Sie sind auf Seite 1von 74

Modelling Biologically Induced Phosphate

Precipitation in Aerobic Granules


S. M. L. Stubbé
Wetsus Academy Joint Degree
MSc. Thesis
We’re all working together, that is the secret
- Sam Walton

Title Modelling Biologically Induced Phosphate Precipitation in Aerobic Granules


Author Stefanie Maria Lousie Stubbé
Studentnumber 910112825040
Daily Supervisor Dr. Mario Pronk, Applied Sciene
Chair Assessment Prof. Dr. Ir. Mark C. M. van Loosdrecht, Applied Science
Assessment Committee Dr. Ir. Merle K. de Kreuk, Civil Engineering
Dr. Ir. Cristian Picioreanu, Applied Science
Dr. Boris M. van Breukelen, Civil Engineering

Project Master of Science Thesis Project


Master Track Watertechnology
Joint Degree by Universities Wageningen University, Groningen University, Technical University of Twente
Publicly Defense Tuesday 27th September 2016, 04:00 PM
Location Kronigzaal, Applied Science, van der Maasweg 9, Delft

Confidentiality of this Report


This study was commissioned by Royal HaskoningDHV. No part of this report may be reproduced and/
or published by print, photocopy, microfilm or by any other means, without the prior written permission
of HaskoningDHV Nederland B.V.; nor may any part of the report be used, without such permission, for
any purposes other than that for which it was produced.
3

Modelling Biologically Induced Phosphate Precipitation in Aerobic Granules

By

Stefanie Maria Louise Stubbé


4
5

Modelling Biologically Induced Phosphate Precipitation in Aerobic Granules

by S.M.L. Stubbe1,2,3 supervised by C. Picioreanu4, M. Pronk3,4, B. M. van Breukelen1, M.C.M. van Loosdrecht2,4, and M.K. de Kreuk1,

¹Delft University of Technology, Civil Engineering, Sanitary Engineering, Stevinweg 1, 2628 CN, Delft, The Netherlands
²Wetsus European Centre of Excellence for Sustainable Water Technology, Oostergoweg 9, 8911 MA, Leeuwarden
³Royal HaskoningDHV, P.O Box 1132, 3800 BC Amersfoort, The Netherlands
4
Delft University of Technology, Kluyvewr Institute for Biotechnology, Van der Maasweg 9, 2629 HZ Delft, The Netherlands

Abstract

The model developed during this study provided a tool to discover the parameter interdependencies that influence biologically
induced phosphate precipitation inside aerobic granules of full-scale installations in the Netherlands. The model included the biological
metabolism of the phosphate accumulating organisms (PAOs), the chemical precipitation kinetics of amorphous calcium phosphate
(ACP), hydroxyapatite (HAP), struvite (STR) and vivianite (VIV) combined with ion transport inside the granule. The model showed that
the resulting mineral content inside aerobic granules is a dynamic process of precipitation and dissolution throughout the process cycle
(feeding and aeration) and over consecutive days. The iron and calcium phosphate minerals were more supersaturated than magnesium
minerals inside the granule in generic Dutch domestic waste water. At the current operational conditions of the full-scale aerobic granular
sludge (AGS) installations in the Netherlands, no significant amounts of HAP, STR and VIV crystals accumulated inside the granules
during various cycles. Only ACP showed noteworthy contents during the simulations. Struvite was undersaturated throughout the process
cycle under the current assumptions. Hydroxyapatite and vivianite had too slow kinetics to precipitate significantly during the cycles. The
operational conditions to enhance the phosphate precipitation inside full-scale AGS systems were discussed.

1. Introduction

Since 2011, aerobic granular sludge (AGS) technology is analogues of the more researched magnesium ammonium
operative at full-scale level in the Netherlands (van der Roest et phosphate (MgNH4PO4.6H2O, STR-NH4) (Banks et al. 1975,
al. 2011, Giesen et al. 2013). There are multiple benefits to AGS Mathew et al. 1982, Graeser et al. 2008). Apart from STR-K, ACP
technology: simultaneously converting chemical oxygen demand and HAP, other precipitates were suggested to be possibly present
(COD) and nutrients (nitrogen and phosphorus) to low effluent within aerobic granules: struvite (MgNH4PO4.6H2O), newberyite
concentrations (Morgenroth et al. 1997), advantageous settling (MgHPO4.3H2O) and brushite (CaHPO4.2H2O) (Barat et al. 2008,
properties (Winkler et al. 2012), less area requirements (De Bruin et Manas et al. 2012). Although it is researched that the phosphate
al. 2004) and is cost-effective compared to conventional activated removal in AGS lab-reactors is done partially through biologically
sludge systems (De Kreuk et al. 2004). The anaerobic release induced phosphate precipitation (Barat et al. 2008, Mañas et al.
of phosphate by phosphate accumulating organisms (PAOs) 2011, Lin et al. 2012), it was shown that this mechanism plays
has proven to facilitate the conditions for biological induced less of a role in full-scale AGS reactors in the Netherlands under
phosphate precipitation inside aerobic granules in lab reactors current conditions (Stubbé et al. 2016). The largest difference
(Mañas et al. 2011, Liu et al. 2016). The release of phosphate inside between the lab and full-scale reactors are the concentrations
spherical granules increases the saturation indices of phosphate and pH. Repeated XRD analyses on aerobic granules from full-
minerals. Maurer et al. (1999) has indicated that part of the scale AGS installations showed dynamic results with different
phosphate in enhanced biological phosphate removal (EBPR) phosphate precipitates (e.g., STR-K, brushite, dolomite)
systems can be removed by means of precipitation. appearing and disappearing in time (Stubbé et al. 2016).
Precipitates in the form of amorphous calcium phosphate The potential of phosphate precipitation does exist
(Ca3(PO4)2.6H2O ACP), Hydroxyapatite (Ca5(PO4)3OH.6H2O, HAP), in full-scale AGS installations, as the studies on lab-scale AGS
potassium struvite (MgKPO4.6H2O, STR-K) and calcite (CaCO3) reactors showed (Barat et al. 2011, Mañas et al. 2011, Lin et al.
were found in aerobic granules harvested from lab reactors (Barat 2012). How the full-scale process needs to be adjusted to exploit
et al. 2008, Mañas et al. 2011, Lin et al. 2012, Liu et al. 2016). STR-K this potential could be evaluated with a computational model
and Na-struvite (MgNaPO4.6H2O, STR-Na) are isomorphous combining the biological, chemical and transport processes.
6

Nomenclature

ACP Amorphous calcium phosphate (Ca3(PO4)2.6H2O) So far, no model aimed to explicitly describe both phosphate
ASM2d Activated Sludge Model No 2d removal mechanisms (biologically and chemical precipitation)
inside an aerobic granule, taking into account pure-system
AGS Aerobic granular sludge
crystallisation kinetics of multiple (Fe2+/3+, Ca2+ and Mg2+)
Bio-P Biological phosphate removal
phosphate minerals. There are models describing the biological
BOD Biochemical oxygen demand conversions of multispecies systems, even inside the granule, but
COD Chemical oxygen demand not including the chemical phosphate precipitation processes
EBPR Enhanced biological phosphorus removal (De Kreuk et al. 2007, Xavier et al. 2007, Kagawa et al. 2015). The
models that did include precipitation either only considered
EDX Energy Dispersive X-ray
one or two cations (mostly Ca2+) or used Monod kinetics for
EPS Extracellular polymeric substances chemical processes and/or omitted the concentration gradient
Fe Iron inside the granule (Maurer et al. 1999, Barat et al. 2011, Mañas
FISH Fluorscence in situ hybridization et al. 2012). More extensive models are presented for plant-
wide application, including numerous biology and chemical
HAP Hydroxyapatite (Ca5(PO4)3OH.6H2O)
conversions, but focused on activated sludge systems (Barat et
Kp Solubility product of mineral phase p
al. 2013, Mbamba et al. 2016). Even in the models for anaerobic
N Nitrogen granules, inclusion of chemical precipitation did not go beyond
NH4 Ammonium calcite and aragonite precipitation (Wu et al. 1997, Van Langerak
NH4 : TP Ammonium : Total Phosphatewwv influent ratio et al. 2000, Batstone et al. 2003, Liu et al. 2003). None of these
models aimed to simulate all processes (biological, chemical and
Nkj Kjeldahl-Nitrogen (organic-N, NH3, NH4)
diffusive transport) inside the granule to assess the precipitation
P Phosphate
potential, possibly because of long computational times or not
PAO Phosphate Accumulating Organism enough available experimental data. The impact of exploiting
PHA Polyhydroxyalkanoates this potential increased considerably since the full-scale AGS
PHB Polyhydroxybutyrate technology is implemented globally (World Water 2015).
The aim of this study was to create a comprehensive
PO4 Ortho-phosphate
model that includes chemical equilibria, biological phosphate
Poly-P Poly-Phosphate conversions, multiple phosphate precipitation processes (Fe2+/3+,
PRT Phosphate retention time Ca2+ and Mg2+) and the ion transport processes inside an aerobic
SEM Scanning Electron Microscopy granule of a full-scale AGS installation in the Netherlands
SI Supersaturation Index (schematically illustrated in Fig. 1). The model could provide
generic insights into the important operating conditions
SRT Sludge Retention Time
and parameter interdependencies that affect the phosphate
STR-NH4 Ammonium magnesium phosphate precipitation in full-scale AGS systems. The study does not aim to
(MgNH4PO4.6H2O) describe or assess the treatment performance of AGS systems. As
STR-K Potassium magnesium phosphate (MgKPO4.6H2O) sufficient data for adequate validation of the model was lacking,
knowledge gaps were identified for further research.
STR-Na Sodium magnesium phosphate (MgNaPO4.6H2O)
SS Suspended solids
TN Total Nitrogen
TP Total Phosphate
TSS Total Suspended Solids
VFA Volatile Fatty Acid
VSS Volatile Suspended Solids
WW Wet Weight
XRD X-ray diffractogram
XRF X-ray fluorescence
7

2. Model Description

PHREEQC software
The model code pH-Redox-Equilibrium in C programming Chemical Equilibria Biological Model Precipitation Model Ion Transport
language (PHREEQC) was used to calculate the solute speciation
equilibria, saturation indices, kinetic rates and ion transport in
each computational cell (Parkhurst et al. 2013). A computational
cell represents a certain biomass thickness (see Ion transport).
The PHREEQC software functions as a split-operator: it
solves the transport equations between the computational
cells (⅓ step) before solving the reaction equation inside each
computational cell (⅓ step), after which the equilibria are
established (⅓ step) (see schematic illustration in Fig. 2).
All computations were done on a personal computer
(Intel® Xeon® CPU E5-1620 v2) with the PHREEQC version 3.3.7
in combination with Notepad++ version 6.6.9. The simulations
took ± 90 hours to compute four consecutive operational
cycles with a total duration 16 h (feeding and aeration).
Fig. 1 Schematic illustration of the model; including chemical
equilibria, biological processes, precipitation processes and ion
Chemical equilibria
transport inside an aerobic granule.
The solute speciation equilibria (PO43-, HPO42-, H2PO4-, etc.) were
considered to behave according to the mass action law (Parkhurst
et al. 2013). The Minteq database was taken as input for the solute I
equilibria as it contained organic solutes such as acetate (Minteq
database 2006). Only the relevant equilibra were included; the
stoichiometry reaction equations and the respective equilibrium
constants of the 61 considered solutes can be found in Table 1 T
and 2. The temperature effects on the solute speciation equilibria
II == Initialization
Initialization
constants were taken into account according to the van ‘t Hoff
equation (Minteq database 2006, Parkhurst et al. 2013).
TT == Transport
Transport Task
Task
RR == Reaction
Reaction Task
Task
N R EE == Equilibria
Equilibria Task
Task
NN == Next
Next time
time step
step Task
Task
CC == Completion
Completion TaskTask
E

Fig. 2 Schematic representation of the split operation by


PHREEQC software (adjusted from Parkhurst et al. (2010))
8

Table
Table1 Equilibrium matrix for
1 Equilibrium solutes
matrix considered
for solutes in this study (I)in
considered this study (I)

Solutes Log K Reference

Solutes H2 H PO43 NH4 CO32 Ca2 Mg2 K Fe2+ Fe3+ Na SO42 NO Cl-
O + - + - + + + + - -

OH- OH- -1 1 -13.997

CaOH+ -1 1 -1 -12.697

MgOH+ -1 1 -1 -11.397

FeOH+ -1 1 -1 -9.387

Fe(OH)2 -2 2 -1 -20.494

Fe(OH)3- -3 3 -1 -28.991

FeOH2+ -1 1 -1 -2.187

Fe(OH)2+ -2 2 -1 -4.594

Fe(OH)3 -3 3 -1 -12.56

Fe(OH)4- -4 4 -1 -21.588

Fe2(OH)24+ -2 2 -2 -2.854

Fe3(OH)45+ -4 4 -2 -6.288

NH3 NH3 1 -1 -9.244

CaNH32+ 1 -1 -1 -9.144 Minteq database (2006)

Ca(NH3)22+ 2 -2 -1 -18.788

PO43- HPO42- -1 -1 12.375

H2PO4- -2 -1 19.573

H3PO4 -3 -1 21.721

MgPO4- -1 -1 4.654

MgHPO4 -1 -1 -1 15.175

MgH2PO4 -2 -1 -1 21.2561

CaPO4- -1 -1 6.46

CaHPO4 -1 -1 -1 15.035

CaH2PO4 -2 -1 -1 20.923

KHPO4- -1 -1 -1 13.255

NaHPO4- -1 -1 -1 13.445

FeHPO4 -1 -1 -1 15.975

FeH2PO4+ -2 -1 -1 22.273

FeH2PO42+ -2 -1 -1 23.8515

FeHPO4+ -1 -1 -1 22.292
9

Table
Table2 Equilibrium matrix matrix
2 Equilibrium for solutes
forconsidered in this study (II)
solutes considered in this study (II)

Solutes Log K Reference

Solutes H2 H PO43 NH4 CO32 Ca2 Mg2 K Fe2+ Fe3+ Na SO42 NO Cl-
O + - + - + + + + - -

CO32- HCO3- -1 -1 10.329

H2CO3 -2 -1 16.681

MgCO3 -1 -1 6.2546

MgHCO3+ -1 -1 -1 11.339

CaCO3 -1 -1 3.2

CaHCO3+ -1 -1 -1 11.599

NaCO 3- -1 -1 1.27

NaHCO3- -1 -1 -1 10.079

FeHCO3 -1 -1 -1 11.429

NO3- CaNO3+ -1 -1 0.5

FeNO 32+ -1 -1 1

SO42- KSO4- -1 -1 0.85

NaSO4- -1 -1 0.73

H(SO4)- -1 -1 1.99

Minteq database (2006)


NH4(SO ) 4 - -1 -1 1.03

CaSO4 -1 -1 2.36

MgSO4 -1 -1 2.26

FeSO4 -1 -1 2.39

FeSO 4+ -1 -1 4.05

Fe(SO4)2- -1 -2 5.38

Acetate- H(Acetate) -1 -1 4.757

Fe(Acetate)+ -1 -1 1.4

Fe(Acetate)2+ -1 -1 4.0234

Fe(Acetate)2+ -1 -2 7.5723

Mg(Acetate)+ -1 -1 1.27

Ca(Acetate)+ -1 -1 1.18

Na(Acetate) + -1 -1 -0.18

K(Acetate)+ -1 -1 -0.1955

Cl- FeCl2+ -1 -1 1.48

FeCl2+ -1 -2 2.13

FeCl3 -1 -3 1.13
10

Biological Model
The biological model developed by Lopez-Vazquez et al. (2009) for 3). As the pH dependence reported by Filipe et al. (2001) was used
Candidatus Accumulibacter phosphatis was adapted. Candidatus in the model of Lopez-Vazquez et al. (2009) and it was determined
Accumulibacter phosphatis is a phosphate accumulating on a more pure PAO culture, this relationship was favoured over
organism (PAO) that embodies a significant amount of the the pH dependence of Smolders et al. (1994).
biomass in AGS systems (Pronk et al. 2015) and facilitates • The Poly-P stoichiometry was adjusted to K0.28Mg0.36H2PO4 in
the biological phosphate removal. The aerobic and anaerobic order to incorporate the production of K+ and Mg2+ during P-release
biological conversions of PAOs included the anaerobic acetate as researched in sequencing batch reactors (SBR) by Barat et al.
up-take (R1), anaerobic maintenance (R2), polyhydroxyalkanoates (2005). H2PO4- was adapted instead of PO3-, since it was assumed
(PHA) degradation (R3), poly-phosphate (Poly-P) formation that the reaction with PO3- and H2O occurs instantaneous inside
(R4), glycogen formation (R5) and aerobic maintenance (R6) (see the bacteria cell (Serralta et al. 2004, Saunders et al. 2007).
Fig. 3 for a schematic illustration and Table 3 for the reaction • The intracellular biopolymer content of the PAOs
stoichiometry). The kinetic rate expressions are shown in Table (Poly-P, PHA and glycogen) was represented in the PHREEQC
4. All the symbols, units and constants used can be found in software by defining new solutes with an extremely small
Addendum 1. The elemental stoichiometry of the biopolymers, diffusion coefficient (10-20 m2 s-1).
biomass, phosphate and acetate are also shown in Fig. 3. • The extra biomass acquired through PHA biodegradation was
Slight alterations were made to the model not included in the consecutive cycles. The biomass concentration
of Lopez-Vazquez et al. (2009). The slight alterations (XPAO) was kept constant.
and additional assumptions are listed here: • No heterotrophic biomass other than the PAO was taken into
• Lopez-Vazquez et al. (2009) did not include the production consideration, as no substrate was available in the aerobic phase
of CO2 in the stoichiometric reaction equations, which might for growth.
influence the pH. The stoichiometric reaction equations used by • No other conversions were implemented (e.g., nitrification or
Lopez-Vazquez et al. (2009) are based on the results of Smolders denitrification).
(1995). Since Smolders et al. (1995) did include the production of • The porosity and biomass concentration were
CO2, these values are adapted instead (see Table 3). considered to be constant over the whole granule.
• The Poly-P degradation was dependent on the pH as described • The bacteria did not embody an electrical charge.
by Filipe et al. (2001) and incorporated in the model (see α in Table

(CH2.09O0.54N0.2P0.015)
(C2H4O2) (H2PO4)

(K0.28Mg0.36H2PO4)

(CH1.5O0.5) (CH1.67O0.83)

Fig.
Fig. XX 3 Metabolic
Metabolic conversionsofofphosphate
conversions phosphate accumulating
accumulating organisms
organisms(PAOs)
(PAOs)depicted schematically
depicted schematicallyduring anaerobic
during and aerobic
anaerobic and
conditions (adjusted with authorization of De Kreuk (2006)). Polyhydroxybutyrate (PHB) belongs to the group of polyhydroxyalkanoates
aerobic conditions (adjusted with authorization of De Kreuk (2006)). Polyhydroxybutyrate (PHB) belongs to the group of
(PHA).
polyhydroxyalkanoates (PHA).
11

Table 3 Stoichiometry equations for the metabolic conversions of the phosphate accumulating organisms*
Table 3 Stoichiometry equations for the metabolic conversions of the phosphate accumulating organisms*
Anaerobic Reference

R1 Acetate up-take 0.5 H(Acetate) + 0.5 Glycogen + (0.5 + α) Poly-P + 0.023 H2O → 1.33 PHA + 0.17 CO2 + Smolders (1995), Barat et al. (2005)
(0.5 + α) H2PO4- + 0.28 (0.5 + α) K+ + 0.36 (0.5 + α) Mg2+

α pH dependence 0.16 . pH – 0.7985 Filipe et al. (2001)

R2 Anaerobic Poly-P + H2O → H2PO4 + 0.28 K + 0.36 Mg- + 2+


Smolders (1995), Barat et al. (2005)
maintenance

Aerobic

R3 PHA degradation 1.37 PHA + 0.20 NH3 + 0.015 H3PO4 + 0.42 O2 → Biomass + 0.37 CO2 + 0.305 H2O Smolders (1995)

R4 Poly-P formation 0.27 PHA + 0.306 O2 + H2PO4 + 0.28 K + 0.36 Mg - + 2+


→ Poly-P + 0.27 CO2 + 1.20 H2O Smolders (1995), Barat et al. (2005)

R5 Glycogen formation 1.12 PHA + 0.26 O2 → Glycogen + 0.12 CO2 + 0.007 H2O Smolders (1995)

R6 Aerobic maintenance PHA + 1.125 O2 → CO2 + 0.75 H2O Smolders (1995)

*The stoichiometry of PHA (group to which polyhydroxybutyrate (PHB) belongs), Poly-P, H(Acetate) and Biomass can be found in Fig. 3 and Addendum 1.

Table
Table XX Kinetic
4 Kinetic rate expressions
rate expressions for the
for the metabolic metabolic
conversions conversions
of the of the phosphate
phosphate accumulating accumulating
organism* organism*

Anaerobic Reference
   
1 Acetate uptake 
 ∗  ∗  ∗ ∗ ∗ 1 −  Lopez-Vazquez et al. (2009)
              

 
2 Maintenance 
 ∗  ∗  ∗ 1 −  Lopez-Vazquez et al. (2009)
      

Aerobic
,  
3 PHA 
 ∗  ∗ , / ∗ ∗  ∗ Lopez-Vazquez et al. (2009)
,           
degradation

4 Poly-P 
 ∗

  ∗
, , 
∗ 

∗ 


 Lopez-Vazquez et al. (2009)
 
formation , , ,               


5 Glycogen 
 ∗

  ∗ , ∗
 , , 
∗ 

∗ 
 Lopez-Vazquez et al. (2009)

formation , , ,           


6 Maintenance 
 ∗  ∗ Lopez-Vazquez et al. (2009)
   

* is the kinetic rate constant of the biomass species i on the solute species r during aerobic (aer) or anaerobic (ana) phase (Cmol Cmol-1 s-1), Xi is biomass concentration of species I
(Cmol L-1), Sr is the concentration of solute species r (mol L-1), Kr is the half saturation constant of solute species r (Cmol L-1) and , is the intracellular fraction of solute species r in
biomass species i (-).
12

Precipitation Model
The equilibria of the mineral phases were considered to be rate expressions into the PHREEQC software, are listed here:
kinetically governed. The kinetic expressions of the following • The nucleation kinetics was not included, as it is assumed
phosphate minerals were included in the model: amorphous that the bacteria cells and the extracellular polysaccharides
calcium phosphate (ACP) and hydroxyapatite (HAP) because of (EPS) provide sufficient nucleation sites to make the
shown presence in lab granules (Barat et al. 2011, Mañas et al. nucleation rate non-limiting (Ferris et al. 1987, Dupraz et al.
2011), vivianite (VIV) because of shown presence in activated 2005, Yilmaz et al. 2008, Lin et al. 2012, Liu et al. 2016).
sludge (Wilfert et al. submitted) and struvite (STR-NH4) • A very small mineral content (10-10, 10-15, 10-7 mol L-1 for ACP, HAP
because of shown presence in lab granules (Lin et al. 2012) and and STR) was inserted to represent the nuclei, since the minerals
the interesting metabolism of PAOs involving PO43-, Mg2+ and cannot precipitate without a nucleus (see Addendum 2). The
K+. In addition, ferrihydrite (Fe(OH)3, FER) is included in the values were chosen in such a way as to not affect the net content of
model because of: (1) having a positive SI in the aqueous phase total precipitated mineral in the time available.
when the equilibria between Fe2+/3+ was included, (2) being an • The same rate expression was used for the precipitation and
important contributor to surface complexation that could be dissolution process of a specific mineral. When the saturation
implemented in the next model upgrade and (3) the black spots index (SI) was negative, the final rate was multiplied with
seen on the granule surface in microscopic pictures, which are -1 to turn around the stoichiometric reaction equation and
suspected to involve iron (see Fig. 4 and Addendum 2 more produce the solutes instead of extracting the solutes from
information). The saturation index (SI) of the minerals was the aqueous phase as happens during precipitation.
calculated according to equation 1 (Parkhurst et al. 1980). • The included phases were studied to have surface-controlled
crystallisation processes, except for vivianite which follows the
spiral growth mechanism of Burton, Cabrera and Frank (BCF)
IAP
SI = log 
 (−) (1) (Mersmann 2001, Appelo et al. 2005, Schwertmann et al. 2008,
K
Madsen et al. 2014, Crutchik et al. 2016).
 • Surface-controlled crystallisation processes were included
log IAP = b, ∗ log a   (−)
(2) in the model by the surface representation of phase p (Ap),
 including the specific crystal surface area (m2 L-1) when that
information was available in literature (see Addendum 2).
where, IAP is the ionic activity product, Kp the equilibrium • An additional conversion was incorporated (CVIV) in the kinetic
constant for the mass action equation for the pth phase, bp,j rate expression for vivianite. The kinetic rate expression for
stoichiometric coefficient of the jth species in the pth mineral vivianite available in literature was expressed in crystal length (m
and aj the activity of the jth species. A positive SI indicates that s-1) (Madsen et al. 2014). The conversion translates the crystal length
the respective mineral can precipitate. The activity coefficientts to molar concentration (mol L-1). This was done based on specific
were calculated according to the extended Debye-Hückel length measurements reported by Madsen et al. (2014).
equation (Truesdell et al. 1974, Parkhurst et al. 2013). • It is accepted that the least thermodynamically stable mineral
The Minteq database was taken as basis for the equilibria (ACP) acts as a precursor for the most stable phosphate
of the phases, but only the relevant phases were included (Minteq calcium mineral (HAP), following the Ostwald ripening
database 2006). The stoichiometry reaction equations and the rule (Koutsoukos et al. 1980, van Kemenade et al. 1987,
respective equilibrium constants of the 22 considered phases can Madsen et al. 1991, Musvoto et al. 2000). This was inserted in
be found in Addendum 2. The following minerals were added as PHREEQC by adjusting the stoichiometry of HAP: one mol of
solid phases to the database: hydroxyapatite (HAP), amorphous ACP was dissolved for each mol HAP precipitated.
calcium phosphate (ACP), struvite (STR-NH4), potassium-struvite • The precipitation process was modelled as a reversible
(STR-K) and sodium-struvite (STR-Na) according to Table 5. crystallisation for all phases except for HAP, which is modelled
The kinetic expressions of the mentioned minerals as an irreversible crystallisation, following Maurer et al. (1999)
were found in literature of pure crystal systems (see Table 6). The and Barat et al. (2011). When the SI of HAP was negative,
expressions were chosen based on the research method (pure no dissolution took place as the rate was set at zero.
aqueous systems) and the fitness to the PHREEQC software, • The formation rates for ferrihydrite are scarcely reported. The
which expresses rates in molar concentrations, not in crystal general kinetic expression of Koutsoukos et al. (1980) was therefore
length. The additional simplifications necessary to incorporate the applied with surface area measurements of Schwertmann et al.
13

Table 5 Solubility products of precipitates at 25oC


Table 5 Solubility products of precipitates at 25oC
Mineral Reaction equation pKp pKp,m* Reference

Amorphous Calcium Phosphate Ca3(PO4)2:6H2O = 3 Ca+2 + 2 PO4-3 + 6H2O 26.52 27.32 Mañas et al. (2011)
(ACP) 26.53 Seckler et al. (1996)

28.92 Minteq database (2006)

Hydroxyapatite Ca5(PO4)3OH:6H2O = 5 Ca+2 + 3 PO4-3 + 1 OH- + 6 H2O 57.5 50.14 Mañas et al. (2011)

(HAP) 48.6 Murray et al. (1996)

44.33 Minteq database (2006)

Struvite NH4MgPO4:6H2O = Mg+2 + PO4-3 + NH4+ + 6H2O 13.26 13.46 Ohlinger et al. (1998)

(STR-NH4) 13.15 Taylor et al. (1963)

13.97 Rahaman et al. (2006)**

13.26 Crutchik et al. (2016)***

Potassium Struvite KMgPO4:6H2O = Mg+2 + PO4-3 + K+ + 6H2O 12.2 11.50 Xu et al. (2015)

(STR-K) 11.68 Luff et al. (1980)

10.62 Taylor et al. (1963)

Sodium Struvite NaMgPO4:7H2O = Mg+2 + PO4-3 + Na+ + 7H2O 11.6 11.60 Xu et al. (2015)

(STR-Na)

Vivianite Fe3(PO4)2:8H2O + 2 H2PO4- = 3 Fe2+ + 4 HPO42- + 8 H2O 35.77 35.88 Al-Borno et al. (1994)

(VIV) 36.00 Minteq database (2006)

*Solubility product (pK) of phase p inserted into the model

**measured at 200C

***measured at pH 8.5

Table
Table 6 Kinetic rate expression
3 Kinetic for precipitation
rate expression and dissolution*
for precipitation of amorphous calcium
and dissolution* phosphatecalcium
of amorphous (ACP), hydroxyapatite (HAP),
phosphate (ACP),
struvite (STR-NH4), vivianite (VIV) and ferrihydrite (FER)**
hydroxyapatite (HAP), struvite (STR-NH4), vivianite (VIV) and ferrihydrite (FER)**
Mineral Reference

ACP precipitation &   ∗  
 /

dissolution =  ∗  ∗   − 1 Mbamba et al. (2015) Nielsen (1984)
 

HAP precipitation only  


=  ∗  ∗   ∗    −   Inskeep et al. (1988) Koutsoukos et al. (1980)
  ∗   ∗  

STR–NH4 precipitation  ∗  ∗ 


/ 

& dissolution =  ∗  ∗   − 1 Mbamba et al. (2015) Nielsen (1984)


 

VIV precipitation &     ∗  


dissolution =  ∗  ∗  − ,  ∗ ln Madsen et al. (2014) Madsen (2002)
 

FER precipitation & 


dissolution =  ∗  ∗    ∗   −  / Koutsoukos et al. (1980), Schwertmann et al. (2008)


*When the saturation index (SI) is negative, dissolution occurs and the rate is multiplied by -1 to reverse the stoichiometric equation, except for HAP, which is modelled as an irreversible
crystallisation process

**Pp is the mineral concentration of phase p (mol L-1), kp is the kinetic constant of phase p (mostly, s-1, depending on rate expression), Ap surface representation of phase p (mostly, m2 L-1,
depending on rate expression), a[r]z is the activity of solute species r with stoichiometric coefficient z (mol L-1), Kp is the solubility constant for phase p (mol L-1) and Sr is the concentration of
solute species r (mol L-1). More specifics are shown in Addendum 2.
14

(2008) and Cornell et al. (2003). A low rate coefficient was adapted, middle of each computational cell to calculate the local
to not overestimate the precipitation of ferrihydrite. concentration before moving into the next transport step.
• The real observed rate might differ from literature due • The end of cell 10 was closed to represent the core of
to density differences and the gel-like ambient inside the the granule as the end of the planar biofilm (see Fig. 5).
granule (Mersmann 2001, Toh et al. 2003). This drawback of Thus, no diffusion was possible further than cell 10.
the model was accepted, since the aim was to discover only the • The software did not allow that a potential difference
parameter interdependencies on phosphate precipitation. remained between computational cells after a transport step, as
• Inhibitor effects of magnesium on ACP and HAP precipitation electron neutrality is a prerequisite in the PHREEQC code.
was not taken into account for simplicity (Boskey et al. 1974, • The porosity was set to 0.5 as is calculated from oxygen
TenHuisen et al. 1997, Kanzaki et al. 2000). The interference of profiles inside aerobic granules using micro-electrodes
calcium on struvite precipitation through the competition for (Picioreanu 2015).
phosphate ions was taken into account through the chemical
equilibria (Bouropoulos et al. 2000, Minteq database 2006).
• Temperature effects on the solutes speciation and mineral
solubility products were taken into account according to the
Van ‘t Hoff equations with the reaction enthalpies stated in
Addendum 2 and the Minteq database (2006).
A more elaborate explanation on the mineral
crystallisation process of the calcium, magnesium and
iron phosphate minerals is available in Addendum 2.

Ion transport
The PHREEQC software calculated the ion transport, taking
into account the multi component diffusion coefficients,
the electrical balance, the activity balance and the porosity
(Parkhurst et al. 2013). The multi component diffusion
coefficients of the relevant solutes (see chemical equilibria) were
transferred from the PHREEQC database into the edited Minteq
Fig. 4 Microscopic picture of an aerobic granule cut in half from
database (PHREEQC database 2016). An overview of the
the full-scale installation of Dinxperlo (see Addedum 3 for more
diffusion coefficients inserted is presented in Addendum 3.
information). Bar indicates 1000 µm
The following simplifications in the model were
necessary in order to comply with the limitations of the
PHREEQC software or to reduce the computational time: 3D Spherical
• The only transport mechanism inside the granule was diffusion,
as the granule consisted of dense biofilm. The diffusion was
modelled in a 1D planar biofilm, as the current PHREEQC
1D Planar
version only allowed for radial diffusion when both mobile (e.g.,
dispersion) and immobile (e.g., clay particle) computational cells
were included (Parkhurst et al. 1999, Parkhurst et al. 2013). In this
1 2 3 4 5 6 7 8 9 10
study, only immobile computational cells were present to represent
the granule. The spherical granule is therefore transformed in
a planar 1D biofilm as illustrated in Fig. 5.
• Ten computational cells where inserted, with each cell
representing one layer of biomass. The computational cell-
length was 10-4 mm for all cells in order to mimic a radius Radius 1 mm Radius 1mm
of 1 mm. It was chosen to simulate a granule with a radius
of 1 mm, as the potential of precipitation is higher in larger Fig. 5 Schematic illustration of the transformation of a spherical
granules than in smaller granules. The software takes the granule into planar diffusion in PHREEQC software
15

Boundary conditions Table


Table7 XX
Dutch raw waste
Dutch water water
raw waste composition
composition
The Dutch waste water composition (see Table 7) was taken as
Parameter Min Max Ave. Reference
input for the boundary conditions at the granule’s surface. The
exact solute concentrations inserted in the model are shown mg/L mg/L mg/L

in Table 8. First, the solute concentrations at the boundary of 524.04 CBS (2015)
COD
the granule were set constant for 1 hour to the averaged values 146 715 506 Pronk et al. (2015)

of the Dutch waste water composition. These conditions are 119.92 Barat et al. (2005)
BOD
the solute concentrations a granule at the bottom of a full-scale 60 420 224 Pronk et al. (2015)

AGS reactor experienced during feeding, when fresh influent Acetate 10.62 19.53 19.47 Garmerwolde, This study
rises continuously from the bottom in plug flow for 1 hour. 8.31 CBS (2015)
Afterwards, the aeration phase started and lasted for 3 hours. TP 1.9 9.7 6.7 Pronk et al. (2015)
3.7 15 10.2 Garmerwolde, This study
The aerobic phase was split in 3 sets of 1 hour, as the solute
concentrations in PHREEQC could not be inserted as a time- 5.1 7.4 5.96 De Kreuk (2006)
PO4
1.5 6.8 4.4 Pronk et al. (2015)
dependent function. The solute concentrations during the aerobic
phase were put constant at the boundary of the granule for each 48.15 CBS (2015)
48.3 69.3 54.13 De Kreuk (2006)
set of 1 hour as shown in Table 8. The solute concentrations TN
14 81 49.4 Pronk et al. (2015)
were determined based on offline cycle measurements by ion 48.7 61.1 56.78 Garmerwolde, This study

chromatography (IC) for the metal concentrations (see Analytical 0.82 63.3 34.75 Watson Database (2016)
Methods). For the other solute concentration, the online database NH4-N 28.3 39.4 33.41 De Kreuk (2006)
13.4 56.5 39 Pronk et al. (2015)
of Royal Haskoning DHV was used (Royal Haskoning DHV
14 140 38.38 Watson Database (2016)
2016). In the end, a typical AGS cycle was represented by 1 hour SO4
19.1 36.1 30.4 Garmerwolde, This study
of anaerobic feeding and 3 hours of aeration. To simulate a rain
2.35 21.3 9.043 Watson Database (2016)
event, the boundary conditions were put four times lower than
Mg 12 18 Schönborn et al. (2001)
the standard cycle solute concentrations (see Table 8). 11.3 13.9 12.8 Garmerwolde, This study
Some additional assumptions were made to insert the 20.7 97.9 61.03 Watson Database (2016)
boundary conditions in PHREEQC: Ca 52 57 Schönborn et al. (2001)
53.9 77.0 68.95 Garmerwolde, This study
• The only present substrate was acetic acid with a negative
charge (Acetate-). It was assumed that all biochemical oxygen 0.405 51.4 3.461 Watson Database (2016)
Fe 1.14 2.68 1.76 Dinxperlo, (Stubbé et al. 2016)
demand (BOD) would become available as acetate during 0.5 1.5 3.93 Hvitved-Jacobsen et al. (2013)
feeding. Little data was presented on the anaerobic digestion
21 33 Schönborn et al. (2001)
products of BOD (butyrate, propionate, etc.) and the respective K
27.6 32.3 29.2 Garmerwolde, This study
uptake rates inside aerobic granules in full-scale AGS
87.50 123 108.3 Garmerwolde, This study
systems. Besides, acetate is reported to be the most common Na
88.6 Utrecht, This study
volatile fatty acid (VFA) in waste water (Wang et al. 2013). F 0 0.17 0.1 Garmerwolde, This study
• The higher average BOD concentration of Dutch waste
Tempera- 12.9 20.6 15.71 De Kreuk (2006)
water was taken for calculating the acetate concentration ture (˚C) 8 23 16 RoyalHaskoningDHV (2016)
(± 225 mg L-1) (see Table 7). A yield of 0.6 mgAc mgBOD -1
7.00 8.6 7.64 Garmerwolde (Stubbé et al. 2016)
was assumed, leading to ± 135 mg L-1 acetate concentration pH (-)
7.18 7.47 7.33 Utrecht (Stubbé et al. 2016)
(Schroeder 1968). The final acetate concentration was Alkalinity 294 506 430 Utrecht, This study
increased to 200 mg L-1 to correct for some biodegradable (HCO3-)
chemical oxygen demand (COD). The measured acetate
concentration (≈ 20 mg L-1) in raw waste water
was deemed too low, as the VFA digestion of
raw waste water can generate more available acetate.
• Ammonium adsorption was taken into account by
adjusting the ammonium concentration 23 % lower in the
first hour of aeration (Stubbé et al. 2016).
• Iron was inserted as total Fe. The kinetics of the Fe2+
16

oxidation to Fe3+ was not taken into account for simplicity Table9 XX
Table Initial
Initial conditions
conditions in the
in the PHREEQC
PHREEQC codecode
for
(Stumm et al. 1996, Morgan et al. 2007). computational cells 1 - 10
for computational cells 1 - 10
• The excess sludge phase was neglected as this phase
time was relatively small (± 10 min) compared to Parameter Value Dimension Reference
the other operational phase (4 - 7 hours). Biomass 56.5 gL -1
Mosquera-Corral et al. (2005)
( ) Beun et al. (2000)

Initial conditions Pha 0.55  CmolPHA Lopez-Vazquez et al. (2009)


The same initial conditions were inserted for all computation Cmolbiomass-1

cells. The granule biomass concentration (XPAO) was set constant at Glycogen 0.27  Cmolglycogen Lopez-Vazquez et al. (2009)
56.5 gbiomass Lbiomass-1 (biomass density), which was the averaged Cmolbiomass-1

value obtained from literature for similar AGS systems. The initial Poly-P 0.30  PmolPoly-P Lopez-Vazquez et al. (2009)
biopolymer content of the biomass (PHA, glycogen and Poly-P) Cmolbiomass-1

was set to the maximal ratios stated in Lopez-Vazquez et al. (2009)


0.55 CxPAO, 0.27 CxPAO and 0.30 CxPAO (Cmol m-3) (see Table 9).

Table 8 Concentration inserted in the PHREEQC model as boundary conditions at the surface of the granule
Table XX Concentration inserted in the PHREEQC model as boundary conditions at the granule’s surface
Component Standard Cycle Rain Cycle (dilution is ± 4)*

Phase Anaerobic Aerobic Anaerobic Aerobic

Duration 1h 1h 1h 1h 1h 1h 2h 3h

Dimension mg L-1 mg L-1 mg L-1 mg L-1 mg L-1 mg L-1 mg L-1 mg L-1

Acetate- 200 0 0 0 50 0 0 0
PO4 -P
3-
10 20 5 0 2.5 5 1 0

NH4 -N
+
35 25 10 5 9 7 2 1

NO3--N 0 0 1.5 3 0 0 0.4 1

SO42--S 30 30 30 30 8 8 8 8

Mg2+ 10 12.5 10 7.5 2.5 3 2.5 2

Ca2+ 60 60 60 60 15 15 15 15

Fe2+/3+ 3 0.8 0.5 0.2 1 0.2 0.1 0.05

K+ 30 35 30 25 8 9 8 7

Na +
90 90 90 90 22.5 22.5 22.5 22.5

Cl (charge)**
-
40 40 40 40 10 10 10 10

F
-
0.1 0.1 0.1 0.1 0.025 0.025 0.025 0.025

O2 0 1 1 3 0 1 1 3

Temperature 15 15 15 15 15 15 15 15

pH 7 6.8 6.8 6.5 7 6.8 6.8 6.5

Alkalinity*** 400 300 250 200 100 80 60 50

*The waste water is diluted by approximately a factor four by urban run-off during rain events

*The Cl- concentration is adjusted by PHREEQC to create a neutral boundary solution, the granule’s internal neutrality is adjusted by pH

**Alkalinity as mgHCO3- L-1, including the contribution of phosphate (2 H+), acetate (H+) and carbonate (2 H+)
17

3. Materials and Methods

Reactor description and 881 Compact IC Pro, the Netherlands) in combination with
Two full-scale aerobic granular sludge (AGSA) installations were MagICNet 2.1 software for data acquisition. The iron concentration
examined during this research: one in Garmerwolde (2 reactors of in the influent was measured with standard kits (Dr. Lange type
9,600 m3 each with a combined flow of 4,200 m3 h-1) and one in Utrecht LCK302, Hach Lange, Dusseldorf, Germany). Organic substances
(1 reactor of 1,050 m3 with a flow of 125 m3 h-1), both located in the in the influent were measured with a High-Performance Liquid
Netherlands. Both installations were operated as a sequence batch Chromatograph (HPLC) in combination with a BioRad Aminex
reactor with cycles between 4 to 10 hours. The first feeding phase was HPx-87H column and a UV/RI detector (waters 2489). The
± 1 hour and always anaerobic, followed by a 3 to 9 hours aerobic alkalinity of the influent was measured according to standard
and settling phase. See the complete description of the full-scale methods for alkalinity titration to pH 4.3 as mgHCO3- L-1 by adding
installation in Stubbé et al. (2016). The influent of the full-scale 0.01M HCl solution with a 702 SM Titrino (Metrohm, Switserland)
AGS system in Dinxperlo was also investigated, as this installation in combination with a 728 Stirrer from Metrohm (APHA 1915).
was located more in the east of the Netherlands. This installation In the same manner, biomass harvested from the aeration phase
was operated in a similar manner as the installations in Utrecht was titrated; both reactor sludge and granule fractions (> 1.6 mm
and Garmerwolde. The detailed description of the AGS system in and 0.2 - 0.6 mm). The granule fractions were made after sieving
Dinxperlo can be found in Addendum 3 (STOWA 2015). and washing the reactor sludge prior to the measurement, within
72 hours of sampling. A slow titration speed (10 - 0.01 µL min-1) was
pH profile in granule and reactor adapted for the biomass to minimize diffusion limitations.
For this study, Unisense pH micro-electrodes (LS18 Unisense, Microscopic pictures were made with the Image
Denmark) and micromanipulator (model MM3M, Unisense, Analyser Leica 80 in combination with Leica software. X-ray
Denmark). Data acquisition is done by ADC216 USB diffraction (XRD) analyses were performed on reactor granules
converter (Unisense, Denmark) and Pyro Profix 4.5 software. (> 0.2 mm) sampled in the aerobic phase and air-dried for > 48
The granules were harvested from the full-scale AGS h. The XRD measurements were done with Bruker D8 Advance
system in Utrecht during the aerobic phase and sieved (> 2 mm). diffractometer Bragg-Brentano geometry, lynxeye position
The analyses were done within 48 hours after sampling and sensitive detector and divergence slit V12 with data evaluation by
the granules were stored in between at 4°C. The profiles were Bruker software Diffrac. EVA vs 4.1. A cobalt tube scattered from
measured with steps of 50 μm, 5 s stabilisation time and Utrecht’s 5° to 90° while sample was spinning, using step size 0.020° and
raw waste water (influent after grit removal) and effluent as counting time 1 s per step.
medium. The medium was sparged with N2 to make the influent
anaerobic and with compressed air to make the effluent aerobic
during at least 15 min prior the measurement. One measurement
in influent and effluent media was done consecutively in the same
granule, directly after each other. The fixed point measurements
were made without motion of the electrode, which was positioned
in the middle of the granule. Different 0.1M Tris buffer media at
pH 6 to 8.5 were used as medium. In both the profile and fixed
point measurements, the flow was set at 80 mL min-1.
The pH of the bulk liquid in the reactor was
measured online during the whole cycle (Online Process
pH sensor DPD1P1, Hach Lange, Germany) from December
2015 to present day in Utrecht and from August 2014 to May
2015 in Garmerwolde. The data was retrieved from Royal
Haskoning DHV’s database (Royal Haskoning DHV 2016).

Analytical methods
For validation of the raw waste water concentrations found in
literature, the cation and anion concentrations in the influent of
full-scale AGS installation in Garmerwolde were measured with
an ion chromatograph (IC) (Metrohm, type 883 Basic IC Plus
18
19

4. Experimental Results

Aerobic granular sludge reactor influent pH compared (± 7.5) to the reactor pH (6.5 – 7). The
A typical cycle of Utrecht and Garmerwolde is depicted in Fig. 6. anaerobic peak in Garmerwolde can only be seen after mixing,
The feeding phase (0:00 - 1:00 h) does not show any concentration since the sensor is hanging at the top of the reactor. The peak in
profiles, as the sensors for phosphate, ammonium, nitrate and Utrecht is considerably larger than in the lab reactors of Serralta
oxygen were hanging at the top of the reactor. After one hour of et al. (2004), probably due to the lower influent pH used (6.5
feeding, the reactors were mixed by aeration (1:00 – 7:00 h). The instead of 7.4 in Utrecht). The pH level in Garmerwolde was
concentration profiles of phosphate and ammonium in time showed significantly lower than in Utrecht, which can be explained
a peak due to the mixing of the influent with the rest of the reactor by: (1) more mature granules releasing more phosphate and
and the released phosphate. Both phosphate and ammonium were (2) a higher influent pH in Garmerwolde (7.6 instead of 7.4 in
taken up by the bacteria, causing the decrease in concentration Utrecht). Firstly, mature granules in Garmerwolde contained
until the cycle ends. Nitrate is formed during aeration and partly more Poly-P, which can be hydrolysed during anaerobic feeding,
converted as the oxygen levels decrease towards the end of the releasing more H2PO4-, and thus reducing the pH. A higher
cycle. The microscopic pictures illustrate the average granule size influent pH increases the energy needed to transport acetic
distribution in the full-scale AGS reactors (see Fig. 7). acid over the cell membrane (Smolders et al. 1994, Filipe et al.
The metal concentration profiles of Ca2+, Mg2+, K+, Fe2+/3+
in time are depicted in Fig. 8 for one cycle (feeding, aeration and Utrecht
settling). The concentration profile of magnesium and potassium
in time followed the phosphate concentration profile, endorsing
the release and uptake of these ions for the Poly-P metabolism
of PAOs (Barat et al. 2005). The calcium concentration profile
in time was rather constant throughout the cycle. This was also
reported by Mañas et al. (2011) when they used weak aeration.
Barat et al. (2008) reported constant calcium concentration profiles
when using low calcium influent concentrations. The total iron
concentration profile (Fe2+/3+) decreased throughout the cycle with
the strongest decline during the anaerobic phase. The cause of
this decline is not directly researched in this study. However, iron
could disappear from the bulk solution by surface complexation,
precipitation (with PO42- and SO42-) or biological incorporation into
the bacteria cells (Ayala-Castro et al. 2008). The model results can
Garmerwolde
indicate the contribution of precipitation to the disappearance of
iron. Although these metal concentration profiles in time were only
measured once, it is believed that these concentration profiles were
representative for a cycle during dry weather conditions.

Bulk pH profile in time


Since the pH is an important parameter for precipitation (Yilmaz
et al. 2008, Mañas et al. 2012, Wan et al. 2015), the bulk pH
profile in time was studied at the full-scale AGS installations in
Utrecht and Garmerwolde (see Fig. 9). The bulk pH profile of
Garmerwolde shown in Fig. 9 does not coincide with the cycle
shown in Fig. 6 since the pH sensor was only operative until
May 2015 at this installation. The bulk pH profile in Utrecht
does coincide with the cycle shown in Fig. 6, but the pH sensor
hanged at sludge bed height. In Garmerwolde, the pH sensor
Fig. 7 The microscopic pictures of sludge in the reactors
hanged at the top of the reactor with the other sensors.
of full-scale AGS installations in Utrecht (above) and
The peak at the beginning of the anaerobic phase in the
Garmerwolde (below) illustrate the average granule
bulk pH profile was caused by acetic acid uptake and a higher
fraction distribution. Bars indicate 1000 µm.
20

Utrecht
35 3

Anaerobic Aerobic
NH4 , NO3 , PO4 concentration (mg L-1)

30
2.5

O2 concentration (mg L-1)


PO4
25
2

20
NH4 O2
1.5
15

1
10

NO3 0.5
5

0 0
0:00 1:00 2:00 3:01 4:01 5:02 6:02 7:03
Garmerwolde Time (hh:mm)
25 25 3
NH4 , NO3 , PO4 concentration (mg L-1)
NH4 , NO3 , PO4 concentration (mg L-1)

Anaerobic
Anaerobic Aerobic
Aerobic 2.5
2.5
20 20
(mgLL-1-1))

22
concentration(mg

NH4
NH4
15 15
2 2concentration

PO4
1.5
1.5
PO4

10
10 O2
O2
1
NO3
NO3 1
OO

5
5 0.5
0.5

0 0
0
0:00 1:00 2:00 3:01 4:01 5:02 6:02 0
0:00 1:00 2:00 3:01 4:01 5:02 6:02
Time (hh:mm)
PO4 NH4
Time (hh:mm)
PO4
NO3 NH4
O2
NO3 O2
Fig. 6 A typical cycle of full-scale AGS systems in Utrecht (above) and Garmerwolde (below). The concentration profiles during feeding
(0:00 – 1:00 h) were rather constant, as the sensors were hanging at the top of the reactor, while influent was introduced at the bottom of
the reactor. After mixing by aeration (1:00 – 7:00 h), the profiles of phosphate, ammonium, nitrate and oxygen appeared.
21

2001). This leads to more Poly-P hydrolysis, more H2PO4- release is dependent on the aeration regime and phosphate uptake rate
and thus lower pH. These combined circumstances lead to a (Serralta et al. 2004). High dissolved oxygen and phosphate
lower pH profile in Garmerwolde compared to Utrecht. concentration induces a pH increase by CO2 stripping and the
The pH profile of the aerobic phase first increased and P-uptake whereas low dissolved oxygen and zero phosphate
afterwards decreased, in both full-scale reactors as lab-reactors concentration induce a faster pH decrease by CO2 accumulation
(Serralta et al. 2004, Barat et al. 2011). The pH increase at the and nitrification (Serralta et al. 2004). The linear, slight increase in
beginning can be caused by the dominating effect of the P-uptake pH at the end of the aerobic phase in Garmerwolde is caused by
and CO2 stripping by aeration over the decreasing pH-effect of denitrification (see Fig. 9).
nitrification. The length of the pH increase in the aerobic phase

70 5

Anaerobic Aerobic 4.5


60
Ca2+
Ca, Mg, K concentration (mg L-1)

Fe concentration (mg L-1)


50 3.5

3
40
2.5
30
K+ 2

20 1.5

Mg2+ 1
10
0.5
Fe2+/3+
0 0
0:00 1:00 2:00 3:01 4:01 5:02 6:02 7:03
Time (mm:hh)

Ca K Mg Fe
8.0concentration during a typical cycle in Garmerwolde
Fig. 8 Metal

8.0 Anaerobic Aerobic


7.8
Anaerobic Aerobic
7.8 x y
PNU 4:29 7.2
7.5
x GMW y 4:10 6.65
PNU 4:29 7.2
7.5 GMW 4:10 6.65

7.3
Utrecht
pH (-)

7.3
Utrecht
pH (-)

7.0
7.0

6.8
6.8
Garmerwolde
Garmerwolde
6.5
6.5
0:00 0:00 1:00 1:00 2:00 2:00 3:00 3:00 4:00 4:00 5:00 5:00 6:00 6:00 7:00 7:00
Time (hh:mm)
Time (hh:mm)
UtrechtUtrecht Garmerwolde
Garmerwolde

Fig. 9 Bulk pH profiles in time in the full-scale AGS installations in Utrecht and Garmerwolde during a typical cycle. The anaerobic
peak in Garmerwolde can only be seen after mixing, since the sensor is hanging at the top of the reactor, whereas in Utrecht the sensor
is hanging at sludge bed height.
22

8,0
Granule pH profile in space Influent
The micro-electrode measurements can indicate the pH profile in (Anaerobe)
7,8
space under different conditions inside the aerobic granule. The
profiles pointed out a pH decrease inside the granule when the Effluent
7,6
granules were submerged in raw (anaerobic) waste water medium (Aerobe)

pH (-)
and a pH increase when submerged in (aerobic) effluent media 7,4

(see Fig. 10). Repeated measurements pointed out that the pH in


7,2
the granules always dropped in anaerobic medium. Possibly, the
pH decrease caused by the metabolic phosphate release, which
7,0
releases 1 or 2 protons, was stronger than the pH increase caused by
Bulk Granule Bulk
consumption of acetate and producing simultaneously K+ and Mg2+ 6,8
ions (Smolders 1995, Serralta et al. 2004, Barat et al. 2005, Saunders 0 1000 2000 3000 4000 5000 6000 7000 8000
et al. 2007, Barat et al. 2011). The acidification could also take Depth (µm)
please due to fermentation of available substrate or precipitation Influent (anaerobe) Effluent (aerobe)

(Viéitez et al. 1999, Dupraz et al. 2009). A pH decrease is reported Fig. 10 Granule pH profiles in space when the granule
when granules are crushed in tap water, demonstrating a slight was embedded in anaerobic media of raw waste water
acidic interior (de Smit et al. 2016). Besides, it is also possible (pink) and when embedded in aerobic media of effluent
that the biofilm contains a pH buffer capacity that delays the water (blue). The granules and media were harvested from
adaptation of the bulk pH. During some aerobic measurements, a the full-scale AGS installation in Utrecht.
pH decrease in aerobic effluent conditions was measured, instead
of a pH increase. This can be explained by the balance of the pH-
influencing processes in this phase: P-uptake and denitrification 8,0
t = 120 min
(pH increase) when no oxygen is present and nitrification (pH t = 90 min
7,8
decrease) when oxygen is present. A change in the dominating t = 60 min
t = 30 min
process could have caused the change in aerobic pH profiles. 7,6 t = 0 min
Multiple profiles in space were made in the same
pH (-)

granule to assess the impact of the micro-electrode piercing the 7,4

biofilm (see Fig. 11). Over the course of 2 hours, five profiles
7,2
were made consecutively in media of raw waste water. Although
the pH of the bulk slightly increased over the course of the 7,0
measurements, due to nitrogen gas sparging, a pH gradient inside Bulk Granule Bulk
6,8
the granule remained. These results showed that the biofilm is not
0 1000 2000 3000 4000 5000 6000 7000 8000
entirely broken by the micro-electrode, which would have had Depth (µm)
let the protons diffuse freely. The decrease in pH gradient over 2
hours (from ± 0.4 to ± 0.2) can be caused by: (1) the depleting pH
Fig. 11 Multiple granule pH profiles in space when the
buffer capacity of the biofilm (2) some increase in porosity caused
granule was embedded in anaerobic media of raw waste
by the micro-electrode or (3) less phosphate release due to the
water. The granule and media were harvested from
depleting Poly-P content of the PAOs (Lemaire et al. 2008).
the full-scale AGS installation in Utrecht.

Granule pH point measurements in time


The buffer capacity of granules was investigated with pH point
measurements in time. The micro-electrode was positioned in the
core of the granule, while changing the bulk pH of the media in
time. The bulk pH shifted slightly during the point measurements
due to the small cascade in the experimental set-up.
Starting with a bulk pH of 7, the granule interior
pH increased towards ± 7.6 in 15 min (see Fig. 12). Possible
explanations for this trend are: (1) an (amorphous) precipitate was
23

dissolving inside the granule, (2) surface complexation equilibria 8,0 8,0
7,8 7,8
are shifting or (3) the EPS structure starts to disintegrate and
7,6 7,6
releases Ca2+ (Pijuan et al. 2009). Biological reactions were ruled Granule pH

Granule pH (-)
7,4 7,4
out, as no substrate was available in the Tris buffer media.

Bulk pH (-)
7,2 7,2
When the bulk solution was changed to pH 7.5, the 7,0 7,0
interior pH also increased, although the pH in time stayed constant. 6,8 6,8
Apparently, an equilibrium had been established inside the granule, 6,6 6,6
that the pH does not increase any longer. When the bulk solution 6,4 6,4
Bulk pH
was changed to pH 6.5, the equilibrium platform of interior pH 6,2 6,2
6,0 6,0
decreased only slightly. The granule was capable of maintaining
0:00 0:14 0:28 0:43 0:57 1:12
a pH difference of ± 1.4 with the bulk for over 30 min. Only when
Time (hh:mm)
the measurement was extended in time (almost 5 h instead of 1
h), a sudden decrease towards the medium pH was measured.
Granule pH
Fig. 12 Point pH measurements BulkofpH
in the core the granule when
This could be indicating a depletion of available (amorphous)
embedded in 0.1M Tris buffer media with pH 7, 7.5 and 6. The
precipitate or complete disintegration of the EPS, ending the surface
gaps indicate the time of changing the media solution with
complexation and Ca2+ release. In the Discussion section the pH
a different pH.
behaviour in time inside the granule is further elaborated .
Overall, the granule pH point measurements in time
showed a certain buffering capacity in aerobic granules, although Utrecht and Garmerwolde
the extent of the capacity was not quantified.
250,0

Alkalinity of biomass
Alkalinity (mgHCO3 L-1 pH-1)

Influent
200,0
The extent of the granule buffering capacity was further
investigated by measuring the alkalinity of reactor granules
and granule fractions compared to the bulk media (influent, 150,0

effluent, reactor, tap water and demi water). In the Netherlands, Effluent Reactor Tap
tap water is used for all water consumption (shower, kitchen, 100,0
water
garden, e.g.) and therefore formed the basis for the waste water
Demi
content. Demi water hardly contained any alkalinity (± 2 mgHCO3 50,0
water
L-1 pH -1) and was therefore used as media for the alkalinity
measurements of the granule fractions. The alkalinity was 0,0
expressed as mgHCO3- L-1 pH-1 or mgHCO3- L-1 mgTSS-1 pH-1 in 0,025
order to make a fair comparison between the measurements. Reactor > 1.6 mm 0.2 -
crushed crushed 0.6 mm
The alkalinity of the influent (± 170 mgHCO3 L-1 pH- 0,020
Alkalinity (mgHCO3 L-1

crushed
1
) was noticeably the highest source of buffer capacity in the Reactor 0.2 -
mgTSS -1 pH-1)

>1.6 mm 0.6 mm
full-scale installations in Utrecht and Garmerwolde (see Fig. 0,015
13). The alkalinity decreased throughout the cycle via VFA
consumption, CO2 stripping by aeration and phosphate uptake 0,010

towards the effluent alkalinity (± 80 mgHCO3 L-1 pH-1). The


reactor alkalinity (± 83 mgHCO3 L-1 pH-1) was measured 10 min 0,005

after aeration started, showing that the VFA consumption was


0,000
mostly governing the alkalinity and not the phosphate uptake.
The CO2 stripping in the first 10 min could have contributed Fig. 13 Alkalinity of the bulk solutions without biomass
slightly to the alkalinity decrease as well. The effluent and (influent, effluent, reactor, tap water and demi water solution)
reactor alkalinity were slightly higher than the alkalinity of from Utrecht and Garmerwolde (above). Alkalinity of (crushed)
tap water (± 64 mgHCO3 L-1 pH -1), which could be caused by mixed granules and sieved granule fractions of Utrecht (below).
residues of phosphate, carbonate and biomass. Demi water The granule fractions were embedded in demi water. The
hardly contained any alkalinity (± 2 mgHCO3 L-1 pH -1). reactor granules were embedded in reactor solution.
24

The crushed reactor granules and granule fractions In January, the XRD of granules from Garmerwolde (> 0.2 mm)
measurements showed higher alkalinity than the non-crushed showed quartz crystals, which could have originated from sand
samples. It is believed this happened due to diffusion limitation particles in the influent (see Fig. 14). In December, however, the
of protons inside the granules. During the measurements it XRD of granules from Garmerwolde showed resemblance to the
became evident that a slower titration rate (0.01µL min-1 instead pattern of potassium struvite (STR-K) (Stubbé et al. 2016). The
of 10 µL min-1 to) resulted in diminishing differences between XRD of granules from the AGS system in Utrecht showed presence
crushed and non-crushed samples. Differences in alkalinity of brushite (CaHPO4.2H2O) in March, which could have formed
between big granules (> 1.6 mm) and small granules (0.2 – 0.6 during the cycle through precipitation or it originated from the
mm) were smaller than expected. The buffer mechanisms (proton influent (see Fig. 14). Brushite could have been formed inside
buffering by organic matter, EPS-trapped phosphate, possible pollen or seeds that reached the AGS installation with rain water
precipitates, e.g.) could have been sufficiently correlated with (Hall et al. 2003, Ajayi et al. 2013, Steinhorst et al. 2013). In January,
the amount of biomass to make the differences less profound the XRD of granules from Utrecht indicated the possible presence
(Appelo et al. 1998, van Breukelen et al. 2004, Zhang et al. 2013, of dolomite (CaMg(CO3)2), magnesium calcite (Mg0.06Ca0.94CO3) and
Stubbé et al. 2016). The alkalinity of reactor granules was, as ammonium phosphate (NH4PO3) (Stubbé et al. 2016).
expected, higher than the alkalinity of granule fractions. The These results point out that the crystal content of a
reactor granules were embedded in reactor solution, which granule from the full-scale installations in Garmerwolde and
had a higher alkalinity than demi water. The high standard Utrecht was dynamic; sometimes there were (different) minerals,
deviation of the reactor granules and small granule fraction sometimes not. Modelling the chemical processes can indicate the
(0.2 – 0.6 mm) was caused by the TSS measurement. conditions which facilitates certain precipitation.

X-Ray diffractograms of full-scale granules


X-ray diffractograms (XRD) can indicate whether crystalline
material is present in aerobic granules harvested from full-scale
AGS installations. The XRDs showed a mountain-shaped base line
and a mixture of peaks. The mountain-shaped base line was caused
by amorphous or organic substances. The peaks indicate crystalline
material. The pattern of the peaks is compared to reference
patterns in the database of known crystalline minerals.
Repeated XRD measurements showed clear peaks
corresponding to the quartz pattern (SiO2) each time, both in
Garmerwolde and Utrecht. The additional peaks in the XRDs
varied in time and corresponded to different mineral patterns.
25

Utrecht

Garmerwolde

Fig. 14 XRD of granules > 0.2 mm from Garmerwolde only showed quartz crystals (above), XRD of reactor granules from Utrecht
showed brushite crystals (below).
26

Cation concentration Phosphate concentration


6 6

4 VIV
4
VIV
2 HAP

SI of mineral phases (-)


SI of mineral phases (-)

ACP HAP
2
0
ACP
STR-NH4 0
-2 STR-NH4
STR-Na
-4
-2 STR-Na
STR-K
-6 STR-K
-4
-8

-6
-10

-12 -8
-100 -50 0 50 100 150 200 250 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Change in cation from raw influent concentration (%) Phosphate concentration (mol L ) -1 # 10-3

Temperature pH
4 8

3 6
VIV

2 4
SI of mineral phases (-)

SI of mineral phases (-)

1 2
ACP
0 0
HAP VIV
-1 -2
STR-NH4
-2 -4
STR-NH4
ACP
-3 -6
STR-Na STR-Na
STR-K
-4 -8
STR-K
HAP
-5 -10
5 10 15 20 25 30 35 5 5.5 6 6.5 7 7.5 8 8.5 9

Temperature ( C) o pH (-)

Fig. 15 Sensitivity analysis of influent solute concentrations, temperature and pH on the mineral SI in the aqueous phase simulated
with PHREEQC software in combination with the Minteq database (2006). Note that the y-axes are not the same in the four graphs
to show the small impact of the temperature on the mineral SI. The cation concentration changes involved Fe2+/3+ for vivianite (VIV),
Ca2+ for amorphous calcium phosphate (ACP) and hydroxyapatite (HAP), Mg2+ and NH4+ for struvite (STR-NH4), Mg2+ and K+ for
potassium struvite (STR-K) and Mg2+ and Na+ for sodium struvite (STR-Na).
27

5. Model Results

5.1 Precipitation in standard conditions

Precipitation in the aqueous phase A sensitivity analysis was performed to investigate


The aqueous phase was investigated in order to assess the the influence of the influent composition on the mineral SI. The
precipitation potential in the bulk liquid of the full-scale AGS solute concentrations, pH and temperature were changed one by
reactors and in the sewer system. Whether precipitation occurs one in the aqueous phase and PHREEQC calculated the change
when a mineral is supersaturated, depends on the mineral in mineral SI (see Fig. 15). The results showed that the pH had
nucleation and precipitation kinetics. The precipitation kinetics the largest influence on the mineral SI, especially on the SI of
of Table 6 on page 13 was not applied in the aqueous phase as HAP. The anion (PO43-) and involved cation (Ca2+, Mg2+, Fe2+, K+,
nucleation kinetics was not included. The saturation indices were NH4+, Na+) concentrations had a smaller impact on the mineral
only used to give an indication of the precipitation potential. SI. Compared to the other mineral SI, the SI of HAP was most
The speciation equilibria gave negative saturation sensitive for the solute concentrations (PO43- and Ca2+). It can be
indices (SI) for all phosphate minerals, except vivianite, in generic seen that a change in the Mg2+ concentration had the smallest
Dutch averaged raw waste water (see Table 10). The extremely effect. The temperature affects the ion speciation and mineral
high SI of vivianite was caused by the usage of HPO42- instead of solubility (Kp) and therefore also the mineral SI (Song et al.
PO43- activity in the stoichiometric equation, following the latest 2002, Yilmaz et al. 2008). The reported range of waste water
research on vivianite kinetics (Madsen et al. 2014). Madsen et al. temperature (10 – 30 °C), however, showed a minor effect on the
(2014) stated that HPO42- is the reactant for vivianite precipitation mineral SI. The SI of HAP and VIV showed a positive relation
in neutral or weakly acid solutions. The SI of vivianite was 3.19 with temperature, indicating endothermal reactions. The SI of
when PO43- was the reactant in the stoichiometric equation. ACP, STR-NH4, STR-K and STR-Na showed a negative relation
During this simulation, it was presumed that the influent total with temperature, indicating exothermal reactions. Nevertheless,
iron was only present as Fe2+, due to the influent pH range (7 – 7.6) the temperature was the parameter with the least influence on the
and oxidation state (-0.4 – 0.1 Volts) (Wilfert et al. 2015). Some iron mineral SI. The reported range of other waste water solutes (NO3-,
oxides (ferrihydrite, goethite, hematite, lepidocrocite, magnetite, SO43-, Cl-, acetate-) showed no or little influence on the mineral SI
maghemite) had positive SI when Fe3+ was taken into account in the aqueous phase. The sensitivity on the assumed Kp is studied
according to the Fe2+/Fe3+ equilibria (see Table 10). It is presumed and explained in the Model parameter sensitivity analysis.
that these iron oxides, except ferrihydrite, do not form in AGS
systems due to the slow kinetics at the pH range, temperatures
and short retention times in the reactor (Schwertmann et
al. 2008). Ferrihydrite can precipitate as it is the least stable
iron oxide and acts as a precursor for other iron oxides.
Table
Table10 10
Mineral saturation
Mineral indices
saturation in the in
indices aqueous phase of
the aqueous
The mineral SI in the aqueous phase were investigated
averaged
phase ofDutch wasteDutch
averaged water waste water
at the top of the sludge bed by taking into account the PO43-, Mg2+
and K+ release by PAOs during the anaerobic phase. Data from the Mineral Stoichimetry SI* SI**
online database of RHDHV pointed out that the concentrations Amorphous calcium Phosphate Ca3(PO4)2:6H2O -0.81 2.55
increased to 70 mg L-1 for PO43-, 25 mg L-1 for Mg2+ and 55 mg L-1 for (ACP)
K+ at the top of the sludge-bed. The pH increased to 7.8 during the Hydroxyapatite (HAP) Ca5(PO4)3OH:6H2O -0.59 5.20
anaerobic phase at the top of the sludge-bed. An exchange ratio of Struvite (STR) MgNH4PO4:6H2O -2.40 -0.30
40% was taken into account (Royal Haskoning DHV 2016, Stubbé
Potassium struvite (STR-K) MgKPO4:6H2O -4.76 -2.39
et al. 2016). These circumstances increased the SI of all phosphate
Sodium struvite (STR-Na) MgNaPO4:6H2O -3.95 -1.85
minerals significantly, although the magnesium minerals
Vivianite (VIV) Fe3(PO4)2:8H2O 13.42 15.93
remained undersaturated (see Table 10). Apparently, calcium
phosphate and iron phosphate minerals can be supersaturated Ferrihydrite (FER) Fe(OH)3 3.80*** 4.65***

in the aqueous phase surrounding the granules in the anaerobic *In raw influent

phase. These results support the granulation theory stating that **At the top of the sludge bed

precipitates can form in the aqueous phase and function as seeding ***When the influent total iron included Fe2+ and Fe3+ according to the equilibria

material for the biomass to grow on (Wan et al. 2015).


28

Short-term precipitation dynamics in aerobic granules


SI of minerals
The biological conversions, precipitation kinetics and ion transport 20
were applied on the inside of an aerobic granule with a radius of 1
mm during one standard cycle (dry weather conditions) (see Fig. 10
16 - 19). The solutes concentrations and mineral content inside the
granule were analyzed for five phosphate minerals (amorphous 0

SI of minerals (-)
calcium phosphate ACP, hydroxyapatite HAP, struvite STR
vivianite VIV, and ferrihydrite FER) in order to see the precipitation -10
ACP AN
potential and parameter interdependencies. The mineral content ACP AE
of the granule at the end of the cycle was the result of the dynamics -20
HAP AN
HAP AE
of both precipitation and dissolution processes. Precipitation took VIV AN
mostly place in the core during the anaerobic phase and dissolution VIV AE
-30 FER AN
occurred mostly in the outerlayer during the aerobic phase. FER AE
The dynamic precipitation and dissolution processes STR-NH 4 AN
-40
were caused by the changing SI during the cycle (see Fig. 16). STR-NH 4 AE

Especially the last hour of the aeration phase showed negative 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
SI (i.e., dissolution potential), due to the lower pH (6.5), very Radius (mm)
low phosphate concentration and for struvite; the absence of
ammonium. Struvite did not precipitate, as the SI was negative
Acetate and pH
throughout the whole cycle. STR-K and STR-Na were also not 3.5 7.4
Acetate AN
supersaturated during the standard cycle.
Acetate AE
Acetate concentration (mmol L -1 )

During the anaerobic phase, the acetate penetration 3 pH AN 7.2


pH AE
depth in the granule was maximum 0.25 mm due to the fast acetate
uptake rate and the abundant supply of biopolymers inside the 2.5 7

biomass (see Fig. 16). If the initial conditions of the glycogen or


2 6.8
Poly-P content would have been set significantly lower, these

pH (-)
biopolymers would deplete during the anaerobic phase and the
1.5 6.6
acetate would have penetrated further. The degradation of Poly-P
caused accumulation of phosphate, magnesium, potassium inside 1 6.4
the granule until the end of the anaerobic phase. The pH profile
showed an increase of approximately 0.2 towards the center of the 0.5 6.2
granule at the end of the anaerobic phase, due to the acetate uptake
and production of K+ and Ca2+ (see Fig. 16). The difference between 0 6
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
the modelled and experimental result (a pH decrease) is further Radius (mm)
elaborated in the Discussion. The non-converted solutes (NH4+,
NO3-, Cl-, Na+, SO42-) diffused into the granule during this phase Fig. 16 Modelling results of one standard cycle, including: the
until equilibrium with the bulk was established (see Fig. 17). PHA saturation index (SI) of amorphous calcium phosphate (ACP),
was formed during the anaerobic phase, mirroring the location of hydroxyapatite (HAP), struvite (STR-NH4), vivianite (VIV)
the degradation of Poly-P and glycogen (see Fig. 18). and ferrihydrite (FER) (top) and pH and acetic acid (top). The
During the aerobic phase, the oxygen penetration depth profiles over the radius of the granule are shown at the end of
was maximum 0.65 mm, due to the O2 consumption for PHA the anaerobic phase (AN, at 01:00 h of the cycle time) and the
degradation, glycogen formation and Poly-P formation (see Fig. end of the aeration phase (AE, at 04:00 h of the cycle time).
18). The O2 penetration depth was higher than the findings of De
Kreuk et al. (2007) (0.24 - 0.4 mm) and Pijuan et al. (2009) (0.1 - 0.25
mm), because O2 consumption of ammonium oxidizing bacteria
(AOB) was not included in this model. De Kreuk et al. (2007)
showed full penetration of O2 in a granule of 0.6 mm radius when
no NH4+ was present. The O2 concentration affects the mineral SI
29

Ion concentration
14 Biopolymers during Rain event
PO 4 , NH 4 , NO 3 concentration (mmol L -1 )

PO AN
2
4 Pha Cycle 1
PO 4 AE Pha Cycle 2

Pha, Polyp and Glycogen (mol L -1 )


12 1.8
Pha Cycle 3
NH 4 AN
Pha Cycle 4
NH 4 AE 1.6
Poly-P Cycle 1
10 NO3 AN Poly-P Cycle 2
1.4 Poly-P Cycle 3
NO AE
3 Poly-P Cycle 4
8 1.2 Glycogen Cycle 1
Glycogen Cycle 2
1 Glycogen Cycle 3
6 Glycogen Cycle 4
0.8

4 0.6

0.4
2
0.2

0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1
Radius (mm) Radius (mm)

Cations concentration Oxygen


7 0.16
Mg AN
O 2 AN
Mg AE
0.14
Cation concentration (mmol L -1 )

6 Ca AN O AE
2
O2 concentration (mmol L -1 )

Ca AE
K AN 0.12
5 K AE
Fe AN
Fe AE 0.1
4
0.08
3
0.06

2
0.04

1 0.02

0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Radius (mm) Radius (mm)

Fig. 17 Modelling results of one standard cycle, including: ion Fig. 18 Modelling results of one standard cycle, including:
concentrations (PO43-, NH4+, NO3-) (top) and cation concentration intracellular biopolymer content of the poly-phosphate
(Mg2+, Ca2+, K+, Fe2+/3+) (below). The concentration profiles over accumulating organisms (PAOs), (polyhydroxyalkanoates
the radius of the granule are shown at the end of the anaerobic (PHA), glycogen and poly-phosphate (Poly-P) (top) and the O2
phase (AN, at 01:00 h of the cycle time) and the end of the concentration profiles over the radius of the granule (below)
aeration phase (AE, at 04:00 h of the cycle time) when the anaerobic feeding phase was finished (AN, at 01:00 h
of the cycle time) and when the aerobic phase was finished (AE,
at 04:00 h of the cycle time).
30

by determining the phosphate uptake rate. When oxygen levels


# 10-6 ACP, HAP, STR # 10-13
are high, PO43-, Mg2+ and K+ ions are rapidly taken up by the PAOs 1.2 1.4
ACP AN
for Poly-P storage. This lowers the ion concentration and thus
ACP AE
1.2
the mineral SI. Also, when oxygen is present, PHA degradation 1 HAP AN
HAP AE

HAP, STR content (mmol)


and glycogen formation produce CO2, which lowers the pH and STR-NH 4 AN
1

ACP content (mmol)


therefore the mineral SI. In the Operational parameter section, 0.8 STR-NH 4 AE

the oxygen concentration is changed to investigate the effect of


0.8
different O2 penetration depth on the phosphate minerals. As long 0.6
as there was O2 available, the PHA concentration decreased rapidly 0.6
and glycogen and Poly-P were formed (see Fig. 18). The formation
0.4
rates of glycogen and Poly-P were not so fast, as the content of 0.4

these biopolymers remained close to the maximum content.


0.2
ACP precipitated the fastest over the whole granule 0.2

during the anaerobic phase, but dissolved in the outerlayer in


0 0
the aerobic phase (see Addendum 4 for more detailed graphs). 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
As the deeper part of the granule always remained anaerobic, Radius (mm)
ACP continued to precipitate in the aerobic phase at the granule’s
core before dissolving slightly. The net accumulation of ACP at # 10-18
# 10-15 FER and VIV
the core of the granule was ± 10-6 mmol per cycle (see Fig. 19). 7 3.2

HAP followed the precipitation pattern of ACP, but with a much 3


6
lower final content level of ± 13 x 10-14 mmol per cycle. Since the
2.8
(slow) dissolution of HAP was not included in the model, this
5
FER content (mmol)

VIV content (mmol)


result only indicated the upper bound of HAP precipitation inside 2.6
the granule. Vivianite is supersaturated at the core of the granule
4 2.4
throughout the whole cycle and therefore could precipitate both
in the anaerobic and aerobic phase. However, the kinetic rate was 3 2.2
so slow, that the resulting content does not exceed 3.2 x 10-18 mmol
2
(see Fig. 19). At the outerlayer, vivianite could dissolve slightly 2
during the aerobic phase. Ferrihydrite precipitates only in the aerobic FER AN 1.8
FER AE
phase when oxygen and Fe3+ are present. As the rate expression 1
1.6
VIV AN
of ferrihydrite was not experimentally proven, the amount of VIV AE

ferrihydrite content should be taken as merely hypothetical. 0 1.4


0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Radius (mm)

Fig. 19 Modelling results of one standard cycle, including:


struvite (STR), amorphous calcium phosphate (ACP),
hydroxyapatite (HAP) (left), vivianite and ferrihydrite (FER)
(right). The content profiles over the radius of the granule were
illustrated when the anaerobic feeding phase was finished (AN,
at 01:00 h of the cycle time) and when the aerobic phase was
finished (AE, at 04:00 h of the cycle time).
31

Vivianite Struvite

Hydroxyapatite Amorphous Calcium Phosphate 200 nm


Source: www.rruff.info for vivianite, struvite and hydroxyapatite. Amorphous calcium phosphate is reproduced from (Mahamid et al. 2008)
32

Long-term precipitation dynamics in aerobic granules aerobic phase. The aerobic pH profile, however, will definitely
The mineral contents and ion concentrations at the core of the be influenced by the (de)nitrification processes in practice,
granule were studied over four consecutive standard cycles in which were not included in this model. In Addendum 5, the
order to study the profile evolution (see Fig. 20). All phosphate biopolymer formation and degradation rates are illustrated.
minerals precipitated fastest in the anaerobic phase (between Simulating more than four consecutive cycles did
straight and dotted black lines), whereas ferrihydrite dissolved not establish a steady state in the biopolymer or precipitation
during the anaerobic phase. The ACP content increased non- dynamics. PHA and Poly-P were consumed for maintenance
linearly in time, because the reaction rate depends on the amount and creation of new biomass during the aerobic phase. After 10
of ACP content; more ACP means more available surface to cycles, the PHA content is depleted for the outer 0.15 mm but new
precipitate on. For hydroxyapatite (HAP), the content was too biomass is formed in these exact same layers. The new biomass
small (± 0.5 x 10-12 mmol) to have an influence on the rate and will consume the acetate before it reaches the inner parts of the
therefore increased more linearly. A horizontal line in the HAP granule. This model can therefore not reach steady state without
content profile represented the periods when the SI of HAP was incorporating the newly formed biomass. The precipitation
negative. Struvite does not precipitate during the four cycles as dynamics are believed to never establish steady state since their
the SI remained negative (i.e., dissolution potential) throughout growth was not limited. The precipitates are believed to remain
the cycles (see Fig. 16 on page 28). The ferrihydrite content at low concentrations before granule breakage or biopolymer
increased almost linearly over four cycles as the content was too depletion occurs. The graphs showing the biopolymer and
low for the surface area to influence the precipitation rate. The precipitation dynamics over 10 standard cycles are shown in
vivianite content increased linearly, since the SI stayed at a high Addendum 6.
level throughout the cycle, keeping the rate at a high level.
The non-converted ions NH4+ and NO3- ions showed
the same concentration profile for one standard cycle as for four
consecutive standard cycles. The PO43- concentration profile
decreased slightly in the aerobic phase with each consecutive
cycle (see Fig. 21). The same trend can be seen in the Mg2+ and
K+ concentration profiles. This occurred due to an increased
Poly-P formation rate as the Poly-P content of the granule moved
further away from the Poly-P saturation levels. The pH profile
showed the same pH increase during each anaerobic phase of
the four standard cycles due to the acetic acid uptake and Mg2+
and K+ production. The pH profile during the aerobic phase
changed during four cycles. It is believed this was caused by the
balance between the biopolymer formation/degradation rates
and the precipitation/dissolution of ACP. In the third and fourth
cycle, the ACP precipitation/dissolution becomes sufficiently
large to impact the pH at the core of the granule during the
33

ACP, HAP and STR # 10-12 Ion concentration at core


0.07 1.5 # 10-3
9 9

, NH +4 , NO -3 concentration (mol L -1 )
ACP PO 4
STR-NH 4 NH
8
0.06 4
HAP NO3 8.5
STR, ACP content (mmoles)

7 pH

HAP content (mmoles)


0.05
1 6 8

0.04
5

pH (-)
7.5
0.03 4

0.5 3 7
0.02
2
6.5
0.01
1
PO 3-
4

0 0 0 6
0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16
Time (hours) Time (hours)

# 10-14 FER and VIV # 10-17 # 10-3 Cation concentration at core


2.5 5 6
2+
FER Ca
VIV 4.5 Mg2+
Cation concentration (mol L -1 )

+
5 K
2 4 Fe
FER content (mmoles)

VIV content (mmoles)

3.5
4
1.5 3

2.5 3

1 2
2
1.5

0.5 1
1
0.5

0 0 0
0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16
Time (hours) Time (hours)

Fig. 20 Mineral content (top) during four consecutive cycles with Fig. 21 Ion concentrations (PO43-, NH4+, NO3-) (top) and cation
standard conditions at the core of the granule, with amorphous concentration (Mg2+, Ca2+, K+, Fe2+/3+) (below) during four
calcium phosphate (ACP), hydroxyapatite (HAP), struvite (STR) consecutive cycles with standard conditions at the core of
and vivianite (VIV). The black lines indicate the start of a new the granule. The black lines indicate the start of a new cycle
cycle (straight) and the end of the anaerobic phase (dashed). (straight) and the end of the anaerobic phase (dashed).
34

5.2 Sensitivity analysis

Model parameter sensitivity analysis concentration was the most influential parameter, more than the
A sensitivity analysis of the model parameters was performed pH. This dependency was also observed during the rain cycles
to assess the robustness of the model. Some parameters are with diluted iron concentrations. The magnesium and calcium
assumed based on literature and could differ per AGS systems, competition for phosphate ions to form precipitates became
such as porosity, biomass concentration, kinetic coefficients clear when analyzing the results: when Mg2+ concentration was
and solubility products. The sensitivity of a certain change (%) increased, the ACP content was suppressed and likewise, when
in a model parameter was demonstrated as the change (%) in the Ca2+ concentration was lowered, STR content increased.
maximum precipitated mineral content compared to the base Changes in the acetate concentration also caused
values. The base values were taken from the standard cycle significant changes in the maximum precipitated mineral
results. Since struvite had a zero base value (no precipitation content. This was caused by the effect of acetate on the pH and
occurred during standard conditions), the maximum precipitated phosphate release: when the acetate concentration was 75 %
STR content was plotted on the right axis and not the change (%). lower (50 mg L-1), the pH at the core increased, facilitating more
Overall, the model was assessed to be less robust than desired. precipitation. When the acetate concentration was 100 % higher
Suggestions to improve the model are listed in the Outlook. (400 mg L-1), more phosphate was released and that increased
The porosity sensitivity was most profound when the the PO43- concentration at the core. In that case, the SI of struvite
porosity was decreased: a 50 % lower porosity resulted in ± 900 became positive and struvite precipitated. The graph also showed
% (or 10 times) more ACP content (see Fig. 22). The maximum the Mg2+/Ca2+ interdependence: when the acetate concentration
precipitated ACP content decreased when the porosity was ± 80 was further increased to 400 % (1000 mg L-1), the pH at the core
% lower, as the phosphate was taken-up by the other precipitated of the granule dropped, causing less STR precipitation and
minerals. The same trend was indicated by the biomass freeing phosphate ions for ACP and VIV precipitation.
concentration sensitivity: a higher biomass concentration (less The bulk phosphate concentration had a positive
porous) increased all the maximum precipitated mineral contents. influence on the precipitated mineral content, although less
The opposite situation (less dense and more porous biomass), profound than the acetate or pH. The SI of struvite did not became
however, had less impact on the maximum precipitated content. positive when changing the bulk phosphate concentration.
With a higher biomass concentration (less porous), the phosphate
diffusion was hindered and that increased the phosphate
concentration at the core of the granule. These results were
supported by the observation that more precipitation occurred
for all minerals when the diffusion coefficient of phosphate (DPO4)
was decreased or no multi-component diffusion coefficients
(MultiD) were taken into account (see addendum 7).

Operational parameter sensitivity analysis


The impact of operational parameters (bulk pH, acetate, cation
and PO43- concentration) on the precipitated mineral content was
investigated by a sensitivity analysis. The set-up of the sensitivity
analysis was the same as for the model parameter sensitivity
analysis. Overall, it became evident that ACP and STR were
the most sensitive precipitates and that HAP and VIV were less
responsive to changes under the current assumptions (see Fig. 23).
The pH was evidently the most sensitive operational
parameter, as a change of 8 % in the pH already causes an
increase of 1,500 % (or 16 times more) in the ACP content.
This effect was larger than the cation concentration, which
was the second most influential parameter. For VIV, the iron
35

Sensitivity of porosity Sensitivity of Biomass concentration # 10-6

Change in maximum precipitated content (%)


Change in maximum precipitated content (%)

1800 0.18 3000 3

Maximum precipitated STR content (mol)


Maximum precipitated STR content (mol)
1600 HAP 0.16
2500 2.5
1400 0.14 STR

1200 0.12 2000 ACP 2

1000 0.1
1500 1.5
ACP
800 0.08
1000 1
600 0.06

400 0.04 500 0.5


STR
200 0.02 HAP
VIV VIV
0 0
0 0

-200 -0.02 -500 -0.5


-100 -80 -60 -40 -20 0 20 40 60 80 100 -100 -80 -60 -40 -20 0 20 40 60 80 100
Change in Porosity (%) Change in Biomass concentration (%)

Sensitivity of kinetic rate coefficient k Sensitivity of Solubility product # 10-6


Change in maximum precipitated content (%)

# 104 1000 10
Change in maximum precipitated content (%)

5 5000

Maximum precipitated STR content (mol)


Maximum precipitated STR content (mol)

STR

4 4000 800 8
ACP
HAP

3 3000 600 6

2 2000 400 4

STR
1 1000 200 2

VIV HAP
0 ACP 0 0 0
VIV

-1 -1000 -200 -2
-200 0 200 400 600 800 1000 -5 -4 -3 -2 -1 0 1 2 3 4 5
Change in kinetic rate coefficient k(%) Change in solubility product (%)

Fig. 22 Sensitivity analysis on the following model parameters: porosity (left top), biomass concentration (right top), kinetic coefficient
of the precipitation rate expressions (kp, left below) and solubility products (Kp, right below) with amorphous calcium phosphate
(ACP), hydroxyapatite (HAP), struvite (STR) and vivianite (VIV). On the x-axis, the changes were calculated as percentages of the
parameter value change compared to the base value. On the y-axis, the changes were calculated as percentages of the maximum value
of precipitated content compared to the base value. The base values were taken from the standard cycle results. Since struvite does not
precipitate in the standard cycle, the maximum STR content are plotted on the secondary y-axis, not the change in maximum content.
36

Sensitivity of Acetate # 10-6 Sensitivity of pH # 10-7

Change in maximum precipitated content (%)


Change in maximum precipitated content (%)

200 2 2000 20

Maximum precipitated STR content (mol)


Maximum precipitated STR content (mol)
1800 18
150 1.5 STR
1600 16

1400 14
100 1 ACP
1200 12

STR 1000 10
50 0.5
800 8
HAP
600 6
0 0
VIV 400 4

-50 -0.5 200 2


HAP
ACP
0 0
VIV
-100 -1 -200 -2
-100 -50 0 50 100 150 200 250 300 350 400 -10 -5 0 5 10 15
Change in Acetate concentration (%) Change in pH (%)

Sensitivity of involved cation concentration # 10-7 Sensitivity of bulk PO 3- concentration


Change in maximum precipitated content (%)

4
Change in maximum precipitated content (%)

5000 5
120 1.2

Maximum precipitated STR content (mol)


Maximum precipitated STR content (mol)

VIV

100 1
4000 ACP 4 ACP
80 0.8

3000 3 60 0.6

40 0.4
2000 2
STR 20 0.2
HAP
STR
1000 1 0 0
VIV
-20 -0.2
0 0
-40 -0.4
HAP
-60 -0.6
-1000 -1 -100 -50 0 50 100 150 200
-200 0 200 400 600 800 1000
Change in involved cation concentration (%) Change in bulk PO 3-
4
concentration (%)

Fig. 23 Sensitivity analysis on operational parameters: acetate concentration (top left), pH (top right), the involved cation concentration
(lower left) and PO43- bulk concentration (lower right) with amorphous calcium phosphate (ACP), hydroxyapatite (HAP), struvite
(STR) and vivianite (VIV). On the x-axis, the changes were calculated as percentages of the parameter value change compared to the
base value. On the y-axis, the changes were calculated as percentages of the maximum value of precipitated content compared to the
base value. The base values were taken from the standard cycle results. Since struvite does not precipitate in the standard cycle, the
maximum STR content are plotted on the secondary y-axis, not the change in maximum content.
37

Effect of operational parameters on the granule pH pH at core in time at different Acetate bulk concentrations
Operational parameters influenced the precipitation potential 9
Acetate = 50 mg L-1
through a change in the pH at the core of the granule. The operational Acetate = 100 mg L -1
8.5
parameters, such as the acetate concentration, bulk pH and Acetate = 200 mg L -1
Acetate = 400 mg L -1
oxygen levels, had a significant impact on the granule pH. 8
Acetate = 1000 mg L -1

The acetate concentration was clearly a sensitive

pH (-)
parameter as it produced the phosphate release and strongly 7.5

influenced the core pH in the anaerobic phase. When the acetate


bulk concentration was changed to 50, 200, 400 and 1000 mg 7

L-1, the pH increased for decreasing acetate concentrations


6.5
(see Fig. 24). When little acetate was available, around 5 times
as less H2PO4- was produced, leading to a higher pH.
6
The bulk pH also influenced the precipitation. As 0 2 4 6 8 10 12 14 16
Time (hours)
the phosphate release and the SI were pH dependent, the pH
pH at core in Time with different bulk O concentrations
gradient in the core also changed. At bulk pH > 7.1, the pH profile 2
9
changed significantly in the aerobic phase due to a faster Poly-P O 2 = 0.1 - 0.3 mg L -1

uptake rate, as more phosphate is released in the anaerobic 8.5 O = 0.5 - 1.5 mg L-1
2
O 2 = 1.0 - 3.0 mg L-1
phase. At a bulk pH between 6.0 and 6.5, the (acidifying) CO2
O 2 = 1.5 - 4.5 mg L-1
production of PHA degradation and glycogen formation 8
O 2 = 2.0 - 6.0 mg L
-1

dominates over the (alkaline) Poly-P uptake rate.


pH (-)

7.5
The oxygen concentration during the aerobic phase
also had a clear effect on the pH at the core of the granule. A 7
higher oxygen concentration caused a lower core pH. The effect
of the oxygen was strongest at the end of the aerobic phase, as the 6.5

oxygen concentration was increased until the maximum. A higher


6
oxygen concentration increased the rate of PHA degradation, 0 2 4 6 8 10 12 14 16

glycogen formation and maintenance, which lowered the pH due Time (hours)

to CO2 production (see Table 4, on page 11). It should be noted that pH at core in Time at different bulk pH
9
the pH profile in the aerobic phase will look different in practice, pH bulk 6.0 - 6.5

due to the contributions of nitrification and CO2 stripping. pH bulk 6.3 - 6.8
8.5 pH 6.5 - 7.0
bulk
pH bulk 7.1 - 7.6
pH bulk 7.3 - 7.8
8
pH bulk 7.5 - 8.0
pH (-)

7.5

6.5

6
0 2 4 6 8 10 12 14 16
Time (hours)

Fig. 24 Influent acetate concentration (top), bulk pH (middle)


and bulk oxygen concentration (below) influence on the pH at
the core of the granule during 4 standard cycles. The range of
bulk pH and oxygen concentration was used during the different
phases within one standard cycle (see also Table 8 on page 16).
The black lines indicate the start of a new cycle (straight) and
the end of the anaerobic phase (dashed).
38

Precipitation potential under rain conditions in aerobic granules


Biopolymers during Rain event
As full-scale reactors in the Netherlands experience a significant 2
Pha Cycle 1
amount of days with rain, the effects of ± four times diluted Pha Cycle 2

Pha, Polyp and Glycogen (mol L -1 )


1.8
Pha Cycle 3
concentrations on the mineral content was studied (see Fig. 25). Pha Cycle 4
1.6
Four cycles are simulated consecutively with rain conditions in Poly-P Cycle 1
Poly-P Cycle 2
cycle 2 and 3 (4:00 – 12:00 h). The precipitation rate of ACP, and 1.4 Poly-P Cycle 3
HAP increased during the rain event, due to a large increase in Poly-P Cycle 4
1.2 Glycogen Cycle 1
the core pH. A pH increase in the anaerobic phase during rainy Glycogen Cycle 2
1 Glycogen Cycle 3
days in the bulk phase was also seen in practice (see Addendum Glycogen Cycle 4
8). The core pH is believed to be increased during the anaerobic 0.8

phase as consequence of the lower acetic acid concentration 0.6


and less H2PO4- production. Because of the increased pH, STR-
0.4
NH4 becomes supersaturated and precipitates at the end of the
anaerobic phase, before dissolving again rapidly. Only the rate of 0.2

ferrihydrite and vivianite precipitation decreased, as the diluted 0


0 0.2 0.4 0.6 0.8 1
Fe2+/3+ concentration dominated over the effect of the increased pH
Radius (mm)
for these minerals. The second pH peak during the aerobic phase
in the rain event was caused by an increased rate of Poly-P uptake Biopolymers during Rain event
(see Addendum 9). Usually in practice, a pH peak during the 2
Pha Cycle 1
aerobic phase was not seen as nitrification damped the pH. Pha Cycle 2
Pha, Polyp and Glycogen (mol L -1 )

1.8
Pha Cycle 3
The Poly-P content decreased less during the rain cycle Pha Cycle 4
1.6
(cycle 2) than during standard cycles, as was already observed in Poly-P Cycle 1
Poly-P Cycle 2
the lower H2PO4- production (see Fig. 26). The glycogen content 1.4 Poly-P Cycle 3
decreased less during the rain cycles than during standard cycles Poly-P Cycle 4
1.2 Glycogen Cycle 1
as less acetic acid was consumed. The PHA content showed a Glycogen Cycle 2
1 Glycogen Cycle 3
large decrease during the rain event. This was caused by the lower
Glycogen Cycle 4
accumulation of PHA during the anaerobic phase since there was 0.8

a lack of acetate. However, the same oxygen levels in the aerobic 0.6
phase caused a similar PHA decrease as in the standard cycles.
0.4
Although the PHA content recovered slightly during the first
following standard cycle (cycle 4), it does not reach the PHA level 0.2

of 4 consecutive standard cycles until five cycles after the last rain 0
cycle. The total recovery time depends on the duration of the rain 0 0.2 0.4 0.6 0.8 1

event, available acetic acid and the distance of the PHA level from Radius (mm)
the saturation level. It is expected that the PHA levels of PAOs in
full-scale AGS installations can be considerably Fig. 26 The biopolymer content during four standard cycles
(top) and four cycles of which the second and third cycle
consisted of rain conditions (below). The profiles over the
radius of the granule are illustrated for the end of the
aerobic phase (at 04:00 h , 08:00 h, 12:00 h and 16:00 h).
39

ACP, HAP and STR-NH4


ACP, HAP and STR # 10-14 FER and VIV # 10-17
# 10-12 # 10-12
1
# 10-3
0.07 2 1.5 1.5 2.5 5
ACP
STR-NH
FER
0.9 0.061.8
HAP
4
VIV 4.5
STR, ACP content (mmoles)
STR-NH4, ACP content (mmoles)

HAP content (mmoles)

0.8 0.051.6 2 4
1

FER content (mmoles)

VIV content (mmoles)


HAP content (mmoles)
0.7 0.041.4 3.5
1

0.6 0.031.2 1.5 3


0.5
0.5 0.02 1 2.5

0.4 0.010.8 1 2

0.3 00.6 0
0.5
1.5
0 2 4 6 8 10 12 14 16
Time (hours) 0.5 1
0.2 0.4

0.5
0.1 0.2

0 0
0 0 0
0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16
Time (hours) Time (hours)

Ion concentration at core # 10-3 Cation concentration at core


10 9 6
, NH +4 , NO -3 concentration (mmol L -1 )

PO 4 Ca 2+
9 NH Mg2+
4
Cation concentration (mol L -1 )

NO 8.5 5 K+
8 3
Fe
pH
7
8 4
6
pH (-)

5 7.5 3

4
7 2
3

2 1
6.5
1
PO 3-
4

0
0 6 0 2 4 6 8 10 12 14 16
0 2 4 6 8 10 12 14 16
Time (hours) Time (hours)

Fig. 25 Mineral content (top), ion concentrations (PO43-, NH4+, NO3-, left below) and cation concentration (Mg2+, Ca2+, K+, Fe2+/3+, right
below) during rain weather conditions in the second and third cycle (4:00 – 12:00 h) at the core of the granule, with amorphous
calcium phosphate (ACP), hydroxyapatite (HAP), struvite (STR) and vivianite (VIV). The black lines indicate the start of a new cycle
(straight) and the end of the anaerobic phase (dashed). Note that the left y-axis of the ACP, HAP and STR-NH4 content is split; the
(blue) left y-axis displays the STR-NH4 content, whereas the right (pink) y-axis isplays the ACP content.
40

Granule size effect on precipitation potential Size effect on ACP


Since the granule size has a large impact on the ion concentration 10-2
1.5 mm

and pH profiles in space, the effect of the granule size on the


10-4
precipitates was studied (see Fig. 27). The radius of a granule 1.0 mm
needed to be ≥ 0.8 mm for the fastest precipitating mineral, ACP,

ACP content (moles)


10-6
to persist throughout the whole cycle. When the granule was
smaller than 0.8 mm, the pH gradient towards the core of the 0.8 mm
10-8
granule was too low and ACP dissolved rapidly (see Fig. 27).
0.5 mm
Struvite (STR) only precipitated when the radius was ≥ 1.5 mm. 10-10
The SI of STR became positive as the pH at the core of the large
0.3 mm
granule increased significantly. The STR profile in time, however, 10-12
showed that this mineral does not persist throughout the cycle.
The effect of the radius size on hydroxyapatite (HAP), vivianite 10-14
0 2 4 6 8 10 12
(VIV) and phosphate concentration profile at the core in time are Time (hours)
shown in Addendum 10. Size effect on STR
10-5

10-6
STR content (moles)

1.5 mm
10-7

10-8

10-9

10-10
0 2 4 6 8 10 12
Time (hours)
Size effect on pH
9
Radius = 1.5 mm
Radius = 1 mm
8.5 Radius = 0.8 mm
Radius = 0.5 mm

8
pH (-)

7.5

6.5

6
0 2 4 6 8 10 12
Time (hours)
Fig. 27 Amorphous calcium phosphate content (ACP, top),
struvite content (STR, middle) and pH at the core of the granules
with different radii: 0.3, 0.5, 0.8, 1 and 1.5 mm. Due to logarithmic
scale on the y-axis, zero values are not shown. The black lines
indicate the start of a new cycle (straight) and the end of the
anaerobic phase (dashed).
41

Radius: 0.1 - 0.2 mm Radius: 0.2 - 0.3 mm

Radius: 0.3 - 0.5 mm Radius: > 1 mm


42
43

6. Discussion

Parameter interdependencies Granule and bulk pH interdependency


At the current operation conditions of full-scale AGS installations Although the micro-electrode measurements in this study
in the Netherlands, no significant phosphate precipitation occurs indicated a decreasing pH gradient inside the granule, other
inside the granules. The differences between the full-scale and studies have pointed out other trends (Lemaire et al. 2008, Winkler
lab reactors, where phosphate precipitation did occur, are mainly et al. 2013). The interdependency of the granule and bulk pH is
focussed on the differnces in influent concentrations and bulk discussed here. In the granule, the following processes are believed
pH. Also, rain events are often not included in lab reactors that to play a major role on the pH in the anaerobic phase:
could effect the precipitation. A more elaboration discussion on • More uptake of acetic acids (> 1000 mg L-1), reduces the pH gradient
the differences in precipitation between lab and full-scale AGS to slightly acidic pH according to the modelling results
installation can be found in (Stubbé et al. 2016). It was investigated (see Fig. 24 on page 37),
whetehr phosphate precipitation could be forced insde an AGS • Precipitation processes (e.g., calcite, vivianite) acidify the
system by creating favourable precipitation conditions. environment by incorporation of PO43- or CO32-, releasing
According to the model, phosphate precipitation in protons (Dupraz et al. 2009, Madsen et al. 2014),
full-scale aerobic granules is stimulated under certain conditions, • H2PO4--release decreases the pH by releasing one or two
such as: a granule size > 0.8 mm, a bulk pH 7 – 7.6 throughout protons, depending on the interior pH (Serralta et al. 2004). As
the cycle, lower oxygen concentrations, metal dosing and a high long as the H2PO4- is not yet transported outside the granule,
phosphate release, or in other words, good functioning biological this acidifies the granule interior. The amount of H2PO4- release
phosphate removal (Bio-P). The model results indicated that at determines the impact of the decreasing pH-effect. The amount
a pH ≤ 6.5, considerably less precipitation or even dissolution of H2PO4- release depends on the biopolymer content of the
occurred. As the bulk pH profile showed a decrease towards the bacteria, the available substrate (VFA or acetate concentration)
end of the aerobic phase (see Fig. 9 on page 21), the starting bulk and the pH of the media (Smolders et al. 1994, Filipe et al. 2001).
pH should be sufficiently high to maintain the pH level above this • The simultaneous release of K+ and Mg2+ ions, during phosphate
level throughout the cycle. It is understood that a certain level of release creates a repulsion of protons to maintain neutrality inside
oxygen concentration is necessary for nitrification. When this level the granule. This would increase the pH inside the granule as
is kept as low as possible, however, it stimulates the precipitation long as the K+ and Mg2+ ions do not diffuse outside the granule.
inside the core of the bigger granules as the phosphate is taken The cations could be transported faster out of the granule than
up more slowly. A good functioning Bio-P means that the phosphate, due to their higher diffusion coefficients (e.g., Ca2+,
biopolymer (PHA, glycogen and Poly-P) balance inside the Na+) (Minteq database 2006).
granules should be sufficient for significant phosphate release to
In the bulk, the following processes are believed to play
initiate precipitation. The metal dosing should consists of ± 50,
a major role on the pH in the feeding phase:
120 and 15 mg L-1 magnesium, calcium and iron to have sufficient
•  Acetic acid uptake increases the pH of the bulk, as protons
impact on the precipitated content of the respective mineral (see
are withdrawn from the bulk,
Fig. 23 on page 36). This could add up to 300, 200 or 30 $ d-1 for
• Incoming influent can have a different pH than the remaining
dosing magnesium, calcium or iron in the full-scale installation
bulk liquid from the previous cycle, which is pushed upwards by
in Utrecht (see Addendum 11 for details). Metal dosing could,
the plug flow of the influent from the bottem. This creates a higher
however, be undesirable for environmental reasons, as mining the
pH around the sludge bed, although some dispersion might create
metal ores produces a lot of waste (Dudka et al. 1997).
a mixture of the remaining bulk and influent liquid.
Phosphate precipitation gets a kick at the start of a rain
• Precipitation processes in the bulk liquid have the same pH-
event, when the biopolymers (PHA, glycogen and Poly-P) are still
effect on the bulk as in the granule mentioned above.
sufficiently present inside the granule (see Fig. 25 on page 39). The
The difference in pH trend inside the granule in this
lower acetate concentration facilitates a high pH increase towards
research compared to previous works can possibly be caused
the core of the granule that triggers phosphate precipitation. This
by the higher media pH used during the measurement: we used
process could continue until the biopolymer content is unable to
7.64 (average pH in Garmerwolde) instead of a pH of 6.7 - 6.9
maintain the necessary phosphate release. Phosphate precipitates
used by Winkler et al. (2013) and 6.5 by Lemaire et al. (2008). The
dissolve when the biopolymers are depleting and the acetate and
higher bulk pH caused more H2PO4- release for acetic acid uptake
oxygen can penetrate deeper into the granule (see Addendum 6).
and possibly more precipitation. These processes acidify the
environment. Lemaire et al. (2008) measured the pH profile inside
44

7,0 7
the granule in a media with high phosphate concentration (three
times higher than in this study). There was no mentioning of metal 6,8 6,8

Granule pH (-)
concentrations or precipitates in the study of Lemaire et al. (2008).

Bulk pH (-)
6,6 6,6
Winkler et al. (2013) measured the pH profile inside the granule
in a media with high acetate concentration (50 % higher than in 6,4 6,4
this study), phosphate concentration (three times higher than
6,2 6,2
in this study) and magnesium concentration (two times higher
than in this study). There was some mentioning of precipitation 6,0 6
in the study of Winkler et al. (2013). The difference in media 0:00 0:07 0:14 0:21 0:28 0:36

composition could have contributed to the difference in the pH Time (hh:mm)


gradient measured inside the granule. Although the pH gradient Granule pH Bulk pH

was different between the mentioned studies, the absolute pH


Fig. 28 Point pH measurement inside the core of a granule
value in the core of the granules did not differ much: a pH of 7.4
pretreated with liquid nitrogen. Bulk pH was 6.84.
at the core of the granule was found in this study, 7.45 in Winkler
et al. (2013) and 7.7 in Lemaire et al. (2008). These values are
surprisingly close to the steady state pH found with the point pH
solutions (Tipping et al. 1992, Su et al. 1997, Appelo et al. 2002, van
measurements (7.6) at the core of the granule in this study.
Breukelen et al. 2004). Since accumulated iron is present in aerobic
granules (Li et al. 2014, Stubbé et al. 2016), the surface complexation
Buffering capacity of granules
could have a significant contribution to the alkalinity. During the
During the micro-electrode measurements and titration of
static pH measurement, protons diffusing into the granule might
granules, it became evident that granules have a certain buffering
be demobilized due to surface complexation on organic material or
capacity. This was also earlier noted by Lemaire et al. (2008).
iron. At the same time, carbonate or hydroxyl ions might be released
The point pH measurement at the core of the granule showed
from different surface complexations and increase the pH.
alkalinity being present or even produced inside the granule over
Diffusion does not only depend on concentration
time. This can be explained by four different processes:
gradients, but also on the charge balance (Parkhurst et al.
1. A (amorphous) precipitate is dissolving, increasing the pH
2013). If the inside of the granule is slightly positively charged,
(Pijuan et al. 2009),
anions like hydroxyl ions could diffuse inside the granule
2. Up-take of residual hydrolyzed COD left inside the granule,
by the driving force of the charge difference. The charge
3. Surface complexation on the bacetrial cells and extracellular
difference can be stimulated by release of cations, which,
polymeric substances (EPS) inside the granule exerts a buffering
for example, can happen with degradation of Ca2+ from the
capacity,
EPS due to starvation suggested by Pijuan et al. (2009).
4. Charge difference as a driving force for diffusion,
A fifth reason for a buffering capacity in the granules
Although no major crystals were seen on the XRD, the
could be the bacteria maintaining their intracellular pH slightly
model and some (dis)appearing precipitates in other XRDs
alkaline for the proton motive force as suggested by Saunders et al.
indicated that some (amorphous) precipitates might still
(2007) and the exported H+ diffuses out of the granule. However,
be present at certain times. These precipitates would create
this would only explain an interior pH equal to the ambient pH,
alkalinity when they dissolve since the released cations
not an elevated interior pH. Yet this process could take place in
will repulse H+ ions and the released PO43- will take up
combination with one of the four processes mentioned above.
protons to form HPO42-/H2PO4- depending on the pH.
Biological processes were limited during the pH
measurements, since no substrate was available in the buffer A point pH measurement was done on a granule pretreated with
media. However, possibly there was some residual complex liquid nitrogen to rule out the buffering effect of biological activity
material inside the granule that was hydrolysed during the time (e.g. residual acetic acid uptake) (see Fig. 28). No buffering capacity
in between sampling and measurement (2-4h). The uptake of the was present any longer. Only diffusion limitation as proposed by
hydrolysed acetic acid would increase the interior pH. Lemaire et al. (2008) remained: the ambient pH was only adapted
The surface complexation of carbonate, protons and after 40 min, whereas a maximum of 6 min was expected with a
hydroxyl ions on organic material, dissolved organic carbon, iron, proton diffusion coefficient of 9.31 x 10-9 m2 s-1, a granule porosity
aluminium or calcite can contribute to the buffering capacity of of 0.5 (-) and a granule radius of 1.7 mm. This could show the
45

activity, viscosity or density differences inside the granule


affect the diffusion and possibly even the crystallization process
(Mersmann 2001). Nevertheless, this result does not prove that a Effluent
charge difference or surface complexation do not contribute to the
buffering capacity. Since liquid nitrogen is a severe method for Nereda

stopping biological activity, it is very likely that the EPS structure


and other complexes were destroyed simultaneously. Influent
Recycle
Matured granules Granules
Practical application
Precipitation can have practical advantages for full-scale AGS
installations; the precipitates increase the settleability of granules Wasted sludge
P-
release
(Peeters et al. 2011, Winkler et al. 2013), which is especially
advantageous in small granules at start-up. Also, controlled
Residual P-rich water
precipitation could prevent unwanted struvite precipitation water
elsewhere in the treatment installation (Ohlinger et al. 1998, Doyle
et al. 2000). In addition, phosphate precipitates could be a new P-
product
pathway to recover phosphate from waste water as easily accessible
and pure phosphate rock reserves are depleting (De Ridder et al.
2012). It depends on the goal of the precipitation which parameters Fig. 29 Schematic illustrations of precipitation scenario 1:
should be used. Two scenarios are discussed here: support biological supporting biological phosphate removal while creating a
phosphate removal (Bio-P) while creating a phosphate recovery phosphate recovery product
product and increase precipitates inside small granules.
Biological phosphate removal in mature granules has ACP at the core of 0.3 mm radius granule
ACP at the core of 0.3 mm radius granule 10-2
a maximum, as saturation is signalized
10-2 in full-scale installations Base
Base
pH > 6.8 pH > 6.8
(Stubbé et al. 2016). Although scientific evidence is lacking, pH > 6.8 + Ca 120 mg/L pH > 6.8 +
pH > 7.4 pH > 7.4
discarding extra mixed reactor 10-4 sludge seemed to maintain the
10-4 pH > 7.4 + Ca 120 mg/L
pH > 7.4 +
required Bio-P capacity. Instead, a side-stream tank could be
ACP content (moles)

installed with the necessary conditions


10-6 for maximum P-release
ACP content (moles)

10-6
(see Fig. 29). The Poly-P content of saturated granules will be
lowered and the granules can 10-8 function properly again in the

main-stream reactor after separation of the phosphate-rich 10-8


water. The substrate for phosphate
10-10
release inside the side-stream
reactor can be provided by raw waste water or other waste
10-12 10-10
streams. The phosphate-rich water will go to an industrial reactor
where a reusable product can be made, such as, but not limited
10-14
to The Crystalactor (2016) of Royal 0 Haskoning
2 DHV.
4 6 8 10 10-12
12
Time (hours)
The reasons to not use the internal gradient of the
granules to force precipitation are multiple: (1) the extraction
10-14
will be much easier if precipitation occurs in the aqueous phase, 0 2 4 6 8 10 12
(2) to certify the phosphate product for reuse is much harder Time (hours)
when pathogenic biomass was close in contact with the mineral, Fig. 30 Schematic illustrations of precipitation scenario 2:
(3) the most environmental friendly and economical valuable enhancing precipitation inside small granules
minerals, STR and HAP are not easy to form inside the Nereda
reactor as the SISTR is to low and the HAP kinetics to slow.
To enhance the precipitation inside small granules,
the combined effects of pH control, metal dosing, increased bulk
phosphate concentrations, decreased acetate concentrations
and longer feeding times were investigated. The increased bulk
46

phosphate concentration, decreased acetate concentration and


longer feeding times produced minor effects on the precipitation
inside granules with 0.3 mm radius. The combination of pH
control and metal dosing however, ensured significant increased
and persistence ACP precipiation in a granule with 0.3 mm
radius (see Fig. 30). The pH control ensured that the bulk pH
would not drop below a certain level which would enhance the
dissolution potential. A bulk pH above 6.4 was not enough to
ensure persistence of the ACP content throughout the whole cycle.
Moreover, only the pH control, even above 7.4, was not enough
to ensure persistence of the ACP content throughout the whole
cycle. The combined effect of pH control and metal dosing could
have the desired precipitation for improved settling properties of
small granules according to the model results. It should be taken
into account however, that ACP is an amorphous material, and
therefore could have less effect on the settling properties. The
model has the capacity to look for other combinations that would
enhance the precipitation of other minerals (STR-NH4 , HAP, VIV
or other minerals).
47

7. Conclusion

The model showed that the resulting phosphate mineral content


of aerobic granules is a dynamic process of precipitation and
dissolution throughout a process cycle and over consecutive
cycles. Overall, no significant net precipitation of struvite,
hydroxyapatie or vivianite was simulated with the developed
model inside aerobic granules embedded in average waste water
concentrations in the Netherlands. Only amorphous material
(amorphous calcium phosphate) showed notable contents
inside the granule during the simulations. During standard
conditions, struvite was undersaturated throughout the whole
process cycle, while hydroxyapatite and vivianite have too
slow kinetics to precipitate significantly during a cycle. The
precipitation and dissolution kinetics of the chemical processes
are decisive for the persistence of the minerals throughout the
process cycle of aerobic granular sludge systems.
The pH inside the granule is the most sensitive
parameter for precipitation, followed by the calcium and iron
concentrations for amorphous calcium phosphate and vivianite.
To a lesser extent, the acetate and bulk phosphate concentrations
are of (indirect) influence on the precipitation inside aerobic
granules. The biological processes influence the precipitation
through phosphate release and acetate consumption, which
both have an effect on the pH. The biopolymer content (PHA,
glycogen and Poly-P) of the granules determines the acetate
penetration depth and potential phosphate release, which
in turn affects the precipitation. A good functioning Bio-P
is therefore believed to have more chances of biologically
induced phosphate precipitation inside aerobic granules. These
results support the obeservation of Stubbé et al. (2016) that
phosphate precipiation does not significantly contribute to the
overall phospahte removal in full-scale AGS systems. The net
phosphate mineral content inside aerobic granules depends on
the granule size, biopolymer content, influent concentrations,
rain events and the duration before a granule breaks.
48

8. Outlook

Model developments taken into consideration that the Cx and porosity might not have
First of all, the limitations of the software PHREEQC version 3.3.2 a uniform profile throughout the granule, as multiple authors
were multiple. It is recommended to continue and/or develop indicate biomass density and porosity profiles inside aerobic
a model where transport processes are easier combined with granules (Tay et al. 2003, Toh et al. 2003, Liu et al. 2009).
chemical processes. Then the model could be extended from Fourthly, the iron chemistry was shrewdly simplified
granule-scale to reactor-scale and include radial diffusion. Radial in this model. The addition of the kinetic rate for Fe2+ oxidation
diffusion would increase the solutes concentrations at the core to Fe3+ would provide more insights in the real precipitation of
of the granule, increasing the precipitation potential. Although ferrihydrite and other iron oxides. These oxides could contribute
the radial diffusion would also increase the movement of the considerably to the surface complexation mentioned earlier.
components out of the granule’s core, the dissolution kinetics can Lastly, the addition of the nitrogen cycle in the model will
dependent on significantly different parameters than precipitation improve the representative pH values and oxygen concentration
(Christoffersen et al. 1979, Margolis et al. 1992, Babić-Ivančić et levels inside the granule. Although the metabolism of PAOs was
al. 2002, Rahman et al. 2014). The planar diffusion model would shown to have a major role in the precipitation process, so did
therefore indicate only the lower bound of the precipitation potential. the pH and oxygen. Therefore, the real experienced phosphate
A possible software alternative for PHREEQC is precipitation in full-scale AGS systems can not fully be assessed
PHAST4WINDOWS (Parkhurst et al. 2010), which includes without the addition(s) of the point(s) mentioned above.
anisotropic dispersion-coefficient tensors, a Cartesian grid for
radial diffusion and a parallel operator to diminish the kinetic Practical application improvement
operation time for 90-99% (Burnett et al. 1987, Dongarra et al. 1990). After the demand of the practical application of the model is
However, to include multi-component diffusion coefficients, confirmed, the sense of urgency for model development should be
PHT3D could be considered more appropriate (Kuder et al. increased to acquire the necessary resources. The online data could
2014). Also, MT3DMS could be considerate when the option for be inserted in (a newer version of) the model and calculate the
spatially varying diffusion coefficients is available. precipitation potential directly for every part of the cycle (feeding,
Second of all, the addition of surface complexation to aeration, setlling). When the potential is high, a XRD analysis of a
the model will provide insights into the impact of this process specific mixed sludge sample can point out whether precipitation
on the precipitation. Adsorption on organic material and iron actually did take place and thereby verify the model.
(III) oxides is suggested by multiple authors to be significant The practical application of the phosphate
(Christoffersen et al. 1981, Schwertmann et al. 2008). Also, bacterial precipitation model can be improved by testing the model
and/or sludge surface charge might cause additional surface results on a lab or bench-scale or even the full-scale AGS
complexation although the last is stated to be less in aerobic reactor in Utrecht. By changing the influent pH, acetate
granules than in flocculent sludge (Zhang et al. 2007). Besides, and metal concentration, precipitation should be able to
bacterial charge might have an impact on the ion diffusion. form during weeks of steady-state operation with standard
Thirdly, the sensitivity analysis for the model parameters conditions. This can be checked by multiple XRDs over time.
indicated the importance of validation of various parameters for
full-scale AGS systems. Currently, the kinetic expressions are Microbial mysteries
obtained from literature where the aqueous systems are often The ‘magical’ interior pH of 7.5 - 7.6 inside the granule, that seems
very different than in aerobic granule systems. The choice of the to be the most stable condition, could be the optimum for the
type of rate expression also influences the outcome seriously; bacteria, the granule structure (e.g. EPS) or both (Lemaire et al.
the kinetic rate of Hydroxyapatite (HAP) of Inskeep et al. (1988) 2008, Pijuan et al. 2009). The bacterial sensitivity to pH has a new
was significantly slower than the HAP kinetic rate of Barat et al. dimension with granular sludge compared to flocculent sludge.
(2011). A kinetic study for aerobic granules specifically would Besides bulk and intracellular pH, there is now a separate, extra
give more insights in the true precipitation rates in AGS systems. pH inside the granule which can be different from the other two.
Besides, the biomass concentration Cx, porosity, and volatile fatty Besides, the activity, viscosity and density differences inside
acid composition of the waste water should be measured in the the granule can have an impact, not only on the diffusion and
full-scale AGS installations to increase the reliability of the model. precipitation, but also on the sensors themselves that measure
Liu et al. (2009) and Xiao et al. (2008) found porosities between activity (for example, the pH micro-electrode). The causes and
0.63 and 0.97 for aerobic granules using confocal laser scanning effects of these mysteries remain unknown.
microscopy and an overall density formula. It should also be
49

9. Acknowledgments

You’ll never work alone – Frank Sinastra shared topic: Ramon Barat (Universidad Politécnica de Valencia) and
Indeed, this project would never have been the same without Hans Erik Lundager Madsen (University of Copenhagen).
the brainstorms, discussions, advices, openness and laughter On a more personal level, I would like to thank the
of everyone around me. The combination of two MSc Thesis people most close to me: my parents, Barbara Bekhof, Sander
projects (The Fate of Phosphate in Aerobic Granular Sludge Systems Scheer, Cees van der Straat and José Vergroesen. You contribute
and Modelling Biologically Induced Phosphate Precipitation in enormously to this project by listening to my stories, maintaining
Aerobic Granules) took almost 14 months to complete. Too much my motivation and making me laugh. Y por último, la persona
has happened to mention here. All I want to say is ‘Thank You’ más importante de todo: mi amor, Gustavo Guerriero. Quiero
to all the people that have contributed to my professional and agradecerte por todo tu apoyo durante este año tan díficil.
personal development during this period. Firstly, a lot of credits Juntos somos un equipo invencible. Contigo es con quien quiero
go to my supervisors Merle K. de Kreuk, Mario Pronk, Cristian pasar todos los días, tomando tu mano, aprendiendo, creciendo,
Picioreanu, Boris M. van Breukelen, Robbert Kleerebezem and viviendo. Te amo.
Mark C. M. van Loosdrecht for their feedback and guidance.
Secondly, I owe a lot to the openness and practical support from
Mark Stevens, Pascal Kamminga, Jan Cloo and Peter de Vries.
You really made this experience more enjoyable for me. At the TU
Delft, I have had help from multiple sides: Armand Middeldorp,
Anke Luijben, Alexander Hendriks, Frans Hamer, Peng Wei,
Cuiji Feng (Civil Engineering), Daine Butterman (Technology,
Policy and Management), Ruud Hendrikx (Mechanical Engineering),
Stefanie Tseggai (Architecture), Bart Joosse, Simon Felz, Danny de
Graaff, Viktor Haaksman, Laurens Welles, Jure Zlopasa, Lorena
Guimaraes, Ingrid Pinel, Laura Valk, Eveline van den Berg,
Leonor Guedes da Silva, Monica Conthe Calvo, Emma Korkakaki,
Maaike Hoekstra, Annie Xing, Maaike Goudriaan, Vincent Poot,
Aina Soler Jofra, Sanne de Smit, Valerie Sels, Ben Norder, Udo van
Dongen, Mitchel Geleijnse, Tom de With, Geert-Jan Witkamp and
Yuemei Lin (Applied Science). Everyone has contributed in one way
or another to this project and for that, I am very grateful. Also at
Wetsus, I found support from multiple people for brainstorming,
editing and even for high-end measurements; Philipp Wilfert,
Prashanth Kumar and Niels ter Hart, I will remember your
kindness. At Royal Haskoning DHV, I found the people with the
practical knowledge on the application issues of the full-scale
aerobic sludge technology. This was indispensable for my learning
curve and motivation. André van Bentem, Edward van Dijk, Ton
Oosterhoff, Jimmy van Opijnen, Arne Boersma, Jean Philippe
and Kerusha Lutchmiah; thank you for your time and openness.
Lastly, special gratitude goes to two international researchers that
I contacted and who responded with in-depth comments on our
50

10. References

Ajayi, I. A. and O. O. Ojelere (2013). "Chemical composition of ten Water Science and Technology 67(7): 1481-1489.
medicinal plant seeds from Southwest Nigeria." ALST 10: 25-32.

Batstone, D. and J. Keller (2003). "Industrial applications of the


Al-Borno, A. and M. B. Tomson (1994). "The temperature IWA anaerobic digestion model No. 1 (ADM1)." Water Science
dependence of the solubility product constant of vivianite." and Technology 47(12): 199-206.
Geochimica et Cosmochimica Acta 58(24): 5373-5378.

Beun, J., et al. (2000). "Aerobic granulation." Water Science &


APHA (1915). Standard methods for the examination of water and Technology 41(4): 41-48.
wastewater, American Public Health Association. American Public
Health Association American Water Works Association Water
Pollution Control Federation Water Environment Federation. Boskey, A. L. and A. S. Posner (1974). "Magnesium stabilization
of amorphous calcium phosphate: A kinetic study." Materials
Research Bulletin 9(7): 907-916.
Appelo, C., et al. (2002). "Surface complexation of ferrous iron
and carbonate on ferrihydrite and the mobilization of arsenic."
Environmental science & technology 36(14): 3096-3103. Bouropoulos, N. C. and P. G. Koutsoukos (2000). "Spontaneous
precipitation of struvite from aqueous solutions." Journal of
Crystal Growth 213(3–4): 381-388.
Appelo, C., et al. (1998). "A hydrogeochemical transport model
for an oxidation experiment with pyrite/calcite/exchangers/
organic matter containing sand." Applied geochemistry 13(2): 257- Burnett, R. and E. Frind (1987). "Simulation of contaminant
268. transport in three dimensions: 2. Dimensionality effects." Water
Resources Research 23(4): 695-705.

Appelo, C. A. J. and D. Postma (2005). Geochemistry, groundwater CBS (2015, 17-03-2015). "Centraal Bureau voor de Statistiek:
and pollution, CRC press. StatLine: Zuivering van stedelijks afvalwater." Retrieved 05-06-
16, 2016, from http://statline.cbs.nl/Statweb/publication/?DM=SL
NL&PA=71476ned&D1=12,15,18,24-28,54,56,58,61,64,67,71&D2=a
Ayala-Castro, C., et al. (2008). "Fe-S cluster assembly pathways &D3=a&HDR=T,G2&STB=G1&VW=T.
in bacteria." Microbiology and Molecular Biology Reviews 72(1):
110-125. Christoffersen, J. and M. R. Christoffersen (1979). "Kinetics of
dissolution of calcium hydroxyapatite: II. Dissolution in non-
stoichiometric solutions at constant pH." Journal of Crystal
Babić-Ivančić, V., et al. (2002). "Precipitation diagrams of struvite Growth 47(5): 671-679.
and dissolution kinetics of different struvite morphologies."
Croatica chemica acta 75(1): 89-106.
Christoffersen, J. and M. R. Christoffersen (1981). "Kinetics of
dissolution of calcium hydroxyapatite: IV. The effect of some
Banks, E., et al. (1975). "Crystal chemistry of struvite analogs of the biologically important inhibitors." Journal of Crystal Growth
type MgMPO4. 6H2O (M+= potassium (1+), rubidium (1+), cesium 53(1): 42-54.
(1+), thallium (1+), ammonium (1+)." Inorganic Chemistry 14(7):
1634-1639.
Cornell, R. M. and U. Schwertmann (2003). The Iron Oxides:
Structure, Properties, Reactions, Occurrences and Uses,
Barat, R., et al. (2008). "Interactions between calcium precipitation Wiley.
and the polyphosphate-accumulating bacteria metabolism." Water
Research 42(13): 3415-3424.
Crutchik, D. and J. M. Garrido (2016). "Kinetics of the reversible
reaction of struvite crystallisation." Chemosphere 154: 567-
Barat, R., et al. (2005). "The role of potassium, magnesium and 572.
calcium in the enhanced biological phosphorus removal treatment
plants." Environmental technology 26(9): 983-992.
De Bruin, L., et al. (2004). "Aerobic granular sludge technology:
an alternative to activated sludge?" Water Science & Technology
Barat, R., et al. (2011). "Modelling biological and chemically 49(11-12): 1-7.
induced precipitation of calcium phosphate in enhanced
biological phosphorus removal systems." Water Research 45(12):
3744-3752. De Kreuk, M. K. (2006). Aerobic granular sludge: scaling up a new
technology, TU Delft, Delft University of Technology.

Barat, R., et al. (2013). "Biological nutrient removal model Nº 2


(BNRM2): a general model for wastewater treatment plants." De Kreuk, M. K., et al. (2007). "Kinetic model of a granular
51

sludge SBR: influences on nutrient removal." Biotechnology and Giesen, A., et al. (2013). Full-scale experiences with aerobic
Bioengineering 97(4): 801-815. granular biomass technology for treatment of urban and
industrial wastewater. Proceedings of the International Water
Week Conference.
De Kreuk, M. K. and M. Van Loosdrecht (2004). "Selection of
slow growing organisms as a means for improving aerobic
granular sludge stability." Water Science & Technology 49(11-12): Graeser, S., et al. (2008). "Struvite-(K), KMgPO4· 6H2O, the
9-17. potassium equivalent of struvite–a new mineral." European
Journal of Mineralogy 20(4): 629-633.
De Ridder, M., et al. (2012). "Risks and Opportunities in the
Global Phosphate Rock Market." Robust Strategies in Times of
Uncertainty. Available online: http://www.phosphorusplatform. Hall, S. R., et al. (2003). "Morphosynthesis of complex
eu/images/download/HCSS_17_12_12_Phosphate.pdf (accessed inorganic forms using pollen grain templates." Chemical
on 18 October 2013). Communications(22): 2784-2785.


Hvitved-Jacobsen, T., et al. (2013). Sewer processes: microbial and
de Smit, S. M., et al. (2016). "BSc Thesis Project - Strenght chemical process engineering of sewer networks, CRC press.
characterization of aerobic granular sludge - Influence of storage
temperature and medium, salt and granule size." TU Delft faculty
Applied Science, Environmental Biotechnology. Inskeep, W. P. and J. C. Silvertooth (1988). "Kinetics of
hydroxyapatite precipitation at pH 7.4 to 8.4." Geochimica et
Cosmochimica Acta 52(7): 1883-1893.
Dongarra, J. J., et al. (1990). Solving linear systems on vector and
shared memory computers, Society for Industrial and Applied
Mathematics. Kagawa, Y., et al. (2015). "Modeling the nutrient removal process
in aerobic granular sludge system by coupling the reactor‐and
granule‐scale models." Biotechnology and Bioengineering 112(1):
Doyle, J., et al. (2000). "Analysis of struvite precipitation in real and 53-64.
synthetic liquors." Process Safety and Environmental Protection
78(6): 480-488.
Kanzaki, N., et al. (2000). "Inhibitory Effect of Magnesium and
Zinc on Crystallization Kinetics of Hydroxyapatite (0001) Face."
Dudka, S. and D. C. Adriano (1997). "Environmental impacts The Journal of Physical Chemistry B 104(17): 4189-4194.
of metal ore mining and processing: a review." Journal of
Environmental Quality 26(3): 590-602.
Koutsoukos, P., et al. (1980). "Crystallization of calcium
phosphates. A constant composition study." Journal of the
Dupraz, C. and P. T. Visscher (2005). "Microbial lithification American Chemical Society 102(5): 1553-1557.
in marine stromatolites and hypersaline mats." Trends in
microbiology 13(9): 429-438.
Kuder, T., et al. (2014). "User's Guide, Integrated Stable Isotop
- Reactive Transport Model Approach for Assessment of
Dupraz, S., et al. (2009). "Experimental and numerical modeling of Chlorinated Solvent Degradation ESTCP Project ER-201029."
bacterially induced pH increase and calcite precipitation in saline
aquifers." Chemical Geology 265(1): 44-53.
Lemaire, R., et al. (2008). "Micro-scale observations of the structure
of aerobic microbial granules used for the treatment of nutrient-
Ferris, F. G., et al. (1987). "Bacteria as nucleation sites for authigenic rich industrial wastewater." The ISME journal 2(5): 528-541.
minerals in a metal-contaminated lake sediment." Chemical
Geology 63(3-4): 225-232.
Lemaire, R., et al. (2008). "Microbial distribution of Accumulibacter
spp. and Competibacter spp. in aerobic granules from a lab‐scale
Filipe, C. D., et al. (2001). "A metabolic model for acetate uptake biological nutrient removal system." Environmental microbiology
under anaerobic conditions by glycogen accumulating organisms: 10(2): 354-363.
stoichiometry, kinetics, and the effect of pH." Biotechnology and
Bioengineering 76(1): 17-31.
Li, Y., et al. (2014). "Aerobic granular sludge for simultaneous
accumulation of mineral phosphorus and removal of nitrogen via
Filipe, C. D., et al. (2001). "Stoichiometry and kinetics of acetate nitrite in wastewater." Bioresource technology 154: 178-184.
uptake under anaerobic conditions by an enriched culture
of phosphorus‐accumulating organisms at different pHs."
Biotechnology and Bioengineering 76(1): 32-43. Lin, Y., et al. (2012). "The contribution of exopolysaccharides
induced struvites accumulation to ammonium adsorption in
52

aerobic granular sludge." Water Research 46(4): 986-992. Manas, A., et al. (2012). "Location and chemical composition of
microbially induced phosphorus precipitates in anaerobic and
aerobic granular sludge." Environmental technology 33(19): 2195-
Lin, Y. M., et al. (2003). "Development and characteristics of 2209.
phosphorus-accumulating microbial granules in sequencing
batch reactors." Applied microbiology and biotechnology 62(4):
430-435. Margolis, H. and E. Moreno (1992). "Kinetics of hydroxyapatite
dissolution in acetic, lactic, and phosphoric acid solutions."
Calcified tissue international 50(2): 137-143.
Liu, X.-W., et al. (2009). "Physicochemical characteristics of
microbial granules." Biotechnology advances 27(6): 1061-
1070. Mathew, M., et al. (1982). "A new struvite-type compound,
magnesium sodium phosphate heptahydrate." Acta
Crystallographica Section B: Structural Crystallography and
Liu, Y.-Q., et al. (2016). "Size-dependent calcium carbonate Crystal Chemistry 38(1): 40-44.
precipitation induced microbiologically in aerobic granules."
Chemical Engineering Journal 285: 341-348.
Maurer, M., et al. (1999). "Kinetics of biologically induced
phosphorus precipitation in waste-water treatment." Water
Liu, Y., et al. (2003). "Mechanisms and models for anaerobic Research 33(2): 484-493.
granulation in upflow anaerobic sludge blanket reactor." Water
Research 37(3): 661-673.
Mbamba, C. K., et al. (2016). "Validation of a plant-wide
phosphorus modelling approach with minerals precipitation in a
Lopez-Vazquez, C. M., et al. (2009). "Modeling the PAO–GAO full-scale WWTP." Water Research 100: 169-183.
competition: effects of carbon source, pH and temperature." Water
Research 43(2): 450-462.
Mbamba, C. K., et al. (2015). "A systematic study of multiple
minerals precipitation modelling in wastewater treatment." Water
Luff, B. B. and R. B. Reed (1980). "Thermodynamic properties of Research 85: 359-370.
magnesium potassium orthophosphate hexahydrate." Journal of
Chemical and Engineering Data 25(4): 310-312.
Mersmann, A. (2001). Crystallization technology handbook, CRC
Press.
Madsen, H. E. L. (2002). "Crystallization of heavy-metal
phosphates in solution—III: nucleation and growth kinetics of Cd
5 H 2 (PO 4) 4· 4H 2 O at 37° C." Journal of Crystal Growth 244(3): Minteq database (2006). "Minteq database version 4, derived from
349-358. MINTEQA2." Database provider USGS.gov.

Madsen, H. E. L. and F. Christensson (1991). "Precipitation of Morgan, B. and O. Lahav (2007). "The effect of pH on the kinetics
calcium phosphate at 40 C from neutral solution." Journal of of spontaneous Fe (II) oxidation by O 2 in aqueous solution–basic
Crystal Growth 114(4): 613-618. principles and a simple heuristic description." Chemosphere
68(11): 2080-2084.

Madsen, H. E. L. and H. C. B. Hansen (2014). "Kinetics of crystal


growth of vivianite, Fe 3 (PO 4) 2· 8H 2 O, from solution at 25, 35 Morgenroth, E., et al. (1997). "Aerobic granular sludge in a
and 45° C." Journal of Crystal Growth 401: 82-86. sequencing batch reactor." Water Research 31(12): 3191-3194.

Mahamid, J., et al. (2008). "Amorphous calcium phosphate is a Mosquera-Corral, A., et al. (2005). "Effects of oxygen concentration
major component of the forming fin bones of zebrafish: Indications on N-removal in an aerobic granular sludge reactor." Water
for an amorphous precursor phase." Proceedings of the National Research 39(12): 2676-2686.
Academy of Sciences 105(35): 12748-12753.

Murray, K. and P. May (1996). Joint Expert Speciation System


Mañas, A., et al. (2011). "Biologically induced phosphorus (JESS) Primer. An international computer system for determining
precipitation in aerobic granular sludge process." Water Research chemical speciation in aqueous and non-aqueous environments.
45(12): 3776-3786. M. Murdoch University, Western Australia and the Division of
Water Technology, CSIR, Pretoria, South Africa.

Mañas, A., et al. (2012). "Parameters influencing calcium


phosphate precipitation in granular sludge sequencing batch Musvoto, E., et al. (2000). "Integrated chemical–physical processes
reactor." Chemical Engineering Science 77: 165-175. modelling—II. Simulating aeration treatment of anaerobic
53

digester supernatants." Water Research 34(6): 1868-1880. Rahman, M. M., et al. (2014). "Production of slow release crystal
fertilizer from wastewaters through struvite crystallization–A
review." Arabian Journal of Chemistry 7(1): 139-155.
Nielsen, A. E. (1984). "Electrolyte crystal growth mechanisms."
Journal of Crystal Growth 67(2): 289-310.
Royal Haskoning DHV (2016). Online database Nereda,
NeredaAnalyse_Garmerwolde. In-company data. t. N. Royal
Ohlinger, K., et al. (1998). "Predicting struvite formation in HaskoningDHV, Watertechnology. Amersfoort.
digestion." Water Research 32(12): 3607-3614.

RoyalHaskoningDHV (2016). Online database Nereda,


Parkhurst, D., et al. (1980). "PHREEQE: A computer program for NeredaAnalyse_Garmerwolde. In-company data. t. N.
Geochemical Calculations, USGS Water Res." Invest. Report: 80- RoyalHaskoningDHV, Watertechnology. Amersfoort.
60.

Saunders, A. M., et al. (2007). "Proton motive force generation


Parkhurst, D. L. and C. Appelo (1999). "User's guide to PHREEQC from stored polymers for the uptake of acetate under anaerobic
(Version 2): A computer program for speciation, batch-reaction, conditions." FEMS microbiology letters 274(2): 245-251.
one-dimensional transport, and inverse geochemical calculations."
U.S. Geological Survey Water-Resource Investigations Report 99-
4259: 312 p. Schönborn, C., et al. (2001). "Stability of enhanced biological
phosphorus removal and composition of polyphosphate
granules." Water Research 35(13): 3190-3196.
Parkhurst, D. L. and C. Appelo (2013). Description of input and
examples for PHREEQC version 3: a computer program for
speciation, batch-reaction, one-dimensional transport, and inverse Schroeder, E. (1968). "Importance of the BOD Plateau." Water
geochemical calculations, US Geological Survey. Chapter 43 of Research 2(11): 803-809.
Section A, Groundwater, Book 6, Modeling Techniques.

Schwertmann, U. and R. M. Cornell (2008). Iron oxides in the


Parkhurst, D. L. K., Kenneth L. and S. R. Charlton (2010). "PHAST laboratory: preparation and characterization, John Wiley &
Version 2—A program for simulating groundwater flow, solute Sons.
transport, and multicomponent geochemical reactions." Modeling
Techniques 6(Chapter 35, section A, Groundwater).
Seckler, M., et al. (1996). "Calcium phosphate precipitation in
a fluidized bed in relation to process conditions: a black box
Peeters, B., et al. (2011). "Quantification of the exchangeable approach." Water Research 30(7): 1677-1685.
calcium in activated sludge flocs and its implication to sludge
settleability." Separation and purification technology 83: 1-8.
Serralta, J., et al. (2004). "An extension of ASM2d including pH
PHREEQC database (2016, 21-4-2016 13:44:52). "PHREEQC calculation." Water Research 38(19): 4029-4038.
(Version 3) -- A computer Program for Speciation, Batch-
Reaction, One-Dimensional Transport, and Inverse Geochemical Smolders, G., et al. (1994). "Model of the anaerobic metabolism of
Calculations." Retrieved 24-4-2016 21:23, 2016, from http:// the biological phosphorus removal process: stoichiometry and pH
wwwbrr.cr.usgs.gov/projects/GWC_coupled/phreeqc/. influence." Biotechnology and Bioengineering 43(6): 461-470.
Picioreanu, C. (2015). Biofilm modeling, Advanced Biofilm Course
10, Delft, 2015. Delft, Technical University Delft. Smolders, G., et al. (1995). "A structured metabolic model for
anaerobic and aerobic stoichiometry and kinetics of the biological
Pijuan, M., et al. (2009). "Effect of long term anaerobic and phosphorus removal process." Biotechnology and Bioengineering
intermittent anaerobic/aerobic starvation on aerobic granules." 47(3): 277-287.
Water Research 43(14): 3622-3632.
Smolders, G. J. F. (1995). A metabolic model of the biological
Pronk, M., et al. (2015). "Full scale performance of the aerobic phosphorus removal. Stoichiometry, kinetics and dynamic
granular sludge process for sewage treatment." Water Research behaviour, TU Delft, Delft University of Technology.
84: 207-217.
Song, Y., et al. (2002). "Effects of solution conditions on the
Rahaman, M., et al. (2006). "Exploring the determination of precipitation of phosphate for recovery: A thermodynamic
struvite solubility product from analytical results." Environmental evaluation." Chemosphere 48(10): 1029-1034.
technology 27(9): 951-961.
54

Steinhorst, L. and J. Kudla (2013). "Calcium - a central regulator van der Roest, H., et al. (2011). "Towards sustainable waste water
of pollen germination and tube growth." Biochimica et Biophysica treatment with Dutch Nereda® technology." Water Practice and
Acta (BBA) - Molecular Cell Research 1833(7): 1573-1581. Technology 6(3).

STOWA (2015). "Nereda®: Nerlands succes van laboratorium tot van Kemenade, M. J. J. M. and P. L. de Bruyn (1987). "A kinetic
wereldwijde toepassing." Stichting Toegepast Onderzoek voor study of precipitation from supersaturated calcium phosphate
Waterbeheer (STOWA). SOON rapport(1). solutions." Journal of colloid and interface science 118(2): 564-
585.

Stubbé, S. M. L., et al. (2016). MSc Thesis The Fate of


Phosphate in Full-scale Aerobic Granular Sludge Systems. C. Van Langerak, E., et al. (2000). "Impact of location of CaCO 3
E. Technical University of Delft, Water management, Sanitary precipitation on the development of intact anaerobic sludge."
Engineering. Water Research 34(2): 437-446.

Stumm, W., et al. (1996). "Aquatic chemistry." Journal of Viéitez, E. R. and S. Ghosh (1999). "Biogasification of solid wastes
Environmental Quality 25(5): 1162. by two-phase anaerobic fermentation." Biomass and Bioenergy
16(5): 299-309.

Su, C. and D. L. Suarez (1997). "In situ infrared speciation of


adsorbed carbonate on aluminum and iron oxides." Wan, C., et al. (2015). "Calcium precipitate induced aerobic
granulation." Bioresource technology 176: 32-37.

Tay, J. H., et al. (2003). "Biomass and porosity profiles in microbial


granules used for aerobic wastewater treatment." Letters in Wang, D., et al. (2013). "Effect of initial pH control on biological
Applied Microbiology 36(5): 297-301. phosphorus removal induced by the aerobic/extended-idle
regime." Chemosphere 90(8): 2279-2287.

Taylor, A., et al. (1963). "Solubility products of magnesium


ammonium and magnesium potassium phosphates." Transactions Watson Database (2016). Emmissieregistratie, Emissies, In- en
of the Faraday Society 59: 1580-1584. effluent van RWZI's, Watson database between 1990-2016,
Microverontreinigingen in influent and effluent van RWZI's. N.
Rijksoverheid.
TenHuisen, K. S. and P. W. Brown (1997). "Effects of magnesium on
the formation of calcium‐deficient hydroxyapatite from CaHPO4·
2H2O and Ca4 (PO4) 2O." Journal of biomedical materials research Wilfert, P., et al. (2015). "The relevance of phosphorus and iron
36(3): 306-314. chemistry to the recovery of phosphorus from wastewater: a
review." Environmental science & technology 49(16): 9400-
9414.
The Crystalactor (2016). "Royal Haskoning DHV,."

Wilfert, P., et al. (submitted). "Vivianite as an important iron


Tipping, E. and M. Hurley (1992). "A unifying model of cation phosphate precipitate in sewage treatment plants." Water
binding by humic substances." Geochimica et Cosmochimica Acta Research.
56(10): 3627-3641.

Winkler, M.-K., et al. (2012). "Temperature and salt effects on


Toh, S., et al. (2003). "Size-effect on the physical characteristics settling velocity in granular sludge technology." Water Research
of the aerobic granule in a SBR." Applied microbiology and 46(16): 5445-5451.
biotechnology 60(6): 687-695.

Winkler, M. H., et al. (2013). "Factors influencing the density of


Truesdell, A. H. and B. F. Jones (1974). "WATEQ, a computer aerobic granular sludge." Applied microbiology and biotechnology
program for calculating chemical equilibria of natural waters." J. 97(16): 7459-7468.`
Res. US Geol. Surv 2(2): 233-248.
World Water (2015). "Nereda technology gains worldwide
van Breukelen, B. M., et al. (2004). "Reactive transport modelling of recognition." January/February. Retrieved 1-9-2019, 2016, from
biogeochemical processes and carbon isotope geochemistry inside https://www.royalhaskoningdhv.com/nereda/~/media/nereda/
a landfill leachate plume." Journal of contaminant hydrology files/public/articles%20and%20papers/public/201501_world%20
70(3): 249-269. water_nereda%20technology%20gains%20worldwide%20
recognition_annemie.pdf?la=en-gb.

55

Wu, M. M. and R. F. Hickey (1997). "Dynamic model for UASB


reactor including reactor hydraulics, reaction, and diffusion."
Journal of environmental engineering 123(3): 244-252.

Xavier, J. B., et al. (2007). "Multi-scale individual-based model


of microbial and bioconversion dynamics in aerobic granular
sludge." Environmental science & technology 41(18): 6410-
6417.

Xiao, F., et al. (2008). "Physical and hydrodynamic properties


of aerobic granules produced in sequencing batch reactors."
Separation and purification technology 63(3): 634-641.

Xu, K., et al. (2015). "The precipitation of magnesium potassium


phosphate hexahydrate for P and K recovery from synthetic
urine." Water Research 80: 71-79.

Yilmaz, G., et al. (2008). "Simultaneous nitrification, denitrification,


and phosphorus removal from nutrient‐rich industrial wastewater
using granular sludge." Biotechnology and Bioengineering 100(3):
529-541.

Zhang, H.-L., et al. (2013). "Phosphorus removal in an enhanced


biological phosphorus removal process: roles of extracellular
polymeric substances." Environmental science & technology
47(20): 11482-11489.

Zhang, L., et al. (2007). "Role of extracellular protein in the


formation and stability of aerobic granules." Enzyme and
Microbial Technology 41(5): 551-557.


•••

56

Addendum 1 Biological Model

Table 11a Symbols, units and constants used in the metabolic conversions
Table XX Symbols, units and constants used in the metabolic conversions
Description Value Dimension Reference

Half-saturation constants

KHAc Coefficient for acetate 0.001 Cmol m-3 Lopez-Vazquez et al. (2009)
Kpp Coefficient for Poly-P 0.01 Pmol m-3 Lopez-Vazquez et al. (2009)

KGly Coefficient for glycogen 0.01 Cmol m-3 Lopez-Vazquez et al. (2009)

KPHA Coefficient for PHA 0.01 Cmol m -3


Lopez-Vazquez et al. (2009)
K PHA Coefficient for fraction of PHA 0.01 Cmmol Cmmol-1 Lopez-Vazquez et al. (2009)

KPO4 Coefficient for orthophosphate 0.01 Pmol m-3 Lopez-Vazquez et al. (2009)

KO2 Coefficient for oxygen 0.01 O2mol m-3 Lopez-Vazquez et al. (2009)

Kinetic constants

 Maximum uptake rate for acetate 0.2 Cmmol Cmmol-1 h-1 1 Lopez-Vazquez et al. (2007) 3

 Anaerobic maintenance coefficient 0.00235 Pmmol Cmmol-1 h-1 1 Smolders et al. (1994) 3

 Aerobic maintenance coefficient 0.004 Pmmol Cmmol-1 h-1 1 Brdjanovic et al. (1997) 3

 Aerobic PHA degradation rate 0.80 Cmmol Cmmol h -1 -1 1
Lopez-Vazquez et al. (2009)

 Aerobic glycogen production rate 0.015 Cmmol Cmmol-1 h-1 1 Lopez-Vazquez et al. (2009)

 Aerobic Poly-P formation rate 0.02 Pmmol Cmmol-1 h-1 1 Lopez-Vazquez et al. (2009)

Solutes and Components


 Acetate concentration multiplied by 22 calc4 Cmol m-3 -
 Orthophosphate concentration calc4 Cmol m-3 -
 Oxygen concentration calc4 O2mol m-3 -
 Density biomass of PAO 51 Cmol m -3
Lopez-Vazquez et al. (2009)

, Maximum fraction of glycogen 0.27 Cmmol Cmmol-1 Wentzel et al. (1989) 3

, Maximum fraction of Poly-P 0.30 Pmmol Cmmol-1 Smolders et al. (1995) 3
 Poly-P concentration calc 4 Pmol m -3 -
 Glycogen concentration calc4 Cmol m-3 -
 PHA concentration calc4 Cmol m-3 -
 Intracellular fraction of PHA calc 4
- -
, =



, = Intracellular fraction of glycogen calc4 - -



, = Intracellular fraction of Poly-P calc4 - -


1 divided by 3600 to adjust to Cmol Cmol-1 s-1 to fit in the PHREEQC software
2 multiplied by 2 to get Cmol m-3, since acetate in PHREEQC contains 2 C’s.

3 These are the original sources of the parameter values, also used in Lopez-Vazquez et al. (2009)

4 calc. means the model calculated the values

2
57

Table
Table XX Symbols,
11b Symbols, units and
units and constants constants
used in the biologicalused
model in the biological model
Desciption Value Dimension Reference

Diffusion-coefficient of biopolymers

DPHA For PHA 1e-20 m2/s

DGly For Glycogen 1e-20 m2/s

Dpp For Poly-P 1e-20 m2/s

Molecular Mass

Poly-P K0.28Mg0.36H2PO4 116.67 g/mol Adjusted Barat et al. (2005)

PHA CH1.5O0.5 21.5 g/mol Smolders (1995)

Glycogen CH1.67O0.83 27.0 g/mol Smolders (1995)

Biomass CH2.09O0.54N0.2P0.015 26.02 g/mol Smolders (1995)

H(Acetate) C2H4O2 60.053 g/mol Minteq database (2006)


58

Addendum 2 Chemical Model

The Dinxperlo installation, designed by Royal HaskoningDHV,


has been in operation since November 2013 (STOWA 2015). Three
tanks are employed to continuously treat incoming wastewater
with an average discharge of 3,100 m3 d-1. Wastewater enters
the installation via a small influent buffer (270 m3) after which it
continues through the screens (3 mm). In the next phase, grit is
removed before entering the intermediate pumping station (640
m3 h-1). The three AGS reactors (height 7.5 m, volume 1,250 m3
each) are followed by three sand filters. The excess sludge of the
reactors is collected in an aerated buffer tank (100 m3) to prevent
phosphate release by the PAOs. After gravitational thickening,
the sludge is collected in another buffer tank (270 m3) before
transportation.

Fig. 31 Microscopic pictures from granules harvested at the full-


scale aerobic granular sludge system in Dinxperlo during the
aeration phase.
•••

59

Table 12a Equilibrium matrix for the phases considered in this study
Table XX Equilibrium matrix for the phases considered in this study
Solutes Log Reference*
Phase H2 H PO43 NH4 CO32 Ca2 Mg2 K Fe2+ Fe3+ Na SO42 NO Cl-
O + - + - + + + + - -

HAP Ca5(PO4)3OH:6H2 7 -1 3 5 -50.14 see Table 5


O
- Ca4(PO4)3:3H2O 3 1 3 4 -47.08

ACP Ca3(PO4)2:6H2O 6 2 3 -27.32 see Table 5

Brushite CaHPO4:2H2O 2 1 1 1 -18.995

STR-NH4 NH4MgPO4:6H2 6 1 1 1 -13.46 see Table 5


O
STR-K KMgPO4:6H2O 6 1 1 1 -11.5 see Table 5

STR-Na NaMgPO4:7H2O 7 1 1 1 -11.6 see Table 5

Vivianite Fe3(PO4)2:8H2O 8 2 3 -35.884 see Table 5

Strengite FePO4:2H2O 2 1 1 -26.4

- Mg3(PO4)2 2 3 -25.4

- MgHPO4:3H2O 3 1 1 1 -18.175

- Fe3(OH)8 8 -8 1 2 20.222

Lepidocrocite FeOOH 2 -3 1 1.371

Minteq database (2006)


Goethite FeOOH 2 -3 1 0.491

Hematite Fe2O3 3 -6 2 -1.418

Magnetite Fe3O4 4 -8 1 2 3.4028

Ferrihydrite Fe(OH)3 3 -3 1 1 3.191

Maghemite Fe2O3 3 -6 2 6.386

Magnesioferrite Fe2MgO4 4 -8 1 2 16.8592

Dolomite CaMg(CO3)2 2 1 1 -16.54

Calcite CaCO3 1 1 -8.48

K-Jarosite KFe3(SO4)2(OH)6 6 -6 1 3 2 -14.8

*For all phases without reference, the solubility products originates from Minteq database (2006)

3
60

Table XX Added phases to the PHREEQC database including solubility products (Kp, log_k) and
Table 12b Added phases to the PHREEQC database including solubility products (Kp, log_k) and enthalpies (ΔH, delta_h)
enthalpies (ΔH, delta_h)
Mineral Equation, Kp and ΔH Reference

Amorphous Calcium Phosphate Ca3(PO4)2:6H2O = 3 Ca + 2 PO4 + 6H2O


+2 -3

(ACP), log_k -27.32 see Table 5

Ca3(PO4)2:6H 2O delta_h 15.95 kJ/mol Kanazawa et al. (1982)

Hydroxyapatite Ca5(PO4)3OH:6H2O = 5 Ca+2 + 3 PO4-3 + 1 OH- + 6 H2O

(HAP), log_k -50.14 see Table 5

Ca5(PO4)3OH:6H2O delta_h 98.85 kJ/mol TenHuisen et al. (1998), Durucan et al. (2000)

Struvite NH4MgPO4:6H2O = Mg + PO4 + NH + 6H2O


+2 -3 4+

(STR-NH4) log_k -13.46 see Table 5

MgNH4PO4:6H2O delta_h 23.62 kcal/mol Bhuiyan et al. (2007)

Potassium Struvite KMgPO4:6H2O = Mg + PO4 + K + 6H2O


+2 -3 +

(STR-K) log_k -11.5 see Table 5

MgKPO4:6H2O delta_h 14.53 kcal Luff et al. (1980)

Sodium Struvite NaMgPO4:7H2O = Mg+2 + PO4-3 + Na+ + 7H2O

(STR-Na) log_k -11.6 see Table 5

MgNaPO4:6H 2O delta_h 96.28 kJ.mol Chauhan et al. (2014)

Vivianite Fe3(PO4)2:8H2O + 2 H2PO = 3 Fe + 4 HPO


4
- 2+
4
2-
+ 8 H2O

(VIV), Log_k -35.884 see Table 5

Fe3(PO4)2:8H 2O delta_h 0 kJ Minteq database (2006)

4
•••

61

Table
Table 13a XX Symbols,
Symbols, units
units and and constants
constants used in themodel
used in the precipitation precipitation model
of amorphous ofphosphate
calcium amorphous calcium
(ACP), phosphate
hydroxyapatite (HAP),
struvite (STR-NH4) (I)
(ACP), hydroxyapatite (HAP), struvite (STR-NH4) (I)
Description Value Dimension Reference

ACP precipitation and dissolution Mbamba et al. (2015)


kACP kinetic constant 0.000875 s-1 1
AACP total ACP concentration, representing surface area PACP + PACP,0 mol L-1

PACP ACP mineral concentration calc2 mol L-1

PACP,0 ACP mineral concentration at time 0 3 1e-10 mol L-1

KACP ACP solubility product 10-27.32 mol L-1

a[Ca2+] Calcium activity calc2 mol L-1

a[PO ]
4 3- Phosphate activity calc 2 mol L-1

HAP precipitation and dissolution Inskeep et al. (1988)


kHAP kinetic constant 173 L mol m s
2 -1 -2 -1

AHAP Surface area AAHAP *TPHAP * MWHAP m2 L-1

AAHAP Specific surface area of crystal 49.7 m2 g-1

MWHAP Molecular weight HAP 502.31 g mol-1

TPHAP Total HAP mineral concentration PHAP + PHAP,0 mol L-1

PHAP HAP mineral concentration calc 2


mol L-1

PHAP,0 HAP mineral concentration at time 03 1e-15 mol L-1

KHAP HAP solubility constant 10 -50.14


mol L-1

a[Ca ]2+ Calcium activity calc 2 mol L-1

a[PO43-] Phosphate activity calc2 mol L-1

a[OH ] - Hydroxyl activity calc 2 mol L-1

STR-NH4 precipitation and dissolution Mbamba et al. (2015)


kSTR kinetic constant 8.89e-4 s -11

ASTR total STR-NH4 concentration, representing surface PSTR + PSTR,0 mol L-1

PSTR Struvite mineral concentration calc2 mol L-1

PSTR,0 Struvite mineral concentration at time 0 3 1e-7 mol L-1

(KSTR STR solubility constant 10-13.26 mol L-1

a[Mg ]2+ Magnesium activity calc 2 mol L-1

a[PO ]
4 3- Phosphate activity calc 2 mol L-1

a[NH+] Ammonium activity calc2 mol L-1


1 converted to s to fit in the PHREEQC software conditions
-1

2 calc. = calculated by the model

3 to represent a nucleus, without affecting the total amount of precipitates in the available time

5
•••

62

Table 13b Symbols, units and constants used in the precipitation model for vivianite (VIV) and ferrihydrite (FER) (II)
Table XX Symbols, units and constants used in the precipitation model for vivianite (VIV) and ferrihydrite
Description Value Dimension Reference

VIV precipitation and dissolution Madsen et al. (2014)


kVIV Rate constant (ratio dependent, on average;) 1.1517e-12 m L mol-1 s-11
PVIV VIV mineral concentration calc 2 mol L-1

SFe Iron (II) concentration calc 2 mol L-1

SFe,equi Equilibrium iron (II) concentration ((H2PO *KVIV)/(HPO ))


4
-
4
- 1/3
mol L-1

KVIV VIV solubility constant 10-36 mol L-1

H2PO 4
-
H2PO concentration
4
-
calc 2
mol L-1

HPO4 HPO4 concentration calc 2 mol L-1

a[Fe2+] Iron (II) activity calc 2 mol L-1

a[PO ]43- Phosphate activity calc 2 mol L-1

CVIV Convertion from length (m L-1) to concentration (mol Ρ * AAVIV / MWVIV mol m-1
 Density of vivianite 2.68 g cm-3 Mindat.org (2016)

AAVIV Specific surface area of crystal per cm growth 1892.25e-12 cm-3 cm

MWVIV Molecular weight vivianite 501.6 g mol -1

FER precipitation and dissolution

kFER kinetic constant 10-12 L2 mol-1 m-2 s-1


AFER Surface area AAFER *TPFER * MWFER m2 L-1

TPFER total FER mineral concentration PFER + PFER,0 mol L-1

PFER Struvite mineral concentration calc 2 mol L-1

PFER,0 Struvite mineral concentration at time 0 3 10 -15 mol L-1

AAFER Specific surface area of crystal 200 m2 g-1 Schwertmann (1991)

MWFER Molecular weight ferrihydrite 106.863 g mol -1

a[Fe2+] Iron (II) activity calc 2 mol L-1

a[OH-] Hydroxyl activity calc 2 mol L-1

KFER FER solubility constant 10 -3.131


mol L-1 Minteq database (2006)
1 converted to s-1 to fit in the PHREEQC software conditions

2 calc. = calculated by the model

3 to represent a nucleus, without affecting the total amount of precipitates in the available time

6
63

Elaborate precipitation model the appearance of vivianite, although this is only researched in
For calcium phosphate minerals, it is accepted that the least lake sediments (Miot et al. 2009). The rate expression proposed
thermodynamically stable mineral (ACP) acts as a precursor for by {Madsen, 2014 #314@@author-year} describes the nucleation
the most stable mineral (HAP), following the Ostwald ripening kinetics as surface-controlled. The precipitation kinetics, however,
rule (Koutsoukos et al. 1980, van Kemenade et al. 1987, Madsen are proposed as a spiral growth mechanism, dependent on the
et al. 1991, Musvoto et al. 2000). In this study, the precursor ACP saturation level. It remains unknown whether the oxidation
was chosen based on the occurrence in lab-scale granules (Mañas by bacteria of vivianite wins from the (slow) formation rate
et al. 2011). The precipitation process was modelled as reversible of vivianite. Lepidocrocite is proposed as a possible oxidation
crystallisation for calcium and phosphate ions to ACP and as product in calcareous medium by Roldán et al. (2002). To the
irreversible crystallisation for ACP to HAP, following Maurer author’s knowledge, no vivianite kinetic rate is reported in
et al. (1999) and Barat et al. (2011). Nevertheless, dissolution molar concentrations. However, Madsen et al. (2014) reported
of HAP can occur although the process is complex and slow vivianite kinetic rates at 25, 35 and 45 oC based on crystal length
(Christoffersen et al. 1979, Margolis et al. 1992, Robinson et (m h-1). This rate was converted to molar concentration based on
al. 2000, Appelo et al. 2005). Since the crystallisation process of density, molecular weight, average measured crystal area and
calcium phosphate minerals is surface-controlled (Appelo et al. the assumption that the crystal growth at the face has the lowest
2005), this was represented in the model by the mineral content rate of advancement. This was done in communication with the
concentration (AACP, AHAP), following Mbamba et al. (2015) (see author of the research at hand (Madsen et al. 2014).
Table 13a). It should be noted that the ACP kinetic parameter For the iron oxides, it is acknowledged that ferrihydrite
was studied on sludge digestate, therefore indicated only the is the precursor for multiple complex iron oxides such as, e.g.,
upper limit of ACP precipitation. The real ACP precipitation will goethite, hematite (Schwertmann 1991). The formation rates
be slower. The inhibitor effect of magnesium on ACP and HAP were however, scarcely reported. The general kinetic expression
precipitation was not taken into account for simplicity (Boskey et of Koutsoukos et al. (1980) was therefore applied with surface
al. 1974, TenHuisen et al. 1997, Kanzaki et al. 2000). area measurements of Schwertmann et al. (2008) and Cornell et
For magnesium phosphate minerals, different kinetic al. (2003). A low rate coefficient was adapted, to not overestimate
expressions are reported. The kinetic expression formulated the precipitation of ferrihydrite. This inaccuracy of the model
by Mbamba et al. (2015) was inserted into the model since this was undesirable, but the only way to include the competition of
expression incorporated a surface-dependency (ASTR). Nelson available Fe ions between phosphate minerals and iron oxides.
et al. (2003) and Le Corre et al. (2007) determined a first-order Also, dissolution of iron oxides was taken into account by
kinetic rate, whereas Türker et al. (2007) developed a second- reversing the rate expression at negative SI, since the dissolution
order kinetic rate expression. An agreement is established that is indicated to be relatively rapid (Schwertmann 1991, Paige et al.
struvite precipitation is a surface-controlled process, or in other 1997). At the same time, the inclusion of ferrihydrite is the first step
words, mostly heterogeneous precipitation occurs (Mersmann towards inclusion of surface complexation, further elaborated on
2001, Crutchik et al. 2016). The dissolution kinetics by Babić- in the Outlook-section.
Ivančić et al. (2002) and Bhuiyan et al. (2007) suggest that
the equilibria are established relatively fast and that they are
dependent on the crystal shape. The interference of calcium on
struvite precipitation through the competition for phosphate
ions was taken into account through the chemical equilibria
(Bouropoulos et al. 2000, Minteq database 2006). Due to a lack
of literature, the same kinetic expression is assumed for SRT-K
and SRT-Na with their respective ion concentrations.
For iron phosphate minerals, the kinetics is less widely
studied for waste water systems. Vivianite is one of the few
iron(II) phosphate minerals, with a very slow crystallization rate
(Postma 1980, Persson et al. 2012, Acton 2013). It is acknowledged
that this crystallization is also surface-controlled rather than a
diffusion mechanism (Emerson et al. 1978, Madsen et al. 2014).
The influence of bacteria on the iron oxidation state can influence
•••

64

Addendum 3 Ion transport

Table 14 Symbols, units and constants used in the ion transport


Table XX Symbols, units and constants used in the ion transport
Desciption Value Dimension Reference

Diffusion-coefficient of solutes

DH H+ 9.31 10-9 m2 s-1

DO2 O2 2.31 10-9 m2 s-1

DCl Cl- 2.03 10-9 m2 s-1

DNH4 NH4+ 1.98 10-9 m2 s-1

PHREEQC database (2016)


DK K+ 1.96 10-9 m2 s-1

DNO3 NO3- 1.9 10-9 m2 s-1

D Na Na+ 1.33 10-9 m2 s-1

DSO4 SO43- 1.07 10-9 m2 s-1

DCO3 CO32- 0.955 10-9 m2 s-1

D Ca Ca2+ 0.793 10-9 m2 s-1

D Fe Fe2+/3+ 0.719 10-9 m2 s-1

DMg Mg2+ 0.705 10-9 m2 s-1

D PO4 PO43- 0.612 10-9 m2 s-1

D Acetate Acetate- 1.1 10-9 m2 s-1 Xavier et al. (2007)


65

Addendum 4

Detailed graphs of the dynamics of the mineral content gradient over a granule of 1 mm radius during a standard cycle

# 10-8 ACP content # 10-16 HAP content


1 1.5
0 min 0 min
15 min 15 min
0.9
30 min 30 min
45 min 45 min
0.8
60 min 60 min
75 min 75 min
0.7

HAP content (mol)


ACP content (mol)

90 min 1 90 min
105 min 105 min
0.6 120 min 120 min
135 min 135 min
0.5 150 min 150 min
165 min 165 min
0.4 180 min 180 min
195 min 0.5 195 min
0.3 210 min 210 min
225 min 225 min
0.2 240 min 240 min

0.1

0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Radius (mm) Radius (mm)
VIV content STR content
# 10-21 1
3.5 0 min
0 min
15 min 15 min
0.8
30 min 30 min
3 45 min
45 min 0.6
60 min 60 min
75 min 75 min
2.5 0.4
STR content (mol)

90 min
VIV content (mol)

90 min
105 min 105 min
0.2 120 min
2 120 min
135 min 135 min
0 150 min
150 min
165 min 165 min
1.5 -0.2
180 min 180 min
195 min 195 min
210 min -0.4 210 min
1
225 min 225 min
240 min -0.6 240 min
0.5
-0.8

0 -1
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Radius (mm) Radius (mm)
pH PO 4 concentration
7.3 # 10-3
0 min 9
0 min
15 min
7.2 15 min
30 min 8 30 min
45 min
PO 4 concentration (mol L -1 )

7.1 45 min
60 min 7 60 min
75 min
7 75 min
90 min
6 90 min
105 min
6.9 105 min
120 min
120 min
5
pH (-)

135 min
6.8 135 min
150 min
150 min
165 min 4 165 min
6.7 180 min 180 min
195 min 3 195 min
6.6 210 min 210 min
225 min 225 min
2
6.5 240 min 240 min

6.4 1

6.3 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Radius (mm) Radius (mm)
66

Addendum 5

The effect of biopolymer formation and degradation rates on the aerobic pH profile during a standard cycle

# 10-5 rate PHA degradation in time at core # 10-6 rate Poly-P formation in time at core
6 9
)

Cycle 1

Poly-P formation rate (mol timestep -1 )


-1

Cycle 1
Cycle 2 8 Cycle 2
PHA degradation rate (mol timestep

5 Cycle 3 Cycle 3
Cycle 4 Cycle 4
7

4 6

5
3
4

2 3

2
1
1

0 0
0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16
Time (hours) Time (hours)
# 10-5 rate Glycogen formation in time at core
1.8
)
-1

Cycle 1
Glycogen formation rate (mol timestep

1.6 Cycle 2
Cycle 3
1.4 Cycle 4

1.2

0.8

0.6

0.4

0.2

0
0 2 4 6 8 10 12 14 16
Time (hours)
67

Addendum 6

Steady State
After 10 standard cycles (dry weather conditions), no steady
state had established. Not for the biopolymers (PHA, Poly-P or
Glycogen), nor for the precipitates (Amorphous calcium phospahte
(ACP), hydroxyapatite (HAP), struvite (STR), ferrihydrite (FER)
and vivianite (VIV). This was due to neglecting newly formed
biomass and limitless precipitation.

PHA Poly-P
1.8 0.8

1.6 0.7
Poly-P concentration (mol L -1 )
Pha concentration (mol L -1 )

1.4
0.6

1.2
0.5
1
Cycle 1 0.4 Cycle 1
0.8 Cycle 2 Cycle 2
Cycle 3 Cycle 3
Cycle 4 0.3
Cycle 4
0.6 Cycle 5 Cycle 5
Cycle 6 0.2 Cycle 6
0.4 Cycle 7 Cycle 7
Cycle 8 Cycle 8
0.2 Cycle 9 0.1 Cycle 9
Cycle 10 Cycle 10
0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Radius (mm) Radius (mm)

Glycogen PO 4 end Feeding


0.85 # 10-3
9
Glycogen concentration (mol L -1 )

0.8 8
PO 4 concentration (mol L -1 )

0.75 7

6
0.7

Cycle 1 5 Cycle 1
0.65 Cycle 2 Cycle 2
Cycle 3 Cycle 3
Cycle 4 4 Cycle 4
0.6 Cycle 5 Cycle 5
Cycle 6 Cycle 6
3
Cycle 7
Cycle 7
0.55 Cycle 8
Cycle 8
Cycle 9 2 Cycle 9
Cycle 10
Cycle 10
0.5
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Radius (mm)
Radius (mm)
68

pH end Feeding # 10-3 ACP


7.25 6
Cycle 1 Cycle 1
Cycle 2 Cycle 2
7.2 Cycle 3 Cycle 3
5
Cycle 4
Cycle 4
Cycle 5
Cycle 5
Cycle 6
ACP content (mol)

7.15 Cycle 6
Cycle 7 4
Cycle 7
Cycle 8
Cycle 8
Cycle 9
pH (-)

Cycle 10 Cycle 9
7.1
3 Cycle 10

7.05
2

7
1

6.95
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0
Radius (mm) 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Radius (mm)

# 10-15 HAP # 10-20 VIV


1.8 3.5
Cycle 1 Cycle 1
1.6 Cycle 2 Cycle 2
3 Cycle 3
Cycle 3
Cycle 4 Cycle 4
1.4
Cycle 5 Cycle 5
2.5 Cycle 6
HAP content (mol)

Cycle 6
VIV content (mol)

1.2 Cycle 7
Cycle 7
Cycle 8 Cycle 8
1 2 Cycle 9
Cycle 9
Cycle 10 Cycle 10
0.8 1.5

0.6
1
0.4

0.5
0.2

0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Radius (mm) Radius (mm)
69

Steady State 20 standard cycles pH in Time at core


8.4
When the simulation time was increased to 20 standard cycles,
the ACP content at the core of the granule started to decrease 8.2

± 40 hours. This was caused by the decrease in pH at the core, as 8


a consequence of further penetration of acetate-. A higher acetate
7.8
penetration depth leads to a lower pH gradient inside the granule
towards the core. At ± 50 hours the acetate concentration reached 7.6

pH (-)
the granule’s core (Notice the logarithmic scale in this graph).
7.4
The acetate penetrated deep into the granule as biopolymers
7.2
depleted and no newly formed biomass is taken into account
in this model. When the acetate penetrated deeper, the Poly-P 7
was also released deeper into the granule. This decreased the
6.8
migration length of the H2PO4-, K+ and Mg2+ ions towards the
core of the granule. A shorter migration length decreased the 6.6
0 10 20 30 40 50 60 70 80
concentration difference between the core and the place of H2PO4- Time (hours)
release. Thus, a higher acetate penetration depth leaded to a lower
pH and less accumulation of phosphate at the core of the granule,
Acetate in Time at core
resulting in a decrease of ACP content in the granule’s core. 100
Although these simulations were done with slightly different
Acetate concentration (mol L -1 )

metal concentrations and no electron neutrality in the boundary


10-5
conditions, it is believed these results would not differ much from
the graphs shown here.
10-10

10-15

10-20

10-25
0 10 20 30 40 50 60 70 80
Time (hours)

# 10-3 ACP in Time at core


4.5 P in Time at core
0.01
4
0.009

3.5 0.008
P concentration (mol L -1 )
ACP content (mol)

3 0.007

2.5 0.006

0.005
2
0.004
1.5
0.003
1
0.002

0.5
0.001

0 0
0 10 20 30 40 50 60 70 80 0 10 20 30 40 50 60 70 80
Time (hours) Time (hours)
70

Addendum 7

Diffusion coeffients # 10-20 Multi-component diffusion coefficients


1.4
The sensitivity analysis of the diffusion coefficient of phosphate
(DPO4) showed that more precipitation occurred when the
1.2
phosphate diffusion was slower. The impact of the DPO4
MultiD ON
change, however, was rather small. The sensitivity analysis of
1
the incorporation of the multi-component diffusion coefficient
(MultiD) feature of PHREEQC showed that without MultiD,
VIV content
0.8
all the phosphate minerals precipitated less. The diffusion
coefficient of all components was 10-9 m2 s-1 when MultiD was
0.6
OFF. The diffusion coefficient of Ca2+, Fe2+/3+ and Mg2+ ions
were therefore increased (see Table 14, on page 64). These cation 0.4 MultiD OFF
concentrations were therefore lower at the core of the granule,
which limited the precipitation potential. 0.2

0
0 2 4 6 8 10 12 14 16
Sensitivity of diffusion coefficient Time (hours)
Change in maximum precipitated content (%)

0.02 0.02
Maximum precipitated STR content (mol)

# 10-16 Multi-component diffusion coefficients


0.015 0.015 8
ACP

0.01 0.01
7
0.005 0.005
STR 6
0 0 MultiD ON
VIV
HAP
5
HAP content

-0.005 -0.005

-0.01 -0.01
4

-0.015 -0.015
3
-0.02 -0.02

2
-0.025 -0.025
-100 -50 0 50 100
Change in diffusion coefficient (%) 1 MultiD OFF

0
# 10-3 Multi-component diffusion coefficients 0 2 4 6 8 10 12 14 16
1
Time (hours)
MultiD ON
0.9

0.8

0.7

0.6
ACP content

0.5

0.4

0.3

0.2

0.1
MultiD OFF
0
0 2 4 6 8 10 12 14 16
Time (hours)
71

Addendum 8 Bulk pH during different


weather conditions
8

7,8

7,6
Bulk pH (-)

7,4

7,2

6,8
:00 24:00
24:00 24:00
24:00 24:00
24:00 24:00
24:00 24:00
24:00 24:00
24:00 24:00
24
Time (hh : mm) Time (hh : mm)
Dry weather Rainy weather
Dry weather Rainy weather

Addendum 9 Poly-P formation rate-


effect on pH in aerobic
phase

# 10-6 rate Poly-P formation during Rain event


9
Poly-P formation rate (mol timestep -1 )

Cycle 1
8 Cycle 2
Cycle 3
7 Cycle 4

0
0 2 4 6 8 10 12 14 16
Time (hours)
72

Addendum 10 Size

Size effect on HAP content at core Size effect on PO 3- at core


10-14 # 10-3 4
Radius = 1.5 mm 9
Radius = 1 mm Radius = 1.5 mm
Radius = 0.8 mm Radius = 1 mm
8 Radius = 0.8 mm
Radius = 0.5 mm

- concentration (mol L -1 )
10-15 Radius = 0.3 mm Radius = 0.5 mm
7 Radius = 0.3 mm
HAP content (moles)

6
10-16
5

4
10-17
3
PO 33

2
10-18
4

10-19 0
0 2 4 6 8 10 12 0 2 4 6 8 10 12
Time (hours) Time (hours)

Size effect on VIV content at core


10-19
Radius = 1.5 mm
Radius = 1 mm
Radius = 0.8 mm
10-20 Radius = 0.5 mm
Radius = 0.3 mm
VIV content (moles)

10-21

10-22

10-23

10-24

10-25
0 2 4 6 8 10 12
Time (hours)
73

Addendum 11 Economics

Table 15 Economical assessment of dosing metals for precipitation

MW Metal price purity needed dosing needed dosing dosing chem volume Total Price Average

g/mol $/ton mgMe/l mgMe/l mmolMe/L mgChem/l m3/d kg/d $/d $/d

1 MgCl2.6H2O 203 1200 99% 50 40 2 328 1164 382 458 272

2 Mg 203 150 47% 50 40 2 497 1164 578 87

3 CaCl2 110 260 94% 300 240 6 700 1164 814 212 200

4 CaCl2 110 200 77% 300 240 6 812 1164 945 189

5 FeCl3 161 500 37% 15 12 0,2 56 1164 66 33 28

6 FeCl3 161 480 99% 15 12 0,2 35 1164 41 20

7 FeCl3 161 780 96% 15 12 0,2 36 1164 42 33

8 MgOH2 42 2900 100% 39810717 29810717 29810717 1254554217 1164 1460301108 4234873213

9 KOH 56 1000 99% 39810717 29810717 29810717,1 1686094157 1164 1962613598 1962613598

10 NaOH 40 390 99% 39810717 29810717 29810717,1 1204352969 1164 1401866856 546728074

Reference

1 http://www.alibaba.com/product-detail/-strong-Magnesium-strong-strong-chloride_60389591188.html?spm=a2700.7906341.35.1.7NhSn6

2 http://www.alibaba.com/product-detail/-strong-magnesium-strong-strong-chloride_60450031911.html?spm=a2700.7906341.35.1.7NhSn6&s=p

3 http://www.alibaba.com/product-detail/Professional-strength-94-pure-CaCl2-pellets_60349308892.html?spm=a2700.7906341.35.1.VKidfG&s=p

4 http://www.alibaba.com/product-detail/Professional-strength-94-pure-CaCl2-pellets_60349308892.html?spm=a2700.7906341.35.1.VKidfG&s=p

5 http://www.alibaba.com/product-detail/superior-quality-37-38-40-ferric_60452348006.html?spm=a2700.7906341.35.1.bxUAg8&s=p

6 http://www.alibaba.com/product-detail/Factory-price-Ferric-strong-chloride-strong_60441811219.html?s=p

7 http://www.alibaba.com/product-detail/-strong-Iron-strong-III-strong_1767584874.html?spm=a2700.7906341.35.1.bxUAg8&s=p

8 http://wholesaler.alibaba.com/product-detail/supply-high-purity-99-8-magnesium_60477071952.html?spm=a2700.7724838.0.0.5ielth

9 http://www.alibaba.com/product-detail/Potassium-Hydroxide_50028731338.html?spm=a2700.7724838.0.0.5ielth

10 http://www.alibaba.com/product-detail/2016-NAOH-purity-99-MIN-sodium_60223224559.html?spm=a2700.7724838.0.0.5ielth
74

Das könnte Ihnen auch gefallen