI can be present, a geometry factor ksim has to be calculated. It is obtained through the following equations:

Φ = wMmiI1 (2.30)

Mmi = (1 + χD)cM (2.31) Φ χD = cM 1 (2.32) wI1 − χm ksim = (2.33) χD kexp = αkksim (2.34) where Φ is the magnetic ﬂux in the secondary coil, w is the windings, Mmi is the mutual inductance between two coils and cM is the constant term of it. The average susceptibility

χD in the measuring domain can be calculated according to the above equations and the susceptibility of nZVI χm can be obtained from the simulation. Therefore, the geometry factor ksim is 4.5. The calibration coeﬃcient αk is determined experimentally.

38 2.5. Measuring Iron Break Through Curves in the Field

2.5.4 Sensor Calibration

A validation of the inductive sensor between simulation and measurement was performed.

Also this experiment was used to determine the calibration coefficient αk. A hollow cylinder box divided in quarters was filled with a mixture of sand and nZVI (Figure 2.18). The compartments were filled each time with a different nZVI concentration and the compartments were arranged around the sensor to fit the numerical simulation. A sine wave generator (DG2021A, Rigol, China) was connected to the reference and measuring coil as described in the previous section. The induced voltage in the secondary coils and the current flowing through the primary coils was measured with a high precision multi meter (DM3064, Rigol, China) and the shape of the AC voltage sine wave was observed with a hand-held oscilloscope (123, Fluke, USA). Based on the measured data, the susceptibility could then be calculated by using the post-processing algorithms. A calibration of the numerical simulation and the post-processing algorithms was performed based on the data and known concentration of nZVI. The result of measurement with the hollow cylinder box showed that the concentration of nZVI was indeed proportional to its susceptibility (Figure 2.19). The inductive sensor can clearly measure a low concentration of nZVI of 1 g/kg over a radius of 10 cm. With the well known concentration of nZVI the coefficient to calibrate the geometry factor ksim was determined (αk = 0.813), thus the geometry factor kexp is 3.658. The post-processing algorithms could be validated within acceptable limits by comparing the measured data to the simulation results (Table 2.1). The larger error for the lowest concentration is acceptable because the assumption that the nZVI is spread exactly homogeneously inside the compartments is experimentally difficult to hold, which will result in a larger error for lower concentrations.

Table 2.1: Susceptibilities measured and calculated through simulation, and the error between both

Concentration Susceptibility Susceptibility Error of nZVI [g/kg] of Simulation [−] of Measurement [−] [%] 1 0.00035 0.00026 25.7 2 0.00070 0.00065 6.5 5 0.00175 0.00190 9.9 10 0.00350 0.00349 0.3

39 2. Detection and Concentration Measurement of nZVI in the Subsurface

Figure 2.18: Measurement of a hollow cylinder box ﬁlled with 10 g/kg nZVI, the measure sensor is placed in the center of the hollow cylinder box. In the background, a sine wave generator and a multi meter. In the foreground a portable oscilloscope and the reference sensor

0.005 Measurements Linear fit

0.004 χ(Stotal) = 0.00045·Stotal

0.003 χ [−] 0.002

0.001

0 0 1 2 5 10

Stotal [g/kg]

Figure 2.19: Susceptibility of nZVI at a concentration of 1, 2, 5 and 10 g/kg

2.5.5 Sensor Test

The second experiment was performed to verify the measuring technique under ﬁeld realistic conditions. The measuring sensor was placed inside a container ﬁlled with sand, after which a nZVI suspension was injected into the container. During the injection, the response of the sensor was recorded. The post processing algorithms described above were

40 2.5. Measuring Iron Break Through Curves in the Field

Figure 2.20: Placement of sensor inside the container during sand packing. Metal plates were used to separate the coarse filter sand from the fine sand filling and are removed after packing was completed used to calculate the concentration of nZVI inside the container throughout the injection. A Plexiglas container (L W H: 50 40 40 cm) was filled with sand (Dorsilit #8, × × × × 0.2 0.8 mm, Dorfner, Germany), a coarse sand was used to make a 2 cm thick filter − screen on the inlet and outlet side of the container. After closing the container with swelling clay and a silicon sealed lid, the container was flushed with degassed tap water. In a reservoir the nZVI suspension (RNIP 10E, Toda Kogyo, Japan) was prepared and dispersed at 10 g/l. The suspension was injected into the container using a constant flux dosing pump (Seepex, Germany) such that the container was flushed at one pore volume in 10 minutes, resulting in a seepage velocity of 8.3 10−4 m/s. · The measuring sensor was installed at the center inside the container while packing it with sand (Figure 2.20). The reference sensor was placed at a location with no iron in a vicinity of 50 cm. The coils of both the measuring and reference sensor were placed inside a plastic mold which was filled with an epoxy resin to avoid water contact with the coils and to stabilize the sensor. Before the injection started a background measurement was performed. During the injection the response of the sensor to the injected nZVI suspension was recorded every

41 2. Detection and Concentration Measurement of nZVI in the Subsurface

Figure 2.21: Master unit with laptop connected to directly visualize the measured data

4 seconds (Figure 2.21). Parallel to the measurement of the susceptibility, the temperature was recorded as well. The transport and final distribution of nZVI as resulted from the experiment was a close resemblance of what would be expected in a real field situation. The nZVI front was moving slower than the injection fluid. Furthermore, a concentration gradient inside the porous media along the length of the container was visually observed (Figure 2.22), and the nZVI distribution was not equal over the whole height of the container. The susceptibility clearly increased during the injection of nZVI into the container. The temperature increased before the susceptibility started to rise (Figure 2.23). This showed that the temperature can be used to prove that the injection fluid arrived at the sensor, it furthermore indicated the velocity difference between water and nZVI. The susceptibility measured in this experiment is different from the one obtained from the previous experiment, because the sensor is in direct contact with the porous medium. Thus a new geometry factor was determined (kexp = 2.573), which will

Figure 2.22: Three photos showing the propagation of nZVI inside the small container, with the time indicated as injected pore volumes

42 2.6. Chemical Measurement of nZVI in Soil & Suspension

S total 32 2 Temperature 30 1.5 28

[g/kg] 26 1 total S 24 Temperature [°C] 0.5 22

20 0 0 0.2 0.4 0.6 0.8 1 Duration [PV] Figure 2.23: Concentration and temperatue during the injection

be the same for a ﬁeld application. The calibration factor (αk = 0.813) obtained from the calibration experiment was used to simulate the new situation. The average concentration at a radius of 10 cm around the sensor inside the container was then derived from the simulation (Figure 2.23).

2.6 Chemical Measurement of nZVI in Soil & Suspension

An important factor in the characterization of nZVI is the percentage of zero valent iron present in the colloids. The nZVI colloids used in this research consist of a zero valent iron core and a shell of iron oxides (Magnetite, F e3O4) [Nurmi et al., 2005]. The percentage of zero valent iron in a particle furthermore determines the reactivity of the particle [Liu and Lowry, 2006]. According to Equation 1.2 the core of zero valent iron corrodes when the colloids get in contact with water and thus the core becomes smaller and the shell thicker. Because the colloids even corrode during storage the exact amount of zero valent iron in a suspension is changing over time. The exact amount of zero valent iron available in a suspension is important when determining the amount of base suspension necessary for an application. A simple and robust gas volumetric based method has been used to develop a set up to measure the zero valent iron content of stored nZVI samples. The setup furthermore

43 2. Detection and Concentration Measurement of nZVI in the Subsurface was developed such that soil samples could be analyzed as well. The latter was mainly applied on samples taken from the large scale container experiments.

2.6.1 Materials & Methods

The nZVI testing material was RNIP-10E (Toda Kogyo, Japan) which is a commercially available nZVI suspension. In a glass vial 4 ml nZVI suspension was injected with a pipette and deep-fried immediately. The vial was weighted (BP 210 S, Satorius, Germany, accuracy 0.1 mg) before and after adding the suspension. Afterwards the vials were dried by sublimation in an Alpha 1-2 (Christ, Germany). Afterwards the vacuum chamber was aerated with technical argon gas (Westfalen, Germany) and a 3 mm septum (Pharma Fix Septum) with an aluminum cap was placed on the vial. The septum and cap were weighted before and the freeze dried sample in the vial with cap was weighted as a total. The air pressure and temperature was measured with a digital barometer (GTD 1100, Greisinger, Germany, accuracy: pressure 1.5 mbar, temperature 1%). Based on ± ± the method described by Elion and Elion [1933], the gas volumetric setup was build. The functional scheme is given in Figure 2.24a and Figure 2.24b shows the ﬁnal set up. The vial was placed in the setup and 3 ml of 32% (10 molar) HCl (Merck, Germany) solution was added to the vial through a Sterican (B-Braun, Germany) needle with a 10 ml plastic syringe (VWR, Germany). To avoid foam from building up due to the presence of surfactant in RNIP-10E, which would enter the tubing of the setup or clog the needle, 90 µl of anti-foaming agent (Drewplus L-674, Ashland, USA) was added with a 100 µl glass syringe (Hamilton, Switzerland). From the volume of water displaced by the hydrogen gas the total moles of zero valent iron can be calculated. The displaced water volume is assumed to be equal to the produced

H2 gas volume [Elion and Elion, 1933].

0 + 2+ F e + 2H F e + H2 (2.35) →

Equation 2.35 shows the production of H2 due to the reaction with an acid, it shows 0 that the total number of H2 molecules produced is equal the number of F e molecules consumed. By using the ideal gas law, the amount of H2 molecules can be calculated from the displaced volume:

P V nH = · (2.36) 2 R T ·

44 2.6. Chemical Measurement of nZVI in Soil & Suspension

(a) Schematic overview of gas volumetric (b) Photograph of the hydrogen measurement set measuring set up, after Elion and Elion up. A, B, C, D, F and G are as described in (a), E [1933]. A is the main water reservoir is a three way valve to switch between E1 and E2, connected through G to B, which is open where E1 is the sample vial and E2 is a 1 l bottle to the air. C is the water outﬂow for large samples or soil samples. Central on the reservoir, pressure between C and A is board a barometric pressure and temperature build up due to hydrogen gas produced measuring device is positioned at E. After the reaction valve D is opened and the water is collected in F Figure 2.24: Measuring set up to chemically determine the amount of zero valent iron inside a sample

where nH2 is the amount of molecules, P the pressure, V the displaced volume, R the gas −1 −1 constant (8.312 Jmol K ) and T the temperature. Since the amount of H2 molecules 0 is equal to the amount of F e molecules, the mass of elementary iron (mF e0 ) in the sample is given by:

F e m 0 = M nH (2.37) F e · 2 where M F e is here the molecular weight of iron (55.845 g/mol). From the mass of elementary iron in the dried suspension the weight ratio of nZVI to total iron was calculated. From this ratio the real concentration of zero valent iron in the suspension can be determined. The zero valent iron content in a soil sample is determined slightly diﬀerent. The soil

45 2. Detection and Concentration Measurement of nZVI in the Subsurface sample is filled into a 1 l glass bottle (Duran, Germany), which is weighted before and after filling. To prevent oxidation of F e0 to air, the head space is flushed with Argon gas. This allows the sample to be stored for a short period. The bottle is placed in the measuring set up (location E2 in Figure 2.24b), through the three-way valve above the bottle acid (HCl, 32 %) is added with a syringe and a stainless steel needle. The three- way valve is turned such that the bottle is open towards E, and the bottle is shaken by hand to get a good mixing of the acid and the soil sample. If the content in the bottle are still black from the iron, new acid is added. This is repeated until the water level in A no longer moves down. The produced gas volume is again used to calculate the content of F e0 in the sample using Equations 2.35-2.37. When the F e0 content of the colloids of the injected suspension was determined prior to the injection, the determined concentration 0 of F e in the soil sample can also be converted into a F etotal concentration

46 3 Fundamentals of Transport of Nano Sized Zero Valent Iron Colloids

3.1 Motivation

The main goal of this work is to get a better understanding and description of the transport of nZVI colloids during the injection into the subsurface. This chapter presents a conceptual transport model based on ideas obtained from the filtration theory and the characterization of the nZVI suspension. Furthermore a method is presented to obtain quantitative measures of transport from column experiments by combining transient concentration profiles, transport equations, an adapted application of the classical filtration theory and numerical modeling. Constitutive relations were sought by looking at the filtration theory, transient concentration profiles of the column experiments and fitting parameters for a numerical model of these experiments.

3.2 Colloid Transport & Filtration in Porous Media

In a porous medium, transport of particles is limited due to filtration effects. Filtration causes particles to be removed from the liquid phase. In this section the filtration theory as described by Tufenkji and Elimelech [2004] is presented. It is based upon Happel’s sphere-in-cell model [Happel, 1958] to describe the fluid flow in a porous media and Prieve & Ruckenstein’s theory [Prieve and Ruckenstein, 1974] which concludes that all individual contributions from transport mechanisms can be summed to calculate the total deposition rate. Many scientist have used this filtration theory (e.g. [Long and Hilpert, 2009; Schrick et al., 2004; Wang et al., 2008; Zhan et al., 2008]).

47 3. Fundamentals of Transport of Nano Sized Zero Valent Iron Colloids

3.2.1 Colloid Filtration Theory

The filtration theory describes the retention of particles through a single-collector efficiency coefficient. This is the rate of particle deposition onto the collector divided by the convective transport of particles towards the collector. The collector is assumed to be a spherical sand grain as is the approaching colloid or particle. The filtration is described by looking at a single spherical grain (collector) and a single colloid (particle) in a porous medium and describing all the interactions between the collector and the particle in detail. There are several models available for the description of flow around packed spherical collectors (e.g. [Brinkman, 1947; Happel, 1958]). Happel’s sphere-in-cell model is used to describe the flow around the collector in the single-collector contact efficiency correlation equation derived by Tufenkji and Elimelech [2004]. Based on the work of Prieve and Ruckenstein [1974], the correlation equation assumes that particle deposition rates can be calculated by a summation of the individual contributions from each transport mechanism to this deposition. The transport contributions accounted for are: ηD, Brownian diffusion which is the filtration of particles not moving along regular flow lines but move by Brownian motion; ηI , the contribution due to interception which occurs when particles move along the sand grain so nearby that they get attracted

(mainly Van der Waals forces) and ηG, the removal of particles due to gravitational eﬀects (Figure 3.1, Equation 3.1).

η0 = ηD + ηI + ηG (3.1)

Figure 3.1: Basic transport mechanisms of particle ﬁltration, after Yao et al. [1971]

48 3.2. Colloid Transport & Filtration in Porous Media

With As being the porosity-dependent parameter from Happel’s ﬂow model [Happel,

1958], and by means of the dimensionless parameters for the aspect ratio NR, the Peclet number P e, the Van der Waals number NV dW , the attraction number NA, and the gravitational number NG, the combined contribution can be rewritten as one equation to deﬁne the single collector eﬃciency η0:

1 3 −0.081 −0.715 0.052 1.675 0.125 η0 = 2.4As NR NP e NV dW + 0.55AsNR NA (3.2) −0.24 1.11 0.053 +0.22NR NG NV dW with: 5 2(1 γ ) 1 As = − with: γ = (1 n) 3 2 3γ + 3γ5 2γ6 − − − 2 dp dp(ρp ρf )g NR = NA = A 2 NG = − dc 3πµdpv 18µv vdc kT NV dW = A NP e = with: Dinf = kT Dinf 3πµdp where dp is the particle diameter, dc the collector (grain) diameter, Dinf the diffusion in an infinite medium, is the Hamaker constant, k is the Boltzmann constant, T is A the absolute temperature of the fluid, µ is the absolute viscosity of the fluid, g is the gravitational constant, ρp is the particle density, and ρf is the fluid density.

3.2.2 Attachment Eﬃciency

The filtration theory predicts with η0 the attachment of the colloids on the sand grain surface at ideal conditions. The theory does not take the real concentration of the suspension or soil chemistry into account. In most natural systems the contact efficiency calculated in Equation 3.2 will provide an overestimation because of repulsive colloidal interactions between particles but also between particles and the collector. These can be influenced by ionic strengths, surfactants and surface charge, where equally signed charges will repel each other. Therefore an empirical attachment efficiency factor α was added:

η = αη0 (3.3)

The attachment eﬃciency factor (α) can be determined in laboratory column experi-

49 3. Fundamentals of Transport of Nano Sized Zero Valent Iron Colloids ments from the transport distance of particles into the porous medium or the concentration of particles in suspension at a specific distance [Tufenkji and Elimelech, 2004]. This follows from the assumption that the particle concentration decreases exponentially with the filter depth in a 1-D flow field [Iwasaki, 1937; Tien, 1989]:

− katt L Ceffluent = Cinpute v (3.4)

3(1 n)αη0v katt = − (3.5) 2dc 2d C α = c ln effluent (3.6) −3(1 n)η0L Cinput − −3 where Ceffluent is the effluent concentration [ML ], Cinput the inflow concentration −3 [ML ], katt the attachment coefficient (also denoted as particle deposition rate coefficient)[ ] and L is the characteristic filtration length [L]. In column experiments, L is − the length of the column, α can be determined fastest by looking at the effluent concentration, the break through curve will show a constant but lower effluent concentration than the input concentration with time (e.g. after 2 PV [Tufenkji and Elimelech, 2005a]).

With the calculated η0, the collector diameter dc and the porosity n, the attachment eﬃciency can be determined.

3.2.3 Sensitivity of Colloid Filtration Theory Parameters

To determine which are affecting the filtration at most and in which way, a parameter sensitivity analysis on different boundary conditions was performed. The constants presented in Table 3.1 were used to perform the sensitivity analysis. The parameters were varied over a range of field realistic values (Table 3.2). The base case value of a parameter was used as constants when another parameter was changed. The base case values were related to the conditions used in the column experiments and large scale container experiments. For each of these parameters the contact efficiency

Table 3.1: Constants for the sensitivity analysis [after Schrick et al., 2004].

Constant Abr. Value Unit −1 Liquid Density ρf 1.010 kg · l Hamaker Constant A 1 · 10−20 J = kg · m2 · s−2 −23 −1 2 −2 −1 Boltzmann Constant k 1.3805 · 10 J · K = kg · m · s · K Gravitational Acceleration g 9.81 m · s−2 Porosity n 0.34 −

50 3.2. Colloid Transport & Filtration in Porous Media

Table 3.2: Base case values and ranges of varied parameters for the sensitivity analysis

Varied Parameter Abr. Base Case Min Max Unit

Particle Diameter dp 0.070 0.001 1000 µm 5 Collector Diameter dc 476 1 1 · 10 µm −1 Particle Densisty ρp 6.7 1 10 kg · l Viscosity µ 1.003 1 10 10−3kg · m−1 · s−1 Seepage Velocity v 1.618 0.003 10 10−3m · s−1 Attachment Factor α 1.0 0.0001 1 − Temperature T 288 275 315 K

(equations 3.2 & 3.3) was calculated using Scilab (INRIA, France). Low contact efficiencies indicate a low tendency to retain particles in the porous medium. Contact efficiencies were calculated as a function of various parameters by changing one boundary condition at a time (Figures 3.2a - 3.2g). The plots all show the same range for the contact efficiency for a better comparison of the influence of each of the parameters on the total contact efficiency. The base case contact efficiency was calculated to be η = 4.409 10−3 and has been plotted in each figure. · dp From the change in contact efficiency by varying the independent parameters it becomes clear that the particle diameter has a strong non-linear effect on the filtration (Figure 3.2c). Especially the increased filtration for smaller particles is an interesting effect. The minimum contact efficiency for the particle diameter indicates that for each set of boundary conditions there is an ideal particle size to be expected. ρp The particle density of nano sized colloids shows to have an insignificant effect on the contact efficiency (Figure 3.2f). α Changing α has the expected linear effect. Since α operates directly on the contact efficiency, the contact efficiency is maximum for α = 1 and zero for α = 0. In some studies α was found to be larger than one. This could be due to extra attracting forces like an extra strong opposing surface charge [Lecoanet et al., 2004]. dc The collector size has a non linear effect on the contact efficiency, though the general trend is the reduction of the contact efficiency with increasing grain sizes. µ An increased viscosity reduces the contact efficiency. It is intuitive that the chance that the particle will deviate from the flow path is hemmed due to the increased viscosity. v The seepage velovity has a non-linear semi-hyperbolical relation to contact efficiency. −5 −5 0.72 For the presented calculation, the relation η0(v) = 8.4 10 +4.2 10 /v gives a near · · perfect fit. Especially very low seepage velocities result in a high contact efficiency while increasing higher seepage velocities results in minimal changes of the contact efficiency.

51 3. Fundamentals of Transport of Nano Sized Zero Valent Iron Colloids

η η 1 1 base case base case

0.1 0.1

0.01 0.01 Contact Efficiency η [−] Contact Efficiency η [−] 0.001 0.001

0.01 0.1 1 1e−07 1e−06 1e−05 0.0001 0.001 0.01 0.1

Attachment Efficiency α [−] Collector Size dc [m]

(a) (b)

η η 1 1 base case base case

0.1 0.1

0.01 0.01 Contact Efficiency η [−] Contact Efficiency η [−] 0.001 0.001

1e−09 1e−08 1e−07 1e−06 1e−05 0.0001 0.001 0.001 0.01

Particle Size dp [m] Viscosity µ [kg/m.s]

Figure 3.2: Contact efficiencies as a function of parameters α, dc, dp and µ calculated by changing one boundary condition at a time (Tables 3.1 & 3.2). The base case solution of η = 4.409 10−3 is shown in each graph. Continues on next page. . . · T Temperature differences are expected to be small, between 10 and 40 ◦C, the contact efficiency shows no significant change. vs. The presented colloid filtration theory was derived for deep bed filtration, ↓where the→ general velocity component is vertical and points in the same direction as the gravitation. It has been assumed that this is transferable to horizontal flow in porous media as well. In which case, the gravitation component stands perpendicular to the flow component and in a one-dimensional system thus has no influence on the particle. The sensitivity of the gravitation component in the semi-analytical solution was looked into by comparing the variation of particle sizes with and without the gravity acting on the particles.

The gravity number NG was found to have a negligible inﬂuence on the contact eﬃciency

52 3.2. Colloid Transport & Filtration in Porous Media

η η 1 1 base case base case

0.1 0.1

0.01 0.01 Contact Efficiency η [−] Contact Efficiency η [−] 0.001 0.001

0.0001 0.001 0.01 1000 2000 3000 4000 7000 10000 Seepage Velocity v [m/s] Particle Density ρ [kg/m3]

(e) (f)

η η 1 1 base case base case η, NG=0 base case, NG=0 0.1 0.1

0.01 0.01 Contact Efficiency η [−] Contact Efficiency η [−] 0.001 0.001

275 285 295 305 315 1e−09 1e−08 1e−07 1e−06 1e−05 0.0001 0.001

Temperature T [K] Particle Size dp [m]

(g) (h)

Figure 3.2: Continued. . . Contact eﬃciencies as a function of parameters v, ρp and T calculated by changing one boundary condition at a time (Tables 3.1 & 3.2). 3.2h shows the contact eﬃciency with and without gravity. The base case solution of η = 4.409 10−3 is shown · in each graph

number for colloids smaller than 0.5 µm (Figure 3.2h). The nano sized colloids used for this study are thus not expected to be inﬂuenced by gravity. Attachment of nano sized colloids thus mainly takes place due to interception, Van der Waals forces and Brownian diﬀusion.

The sensitivity analysis on the contact efficiency can be a useful tool to investigate the effect of a certain parameter on the filtration of particles during the injection into a porous medium. The sensitivity analysis can furthermore be useful to investigate an ideal (though simplified) scenario for injection. This can be used when approached from different point-

53 3. Fundamentals of Transport of Nano Sized Zero Valent Iron Colloids of-views. (I) One approach would be to find out which particle size would be most suitable for a given aquifer. This is a valuable tool that can be taken into account while designing a field application. (II) To indicate the range of porous media types or injection conditions that are most suitable for a given particle. This for example is useful when the chemical behavior of a certain given particle is preferred but it is not possible to change the size of that particle. To be able to predict the transport distance of particles during an injection in the field a non-linear numerical transport model should be used to be able to account for the suspension concentration and non-constant seepage velocity around the injection well.

3.3 Characteristics of the nZVI Suspension

Classic filtration theories and most classic research on colloid transport focuses on other colloids than nZVI. Bacteriophages [Schijven et al., 2002], viruses [Schijven and Has- sanizadeh, 2002], latex particles [Tosco et al., 2009] and many more have all been reasonably described using the classic filtration theory by approaching them as single colloids in suspension. Furthermore, most descriptions focus on suspensions with very low particle concentrations. The colloids used here are very dense (approximately seven times the density of water), are ferromagnetic and are applied in very high concentrations. During preliminary research [de Boer, 2007], it was also found that a freshly received suspension behaves different from an older one, which can be approached as an aging effect. There- fore, it was necessary to focus also on characterizing the colloidal suspension itself when looking for a good description of the transport behavior. This helps to understand the behavior of the colloids in suspension, and could lead to verification or extension of the available theories to account for the behavior observed.

3.3.1 nZVI Suspension

For this research the nano sized zero valent iron colloids (nZVI, RNIP 10E) were obtained from Toda Kogyo Corp., Japan. It was delivered as slurry with a concentration of 214 g/l. The colloids contained according to the data sheet 60% zero valent iron (F e0) and 40%

Magnetite (F e3O4). For the experiments the slurry was diluted with degassed tap water (dissolved O2 . 1 mg/l, Total Organic Carbon: 0.4 g/l, pH: 8.0, Electrical Conductivity: 8.9 S/m) to the desired suspension concentration. To break up aggregated colloids after ∼ the suspension was taken from storage, a disperser with a ﬂow through chamber (UTL25

54 3.3. Characteristics of the nZVI Suspension

Ultra Turrax with S25-IL, IKA, Germany) was used.

3.3.2 Aging of nZVI During Storage

The aging aspect (longevity) due to anaerobic corrosion (Equation 1.2) is a strong criterion for the feasibility of using nZVI for In-Situ remediation. Furthermore, it impacts the conversion factor to convert the measured susceptibility into a concentration of the total iron, because the zero valent iron content is responsible for the high susceptibility of the suspension. Measurements were performed monthly over a period of five years to get real long term data sets. The aging was determined by measuring the zero valent iron weight percentage of the total-iron mass in the sample (Section 2.6.1). Since these measurements started, similar work has been published (e.g. Liu and Lowry [2006]). For the presented measurements three different lots were used (Figure 3.3), each lot was produced in a different year, 2006, 2007 and 2008. The first measurement was performed in 2008 for all three charges. The iron was stored in a fridge at 8 ◦C, was not in contact with any other reactants beside water and this water was not refreshed during the period of observation. After each sampling the containers were aerated with nitrogen gas to remove all oxygen in the gas atmosphere. The following exponential equation was fitted to the measurement results:

−kaging·t C 0 (t) = C 0 (0) e (3.7) F e F e ·

0 where CF e0 is the F e weight percentage of the total F e mass, kaging the aging coeﬃcient and t the days of storage. Table 3.3 shows ﬁtted the constants of Equation 3.7 for the three lots and the three combined, assuming that initial F e0 weight percentage was variable instead of the 65% presented on the data sheets of the lots. Based on these equations the content of zero valent iron inside the colloids could be determined for all the experiments when the total iron was determined using a photometer.

Table 3.3: Aging coeﬃcient and initial concentration of the diﬀerent RNIP lots. According to Equation 3.7

Production Year CF e0 (0) [w%] kaging 2006 77.1 1.08 · 10−3 2007 59.3 1.21 · 10−3 2008 56.9 9.42 · 10−4 Combined 55.6 9.40 · 10−4

55 3. Fundamentals of Transport of Nano Sized Zero Valent Iron Colloids

80 RNIP−10E 2006 (storage at pH 12.2) RNIP−10E 2007 (storage at pH 12.2) 70 RNIP−10E 2008 (storage at pH 12.2) fitted curve on all data 60 2006 fitted curve 2007 fitted curve 50 2008 fitted curve

30 Content of Colloids [w%]

0 20 Fe 10

0 0 500 1000 1500 2000 Days of Storage Figure 3.3: F e0 weight percentage in the stored samples, starting concentration ﬁtted

By using this equation, it was also possible to determine the total mass of iron based on the hydrogen production in soil samples, as described in 2.6.1.

Although the reduction exponent kaging is approximately the same for all three lots, the charges all show an offset. The following arguments for the discrepancy could be reasoned, the first being most likely, the last less: (I) The lots are differently reactive. Particle size, surface area, amount of surfactants in the suspension or a Monday morning production could be some of the reasons. (II) The lots were delivered in different packaging. The 2006 lot is packed from the factory in a small PP-bottle (0.5 l) and the 2007 lot in a large one (1.5 l), whereas the 2008 lot was delivered in a PP-bag (10 l). The bag might let more oxygen through, oxygen diffusion could cause aerobic corrosion additionally to the anaerobic corrosion. Inside the bottle also a higher pressure built up due to hydrogen production can occur, which could cause the pH of the 2006 lot to be slightly higher, reducing the corrosion rate. (III) The nitrogen flushing before closing the bottle and bag could be more effective in removing the air from the bottle. (IV) The initial F e0 of the lots were different. There are no measurements or exact data available of the initial state before shipping except for the information on the data sheet. (V) The production date provided in the data sheet was erroneous.

56 3.3. Characteristics of the nZVI Suspension

3.3.3 Aging of nZVI after Dilution

Diluted suspensions quickly (< 24h) turn into a dark brown suspension (Figure 3.4), indicating rapid oxidation and high iron concentrations in solution. Even in degassed water the oxygen level is still higher than in the storage water. Degassed tap water furthermore has a neutral pH of approximately 7, whereas the storage water is around pH12, causing a signiﬁcant increase in the oxidation rate.

Figure 3.4: Oxidation eﬀect on nZVI suspensions after dilution. Left: fresh suspension, clear, Right: older suspension with dark oxidized water color. Both vials were closed and left untouched to let all nZVI settle out of suspension

3.3.4 Size of nZVI in Suspension using Stokes’ Law

During preliminary transport experiments, it was shown that older suspensions clogged the pores quicker and there appeared to be more and larger aggregates in the suspension. It was also found out that by applying mechanically high shear forces with a dispersing unit the older suspension behaved from a transport point of view again very similar to a very fresh sample [de Boer, 2007]. Similar results were presented by Phenrat et al. [2007], though, they used ultrasoniﬁcation instead. The size of the particles (or aggregates) in suspension before and after dispersion was determined through a sedimentation experiment.

57 3. Fundamentals of Transport of Nano Sized Zero Valent Iron Colloids

In order to determine the particle size, first the sedimentation rate was determined using a hydrometer (VWR NFB 35511, BS 718, ISO 649), according to the method of [Dunford and Lorentz, 1994, Ch. 3.3]. From Stokes’ law, the settling velocity (or terminal velocity) for a specific particle size can be calculated (Equation 3.8) or the particle size can be derived for a given settling velocity (Equation 3.9). Stokes derived an expression for the frictional force exerted on spherical objects with very small Reynolds numbers (e.g. nano or micro sized particles) in a continuous viscous fluid. Stokes’ law assumes that: (I) The terminal velocity is attained as soon as settling begins, (II) settling and resistance are entirely due to viscosity of the fluid, (III) particles are smooth and spherical and (IV) there is no interaction between individual particles in the solution [ ¸Cengel and Cimbala, 2006; Dunford and Lorentz, 1994].

g(ρp ρf ) 2 vp = − d (3.8) 18µ p s 18µvp dp = (3.9) g(ρp ρf ) − where vp is the particle settling velocity, g is the gravitational acceleration, ρp the density of the particle, ρf is the density of the fluid, dp the diameter of the particle and µ the fluid viscosity. From Equation 3.8 it follows that for a nZVI colloid of 70 nm and using 2 3 3 −3 g = 9.81 m/s , ρp = 6700 kg/m , ρf = 1010 kg/m and µ = 1.003 10 kg/m s, −8 · · the settling velocity is: vp = 1.51 10 m/s. · The glass vessel used for the sedimentation test was 0.5 l with a diameter of 50 mm and a height of 250 mm. During the experiment at different stages of the sedimentation, the density was measured. The sedimentation curve was constructed by plotting the normalized concentration against time. Both a dispersed suspension and a non-treated only stirred suspension of 40 g/l were tested. The concentration of the stirred suspension decreased very rapid (Figure 3.5a). Within 20 minutes almost all the nZVI was settled out. The dispersed suspension took 2 hours to reach its minimum value. Comparable sedimentation curves were also presented by Saleh et al. [2006]. To determine the observed settling velocity, the height of the glass vessel (250 mm) was divided by the time it took the normalized concentration to go from 1 to 0 (Fig- ure 3.5a). The observed settling velocity for the only stirred suspension was approximately

58 3.3. Characteristics of the nZVI Suspension

Stirred nZVI Dispersed nZVI 1 Dispersed nZVI 1 I Stirred fit II Dispersed fit III 0.8 Stirred: f(x) = -3.3 x + 1 0.8 I: f(x) = -0.0092 x + 1 Dispersed: g(x) = -0.49 x + 1 II: g(x) = -0.86 x + 1.3 Stirred: vp = 2.32e-04 m/s III: h(x) = -0.26 x + 0.61 Dispersed: vp = 3.39e-05 m/s I: vp = 6.42e-07 m/s 0.6 Stirred: dp = 8.7 µm 0.6 II: vp = 5.96e-05 m/s Dispersed: dp = 3.3 µm III: vp = 1.78e-05 m/s d = µ [-] [-] I: p 0.46 m 0 0 II: dp = 4.4 µm 0.4 0.4 III: dp = 2.4 µm C/C C/C

0.2 0.2

0 Stirred: ∆t = 0.2994 h 0 I: ∆t = 108.2 h Dispersed: ∆t = 2.05 h II: ∆t = 1.165 h III: ∆t = 3.901 h

0 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.25 2.5 0 2 4 6 Time [h] Time [h] (a) (b) Figure 3.5: Concentration change due to sedimentation of nZVI. Both suspensions had an initial concentration of 40 g/l vstirred = 2.32 10−4 m/s and for the dispersed suspension vdisp = 3.39 10−5 m/s. obs · obs · These observed fall velocities are not even close to the calculated one. The assumption that the colloids fall as single colloids has thus to be declined. With the assumption that the colloids attach to each other while in suspension and colloid-colloid interaction thus results in the formation of aggregates [Phenrat et al., 2007], an approximation of the aggregate size can be made based on the observed fall velocity. The determined velocities were used as the settling velocity in Stokes’ Law, to es- timate the diameter of the aggregates (Equation 3.9). For the non-dispersed suspension this is: dstirred = 8.7 10−6 m = 8.7 µm, and for the dispersed suspension: p · ddisp = 3.3 10−6 m = 3.3 µm. This would mean that mean diameter of the aggre- p · gates is approximately 50 100 times the diameter of the primary colloids. − The above calculated diameters were based on the average sedimentation time. The sedimentation curve of the dispersed suspension though shows three phases (Fig- ure 3.5a) of sedimentation. (I) The first half hour shows almost no change in the concentration, (II) then a rapid change occurs for about 40 minutes, (III) followed by a slower change. It can be reasoned that the differences are caused by aggregation effects and this aggre-

59 3. Fundamentals of Transport of Nano Sized Zero Valent Iron Colloids gation is not instantaneous. (I) In the first stage there are mainly single colloids or very small aggregates which are very stable, (II) in the second phase large aggregates form and settle out quickly and (III) in the last phase there are only intermediate sized aggregates and single colloids left, or as proposed and observed by Phenrat et al. [2007], interconnected aggregates have formed networks (gelation) which results in a network fluid which can hold its shape under motionless conditions [Witten and Pincus, 2004]. Since the observed third phase here is not completely stable and still sedimentation occurs, it could be assumed that either there are still single aggregates present which settle out or that the minimal vibrations in the room were enough to induce motion. On the sedimentation curve of the dispersed suspension three different lines were fitted to the three phases (Figure 3.5b). The time it took the normalized concentration to go from 1 to 0 was determined by assuming that the fitted line was valid for a suspension of equally sized particles without aging effect. The lines were then extrapolated to C/C0 = 1 and the ∆t between this intersection and the intersection with C/C0 = 0 was calculated. Next the height of the vessel 250 mm was divided by the calculated ∆t to determine the fall velocity and subsequent the aggregate (or colloid) diameter. Under the assumption that each phase represents a different aggregation stage with differently sized aggregates, the estimated average colloid and aggregate sizes of each I II III phase are: dp = 0.46 µm, dp = 4.4 µm and dp = 2.4 µm. The assumption that in the first half hour after dispersion the aggregates are fully disintegrated into single colloids does not hold by this determination. The hydrometer reading though is not extremely accurate and the first 20 minutes no hydrometer change was observed, which would result in a fully horizontal tangent, and thus in zero-sized colloids. Therefore, the first reading with a difference was included for the fitting of the first tangent. The assumption that dispersion breaks up the aggregates into single particles might still hold, though the aggregation is quick and the single particles are present only shortly. An aggregate diameter of 460 nm still is a very small colloid and could stay in suspension with the absence of further aggregation, as calculated for over 100 hours. The diameters of the second and third phase do not differ that much, though already show a distinctly different sedimentation velocity.

3.3.5 Size of nZVI in Suspension using a Laser Detector

From the data sheet as well as from some publications an average diameter of the primary colloids was determined at 70 nm [e.g. Nurmi et al., 2005].

60 3.4. Conceptual Model for Transport of nZVI

0.6 top upper middle 0.5 lower middle bottom 0.4

0.3

0.2

Particle size count [%] 0.1

0 0 50 100 150 200 250 Particle size [nm] Figure 3.6: RNIP 10-E nano particles measured with the Mastersizer 2000 (measured 3.5 years after production)

The method after Stokes law showed to be fairly unsuitable to determine the size of the primary particles of the nZVI suspension because the primary particles aggregate quickly after the measurement started and a high concentration of these primary particles are needed for the test. Therefore the suspension was analyzed through Mie scattering of a Helium neon laser (Mastersizer 2000 with Hydro G, Malvern, United Kingdom). A diluted suspension of 40 g/l of nZVI was prepared and dispersed. After preparation it was set to rest in a 40 cm high 1 l vessel to let all aggregates settle out of suspension. After three hours the (visually clear) suspension above the settled out nZVI was sampled by taking samples of 160 ml at a time. The particle size counts in percents are given for four diﬀerent levels in Figure 3.6 for the nano range. The measurement showed that nano sized particles are still in suspension, even though the concentration of stable nano particles is very low. Most iron settled out of suspension. The sample taken from the bottom shows no nano sized particles, indicating that all settled iron must have been in aggregated form or at least has aggregated once it settled out.

3.4 Conceptual Model for Transport of nZVI

Based on the characterization of the suspension and the colloid filtration theory a conceptual model for the transport of nZVI in porous media could be constructed. From the aging effects, it can be concluded that the suspension of nZVI does not behave as a typical colloid suspension for which classical deep bed filtration theories were derived, nor can Stokes’ law be easily applied to determine particle sizes. Aggregation

61 3. Fundamentals of Transport of Nano Sized Zero Valent Iron Colloids of nZVI colloids due to magnetic attraction is expected to be the main cause for the discrepancy. Collisions between colloids result in clustering and continuously increasing size of aggregates. Therefore, the suspension is under continuous change and particles of different sizes are present within the suspension. The dispersed suspension clearly showed to contain a mixture of nano sized colloids and micro sized aggregates, both behave quite differently according to the filtration theory, this has to be regarded to get a good description of the transport behavior. Roughly the suspension can be accounted to consist of particles of two different sizes, nano sized colloids and micro sized aggregates. According to the filtration theory, nano sized colloids are not influenced by gravity and are mainly filtered out of suspension due to interception, Van der Waals forces and Brownian diffusion. Micro sized aggregates on the other had are under the influence of gravity, which causes them to sediment and due to their size can result in straining at narrow pore throats. The proposed conceptual model assumes that the suspension consists of particles of two different sizes which can be determined from the sedimentation tests for the aggregates and particle size detector for the nano sized colloids. The aging effect causing increasing aggregate sizes with time is not taken into account because the injection duration is fairly short and in contrast to the undisturbed sedimentation experiments the suspension before the injection is kept mechanically in continuous disturbance and during the flow in the porous media is in movement as well. The removal of nZVI from the suspension and the deposition onto the porous medium affects both the colloidal and aggregated form of nZVI. The concept assumes that the removal of the colloids is fully described by the filtration theory and can only take place for as long there is enough available space on the grain surfaces. The removal of aggregates is mainly a sedimentation process which can also be described by the filtration theory. Removal of aggregates is not limited by the available surface area because the removal from the suspension is almost purely due to gravitation and the aggregates will mainly occupy low velocity corners in the porous media and the bottom of pores. A maximum removal of aggregates would be the available pore space, this though is not taken into account because of the short injection durations. The removed aggregates are assumed to have no influence on the maximum attached colloids on the surface of the sand grains. The nano sized colloids are expected to reach the grain surface before the aggregates do and the aggregates mainly occupy different portions of the pore space. Furthermore, for relatively short injection durations, the removal is assumed to have no influence on the porosity of the porous medium

62 3.4. Conceptual Model for Transport of nZVI

meso scale

1 m

macro scale

1 cm

micro scale

100 µm

Figure 3.7: Conceptual model for the transport and retention of nZVI during the injection

63 3. Fundamentals of Transport of Nano Sized Zero Valent Iron Colloids and the associated seepage velocity. The conceptual model is graphically depicted in Figure 3.7, the model can be summarized following:

J Two sizes of particles: nano colloids, micro aggregates

J Grain surface attachment of nano colloids, with a maximum due to availability of surface area

J Sedimentation and straining of micro aggregates, no maximum amount of aggregates retained in porous media

J The ﬁltration theory of Tufenkji and Elimelech [2004] can describe the removal of both the colloids and aggregates if regarded separately

3.5 Mathematical Model for Transport of nZVI

In the search for a good description of the transport of nZVI in porous media a robust mathematical description of the transport is preferred. A robust mathematical description in 1-D can easily be verified and fitted to column data and could provide the basis for the prediction of nZVI transport in more dimensions and larger scales. The presented mathematical model here describes the removal of nZVI from the suspension in colloidal and aggregated form and its deposition onto the porous medium. The approach assumes that the concentration of removed aggregates is a continuous addition to the total concentration of nZVI removed from the suspension. The removed aggregates are assumed to have no influence on the maximum attached colloids on the surface of the sand grains. Therefore, the removed aggregates are added to the maximum attachable colloid concentration, and thus, there is no maximum of nZVI in the porous media. This limits the applicability of the model to relatively short injection durations (or respectively small removal rates) since a prolonged injection (or high removal rate) would induce so much solid phase to the system that the alteration of the porous medium would no longer be negligible.

3.5.1 Groundwater Flow and Transport

The conservative transport of mass in a porous medium can be described by the governing equation of mass transport combined with a continuity equation including hydrodynamic

64 3.5. Mathematical Model for Transport of nZVI dispersion and advection: ∂C n + C q + q C (nD C) = 0 (3.10) ∂t ∇ · ∇ − ∇ · ∇ where C is the concentration of nZVI in suspension [ML−3], q the Darcy velocity vector [LT −1], t time [T ], n porosity [ ] and D the hydrodynamic dispersion tensor [L2T −1]. − Taking into account only the component in the x-direction the conservative advection dispersion equation becomes:

∂C ∂C ∂2C + v Dl = 0 (3.11) ∂t ∂x − ∂x2

−1 where v is the seepage velocity [LT ] and Dl the longitudinal hydrodynamic dispersion coeﬃcient [L2T −1].

3.5.2 Transport and Kinetic Removal

Non-conservative transport of mass in a porous medium can be described by the mass balance equation for the liquid phase (including hydrodynamic dispersion, advection and a sink term) combined with the mass balance equation for the solid phase, that describes the sink term of the ﬁrst one. Assuming that the porosity is constant and the transport only occurs in the direction of ﬂow, the equation becomes:

2 ∂C ∂C ∂ C ρb ∂S + v Dl = (3.12) ∂t ∂x − ∂x2 − n ∂t

−1 where S is the mass of nZVI per dry unit weight of sand [MM ] and ρb the bulk density

(ρb = (1 n)ρs, with n the porosity and ρs the density of the sand). −

3.5.3 nZVI Removal in Porous Media

The removal of colloids from the suspension, and with that the associated concentration of colloids at the solid phase S, is here treated to be not only caused by the attachment on the grain surface but also by the gravitational settling and straining eﬀects of the colloids and aggregates in the porous medium during transport. It is here proposed to build up the removal of nZVI from the liquid phase in two parts, col katt for the attachment on the grain surface due to colloid ﬁltration of single colloids and agg katt to account for the removal of aggregates from the suspension. The sink term from

65 3. Fundamentals of Transport of Nano Sized Zero Valent Iron Colloids

Equation 3.12) is described by a kinetic version of the Langmuir sorption isotherm for the removal of colloids due to filtration and a term to account for the continuous filtration of the aggregates. This results in the following definition of the sink term: ∂S n col S n agg = katt 1 C + katt C kdS (3.13) ∂t ρb − Smax ρb −

col agg where katt and katt represent the attachment- and kd the detachment coefficients [ ] and −1− Smax the maximum amount of nZVI that can be attached at the solid phase [MM ]. The aggregates removed from the suspension are assumed to have no effect on the maximum of colloids that can be attached on the grain surface, thus they were chosen to be added to Smax, which was therefore defined by:

0 Smax = Smax + Sr (3.14)

−1 agg where Sr represents the removed aggregates from the suspension (i.e. Sr = nρb katt C). Once these equations are substituted into Equation 3.13, it becomes clear that the total mass of removed colloids will show an asymptotic behavior. Concentrations taking in more space than available by the pore space are in theory possible. Though, it is assumed that the injection duration is relatively short or the removal rates are respectively small, avoiding the system to ever reach concentrations which could significantly fill up the pore space. The removal of aggregates is proposed to also be described by using the contact efficiency coefficient:

agg agg agg 3(1 n)α η0 v katt = − (3.15) 2dc

agg agg where, η0 is the contact efficiency coefficient calculated for the aggregates and α is the aggregate removal efficiency. Since the aggregates are much bigger, the removal will be mainly caused by sedimentation and straining and thus the available surface area of the sand grains is less of importance, hence, a Langmuir isotherm is not expected to be necessary for the removal of the aggregates. It is assumed that furthermore for the transport of nZVI colloids, the desorption is negligibly small and thus further has not to be taken into account. Together these equations form the following flow, transport and kinetic removal equation:

66 3.6. Transport Experiments (1D)

2 ∂C ∂C ∂ C 3(1 n)αη0v S = v + Dl 2 − 1 0 C ∂t − ∂x ∂x − 2dc − Smax + Sr − 3(1 n)αaggηaggv − 0 C (3.16) 2dc

This mathematical description takes all the assumptions of the conceptual model into account. Whether this transport equation and the ﬁltration theory are capable of fully describing the removal of both colloids and aggregates can only be veriﬁed by combining experimental with numerical work.

3.6 Transport Experiments (1D)

Transport experiments in columns were performed to increase the understanding of the occurring transport phenomena during the injection of a nZVI suspension into porous media. The column experiments were performed in horizontal two meter long columns to have transport conditions which are closer to those of a ﬁeld scale injection (Figure 3.8).

Figure 3.8: Column experiment set up

67 3. Fundamentals of Transport of Nano Sized Zero Valent Iron Colloids

The raw data of the column experiments presented here was obtained through the work of Xu [2009] and Steiert [2008]. Both descriptive and quantitative methods were applied to compare column experiments of different condition with each other and obtain a way of identifying conditions resulting in better or poorer transport. The measuring technique as described in section 2.3 was used to measure transient concentration profiles from which the information was obtained to compare the experiments among each other. Furthermore, these datasets were to be used as fitting input of the transport equation in a numerical model. Because the filtration theory does not directly take the concentration of the suspension into account, the set of experiments presented here was performed with different input concentrations of the injected suspension. It was sought if the transport is independent of the concentration, or if there is a dependency and if so, if a constitutive relationship could be obtained from these experiments.

3.6.1 Materials and Methods

Porous Medium

For the porous medium a quartz sand (Dorsilit nr. 8, d10 = 0.3 mm, d50 = 0.47 mm, d90 = 0.8 mm, Dorfner GmbH, Germany) was used. This quartz sand was delivered ﬁre dried and had a very low iron content (< 1 mg/kg). The used columns consisted of transparent Plexiglas (Acrylglas XT, Ernst Kienzle GmbH, Germany) and were 2 m long and had a 36 mm inner-diameter. The in- and outlet plugs (IWS, University of Stuttgart, Germany) were made of PTFE and were placed inside the column. A PTFE gauze (Sefar Nytal, PA5-HD225, mesh-opening 225 µm) was pulled over the front of the plugs to create a good contact with the porous medium. The screws, valves, tube-nipples and other accessories were made of various plastics. The column ﬁlling was done by dry packing using a sand raining construction with two wire gauzes placed on a 45 ◦ angle to each other, according to Rad and Tumay [1985].

After ﬁlling, the top side was closed with the outlet plug. From bottom to top CO2 gas was pushed through (6 pore volumes), which replaced the air. Next the column was saturated and ﬂushed with 6 pore volumes from bottom to top with degassed tap water

( 1 mg O2/l). ∼ Placed horizontally in the framework, the hydraulic conductivity of the column was determined by attaching two constant head containers (the one at the inlet side placed

68 3.6. Transport Experiments (1D) approximately 90 cm higher then the one at the outlet) and measuring the discharge through the column by weighting the outflow on a scale. The average packing density was 1.55 g/cm3 and the porosity was 0.34. The hydraulic conductivity was on average 1.28 10−3 m/s. · Tracer break through curves were analyzed to further characterize the porous medium in the column. The conservative tracer Uranine (C20H10O5Na2) which gives a green solution was injected and measured at the outlet of the column with a flow through cell (Hellma, 175.000 QS) inside a photometer (Lambda 20, Perkin Elmer, Germany) the − extinction at 490 nm. A Uranine pulse of 10 ml (assumed to represent a Dirac pulse) with a concentration of 8.1 mg/l in degassed tap water was injected with a syringe at the inlet of the column. A flux of 70 ml/h was applied with a peristaltic pump (Ismatec IPC 8, Novodirect, Germany).

The longitudinal dispersion length αl which is a seepage velocity independent characteristic of the porous medium was determined from the break through curve according to the following equation [Bear, 1972]:

2 mtracer (x x¯) Cmax(t) = exp − (3.17) 2Acolumnn√πDlt − 4Dlt

−3 where Cmax is the maximum concentration [ML ] at time t [T ], mtracer the mass of 2 tracer [M], Acolumn the cross-sectional area of the column [L ], n the porosity [ ], Dl − the longitudinal dispersion coeﬃcient [L2T −1], x is the observation point seen from the column inlet [L] andx ¯ the transported distance of the center of the pulse [L]. Because a conservative tracer is used and the arrival time of the center of the pulse is equal to the duration of one pore volume, x =x ¯, Equation 3.17 can be following rewritten to determine the longitudinal dispersion coeﬃcient Dl and the associated longitudinal dispersion length

αl:

m 2 1 Dl = (3.18) AnCmax,t 4πt D α = l (3.19) l v

−1 where v is the seepage velocity [LT ] and αl the dispersion length [L]. A maximum concentration of 0.594 mg/l was measured after 614.5 minutes (Figure 3.9), −4 thus the longitudinal dispersion length is αl = 5.18 10 m. The determined longitudinal · −4 dispersion length is very close to the d50 of the sand used (4.7 10 m). This supports a ·

69 3. Fundamentals of Transport of Nano Sized Zero Valent Iron Colloids

0.6 100

0.5 80 Breakthrough curve 0.4 Cumulative mass Breakthrough of 50% 60 [mg/l] 0.3

Uranine 40 C 0.2 Cumulative [%]

0.1 20

0 0 500 600 700 800 900 1000 Time [min] Figure 3.9: Tracer experiment data to determine the longitudinal hydrodynamic dispersion coeﬃcient

common rule of thumb that for a sand with a very steep sieve curve d50 can be used as the longitudinal dispersion length.

Suspension

The nZVI base suspension for all column experiments was RNIP 10-E (2007 lot, Toda Kogyo, Japan). A detailed characterization of the suspension is described in section 3.3.1. The diluted suspension was dispersed mechanically by applying high shear forces. The Dispersing unit circulated the mixed suspension inside the reservoir while breaking up the aggregates (Figure 3.10). At the start of the injection, the disperser was switched oﬀ and the mixer was kept on to keep the colloids in suspension.

Experimental Set Up

An overview of the set up is given in Figure 3.11. The whole setup of the experiment including the column and plugs was constructed without any metal parts within 1 m distance from the column to avoid interference on the metal detector reading. The framework holding the column was made of wood and placed the column 1.5 m above the metal reinforced concrete ﬂoor.

70 3.6. Transport Experiments (1D)

Figure 3.10: Disperser connected to the nZVI reservoir with mixer

The constant head container at the outflow side was kept connected during the transport experiment, the constant head at the inlet side was disconnected to connect a peristaltic pump (Watson Marlow 323, pump heads 318MC and 314X) for the injection of the nZVI suspension. This pump could deliver a constant flux until a pressure of approximately 0.2 bar (i.e. a gradient of 1 in this system). At higher pressures the discharge reduces and eventually stops. To know the exact discharge throughout the experiment the outflow was collected and weighted on a scale. The pressure was measured at the inlet and outlet side with pressure sensors (DRTR 0 1.6 bar, HYGROSENS, Germany). −

The hydraulic conductivity of the column could be measured by selecting the constant head instead of the pump at the inlet using the 3-way valve. To inject the suspension from the nZVI reservoir the 3-way valve was turned towards the pump. The water at the outlet was collected and continuously weighed using a balance (DE 12K1A, Kern, Germany) to know the exact discharge during the hydraulic conductivity measurement and the injection. The pressure sensors (PS1 and PS2) were connected in the vicinity of the in- and outlet plugs.

71 3. Fundamentals of Transport of Nano Sized Zero Valent Iron Colloids

Mixer Constant Head Tank (inlet)

Constant Head Metall Detector Tank (outlet) Pump Disperser Column 3-Way Valve PS 1 PS 2 Figure 3.11: Schematic overview of the column experiment set up

The total-iron concentration of each suspension prepared for injection was after dissolution with HCl determined using the phenanthroline colorimetric method and a photo meter (Lambda 12, Perkin Elmer, Germany).

3.6.2 Conditions

The column experiments were performed with diﬀerent suspension concentrations while maintaining the seepage velocity inside the column constant. The total injected nZVI mass was not kept constant, but the injected pore volumes were adjusted to get the total injected mass closer together. The conditions that were kept constant throughout all the column experiments were: porosity n = 0.34, hydraulic conductivity K = 1.3 10−3 m/s, porous media length · −4 3 L = 1.9 m, mass of sand msand = 3.2 kg, pore volume 1PV = 6.86 10 m . The varied · input conditions that were measured just before or during each experiment are given in Table 3.4. The average seepage velocity is presented in Table 3.4. The initially set Table 3.4: Conditions used for the column experiments. Provided are: the suspension concentration (C0), amount of injected pore volumes (PV ), volume of injected nZVI suspension (VF e), injection duration (t), total mass of injected nZVI (MF e) and the average seepage velocity (v)

−3 C0(g/l) PV (−) VF e(l) t(min) MF e(g) v(·10 m/s) 0.08 20.35 39.21 500 1.07 1.28 0.81 10.17 20.23 240 5.55 1.38 1.50 9.88 19.66 240 10.00 1.34 5.53 6.95 12.39 200 23.24 1.01 9.26 7.27 14.38 250 45.20 0.94 11.26 4.45 3.00 150 33.69 0.96

72 3.6. Transport Experiments (1D) seepage velocity for all experiment was 1.6 10−3m/s, nevertheless, the peristaltic pump · could not provide a very constant ﬂux and thus the seepage velocity was obtained from the measured discharge.

3.6.3 Results & Discussion

For one experiment successive photos with a 10 min interval are shown in Figure 3.12 Transient concentration profiles are shown in Figure 3.13 of column experiments performed with different input concentrations. The plots present subsequent concentration profiles along the column, the time in the plots is presented by injected pore volumes. The time interval between each of the measurements was ten minutes which accounts for approximately half a pore volume. Beside the presented experiments in these plots more experiments (some as duplicates) were performed and used for the further analysis of the transport behavior. The first measurement has the lowest concentrations (bottom left corner of each plot), the last measurement is always the outside profile with the highest concentrations. During the injection the lateral transport progressed while the concentration of retained colloids inside the porous medium continuously increased. In none of the experiments a maximum concentration was observed. In non of the experiments a concentration of iron occupying a volume close to that of the pore space was reached. The maximum measured volume ratio occupied by the iron was 3%. Nevertheless, the pressure gradient did change during the injection (Figure 3.14). Especially for the experiments with the highest concentrations the gradient increased significantly. This indicates that the hydraulic conductivity changed during the injection. The hydraulic conductivity of the whole system though can be lowered by a reduced conductivity at a single place. Bridging effects [Bradford and Torkzaban, 2008; de Zwart, 2007] and filter cake building [Al-Abduwani et al., 2005] directly at the inlet filter or first few sand grains after the filter would be the most likely cause of the significant hydraulic conductivity reduction, while the rest of the column was not affected. For the lower injected concentrations no significant change in the pressure gradient was observed.

73 3. Fundamentals of Transport of Nano Sized Zero Valent Iron Colloids

(a) 0 min

(b) 2 min

(d) 22 min

(e) 32 min, 1 PV

(f) 42 min

(g) 52 min

(h) 62 min, 2 PV

(i) 72 min

(j) 82 min

(k) 92 min, 3 PV

(l) 102 min

(m) 112 min

(n) 122 min, 4 PV Figure 3.12: Injection on nZVI into a column. A small volume of dyed water was injected ﬁrst to show the injection front propagation (the dark color on the right end of the column is a lightning artifact)

74 3.6. Transport Experiments (1D)

2 C0: 0.08 g/l 0.45 5.44 9.55 13.37 16.96 1.5 1.00 5.89 9.94 13.71 17.34 1.55 6.31 10.33 14.04 17.72 2.08 6.71 10.72 14.42 18.09 2.58 7.13 11.11 14.79 18.47 [g/kg] 1 3.09 7.54 11.49 15.14 18.85 3.58 7.95 11.87 15.48 19.22

total 4.05 8.35 12.25 15.83 19.60

S 0.5 4.52 8.75 12.63 16.21 19.98 4.98 9.15 13.00 16.59 20.35 0 0 500 1000 1500 2000

6 C0: 0.81 g/l 5 0.48 4.15 7.44 0.96 4.57 7.82 1.46 4.99 8.19 4 1.93 5.40 8.55 2.36 5.81 8.90 [g/kg] 3 2.81 6.22 9.23 3.26 6.63 9.56

total 2 3.71 7.04 9.88 S 1 0 0 500 1000 1500 2000

30 C0: 5.53 g/l 25 0.35 3.95 0.71 4.29 1.07 4.63 20 1.44 4.97 1.81 5.31 [g/kg] 15 2.17 5.65 2.53 5.97

total 10 2.89 6.30

S 3.24 6.62 5 3.60 6.95 0 0 500 1000 1500 2000

30 C0: 11.26 g/l 25 0.33 2.92 0.65 3.19 1.01 3.46 20 1.35 3.72 1.69 3.97 [g/kg] 15 2.02 4.21 2.33 4.45

total 10 2.63 S 5 0 0 500 1000 1500 2000 Distance from inlet [mm]

Figure 3.13: Concentration proﬁles for nZVI (Stotal, retained and in suspension combined) along the column with 10 minutes interval, the legend shows the injected pore volumes [PV ]

75 3. Fundamentals of Transport of Nano Sized Zero Valent Iron Colloids

1 0.08 g/l 0.81 g/l 5.53 g/l 0.8 11.26 g/l

0.6

0.4 Pressure gradient [-] 0.2

0 0 1 2 3 4 5 6 7 8 Injected pore volumes [PV] Figure 3.14: Pressure gradient development during the injection. The gradient is the pressure at the inlet divided by the porous media length

Water Flushing

For several experiments the effect of water flushing after the injection of iron was tested. In Figure 3.15 the difference of the iron distribution inside a column is shown for an experiment where after the injection of iron the column was flushed with fresh water at the same injection rate. The pump was kept on running during the switching of the reservoirs, with two valves the reservoir switch can be performed without pressure pulses. The removal of colloids showed to be nearly irreversible under the applied flow conditions from these measurements. The flushing applied to the other experiments (not presented here), showed comparable results. These measurements support the choice of an irreversible sink term and the negligence of the detachment coefficient in the transport equation (Equation 3.16). It can further be assumed based on these observations that the colloids will not remobilize after the injection has finished. Once after the injection, the natural ground water again flows through the injected nZVI, the natural ground water velocities are orders of magnitude smaller than the flow velocities induced during the injection, which will further minimize the chance of remobilization.

76 3.6. Transport Experiments (1D)

8 End of 5 PVs Fe Injection After 1 PV H O Flushing 7 2 Difference 6

4 [g/kg] 3 total S 2

0 0 500 1000 1500 2000 -1 Distance from the inlet [mm] Figure 3.15: Concentration change within the column after a water ﬂushing of 1 pore volume. The discharge was not interrupted between the nZVI injection and H2O ﬂushing

Column Data Analysis

Quantitative methods to describe and compare the column experiments were developed based on the datasets of the transient concentration profiles. To evaluate the concentration change within the column, the concentration at several locations within the column were plotted against the time. For a selection of the experiments these are presented for 50 cm from the inlet in Figure 3.16. The curves show that no steady state was reached during the injection duration, since the concentrations continuously increased throughout the whole injection time. From the transient concentration profiles, the distances of different mass fractions (e.g. 25, 50 or 95%) of the injected iron were determined by first integrating the concentration profile and then finding the distance where the cumulative mass reaches the searched mass fraction. To compare these transport distances with each other over time, but also from one experiment to the other, the travel distance (xF ) was normalized by the displacement of the advective front, where only the representative fraction of the advective front displacement was taken into account (Equation 3.20).

x (t) X (t) = F (3.20) F F v t · ·

77 3. Fundamentals of Transport of Nano Sized Zero Valent Iron Colloids

25 0.08 g/l 0.81 g/l 5.53 g/l 20 11.26 g/l

at 50 cm [g/kg] 10 total S 5

0 0 1 2 3 4 5 6 7 8 Injected pore volumes [PV] Figure 3.16: Breakthrough curves within the column at 50 cm during the injection. Shown are experiments with diﬀerent suspension concentrations

here XF (t) is the normalized travel distance of the mass fraction [ ], F is the mass − fraction (e.g. 0.5 for the center of mass or 0.95 for the 2σ position) of the total injected nZVI mass [ ] and xf (t) the distance where the cumulative concentration reaches the − mass fraction [L]. It follows that the advective front (F v t) can be ﬁctive when it · · becomes larger than the porous media length. This procedure is similar to the procedure used for calculating the retardation factor as applied in solvent transport. For solvent transport, the retardation describes a process of adsorption and desorption [Fetter, 1999], whereas here colloids attach and do not detach, and beside the attachment of colloids a general removal of the aggregates takes place. The retardation factor in solvent transport is applied on the concentration in the liquid phase, here the normalized travel distance describes the mass distribution at the solid phase and liquid phase combined. For the application of nZVI as in-situ remediation technique, the mass of iron in the non-mobile phase is of main interest, hence the quality of transport was chosen to be at best described by the here presented normalized transport distance. The normalized travel distance of the 50% mass fraction (center of mass) as a function of time is presented in Figure 3.17, the 50%, 75% and 98% mass fraction values at 2 pore volumes are given in Table 3.5. Values closer to 1 represent better transport, for conservative transport X50, X75 and X98 would all be equal to 1. From these values it can be seen that the normalized transport distance

78 3.6. Transport Experiments (1D)

Table 3.5: Transport results of the 1-D ﬂow experiments. X50 and X98 are the normalized transport distances at 2 pore volumes

C0 X50 X75 X98 0.08 0.024 0.025 0.029 0.81 0.075 0.093 0.38 1.50 0.077 0.11 0.47 5.53 0.087 0.096 0.14 9.26 0.15 0.16 0.18 11.26 0.12 0.13 0.14 presents a quick method to compare different experiments with very different conditions to each other. The overall normalized travel distances are largest for the experiments around 10 g/l. Experiments at higher concentrations were initially also performed but are not presented here because they showed serious clogging effects (pressure gradient larger than 1) before one pore volume could be injected. A general transport behavior becomes visible from Figures 3.17 and 3.16. The transport shows two phases in the transport, first a phase of quick movement along with the injection fluid is seen, followed by a phase during which the iron is being further

0.3 0.08 g/l 0.81 g/l 0.25 5.53 g/l 11.26 g/l

0.2

0.15

0.1

0.05 Normalized transport distance [-]

0 0 1 2 3 4 5 6 7 8 Injected pore volumes [PV] Figure 3.17: Normalized transport distances as a function of injected pore volumes for 50 % of the injected nZVI mass (center of mass). Presented are experiments with diﬀerent suspension concentrations

79 3. Fundamentals of Transport of Nano Sized Zero Valent Iron Colloids deposited over the distance of this first transport phase. During the second phase only minimal transport in the lateral direction was taking place. This effect was also visually observed during these and other (e.g. de Boer [2007]) experiments. In Figure 3.16 this is clearly observed for the 11.26 g/l injection through the two different inclina- tions of the concentration change over time. The normalized transport distance shows a more smooth change (Figure 3.17), though the normalized transport distance shows a fast reduction in the beginning followed by a slow continuous reduction during the rest of the injection. When aiming for an injection to result in high concentration (e.g. 25 g/kg) without clogging the porous medium, from this data analysis the injections performed at a concentration around 10 g/l performed best. If much larger distances are aimed and the concentration of nZVI retained in the porous medium does not have to be that high (e.g. 1 2 g/kg) an injection around 1 g/l should be applied based on these results. To reach ∼ the larger distances off course also much larger injection volumes have to be applied.

3.7 Numerical Model

The transport was modeled using the presented transport equation (Equation 3.16) in a numerical finite elements code in MATLAB (the Mathworks, Novi, MI), the finite element base of the code and the fitting algorithm were provided by D.M. O’Carrol (RESTORE, UWO, Canada), which was also used in Liu et al. [2009]. In their application the model was fitted to the break through curves of the effluent concentration of the columns. In the work presented here the model was adapted to produce the concentration profiles for the total of attached and suspended iron along the full length of the column. The flow and transport equation was replaced by the proposed one, and the model was also adjusted to produce real concentrations instead of normalized concentrations. The model was then 0 agg applied for inverse fitting to the measured data over Smax, α and α on three successive concentration profiles of each injection experiment.

3.7.1 Numerical Simulation with Fitting on Transient Column Data

The flow and transport equation was solved in the 1-D finite elements code and inversely fitted on the transient concentration profiles of the column experiments. In Figure 3.18 the measured concentration profiles and their model results are presented for four successive concentration profiles of each experiment.

80 3.7. Numerical Model

2 C0: 0.08 g/l Data 1.5 Simulation

[g/kg] 1 total

S 0.5

0 0 500 1000 1500 2000

6 C0: 0.81 g/l 5 Data Simulation 4

[g/kg] 3

total 2 S 1 0 0 500 1000 1500 2000

30 C0: 5.53 g/l 25 Data Simulation 20

[g/kg] 15

total 10 S 5 0 0 500 1000 1500 2000

30 C0: 11.26 g/l 25 Data Simulation 20

[g/kg] 15

total 10 S 5 0 0 500 1000 1500 2000 Distance from inlet [mm] Figure 3.18: Data and numerical simulation for four concentration profiles of four selected experiments with different input concentrations. Fitted was on three profiles, the fourth (most outer) profile is always the forward model result

81 3. Fundamentals of Transport of Nano Sized Zero Valent Iron Colloids

agg 0 agg Table 3.6: Calculated (η0, η0 ) and ﬁtted (α, Smax, α ) values of the simulations.

0 agg agg C0(g/l) η0(−) α(−) Smax(g/kg) η0 (−) α (−) 0.08 5.30 · 10−3 1.0 6.08 · 10−4 2.77 · 10−2 0.038 0.81 5.00 · 10−3 0.9 2.53 · 10−3 2.56 · 10−2 0.016 1.50 5.10 · 10−3 0.3 4.13 · 10−3 2.66 · 10−2 0.010 5.53 6.20 · 10−3 1.0 8.00 · 10−3 3.46 · 10−2 0.019 9.26 6.70 · 10−3 1.0 5.00 · 10−3 3.92 · 10−2 0.018 11.26 6.50 · 10−3 1.0 1.29 · 10−2 3.71 · 10−2 0.010 Averages 0.0058 0.86 Equation 3.21 0.0318 0.018

The fitting was performed on three of these concentration profiles, the fitting routine searches the minimum of the normalized summed root mean square error (RMSE) of the calculated three concentration profiles and the measured concentration profiles. The fourth profile shown in each plot was not included for the inverse fitting but is a continued calculation to see how well the model predicts the further progress of the experiment. In Table 3.6 the fitting results are presented for all experiments performed at different input concentrations. The transport including the observed continuous increase of the concentration inside the column could be accurately described due to the addition of the separate aggregate agg removal term (katt ) and the continuous addition of these removed aggregates to the denominator of the Langmuir weighting term (Equation 3.16). 0 Smax increases with higher input concentration (C0) (Figure 3.19, Table 3.6), the amount of colloids attachable on the grain surface thus showed to be a function of the input concentration. A linear constitutive relationship could be reasonably fitted to this dependency:

0 −4 −3 S (C0) = 7.81 10 C0 + 1.83 10 (3.21) max · · For α and αagg there does not seem to be any relationship towards the input concentration, they both are almost identical for all experiments and an average value should be capable of describing the transport independent of the input concentration. agg η0 was strongly over predicted by the filtration theory which results in a correction of 0.018 via the attachment efficiency factor αagg. The most likely causes for the discrepancy would be the non-spherical shape of the aggregates which results in different behavior and a lower density of the aggregates. Aggregates thus clearly get much better transported than would be expected from the sedimentation experiments and filtration theory alone. Another reason could be that the aggregates were more than 0.5 µm, in

82 3.8. Summarizing the Findings

0.016 0 Smax S0 (C ) 0.014 max 0

0.012

0.01

[g/kg] 0.008 0 max

S 0.006 0 Smax(C0) = a C0 + b 0.004 a =7.81e-04 0.002 b =1.83e-03

0 0 2 4 6 8 10 12 14 16

C0 [g/l]

0 Figure 3.19: Smax ﬁtted for experiments with diﬀerent input concentrations

Section 3.2.3 it was already mentioned that the assumption of horizontal flow instead of vertical would probably not hold for colloids larger than 0.5 µm. The larger aggregates thus will be influenced by gravity and the filtration theory assumes that the flow direction is in the same direction as gravity, which is actually not the case for the horizontal column experiments.

3.8 Summarizing the Findings

Through a sensitivity analysis of a filtration theory and the characterization of the nZVI suspension a better understanding of the expected transport behavior of nZVI could be constructed. The result is a conceptual model that describes the transport of nZVI by approaching the suspension to be consisted of two main particle sizes (nano sized colloids and micro sized aggregates) of which the size can be determined through sedimentation experiments for the aggregates and a commercial particle size detector for the nano sized colloids. A transport equation was derived to fully describe the concept. This transport equation was verified and where necessary fitted in a numerical model to transient concentration profiles of column experiments. The transport equation, measured parameters characterizing the suspension and the fitted parameters of the numerical code will be used for the upscaling and prediction of transport in more dimensions in the following chapter.

83 3. Fundamentals of Transport of Nano Sized Zero Valent Iron Colloids

84 4 Colloid Transport in a Radial Flow Field

4.1 Motivation

Column experiments as described in Section 3.6 give an indication on how nZVI is being transported in a 1-D geometry. They cannot be directly used to predict field conditions or to design a pilot or full remediation. Therefore, research of two and three dimensional transport behavior at a near field scale would be preferable. So far, in the literature, most of the research, has been done on column experiments, and transport was characterized in only one dimension. Less research has been done into the two and three dimensional description of transport of nZVI. Most of the studies involving two or three dimensions, use small scale flume experiments [Kanel et al., 2008; Phenrat et al., 2010, 2011]. Although the basic transport mechanisms of nZVI transport in porous media must be understood before accurate predictions of transport can be made, one dimensional models and experiments are insufficient to describe the rapid decrease of the seepage velocity in a radial flow field around an injection well. Furthermore, flume experiments at small scales do not sufficiently reproduce the results of field scale injections. Here, three different methods were developed to get a better description and prediction possibility of nZVI transport in radial full scale flow fields. (I) Large near field scale container experiments, (II) experiments using multiple columns and boundary conditions to experimentally discretize a radial flow field, and (III) the transfer of calibrated parameters from the 1-D model (Section 3.7) to run forward simulations of radial flow.

85 4. Colloid Transport in a Radial Flow Field

4.2 Flow and Transport in Radial Systems

4.2.1 Concept

During injection into a well, the injected fluid has to flow through a continuously grow- ing cross sectional area (Figure 4.1a) The seepage velocity around the well is therefore no longer constant (as in a column experiment) but decreases hyperbolically (for confined, fully screened and homogeneous conditions) with increasing distance from the well (Figure 4.1b).

8 7 6 5 4 3 2

Seepage velocity [mm/s] 1 0 0 500 1000 1500 2000 Radial distance [mm] (a) After Payne et al. [2008] (b) Figure 4.1: (a) The cross sectional area increases while the injection ﬂuid progresses farther from the injection well. (b) Hyperbolically decreasing seepage velocity around a fully screened well in a conﬁned aquifer of 0.6 m thick with an injection rate of 6000 l/h

4.2.2 Mathematical Description

The groundwater flow and transport equations have been presented in the previous chapter in detail for 1-D flow fields. Here, the equations will be presented for radial flow fields, which occur around a fully screened well over the height of a confined aquifer.

Steady State Groundwater Flow Equation

Steady state groundwater flow can be presented in two different ways in cylindrical coordinates when assuming there is no influence of the vertical z-direction. (I) As a function of the hydraulic conductivity and the change in hydraulic head in all directions [Schwartz and Zhang, 2003]. The discharge per unit width of aquifer is then given by the

86 4.2. Flow and Transport in Radial Systems

Darcy velocity:

1 Z b q = K˜ h with: K˜ = K(z)dz (4.1) ∇ b 0 1 ∂h ∂h q = K˜ + (4.2) r ∂θ ∂r where θ and r are the cylindrical coordinates and q is the Darcy velocity [LT −1], K˜ the average hydraulic conductivity over the thickness of the aquifer [LT −1], b the base height of the conﬁned aquifer [L] in the z direction and h the hydraulic head [L]. This can be further simpliﬁed when assuming that there is no change of the hydraulic head over θ and the hydraulic conductivity is everywhere the same in the whole domain:

∂h q = K (4.3) ∂r with K being the hydraulic conductivity. (II) The Darcy velocity can furthermore be calculated from the pumping rate at the well, using cylindrical coordinates:

Q q = (4.4) 2πbr where Q is ﬂow rate [L3T −1] and r is radius [L]. The Darcy velocity can then be divided by porosity, to obtain the seepage velocity:

Q v = (4.5) 2πnbr where v is the seepage velocity [LT −1] and n the porosity [ ]. −

Groundwater Flow, Transport and Filtration

The flow and transport equation for an incompressible fluid (Equation 3.10) was extended with kinetic removal: ∂C ∂S n + C q + q C (nD C) = ρb (4.6) ∂t ∇ · ∇ − ∇ · ∇ − ∂t Following the procedure described by equations 3.13-3.16, but taking into account that the velocity (v), the hydrodynamic dispersion (D), the collector efficiency for colloids (η0) agg and the collector efficiency for aggregates (η0 ) are no longer constants, but vary over

87 4. Colloid Transport in a Radial Flow Field space and have to be kept in the equation as a tensor, the kinetic removal term is deﬁned following:

agg agg ∂S 3(1 n)αη0v S 3(1 n)α η0 v = − 1 0 C − C (4.7) ∂t ρb2dc − Smax + Sr − ρb2dc

Assuming that the porosity is constant, the following partial diﬀerential equation can be deﬁned: ∂C = C v v C + (D C) ∂t − ∇ · − ∇ ∇ · ∇ − 3(1 n)αη0v S − 1 0 C 2dc − Smax + Sr − 3(1 n)αaggηaggv − 0 C (4.8) 2dc

4.3 Container Experiments with a Radial Flow Field

4.3.1 Motivation

The Darcy velocity, as well as other quantities and variables such as mass flux density, decrease hyperbolically with distance from the well, which cannot be represented by a single column experiment. Before field trials are started, the effect of this type of flow field was to be investigated. To be able to get a better understanding of the transferability of the description derived from batch and column experiments and filtration theory into the field scale a large scale experiment was developed. At this large scale it also became possible to test real pumping techniques and the preparation of large volumes suspension. The experiment consisted of a triangular container which represents a 60° segment of a full cylinder. Assuming radially symmetrical flow, the 60° segment can represent a full cylinder without loss of generalization. Moreover, a window installed along the side wall of the tank allows for a visual observation of flow and transport along a stream line (Figure 4.2). The main questions to be answered with the four experiments presented here were: Which transport distances are possible for nZVI colloids under field realistic condi- ◦ tions? Are the results reproducible? ◦ What is the influence of injection rate on the transport distance? ◦ Which pumping mechanism is better for the transport; pulsating or continuous flux? ◦

88 4.3. Container Experiments with a Radial Flow Field

Figure 4.2: Overview photo of the container experiment set up

4.3.2 Material & Methods

ZVI Suspensions

For the experiments presented here, nZVI colloids (RNIP 10E, Lot 2008, Toda Kogyo Corp., Japan) were used. Detailed speciﬁcations are provided in sections 3.3 & 3.6.1 The delivered slurry was diluted with degassed tap water to achieve the desired suspension concentration for each experiment. The suspension was prepared using a mixing and dispersing unit (Figure 3.10) at a high concentration ( 75 g/l), and was then ∼ pumped into a large 1000 liter reservoir for further dilution. A circular mixing unit in the reservoir was also used to keep the colloids in suspension throughout the duration of the experiment. The IKA UTL25 Ultra Turrax was used for the ﬁrst three experiments, because in total 1000 liters of suspension needed to be prepared for each experiment, it was replaced by a larger dispersing unit. The new dispersing unit (DK40, CAT, Germany) could process up to 6000 l/h, versus 600 l/h for the IKA UTL25. The desired suspension concentration was 10 g/l, the measured concentrations varied between 8.3 and 12.7 g/l.

89 4. Colloid Transport in a Radial Flow Field

Porous Medium

For all experiments the same quartz sand as described in Section 3.6.1 was used. The pore volume of the porous medium between the filter screens was approximately 300 liters. Porosity in the column experiments was approximately 0.35, and approximately 0.4 in the container experiments, due to different packing methods. There is also a difference in anisotropy when comparing the porous medium of the columns to containers. The columns were filled perpendicular to the direction of the flow, whereas the container was filled parallel to the direction of flow. The container was filled with sand using buckets and spreading the sand with a rake. The sand ( 5 cm layer) was compacted with a heavy pestle. ∼ During packing, the sensors for measuring the nZVI break through were installed. The sensors were first filled with sand and then placed horizontally in line with the direction of flow. A constant head boundary condition was implemented with a constant head tank at a height of 68 cm that collects the outflow from the outer edge of the porous medium.

Experimental Set Up

The container constructed to simulate the radial symmetrically flow field inside a confined aquifer was chosen to be triangularly shaped. The triangle represents a 60° segment of a full cylinder to make it possible to get a cross-sectional view of the experiment through a glass plate (40 mm security glass, Sudwest¨ Glas, Germany) as well as to reduce the total size. By using only a segment of the full cylinder, this scales the aquifer volume, necessary suspension volume and injection rate but not the concentration and seepage velocity and

Figure 4.3: Schematic overview of the container experiment set up

90 4.3. Container Experiments with a Radial Flow Field thus the transport behavior. The segment represents one sixth of a full cylinder, therefore, an injection of 1000 l/h into the container is equivalent to an injection of 6000 l/h at a full 360° system. The dimensions of the container (inside: b = 0.6 m, r = 1.78 m) were chosen to be comparable to a real field situation. Sand was packed into the container to create the porous medium as describe above. A reinforced lid provided an upper seal of the container. To ensure confined conditions and to prevent preferential flow along the lid a swelling bentonite (QSM200, Eijkelkamp, The Netherlands) clay layer was packed on top of the sand. Additionally the lid was air tight closed with silicon (Ottoseal S10, Otto- Chemie, Germany). The whole container was built from Polypropylene (PP) and no iron was used within at least 50 cm from the container wall to reduce the background magnetic susceptibility and thus avoid interference of the container material on the electromagnetic measurement system (Section 2.4). The injection well (∅ = 2.5 cm) was installed inside another well screen (∅ = 12 cm), to allow for the emplacement of filter sand (Dorsilit, nr. 3, 2 3.5 mm, Dorfner, Germany) to guarantee a good transfer from the well into the − porous medium and to avoid the finer porous media sand to enter the well (Figure 4.4a). At the outflow side drainage pipes (∅ = 10 cm, DIY store) were placed to create a filter screen. Around the drainage pipes filter sand was placed to prevent the washing out of sand grains, to avoid during the filling mixing of the sand and filter sand a bended metal plate of 10 cm height was pulled up 5 cm after each sand layer was packed (Figure 4.4b).

(a) Top view of the double injection filter screen (b) Top: drainage pipes used for the outflow filter with the coarse filter sand around the inner screen. Metal bended plate was moved up during injection filter. Photo taken after the injection, filling to fill the area around the drainage pipes nZVI clearly seen in sand outside of second filter with filter sand. Bottom: top view of the filter screen, and filter sand is almost clean sand when the whole container was filled Figure 4.4: Injection filter and outlet filter of the container experiment

91 4. Colloid Transport in a Radial Flow Field

Table 4.1: Set up of the container experiments 1, 2, 3 and 4, which represent a 60° wedge from a conﬁned fully saturated aquifer

Experiment 1 2 3 4 Distance between ﬁlter screens [m] 1.78 1.78 1.78 1.78 Well radius [m] 0.065 0.065 0.065 0.065 Base height [m] 0.6 0.6 0.6 0.6 Volume of sand [l] 833 833 833 833 Porosity [−] 0.4 0.4 0.4 0.4 1 PV [l] 333 333 333 333

The outﬂow of the drainage pipes was connected to a constant head tank (Figure 4.3). The suspension was injected into the container using two in parallel connected dosing pumps (Seepex, Germany) for all experiments except one, which used a piston-diaphragm pump (Figure 4.5, Grundfos Alldos). The pumping rates were kept constant throughout the experiment, creating a constant ﬂux boundary condition at the injection well. Soil samples were taken from the interior of each nZVI sensor and a representative sample was taken at the same distance but at an angle of 45◦ from the glass plate where no nZVI sensors were placed. The samples were analyzed by using the method described in Section 2.6 and the nZVI concentration was calculated based on the age of the suspension and Equation 3.7.

Figure 4.5: Two different types of pumps used. Left: two parallel connected continuous flux dosing pumps. Right: pulsating flux piston-diaphragm pump

92 4.3. Container Experiments with a Radial Flow Field

Installed Measuring Techniques

In total 27 nZVI sensors (see Section 2.4) were installed to measure the concentration of iron at different locations within the container during the injection. The sensors were placed at five different locations along the radial direction (measured from the center of the injection well: 0.2, 0.4, 0.8, 1.1 and 1.5 m), at two angular directions (seen from the glass plate at 5 and 30◦) and at three different levels (10, 30 and 50 cm). At the distance of 0.2 m only one sensor at each level was placed, because the space at this distance from the injection well was small for placing two sensors next to each other. They were buried in the sand during the container filling and placed in line with the flow field (Figure 4.6 & 4.7). Each sensor was connected through two coaxial cables (RG 58) with a corresponding channel of the electronic measurement system. The measured voltages were digitalized and logged by a computer. Finally, the data was evaluated to obtain the change in magnetic susceptibility, which is directly related

Figure 4.6: 3-D Overview of sensor locations inside the container

93 4. Colloid Transport in a Radial Flow Field

Figure 4.7: View into the triangular container during filling. Sensors of the first layer are in place. The injection well is located in the center end (blue) and the filter pipes are located in the front (yellow) for the constant head boundary (clearly shown in Figure 4.6). The total flow distance between injection (blue) and extraction (yellow) is 1.8 m, the inside height is 0.6 m and the angle is 60° to the nZVI concentration, as described in Section 2.4. In order to convert this magnetic susceptibility into iron concentration the system needed to be calibrated retrospectively after the experiment was excavated and the iron concentration within the coils were determined analytically (Figure 4.8 and Section 2.6). For the third and fourth transport experiment in the container the sensors installed were coated with an epoxy resin (Figure 2.7) to avoid the influence of water chemistry on the measured signal. During the first two tests in the container the signal of the metal detection sensors was strongly influenced by the water chemistry (mainly the surfactants inside the water which were added by the producer to prevent the iron colloids from strong aggregation). Several preliminary tests were performed as described in Section 2.4.3, these showed that the influence of the water chemistry was completely removed due to the coating. The coated sensor only showed a change in the measured signal when ferromagnetic material was inside the sensor.

94 4.3. Container Experiments with a Radial Flow Field

Figure 4.8: Samples taken out of the sensors for chemical analysis

4.3.3 Conditions

To answer the four questions posed in the motivation above, four experiments were performed and compared to each other. (Exp. 1) To determine the possible transport distance under field realistic conditions, the field trial description of Muller¨ et al. [2006a] was used as basis. An injection rate of 1000 l/h in the container experiment results in approximately the seepage velocities of their field trial. (Exp. 1 & 4) The first experiment was repeated to test the reproducibility of the experiments. (Exp. 3 & 4) An other experiment was performed at half the injection rate (Exp. 3) to see the influence of the injection rate on the transport distance. As mentioned before, in a radial flow field the seepage velocity decreases hyperbolically with distance to the injection well. Due to that the seepage velocity drops fast in the direct vicinity of the well. Recall that from the analysis of the filtration theory it was shown that the relation between the collector efficiency and the seepage velocity was non linear and could be described by a semi-hyperbolic function (Section 3.2.3). It was hypothesized that the injection rate might have only a minimal effect on the transport distance in a radial flow field. When changing the flow rate, the absolute difference in seepage velocity close to the well is large, while further from the injection well the absolute difference is much smaller. Experiment 3 was performed at 500 l/h and experiment 4 at 1000 l/h (Table 4.2), both experiments used the same

95 4. Colloid Transport in a Radial Flow Field

continuous flux dosing pump and the total amount of nZVI suspension injected was the same for both experiments. (Exp. 2 & 3) For a field application different types of pumps can be used. Two extremes were compared here, a pump providing a very stable and continuous flux (Exp. 3) through a rotor stator technique and one that provides a pulsating discontinuous flux (Exp. 2) through a large single diaphragm (Table 4.2). Both pumps are shown in Figure 4.5. Both experiments were performed at 500 l/h. The comparison of the two pumping techniques should give an indication if the pulsating effect provides a better or worse transport. The fluctuations in shear rate as a result of pulsation could improve the transport of the colloids since colloids that attach to a sand grain can potentially be remobilized by the high shear rates exited by the high flow rate at the beginning of each pulse. It would though be worse if the increased flow rate at the beginning of each pulse is not high enough to remobilize the colloids and the lower flow rates (down to zero-flux) would result in a stronger removal of colloids.

Each of the experiments consisted of three injection phases. (I) Five liters of tracer solution (Uranine, 0.5 g/l) were injected followed by (II) 900 liters of nZVI suspension. The aimed suspension concentration was for all four experiments 10 g/l. (III) An injection of 900 liters of degassed tap water (DO: 1 mg/l, pH: 7.0).The conditions of ∼ the four experiments are presented in Table 4.2. Experiments 1 and 4 were performed as duplicates, experiments 2 and 3 were performed at half the injection rate, and 2 was performed with a diﬀerent pump.

Table 4.2: The parameters used in the container experiments 1, 2, 3 and 4, which represent a 60° wedge from a conﬁned fully saturated aquifer

Experiment 1 2 3 4 Experiment Injection Rate [l/h] 1000 500 500 1000 Pump type Continuous Pulsating Continuous Continuous Number of PVs injected [−] 2.7 2.7 2.7 2.7 Duration of injection [min] 55 110 110 55 nZVI age [days] 184 338 556 921

Cinput set [g/l] 10 10 10 10

Cinput measured [g/l] 8.3 9.9 10.5 12.7

96 4.3. Container Experiments with a Radial Flow Field

4.3.4 Results and Discussion

In total four experiments were carried out to test the transport of nZVI colloids in a confined aquifer using the container setup. Only the results of the nZVI sensors from experiment 1 are shown, the strong influence of the water chemistry on the measurement resulted the measurements of experiment 2 to go off scale. For experiments 3 and 4 the coated nZVI sensors were installed, unfortunately, due to the resin around the sensors the total signal was also reduced and during experiment 3 the sensitivity of the measuring system was not enough anymore to get representative iron break through curves, this was solved for experiment 4 but there the datafile unfortunately got corrupted during saving. Each of the injections of nZVI was followed by an equal volume of fresh water, visually no change in the distribution could be observed (compare Figures 4.10d and 4.10e), also in the measurements with the nZVI sensors during the water flushing no change was observed (Figure 4.11a). The presence of the surfactants in the suspension (although the concentration of surfactants is low in the strongly diluted suspension) could enhance the transport during the suspension injection and the lack of these surfactants during the water flushing would then stop the transport. For the container experiments the Reynolds number (Re) was calculated to test the validity of assuming laminar flow throughout the whole container and thus the applicability of Darcy’s Law to describe the flow field in the container. Reynolds Number is a dimensionless indicator for the type of flow, the exact boundaries of the flow types laminar and turbulent are case specific and also depend on the exact choice of the characteristic length, Bear [1972] describes four different characteristic lengths and states that all of them result in an upper limit for laminar flow at a value of Re between 1 and 10. Thus, for Re values below 1 10 there correlation between pressure and Darcy velocity is linear − and for Re values between 1 10 and 100 a transition from laminar to turbulent flow − takes place because inertial forces start to predominate the flow. Reynolds number is calculated by:

ρ qL Re = f (4.9) µ

−3 −1 where ρf is the fluid density [ML ], q the Darcy velocity (or specific discharge) [LT ], L the characteristic length (often taken to be the representative grain diameter for the −1 −1 porous medium, d50)[L] and µ the viscosity [ML T ] [Bear, 1972]. Figure 4.9 shows the resulting function for L = d50 = 0.6 mm. In the experiment the filter screens were positioned at 6.5 cm and 184.4 cm from the center of the injection well which are indicated

97 4. Colloid Transport in a Radial Flow Field

6 m3/h 3 10 3 m /h

1 Reynolds number [-]

0.1

0 0.5 1 1.5 2 Distance from injection well center [m] Figure 4.9: Dependency of Reynolds number on r between the two ﬁlter screens of the container. Used parameters were: −4 3 3 d50 = 6 10 m, b = 0.6 m, n = 0.4, ρf = 1 kg/l, µ = 1.003 g/m s, Q1 = 6 m /h, Q2 = 3 m /h, · · which represent container experiments 1 and 4 for Q1, and 2 and 3 for Q2 by the blue and yellow vertical lines in Figure 4.9. The graph shows that for both injection rates the Reynolds number is below 10 in the whole domain, although close to the well the injection rate of 1000 l/h is only just below 10. 60 cm from the injection well screen it is below 1 for the injection at 1000 l/h and for the injection rate of 500 l/h this is already after 27 cm.

Injection at 1000 l/h (Exp. 1)

During experiment 1 (Table 4.2) it was possible to get the iron colloids transported over the full length of the container (Figure 4.10). While visually the window was black throughout, the sensors showed a large gradient in concentration along the ﬂow path. Vi- sual observation through the glass plate showed signiﬁcant retention of nZVI compared to the injection front which was highlighted by a conservative tracer (Uranine, Figure 4.10b). The iron front visibly arrived at the outlet side after three pore volumes of suspension were injected.

98 4.3. Container Experiments with a Radial Flow Field

(a) t = 3 min, Uranine front (green) directly in front of nZVI front (black)

(b) t = 1 PV , Uranine front (green) at outlet, nZVI front (black) progressing slower

(d) t = 3 PV , nZVI front (black) close to outﬂow

(e) t = 3 P V nZV I + 3 PVH2O, nZVI front (black) very stable during water ﬂush Figure 4.10: Uranine and nZVI front during the injection at diﬀerent time intervals. Shown is experiment 1 at 1000 l/h

99 4. Colloid Transport in a Radial Flow Field

In Figure 4.11a the magnetic susceptibility is shown as a function of the injection time for five successive locations inside the container. It shows that at each location the magnetic susceptibility initially increases strongly followed by a slight, steady increase until the injection of the nZVI suspension stopped. This steady increase shows the real break through curve of nZVI. After the injection with nZVI, the suspension fluid was chased by degassed tap water, resulting in a drop of the magnetic susceptibility. This drop was caused due to an influence of the surfactants on the capacity of the coils (Section 2.4.3). As mentioned before, later the coils were covered with an epoxy resin, which prevents this jump and drop in the signal. The susceptibility during the water flushing stage re- mained stable, indicating that the colloids no longer moved and, hence, that the iron concentration inside the container did not change any more. The magnetic susceptibility is directly linearly related to the nZVI concentration, as can be seen by comparing the final magnetic susceptibility after water flushing (Figure 4.11a) with the average of the measured concentrations through chemical analysis (Figure 4.11b). Therefore, it is also possible to plot the breakthrough curves directly in concentrations, which was presented in theory in Section 2.2. It was chosen not to present this for this data set because the measured susceptibility was biased due to the water chemistry and the concentrations calculated would easily lead to misinterpretation of the plot.

1.05 r=0.24 m r=0.43 m 1.04 r=0.80 m r=1.17 m 30 r=1.54 m r=0.24 m r=0.43 m 25 r=0.80 m 1.03 r=1.17 m r=1.54 m 20 χ [-] 1.02

, [g/kg] 15 total S 10 1.01 5

1 0 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 0 0.5 1 1.5 2 Time [h] Distance from the injection well [m] (a) (b) Figure 4.11: (a) Magnetic susceptibility changes during the injection of nZVI (0 1 h) and − water flushing (1 2 h, not all shown) of experiment 1. Each coil pair responded differently to − the surfactants, after the water flushing the sensors showed the actual susceptibility without influence from water chemistry. (b) nZVI concentration inside the coils obtained by chemical analysis. From left to right the values represent the sensors at different distances; the final susceptibilities are directly related to the average values of the chemical analysis

100 4.3. Container Experiments with a Radial Flow Field

Reproducibility of the container experiments (Exp. 1 & 4)

The conditions of experiment 4 were chosen identical to the conditions of experiment 1 to test the reproducibility of the experiments. The boundary conditions of both experiments were not exactly the same, it is with experimental investigations nearly impossible to perform an experiment twice with exactly the same boundary conditions. The main deviations determined for these experiments were the dispersing unit and the input concentration of the nZVI suspension. In a visual comparison (Figure 4.12), the transport distances of both experiments are similar but not identical, experiment 4 shows higher concentrations and a sharper front. Unfortunately, a comparison of the break through curves of the inductive iron detectors could not be performed because the datasets were corrupted during saving. A comparison of the concentrations obtained in the chemical analysis of the soil samples taken during excavation is shown in Figure 4.13. Looking at the average of these measurements, it can be seen that the transport of experiment 4 to a large extend repli- cated the results of experiment 1. The measured values all are on average a little higher, this is mainly due to the higher concentration of nZVI injected. During experiment 1, the dispersing was done using the much smaller IKA machine as compared to the large CAT machine used in experiment 4. In experiment 4 the concentration at three different levels inside the filter sand of the injection well was measured additionally. Even though some iron is located inside the filter sand, the concentration is low compared to the concentrations at 20 cm from the injection well. A risk of well clogging is based on these data minimal for this material, and the filter sand worked well to provide a good connection between the injection well and the porous medium. It can be seen that the measuring technique to analytically determine the nZVI concentrations inside the soil samples (Figure 4.13) has been improved significantly since the first experiment, the measured concentrations for each location have a much smaller deviation. During the first experiment this measuring set up was still in a testing state, whereas from experiment 2 on the set up was replaced by the one described in Section 2.6.

Inﬂuence of Diﬀerent Injection Rates (Exp. 3 & 4)

Experiments 3 and 4 are here compared to discuss the influence of the injection rate on the transport in a radial flow field. In Figure 4.14 the final iron distribution can be seen for both experiments. In experiment 4 (1000 l/h) the iron is transported slightly further.

101 4. Colloid Transport in a Radial Flow Field

Figure 4.12: Final transport results of the two container experiments performed at 1000 l/h, top: experiment 1, bottom: experiment 4

30 Exp. 1: 1000 l/h Exp. 4: 1000 l/h 25

[g/kg] 15 total S 10

0 0 0.5 1 1.5 2 Distance from injection well [m] Figure 4.13: Chemically analyzed iron concentrations at diﬀerent locations in the container obtained from soil samples taken during excavation

102 4.3. Container Experiments with a Radial Flow Field

Looking at the chemically analyzed soil samples (Figure 4.15) the injection at experiment 3 (500 l/h) shows higher concentrations closer to the well. Further from the well, the concentrations are very similar. The diﬀerences in concentrations close to the well are in accordance with the hypothesis that the transport of nZVI is mainly inﬂuenced in this region. Nevertheless, since the concentrations in experiment 4 (1000 l/h) are lower in the vicinity of the well and thus the danger of clogging is reduced, this higher injection rate has to be preferred.

Pulsating Flux Pump vs. Continuous Flux Pump (Exp. 2 & 3)

The flow rate in experiment 2 was discontinuous and the pressure relaxed from 10 to 0 bar during each pulse, therefore the flow rate of each pulse was higher at the beginning of the pulse and lower at the end. This resulted in a very erratic flow regime. Although the high pressure that was induced at each pulse of the membrane pump ( 10 bar) was ∼ higher than the designed limit of the container (1 bar), this pressure only existed for a very short time at the beginning of each pulse and no leakage problems arose. In both the photographs in Figure 4.16 and the chemical measurements presented in Figure 4.17 it can be seen that the pulsating injection provided a poorer transport result. From the pulsating flux pump (Exp. 2), nZVI was mainly retained close to the well and there was a fairly sharp front. Applying the continuous flux pump (Exp. 3), the transport of nZVI was better and the front was less sharp. From these results it can be concluded that the pulsating pumping technique did not provide the improved transport of nZVI as was hypothesized. A possible cause could be that in the interval between the pulses the colloids can sediment out of the suspension faster due to the lack of shear forces and that the increased shear rate at the beginning of each pulse was not high enough to remobilize the attached or sedimented colloids.

Summarizing the Findings

The comparison of experimental results was difficult because beside the desired change of one boundary always other boundary conditions changed as well. The main discrepancy was the inconsistency of the input concentrations, which was expected to have affected the final results significantly. The injections showed that transport over a distance of nearly two meters was possible. Each of the injections of nZVI was followed by an equal volume of fresh water, visually no change in the distribution could be observed, further research could possibly provide a more conclusive statement with respect to the exact reason for

103 4. Colloid Transport in a Radial Flow Field

Figure 4.14: Comparison of the ﬁnal transport result of the container experiments performed at 500 l/h (Exp. 3) and 1000 l/h (Exp. 4)

40 Exp. 3: 500 l/h Exp. 4:1000 l/h 35

[g/kg] 20 total

S 15

0 0 0.5 1 1.5 2 Distance from injection well [m] Figure 4.15: Chemically analyzed ﬁnal nZVI concentration distribution of two diﬀerent injection rates

104 4.3. Container Experiments with a Radial Flow Field

Figure 4.16: Comparison of the ﬁnal result after three pore volumes of injected suspension in the container experiments with the pulsating membrane pump and the continuous ﬂux pump. Both performed at 500 l/h

45 Exp. 2: Pulsating 40 Exp. 3: Continuous

25 [g/kg] 20 total S 15

0 0 0.5 1 1.5 2 Distance from injection well [m] Figure 4.17: Chemically analysed ﬁnal nZVI concentration distribution of two diﬀerent pumping techniques

105 4. Colloid Transport in a Radial Flow Field the immobilization of nZVI during water flushing. It was also shown that a continuous flux is preferable over a pulsating flux for injection. The influence of the injection rate is less pronounced in the large scale experiments than in the column experiments due to the hyperbolically decreasing seepage velocity in a radial flow field. Nevertheless the injection at the higher injection rate performed slightly better. Especially the concentrations close to the well were lower, avoiding the risk of clogging.

4.4 Discretized Radial Flow Field Reproduced with Columns

4.4.1 Motivation

The container experiment is a very useful tool to provide insight into the impact of real field scale methods like pumps used and large suspension volumes preparation on the transport of nZVI. The experiments on the other hand are also very labor intense and costly. For a quick screening of applicability of a certain nZVI, porous medium or injection rate, the container experiment would simply take too much time and effort. Therefore simplified quick screening methods were developed. The first to be developed was an experimental method to approximate the results of full scale radial nZVI injections by discretizing the full radial flow field into hollow cylinder segments and representing each of these with 1-D column experiments using sets of multiple subsequent column experiments with the appropriate boundary conditions [de Boer et al., 2008; Estrella, 2011a]. The measured concentration profile from each column was assumed to represent the concentration profile of that section of the radial flow field. This would allow for the quick approximation of a nZVI injection using only column experiments. The method was tested and validated against the results of two of the near field scale container experiments performed with the continuous flux dosing pump described before (experiments 1 and 3).

4.4.2 Concept

Equations were derived to discretize a two dimensional radial flow field into a series of sections (as presented in Figure 4.18). These equations were derived so that the boundary conditions from a full scale radial flow system could be applied to a discretization with any number of columns.

106 4.4. Discretized Radial Flow Field Reproduced with Columns

To implement the boundary conditions of radial flow to one dimensional columns, a set of equations was set up. These apply the average Darcy velocity across sections of the flow field to columns of the same length as the distance between the two radii of each section (as presented in Figure 4.19).

Derivation

An injection well is chosen to be fully screened over the thickness of a confined aquifer 1 and therefore the velocity decreases by r with increasing distance r from the injection point, resulting in a radial flow field, as follows from continuity (Equation 4.4). The derivation is graphically depicted including the nomenclature of all the notations used in Figure 4.18. The 2-D flow field was discretized in ring-shaped sections (Figure 4.18). It was chosen to do this such that the average velocity over each section maintains the same ratio, ξ, to the velocity of the next section (the ratio ξ does not have to be a constant, refinements of the discretization are for example possible but not aimed for in this derivation).

q˜i = ξ q˜i+1 (4.10)

Where i refers to one section and i + 1 refers to the next one (away from the injection well). The average velocity in each section can be calculated as an arithmetic mean of the integrated continuity equation (eq. 4.4).

1 Z ri+1 q˜i = q(r) dr (4.11) ri+1 ri − ri the radii ri+1 refers to the boundary between adjacent segments i and i + 1. To describe the transport of nZVI the mass ﬂux, M˙ , is considered:

∆M M˙ = (4.12) ∆t where ∆M is the total mass transported in time ∆t. Mass ﬂux can also be described as a function of concentration and volume ﬂux:

˙ ˙ Mi = Ci Vi (4.13)

˙ where Ci is the concentration of the suspension at the boundary ri and Vi the volume

ﬂux through the boundary at ri.

107 4. Colloid Transport in a Radial Flow Field

in CN out J Ci+1 i+1 6

Cin i+1 out Ci rN Ji 6 in r Ci i+1 Vi+1 ri ˙ ˙ M˙ N Vi -Mi -Mi+1 - r0

∆ Mi ∆ t q˜i i R

q˜ ∆ Mi+1 R i+1 ∆ ti+1

q˜ R N

out in out in in Ci Ci+1 Ci+1 C -Ci - - -- N -M˙ i -M˙ i+1 -M˙ N Vi Vi+1 -Ji -Ji+1 -JN q˜ q˜ q˜ ∆Mi -i ∆Mi+1 -i+1 -N ∆ti ∆ti+1 r0 ri ri+1 rN

Figure 4.18: Graphical overview of the nomenclature of all the notations used for the discretization of transport in a radial flow field. The rectangle shows the cross section of the top view, relation of both figures is 1:1

108 1 4.4. Discretized Radial Flow Field Reproduced with Columns

The volume ﬂux can also be rewritten in terms of the Darcy velocity due to continuity:

˙ Mi = Ci Q (4.14) ˙ Mi = Ci qi Ai (4.15) where Ai is the cross-sectional area of the boundary at ri. Although the mass flux is a constant throughout the system, the mass flux density (J) is not. Mass flux density can be described as follows:

M˙ J = (4.16) A The ratio of the mass ﬂux density from one radius to the next can be derived as follows:

˙ ˙ Mi = Mi+1 (4.17)

JiAi = Ji+1Ai+1 (4.18) Ai Ji+1 = Ji (4.19) Ai+1 2πbri Ji+1 = Ji (4.20) 2πbri+1 ri Ji+1 = Ji (4.21) ri+1 Because the mass ﬂux density (Equation 4.16) can also be described by the Darcy velocity:

CQ J = = Cq (4.22) A Equation 4.21 can be rewritten as (for this derivation it was assumed that the mass ﬂux is conservative, otherwise sink terms would have to be added to this set of equations):

˜ ri ˜ Ji+1 = Ji (4.23) ri+1 ri Cq˜i+1 = Cq˜i (4.24) ri+1 ri q˜i+1 = q˜i (4.25) ri+1 ri+1 q˜i = q˜i+1 (4.26) ri

109 4. Colloid Transport in a Radial Flow Field

The ratio, ξ, as introduced in Equation 4.10 is thus given by: r ξ = i+1 (4.27) ri

Experimental Application

Calculation of the appropriate experimental conditions is necessary to accurately discretize a radial flow field of a field application or a large scale radial flow experiment into 1-D column experiments. Equation 4.4 was used to calculated the average Darcy velocity between two radii as follows: 1 Z ri+1 Q q˜i = dr ri+1 ri 2πrb − ri Q Z ri+1 1 = dr 2πb(ri+1 ri) r − ri Q h iri+1 = ln r 2πb(ri+1 ri) | | ri − Q r = ln i+1 2πb(ri+1 ri) | ri | − Q = ln ξ (4.28) 2πb(ri+1 ri) | | − An example of the average Darcy velocity over three different sections in a radial flow field is also given in Figure 4.19a. The radii for each column can now be calculated such that the ratio ξ, is consistent throughout the entire domain. Given the well radius r0 (thus r0 > 0), the outer radius of the full domain rN and the desired number of discretization sections or columns N, ξ can be calculated. From the definition of Equation 4.27, the following definition of ξ can be derived using r0, rN and N, to explain, first r0, r1 and r2 are substituted: r 0 = ξ r1 r r = 0 1 ξ r r0 r r = 1 = ξ = 0 (4.29) 2 ξ ξ ξ2

110 4.4. Discretized Radial Flow Field Reproduced with Columns thus: r r = 0 and: N ξN 1 r N ξ = 0 (4.30) rN

Using Equation 4.28 and the radii, r0 . . . rN , the average Darcy velocity in each column column can be determined. To determine the required ﬂow rate for each column, Qi , the average Darcy velocity must be multiplied by the cross sectional area of the column,

Acolumn:

column Qi =q ˜iAcolumn (4.31)

The volume of the solution injected into the full radial system is determined by the number of pore volumes. The time for one pore volume (PV ) to be injected at the given ﬂow rate Q is calculated as follows:

V 1PV = t (4.32) Q 1PV

Where V1PV is the pore volume of the full system (from r0 to rN ), and t1PV is the time required for the injection of one full system pore volume. The total time of injection for the ﬁrst segment is determined by multiplying the time for one pore volume by the desired number of pore volumes. Each subsequent segment will have a shorter injection duration, the duration of the previous segment minus the residence time of that previous segment. Time for subsequent segments can thus be calculated as follows:

V1PVi ti+1 = ti (4.33) − Q

where ti is injection time for segment i and ti+1 that of the subsequent segment and V1PVi the pore volume of segment i. The injection volumes for each column that represents a segment can be calculated by combining Equation 4.31 to calculate the injection rate of each column and Equation 4.33 to calculate the injection duration for each column:

column Vi = Q ti (4.34) i ·

111 4. Colloid Transport in a Radial Flow Field

where Vi is the suspension volume to be injected into column i. Either the initial concentration of the first column or the effluent concentration of the previous column should be used as the input concentration. This then replicates the concentration reduction over a given radius and satisfies mass conservation throughout the system.

4.4.3 Methods

The porous medium description and suspension preparation of Section 3.6.1 are applicable for these experiments as well. The column experiment set up was identical to set up described in Section 3.6.1, except for the lengths of the columns which were changed to fit the length of each of the discretized sections (Figure 4.19b). The first column of a set was injected for the full duration of a full radial injection, each subsequent column injection was shorter according to Equation 4.33. To satisfy the mass balance, the correct suspension concentration must be injected into each column in the set, the volume necessary for each subsequent column is smaller though. When thinking about the discretization, the total mass of nZVI in the effluent of each segment goes into the next segment. Therefore, the effluent of a column was collected and well mixed, to get a homogeneous suspension of which the necessary volume for the next column in the set was taken. Liquid samples were taken for chemical analysis prior to injection at the inlet of the first column and after each column injection from the homogenized effluent. The radial flow container experiment is a 60° wedge of a full radial flow system, and

0.007 Darcy velocity Average Darcy velocities 0.006

0.005

0.004

0.003

0.002 Darcy velocity [m/s]

0.001

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 Distance from the injectionwell [m] (a) (b) Figure 4.19: (a)Darcy velocity in a radial flow field and the average Darcy velocities in a discretization into 3 sections. (b)Three columns of different length

112 4.4. Discretized Radial Flow Field Reproduced with Columns

1 therefore the injected flow rate is only 6 th of a full size injection. When using the presented equations to replicate the radial flow experiment, the full size flow rate Q, must be used for calculation of the proper Darcy velocity instead of the experiment size flow rate Qexp, thus: 6 Qexp = Q. · While packing the column experiments and container experiment with exactly the same sand, the two different packing methods result in differences in porosity. The packing method for the columns has provided very consistent porosities in the columns of 0.35, the porosity in the radial flow container experiment is less consistent though with 0.40 on average a bit higher. This means that with equal Darcy velocities, the seepage velocity will be different in both experimental set ups. Because the filtration theory showed that the transport of nZVI depends on the seepage velocity (Section 3.2.3), the Darcy velocities in the column experiments must be calculated such that the seepage velocities are the same as those of the container experiment. The average Darcy velocity within one segment is calculated according to Equation 4.28, next this value is divided by the porosity of the container and multiplied by the porosity of the column (Equation 4.35), the rest of the equations use the new adjusted Darcy velocity.

qcontainer qcolumn = ncolumn (4.35) ncontainer

The metal detector, as described in Section 2.3, was used to record the concentration proﬁles of each of the columns in the set.

4.4.4 Conditions

In preliminary tests it was determined that a discretization of three columns provided the most accurate representation. Experiments with more columns (up to seven in one set was tested) introduce too many boundary eﬀect at the column in- and outlet, and the experimental duration becomes too long. Because each column needs to be injected for the same duration, the experimental duration becomes as many times the full scale duration, and it can no longer be assumed that the suspension is fresh in the last columns. Using only one column was also tested to represent the whole full scale system, this showed to be non mass conservative and provided a high overestimation of the transport distances. Therefore the comparison of the discretized column experiments were performed using sets of three columns. The boundary and initial conditions of radial ﬂow container experiments 1 and 3 were used as previously described in Section 4.3, with all the conditions in Table 4.2. The

113 4. Colloid Transport in a Radial Flow Field parameters of each of the column sets to represent the container experiments are given in Tables 4.3 and 4.4.

Table 4.3: Parameters for column set 1 (number of discretization segments: N = 3) representing container experiment 1 with Qexp = 1000 l/h

Column number 1a 1b 1c

Start radius [m] r0 = 0.065 r1 = 0.196 r2 = 0.591

End radius [m] r1 = 0.196 r2 = 0.591 r3 = 1.780 Column length [m] 0.131 0.395 1.189 −3 −3 −3 Average seepage velocity [m/s] v˜1 = 9.31 · 10 v˜2 = 3.09 · 10 v˜3 = 1.03 · 10 column −3 column −3 column −4 Average Darcy velocity column [m/s]q ˜1 = 3.26 · 10 q˜2 = 1.08 · 10 q˜3 = 3.59 · 10 container −3 container −4 container −4 Average Darcy velocity container [m/s]q ˜1 = 3.73 · 10 q˜2 1.24 · 10 q˜3 4.10 · 10 column column column Column ﬂow rate [ml/min] Q1 = 227.5 Q2 = 75.5 Q3 = 25.0

Injection time [min] t1 = 56.4 t2 = 56.1 t3 = 54.0

Injection volume [l] V1 = 11.22 V2 = 3.71 V3 = 1.18 Pore volume of column [l] 0.05 0.14 0.42 Number of pore volumes injected [−] 240.59 26.37 2.79 Porosity of column [−] 0.35 0.35 0.35

Table 4.4: Parameters for column set 3 (number of discretization segments: N = 3) representing container experiment 3 with Qexp = 500 l/h

Column number 3a 3b 3c

Start radius [m] r0 = 0.065 r1 = 0.196 r2 = 0.591

End radius [m] r1 = 0.196 r2 = 0.591 r3 = 1.780 Column length [m] 0.131 0.395 1.189 −3 −3 −4 Average seepage velocity [m/s] v˜1 = 4.66 · 10 v˜2 = 1.55 · 10 v˜3 = 5.13 · 10 column −3 column −4 column −4 Average Darcy velocity column [m/s]q ˜1 = 1.63 · 10 q˜2 = 5.41 · 10 q˜3 = 1.79 · 10 container −3 container −4 container −4 Average Darcy velocity container [m/s]q ˜1 = 1.86 · 10 q˜2 = 6.18 · 10 q˜3 = 2.05 · 10 column column column Column ﬂow rate [ml/min] Q1 = 99.55 Q2 = 33.03 Q3 = 10.96

Injection time [min] t1 = 112.7 t2 = 112.3 t3 = 108.0

Injection volume [l] V1 = 11.22 V2 = 3.71 V3 = 1.18 Pore volume of column [l] 0.05 0.14 0.42 Number of pore volumes injected [−] 240.59 26.37 2.79 Porosity of column [−] 0.35 0.35 0.35

114 4.4. Discretized Radial Flow Field Reproduced with Columns

4.4.5 Results & Discussion

The results of both the radial flow experiments and the columns experiments are displayed in Figures 4.20 and 4.21. The values from the container experiments are the average values of the measurements taken at each radius analyzed using the hydrogen production method (Section 2.6). The error bars represent the minimum and maximum values measured at each point. The inflow concentrations of the suspensions injected into the container and the first column of each set are also provided in these plots, because the inflow concentrations were not always successfully matched (Table 4.5). Mainly this was related to the concentration of the base suspension, which was subject to heterogeneities and uneven distribution during storage. The measured iron content (Stotal) was nevertheless chosen to be shown instead of normalized values because normalized plots would not show the predictive capacity of the method. The iron concentration profiles were reasonably close to those of the container, although not exact or always fully within the standard deviation of the measured values. The most likely cause for these discrepancies would be the differences in the input concentrations. This effect is clearly observed when comparing the results of the last columns from both sets. In column set 1 the input concentration was higher than in the container, as was the profile of the last column, whereas in column set 3 the input concentration was lower, and lower values were observed.

Table 4.5: Chemically measured input and eﬄuent concentrations of the container and column experiments

Experiment Cinput [g/l] Ceffluent [g/l] Container 1 8.3 N.N. Column 1a 10.12 9.97 Column 1b 9.97 8.83 Column 1c 8.83 0.21

Container 3 10.5 N.N. Column 3a 8.00 5.69 Column 3b 5.69 2.71 Column 3c 2.71 0.20

115 4. Colloid Transport in a Radial Flow Field

30 3 Column Experiment (Cinput = 10.1 g/l) Container Exp. 1 (Cinput = 8.3 g/l) 25

[g/kg] 15 total S 10

0 0 200 400 600 800 1000 1200 1400 1600 1800 2000 Distance from the injection well [mm] Figure 4.20: nZVI concentration profiles from column set 1 compared to measurements from radial flow experiment 1. Both experiments simulated a field injection rate of 6000 l/h using RNIP nZVI suspensions into a confined aquifer. From left to right: column 1a, 1b and 1c

45 3 Column Experiment (Cinput = 8.0 g/l) 40 Container Exp. 3 (Cinput = 10.5 g/l)

25 [g/kg] 20 total S 15

0 0 200 400 600 800 1000 1200 1400 1600 1800 2000 Distance from the injection well [mm] Figure 4.21: nZVI concentration profiles from column set 3 compared to measurements from radial flow experiment 3. Both experiments simulated a field injection rate of 3000 l/h using RNIP nZVI suspensions into a confined aquifer. From left to right: column 3a, 3b and 3c

116 4.4. Discretized Radial Flow Field Reproduced with Columns

Conservation of Mass

To prove that the discretization method is valid, it is necessary to show that the system is mass conservative and the storage and transport value represent accurately the ratios in an actual radial flow system. The total mass in an actual radial system was calculated two different ways: using the metal detector measurements and using the chemical data. First the mass in each column was calculated using the volumes and concentrations of the in- and outflow. This total mass was then divided by the dry mass of soil within the column to obtain a ratio in grams of iron to kilograms of soil. This ratio was then multiplied by the mass of soil in that section of a 0.6 m thick radial system. This yields the total iron storage within a full radial system. Volume and concentration of the outflow were then used to calculate the total mass of iron to leave the system. The storage and outflow mass were then summed and compared to the injected mass, or the injected volume of 6000 l multiplied by the measured input concentration. The same method was used a second time, but the iron to dry soil ratio was calculated from the average metal detector reading in that section. It should be noted that this calculation is still partially dependent on the chemical analysis because it uses the measured input concentration and the effluent concentration to close the system. input The calculation of Mtotal iron was performed by multiplying the input concentration by the theoretical full scale injection volume minus the mass lost in the effluent.

input full scale M = (Cinput Ceffluent)Q t (4.36) total iron −

stored where The calculations of Mtotal iron were performed by multiplying the total measured iron content with respect to the column volume by the total volume in a full radial system.

P π 2 ! Stotalρb d ∆x stored 2 2 X bcolumn,i 4 column Mtotal iron = πb rN r0 π 2 (4.37) − d Lcolumn,i 1→N 4 column where N is the number of columns, dcolumn the diameter of the column and Lcolumn,i is the length of column i, bcolumn,i is the total number of data points of the concentration profile along column i, and ∆x the distance between two data points. The coefficient to convert the metal detector readings to iron content was the same for all experiments, therefore the measured iron content is not dependent on the measured liquid concentrations and these two values can be compared. The results of the mass balance calculations confirm that the iron content profiles

117 4. Colloid Transport in a Radial Flow Field satisfy mass balance when converted to a full radial system (Table 4.6).

Table 4.6: Mass balance calculations compared the calculated mass storage in a full radial system from both the liquid concentrations of the in and outﬂow, and the metal detector readings

input stored Column Set Mtotal iron [g] Mtotal iron [g] Percent Error

from inﬂow - outﬂow from Stotal 1 53526 53606 0.10% 3 42186 41838 -0.80%

Errors in the Concentration Proﬁles

Measurements of the iron concentration profiles with the metal detector result in some errors, especially associated with the boundaries of columns, which for short columns are much more prominent and can unfortunately dominate the measurement. The most prominent error is (I) the peak in nZVI concentration in the direct vicinity of the inlet, as opposed to directly at the inlet, where the peak concentration would be expected. As the metal detector moves from the filter to the porous media, it detects nZVI only on one side, therefore the reading will be lower than when it is one or two centimeters farther and is completely surrounded by sand that contains iron. This is most prominent with very short columns where the first two centimeters accord to a significant portion of the column, in columns of one meter or more this effect is negligible. So far no reliable methods were found to automatically and scientifically acceptable convert the recordings in the vicinity of the inlet and outlet. Therefore it was chosen to keep the data as it was recorded. For long columns the induced error is negligible, only with very short column this measuring artefact can induce a significant error. For short columns it could be discussed if manual adjustment of the recordings as shown in Figure 4.22 is justifiable. The metal detector reading was (II) transformed using data from the chemical anal- yses for each experiment, errors in the chemical analysis thus directly transfer into the concentration profiles and mass balances. The transformation factor was calculated from the input and effluent concentration of the second column of each set and this factor was then applied to all columns of the set, which for these two experiments was identical. The chemical analysis using the photometer is highly accurate, though, the method as a whole does have some error associated with it. The photometer itself (given it is correctly calibrated) has negligible error, but the method has some error associated with

118 4.5. Numerical Simulations

3 Column Experiment (C = 8.0 g/l) 60 input Adjusted 3 Column Experiment (Cinput = 8.0 g/l) Container Exp. 3 (Cinput = 10.5 g/l) 50

40 [g/kg] 30 total S 20

0 0 50 100 150 200 250 Distance from the injection well [mm] Figure 4.22: Manual adjustment of the recording at the vicinity of the inlet and outlet of the first column of set 3. Both profiles represent the same total mass it because very small volumes from the original sample (100µl) can only be tested in the photometer. The difficulty of this method thus lays in taking a fully representable sample. This error is even more prevalent in samples with higher concentrations, such as the samples from the base solution, since they have to be diluted in several steps, each step with an error associated due to the difficulty of taking a representative sample. For samples with lower concentrations, such as those of the effluent, the method would thus be best suited.

4.5 Numerical Simulations

4.5.1 Motivation

The third method is a combination of numerical simulations and column experiments to determine the transport distance and concentration profile in a radial flow field. The transfer of the numerical model with fitting parameters from the column experiments of Section 3.7 into a radial flow system was therefore explored [de Boer et al., 2008; Estrella, 2011b]. The basis of the new injection model for a radial flow field was the transport description in 1-D as described in Section 3.5, combined with the fitted parameters and relationships obtained from fitting the 1-D finite difference model (implemented in

119 4. Colloid Transport in a Radial Flow Field

Matlab) on the concentration profiles of column experiments. The Matlab model used three fitting parameters to produce the transient nZVI concentration profiles along the two meter long columns.

4.5.2 Solver

In order to broaden the application and robustness of the mathematical model and to obtain more stable solutions in a more complex geometry the 1-D finite elements Matlab model used for the calibration of column experiments was transferred to a solver which can deal with 2-D and 3-D geometries. The solver also uses the finite element method and can deal very well with curves and grid refinements in the more critical areas and therefore provides stable solutions in a radial flow field. Furthermore the transfer into this solver provides to possibility to run the numerical model on much more complex geometries or hydraulic conductivity fields. For the implementation of this mathematical model, FEniCS was selected. The FEniCS Project focusses on the solution of differential equations by the finite element methods [Logg et al., 2011].

Implementation in FEniCS

In general, implementation of a mathematical problem in FEniCS can be performed by the following steps [Logg et al., 2011]: (I) Define the partial differential equation (PDE) and the appropriate boundary conditions. (II) Reformulate the PDE as a variational problem. (III) Define all initial and boundary conditions, and a mesh for the domain Ω. (IV) Add any necessary statements for the solution of the variational problem and calculate all necessary information for presentation. (V) Graphically represent the results using ParaView and Gnuplot.

Deﬁnition of the PDE and Boundary Conditions

There are two PDEs that must be solved for this problem: (I) the flow problem with the appropriate boundary conditions, and (II) the transport problem using the velocity calculated from the flow problem.(I) The flow problem is defined by equation 4.1. Two boundary conditions were implemented for the flow problem. The first is a Neumann, constant flux, boundary condition applied at the injection well using the Darcy velocity at the well screen (6.5 cm from the center of the well) corresponding to the given flow

120 4.5. Numerical Simulations rate. A Dirichlet, constant head, boundary condition was applied to the outer edge of the domain at a radial distance of 1.78 m from the well screen. (II) Next the PDE for nZVI transport was defined after Equation 4.8. The only difference is that in this case the seepage velocity v, is calculated from the gradient of h, the solution of the flow problem.

∂C K K = C h h C + (D C) ∂t − ∇ · n ∇ − n ∇ ∇ ∇ · ∇ − K ! 3(1 n)αη0 n h S − ∇ 1 0 C 2dc − Smax + Sr − ! 3(1 n)αaggηagg K h − 0 n ∇ C (4.38) 2dc

The boundary condition at the injection well screen applied to this problem was a Dirichlet, constant concentration:

C(t) r = Cinput (4.39) | 0 The boundary condition at the outer boundary was Neumann, zero normal derivative of the concentration:

C n = 0 (4.40) ∇ · where n is a normal positive outward unit vector.

Reformulation of the PDEs as Variational Problems

The general method for reformulating a PDE into a variational problem consists of the following steps [Donea and Huerta, 2003; Logg et al., 2011]: (I) Multiply the whole PDE by the test function, here denoted by ω. (II) Integrate the PDE over the domain Ω (III) Perform integration by parts on terms with second order derivatives The new condition defined for the variational formulation was the Neumann boundary condition of prescribed flux fn, which is defined as a piecewise function on the domain boundary (fn = fn(r, θ, z)) which was integrated over the surface of the boundary (Sb). The variational formulations are given in equations 4.41 and 4.42. Equation 4.41 was

121 4. Colloid Transport in a Radial Flow Field solved as a steady state problem, so there is no time dependency. Z Z ωK h dΩ = ωfn dSb (4.41) Ω ∇ Sb

Equation 4.42 is time dependent, therefore the solution of C is deﬁned for both the current and previous time steps as Ct and Ct−1 ! Z Z Z K ωCt−1dΩ = ωCt dΩ + ω h Ct dΩ Ω Ω Ω n ∇ · ∇ Z ω (D Ct) dΩ − Ω ∇ · ∇ Z K t−1 ! 3(1 n)αη0 n h S t + ω − ∇ 1 0 t−1 C dΩ Ω 2dc − Smax + Sr Z 3(1 n)αaggηagg K h + ω − 0 n ∇ Ct dΩ (4.42) Ω 2dc

Additional Equations and Statements

Several additional calculations are made within the model, both for the solving of the PDE, as well as presentation of the results. At each time step and location within the domain, the values of S and Sr are calculated based on the removal of nZVI during the previous time step. Sr is now calculated following:

n agg t−1 t−1 Sr = katt C dt + Sr (4.43) ρb

−3 agg where ρb is the bulk density of the soil [ML ], and katt is the removal coeﬃcients of the aggregates [ ] as presented in Equation 3.15. The calculation of S is calculated similarly − such that the removal of iron due to the constant removal and colloid terms are combined, yielding equation 4.44.

S = Sc + Sr (4.44) where

! t−1 t−1 ! t−1 n 3(1 n)αη0v Sc + Sr t−1 Sc = Sc + − 1 0 t−1 C dt (4.45) ρb 2dc − Smax + Sr

122 4.5. Numerical Simulations

0 These equations are formulated in this manner because Smax corresponds to the maximum possible removal due to the colloid removal term acting on the colloids. A conditional 0 statement was then implemented to limit the value of Sc at Smax. For the presentation of the ﬁnal iron content proﬁles, C was converted into giron kgdrysoil and added to S. This was necessary because the results of the column and container experiments, the total iron, including that of the liquid phase, was measured. Therefore,

Stotal is calculated as shown in Equation 4.46. ! n Stotal = C + S 1000 (4.46) ρb where the factor 1000 accounts for the conversion form kg to g.

Deﬁnition of the Mesh

The mesh (grid) was deﬁned according to the dimensions of the porous medium in the container experiment. Divisions were made in the radial and angular directions. An example of the layout of the mesh is displayed in Figure 4.23, this mesh shows 25 divisions with respect to radius and 10 with respect to angle. All simulations were performed on a mesh of 500 divisions in the radial direction and 50 in the angular.

Figure 4.23: Example mesh with 25 divisions in the r direction and 10 in the θ direction, the calculations were performed at a much ﬁner discretization (500 in r, 50 in θ)

123 4. Colloid Transport in a Radial Flow Field

4.5.3 FEniCS vs. Matlab

Before the FEniCS model was applied to radial ﬂow around an injection well, simulations were run with a constant velocity in 1-D in order to compare, while using the same input parameters, the results from the FEniCS model to those of the Matlab model. Using the parameters in Table 4.7, simulations were performed with both models and the results are plotted against each other in Figure 4.24. This proves that the solution in both models are the same within reasonable error, while using very diﬀerent numerical solvers.

Table 4.7: Parameters used for the comparison simulation of the models in Matlab and FEniCS

Parameter Value Input concentration [g/l] 11.25 Column length [mm] 2000 Grid size [mm] 1 Porosity [−] 0.34 Seepage Velocity [m/s] 1.43 · 10−3 0 −2 Smax [g/kg] 1.06 · 10 α [−] 8.63 · 10−1 αagg [−] 1.8 · 10−2 ∆t [s] 30

30 FEniCs Simulation, t=3000s, t=6000s, t=9000s Matlab Simulation, t=3000s, t=6000s, t=9000s 25 20.1

20 20

19.9 [g/kg] 15 800 805 810 total S 10

0 0 200 400 600 800 1000 1200 1400 1600 1800 2000 Distance from the injection well [mm] Figure 4.24: Comparison between three transient nZVI concentration proﬁles of the Matlab and FEniCS models, with an inset zooming in on the results

124 4.5. Numerical Simulations

4.5.4 Results

Using the FEniCS model, forward simulations were run with the initial parameters and boundary conditions of container experiments 1, 3 and 4. Container experiment 2 was not simulated because that was performed with a pulsating ﬂux pump.

Container Experiment 1

The input parameters for experiment 1 can be found in Table 4.2. These parameters were then set in the FEniCS model in order to compare the forward simulation results to those of the container experiment. As previously discussed, the FEniCS model uses both a flow and transport equation. The flow equation was solved for head distribution, then the vector field of the gradient of h was used in the advective term of the transport equation. This head distribution can be found in Figure 4.25. The result of the final nZVI distribution can be found in Figure 4.26.

In order to compare the results shown in Figure 4.26 to the container experiment, Stotal was plotted with respect to radial distance from the center of the well (Figure 4.27). The results with respect to radial distance are plotted with the average, minimum, and maximum values of measured nZVI concentrations from container experiment 1. Based on the comparison in figure 4.27, the FEniCS simulation fits very well within the measurements. The largest residuals seem to be between the values at 800 mm, but the simulation results still lie well within the measured values. It is also to important to note that the final transport distance seems to be very close to the observed values.

Figure 4.25: Head distribution from FEniCS simulation of container experiment 1

125 4. Colloid Transport in a Radial Flow Field

Figure 4.26: nZVI distribution from FEniCS simulation of container experiment 1

30 FEniCs Simulation Container Experiment 25

[g/kg] 15 total S 10

0 0 200 400 600 800 1000 1200 1400 1600 1800 2000 Distance from the injection well [mm] Figure 4.27: FEniCS simulation and results from container experiment 1

126 4.5. Numerical Simulations

Container Experiment 3

The input parameters for experiment 3 can be found in Table 4.2. These parameters were then set in the FEniCS model in order to compare the forward simulation results to those of the container experiment. The head distribution calculated by FEniCS can be found in Figure 4.28. The result of the ﬁnal nZVI distribution can be found in Figure 4.29.

In order to compare the results shown in Figure 4.29, Stotal was plotted with respect to radial distance from the center of the well. In Figure 4.30, these results with respect to radial distance are plotted with the average, minimum, and maximum values of measured nZVI concentrations from container experiment 3. The FEniCS simulation ﬁts the observed data for experiment 3 very well The largest discrepancy is located at 400 mm. The results from the other measuring points in comparison to the model are very close to each other. The ﬁnal transport distance calculated appears to be a bit shorter since around 1200 mm the measurements from the container still showed to contain iron where the model already is almost at zero. A slight underes- timation of the front progression thus occurred in this simulation.

Figure 4.28: Head distribution from FEniCS simulation of container experiment 3

127 4. Colloid Transport in a Radial Flow Field

Figure 4.29: nZVI distribution from FEniCS simulation of container experiment 3

35 FEniCs Simulation Container Experiment 30

20 [g/kg]

total 15 S

0 0 200 400 600 800 1000 1200 1400 1600 1800 2000 Distance from the injection well [mm] Figure 4.30: Fenics simulation and results from container experiment 3

128 4.5. Numerical Simulations

Figure 4.31: nZVI distribution from FEniCS simulation of container experiment 4

30 FEniCs Simulation Container Experiment 25

[g/kg] 15 total S 10

0 0 200 400 600 800 1000 1200 1400 1600 1800 2000 Distance from the injection well [mm] Figure 4.32: FEniCS simulation and results from container experiment 4

129 4. Colloid Transport in a Radial Flow Field

Container Experiment 4

The input parameters for experiment 4 can again be found in Table 4.2. The head distribution calculated by FEniCS was identical to the head distribution of container experiment 1, and can thus be found in Figure 4.25. The result of the ﬁnal nZVI distribution can be found in Figure 4.31. The results from the simulation of experiment 4 are very close to the observed values. The largest residuals are located at around 200 mm and 400 mm, but they are still within the error limits of the chemical measurements. At all other points of the simulation data not only ﬁt within the maximum and minimum values, but intersect the average measurement values.

4.5.5 Discussion

The implementation of the governing nZVI transport equation into FEniCS was successful, and the comparison of the results to the calculations from the Matlab model clearly demonstrated that the solution calculated by both solvers was identical. The subsequent transfer of the FEniCS model to two dimensions assumes that the calibrated parameters hold true as the flow properties change. This assumption is also supported by the lack of any clear correlations between the fitting parameters and seepage velocity. This indicated that changes in velocity are accurately accounted for by the function for collector efficiency, η0 and ηagg. Based on the good correlations between the forward FEniCS simulations and the measured data from the container experiments, it can be stated that these calibrated parameters are indeed transferable to different conditions. Another important point of discussion is the further application of these calibration parameters to other materials. These calibrations were performed only with one combination of sand and iron particles. Some of the soil and iron suspension properties, such as particle size, density, and viscosity, are accounted for in the equation for single particle collector efficiency, η0 and ηagg, but the effects of these changes on the calibration parameters is not known. Future research with different nZVI suspensions and soils would have to determine the dependence of these fitting parameters on the nZVI-soil combinations. This does not mean that this model cannot be used for other materials; it means only that these parameters may need to be derived from a set of column experiments, where parameter correlations can be investigated. Information for a specific soil-nZVI combination can be obtained by 3 6 two meter column experiments with concentrations ranging − from 1 30 g/l. −

130 4.6. Comparison of Radial Flow Simulations Methods

4.6 Comparison of Radial Flow Simulations Methods

All three techniques presented in this chapter have their pros and cons. The ﬁnal transport distance is one of the most important pieces of information because, in application of nZVI as a remediation technique, it determines the required distance to the next well while maintaining a minimum nZVI concentration in the soil. The apparent accuracy of the transport distance of all methods is promising for the application of such techniques for a ﬁeld application design.

Radial Flow Field Container

The simulation of a full scale radial system using the large scale container experiments is an exclusive and useful method. Real field scale techniques like the pumps and volume of suspension used can be tested with this method, which is otherwise only possible in a field pilot test. The big advantage of the container experiment is in the possibility to emplace sensors for direct measurement of different parameters. The sensors developed in this work could be used to get In-Situ live break trough curves of nZVI. Sensors for tracer detection could aid in characterizing the hydraulic conductivity of the system and pressure sensors could be used to show whether clogging occurs at specific locations. Sample points could be installed to measure nZVI concentrations or colloid and aggregate sizes in liquid samples. Furthermore, the container experiments showed to be a perfect tool to provide near field scale data to compare the other simulation techniques to.

Sets of Columns Representing Radial Flow

The experimental approximation method for injected nZVI transport showed adequate accuracy in comparison to the measured data from the radial flow container experiments, but it does not quantify transport properties. Mass balance conditions for experiments of three columns were satisfied within reasonable error in all cases. Simulations using column experiments showed to be difficult because the aimed input concentration was never without error. Comparison of the results from columns between each other is difficult due to the differences in input concentration, but also the prediction of a full radial system based on column experiments is only valid when the same input concentration in the full system is matched. Comparison of column experiments between each other with differences in other conditions like the injection rate is furthermore difficult because in most cases not only the injection rate will be changed but also the input concentration.

131 4. Colloid Transport in a Radial Flow Field

Nevertheless, the discretization using column experiments to represent the radial flow field can be a useful and quick tool to do qualitative screening of differences in several conditions (e.g. porous medium material, suspension chemistry, nZVI pretreatment or ZVI type), given that more care is taken into getting the aimed input concentration to make the experiments comparable. This could be done by chemically analyzing the suspension before the injection is started, the chemical analysis is not very quick so the suspension would age already a bit before the experiment is started. When nZVI would be dried (e.g. through sublimation) and exact masses could be weighted under an anaerobic environment (glove box) or when µZVI would be used which is air stable, the exact aimed input concentration could be accurately set. Furthermore, this technique can be used to test the transport behavior of other nano materials in a radial flow field. The metal detector might then not be usable (e.g. when not using ferro-magnetic materials), but still the outflow concentrations of each of the columns in a set will be able to show the transport distance.

Numerical Simulations of Transport in Radial Flow Fields

Numerical aided predictions proved to work best when compared to the observed experimental data. Though, it is the most elaborate method and knowledge of both the experimental work as well as numerics is needed. The big advantage is that there is no problem with the input concentration and discharge in the column experiments performed to find the fitting parameters. A range of different input concentrations are needed, but the exact concentration of the individual column experiments is of no importance. Fur- thermore the boundary effects that occur with the short columns are insignificant with the two meter long columns used for the model fitting. Given the relationships determined with the column experiments many different injection scenarios could be simulated. This method could further reduce the amount of experimental work when investigating injection parameters, and broaden the application of these parameters to a wider range of systems and boundary conditions. The total amount of resources used for injection optimization would be greatly reduced.

132 5 Feasibility of Injecting nZVI into the Subsurface

5.1 Motivation

The results presented in the previous chapters provide a better scientific understanding of the transport of nano sized zero valent iron (nZVI) during the injection into the subsurface. It was shown that the transport of nZVI is limited by filtration of primary and aggregated particles, therefore the progression of the nZVI front is much slower than the injection front. The study is not just theoretical but the developed relationships and models can also be used in order to understand the fate of nZVI in the field. It was shown that given the right conditions, nZVI can be transported in a radial flow field over a distance of almost two meters. Long term measurements showed that after five years in storage the material still contains the reactive zero valent iron (ZVI), though significantly less than in the beginning. In the field, dilution and continuous contact with fresh water will further reduce the longevity. The main impact will be the methods developed in this work which can be used for testing the feasibility and expected success of using nZVI for the remediation of contaminated aquifers. In the following the assumption was again made that transport of nZVI has to be due to permeation and the carrier liquid is a near newtonian fluid. Other techniques like fracturing the subsurface, addition of chemical colloid stabilizing compounds or using non- newtonian carrier fluids were not taken into account in this work, and thus no predictions can be made for those techniques.

5.2 Prerequisites for Success

Before starting to use nZVI for the remediation of a contaminated aquifer, it should be know if the application is likely to lead to a successful remediation. In order to decide if

133 5. Feasibility of Injecting nZVI into the Subsurface the remediation using nZVI will be successful, several questions will have to be answered first. The questions of main importance are listed and discussed below. (a) At which depth is the source located, and is it located in the saturated or unsaturated zone? (b) Is the source of contamination concentrated in one point, is the source smeared over a large area and/or depth or is generally little known about the exact location, shape and dimension of the source? (c) Which type of aquifer material is present. Porous or fractured aquifer; consolidated or unconsolidated sediments; coarse or fine material; heterogeneous or homogeneous; lenses of fine or coarse material present? (d) Is nZVI capable of remediating the contaminant under consideration? (e) What is the average hydraulic conductivity of the aquifer and what is the average seepage velocity of the groundwater through the contaminated area (hydraulic head gradient)? (f) What is the aim of remediation and how will this result be tested? (g) Which amount of injection points is possible or affordable? (h) Which concentrations in the subsurface (in suspension and attached to the aquifer

material, i.e. Stotal) are required to emplace the necessary total mass of nZVI. Is this technically possible to reach with a single injection?

Already based on these questions a decision on using nZVI or not can be made. The following discussion points with respect to the open questions will have to be taken into account for that decision. (a) The application of nZVI in the unsaturated zone is not possible, nZVI corrodes rapidly when in contact with free oxygen. The depth will provide an indication of the injection pressures that can be build up before fracturing or daylighting is expected to occur. (b) With a contamination spread of a very large aquifer volume at a fairly low concentration, the use of nZVI is likely to become very costly because many injection points and especially at many levels will be needed, furthermore, the consumption on nZVI due to the reaction with other compounds will be too abundant. (c) If the aquifer material is consolidated, very fine, or the contaminants are mainly located inside fine sand lenses, the injection of nZVI is unlikely to be successful. Injection into consolidated or very fine material will likely results in fracturing and when lenses with fine material are present, nZVI will be transported around these lenses. In these situations, nZVI won’t be able to reach the contaminant source.

134 5.2. Prerequisites for Success

(d) Many contaminants can be remediated by nZVI, though some better than others, preliminary batch experiments might therefore be necessary. (e) With an unconsolidated fairly coarse sand with good hydraulic conductivity (K > 1 10−4 m/s) the injection of nZVI is possible. Lower hydraulic con- ∼ · ductivities could be feasible but experimental proof first has to be obtained. Other stabilizing agents are likely to be necessary for lower hydraulic conductivities, which was not investigated in this study. The average groundwater seepage velocity through the source zone will indicate if the contact time is long enough for the nZVI to fully reduce the contaminant. Also, at too high flow rates the side reactions (e.g. anaerobic corrosion) are likely to become too abundant resulting in significant reduction of the longevity of nZVI. (f) Required fast remediation (e.g. less than two years) in combination with the right subsurface conditions would make nZVI a suitable technique. Caution must be taken when testing the success, the techniques used should suitable for testing the success of the remediation due to nZVI. Consumption of nZVI could for example be monitored by measuring the susceptibility change throughout the whole remediation process. A mass flux calculation based on a large cross section down stream can also be used to show that the contaminant flux reduces, this measuring net should then be set up abundantly long early before the injection of nZVI takes place and operated long enough afterwards to make sure all dilution effects due to the injection of the carrier fluid are gone. The seepage velocity through the source zone is likely to be lower once nZVI has been injected because the hydraulic conductivity is slightly reduced. Therefore, tracer tests before and after the injection are necessary to determine the dilution effect caused by more groundwater flowing around the source zone than through it. (g) The minimum or maximum distance between the injection points is needed, these might be restricted due to on-site building constructions or infrastructure (above- and underground). The necessary transport distance of nZVI will be bounded by the size of the treatment volume and distances between injection wells.

(h) The minimal concentration of nZVI necessary in the subsurface (Stotal minimal) can | be calculated based on stoichiometric mass balance, but an extra amount of nZVI is needed to cover the loss of nZVI due to side reactions. With high groundwater ﬂow rates, side reactions are likely to become more abundant. It might be necessary to perform multiple injections if the concentration can not be reached with a single injection or the longevity of nZVI is not enough for the expected remediation duration.

135 5. Feasibility of Injecting nZVI into the Subsurface

Next it is necessary to determine what is needed to be able to speak about a successful field injection and site remediation. The main issues to be taken care of are: (I) Proof of transport distance. A successful injection can only be proven by using the correct detection method. In the introduction several currently used methods are described, these all have their limits and possible lack of conclusiveness. The electromagnetic measurement of the susceptibility change in the subsurface is the chosen method to provide a conclusive statement about the transport distance in the subsurface. The installation of the detection sensors has to be done at those locations that are thought of to be critical. The amount of injection points determines the minimal transport distance that nZVI should reach. Furthermore, a certain concentration needs to be reached to ensure that the reactivity will not diminish before accomplishing the degradation of the contaminant. Emplacement of the sensors at the center between two injection wells (Figure 5.1) will show whether the nZVI concentration is high enough at these critical locations. When first a pilot trial injection is performed, the amount of sensors around the injection well should be increased, then a better understanding of the flow field and transport distances can be obtained. (II) Proof of removal of contaminants. Before starting sampling, the dilution effect due to the injection of the nZVI carrier fluid has first to be taken into account. Hence, enough time must be allowed to pass before the concentration measurements are taken. (III) Proof of consumption of nZVI due to reaction with contaminants. The electromagnetic sensors will be able to show the reduction of nZVI in the subsurface. This alone is not yet enough to prove that nZVI is reducing the contaminant concentration. Therefore, also other reaction end products will have to be measured, for example Ethane which is an end product of PCE when remediated by nZVI.

Well

Sensor

Figure 5.1: Overview of minimal amount of sensor locations in an injection ﬁeld with multiple injection wells to prove the continuity of nZVI between the injection wells

136 5.3. Determination of Necessary Amount of nZVI

5.3 Determination of Necessary Amount of nZVI

Given that the previous questions lead to the choice of using nZVI for the remediation, a more precise prediction of the feasibility can be made. First the chemical applicability of the technique has to be veriﬁed. In order to decide if the contaminants can be targeted with the speciﬁc nZVI available, batch experiments need to be performed. Reactivity testing methods were not described in this work, but can be found in literature (among many others e.g.: Lien and Zhang [2001]; Liu and Lowry [2006]; Steiert [2008]; Zhang et al. [2009]). Furthermore the necessary mass of ZVI should be determined based on stoichiometry and results from the reactivity tests. The use of site own aquifer material and ground water compared to tests using clean quartz sand and demineralized water can help to determine the amount of ZVI loss due to side reactions. For example, based on Equation 5.1 [Lien and Zhang, 2001], where the transfer from PCE into Ethane is described, the necessary stoichiometric mass of zero valent iron can be determined. According to Equation 5.1, and the molecular masses, the mass ratio is approximately 1:1, so for each unit mass of PCE one unit mass of ZVI is needed.

0 + + − C2Cl4 + 5F e + 6H C2H6 + 5F e + 4Cl (5.1) → Adding to this the fraction of ZVI lost due to the side reactions determined from the batch (or column) experiments, a minimal mass ratio of ZVI can be calculated.

Mass of ZVI Stoichiometric mass of ZVI Mass of ZVI for side reactions = + (5.2) Unit mass PCE Unit mass PCE Unit mass PCE Using the nZVI concentration storage equation (Equation 3.7), the mass of ZVI can be transfered into the actual mass of nZVI. Mass of nZVI Mass of ZVI 1 = (5.3) Unit mass PCE Unit mass PCE × CF e0 (t) recall from Equation 3.7 that CF e0 (t) is the mass ratio of ZVI in nZVI at age t. Given the contaminated volume of aquifer, and the total mass of contaminants inside this volume, the minimal nZVI concentration at the porous medium (Stotal) can be calculated. Because the actual total mass and exact distribution of contaminants in the subsurface is never completely known, a two-fold safety factor for the total mass should at least be taken

137 5. Feasibility of Injecting nZVI into the Subsurface into account.

Mass of nZVI Unit mass PCE Total mass of PCE Stotal minimal = 2 × (5.4) | × ρbulk Volume of contaminated aquifer × 5.4 Application of Feasibility Test Methods

The next step is to determine how to reach this minimal concentration of nZVI

(Stotal minimal) best over the whole source zone. Since nZVI does not move with the | same velocity as the injected carrier fluid, a gradient of nZVI concentration at the porous medium (Stotal) will develop around the injection well. The aimed concentration though has to be reached over a certain distance (i.e. half of the aimed distance between injection wells, raimed). The total nZVI necessary will thus be higher due to the higher concentrations close to the injection well. The final Stotal distribution shape can result in different amounts of excess nZVI, compare the Stotal distributions in the graphs in Figure 5.2. Excesses close to the well are less significant when compared to excesses at a large radial distance, since the latter affects a much larger aquifer volume.

The minimum of excess nZVI necessary to reach Stotal minimal at the aimed radial dis- | tance thus has to be found. Using one of the methods described in the previous chapters the input concentration and injection rate can be determined to reach this. Based on the time and funding available but also the complexity of the subsurface conditions, one or more of the presented methods can be chosen. The first and simplest method is the prediction based on (I) the discretization of the full scale flow field using sets of column experiments (Section 4.4). This method is ideal for fairly simple aquifer conditions (i.e. there should be no large heterogeneities), the experiments could be performed with real aquifer material obtained from core samples. Not much material is needed, one liter of aquifer material is already enough to perform

Figure 5.2: Three diﬀerently possible Stotal distributions resulting in diﬀerent amounts of excess mass

138 5.4. Application of Feasibility Test Methods one set with three columns. Between three and six experiment sets will be needed to be able to determine the optimal combination of input concentration and injection rate for reaching the aimed concentration of nZVI at a specific distance. With more complex aquifer conditions and the desire to present full complexity predictions of the final nZVI distribution over the source zone, (II) a computer simulation aided prediction is to be chosen (Sections 3.6, 3.7 & 4.5). This method is more elaborate and complex and therefor will need more time, but it provides a better prediction and higher flexibility. The nZVI to be used will have to be characterized to determine the size distribution of the primary colloids and the aggregates, although, when RNIP (Toda Kogyo, Japan) is to be used, the parameters from this research can be used. Also here the use of real aquifer material is possible. Between four and seven horizontal two meter long column experiments are to be performed, each with a different input concentration, varying between 0.1 and 10 g/l. The injection rate should be tried to keep constant between the experiments, but this is not very critical since the model can compensate the velocity effects, given that the exact discharge of each experiment is recorded. The injection rate should be chosen such that the seepage velocity is slower than the forward velocity of the wagon carrying the metal detector (which for this research moved at 10 mm/s). For a good numerical fit, at least three transient concentration profiles should have been recorded before the nZVI front passes the outflow filter of the column. The fitting parameters, obtained from the numerical fitting algorithm on the transient concentration profiles, can be used for the prediction of the nZVI distribution in a full radial flow field. Depending on the amount of data available from the field site, different scenarios or computations with measured hydraulic conductivity fields can be performed. Again the aim is to find the right injection rate and input concentration that results in the least consumption of nZVI while the minimal necessary concentration (Stotal) is reached at the largest possible (or desired) distance from the injection well.

139 5. Feasibility of Injecting nZVI into the Subsurface

Contaminated Site

Don't use nZVI Search for other method Contaminants No to treat are in the saturated zone ? No

nZVI could be used, Yes but more cost eﬃcient Continue ? methods might be avaiable

Yes

Contaminants are located deep (> 10 m), or Unwise to use nZVI Infrastructure obstructs No Lab tests necessary if continuing No excavation method ? Continue ?

Yes Yes

Choose experimental technique

Aquifer material has a high hydraulic conductivity Quick Flexible Real data (K > 1e-04 m/s) ? & Column Computer aided Large scale Contaminants are not located in No discretization simulations with container low conductivity zones of radial ﬂow column data input experiments (e.g. clay lenses) ? ﬁeld

Yes Good Bad Transport Transport

Aquifer material is well sorted ? & nZVI already characterized ? No Charcterization with sedimentation tests and particle size detector: size of aggregates and colloids

Yes

Run horizontal column tests (L > 1 m) Use previous or new column experiment results. at least 5 experiments, C varied: 0.1 - 10 g/l Run 2-D model to determine Cinput based on aimed S_total and associated transport Numerical ﬁtting algorithm on column data distance (well separation)

Calculate volume of suspension and C needed for ﬁeld, Start on-site Field based on extrapolation of column discretization Phase or numerical simulations

Figure 5.3: Decision chart to determine the applicability of nZVI

140 5.5. Application of Methods in the Field

The third experimental method provided in this work is (III) the large scale container experiment, which is likely to be too labor intensive to be used for a feasibility study on itself. Like with the discretized column sets approach, between three and six experiments would be necessary to determine the optimal combination of input concentration and injection rate. This results in something between six and twelve months of non-stop work on the container experiments. Therefore it is here suggested that this method is optionally added to either one of the two other methods, to verify if the found combination of input concentration and injection rate indeed provide the predicted result. This method can furthermore ideally be used to test the pumping technique, different types of sensors, sampling systems, suspension preparation technique, among other techniques that would otherwise need a much more expensive and less controlled field trial. A flow chart was drawn (Figure 5.3) to help guiding the decision making of whether to use nZVI and if so, which feasibility testing methods should be used.

5.5 Application of Methods in the Field

In the presented work also technical problems had to be overcome, of which several resulting techniques are useful and recommended to be used in the field as well. Agglomeration is a not to be overlooked problem of nZVI, (I) dispersing nZVI using mechanically induced high shear forces was found to be an effective method to partly undo the aggregation. In-line dispersers were found to be best suited. The in-line operation avoids the introduction of air (free oxygen leads to corrosion) into the suspension. When performing the small flume experiment at the end of the research (Section 2.5.5) it was chosen to use (II) a funnel shaped nZVI reservoir (Figure 5.4). The outlet at the bottom first went through the dispersing unit and then into a small flow-through reservoir

Ventilation

Floater nZVI Reservoir Pump

Disperser Figure 5.4: Drawing of a funnel shaped nZVI reservoir with the connected disperser and pump

141 5. Feasibility of Injecting nZVI into the Subsurface from which the dosing pump took the suspension, all surplus suspension automatically went back into the top op the funnel shaped reservoir. By using a floater, the well dispersed suspension was injected just under the water table, providing a very good and homogeneous suspension inside the reservoir. It is highly recommended that a similar set up be used for field applications. One of the tests performed in the container experiments was to determine which type of pump performed better for an injection. (III) The continuous flux dosing pump was found to be best suitable. This type of pump can provide a very continuous flux and can build up high pressures while maintaining the same flux. Also for the column experiments a small version of such a dosing pump could be used to maintain a continuous flux, also at higher injection pressures. (IV) The susceptibility based monitoring technique to observe the transport during injection should be installed to make sure that nZVI arrived at the calculated concentration at the desired distance from the injection well. For long term observation of the consumption of nZVI it is advisable to keep the sensors and electronics installed. While installing the electro magnetic sensors, this can easily be (V) combined with the installation of a whole range of other sensors and sampling ports to support the monitoring during and after the injection.

142 Conclusion

Application of nano sized zero valent iron (nZVI) to remediate contaminated aquifers is a promising method. By applying nZVI in the source zone and reduce the concentration of contaminants directly in-situ, the remediation duration and expenses are expected to be significantly reduced. The main goal of this work was to get a better understanding and description of the transport of nZVI during the injection into the subsurface. Chemical feasibility was in the context of this work assumed to be granted and was not further investigated. Four research questions were set up which could all be answered throughout this work. (a) Is it possible to detect In-Situ the concentration of iron and determine the transport distance? The first problem faced was the lack of available measuring techniques to determine non-destructively the concentration of nZVI inside the porous medium. A measuring technique based on the magnetic susceptibility of zero valent iron inside nZVI colloids was developed. The technique was at first worked out for column experiments, making it possible to measure transient concentration profiles along the full length of a column. Next the technique was further developed to make in-situ measurements of nZVI breakthrough possible in a large scale container experiment. This technique was then further upscaled to a field ready technique, which could be installed in boreholes at a pilot test or full scale field application. (b) Which conditions have the largest effect on mobility and how can they be optimized? By looking into a classic filtration theory in detail, it was determined that the colloid size is the main factor affecting the transport of nZVI. Through characterization of the chosen nZVI material in this study (RNIP 10-E, Toda Kogyo, Japan), it was found out that colloidal aggregation is very abundant, which following the filtration theory significantly reduces the transport. In order to undo the aggregation, mechanically induced high shear forces were applied by using an in-line dispersing unit, the aggregation was found not be fully undone, and smaller aggregates maintained. Column experiments with different input suspension concentrations were performed

143 Conclusion

because the colloid filtration theory does not take the suspension concentration into account. It was found out that the suspension concentration was a further factor that significantly influences the transport of nZVI. Based on these finding, the minimal alterations that had to be applied to the sup- plied nZVI suspension were dilution and dispersion. Further methods of optimizing the suspension were not investigated because the aim was to use the commercially available material with as little changes as possible. (c) What is the transport distance in a radial flow field? The transport in a radial flow field was investigated first by performing transport experiments in a large triangular shaped container, representing a wedge out of a full cylinder. nZVI was transported over a distance of almost two meters (distance of one pore volume) after injecting three pore volumes of suspension at 10 g/l, at an injection duration of one hour. A transport distance of two meters is a very reasonable distance for field applications. (d) Which method can test the feasibility of injection into the subsurface? Two methods were developed to predict the transport of nZVI in a radial flow field with a high accuracy. A computer simulation aided prediction based on input from column experiments, and a discretization of the radial flow field into segments that could be represented by a set of column experiments of various length. An extended transport and filtration equation was derived and used in the numerical simulations, both for the column and container experiments. The results of both predictive methods were compared to the results obtained from the container experiment. An overview on how to decide in favor of using nZVI and how the predictive methods can help to design a field application has finally been presented.

The results of this work are very promising and are a sound scientiﬁc basis for further investigations and case studies on nZVI based remediation of contaminated aquifers.

144 Outlook

Already the methods developed here showed to be very useful and the understanding of the transport of nZVI was significantly improved. Nevertheless, more research could be performed to further improve the methods and to make the extrapolation to real field situations solid. The following research items could be taken into account: Further improve the field sensor to make measurements without reference sensor pos- ◦ sible, as was the case with the sensor developed for the container. Record breakthrough curves during the container experiments with the container sen- ◦ sors but using the much more sensitive electronics that were developed for the field sensors. The breakthrough curves could then be used to verify the transient results of the numerical simulations. Transient concentration profiles in the container based on photo series from the con- ◦ tainer experiments calibrated on the final measured nZVI distribution might be possible. This could then also be used to test the transient results of the numerical simulations instead of just the final results. Identify the mechanisms responsible for the fairly low attachment efficiency of the ◦ aggregates when predicted by the filtration theory. Improve the filtration theory, which is based on deep bed filtration, because the flow ◦ direction is not only equal to the vertical gravity force component during an injection in the subsurface. Perform simulations in a radial flow field by using hydraulic conductivity fields obtained ◦ from tracer breakthrough curves of the container experiments. Test the simulation with other hydraulic conductivities, include lenses with a lower ◦ or higher hydraulic conductivity, extend the model to full 3-D, include variation in hydraulic conductivity in the z-direction and simulate multiple injection wells. Run the numerical simulation for different flow fields, e.g. using an injection-extraction ◦ configuration of wells (e.g. 2-spot or 5-spot). Testing of characterization and prediction methods on other nZVI products and other ◦ soil material.

145 Outlook

Use a real ﬁeld site as a case study to determine the input concentration and injection ◦ rate by using the methods developed in this work.

146 Bibliography

Al-Abduwani, F., Bedrikovetsky, P., Farajzadeh, R., van den Broek, W., and Currie, P. (2005). External ﬁlter cake erosion: Mathematical model and experimental study. In SPE 6th European Formation Damage Conference, SPE 94635. Society of Petroleum Engineers Inc.

Auslander, K., Fabris, H.-J., Mock, K.-H., Patwald, W., and Seichter, H. (1991). Elec- tromagnetic shield for an inductive search coil assembly. Patent: US5068612.

Bartzas, G., Komnitsas, K., and Paspaliaris, I. (2006). Laboratory evaluation of F e0 barriers to treat acidic leachates. Minerals Engineering, 19:505–514.

Baumann, T. and Werth, C. (2005). Visualization of colloid transport through heterogeneous porous media using magnetic resonance imaging. Colloids and Surfaces A: Physicochemical and Engineering Aspects, 265(1-3):2–10.

Bear, J. (1972). Dynamics of Fluids in Porous Media. Elsevier (New York).

Bradford, S. and Bettahar, M. (2006). Concentration dependent transport of colloids in saturated porous media. Journa of Contaminant Hydrology, 82(1-2):99–117.

Bradford, S. and Torkzaban, S. (2008). Colloid transport and retention in unsaturated porous media: A review of interface-, collector-, and pore-scale processes and models. Vadose Zone Journal, 7(2):667–681.

Bregar, V. B. and Pavlin, M. (2004). Eﬀective-susceptibility tensor for a composite with ferromagnetic inclusions: Enhancement of eﬀective-media theory and alternative ferromagnetic approach. Journal of Applied Physics, 95(11):6289 –6293.

Brinkman, H. (1947). A calculation of the viscous force exerted by a flowing fluid on a dense swarm of particles. Applied Scientific Research, A1:27.

147 Bibliography

Buchau, A., Rucker, W., de Boer, C., and Klaas, N. (2010). Inductive detection and concentration measurement of nano sized zero valent iron in the subsurface. IET Science, Measurement & Technology, 4(6):289–297.

Cantrell, K. and Kaplan, D. (1997). Zero-valent iron colloid emplacement in sand columns. Journal of Environmental Engineering, May:499–505.

¸Cengel, Y. and Cimbala, J. (2006). Fluid Mechanics, Fundamentals and Applications. Graw-Hill, New York. de Boer, C. V. (2007). Characteristics and mobility of zero-valent nano-iron in porous media. Master’s thesis, Utrecht University. de Boer, C. V., Braun, J., Klaas, N., and Steiert, S. (2008). Injection of nano-scale iron for the in-situ remediation of chlorinated hydrocarbons in soil and groundwater. In Flow & Transport in Permeable Media, Gordon Research Conferences, 13 - 18 July, Magdalan College, Oxford, United Kingdom. de Zwart, B.-R. (2007). Investigation of clogging processes in unconsolidated aquifers near water supply wells. PhD thesis, Delft University of Technology.

Donea, J. and Huerta, A. (2003). Finite Element Methods for Flow Problems. Wiley.

Dunford, D. and Lorentz, S. (1994). Tests of hydraulic and leaching properties of the vadose zone and groundwater. Department of Agricultural and Chemical Engineering, Fort Collins, CO 80523.

Elion, E. and Elion, L. (1933). Eine einfache gasvolumetrische methode. Laboratorium fur¨ G¨arungstechnik und Angewandte Chemie.

Elliott, D. and Zhang, W.-X. (2001). Field assessment of nanoscale bimetallic particles for groundwater treatment. Environmental Science & Technology, 35:4922–4926.

Estrella, D. (2011a). Transport of nano-sized zero-valent iron in porous media: Transfor- mation of 2D container experiments into 1D column experiments. Independent study, Institut fur¨ Wasserbau, Universit¨at Stuttgart.

Estrella, D. (2011b). Experimental and numerical approximation methods for zero-valent iron transport around injection wells. Master’s thesis, Institut fur¨ Wasserbau, Univer- sit¨at Stuttgart.

148 Ferguson, J. F. and Pietari, J. M. H. (2000). Anaerobic transformations and bioremediation of chlorinated solvents. Environmental Pollution, 107(2):209 – 215.

Fetter, C. (1999). Contaminant Hydrogeology, Second Edition. Prentice Hall, NJ.

Getzlaﬀ, M. (2008). Introduction. In Fundamentals of Magnetism, pages 1–6. Springer Berlin Heidelberg.

Gillham, R. and O’Hannesin, F. (1992). Metal-catelysed abiotic degradation of halo- genated organic compounds. In IAH Conference Modern trends in Hydrogeology.

Glazier, R., Venkatakrishnan, R., Gheorghiu, F., Walata, L., Nash, R., and Zhang, W.-X. (2003). Nanotechnology takes root. Civil Engineering, 73(5):64–67.

G¨odeke, S. (2011). Contaminant hydrogeology. Online. http://www.eugris.info.

Hahn, R. (2006). Personal Communication. Landesanstalt fur¨ Umwelt, Messungen und Naturschutz Baden-Wurttemberg¨ (LUBW).

Happel, J. (1958). Viscous ﬂow in multiparticle systems. slow motion of ﬂuids relative to beds of spherical particles. AIChE Journal, 4:197.

Higgs, S. (2009). Mama, Dada, and Nano? Subparticles may be toxic for kids. The Progressive, 74:18–20.

Hong, Y., Honda, R. J., Myung, N. V., and Walker, S. L. (2009). Transport of iron- based nanoparticles: Role of magnetic properties. Environmental Science & Technology, 43(23):8834–8839.

Huang, Y., Zhang, T., Shea, P., and Comfort, S. (2003). Eﬀects of oxide coating and selected cations on nitrate reduction by iron metal. Journal of Environmental Quality, 32:1306–1315.

Iwasaki, T. (1937). Some notes on sand ﬁltration. Journal of the American Water Works Association, 29:1591–1602.

Kamat, P. V., Huehn, R., and Nicolaescu, R. (2002). A sense and shoot approach for photocatalytic degradation of organic contaminants in water. The Journal of Physical Chemistry B, 106(4):788–794.

149 Bibliography

Kanel, S. R., Goswami, R. R., Clement, T. P., Barnett, M. O., and Zhao, D. (2008). Two dimensional transport characteristics of surface stabilized zero-valent iron nanoparticles in porous media. Environmental Science & Technology, 42(3):896–900.

Klaas, N. C., Buchau, A., and de Boer, C. V. (2010a). Induktive in-situ-Messtechnik fur¨ elementare Metalle bei der Sanierung von Grundwasserkontaminationen mit Metallkol- loiden. Patent: DE102009047858A1.

Klaas, N. C., Steiert, S., de Boer, C. V., and Ruck, W. (2010b). Stabilisierung des ph- Werts bei w¨assrigen Suspensionen von Metallpartikeln. Patent: DE102009012834A1.

Koch, D. (2007). Untersuchungen zu den Transporteigenschaften von nullwertigen Fe- Nanopartikeln in S¨aulenversuchen. Master’s thesis, Fachbereich Geowissenschaften, Universit¨at Bremen.

Lecoanet, H., Bottero, J.-Y., and Wiesner, M. (2004). Laboratory assessment of the mobility of nanomaterials in porous media. Environmental Science & Technology, 38:5164– 5169.

Ley, B. (2010). Diameter of a human hair. Online. http://hypertextbook.com/facts/1999/BrianLey.shtml.

Li, H., de Boer, C. V., Buchau, A., Klaas, N. C., Rucker, W. M., and Hermes, H. (2012). Development of an inductive concentration measurement sensor of nano sized zero valent iron. In Conference of 9th International Multi-Conference on Systems, Signals and Devices, SSD 2012.

Lien, H.-L. and Zhang, W.-X. (2001). Nanoscale iron particles for complete reduction of chlorinated ethenes. Colloids and Surfaces, 191:97–105.

Liu, X., O’Carroll, D. M., Petersen, E. J., Huang, Q., and Anderson, C. L. (2009). Mobility of multiwalled carbon nanotubes in porous media. Environmental Science & Technology, 43(21):8153–8158.

Liu, Y. and Lowry, G. V. (2006). Eﬀect of particle age (Fe-0 content) and solution pH on NZVI reactivity: H-2 evolution and TCE dechlorination. Environmental Science & Technology, 40(19):6085–6090.

150 Liu, Y., Majetich, S., Tilton, R., Sholl, D., and Lowry, G. (2005). TCE Dechlorination Rates, Pathways, and Eﬃciency of Nanoscale Iron Particles with Diﬀerent Properties. Environmental Science & Technology, 39:1338–1345.

Logg, A., Mardal, K.-A., and Wells, G. (2011). Automated Solution of Diﬀerential Equa- tions by the Finite Element Method. Free Software Foundation.

Long, W. and Hilpert, M. (2009). A correlation for the collector eﬃciency of Brownian particles in clean-bed ﬁltration in sphere packings by a Lattice-Boltzmann method. Environmental Science & Technology, 43(12):4419–4424.

Mattigod, S. V., Fryxell, G. E., Alford, K., Gilmore, T., Parker, K., Serne, J., and Engelhard, M. (2005). Functionalized TiO2 nanoparticles for use for in situ anion immobilization. Environmental Science & Technology, 39(18):7306–7310.

Muller,¨ C., L¨obel, E., and Rissing, P. (2006a). Sanierung mit Nano-Eisen - Stand der Technik. Altlasten Spektrum, 2:75–83.

Muller,¨ C., Rissing, P., Widmayer, F., and Wischott, M. (2006b). Nano-Eisen Feldversuch: Strategie, Durchfuhrung,¨ Ergebnisse und Auswertung. Altlasten Spektrum, 3:137–147.

Nurmi, J., Tratnyek, P., Srathy, V., Baer, D., Amonette, J., Pecher, K., Wang, C., Linehan, J., Matson, D., Penn, R., and Driessen, M. (2005). Characterization and properties of metallic iron nanoparticles: spectroscopy, electrochemistry, and kinetics. Environmental Science & Technology, 39:1221–1230.

Orlov, A., Jeﬀerson, D. A., Tikhov, M., and Lambert, R. M. (2007). Enhancement of MTBE photocatalytic degradation by modiﬁcation of TiO2 with gold nanoparticles. Catalysis Communications, 8(5):821 – 824.

Payne, F., Quinnan, J., and Potter, S. (2008). Remediation Hydraulics. Boca Raton: CRC Press.

Phenrat, T., Cihan, A., Kim, H.-J., Mital, M., Illangasekare, T., and Lowry, G. V. (2010). Transport and deposition of polymer-modiﬁed Fe0 nanoparticles in 2-D heterogeneous porous media: Eﬀects of particle concentration, Fe0 content, and coatings. Environ- mental Science & Technology, 44(23):9086–9093.

Phenrat, T., Fagerlund, F., Illangasekare, T., Lowry, G. V., and Tilton, R. D. (2011). Polymer-modiﬁed Fe0 nanoparticles target entrapped NAPL in two dimensional porous

151 Bibliography

media: Eﬀect of particle concentration, NAPL saturation, and injection strategy. En- vironmental Science & Technology, 45(14):6102–6109.

Phenrat, T., Liu, Y., Tilton, R., and Lowry, G. (2009). Adsorbed polyelectrolyte coatings decrease Fe0 nanoparticle reactivity with TCE in water: Conceptual model and mechanisms. Environmental Science & Technology, 43(5):1507–1514.

Phenrat, T., Saleh, N., Sirk, K., Kim, H.-J., Tilton, R., and Lowry, G. (2008). Stabi- lization of aqueous nanoscale zerovalent iron dispersions by anionic polyelectrolytes: adsorbed anionic polyelectrolyte layer properties and their eﬀect on aggregation and sedimentation. Journal of Nanoparticle Research, 10(5):795–814.

Phenrat, T., Saleh, N., Sirk, K., Tilton, R., and Lowry, G. (2007). Aggregation and sedimentation of aqueous nanoscale zerovalent iron dispersions. Environmental Science & Technology, 41(1):284–290.

Prieve, D. C. and Ruckenstein, E. (1974). Eﬀect of london forces upon the rate of deposition of brownian particles. AIChE Journal, 20(6):1178–1187.

Quinn, J., Geiger, C., Clausen, C., et al. (2005). Field demonstration of DNAPL de- halogenation using emulsiﬁed zero-valent iron. Environmental Science & Technology, March 1(39):1309–1318.

Rad, N. S. and Tumay, M. T. (1985). Factors aﬀecting sand specimen preparation by raining. Technical report, American Society for Testing and Materials.

Saleh, N., Sirk, K., Liu, Y., Phenrat, T., Dufour, B., Matyjaszewski, K., Tilton, R., and Lowry, G. V. (2006). Surface modiﬁcations enhance nanoiron transport and dnapl targeting in saturated porous media. Environmental Engineering Science, in press.

Schijven, J. and Hassanizadeh, S. (2002). Virus removal by soil passage at ﬁeld scale and ground-water protection of sandy aquifers. Water Science & Technology, 46(3):123–129.

Schijven, J. F., Hassanizadeh, S., and de Bruin, R. H. (2002). Two-site kinetic modeling of bacteriophages transport through columns of saturated dune sand. Journal of Contaminant Hydrology, 57(34):259–279.

Schrick, B., Hydutsky, B., Blough, J., and Mallouk, T. (2004). Delivery vehicles for zerovalent metal nanoparticles in soil and groundwater. Chemistry of Materials, 16:2187– 2193.

152 Schwartz, F. and Zhang, H. (2003). Fundamentals of Ground Water. John Wiley & Sons, Inc., New York.

Steiert, S. (2008). Reaktivität und Langzeitstabilität von nullwertigem Nano-Eisen bei der Sanierung von CKW-Schadensfällen in Grundwasserleitern. Diplomarbeit, Versuch- seinrichtung zur Grundwasser- und Altlastensanierung VEGAS, Institut fur¨ Wasserbau, Universität Stuttgart.

Swartjes, F. A. (2011). Introduction to contaminated site management. In Swartjes, F. A., editor, Dealing with Contaminated Sites, pages 3–89. Springer Netherlands.

Texas Instruments (2002). Analysis of the Sallen-Key Architecture, Application Report. Texas Instruments, Dallas, Texas 75265, SLOA024B edition.

Tien, C. (1989). Granular Filtration of Aerosols and Hydrosols. Butterworth, Stoneham, MA.

Tiraferri, A., Chen, K. L., Sethi, R., and Elimelech, M. (2008). Reduced aggregation and sedimentation of zero-valent iron nanoparticles in the presence of guar gum. Journal Of Colloid And Interface Science, 324(1-2):71–79.

Tosco, T., Tiraferri, A., and Sethi, R. (2009). Ionic strength dependent transport of microparticles in saturated porous media: Modeling mobilization and immobilization phenomena under transient chemical conditions. Environmental Science & Technology, 43(12):4425–4431.

Tratnyek, P. and Johnson, R. (2006). Nanotechnologies for environmental cleanup. nan- otoday, 1(2):44–48.

Tufenkji, N. and Elimelech, M. (2004). Correlation equation for predicting single-collector eﬃciency in physicochemical ﬁltration in saturated porous media. Environmental Sci- ence & Technology, 38(2):529–536.

Tufenkji, N. and Elimelech, M. (2005a). Breakdown of colloid ﬁltration theory: Role of the secondary energy minimum and surface charge heterogeneities. Langmuir, 21(3):841– 852.

Tufenkji, N. and Elimelech, M. (2005b). Response to comment on correlation equation for predicting single-collector eﬃciency in physicochemical ﬁltration in saturated porous media. Environmental Science & Technology, 39(14):5496–5497.

153 Bibliography

VanStone, N., Przepiora, A., Vogan, J., Lacrampe-Couloume, G., Powers, B., Perez, E., Mabury, S., and Lollar, B. S. (2005). Monitoring trichloroethene remediation at an iron permeable reactive barrier using stable carbon isotopic analysis. Journal Of Contaminant Hydrology, 78(4):313–325.

Vogan, J. L., Focht, R. M., Clark, D. K., and Graham, S. L. (1999). Performance evaluation of a permeable reactive barrier for remediation of dissolved chlorinated solvents in groundwater. Journal of Hazardous Materials, 68(1-2):97–108.

Wang, C. B. and Zhang, W. X. (1997). Synthesizing nanoscale iron particles for rapid and complete dechlorination of TCE and PCBs. Environmental Science & Technology, 31(7):2154–2156.

Wang, Y., Li, Y., Fortner, J. D., Hughes, J. B., Abriola, L. M., and Pennell, K. D. (2008). Transport and retention of nanoscale C60 aggregates in water-saturated porous media. Environmental Science & Technology, 42(10):3588–3594.

Witten, T. and Pincus, P. (2004). Structured Fluids. Polymers, Colloids, Surfactants. Oxford.

Xu, W. (2009). Ermittlung der Transporteigenschaften von reaktiven Nano-Eisen- Partikel mittels horizontaler S¨aulenversuche. Diplomarbeit, Versuchseinrichtung zur Grundwasser- und Altlastensanierung VEGAS, Institut fur¨ Wasserbau, Universit¨at Stuttgart.

Yang, S.-Z., Jin, H.-J., Wei, Z., He, R.-X., Ji, Y.-J., Li, X.-M., and Yu, S.-P. (2009). Bioremediation of oil spills in cold environments: A review. Pedosphere, 19(3):371 – 381.

Yao, K.-M., Habibian, M. T., and O’Melia, C. R. (1971). Water and waste water ﬁltration. concepts and applications. Environmental Science & Technology, 5(11):1105–1112.

Zhan, J., Zheng, T., Piringer, G., Day, C., McPherson, G. L., Lu, Y., Papadopoulos, K., and John, V. T. (2008). Transport characteristics of nanoscale functional zerovalent iron/silica composites for in situ remediation of trichloroethylene. Environmental Science & Technology, 42(23):8871–8876.

Zhang, W. (2003). Nanoscale iron particles for environmental remediation: An overview. Journal of Nanoparticle Research, 5:323–332.

154 Zhang, X., Lin, Y.-M., and Chen, Z.-l. (2009). 2,4,6-Trinitrotoluene reduction kinetics in aqueous solution using nanoscale zero-valent iron. Journal of Hazardous Materials, 165(1-3):923–927.

155 Curriculum Vitae

Cjestmir Volkert de Boer, M.Sc.

Personal Details

Gender: Male Date of birth: 27th of July, 1980 Place of birth: Langedijke, Ooststellingwerf, The Netherlands Citizenship: Dutch Email: [email protected]

Language Knowledge

Dutch: native English: ﬂuent German: ﬂuent

Education

09/2000–06/2007 Bachelor-/ Master study in Earth Sciences with specialization in En- vironmental Hydrogeology at the department of Geo Sciences of the Utrecht University, The Netherlands

09/1992–07/1999 High-school (Atheneum) at Stellingwerf College in Oosterwolde, Fries- land, The Netherlands

Working Experience

04/2007–03/2012 Research assistant at VEGAS research facility, department of Hy- draulic Engineering, Stuttgart University, Germany (full-time) 09/2003–08/2004 Help desk assistance at the Faculty ICT Service, Earth Sciences, Utrecht University (part-time)

04/2003–09/2009 First-line help desk Technical Supporter for YourName Webhosting, Nijmegen, The Netherlands (part-time)

05/2000–09/2000 Lock-keeper in Friesland, Provinsje Fryslˆan, The Netherlands (full- time)

02/2000–05/2000 Assemblage worker at Aludon, Donkerbroek, The Netherlands (full- time)

Research Projects

05/2009–03/2012 Development of rehabilitation technologies and approaches for multi- pressured degraded waters and the integration of their impact on river basin management. Funded by the European Commission through the 7th Research Framework Programme: FP7 ENV 2008.3.1.1.1, AQUAREHAB

08/2007–12/2008 Injection of Nano-Scale Iron for the In-Situ Remediation of Chlo- rinated Hydrocarbons in Soil and Groundwater (Investigations on Transport in large-scale Experiments and in Field Applications). Funded by the State of Baden-Wurttemberg,¨ Germany under the project number 111-047588.6 / 973.049977.9, BUT 013

05/2007–08/2007 Surfactant Flushing in Column and Flume Experiments for the Re- moval of Residual Chloronaphthaline.

10/2005–05/2007 Feasibility Study on the In-Situ Application of Reactive Nano-Sized Iron Particles to Remediate CHC-Contaminations. Funded by the State of Baden-Wurttemberg,¨ Germany through BW-PLUS under the project number BWR25001

Institut für Wasser- und Umweltsystemmodellierung

Universität Stuttgart

Pfaffenwaldring 61 70569 Stuttgart (Vaihingen) Telefon (0711) 685 - 64717/64749/64752/64679 Telefax (0711) 685 - 67020 o. 64746 o. 64681 E-Mail: [email protected] http://www.iws.uni-stuttgart.de

Direktoren Lehrstuhl für Wasserbau und Prof. Dr. rer. nat. Dr.-Ing. András Bárdossy Wassermengenwirtschaft Prof. Dr.-Ing. Rainer Helmig Leiter: Prof. Dr.-Ing. Silke Wieprecht Prof. Dr.-Ing. Silke Wieprecht Stellv.: PD Dr.-Ing. Walter Marx, AOR

Versuchsanstalt für Wasserbau Vorstand (Stand 01.04.2009) Leiter: Dr.-Ing. Sven Hartmann, AOR Prof. Dr. rer. nat. Dr.-Ing. A. Bárdossy Prof. Dr.-Ing. R. Helmig Lehrstuhl für Hydromechanik Prof. Dr.-Ing. S. Wieprecht und Hydrosystemmodellierung Jürgen Braun, PhD Leiter: Prof. Dr.-Ing. Rainer Helmig Dr.-Ing. H. Class Stellv.: Dr.-Ing. Holger Class, AOR Dr.-Ing. S. Hartmann Dr.-Ing. H.-P. Koschitzky Lehrstuhl für Hydrologie und Geohydrologie PD Dr.-Ing. W. Marx Leiter: Prof. Dr. rer. nat. Dr.-Ing. András Bárdossy Dr. rer. nat. J. Seidel Stellv.: Dr. rer. nat. Jochen Seidel

Emeriti VEGAS, Versuchseinrichtung zur Prof. Dr.-Ing. habil. Dr.-Ing. E.h. Jürgen Giesecke Grundwasser- und Altlastensanierung Prof. Dr.h.c. Dr.-Ing. E.h. Helmut Kobus, PhD Leitung: Jürgen Braun, PhD Dr.-Ing. Hans-Peter Koschitzky, AD

Verzeichnis der Mitteilungshefte

1 Röhnisch, Arthur: Die Bemühungen um eine Wasserbauliche Versuchsanstalt an der Technischen Hochschule Stuttgart, und Fattah Abouleid, Abdel: Beitrag zur Berechnung einer in lockeren Sand geramm- ten, zweifach verankerten Spundwand, 1963

2 Marotz, Günter: Beitrag zur Frage der Standfestigkeit von dichten Asphaltbelägen im Großwasserbau, 1964

3 Gurr, Siegfried: Beitrag zur Berechnung zusammengesetzter ebener Flächen- tragwerke unter besonderer Berücksichtigung ebener Stauwände, mit Hilfe von Randwert- und Lastwertmatrizen, 1965

4 Plica, Peter: Ein Beitrag zur Anwendung von Schalenkonstruktionen im Stahlwas- serbau, und Petrikat, Kurt: Möglichkeiten und Grenzen des wasserbaulichen Ver- suchswesens, 1966 2 Institut für Wasser- und Umweltsystemmodellierung * Universität Stuttgart * IWS

5 Plate, Erich: Beitrag zur Bestimmung der Windgeschwindigkeitsverteilung in der durch eine Wand gestörten bodennahen Luftschicht, und Röhnisch, Arthur; Marotz, Günter: Neue Baustoffe und Bauausführungen für den Schutz der Böschungen und der Sohle von Kanälen, Flüssen und Häfen; Geste- hungskosten und jeweilige Vorteile, sowie Unny, T.E.: Schwingungs- untersuchungen am Kegelstrahlschieber, 1967

6 Seiler, Erich: Die Ermittlung des Anlagenwertes der bundeseigenen Bin- nenschiffahrtsstraßen und Talsperren und des Anteils der Binnenschiffahrt an diesem Wert, 1967

7 Sonderheft anläßlich des 65. Geburtstages von Prof. Arthur Röhnisch mit Beiträ- gen von Benk, Dieter; Breitling, J.; Gurr, Siegfried; Haberhauer, Robert; Hone- kamp, Hermann; Kuz, Klaus Dieter; Marotz, Günter; Mayer-Vorfelder, Hans-Jörg; Miller, Rudolf; Plate, Erich J.; Radomski, Helge; Schwarz, Helmut; Vollmer, Ernst; Wildenhahn, Eberhard; 1967

8 Jumikis, Alfred: Beitrag zur experimentellen Untersuchung des Wassernachschubs in einem gefrierenden Boden und die Beurteilung der Ergebnisse, 1968

9 Marotz, Günter: Technische Grundlagen einer Wasserspeicherung im natürlichen Untergrund, 1968

10 Radomski, Helge: Untersuchungen über den Einfluß der Querschnittsform wellen- förmiger Spundwände auf die statischen und rammtechnischen Eigenschaften, 1968

11 Schwarz, Helmut: Die Grenztragfähigkeit des Baugrundes bei Einwirkung vertikal gezogener Ankerplatten als zweidimensionales Bruchproblem, 1969

12 Erbel, Klaus: Ein Beitrag zur Untersuchung der Metamorphose von Mittelgebirgs- schneedecken unter besonderer Berücksichtigung eines Verfahrens zur Bestim- mung der thermischen Schneequalität, 1969

13 Westhaus, Karl-Heinz: Der Strukturwandel in der Binnenschiffahrt und sein Einfluß auf den Ausbau der Binnenschiffskanäle, 1969

14 Mayer-Vorfelder, Hans-Jörg: Ein Beitrag zur Berechnung des Erdwiderstandes unter Ansatz der logarithmischen Spirale als Gleitflächenfunktion, 1970

15 Schulz, Manfred: Berechnung des räumlichen Erddruckes auf die Wandung kreis- zylindrischer Körper, 1970

16 Mobasseri, Manoutschehr: Die Rippenstützmauer. Konstruktion und Grenzen ihrer Standsicherheit, 1970

17 Benk, Dieter: Ein Beitrag zum Betrieb und zur Bemessung von Hochwasser- rückhaltebecken, 1970 Verzeichnis der Mitteilungshefte 3

18 Gàl, Attila: Bestimmung der mitschwingenden Wassermasse bei überströmten Fischbauchklappen mit kreiszylindrischem Staublech, 1971,

19 Kuz, Klaus Dieter: Ein Beitrag zur Frage des Einsetzens von Kavitationserschei- nungen in einer Düsenströmung bei Berücksichtigung der im Wasser gelösten Ga- se, 1971,

20 Schaak, Hartmut: Verteilleitungen von Wasserkraftanlagen, 1971

21 Sonderheft zur Eröffnung der neuen Versuchsanstalt des Instituts für Wasserbau der Universität Stuttgart mit Beiträgen von Brombach, Hansjörg; Dirksen, Wolfram; Gàl, Attila; Gerlach, Reinhard; Giesecke, Jürgen; Holthoff, Franz-Josef; Kuz, Klaus Dieter; Marotz, Günter; Minor, Hans-Erwin; Petrikat, Kurt; Röhnisch, Arthur; Rueff, Helge; Schwarz, Helmut; Vollmer, Ernst; Wildenhahn, Eberhard; 1972

22 Wang, Chung-su: Ein Beitrag zur Berechnung der Schwingungen an Kegelstrahl- schiebern, 1972

23 Mayer-Vorfelder, Hans-Jörg: Erdwiderstandsbeiwerte nach dem Ohde- Variationsverfahren, 1972

24 Minor, Hans-Erwin: Beitrag zur Bestimmung der Schwingungsanfachungs- funktionen überströmter Stauklappen, 1972,

25 Brombach, Hansjörg: Untersuchung strömungsmechanischer Elemente (Fluidik) und die Möglichkeit der Anwendung von Wirbelkammerelementen im Wasserbau, 1972,

26 Wildenhahn, Eberhard: Beitrag zur Berechnung von Horizontalfilterbrunnen, 1972

27 Steinlein, Helmut: Die Eliminierung der Schwebstoffe aus Flußwasser zum Zweck der unterirdischen Wasserspeicherung, gezeigt am Beispiel der Iller, 1972

28 Holthoff, Franz Josef: Die Überwindung großer Hubhöhen in der Binnenschiffahrt durch Schwimmerhebewerke, 1973

29 Röder, Karl: Einwirkungen aus Baugrundbewegungen auf trog- und kastenförmige Konstruktionen des Wasser- und Tunnelbaues, 1973

30 Kretschmer, Heinz: Die Bemessung von Bogenstaumauern in Abhängigkeit von der Talform, 1973

31 Honekamp, Hermann: Beitrag zur Berechnung der Montage von Unterwasserpipe- lines, 1973

32 Giesecke, Jürgen: Die Wirbelkammertriode als neuartiges Steuerorgan im Was- serbau, und Brombach, Hansjörg: Entwicklung, Bauformen, Wirkungsweise und Steuereigenschaften von Wirbelkammerverstärkern, 1974

4 Institut für Wasser- und Umweltsystemmodellierung * Universität Stuttgart * IWS

33 Rueff, Helge: Untersuchung der schwingungserregenden Kräfte an zwei hinterein- ander angeordneten Tiefschützen unter besonderer Berücksichtigung von Kavita- tion, 1974

34 Röhnisch, Arthur: Einpreßversuche mit Zementmörtel für Spannbeton - Vergleich der Ergebnisse von Modellversuchen mit Ausführungen in Hüllwellrohren, 1975

35 Sonderheft anläßlich des 65. Geburtstages von Prof. Dr.-Ing. Kurt Petrikat mit Bei- trägen von: Brombach, Hansjörg; Erbel, Klaus; Flinspach, Dieter; Fischer jr., Ri- chard; Gàl, Attila; Gerlach, Reinhard; Giesecke, Jürgen; Haberhauer, Robert; Haf- ner Edzard; Hausenblas, Bernhard; Horlacher, Hans-Burkhard; Hutarew, Andreas; Knoll, Manfred; Krummet, Ralph; Marotz, Günter; Merkle, Theodor; Miller, Chris- toph; Minor, Hans-Erwin; Neumayer, Hans; Rao, Syamala; Rath, Paul; Rueff, Hel- ge; Ruppert, Jürgen; Schwarz, Wolfgang; Topal-Gökceli, Mehmet; Vollmer, Ernst; Wang, Chung-su; Weber, Hans-Georg; 1975

36 Berger, Jochum: Beitrag zur Berechnung des Spannungszustandes in rotations- symmetrisch belasteten Kugelschalen veränderlicher Wandstärke unter Gas- und Flüssigkeitsdruck durch Integration schwach singulärer Differentialgleichungen, 1975

37 Dirksen, Wolfram: Berechnung instationärer Abflußvorgänge in gestauten Gerin- nen mittels Differenzenverfahren und die Anwendung auf Hochwasserrückhalte- becken, 1976

38 Horlacher, Hans-Burkhard: Berechnung instationärer Temperatur- und Wärme- spannungsfelder in langen mehrschichtigen Hohlzylindern, 1976

39 Hafner, Edzard: Untersuchung der hydrodynamischen Kräfte auf Baukörper im Tiefwasserbereich des Meeres, 1977, ISBN 3-921694-39-6

40 Ruppert, Jürgen: Über den Axialwirbelkammerverstärker für den Einsatz im Was- serbau, 1977, ISBN 3-921694-40-X

41 Hutarew, Andreas: Beitrag zur Beeinflußbarkeit des Sauerstoffgehalts in Fließge- wässern an Abstürzen und Wehren, 1977, ISBN 3-921694-41-8,

42 Miller, Christoph: Ein Beitrag zur Bestimmung der schwingungserregenden Kräfte an unterströmten Wehren, 1977, ISBN 3-921694-42-6

43 Schwarz, Wolfgang: Druckstoßberechnung unter Berücksichtigung der Radial- und Längsverschiebungen der Rohrwandung, 1978, ISBN 3-921694-43-4

44 Kinzelbach, Wolfgang: Numerische Untersuchungen über den optimalen Einsatz variabler Kühlsysteme einer Kraftwerkskette am Beispiel Oberrhein, 1978, ISBN 3-921694-44-2

45 Barczewski, Baldur: Neue Meßmethoden für Wasser-Luftgemische und deren An- wendung auf zweiphasige Auftriebsstrahlen, 1979, ISBN 3-921694-45-0 Verzeichnis der Mitteilungshefte 5

46 Neumayer, Hans: Untersuchung der Strömungsvorgänge in radialen Wirbelkam- merverstärkern, 1979, ISBN 3-921694-46-9

47 Elalfy, Youssef-Elhassan: Untersuchung der Strömungsvorgänge in Wirbelkam- merdioden und -drosseln, 1979, ISBN 3-921694-47-7

48 Brombach, Hansjörg: Automatisierung der Bewirtschaftung von Wasserspeichern, 1981, ISBN 3-921694-48-5

49 Geldner, Peter: Deterministische und stochastische Methoden zur Bestimmung der Selbstdichtung von Gewässern, 1981, ISBN 3-921694-49-3,

50 Mehlhorn, Hans: Temperaturveränderungen im Grundwasser durch Brauchwas- sereinleitungen, 1982, ISBN 3-921694-50-7,

51 Hafner, Edzard: Rohrleitungen und Behälter im Meer, 1983, ISBN 3-921694-51-5

52 Rinnert, Bernd: Hydrodynamische Dispersion in porösen Medien: Einfluß von Dich- teunterschieden auf die Vertikalvermischung in horizontaler Strömung, 1983, ISBN 3-921694-52-3,

53 Lindner, Wulf: Steuerung von Grundwasserentnahmen unter Einhaltung ökologi- scher Kriterien, 1983, ISBN 3-921694-53-1,

54 Herr, Michael; Herzer, Jörg; Kinzelbach, Wolfgang; Kobus, Helmut; Rinnert, Bernd: Methoden zur rechnerischen Erfassung und hydraulischen Sanierung von Grund- wasserkontaminationen, 1983, ISBN 3-921694-54-X

55 Schmitt, Paul: Wege zur Automatisierung der Niederschlagsermittlung, 1984, ISBN 3-921694-55-8,

56 Müller, Peter: Transport und selektive Sedimentation von Schwebstoffen bei ge- stautem Abfluß, 1985, ISBN 3-921694-56-6

57 El-Qawasmeh, Fuad: Möglichkeiten und Grenzen der Tropfbewässerung unter besonderer Berücksichtigung der Verstopfungsanfälligkeit der Tropfelemente, 1985, ISBN 3-921694-57-4,

58 Kirchenbaur, Klaus: Mikroprozessorgesteuerte Erfassung instationärer Druckfelder am Beispiel seegangsbelasteter Baukörper, 1985, ISBN 3-921694-58-2

59 Kobus, Helmut (Hrsg.): Modellierung des großräumigen Wärme- und Schadstoff- transports im Grundwasser, Tätigkeitsbericht 1984/85 (DFG-Forschergruppe an den Universitäten Hohenheim, Karlsruhe und Stuttgart), 1985, ISBN 3-921694-59-0,

60 Spitz, Karlheinz: Dispersion in porösen Medien: Einfluß von Inhomogenitäten und Dichteunterschieden, 1985, ISBN 3-921694-60-4,

61 Kobus, Helmut: An Introduction to Air-Water Flows in Hydraulics, 1985, ISBN 3-921694-61-2 6 Institut für Wasser- und Umweltsystemmodellierung * Universität Stuttgart * IWS

62 Kaleris, Vassilios: Erfassung des Austausches von Oberflächen- und Grundwasser in horizontalebenen Grundwassermodellen, 1986, ISBN 3-921694-62-0

63 Herr, Michael: Grundlagen der hydraulischen Sanierung verunreinigter Poren- grundwasserleiter, 1987, ISBN 3-921694-63-9

64 Marx, Walter: Berechnung von Temperatur und Spannung in Massenbeton infolge Hydratation, 1987, ISBN 3-921694-64-7

65 Koschitzky, Hans-Peter: Dimensionierungskonzept für Sohlbelüfter in Schußrinnen zur Vermeidung von Kavitationsschäden, 1987, ISBN 3-921694-65-5

66 Kobus, Helmut (Hrsg.): Modellierung des großräumigen Wärme- und Schadstoff- transports im Grundwasser, Tätigkeitsbericht 1986/87 (DFG-Forschergruppe an den Universitäten Hohenheim, Karlsruhe und Stuttgart) 1987, ISBN 3-921694-66-3

67 Söll, Thomas: Berechnungsverfahren zur Abschätzung anthropogener Tempera- turanomalien im Grundwasser, 1988, ISBN 3-921694-67-1

68 Dittrich, Andreas; Westrich, Bernd: Bodenseeufererosion, Bestandsaufnahme und Bewertung, 1988, ISBN 3-921694-68-X,

69 Huwe, Bernd; van der Ploeg, Rienk R.: Modelle zur Simulation des Stickstoffhaus- haltes von Standorten mit unterschiedlicher landwirtschaftlicher Nutzung, 1988, ISBN 3-921694-69-8,

70 Stephan, Karl: Integration elliptischer Funktionen, 1988, ISBN 3-921694-70-1

71 Kobus, Helmut; Zilliox, Lothaire (Hrsg.): Nitratbelastung des Grundwassers, Aus- wirkungen der Landwirtschaft auf die Grundwasser- und Rohwasserbeschaffenheit und Maßnahmen zum Schutz des Grundwassers. Vorträge des deutsch-franzö- sischen Kolloquiums am 6. Oktober 1988, Universitäten Stuttgart und Louis Pas- teur Strasbourg (Vorträge in deutsch oder französisch, Kurzfassungen zwei- sprachig), 1988, ISBN 3-921694-71-X

72 Soyeaux, Renald: Unterströmung von Stauanlagen auf klüftigem Untergrund unter Berücksichtigung laminarer und turbulenter Fließzustände,1991, ISBN 3-921694-72-8

73 Kohane, Roberto: Berechnungsmethoden für Hochwasserabfluß in Fließgewäs- sern mit überströmten Vorländern, 1991, ISBN 3-921694-73-6

74 Hassinger, Reinhard: Beitrag zur Hydraulik und Bemessung von Blocksteinrampen in flexibler Bauweise, 1991, ISBN 3-921694-74-4,

75 Schäfer, Gerhard: Einfluß von Schichtenstrukturen und lokalen Einlagerungen auf die Längsdispersion in Porengrundwasserleitern, 1991, ISBN 3-921694-75-2

76 Giesecke, Jürgen: Vorträge, Wasserwirtschaft in stark besiedelten Regionen; Um- weltforschung mit Schwerpunkt Wasserwirtschaft, 1991, ISBN 3-921694-76-0 Verzeichnis der Mitteilungshefte 7

77 Huwe, Bernd: Deterministische und stochastische Ansätze zur Modellierung des Stickstoffhaushalts landwirtschaftlich genutzter Flächen auf unterschiedlichem Skalenniveau, 1992, ISBN 3-921694-77-9,

78 Rommel, Michael: Verwendung von Kluftdaten zur realitätsnahen Generierung von Kluftnetzen mit anschließender laminar-turbulenter Strömungsberechnung, 1993, ISBN 3-92 1694-78-7

79 Marschall, Paul: Die Ermittlung lokaler Stofffrachten im Grundwasser mit Hilfe von Einbohrloch-Meßverfahren, 1993, ISBN 3-921694-79-5,

80 Ptak, Thomas: Stofftransport in heterogenen Porenaquiferen: Felduntersuchungen und stochastische Modellierung, 1993, ISBN 3-921694-80-9,

81 Haakh, Frieder: Transientes Strömungsverhalten in Wirbelkammern, 1993, ISBN 3-921694-81-7

82 Kobus, Helmut; Cirpka, Olaf; Barczewski, Baldur; Koschitzky, Hans-Peter: Ver- sucheinrichtung zur Grundwasser und Altlastensanierung VEGAS, Konzeption und Programmrahmen, 1993, ISBN 3-921694-82-5

83 Zang, Weidong: Optimaler Echtzeit-Betrieb eines Speichers mit aktueller Abflußre- generierung, 1994, ISBN 3-921694-83-3,

84 Franke, Hans-Jörg: Stochastische Modellierung eines flächenhaften Stoffeintrages und Transports in Grundwasser am Beispiel der Pflanzenschutzmittelproblematik, 1995, ISBN 3-921694-84-1

85 Lang, Ulrich: Simulation regionaler Strömungs- und Transportvorgänge in Karst- aquiferen mit Hilfe des Doppelkontinuum-Ansatzes: Methodenentwicklung und Pa- rameteridentifikation, 1995, ISBN 3-921694-85-X,

86 Helmig, Rainer: Einführung in die Numerischen Methoden der Hydromechanik, 1996, ISBN 3-921694-86-8,

87 Cirpka, Olaf: CONTRACT: A Numerical Tool for Contaminant Transport and Chemical Transformations - Theory and Program Documentation -, 1996, ISBN 3-921694-87-6

88 Haberlandt, Uwe: Stochastische Synthese und Regionalisierung des Niederschla- ges für Schmutzfrachtberechnungen, 1996, ISBN 3-921694-88-4

89 Croisé, Jean: Extraktion von flüchtigen Chemikalien aus natürlichen Lockergestei- nen mittels erzwungener Luftströmung, 1996, ISBN 3-921694-89-2,

90 Jorde, Klaus: Ökologisch begründete, dynamische Mindestwasserregelungen bei Ausleitungskraftwerken, 1997, ISBN 3-921694-90-6,

91 Helmig, Rainer: Gekoppelte Strömungs- und Transportprozesse im Untergrund - Ein Beitrag zur Hydrosystemmodellierung-, 1998, ISBN 3-921694-91-4, 8 Institut für Wasser- und Umweltsystemmodellierung * Universität Stuttgart * IWS

92 Emmert, Martin: Numerische Modellierung nichtisothermer Gas-Wasser Systeme in porösen Medien, 1997, ISBN 3-921694-92-2

93 Kern, Ulrich: Transport von Schweb- und Schadstoffen in staugeregelten Fließge- wässern am Beispiel des Neckars, 1997, ISBN 3-921694-93-0,

94 Förster, Georg: Druckstoßdämpfung durch große Luftblasen in Hochpunkten von Rohrleitungen 1997, ISBN 3-921694-94-9

95 Cirpka, Olaf: Numerische Methoden zur Simulation des reaktiven Mehrkomponen- tentransports im Grundwasser, 1997, ISBN 3-921694-95-7,

96 Färber, Arne: Wärmetransport in der ungesättigten Bodenzone: Entwicklung einer thermischen In-situ-Sanierungstechnologie, 1997, ISBN 3-921694-96-5

97 Betz, Christoph: Wasserdampfdestillation von Schadstoffen im porösen Medium: Entwicklung einer thermischen In-situ-Sanierungstechnologie, 1998, ISBN 3-921694-97-3

98 Xu, Yichun: Numerical Modeling of Suspended Sediment Transport in Rivers, 1998, ISBN 3-921694-98-1,

99 Wüst, Wolfgang: Geochemische Untersuchungen zur Sanierung CKW- kontaminierter Aquifere mit Fe(0)-Reaktionswänden, 2000, ISBN 3-933761-02-2

100 Sheta, Hussam: Simulation von Mehrphasenvorgängen in porösen Medien unter Einbeziehung von Hysterese-Effekten, 2000, ISBN 3-933761-03-4

101 Ayros, Edwin: Regionalisierung extremer Abflüsse auf der Grundlage statistischer Verfahren, 2000, ISBN 3-933761-04-2,

102 Huber, Ralf: Compositional Multiphase Flow and Transport in Heterogeneous Po- rous Media, 2000, ISBN 3-933761-05-0

103 Braun, Christopherus: Ein Upscaling-Verfahren für Mehrphasenströmungen in po- rösen Medien, 2000, ISBN 3-933761-06-9

104 Hofmann, Bernd: Entwicklung eines rechnergestützten Managementsystems zur Beurteilung von Grundwasserschadensfällen, 2000, ISBN 3-933761-07-7

105 Class, Holger: Theorie und numerische Modellierung nichtisothermer Mehrphasen- prozesse in NAPL-kontaminierten porösen Medien, 2001, ISBN 3-933761-08-5

106 Schmidt, Reinhard: Wasserdampf- und Heißluftinjektion zur thermischen Sanie- rung kontaminierter Standorte, 2001, ISBN 3-933761-09-3

107 Josef, Reinhold:, Schadstoffextraktion mit hydraulischen Sanierungsverfahren unter Anwendung von grenzflächenaktiven Stoffen, 2001, ISBN 3-933761-10-7

Verzeichnis der Mitteilungshefte 9

108 Schneider, Matthias: Habitat- und Abflussmodellierung für Fließgewässer mit un- scharfen Berechnungsansätzen, 2001, ISBN 3-933761-11-5

109 Rathgeb, Andreas: Hydrodynamische Bemessungsgrundlagen für Lockerdeckwer- ke an überströmbaren Erddämmen, 2001, ISBN 3-933761-12-3

110 Lang, Stefan: Parallele numerische Simulation instätionärer Probleme mit adapti- ven Methoden auf unstrukturierten Gittern, 2001, ISBN 3-933761-13-1

111 Appt, Jochen; Stumpp Simone: Die Bodensee-Messkampagne 2001, IWS/CWR Lake Constance Measurement Program 2001, 2002, ISBN 3-933761-14-X

112 Heimerl, Stephan: Systematische Beurteilung von Wasserkraftprojekten, 2002, ISBN 3-933761-15-8,

113 Iqbal, Amin: On the Management and Salinity Control of Drip Irrigation, 2002, ISBN 3-933761-16-6

114 Silberhorn-Hemminger, Annette: Modellierung von Kluftaquifersystemen: Geosta- tistische Analyse und deterministisch-stochastische Kluftgenerierung, 2002, ISBN 3-933761-17-4

115 Winkler, Angela: Prozesse des Wärme- und Stofftransports bei der In-situ- Sanierung mit festen Wärmequellen, 2003, ISBN 3-933761-18-2

116 Marx, Walter: Wasserkraft, Bewässerung, Umwelt - Planungs- und Bewertungs- schwerpunkte der Wasserbewirtschaftung, 2003, ISBN 3-933761-19-0

117 Hinkelmann, Reinhard: Efficient Numerical Methods and Information-Processing Techniques in Environment Water, 2003, ISBN 3-933761-20-4

118 Samaniego-Eguiguren, Luis Eduardo: Hydrological Consequences of Land Use / Land Cover and Climatic Changes in Mesoscale Catchments, 2003, ISBN 3-933761-21-2

119 Neunhäuserer, Lina: Diskretisierungsansätze zur Modellierung von Strömungs- und Transportprozessen in geklüftet-porösen Medien, 2003, ISBN 3-933761-22-0

120 Paul, Maren: Simulation of Two-Phase Flow in Heterogeneous Poros Media with Adaptive Methods, 2003, ISBN 3-933761-23-9

121 Ehret, Uwe: Rainfall and Flood Nowcasting in Small Catchments using Weather Radar, 2003, ISBN 3-933761-24-7

122 Haag, Ingo: Der Sauerstoffhaushalt staugeregelter Flüsse am Beispiel des Ne- ckars - Analysen, Experimente, Simulationen -, 2003, ISBN 3-933761-25-5

123 Appt, Jochen: Analysis of Basin-Scale Internal Waves in Upper Lake Constance, 2003, ISBN 3-933761-26-3

10 Institut für Wasser- und Umweltsystemmodellierung * Universität Stuttgart * IWS

124 Hrsg.: Schrenk, Volker; Batereau, Katrin; Barczewski, Baldur; Weber, Karolin und Koschitzky, Hans-Peter: Symposium Ressource Fläche und VEGAS - Statuskol- loquium 2003, 30. September und 1. Oktober 2003, 2003, ISBN 3-933761-27-1

125 Omar Khalil Ouda: Optimisation of Agricultural Water Use: A Decision Support System for the Gaza Strip, 2003, ISBN 3-933761-28-0

126 Batereau, Katrin: Sensorbasierte Bodenluftmessung zur Vor-Ort-Erkundung von Schadensherden im Untergrund, 2004, ISBN 3-933761-29-8

127 Witt, Oliver: Erosionsstabilität von Gewässersedimenten mit Auswirkung auf den Stofftransport bei Hochwasser am Beispiel ausgewählter Stauhaltungen des Ober- rheins, 2004, ISBN 3-933761-30-1

128 Jakobs, Hartmut: Simulation nicht-isothermer Gas-Wasser-Prozesse in komplexen Kluft-Matrix-Systemen, 2004, ISBN 3-933761-31-X

129 Li, Chen-Chien: Deterministisch-stochastisches Berechnungskonzept zur Beurtei- lung der Auswirkungen erosiver Hochwasserereignisse in Flussstauhaltungen, 2004, ISBN 3-933761-32-8

130 Reichenberger, Volker; Helmig, Rainer; Jakobs, Hartmut; Bastian, Peter; Niessner, Jennifer: Complex Gas-Water Processes in Discrete Fracture-Matrix Systems: Up- scaling, Mass-Conservative Discretization and Efficient Multilevel Solution, 2004, ISBN 3-933761-33-6

131 Hrsg.: Barczewski, Baldur; Koschitzky, Hans-Peter; Weber, Karolin; Wege, Ralf: VEGAS - Statuskolloquium 2004, Tagungsband zur Veranstaltung am 05. Oktober 2004 an der Universität Stuttgart, Campus Stuttgart-Vaihingen, 2004, ISBN 3- 933761-34-4

132 Asie, Kemal Jabir: Finite Volume Models for Multiphase Multicomponent Flow through Porous Media. 2005, ISBN 3-933761-35-2

133 Jacoub, George: Development of a 2-D Numerical Module for Particulate Con- taminant Transport in Flood Retention Reservoirs and Impounded Rivers, 2004, ISBN 3-933761-36-0

134 Nowak, Wolfgang: Geostatistical Methods for the Identification of Flow and Trans- port Parameters in the Subsurface, 2005, ISBN 3-933761-37-9

135 Süß, Mia: Analysis of the influence of structures and boundaries on flow and transport processes in fractured porous media, 2005, ISBN 3-933761-38-7

136 Jose, Surabhin Chackiath: Experimental Investigations on Longitudinal Dispersive Mixing in Heterogeneous Aquifers, 2005, ISBN: 3-933761-39-5

137 Filiz, Fulya: Linking Large-Scale Meteorological Conditions to Floods in Mesoscale Catchments, 2005, ISBN 3-933761-40-9

Verzeichnis der Mitteilungshefte 11

138 Qin, Minghao: Wirklichkeitsnahe und recheneffiziente Ermittlung von Temperatur und Spannungen bei großen RCC-Staumauern, 2005, ISBN 3-933761-41-7

139 Kobayashi, Kenichiro: Optimization Methods for Multiphase Systems in the Sub- surface - Application to Methane Migration in Coal Mining Areas, 2005, ISBN 3-933761-42-5

140 Rahman, Md. Arifur: Experimental Investigations on Transverse Dispersive Mixing in Heterogeneous Porous Media, 2005, ISBN 3-933761-43-3

141 Schrenk, Volker: Ökobilanzen zur Bewertung von Altlastensanierungsmaßnahmen, 2005, ISBN 3-933761-44-1

142 Hundecha, Hirpa Yeshewatesfa: Regionalization of Parameters of a Conceptual Rainfall-Runoff Model, 2005, ISBN: 3-933761-45-X

143 Wege, Ralf: Untersuchungs- und Überwachungsmethoden für die Beurteilung na- türlicher Selbstreinigungsprozesse im Grundwasser, 2005, ISBN 3-933761-46-8

144 Breiting, Thomas: Techniken und Methoden der Hydroinformatik - Modellierung von komplexen Hydrosystemen im Untergrund, 2006, 3-933761-47-6

145 Hrsg.: Braun, Jürgen; Koschitzky, Hans-Peter; Müller, Martin: Ressource Unter- grund: 10 Jahre VEGAS: Forschung und Technologieentwicklung zum Schutz von Grundwasser und Boden, Tagungsband zur Veranstaltung am 28. und 29. Sep- tember 2005 an der Universität Stuttgart, Campus Stuttgart-Vaihingen, 2005, ISBN 3-933761-48-4

146 Rojanschi, Vlad: Abflusskonzentration in mesoskaligen Einzugsgebieten unter Berücksichtigung des Sickerraumes, 2006, ISBN 3-933761-49-2

147 Winkler, Nina Simone: Optimierung der Steuerung von Hochwasserrückhaltebe- cken-systemen, 2006, ISBN 3-933761-50-6

148 Wolf, Jens: Räumlich differenzierte Modellierung der Grundwasserströmung allu- vialer Aquifere für mesoskalige Einzugsgebiete, 2006, ISBN: 3-933761-51-4

149 Kohler, Beate: Externe Effekte der Laufwasserkraftnutzung, 2006, ISBN 3-933761-52-2

150 Hrsg.: Braun, Jürgen; Koschitzky, Hans-Peter; Stuhrmann, Matthias: VEGAS- Statuskolloquium 2006, Tagungsband zur Veranstaltung am 28. September 2006 an der Universität Stuttgart, Campus Stuttgart-Vaihingen, 2006, ISBN 3-933761-53-0

151 Niessner, Jennifer: Multi-Scale Modeling of Multi-Phase - Multi-Component Pro- cesses in Heterogeneous Porous Media, 2006, ISBN 3-933761-54-9

152 Fischer, Markus: Beanspruchung eingeerdeter Rohrleitungen infolge Austrocknung bindiger Böden, 2006, ISBN 3-933761-55-7 12 Institut für Wasser- und Umweltsystemmodellierung * Universität Stuttgart * IWS

153 Schneck, Alexander: Optimierung der Grundwasserbewirtschaftung unter Berück- sichtigung der Belange der Wasserversorgung, der Landwirtschaft und des Natur- schutzes, 2006, ISBN 3-933761-56-5

154 Das, Tapash: The Impact of Spatial Variability of Precipitation on the Predictive Uncertainty of Hydrological Models, 2006, ISBN 3-933761-57-3

155 Bielinski, Andreas: Numerical Simulation of CO2 sequestration in geological formations, 2007, ISBN 3-933761-58-1

156 Mödinger, Jens: Entwicklung eines Bewertungs- und Entscheidungsunterstüt- zungssystems für eine nachhaltige regionale Grundwasserbewirtschaftung, 2006, ISBN 3-933761-60-3

157 Manthey, Sabine: Two-phase flow processes with dynamic effects in porous media - parameter estimation and simulation, 2007, ISBN 3-933761-61-1

158 Pozos Estrada, Oscar: Investigation on the Effects of Entrained Air in Pipelines, 2007, ISBN 3-933761-62-X

159 Ochs, Steffen Oliver: Steam injection into saturated porous media – process analysis including experimental and numerical investigations, 2007, ISBN 3-933761-63-8

160 Marx, Andreas: Einsatz gekoppelter Modelle und Wetterradar zur Abschätzung von Niederschlagsintensitäten und zur Abflussvorhersage, 2007, ISBN 3-933761-64-6

161 Hartmann, Gabriele Maria: Investigation of Evapotranspiration Concepts in Hydro- logical Modelling for Climate Change Impact Assessment, 2007, ISBN 3-933761-65-4

162 Kebede Gurmessa, Tesfaye: Numerical Investigation on Flow and Transport Char- acteristics to Improve Long-Term Simulation of Reservoir Sedimentation, 2007, ISBN 3-933761-66-2

163 Trifković, Aleksandar: Multi-objective and Risk-based Modelling Methodology for Planning, Design and Operation of Water Supply Systems, 2007, ISBN 3-933761-67-0

164 Götzinger, Jens: Distributed Conceptual Hydrological Modelling - Simulation of Climate, Land Use Change Impact and Uncertainty Analysis, 2007, ISBN 3-933761-68-9

165 Hrsg.: Braun, Jürgen; Koschitzky, Hans-Peter; Stuhrmann, Matthias: VEGAS – Kolloquium 2007, Tagungsband zur Veranstaltung am 26. September 2007 an der Universität Stuttgart, Campus Stuttgart-Vaihingen, 2007, ISBN 3-933761-69-7

166 Freeman, Beau: Modernization Criteria Assessment for Water Resources Plan- ning; Klamath Irrigation Project, U.S., 2008, ISBN 3-933761-70-0 Verzeichnis der Mitteilungshefte 13

167 Dreher, Thomas: Selektive Sedimentation von Feinstschwebstoffen in Wechselwir- kung mit wandnahen turbulenten Strömungsbedingungen, 2008, ISBN 3-933761-71-9

168 Yang, Wei: Discrete-Continuous Downscaling Model for Generating Daily Precipi- tation Time Series, 2008, ISBN 3-933761-72-7

169 Kopecki, Ianina: Calculational Approach to FST-Hemispheres for Multiparametrical Benthos Habitat Modelling, 2008, ISBN 3-933761-73-5

170 Brommundt, Jürgen: Stochastische Generierung räumlich zusammenhängender Niederschlagszeitreihen, 2008, ISBN 3-933761-74-3

171 Papafotiou, Alexandros: Numerical Investigations of the Role of Hysteresis in Het- erogeneous Two-Phase Flow Systems, 2008, ISBN 3-933761-75-1

172 He, Yi: Application of a Non-Parametric Classification Scheme to Catchment Hy- drology, 2008, ISBN 978-3-933761-76-7

173 Wagner, Sven: Water Balance in a Poorly Gauged Basin in West Africa Using At- mospheric Modelling and Remote Sensing Information, 2008, ISBN 978-3-933761-77-4

174 Hrsg.: Braun, Jürgen; Koschitzky, Hans-Peter; Stuhrmann, Matthias; Schrenk, Vol- ker: VEGAS-Kolloquium 2008 Ressource Fläche III, Tagungsband zur Veranstal- tung am 01. Oktober 2008 an der Universität Stuttgart, Campus Stuttgart- Vaihingen, 2008, ISBN 978-3-933761-78-1

175 Patil, Sachin: Regionalization of an Event Based Nash Cascade Model for Flood Predictions in Ungauged Basins, 2008, ISBN 978-3-933761-79-8

176 Assteerawatt, Anongnart: Flow and Transport Modelling of Fractured Aquifers based on a Geostatistical Approach, 2008, ISBN 978-3-933761-80-4

177 Karnahl, Joachim Alexander: 2D numerische Modellierung von multifraktionalem Schwebstoff- und Schadstofftransport in Flüssen, 2008, ISBN 978-3-933761-81-1

178 Hiester, Uwe: Technologieentwicklung zur In-situ-Sanierung der ungesättigten Bo- denzone mit festen Wärmequellen, 2009, ISBN 978-3-933761-82-8

179 Laux, Patrick: Statistical Modeling of Precipitation for Agricultural Planning in the Volta Basin of West Africa, 2009, ISBN 978-3-933761-83-5

180 Ehsan, Saqib: Evaluation of Life Safety Risks Related to Severe Flooding, 2009, ISBN 978-3-933761-84-2

181 Prohaska, Sandra: Development and Application of a 1D Multi-Strip Fine Sedi- ment Transport Model for Regulated Rivers, 2009, ISBN 978-3-933761-85-9 14 Institut für Wasser- und Umweltsystemmodellierung * Universität Stuttgart * IWS

182 Kopp, Andreas: Evaluation of CO2 Injection Processes in Geological Formations for Site Screening, 2009, ISBN 978-3-933761-86-6

183 Ebigbo, Anozie: Modelling of biofilm growth and its influence on CO2 and water (two-phase) flow in porous media, 2009, ISBN 978-3-933761-87-3

184 Freiboth, Sandra: A phenomenological model for the numerical simulation of multiphase multicomponent processes considering structural alterations of porous media, 2009, ISBN 978-3-933761-88-0

185 Zöllner, Frank: Implementierung und Anwendung netzfreier Methoden im Kon- struktiven Wasserbau und in der Hydromechanik, 2009, ISBN 978-3-933761-89-7

186 Vasin, Milos: Influence of the soil structure and property contrast on flow and transport in the unsaturated zone, 2010, ISBN 978-3-933761-90-3

187 Li, Jing: Application of Copulas as a New Geostatistical Tool, 2010, ISBN 978-3- 933761-91-0

188 AghaKouchak, Amir: Simulation of Remotely Sensed Rainfall Fields Using Copu- las, 2010, ISBN 978-3-933761-92-7

189 Thapa, Pawan Kumar: Physically-based spatially distributed rainfall runoff modelling for soil erosion estimation, 2010, ISBN 978-3-933761-93-4

190 Wurms, Sven: Numerische Modellierung der Sedimentationsprozesse in Retenti- onsanlagen zur Steuerung von Stoffströmen bei extremen Hochwasserabflusser- eignissen, 2011, ISBN 978-3-933761-94-1

191 Merkel, Uwe: Unsicherheitsanalyse hydraulischer Einwirkungen auf Hochwasser- schutzdeiche und Steigerung der Leistungsfähigkeit durch adaptive Strömungs- modellierung, 2011, ISBN 978-3-933761-95-8

192 Fritz, Jochen: A Decoupled Model for Compositional Non-Isothermal Multiphase Flow in Porous Media and Multiphysics Approaches for Two-Phase Flow, 2010, ISBN 978-3-933761-96-5

193 Weber, Karolin (Hrsg.): 12. Treffen junger WissenschaftlerInnen an Wasserbauin- stituten, 2010, ISBN 978-3-933761-97-2

194 Bliefernicht, Jan-Geert: Probability Forecasts of Daily Areal Precipitation for Small River Basins, 2011, ISBN 978-3-933761-98-9

195 Hrsg.: Koschitzky, Hans-Peter; Braun, Jürgen: VEGAS-Kolloquium 2010 In-situ- Sanierung - Stand und Entwicklung Nano und ISCO -, Tagungsband zur Veran- staltung am 07. Oktober 2010 an der Universität Stuttgart, Campus Stuttgart- Vaihingen, 2010, ISBN 978-3-933761-99-6

Verzeichnis der Mitteilungshefte 15

196 Gafurov, Abror: Water Balance Modeling Using Remote Sensing Information - Fo- cus on Central Asia, 2010, ISBN 978-3-942036-00-9

197 Mackenberg, Sylvia: Die Quellstärke in der Sickerwasserprognose: Möglichkeiten und Grenzen von Labor- und Freilanduntersuchungen, 2010, ISBN 978-3-942036-01-6

198 Singh, Shailesh Kumar: Robust Parameter Estimation in Gauged and Ungauged Basins, 2010, ISBN 978-3-942036-02-3

199 Doğan, Mehmet Onur: Coupling of porous media flow with pipe flow, 2011, ISBN 978-3-942036-03-0

200 Liu, Min: Study of Topographic Effects on Hydrological Patterns and the Implica- tion on Hydrological Modeling and Data Interpolation, 2011, ISBN 978-3-942036-04-7

201 Geleta, Habtamu Itefa: Watershed Sediment Yield Modeling for Data Scarce Ar- eas, 2011, ISBN 978-3-942036-05-4

202 Franke, Jörg: Einfluss der Überwachung auf die Versagenswahrscheinlichkeit von Staustufen, 2011, ISBN 978-3-942036-06-1

203 Bakimchandra, Oinam: Integrated Fuzzy-GIS approach for assessing regional soil erosion risks, 2011, ISBN 978-3-942036-07-8

204 Alam, Muhammad Mahboob: Statistical Downscaling of Extremes of Precipita- tion in Mesoscale Catchments from Different RCMs and Their Effects on Local Hydrology, 2011, ISBN 978-3-942036-08-5

205 Hrsg.: Koschitzky, Hans-Peter; Braun, Jürgen: VEGAS-Kolloquium 2011 Flache Geothermie - Perspektiven und Risiken, Tagungsband zur Veranstaltung am 06. Oktober 2011 an der Universität Stuttgart, Campus Stuttgart-Vaihingen, 2011, ISBN 978-3-933761-09-2

206 Haslauer, Claus: Analysis of Real-World Spatial Dependence of Subsurface Hydraulic Properties Using Copulas with a Focus on Solute Transport Behav- iour, 2011, ISBN 978-3-942036-10-8

207 Dung, Nguyen Viet: Multi-objective automatic calibration of hydrodynamic models – development of the concept and an application in the Mekong Delta, 2011, ISBN 978-3-942036-11-5

208 Hung, Nguyen Nghia: Sediment dynamics in the floodplain of the Mekong Delta, Vietnam, 2011, ISBN 978-3-942036-12-2

209 Kuhlmann, Anna: Influence of soil structure and root water uptake on flow in the unsaturated zone, 2012, ISBN 978-3-942036-13-9 16 Institut für Wasser- und Umweltsystemmodellierung * Universität Stuttgart * IWS

210 Tuhtan, Jeffrey Andrew: Including the Second Law Inequality in Aquatic Ecody- namics: A Modeling Approach for Alpine Rivers Impacted by Hydropeaking, 2012, ISBN 978-3-942036-14-6

211 Tolossa, Habtamu: Sediment Transport Computation Using a Data-Driven Adap- tive Neuro-Fuzzy Modelling Approach, 2012, ISBN 978-3-942036-15-3

212 Tatomir, Alexandru-Bodgan: From Discrete to Continuum Concepts of Flow in Fractured Porous Media, 2012, ISBN 978-3-942036-16-0

213 Erbertseder, Karin: A Multi-Scale Model for Describing Cancer-Therapeutic Trans- port in the Human Lung, 2012, ISBN 978-3-942036-17-7

214 Noack, Markus: Modelling Approach for Interstitial Sediment Dynamics and Repro- duction of Gravel Spawning Fish, 2012, ISBN 978-3-942036-18-4

215 De Boer, Cjestmir Volkert: Transport of Nano Sized Zero Valent Iron Colloids during Injection into the Subsurface, 2012, ISBN 978-3-942036-19-1

Die Mitteilungshefte ab der Nr. 134 (Jg. 2005) stehen als pdf-Datei über die Homepage des Instituts: www.iws.uni-stuttgart.de zur Verfügung.