modelling the reactive transport of antimony in river...

Faculteit Bio-ingenieurswetenschappen

Academiejaar 2015-2016

Modelling the reactive transport of antimony in riversystems polluted by mining

Lisa LanduytPromotor: Prof. Dr. ir. Piet SeuntjensTutor: Jef Dams

Masterproef voorgedragen tot het behalen van de graad vanMaster in de bio-ingenieurswetenschappen: land- en waterbeheer

De auteur en promotor geven de toelating deze masterproef voor consultatie beschikbaar testellen en delen ervan te kopiëren voor persoonlijk gebruik. Elk ander gebruik valt onderde beperkingen van het auteursrecht, in het bijzonder met betrekking tot de verplichtinguitdrukkelijk de bron te vermelden bij het aanhalen van resultaten uit deze scriptie.

The author and promoter give the permission to use this thesis for consultation and to copyparts of it for personal use. Every other use is subject to the copyright laws, more specificallythe source must be extensively specified when using from this thesis.

Ghent, June 2016

De promotor De auteur

Prof. dr. ir. Piet Seuntjens Lisa Landuyt

Preface

This master thesis was one of the last hurdles to take in my career as a student. During theprocess of writing this work I acquired some competences which will be indispensable for myfurther career as a bio-engineer - independence and persistence just to name a few.

However, this work could not have been established without the help of some people andinstances. First of all I would like to thank my promotor, prof. dr. ir. Piet Seuntjens,for the many moments of consultation, the guidance and for reviewing this work. I wouldalso like to offer my special thanks to Jef, my tutor, for the quick responses to my questionsand the constructive comments with regard to this work. VITO, the Flemish Institute forTechnological Research, provided me the professional instruments needed to perform thenecessary simulations, which was greatly appreciated. Furthermore I would like to thankEva and Zarah, who did their researches in the framework of the same project and went toChenzhou for sample collection. They provided me the results of their analyses and we hadsome consultations who gave me new insights thanks to the multidisciplinary of our works.

Last but not least I would like to thank my family for their unconditional support andenthusiasm, even when mine was far. A grateful thanks also goes out to my friends andparticularly Erika, for the nice evenings, coffee breaks and contributions to my quality of life.

Lisa LanduytGhent, June 2016

Contents

List of abbreviations iii

Abstract v

Samenvatting vii

1 Introduction 11.1 The METALert project . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.1.1 Project description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1.2 Pilot Sites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.2 The WASP model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31.3 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2 Literature review 52.1 Antimony . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2.1.1 General characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . 52.1.2 Toxicity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

2.2 Chemical processes influencing the fate and transport of heavy metals . . . . 62.2.1 Aqueous complexation . . . . . . . . . . . . . . . . . . . . . . . . . . . 62.2.2 Precipitation and dissolution . . . . . . . . . . . . . . . . . . . . . . . 82.2.3 REDOX processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82.2.4 Surface complexation . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

2.3 Physical processes influencing the fate and transport of heavy metals . . . . . 182.3.1 Water transport: advection, diffusion, dispersion . . . . . . . . . . . . 182.3.2 Sedimentation and resuspension . . . . . . . . . . . . . . . . . . . . . 202.3.3 Interaction between physical and geochemical processes in the river system 21

2.4 Geochemical models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

3 Materials and methods 253.1 Software . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

3.1.1 PHREEQC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

i

ii

3.1.2 PhreePlot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283.1.3 MATLAB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

3.2 Model construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293.2.1 Speciation in the aqueous phase: pe-pH diagram . . . . . . . . . . . . 293.2.2 Interaction with the solid phase: sorption edge . . . . . . . . . . . . . 293.2.3 Transport . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

4 Results and discussion 354.1 Speciation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 354.2 Sorption behaviour . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 364.3 Transport scenarios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

4.3.1 Pure advective transport . . . . . . . . . . . . . . . . . . . . . . . . . 404.3.2 Advective transport in a perfectly mixed water column with pH-independent

sorption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 414.3.3 Advective transport in a perfectly mixed water column with pH-dependent

sorption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 434.3.4 Advective transport with diffusive exchanges with the benthic segments 474.3.5 Advective transport in the presence of mobile suspended matter . . . 50

4.4 Method choice, model simplifications and shortages . . . . . . . . . . . . . . . 514.4.1 Method choice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 514.4.2 Model simplifications and shortages . . . . . . . . . . . . . . . . . . . 52

5 Conclusion and perspectives 55

Bibliography 57

A PHREEQC code 65A.1 Advection in a perfectly mixed water column, pH-independent sorption . . . 65A.2 Advection in a perfectly mixed water column, pH-dependent sorption . . . . . 67A.3 Advective transport with diffusive exchanges with the benthic segments . . . 70A.4 Desorption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73A.5 Colloid transport . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76

List of abbreviations

DLM Diffuse Layer Model

DOC Disssolved Organic Carbon

DOM Dissolved Organic Matter

ERS emergency responce system

EU European Union

HFO Hydrous Ferric Oxides

JNC Japan Nuclear Cycle Organisation

K equilibrium constant

KD partition coefficient

LFER linear free energy relationship

pe reduction potential

pKa logarithmic acid dissociation constant

pO2 partial pressure of oxygen

PZC point of zero charge

TLM Tripel Layer Model

TMDL total maximal daily loads

RMSE root mean square error

US EPA United States Environmental Protection Agency

USGS United States Geological Survey

iii

Abstract

With its impressive industrial development of the recent decades, China has become oneof the main players in the world’s economic and political scene. This growth however, hasalso resulted in increasing environmental issues like heavy metal pollution of surface andgroundwater due to e.g. mining activities. As a consequence, there is a rising interest inreactive-transport models, which can be used to assess the impact of accidental discharges onthe environment. This research explored the possibilities of geochemical models to improvethe currently used reactive-transport models.

More particularly, this research focussed on antimony, a heavy metal and pollutant of risingconcern. Despite its wide application in industrial products like flame retardants and leadalloys for lead-acid batteries and ammunition, the knowledge about its geochemistry andtoxicity is limited. It has been shown to be carcinogenic in female rats and is often assumedto have a lot of characteristics in common with arsenic.

A sorption model was constructed by means of the geochemical model PHREEQC, basedon reported experimental data. The resulting partition coefficients varied with two orders ofmagnitude within the range of environmentally relevant pH values for river systems (pH 6 - 8),indicating the need to incorporate geochemical mechanisms in water quality models for theassessment of antimony. This sorption model was linked to the PHREEQC transport moduleand some selected scenarios were simulated, ranging from a perfectly mixed water column todiffusive exchanges with benthic sediments and colloid transport.

v

Samenvatting

De voorbije decennia heeft China zich ontpopt tot één van de grotere machthebbers opeconomisch en politiek vlak. Een van de belangrijkste oorzaken hiervan is de indrukwekkendeindustriële groei die het land liet optekenen, al heeft deze ook voor onder andere ecologischeproblemen zoals oppervlakte- en grondwatervervuiling door mijnbouw gezorgd. Bijgevolg iser een stijgende vraag naar waterkwaliteitsmodellen die een inschatting kunnen geven vande impact van bijvoorbeeld accidentele lozingen. Deze masterproef had als doel het poten-tieel van geochemische modellen te onderzoeken om de huidige waterkwaliteitsmodellen teverbeteren.

De focus van deze masterproef ligt op het zwaar metaal antimoon. Rond dit element hangt nogeen waas van onwetendheid, ondanks de talrijke toepassingen in industriële producten zoalsvlamvertragers en loodlegeringen voor batterijen en munitie. Onderzoek heeft aangewezen dathet kankerverwekkend is in vrouwelijke ratten en het wordt vaak gelinkt aan arseen wegensde vele gelijkenissen op chemisch vlak.

Op basis van experimentele data uit een ander onderzoek werd een sorptiemodel opgesteld metbehulp van het geochemisch model PHREEQC. De verdelingscoefficiënten verkregen op basisvan dit model verschilden tot twee grootteordes binnen de range van relevante pH-waardenvoor riviersystemen (pH 6 - 8), waarmee de nood aan het incorporeren van geochemischemechanismen in waterkwaliteitsmodellen werd aangetoond. Dit sorptiemodel werd vervolgensgelinkt aan de PHREEQC transportmodule waarmee verschillende scenario’s gesimuleerdwerden. Ten eerste werd er uitgegaan van een perfect gemengde waterkolom, maar ook diffuseuitwisseling met de waterbodem en het transport van zwevend stof werden gesimuleerd.

vii

1 | Introduction

1.1 The METALert project

1.1.1 Project description

With its impressive economic growth and industrial development of recent years, China hasbecome one of the main players in the world’s economic and political scene. But this intensiveindustrialization has resulted in an increase of environmental threats, with growing pollutionissues. In the last years, different heavy metal pollution incidents have occurred, of whichsome have been extreme environmental accidents that have affected public health and socialstability in the region. For example, the antimony pollution incident that occurred in June2011 in Chenzhou, Hunan Province, caused severe problems with regard to the drinking waterof the cities along the Beijiang River, and these problems lasted for months. Incidents like thisillustrate the need of establishing prevention and remediation systems, emergency responsesystems (ERS), and environmental and health risk assessment systems for heavy pollutedsites, in addition to constructive long-term solutions. Within this framework, the EU-fundedMETALert project (2014-2016) was established. It should assist authorities in determining thelocation, spread and concentration of a pollution incident. Short-term forecasts for exampleshould enable authorities to take appropriate measures, such as discouraging the populationto use river water.

Sponsored by the EU-China Environmental Sustainability Programme (ESP) and supportedby the Ministery of Environmental Protection (MEP) and the EU-Delegation in China, MET-ALert is a collaboration between the Flemish Institute for Technological Research (VITO), theDutch Organization for Applied Scientific Research (TNO) and two local partners, the Beijing-based Chinese Academy for Environmental Planning (CAEP) and the Chenzhou ProvincialEnvironmental Science Institute (CPESI), located in the city of Chenzhou.

The primary goal of the project is the implementation of a generic ERS for accidental pol-lution incidents caused by key heavy metal related industries in China. The formation of asound technical basis for the ERS based on environmental models for source tracking, forecast,simulation and visualisation of heavy metal pollution is intended, as well as the implemen-

1

2

tation, evaluation and demonstration of the ERS at two representative pilot sites. At last,the project intends to set up the basis for replicating the ERS beyond the case studies of theproject by training target groups in using the ERS and transferring knowledge to final bene-ficiaries. Two representative pilot sites were selected based on natural geological background,the development of mining and smelting industries in Hunan province and analyses of thepollution status of the environment (Zhang et al., 2014a).

1.1.2 Pilot Sites

Two representative pilot sites were chosen in the surroundings of Chenzhou city and the 36Bay mining area, situated in the southeast of Hunan province. The region is located inthe subtropical climate zone, but local variations occur. The area, entirely surrounded bymountains, consists out of a densely distributed river network with a radial pattern. A largevariety of mineral resources can be found here, in particular non-ferrous metals of whomreserves of more than 6 million tons exist. Especially tungsten (W), bismuth (B), zinc (Zn),tin (Sn) and fluorite (CaF2) are very abundant.

1.1.2.1 Mining areas in Wushui and Yuxi River

The mining areas around the Wushui and Yuxi River, presented on figure 1.1, were chosen asthe representative mining pilot site. The rivers are situated in respectively Linwu and Yizhangcounty, on the border of the Hunan and Guangdong provinces, where multiple enterprisescausing severe As, Cd and Sb pollution are situated. Since the introduction of the ERS,non-point source contamination from historical activities has become the most importantcontribution to Sb in surface water.

Figure 1.1: Location of the pilot sites in the South of Hunan province: overview of the Wushui (cyan blue) and Yuxi(light blue) rivers and their tributaries (dark blue) as well as the approximate location of the mining areas (red).

Scale: 1:1,860,000 (retrieved from Google Earth).

Chapter 1. Introduction 3

Upstream of the Wushui River, 20 mining factories can be found with lead-zinc mining asprincipal activity. Analyses on heavy metals in abandoned ores, tailing sands and raw ma-terials showed mainly zinc and lead ores occur, with a severe problem of associated arsenicand varying associated antimony. Sb concentrations rise up to 90 - 1500 mg kg-1 in rawores, 20 - 350 mg kg-1 in sediments of tributaries, 0.01 - 0.4 mg L-1 in surface water and0.05 - 1.5 mg L-1 in mine water.

There are 4 non-ferrous mines and 4 dressing factories in Yizhang county, where the YuxiRiver watershed is located, and an antimony mine in Lechang City in Guangdong province.Sb concentrations in the range of 0.2-5.0 mg L-1 were reported in mine and surface water,while concentrations between 20 and 11,400 mg kg-1 were observed in tailing sands and rawores. Also in farmland soils concentrations up to 300 mg kg-1 were measured (Zhang et al.,2014a).

1.2 The WASP model

WASP 7.52, developed by the U.S. Environmental Protection Agency (US EPA), is the waterquality model that is currently being used in the METALert project. WASP stands for Waterquality Analysis Simulation Program and is a dynamic compartment-modelling program foraquatic systems, including both the water column and the underlying benthos. It can takeinto account advection, dispersion, point and diffuse mass loading, and boundary exchanges -all variable in time - to simulate the transport and fate of toxicants in the water column andin the bed (Wool et al., 2006).

To assess the water quality with respect to heavy metals, the software is used in the SimpleToxicant or TOXI module which considers processes like sorption, volatilization, biodegrada-tion, hydrolysis, photolysis and oxidation. Sorption is described as an equilibrium reaction.The module allows to simulate advection and dispersion of dissolved chemicals among watersegments, dispersive mixing to benthic segments and migration up and down the bed by porewater diffusion. For sorbed chemicals, settlement to and erosion from benthic segments canbe simulated as well as up- and downward migration within the bed through sedimentationand resuspension. A major limitation of TOXI is that chemical concentrations should benear trace levels, since the assumptions of linear partitioning do not apply at higher concen-trations. Furthermore, environmental characteristics like pH or bacterial populations, andconsequently transformation rates, can be altered by larger concentrations.

WASP produces reasonable simulation results, which can easily be visualised. The majordrawback of the program is the use of lumped KD-values to describe the partitioning ofthe toxicant between the dissolved and the particulate phase. Consequently, changes inpartitioning due to changes in the solution composition cannot be taken into account.

4

1.3 Objectives

This research aims at exploring the possibilities of alternative geochemical models to assessthe water quality with respect to heavy metals that exhibit a strong pH-dependent sorptionbehaviour, with the focus on antimony. A geochemical sorption model will be constructedand linked to a transport module. Several scenarios will be simulated and compared to theresults obtained with fixed partition coefficient, as used in the WASP model.

2 | Literature review

2.1 Antimony

2.1.1 General characteristics

Antimony, presented by the symbol Sb, is a toxic metalloid. This term indicates it has proper-ties in between those of metals and non-metals. In the Periodic System of the Elements it canbe found in group 15 just below the metalloid arsenic. The two have a lot of similar chemicalcharacteristics, i.e. electro-negativity and electron structure, and exhibit similar behaviour,like the formation of tri- and pentavalent species or oxyanions, and toxicity (Flakova et al.,2012). However, due to some differences like the atomic size some differences in biogeochemi-cal behaviour can also be expected (Fu et al., 2016). Furthermore, environmental health risksare probably less for antimony than for arsenic since human carcinogenicity of Sb has notbeen proven and the environmental distribution of Sb is low (Gebel, 1997).

The metalloid can exist in different oxidation states, i.e. -III, 0, +III and +V, but Sb(III)and Sb(V) are the most abundant in the natural environment. In the solid phase, Sb mainlyoccurs as stibnite (Sb2S3) or its transformation product valentinite (Sb2O3). When dissolvedin aqueous solutions, Sb is present as an oxyyanion, i.e. a polyatomic ion containing oxygen.In oxic environments thermodynamic modelling points out the oxidised form Sb(V) as thepredominant species, while the reduced form Sb(III) is more stable under anoxic conditions(Filella et al., 2002a; Leuz et al., 2006b). Antimony compounds naturally occur in the earth’scrust. They are found in ores of copper, silver and lead, and are also a common componentof coal and petroleum. Antimony can be present in the aquatic environment due to rockweathering, soil runoff and anthropogenic activities. The latter is the most important causeof elevated Sb-concentrations in aquatic systems, prior anthropogenic sources are mining,industrial emissions and emissions from vehicles (Filella et al., 2002a; He et al., 2012).

Production of antimony primarily takes place through mining. Total reserves were estimatedaround 2 million tons in 2015, with ores principally located in Bolivia, China, Russia, SouthAfrica and Mexico. In 2013, the world antimony production was around 154,000 tons withChina accounting for about 78% of the total (Guberman, 2015). Major applications of anti-

5

6

mony include the production of alloys, flame retardants, munitions, medicines for some trop-ical diseases, batteries, glasses, and more recently diodes, inferred detectors and Hall-effectdevices (Filella et al., 2002a; Krachler et al., 2001).

2.1.2 Toxicity

Both the US EPA and the European Union (EU) have pointed out antimony and its com-pounds as pollutants of priority interest. In its toxicological behaviour, antimony is compara-ble to arsenic and bismuth (Grund et al., 2012). The EPA has set both maximum contaminantlevel goal as maximum contaminant level in drinking water to 6 µg L-1 (U.S. EPA, 2001),while the EU allows a maximum admissible Sb concentration in drinking water of 5 µg L-1

(Counsil of the European Communities, 1998). Typical concentrations in unpolluted areasare lower than 1 µg L-1, but in the proximity of antropogenic sources concentrations can be upto 100 times higher (Filella et al., 2002a). For example, concentrations up to 6.4 mg L-1 havebeen reported in well water around the world’s largest antimony mine located in Xikuangshan,Hunan province, China (Wang et al., 2011).

It has been shown that exposure to antimony compounds can lead to respiratory effectslike bronchitis and inactive tuberculosis, cardiovascular effects like increased blood pressureand electrocardiography abnormalities, gastrointestinal effects like diarrhea and vomiting,and dermal effects like ’antimony spots’ and eczematous eruptions. In addition, antimonytrioxide and trisulphide are assumed carcinogenic to humans. Data concerning the possiblecarcinogenity of antimony in humans are limited but Groth et al. (1986) have demonstratedantimony trioxide causes lung cancer in female rats. Moreover, antimony has been shown tobe mutagenic in bacteria and phages and to induce chromosomal aberrations and abnormalcell divisions in animal and plant cells (Sundar and Chakravarty, 2010; Grund et al., 2012).Its toxicity strongly depends on speciation, similar as it applies to arsenic. Antimonites, i.e.Sb(III) species, are considered most toxic, followed by antimonates, i.e. Sb(V) species, androganoantimonials like methylated species. Trivalent compounds are stated to be about tentimes more poisonous then pentavalent species (Guo et al., 2014).

2.2 Chemical processes influencing the fate and transport ofheavy metals

2.2.1 Aqueous complexation

2.2.1.1 Theoretical background

Complex formation or coordination refers to any combination of cations with molecules oranions containing free pairs of electrons. The latter, referred to as ligands, act as bases,

Chapter 2. Literature review 7

leading to a bounding that can be electrostatic, covalent, or a mixture of both. A distinctionis made between outer-sphere complexes, temporary partnerships in which the metal ionand the ligands are separated by one or more water molecules, and inner-sphere complexes,more stable entities with the interacting ligands immediately adjacent to the metal cation.These complexes can be formed in the aqueous phase or as interaction with a particulatephase (Stumm and Morgan, 1996). In the latter case, the process is referred to as surfacecomplexation (cfr. section 2.2.4).

Metal ions exist in the water column as hydrates with hydration shells of up to six H2Omolecules per ion. The latter can act as a weak acid, with an acidity much larger than thatof free water, leading to hydrolysis and the formation of aqueous (hydr)oxy complexes. Theenhancement in acidity of the bounded water can be explained by the repulsion of the protonsof the H2O molecules by the positive charge of the metal ion (Stumm and Morgan, 1996).

2.2.1.2 Complexes of Sb

The principal oxidation states of antimony, i.e. Sb(III) and Sb(V), undergo hydrolysis result-ing in Sb(OH)3 and Sb(OH) –

6 as predominant forms. This makes it difficult to keep antimonyions stable in solution except in highly acidic media. Trivalent antimony can also occur asmethylated species but in natural waters concentrations generally remain smaller than 1% ofthe total dissolved Sb(III) concentration (Filella et al., 2002a).

Studies have shown the formation of organic complexes between Sb(III) and oxalic acid,EDTA and a large series of ligands with oxygen- and sulphur-containing groups. Being clas-sified as a borderline metal, Sb interacts with both soft ligands such as -SH and hard ligandslike -COOH. Although significant complexation has been observed at acidic pH values, theseligands generally do not prevent Sb(III) hydrolysis from occurring (Leuz, 2006; Ozer andBogucki, 1971). With regard to Sb(V), complexes with polyhydric alcohols, polyhydric phe-nols, α-hydroxy acids have been reported, but also in low amounts (Guy et al., 1998; Filellaet al., 2002b).

Both Sb(III) and Sb(V) may form precipitating inorganic complexes with chlorides and sul-phides, that are stable under respectively acidic and alkaline conditions. However, chloridecomplexes of both Sb(V) and Sb(III) have only been reported in very acidic solutions andat relatively high chloride concentrations, their presence has not been confirmed at typicalenvironmental concentrations (Takayanagi and Cossa, 1997). The interaction between anti-mony and sulphides has been quite extensively studied due to its relevance for hydrothermalsolutions. Yet, the relevance for fresh water systems is low and the stoichiometry of theformed species still remains uncertain with important discrepancies between experimentalresults (Filella et al., 2002b). For both ligands, only few research was done to complexes

8

with Sb(V). These complexes are weak and the most abundant form of Sb(V) is the hydroxocomplex Sb(OH) –

6 (Filella et al., 2002b; Leuz, 2006).

2.2.2 Precipitation and dissolution


Precipitation and dissolution processes form the link between the hydrological cycle and thecycle of rocks. The extent of the dissolution or precipitation reaction for systems that attainequilibrium can be estimated by means of equilibrium constants. From this thermodynamicinformation, diagrams can be constructed that describe the stability boundaries of the solidphases (cfr. section 2.2.3.3). An actual ion activity product higher than the solubility con-stant at a given temperature indicates oversaturation, leading to precipitation, and vice versa(Stumm and Morgan, 1996).

2.2.2.2 Precipitation and dissolution of Sb

Both Sb(III) and Sb(V) form oxides with a rather low solubility. The solubilities of theseoxides, i.e. Sb2O3 and Sb2O5, differ significantly with values up to 60 µg L-1 and 20 mg L-1

respectively. The higher solubility of Sb2O5 is one of the explanations for the predominanceof Sb(V) in the water column (Leuz, 2006). Other precipitation and dissolution processesthat influence Sb mobility include the dissolution of stibnite (Sb2S3) and de precipitationof some Sb(III) sulphates (Filella et al., 2009). Furthermore, researchers have reported theco-precipitation of both Sb(III) and Sb(V) by hydrous ferric oxides (HFO) (Guo et al., 2014).However, precipitation and co-precipitation processes are generally only important in areasof gross contamination where Sb concentrations are high enough to initiate precipitation andsecondary mineral formation (Wilson et al., 2010).

2.2.3 REDOX processes


Oxidation-reduction or redox reactions are reactions involving changes of oxidation states ofthe reactants. Such reactions are the sum of two half-reactions, a reduction half-reaction inwhich an ion accepts one or more electrons and an oxidation half-reaction in which an ionloses one or more electrons. These transfers of electrons between species have an importantinfluence on the nature of solute species in the water column (Manahan, 2000).

The equilibrium of a redox reaction can be deduced from its constituent half-reactions. If anelectrochemical cell is considered, the difference between the potentials of both half-cells is ameasure of the driving force of the overall reaction. The potential of an electrode compared tothe standard hydrogen electrode is called the electrode potential Eh. In order to more clearly


express redox equilibria over different orders of magnitude, electron activity is often expressedas the reduction potential pe. The conversion between Eh and pe is made as follows:

pe = Eh2.303RT

F

= Eh

0.0591 (2.1)

with: Eh the electrode potential, R the molar gas constant (J K-1 mol-1), T the absolutetemperature (K) and F the Faraday constant (C mol-1) (Appelo and Postma, 2005).

The pe concept can also be linked to the redox reaction itself and the associated equilibriumconstant. Consider a redox reaction:

bBred + cCox → dDox + gGred (2.2)

with: Bred the reductant and Dox its oxidised form, Cox the oxidant and Gred its reducedform and b, c, d and g stoichiometric constants.

The two half-reactions are both associated with an equilibrium constant K and can be pre-sented as followed:

bBred → dDox + n e− (2.3)cCox → gGred + n e− (2.4)

Conform to the law of mass action, the equilibrium constant of each reaction equals theratio between the product of the product concentrations and the product of the reactantconcentrations if the system is in equilibrium (Appelo and Postma, 2005):

K = [Dox]d[e−]n

[Bred]b(2.5)

log K = d· log[Dox]− pe− b· log[Bred] (2.6)

From equation 2.6 it can be deduced that the oxidation state of an element will strongly beinfluenced by the prevailing reduction potential and, if protons are produced or consumed,the prevailing pH conditions. This relation will be further elaborated in section 2.2.3.3.

2.2.3.2 Oxidation of Sb

As mentioned earlier, antimony has 2 predominant oxidation states: trivalent (+III) andpentavalent (+V). According to thermodynamic data, Sb(V) will be reduced to Sb(III) afterthe reduction of Fe(III) to Fe(II) but before the reduction of sulphate to sulphide (Wilsonet al., 2010). Apart from oxygen and hydrogen peroxide, also humic acids, as well as Mn(IV)and Fe(III) (hydr)oxides, have been reported to oxidize Sb(III) to Sb(V) by respectively

10

Belzile et al. (2001), Buschmann et al. (2005) and Leuz (2006). Due to the higher solubilityof Sb(V) and the difference in sorption behaviour between the two oxidation states, oxidationprocesses are very important with regard to Sb mobility.

Oxidation of Sb by O2 Oxygen alone does not appear to be strong enough to provoke sig-nificant Sb oxidation rates within reasonable timespans, except under extreme pH conditions,and is consequently not considered as a significant oxidant under environmental conditions(Belzile et al., 2001; Leuz and Johnson, 2005). Leuz and Johnson (2005) observed significantoxidation of Sb(III) by O2 in the pH range between 10.9 and 12.9 within 3 to 420 days. Byextrapolating the trend of pH dependence to lower pH values, a half-life of about 220 yearsis to be expected at pH 8.5. This trend can be explained by the pH-dependent speciationof Sb(III), that can occur as Sb(OH)3 or as Sb(OH) –

4 after hydrolysis. The latter is moreabundant in alkaline conditions (cfr. section 2.2.3.3) and thermodynamically more easily ox-idised due to its increased metal basicity and reducing power, caused by the electron densitydonated by the additional OH– ligand (Belzile et al., 2001; Leuz and Johnson, 2005).

Oxidation of Sb by H2O2 The oxidation of Sb(III) by hydrogen peroxide (H2O2) appearsto be more significant under environmental conditions, with half-lives varying between 32 yearsand 117 days for a Sb(III) concentration ranging from 10−8 M to 10−6 M at pH 8. Thisreaction also shows a strong first order pH-dependence for pH values below pH 11.7, withincreasing oxidation for increasing pH. The pH value of 11.7 corresponds to the dissociationconstants of Sb(OH)3 and H2O2. The latter being respectively 11.8 and 11.6, both reactionscould be rate limiting and no kinetic distinction can be made (Leuz and Johnson, 2005).Also the ionic strength has a great influence on the oxidation rates of Sb(III) by H2O2, withrate coefficients increasing by a factor of 1.9 between ionic strengths of 0.001 M and 1.0 Mat pH 10. This effect can be caused by the shift of pKa values to lower pH values and thusthe increase in the fraction of deprotonated species at a given pH, provoked by the increasedionic strength. Another explanation can be found in the shielding effect of the increasing ionicstrength, decreasing repulsion between ions of the same charge (Leuz and Johnson, 2005).

Fe(II)-mediated oxidation of Sb Leuz et al. (2006a) studied the possible co-oxidationof Sb(III) with Fe(II) by both O2 and H2O2, which showed to be a lot faster than theoxidation of Sb(III) by O2 and H2O2 alone. The Fe(II)-mediated oxidation of Sb(III) with O2is strongly pH dependent for pH values between 5 and 7.6, while for pH values between 2.2and 3.6 rate coefficients are slow and almost pH-independent. Furthermore, oxidation ratesincrease for an increasing Fe:Sb ratio. Leuz et al. (2006a) observed an optimal ratio of 3, forwhich all Sb(III) was oxidised using O2 as an oxidant. At higher ratios the oxidised fractiondecreased due to the competition for reaction intermediates between Fe(II) and Sb(III). In thepresence of H2O2 and Fe(II), a fast and pH dependent oxidation of Sb(III) occurs. As with


O2, reaction rates increase with increasing pH. Leuz et al. (2006a) observed that all Sb(III)was oxidised for a Fe:Sb ratio above 0.1, so only small Fe(II) concentrations can initiate thereaction mechanism. In a chain of reactions, Fe(II) first reacts with H2O2 to Fe(III) and OH·

or a reactive intermediate. Subsequently, OH· or the reactive intermediate oxidizes Sb(III) toSb(IV), which reacts with O2 to result in Sb(V) and O ·–

2 . The mechanism is propagated bythe reduction of Fe(III) with O ·–

2 (Leuz et al., 2006a). Due to their strong pH dependence,both pathways can be important under environmental conditions. The reactions can occur atthe oxidizing end of redox gradients, e.g. in saturated soils and lake sediments. Analysis ofpore water profiles in the freshwater lake Baldegger (Switzerland) showed an enrichment ofSb(V) and a minimum Sb(III) content, suggesting Sb mobilization due to oxidation processeswith Fe(II)-oxidation or Mn oxides as catalyser (Leuz et al., 2006a).

Oxidation of Sb by Fe (hydr)oxides The results Belzile et al. (2001) obtained with Feoxyhydroxides(1) collected from natural sediments, suggest these compounds can effectivelyoxidize Sb(III) to Sb(V) when the former is diffusing from the reducing layers of the sedimentsalong with Fe(II) to the oxidised sediment layer containing amorphous Fe oxyhydroxides. Thesame observation was made by other researchers (Leuz et al., 2006b; Guo et al., 2014). Inthe presence of natural Fe oxyhydroxides, 6 to 8 days were required to reach the completetransformation of Sb(III) into Sb(V), while in the presence of synthetic Fe oxyhydroxides oneor two days less were needed (Belzile et al., 2001). Sb(III) is oxidised for pH values between3 and 12 with a pH-dependent rate but Sb(V) is only desorbed at higher pH-values conformthe sorption edge (cfr. section 2.2.4). The oxidation rate slightly decreases under moderatelyacidic conditions (pH < 6), which can be attributed to lower stability of Fe oxyhydroxides inthese conditions. A possible overall reaction is given by:

2 Fe(OH)3 + Sb(OH)3 −−→ 2 Fe(OH)2 + H3SbO4 + H2O (2.7)

The mechanism of this oxidation reaction, as suggested by Belzile et al. (2001), consists out ofthe adsorption of antimonite by the Fe(III) oxyhydroxide, the formation of a surface complexwith one Sb(III) bounded on two Fe(III) sites, the transfer of two electrons from antimonyto two iron atoms and the release of oxidised antimony and reduced Fe(II). A schematicrepresentation of this mechanism is presented in figure 2.1. However, different mechanismshave been suggested by other researchers like Watkins et al. (2006).

Based on their results, Belzile et al. (2001) suggested that considering only the oxidised formof Sb for the adsorption of inorganic Sb on natural Fe and Mn oxyhydroxides should beacceptable since the Sb(III) can rapidly be transformed into Sb(V) in their presence. Yet,(1)Iron oxyhydroxides are chemical compounds that commonly form in aqueous environments with different

content in iron cations (Fe2+ and Fe3+), oxygen, hydroxyl, water and some amounts of SO 2–4 , CO 2–

3 andCl– . The different species are characterised by differences in ion content and mineral structure. Examplesinclude F2O3 (hematite), FeOOH (goethite) and Fe2O3 · (H2O)0.5 (ferrihydrite) (Fernández-Remolar, 2011).

12

+ Sb(OH)3

adsorption of Sb(III)

electron transfer

Fe OH

Fe OH

Fe OH

Fe OH

Fe OH

Fe O

Fe O

Fe OH

Sb

Fe OH

Fe O

Fe O

Fe OH

Sb OH

OH

O

release of Sb(V) and Fe(II)

2 Fe(OH)2 + Sb O OH

Figure 2.1: Schematic representation of a possible mechanism for the oxidation reaction of Sb(III) byFe oxyhydroxides, suggested by Belzile et al. (2001).

Guo et al. (2014) stated that the immediate oxidation of Sb(III) to Sb(V) due to the oxidativereactivity of Fe(III) should be very unlikely, since this would result in similar behaviour ofSb(III) and Sb(V) during the adsorption process whereas they observed distinctly differentbehaviour. They suggested Sb(III) is slowly oxidised in the presence of O2, probably mediatedby its coordination to the HFO as it was stated by Leuz et al. (2006b).

Oxidation of Sb in the presence of humic acids Sb(III) bound to natural organicmatter is more easily oxidised to Sb(V) by photo-oxidants than the free complex, in analogyto Sb(OH) –

4 that is more easily oxidised by H2O2 than Sb(OH)3 because of the extra electrondonating ligand. Buschmann et al. (2005) observed photosensitised oxidation of Sb(III) toSb(V) by humic acids in sunlight, with a half-life of only 17 minutes for Sb(III) at pH 7.2 and5 mg L-1 dissolved organic carbon (DOC). During the reaction of dissolved organic matter(DOM) and oxygen, different photo-oxidants are produced. Thereof, excited triplet statesand phenoxyl radicals are relevant in the oxidation process of Sb(III) to Sb(V). The reactionrates depend on both Sb(III) concentrations and pH. Oxidation rate coefficients decrease forincreasing initial Sb(III) concentrations with other conditions kept equal, while increasingpH values increase the reaction rates. This pH dependency could be explained by bindingsites like phenolic or thiol entities having pKa-values higher than 7, or by the presence of


a photo-oxidant that becomes more reactive at more alkaline conditions (Buschmann et al.,2005).

2.2.3.3 Eh-pH diagrams of Sb

As stated in section 2.2.3.1, both pe and pH can influence the stability of dissolved speciesand mineral phases due to their interference in redox reactions. Eh-pH diagrams are a usefultool to obtain a clear overview of the speciation of an element and the stability of relateddissolved species and mineral phases as a function of pH and Eh (or pe). However, reducedspecies can occur in oxic systems, while oxidised species may persist in anoxic environments,due to kinetic stabilisation and/or slow reaction rates (Takayanagi and Cossa, 1997).

Figure 2.2.A presents the Eh-pH diagram of the system Sb-O-H for a temperature of 25◦C anda pressure 105 Pa, constructed by Takeno (2005) based on the thermodynamic database of theJapan Nuclear Cycle Organisation (JNC). In figure 2.2.B, the Eh-pH diagram for the systemSb-S-O-H at 25◦C and a pressure of 105 Pa, constructed by Takayanagi and Cossa (1997)based on thermodynamic data by Pourbaix (1966) and adapted by Filella et al. (2002b), isshown.

B

Figure 2.2: A: Eh-pH diagram for the system Sb-O-H at 25◦C and 105 Pa, based on the JNC-TDB database. Thesummed concentration of all the Sb species present equals 10-10 M (Takeno, 2005). B: Eh-pH diagram for the

Sb-S-O-H system at 25◦C and 105 Pa. The dissolved Sb concentration equals 10−8 M, the concentration of dissolvedsulphur 10−3 M (Takayanagi and Cossa, 1997; Filella et al., 2002b).

These diagrams demonstrate the two most abundant forms, i.e. Sb(OH)3 and Sb(OH) –6 , are

14

stable over a very broad pH range. Sb(OH)3 is most present in a more reducing environment,except at extreme pH values, while Sb(OH) –

6 is most stable in oxic environments, except invery acidic conditions. The diagram of Takayanagi and Cossa (1997) furthermore includesthe important interaction with stibnite (Sb2S3), formed in acidic to neutral conditions, andthe dominance of the soluble Sb(III) complex SbS –

2 in alkaline reducing conditions.

In freshwater streams, Eh and pH typically vary within the range of respectively 0.3 to0.5 V, and 6.5 to 8.5. In acid mine drainage, Eh varies between 0.6 and 0.8 V while pH ismostly below 5 (Sanders, 1998). Given these environmental conditions, it can be concludedthat in river systems Sb(V) will be the thermodynamically stable Sb species. However, ingroundwater and anoxic pore water more reducing conditions reveal with Eh varying between-0.2 and 0.1 V and pH values between 6 and 8.5 (Sanders, 1998). Under these conditionsSb(III) will predominate but, as described in this section, Sb oxidation catalysed by Fe(II) orFe oxides can occur. Moreover, in the presence of humic acids and sunlight, Sb oxidation isfast.

2.2.4 Surface complexation


The accumulation of matter at the solid-water interface due to surface complexation reactions,electric interactions and adsorption of surfactants or polymers is referred to as adsorption.When considering the water column, the solid phase can include bottom sediments or colloidssuspended in the water column. Colloids range in diameter from 0.001 µm to about 1 µm andcan be composed of a variety of inorganic materials, organic substances, including humic sub-stances, and pollutants (Stumm and Morgan, 1996). If finely divided, the solid surface tendsto have excess surface energy due to an imbalance of chemical forces among surface atoms,ions and molecules. This energy level may be lowered by a reduction in surface area, obtainedby aggregation of particles or sorption of solute species. Both hydrous oxide and aluminiumsilicate surfaces, as well as organically coated and organic surfaces, contain functional surfacegroups like hydroxyl and/or carboxyl groups that are able to act as coordinating sites of thesurface and can undergo acid-base and other coordinative interactions (Stumm and Morgan,1996; Manahan, 2000).

2.2.4.2 Surface complexation of Sb

Adsorption seems to be the most principal natural attenuation process for Sb in the environ-ment. Iron and aluminium hydroxides, formed after the release of Fe and Al from the ores, aswell as other suspended particles and sediments have strong adsorption abilities for antimony(Liu et al., 2010).


Surface complexation of Sb by Fe oxyhydroxides Crecelius et al. (1975) were amongstthe first to point out the association of Sb compounds with extractable Fe and Al compounds.Iron oxyhydroxides have been reported to adsorb 40 to 75 % of total Sb concentrations retainedin the soil (Thanabalasingam and Pickering, 1990; Guo et al., 2014), so the focus of this sectionwill be on the interaction between Sb and these compounds. However, similar sorption trendshave been observed for hydrous oxides of both Al and Mn (Thanabalasingam and Pickering,1990).

Sorption of Sb(V) by amorphous Fe(III) hydroxide decreases with increasing pH, except forsmall Sb concentrations for which sorption ratios are close to 100 % for pH values between 3and 7 (Tighe et al., 2005). The adsorption edges of Sb(V) onto goethite and HFO show similarpH dependent trends (Guo et al., 2014; Leuz et al., 2006b). This pH-dependence might beexplained by the increasingly negative surface charge of Fe(III) oxyhydroxides for increasingpH values, since Sb(V) is most abundant as the oxyanion Sb(OH) –

6 (Flakova et al., 2012).Maximum sorption of Sb(V) occurs around pH 3.5, below the point of zero charge (PZC) ofboth goethite (α-FeOOH) and Fe(OH)3, being respectively around 7.4 (Cristiano et al., 2011)and 7.1 (Tighe et al., 2005). Thus, Tighe et al. (2005) suggested an increased ionic strengthmay increase the amount of Sb(V) that is adsorbed by Fe(OH)3, while Leuz et al. (2006b)observed a decrease in sorption of Sb(V) by goethite for increasing ionic strengths. The latterattributed this to the formation of KSb(OH)6 ion pairs, reducing the activity of Sb(OH) –

6 insolution. The sorption edge they fitted to their experimental data is presented in figure 2.3.A.At pH values higher than 3.5 to 4, some desorption of Sb(V) from Fe(OH)3 may occur, whileadsorption of Sb(V) onto goethite and HFO only starts to decrease at pH values higher than6.1 to 6.8 (Leuz et al., 2006b; Guo et al., 2014; Tighe et al., 2005).

Figure 2.3: Sorption edges of Sb(V) (A, squares) and Sb(III) (B triangles) on goethite at I = 0.01 M (solid symbols)and I = 0.1 M (open symbols). Experimental conditions: [Sb(V)]0 = 4.15 µM or [Sb(III)]0 = 2.2 µM, 0.5 g L−1

goethite, 25◦C. The dashed lines shows predicted adsorbed arsenate and phosphate for the same initial concentrationsof anions and goethite, calculated based on the findings of Dixit and Hering (2003) and Nilsson et al. (1992). Crosses

show corrections for Sb(III) oxidation (Leuz et al., 2006b).

16

Sb(III) is most abundant as the uncharged species Sb(OH)3 and thus differs from Sb(V) withregard to its sorption behaviour. Sb(III) retention by both goethite and HFO is maximalat neutral pH values (Guo et al., 2014; Leuz et al., 2006b; Thanabalasingam and Picker-ing, 1990). The process is almost pH-independent, with more than 80% of Sb(III) adsorbedbetween pH 1 and 12. Adsorption decreases for very low pH-values due to increasing concen-trations of Sb(OH) +

2 , while for very high pH-values the decrease can be attributed to Sb(III)oxidation and Sb(V) desorption (Leuz et al., 2006b). The sorption edge of Sb(III) constructedby Leuz et al. (2006b) is shown in figure 2.3.B. In contrast to the sorption behaviour of Sb(V),(Thanabalasingam and Pickering, 1990) observed decreasing adsorption rates of Sb(III) ontogoethite for increasing ionic strengths due to repulsion effects between the Sb(OH)3 speciesand the added ions. Leuz et al. (2006b) on the other hand observed no ionic strength de-pendence for Sb(III) sorption onto goethite over the whole pH range. This observation madethem conclude Sb(III) forms inner-sphere complexes at the surface, an assumption followedby later researchers like Guo et al. (2014).

Multiple attempts have been made to model the sorption process of Sb onto Fe oxyhydroxides.Whereas Leuz et al. (2006b) tried to achieve this using a modified triple-layer-model (TLM),Guo et al. (2014) obtained good results using a diffuse layer model (DLM). The differencebetween both models is illustrated on figure 2.4. In the DLM, the diffuse layer starts immedi-ately at the charged surface, while in the TLM three different layers are considered. Betweenthe layer at the surface, which is called the Stern layer, and the one corresponding withthe diffuse double layer, a layer starting at the closest approach distance is situated. Thesethree layers all have different capacitances, which makes the TLM more complex to describebut also makes it more capable to account for mechanistic details of the sorption process.However, most chemical observations such as sorption edges and ionic strength effects can bevery well described by both of the models. Moreover, Dzombak and Morel (1990) derived adatabase for sorption onto ferrihydrite with the DLM which fits many observations (Appeloand Postma, 2005). Thus, the DLM often serves as the standard model for heavy metalsorption modelling in natural environments but the flexibility of the TLM can be helpful formodelling detailed laboratory experiments, as performed by Leuz et al. (2006b).

Figure 2.4: The Gouy-Chapman model or DLM (left) and the Stern-Grahame model or TLM (right), used forcalculating the electrostatic contribution to surface complexation (Appelo and Postma, 2005).


Surface complexation of Sb by humic acids The retention of Sb in polluted soilsor sediments can also be attributed to interaction with humic acids. According to Pilarskiet al. (1995), Sb(V) is only sorbed when present in higher concentrations, i.e. > 10 µM, whileSb(III) sorption is also effective for smaller concentrations. They explained this limited affinitybetween Sb(V) and humic acids by the stability and structure of the hexa-hydroxy species aswhich Sb(V) is present. Filella et al. (2002b) also suggested that due to its anionic characteras hydroxo complex, Sb(V) in solution is less likely to interact with organic material havinga predominantly negative charge at the pH of natural waters. However, Tighe et al. (2005)observed significant sorption of Sb(V) at concentrations less than those used by the former,i.e. starting from 0.23 µM. Two possible explanations apply. First, Pilarski et al. (1995)used acid washed, commercially supplied humic acid and, as Thanabalasingam and Pickering(1986) found, acid washing can decrease the sorption capacity of humic acid. Furthermore,Pilarski et al. (1995) did not report the ionic strengths used in their experiments, whilebelow the PZC an increased ionic strength will lead to a decreased adsorption of inner-spherecomplexing ions with the opposite occurring at pH values above the PZC. Thus, it could beconcluded that sorption by humic acids is relevant for both Sb(III) and Sb(V), starting fromsmall Sb concentrations in the order of µM, but confirmation by further research is desirable.

Tighe et al. (2005) observed that sorption of Sb(V) onto humic acids decreases linearly forincreasing pH values within the pH range of 3 - 7, reaching values of 0 around pH 6 - 6.5.This trend can be explained by the dissolution of the humic acid at these pH values, thedeprotonation of weak acids or the competition with hydroxyls for anion exchange sites atincreased pH values. Sorption of Sb(III) on the other hand, follows a parabolic trend betweenpH 3 and 8, reaching a maximum around pH 6. This pH-dependency can be explained by theacid/base characteristics of the humic acids. At pH values lower than 6, competition with H+

for binding sites and possibly a partial dissolution of the humic acids will decrease the sorbedamounts, while increasing concentrations of OH– shift the equilibria of bounding mechanismsto non-bound Sb(III) (Buschmann and Sigg, 2004; Pilarski et al., 1995).

It is important to note that the term humic acid refers to a complex group of different acidscontaining different functional groups like carboxyl and phenolate groups. Since probablydifferent functional groups are involved in the binding process and different degrees of sta-bilization by chelation or H-bridges exist, the affinity of Sb for humic acids depends on theindividual humic acid (Buschmann and Sigg, 2004).

Surface complexation of Sb by bulk soil and sediments In natural soil systems, bothmultiple ions and different adsorbents, between which competitive effects may occur, arepresent. Different researchers have reported decreased sorption of arsenic in the presence ofphosphates and sulphates, although this effect is severely reduced in soils with large amountsof free Fe oxides (Smith et al., 2002; Livesey and Huang, 1981). The inhibiting effect of

18

phosphates has been reported for antimony sorption as well, and due to the similarities ingeochemistry it could be assumed that the same mechanisms apply (Tighe et al., 2005; Kilgouret al., 2008). The two most important adsorbents present in the soil solid phase are organicmatter, like humic acids, an non-crystalline hydroxides. Furthermore, increasing sorption hasbeen observed for increasing silt and clay fractions (Wilson et al., 2010). Clay minerals canprovide significant reactive surface areas in soils and river sediments, since they commonlycontain Fe impurities in the form of co-precipitated Fe-oxides or structural Fe substitutingfor Al in the octahedral sheet. Ilgen and Trainor (2012) observed significant sorption of bothSb(III) and Sb(V) by different clay minerals.

Considering all environmentally relevant adsorbents, sorption of Sb is generally higher inmineral soils compared to organic rich soils (King, 1998). In accordance, Tighe et al. (2005)concluded that sorption of Sb(V) by humic acids in acidic soils with high proportions oforganic matter may be relevant, but the relevance probably decreases in e.g. soils withhigh amounts of non-crystalline hydroxides due to the strong Sb(V) scavenging potentialof Fe oxyhydroxides. Within the framework of this research, the most important retentionprocess to consider is surface complexation onto Fe oxyhydroxides. As stated by Appelo andPostma (2005), the DLM is an appropriate model to simulate this process.

2.3 Physical processes influencing the fate and transport ofheavy metals

2.3.1 Water transport: advection, diffusion, dispersion

2.3.1.1 Advection

The co-transport of substances with the flowing medium they are in is called advection orconvection. The advective flux Nadv (kg s−1) can be described as followed (Baumgarten et al.,1998):

Nadv = Jadv ·A = A·u·C (2.8)

with: Jadv the net substance flux through a unit area (kg s−1 m−2), A the cross-sectionalarea (m2), u the water flow velocity (m s−1) and C the concentration of the dissolved solid(kg m−3).

Considering a spatially continuous flow and segments in which the flow velocity can be as-sumed constant, the concentration change in time due to advective transport can be expressedby means of the following formula:

∂C

∂t= −u ∂C

∂x(2.9)

Advective flow directly controls the transport of dissolved and particulate pollutants in many


water bodies. In most water quality models, this process is described by means of a hydrody-namic model. Furthermore, pore water advection can have a significant influence on pollutantconcentrations, especially in the benthic part of the water column. This process can act as asource or a sink for the water column, depending on the flow direction and the source of thepollutants (Wool et al., 2006).

2.3.1.2 Diffusion

Diffusion is considered as the movement of particles or molecules from places of higher con-centration to places of lower concentration. This process, driven by concentration gradients,was first mathematically described by Fick (Baumgarten et al., 1998):

Ndiff = Jdiff ·A = −A·Ddiff · ∆C∆x (2.10)

with: Ndiff the net substance flux (kg s−1), Jdiff the net substance flux through a unit area(kg s−1 m−2), A the cross-sectional area (m2), Ddiff the diffusion coefficient (m2 s−1) and∆C∆x the concentration gradient (kg m−3 m−1).

The equation, referred to as Fick’s first law, assumes an isotropic medium and a flux per-pendicular to the cross section area, driven by a constant concentration gradient. Fick alsodescribed how diffusion causes the concentration to change in time. Assuming an isotropicmedium, the diffusion coefficient Ddiff can be considered to be constant and Fick’s secondlaw in one dimension can be presented as follows:

∂C

∂t= Ddiff

∂2C

∂x2 (2.11)

Besides diffusion in the water column, pore water diffusion can also significantly influencepollutant concentrations. This is particularly the case in the benthic part of the water columnand in water bodies with little sediment loading (Wool et al., 2006).

2.3.1.3 Dispersion

Whereas diffusion is a property of the molecule, with the diffusion coefficient strongly de-pending on the aggregate state, dispersion is a mixing process caused by the turbulence ofthe surrounding medium which is independent of molecular properties. The mathematicaldescription of this process is very similar to molecular diffusion:

Ndisp = Jdisp ·A = −A·Ddisp · ∆C∆x (2.12)

∂C

∂t= Ddisp

∂2C

∂x2 (2.13)

20

with: Ndisp the net substance flux (kg s−1), Jdisp the net substance flux through a unit area(kg s−1 m−2), A the cross-sectional area (m2), Ddisp the dispersion coefficient (m2 s−1) and∆C∆x the concentration gradient (kg m−3 s−1).

Due to turbulent mixing concentration gradients disappear, just like due to molecular dif-fusion. However, dispersion is always connected to flow processes. Dispersive water columnexchanges, especially in the longitudinal direction, can significantly influence the transport ofpollutants. These exchanges may have an important role in e.g. diluting peak concentrationsthat may result from unsteady loads or spills (Wool et al., 2006; Baumgarten et al., 1998).

2.3.2 Sedimentation and resuspension

Since many heavy metals strongly sorb onto sediment, sediment transport also influences theirtransport and fate. Sorbed pollutants can undergo both settling and sedimentation as well asscour. In general, the stream transport capacity for sediment is mostly in excess of the actualload and resuspension or erosion occurs. However, in low flow regions the stream transportcapacity may drop enough to allow net deposition and strongly sorbed pollutants can buildup (Wool et al., 2006).

The resulting transport regime strongly depends on both the stream velocity and the particlesize, as shown in figure 2.5. Erosion only occurs starting from a threshold flow velocity whichis lowest for the silt fraction. The threshold increases for increasing and decreasing particlesize diameters due to respectively increasing weight and increasing cohesiveness. Furthermore,also porosity influences the extent to which sediments resuspend (Alkhatib and Castor, 2000).Sedimentation only occurs for particles of the coarser silt and the sand fraction and for streamvelocities below a threshold velocity, which increases with increasing particle diameter size.

Figure 2.5: Relationship between stream velocity, particle size and the regimes of erosion, transportation anddeposition (Graf, 1971).


2.3.3 Interaction between physical and geochemical processes in the riversystem

The whole of interacting geochemical and physical processes can be summarised as in fig-ure 2.6. Sorption and desorption can occur between colloids and the water column, as well asbetween bed sediments and pore water. Furthermore, scour and sedimentation of sedimentsand diffusion of dissolved metals between the water column and the pore water can take place.The interaction between these processes can be mathematically described by means of the

Figure 2.6: Overview of the different interacting processes in and between the water column and the bed sediment.

Advection-Reaction-Dispersion (ARD) equation. Considering one-dimensional transport, theequation yields (Appelo and Postma, 2005):

∂C

∂t= −u ∂C

∂x+DL

∂2C

∂x2 −∂q

∂t(2.14)

with: C the concentration of the constituent (kg m−3), t the time (s), u the water flow velocity(m s−1), x the distance (m), DL the hydrodynamic dispersion coefficient (m2 s−1) and q theconcentration in the solid phase (kg m−3). In the right hand side of the equation, the first termrepresents advective transport, the second term is the combined representation of diffusiveand dispersive transport, and the third term represents the change in concentration in thesolid phase due to reactions.

2.4 Geochemical models

In order to model the fate of heavy metals in groundwater and river systems, a variety of geo-chemical models are available that allow to model chemical reactions, e.g. sorption, in moredetail than it is done in water quality models like WASP. When choosing the appropriate

22

model, often a trade-off has to be made between the extent up to which geochemical mecha-nisms are modelled and the extent up to which physical aspects, e.g. sediment transport, areconsidered.

WATEQ4F Chemical speciation in natural waters can be calculated using WATEQ4F,provided by the USGS (Ball and Nordstrom, 1991). It is a further development of the WA-TEQ (WATer EQuilibrium) computer program, created in 1973. Based on the environmentalconditions, described by temperature, pH, Eh, dissolved oxygen, alkalinity and solution com-position, the distribution of aqueous species, ion activities and mineral saturation indices arecalculated. The resulted species may be used for reaction modelling with e.g. PHREEQC(see further).

MINTEQ and successors MINTEQ is a computer program, developed by the US EPA,which serves to calculate geochemical equilibria and was developed combining features of twoexisting models, i.e. MINEQL (MINeral EQiLibrium) and WATEQ3. MINEQL served asthe basic mathematical structure for the program, to which some WATEQ3 features wereadded. For example, a well referenced thermodynamic database, temperature correction forequilibrium constants and six different adsorption algorithms were incorporated in the newprogram (Felmy et al., 1984). This program originates from 1984 and was used as a basisfor the development of other programs in the years thereafter. The major shortage of theMINTEQ program and its successors for the purpose of this research is probably the fact thatit lacks a transport module.

A first successor of MINTEQ is MINTEQA2. It differs from its predecessor in the way calcu-lations are implemented and in the thermodynamic database used. Apart from equilibriumcompositions of aqueous solutions, it can be used to calculate the mass distribution betweenthe dissolved, adsorbed an solid phases under a variety of conditions as it includes sevenadsorption models (Allison et al., 1991). The most recent version, version 4.03, was releasedin 2006.

The most recent MINTEQ successor is Visual MINTEQ. The code, originally built basedon the MINTEQA2 software, is still maintained and updated. Version 3.1, the most recentversion, was released in 2013. There are many similarities with the MINTEQA2 code, butthroughout its development differences gradually increased. For example, the thermodynamicdatabase was expanded and revised, and new options for modelling surface complexation andorganic complexation were included (Gustafsson, 2011). The model comes with a user friendlygraphical user interface.

PHREEQC PHREEQC, acronym for pH-REdox-EQuilibrium, is a computer program de-veloped by the USGS to perform geochemical calculations. The program is based on equi-


librium chemistry of aqueous solutions interacting with minerals, gases, solid solutions, ex-changers and sorption surfaces, but is also capable of modelling kinetic reactions and one-dimensional transport. Written in the C and C++ programming languages, it includes dif-ferent aqueous models. Based on these aqueous models, speciation and saturation-indexcalculations can be performed as well as batch-reaction and one-dimensional transport calcu-lations with reversible and irreversible reactions like surface-complexation and ion-exchangeequilibria, specified mole transfers of reactants, kinetically controlled reactions and mixing ofsolutions. It lacks the ability of modelling sorption of heavy metals onto organic compounds,but surface complexation onto iron compounds can be simulated into detail and the transportmodule is developed quite well. Furthermore, inverse modelling is possible. More particular,the program allows the calculation of mineral and gas mole transfers that account for differ-ences in composition between waters within specified compositional uncertainty limits. Asa result, it is possible to identify the reactions accounting for the observed changes in watercomposition (Caruso et al., 2008; Parkhurst and Appelo, 1999).

OTEQ Developed by the USGS, OTEQ is the acronym for One-Dimensional Transportwith Equilibrium Chemistry. By coupling the solute transport model OTIS and the chemicalequilibrium model MINTEQ, it can simulate the fate and transport of water borne solutes instreams and rivers. Both the physical processes of advection, dispersion, lateral inflow andtransient storage, as speciation of aqueous species, precipitation/dissolution and sorption areconsidered. A conceptual representation of the system is provided in figure 2.7, in which alsothe different processes influencing the component concentration are indicated. The couplingof transport processes and chemical equilibrium calculations results in a simultaneous set ofalgebraic and partial differential equations, that is solved by means of the sequential iterationapproach. Each time step is divided into a reaction step and a transport step. Similar tothe PHREEQC procedure, each segment is considered as a batch reactor in which chemicalequilibrium is assumed. Based on the resulting dissolved, precipitated and sorbed forms, atransport step is performed for the mobile phases of each solute. The procedure is iterateduntil a specified level of convergence is achieved (Runkel, 2010).

ORCHESTRA ORCHESTRA, which stands for Objects Representing CHEmical Speci-ation and TRAnsport, is another geochemical model, but it differentiates itself from otherchemical equilibrium algorithms by the fact that the model equations are not hard-coded inthe source code but defined as objects in text format. As a consequence, the program is verytransparent and model definitions are easily accessible and extendible by users without havingto change the source code. Furthermore, the calculation kernel does not contain any infor-mation on chemical models and is consequently very compact and easily used as a submodelwithin mass transport or kinetic models (Meeussen, 2003).

24

Figure 2.7: Conceptual representation of the surface-water system as considered in OTEQ. The total componentconcentration consists of a dissolved fraction (C), mobile precipitate (Pw), immobile precipitate (Pb), mobile

sorbed (Sw) and immobile sorbed (Sb) phases. The dissolved fraction and mobile phases are subject to transport,indicated by L(). Mass transfer between the different phases occurs due to dissolution from the immobile

substrate (fb), precipitation/dissolution from the water column (fw), (de)sorption from the immobile substrate (gb),(de)sorption from the water column (gw), external gains and losses (sext) and settlement (v1) (Runkel, 2010).

Other models The models described above are some, but certainly not all, of the availablegeochemical models that can be used to describe the fate and transport of heavy metalsin hydrogeological systems. Some other examples are ECOSAT (Keizer and van Riemsdijk,2009), the Windermere Humic Aqueous Model (Centre for Ecology & Hydrology, 2016) andthe Geochemist’s Workbench (Bethke and Yeakel, 2016).

3 | Materials and methods

3.1 Software

3.1.1 PHREEQC

The geochemical model PHREEQC is used for all geochemical equilibrium and transportcalculations performed in the framework of this research. This model was chosen for itsflexibility, e.g. different databases can be chosen and modified, and easy workable output.Its most important advantage however is the fact that it incorporates a transport modulewith a lot of possibilities, e.g. for the transport of solid phases and diffusive exchanges withstagnant regions. In the purpose of this research, version 3 of the PHREEQC software wasused (Parkhurst and Appelo, 1999).

3.1.1.1 PHREEQC keywords

In this section, the keywords used for the PHREEQC simulations made in the framework ofthis research will be highlighted.

DATABASE This keyword is used to specify the database that should be used for thesimulation. Different databases are provided in the PHREEQC software package, includ-ing PHREEQC.dat, MINTEQ.dat, WATEQ4F.dat, PITZER.dat, LLNL.dat, minteq.v4.dat,frezchem.dat, iso.dat and sit.dat. These all include different reactions with possibly differentequilibrium constants. In the framework of this research, the minteq.v4 database is modifiedand used.

SURFACE_MASTER_SPECIES & SURFACE_SPECIES The data block namedSURFACE_MASTER_SPECIES is used to define a new surface master species, while SUR-FACE_SPECIES is used to define or modify the reactions and associated log K values foreach of the surface species. In the framework of this research, the SURFACE_SPECIES datablock is used to modify the already existing sorption reactions for Sb onto HFO. Moreover, anew surface assemblage named Sorb is defined in order to simulate pH-independent sorption.The K values associated with the sorption reactions for this surface species are calculated as

25

26

followed:K = [SorbSb]

[Sorb] [Sb] = KD

[Sorb] (3.1)

with: [Sorb] the concentration of the solid phase onto which no sorption occurred (mmol kgw−1)(2),[SorbSb] the concentration of Sb sorbed onto the solid phase (mmol kgw−1), [Sb] the dis-solved Sb concentration (mmol kgw−1) and KD the dimensionless partition coefficient (-).The amount of solid phase present in solution is set to 10100, so that it can be consideredas constant and the number of sorption sites is limitless. Consequently, the log K value canbe calculated by subtracting the value of 100 of the logarithm of the dimensionless partitioncoefficient.

SURFACE_MASTER_SPECIESSorb Sorb

SURFACE_SPECIESSorb = Sorblog_k 0Sb(OH)3 + Sorb = SorbSb(OH)3log_k -98.9063Sb(OH)6- + Sorb = SorbSb(OH)6-log_k -98.9063

Frame 3.1: Definition of a new surface assemblage Sorb and the associated sorption reactions, to which a KD of124 L kg-1 applies. The log K values are calculated based on equation 3.1

PHASES Gas components and minerals used in the speciation, reaction and transportcalculations are defined under the PHASES data block. This block can be used to modifyalready defined phases or to define phases that are not yet defined in the thermodynamicdatabase used. To be able to fix the pH to the desired value, a new phase Fix_H+ is definedin the framework of this research. The log K value of the reaction is set to 0.

PHASESFix_H+

H+ = H+log_k 0.

Frame 3.2: Definition of a new phase Fix_H+ that allows to fix the pH in a solution.

SOLUTION In the SOLUTION data block, the temperature and initial chemical compo-sition of a solution can be defined. Default concentration units are mmol kgw−1 but thesecan be changed by the user.(2)mmol per kg water

Chapter 3. Materials and methods 27

SOLUTION 1pH 7Na 10Cl 10Sb(V) 1e-01END

Frame 3.3: Definition of a solution with a pH of 7, containing 0.01 mol kgw−1 NaCl and 100 µmol kgw−1 Sb(V).

SURFACE The SURFACE data block is used to define the amount and composition ofa solid phase present in a solution. Different complexation models are provided, but in theframework of this research the DLM is used. The databases included with PHREEQC containdata for a DLM surface named Hfo, referring to hydrous ferric oxide, derived from Dzombakand Morel (1990). This surface has strong binding sites, referred to as Hfo_s, and weakbinding sites, referred to as Hfo_w. Dzombak and Morel (1990) used 0.2 mol weak sites and0.005 mol strong sites per mole Fe, a surface area of 5.33 104 m2 mol−1 Fe and a molar weightof 89 g mol−1. To be consistent with their model, the relative number of strong and weaksites should be kept constant as the total number of sites varies.

SURFACE 1Hfo_s 5.627e-05 600 1Hfo_w 0.0025END

Frame 3.4: Definition of a surface with a mass of 1 g, 5.627 10−5 mol strong sites, 0.0025 mol weak sites and aspecific area of 600 m2 g−1.

EQUILIBRIUM_PHASES The EQUILIBRIUM_PHASES data block is used to definemineral phases with a fixed saturation index and gases with fixed partial pressures that canreact reversibly with the aqueous phase. In the purpose of this research, the EQUILIB-RIUM_PHASES data block is used to fix the pH and/or the partial pressure of oxygen.

EQUILIBRIUM_PHASES 1-40Fix_H+ -7 NaOHO2(g) -0.67778

Frame 3.5: Fixation of the phase Fix_H+ to a saturation index of -7, which leads to a pH of 7, and of the log pO2 to-0.68 in cell 1 to 40. In order to reach a pH value of 7, NaOH will be added or removed.

TRANSPORT In order to simulate one-dimensional transport of both solutes, water, col-loids and heat, the TRANSPORT data block can be used. By means of this data block, bothadvection, dispersion and diffusion are considered. An expansion towards three-dimensionaldiffusion is possible by means of stagnant zones or immobile cells associated with the mobile

28

cells in which advection occurs. The immobile cells associated with a mobile cell i are givena number n according to the following formula:

n = m· cells+ 1 + i (3.2)

with: cells the number of mobile cells and m a number ranging from 1 to the number ofimmobile cells associated with the mobile cell i. For each immobile cell the exchange factorα associated with the first-order exchange model should be defined, which can be calculatedusing the following formula (Van Genuchten, 1985):

α = De · εim(a· f)2 (3.3)

with: De the effective diffusion coefficient in the stagnant zone, a the effective diffusion length,εim the porosity in the immobile cell and f a shape factor. Both the effective diffusion lengthand the shape factor depend on the shape of the stagnant region. For a stagnant region withthe geometry of a plane sheet, the effective diffusion length equals half the thickness of thesheet while the shape factor equals 0.533.

TRANSPORT-cells 40-shifts 100-time_step 312.5-flow_direction forward-boundary_conditions flux-lengths 250-dispersivities 0-diffusion_coefficient 0-stagnant 0END

Frame 3.6: Definition of 1D transport without dispersion and diffusion over 40 cells of 250 m. The flow direction isforward and the Cauchy boundary condition is imposed. In order to impose a flow velocity of 0.8 m s-1 for cells with alength of 250 m, each time step lasts 312.5 s. The duration of this transport simulation is 100 time steps. No stagnant

zones are considered.

3.1.2 PhreePlot

PhreePlot is a program developed by Kinniburgh and Cooper (2011) in order to create geo-chemical plots and to fit geochemical models using PHREEQC. The program contains an em-bedded version of PHREEQC and makes it relatively easy to perform repetitive PHREEQCcalculations, e.g. to make graphs. In the framework of this research, PhreePlot is used togenerate predominance diagrams and to fit PHREEQC models to observations by non-linearleast squares.


3.1.3 MATLAB

MATLAB, a commercial matrix-based computer language, is used in the framework of thisresearch for processing model output and constructing explanatory graphs (MATLAB, 2015).

3.2 Model construction

The purpose of this research is to construct a generic model simulating the transport ofantimony in a river system. Since eventually no data applying for the pilot sites used inthe METALert project were provided, some general environmentally relevant scenarios areevaluated. Given the importance of Fe oxyhydroxides on the one hand, and the complexityof organic matter on the other hand, only surface complexation by Fe oxyhydroxides will beconsidered. As stated by Buschmann and Sigg (2004), the affinity of Sb for e.g. humic acidsdepends on the individual humic acid since different functional groups are involved in thebinding process. A pH-range of 3 to 11 is considered, since the environmental relevance ofvalues outside this range is low.

3.2.1 Speciation in the aqueous phase: pe-pH diagram

First, the speciation resulting from chemical equilibrium calculations based on the minteq.v4database is investigated. Calculations are done for a Sb-O2-H2O system containing 10−6 M Sband 10−3 M NaCl as background electrolyte. The partial pressure of O2, linked to the pethrough the reduction and oxidation reaction of H2O, and the pH are varied within the rangesof 10−90 to 1 atm and 3 to 11 respectively. Based on the obtained results, changes are madeto the database considering the Eh-pH diagrams presented in literature. The Eh-pH diagramis constructed using PhreePlot, by means of the hunt & track approach.

3.2.2 Interaction with the solid phase: sorption edge

In order to model the geochemical behaviour of antimony with regard to surface complexation,sorption reactions for both Sb(III) and Sb(V) onto HFO are defined. The sorption edgesobtained by both Leuz et al. (2006b) and Guo et al. (2014) are used as a benchmark. Theseresearchers observed a pH-independent high sorption of Sb(III), while for Sb(V) a strongpH-dependency was reported with complete sorption up to pH 6 but decreasing sorption forincreasing pH values above pH 6. The sorption edges for Sb(III) and Sb(V) obtained by Leuzet al. (2006b) are presented in figure 2.3.

The HFO model is constructed based on data obtained by Leuz et al. (2006b) for sorptiononto goethite, as summarised in table 3.1, since these two solid phases are very similar (Guoet al., 2014). A solution of 1 kg water containing 4.15 µM Sb(V) or 2.2 µM Sb(III) and 0.01M NaCl is defined. By means of the SURFACE data block, an amount of 0.5 g HFO is defined

30

with a surface area of 600 m2 g−1, 1.125 · 10−3 mol weak sites and 2.814 · 10−5 mol strongsites. The pH is fixed to the desired value by means of the phase Fix_H+. Furthermore,the log pO2 is fixed to 0.6778 for Sb(V) and -70 for Sb(III), in order to avoid oxidation orreduction to the other valence state.

Table 3.1: Experimental sorption data obtained by Leuz et al. (2006b). Experimental conditions:[Sb(V)]0 = 4.15 µM or [Sb(III)]0 = 2.2 µM, I = 0.01 M, 0.5 g L−1 goethite, 25◦C.

sorption data for Sb(V)pH amount of Sb(V) sorbed(-) (µM)3.04.16.66.88.48.48.68.68.99.110.010.110.110.910.912.012.0

4.154.154.054.002.791.922.002.232.002.061.291.211.390.920.700.500.38

sorption data for Sb(III)pH amount of Sb(III) sorbed(-) (µM)2.92.93.04.05.46.07.07.8

10.011.012.0

1.922.002.102.102.102.152.152.202.152.152.15

After a first rough estimation based on trial and error, the equilibrium constants of theconcerned soption reactions are determined by means of a geochemical model fit using thePhreePlot software. The fitting is performed on the sorbed percentages, in order to eliminateeffects of oxidation or reduction. A convergence criterion of 10−5, an initial step size of 10−2

and an iteration maximum of 5000 are used.

The partition coefficient KD (L kg-1) is calculated for different sediment concentrations anddifferent pH values by means of the following formula:

KD = CsCw

(3.4)

with: Cs the concentration in the solid phase (mol kg-1) and Cw the dissolved concentration(mol L-1). The obtained values are compared with the value used in the WASP model, i.e.124 L kg-1.


3.2.3 Transport

The transport scenarios considered in this research comprise a system consisting out of 40cells, each with a length of 250 m and a depth of 1m. Each cell is assumed to have a surfacepart of 0.9 m deep in which 80 mg L-1 suspended matter is present and a benthic part of0.1 m deep containing 1.0 kg L-1 of sediments. These values are chosen based on the valuesused in the WASP model and correspond to values reported in literature (Chen, 2000). First,it is assumed that the benthic sediments and the suspended matter are perfectly mixed overthe entire water column, which results in a solids concentration of 0.10 kg-1. A conceptualrepresentation of this scenario is given in figure 3.1A. Next, the benthic part and a surfacepart without suspended matter are considered, as presented in figure 3.1B. At last, only thesurface part considered in which suspended matter is present (figure 3.1C).

. . .

cell 2 cell i cell 39 cell 40

1 m

. . .

cell 1

250 m

0.10 kg L-1 A

1.0 kg L-1

. . .

cell 1 cell 2 cell i cell 39 cell 40

0.9 m

. . .

0.1 m

250 m

B

. . .

cell 1 cell 2 cell i cell 39 cell 40

0.9 m

. . .

250 m

80 mg L-1 C

Figure 3.1: A: conceptual representation of the river system assuming a perfectly mixed water column. The systemconsists out of 40 cells with a length of 250 m and a depth of 1 m, containing 0.10 kg L-1 solids. B: conceptual

representation of the river system when distinction is made between the benthic and surface part. The system consistsout of 40 cells with a length of 250 m, a surface part with a depth of 0.9 m and a benthic part with a depth of 0.1 mand a sediment concentration of 1.0 kg L-1. C: conceptual representation of the river system when only considering

the surface part. The system consists out of 40 cells with a length of 250 m, a depth of 0.9 m and a suspended matterconcentration of 80 mg L-1.

32

A water flow velocity of 0.8 m s−1 is imposed, in analogy to the flow values resulting from theWASP model, in particular for the point where the Yuxi River enters the Wushui River (cfr.figure 1.1). Since the river segments have a length of 250 m, each time step corresponds to atime interval of 312.5 s or about 5 min. The incoming aquatic solution contains 1 µM Sb and0.01 M NaCl as background electrolyte in case of a step input or 100 µM Sb and 0.01 M NaClas background electrolyte in case of a pulse input. These Sb concentrations fall within therange of environmentally relevant values, as reported by Wang et al. (2011). Unless statedotherwise, the 40 cells initially only contain 0.01 M NaCl as background electrolyte. Thediffusion coefficient is set to 10−9 m2 s-1 for all solutes and to 10−13 m2 s-1 for suspendedmatter. Dispersion is not considered. For the transport of suspended matter, the diffuse layeris calculated.

The solid phase is considered to have a HFO content of 10.59 g kg-1, which is the median of theamorphous Fe contents of sediments from the Lianxi river (Hunan, China) measured by Wanget al. (2011). This value furthermore corresponds well to the oxalate extractable Fe contentof sediments from the Wushui river, equalling 9 g kg-1, as measured by Verbrugghe (2016).The HFO concentration, number of strong sites and number of weak sites corresponding toeach of the solids concentrations considered in the framework of this research are summarisedin table 3.2. To determine the log K value of the pH-independent sorption reactions of Sbonto the surface Sorb, the same partition coefficient of 124 L kg-1 is used as in the WASPmodel. By means of equation 3.1, this results in a log K value of -98.91.

Table 3.2: HFO concentration, number of weak binding sites and number of strong binding sites associated to thedifferent solids concentrations considered in the framework of this research.

solids concentration HFO concentration number of weak sites number of strong sites(kg L-1) (g L-1) (mol) (mol)

1.01.0· 10−1

8.0· 10−5

1.06· 101

1.068.47· 10−4

2.38· 10−2

2.38· 10−3

1.90· 10−6

5.95· 10−4

5.95· 10−5

4.76· 10−8

In the case retardation occurs, the retardation factor R is calculated by means of the followingformula:

R = 1 + ρbθKD (3.5)

with: ρb the bulk density of the sediment (kg L-1), θ the porosity (-) and KD the partitioncoefficient (L kg-1).

For simulations in which the surface and benthic part of the water column are consideredseparately, an exchange factor describing the diffusive transfer between the surface and thebenthic part of the water column is calculated by means of equation 3.3. For an effectivediffusion coefficient of 5.32 · 10−10 m2 s−1 and an immobile zone with the geometry of a


plane sheet, an effective diffusion length of 0.05 m and a porosity of 0.6226, this factor equals4.661 · 10−7 s−1.

For each scenario, the influence of both pH and oxygen content is investigated. pH values arevaried between 3 and 11, while log pO2 is set alternately to -0.68, where Sb(V) is dominant,-45, where both Sb(III) and Sb(V) are present, and -70, where Sb(III) predominates.

4 | Results and discussion

4.1 Speciation

From the minteq.v4 database provided in the PHREEQC software package, all speciationreactions resulting in other aqueous Sb species then Sb(OH) –

6 or Sb(OH)3 were removed, inorder not to make the model too complicated and given the low relevance of these species inthe pH range considered. The log K value of the reduction reaction of Sb(V) was modifiedfrom a value of 24.31 to a value of 25.15, in accordance with what Filella and May (2003)published:

Sb(OH) −6 + 3 H+ + 2 e− −−→ Sb(OH)3 + 3 H2O (log K 25.15) (4.1)

The resulting Eh-pH diagram is shown in Figure 4.1.

Figure 4.1: Eh-pH diagram for the system Sb-O-H at 25◦C and 101, 325 Pa, constructed based on a modifiedminteq.v4 database. The diagram was constructed for 10−6 M Sb and a background electrolyte of 10−3 M.

This diagram corresponds rather well with the Eh-pH diagram presented by Takeno (2005),shown in figure 2.2.A. The aqueous species Sb(OH) +

2 , Sb(OH) –4 and Sb(OH)5 are not present

in this diagram since they were removed from the database. However, as can be seen on fig-

35

36

ure 2.2, neither of them is dominant within the pH range of 3 to 11. From the diagram it canbe deduced that for a partial oxygen pressure of 0.21 atm, which corresponds to an electrodepotential of 0.85 V at a pH of 7, Sb(V) is dominant. This condition is representative for afully oxygenated freshwater system. For a partial oxygen pressure of 10−70 atm, correspond-ing to an electrode potential of -0.18 V at pH 7, Sb(III) predominates. This condition isrepresentative for a ground water system as stated by Sanders (1998). When the partial oxy-gen pressure is around 10−45 atm both Sb(III) and Sb(V) are present. This partial oxygenpressure corresponds to an electrode potential of 0.15 V at pH 7 and is representative fortransient conditions.

4.2 Sorption behaviour

The sorption process of Sb onto HFO is described by three sorption equations:

Hfo_OH + Sb(OH) −6 + H+ = Hfo_SbO(OH)4 + 2 H2O (4.2)

Hfo_OH + Sb(OH) −6 = Hfo_OHSbO(OH) −

4 + H2O (4.3)Hfo_OH + Sb(OH)3 + H+ = Hfo_SbO(OH)2 + H2O (4.4)

The two equations describing the sorption behaviour of Sb(V) were already defined in theminteq.v4 database, the equation describing the sorption behaviour of Sb(III) was added tothe database based on the equations presented by Guo et al. (2014). The log K values obtainedbased on the data of Leuz et al. (2006b) equal 4.478 for equation 4.2, 3.735 for equation 4.3 and4.957 for equation 4.4. The obtained log K values differ from both the values present in theminteq.v4 database, i.e. 8.4 for equation 4.2 and 1.3 for equation 4.3, and the values presentedby Guo et al. (2014), i.e. 19.15 for equation 4.4. However, Guo et al. (2014) presented thesevalues based on sorption experiments with Sb concentrations of 101 and 202 µM and useda different set of equations, so the comparability could be questioned. Dzombak and Morel(1990) on the other hand used linear free energy relationships (LFER) to determine theintrinsic complexation constant of Sb(OH) –

6 based on its acidity constant pKa, assumingthat the site affinity of an anion reflects its affinity for protonation in solution. With their logK values, sorption starts to decrease from pH 4, in contrast to what is experimentally observed.Possibly, this calculation method cannot take into account all occurring mechanisms.

The resulting sorption edges, fitted to the data presented in table 3.1 are presented in fig-ure 4.2. Leuz et al. (2006b) used a goethite concentration of 0.5 g L-1 for their experiments,which corresponds to a sediment concentration of 0.04722 kg L-1 assuming a goethite contentof 10.59 g kg-1. The RMSE’s(3) of these sorption edges are respectively 3606 % for Sb(V)(3)The root mean square error is a measure to express the deviation of predictions from observations and is

calculated as followed: RMSE =

√∑n

i=1(yi − yi)2

n, for n pairs of observations yi and predictions yi.

Chapter 4. Results and discussion 37

and 2 % for Sb(III). The obtained sorption edge for Sb(V) matches rather well with the edgepresented by Leuz et al. (2006b) for pH values up to 9. However, the latter shows a slightchange in trend around this pH value which is not visible in the sorption edge presented in fig-ure 4.2.A. For pH values above 10, the modelled values overestimate the measured values withdifferences up to 1 µM. These deviations are the cause of the very high RMSE value and couldprobably be explained by the fact that in the framework of this research the DLM was usedwhile Leuz et al. (2006b) made use of a modified TLM. These researchers did not model thesorption behaviour of Sb(III), but the slight pH-dependence they observed for small pH-valuescan be described by means of equation 4.4. Figure 4.2 also presents the sorption edges for thesolids concentrations used for the simulations in the framework of this research. Especiallythe sorption edges onto suspended matter, present in a concentration of 8.0 · 10−5 kg L-1,draw the attention. For such a low sediment concentration the number of binding sites isaround 2 · 10−6 mol, which is lower than the total Sb concentration. The sorption capacityof the sediments is thus limiting, which results in distinctly different sorption edges than forhigher sediment concentrations.

Figure 4.2: Data points of Leuz et al. (2006b) and modelled sorption edges of Sb(V) (A) and Sb(III) (B) for0.5 g L-1 goethite or 0.04722 kg L-1 sediments, as well as the sorption edges in a system with resp. 0.10 and

8.0 · 10−5 kg L-1 solids.

38

Verbrugghe (2016) performed desorption experiments onto sediments collected from theWushuiriver and tailings collected from a mining site closeby. The influence of pH was investi-gated, although only in small ranges. Figure 4.3 shows the results obtained by Verbrugghe(2016), compared to the sorption data obtained by Leuz et al. (2006b) for sorption of Sb(V)onto goethite. The results were obtained from a sediment sample which initially contained34 mg kg-1 Sb and a tailing sample which initially contained 320 mg kg-1 Sb . For both sam-ples about 90 % was present as Sb(V) and 10 % as Sb(III). As can be seen on the graph, thesedata fit well into the edge presented by Leuz et al. (2006b), although only small pH rangeswere considered. Furthermore it has to be noted the Fe contents of the sediment (9 g kg-1)and the tailing sample (34.7 g kg-1) differed significantly. The agreement between the twodata series could indicate that the sorption model based on goethite is a good approximationof sorption onto river sediments.

Figure 4.3: Data points of Verbrugghe (2016) for sorption of Sb onto sediments and data of Leuz et al. (2006b) forsorption of Sb(V) onto goethite.

The partition coefficients for sorption of Sb(V) onto HFO were calculated by means of equa-tion 3.4 and are presented in table 4.1. For a sediment concentration of 1.0 kg L-1, no valueswere included for pH values below 5, since for these pH values no convergence could be ob-tained with the PHREEQC model. Possibly the problem arose due to the very high sorptionand consequently the very low dissolved Sb concentration values for these pH values, sincecalculations with very small concentrations can be problematic in PHREEQC. As observedin figure 4.2.A, sorption decreases with increasing pH and with decreasing sediment concen-tration. Maximal sorption is observed around pH 3.5, which is in correspondence with whatTighe et al. (2005) observed for Sb(V) sorption onto Fe(OH)3.

The obtained values are consistent with the values reported by other researchers, althoughthose cover a very large range. For example, Wang et al. (2011) reported values ranging from65 to 6.4 · 104 L kg-1 for suspended particulate matter while Resongles et al. (2013) measured


Table 4.1: KD values (L kg-1) of Sb(V) for a range of pH values and different sediment concentrations: 1.0 kg L-1,0.10 kg L-1, 0.04722 kg L-1, 8.0 · 10−5 kg L-1. Conditions: log pO2 = -0.68, [Sb(V)]0 = 4.15 µM.

pH KD for KD for KD for KD for1.0 kg L-1 solids 0.1 kg L-1 solids 0.04722 kg L-1 solids 8.0 · 10−5 kg L-1 solids

(-) (L kg-1) (L kg-1) (L kg-1) (L kg-1)3.03.54.05.06.07.08.09.0

10.011.0

3.46 · 104

6.77 · 103

9.70 · 102

1.16 · 102

1.37 · 101

1.863.23 · 10−1

5.54 · 104

7.39 · 104

6.91 · 104

2.88 · 104

6.44 · 103

9.56 · 102

1.15 · 102

1.36 · 101

1.833.07 · 10−1

5.00 · 104

6.90 · 104

6.61 · 104

2.83 · 104

6.36 · 103

9.46 · 102

1.15 · 102

1.36 · 101

1.833.06 · 10−1

1.92 · 103

2.18 · 103

2.34 · 103

2.21 · 103

1.46 · 103

5.66 · 102

1.06 · 102

1.36 · 101

1.833.05 · 10−1

KD values between 2.5 · 102 L kg-1 and 7.9 · 104 L kg-1. The IAEA (2010) reported valuesranging from 5.5 · 102 to 4.6 · 104 L kg-1 for freshwater systems. The large ranges in whichthe values vary are a consequence of the strong pH-dependency of the sorption process andthe influence of some parameters, e.g. dissolved Sb and sediment concentration, of whichthe values were not reported. In the WASP model a value of 124 L kg-1 is used, whichcorresponds to the values obtained for a pH slightly lower than 8. This is reasonable sinceneutral to slightly alkaline pH values are common in freshwater systems (Wang et al., 2011;Routh and Ikramuddin, 1996). The obtained values are of the same order of magnitude for thedifferent sediment concentrations, except for acidic pH values and a sediment concentrationof 8.0 · 10−5 kg L-1. In this case, the number of available binding sites is limiting.

Table 4.2 presents the KD values for sorption of Sb(III) onto HFO considering a range of pHvalues and different sediment concentrations. Values are of the same order of magnitude forpH values within the range of 5 to 10, with maximal sorption around pH 8. For pH valuesbelow 5, a pH-dependent trend of decreasing sorption with decreasing pH exists, similar towhat can be observed in figure 4.2.B. Also for a pH above 10 sorption slightly decreases.

The obtained values all fall within the range of 0.6 - 2.1 · 103 L kg-1, which is the expectedrange of KD values for the soil solid phase to the soil liquid phase in temperate environmentsas reported by the IAEA (2010). As stated earlier, Sb(III) is more likely to be present underthese conditions. Sediment concentration has a small influence for concentration values of0.10 kg L-1 and higher, but for a sediment concentration of 8.0 · 10−5 kg L-1 values areremarkably lower. This can again be explained by the limited number of binding sites for lowsediment concentrations.

40

Table 4.2: KD values (L kg-1) of Sb(III) for a range of pH values and different sediment concentrations: 1.0 kg L-1,0.10 kg L-1, 0.04722 kg L-1, 8.0 · 10−5 kg L-1. Conditions: log pO2 = -70, [Sb(III)]0 = 2.2 µM.

pH KD for KD for KD for KD for1.0 kg L-1 solids 0.1 kg L-1 solids 0.04722 kg L-1 solids 8.0 · 10−5 kg L-1 solids

(-) (L kg-1) (L kg-1) (L kg-1) (L kg-1)3.04.05.06.07.08.09.0

10.011.0

1.47 · 103

1.64 · 103

1.69 · 103

1.70 · 103

1.69 · 103

1.65 · 103

1.43 · 103

3.16 · 103

9.16 · 103

1.42 · 103

1.63 · 103

1.69 · 103

1.69 · 103

1.69 · 103

1.65 · 103

1.46 · 103

3.03 · 103

9.04 · 103

1.41 · 103

1.63 · 103

1.69 · 103

1.69 · 103

1.69 · 103

1.65 · 103

1.46 · 103

2.71 · 102

8.00 · 102

1.24 · 103

1.43 · 103

1.48 · 103

1.49 · 103

1.49 · 103

1.45 · 103

1.29 · 103

4.3 Transport scenarios

4.3.1 Pure advective transport

As a benchmark, pure advective transport, in the absence of any retarding agent, was sim-ulated. In this case, the Sb species are transported at the water flow velocity. In figure 4.4,the resulting dissolved Sb concentrations over the river stretch are shown 5, 15, 25 and 35time steps after the start of the simulation. A pulse of 5 time steps or 25 minutes, containing100 µM Sb, was entered into the system at the start of the simulation. A pH of 7 and a pO2of 10−45 atm were imposed. The pulse breaks through without any retardation and leavesthe system after 40 time steps or about 3 hours and 30 minutes.

Figure 4.4: Breakthrough curves of an advective pulse front with a Sb concentration of 100 µM Sb after 5, 15, 25 and35 time steps (25 min, 1 h 20 min, 2 h 10 min and 3 h). The duration of the pulse is 5 time steps or about 25 min.

Conditions: pH = 7, log pO2 = -45.

This figure also reflects the aqueous speciation of Sb as presented in figure 4.1. For a pH of


7 and a log pO2 of -45, which corresponds to an Eh of 0.1491 V, about 10 % of the total Sbconcentration is present as Sb(III) while about 90 % is present as Sb(V). In contrast to whatSanders (1998) stated, Wang et al. (2011) reported pH values between 6.5 and 8.6, and Ehvalues between 0.1128 and 0.2421 V in water samples of three rivers in the proximity of the Sbmine in Xikuangshan, China. These conditions correspond to partial oxygen pressures around10−45 atm, assuming that the reduction/oxidation of water is the dominant redox reactionin the system. Wang et al. (2011) reported Sb speciation in water samples ranging from nosignificant amounts of Sb(III) to 27 % as Sb(III) and 73 % as Sb(V), while Verbrugghe (2016)found that in sediments from the Wushui river 90 % of the Sb content was present as Sb(V)and 10 % as Sb(III).

Serafimovska et al. (2011) measured Sb(V) to total Sb ratios in the range of 62 to 98 %in river waters in Bulgaria and Macedonia. They suggested the presence of Sb(III) can belinked to higher concentrations of organic matter in industrially polluted regions, with theprobability of organic ligands forming Sb(III) complexes and thus stabilising this oxidationstate. In the results presented by Wang et al. (2011) no remarkable link is observed betweenthe TOC content and the presence of Sb(III), but the elevated Sb(III) concentrations couldcertainly point out an environmental influence since the sampling points with high Sb(III)contents were all located close to the mining areas. The obtained speciation, as presentedin figure 4.4, thus falls within the range of reported Sb(III) and Sb(V) contents, althoughthis range is large. Moreover it should be noted that in reality the resulting dissolved Sbconcentrations will be lower than those obtained in this simulation. Not only sorption ontoFe-bearing phases and organic matter will occur, also (co-)precipitation and complexation,e.g. with organic ligands, are possible.

4.3.2 Advective transport in a perfectly mixed water column with pH-independent sorption

Next, advection was simulated in the presence of a surface assemblage Sorb, onto which pH-independent sorption of Sb, with a partition coefficient of 124 L kg−1, takes place. Consideringa step input of 1 µM Sb, the dissolved and sorbed Sb concentrations over the river stretch100, 300 and 500 time steps after the start of the simulation are presented in figure 4.5. Overthe whole river system, the pH was kept equal to 7 and the pO2 equal to 0.21 atm. Comparedto the advective front resulting from advection in the absence of a solid phase (cfr. figure 4.4),the advective front presented in this figure takes a lot longer to break through. After 337 timesteps or about 29 h, the dissolved Sb concentration in cell 28 equals 0.5 µM while in the caseof pure advection it only took 28 time steps (2 h 25 min) to reach cell 28. In other words, theretardation factor corresponding to the partition coefficient of 124 L kg-1, calculated basedon the differences in flow velocity, equals 12.04. This value deviates from the value calculatedby means of equation 3.5. Given a sediment bulk density of 0.10 kg L-1, a porosity of 0.9622

42

and a partition coefficient of 124 L kg-1, this formula returns a retardation factor of 13.90.However, also the resulting partition coefficient between the solid and the dissolved phase,equalling 111.9 L kg-1, deviates from the value imposed. This difference could be attributedto numerical inaccuracy of the model with regard to the way the surface Sorb was defined,and is probably the cause of the deviation of the retardation factor. Indeed, the retardationfactor calculated by means of equation 3.5 based on this value equals 12.63.

Figure 4.5: Breakthrough curves of an advective step front with an Sb concentration of 1 µM Sb after 100, 300 and500 time steps (8 h 40 min, 26 h, 43 h 20 min) in the presence of a solid phase with pH-independent sorption

(0.10 kg L-1 sediment, KD = 124 L kg-1. Conditions: pH = 7, log pO2 = -0.68.

Due to the way the surface assemblage Sorb was defined, the sorption process onto thissolid phase is pH-independent. This is illustrated on figure 4.6, which shows the sorbed Sbconcentrations in the river system 300 time steps after a 100 time steps during pulse containing100 µM Sb was sent into the system, for different pH values. A partial oxygen pressure of0.21 atm was imposed.

Figure 4.6: Sorbed Sb concentrations over the river stretch 300 time steps (26 h) after a pulse containing 100 µM,with a duration of 100 time steps (8 h 40 min), was sent into the system. Conditions: pH 4, 7, 10; log pO2 = -0.68.


As can be seen on this figure, concentrations are equal for all three values considered. Ofcourse, this is a strong simplification of reality since the sorption process onto Fe oxyhydrox-ides has been reported to be strongly pH-dependent (Leuz et al., 2006b; Guo et al., 2014).

4.3.3 Advective transport in a perfectly mixed water column with pH-dependent sorption

A closer approximation of reality is obtained when the transport simulation is done in presenceof the HFO surface assemblage, to which equation 4.2, 4.3 and 4.4 apply. In figure 4.7, thesorbed and dissolved Sb concentrations over the river stretch are presented 100, 300 and 500time steps after the start of the simulation. A step input of 1 µM Sb was sent into the system,a pH of 8 and a pO2 of 0.21 atm were imposed. It was assumed that the water column isperfectly mixed and the solid phase is static, i.e. it is not transported with the solution.The similarities with figure 4.5 are large, which is to be expected since there is only a smalldifference between the partition coefficient of HFO at pH 8 in aerobic conditions and the KD

value used in the WASP model.

Figure 4.7: Breakthrough curves for a step input containing 1 µM Sb after 100, 300 and 500 time steps (8 h 40 min,26 h, 43 h 20 min) in a perfectly mixed water with 0.10 kg L-1 solids, containing 10.59 g kg-1 HFO. Conditions:

pH = 8, log pO2 = -0.68.

However, the sorption behaviour of Sb onto HFO is strongly pH-dependent. This is illustratedon figure 4.8, which shows the sorbed Sb concentrations for different pH values in the riversystem 3000 time steps or about 11 days after the start of the simulation with a step inputof 1 µM Sb and a constant pO2 of 0.21 atm. Under this partial oxygen pressure, Sb(V) isthe dominant species. The Sb front has passed the whole river stretch for pH values of 8 andabove, while for a pH of 5 or below it has only migrated for about 2 km. This is a consequenceof the sorption capacity of the solid phase: sorption is highest at pH 5, while at pH 9 and 11almost no sorption occurs. The largest differences occur between pH 6 and pH 9, which arethe pH values situated on the steepest part of the sorption edge (cfr. figure 4.2).

44

Figure 4.8: Sorbed Sb concentrations for different pH values, 3000 time steps (260 h or about 11 days) after the startof the simulation in which a step input containing 1 µM Sb was sent into the system. Conditions: log pO2 = -0.68.

Some important differences with the sorption edge (figure 4.2.A) and partition coefficients(table 4.1) for Sb(V) should be noted. First of all, the sorption capacity of the solid phaseis higher at pH 5 then at pH 3. Furthermore, the partition coefficients resulting from thistransport simulation are remarkably lower than the ones presented in table 4.1, with increasingdifferences for decreasing pH values. For pH values of 8, 9 and 11 the KD values calculatedbased on the transport simulation are respectively 113.6, 13.66 and 0.3021 L kg-1. These valuescorrespond well to the values presented in table 4.1. However, for pH values of 3, 5 and 7 theKD values equal respectively 5213, 5748 and 3076 L kg-1, which is about a factor 10, a factor4 and a factor 2 smaller than the values based on the sorption edge. This difference cannotbe caused by the transport process, since instantaneous equilibrium is assumed. However,there is an important difference in concentration: the values presented in table 4.1 apply for atotal concentration of 4.15 µM , while the values calculated based on the transport simulationapply for a dissolved concentration of 1 µM and a variable total concentration. For pH valuesof 5 or below, this total concentration is higher than 500 µM. Thus, this deviation from thevalues presented in table 4.1 is an indication of non-linear sorption which is more pronouncedat lower pH values. This observation corresponds with what Zhang et al. (2014b) found, i.e.that adsorption of Sb onto soils is non-linear. It can furthermore explain the large rangeof KD values reported in literature, since the degree of contamination, i.e. the dissolved Sbconcentration, will thus have an important influence on the relative amount sorbed.

The resulting concentrations are furthermore influenced by the prevailing partial oxygen pres-sure. Figure 4.9 presents the sorbed Sb concentrations in the river system 3000 time steps(11 days) after the start of a simulation with a step input of 1 µM Sb, for different pH values.A pO2 of 10−45 atm was imposed. Under this partial oxygen pressure, Sb(III) and Sb(V) arepresent in comparable amounts. Considering the whole pH range, maximal sorption is lessthan under a partial oxygen pressure of 0.21 atm, with a maximal sorption of 2594 µmol kg-1


at pH 6, compared to 5748 µmol kg-1 at pH 5 under a partial oxygen pressure of 0.21 atm.This also results in a lower retardation. However, for pH values above 7, a higher fractionof the antimony present will be sorbed. Where under a partial oxygen pressure of 0.21 atmsorption in general decreased with increasing pH, sorption is highest for slightly acidic toneutral pH values under these conditions. The difference between the two partial oxygenpressures is highest for pH 3, which is to be expected given the sorption behaviour of Sb(III)at this pH value (cfr. figure 4.2.B).

Figure 4.9: Sorbed Sb concentrations for different pH values, 3000 time steps (260 h or about 11 days) after the startof the simulation in which a step input containing 1 µM Sb was sent into the system. Conditions: log pO2 = -45.

Given the fact that incidental discharges occur, it is interesting to see how the solid phaserecovers from a pulse input. Figure 4.10 and 4.11 present the sorbed Sb concentrations in theriver system for different pH values respectively 100 and 1000 time steps after the start of asimulation with a pulse input of 100 time steps, containing 100 µM Sb. A pO2 of 10−45 atmwas imposed. Figure 4.10 illustrates how far the front has proceeded at the end of the pulse.Only for pH 9 and 11 the front has reached the end of the river system. Compared to thesorbed concentrations resulting from a 1 µM Sb step input (figure 4.9), the partition betweenthe dissolved and the solid phase tends more towards the dissolved phase. Again this is aconcentration effect indicating non-linear sorption.

As can be seen on figure 4.11, the rate at which the system recovers from the contaminatingfront strongly depends on pH. After about 4 days, no traces of the front are visible anymorein alkaline conditions while in slightly acidic conditions (pH 5, 6) almost all sorbed Sb is stillsituated in the first half of the transect. The concentration curves have a geometry skewedtowards the right, due to the fact that the clean incoming solution is equilibrated with eachsegment during the simulation and thus becomes contaminated. Consequently, an equilibriumwhich tends more towards the solid phase prevails in the cells which are located further awayfrom the start of the transect.

46

Figure 4.10: Sorbed Sb concentrations for different pH values, 100 time steps (8 h 40 min) after the start of thesimulation in which a pulse input containing 100 µM Sb was sent into the system. Conditions: log pO2 = -45.

Figure 4.11: Sorbed Sb concentrations for different pH values, 1000 time steps (about 4 days) after the start of thesimulation in which a pulse input containing 100 µM Sb was sent into the system. Conditions: log pO2 = -45.

Under a partial oxygen pressure of 0.21 atm, sorption in acidic conditions is stronger (cfr.figure 4.2) and thus recovery even slower. For example, given a pH of 6 it takes about 45 daysbefore the sorbed concentration over the whole stretch drops below 100 µmol kg-1 under apartial oxygen pressure of 10−45 atm after a 100 µM pulse of about 9 hours, while under apartial oxygen pressure of 0.21 atm this takes about 64 days. The corresponding dissolvedconcentrations are respectively 0.30 µM (or 36.5 µg L-1) and 0.19 µM (or 23.4 µg L-1),which is still largely above the EU drinking water limit of 5 µg L-1 (Counsil of the EuropeanCommunities, 1998). Given a pH of 7, the recovery rate is similar for both partial oxygenpressures. It takes about 18 days for the dissolved Sb concentration to drop below the drinkingwater limit of 5 µg L-1 under a partial oxygen pressure of 0.21 atm, while under a partialoxygen pressure of 10−45 atm this takes about 19 days.


It is important to note that this simulation scenario relies on some heavy simplifications. Firstof all, it was assumed that the suspended matter present in the surface part and the sedimentspresent in the benthic part are perfectly mixed over the water column. In reality, the amountof solid phase in direct contact with the advective flow will be smaller. Secondly, instantaneouschemical equilibrium was assumed. In practice, the kinetics of the sorption process will be animportant influencing factor. Due to the water flow, no sorption equilibrium will be reachedand both sorption and desorption will be slower than calculated by this model. Thus, thismodel probably overestimates the retention of Sb by the solid phase. Indeed, Verbrugghe(2016) reported a Sb content of 34 mg kg-1 for sediments collected from the Wushui riverwith a pH of about 6.5, while the simulated scenarios resulted in sorbed concentrations of115.6 mg kg-1 (pH 7, log pO2 -45) to 365.3 mg kg-1 (pH 6, log pO2 -0.68) for pH values between6 and 8 and log pO2 values between -0.68 and -45. However, a dissolved Sb concentrationpresent was not reported and based on the sorbed Sb concentrations measured by Wang et al.(2011), i.e. 57.11 to 7316 mg kg-1 for sediments and 30 to 11,888 µg kg-1 for suspendedmatter, it could be stated that the concentrations resulting from the simulations fall withinthe range of values reported in literature.

In periods of elevated pollution, e.g. due to an accidental discharge or intense precipitationand the consequent resuspension of sorbed pollutants, local water managers increase the pHof the Wushui river system. Based on the obtained simulation results, it could be concludedthat this measure is effective in decreasing the amounts sorbed and consequently the impactof the pollution incident on the long term. However, the effect of a pH increase on e.g. biotashould be taken into consideration as well.

4.3.4 Advective transport with diffusive exchanges with the benthic seg-ments

A water column typically exists out of a surface part in which suspended matter is presentand a benthic part with a high sediment content. The latter could also be interpreted as akind of water soil, which is completely saturated and has a high porosity. In the framework ofthis research, a water soil with a thickness of 0.1 m and a porosity of 0.6226 was considered.A simulation was done considering this benthic part and assuming no solid phase in thesurface part, as conceptually presented in figure3.1.C. Figure 4.12 shows the resulting Sbconcentrations in both the surface and the benthic part 5000 and 10000 time steps after astep input was sent into the system with a pH of 7. In the assumption not a lot whirling upoccurs and the benthic part is completely anoxic, partial oxygen pressures of 0.21 atm and10−70 atm were imposed for respectively the surface and the benthic part. Due to diffusivetransfer, the benthic part acts as an antimony sink. The dissolved Sb concentration in thispart remains close to 0, while the sorbed concentration increases slowly but continuously.The dissolved Sb concentration in the surface part remains close to 1 µM. After 5000 time

48

steps or about 18 days, the sediments have adsorbed about 1.155 µmol kg-1 (0.14061 mg kg-1)and after 10 000 time steps this has increased up to 2.317 µmol kg-1 (0.2821 mg kg-1). Noremarkable pH-effect occurs under these conditions, which is due to the fact that Sb(III)predominates in the benthic part.

Figure 4.12: Sb concentrations 5000 and 10 000 time steps (18 and 36 days) after the start of the simulation with astep input of 1 µM Sb. Conditions: pH = 7, log pO2 = -0.68 in surface part, log pO2 = -70 in benthic part.

The sorbed concentrations resulting after 10 000 time steps or about 36 days are low comparedto sorbed concentrations measured in the sediments of river systems. Wang et al. (2011)reported Sb contents in the sediments within the range of 57.11 to 7316 mg kg-1 and contentsup to 350 mg kg-1 were reported in sediments from the Wushui river and its tributaries(Zhang et al., 2014a). The sediment sample from the Wushui river analysed by Verbrugghe(2016) contained 34 mg kg-1 Sb. However, this simulation started from clean sediments andsorption increased linearly with time, as shown in figure 4.13. Given the rate at which thesediments adsorb within the timespan of the simulation, i.e. 7.755 µg day-1, it would takeapproximately 123 years to reach a sorbed concentration of 350 mg kg-1. This is a simplifiedapproximation, not taking into account kinetics, aeration, sedimentation or resuspension ofsediments nor any decrease in adsorption rate due to saturation. Furthermore, a constantdissolved Sb concentration is assumed while in reality this value will fluctuate, but it givesan indication of how important historical contamination is.

In order to estimate how long it takes for contaminated sediments to recover, a simulationwas done in which the initial Sb content of the sediments in the benthic part was set to2.873 · 103 µmol kg-1 or 350 mg kg-1. A step input with pure water was imposed, so desorptionwas provoked. However, as can be seen on figure 4.14, this process is very slow. This graphpresents the evolution of the amount of Sb sorbed onto the solid phase in time, which decreaseslinearly during the simulation with a rate of 15.17 µg day-1. The dissolved Sb concentrationin the benthic part initially is 1.932 µM and also decreases very slowly to a value of 1.929 µM


Figure 4.13: Evolution of the Sb concentration over time in both the surface and the benthic part of cell 10,2.375 km far in the stretch, given a step input of 1 µM Sb. Conditions: pH = 7, log pO2 = -0.68 in surface part,

log pO2 = -70 in benthic part.

after 36 days. The dissolved Sb concentration in the surface part remains constant at about2.807 · 10−3 µM or 0.3418 µg L-1, and thus does not exceed the drinking limit of 5 µg L-1

imposed by the EU (Counsil of the European Communities, 1998). Assuming linear desorptionwould continue to occur at the same continuous rate as during the simulation, it would takeabout 63 years for the sediments to be ’clean’ again. This is an underestimation since thekinetics of the reaction will be limiting. It was even stated that adsorption of Sb onto soilsis irreversible (Zhang et al., 2014b). Furthermore the desorption rate will decrease withdecreasing Sb contents in the sediments and the inflowing dissolved Sb concentration is likelyto be above 0.

Figure 4.14: Evolution of the sorbed Sb concentration over time in the benthic part of cell 10, 2.375 km far in thestretch, given an initial sorbed Sb concentration of 2.873 · 103 µmol kg-1 and an inflow containing no Sb. Conditions:

pH = 7, log pO2 = -0.68 in surface part, log pO2 = -70 in benthic part.

50

4.3.5 Advective transport in the presence of mobile suspended matter

In all previous scenarios it was assumed that the solid phase is stagnant, i.e. is not transportedwith the solution. However, in river systems, part of the solid phase is present as suspendedmatter which is transported due to the force of the flowing water. A simulation was donein which the water column contains 80 mg L-1 of suspended matter with a HFO content of10.59 mg kg-1, that is transported at the same advective rate as the solutes. It was assumed noresuspension or sedimentation occurs. Figure 4.15 presents the Sb concentrations in the riversystem 10 and 130 time steps after a pulse containing 100 µM Sb, with a duration of 100 timesteps, was sent into the system with a pH of 7 and a pO2 of 0.21 atm. These concentrationsshow a lot of similarities with the concentrations resulting from advective transport withouta solid phase (cfr. figure 4.4). The only difference is the presence of suspended matter, whichacts as a minor sink. Consequently, the resulting dissolved Sb concentration equals 99.38 µMinstead of 100 µM. The influence of the colloids is thus very small, due to the low amountof binding sites for such a small concentration of the solid phase. Similar to previous resultssorption is pH-dependent, with e.g. 7759 µmol kg-1 sorbed at pH 7 and 9550 µmol kg-1 atpH 3. However, the influence on the resulting dissolved concentration is small (99.38 µM atpH 7 versus 99.23 µM at pH 3).

Figure 4.15: Sb concentrations in the river system 10 and 130 time steps (50 min and 11 h 15 min) after the start ofthe simulation. A pulse input of 100 time steps (8 h 40 min), containing 100 µM Sb, was sent into the system.

Conditions: pH = 7, log pO2 = -0.68.

Wang et al. (2011) reported Sb contents of particulate matter varying between 30 and11, 888 µg g-1, with a median of 229.5 µg g-1, throughout the three rivers from which theysampled. As a measure of comparison, the sorbed Sb concentration in equilibrium with adissolved concentration of 10 µg L-1, a value which they measured in 4 of the samples, wascalculated. The corresponding Sb contents of the particulate matter reported by Wang et al.(2011) equal 173, 214, 50 and 78 µg g-1, while a value of 90.75 µg g-1 was obtained by themodel. Thus, the model prediction falls within the range of measured values, although some


important parameters like the suspended matter content were unknown.

Of course, also this scenario is a simplification of reality. In practice, both settlement andresuspension can occur so the solid phase can act as source or a sink for the dissolved phase.The extent to which these processes occur depends on the water flow velocity and the size ofthe particles (cfr. figure 2.5).

4.4 Method choice, model simplifications and shortages

4.4.1 Method choice

When considering the results presented in the above sections, it has to be noted that re-active transport modelling in river systems is a rather rare application of PHREEQC. ThePHREEQC model is typically used for reactive transport modelling of groundwater flow orfor simulating laboratory tests like leaching tests. For example, Genc-Fuhrman et al. (2004)used PHREEQC to predict As(V) adsorption onto activated neutralized red mud, Halim et al.(2005) used it to model leaching of Pb, Cd, As and Cr from cementitious waste, Ettler et al.(2004) modelled the leaching of lead from metallurgical slag in citric solutions by means ofPHREEQC and Liu et al. (2010) used it to evaluate the Sb speciation in the water samplesthey analysed. However, no example was found in which PHREEQC was used to modelreactive transport in a water stream. The results presented in the previous sections indicatethe possibilities of the PHREEQC transport module, although it is not optimal to includedynamic hydraulic properties. For example, hydraulic aspects like the cross-sectional area,bottom roughness, slope, etc. cannot be considered by means of the PHREEQC transportmodule.

Besides PHREEQC, other models could have been used as well. For example, two applicationsof the OTEQ model (cfr. section 2.4) for reactive transport in water streams have beenpresented on respectively the 5th International Conference on Acid Rock Drainage (ICARD2000) and the 10th International Symposium on Water-Rock Interaction (Ball et al., 2000,2001). Caruso (2005) on the other hand, used the WASP 5 model to simulate metal transportin a mining-impacted stream, with the purpose of developing metals total maximum dailyloads (TMDL) and evaluate different remediation alternatives. However, he considered As,Cd, Cu, Pb and Zn, all metals of which the sorption behaviour is rather pH-independent inthe range of freshwater pH values. As mentioned before, Sb(V) shows the largest changes insorption behaviour between pH 6 and pH 9, which corresponds more or less to the range ofenvironmentally relevant pH values in freshwater systems. Caruso (2005) indicated himselfthat the use of lumped KD-values to simulate the distribution of metals between the dissolvedand particulate phases is only justified in cases where metal sorption is pH-independent inthe pH range considered. He also indicated more detailed geochemical modelling is needed

52

for cases to which this does not apply. In response to this shortage, the US EPA developedthe META4 module for WASP. This module is described as capable of simulating equilibriumreactions and slower kinetic processes, including metals adsorption/desorption, precipitation,ion exchange and complexation. Furthermore it is stated that environmental controls onthese processes may be included, such as site-specific mineral and sediment characteristics,and concentrations of iron oxyhydroxides for sorption of metals and pH. This module seemsto have all the tools needed to properly simulate the reactive transport of heavy metals likeSb, but is unfortunately not available in the general WASP release.

4.4.2 Model simplifications and shortages

A number of simplifications were made in the purpose of this research, some of them dueto the lack of data, some of them due the limited possibilities of PHREEQC. First of all,a geochemical equilibrium model like PHREEQC requires detailed geochemical informationlike pH and solution composition which, in the framework of this research, was not available.Thus, a basic scenario was elaborated in which the solution only contained Sb and 0.01 MNaCl as background electrolyte. In reality, the water column will have a much more complexand variable composition. Wang et al. (2011) reported variable amounts of TOC, sulphates,chlorides, iron and manganese in a comparable river system. These are all parameters whowill influence the geochemical behaviour of antimony due to e.g. complexation, precipitationand catalysis of the oxidation process. Furthermore, both the pO2 and the pH were fixatedover the whole river stretch, while in reality these parameters will be both time and spacedependent. For example, pH will vary due to changes in dissolved Sb concentration or othervariations in solution composition.

Secondly, considerable simplifications were made with regard to the considered processes.All kinetic aspects with regard to surface complexation onto Fe oxyhydroxides were ignored,although the implementation of kinetic reactions is possible in PHREEQC, and instantaneousequilibrium was assumed. However, in the case of flowing water, this equilibrium is mostlynot reached within the contact time due to kinetic limitations and both accumulation andrelease of the contaminant will be much slower than calculated by this model. Moreover, purehydrous ferric oxide was considered by using the model of Dzombak and Morel (1990). Inreality this mineral will occur imbedded in sediments and thus will be less available, leadingto lower sorption amounts. On the other hand, both surface complexation by humic acidsand Al and Mn were not considered which could lead to underestimations of the sorbed Sbamounts. Also other processes and interactions, e.g. (co-)precipitation and competition forbinding sites with other ions, were not considered.

With regard to physical processes, both erosion, sedimentation and resuspension were notstudied. Under the flow conditions which prevail in the Wushui watershed, probably mostly


erosion occurs. Moreover, hydraulic properties were not considered and for all simulations, aconstant flow velocity was assumed over the entire simulation time. It is very unlikely to havea constant flow for periods up to 10 days under natural conditions, except in precipitationfree periods. However, the transport simulations elaborated in the framework of this researchserved mostly as an illustration of how the sorption model could be incorporated in a transportmodule.

Another aspect that was not considered in the framework of this research is the interactionwith microbiota. Antimony is a non-essential element and toxic to most organisms at elevatedlevels. Biomethylation is the only in-vivo reaction process of Sb which has been reported anddescribed with analytical confidence. Very little information exists on other processes likeredox reactions and biosorption reactions (Filella et al., 2007). Throughout recent years, re-search has been going on with regard to biosorption of antimony. Madrid et al. (1998) forexample investigated biosorption of Sb by cyanobacteria Spirulina platensis. Wu et al. (2012)used cyanobacteria Microcystis for biosorption of Sb(III) from aqueous solutions, while Sunet al. (2014) investigated the biosorption properties of Sb(V) by cyanobacteria Microcystisunder environmentally relevant conditions. However, these interactions were mostly investi-gated with the purpose of removing Sb from effluents and are probably less relevant undernatural conditions. Filella et al. (2007) also stated that, considering how low the concentra-tions of methylated antimony compounds detected in the environment are, it seems unlikelythat they would be of great concern from an environmental perspective.

Despite all these simplifications and assumptions, the obtained results are in line with valuesreported in literature, although those cover a large range. However, no real validation couldbe performed due to the lack of data. This is probably the most important shortage ofthe obtained model. Since no data concerning the Wushui watershed were provided, valuesreported by other studies were used as a means of comparison. However, there were alwayssome parameters lacking to allow a full comparison.

5 | Conclusion and perspectives

The purpose of this research was to investigate the possibilities of geochemical models toimprove reactive transport modelling of heavy metals in polluted river systems, with the focuson antimony. This metal appeared to exhibit a strong pH-dependent sorption behaviour, withpartition coefficients decreasing with two orders of magnitude for a pH increase from 6 to 9.

Antimony speciation reactions were modified in order to obtain a speciation model in linewith those presented in literature. Speciation reactions resulting in less relevant species wereremoved from the database in order not to make the model too complicated. A model forsorption of Sb onto HFO was set up by means of the geochemical model PHREEQC andparametrised based on experimental data for sorption of Sb onto goethite, obtained by Leuzet al. (2006b). These experimental data were approached rather well, with RMSE values of3606 % and 2 % for respectively Sb(V) and Sb(III). The elevated RMSE for Sb(V) sorptionis caused by the deviations for pH values above 9. For these pH values, the sorption modelof Leuz et al. (2006b) showed a change in trend which could not be modelled by the DLMused in the framework of this research. This range of pH values is however less relevant forfreshwater conditions. Given a log pO2 of -0.68 and a dissolved Sb concentration of 1 µM, themodel led to KD values ranging from 114 L kg-1 to 3076 L kg-1 in the pH range relevant forfreshwater conditions. These values correspond to KD values reported in literature, althoughthose vary in a large range. Both the strong pH-dependency and the non-linearity of Sbsorption, as well as the difference in sorption behaviour between the two valence states, canbe the cause of these large variations.

The obtained sorption model was linked to the transport module imbedded in PHREEQC anddifferent scenarios were simulated. As a benchmark, advective transport with pH-independentsorption was simulated. Next, advective transport with pH-dependent sorption was simulatedin a perfectly mixed water column and the obtained results were compared with the resultsof the benchmark scenario. This comparison showed the importance of considering the pH-dependent sorption behaviour of Sb, with sorbed amounts ranging from 4.788 mg kg-1 (pH8.5) to 211.9 mg kg-1 (pH 6.5) within the range of typical pH values in freshwater streams asstated by Sanders (1998), given a perfectly mixed, oxic water column with a solids content of0.10 kg L-1 in which 1 µM dissolved Sb is present. In comparison, the WASP equivalent with

55

56

a fixed partition coefficient resulted in a sorbed amount of 13.83 mg kg-1 and significantlyunderestimated retardation in slightly acidic to neutral conditions. Some other scenarios wereelaborated in the purpose of obtaining a closer approximation of reality. Diffusive exchangesbetween the water column and benthic sediments were simulated, as well as colloid transport.These simulations led to the conclusion that particulate matter only has a small influenceon dissolved concentrations while benthic sediments have very high sorption capacities butaccumulation and release of pollutants are very slow processes.

However, for all elaborated scenarios some important assumptions and simplifications weremade. Some processes, like resuspension and precipitation, were not considered and someparameters, like pH and flow velocity, were given a constant value while in reality they areboth time and space dependent. Despite these simplifications, the obtained results were inline with measurements reported in literature. In comparison to the WASP model, a closerapproximation of reality was obtained with regard to geochemical processes while WASPperforms better with regard to physical aspects. Thus, it seems a trade off has to be madebetween these two types of processes. In river systems, physical properties and processes aretypically of great importance. However, the results obtained in this research indicate thatit would be a (too) strong simplification not to consider the complexity of the geochemicalbehaviour of antimony.

For further improvements, it would be interesting to see how other models like OTEQ or theWASP META4 module perform. In addition, some aspects of PHREEQC could be furtherelaborated. The model allows to include kinetic reactions, which could be used to implemente.g. kinetic (de)sorption or resuspension/sedimentation. It could also be interesting to explorethe possibilities of linking PHREEQC to a solute transport model, as it was done in OTEQ.

The aspect which should probably receive most attention to enable further improvements,is data acquisition. Since no data were available concerning the Wushui watershed, theconstructed model could not be validated. The lack of data, and the lack of consistency in theavailable data, is something which is mentioned as a bottleneck in many studies concerningantimony. Particularly studies performed under conditions relevant to freshwater systemsare scarce (Filella et al., 2002b). More data would enhance both model validation and theunderstanding of the occurring mechanisms. In the framework of this research, a completedataset would consist out of water flow data, solution composition data, including pH andEh, and data concerning the solid phase present, i.e. suspended matter content, porosity ofthe benthic segments, Fe and OM content, sorbed metal concentrations. In the ideal case, ameasurement campaign of at least a year would be set up in which the mentioned parameterswould be measured on different sampling locations at regular time intervals. This would allowto evaluate variations in time and to link those variations to external influences like dischargesor precipitation events.

Bibliography

Alkhatib, E. and Castor, K. (2000). Parameters influencing sediments resuspension andthe link to sorption of inorganic compounds. Environmental Monitoring and Assessment,65(3):531–546.

Allison, J. D., Brown, D. S., and Novo-Gradac, K. J. (1991). MINTEQA2/PRODEFA2, ageochemical assessment model for environmental systems: version 3.0 user’s manual.

Appelo, C. and Postma, D. (2005). Geochemistry, groundwater and pollution. London :Taylor and Francis, 2nd ed. edition.

Ball, J. W. and Nordstrom, D. K. (1991). User’s manual for WATEQ4F, with revised ther-modynamic data base and test cases for calculating speciation of major, trace and redoxelements in natural waters.

Ball, J. W., Nordstrom, D. K., Runkel, R. L., Icard, and Icard (2000). Reactive and non-reactive transport modeling for wightman fork, summitville mine, colorado: Application ofthe otts/oteq model to a low-flow synoptic study.

Ball, J. W., Runkel, R. L., and Nordstrom, D. K. (2001). Reactive transport modeling athigh-flow - Wightman Fork/Alamosa River, USA. Water-Rock Interaction, Vols 1 and 2.A a Balkema Publishers, Leiden.

Baumgarten, G., Matthies, M., Reiter, B., Scheil, S., and Trapp, S. (1998). Chemodynamicsand environmental modeling : an introduction. Springer, Berlin.

Belzile, N., Chen, Y. W., and Wang, Z. J. (2001). Oxidation of antimony(III) by amorphousiron and manganese oxyhydroxides. Chemical Geology, 174(4):379–387.

Bethke, C. M. and Yeakel, S. (2016). GWB - Reactive transport modeling guide. AqueousSolutions LLC.

Buschmann, J., Canonica, S., and Sigg, L. (2005). Photoinduced oxidation of antimony(III)in the presence of humic acid. Environmental Science & Technology, 39(14):5335–5341.

57

58

Buschmann, J. and Sigg, L. (2004). Antimony(III) binding to humic substances: influence ofpH and type of humic acid. Environmental Science & Technology, 38(17):4535–4541.

Caruso, B. S. (2005). Simulation of metals total maximum daily loads and remediation in amining-impacted stream. Journal of Environmental Engineering-Asce, 131(5):777–789.

Caruso, B. S., Cox, T. J., Runkel, R. L., Velleux, M. L., Bencala, K. E., Nordstrom, D. K.,Julien, P. Y., Butler, B. A., Alpers, C. N., Marion, A., and Smith, K. S. (2008). Metalsfate and transport modelling in streams and watersheds: state of the science and USEPAworkshop review. Hydrological Processes, 22:4011–4021.

Centre for Ecology & Hydrology (2016). Windermere Humic Aqueous Model (WHAM). http://www.ceh.ac.uk/services/windermere-humic-aqueous-model-wham [Accessed: febru-ary 2016].

Chen, K. (2000). Modeling the effect of sedimentation on Cesium transport in FourmileBranch. Report, U.S. Department of Energy.

Counsil of the European Communities (1998). Council directive 98/83/ec of 3 november 1998on the quality of water intended for human consumption.

Crecelius, E. A., Bothner, M. H., and Carpenter, R. (1975). Geochemistry of arsenic, anti-mony, mercury and related elements in sediments of puget sound. Environmental Science& Technology, 9(4):325–333.

Cristiano, E., Hu, Y. J., Siegfried, M., Kaplan, D., and Nitsche, H. (2011). A comparison ofpoint of zero charge measurement methodology. Clays and Clay Minerals, 59(2):107–115.

Dixit, S. and Hering, J. G. (2003). Comparison of arsenic(V) and arsenic(III) sorption ontoiron oxide minerals: Implications for arsenic mobility. Environmental Science & Technology,37(18):4182–4189.

Dzombak, D. A. and Morel, F. M. M. (1990). Surface Complexation Modeling: Hydrous FerricOxide. John Wiley & Sons.

Ettler, V., Komarkova, M., Jehlicka, J., Coufal, P., Hradil, D., Machovic, V., and Delorme,F. (2004). Leaching of lead metallurgical slag in citric solutions - implications for disposaland weathering in soil environments. Chemosphere, 57(7):567–577.

Felmy, A., Girvin, D., and Jenne, E. (1984). MINTEQ–A computer program for calculatingaqueous geochemical equilibria.

Fernández-Remolar, D. C. (2011). Iron Oxyhydroxides, pages 855–857. Springer Berlin Hei-delberg, Berlin, Heidelberg.

http://www.ceh.ac.uk/services/windermere-humic-aqueous-model-wham

http://www.ceh.ac.uk/services/windermere-humic-aqueous-model-wham

Bibliography 59

Filella, M., Belzile, N., and Chen, Y.-W. (2002a). Antimony in the environment: a reviewfocused on natural waters. I. occurrence. Earth-Science Reviews, 57:125–176.

Filella, M., Belzile, N., and Chen, Y.-W. (2002b). Antimony in the environment: a reviewfocused on natural waters. II. relevant solution chemistry. Earth-Science Reviews, 59:265–285.

Filella, M., Belzile, N., and Lett, M. C. (2007). Antimony in the environment: A reviewfocused on natural waters. III. microbiota relevant interactions. Earth-Science Reviews,80(3-4):195–217.

Filella, M. and May, P. M. (2003). Computer simulation of the low-molecular-weight inorganicspecies distribution of antimony(III) and antimony(V) in natural waters. Geochimica EtCosmochimica Acta, 67(21):4013–4031.

Filella, M., Philippo, S., Belzile, N., Chen, Y., and Quentel, F. (2009). Natural attenuationprocesses applying to antimony: a study in the abandoned antimony mine in goesdorf,luxembourg. Science of the Total Environment, 407(24):6205–16.

Flakova, R., Zenisova, Z., Sracek, O., Krcmar, D., Ondrejkova, I., Chovan, M., Lalinska,B., and Fendekova, M. (2012). The behavior of arsenic and antimony at Pezinok miningsite, southwestern part of the Slovak Republic. ENVIRONMENTAL EARTH SCIENCES,66(4):1043–1057.

Fu, Z., Wu, F., Mo, C., Deng, Q., Meng, W., and Giesy, J. P. (2016). Comparison of arsenicand antimony biogeochemical behavior in water, soil and tailings from Xikuangshan, China.SCIENCE OF THE TOTAL ENVIRONMENT, 539:97–104.

Gebel, T. (1997). Arsenic and antimony: comparative approach on mechanistic toxicology.Chemico-Biological Interactions, 107(3):131–144.

Genc-Fuhrman, H., Tjell, J. C., and McConchie, D. (2004). Adsorption of arsenic from waterusing activated neutralized red mud. Environmental Science & Technology, 38(8):2428–2434.

Graf, W. (1971). Hydraulics of Sediment Transport. McGraw-Hill, New York.

Groth, D. H., Stettler, L. E., Burg, J. R., Busey, W. M., Grant, G. C., and Wong, L. (1986).Carcinogenic effects of antimony trioxide and antimony ore concentrate in rats. Journal ofToxicology and Environmental Health, 18(4):607–626.

Grund, S. C., Hanusch, K., Breunig, H. J., and Wolf, H. U. (2012). Antimony and AntimonyCompounds, pages 11–42. Weinheim, Germany John Wiley & Sons, Inc., 6th ed. edition.

Guberman, D. (2015). Antimony. U.S. Geological Survey Mineral Commodity Summaries.

60

Guo, X. J., Wu, Z. J., He, M. C., Meng, X. G., Jin, X., Qiu, N., and Zhang, J. (2014). Ad-sorption of antimony onto iron oxyhydroxides: Adsorption behavior and surface structure.Journal of Hazardous Materials, 276:339–345.

Gustafsson, J. (2011). Visual MINTEQ 3.0 user guide.

Guy, A., Jones, P., and Hill, S. J. (1998). Identification and chromatographic separation ofantimony species with alpha-hydroxy acids. Analyst, 123(7):1513–1518.

Halim, C. E., Short, S. A., Scott, J. A., Amal, R., and Low, G. (2005). Modelling the leachingof Pb, Cd, As, and Cr from cementitious waste using PHREEQC. Journal of HazardousMaterials, 125(1-3):45–61.

He, M., Wang, X., Wu, F., and Fu, Z. (2012). Antimony pollution in China. SCIENCE OFTHE TOTAL ENVIRONMENT, 421:41–50.

IAEA (2010). Handbook of parameter values for the prediction of radionuclide transfer interrestrial and freshwater environments. Report, International Atomic Energy Agency.

Ilgen, A. G. and Trainor, T. P. (2012). Sb(III) and Sb(V) sorption onto Al-Rich phases: Hy-drous Al xxide and the clay minerals kaolinite KGa-1b and xxidized and reduced nontroniteNAu-1. Environmental Science & Technology, 46(2):843–851.

Keizer, M. G. and van Riemsdijk, W. H. (2009). ECOSAT: a computer program for thecalculation of speciation and transport in soil-water systems. Department of Soil Quality,Wageningen University.

Kilgour, D. W., Moseley, R. B., Barnett, M. O., Savage, K. S., and Jardine, P. M. (2008).Potential negative consequences of adding phosphorus-based fertilizers to immobilize leadin soil. Journal of Environmental Quality, 37(5):1733–1740.

King, D. W. (1998). Role of carbonate speciation on the oxidation rate of Fe(II) in aquaticsystems. Environmental Science & Technology, 32(19):2997–3003.

Kinniburgh, D. G. and Cooper, D. M. (2011). PhreePlot - Creating graphical output withPHREEQC.

Krachler, M., Emons, H., and Zheng, J. (2001). Speciation of antimony for the 21st century:promises and pitfalls. Trac-Trends in Analytical Chemistry, 20(2):79–90.

Leuz, A. K. (2006). Redox reactions of antimony in the aquatic and terrestrial environment.Thesis, Swiss Federal Institute of Technology Zurich.

Leuz, A. K., Hug, S. J., Wehrli, B., and Johnson, C. A. (2006a). Iron-mediated oxidationof antimony(III) by oxygen and hydrogen peroxide compared to arsenic(III) oxidation.Environmental Science & Technology, 40(8):2565–2571.

Bibliography 61

Leuz, A. K. and Johnson, C. A. R. (2005). Oxidation of Sb(III) to Sb(V) by O2 and H2O2in aqueous solutions. Geochimica Et Cosmochimica Acta, 69(5):1165–1172.

Leuz, A. K., Mönch, H., and Johnson, C. A. (2006b). Sorption of Sb(III) and Sb(V) togoethite: influence on Sb(III) oxidation and mobilization. Environmental Science & Tech-nology, 40(23):7277–7782.

Liu, F., Le, X. C., McKnight-Whitford, A., Xia, Y., Wu, F., Elswick, E., Johnson, C. C.,and Zhu, C. (2010). Antimony speciation and contamination of waters in the Xikuang-shan antimony mining and smelting area, China. Environmental Geochemistry and Health,32(5):401–413.

Livesey, N. T. and Huang, P. M. (1981). Adsorption of arsenate by soils and its relation toselected chemical properties and anions. Soil Science, 131(2):88–94.

Madrid, Y., Barrio-Cordoba, M. E., and Camara, C. (1998). Biosorption of antimony andchromium species by Spirulina platensis and Phaseolus: applications to bioextract antimonyand chromium from natural and industrial waters. Analyst, 123(7):1593–1598.

Manahan, S. E. (2000). Environmental Chemistry. Boca Raton (Fla.), 7th. edition.

MATLAB (2015). version 8.6 (R2015b). The MathWorks Inc., Natick, Massachusetts.

Meeussen, J. C. L. (2003). ORCHESTRA: An object-oriented framework for implementingchemical equilibrium models. Environmental Science & Technology, 37(6):1175–1182.

Nilsson, N., Lovgren, L., and Sjoberg, S. (1992). Phosphate complexation at the surface ofgoethite. Chemical Speciation and Bioavailability, 4(4):121–130.

Ozer, U. Y. and Bogucki, R. F. (1971). Equilibrium studies of antimony(III) chelates inaqueous solution. Journal of Inorganic & Nuclear Chemistry, 33(12):4143–4153.

Parkhurst, D. L. and Appelo, C. A. J. (1999). User’s guide to PHREEQC (Version 2) : acomputer program for speciation, batch-reaction, one-dimensional transport, and inversegeochemical calculations.

Pilarski, J., Waller, P., and Pickering, W. (1995). Sorption of antimony species by humic-acid.Water Air Soil Pollution, 84:51–59.

Pourbaix, M. (1966). Atlas of Electrochemical Equilibria in Aqueous Solutions. PergamonPress, London.

Resongles, E., Casiot, C., Elbaz-Poulichet, F., Freydier, R., Bruneel, O., Piot, C., Delpoux,S., Volant, A., and Desoeuvre, A. (2013). Fate of Sb(V) and Sb(III) species along agradient of pH and oxygen concentration in the Carnoules mine waters (Southern France).ENVIRONMENTAL SCIENCE-PROCESSES & IMPACTS, 15(8):1536–1544.

62

Routh, J. and Ikramuddin, M. (1996). Trace-element geochemistry of Onion Creek near VanStone lead-zinc mine (Washington, USA) - Chemical analysis and geochemical modeling.Chemical Geology, 133:211–224.

Runkel, R. L. (2010). One-Dimensional Transport with Equilibirum Chemistry (OTEQ): AReactive Transport Model for Streams and Rivers, book section Chapter B6, page 101.

Sanders, L. L. (1998). A Manual of Field Hydrogeology. Prentice Hall, New Jersey.

Serafimovska, J. M., Arpadjan, S., and Stafilov, T. (2011). Speciation of dissolved inor-ganic antimony in natural waters using liquid phase semi-microextraction combined withelectrothermal atomic absorption spectrometry. Microchemical Journal, 99(1):46–50.

Smith, E., Naidu, R., and Alston, A. M. (2002). Chemistry of inorganic arsenic in soils: II.Effect of phosphorus, sodium, and calcium on arsenic sorption. Journal of EnvironmentalQuality, 31(2):557–563.

Stumm, W. and Morgan, J. J. (1996). Aquatic chemistry: chemical equilibria and rates innatural waters. Environmental science and technology. New York: Wiley, 3rd ed. edition.

Sun, F. H., Yan, Y. B., Liao, H. Q., Bai, Y. C., Xing, B. S., and Wu, F. C. (2014). Biosorptionof antimony(V) by freshwater cyanobacteria Microcystis from Lake Taihu, China: effects ofpH and competitive ions. Environmental Science and Pollution Research, 21(9):5836–5848.

Sundar, S. and Chakravarty, J. (2010). Antimony toxicity. International Journal of Environ-mental Research and Public Health, 7(12):4267–4277.

Takayanagi, K. and Cossa, D. (1997). Vertical distributions of Sb(III) and Sb(V) in PavinLake, France. Water Research, 31(3):671–674.

Takeno, N. (2005). Atlas of Eh-pH diagrams: Intercomparison of thermodynamic databases.Report, National Institute of Advanced Industrial Science and Technology.

Thanabalasingam, P. and Pickering, W. F. (1986). Arsenic sorption by humic acids. Envi-ronmental Pollution Series B-Chemical and Physical, 12(3):233–246.

Thanabalasingam, P. and Pickering, W. F. (1990). Specific sorption of antimony(III) by thehydrous oxides of Mn, Fe and Al. Water Air and Soil Pollution, 49(1-2):175–185.

Tighe, M., Lockwood, P., and Wilson, S. (2005). Adsorption of antimony(V) by floodplainsoils, amorphous iron(III) hydroxide and humic acid. J Environ Monit, 7(12):1177–85.

U.S. EPA (2001). National primary drinking water standards.

Van Genuchten, M. T. (1985). A general approach for modeling solute transport in structuredsoils. Memoires IAH, 17(2):513–526.

Bibliography 63

Verbrugghe, E. (2016). Antimony speciation and sorption in soils, sediments and tailings fromthe Chenzhou region in China. Master’s thesis, Ghent University.

Wang, X., He, M., Xi, J., and Lu, X. (2011). Antimony distribution and mobility in riversaround the world’s largest antimony mine of Xikuangshan, Hunan Province, China. Mi-crochemical Journal, 97(1):4–11.

Watkins, R., Weiss, D., Dubbin, W., Peel, K., Coles, B., and Arnold, T. (2006). Investigationsinto the kinetics and thermodynamics of Sb(III) adsorption on goethite (alpha-FeOOH).Journal of Colloid and Interface Science, 303(2):639–646.

Wilson, S. C., Lockwood, P. V., Ashley, P. M., and Tighe, M. (2010). The chemistry andbehaviour of antimony in the soil environment with comparisons to arsenic: A criticalreview. Environmental Pollution, 158(5):1169–1181.

Wool, T. A., Ambrose, R. B., Martin, J. L., and Comer, E. A. (2006). Water Quality AnanlysisSimulation Program (WASP) Version 6.0 DRAFT: User’s Manual.

Wu, F. C., Sun, F. H., Wu, S., Yan, Y. B., and Xing, B. S. (2012). Removal of antimony(III)from aqueous solution by freshwater cyanobacteria Microcystis biomass. Chemical Engi-neering Journal, 183:172–179.

Zhang, H., Dong, J., Li, C., and Seuntjens, P. (2014a). Pilot area investigation report. Report,METALert.

Zhang, H., Li, L., and Zhou, S. (2014b). Kinetic modeling of antimony(V) adsorption-desorption and transport in soils. CHEMOSPHERE, 111:434–440.

A | PHREEQC code

A.1 Advection in a perfectly mixed water column, pH-independentsorption

With the code below, advective transport is simulated in a perfectly mixed water columnin which 0.10 kg L-1 of the surface Sorb is present. Sorption of Sb onto this surface is pH-independent. A step input is imposed, the inflowing solution contains 1 µM Sb. A pH of 7and a log pO2 of -0.68 are imposed over the entire system. A time span of 2000 time steps issimulated.

database C: \ phreeqc \ database \adapted_minteq . dat

SURFACE_MASTER_SPECIES # new sp e c i e sSorb Sorb

SURFACE_SPECIES # sorp t i on r e a c t i o n s new sp e c i e sSorb = Sorblog_k 0Sb(OH)3 + Sorb = SorbSb (OH)3−no_check−mole_balance SorbSb (OH)3log_k −98.90626574 # ca l cu l a t ed by means o f equat ion 3 .1Sb(OH)6− + Sorb = SorbSb (OH)6−−no_check−mole_balance SorbSb (OH)6−log_k −98.90626574

PHASESFix_H+H+ = H+log_k 0 .

65

66

SOLUTION 0pH 7Na 10Cl 10Sb 1E−03 # un i t s : mmol/kgwEQUILIBRIUM_PHASES 0Fix_H+ −7 NaOH # pH in s o l u t i o n 0 f i x ed to 7−f o r c e_equa l i t y t rueO2( g ) −0.67778 # log pO2 in s o l u t i o n 0 f i x ed to −0.68−f o r c e_equa l i t y t rueSAVE SOLUTION 0 # save s o l u t i o n 0 with f i x ed pH and log pO2END

SOLUTION 1−40Na 10Cl 10pH 7EQUILIBRIUM_PHASES 1−40Fix_H+ −7 NaOH−f o r c e_equa l i t y t rueO2( g ) −0.67778−f o r c e_equa l i t y t rueSURFACE 1−40Sorb 1e100 1 .0 1 e100 # l i m i t l e s s s o rp t i on s i t e sSAVE SOLUTION 1−40END

TRANSPORT− c e l l s 40− s h i f t s 2000−time_step 312 .5 # to impose a v e l o c i t y o f 0 . 8 m/ s

f o r c e l l s o f 250 m−f l ow_d i r e c t i on forward−boundary_condit ions f l u x # constant f l u x−l eng th s 250− i n i t i a l_ t ime 0−d i s p e r s i v i t i e s 0−d i f f u s i o n_ c o e f f i c i e n t 0

Appendix A. PHREEQC code 67

−stagnant 0 # no immobile c e l l s−p r i n t_c e l l s 1−40−pr int_frequency 100−punch_cel l s 1−40 # r e s u l t s are pr in ted f o r c e l l 1−40−punch_frequency 100 # r e s u l t s are pr in ted every 100 time s t ep s

SELECTED_OUTPUT− f i l e advection_sorb_pH7_O2−068_step . s e l−r e s e t f a l s e−s tep−h igh_prec i s i on true

USER_PUNCH % user s p e c i f i e d parameters to save in output f i l e−headings t iming Sb_diss Sb_sorb Sb_tot pe pH−s t a r t10 microSb=tot ( " Sb " )∗1 e+06 # d i s s o l v ed Sb20 sorbedSb=SURF( " Sb " , " Sorb " )∗1 e+06 # sorbed Sb30 totSb=SYS( " Sb " )∗1 e+06 # t o t a l Sb40 PUNCH tota l_t ime /3600 # time (h)50 PUNCH microSb60 PUNCH sorbedSb70 PUNCH totSb80 PUNCH −l a ( " e−") , −l a ( "H+") # pe and pH−end

END

A.2 Advection in a perfectly mixed water column, pH-dependentsorption

The code below simulates advective transport in a perfectly mixed water column in which0.10 kg L-1 sediments, containing 10.59 g kg-1 HFO, is present. Sorption of Sb onto HFO ispH-dependent. A pulse input is imposed, the inflowing solution contains 1 µM Sb during 100time steps. A pH of 6 and a log pO2 of -45 are imposed over the entire system. A time spanof 40,100 time steps is simulated.


PHASES

68

Fix_H+H+ = H+log_k 0 .

KNOBS−t o l 1e−20 # reduce t o l e r an c e to avoid convergence problems

SOLUTION 0pH 6Na 10Cl 10Sb 1E−01 # un i t s : mmol/kgwEQUILIBRIUM_PHASES 0Fix_H+ −6 NaOH−f o r c e_equa l i t y t rueO2( g ) −45−f o r c e_equa l i t y t rueSAVE SOLUTION 0END

SOLUTION 1−40Na 10Cl 10pH 6EQUILIBRIUM_PHASES 1−40Fix_H+ −6 NaOH # pH in s o l u t i o n 0 f i x ed to 6−f o r c e_equa l i t y t rueO2( g ) −45 # log pO2 in s o l u t i o n 0 f i x ed to −45−f o r c e_equa l i t y t rueSURFACE 1−40Hfo_s 5.9533E−05 600 1.059687135# 0 .1 kg s o l i d s −> 1.06 g HFO, spec . area 600 m2/g , 5 .9533E−05 mol s i t e sHfo_w 0.002381319# 0.002381319 s i t e s# no . o f s i t e s : 0 . 2 mol/mol Fe weak , 0 .005 mol/mol Fe s t rong (Dzombak & Morel )SAVE SOLUTION 1−40SAVE SURFACE 1−40END


SELECTED_OUTPUT− f i l e advection_hfo_pH6_O2−45_pulse . s e l−r e s e t f a l s e−s tep−s o l u t i o n−h igh_prec i s i on true

USER_PUNCH−headings t iming Sb_diss Sb (3 ) _diss Sb (5 ) _diss Sb_sorb Sb (3 ) _sorb Sb (5 ) _sorbSb_tot pe pH−s t a r t10 microSb=tot ( " Sb " )∗1 e+0620 sorbedSb=SURF( " Sb " , " Hfo " )∗1 e+0630 totSb=SYS( " Sb " )∗1 e+0640 PUNCH tota l_t ime /360050 PUNCH microSb60 PUNCH tot ( " Sb ( 3 ) " )∗1 e+06, to t ( " Sb ( 5 ) " )∗1 e+0670 PUNCH sorbedSb80 PUNCH SURF( " Sb ( 3 ) " , " Hfo " )∗1 e+06, SURF( " Sb ( 5 ) " , " Hfo " )∗1 e+0690 PUNCH totSb100 PUNCH −l a ( " e−") , −l a ( "H+")−end

TRANSPORT− c e l l s 40− s h i f t s 100−time_step 312 .5 # to impose a v e l o c i t y o f 0 . 8 m/ s

f o r c e l l s o f 250 m−f l ow_d i r e c t i on forward−boundary_condit ions f l u x # constant f l u x−l eng th s 250− i n i t i a l_ t ime 0−d i s p e r s i v i t i e s 0−d i f f u s i o n_ c o e f f i c i e n t 0−stagnant 0 # no immobile c e l l s−p r i n t_c e l l s 40−pr int_frequency 1−punch_cel l s 40 # r e s u l t s pr in ted only f o r c e l l 40−punch_frequency 1 # r e s u l t s pr in ted every time step

70

END

SOLUTION 0pH 6Na 10Cl 10EQUILIBRIUM_PHASES 0Fix_H+ −6 NaOH−f o r c e_equa l i t y t rueO2( g ) −45−f o r c e_equa l i t y t rueSAVE SOLUTION 0END

TRANSPORT− c e l l s 40− s h i f t s 40000−time_step 312 .5−f l ow_d i r e c t i on forward−boundary_condit ions f l u x # constant f l u x−l eng th s 250−d i s p e r s i v i t i e s 0−d i f f u s i o n_ c o e f f i c i e n t 0−stagnant 0 # no immobile c e l l s−p r i n t_c e l l s 40−pr int_frequency 1−punch_cel l s 40 # r e s u l t s are only pr in ted f o r c e l l 40−punch_frequency 1 # r e s u l t s are pr in ted every time stepEND

A.3 Advective transport with diffusive exchanges with thebenthic segments

This code simulates advective transport with diffusive exchanges with benthic immobile seg-ments. These exchanges are defined by means of an exchange factor (cfr. equation 3.3). Inthe surface part no solid phase is present, the benthic part contains 1.0 kg L-1 sediments. Astep input containing 1 µM Sb is sent into the system. A pH of 5 and log pO2 values of -0.68(surface) and -70 (benthic) are imposed. A time span of 10,000 time steps is simulated.



PHASESFix_H+H+ = H+log_k 0 .Fix_pee− = e−log_k 0 .

KNOBS−t o l 1e−20

SOLUTION 0pH 5Na 10Cl 10Sb 1E−03 # un i t s : mmol/kgwEQUILIBRIUM_PHASES 0Fix_H+ −5 NaOH # pH in s o l u t i o n 0 f i x ed to 5−f o r c e_equa l i t y t rueO2( g ) −0.67778 # log pO2 in s o l u t i o n 0 f i x ed to −0.68−f o r c e_equa l i t y t rueSAVE SOLUTION 0END

SOLUTION 1−40 # su r f a c e c e l l spH 5Na 10Cl 10EQUILIBRIUM_PHASES 1−40Fix_H+ −5 NaOH−f o r c e_equa l i t y t rueO2( g ) −0.67778−f o r c e_equa l i t y t rueSAVE SOLUTION 1−40SAVE SURFACE 1−40END

72

SOLUTION 42−81 # immobile benth ic c e l l s ( c f r . equat ion 3 . 2 )pH 5Na 10Cl 10EQUILIBRIUM_PHASES 42−81Fix_H+ −5 NaOH−f o r c e_equa l i t y t rueO2( g ) −70−f o r c e_equa l i t y t rueSURFACE 42−81Hfo_s 0.000594902 600 10.58924709 # 10 .6 g HFO# 1 kg s o l i d s −> 10.6 g HFO, spec . area 600 m2/g , 0 .00059 mol s i t e sHfo_w 0.023796061# 0.0238 mol s i t e s# no . o f s i t e s : 0 . 2 mol/mol Fe weak , 0 .005 mol/mol Fe s t rong (Dzombak & Morel )SAVE SOLUTION 42−81SAVE SURFACE 41−81END

TRANSPORT− c e l l s 40− s h i f t s 10000−time_step 312 .5−f l ow_d i r e c t i on forward−boundary_condit ions f l u x # constant f l u x−l eng th s 250− i n i t i a l_ t ime 0−d i s p e r s i v i t i e s 0−d i f f u s i o n_ c o e f f i c i e n t 0−stagnant 1 4 .661E−07 0.999969811 0.622641509# exchange fcn : 1 immobile c e l l per mobile c e l l , exchange f a c t o r 4 .661E−07# ( c f r . eq 3 . 3 ) , po ro s i t y mobile c e l l 0 .99997 , po ro s i t y immobile c e l l 0 .6226−p r i n t_c e l l s 1−81−pr int_frequency 10−punch_cel l s 1−81 # r e s u l t s are pr in ted f o r a l l c e l l s−punch_frequency 10 # r e s u l t s are pr in ted every 10 time s t ep s

SELECTED_OUTPUT


− f i l e advection_benthsurf_step_O2−068−70_pH510000 . s e l−r e s e t f a l s e−h igh_prec i s i on true−s tep true−s o l u t i o n true

USER_PUNCH−headings t iming Sb_diss Sb3_diss Sb5_diss Sb_sorbed Sb_tot pH pe−s t a r t10 microSb=tot ( " Sb " )∗1 e+0620 sorbedSb=SURF( " Sb " , " Hfo " )∗1 e+0630 totSb=SYS( " Sb " )∗1E+0640 PUNCH tota l_t ime /360050 PUNCH microSb60 PUNCH TOT( " Sb ( 3 ) " )∗1E+0670 PUNCH TOT( " Sb ( 5 ) " )∗1E+0680 PUNCH sorbedSb , totSb90 PUNCH −l a ( "H+")100 PUNCH −l a ( " e−")−end

END

A.4 Desorption

The code below allows to simulate desorption over time, starting from a sorbed Sb concen-tration of 2.873 · 103 µmol kg-1 or 350 mg kg-1 onto the sediments in the benthic part of thewater column. The inflowing solution contains no Sb. A pH of 7 and log pO2 values of -0.68(surface) and -70 (benthic) are imposed. A time span of 10,000 time steps is simulated. Theduration of the simuation equals 200 time steps.


PHASESFix_H+H+ = H+log_k 0 .Fix_pee− = e−

74

log_k 0 .

KNOBS−t o l 1E−20

SOLUTION 42−81 # immobile benth ic c e l l s ( c f r . equat ion 3 . 2 )pH 7Na 10Cl 10Sb 2 .877 # concent ra t i on needed to obta in 350 mg/kg sorbedEQUILIBRIUM_PHASES 42−81Fix_H+ −7 NaOH−f o r c e_equa l i t y t rueO2( g ) −70−f o r c e_equa l i t y t rueSURFACE 42−81Hfo_s 0.000594902 60 10.58924709# 1 kg s o l i d s −> 10.6 g HFO, spec . area 600 m2/g , 0 .00059 mol s i t e sHfo_w 0.023796061# 0.0238 mol s i t e s# no . o f s i t e s : 0 . 2 mol/mol Fe weak , 0 .005 mol/mol Fe s t rong (Dzombak & Morel )SAVE SURFACE 41−81 # save ’ contaminated ’ s u r f a c eEND

SOLUTION 0pH 7Na 10Cl 10EQUILIBRIUM_PHASES 0Fix_H+ −7 NaOH−f o r c e_equa l i t y t rueO2( g ) −0.67778−f o r c e_equa l i t y t rueSAVE SOLUTION 0END

SOLUTION 1−40pH 7Na 10


Cl 10EQUILIBRIUM_PHASES 1−40Fix_H+ −7 NaOH−f o r c e_equa l i t y t rueO2( g ) −0.67778−f o r c e_equa l i t y t rueSAVE SOLUTION 1−40 # save c l ean s o l u t i o nEND

SOLUTION 42−81pH 7Na 10Cl 10EQUILIBRIUM_PHASES 42−81Fix_H+ −7 NaOH−f o r c e_equa l i t y t rueO2( g ) −70−f o r c e_equa l i t y t rueSAVE SOLUTION 42−81END

TRANSPORT− c e l l s 40− s h i f t s 10000−time_step 312 .5−f l ow_d i r e c t i on forward−boundary_condit ions f l u x # constant f l u x−l eng th s 250− i n i t i a l_ t ime 0−d i s p e r s i v i t i e s 0−d i f f u s i o n_ c o e f f i c i e n t 0−stagnant 1 4 .661E−07 0.999969811 0.622641509# exchange fcn : 1 immobile c e l l per mobile c e l l , exchange f a c t o r 4 .661E−07# ( c f r . eq 3 . 3 ) , po ro s i t y mobile c e l l 0 .99997 , po ro s i t y immobile c e l l 0 .6226−p r i n t_c e l l s 10 51−pr int_frequency 1−punch_cel l s 10 51 # r e s u l t s are only pr in ted f o r c e l l 10−punch_frequency 1 # r e s u l t s are pr in ted every time step

76

SELECTED_OUTPUT− f i l e advection_benthsurf_desorption_O2−068−70_pH710000t . s e l−r e s e t f a l s e−h igh_prec i s i on true−s tep true−s o l u t i o n true

USER_PUNCH−headings t iming Sb_diss Sb3_diss Sb5_diss Sb_sorbed Sb_tot pH pe−s t a r t10 microSb=tot ( " Sb " )∗1 e+0620 sorbedSb=SURF( " Sb " , " Hfo " )∗1 e+0630 totSb=SYS( " Sb " )∗1E+0640 PUNCH tota l_t ime /360050 PUNCH microSb60 PUNCH TOT( " Sb ( 3 ) " )∗1E+0670 PUNCH TOT( " Sb ( 5 ) " )∗1E+0680 PUNCH sorbedSb , totSb90 PUNCH −l a ( "H+")100 PUNCH −l a ( " e−")−end

END

A.5 Colloid transport

With the code below, advective transport in the presence of mobile colloids is simulated. Thediffusion coefficient equals 10−9 m2 s-1 for the solutes and 10−13 m2 s-1 for the solid phase. Apulse of 100 time steps, containing 1 µM Sb, is sent into the system. A pH of 3 and a log pO2of -0.68 are imposed.


PHASESFix_H+H+ = H+log_k 0 .Fix_pee− = e−


log_k 0 .

KNOBS−t o l 1e−20

SOLUTION 0pH 3Na 10Cl 10N(3) 1E−06 chargeSb 1E−01 # un i t s : mmol/kgwEQUILIBRIUM_PHASES 0Fix_H+ −7 NaOH−f o r c e_equa l i t y t rueO2( g ) −0.67778−f o r c e_equa l i t y t rueSURFACE 0Hfo_s 4.75921E−08 600 0.00084714 Dw 1e−13# 80 mg s o l i d s −> 0.0008 g HFO, spec . area 600 m2/g , 4 .76 mol s i t e s# d i f f u s i o n c o e f f i c i e n t = 1e−13 m2/ sHfo_w 1.90368E−06# 1.90368E−06 mol s i t e s# no . o f s i t e s : 0 . 2 mol/mol Fe weak , 0 .005 mol/mol Fe s t rong (Dzombak & Morel )−d i f f u s e_ l ay e r # d i f f u s e l ay e r c a l c u l a t ed f o r t ranspo r tSAVE SOLUTION 0SAVE SURFACE 0END

SOLUTION 1−40pH 3Na 10Cl 10N(3) 1E−06 chargeEQUILIBRIUM_PHASES 1−40Fix_H+ −7 NaOH−f o r c e_equa l i t y t rueO2( g ) −0.67778−f o r c e_equa l i t y t rueSURFACE 1−40

78

Hfo_s 4.75921E−08 600 0.00084714 Dw 1e−13Hfo_w 1.90368E−06−d i f f u s e_ l ay e rSAVE SURFACE 1−40SAVE SOLUTION 1−40END

SELECTED_OUTPUT− f i l e mobi lecol lo ids_pulse_O2 −068_pH7 . s e l−r e s e t f a l s e−h igh_prec i s i on true−s o l u t i o n true−s tep true

USER_PUNCH−headings t iming Sb_diss Sb (5 ) _diss Sb (3 ) _diss Sb_sorb Sb_tot pe pH−s t a r t10 microSb=TOT( " Sb " )∗1 e+0620 sorbedSb=SURF( " Sb " , " Hfo " )∗1 e+0630 totSb=SYS( " Sb " )∗1 e+0640 PUNCH tota l_t ime /360050 PUNCH microSb60 PUNCH TOT( " Sb ( 5 ) " )∗1E+0670 PUNCH TOT( " Sb ( 3 ) " )∗1E+0680 PUNCH sorbedSb , totSb90 PUNCH −l a ( " e−")100 PUNCH −l a ( "H+")−end

TRANSPORT− c e l l s 40− s h i f t s 100−time_step 312 .5−f l ow_d i r e c t i on forward−boundary_condit ions f l u x # constant f l u x−l eng th s 250− i n i t i a l_ t ime 0−d i s p e r s i v i t i e s 0−d i f f u s i o n_ c o e f f i c i e n t 0


−stagnant 0 # no immobile c e l l s−multi_d true 1E−9 0.999969811 0 .05 1 .33−p r i n t_c e l l s 1−40−pr int_frequency 1−punch_cel l s 1−40 # r e s u l t s pr in ted f o r a l l c e l l s−punch_frequency 1 # r e s u l t s pr in ted every time stepEND

SOLUTION 0pH 3Na 10Cl 10N(3) 1E−06 chargeEQUILIBRIUM_PHASES 0Fix_H+ −7 NaOH−f o r c e_equa l i t y t rueO2( g ) −0.67778−f o r c e_equa l i t y t rueSURFACE 0Hfo_s 4.75921E−08 600 0.00084714 Dw 1e−13Hfo_w 1.90368E−06−d i f f u s e_ l ay e rSAVE SOLUTION 0SAVE SURFACE 0END

TRANSPORT− c e l l s 40− s h i f t s 100−time_step 312 .5−f l ow_d i r e c t i on forward−boundary_condit ions f l u x # constant f l u x−l eng th s 250−d i s p e r s i v i t i e s 0−d i f f u s i o n_ c o e f f i c i e n t 0−stagnant 0 # no immobile c e l l s−multi_d true 1e−9 0.999969811 0 .05 1 .33# d i f f e r e n t d i f f u s i o n c o e f f i c i e n t s : 1e−9 de fau l t , mobile po ro s i t y 0 .99997 ,# t r e sho l d po ro s i t y f o r d i f f u s i o n = 0 .05 , exponent to c a l c u l a t e Def f = 1 .33

80

−p r i n t_c e l l s 1−40−pr int_frequency 1−punch_cel l s 1−40 # r e s u l t s pr in ted f o r a l l c e l l s−punch_frequency 1 # r e s u l t s pr in ted every time stepEND

modelling the reactive transport of antimony in river...

Documents