Professional Documents
Culture Documents
PhDThesis SamEggermont
PhDThesis SamEggermont
Electroprecipitation of metal
ions from aqueous
electrolytes
Kinetic study and process performance
Sam Eggermont
May 2020
Electroprecipitation of metal ions from aqueous
electrolytes
Kinetic study and process performance
Sam EGGERMONT
May 2020
© 2020 KU Leuven – Faculty of Engineering Science
Uitgegeven in eigen beheer, Sam Eggermont, Kasteelpark Arenberg 44 - box 2450, B-3001 Leuven (Belgium)
Alle rechten voorbehouden. Niets uit deze uitgave mag worden vermenigvuldigd en/of openbaar gemaakt worden
door middel van druk, fotokopie, microfilm, elektronisch of op welke andere wijze ook zonder voorafgaande
schriftelijke toestemming van de uitgever.
All rights reserved. No part of the publication may be reproduced in any form by print, photoprint, microfilm,
electronic or any other means without written permission from the publisher.
Preface
The accomplishments described in this PhD thesis were not mine alone. Here, I
would like to take the opportunity to acknowledge the people that guided and
helped me throughout the last 5 years.
I thank Jan Fransaer and Xochitl Dominguez Benetton, for their financial
support and mentoring. Endless discussions about (electro)chemistry, and
diving deep into the mechanism of GDEx, helped to make this PhD thesis a
success. Rafael Prato, for being my closest colleague, although we were miles
apart thanks to VITO’s extremely strategic location. You were my partner-
in-crime while opening Pandora’s box of the GDEx process. Talking to you
about our common struggles in understanding GDEx always brought us closer
to the answers. Guillermo Pozo and Omar Martinez, for being my other team
members in the GDEx squad. Our monthly meetings would not have been the
same without you.
I would like to thank the members of the examination committee for investing
time and effort to read and comment on my manuscript. I hope you are satisfied
with how I addressed your remarks in this final version.
Special thanks to all of the administrative and technical support personnel at
MTM, in particular, Huberte Cloosen for her endless enthusiasm and patience,
and for being the rock-solid foundation of our department. Tom Van der
Donck, for the many excellent SEM photographs and for training me in the skill
of investigating the microscopic world. Joop van Deursen deserves a special
acknowledgment for assisting me with the many technical practicalities that I
struggled with, as well as for building my electrochemical flow-cell reactor from
scratch using the designs that came from hours of interesting discussions.
I would like to pay tribute to all of my colleagues, friends, and family that were
there when I needed support. Gijs and Nicolò, for being there in the beginning
and for becoming friends rather than colleagues over the years. Isadora, for
being my PhD-mate, going through the same struggles at the same time, and
i
ii PREFACE
helping one another when all hope seemed lost. Of course, I cannot thank
her enough for our coffee moments. Liuba, Basit, Xuan, Kai, Pieter, Palan,
Samy, Joao, among others, for being part of my journey and lighting up my
days. Senne, for cooperating along the way, and especially at the end, when you
were very busy yourself, finishing your PhD thesis. You inspired me to remain
interested in the mathematics that were necessary in the search for answers to
questions of the GDEx process.
MTM would not be the same without its yearly departmental trip. The
Departmental Trip Organizing Committee (DTOC) was by far my most favorite
committee to have been part of during these five years. I thank all of the current
and previous members to help make the departmental trip a yearly success.
I would like to thank my West-Flemish friends that, at some point or another,
co-migrated to Leuven: Mathieu, Karel, Lukas, Thomas, and Xander. It is an
absolute privilege to call them (and their partners) my friends. They made
my years in Leuven both unforgettable and sometimes somewhat forgettable.
Whatever the future brings, I will always be grateful for those moments. I will
not forget.
I thank my mom, my dad, Tom and Annabel, and Lies, for making Annabelle
and me feel at home after our (re)migration to Bruges, as well as for any other
support during the last five years and before.
Most importantly I would like to thank my fiancee Annabelle. She was at
my side during the entire journey. Being young (and somewhat lost) together
has been extremely interesting, educational and inspiring. She helped me to
rediscover the importance of family and taught me so many things about myself.
I love her with all my heart, and I am grateful for everything we have been able
to share.
Thank you.
Sincerely,
Sam
PREFACE iii
Acknowledgment
This research was funded by the Strategic Initiative Materials (SIM) in Flanders,
within the recyclable materials MaRes program, under grant agreement n◦
150626 (Get-A-Met project). Jan Fransaer and Xochitl Dominguez-Benetton
acknowledge funding from the European Union’s Horizon 2020 research and
innovation program under grant agreement n◦ 654100 (CHPM2030 project).
Abstract
The purpose of this work was to discover the fundamental aspects of gas-diffusion
electrocrystallization (GDEx). GDEx is defined as the reactive precipitation
of intermediaries issued from an electrochemical reaction at a gas-diffusion
electrode (GDE), with metal precursors provided in the electrolyte. It can be
used as a synthesis method to prepare (nano)particles or as a metal removal
and recovery process. An exploratory study of the GDEx process was carried
out to demonstrate the similarities and differences of GDEx with alkaline
titration, a well-known precipitation method. The motivation to investigate
GDEx came from preliminary experiments at VITO which demonstrated the
potential of GDEx for producing crystalline metal oxide (nano)particles from
metal-containing waste streams, even for low metal ion concentrations. The
combination of synthesis and metal recovery is desirable because society relies
increasingly on the development of new functional materials preferably from
recyclable sources. In light of this, this work focused on understanding the GDEx
process with regards to how metal ions are recovered, and how and why they
precipitate as nanoparticles. The focus was on air-fed GDEs for the formation
of metal (oxy)(hydr)oxide nanoparticles. In summary, the key contributions of
this PhD were: (1) demonstrating that although it is an electrochemical process
and therefore inherently heterogeneous, GDEx can have the appearance of a
homogeneous process, (2) the comparison of GDEx to alkaline titration, (3)
the discovery of oxidation-assisted alkaline precipitation and its role in GDEx
as a process to synthesize metal (oxy)(hydr)oxide nanoparticles, and (4) the
investigation of the accumulation of species in the porous GDE, a phenomenon
observable during the GDEx process in batch mode. Future work includes the
development of an all-encompassing model to reach the goal of GDEx as a
process for synthesis-by-design.
v
Beknopte samenvatting
vii
Abbreviations
AM analytical model
AP alkaline precipitation
AT (traditional) alkaline titration
CSTER continuously stirred tank electrochemical reactor
CSTR continuously stirred tank reactor
Cu-ISE Cu2+ -ion selective electrode
CNT classical nucleation theory
DLS dynamic light scattering
GDE gas-diffusion electrode
GDEx gas-diffusion electrocrystallization
Get-A-Met Groundbreaking Extraction Technology for critical Metals
and metalloids from industrial wastewaters
(fundamental research project within the SIM MaRes Programme)
MaRes Materials from solid and liquid industrial process Residues
NM numerical model
NP nanoparticle
ORR oxygen reduction reaction
Ox-AP oxidation-assisted alkaline precipitation
PTC precipitation titration curve
PTFE polytetrafluoroethylene
VITO Vlaamse Instelling voor Technologisch Onderzoek
(Flemish Institute for Technological Research)
SEM scanning electron microscopy
SIM Strategic Initiative Materials
sp ICP-MS single-particle inductively coupled plasma mass spectrometry
VLAIO Vlaams Agentschap Innoveren en Ondernemen
XRD X-ray diffraction
ix
List of symbols
Latin letters
xi
xii LIST OF SYMBOLS
t [s] time
Vj [m3 ] volume with specifier j
wGDE [m] penetration depth of the GDE
x [m] x-coordinate
Greek letters
α [-] fitting model correction factor
γi [-] activity coefficient of species i
δ [m] diffusion layer thickness
[-] hydrophilic porosity of the GDE
ηi [-] production efficiency of species i
ψ [-] porous media coverage of the GDE
τi [s] characteristic diffusion time in the GDE of species i
0 initial
app applied
b bulk
dl diffusion layer
f final
GDE gas-diffusion electrode
geo geometric
het heterogeneous
hom homogeneous
i of species i
phil hydrophilic porosity
phob hydrophobic porosity
q with regards to flow rate q
ss steady-state
tot total
Contents
Abstract v
Abbreviations ix
List of symbols xi
Contents xiii
1 Introduction 1
1 Problem statement . . . . . . . . . . . . . . . . . . . . . . . . . 2
2 Aim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3 Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
xiii
xiv CONTENTS
Bibliography 159
Introduction
This PhD thesis is the result of a close collaboration between KU Leuven and
VITO within the framework of the Get-A-Met project (funded by SIM, VLAIO,
within the MaRes program). The goal of this PhD was to obtain a fundamental
understanding of Gas-Diffusion Electrocrystallization (GDEx), as a new process
consisting of the electrochemically-driven reactive precipitation of metals or
metalloids. The PhD thesis aims to demonstrate the results of the investigation
of GDEx and to provide the reader with a better understanding of the process
in general, and more specifically of the fundamental principles that guide it.
This introductory chapter defines the research questions and how the problem
statement of this PhD thesis originated (section 1), it states the specific aims of
the PhD within the framework of the Get-A-Met Project (section 2), and the
outline section clarifies the structure of the PhD thesis (section 3).
1
2 INTRODUCTION
1 Problem statement
Within the framework of the Get-A-Met project, this PhD thesis focused more
on the fundamental aspects of the process. Therefore, the actual essay of these
expectations with the current best-practices was excluded from the scope of
this work.
4 INTRODUCTION
2 Aim
3. discover how selective and how universal the GDEx process is;
4. identify the role of the GDEx process regarding the oxidation of the metal
ions in the catholyte;
5. develop mathematical models that can help to understand specific sub-
processes of the GDEx process or the GDEx process as a whole;
Answers to all of these questions, and more, were sought during work done for
this PhD. The scope of this PhD was limited to metal removal and nanoparticle
formation in aqueous solutions. Concerning the formation of nanoparticles, the
focus was on metal (oxy)(hydr)oxide nanoparticles. Moreover, this work only
considers the GDEx process conducted with the GDE as an air-fed cathode in
aqueous electrolytes, hereinafter simply referred to as the GDEx process.
OUTLINE 5
3 Outline
This PhD thesis consists of eight chapters. After this first chapter, Chapter 2
contains a literature review relevant to the study of GDEx and situates GDEx
among the relevant metal removal processes and nanoparticle synthesis methods.
Chapter 3 provides an exploratory study of the GDEx process and the different
key parameters. In several instances, GDEx is assessed via a comparison with
traditional alkaline titration (AT), a method known for being used as a metal
removal process and a nanoparticle synthesis method. The oxygen reduction
reaction (ORR) and the production of H2 O2 are discussed and their effect
on the redox potential of the catholyte solution. In addition, it demonstrates
the potential and limitations of the GDEx process in terms of selectivity and
universality. It also elaborates on the power consumption of GDEx.
Chapter 4 contains the comparison of modeled and experimental titration
curves for the traditional alkaline precipitation of Zn(OH)2 . The modeled and
experimental titration curves show good agreement and demonstrate that AT
of ZnCl2 solutions for AP of Zn(OH)2 can serve as a good benchmark for other
metal removal or recovery processes, including GDEx. Specifically in terms of
the pH evolution during the precipitation process.
Chapter 5 contains the experimental results of controlled experiments to
demonstrate the role of H2 O2 in the formation process of metal (oxy)(hydr)oxide
nanoparticles and the effect of H2 O2 on the size of these nanoparticles. It
introduces the concept of oxidation-assisted alkaline precipitation (Ox-AP).
Chapter 6 compares experimental results for the AP process and the GDEx
process of solutions with decreasing oxidation potential (i.e., ZnCl2 , MnCl2 , and
FeCl2 ). It confirms the initial discrepancy of AT and GDEx introduced in the
second chapter. It also confirms AT of ZnCl2 as a good reference process and
how solutions of MnCl2 and FeCl2 deviate from this. In addition, it provides a
practical example of Ox-AP, because GDEx of FeCl2 produces very small Fe3 O4
nanoparticles due to GDEx-elicited oxidation-assisted alkaline precipitation, in
contrast to the much larger FeO nanoparticles produced by AT-elicited alkaline
precipitation.
Chapter 7 provides an answer to why a discrepancy is observed between AT and
GDEx. It contains an analytical model that demonstrates the accumulation
of OH− and HO− 2 within the pores of the GDE and connects these results to
experimental data with the GDE electrochemical flow-cell reactor.
Chapter 8 contains the general conclusion of this work. It also elaborates on
open questions and opportunities for future work.
Chapter 2
7
8 GENERAL INTRODUCTION TO THE GDEX PROCESS
also shortly situates GDEx, as a method that combines metal removal with
nanoparticle synthesis, among other comparable methods. To avoid repetition,
several paragraphs of this chapter will refer to later chapters which discuss the
introduced concepts in more detail, related to the (experimental) research there
discussed.
1 Electrochemical synthesis of
metal (oxy)(hydr)oxides
A U.S. patent from 1904 by Hunt [4] was the oldest record on electrochemical
synthesis of metal (oxy)(hydr)oxides found in this literature investigation. Since
then, a large body of scientific literature on the electrochemical synthesis of
metal (oxy)(hydr)oxides has developed, most of it at least partially rely on the
cathodic electrogeneration of a base [5–10].
This scientific literature study does not pretend to be a complete review of the
topic. As a rule, the cited references in the following paragraphs are the ones
that were decided to be the most relevant or complete sources for the reader
to further investigate the topic of interest. The study is limited to aqueous
solutions, although metal oxides can be produced in non-aqueous solutions as
well [11–17].
In this section, the precipitation mechanisms of metal (oxy)(hydr)oxides are
briefly discussed, followed by a discussion on electrochemical synthesis of metal
(oxy)(hydr)oxides, relying on the cathodic electrogeneration of base.
more depth in Chapter 5 (p. 81) with appropriate examples. Especially the
book of Jolivet et al. [19] deserves to be mentioned, for discussing the formation
of metal (oxy)(hydr)oxides in depth and width. In addition, Chapter 4 discusses
the literature and results on how to follow the chemical formation of sparingly
soluble Zn(OH)2 (phase 1), by modeling or measuring the pH.
Especially, diagrams representing electrochemical equilibria in aqueous solutions
(e.g., Pourbaix diagrams, E-pH diagrams) are useful tools in aqueous
(electro)chemistry, to determine the thermodynamic feasibility of certain phases
[20]. They indicate, for certain values of the redox potential and the pH, which
are the stable phases, and referring to a 1 M concentration of the targeted
element they are referred to as Pourbaix diagrams [20]; however, they can
be calculated for any pressure and temperature. Figure 2.1 demonstrates
a simplified E-pH diagram for a system with 0.001 M Zn and water, at 1
atmosphere and 25 ◦ C. More sophisticated diagrams can be calculated from
thermodynamic data, containing zones of multi-species ions and taking into the
anion and supporting electrolyte. The dominating phase of a species depends
on the pH and redox potential and can vary by sufficiently changing either
parameter.
-2 0 2 4 6 8 10 12 14 16
2.2 2.2
1.8 1.8
1.4 1.4
1 1
E vs. SHE [V]
0.6 0.6
0.2 Zn2+ 0.2
Zn(OH)2
-0.2 -0.2
-0.6 -0.6
-1 -1
-1.4 Zn -1.4
-1.8 -1.8
-2 0 2 4 6 8 10 12 14 16
pH
2H2 O + 2e−
H2 + 2OH− (E◦ = −0.828 V) (2.1)
NO− − −
3 + H2 O + 2e
NO2 + 2OH
−
(E◦ = 0.01 V) (2.2)
Except for one 1978 U.S. patent [1], one 2015 U.S. patent [2] and a recent article
(2017, published after the initial literature investigation for this thesis) [3], no
other literature was discovered where an air-cathode (gas-diffusion electrode) is
suggested to carry out oxygen reduction reactions (see equations (7.3), (7.4),
(7.1) and (7.2) p. 108) for the electrogeneration of a base to synthesize metal
(oxy)(hydr)oxides. Although the methods used in these publications share similar
characteristics with the GDEx process (i.e., use of a gas-diffusion electrode
passing air, precipitating metal (oxy)(hydr)oxides), they are not carried out
GAS-DIFFUSION ELECTRODES AND THE OXYGEN REDUCTION REACTIONS 11
in a two-compartment cell [2, 3], or are carried out at much lower current
densities [3], or are using a sacrificial anode [1, 2]. Moreover, none of them
discuss the cogeneration of hydrogen peroxide during the process. For the cases
using sacrificial anodes [1, 2], the steady-state concentration of the metal ions
in solution was not specifically mentioned. However, this is expected to be
low, because the amount of metal from the anode that is dissolving as metal
ions would be countered with a corresponding amount of hydroxyl ions. The
resulting precipitation reaction would keep the overall metal ion concentration
low. In the other case [3] the concentration was in the 5 mM to 10 mM range.
®
designs are possible and the design depends on how single fuel cells are stacked
together (e.g., monopolar [55, 56] or bipolar [57]). The VITO CoRE GDE
12 GENERAL INTRODUCTION TO THE GDEX PROCESS
design will be discussed and most closely resembles the designs of monopolar
AFC GDEs.
®
In the VITO CoRE GDE, the three-phase boundary between electrode,
electrolyte, and gas is provided using a multi-layer design. The first layer (1) is
a porous hydrophobic layer. This gas-diffusion layer (GDL), forms the barrier
between the gas compartment and the electrolyte compartment. It consists
of porous PTFE and allows gas diffusion but resists liquid water to pass to
the gas compartment. The second layer (2) is partially hydrophobic, partially
hydrophilic. This carbon/PTFE layer (C/PTFE), is a layer made of a mixture
of hydrophilic porous activated carbon particles and hydrophobic PTFE, and a
catalyst (optional). It is a hybrid material consisting of two networks, with a
very large interface: a network of hydrophobic PTFE channels and hydrophilic
porous carbon particles. Alvarez-Gallego et al. determined the overall porosity
of an 80/20 C/PTFE layer to be 67 %, and the hydrophilic and hydrophobic
porosity to be 29 % and 38 %, respectively [58]. The intrinsic BET surface area
of the active carbon is 745 m2 g−1 , however, the overall BET surface area in the
80/20 C/PTFE layer was determined to be 454 m2 g−1 [58]. The porous carbon
network is penetrated with electrolyte. The three-phase boundary emerges
at the interface with the hydrophobic PTFE channels and the porous carbon
network. The carbon network is provided with electrons via a metal-mesh
current collector (3), in this case, 316L stainless steel. Figure 2.2 shows a
schematic of a GDE and the hydrophilic and hydrophobic networks.
gas
electrolyte
porous carbon particle network
binder network
hydrophobic GDL
active layer
current collector
electrolyte
gas
Figure 2.2: Schematic of GDE with inset showing hydrophilic (electrolyte) and
hydrophobic (gas) channels.
Although different gases can be used to feed the GDEx process, in this thesis,
especially the oxygen reduction reactions were of interest. The low availability
of oxygen at the electrode surface, as a result of the low oxygen solubility
in water, is often a limiting factor in the electrochemistry of oxygen [60–62].
However, this limitation is resolved by the design of the internal structure of
the GDE (section 2.1). A large amount of oxygen reduction reactions exist and
are discussed in detail in several reviews and books [63–65]. In general, the
two overall pathways resulting in the most stable products are the four-electron
pathway and the two-electron pathway (written for alkaline media) [63]:
O2 + 2H2 O + 4e−
4OH− (E◦ = 0.401 V) (2.3)
O2 + H2 O + 2e−
HO−
2 + OH
−
(E◦ = −0.065 V) (2.4)
The oxygen reduction reactions are useful for the electrogeneration of a base.
Moreover, the two-electron pathway yields hydrogen peroxide (deprotonated
14 GENERAL INTRODUCTION TO THE GDEX PROCESS
in alkaline media). The key to answer one of the research questions related to
the oxidation of metal ions in the cathode lies in the formation of hydrogen
peroxide in the GDE as a result of the oxygen reduction reactions.
Besides the reduction of oxygen from pure oxygen, air or other mixtures, the
reduction of other gasses is widely researched. Especially N2 and CO2 deserve
special attention due to their potential to help reach some of the sustainable
development goals [66]. The electrochemical reduction of N2 to ammonia [67–72],
or the reduction of CO2 to urea in nitrate/nitrite solutions [73–79] could be the
key for sustainable production of artificial fertilizers for food production, while
the reduction of CO2 could help to reduce the concentration of this greenhouse
gas in the atmosphere and provide a solution for the sustainable production of
green carbon-based fuels [80–86].
REACTIVE PRECIPITATION: COMBINING METAL REMOVAL AND NANOPARTICLE SYNTHESIS 15
This chapter is an exploratory study of the GDEx process. The most important
process characteristics of the GDEx process are introduced. We highlight the
key process parameters, and by exploring the process, research questions 1, 2,
and 5 posed in Chapter 1 are answered. After this chapter, the reader should
have a clear overview of what the GDEx process is, and why the following
chapters help to answer the other research questions. The first section discusses
the GDEx process, and how it relates to the alkaline titration process (AT)
and the key process parameters. The second section explores the contradictory
homogeneous appearance of the GDEx process. The third section shows how
GDEx, the oxygen reduction reaction, hydrogen peroxide, and the solution
redox potential are connected. The fourth section discusses the universality and
selectivity of GDEx as a metal removal/recovery process. In the last section,
the energy considerations of the GDEx process are investigated.
For the remainder of this thesis, we redefine GDEx more specifically as the
reactive precipitation of metal ions in the catholyte with OH− , and possibly
H2 O2 or HO− 2 issued from the oxygen reduction reaction (ORR) at an air-fed
GDE. In contrast, AT can be defined as the reactive precipitation of metal
17
18 EXPLORATORY STUDY OF THE GDEX PROCESS
ions in the electrolyte with OH− issued from a titration with NaOH. Figure
3.1 demonstrates schematic drawings of the GDEx and AT processes for the
metal hydroxide precipitation case. It provides a rudimentary summary of the
expected precipitation mechanism for each process.
e-
OH- OH-
HO2-
HO2 - Mn+
O2 O HO2- H +
Mn+
- OH-
2 -
OH HO M n+
H +
O2 e- OH- OH-
2
O2 H + H+
H2O OH -
Mn+
n*>n
Alkaline Titration (AT)
M(OH)n H2O M(OH)n
titrator tip bulk H2O precipitates
M(OH)n
OH- + OH-
Na Mn+
Na+ H+ Mn+
Na+OH- OH- M H+
n+
OH- + Na+ H+
Na H+
OH- Mn+
Ar
magnetic
stirrer bar
magnetic stirrer
gas
b
anion exchange membrane
Vito Core cathode Pt anode
outlet
gas channel
Front-view cross-section of
catholyte
anolyte
inlet
1 cm
c Ar
titrator tip
pH sensor
NaOH
Figure 3.2: Schematic drawing of the GDEx flow-cell reactor set-up (a) and
cross-sections of the flow-cell reactor (b), and the AT reactor set-up (c).
20 EXPLORATORY STUDY OF THE GDEX PROCESS
The anode was a platinized tantalum plate (with a 10 µm Pt layer plated with
®
the molten salt process by Umicore Galvano), further referred to as the Pt
electrode. In general, the cathode was a VITO CoRE GDE [58], but some
experiments were carried out with a Pt electrode as both the anode and the
cathode. Generally, when the Pt electrode is mentioned, it will be in the capacity
of the cathode, because in this PhD, the anode and anolyte compartment only
served to provide the counter-reactions for the requested currents. Otherwise, a
counter electrode will be simply addressed.
Figure 3.2b shows the electrochemical flow-cell reactor. The two compartments
®
(i.e., the anolyte and catholyte compartments) were separated with an anion-
exchange membrane (FUMASEP FAP-4130-PK). Each compartment had
a volume of 10−5 m3 (10 mL). The compartments had a cylindrical shape,
with a length h of 0.01 m and a base surface area of 0.001 m2 (10 cm2 ). The
cylinder axis was perpendicular to the electrode surfaces, giving the electrodes
(both anode and cathode) a geometrical surface area of 0.001 m2 , equal to
the cylinder base area. Each compartment had radial in- and outflow, against
®
gravity, through in - and outlets with diameters of 0.002 m. A peristaltic pump
(Cole- Palmer MasterflexTM L/s 7551-00 with Easyload II 77200-60 pump
heads) was used to provide the flow rate.
The GDEx process was operated as a batch process. Eventually, if GDEx would
become an industrial process, it is more likely that it would be carried out
as a continuous process rather than a batch process. However, at this stage
research, time-dependent data collection is easier for a batch process. Moreover,
it allows to focus on thermodynamic aspects without these being obscured with
kinetic phenomena or noise related to a fast continuous process. Operating
GDEx as a batch process means that the anolyte and catholyte compartments
of the flow cell reactor connect to the anolyte and catholyte reservoirs outside
the reactor, respectively. The electrolyte was pumped from the reservoir to
the flow cell compartment and back. The four connection tubes (Cole-Palmer
MasterflexTM VersilonTM L/s 06475-16) each had a diameter of 0.0032 m and a
length of 0.9 m. The anolyte and catholyte reservoirs were 250 mL DURAN ®
bottles with three-port screw caps. The bottles were adapted with additional
screw-cap necks in an in-house glass-blowing workshop. The additional necks
were needed to accommodate measuring probes (e.g., pH probe, redox probe)
and for sampling material.
When GDEx is operated as a batch process, the catholyte reservoir can be
considered to be a batch reactor, rather than being the recirculation reservoir
of a flow-cell reactor. This is useful because it makes it easier to benchmark
GDEx against other batch reactor processes (e.g., alkaline titration). To operate
GDEx as a batch reactor, the default flow rate was chosen equal to 3.33 10−5
m3 s−1 (200 mL min−1 ). This flow rate was carefully chosen to be high enough
GDEX PROCESS, AT PROCESS AND KEY PROCESS PARAMETERS 21
so the GDEx reactor could be considered as a batch reactor, while also avoiding
to create turbulence in the connection tubes (Re = 1326). At this flow rate, the
combined residence time in the flow cell compartment (3 s), the inlet tube (1
s), and the outlet tube (1 s) was 5 seconds, which is the time-resolution of the
pH measurements. As such, from the eye of the observer (the pH meter), the
species produced heterogeneously by the electrochemical reaction at the surface
of the cathode might as well have been produced or added homogeneously in
the bulk of the catholyte reservoir.
In this work, the GDEx process was by default carried out galavanostatically.
The default cathodically applied current at the GDE was 100 mA (100 A m−2 )
provided with a potentiostat (Biologic Biostat SP-300). In addition, the focus
was on the GDEx process using an air-fed GDE. The default airflow at the back
of the electrode was chosen to be 100 mL min−1 . A back pressure of 30 mbar
was established by putting the end of the gas outlet tube at the bottom of a 30
cm water column. If the process parameters deviated from these default values,
it is pertinently clarified within this work.
The AT process. The AT process could be summarized as follows: during
AT, a concentrated solution of NaOH is slowly added to the electrolyte; the
OH− ions move into the solution and react with the available metal ions to
form sparingly soluble metal (oxy)(hydr)oxides (or metal hydroxysalts); in some
cases, these sparingly soluble species precipitate as nanoparticles (figure 3.1).
®
The AT process was carried out in one 250 mL DURAN bottle (equivalent
®
to the ones used for the GDEx reservoirs) with a titrator (Methrohm 702SM
Titrino). By default, the titrant was a 0.1 M NaOH solution (Titrisol , Merck).
The default titration rate was 0.62 mL min−1 , the equivalent of 1.036 10−6 mol
NaOH s−1 , which is the theoretical addition rate of OH− in the default GDEx
process at 100 A m−2 using a GDE fed with air.
Default electrolyte solutions. The volume of electrolyte solutions used
during GDEx and AT was 100 mL by default. Two solutions were used most
commonly: (1) a default blank solution without metal ions (except for Na+ ),
hereinafter referred to as solutions without metals, and (2) a default solution
with metal ions (other than Na+ ), further referred to as solutions with metals.
Except if specified otherwise, all solutions were made with a supporting
electrolyte of 0.5 M NaCl and experiment-dependent HCl concentrations (most
often 0.001 M or 0.01 M) to establish the initial pH (e.g., pH 3 or pH 2,
respectively).
The composition of the solutions without metals was a 0.5 M NaCl supporting
electrolyte with HCl. Generally, these solutions were made in 1 L volumetric
flasks. 100 mL of a 5.00 M NaCl stock solution together with the chosen amount
22 EXPLORATORY STUDY OF THE GDEX PROCESS
Investigating the GDEx process in more detail was possible by having established
these default process parameters. Despite its operational simplicity, GDEx is
a complex process and additional parameters (not investigated here) can be
manipulated, such as the gas partial pressure (e.g., amount of O2 in an O2 -N2
mixture) and flow rate, and the size and shape of the flow-reactor, among
others. In the next section AT and the batch GDEx process are demonstrated
to indeed have many similarities, both as metal ion removal/recovery processes
and nanoparticle synthesis processes.
HOMOGENEOUS APPEARANCE OF THE GDEX PROCESS 23
The first aim of this thesis was to discover the homogeneous and heterogeneous
processes of GDEx. Since GDEx is an electrochemical process, it is inherently
governed by heterogeneous electrochemical reactions at the surface of the GDE.
However, the precipitation reactions can happen at the surface of the electrode
(heterogeneous reaction) or in the bulk electrolyte solution (homogeneous
reaction). Here, model and experiment are compared to investigate if these
precipitation reactions are heterogeneous or homogeneous.
In a homogeneous process, the desirable products result from homogeneous
reactions (i.e., everywhere) in the solution. In a heterogeneous process, the
desirable products result from heterogeneous reactions at a surface (i.e., in this
case, the GDE surface). This section demonstrates a simple time-dependent
model of the concentration in the solution for both a heterogeneous and a
homogeneous process. Then, the data of the time-dependent Cu2+ concentration
and the pH during AT and GDEx processes of a CuSO4 electrolyte are compared,
to investigate the nature of the precipitation reactions in GDEx.
Consider 3 species: Cu2+ , OH− , and Cu(OH)2 . The chemical reaction between
the three species can be written as:
Cu2+ + 2OH−
Cu (OH)2 (3.1)
When all species are in equilibrium, their chemical activity is related to the
equilibrium constant constant KCu(OH)2 [mol3 L−3 ] of Cu(OH)2 :
2 2
KCu(OH)2 = aCu2+ (aOH− ) = γCu2+ cCu2+ (γOH− cOH− ) (3.2)
with aj [-] the activity of species j, cj [-] the concentration of species j, γj [-]
the activity coefficient of species j. For activity coefficients equal to 1, equation
(3.2) simplifies to:
2
KCu(OH)2 = cCu2+ (cOH− ) (3.3)
Assuming equilibrium, every time two OH− ions are added to the solution,
these would quickly react with a Cu2+ and these three ions would be removed
from solution in the form of Cu (OH)2 .
Because electrochemical reactions are inherently heterogeneous, the real question
is whether the precipitation reactions happen at (or in the vicinity of) the surface
of the electrode (heterogeneous process), or in the bulk of the electrolyte solution
(homogeneous process). The Cu2+ concentration in the bulk electrolyte was
determined experimentally and the data was compared to models for each of the
possible scenarios (i.e., homogeneous or heterogeneous precipitation reactions).
24 EXPLORATORY STUDY OF THE GDEX PROCESS
Figure 3.3 demonstrates the boundary conditions for both the homogeneous
and heterogeneous scenarios. It is necessary to look at what happens in the
vicinity of the electrode (model domain 1, figure 3.3), to be able to model the
time-dependent Cu2+ concentration in the bulk electrolyte (model domain 2,
figure 3.3). Considering a well-mixed batch reactor, it is possible to define a
diffusion layer of average thickness δ [m] at the surface of the electrode. At this
distance δ away from the electrode surface, the concentration of the species in
solution could be considered equal to that of the bulk electrolyte. Between the
surface of the electrode and the distance δ, the concentration can be calculated
from the time-dependent constitutive equation for mass-transfer. Assuming that
no reactions occur within the diffusion layer (i.e., either the chemical reaction
happens exactly at the surface, for a heterogeneous process, or the chemical
reaction happens outside of the diffusion layer, for a homogeneous process), the
equation is written as:
∂cdl
j (x, t) ∂ 2 cdl
j (x, t)
− Dj =0 (3.4)
∂t ∂x2
−3
with cdl
j (x, t) [mol m ] the concentration of species j in the diffusion layer, Dj
2 −1
[m s ] the diffusion coefficient of species j, and t [s] and x [m] the variables
of time and space, respectively. As a further simplification, it is possible to
assume that within this layer, a steady-state is quickly arrived at compared to
the time-dependent changes in the bulk:
∂cdl
j (x, t)
≈0 (3.5)
∂t
A time-independent concentration profile is then easily found from equation
(3.4), which is assumed to be valid at any instant of time. Setting the time-
derivative to zero in equation (3.4) then results in:
d2 cdl
j (x)
=0 (3.6)
dx2
with a general solution of the form:
cdl
j (x) = C1 x + C2 (3.7)
with C1 and C2 integration constants that follow from the boundary conditions.
The boundary conditions depend on whether a homogeneous or a heterogeneous
process is considered. This is discussed in the following paragraphs and
summarized in figure 3.3.
HOMOGENEOUS APPEARANCE OF THE GDEX PROCESS 25
𝑏
𝑑𝑙
d2 𝑐𝐶𝑢 2+ 𝑥 d𝑐𝐶𝑢 2+ 𝑡 d𝑐 𝑑𝑙 2+ 𝑥
=0 𝑉 = −𝐴𝑔𝑒𝑜 𝐷𝐶𝑢2+ 𝐶𝑢
d𝑥 2 d𝑡 d𝑥 𝑥=
heterogeneous 𝑥 = 0 𝑥=
homogeneous 𝑏 −
d𝑐𝑂𝐻 𝑡 𝑑𝑙 −
d𝑐𝑂𝐻 𝑥
𝑑𝑙 −
d2 𝑐𝑂𝐻 𝑥 𝑉 = 𝐴𝑔𝑒𝑜 𝐷𝑂𝐻− − 2𝑅𝐶𝑢(𝑂𝐻)2
=0 d𝑡 d𝑥 𝑥=
𝑏
d𝑥 2 d𝑐 2+ 𝑡
𝑉 𝐶𝑢 = −𝑅𝐶𝑢(𝑂𝐻)2
𝑥=0 𝑥= d𝑡
𝑑𝑙 −
d𝑐𝑂𝐻 𝑥 𝑂𝐻 − 𝐼𝑎𝑝𝑝
=−
𝑑𝑙 −
𝑐𝑂𝐻 ~𝑐𝑂𝐻
𝑏 −
𝑡
d𝑥 𝐴𝑔𝑒𝑜 𝐷𝑂𝐻 − 𝐹
𝑥=0
cdl
Cu2+ (0) = 0 (3.10)
cdl b
Cu2+ (δ) = cCu2+ (t) (3.11)
This will be the case for thin boundary layers (and for relatively large
bulk volumes, because then the concentration changes more slowly). Using
equations (3.10) and (3.11), the quasi-steady state approximation of the Cu2+
concentration profile at the electrode surface in the heterogeneous case can thus
be written as follows:
x
cdl b
Cu2+ (x) = cCu2+ (t) δ (3.13)
To find the explicit expression of cbCu2+ (t), it is necessary to solve the mass
balance over the bulk solution (model domain 2):
Using equation (3.14) and (3.15), the mass balance equation can be written as
a first-order ordinary differential equation:
more easily defined. The boundary condition at the surface of the electrode
(x = 0) is now a Neumann boundary condition instead of a Dirichlet boundary
condition, while the boundary condition at x = δ remains a Dirichlet boundary
condition.
The boundary condition at the electrode surface is calculated from the
production rate of OH− at the surface of the electrode, POH− [mol s−1 ], given
by:
η − Iapp
POH− = OH (3.18)
F
with ηOH− [-] the efficiency for OH− production, Iapp the applied current [A], and
F = 96, 485 mol C−1 the Faraday constant. The model defines the interaction
between the bulk and the electrode. Because of this only the geometric surface
area of the electrode is of interest, rather than the effective wetted surface
of the electrode. Using the geometric surface area of the electrode Ageo , it is
possible to define the flux of OH− at the surface as R/Ageo and set it equal to
dc − (0)
the flux of OH− caused by the concentration gradient −DOH− OH dx at the
surface. From this follows the Neumann boundary condition at the surface of
the electrode:
dcdl
OH−
(x) POH−
=− (3.19)
dx
x=0 DOH− Ageo
cdl b
OH− (δ) = cOH− (t) (3.20)
Using equation (3.24) and (3.25), the mass balance equation can be written as
an ordinary differential equation:
dcbOH− (t)
Vtot = POH− − 2RCu(OH)2 (3.26)
dt
A similar mass balance can be written for Cu2+ :
dcbCu2+ (t)
Vtot = −RCu(OH)2 (3.27)
dt
dcbCu2+ (t)
From equations (3.26) and (3.27), the following relationship between dt
dcbOH− (t)
and dt is obvious:
5
𝐴𝑔𝑒𝑜 𝐷Cu2+
− 𝑡
4.5 𝑏
𝑐Cu 𝑏
2+ 𝑡 = 𝑐Cu2+ 0 𝑒
𝑉𝑡𝑜𝑡
3
2.5
2
1.5
1
0.5 heterogeneous
homogeneous
0
0 500 1000
time t [s]
reactor at high flow rate and long time frames. The method for measuring the
Cu2+ concentration is described in appendix A.
Figure 3.5 shows the experimental results of both experiments. For comparison,
the modeled curves for the homogeneous and heterogeneous reactions are also
provided in the figure. It is immediately obvious that both the AT and GDEx
processes behave in the same way. As expected, the AT process behaves like
a homogeneous process, demonstrating a time-dependent Cu2+ concentration
similar to that of figure 3.4, with similar process times (i.e., approximately
1000 seconds to complete removal). The GDEx process also demonstrates this
behavior. This proves the homogeneous nature of the precipitation reactions in
the GDEx process, under the given process conditions.
Despite the obvious similarities between model and experiment, some differences
should be pointed out. The first observed difference (1) is that the experimental
Cu2+ concentration does not immediately start to decrease. This is because
for the experiments the value of the initial pH was chosen below the value
where the Cu(OH)2 equilibrium is valid. This was done deliberately because it
allows a clear-cut initiation of the precipitation process. Starting the process
at this lower pH delays the initial time where the Cu2+ concentration starts
decreasing. To completely understand the shape of the pH curves, Chapter
4 is dedicated entirely to matching the model and the experiment of the AT
process for a reference case of ZnCl2 . The second observed difference (2) is
that, even after taking into account the delay, the experimental time-frame
is approximately 75% shorter than what was expected theoretically. This is
caused due to a discrepancy between the theoretically predicted precipitate (i.e.,
Cu(OH)2 ) and the experimentally observed precipitate (i.e., Cu4 (OH)6 SO4 ).
The theoretical stoichiometry of Cu(OH)2 is 1 Cu2+ for every 2 OH− , while
the experimental stoichiometry of Cu4 (OH)6 SO4 is 1 Cu2+ for every 1.5 OH− .
Therefore, the value of the theoretical removal rate of Cu should be 75% (i.e.,
1.5 divided by 2) of the value of the theoretical removal rate. In other words,
the experimental removal time should be 75% of the theoretical removal time,
which is confirmed in this case. The third observed difference (3) is the slight
curvature of the relatively linear decrease of the experimental Cu2+ . Given
that the curves of AT and GDEx run rather parallel, this might be the result of
calibration errors, which can be significant when the concentration changes are
small (i.e., concentration changes within the same order of magnitude) because
the measured potential is a logarithmic function of the Cu2+ concentration.
Lastly, one additional observation can be made, which is (4) the discrepancy
between AT and GDEx regarding both the beginning and ending times of the
removal/recovery process, indicated by the orange lines in figure 3.5. Moreover,
the difference increases over time (i.e., the difference between the beginning time
for AT and GDEx is smaller than the difference between the ending time for
32 EXPLORATORY STUDY OF THE GDEX PROCESS
5 12
4.5
10
pH
2.5 6
2
4
1.5
GDEx
1 AT 2
0.5 het.
hom.
0 0
0 500 1000 1500
added NaOH or charge [mol]
One of the aims of this PhD was to discover what caused oxidation of Ce(III)
to Ce(IV) in the catholyte side during GDEx. The oxidation of metal ions in
the catholyte side is remarkable since at the cathode a reduction reaction is
carried out, which should expectedly yield species that decrease the solution
redox potential and cause reduction reactions, rather than oxidation reactions.
Gas-diffusion electrodes are designed for electrochemical processes wherein at
least one of the reagents is in the gas form (i.e., not dissolved in solution).
They provide large surface areas to increase the triple-phase boundary of
gas/electrolyte/electrode and as such provide an answer to the mass-transfer
limitations posed by the combination of a small electrode surfaces and the
relatively low solubility of most gasses in water [105]. Passing air (or oxygen)
through a GDE is a well-known method to carry out the 4-electron and 2-
electron oxygen reduction reactions (ORRs), among others. The latter reaction
reduces two O(0) in O2 to two O(-I) in H2 O2 (in acid solutions) or HO− 2 (in
alkaline solutions), both well-known oxidants:
The formation of, what are considered, oxidants with a reduction reaction is
odd. In this case, it results from the fact that many species (e.g., many of the
transition metal ions) have a lower oxidation potential than H2 O2 , even though
it is the product of a reduction reaction of O2 . At the electrode surface, the
potential is chosen sufficiently low for the reduction of O2 to H2 O2 (or HO−
2 in
alkaline media) [20]. However, once in the bulk electrolyte, H2 O2 can oxidize
species with lower oxidation potentials.
®
Figure 3.6 demonstrates the change in pH and H2 O2 concentration as a result
®
of carrying out ORRs at the VITO CoRE GDE [58]. The H2 O2 concentration
was measured with QUANTOFIX measuring strips (details in chapters 6
and 7). Remark that the H2 O2 concentration is measured and not the HO− 2
concentration because, even if HO− 2 would be produced at the electrode, for
the practical pH ranges of the bulk electrolyte, it would be protonated to H2 O2
(pKH2 O2 = 11.62 [20]).
34 EXPLORATORY STUDY OF THE GDEX PROCESS
14
100% efficiency
14
12
pH
30% efficiency
6
6
4 4
2 H 2O2
Series2 2
pH
0 0
0 500 1000 1500
time [s]
0.8
10
0.8
0.6
0.6
8
0.4 0.4
pH
0.2 6 0.2
0 0
4
-0.2
-0.2
-0.4
2
-0.4 -0.6
-0.6 0 -0.8
0 500 1000 1500 2 3 4 5 6 7 8 9 10 11
time [s] pH
The discrepancy between both experiments is very clear. Using the Pt electrode
creates a reducing environment (potentials below 0 V vs. SHE). In contrast,
using the GDE creates an oxidizing environment (potentials above 0 V vs. SHE).
The theoretical relationship between the potential and the pH (-0.059 V per pH
unit) is relatively well observed in the experiments. The agreement at low pH in
the case of the platinum electrode is especially remarkable. For the Pt electrode,
the redox potential drops rapidly at the beginning of the experiment (figure
3.7a). This is likely caused by dissolved H2 in solution rapidly establishing a
potential close to the theoretical redox potential (figure 3.7b). In the case of
the GDE experiments, the increase is rather limited, but the potential decrease
with increasing pH is barely present. This is likely caused by the increasing
concentration of H2 O2 . The delay in pH increase between the experiment with
the GDE and the Pt electrode is remarkable and shows strong similarity to that
observed in section 2 in figure 3.5. This delay is discussed in Chapter 7.
In addition to the above observations, it is worth noting that the measured H2 O2
36 EXPLORATORY STUDY OF THE GDEX PROCESS
14
12
10
8
pH
6
Cu
CuCl2
Zn
ZnCl2
4
Fe
FeCl2
Co
CoCl2
2 Ni
NiCl2
Mn
MnCl2
0
0 0.001 0.002
added NaOH [mol]
12 12
(a) CuCl2 and ZnCl2 (b) NiCl2 and CoCl2
10 10
8 8
pH
pH
6 6
4 4
2 2
0 0
0 0.0005 0.001 0.0015 0 0.0005 0.001 0.0015
added charge [mol] added charge [mol]
been experimentally tested and the results are shown in figure 3.9. Mixtures of
equal amounts of CuCl2 and ZnCl2 , and of NiCl2 and CoCl2 , were subjected
to the GDEx process (see table 3.1 for process conditions). While GDEx of
the copper-zinc mixture demonstrates a clear separation of the precipitation
plateaus, such separation is indistinguishable for GDEx of the nickel-cobalt
mixture. Then again, it has been known for a very long time that nickel-cobalt
mixtures pose difficulties for such type of separation. For the interested reader,
Sole’s recent review is an excellent starting point [106]. Nevertheless, other
processes and mechanisms are available for such cases. Some of which might be
possible using GDEs (e.g., careful choice of cathode potential, use of different
gas, etc.). However, their discussion would lead us too far afield.
In addition, the selectivity of the metal hydroxides is complicated by the fact
that it depends on the individual concentrations of the dissolved metals. The
limitation of metal hydroxide separation can be demonstrated mathematically
as shown in the next paragraphs.
Assume a solution with a sufficiently low pH to have a mixture of two completely
dissolved metal ions: defining metal 1 as the metal having the lowest - and
metal 2 as the metal having the highest metal hydroxide solubility in water. For
infinitely slow (thermodynamic) reactive precipitation it is possible to determine
UNIVERSALITY AND SELECTIVITY OF AT AND GDEX PROCESSES 39
This removal efficiency can be easily calculated with the necessary assumptions.
To demonstrate the point, the calculations are made assuming that only the
metal hydroxides are precipitating and no other reactions occur (e.g., oxidation,
etc.). Another assumption is to ignore the occurrence of multi-species ions, which
is, in general, a reasonable assumption (see figure 4.2 on page 54 for an example
of the speciation of aqueous zinc). An arbitrary metal hydroxide equilibrium
can be defined using the equilibrium constant of that metal hydroxide in water:
ni
KM(OH)n = a n + (aOH− ) (3.34)
i M i
i
aj = γj cj (3.35)
An expression for the water equilibrium is also needed. For the value of the
activity coefficients of H+ and OH− equal to 1, the water equilibrium constant
KH2 O [mol2 L−2 ] is defined as:
! !
n2 n1
1 c2,0 (KH2 O ) 1 c1,f (KH2 O )
pH2,0 = − log = − log = pH1,f
n2 KM(OH)n n1 KM(OH)n
2 1
(3.38)
Using all of the prior definitions, the removal efficiency can be calculated as:
1 n
(c2,0 ) n2 KM(OH)n
η1 = 1 − 1
nn1 (3.39)
c1,0
2
KM(OH)n
2
The valences and the equilibrium constants are fixed for two chosen metals.
When considering two metals with equal valences, for example, η1 is a direct
function of the ratio of the two initial metal ion concentrations. The process
is more selective for one mixture compared to another mixture, if the removal
efficiency is higher for a given ratio of the initial concentrations c2,0 /c1,0 . In
figure 3.10, a higher selectivity for the case of Cu/Zn in comparison to the case
of Ni/Co can thus easily be recognized. This figure also suggests that at a
50/50 ratio, the nickel removal efficiency can still be as high as 45%, contrary to
what figure 3.9 might suggest. Therefore, in theory, stopping the precipitation
process when the pH reaches the initial precipitation pH of cobalt hydroxide,
45% of the nickel would have already been precipitated. In practice, this is
much harder to achieve, because, at the location where the base is introduced, it
would be difficult to control the pH well enough to selectively precipitate nickel
but not cobalt. Although the experimental verification would be interesting, it
was not carried out for this PhD.
In the previous sections, many aspects of the GDEx process have been introduced.
GDEx was demonstrated to be a partly homogeneous process, very similar to
AT. Therefore, the precipitation of metal (oxy)(hydr)oxide or metal hydroxysalt
(nano)particles with GDEx, is essentially caused by the production of OH− ,
which increases the pH. However, in section 3, it was demonstrated that by
using the GDE to carry out oxygen reduction reactions H2 O2 is produced,
which causes a high redox potential. This can cause the oxidation of metal ions
and affect the final characteristics of the produced nanoparticles. The effect of
oxidation is discussed in detail in chapters 5 and 6.
Here, the GDEx process demonstrates another virtue: its potential to work
at lower power consumption than alternative electrochemical OH− producing
ENERGY CONSIDERATIONS OF GDEX AND EFFECT OF SUPPORTING ELECTROLYTE 41
100%
90%
®
compartment flow cell reactor experiments with a Pt plate electrode compared
to an air-fed VITO CoRE GDE. Default process conditions.
Therefore, the removal time is directly proportional to the amount of cbCu2+ (0)
that is present, and the total volume Vtot . This is expected since it is obvious
that higher volumes with more species to be removed will take a longer time for
a given rate. The last important proportionality factor is the factor 2, which is
the amount of OH− necessary to remove 1 Cu2+ ion. This factor is equal to
the valence of the metal ion if the precipitate is the metal hydroxide or metal
ENERGY CONSIDERATIONS OF GDEX AND EFFECT OF SUPPORTING ELECTROLYTE 43
12
10
pH 8
2
NaCl
Na2SO4
Na2SO4
0
0 0.0005 0.001
added NaOH [mol]
oxide. Therefore, to precipitate all of the metal ions, solutions with metal ions
with a valence of 3 will need 1.5 times more OH− (and thus time, and energy)
than solutions with metal ions with a valence of 2. This factor is different
from the valence if negative anions participate in the precipitated molecule
(e.g., Cu4 (OH)6 SO4 ). This was demonstrated for the case of a 0.005 M CuSO4
solution (figure 3.5 on page 32), where this factor is lower than the valence of
Cu2+ (e.g., 1.5 instead of 2). In contrast, if a cation would be incorporated (e.g.,
Na+ ), this factor would be higher than the valence. Therefore, the available
cations and anions in the supporting electrolyte can be very important in the
consumption of OH− ions in traditional AT as well as in the GDEx process.
Anions can play another important role in the consumption of OH− ions. If
anions are conjugate bases of polyprotic acids they can act as buffers in certain
pH ranges. This is demonstrated in figure 3.12 by comparing AT of two different
supporting electrolytes without metals at the same initial pH of 3: 0.5 M NaCl
and 0.25 M Na2 SO4 . In the Na2 SO4 case, it takes up to 5 times more NaOH to
reach the initiation of the water inflection point than in the NaCl case. This
is caused by the high amount of HSO− 4 in the solution, from which protons
need to be removed before a significant pH change can be obtained. This high
amount of HSO− 4 is caused by the high amount of Na2 SO4 , which disturbs the
sulphate-bisulphate equilibrium. Thus, electrolyte solutions containing buffers
44 EXPLORATORY STUDY OF THE GDEX PROCESS
6 Conclusions
45
46 METAL REMOVAL FROM AQUEOUS SOLUTIONS: INSIGHTS FROM MODELING PRECIPITATION
TITRATION CURVES
Abstract
To provide a good reference for many metal ion removal processes, we developed
an analytical model (AM) and used the PHREEQC computer program to
numerically model (NM) arbitrary metal hydroxide precipitation titration curves
(PTCs) and compared these with slow titration experiments of ZnCl2 with
NaOH to precipitate Zn(OH)2 for validation. The AM effectively demonstrates
how initial pH, metal ion concentration and - valence and metal hydroxide
equilibrium constant affect the shape of the PTCs. A comparison of AM and
NM exposed AM limitations that include the effect of NaCl concentration,
activity coefficients, and multi-species ions, on the shape of the PTCs. The
experimental results agree well with either model for different concentrations of
ZnCl2 at slow titration rates. However, the models and experiments deviated
at higher titration rates. These deviations were attributed to inhomogeneities
in the solution concentration, which caused precipitation at a bulk pH below
the theoretical precipitation pH (i.e., pre-precipitation).
48 METAL REMOVAL FROM AQUEOUS SOLUTIONS: INSIGHTS FROM MODELING PRECIPITATION
TITRATION CURVES
1 Introduction
Heavy metal contamination of soil, ground - and drinking water, and food
is of major global concern [91, 114, 115]. Chemical precipitation, chemical
coagulation/flocculation, electrochemical methods, membrane filtration, ion-
exchange, bioremediation, and adsorption are the most current techniques for
heavy metal removal [91–96]. Metal hydroxide precipitation is one of the most
frequently used chemical precipitation methods to remove heavy metal ions from
aqueous solutions [98, 116–118]. Most of the existing research on metal removal
with metal hydroxide precipitation focuses on the end-result (i.e., effective metal
removal) but there is a dearth in the literature on the path towards metal
removal using metal hydroxide precipitation. Information on the path would
help in the comparison of different precipitation techniques. To bridge this
knowledge gap, an in-depth investigation of the metal hydroxide precipitation
path imposed itself. Such investigation was the goal of this work and was carried
out through the in-depth analysis of modeled and experimental precipitation
titration curves. The aim was to demonstrate easy-to-use models and to provide
an ideal case for future reference.
Precipitation titrations are traditionally used for the determination of halide
concentrations [119]. However, their usefulness has also been demonstrated for
metal hydroxide precipitation, for example when used for the characterization
of acid mining waters [120, 121].
During the metal hydroxide precipitation titration, the pH is measured and
plotted as a function of the added amount of precipitant (e.g., NaOH). Such a
pH plot can act as the blueprint of the metal removal path because the proton
concentration is indirectly related to the metal ion concentration via chemical
equilibria. A mathematical model of such precipitation titration curve can
demonstrate the effect of different parameters on the precipitation behavior.
Some authors have presented extensive work on the mathematical modeling of
titration curves, but in such cases, precipitation titrations are generally left out
of scope, or it is not the pH that is being modeled [122, 123].
Several parameters were identified as important: (1) the initial pH, (2) the initial
metal ion concentration, (3) the metal ion valence, (4) the metal hydroxide
equilibrium constant, (5) the supporting electrolyte concentration, (6) the
activity coefficients, and (7) the multi-species ions. We developed an analytical
model (AM) because its simple equations reveal the effect of the parameters
that most strongly influence the major features of the titration curve (i.e.,
parameters 1 to 4). We developed a numerical model (NM) with PHREEQC
that was able to include the effect of all parameters (i.e., parameters 1 to 7). A
comparison of AM and the NM demonstrated how parameters 5 to 7 influenced
EXPERIMENTAL 49
2 Experimental
®
initial pH of the solutions (i.e., pH = 3) was obtained via small additions (i.e.,
approximately 0.1 mL) of 1 M HCl solution (Titrisol , Merck). All experiments
were carried out at room temperature (i.e., 22 ◦ C ± 2 ◦ C). The pH was measured
every 5 seconds with a Metrohm 781 pH/ion meter equipped with a Metrohm
Unitrode pH electrode. The electrode was calibrated with 4 Merck standard
solutions (pH 4.01, pH 7.00, pH 10.00, and pH 12.00).
3 Models
The titration curve models in this work aim to simulate a traditional volumetric
titration (figure 3.2c). In volumetric titrations, a reagent is slowly added to the
analyte. This overall process is equivalent to a semi-batch reaction because the
total analyte volume increases (slightly) over time. Therefore, by extension, the
titration model is also a semi-batch reactor model.
50 METAL REMOVAL FROM AQUEOUS SOLUTIONS: INSIGHTS FROM MODELING PRECIPITATION
TITRATION CURVES
Figure 4.1 summarizes the stages of metal hydroxide precipitation. It only shows
the most important species. From the shape of the titration curve, one can
easily distinguish three stages during volumetric titration (figure 4.1b). Stage 1
before precipitation, stage 2 during precipitation (i.e., the so-called precipitation
plateau) and stage 3 after precipitation.
The titration model is calculated in two steps: first, the mass and charge
balances are solved to give time-dependent values of the concentrations and
activity coefficients of the protons and other species, then, the pH is calculated
as a function of the proton activity:
(a) (b)
Figure 4.1: (a) Simplified scheme of the different stages of the titration process.
The most prominent species at each stage are indicated. Upwards or downwards
arrows respectively indicate a strong concentration increase or decrease. (b)
An arbitrary example of a precipitation titration curve with the corresponding
three stages.
Table 4.1: The different solution species that are considered in the ZnCl2
reference system (minteqv4.dat database [125,126]), and corresponding values for
the equilibrium constants Ki (two or three significant figures are reprinted) [126]
Master H log Ki Zn log Ki Na log Ki Cl log Ki
species
Solution H+ 1.0 Zn2+ 1.0 Na+ 1.0 Cl− 1.6
species OH− 14 ZnCl+ 0.40 ZnCl+ 0.40
H2 O 1.0 ZnCl2 0.60 ZnCl2 0.60
ZnCl−
3 0.50 ZnCl−
3 0.50
ZnCl2−
4 0.20 ZnCl2−
4 0.20
ZnOHCl -7.5 ZnOHCl -7.5
ZnOH+ -9.0
Zn(OH)2 (aq) -17.8
Zn(OH)−3 -28.1
Zn(OH)2−
4 -40.5
In the precipitation reactor, solid and liquid are in (dynamic) equilibrium. The
species in solution are in equilibrium with the species in the solid. To capture the
conservation of mass for each species across the two phases, we define 4 master
species in our system (i.e., H, Zn, Na, and Cl) [124]. All (non-master) solution
species are defined in terms of four solution master species (i.e., respectively
H+ , Zn2+ , Na+ , and Cl− ) [124]. For example, a reaction for ZnCl− 3 would be
defined as Zn2+ + 3 Cl− = ZnCl− −
3 instead of as ZnCl2 + Cl = ZnCl3 . For
−
our zinc chloride reference system the different solution species for each master
species are given in table 4.1. We also define pure phases (i.e., solids) that are
in equilibrium with the solution (i.e., Zn(OH)2 (s)).
Each solution species i is defined by its activity [124]:
ai = γi mi (4.2)
with ai [mol kg−1 ] the activity of species i, γi [-] the activity coefficient of
species i, and mi [mol kg−1 ] the molality of species i and its total number of
moles in solution [124]:
ni = mi Waq (4.3)
with ni [mol] the total number of moles of species i, and Waq [kg] the total mass
of solvent water in solution [124]. For each aqueous species i, a mass-action
equation is defined [124]:
Maq
Y
Ki = ai a−c
m
m,i
(4.4)
m
Maq
Y
acmm,i
m
ni = mi Waq = Ki Waq (4.6)
γi
Maq
Y
Kp = acmm,p (4.7)
m
Remark that this definition differs from the definition of the equilibrium product
constant of metal hydroxides given in other parts of this PhD thesis. More
commonly, the equilibrium product constant is defined as KZn(OH)2 (s),common =
2
aZn2+ (aOH− ) . The other definition here is needed because all species are
referenced against H+ in PHREEQC. Of course, the common definition and
the definition of PHREEQC are intricately linked via the water equilibrium:
2
KZn(OH)2 (s),common = KOH − KZn(OH) (s) .
2
MODELS 53
with the function fm = 0 when the mole balance is achieved, Tm [mol] the
number of moles of master species m, Naq , and Np the number of aqueous
species and the number of solid phases (i.e., precipitates) respectively, bm the
moles of master species m in one mole of a given entity i or p. bm is usually,
but not always, equal to cm (the coefficient of the master species for m in
the mass-action equations) [124]. The charge balance, in our case, is written
as [124]:
Naq
X
fz = − zi ni (4.10)
i
with the function fz = 0 when the charge balance is achieved, and zi equal to
the charge of species i [124]. With all the above equations, it is possible to find
both an analytical or numerical solution for the proton concentration and to
solve equation (4.1).
Generally, numerical modeling is the go-to option for solving problems dealing
with metal ions in solution because of the large number of variables and
their mathematical interdependence. The added value of the analytical
model lies in its simplicity. It can easily demonstrate the effect of the most
important parameters on the path towards metal hydroxide precipitation. These
parameters are the initial solution pH, initial metal ion concentration, metal
ion valence, and metal hydroxide equilibrium constant. Six assumptions are
necessary to obtain an analytical solution to the system of mass- and charge
balance equations: (1) consider a limited amount of species, (2) assume constant
activity coefficients, (3) the values of all fi are equal to zero, (4) introduce the
precipitation pH as a new parameter, (5) use the charge balance as the master
equation, (6) define Zn2+ as a piece-wise function in relation to the precipitation
pH.
Assumption 1 states to only consider a few of the species from table 4.1(i.e., five
in total: H+ , OH− , Na+ , Cl− , Zn2+ ) and one pure phase (i.e., Zn(OH)2 (s)). In
essence, the former is equivalent to combining all Zn-containing solution species
into a single species Zn2+ . It should be evaluated separately for every system
(e.g., if other supporting electrolytes would be considered such as NaNO3 and
Na2 SO4 ) if this assumption is reasonable. If the assumption would appear
54 METAL REMOVAL FROM AQUEOUS SOLUTIONS: INSIGHTS FROM MODELING PRECIPITATION
TITRATION CURVES
-2 -2
totZn(aq)
tot Zn(aq) tot
totZn(aq)
Zn(aq)
-3 Zn2+
Zn(2+) -3 Zn(2+)
Zn2+
log(Concentration [mol L-1])
-10 -10
3 4 5 6 7 8 9 10 11 12 3 4 5 6 7 8 9 10 11 12
pH pH
Assumption 2 states that all the activity coefficients should be constant, instead
of concentration-dependent. Without this assumption, the system of equations
cannot be solved analytically. For very dilute systems, a value of 1 for the activity
coefficients is acceptable. At higher concentrations, the activity coefficients are
often very different from 1 and it is more suitable to use other values than 1.
Assumption 3 states that all the values of fi (equation (4.9)) should be fixed
and equal to zero. These functions only have non-zero values in the numerical
model (i.e., their values are iteratively decreased until a certain precision is
reached).
Assumption 4 introduces a new parameter, the so-called precipitation pH.
The precipitation pH, pHprec , is the lowest pH value at which M(OH)n starts
MODELS 55
2
!
γZn2+ Zn2+ (aH2 O )
1
pHprec = − log (4.11)
2 KZn(OH)2 (s)
with Zn2+ the concentration of Zn2+ [M].
2+ TZn CZnCl2 V0
Zn = = (pH < pHprec ) (4.16)
V0 + rV t V0 + rV t
2
2+ KZn(OH)2 (s) (aH+ )2 KZn(OH)2 (s) γH+ H+
Zn = 2 = 2 (pH ≥ pHprec )
(aH2 O ) γZn2+ (aH2 O ) γZn2+
(4.17)
56 METAL REMOVAL FROM AQUEOUS SOLUTIONS: INSIGHTS FROM MODELING PRECIPITATION
TITRATION CURVES
q
K a
2
CHCl + (CHCl ) + 4 γOH+−γ H2−O
H OH
pHi = − log ≈ − log (CHCl ) (4.20)
2
The numerical model is made with PHREEQC and the reader is guided to
the manual, where a very detailed description of the implementation can be
found [124].
RESULTS AND DISCUSSION 57
The most important functions fi that are solved by PHREEQC were shown
before in the basic equations section. The numerical model is solvable without
the assumptions made for the analytical model. PHREEQC can take into
account all species in table 4.1 and it can calculate all of the concentration-
dependent activity coefficients. The functions fi are iteratively solved via a
Newton-Raphson method. The PHREEQC algorithms run according to the
instructions of a user-defined input file. The complete input file of the numerical
model implementation with PHREEQC is found in appendix B.
Hereunder follows the modeling sequence in words. A certain solution with
an initial composition is chosen (i.e., an initial volume of water, a given initial
metal ion concentration, a given initial NaCl concentration, and initial acid
concentration). PHREEQC takes the thermodynamic data (e.g., the equilibrium
constant for the mass-action equation or some parameters to help solve the
activity coefficients) of the species in solution (i.e., all the species containing
only those elements provided in the initial composition) from the provided
database. PHREEQC iteratively calculates the initial number of moles of the
different species in solution. This determines the simulated initial experimental
electrolyte composition. From this point on, a chosen titrant solution (i.e., 0.1
M NaOH) is added in a fixed amount of discrete steps. The same iterative
process is used after every addition step. For each step, the electrolyte pH is
plotted as a function of the added amount of NaOH and the corresponding data
points are stored.
The analytical model (AM) and the numerical model (NM) were both approached
in two ways. The first version of the analytical model (AM1) assumes all activity
coefficients γi equal to 1. It is the most stripped version of a titration model and
is considered the baseline. The first version of the numerical model (NM1) takes
into account concentration-dependent activity coefficients but does not take into
account multi-species ions. NM1 can also calculate the concentration-dependent
activity coefficients at every titration step. For each species, a time-averaged
activity coefficient can be calculated. The second version of the analytical model
(AM2) uses these average values of the activity coefficients instead of a value
equal to 1. The second version of the numerical model (NM2) takes into account
concentration-dependent activity coefficients as well as multi-species ions.
58 METAL REMOVAL FROM AQUEOUS SOLUTIONS: INSIGHTS FROM MODELING PRECIPITATION
TITRATION CURVES
Figure 4.3 shows the results of AM1 and the effect of the initial acid concentration
(i.e., initial pH, equation (4.20)), the metal ion concentration, the metal ion
valence, and the equilibrium constant on the different stages of the precipitation
titration process. In every titration curve, the three stages are apparent. Stage
1 is mathematically described by equation (4.18), while stage 2 and stage 3 are
described by equation (4.19).
Stage 1 is determined by the initial HCl concentration (i.e., initial pH, equation
(4.20)), see figure 4.3a. Stage 1 ends when CHCl V0 − CNaOH rV t reaches a value
of 0 (equation (4.18)). Stage 2 is called the precipitation plateau and is the most
important stage during the precipitation titration process. Stage 2 is determined
by its length and its (initial) height. Its length is influenced by the initial metal
ion concentration and the metal ion valence, see figure 4.3b and figure 4.3c,
respectively. It ends when CHCl V0 + 2CZnCl2 V0 − CNaOH rV t reaches a value of
0. Its initial height is equal to the value of the precipitation pH (see equation
(4.11)). The initial height is influenced by the metal ion valence (i.e., a lower
valence gives a higher plateau height, figure 4.3c), the metal ion concentration
(i.e., a lower concentration gives a higher plateau height, figure 4.3b) and metal
hydroxide equilibrium constant (i.e., a higher value gives a higher plateau height,
figure 4.3d). Because of our assumptions, stage 3 is not significantly influenced
by the chosen parameters. To avoid the redissolution of formed precipitates at
higher pH, most metal removal or metal recovery processes should choose an
endpoint somewhere at the beginning of stage 3. This is also clear from figure
4.2, which demonstrates an increase in the concentration of soluble Zn(OH)− 3
and Zn(OH)2− 4 at high pH.
The process costs of a batch metal removal process is strongly correlated with
the amount of NaOH necessary to reach the second inflection point. Therefore,
because of their effect on the position of the second inflection point, the initial pH,
the metal ion concentration, and the metal ion valence are the most important
parameters concerning cost predictions.
RESULTS AND DISCUSSION 59
14 14
(a) (b)
12 12
10 10
8 8
pH
pH
6 6
4 4
2 pH 2 2 1 mM
pH 3 2 mM
pH 4 5 mM
0 0
0 0.0005 0.001 0.0015 0.002 0 0.0005 0.001 0.0015 0.002
NaOH added/ CNaOH rV t [mol] NaOH added/ CNaOH rV t [mol]
14 14
(c) (d)
12 12
10 10
8 8
pH
pH
6 6
4 4
2 2 2 10
3 11
4 12
0 0
0 0.0005 0.001 0.0015 0.002 0 0.0005 0.001 0.0015 0.002
NaOH added/ CNaOH rV t [mol] NaOH added/ CNaOH rV t [mol]
The effect of activity coefficients γi , the initial NaCl concentration CNaCl , and
the multi-species ions was investigated by comparing the analytical model and
the numerical model. Figure 4.4 compares the four model versions (i.e., AM1,
AM2, NM1, NM2) for specific sets of variables.
14 14 14
AM1 no NaCl AM2 no NaCl AM2 NaCl
12 AM2 no NaCl 12 AM2 NaCl 12 NM1 NaCl
NM1 no NaCl NM2 no NaCl NM2 NaCl
10 NM2 no NaCl 10 NM2 NaCl 10
8 8 8
pH
pH
pH
6 6 6
4 4 4
2 2 2
0 0 0
0 0.0005 0.001 0.0015 0.002 0 0.0005 0.001 0.0015 0.002 0 0.0005 0.001 0.0015 0.002
NaOH added/ CNaOH rV t [mol] NaOH added/ CNaOH rV t [mol] NaOH added/ CNaOH rV t [mol]
(a) (b) (c)
Figure 4.4: Comparison of the analytical model and numerical model. The
parameter values are V0 = 100 mL, CMCln = 5.0 mM, n = 2, KM(OH)2 =
1012 , KW = 10−14 , pHi = 3, CNaOH = 0.1 M. The constant average activity
coefficients for AM2 without NaCl are γH+ = 0.89, γOH− = 0.88 and γZn2+ =
0.63. The constant average activity coefficients for AM2 with NaCl are γH+ =
0.82, γOH− = 0.62 and γZn2+ = 0.23. (a) Influence of the activity coefficients
(equal to 1 for AM1, different from 1 for AM2, NM1, and NM2, see text for
more details) and influence of multi-species ions (species included in NM2, not
in AM1, AM2, and NM1) for electrolytes without NaCl. (b) Influence of NaCl
and its effect on the influence of the activity coefficients. (c) Influence of the
multi-species ions in solutions with NaCl.
In light of the complete precipitation titration curve, the models differ only
slightly. The inserts of figure 4.4 demonstrate the subtle differences. The largest
difference exists between AM1 and the other three models (figure 4.4a) because
AM1 does not take into account activity coefficients. This demonstrates that
the activity coefficients play a certain role, albeit much less pronounced than
the role of the parameters discussed in the previous section. This is especially
clear, taking into account that using average activity coefficients (AM2) almost
completely resolves the difference with the numerical model. A comparison of
NM1 and NM2 demonstrates the rather insignificant effect of multi-species ions
RESULTS AND DISCUSSION 61
1012 is used, rather than the, by the Minteq v4 database, provided value of
1011.334 . This confirms the need, expressed by the authors of the PHREEQC
manual [124], to cautiously use the constants provided in the database and
where necessary verify them experimentally.
14
12
10
8
pH
6
Exp. 1 mM
4 NM2 1 mM
Exp. 2 mM
2 NM2 2 mM
Exp. 5 mM
0 NM2 5 mM
0 0.0005 0.001 0.0015 0.002
NaOH added / CNaOHrVt [mol]
both the data of the modeled pH and experimental pH (i.e., the modeled
and experimental proton activities), it is possible to estimate the amount
of pre-precipitated metal ions. From figure 4.6, the estimated amount of
metal precipitated for 2 mL min−1 , 0.2 mL min−1 and 0.02 mL min−1 are
respectively 0.5 10−5 M, 0.5 10−6 M and 0.5 10−7 M. These are back of the
envelope calculations because it is difficult to determine the exact start of
pre-precipitation with experiments conducted for this study. Although the
calculations are only estimates, it is clear that a pH meter is a very sensitive
instrument to measure pre-precipitation, given it can detect pre-precipitated
amounts of 10−6 M and less. We leave to further developments the evaluation
of the time-dependent redissolution behavior.
64 METAL REMOVAL FROM AQUEOUS SOLUTIONS: INSIGHTS FROM MODELING PRECIPITATION
TITRATION CURVES
14
12
10
8
pH
4
NM2
NM2
2 0.02 mL min-1
0.02mL/min
0.2 mL min-1
0.2mL/min
0 mL min-1
22mL/min
0 0.0005 0.001 0.0015 0.002
NaOH added/ CNaOH rV t [mol]
At the end of precipitation (i.e., at the second inflection point), the model agrees
more with the experimental titration curves at faster precipitation rates than
with those of the 0.02 mL min−1 titration rate (figure 4.6). This anomaly was
also observed at different ZnCl2 concentrations (not shown here). It has yet
to be confirmed what causes this behavior, but we believe the long time-frame
(1000 min) causes disturbances in the pH measurement.
The influence of the NaCl concentration was also investigated. For industrial
wastewaters of low initial pH, the supporting electrolyte concentration can
become substantial when treated with NaOH. Figure 4.7 compares model NM2
with the experimental titration curves for electrolytes with and without 0.5 M
NaCl. In stage 3, the experimental pH is much lower in the case with NaCl than
in the case without NaCl. Although this is also observed by a comparison of
the models (figure 4.4b), the difference is much more significant experimentally.
This is attributed to the interference of sodium ions with the pH measurement,
the so-called alkali error [130]. At high sodium concentrations, and high pH
(i.e., low proton concentrations), the pH meter has difficulties in distinguishing
sodium ions from protons. This results in an overestimation of the proton
concentration (and thus the proton activity), which corresponds with a lower
RESULTS AND DISCUSSION 65
pH. Another important observation is the fact that in the case of 0.5 M NaCl, the
experimental precipitation plateau is significantly lower than that of the plateau
calculated by the model. It is practically equal to that of the case without 0.5
M NaCl. This might be the result of the choice of Pitzer coefficients for ZnCl2
solutions (see the PHREEQC input file in appendix B), taken from scientific
literature [131]. These coefficients might not be suitable in the concentration
range used, or as indicated before, might be unreliable. Additionally, at initial
precipitation, the supersaturation behavior in the case with and without NaCl
differs significantly. This might suggest the possible precipitation of a different
species, before the actual precipitation of Zn(OH)2 .
14
12
10
8
pH
4
Exp. no NaCl
2 Exp. NaCl
NM2 no NaCl
NM2 NaCl
0
0 0.0005 0.001 0.0015 0.002
NaOH added/ CNaOH rV t [mol]
5 Conclusions
Oxidation-assisted alkaline
precipitation: effect of H2O2
on the size of CuO and
FeOOH nanoparticles
One of the objectives of this PhD was to identify the role of the GDEx
process regarding the oxidation of metal ions in the catholyte. Chapter 3
demonstrated the production of an oxidant, H2 O2 , during GDEx with
an air-fed GDE and the role of H2 O2 on increasing the redox potential.
As such, the oxidant was identified, but not yet its fundamental role in
the precipitation mechanism of the metal (oxy)(hydr)oxide nanoparticles.
This chapter clarifies this. An experimental procedure was specifically
designed to isolate the effect of H2 O2 on the precipitation behavior of
metal (oxy)(hydr)oxides. As such, the mechanism of oxidation-assisted
alkaline precipitation (Ox-AP) was discovered.
67
68 OXIDATION-ASSISTED ALKALINE PRECIPITATION: EFFECT OF H2 O2 ON THE SIZE OF CUO AND
FEOOH NANOPARTICLES
This chapter has been published as: Eggermont S.G., Rua-Ibarz A., Tirez K.,
Dominguez Benetton X., Fransaer J. Oxidation-assisted alkaline precipitation:
the effect of H2 O2 on the size of CuO and FeOOH nanoparticles. RSC Advances.
2019;9(51):29902-8. dio: 10.1039/C9RA03086G
A large part of the electronic supporting information of the original publication
was merged into the text, rather than leave it as an appendix, for a more unified
approach in the PhD thesis.
OXIDATION-ASSISTED ALKALINE PRECIPITATION: EFFECT OF H2 O2 ON THE SIZE OF CUO AND
FEOOH NANOPARTICLES 69
Abstract
H2 O2 was demonstrated to narrow the size distribution and decrease the size of
CuO and hydrous FeOOH (2-line ferrihydrite) nanoparticles at conditions of high
supersaturation. We introduce oxidation-assisted alkaline precipitation (Ox-AP)
and compare it to traditional alkaline precipitation (AP). While for AP, a metal
salt solution (e.g., CuCl2 ) is mixed with an alkali (e.g., NaOH), for Ox-AP, the
more reduced form of that metal salt solution (e.g., CuCl) is simultaneously
mixed with that alkali and an oxidant (e.g., H2 O2 ). The resulting precipitates
were characterized by SEM, XRD, DLS and single-particle ICP-MS and showed
to be nanoparticles (NPs). Ox-AP CuO NPs were up to 3 times smaller than AP
NPs. Ox-AP FeOOH NPs were up to 22.5% smaller than AP NPs. We discuss
and propose a possible mechanism of Ox-AP through careful consideration of
the known reaction chemistry of iron and copper. We propose that an increased
monomer formation rate enhances the nucleation rate, which ultimately results
in smaller particles with a more narrow distribution. The more distinct effect
of Ox-AP on copper was attributed to the fast formation of the stable CuO
monomer compared to AP, where the Cu(OH)2 and/or Cu2 (OH)3 Cl monomers
are more likely. Although the exact mechanism of Ox-AP needs experimental
confirmation, our results nicely demonstrate the potential of using Ox-AP to
produce smaller NPs with a more narrow distribution in comparison to using
AP.
70 OXIDATION-ASSISTED ALKALINE PRECIPITATION: EFFECT OF H2 O2 ON THE SIZE OF CUO AND
FEOOH NANOPARTICLES
1 Introduction
with Ksp,Cu(OH)2 :
Ksp,Cu(OH)2 = aCu2+ ,eq a2OH− ,eq (5.6)
and ai,eq the equilibrium activities of species i.
If S > 1, the amount of reactants is in excess for equilibrium and precipitation
can occur. Often S needs to be many times higher than 1 before precipitation
occurs, depending on other limiting factors (e.g., surface energy of the precipitate,
temperature, etc.). If S < 1, the amount of reactants is limiting for precipitation
and hence the reactants stay in solution. According to CNT, a higher
supersaturation level increases the nucleation rate which results in smaller
particles [135, 137–139].
The (homogeneous) nucleation rate J [s−1 ] depends on supersaturation via a
power law, which in the simplest form is written as:
J −2
= 10−A[log S] (5.7)
Jmax
with Jmax [s−1 ] the nucleation rate at infinitely high supersaturation and A [-]
a collection of variables [135, 137–139]:
4βa3 γ 3 V 2
A= (5.8)
[27βv2 (kB T ln 10)3 ]
with βa [-] the form factor of the particle surface , γ [J m−2 ] the surface free
energy per unit area, V [m3 ] the molecular volume, βv [-] the form factor of
the particle volume, kB [m2 kg s−2 K−1 ] the Boltzmann constant and T [K] the
temperature.
A more elaborate approach, including for example heterogeneous nucleation
and other important phenomena during precipitation, is found in scientific
72 OXIDATION-ASSISTED ALKALINE PRECIPITATION: EFFECT OF H2 O2 ON THE SIZE OF CUO AND
FEOOH NANOPARTICLES
CNT literature [135, 137–139, 141]. Although CNT demonstrated its usefulness
in the past, recent studies increasingly demonstrate the limitations of CNT.
These studies discovered different nonclassical crystallization routes and the
importance of chemical kinetics [140, 142].
In this study, we refute our original hypothesis, by demonstrating for two cases
that Ox-AP yielded smaller nanoparticles with a more narrow size distribution
than AP. Ox-AP achieves these feats with as little as possible alteration of the
supersaturation. Contrary to recent discoveries of non-classical crystallization
routes, which are related to steps (3) and (4) of the precipitation process, we
believe that the effect of Ox-AP is related to steps (1) and (2) of the precipitation
process. We discuss the possible nature of Ox-AP of CuO and FeOOH through
careful consideration of the known reaction chemistry of iron and copper.
2 Experimental
When fully dissolved, the solution was diluted with the required amount of
demineralized water to complete a 500 mL volumetric flask. The NaOH
electrolytes were prepared to match the concentration and the valence of the
metal ions in the metal chloride solutions. This was deemed important because
Jolivet and coworkers demonstrated the importance of pH on the size of metal
oxide nanoparticles [143]. The necessary amount of OH− to precipitate all
EXPERIMENTAL 73
of the metal ions as hydroxides were taken as the starting value (n CM Cln ).
Because the metal chloride solutions had 0.1 M HCl, an additional 0.1 M NaOH
was necessary to overcome the water equilibrium. The end pH of the solution
after mixing was chosen to be pH 12 (i.e., 0.01 M OH− ), therefore an additional
0.02 M NaOH (in the 250 mL initial solution to become 0.01 M in 500 mL final
solution) was necessary. In the case of FeCl2 and CuCl, 10 mM H2 O2 in the
form of 0.283 mL of 30% H2 O2 (Sigma-Aldrich), an excess of 100%, was added
to the 250 mL NaOH solution just before mixing (to avoid H2 O2 decomposition
at high pH). The composition of the NaOH solutions was:
To have a consistent reaction front and consistent mixing, the solutions were
mixed in a Y-junction by pumping them separately at high equal volumetric
rates, after which the mixture was collected in a glass beaker. The experimental
setup used a Watson-Marlow 323 peristaltic pump, three 90 cm Watson Marlow
1.6/3.2 mm Marprene 902 tubes and a Y-junction Hibiki Y-1 3mm. The Y-
junction was connected to the ends of each tube. The two tubes connected to
the upper legs of the Y-junction were each put into a peristaltic pump head
at around the 75 cm mark. One of the inlets went into the NaOH solution,
the other inlet into the metal chloride solution. The outlet went into a beaker
with a WTW pH electrode Sentix 61 sensor (WTW 340i pH meter) calibrated
with a two-point calibration (pH 4 and pH 7), which was used to measure the
final steady-state pH. Both the metal chloride solution and the NaOH solution
were pumped by an equal rate of 100 mL min-1 through the Y-junction into
the beaker. Depending on the composition of the influents, the effluent was
colorless with differently colored precipitates.
A summary of the electrolytes and a schematic drawing of the Y-junction
tubular mixing-reactor are shown in figure 5.1.
74 OXIDATION-ASSISTED ALKALINE PRECIPITATION: EFFECT OF H2 O2 ON THE SIZE OF CUO AND
FEOOH NANOPARTICLES
For each of the samples, the post-treatment was equal. 250 mL of the effluent
was kept for storage and possible later use. The other 250 mL was centrifuged
with a Jouan CR422 in 6 centrifuge tubes of 50 mL capacity for 10 min at 3000
RPM. The supernatant of the tubes was removed and the remaining precipitate
was washed with demineralized water and using a vortex (IKA VF2). The
resulting solution was diluted to 50 mL into a 50 mL tube. This tube was
centrifuged for 10 min at 14000 RPM (Sorvall LYNX 6000 Superspeed Centrifuge
- rotor for 12 50 mL tubes), the supernatant removed and subsequently washed
in 50 mL demineralized water. These steps were repeated once more, but adding
only 20 mL demineralized water instead of 50 mL. From this 20 mL, 1 mL was
used for dilutions for the single-particle ICP-MS measurements. 9 mL was used
for dilutions of the DLS measurements. The other 10 mL was dried for dry
analysis.
Table 5.1: pH resulting after mixing the different solutions (the theoretical pH
value anticipated for all instances is pH 12)
Solution pH measured after mixing
CuCl2 11.67 ± 0.02
CuCl + H2 O2 11.72 ± 0.04
FeCl3 11.49 ± 0.05
FeCl2 + H2 O2 11.54 ± 0.06
3 Results
Figure 5.2 demonstrates the dispersed precipitates just after mixing, stored in
plastic storage tubes. When H2 O2 was used, some gas evolution was observed,
in the form of bubbles, suggesting the decomposition of excess H2 O2 to oxygen
gas and water (see insets of figures 5.2b and 5.2d). Table 5.1 shows the pH
measured after mixing, which was lower than the expected value of 12. This is
most likely due to the sodium error at 0.5 M NaCl. The pH difference between
the copper case and the iron case was approximately 1 mM OH− , and thus
considered insignificant for this study, because it is less than 1 % of the amount
of the total OH− added. The copper samples had a distinct color difference
(figure 5.2). Initially, the Cu2+ sample was distinctly blue but developed into
a dark brown color by the end of the washing process. The Cu+ (with H2 O2 )
sample was dark greenish-brown for a very short instance, but even before the
washing step started its color changed to dark brown. Both the Fe3+ and the
76 OXIDATION-ASSISTED ALKALINE PRECIPITATION: EFFECT OF H2 O2 ON THE SIZE OF CUO AND
FEOOH NANOPARTICLES
Figure 5.2: Color of the precipitates obtained directly after precipitation. From
left to right: CuCl2 (blue/turquoise), CuCl with H2 O2 (dark green/brown),
FeCl3 (bright orange) and FeCl2 with H2 O2 (bright orange). Insets for better
appreciation of the gas bubbles formed in the cases with H2 O2 .
Fe2+ (with H2 O2 ) samples were bright orange, with little observable change
over time.
Single-particle ICP-MS and DLS were used to determine the size of the
nanoparticles [144]. In all cases, spherical shape was assumed, and thus, the
particle spherical equivalent diameter is reported. For rod-shaped particles
of similar length over width ratio, the comparison of particles with different
sizes remains valid [145]. Figure 5.3 demonstrates the particle size distributions
calculated from the sp ICP-MS measurements for the different cases. In the case
of copper, the size of the particles shows a remarkable difference, contradicting
our initial hypothesis of size consistency between AP and Ox-AP for the same
supersaturation value. In the case of iron, the size difference follows a similar
trend, although much less pronounced. Figures 5.4a and 5.4b respectively
compare the DLS and sp ICP-MS size measurements. Both measurements
demonstrate that the Ox-AP yields smaller particles than the traditional AP.
In the case of copper, the average size of the oxidation-induced precipitates is
3 times smaller (i.e., 30 nm instead of 90 nm) according to sp ICP-MS, and
over 2 times smaller (i.e., 100 nm instead of 250 nm) according to DLS. In the
case of iron, the results are less pronounced, yet inconsistent with our original
hypothesis: the average size of the oxidation-induced precipitates is 12.5%
RESULTS 77
20 20
CuCl2 FeCl3
CuCl + H2O2 FeCl2 + H2O2
relative frequency [%]
10 10
5 5
0 0
10 30 50 70 90 110 130 150 10 20 30 40 50 60 70 80
equivalent spherical diameter [nm] equivalent spherical diameter [nm]
Table 5.2: Polydispersity index values from single-particle ICP-MS and DLS
measurements of the FeOOH and CuO precipitates formed by Ox-AP and AP
CuCl2 CuCl + H2 O2 FeCl3 FeCl2 + H2 O2
sp ICP-MS 0.32 0.08 0.12 0.07
DLS 0.38 0.28 0.23 0.20
250 100
a 95 b
225
90
80
150 40
125 35
30
100 25
75 20
50 15
10
25 5
0 0
CuCl2 FeCl3 CuCl2 FeCl3
CuCl + H2O2 FeCl2 + H2O2 CuCl + H2O2 FeCl2 + H2O2
Figure 5.4: Average diameters of the formed nanoparticles for the cases of
copper and iron derived from distributions of (a) z-average diameter from DLS
measurements and (b) average equivalent spherical diameter from single-particle
ICP-MS measurements. Error bars indicate the standard deviation of the 3
measured average particle sizes (i.e., not from the particle distributions).
The XRD-patterns for the copper and iron cases are shown in figures 5.5a
and 5.5b, respectively. CuO is the main phase of the copper case. The CuCl2
case shows two additional peaks, indicating an additional Cu2 (OH)3 Cl phase.
This indicates that at least part of the precipitated phase is the Cu2 (OH)3 Cl
instead of the Cu(OH)2 and besides the reactions described in equation (5.1),
the following overall reactions are likely:
■ Cu2(OH)3Cl a • FeOOH b
◻ CuO ◻ •
◻ FeCl2 + H2O2
CuCl + H2O2
◻
◻ ◻ ◻ ◻ ◻◻
a.u. [-]
a.u. [-]
• FeCl3
◻ CuCl2
■ ◻ ■ ◻ ◻
◻ ◻ ◻◻
•
15 20 25 30 35 40 45 50 55 60 65 70 15 20 25 30 35 40 45 50 55 60 65 70
2θ [�] 2θ [�]
It is known that, with time ferrihydrite can convert into more crystalline
products such as goethite and hematite [152].
Figure 5.6 shows the SEM images for the different precipitates. For the copper
case, separate particles are easily distinguished for AP and Ox-AP. Ellipsoidal
needle-shaped NPs are observable in either case, but the Ox-AP NPs are
smaller than the AP NPs. For the AP case, the existence of a second, less
abundant, phase (i.e., Cu2 (OH)3 Cl) is also clear. For the iron case, even at
a higher magnification than for the copper case, separate particles are hardly
distinguishable for AP and Ox-AP (figures 5.6c and 5.6d). The nanoparticles
form large aggregates of smaller nanoparticles. With digital magnification, it was
possible to identify a few separate nanoparticles (see insets figure5.6). However,
these do not provide conclusive evidence on the size differences between AP
and Ox-AP for the iron case.
80 OXIDATION-ASSISTED ALKALINE PRECIPITATION: EFFECT OF H2 O2 ON THE SIZE OF CUO AND
FEOOH NANOPARTICLES
4 Discussion
The observations can be summarized as follows: (1) the experiments with iron
result in smaller nanoparticles with more narrow size distributions than the
experiments with copper, both for Ox-AP and AP; (2) Ox-AP results in smaller
nanoparticles with more narrow size distributions than AP, both for the copper
and the iron cases; (3) the effect of Ox-AP is much more pronounced in the
case of copper than in the case of iron.
The observations can be discussed in the light of the particle formation process.
As discussed above, the particle formation process via hydrolysis proceeds in
four steps:(1) formation of zero-charge precursors, (2) creation of nuclei via the
olation and/or oxolation of zero-charge precursors, (3) growth of the nuclei via
addition of matter via olation and/or oxolation and (4) aging of the primary
particles via Ostwald Ripening and/or aggregation [19]. Step (1) depends on the
solubility of the species and the concentration of the reagents and the available
ligands (i.e., especially on the pH). Moreover, it is also limited by the lability
of the water ligand in the first coordination shell of the metal ion. Step (2)
depends on the concentration of the zero-charge species and the water lability
as well, because the lability co-dictates the rates at which olation and oxolation
can occur. Step (3) and step (4) are still intensively debated in current literature
and strongly depend on step (1) and step (2), and any of the other process
parameters [140].
In the AP process, OH− ions compete with H2 O and Cl− ions to enter the first
hydration sphere of the metal aquo complex. The strong nucleophilic character
of OH− allows it to form more stable complexes with the metal ion, and at the
right OH− concentration, the most stable form is a sparingly soluble zero-charge
monomer. This monomer can form polymers via olation and oxolation until a
critical size is reached and they are stable in solution. The critical size strongly
depends on the inherent chemical nature of the monomer (i.e., the metal ion,
the ligands, etc.), the olation and oxolation kinetics to which it is subjected,
and also on the level of supersaturation (i.e., the concentration of the precursor
ions).
Supersaturation S is much higher for the iron case than for the copper case
(table 5.3), more detailed calculations are given in appendix D, and in accordance
with CNT, the NPs in the iron case should be smaller than in the copper case.
Since the final pH, supporting electrolyte concentration, metal ion concentration
and process operating conditions were the same for the copper and iron cases,
the inherent differences between aqueous behavior of Cu2+ and Fe3+ (i.e., the
case-specific values of Jmax and A from equation (5.7)) likely cause the observed
size difference between the CuO and hydrous FeOOH nanoparticles. The values
82 OXIDATION-ASSISTED ALKALINE PRECIPITATION: EFFECT OF H2 O2 ON THE SIZE OF CUO AND
FEOOH NANOPARTICLES
Table 5.3: Solubility products and supersaturation values for the hydroxides of
copper and iron acting as a precursor for FeOOH and CuO
Precursor Ksp S for end-pH 12
Cu(OH)2 2.20 10−20 [mol3 L−3 ]( [107]) 9.1 1013
−35 6 −6 ( [153])
Cu2 (OH)3 Cl 2.34 10 [mol L ] 5.2 1024
−39 4 −4 ( [107])
Fe(OH)3 2.79 10 [mol L ] 2.8 1031
of Jmax and A depend on the chemical nature of the monomers and the olation
and oxolation kinetics of those monomers. This is inherently different for copper
and iron. Discussion of these differences would lead us too far afield.
More interesting is to compare the Ox-AP cases of copper and iron with the AP
cases of copper and iron, respectively. Our original hypothesis was that, for the
same supersaturation values, AP and Ox-AP would yield nanoparticles with
the same characteristics (i.e., composition, size, morphology, etc.). Imagine
infinitely fast oxidation of the metal ions in the Ox-AP case. This would cause
complete oxidation prior to any other reaction (i.e., any of the four steps in the
particle formation process), OH− would be formed according to the following
equations:
2Cu+ + H2 O2 + 2OH− → 2Cu2+ + 2OH−
Cu2+
+ 2OH− → 2CuO + 2H2 O (5.11)
suppress the redissolution of smaller particles in favor of the larger particles (i.e.,
Ostwald ripening). It is also unlikely that hydrogen peroxide would interfere in
these stages, especially because most of the hydrogen peroxide reacts away in
an oxidation reaction with the metal, and the excess quickly decomposes (i.e.,
bubbles are visible in the final solution). Therefore the interference of H2 O2
is most likely occurring during step (1) and/or step (2) of the precipitation
process. The olation and oxolation reactions in step (2) depend strongly on the
characteristics of the zero-charge monomers formed in step 1. Therefore, we
argue that H2 O2 takes an important role in the formation of the monomers and
the subsequent olation and/or oxolation reactions.
In Ox-AP, both OH− and HO− 2 (i.e., the deprotonated form of H2 O2 at high pH
values [20]) act as strong nucleophiles (i.e., HO− −
2 even more so than OH [154])
−
and compete with water and Cl to enter the first hydration sphere of the metal
ion. There are many possible reaction pathways for HO− 2 (i.e., H2 O2 ) with the
metal ion, known in scientific literature as Fenton(-like) reactions [155,156]. The
exact pathways are still subject to much research and heavily system dependent.
Here, we propose, both for copper and iron, only one of the possible pathways.
For clarity, we do not take into account the many possible side-reactions and
intermediaries.
Without hydrogen peroxide, the expected zero-charge monomer of Cu+ is CuOH
at elevated pH, with a subsequent oxolation reaction to Cu2 O:
Cu+ + HO−
2 → CuOOH (5.14)
5 Conclusion
Oxidation-assisted alkaline
precipitation of nanoparticles
using gas-diffusion electrodes
87
88 OXIDATION-ASSISTED ALKALINE PRECIPITATION OF NANOPARTICLES USING GAS-DIFFUSION
ELECTRODES
Abstract
1 Introduction
1 2 3
precipitation plateau
pH
precipitation pH
2 Experimental
A NaCl stock solution of 5.00 M NaCl was made with 292.2 g of anhydrous NaCl
(Merck) in a volumetric flask of 1 L and filled with the appropriate amount
of demineralized water. An acid stock solution of 0.01 M HCl was made by a
two-step 1/10 dilution of a 1 M HCl solution (Chem-Lab) with demineralized
water. Metal chloride stock solutions of 0.500 L with a concentration of 0.200 M
of the individual metal chloride were made, with ZnCl2 (anhydrous, Chem-lab),
MnCl2 ·4H2 O (Sigma Aldrich) or FeCl2 ·4H2 O (Arcos organics). First, 110% of
the necessary mass was weighed and added to the required amount of water to
complete a 0.500 L volumetric flask. The FeCl2 solution was filtered to remove
any oxidized iron-containing particles. Subsequently, ICP-OES (Varian 720-ES)
was used to determine the concentration of Zn, Mn, and Fe. The values of these
concentrations were used to calculate the appropriate dilution factors needed to
establish 1 L metal chloride stock solutions with a final composition of 0.200 M
ZnCl2 , MnCl2 or FeCl2 , respectively. The GDEx catholyte solutions and AT
solutions were made by diluting 0.025 L of the metal chloride stock solution
with 0.100 L of the acid stock solution, 0.100 L of NaCl stock solution, and
0.775 L of demineralized water. Each experiment was carried out with 0.100
L of these solutions, with a composition of 1 mM HCl with 0.50 M NaCl and
with 0.005 M ZnCl2 , MnCl2 or FeCl2 . The GDEx anolyte solution was a 0.100
L 0.50 M NaCl solution.
The complete GDEx set-up is shown in figure 3.2a and a schematic of the design
of a representative GDEx flow-cell reactor in figure 3.2b. The volume of the
flow-cell reactor compartments was 10 mL.
®
The cathode was a VITO CoRE gas-diffusion electrode (GDE) [58]. The
anode was a flat platinized tantalum plate. The geometric surface areas of
the anode and cathode were 10 cm2 , respectively. In between the anolyte and
catholyte compartment, there was an anion exchange membrane (FUMASEP ®
94 OXIDATION-ASSISTED ALKALINE PRECIPITATION OF NANOPARTICLES USING GAS-DIFFUSION
ELECTRODES
®
through the membrane. The anolyte and catholyte reservoirs were 250 mL
DURAN laboratory bottles, with additional flask openings to allow for a
pH-meter. The cap was supplemented with openings for argon purging. The
solutions in the reservoir bottles were stirred at approximately 700-800 rpm
with a magnetic stirring bar of 3 cm, on an IKA RCTbasic S1 stirrer at rotation
level 7. The electrolytes were circulated with a double-headed peristaltic pump
(Cole-Palmer MasterflexTM L/s 7551-00 with Easyload II 77200-60 pump heads)
at 200 mL min−1 through 90 cm tubes with a diameter of 3.2 mm (Cole-Palmer
MasterflexTM VersilonTM L/s 06475-16). For 200 mL min−1 , the average
residence time is 3 seconds in the flow-cell and 1 second in a connection tube.
At the back of the GDE, an airflow of 50 mL min−1 was provided. The exit tube
of the gas channel was put into a 30 cm high water column to provide a gas back-
pressure of approximately 30 mbar. GDEx was carried out galvanostatically
at 100 A m−2 (Biologic Biostat SP-300) for 1926 seconds (i.e., equivalent to
the total theoretical addition of 0.002 mol of OH− ). At this current density,
the polarization potential is approximately -0.07 V vs. SHE which guarantees
that only oxygen in the air is reduced (and not, for example, CO2 ) and that
the hydrogen evolution reaction (HER) is avoided [20]. A 3 M KCl Ag/AgCl
reference electrode (Radiometer Analytical REF321) was used and in connection
with the working electrode (i.e., GDE) through a Luggin capillary filled with
500 mM NaCl dilution stock solution. All experiments were carried out at room
temperature (i.e., 22 ◦ C ± 2 ◦ C).
2.3 AT process
®
was carried out with a titrator (Methrohm 702 SM Titrino). The batch reactor
was the same DURAN bottle as described before.
®
A total of 20 mL of a 0.1 M NaOH solution (Titrisol , Merck) was added
at a titration rate of 0.62 mL min−1 . This titration volume and titration
rate were chosen to add a total of 0.002 mol OH− at an equivalent rate of
100% efficient OH− production rate using GDEx current density of 100 mA
m−2 (i.e., the chosen galvanostatic current density for the GDEx process).
The relatively low concentration of NaOH in the titrant was chosen to avoid
concentration gradients in the solution as much as possible. This advantage
goes hand in hand with the disadvantage of increasing the total reaction volume
over time. However, this volume increase does not influence the time axis and
only insignificantly influences the pH values compared to the case of infinitely
small additions of highly alkaline NaOH solutions. The pH measurements were
EXPERIMENTAL 95
carried out equivalently to GDEx. All experiments were carried out at room
temperature (i.e., 22 ◦ C ± 2 ◦ C).
Both during AT and GDEx, the pH was measured every 5 seconds with a
Metrohm 781 pH/ion meter equipped with a Metrohm Unitrode pH electrode.
The pH electrode was calibrated with 4 Merck standard solutions (pH 4.01, pH
7.00, pH 10.00, and pH 12.00).
®
®
The hydrogen peroxide concentration was measured every minute (QUANTOFIX
Relax with QUANTOFIX Peroxide 25 and 100 test strips) during the GDEx
experiments. Additionally, the peroxide concentration was also measured for
a GDEx experiment of a blank 0.5 M NaCl with 0.01 M HCl solution as a
benchmark.
After completion of the GDEx or AT process, the precipitates were washed and
dried. Washing was done in 5 subsequent washing steps. In the first washing
step, the electrolyte was divided into two samples and centrifuged for 5 minutes
at 10,000 rpm (Hettich Rotina 35). The liquid supernatant was decanted and 80
mL of demineralized water was added to each of the two samples to redisperse
and wash the precipitates. The subsequent two washing steps repeated the
centrifuging, decanting and redispersion steps. In the fourth washing step, the
samples were redispersed into 40 mL of water each and joined into one sample
and only then centrifuged. This allowed for a better collection at the end of
the post-processing. In the last washing step, the 80 mL supernatant of the
single sample was removed and 20 mL of ethanol was added to redisperse the
particles. The last centrifugation was completed and the ethanol supernatant
removed. The humid precipitate was collected from the centrifugation container
and put on a weighing boat and left to dry in a desiccator for 12 hours. The
last step in ethanol was carried out to allow for much quicker drying than if
only water was used, avoiding the unnecessary complications of the particles
being in humid conditions for too long.
The composition of the dried precipitates were characterized with XRD (Bruker
D2 phaser, Cu source Kα,avg = 1.54, line focus, LynxEye detector). The size
and morphology of the dried precipitate were characterized by SEM (Philips
XL 30 FEG).
96 OXIDATION-ASSISTED ALKALINE PRECIPITATION OF NANOPARTICLES USING GAS-DIFFUSION
ELECTRODES
Zn Mn Fe
3.2 Production of H2 O2
7
blank 100% efficiency
Zn
2 24% efficiency
0
0 0.0005 0.001 0.0015
added charge [mol]
The expected shape of the dynamic pH curves was briefly discussed in the
introduction (figure 6.1). Figure 6.4 compares the dynamic pH measurements of
the galvanostatic GDEx experiments with the AT experiments for the solutions
with ZnCl2 , MnCl2 , and FeCl2 . Figures 6.4a and 6.4b compare the results of
AT and GDEx respectively for solutions with ZnCl2 , MnCl2 , or FeCl2 . The
shapes of the dynamic pH curves are all in line with the expectations. Figures
6.4c, 6.4d, and 6.4e compare GDEx and AT for solutions with ZnCl2 , MnCl2 ,
or FeCl2 respectively.
For AT, stage 1 and 2 ended as expected upon the addition of respectively
10−4 mol and 1.1 10−3 mol of OH− (figure 6.4a). Also, the theoretical shape
of the curve corresponds to the experimental shape of the curve for the cases
of Zn and Mn. For Fe, a small deviation of this shape is observable: besides
the main precipitation plateau, a short inclination before the main plateau is
visible (figure 6.4a). This is attributed to the fact that small amounts of Fe3+
may have remained or may have formed in the prepared solution. These Fe3+
RESULTS AND DISCUSSION 99
14
a AT b GDEx
12
10
pH 8
6
4
Zn Zn
2 Mn Mn
Fe Fe
0
0 0.001 0.002 0 0.001 0.002
added NaOH [mol] added charge [mol]
14
c Zn d Mn e Fe
12
10
8
pH
6
4
2 AT AT AT
GDEx GDEx GDEx
0
0 0.001 0.002 0 0.001 0.002 0 0.001 0.002
added NaOH or charge [mol] added NaOH or charge [mol] added NaOH or charge [mol]
Figure 6.4: Comparison of the dynamic pH curves for ZnCl2 , MnCl2 and FeCl2
for AT (a) and GDEx (b). Comparison of the dynamic pH curves for AT and
GDEx for ZnCl2 (c), MnCl2 (d) and FeCl2 (e).
350 10
a Zn(OH)2 Zn b Mn3O4 Mn c Fe3O4 Fe
9
300 □ ZnO
8
250 7
GDEx GDEx GDEx
intensity [a.u.]
intensity [a.u.]
intensity [a.u.]
6
200
5
□
150 □□ □
□ □ □ 4
0 0
15 20 25 30 35 40 45 50 55 60 65 70 15 20 25 30 35 40 45 50 55 60 65 70 15 20 25 30 35 40 45 50 55 60 65 70
2 [] 2 [] 2 []
Figure 6.5: XRD patterns for AT and GDEx for solutions with ZnCl2 (a),
MnCl2 (b) or FeCl2 (c). Only for GDEx of ZnCl2 a phase mixture is observed
and the peaks of ZnO are indicated with a -symbol for clarity.
in both the AT and GDEx case (figure 6.5). However, in the GDEx case, some of
the peaks corresponded to the dehydrated ZnO phase [180] (, figure 6.5a). The
XRD pattern for manganese could be identified as Mn3 O4 [166, 181] for both
the AT and GDEx cases (figure 6.5b). The XRD pattern for iron was identified
as that of FeO [182–184] in the case of AT and as that of Fe3 O4 [182, 185] in the
case of GDEx (figure 6.5c). In the case of zinc, the composition was as expected,
with Zn(OH)2 as the main phase with or without some dehydrated ZnO. In
the case of manganese, the observation of Mn3 O4 rather than Mn(OH)2 is in
line with the color observations during the post-treatment washing steps. In
the case of iron, visual observations of the precipitates after processing already
hinted at the occurrence of a different composition. For AT and GDEx, this
deviation from the expected composition of Fe(OH)2 was attributed to different
phenomena. In the case of AT, the dehydrated FeO showed light orange color,
most likely caused by superficial oxidation to Fe2 O3 , which cannot be observed
with XRD. In the case of GDEx, Fe(II) is oxidized by the H2 O2 produced
during the process. This is in agreement with the readily observed dark brown
precipitates and the fact that no H2 O2 could be measured during the process.
Scanning Electron Microscopy (SEM) was used to visualize the size and
morphology of the precipitates (Figure 6.6). In all cases, the precipitate was
highly agglomerated. For the cases of Zn and Mn, the sizes and morphology
of the particles processed by GDEx differ little with those processed by AT
(figure 6.6a and 6.6d for Zn, and figure 6.6b and 6.6e for Mn). In the case of
102 OXIDATION-ASSISTED ALKALINE PRECIPITATION OF NANOPARTICLES USING GAS-DIFFUSION
ELECTRODES
Zn, irregular sheetlike particles were identifiable, albeit difficult. In the case of
Mn, spherical nanoparticles were easily observed. These similarities are in line
with the other observations: similar precipitate color, similar composition, and
similar precipitation behavior observed by dynamic pH measurements. Except
for the fact that no H2 O2 could be measured in the case of GDEx of MnCl2 ,
it appears that the GDEx and AT processes are equivalent in the cases of the
synthesis of Zn(OH)2 and Mn3 O4 nanoparticles, within the confines of our
process conditions. Given that AT is used as a benchmark process for alkaline
precipitation and GDEx is equivalent to AT for our solutions of MnCl2 and
ZnCl2 , GDEx is considered only to elicit alkaline precipitation in these cases,
and not oxidation-assisted alkaline precipitation. The fact that no H2 O2 could
be measured in the case of Mn is not entirely clear, but it might be explained
by the effective catalytic decomposition of H2 O2 on Mn(OH)2 [186].
In contrast to the Zn and Mn cases, for the case of Fe, the particles processed
by AT and GDEx differed significantly in size and morphology (figure 6.6c and
6.6f). While the FeO particles synthesized with AT are tile-shaped and with a
long-axis length in the order of 200 nm (figure 6.6c), the Fe3 O4 nanoparticles
synthesized by GDEx are spherical with a diameter in the order of 20 nm or
lower(figure 6.6f). These particles are only just visible by SEM (inset figure 6.6f).
This large size and morphology difference is a strong indication for the fact that
GDEx elicits oxidation-assisted alkaline precipitation, rather than only alkaline
precipitation. This is in line with our expectations, given our experience with
the Ox-AP of Fe2+ in a chemical Y-junction reactor [164] and the controllable
synthesis of iron oxide nanoparticles with GDEx [169]. In our exploratory
study on AP and Ox-AP [164], the size decrease between both processes was
assigned to improved condensation (olation and oxolation) kinetics as the result
of oxidation of the metal ion during the precipitation process. Nevertheless,
such a large size difference between alkaline precipitation and oxidation-assisted
alkaline precipitation was not yet observed. One possible explanation for this
would be the fact that here, Ox-AP and AP result in precipitates of species with
other compositions, while in the previous work [164] the precipitates had the
same composition. Species with different compositions have different solubility
products. Because Fe3 O4 (Ksp,Fe3 O4 = 10−108.6 [187]) has a much lower solubility
than FeO (Ksp,FeO = 10−14.5 [188]), the supersaturation in the GDEx process
will be significantly higher than in the AT process, as a result of the oxidation
of Fe(II) to Fe(III). It has long been established from classic nucleation theory
that higher supersaturation results in smaller (nano)particles [135, 137, 139].
Therefore, the large size difference, between the nanoparticles synthesized by AT
(eliciting AP) and the nanoparticles synthesized with GDEx (eliciting Ox-AP),
could be explained by the combined effect of the improved condensation kinetics
and the increased supersaturation in the GDEx process.
RESULTS AND DISCUSSION 103
a Zn b Mn c Fe
50 nm
Figure 6.6: SEM images of the precipitates synthesized with AT (a, b, c) and
GDEx (d, e, f) for the cases of ZnCl2 (a, d), MnCl2 (b, e), and FeCl2 (c, f).
104 OXIDATION-ASSISTED ALKALINE PRECIPITATION OF NANOPARTICLES USING GAS-DIFFUSION
ELECTRODES
4 Conclusions
Accumulation of ORR
reaction products in porous
gas-diffusion electrodes
105
106 ACCUMULATION OF ORR REACTION PRODUCTS IN POROUS GAS-DIFFUSION ELECTRODES
®
and everything related to MATLAB . X. Dominguez Benetton and J. Fransaer
helped with funding, discussions, and proof-reading.
This chapter is prepared for submission to Journal of Electrochemical Society as:
Eggermont S.G., Fransen S., Dominguez Benetton X., Fransaer J. Accumulation
of ORR reaction products in porous gas-diffusion electrodes.
ACCUMULATION OF ORR REACTION PRODUCTS IN POROUS GAS-DIFFUSION ELECTRODES 107
Abstract
In this work, the accumulation of OH− and HO− 2 in a porous air-fed gas-diffusion
electrode was investigated. An analytical steady-state model was derived for the
concentration profiles in the GDE and an analytical time-dependent model was
derived for the total accumulation of OH− and HO− 2 in the GDE. Measuring the
bulk electrolyte pH in a one-compartment electrochemical flow-cell reactor, it
was possible to experimentally determine the accumulation of these species in the
GDE. Other experiments were carried out to determine the penetration depth of
the electrolyte in the GDE, the efficiency of hydrogen peroxide production, and
the mass-transfer coefficients in the flow-cell reactor. Using these parameters,
the model and the experimental data of the pH were fitted to obtain a value for
the effective diffusion coefficient of OH− and HO− 2 in the GDE. The analytical
model is limited by its assumption that no migration of OH− and HO− 2 occurs.
This fact, and the occurrence of an undetermined parasitic side-reaction in the
GDE, were considered to be the two main contributing factors for mismatch
between the model and the experiments. Overall, this work succeeded in
demonstrating the existence of and quantifying the accumulation of species
in porous air-fed gas-diffusion electrodes, as well as give an estimate of the
concentration profile of OH− and HO− 2 in such porous electrodes.
108 ACCUMULATION OF ORR REACTION PRODUCTS IN POROUS GAS-DIFFUSION ELECTRODES
1 Introduction
alkaline : O2 + H2 O + 2e−
HO−
2 + OH
−
(E◦ = −0.065 V) (7.4)
Several other ORRs exist [63], for example, the further reduction of HO− 2 to
OH− , but in general, the most stable generated species are OH− , H2 O2 , and
HO−2 . These species are the focus of this work.
H+ + HO−
2
H2 O 2 (7.6)
MODELING THE GDE IN ONE SPATIAL DIMENSION 109
In practice, far from the equilibrium, only one of the species is dominating.
In the case of H2 O, below pH 7, the dominant species is the H+ ion, while
above pH 7, the dominant species is the OH− ion. In the case of H2 O2 , the
dominant species is H2 O2 , below pH 11.62, and the HO− 2 ion, above pH 11.62
(pKH2 O2 = 11.62 [20]). This has interesting consequences for the concentrations
of these species in solution. For example, in a solution of low pH, when OH−
is added, the concentration of OH− will barely increase (in absolute terms),
and instead, the H+ concentration will decrease with an amount (practically)
equivalent to the added amount of OH− ions. Similarly, adding HO− 2 to a
solution of low pH will hardly change the HO− 2 concentration and will instead
increase the H2 O2 concentration, while decreasing the H+ concentration with
amounts (practically) equivalent to the added amount of HO− 2 ions.
Carrying out GDEx, we discovered that the production rate of OH− (i.e., the
removal rate of H+ ) was below the predicted rate (Chapter 3). This is odd
because any electron should lead to an equivalent decrease in the amount of
H+ (due to the water equilibrium). One could argue that part of the electrons
would produce the negative HO− 2 species, which is correct, but as discussed
before, at sufficiently low pH, the HO−2 completely protonates, removing H
+
nonetheless.
We hypothesize that this apparently-lower production rate is caused by the initial
accumulation of OH− in the pores of the GDE. Initially, this work aimed to
model this accumulation and verify the model with experimental data. However,
obtaining realistic values for all of the model parameters proved difficult. Instead,
using the model and empirical data of all but one modeling parameter, the last
modeling parameter could be estimated: the effective diffusion coefficient of the
accumulating species. Therefore, in addition to our overarching work, this work
demonstrates a new approach for estimating the effective diffusion coefficient of
OH− and HO− 2 in air-fed porous gas-diffusion electrodes. Adaptations of the
method are believed to be useful for obtaining effective diffusion coefficients of
other species in other porous electrodes and reactor designs.
(i.e., an exact 3-dimensional model is unnecessary). The same holds for the
diffusion layer at the surface of the GDE. Figure 7.1 demonstrates a schematic
representation of the GDE and the diffusion layer.
𝜕𝑐𝑖 𝑤𝐺𝐷𝐸 ,𝑡
𝜕𝑐𝑖 0, 𝑡 𝜖𝜓𝒟𝑖 = 𝑘𝑖 𝑐𝑖 𝑤𝐺𝐷𝐸 , 𝑡 − 𝑐𝑖𝑏 𝑡
=0 𝜕𝑥
𝜕𝑥
wetted active carbon layer
gas−diffusion layer
diffusion layer
of GDE
𝜕𝑐𝑖 𝑥, 𝑡 𝜕 2 𝑐𝑖 𝑥, 𝑡
− 𝒟𝑖 = 𝑃𝑖 − 𝑅𝑖,𝑗
𝜕𝑡 𝜕𝑥 2
𝑗
𝑥=0 𝑥 = 𝑤𝐺𝐷𝐸
Figure 7.1: Schematic figure of the model geometry indicating the governing
equation and the boundary conditions in the porous active carbon layer of the
GDE. The curve demonstrates the shape of the expected profile of the OH−
concentration.
with Vmetal [m3 ] the volume of the stainless steel mesh, Vcarbon [m3 ] the volume
of the carbon and PTFE in the porous media, Vphob [m3 ] the volume of the
hydrophobic pores, Vphil [m3 ] the volume of the hydrophilic pores filled with
liquid electrolyte, wGDE [m] the electrolyte penetration depth in the GDE,
and Ageo [m2 ] the total geometric electrode surface area. Because only the x
dimension is considered in the model, it is necessary to define an average surface
MODELING THE GDE IN ONE SPATIAL DIMENSION 111
through which species in solution can move. This is the average surface area of
the hydrophilic pores Aphil , projected on the yz-plane. Aphil can be related to
the total geometric electrode surface area Ageo [m2 ], which is defined as:
with Ametal [m2 ] the average projected surface area of the stainless steel mesh,
Acarbon [m3 ] the average projected surface area of the carbon and PTFE in the
porous media, Aphob [m3 ] the average projected surface area of the hydrophobic
pores, Vphil [m3 ] the average projected surface area of the hydrophilic pores
filled with liquid electrolyte. The hydrophilic porosity of the GDE is defined as:
Vphil
= (7.9)
Vphob + Vphil + Vcarbon
The portion of the electrode covered by porous carbon/PTFE (i.e., the total
volume of the electrode minus the mesh), the porous media coverage, is defined
as:
Vphob + Vphil + Vcarbon
ψ= (7.10)
Vtot
The hydrophilic pore volume is thus related to the total volume via the following
relationship:
Vphil = ψVtot (7.11)
A similar calculation could be made for the surface area, however, because of
the reduction to 1 dimension, the equivalent for the surface area can be written
as:
Aphil = ψAgeo (7.12)
The model assumes the electrolyte in the active layer to be stagnant, without
convective mass transport. In addition, the model assumes that the electrolyte
is well-supported with ions that do not participate in the (electro)chemical
reactions, as such the current is considered to be carried only by the ions in the
supporting electrolyte. In a well-supported electrolyte, the potential gradient
in the electrolyte is small, electromigration can be neglected for species with a
low concentration compared to the concentration of the ions of the supporting
electrolyte and the Nernst-Planck and Poisson equations can be simplified,
reducing the problem to that of diffusion-only [189]. The model considers 4
species: H+ , OH− , HO− 2 , and H2 O2 . The general 1D mass-balance equation for
a species i in the porous GDE can be stated as:
P
Ri,j
∂ci ∂ 2 ci Pi j
− Di 2 = − (7.13)
∂t ∂x Vphil Vphil
112 ACCUMULATION OF ORR REACTION PRODUCTS IN POROUS GAS-DIFFUSION ELECTRODES
with ci [mol m−3 ] the concentration of species i and i equal to H+ , OH− , HO− 2,
or H2 O2 , Di [m2 s−1 ] the effective diffusion coefficient of species i, P i [mol
s−1 ] the production rate of species i and
P
Ri,j the sum of the individual
j
homogeneous chemical reaction rates Ri,j of species i with species j. The
species-specific production
Pterm Pi can be written as a function of the total
production term Ptot = Pi [mol s−1 ], which is a function of the applied
i
current Iapp [A]:
ηi Iapp
Pi = ηi Ptot = (7.14)
F
with F Faraday’s constant and ηi [-] the efficiency factor that expresses the
percentage of the current that is used to produce species i. Looking more closely
to the reactions in equations (7.1), (7.2), (7.3), and (7.4), the maximum value
of ηOH− is 1, because if only the 4-electron ORR occurs, each electron is used to
produce OH− . The maximum value of ηH2 O2 or ηHO− is 0.5 because, if only the
2
acidic or the alkaline 2-electron ORR occur, each electron only yields 0.5 H2 O2
or HO− 2 . Given that in reality a linear combination of the 4-electron ORR, the
2-electron ORR, and possibly some other reactions takes place, the values of ηi
will need experimental validation.
In light of this, the model assumes that the current produces only OH− and
HO− 2 with a certain efficiency and that H
+
and H2 O2 do not participate in
the electrochemical reaction and therefore have no production term. This is a
reasonable assumption if the concentration of H+ is low. However, two chemical
equilibrium reactions are allowed to occur (equation (7.5) and (7.6)), which
couple the concentrations of H+ , OH− , HO− 2 , and H2 O2. These equations
imply
the
existence
of two homogeneous reaction terms Ri,j : RH+ ,OH− = RH2 O and
RH+ ,HO− = RH2 O2 . Equation (7.13) thus specifies a system of four equations
2
∂ HO−
∂ 2 HO−
ηHO− I (x) (7.15)
2 2
− DHO− = 2
− RH2 O2
∂t 2 ∂x2 ψAgeo wGDE F
∂ [H2 O2 ] ∂ 2 [H2 O2 ]
− DH2 O2 = RH2 O2
∂t ∂x2
KH2 O = H+ OH−
(7.16)
HO−
+
H 2
KH2 O2 = (7.17)
[H2 O2 ]
In addition to the equilibrium equations, solving the differential equations
requires an initial condition and two boundary conditions for each species.
Figure 7.1 shows a scheme with the model and the boundary conditions. The
initial condition is:
ci (x, 0) = ci,0 (7.18)
with ci,0 [mol m−3 ] the initial concentration of species i everywhere in the liquid.
The first boundary condition is a zero-flux boundary condition. It is defined
at the boundary between the liquid phase and the gas phase (i.e., the internal
wetted boundary in the GDE) because the chemical species are assumed not to
move into the gas phase. The first boundary condition (at x = 0, figure 7.1) is
therefore defined as:
∂ci (0, t)
− Di =0 (7.19)
∂x
The second boundary condition is defined at the surface of the electrode facing
the bulk of the flow-cell reactor (at x = wGDE , figure 7.1). At this boundary,
the number of species coming from the pores needs to be equal to the amount
of species going into the bulk solution:
with Npi [mol m−2 s−1 ] the flux with respect to the pore surface and Ndl i [mol
m−2 s−1 ] the flux term with respect to the total surface contacting the diffusion
114 ACCUMULATION OF ORR REACTION PRODUCTS IN POROUS GAS-DIFFUSION ELECTRODES
∂ci (wGDE , t)
Nip (wGDE , t) = −Di (7.21)
∂x
Nidl (wGDE , t) = ki ci (wGDE , t) − cbi (t)
(7.22)
with ki [m s−1 ] the mass transfer coefficient of species i, and cbi [mol m−3 ]
the concentration in the bulk solution. Remark that, although no convective
mass transfer is occurring in the porous electrode, equation (7.22) introduces a
dependency on the convective mass transport in the flow-cell reactor via the
mass transfer coefficient. Using previous definitions (equations (7.12), (7.20),
(7.21), (7.22)), the second boundary condition simplifies to:
∂ci (wGDE , t) ki
ci (wGDE , t) − cbi (t)
=− (7.23)
∂x ψDi
With the initial condition and boundary conditions defined, the system
of equations (7.15) could be solved numerically but it needs additional
simplification to be solved analytically. An analytical model is preferred
over a numerical model in this work because the focus is on determining the
contributions of the different key parameters to the accumulation of chemical
species in the GDE. This is more easily achieved with an analytical model than
with a numerical model.
with ci,0 [mol m−3 ] the initial bulk concentration as well as at any position in
the GDE. Moreover, the chemical equilibria introduce a large deal of complexity.
Assuming working conditions away from the water equilibrium (i.e., below H+
concentrations of 10−4 mol m−3 or thus above pH = 7), the water reaction
would have a negligible effect on the values of the OH− concentration and
the reaction terms could be neglected. The H+ concentration could then be
calculated from the H2 O equilibrium alone. As a consequence of working at low
initial H+ concentration, and assuming no H2 O2 to be present initially, also
MODELING THE GDE IN ONE SPATIAL DIMENSION 115
KH2 O HO−
2
[H2 O2 ] =
KH2 O2 OH−
Using equations (7.25) and the boundary conditions equations, (7.19) and (7.23),
the general analytical solution for the steady-state concentration of i equal to
OH− or HO− 2 is found to be given by:
" 2 #!
2
Pi wGDE ψ 1 wGDE x
ci,ss (x) = ci,0 + + 1− (7.26)
Vphil ki 2 Di wGDE
The concentration of species in the GDE is the sum of an initial term, a reactor
design term (i.e., defined by the mass-transfer coefficient ki ) and a GDE design
term (see figure 7.2). The initial term defines the initial concentration in the
pores. The reactor design term defines the concentration that can be reached
at the surface of the electrode given a certain mass-transfer coefficient. The
GDE design term determines how high the concentration can increase above
this surface concentration value, given the electrolyte penetration depth of the
GDE and the effective diffusion coefficient. In addition, both the reactor design
term and the GDE design term depend linearly on the Faraday-compensated
Pi η Iapp
current in the pores Vphil = Viphil F in front of the brackets. The concentration
in the GDE is a quadratic function of the spatial variable x with a maximum
value at the back of the wetted part of the active carbon layer.
116 ACCUMULATION OF ORR REACTION PRODUCTS IN POROUS GAS-DIFFUSION ELECTRODES
2 2
𝑃𝑖 𝑤𝐺𝐷𝐸 𝜓𝜖 1 𝑃𝑖 𝑤𝐺𝐷𝐸 𝑤𝐺𝐷𝐸
𝑐𝑖 𝑥 = 𝑐𝑖,0 + + 1−
𝑉𝑝ℎ𝑖𝑙 𝑘𝑖 2 𝑉𝑝ℎ𝑖𝑙 𝒟𝑖 𝑥
wetted active carbon layer
gas−diffusion layer
diffusion layer
2
1 𝑃𝑖 𝑤𝐺𝐷𝐸
of GDE
2 𝑉𝑝ℎ𝑖𝑙 𝒟𝑖
𝑃𝑖 𝑤𝐺𝐷𝐸 𝜓𝜖
𝑐𝑖,0 𝑉𝑝ℎ𝑖𝑙 𝑘𝑖
𝑥=0 𝑥 = 𝑤𝐺𝐷𝐸
Figure 7.2: Schematic figure of the model geometry indicating the different
contributions to the steady-state concentration of a produced species in the
GDE
with τi [s] the characteristic diffusion time in the GDE and Shi [-] a Sherwood-like
number equal to:
w2
τi = GDE (7.28)
Di
ki wGDE
Shi = (7.29)
ψDi
The characteristic diffusion time τi of chemical species i is an indicative measure
of the time necessary to diffuse over a distance equal to the penetration depth.
The dimensionless number Shi has the appearance of a Sherwood number and
is equal to the ratio of the convective mass transfer in the bulk and the diffusive
mass transfer term in the GDE. If it is large, it indicates that convective mass
transfer in the bulk is much faster than diffusive mass transfer in the GDE.
Because the convective mass transfer term and the diffusive mass transfer term
relate to different locations (i.e., the bulk and the GDE, respectively), Shi is not
an authentic Sherwood and is therefore defined as a Sherwood-like number. The
concentrations in the bulk depend linearly on the production rate Pi and the
characteristic diffusion time τi , and depend inversely on the pore volume Vphil
MODELING THE GDE IN ONE SPATIAL DIMENSION 117
and the Sherwood-like number Shi (to a lesser extent). Integrating equation
(7.27) over the hydrophilic pore volume of the GDE gives an expression for
the steady-state accumulation of species in the pores of the GDE that contain
electrolyte:
Rw
nGDE
RRR
i,ss = c (x) dV = ψAgeo 0 GDE ci,ss (x) dx
Vphil i,ss
(7.30)
1 1
= ci,0 Vphil + Pi τi +
Shi 3
With this notation, it is clear that higher production (i.e., higher applied
current) and higher characteristic diffusion times (i.e., thicker electrodes or
lower effective diffusion coefficients) result in more accumulation in the electrode.
Additionally, the convective behavior of the flow-cell reactor becomes increasingly
less important for values of Shi above 3.
∂ci ∂ 2 ci Pi
− Di 2 = (7.31)
∂t ∂x Vphil
This equation resembles the heat equation with a constant source term and is
often solved with separation of variables. Although this approach can potentially
give the exact solution, its form is needlessly complex and does not comply
with a vision to provide a straightforward solution. Because the very short
time range is not of special interest, but more the overall behavior, and more
specifically the times approaching the steady-state, Laplace transforms were
opted to solve the equation instead. As such, while solving the equations in the
Laplace domain, it is possible to solve the equations for the asymptotic behavior
(i.e., the limiting situation when s approaches zero, more or less equivalent to
118 ACCUMULATION OF ORR REACTION PRODUCTS IN POROUS GAS-DIFFUSION ELECTRODES
3 Experimental
®
demineralized water. This solution was added to 0.050 L of 0.05 M H2 SO4
(Titrisol ) and the required amount of demineralized water to complete a 0.500
L volumetric flask. A copper salt was chosen because back-scattered electrons
in a scanning electron microscope give good contrast with the carbon phase in
the active carbon layer of the GDE.
0.0001 M HCl) were made by diluting the solution of 2 M KNO3 with 0.01 M
HCl with the 2 M KNO3 stock solution in the correct ratios. Using pH values
of 2, 2.301, 2.602, 3, and 4 to calibrate the 2 M KNO3 solutions of respective
HCl concentrations of 0.01 M, 0.005 M, 0.0025 M, 0.001 M, and 0.0001 M, it
is possible to artificially force the pH meter to measure proton concentrations,
rather than proton activities, in the highly supported electrolyte.
®
reactor. The flow-cell reactor compartment volumes are 0.010 L. The cathode
was a VITO CoRE gas-diffusion electrode. The anode was a flat platinized
tantalum plate. The geometric surface area of the anode and cathode were
10 cm2 . The anolyte and catholyte compartment are separated with an anion
®
exchange membrane (FUMASEP FAP-4130-PK). This type of membrane and
a sufficiently high chloride concentration allow a closed electric circuit. The
anolyte and catholyte recirculation reservoirs were 250 mL DURAN laboratory
bottles, with additional flask openings. The cap has openings to allow argon
purging. The reservoir solutions were stirred with a magnetic stirring bar of
3 cm on an IKA RCTbasic S1 stirrer at rotation level 7 (at rotation rates of
approximately 700-800 rpm). A double-headed peristaltic pump (Cole-Palmer
MasterflexTM L/s 7551-00 with Easyload II 77200-60 pump heads) was used to
EXPERIMENTAL 121
front side
electrolyte side
active carbon layer stainless steel mesh
model
®
Figure 7.3: SEM photographs of the front view and cross-section of the VITO
CoRE electrode. The front side view demonstrates the stainless steel mesh
and the active carbon phase. The cross-section demonstrates the PTFE foam
layer (lighter phase), the carbon/PTFE active layer (darker phase) and the
stainless steel mesh (very light disks). The model is limited to the carbon/PTFE
active layer as indicated by the white line.
122 ACCUMULATION OF ORR REACTION PRODUCTS IN POROUS GAS-DIFFUSION ELECTRODES
pump the solutions in anolyte and catholyte around at 500 mL min−1 through
90 cm tubes with a diameter of 3.2 mm (Cole-Palmer MasterflexTM VersilonTM
L/s 06475-16). For 500 mL min−1 , the average residence time is 2.4 seconds
in the flow-cell and 0.8 seconds in a connection tube. The GDE was fed with
air at its hydrophobic gas-diffusion layer side with an airflow of 0.100 L min−1 .
The air back-pressure of approximately 30 mbar was provided by putting the
exit tube of the gas channel into a 30 cm high water column. All experiments
were carried out at room temperature (i.e., 22 ◦ C ± 2 ◦ C).
® ®
every 5 seconds. The H2 O2 concentration was measured every minute using a
QUANTOFIX Relax with QUANTOFIX Peroxide 25 and QUANTOFIX ®
Peroxide 100 test strips. 0.100 L solutions with 0.050 M, 0.02 M and 0.01 M
initial proton concentration were respectively treated with current densities
of 500 A m−2 , 200 A m−2 , and 100 A m−2 . As such, each experiment could
be carried out for 1000 seconds before reaching high values of the pH (due
to the co-production of OH− in the air-fed GDE). The slope of the measured
d [H2 O2 ]
concentration of H2 O2 as a function of time [mol s−1 ], is equal to the
dt
H2 O2 production rate:
d [H2 O2 ]
PH2 O2 = (7.34)
dt
EXPERIMENTAL 123
Because of the sufficiently low pH in the bulk, all HO− 2 reaching the bulk would
react to H2 O2 as a result of the equilibrium reaction (equation (7.6)). Therefore
the production rates of H2 O2 and HO− −
2 are equivalent. The HO2 production
efficiency ηHO− is then calculated as follows:
2
PHO− PH2 O2 F
ηHO− = 2
= (7.35)
2 Ptot Iapp
I− −
3 + 2e
3I
−
(7.36)
The mass-transfer coefficient of triiodide was then calculated from the limiting
current as [190]:
q
Ilim
kIq− = (7.37)
3 2Fc∞,I− Ageo
3
124 ACCUMULATION OF ORR REACTION PRODUCTS IN POROUS GAS-DIFFUSION ELECTRODES
with kIq− [m s−1 ] the convective mass-transfer coefficient of triiodide for a given
3
q
flow rate q [m3 s−1 ], Ilim [A] the measured limiting current for a given flow
rate q and c∞,I− [mol m−3 ] the bulk concentration of triiodide. To recalculate
3
the mass-transfer coefficient of triiodide to OH− and HO−
2 , the definition of
the dimensionless average Sherwood number Shavg [-] of the electrochemical
flow-cell can be used:
kiq de kIq− de q
kOH− de
q
kHO− de
Shavg = = 3
= = 2
(7.38)
Di DI− DOH− DHO−
3 2
with kiq the mass-transfer coefficient of species i for a given flow rate q, de [m]
a characteristic length of the flow-cell reactor, and Di [m s−1 ] the diffusion
q q
coefficient of species i. kOH− and k
HO−
can be estimated for different flow rates
2
using the measured values of kIq− and literature values for DI− = 9.60 10−10
3 3
m2 s−1 [190], DOH− = 4.56 10−9 m2 s−1 [191] and DHO− = 1.71 10−9 m2 s−1
2
(roughly estimated by equalizing it to the value ofDH2 O2 from [192] due to the
lack of reliable sources for HO−
2 ).
To estimate the accumulation of OH− in the pores of the GDE, the pH of the
bulk electrolyte was measured in a one-compartment configuration. Experiments
were conducted with 0.050 L solutions of the 2 M KNO3 stock solution for
different values of the current densities Iapp /Ageo of 50 A m−2 , 100 A m−2 , 200
A m−2 , 350 A m−2 , and 500 A m−2 . Because of the electrolyte choice of 2 M
KNO3 , it is assumed that no side reactions occurred (e.g., Cl2 evolution) and
therefore assumed a 100% efficiency for the production of H+ at the anode (i.e.,
every electron that passes through the electrochemical cell yields a proton at
the anode). At the GDE, the cathode, it is assumed that the ORRs produce
OH− and HO− 2 . Given a sufficiently low pH in the bulk electrolyte (below
pH 7), every such species would subsequently remove an H+ as the result
of their equilibrium reactions (equations (7.5) and (7.6)). The removal and
production rates of protons at anode and cathode, respectively, are theoretically
equivalent. Because of the one-compartment configuration, given the hypothesis
of accumulation of species in the GDE, an initial mismatch would exist between
the number of protons reaching the bulk via production at the anode and the
amount protons exiting the bulk via the equilibrium reactions with the OH− and
HO− 2 ions produced at the cathode (GDE). This mismatch would be observed
by a decrease in the pH (increase of the proton concentration) in the bulk
RESULTS AND DISCUSSION 125
−5
500
kOH − [10 m s−1 ] 37.5
−5
500
kHO − [10 m s−1 ] 14.1
2
DOH− [10 −10
m s2 −1
] 2.36 2.31 3.05 2.87 3.09
DOH−
δD = [-] 2.75
DHO−
2
E [V] 0.041 0.077 0.132 0.185 0.242
0.25
0.15
0.1
0.05
0
10 100 1000
𝐼𝑎𝑝𝑝
current density [A m-2]
𝐴𝑔𝑒𝑜
The average efficiency for the production of HO− 2 , ηHO− , was calculated from
2
the bulk H2 O2 concentration measurements during galvanostatic experiments
in a two-compartment electrochemical flow-cell reactor using an air-fed GDE.
Figure 7.4 plots this hydrogen peroxide efficiency as a function of the current
density. The relationship is very well represented by the function ηHO− =
2
0.138 log (Iapp /Ageo )−0.190. The relationship between the production efficiency
and the current density is most probably the result of a combination of factors.
Such factors might be a higher Faradaic current efficiency and a higher degree
of migration of the HO− 2 (in insufficiently well-supported electrolytes), at higher
current densities, therefore increasing the outward flux of these species. The
latter might have an additional effect because a higher outward flux might
decrease the amount of subsequent reduction of HO− −
2 to OH in the GDE.
Figure 7.5 shows the penetration depth of electrolyte by showing the position of
CuSO4 salt crystals in the active carbon/PTFE layer of the GDE. The occurrence
RESULTS AND DISCUSSION 127
50 m
200 m
of CuSO4 salt crystals, at the interface of the porous PTFE gas-diffusion layer
and the active carbon/PTFE layer, indicate the complete penetration of the
active layer by the electrolyte. This estimates the penetration depth as the
total depth of the active layer (i.e., on average approximately 300 µm).
128 ACCUMULATION OF ORR REACTION PRODUCTS IN POROUS GAS-DIFFUSION ELECTRODES
9
8
Figure 7.7 shows the observed amount of protons in the bulk nbH+ as a function
of time. nbH+ [mol] is calculated from the measured pH values as:
0.4
500 mA
A m-2
slope = rpar
350
350 mA
A m-2
A m-2
200 mA
[10-3 mol]
A m-2
100 mA
0.3
5050mA
A m-2
protons in the bulk
0.2
0.1
slope determination
0
0 500 1000
time t [s]
Figure 7.7: The observed amount of protons in the bulk electrolyte nbH+ as a
function of time for different values of the applied current density Iapp /Ageo .
with Vbulk [L] the volume of bulk electrolyte (0.05 L in the experiments). From
the assumptions, this would be equal to the total amount of species accumulated
in the GDE nGDEtot,ss [mol]:
nGDE nGDE
P
tot,ss (t) = i,ss (t) (7.40)
i
with nGDE
i,ss [mol] the accumulated amount of species i. The curves in figure
7.7 clearly show that a constant value is not reached within the observed time-
frame. Additional experiments were carried out to confirm that a constant
value steady-state was never reached, even for experiments up to 2000 seconds
(appendix F). However, the slope of the curve remained constant for times
higher than 800 seconds. Such constant sloped steady-state would be in line
with a consistent cathode H+ removal efficiency lower than the assumed 100%.
This possibly indicates a parasitic reaction in the GDE consuming electrons not
resulting in the removal of a proton.
To confirm the existence of a parasitic reaction, a linear scan voltammetry at
1 mV s−1 in the 2 M KNO3 stock solution was carried out. Figure 7.8 shows
the voltammogram, which demonstrates the occurrence of a parasitic reaction
resulting in an observable current before reaching the potential at which the
ORR starts. This parasitic current depends more or less linearly on the applied
130 ACCUMULATION OF ORR REACTION PRODUCTS IN POROUS GAS-DIFFUSION ELECTRODES
0
𝐼𝑝𝑎𝑟
= 81 𝐸 − 25.1
𝐴𝑔𝑒𝑜
-200
500 A m-2
-800
-1000
-1200
-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2
GDE potential E vs. SHE [V]
Figure 7.8: Linear scan voltammogram for a scan rate of 1 mV s−1 . The
different values for the applied current density Iapp /Ageo are indicated with
diamonds. The black line indicates the slope of the line for the current not
originating from the ORRs.
potential. The observed amount of protons in the bulk nbH+ is thus not equal to
the accumulation of species in the GDE, but a combination of the accumulation
of the species in the GDE and a linear steady-state part, caused by a parasitic
reaction:
Ipar
nbH+ (t) = nGDE GDE
ss,tot,exp (t) + rpar t = nss,tot,exp (t) + t (7.41)
F
with rpar = Ipar /F [mol s−1 ] a constant rate of the parasitic reaction and Ipar
[A] the parasitic current (figure 7.8). The accumulation of species in the GDE
nGDE
tot,ss,exp can then supposedly be obtained by subtracting the linear part from
the total observed amount of protons in the bulk nbH+ (figure 7.10). The data
obtained by this treatment now represents the total accumulation of species in
the GDE, equal to the sum of the accumulation of OH− and the accumulation
of HO− 2 ions:
nGDE
ss,tot,exp (t) = nGDE
ss,OH− ,exp
(t) + nGDE
ss,HO−
(t)
2 ,exp
1
GDE GDE
(7.42)
= α n ss,OH− ,mod
(t) + n ss,HO− ,mod
(t)
2
to handle the miss-match between the model and experiment which occurs as a
result of additional phenomena that are not accounted for in the model (e.g.,
migration of species, etc.). For values of α = 1, the modeled and experimental
accumulation would fit perfectly and the model would take into account all
occurring phenomena. Other values would indicate the occurrence of phenomena
not taken into account.
Both accumulation terms depend heavily on the effective diffusion coefficient of
the respective species (equation (7.32)). First, a parameter δD is defined as the
ratio of the two effective diffusion coefficients:
DOH−
δD = (7.43)
DHO−
2
Now, the model can be fitted using 3 fitting parameters. The first parameter
is the effective diffusion coefficient of OH− , DOH− , the second parameter is
the ratio of the effective diffusion coefficients of OH− and HO− 2 , δD , and the
third fitting parameter is the correction factor α. The procedure fits α and
DOH− in the model to experimental accumulation data for the 5 previously
discussed current densities. This is done for different values of δD (i.e., from 1
to 10). For each value of δD , the root mean square (RMS) of the residuals of
each current density experiment is calculated and summed up. These sums are
plotted as a function of δD to find the most appropriate value for δD . Figure
7.9 demonstrates these sums as a function of δD , demonstrating the best fit
for a value of δD = 2.75. The value for the ratio of the diffusion coefficients
found in the literature [191, 192] is equal to 2.67 (DOH− /DHO− = 4.56 10−9 m2
2
s−1 /1.71 10−9 m2 s−1 = 2.67). Therefore, this value for δD could be considered
reasonable. Figure 7.10 shows the treated experimental data from figure 7.7 as
the total accumulation in the GDE and compares it to the fitted model for a
value of δD = 2.75.
The fitted values of α and DOH− are plotted for the different current densities in
figure 7.11. The values of the effective diffusion coefficients (figure 7.11a) show
a slightly upwards trend with current density, but given the many assumptions,
it is probably safer to estimate the effective diffusion coefficient of OH− with
the average of these values. As such, the average effective diffusion coefficient of
OH− , DOH− , is equal to (2.73 ± 0.38) 10−10 m2 s−1 . Using δD and DOH− , it is
possible to estimate the average effective diffusion coefficient of HO−
2 to be equal
to (9.94 ± 0.14) 10−11 m2 s−1 . Using equation (7.29) and the experimentally
obtained values of kiq , wGDE , ψ and Di , the Sherwood-like numbers of OH−
and HO− 2 in the flow-cell at a flow rate q of 500 mL min
−1
can be estimated to
be ShOH− = 2902 and ShHO− = 2989, both significantly higher than 2 and 3,
2
demonstrating convection in the flow-cell reactor has very little influence on
the accumulation of species in the GDE.
132 ACCUMULATION OF ORR REACTION PRODUCTS IN POROUS GAS-DIFFUSION ELECTRODES
1.6
1.4
0.8
0.6
minimum at 2.75
0.4
0.2
0
0 1 2 3 4 5 6 7 8 9 10
D[-]
Figure 7.9: Sum of the 5 root mean squared (RMS) residuals of the fit
of 5 experimental and modeled accumulation curves as a function of δD =
DOH− /DHO− , the ratio of the effective diffusion coefficient of OH− to the
2
effective diffusion coefficient of HO−
2 . The lowest residuals are found for a ratio
of δD = 2.75.
The values of α show a significant dependence on the current density and are not
equal to 1 (figure 7.11b). This indicates the occurrence of phenomena not taken
into account by the model. It is clear that at higher currents, the corrected
accumulation (α = 0.3) is significantly smaller than 1. One very suitable
explanation is the fact that migration is not taken into account, something
increasingly important at higher current densities. This is especially true if the
concentration of the ions in the supporting electrolyte would have similar values
as the concentration of OH− or HO− 2 . To validate this, it is possible to calculate
the concentration profiles of OH− and HO− 2 in the GDE with equation (7.26)
using the values obtained from the experiments (table 7.1). This is shown in
Figure 7.12. In this case, the values of α were used as a correction factor to
the production term of equation (7.26). As such, the concentration profile is
corrected for the observed deviations for the accumulation in the GDE. From
figure 7.12 it is clear that for applied current density values between 100 A m−2
and 200 A m−2 , the maximum concentrations of OH− in the GDE surpasses
the 2 M mark, which is the concentration of the supporting electrolyte. In such
a case, the assumption that migration of these species is negligible is incorrect.
RESULTS AND DISCUSSION 133
0.25
500 mA
500 A m-2
0.1
0.05
0
0 200 400 600 800 1000
time t [s]
10 0.7
a b
eff. diffusion coeff. 𝒟OH-[10-10 m2 s-1]
0.6
correction factor [-]
0.5
0.3
1 0.2
0 200 400 600 0 200 400 600
𝐼𝑎𝑝𝑝 𝐼𝑎𝑝𝑝
current density [A m-2] current density [A m-2]
𝐴𝑔𝑒𝑜 𝐴𝑔𝑒𝑜
Figure 7.11: Values of the effective diffusion coefficient of OH− (a) and correction
factor α (b) as a function of the applied current density Iapp /Ageo
134 ACCUMULATION OF ORR REACTION PRODUCTS IN POROUS GAS-DIFFUSION ELECTRODES
Therefore, it is not surprising that the value of α differs more from 1 if the
applied current density increases.
The accumulation of OH− above concentrations of 1 M results in a noticeable
decrease in the apparent OH− production efficiency. This translates itself
in observable delays in the time-dependent pH increase during galvanostatic
experiments. When GDEx is carried out in batch mode, this can result in
a significant increase of the time to remove all metal ions (as precipitates
or nanoparticles). The additional time necessary results in a higher power
consumption per amount of removed metal ion or per amount of produced
nanoparticle, which is undesirable. However, because the accumulation reaches
a steady state when GDEx would be carried out in continuous mode, the
accumulation could be considered part of the so-called start-up time, which
from our experiments appeared to be around 800 seconds. As stated before, to
become an industrially relevant process, GDEx must be carried out in continuous
mode, therefore the initial accumulation causes little harm to the overall process
efficiency. In addition to the accumulation, certain parasitic electrochemical
reactions also decrease the production efficiency of OH− . In contrast to the
initial accumulation, such reactions continuously weigh on the process efficiency.
Therefore, to optimize the process, special attention must be paid to avoid such
parasitic reactions as much as possible.
Although the initial accumulation initially decreases the apparent production
efficiency, it also results in very high OH− concentrations in the GDE (figure
7.12). The resulting high pH causes a very stable work potential of the GDE
(figure 7.13) for a very large range of current densities. This is caused by the
fact that the reduction potential of O2 strongly depends on the pH. Because the
pH is high, it remains stable over time. A stable potential is beneficial, because
it needs little control and results in less electric losses.
RESULTS AND DISCUSSION 135
0 x→ wGDE
5
OH-
4.5 HO2-
1.5
0.5
0
0 100 200 300
position in active layer of GDE [m]
Figure 7.12: Concentration profiles of OH− and HO−2 in the active layer of the
GDE for different values of the applied current density Iapp /Ageo , calculated
from the experimentally obtained parameters and the values of the fitted model
parameters (see table 7.1).
0.2
500 A m-2
Series3
350 A m-2
Series4
Series5
200 A m-2
0.1
GDE potential E vs. SHE [V]
Series6
100 A m-2
Series7
50 A m-2
-0.1
-0.2
-0.3
0 500 1000
time t [s]
Figure 7.13: GDE potential as a function of time for different values of the
applied current density Iapp /Ageo .
136 ACCUMULATION OF ORR REACTION PRODUCTS IN POROUS GAS-DIFFUSION ELECTRODES
5 Conclusions
In this work, the accumulation of OH− and HO− 2 in a porous air-fed gas-diffusion
electrode was investigated. An analytical steady-state model was derived for the
concentration profiles in the GDE and an analytical time-dependent model was
derived for the total accumulation of OH− and HO− 2 in the GDE. Experimental
data was gathered for the time-dependent accumulation for different current
densities. These data suggested the existence of parasitic reactions besides
the desired oxygen reduction reactions (ORRs). This existence was verified by
linear scan voltammetry. The experimental data could be fitted to the modeled
accumulation to obtain an estimate of the effective diffusion coefficients of OH−
and HO− 2 . The best fits were obtained for a ratio of the diffusion coefficients of
OH− and HO− 2 equal to 2.75, close to what values found in literature. Good
fits were only obtained when a correction factor was used, which indicated the
limitation of the analytical model to incorporate all experimental phenomena.
One such phenomenon is the fact that migration was not taken into account.
Since the model predicts concentrations of OH− and HO− 2 equal to or higher
than the concentration of the ions of the supporting electrolyte, this assumption
is invalid at high current densities. Overall, this work demonstrates the existence
of accumulating species and quantifying this accumulation in porous air-fed gas-
diffusion electrodes. It was possible to estimate the effective diffusion coefficients
of these species in relatively simple experiments.
Chapter 8
General conclusions
The general aim of this dissertation was to discover the fundamental aspects
of gas-diffusion electrocrystallization (GDEx). To achieve this purpose (1) an
extensive exploratory study of the GDEx process was carried out to demonstrate
similarities and differences with alkaline titration, a well-known method for
nanoparticle synthesis in addition to several other more in-depth investigations
to further develop the understanding of the physicoelectrochemical processes
in the GDEx process. These investigations were (2) the mathematical and
experimental analysis of alkaline titration of metal ions to reveal the effect of
different system parameters on the time-dependent behavior of metals during
alkaline precipitation, (3) the investigation of the effect of oxidants on the
precipitation behavior of oxidizable metal ions during alkaline precipitation,
discovering the phenomenon of oxidation-assisted alkaline precipitation, (4)
the demonstration of this newly discovered phenomenon in the confines of
the GDEx process by producing and comparing nanoparticles from different
metal-containing solutions and (5) the mathematical and experimental analysis
of the accumulation of electrochemically produced species in the pores of the
porous gas-diffusion electrode to increase understanding of how the processes in
the porous electrode affect the observations in the bulk electrolyte.
The motivation to investigate GDEx came from preliminary experiments at
VITO that demonstrated the potential of GDEx to fastly produce crystalline
metal oxide nanoparticles from metal-containing waste streams, even for streams
with low concentrations of metal ions. The combination of synthesis and metal
removal or recovery is desirable given the fact that society relies increasingly on
the development of new functional materials [90] and environmental concerns are
put higher on the political agendas by every new election [89]. In addition to this,
137
138 GENERAL CONCLUSIONS
On the one hand, an essential part of the GDEx process has been demystified.
Fortunately, or unfortunately, depending on the scientist’s frame of mind, many
open questions and ideas remain, as well as a few unfinished projects that
deserve some attention in this concluding section.
Although all of the work, related to GDEx, presented in this dissertation was
carried out in a flow-cell reactor operating in recirculation mode, GDEx would
be more practical as a continuous process. The first steps toward this were made
together with my close colleague Rafael Prato, who demonstrated the usefulness
of the GDEx process to controllably synthesize iron oxide nanoparticles with
different characteristics [169]. This work of him focused on the effect of the
initial iron concentration. However, besides the initial metal ion concentration,
the background electrolyte, the flow rate, and the current-density are expected to
also have important effects on the characteristics of the synthesized nanoparticles.
Especially because of their effect on the supersaturation of the precipitating
species. Additionally, the flow rate and current-density will have a significant
effect on the relative amount of H2 O2 available in solution, the species responsible
for eliciting Ox-AP in the GDEx process and thus the root cause for the
formation of nanoparticles with smaller sizes than obtainable with traditional
alkaline precipitation.
GENERAL CONCLUSIONS 141
which would help to improve the agreement between model and experiment. It
would also provide the possibility to model the continuous process and provide
the first steps towards choosing the right process parameters in function of,
among other things, the desired supersaturation, and H2 O2 concentration in
the reactor.
Besides these unfinished projects closest to my own interests, many more
applications can be imagined. The potential for GDEx as a new green
electrochemical process is high, both from a removal and a synthesis perspective,
with many opportunities for GDEx as a sustainable alternative to existing
processes. It is my hope and conviction that some of the ideas and concepts
that were worked out in this PhD are useful in future work regarding the
GDEx process specifically, but the field of (green) chemical and electrochemical
synthesis more generally.
Appendix A
2+ 0.005
Cu (t) = 1029.4EISE (t)+3.3
29.4EISE (0)+3.3
(A.1)
10
with the multiplication factor 0.005/1029.4EISE (0)+3.3 a correction using the
firstly measured term to always have the correct initial concentration (necessary
143
144 MEASURING THE COPPER CONCENTRATION WITH A CU-ISE
-2.1
log [Cu2+] = 29.4 ECu-ISE +3.3
-2.3
-2.5
log [Cu2+]
-2.7
-2.9
-3.1
-3.3
-3.5
-0.23 -0.22 -0.21 -0.2 -0.19 -0.18
Cu-ISE potential vs. Hg/HgSO4 EISE [V]
Figure A.1: Calibration curve of the Cu-ISE, plotting the logarithm of the
known concentration of Cu2+ against the measured potential of the Cu-ISE.
-0.15 6
Cu concentration [Cu2+] [10-3 mol L-1]
-0.2 5
Cu-ISE potential EISE [V]
-0.25 4
-0.3 3
-0.35 2
-0.4 1
-0.45 0
0 500 1000 1500
time t [s]
Figure A.2: Calibration curve of the Cu-ISE, plotting the logarithm of the
known concentration of Cu2+ against the measured potential of the Cu-ISE.
Appendix B
#######################################################################################
##### This PHREEQC (.pqi) input file is developed by Sam Eggermont (2019). #####
##### It allows the calculation of the titration of a ZnCl2 solution with #####
##### a NaOH solution in a fixed amount of titration steps. It uses the #####
##### pitzer.dat database. In this input file the unavailable but necessary #####
##### additional phases and species are added from the Minteq.v4.dat database #####
##### This version is the adaptation for 100 mL 5 mM ZnCl with 0.5 M NaCl, #####
##### titrated with 20 mL of 0.1 M NaOH solution in 1222 steps. #####
##### #####
##### This file is easily adopted to other solutions allowing for different #####
##### metal ions, different anions, different salts, different titrant, etc. #####
#######################################################################################
#######################################################################################
##### Part 1: adding species that are not available in the Pitzer database #####
#######################################################################################
Zn+2 = Zn+2
log_k 0
Zn+2 + H2O = ZnOH+ + H+ ### This block is excluded for NM1 and included for NM2
log_k -8.997
145
146 PHREEQC INPUT FILE FOR MODELING PTCS OF ZN(OH)2
delta_h 55.81 kJ
-gamma 0 0
Zn+2 + 2H2O = Zn(OH)2 + 2H+ ### This block is excluded for NM1 and included for NM2
log_k -17.794
delta_h 0 kJ
-gamma 0 0
Zn+2 + 3H2O = Zn(OH)3- + 3H+ ### This block is excluded for NM1 and included for NM2
log_k -28.091
delta_h 0 kJ
-gamma 0 0
Zn+2 + 4H2O = Zn(OH)4-2 + 4H+ ###This block is excluded for NM1 and included for NM2
log_k -40.488
delta_h 0 kJ
-gamma 0 0
Zn+2 + Cl- = ZnCl+ ###This block is excluded for NM1 and included for NM2
log_k 0.4
delta_h 5.4 kJ
-gamma 4 0
Zn+2 + 2Cl- = ZnCl2 ###This block is excluded for NM1 and included for NM2
log_k 0.6
delta_h 37 kJ
-gamma 0 0
Zn+2 + 3Cl- = ZnCl3- ###This block is excluded for NM1 and included for NM2
log_k 0.5
delta_h 39.999 kJ
-gamma 4 0
Zn+2 + 4Cl- = ZnCl4-2 ###This block is excluded for NM1 and included for NM2
log_k 0.199
delta_h 45.8566 kJ
-gamma 5 0
Zn+2 + H2O + Cl- = ZnOHCl + H+ ###This block is excluded for NM1 and included for NM2
log_k -7.48
delta_h 0 kJ
-gamma 0 0
ZincHydroxide
Zn(OH)2 + 2H+ = Zn+2 + 2H2O
log_k 12 ### 12 is the used value in the model for this publication,
delta_h -89.62 kJ ### the original value was 11.334 in Minteq.v4.dat database
PITZER ### Adding the Pitzer coefficients of Zn+2 which are missing
### in the pitzer.dat database
PHREEQC INPUT FILE FOR MODELING PTCS OF ZN(OH)2 147
-B0
Zn+2 Cl- 0.2602 ### ref. 1: Simoes et al. Estimation of the Pitzer
Parameters
-B1 ### ... for 1--1, 2--1, 3--1, 4--1, and 2--2 Single
### Electrolytes at 25* C
Zn+2 Cl- 1.6425 ### ref. 1
-C0
Zn+2 Cl- -0.088 ### ref. 1
END
###########################################################################################
##### Part 2: define initial solution - water with ZnCl2 and with or without NaCl #####
###########################################################################################
USE SOLUTION 1
EQUILIBRIUM_PHASES ### Phases in equilibrium with water, to remove
### a phase delete the program line or add "#"
### at line start
Halite 0 0.05 ### 0.5 M NaCl in 100 mL
Melanothallite 0 0.0005 ### 0.005 M ZnCl2 in 100 mL
fix_pH -3 HCl ### Virtual phase to change initial pH to 3 with HCl
SAVE SOLUTION 1
END
##########################################################################################
##### Part 3: modeling stepwise addition of 20 mL of 0.1 M NaOH #####
##### solution, allow precipitation of Zn(OH)2 #####
##########################################################################################
USE SOLUTION 1
REACTION 2 ### Adding a total of 20 mL of 0.1 M NaOH via small
Na 0.1 ### steps
OH 0.1
H2O 55.55
T 0.1 ### Artificial counter to register the addition steps
0.02 moles in 1200 steps ### Steps are arbitrary, here they mimic 0.2 mL/min
The samples obtained after the post-treatment explained in chapter 5 were first
100-fold diluted with 18MΩ ultra-pure water from a Milli-Q system (Millipore,
USA). Before analysis, the nanoparticle (NP) suspensions were sonicated using
a sonicator, ultrasonic processor (Qsonica LLC, USA) equipped with a microtip
1/16” working in pulse mode and with a maximum amplitude of 40% during 10
minutes to avoid NP agglomeration. After sonication, different sample dilutions
were prepared using ultra-pure water, followed by a gently stirred.
All measurements were carried out using a high-resolution sector-field ICP-MS
(HR-SF-ICP-MS) instrument (Nu AttoM ES, Nu Instruments, UK) equipped
with a conventional sample introduction system comprising a glass concentric
nebulizer mounted onto a baffled quartz cyclonic spray chamber. The instrument
was operated in single-particle (SP) mode with a dwell time of 50 µs. To
overcome spectral overlap, pseudo-medium resolution mode was selected for the
monitoring of 56 Fe, while a low resolution was used in the case of 65 Cu (see Table
S2). To calibrate the transport efficiency (TE), gold nanoparticles (AuNPs),
stabilized in a citrate buffer, of known diameter and ionic standards solutions
of Au were used. Additionally, the analyte sensitivity was also measured by
using Cu and Fe ionic standard solutions. Appropriate dilutions factors for each
of the samples were selected to obtain good statistics without compromising
the occurrence of double events (Poisson statistics). Data treatment was
149
150 SAMPLE TREATMENT FOR AND CHARACTERIZATION WITH SP ICP-MS
performed using the Nu Quant software (Nu Instruments, UK). This software
can automatically differentiate the signal spikes corresponding to individual
NPs and the continuous background. In this way, particle sizes can be obtained
using the calculation’s editor built into the software and based on different
algorithms [194]. Additional information required for data treatment, such as
shape, chemical composition and density of the particles, is shown in table C.1.
In all cases, a spherical shape was assumed, and thus, the particle spherical
equivalent diameter is reported.
Calculating supersaturations
At the Y-junction the metal chloride solution and the NaOH solution are mixed.
Just before the precipitation reaction takes place, the metal ion concentration
and the hydroxyl ion concentration can be calculated for the total effluent
volume. Using the solubility products of the assumed precursors (Cu(OH)2 and
Fe(OH)3 ), a theoretical value for the supersaturation can be calculated. The
metal ion concentration is equal to the metal ion concentration of the influent
divided by 2. The hydroxyl ion concentration is calculated similarly in the
case of CuCl2 and FeCl3 but subtracting the amount of NaOH to neutralize
the acidic pH and overcome the water equilibrium. For the cases of CuCl and
FeCl2 , where H2 O2 is added, the theoretical calculation should also consider
the hydroxyl ions produced from the oxidation reaction between the metal ion
and hydrogen peroxide (maximally 1 time the metal concentration; because
each metal ion is assumed to oxidize to only one unit of valence higher). For
CuCl2 , CuCl, FeCl3 and FeCl2 the metal ion concentration is 0.005 M and the
OH- concentrations are respectively:
As observable, the OH− concentration for the copper cases is equal as well as the
OH− concentrations for the iron cases. The end OH− concentration in the iron
cases is 0.005 M more than the copper case. This is expected because the same
151
152 CALCULATING SUPERSATURATIONS
end pH is desired but for each Fe3+ ion, 1 more OH− -ion needs to react with it
compared to the Cu2+ -ion, and the metal concentration after
n mixing is 0.005
M. With the formula for supersaturation S = Mn+ OH− /Ksp,M(OH)n , the
∂ci ∂ 2 ci Pi
− Di 2 = (E.1)
∂t ∂x Vp
This equation resembles the heat equation with a constant source term and
is often solved with separation of variables. Although this approach gives the
exact solution, its form is needlessly complex and does not comply with our
vision to provide a straightforward solution. Because we are not especially
interested in the very short time range, but more in the overall behavior, and
more specifically at times approaching the steady state, we opted to solve the
equation using Laplace transforms. As such, while solving the equations in the
Laplace domain, it is possible to solve the equations for the limiting situation
when s approaches zero (equivalent to the time t going to infinity). The Laplace
transform of equation (E.1) is:
∂ 2 ĉi (x, s) Pi
sĉi (x, s) − ci (x, 0) − Di 2
= (E.2)
∂x Vp s
153
154 DERIVATION OF THE ANALYTICAL TIME-DEPENDENT MODEL OF THE TOTAL
ACCUMULATION IN THE GDE
with s the Laplace complex number frequency parameter, ĉi the Laplace
transformed concentration, and ci (x, 0) = ci,0 , a space independent initial
concentration equal to that of the initial bulk concentration ci,∞ (0). The
general solution to this differential equation is:
Pi + Vp ci,0 s √ √
ĉi (x, s) = + C1 cosh √ s x + C2 sinh √ s x (E.3)
Vp s2 Di Di
∂ĉi (0, s)
−Di =0 (E.4)
∂x
∂ĉi (wGDE , s) −ki ĉi,∞ (s)
= ĉi (wGDE , s) − (E.5)
∂x ψDi s
From the first boundary condition (equation (E.4)) we find C2 = 0:
∂ĉi (0, s) √ √ √ √
−Di = C1 √Ds sinh √s 0
Di
+ C2 √Ds cosh √s 0
Di
∂x i i
(E.6)
= C1 0 + C2 1 = C2 = 0
Using the second boundary condition (equation (E.5)), we can find the value of
C1 to be:
ki Pi +Vp ĉi (wGDE ,0)s−Vp ĉi,∞ (0)s
− ψD i V p s2
C1 = √ √ √ (E.7)
√ s sinh √ s wGDE + ki cosh √ s wGDE
D i D i ψDi D i
We can integrate ĉi (x, s) over the volume of the pores Vp to find an expression
of the Laplace-transformed accumulation in the GDE n̂GDE i (s):
ci,0 Rw
n̂GDE
R
i (s) = ĉ
Vp i
dVp = Vp + ψAgeo 0 GDE ĉi (x, s) dx
s
ci,0 Pi
= Vp + ψAgeo wGDE ...
s Vp s2
√ √
ki
√Di sinh √ s wGDE
ψDi wGDE s Di
1 −
... √ √ √
√s sinh √Ds wGDE + ψDki
cosh √Ds wGDE
Di i i i
(E.9)
with Vp = ψAgeo wGDE , the expression for n̂GDE
i (s) becomes:
ci,0
n̂GDE
i (s) = Vp + ...
s
√ √
ki
√Di sinh √ s wGDE
Pi ψDi wGDE s Di
... 1 − √ √ √
s2 √s sinh √Ds wGDE + ψDki
cosh √Ds wGDE
Di i i i
(E.10)
Since we are interested in the long time-frames, approaching steady-state, we
can simplify the expression for n̂GDE
i (s) by looking at the asymptotic behavior
and taking the limit for s going to 0 (something equivalent to taking the limit
for t to infinity) of the sinh and cosh terms:
√ √ 3/2
lim sinh √Ds wGDE = √Ds wGDE + s 3/2 wGDE 3
+ ... (E.11)
s→0 i i Di
√ 2
swGDE
lim cosh √ s wGDE =1+ + ... (E.12)
s→0 Di Di
In figure F.1 we demonstrate the long time experiments of the observed amount
of protons in the bulk. It is easily observed that the steady-state has a constant
slope after approximately 800 seconds.
0.7
Series3
500 A m-2
Series4
350 A m-2
protons in the bulk nbulk,H+ [10-3 mol]
0.6 Series5
200 A m-2
Series6
100 A m-2
0.5 Series7
50 A m-2
0.4
0.3
0.2
0.1
0
0 500 1000 1500 2000
time t [s]
157
Bibliography
[7] Gao-Ren Li, Han Xu, Xue-Feng Lu, Jin-Xian Feng, Ye-Xiang Tong, and
Cheng-Yong Su. Electrochemical Synthesis of Nanostructured Materials
for Electrochemical Energy Conversion and Storage. Nanoscale, 5(10):4056–
4069, 2013.
[8] Lok-kun Tsui and Giovanni Zangari. Electrochemical Synthesis of Metal
Oxides for Energy Applications. Springer, New York, US, 2014.
159
160 BIBLIOGRAPHY
[19] Jean-Pierre Jolivet, Marc Henry, and Jacques Livage. Metal Oxide
Chemistry and Synthesis: From Solution to Solid State. Wiley-Blackwell,
Chichester, UK, 2000.
[42] Jacqueline E Cloud, Tara S Yoder, Nathan K Harvey, Kyle Snow, and
Yongan Yang. A Simple and Generic Approach for Synthesizing Colloidal
Metal and Metal Oxide Nanocrystals. Nanoscale, 5(16):7368–7378, 2013.
[43] Fabian Kubannek, Thomas Turek, and Ulrike Krewer. Modeling Oxygen
Gas Diffusion Electrodes for Various Technical Applications. Chemie
Ingenieur Technik, 91(6):720–733, 2019.
[44] Yanguang Li and Hongjie Dai. Recent Advances in Zinc–Air Batteries.
Chemical Society Reviews, 43(15):5257–5275, 2014.
[45] Linden David and B Reddy Thomas. Handbook of Batteries. McGraw-Hill
Professional, New York, US, 2001.
[46] F Bidault, DJL Brett, PH Middleton, and NP Brandon. Review of Gas
Diffusion Cathodes for Alkaline Fuel Cells. Journal of Power Sources,
187(1):39–48, 2009.
[47] L Cindrella, Arunachala Mada Kannan, JF Lin, K Saminathan, Y Ho,
CW Lin, and J Wertz. Gas Diffusion Layer for Proton Exchange Membrane
Fuel Cells—a Review. Journal of Power Sources, 194(1):146–160, 2009.
[48] Sehkyu Park, Jong-Won Lee, and Branko N Popov. A Review of Gas
Diffusion Layer in PEM Fuel Cells: Materials and Designs. International
Journal of Hydrogen Energy, 37(7):5850–5865, 2012.
[49] Jaeman Park, Hwanyeong Oh, Taehun Ha, Yoo Il Lee, and Kyoungdoug
Min. A Review of the Gas Diffusion Layer in Proton Exchange Membrane
Fuel Cells: Durability and Degradation. Applied Energy, 155:866–880,
2015.
[50] George W Heise. Air-Depolarized Primary Battery, February 28 1933.
US Patent 1,899,615.
[51] Roy J Plunkett. Tetrafluoroethylene Polymers, February 4 1941. US
Patent 2,230,654.
[52] Walter G Taschek. Electrode for Fuel Cell, July 23 1963. US Patent
3,098,772.
[53] Charles E Thompson. Electrodeposition of Polymers in Porous Electrodes,
December 3 1963. US Patent 3,113,048.
[56] Eddy De Geeter, Michael Mangan, Staf Spaepen, William Stinissen, and
Georges Vennekens. Alkaline Fuel Cells for Road Traction. Journal of
Power Sources, 80(1-2):207–212, 1999.
[57] K Tomantschger, F McClusky, L Oporto, A Reid, and K Kordesch.
Development of Low Cost Alkaline Fuel Cells. Journal of Power Sources,
18(4):317–335, 1986.
[58] Yolanda Alvarez-Gallego, Xochitl Dominguez-Benetton, Deepak Pant,
Ludo Diels, Karolien Vanbroekhoven, Inge Genné, and Philippe Vermeiren.
Development of Gas Diffusion Electrodes for Cogeneration of Chemicals
and Electricity. Electrochimica Acta, 82:415–426, 2012.
[66] UN Desa et al. Transforming Our World: The 2030 Agenda for Sustainable
Development. 2016.
[67] Aling Chen and Bao Yu Xia. Ambient Dinitrogen Electrocatalytic
Reduction for Ammonia Synthesis. Journal of Materials Chemistry A,
7(41):23416–23431, 2019.
[68] Xiaoxi Guo, Huitong Du, Fengli Qu, and Jinghong Li. Recent Progress in
Electrocatalytic Nitrogen Reduction. Journal of Materials Chemistry A,
7(8):3531–3543, 2019.
[69] Lin Hu, Zhuo Xing, and Xiaofeng Feng. Understanding the
Electrocatalytic Interface for Ambient Ammonia Synthesis. ACS Energy
Letters, 2020.
[70] Antonio José Martı́n, Tatsuya Shinagawa, and Javier Pérez-Ramı́rez.
Electrocatalytic Reduction of Nitrogen: From Haber-Bosch to Ammonia
Artificial Leaf. Chem, 5(2):263–283, 2019.
[71] Xiaoyang Cui, Cheng Tang, and Qiang Zhang. A Review of
Electrocatalytic Reduction of Dinitrogen to Ammonia Under Ambient
Conditions. Advanced Energy Materials, 8(22):1800369, 2018.
[72] Bryan HR Suryanto, Hoang-Long Du, Dabin Wang, Jun Chen, Alexandr N
Simonov, and Douglas R MacFarlane. Challenges and Prospects in the
Catalysis of Electroreduction of Nitrogen to Ammonia. Nature Catalysis,
page 1, 2019.
[73] Masami Shibata, Kohji Yoshida, and Nagakazu Furuya. Electrochemical
Synthesis of Urea at Gas-diffusion Electrodes I. Simultaneous Reduction
of Carbon Dioxide and Nitrite Ions at Zn Catalysts. Denki Kagaku Oyoki
Kogyo Butsuri Kagaku, 64(10):1068–1073, 1996.
[74] Masami Shibata, Kohji Yoshida, and Nagakazu Furuya. Electrochemical
Synthesis of Urea at Gas-Diffusion Electrodes: Part II. Simultaneous
Reduction of Carbon Dioxide and Nitrite Ions at CU, Ag and AU Catalysts.
Journal of Electroanalytical Chemistry, 442(1-2):67–72, 1998.
[75] Masami Shibata, Kohji Yoshida, and Nagakazu Furuya. Electrochemical
Synthesis of Urea at Gas-Diffusion Electrodes III. Simultaneous Reduction
of Carbon Dioxide and Nitrite Ions with Various Metal Catalysts. Journal
of the Electrochemical Society, 145(2):595–600, 1998.
[76] Masami Shibata, Kohji Yoshida, and Nagakazu Furuya. Electrochemical
Synthesis of Urea at Gas-Diffusion Electrodes IV. Simultaneous Reduction
of Carbon Dioxide and Nitrate Ions with Various Metal Catalysts. Journal
of the Electrochemical Society, 145(7):2348–2353, 1998.
166 BIBLIOGRAPHY
[79] Duraisamy Saravanakumar, Jieun Song, Sunhye Lee, Nam Hwi Hur, and
Woonsup Shin. Electrocatalytic Conversion of Carbon Dioxide and Nitrate
Ions to Urea by a Titania-Nafion Composite Electrode. ChemSusChem,
10(20):3999–4003, 2017.
[80] Ronald L Cook, Robert C MacDuff, and Anthony F Sammells. High Rate
Gas Phase CO2 Reduction to Ethylene and Methane Using Gas Diffusion
Electrodes. Journal of the Electrochemical Society, 137(2):607–608, 1990.
[81] Maria Jitaru. Electrochemical Carbon Dioxide Reduction-Fundamental
and Applied Topics. Journal of the University of Chemical Technology
and Metallurgy, 42(4):333–344, 2007.
[82] Devin T Whipple and Paul JA Kenis. Prospects of CO2 Utilization via
Direct Heterogeneous Electrochemical Reduction. The Journal of Physical
Chemistry Letters, 1(24):3451–3458, 2010.
[83] Jinli Qiao, Yuyu Liu, Feng Hong, and Jiujun Zhang. A Review of Catalysts
for the Electroreduction of Carbon Dioxide to Produce Low-Carbon Fuels.
Chemical Society Reviews, 43(2):631–675, 2014.
[84] Sichao Ma, Masaaki Sadakiyo, Raymond Luo, Minako Heima, Miho
Yamauchi, and Paul JA Kenis. One-Step Electrosynthesis of Ethylene
and Ethanol from CO2 in an Alkaline Electrolyzer. Journal of Power
Sources, 301:219–228, 2016.
[85] Jun Li, Guangxu Chen, Yangying Zhu, Zheng Liang, Allen Pei, Chun-Lan
Wu, Hongxia Wang, Hye Ryoung Lee, Kai Liu, Steven Chu, et al. Efficient
Electrocatalytic CO2 Reduction on a Three-Phase Interface. Nature
Catalysis, 1(8):592, 2018.
[98] Fenglian Fu and Qi Wang. Removal of Heavy Metal Ions from Wastewaters:
A Review. Journal of Environmental Management, 92(3):407–418, 2011.
[99] Arezoo Azimi, Ahmad Azari, Mashallah Rezakazemi, and Meisam
Ansarpour. Removal of Heavy Metals from Industrial Wastewaters: A
Review. ChemBioEng Reviews, 4(1):37–59, 2017.
[100] Jian-Feng Chen, Yu-Hong Wang, Fen Guo, Xin-Ming Wang, and Chong
Zheng. Synthesis of Nanoparticles with Novel Technology: High-Gravity
Reactive Precipitation. Industrial & Engineering Chemistry Research,
39(4):948–954, 2000.
[101] Marcos Fernández-Garcı́a and José A Rodriguez. Metal Oxide
Nanoparticles. Wiley, New Jersey, 2011.
[102] SK Tripathi, Ramneek Kaur, and Mamta Rani. Oxide Nanomaterials and
Their Applications as a Memristor. In Solid State Phenomena, volume
222, pages 67–97. Trans Tech Publ, 2015.
[103] Margherita Izzi, Maria Chiara Sportelli, Nicoletta Ditaranto, Rosaria Anna
Picca, Massimo Innocenti, Luigia Sabbatini, and Nicola Cioffi. On the
Pros and Cons of the Sacrificial Anode Electrolysis for the Preparation of
Transition Metal Colloids. ChemElectroChem, 2019.
[104] David R Lide and Henry V Kehiaian. CRC Handbook of Thermophysical
and Thermochemical Data. CRC Press, Boca Raton, FL, 1994.
[105] GWC Kaye and TH Laby. Tables of Physical and Chemical Constants
15th ed. Longman, New York, 1986.
[106] Kathryn C Sole. The Evolution of Cobalt–Nickel Separation and
Purification Technologies: Fifty Years of Solvent Extraction and Ion
Exchange. Springer, Cham, 2018.
[107] Pradyot Patnaik. Handbook of Inorganic Chemicals. McGraw-Hill, New
York, 2003.
[108] David Lide. CRC Handbook of Chemistry and Physics, 90th Edition. Boca
Raton, FL, 2009.
[109] Jerzy Chlistunoff. Advanced Chlor-Alkali Technology. Final Technical
Report, 2004.
[110] Imad Moussallem, Jakob Jörissen, Ulrich Kunz, Stefan Pinnow, and
Thomas Turek. Chlor-Alkali Electrolysis with Oxygen Depolarized
Cathodes: History, Present Status and Future Prospects. Journal of
Applied Electrochemistry, 38(9):1177–1194, 2008.
BIBLIOGRAPHY 169
[112] Johannes Jung, Sarah Postels, and André Bardow. Cleaner Chlorine
Production Using Oxygen Depolarized Cathodes? A Life Cycle
Assessment. Journal of Cleaner Production, 80:46–56, 2014.
[113] Juergen Kintrup, Marta Millaruelo, Vinh Trieu, Andreas Bulan, and
Ernesto Silva Mojica. Gas Diffusion Electrodes for Efficient Manufacturing
of Chlorine and Other Chemicals. The Electrochemical Society Interface,
26(2):73–76, 2017.
[114] Fourth Edition. Guidelines for Drinking-Water Quality. WHO Chronicle,
38(4):104–8, 2011.
[132] Qiaobao Zhang, Kaili Zhang, Daguo Xu, Guangcheng Yang, Hui Huang,
Fude Nie, Chenmin Liu, and Shihe Yang. Cuo Nanostructures: Synthesis,
Characterization, Growth Mechanisms, Fundamental Properties, and
Applications. Progress in Materials Science, 60:208–337, 2014.
[133] M Mohapatra and S Anand. Synthesis and Applications of Nano-
Structured Iron Oxides/Hydroxides–A Review. International Journal
of Engineering, Science and Technology, 2(8), 2010.
[134] Victor K LaMer and Robert H Dinegar. Theory, Production and
Mechanism of Formation of Monodispersed Hydrosols. Journal of the
American Chemical Society, 72(11):4847–4854, 1950.
[135] Arne E Nielsen. Kinetics of Precipitation, volume 18. Pergamon, Oxford,
New York, 1964.
[136] Tadao Sugimoto. Preparation of Monodispersed Colloidal Particles.
Advances in Colloid and Interface Science, 28:65–108, 1987.
[137] JA Dirksen and TA Ring. Fundamentals of Crystallization: Kinetic Effects
on Particle Size Distributions and Morphology. Chemical Engineering
Science, 46(10):2389–2427, 1991.
[138] JP Klein and R David. Reaction Crystallization. Marcel Dekker, New
York, 1995.
[139] GP Demopoulos. Aqueous Precipitation and Crystallization for the
Production of Particulate Solids with Desired Properties. Hydrometallurgy,
96(3):199–214, 2009.
[140] James J De Yoreo, Pupa UPA Gilbert, Nico AJM Sommerdijk, R Lee
Penn, Stephen Whitelam, Derk Joester, Hengzhong Zhang, Jeffrey D
Rimer, Alexandra Navrotsky, Jillian F Banfield, et al. Crystallization by
Particle Attachment in Synthetic, Biogenic, and Geologic Environments.
Science, 349(6247):aaa6760, 2015.
[141] Nguyen TK Thanh, N Maclean, and S Mahiddine. Mechanisms of
Nucleation and Growth of Nanoparticles in Solution. Chemical Reviews,
114(15):7610–7630, 2014.
[142] Renguo Xie, Zheng Li, and Xiaogang Peng. Nucleation Kinetics vs.
Chemical Kinetics in the Initial Formation of Semiconductor Nanocrystals.
Journal of the American Chemical Society, 131(42):15457–15466, 2009.
[143] Jean-Pierre Jolivet, Sophie Cassaignon, Corinne Chanéac, David Chiche,
and Elisabeth Tronc. Design of Oxide Nanoparticles by Aqueous Chemistry.
Journal of Sol-Gel Science and Technology, 46(3):299–305, 2008.
172 BIBLIOGRAPHY
[144] Manuel D Montaño, John W Olesik, Angela G Barber, Katie Challis, and
James F Ranville. Single Particle ICP-MS: Advances Toward Routine
Analysis of Nanomaterials. Analytical and Bioanalytical Chemistry,
408(19):5053–5074, 2016.
[145] A Ortega and J Garcıa de la Torre. Hydrodynamic Properties of Rodlike
and Disklike Particles in Dilute Solution. The Journal of Chemical Physics,
119(18):9914–9919, 2003.
[146] Richard C Murdock, Laura Braydich-Stolle, Amanda M Schrand, John J
Schlager, and Saber M Hussain. Characterization of Nanomaterial
Dispersion in Solution Prior to in Vitro Exposure Using Dynamic Light
Scattering Technique. Toxicological Sciences, 101(2):239–253, 2008.
[147] Puthusserickal A Hassan, Suman Rana, and Gunjan Verma. Making
Sense of Brownian Motion: Colloid Characterization by Dynamic Light
Scattering. Langmuir, 31(1):3–12, 2014.
[148] Harry B Weiser, WO Milligan, and EL Cook. Hydrous Cupric Hydroxide
and Basic Cupric Sulfates. Journal of the American Chemical Society,
64(3):503–508, 1942.
[149] Yannick Cudennec and André Lecerf. The Transformation of Cu(OH)2
Into CuO, Revisited. Solid State Sciences, 5(11-12):1471–1474, 2003.
[150] Hui Liu, Ping Li, Bin Lu, Yu Wei, and Yuhan Sun. Transformation of
Ferrihydrite in the Presence or Absence of Trace Fe (II): The Effect of
Preparation Procedures of Ferrihydrite. Journal of Solid State Chemistry,
182(7):1767–1771, 2009.
[151] Bai-Sheng Zhu, Yong Jia, Zhen Jin, Bai Sun, Tao Luo, Ling-Tao Kong,
and Jin-Huai Liu. A Facile Precipitation Synthesis of Mesoporous 2-Line
Ferrihydrite with Good Fluoride Removal Properties. RSC Advances,
5(103):84389–84397, 2015.
[152] Soumya Das, M Jim Hendry, and Joseph Essilfie-Dughan. Transformation
of Two-Line Ferrihydrite to Goethite and Hematite as a Function of pH
and Temperature. Environmental Science & Technology, 45(1):268–275,
2010.
[153] James W Ball, D Kirk Nordstrom, et al. User’s Manual for wateq4f,
with Revised Thermodynamic Data Base and Test Cases for Calculating
Speciation of Major, Trace, and Redox Elements in Natural Waters. 1991.
[154] John O Edwards and Ralph G Pearson. The Factors Determining
Nucleophilic Reactivities. Journal of the American Chemical Society,
84(1):16–24, 1962.
BIBLIOGRAPHY 173
[166] Balasubramaniam Gnana Sundara Raj, Abdullah M Asiri, Jerry J Wu, and
Sambandam Anandan. Synthesis of Mn3 O4 Nanoparticles via Chemical
Precipitation Approach for Supercapacitor Application. Journal of Alloys
and Compounds, 636:234–240, 2015.
[167] SF Hasany, I Ahmed, J Rajan, and A Rehman. Systematic Review of the
Preparation Techniques of Iron Oxide Magnetic Nanoparticles. Nanosci.
Nanotechnol, 2(6):148–158, 2012.
[168] Sam GF Eggermont, Rafael Prato, Xochitl Dominguez-Benetton, and
Jan Fransaer. Metal Removal from Aqueous Solutions: Insights from
Modeling Precipitation Titration Curves. Journal of Environmental
Chemical Engineering, page 103569, 2019.
[169] Rafael A Prato, Vincent Van Vught, Sam Eggermont, Guillermo Pozo,
Pilar Marin, Jan Fransaer, and Xochitl Dominguez-Benetton. Gas
Diffusion Electrodes on the Electrosynthesis of Controllable Iron Oxide
Nanoparticles. Scientific Reports, 9(1):1–11, 2019.
[170] Yohannes Kiros, Massoud Pirjamali, and Martin Bursell. Oxygen
Reduction Electrodes for Electrolysis in Chlor-Alkali Cells. Electrochimica
Acta, 51(16):3346–3350, 2006.
[171] Peter Woltering, Philipp Hofmann, Frank Funck, Randolf Kiefer, Ulf-
Steffen Baeumer, Dmitri Donst, and Carsten Schmitt. Energy-Saving
Chlorine Production. Chlor-Alkali Electrolysis Using Innovative Cathode
Technology; Energiesparende Chlorpoduktion. Chlor-Alkali-Elektrolyse
Mit Innovativer Kathoden-Technologie. 2013.
[172] M Paidar, V Fateev, and K Bouzek. Membrane Electrolysis—History,
Current Status and Perspective. Electrochimica Acta, 209:737–756, 2016.
[173] Hella Boumaiza, Romain Coustel, Ghouti Medjahdi, Christian Ruby, and
Latifa Bergaoui. Conditions for the Formation of Pure Birnessite During
the Oxidation of Mn (II) Cations in Aqueous Alkaline Medium. Journal
of Solid State Chemistry, 248:18–25, 2017.
[174] Zheng-Rong Tian, Wei Tong, Jin-Yun Wang, Nian-Gao Duan,
Venkatesan V Krishnan, and Steven L Suib. Manganese Oxide
Mesoporous Structures: Mixed-Valent Semiconducting Catalysts. Science,
276(5314):926–930, 1997.
[175] Darminto, Machida N Cholishoh, Feby A Perdana, Malik A Baqiya,
Mashuri, Yoyok Cahyono, and Triwikantoro. Preparing Fe3 O4
Nanoparticles from Fe2+ Ions Source by Co-Precipitation Process in
Various pH, volume 1415. 2011.
BIBLIOGRAPHY 175
[178] Mingsong Wang, Lingxia Jiang, Eui Jung Kim, and Sung Hong
Hahn. Electronic Structure and Optical Properties of Zn(OH)2 :
LDA+U Calculations and Intense Yellow Luminescence. RSC Advances,
5(106):87496–87503, 2015.
[179] A Khorsand Zak, WH Abd Majid, M Ebrahimizadeh Abrishami, and
Ramin Yousefi. X-Ray Analysis of ZnO Nanoparticles by Williamson–Hall
and Size–Strain Plot Methods. Solid State Sciences, 13(1):251–256, 2011.
[180] Mingsong Wang, Yajun Zhou, Yiping Zhang, Sung Hong Hahn, and
Eui Jung Kim. From Zn(OH)2 to Zno: A Study on the Mechanism of
Phase Transformation. CrystEngComm, 13(20):6024–6026, 2011.
[184] Wei Jiang, Fei Liang, Jianwei Wang, Lei Su, Yaoming Wu, and Limin
Wang. Enhanced Electrochemical Performances of FeO X–Graphene
Nanocomposites as Anode Materials for Alkaline Nickel–Iron Batteries.
RSC Advances, 4(30):15394–15399, 2014.
[185] Jing Xu, Haibin Yang, Wuyou Fu, Kai Du, Yongming Sui, Jiuju Chen,
Yi Zeng, Minghui Li, and Guangtian Zou. Preparation and Magnetic
Properties of Magnetite Nanoparticles by Sol–Gel Method. Journal of
Magnetism and Magnetic Materials, 309(2):307–311, 2007.
176 BIBLIOGRAPHY
[188] J Johnson, G Anderson, and D Parkhurst. Database “thermo. com. V8. R6.
230,” Rev. 1-11. Lawrence Livermore Natl. Lab., Livermore, California,
2000.
[189] John Newman and Karen E Thomas-Alyea. Electrochemical Systems.
John Wiley & Sons, New York, 2012.
[192] David MH Kern. The Polarography and Standard Potential of the Oxygen-
Hydrogen Peroxide Couple. Journal of the American Chemical Society,
76(16):4208–4214, 1954.
[193] Noam Agmon. The Grotthuss Mechanism. Chemical Physics Letters,
244(5-6):456–462, 1995.
Personalia
Eggermont Sam
1989/08/04
Belgium
Education
2016−2020 KU Leuven Doctoral student of Engineering - Materials Engineering
Positions
2015−2016 KU Leuven Academic researcher
2013−2014 Jan De Nul QHSSE advisor of dredging activities
177
178 CURRICULUM VITAE
Teaching activities
Institutional responsibilities
This is a list of all publications to which the author has contributed up until
now:
This is a list of all the conferences to which the author has contributed up until
now:
179
180 LIST OF PUBLICATIONS