Phillip Loldrup Fosbøl - 87-91435-89-7

Philip Loldrup Fosbl
the problem is presented. The chemical system in the pipelines consists of

CO2-Na2CO3-NaHCO3-MEG-H2O. Sodium is injected as aqueous NaOH or
NaHCO3 in order to pH-stabilize and avoid corrosion and MEG is injected in
order to prevent gas hydrates.
There are a great number of models available in the literature which may
predict CO2 corrosion. These models are not accurate and assume ideality in
the main part of the equations. This thesis deals with aspect of improving the
models to account for the non-ideality.
Theories on electrolyte thermodynamics, electrolyte mass transport, and
electrochemical kinetics are presented. A literature overview of CO2 corrosion
is given and possible extensions of existing models are discussed.
A literature review presented in this work shows that FeCO3 plays a main part
in the protection of steel. Especially the solubility of FeCO3 is an important
factor. The thermodynamic properties of FeCO3 are discussed and validated.
A consistent set of FeCO3 properties are given.
A mixed solvent electrolyte model is regressed for the CO2-Na2CO3-NaHCO3MEG-H2O system. Parameters of the extended UNIQUAC model is fitted to
literature data of VLE, SLE, heat excess, and validated against heat capacity
data. The model is also fitted to experimental data produced in this work for
SLE in the Na2CO3-NaHCO3-MEG-H2O system.
The application of the above model is shown. It demonstrates the calculated
thermodynamic correction factors. These show how the diffusion process in
CO2 corrosion models deviate from the ideal case.
Department of Chemical
and Biochemical Engineering
DTU Building 229
Sltofts Plads
DK-2800 Kgs. Lyngby
www.kt.dtu.dk
Carbon Dioxide Corrosion: Modelling and Experimental Work Applied to Natural Gas Pipelines 2008
The focus of this study is carbon dioxide corrosion. CO2 corrosion is observed in offshore natural gas transportation pipelines. A general overview of
Carbon Dioxide Corrosion

Modelling and Experimental Work Applied to Natural Gas Pipelines

2008
Carbon Dioxide Corrosion:

Modelling and Experimental Work
Applied to Natural Gas Pipelines
Ph.D. Thesis
2008
IVC-SEP
Department of Chemical Engineering
Technical University of Denmark
DK-2800 Kongens Lyngby
Denmark
Technical University of Denmark (DTU)

Department of Chemical and Biochemical Engineering (DTU Kemiteknik)
Center for Phase Equilibria and Separation Processes (IVC-SEP)
Sltofts Plads Building 229
DK-2800 Kongens Lyngby
Denmark
Phone: +45 4525 2800
Fax: +45 4588 2258
Contact: kt@kt.dtu.dk
http://www.kt.dtu.dk
http://www.ivc-sep.kt.dtu.dk
Front cover by Philip L. Fosbl: Severe CO2 attacked pipeline.

Handed in October 2007.
Defended at DTU building 101 room 1, 12th of February 2008 at 13:00.
Corrections and additions included in the manuscript after the defence.
Copyright Philip Loldrup Fosbl, 2008
ISBN-10: 87-91435-89-7
ISBN-13: 978-87-91435-89-8
Printed in Denmark by
FRYDENBERG A/S
Baldersgade 12 - 16
DK-2200 Copenhagen N
Introduction
Preface
This thesis is published as a partial fulfilment of the requirements for the Ph.D. degree
at the Technical University of Denmark (DTU). The work was carried out in the
research group IVC-SEP at the Department of Chemical Engineering under
supervision of Kaj Thomsen and Erling H. Stenby. The project was started February
2004 and ended October 2007, interrupted by a two month leave for another project
during 2006.
The project was financed partly by Mrsk Oil and Gas A/S, DTU, and the Danish
Research Training Council trough Molecular Products and Product Technology
(MP2T) Graduate School in Chemical Engineering.
I would like to thank Erling H. Stenby for making this project possible, mainly the
industrial contact, but also for the suggestions and comments during the years.
Especially Kaj Thomsen has helped me in all aspects of the theory and experimental
work. I am grateful to the theory, database, programs, and library he has developed
over the years. It is a core part of this project. Thank you for being patient and letting
me have my own opinions. I look back at our fruitful discussions and hope to have
even more in the future.
I acknowledge gratefully that Mrsk Oil and Gas A/S let me use their production data
and internal knowledge unrestricted, but also the communication to the company
through Kim Rasmussen and Steffen Fredberg Hansen.
I appreciate the weeks Hans Wesselingh stayed at DTU for teaching, where he
introduced me to the Maxwell and Stefan diffusion theory. I recognise the importance
of the few discussions Heinz Gamsjger and I had during ISSP on FeCO3 and related
solubility experiments.
In the IVC-SEP group I would like to thank the technical personal for helping me set
up the equipment and sort out the difficulties, but also the four students who
performed some of the experiments. I am very happy for the time that I spend with the
colleagues in IVC-SEP and for the social part, let it continue.
I am thankful to Martin P. Breil and Kaj Thomsen who proof read parts of the thesis.
Thank you Jan Kamyno Rasmussen for giving me useful guidance and support during
the last months of writing. I am glad that all my friends and family understood the
problems that I had. Last but not least I owe it to you Marie. Hanging in there, even in
the most difficult times.
Hopefully this work will be used in the corrosion related research institutions at IFE
and the Ohio University or any other scientific research group out there.
___________________________
Philip Loldrup Fosbl,
Kongens Lyngby, October 2007
- iii -
Introduction
Summary
CO2 corrosion is a general problem in the industry and it is expensive. The focus of
this study is an oil gas production related problem. CO2 corrosion is observed in
offshore natural gas transportation pipelines. A general overview of the problem is
presented in chapter 1. The chemical system consists mainly of CO2-Na2CO3NaHCO3-MEG-H2O. Sodium is injected in the pipelines as NaOH in order to pHstabilize the pipeline to avoid corrosion and MEG is injected in order to prevent gas
hydrates.
There are a great number of models available in the literature which may predict CO2
corrosion. These models are not very accurate and assume ideality in the main part of
the equations. This thesis deals with aspect of improving the models to account for the
non-ideality.
A general overview and extension of the theory behind electrochemical corrosion is
presented in chapter 2 to 4. The theory deals with the basic thermodynamics of
electrolytes in chapter 2, the general description and extension of electrolyte mass
transport in chapter 3, and the electrochemical kinetics of corrosion in chapter 4. A
literature overview of CO2 corrosion is shown in chapter 5 and possible extensions of
the models are discussed. A list of literature cited is given in chapter 6.
The literature review in chapter 5 shows how FeCO3 plays a main part in the
protection of steel. Especially the solubility of FeCO3 is an important factor. Chapter
7 discusses and validates the thermodynamic properties of FeCO3. The study shows
that there is a discrepancy in the properties of FeCO3. Sets of consistent
thermodynamic properties of FeCO3 are given.
A mixed solvent electrolyte model is regressed in chapter 8 for the CO2-Na2CO3NaHCO3-MEG-H2O system. Parameters of the extended UNIQUAC model is fitted to
literature data of VLE, SLE, heat excess, and validated against heat capacity data. The
model is also fitted to experimental data produced and shown in chapter 8 for SLE in
the Na2CO3-NaHCO3-MEG-H2O system.
The application of the above model is shown in chapter 9. Here the thermodynamic
correction factors are calculated. These show how the diffusion process in CO2
corrosion models deviate from the ideal case. Conclusion and suggestion for future
work are presented in chapter 10 and 11.
-v-
Introduction
Resum p Dansk
CO2 korrosion er et udbredt og dyrt problem for industrien. I dette arbejde sttes
fokus p et olie og gas relateret produktionsproblem, som findes i offshore
naturgasrr. Problemet prsenteres kort i kapitel 1. Det kemiske system bestr af
CO2-Na2CO3-NaHCO3-MEG-H2O og natrium indfres i rrledningerne i form af
NaOH for a pH-stabilisere og p den mde forhindre korrosion. MEG tilfres for at
undg gashydratdannelse. I litteraturen findes et stort antal modeller som kan
forudsige CO2 korrosion. De er dog ikke prcise og idealitet er antaget i store dele af
modellerne. Kapitlerne i denne tese gennemgr hvordan teorien og de eksisterende
modeller kan forbedres for at kunne hndtere ikke-idealiteten i gas og vske fasen.
Et overblik og en udvidelse af teorien bag elektrokemisk korrosion prsenteres i
kapitel 2 til 4. Teorien for elektrolytters termodynamik prsenteres i kapitel 2, en
udvidelse af teorien for elektrolytisk masse transport prsenteres i kapitel 3 og
udvidelsen af elektrokemisk kinetik prsenteres i kapitel 4. Et resume over CO2
relateret korrosions litteratur prsenteres i kapitel 5 og det diskuteres hvordan
modeller kan forbedres. De anvendte referencer er givet i kapitel 6.
Diskussionen i kapitel 5 viser at FeCO3 spiller en vigtig rolle for beskyttelsen mod
korrosion og specielt FeCO3s oplselighed er en vigtig parameter. I kapitel 7
diskuteres og valideres de termodynamiske egenskaber for FeCO3. Her vises at der er
en stor usikkerhed i FeCO3s egenskaber. De tilgngelige egenskaber valideres og
st af konsistente egenskaber gives.
En termodynamisk model for det blandede solvent system CO2-Na2CO3-NaHCO3MEG-H2O prsenteres i kapitel 8. Parametrene for den udvidede UNIQUAC model
tilpasses litteratur data for gas-vske, fastof-vske, entalpi-excess og varmekapacitet.
Modellen tilpasses ogs eksperimentelle fastof-vske ligevgts data for Na2CO3NaHCO3-MEG-H2O systemet, som er blevet mlt i dette studie og prsenteret i
kapitel 8.
Den udvidede UNIQUAC models anvendelse vises i kapitel 9. Her prsenteres de
beregnede termodynamiske faktorer. De viser hvordan diffusions processer i CO2
korrosion afviger fra ideal diffusion og at man derfor br tage hjde for dette i
korrosions modeller. Konklusionen og forslag til fremtidig arbejde gives i kapitlerne
10 og 11.
- vii -
Introduction
Table of content
1
1.1
1.2
Introduction
1
Corrosion definition ................................................................................................ 3
Motivation and method .......................................................................................... 3
2.1
2.2
2.3
2.4
2.5
2.6
2.7
2.8
2.9
Theory of electrolyte thermodynamics

5
Standard states and reference (T, P) .................................................................. 5
The thermodynamic framework............................................................................. 6
Activity .................................................................................................................... 7
Standard states and concentration scales............................................................. 8
Liquid phase fugacities, J approach....................................................................... 9
Vapour phase fugacities, I approach ..................................................................... 9
Liquid phase fugacity, I approach ....................................................................... 10
The thermodynamic equilibrium criteria............................................................ 11
Phase equilibrium................................................................................................. 11
3.1
3.2
3.3
3.4
Theory of electrolyte mass transport

13
The volumetric component mol balance .............................................................. 13
The thermodynamic driving force........................................................................ 14
The Maxwell-Stefan friction theory..................................................................... 18
Setting up the diffusion problem ......................................................................... 20
3.4.1
3.4.2
3.4.3
3.4.4
3.4.5
4
Fick diffusion .................................................................................................... 21

Fick diffusion expressed in terms of activities ............................................... 22
The Nernst-Planck equation............................................................................ 23
Simple Extension of the Nernst-Planck equation .......................................... 25
Estimating missing diffusivities...................................................................... 27
Theory of electrochemical kinetics and corrosion

4.1
4.2
4.2.1
4.2.2
4.2.3
4.2.4
4.2.5
4.3
General activation controlled kinetics ............................................................ 37

The Tafel law .................................................................................................... 37
Stern and Gearys theory ................................................................................. 38
Corrosion rate (mm/y) from current measurements ...................................... 40
Extension of Stern and Geary theory for activation control.......................... 41
Concentration & activation polarization............................................................. 43
4.3.1
4.3.2
4.3.3
4.3.4
4.4
General activation and concentration polarization........................................ 46

Stern and Geary theory with concentration polarization .............................. 46
A general Stern and Geary theory .................................................................. 47
Comparison of Stern and Geary equations..................................................... 48
Reaction polarization and IR drop....................................................................... 50
CO2 corrosion
5.1
5.2
5.3
5.4
5.5
5.6
29
Electrochemical kinetics....................................................................................... 29
Activation polarization ......................................................................................... 32
51
CO2 corrosion types............................................................................................... 53

Empirical or purely correlative models ............................................................... 54
Semi-empirical models ......................................................................................... 55
Mechanistic models............................................................................................... 56
Electrochemical models ........................................................................................ 64
Improving CO2 corrosion models ......................................................................... 65
- ix -
Introduction
5.6.1
5.6.2
5.6.3
6
Assumptions of mechanistic models................................................................ 65

Thermodynamics of the fluid phases in CO2 corrosion .................................. 65
Properties of FeCO3 .......................................................................................... 66
Literature cited
69
7
Thermodynamic model and experiments in the mixed solvent electrolyte system
79
CO2-Na2CO3-NaHCO3-MEG-H2O
7.1
Thermodynamic modelling details ...................................................................... 80
7.2
Experimental methods and apparatus................................................................ 83
7.2.1
7.2.2
7.3
Chemicals.......................................................................................................... 83
Experimental setup .......................................................................................... 83
Experimental procedure ....................................................................................... 85
7.3.1
7.3.2
7.3.3
7.4
Previous work on analysis for sodium and carbonate.................................... 85

Interpreting the titration results .................................................................... 87
The reverse Schreinemakers method.............................................................. 89
Experimental results ............................................................................................ 94
7.4.1
7.5
Meta stable solutions ..................................................................................... 101

Parameter fitting ................................................................................................ 104
7.5.1
7.6
Parameter estimation procedure................................................................... 106

Results and discussion ....................................................................................... 106
7.6.1
7.6.2
7.7
7.8
8
Correlation of density..................................................................................... 106

Model of the Na2CO3-NaHCO3-MEG-H2O system ....................................... 112
Conclusion ........................................................................................................... 135
Literature cited ................................................................................................... 136
Review and recommended thermodynamic properties of FeCO3 corrosion films 143

8.1
8.2
Introduction......................................................................................................... 143
Standard state properties of FeCO3 .................................................................. 145
8.2.1
8.2.2
8.2.3
8.2.4
8.2.5
8.2.6
8.2.7
8.2.8
8.2.9
8.2.10
8.2.11
8.3
Experimental literature data on FeCO3 ............................................................ 168
8.3.1
8.3.2
8.3.3
8.3.4
8.4
8.5
8.6
Databases of standard state formation properties....................................... 145

Standard state heat capacity of FeCO3 ......................................................... 145
Standard state entropy of FeCO3 .................................................................. 146
Fe2+ and CO32- standard state properties...................................................... 147
Gibbs energy of formation of FeCO3 .............................................................. 148
Enthalpy of formation of FeCO3 .................................................................... 151
Solution enthalpy of FeCO3, reaction (252) .................................................. 152
Solubility constants of FeCO3 ........................................................................ 153
Equilibrium constant of FeHCO3+, FeCO3(aq), and Fe(CO3)22- ................... 162
Supersaturation.............................................................................................. 165
Comment on equilibrium constants of iron hydroxides ............................... 167
Aqueous solubility .......................................................................................... 168
Solid phases .................................................................................................... 171
Specific heat capacity of FeCO3 ..................................................................... 171
Thermal decomposition of FeCO3 .................................................................. 172
Synthesis of FeCO3 ............................................................................................. 172

Discussion and conclusion.................................................................................. 176
Literature cited ................................................................................................... 179
Improving diffusion models
189
10
Conclusion
195
11
Future, short term long term
197
-x-
10
Introduction
12
List of figures and tables

199
12.1
Figures................................................................................................................. 199
12.2
Tables .................................................................................................................. 203
13
A.
205
205
A.1
A.2
A.3
Converting between activity coefficient types .................................................. 205

Relations between standard state fugacities .................................................... 210
Converting standard properties between standard states............................... 215
B.1
B.2
B.3
B.4
Theory of electrochemistry
217
Isolation of net current in equation (171) ......................................................... 217
Derivation of equation (163)............................................................................... 217
C.1
Theory of diffusion
223
Porous media....................................................................................................... 223
B.
C.
D.
Model details of Na2CO3-NaHCO3-MEG-water

D.1
D.2
D.3
14
Appendix
225
Correlation matrix and confidence intervals .................................................... 225

The extended UNIQUAC model ........................................................................ 227
The extended UNIQUAC parameters ............................................................... 230
Index
233
- xi -
11
12
Introduction
Introduction
Process equipment is exposed to corrosion on a daily basis. Maintenance and repair

cost make up a significant part of the economy. CorrosionCost.com estimates that
$276 billion dollars/year is lost and spend in relation to corrosion only in the US
which is 2.5% of the US gross national product. $12 billion dollars/year are related
directly to the transportation lines of the oil and gas industry1.
The focus of this project is the natural gas transportation lines between platforms in
the North Sea. A crude separation of oil, gas, and water is carried out close to the
wellhead on an unmanned platform. The produced natural gas is transported 10-20 km
in 16 pipelines downstream for post-processing, see figure 1.
16
CO2
H2O
Natural gas
CO2
H2O
NaOH
Glycol
Wet GAS Line
Well head
Figure 1: An overview of the gas transportation systems.
A pipeline transports on average 4 million m3 natural gas/day. A breakdown will be

costly. Unfortunately the oil industry often deal with the problems as they arise,
instead of applying know-how and understanding of the system to prevent potential
problems.
Table 1 shows a typical composition of the alloy used for the offshore pipelines. It is a
mild carbon steel which indicate that the carbon content is relatively low. High carbon
steels are up to 1 wt% carbon. A corrosion resistant alloy as e.g. duplex steel, has a
high content of chromium and nickel. These types of steels are not used for gas
transportations since it adds to the cost of pipeline and makes the gas production
uneconomical.
Table 1: Composition of mild carbon steel, X65, used for pipelines.
X65
wt %
C
Si
Mn
S
P
Cr
Ni
V
Mo Cu
Al
0.057 0.22 1.56 0.002 0.013 0.05 0.04 0.04 0.02 0.01 0.041
Sn
0.001
Table 2 lists a typical composition of the gas phase at the production site. The total
pressure is 60 to 70 bar in the pipe inlet and the pressure drop is approximately 5 bar
across the pipelines.
-1-
13
Introduction
Table 2: Composition of a typical natural gas transported in pipelines in focus of this study.
Component
Methane
Ethane
Propane
Butane
CO2
N2
H2O
Higher alkanes
Composition (mol%)
88
6
2
1
1.6 (approx. 1 bar)
0.2
0.1
Residual
10
450
400
0.01
0.001
0.0001
1E-05
1E-06
1E-07
1E-08
350
300
250
200
I
mFeCO (g)
pCO (bar)
mNaOH(g)
150
100
50
1E-09
1E-10
0
5
8 pH
10
11
Figure 2: Ideal calculation of FeCO3 solubility in water. The liquid composition at pH >8 is a
hypothetical since NaHCO3 is most likely to precipitate. It was not included in the above calculations.
-2-
14
g NaOH/kg H 2O
0.1
Optimal Operating Window
pCO 2(bar),FeCO 3(g/kg H 2O)
Table 2 shows that small amounts of CO2 is found in natural gas and it is corrosive
like O2. The CO2 partial pressure is close to 1 bar. Produced natural gas is saturated
with water and it is therefore also known as wet gas. The water condenses at the cold
pipe wall during sub-sea transportation. Liquid mono ethylene glycol (MEG) or
methanol is injected at the inlet to prevent gas hydrates. The concentration is 95 wt%
in the inlet and approximate 30 wt% in the outlet due to dilution by the condensed
water. CO2 dissolves in the water-MEG liquid phase and the electrolytes will corrode
the bottom of the pipeline. pH is increased at the production site by adding NaOH,
NaHCO3 or a similar base in order to lower corrosion. The concentration is typically
10 mol NaOH/kg water in the inlet and 0.1 in the outlet. This shows that the liquid
phase is not simple, it is thermodynamically non-ideal and it is a multi solvent
electrolyte system.
Various protective corrosion products are produced, depending on the chemical
environment. FeCO3(s) is formed as a corrosion product. It can protect against
corrosion depending on pH, inhibitors and the bulk composition. Figure 2 shows the
ideal solubility of FeCO3 in water. When no pH-stabilizer of NaOH is added, then pH
is approximately 5 and a high solubility of FeCO3 is observed.
Introduction
When small amounts of NaOH is added, pH increases fast and at an addition of

approximately 25 g NaOH/kg H2O the pH is 7.5. The solubility of FeCO3 decreases
rapidly in the pH interval 5 to 7.5 from 0.1g/kg H2O to 10-6g/kg H2O. It is important
to notice that the partial pressure of CO2 stays almost constant in the interval pH=5 to
7.5 while above pH=8 significant amounts of CO2 is dissolved into the liquid phase. It
neutralizes the NaOH and it is not effectively used for pH-stabilization. The operating
window, shown in figure 2, gives the correct amount of NaOH to inject. This is done
to prevent iron dissolution and to keep the use of NaOH down to an economical level
but also to prevent dissolution of CO2 into the liquid phase. The ionic strength (I) is
shown in the figure. It rises from approximately 0.005 up to 10 mol/kg H2O in the
shown pH interval. The high ionic strength may never be reached due to saturation by
NaHCO3, though it may approach 4 mol/kg H2O.
Another way to avoid CO2 corrosion is to remove CO2 from the gas or dry it. This is
not possible since the process is offshore and production platform is unmanned.
Separation equipment is basically too expensive and highly cost-ineffective.
1.1
Corrosion definition
Different types of corrosion exist and the phenomenon is not a simple process. The
reaction is typically an interaction between many mechanisms. Corrosion is generally
defined as the irreversible deterioration of a substance or its properties because of a
reaction with its environment. Some of the known types of corrosion are
electrochemical corrosion caused by electron transfer, cracking caused by tensile
stress, flow assisted corrosion caused by fluid flow, fretting caused by friction and
rubbing and high temperature corrosion caused by alloy melting.
CO2 corrosion is an electrochemical process which is discussed in this study and the
remaining types of corrosion are subsequently not discussed.
1.2
Motivation and method
The motivation of this project has been to build a CO2 corrosion model in order to
predict corrosion rates. A great number of these models are available in the open
literature or as commercial products. Instead of reproducing their result the focus has
been to improve the existing models and discuss the main assumption. The method
has been to improve the thermodynamic description through experimental work and
modelling. The model can be plugged into existing corrosion models in order to
improve the description of the liquid phase. The protective corrosion product layers
are discussed in order to present a reliable set of thermodynamic properties for these
compounds.
-3-
15
16
2
2.1

Standard states and reference (T, P)
Thermodynamic properties like G, H, and S, etc. (Gibbs energy, enthalpy, and

entropy) are always modelled relative to a well-defined origin. In electrolyte systems
this implies modelling properties relative to the H+(aq) ion. Thermodynamic
properties of H+(aq) are arbitrarily defined as 0 at T=298.15K and P=1 bar. The
standard state properties at e.g. 313K and 1.5 bar are modelled relative to the defined
properties at (T, P). (T, P) are called the reference temperature and pressure. Most
standard state tables give properties at (T, P) but other conditions are seen. This
could for example be 350K and 2 bar, but the reference temperature and pressure are
still 298.15K and 1 bar.
Some old tables use a (T, P) of 273.15K and 1 atm. This implies that the properties
of the H+(aq) ion are arbitrarily set to 0 at this (T, P). It is not difficult to use
properties at other (T, P), but the user must be aware. The final result is always
independent of (T, P), since all of the thermodynamic properties that we are
interested in are basically changes from one state to another. The choice of (T, P) is
not crucial. Databases which uses different (T, P) may not be combined without
converting to the same (T, P).
The standard state is an arbitrary well defined chemical state. It is chosen for
convenience of the particular thermodynamic calculations. It is a well-defined
chemical state which is used as an origin for the remaining calculations. The
properties of the mixture are modelled relative to the defined standard state and it is a
set point of the calculations. The final result is independent of the particular standard
state. It can be chosen intelligently in order to facilitate the calculations. The
temperature, pressure, and composition may change, but the type of standard state
does not change.
There is a new defined standard state for every temperature and pressure in the
system.
There is a great misunderstanding of the principles in the literature on standard states
and reference (T, P). Standard states are given at every (T,P). Standard state
properties are always listed relative to reference (T, P). This is why properties at
T , P are always calculated from well known values at (T, P).
A standard state of ideal gas is typically used for gas phases. In standard state
property tables this is indicated by (g). The standard state Gibbs energy of formation,
' f G Dig , at a reference temperature 298.15K and 1 bar, is the Gibbs energy a
compound would have, if it were an ideal gas. The non-ideality arises if the mixture
suddenly does not behave as an ideal gas.
Liquids do not behave as ideal gasses and therefore other types of standard states are
used. Three schemes are typically applied, one which assumes the liquid phase is a
very non-ideal gas phase, second a standard state which uses pure liquid properties
and third a standard state which uses the properties in a well-defined mixture. The
-5-
17
third standard state is typically used for components which can not exist as a pure
component, e.g. ions.
2.2
The thermodynamic framework
From the first law of thermodynamics for a closed system, the following is true for the
internal energy, U:
dU
TdS Pdv
(1)
where v is molar volume of the system. The enthalpy and Gibbs energy per mol are
defined by:
H { U Pv
(2)
G { H TS
(3)
differentiation of the above gives the following relations:

dH { dU Pdv vdP
(4)
dG { dH TdS SdT
(5)
By substitution (4) into (5) and (1) into (4) reveals dG given by:
dG
vdP SdT
(6)
Equation (6) reduces to the following, at constant T:

dG
vdP (constant T)
(7)
For the ideal system v is substituted by RT/P and the above is therefore:
dG
RT
dP
P
RTd ln P (constant T)
(8)
Fluids are not ideal and in the real mixture, the pressure P is given by an observed
total pressure, f, the fugacity. By assuming that the equation for a non-ideal solution
looks like (8). A relation can be defined between the total fugacity and the total molar
Gibbs energy, G, in J/mol by:
dG { RTd ln f
(9)
Every component in the mixture contributes to the total Gibbs energy. The fractions
that they provide are the derivative of the total Gibbs energy with composition:
Pi { Gi
dnG
dni T , P ,niz j
-6-
18
(10)
this quantity is defined as the partial molar Gibbs energy, Gi , also called the chemical
potential, Pi . Similar to equation (9), it may be assumed that the real pressure of a
component i in a mixture of more than one component also follows a relation
equivalent to (9):
dGi { RTd ln fi S
(11)
The hat signifies the mixture and the fugacity of component i in phase S, fi S , is not a
partial molar property like the chemical potential but a definition given by (11).
Substituting (10) in (11) for the partial Gibbs energy shows that:
RTd ln fi S
d Pi
(12)
Integration over the chemical potential and the fugacity from a well defined standard
state to the state of the system gives:
Pi4 s RT ln
Pi
k l
fi S
f 4k sl
(13)
Pi4 s and fi 4 s represent the chemical potential and the solution fugacity in the
k l
k l
arbitrary standard state 4 with a concentration scale of s. s can be mol fraction scale
x also called rational scale, molal scale b (designated by superscript m) and molar
scale c. The indices k and l indicates the standard state type k and concentration scale
type, l. This sub-indexing of scale and standard state is used since different types can
be chosen for the same component in the varying phases.
An equation that converts from one standard state to another is derived from (12)
similar to (13). An integration of (12) from a standard state 41 of a scale s1 to a
standard state 42 of a scale s2 shows:
Pi4 s Pi4 s
2
RT ln
fi 42 s
f 41s
(14)
Appendix A.3 shows how to convert the chemical potential between standard state
types.
2.3
Activity
Equation (13) may be written in terms of the activity of component i in phase S, ai4,Sk sl :
Pi
Pi4 s RT ln ai4,S s
k l
k l
(15)
The activity is defined as the deviation of the fugacity from the standard state fugacity
in the liquid phase:
-7-
19
ai4s T , P, n
fi L T , P, n
fi 4s T , P, n
J i4s T , P, n si si4
(16)
which indicate that the activity is 1 in the standard state, fi 4s . ai4s is also related to
the activity coefficient J i4s and the concentration scale si, where si4 is the constant
concentration in the standard state as shown by (16). T , P, n signifies the variable is
a function of temperature, pressure, and composition. fi 4s is shown to be a function
of composition even though it in most thermodynamic frameworks is solely a function
of T and P. Appendix A.1 shows the relation between different types of activity
coefficients. A component expresses ideal behaviour, when J i4s T , P, n | 1 and the
activity of a component which behaves ideally is therefore:
ai4s ,id T , P, n
si si4 (ideal)
(17)
where id indicates ideality. Equation (16) shows that activity is ai4s =1 when the
solution fugacity, f L T , P, n , is identical to the standard state fugacity,
i
fi 4s T , P, n . Equation (17) shows that ai4s ,id |1 if a component behaves ideal and
si | si4 . The concentration relative to the standard state concentration is defined by:
si si4
si ,r
(18)
denoted by subscript r.
2.4
Standard states and concentration scales
The symbol 4 in equation (13) signifies an arbitrary standard state. There are two
well know standard states, the symmetrical and the unsymmetrical standard state.
There are other types of standard states. Recently a mixed-solvent electrolyte standard
state is being applied in the open literature2.
The symmetrical standard state is denoted by a superscript . The symmetrical
standard state fugacity, fi v s , is defined by:
lim
xi o1
fi si4
si
lim J iv s
xi o1
fi v s
(19)
(20)
The symmetrical standard state concentration is purity. This indicates that si4 on a
mol fraction scale is si4 xi4 =1 but at molal scale it is si4 bi4 =f. For mol fraction
scale the above implies that the fugacity move towards the symmetrical standard state
fugacity as its composition approaches purity. The unsymmetrical standard state is
-8-
20
denoted by superscript *. The unsymmetrical standard state fugacity, fi *s , is defined

by:
lim
xi o0
xsolv o1
fi s 4
si
lim J i*s
xsolv o1
f i *s
(21)
(22)
The standard state concentration varies with scale. It is 1 molal for molal scale, 1
molar for molar scale etc. fi *m for example is also known as the unsymmetrical
standard state fugacity in a 1 molal solution. The fugacity approaches fi *s when
sio si4 and J i*s =1. This is a very hypothetical state since J i*s z1. The activity
coefficient is defined as 1 at infinite dilution in one solvent, solv, as stated by (22).
Solv may be an arbitrary solvent, but water is almost always used. If a second solvent
is present in the solution, then it is treated as a solute in water. Appendix A.2 shows
how to convert fugacities between different types of standard states.
If the mixed-solvent standard state is used, then there is no longer one solvent, but the
solvents are treated as separate solvents. This complicates the thermodynamic
description and the ion activity coefficient becomes one in the mixture of solvents.
This is not used in the combination with unsymmetrical standard state.
2.5
Liquid phase fugacities, J approach
The fugacity of component i in the liquid phase at (T,P,n), fi L , where n and s

signifies a component vector, can be expressed by one of the following relations from
(16):
fi L T , P, n J i4s T , P, s si fi 4s T , P, s si4
J iD x T , P, x xi fi D x T , P xiD
J i* x T , P, x xi f i * x T , P xi*
(23)
T , P, m bi fi T , P b
J i*c T , P, c ci fi *c T , P ci*
J
*m
i
*m
*
i
The calculated value of fi L is independent of the chosen standard state and

concentration scale. Appendix A.2 shows how to convert standard state fugacities
between standard states and concentration scales.
2.6
Vapour phase fugacities, I approach
By the I approach indicates that fugacities are modelled using fugacity coefficients
instead of activities. Instead of applying an activity coefficient and defining many
-9-
21
different standard states, then only one standard state is typically used, the rational
symmetrical ideal gas standard state, denoted by superscript Dig and y for mol
fraction scale. This is normally the only standard state applied for gas phases and it is
also known as the ideal gas standard state. This standard state is very different than
the liquid phase standard states since it depends on composition.
The fugacity coefficient IiV T , P, n defines component is deviation from the ideal
gas law at (T,P,n) in the mixture:
fiV T , P, n IiV T , P, n fi Digy T , P, n
(24)
fiV T , P, n is the fugacity in the non-ideal gas phase and fi Digy T , P, n is the
standard state fugacity. The fugacity is essentially modelled like the J approach.
Instead the standard state is a function of composition. fi Digy T , P, n as given by:
fi Digy T , P, n
yi P
(25)
Which inserted in (24) is:

fiV T , P, n IiV T , P, n yi P
2.7
(26)
Liquid phase fugacity, I approach
The I approach for the liquid phase is much like the approach for the gas phase,
shown above. The fugacity is given by:
fi L T , P, n IiL T , P, n f i Digx T , P, n
(27)
Here the standard state fugacity is modelled as if the liquid is a very non-ideal gas.
The liquid phase fugacity coefficient IiL T , P, n is defined as the deviation from the
ideal gas law. The standard state fugacity is essentially defined equivalent to (25),
applying the liquid composition instead of the gas-phase composition:
fi Digx T , P, n
xi P
(28)
Which inserted in (27) becomes:

fi L T , P, n IiL T , P, n xi P
(29)
and reduces to the pure liquid fugacity, fi L , when xi=1 and IiL T , P, n IiL T , P :
fi L T , P IiL T , P f i Digx T , P, n , (for xi=1)
Setting (30) equal to (23), gives the following relation of fi D x , since (20) is valid:
- 10 -
22
(30)
fi D x T , P IiL T , P P
(31)
Combining (23), (29), and (31) gives a relation between the rational symmetrical
activity coefficient and the fugacity coefficient:
J iD x T , P, n IiL T , P, n xiD IiL T , P
(32)
J iD x T , P, n is typically calculated using a Gibbs excess model and IiL T , P, n is

calculated using an equation of state (EoS). An EoS may therefore be used for
calculating the activity through (16). xiD is the standard state mol fraction which is
usually xiD =1.
2.8
The thermodynamic equilibrium criteria
The criterion for equilibrium needs to be defined in order to calculate the equilibrium
composition. Imagine a number of containers, each holding a pure chemical. They are
mixed in a closed tank of fixed volume. Equilibrium is reached at some point and a
number of distinct phases are created, e.g. vapour, liquid, solid or other phases.
The temperature and pressure (T, P) have reached uniform values in every phase. The
general equilibrium criterion is:
PiD T , P, n PiE T , P, n ... PiS T , P, n
(33)
which states, that the chemical potential of component i at (T,P,n), PiS T , P, n , is

equivalent in all phases from D to S. By inserting (13) in (33) reveals the following
relation for a two phase system:
Pi4 s RT ln
1 1
fiD
f 41s1
Pi4 s RT ln
2 2
fi E
f 42 s2
(34)
Inserting (14) in (34) reveals the equilibrium criteria expressed in term of fugacities:
fiD
fi E
(35)
which can be proved for an arbitrary number of phases:

fiD
fi E
...
fi S
(36)
(33) and (36) are therefore equivalent formulation of the equilibrium criteria.
2.9
Phase equilibrium
Equation (33) or (36) is applied in a speciation routine in order to calculate the

equilibrium conditions. The electrolyte system is a system which contains a number
- 11 -
23
of reactions. The equilibrium condition in a reacting system is given by the following

relation obtained from (33):
' k P T , P ' k G T , P
k 1...N reaction
(37)
Nreaction is the number of reactions. This equilibrium criteria is applicable for all
systems including multiple phase systems. ' k P is the change in chemical potential of
reaction k at constant T and P. Numerous reaction may be defined in the system but
rarely all are solved in the same calculation. The reason is that the systems must obey
the Gibbs phase rule, F=C+2-P. This indicates that even though 10 solid phases may
potentially precipitate, then only the reactions related to the specific salts in interest
are included in the calculations. Inserting (15) in (37) gives:
' k P T , P Q i ,k Pi ,k
N comp
Q P
i ,k
4s
i, p k
(38)
RT ln a
4s
i, p k
k 1...N reaction
Subscript p signifies phase number and k is the reaction number. Therefore the
reaction may be homogeneous dissociation in the liquid phase or heterogeneous
equilibria across phase boundaries in for example SLE and VLE calculations. The
above rearranges to:
' r ,k G 4
RT ln K k
k 1...N reaction
(39)
' r , k G 4 is the standard reaction Gibbs energy and K k is the calculated equilibrium
constant of the k'th reaction defined by:

' r ,k G 4
N comp
N comp
Q i,k Pi4,ks ,
Kk
4s Q i ,k
i, p k
k 1...N reaction
(40)
Pi4s is found in open literature as tabulated values of standard Gibbs formation energy
or as correlation of the equilibrium constant as function of temperature. A quantity
called solubility index or saturation index is defined by:
N comp
SI k
4s Q i ,k
i, p k
Kk
(41)
It is typically used in a speciation routine in order to calculate the equilibrium

composition from the number of defined reactions. ai4, ps is obtained from the
thermodynamic model and Kk is calculated from tabulated standard state properties
through (39). The solubility of a salt is found at SIk=1 which is the criteria for
equilibrium saturation. It is determined by varying the liquid composition. Only the
reactions for the salts which saturate the solution are included in the equilibrium
calculation.
- 12 -
24
Some of the theories presented below are known from existing literature, as for
example presented by Taylor and Krishna3. The theory presented in section 3.4.4 has
not previously been presented and should be regarded as extension of the current
understanding of mass transport in electrolyte systems.
3.1
The volumetric component mol balance
The non-steady state mol balance across a control volume of section area A m 2 ,
volume V A 'z m3 and length 'z > m @ of component i is given by equation (42)
. The balance is given as the molar flux in, N i
mol m 2 s , plus production from
reaction, Ri mol m3 s , is equal to flux out, N i

time, 't > s @ , with concentration, ci mol m3 :
z 'z
, plus accumulation over the
't A Ni z 't Ri V 't A Ni z 'z ci t 't ci t V

Ni
and N i
z 'z
(42)
are perpendicular to the section area A. Dividing (42) by A't 'z at
'z o 0 and 't o 0 rearranges the above to
wci
wt
wN i
Ri
wz
(43)
which is the component mol balance of the volume shown in figure 3 over time.
Ni
Ri
z
ci
A
z
Ni
z+z
z+z
Figure 3: Mol balance of component i over time
The mol balance across a differential volume element written equivalent to (43) but
for the general case in 3D is given by:
wci
wt
N i Ri
- 13 -
25
(44)
Here the differential element wN i wz is substituted by the gradient vector of the molar
fluxes N i . The del or Nabla operator, , is the gradient vector
w
wx
w
wz
w
wy
(45)
and the molar flux is similarly given by:

Ni
N i , x
N i , z
Ni, y
(46)
Which is a vector of the molar fluxes in the three directions x, y, z. N i is therefore

a scalar given by:
Ni
wN i , x
wx
wN i , y
wy
wN i , z
(47)
wz
Usually N i is in it self a function of the del operator and consequently the mass
balance (44) results in an equation consisting of the Laplacian operator:
fi
w 2 fi , x
2 fi
wx 2
w 2 fi , y
wy 2
w 2 fi, z
wz 2
(48)
and the model becomes a parabolic second order partial differential equation (PDE)
which describes the un-steady state diffusion problem.
The steady state solution may be found by setting wci wt 0 and (44) becomes:
Ni
Ri
(steady state)
(49)
A sufficient amount of boundary conditions or bootstraps are required in order to

solve these equations and they shape the final solution of the problem.
3.2
The thermodynamic driving force
The following section goes through defining the driving force for diffusion and lays
out the thermodynamic equations related to diffusion.
Imagine a system which contains an amount of stored energy; it may be released as
work or heat. The energy is a difference between two states of the system. One state is
lower in energy. Take for example a ball on a desk. One scenario is the ball on the
desk and a second the ball on the floor. The energy released by the system, on moving
the ball from one state to the other, is the stored energy. The difference is also known
as the potential energy of the system. The energy is released on applying activation
energy, which in our example would be equivalent to rolling the ball over the edge of
the desk.
In chemical systems the stored energy is also related to a potential energy, the
chemical potential, Pi n, T , P in > J mol Nm mol @ . The chemical potential
- 14 -
26
describes the energy required to move the system from the low energy state to the
high energy state.
The potential is the integration over all the forces required. In case of the falling ball it
would be the gravity. The forces always point from a high energy level to a low level.
The driving force, didriv , is therefore the negative of the chemical potential gradient
didriv { Pi T , P, n
(50)
The force is given in Newton per mol > N mol @ . Certain authors3,4 decide to define the
xi Pi RT , this is principally wrong since it has a dimension
driving force as didriv
of >1 m @ and therefore it is not a force.

Contributions to the chemical potential can be temperature, pressure, concentration, or
electrical gradients. Other forces could be centrifugal forces, gravity, or related to
surface tension. All contributions are contained in the chemical potential, but they are
only included if they have a gradient. The temperature gradient is for example zero if
the diffusion layer has a uniform temperature.
The electrolytic diffusion process is influenced by the difference in the charge of the
components. An ion will influence the diffusion of other ions due to charge
interaction. The slower ion drags the faster due to pulling/pushing from the ionic
charge difference. For example a big slow diffusing negative ion will drag smaller
positive ions and the small positive ion will pull the slow negative ions.
The chemical potential at constant T, P is given by (15) or in the appendix as (281) to
(285). It is valid for electrolytic phase equilibria, but not for diffusion processes. The
mean ion charges of a liquid is zero at equilibrium and the ion charges has no
influence. Diffusion is not at equilibrium and the distribution of charges becomes
important. The resulting chemical potential at constant T, P is known as the
electrochemical potential similar to (15):
Pi T , P, n Pi4s T , P RT ln ai4s T , P, n zi FI
(51)
The difference is a contribution from the electrical potential, I . Where zi is the

charge number of the ion and F is the Faraday constant, 96485coul mol . Any
arbitrary standard state can be chosen.
There are only n-1 independent driving forces due to the Gibbs-Duhem relation at
constant T, P: xi dichem xi d Pi 0 . Inserting (51) in (50) gives the driving force:
didriv
Pi4s T , P RT ln ai4s T , P, n zi FI
P
4s
i
T ,P
T , P T , P RT ln a T , P, n T , P zi F T , PI T , P
4s
i
(52)
The standard state chemical potential does not change with position in the system and
the gradient is zero, Pi4s T , P T , P =0. This is only true at constant T, P since the
standard state chemical potential is a function of T, P. Constant T, P can be assumed
in most applications. When pressure gradients or temperature profiles are observed,
then the gradient of the standard state is a function of T, P and Pi4s T , P T , P z0.
- 15 -
27
Using the rational symmetrical standard state reduces the complexity of the driving
force considerably. Therefore Pi4s T , P PiD x T , P and ai4s aiD x . If the
standard molal unsymmetrical standard was used, it would have posed no problem
since:
PiD x
Pi*m RT ln M sJ ifx
Pi*m RT ln M sJ ifx
(53)
0
because the gradient of Pi*m and J ifx is zero and they are independent of the position
in the system, see equation (331) in the appendix. This indicates that even though the
rational symmetrical standard state is chosen, it is equivalent to molal unsymmetrical
standard state or any other state as long as the gradient of the conversion factor does
not change. The rational symmetrical standard state is therefore used as an example
system in order to set up a scheme where any standard state may be used. This gives a
general thermodynamic formulation of the driving force and the description becomes
independent of standard state properties. This reduces equation (52) to:
didriv
RT ln aiD x T , P, n
T ,P
zi F I T , P
(54)
Equation (54) may be rewritten as the following sum of partial derivatives using the
chain rule for multivariate functions:
didriv
n w ln a D x T , P, n
i
x z F I
RT
i
T , P
j
wx j
j 1
T , P , x jzi
(55)
since the derivative of a multivariate function is given by:

f
wf
wf
wf
xn (constant T,P)
x1
x2 ...
wx1 xiz1
wx2 xiz2
wxn xizn
(56)
note that the above equation is only valid at constant T, P and the partial derivative of
T, P is neglected. The subscript x j zi signifies the partial derivative with respect to
constant x j except for j=i. The gradients sum to zero:
n
(57)
i 1
this is equivalent to the Gibbs-Duhem relation above and implies again that only n-1
independent driving forces exist. Inserting (16) in (55) gives:
- 16 -
28

n w ln J D x T , P, n x T , P, n x D
i
i
i
RT
x j zi F I
wx j
j 1
T , P , x jzi
didriv
w ln xi T , P, n xiD w ln J iD x T , P, n

RT
x j zi F I
wx j
wx j
j 1
T , P , x jzi
(58)
Which is converted to the following since
1 dy
y dx
using a relative concentration
D
i
scale, xi ,r
d
d ln y
dx
xi x , from (18):
1 wxi , r T , P, n w ln J iD x T , P, n

x j zi F I
RT
wx j
wx j
j 1 xi , r
T , P , x jzi
n
driv
i
RT
xi ,r
wxi ,r T , P, n
w ln J iD x T , P, n

x
x j zi F I
i ,r
wx j
wx j
j 1
T , P , x jzi
(59)
If rational symmetrical activity coefficient is not applied as the activity coefficient

model in (52), then an appropriate conversion can be performed by using (286) from
the appendix:
wx j
w ln J i*m
wx j
w ln J ifx
wx j
w ln xs
wx j
w ln J i*m
wx j
(60)
It shows that it does not matter which activity coefficient model is used as long as the
conversion factor is independent of xj.
The relative concentration scale, si , r , in the first term of the sum in equation (59), may
be a function of T, P, and n. The problem is that some si ,r are dependent on T, P, and
n as for example molar scale. It is convenient to chose rational scale since the partial
derivative with respect to mol fraction at constant composition is 1 if i=j and 0 if izj.
The derivative may therefore be substituted by the Kronecker delta function:
wsi ,r T , P, n
wx j
x j zi
G ij
(61)
If any other scale is used then the function becomes complicated. This is one of the
primary reasons for applying the rational scale.
The content of the parenthesis in the sum of (59) is defined by:
*ij
wx j
x j zi
G ij xi ,r
also known as the thermodynamic factor. Examples for a binary mixture are:
- 17 -
29
(62)
w ln J 1D x T , P, n
*11 1 x1,r
*12
x1,r
(63)
wx1
w ln J 1D x T , P, n
(64)
wx2
At ideal condition it becomes the unit matrix:

*ii
1 , *ij ,i z j
(65)
An example of component i=1 in an ideal n component mixture:

*11 1 , *12
*13
... *1n
(66)
Inserting (62) in (59) using (61) reduces the driving force to:
didriv
RT
xi ,r
* x
ij
zi F I T , P
(67)
j 1
The driving force for moving an ion, is the sum of the thermodynamic factors times
the species gradients. At ideal condition this reduces significantly by using (65) in
(67):
didriv
RT
xi zi F I
xi ,r
(ideal)
(68)
The above driving force is essentially a term for transport due to a concentration
gradient and a term for the ionic charge effect. Equation (67) is used in the following
section to show the relation to diffusion theory.
3.3
The Maxwell-Stefan friction theory
With no further hesitation we present the generalized Maxwell-Stefan equations. For a

thorough discussion of the related subject please see Taylor and Krishna3, Medvedev4,
Newman5, and Wesselingh and Krishna6:
n
difriction
RT
j 1
ij
x j ui u j
(69)
difriction is the friction force which has a dimension of > N mol @ , u i is the velocity
> m s @ and ij
is the Maxwell-Stefan diffusivity, m 2 s . Diffusivity is a

concentration dependent diffusion coefficient which is related to the drag between
components. For a binary system it becomes equal to the Fickian diffusion coefficient
at infinite dilution. difriction originate from the frictional drag of molecules of
component i moving past or through the remaining components. This is proportional
- 18 -
30
to the velocity difference, ui u j , and the drag coefficients, RT ij . ij is therefore

the reverse proportional to the friction between component i and j. ij may also be
known as the drag between species. The diffusivities are symmetrical, ij ji , and
1
2
n n 1 diffusivities are required for a n component system. For multi component
Fick diffusion coefficient are not symmetrical and requires n 1 n 1 coefficients

for a n component system.
It is clear that the i j contribution to the sum in (69) is zero because the i j
velocities cancel out. Therefore ii is not defined for any system. The molar flux, Ni
mol m 2 s , is defined by:
Ni { ci ui
(70)
Where ci is the molar concentration, mol m3 . The total concentration is given by:
n
ct
ci
xi ct
(71)
i 1
Inserting (71) and (70) in (69) gives the friction force as a function of molar flux:
RT
xi ct
difriction
x j N i xi N j
j 1
ij
(72)
In the treatment of non-ionic systems it is usual to define diffusion fluxes, J i , with

respect to a reference velocity. This way all fluxes are treated relative to that velocity
and no absolute values are given. Three types of relative velocity frameworks are
normally applied: mass average, molar average and volume average velocity. This
implicates that the fluxes are relative to the mean velocity weighted either by mass,
molar, or volume. Conversion between the different frameworks can be done with a
certain amount of difficulty and it may imply using molar volumes which for most
thermodynamic models are inaccurate. ij remains independent of the framework.
For diffusion in electrolytes the most commonly used reference velocity is not one of
the above, but the solvent velocity, u s . The diffusion flux of component i is therefore:
J is
N i ci u s
(73)
The superscript s of the flux signifies solvent reference velocity framework. The
diffusive flux of solvent is zero.
J is
Ns
cs u s
Which is obtained by substituting (70) in (73). Inserting (73) in (72) reveals the
generalized Maxwell-Stefan in terms of diffusion flux:
- 19 -
31
(74)
RT
xi ct
difriction
x j J is xi J sj
ij
j 1
(75)
The above is generalised to:

d ifriction
RT
xi ct
B J
s
ij
s
j
(76)
j 1
Where B s is defined by:

n
xj
j 1,i z j
ij
Biis
xi
for i z j
ij
Bijs
(77)
Examples of B s are:
x
x2
x
2 ... n
12 13
1n
B11s
x2
21
B21s
x2
12
(78)
(79)
In vector notation (76) is:
RT 1 s
X BJ
ct
friction
where X is a diagonal matrix of xi and X
1
(80)
is a diagonal matrix of 1/xi. It must be
underlined, that there is only n-1 independent J is and difriction and the last equation is a
linear combination of the others. This is identified immediately by equation (74)
s
which states that the relative flux of solvent is zero. Solving (80) for J reveals:
ct 1
friction
B Xd
RT
(81)
The problem of applying the above equations lies in the missing diffusivities between
ions, which are rarely known. This is discussed later. The above equations are used in
the following sections to derive the theory of diffusion.
3.4
Setting up the diffusion problem
At steady diffusion the driving force is equal to the friction and therefore
- 20 -
32
didriv
difriction
(82)
Inserting the derived equations from (67) and (76) gives the link between the
thermodynamic description and the diffusivities:
RT
xi
d i
* x
ij
zi F I
j 1
RT
xi ct
B J
s
ij
s
j
(83)
j 1
Simplifying (83) may be related to the thermodynamic part, the diffusivities or both.
It may neglect some important physical phenomena and the assumption needs to be
reconsidered before the final conclusion may be drawn on the model. In the following
sections it is shown how (83) can be simplified to some of the popular representations
of diffusion descriptions in the literature. Suggestions are given to approach a correct
description of the diffusion process but still maintaining accuracy.
Harned and Owen7 and Robinson and Stokes8 discuss the theory of electrolytic
diffusion. They set up the equations for diffusion in solutions containing one salt.
Their equations include all correction to the infinite dilution activity coefficient. This
includes thermodynamic factor determined from experimental data. They conclude
that there is a discrepancy of a few percent between the calculated and the measured
observed diffusion coefficient. Ascribing the effect to some of the same phenomena
as the electrical potential, I , which is the pulling and pushing of ions due to charge
difference thereby maintaining electro neutrality. The effect is referred to as the
electrophoretic effect which is partly the concentration dependency of the diffusion
coefficients. They also discuss the effect of viscosity. These effects are not shown in
the derivation below and are assigned to the concentration dependence of the
diffusivities.
3.4.1
Fick diffusion
To prove that Ficks diffusion equation is a subproblem of (83) we need to do an

amount of assumptions. For a mixture of non-charged molecules, the potential is zero,
I 0 . Assume ideal thermodynamic condition. This results in the thermodynamic
factors are *ii 1 and *ij ,i z j 0 . The composition of the solutes, xi, are assumed to be
close to zero, xi|0. The solvent composition, xs, is assumed to be close to one, xs|1.
isf are the diffusivities at infinite dilution in solvent s. This reduces (77) to
Biis
1
, Bijs
isf
(84)
Written for a binary system this results in the following relation from (83)
d1
RT
*11x1 *12x2
x1
RT s s
B11J1 B12s J 2s
x1ct
(85)
Only one independent equation may be written for the binary system. Equation (83)
reduces to the following relation by applying the above assumptions:
- 21 -
33
xi
J
J is
ct isf
(86)
f
is
ci
s
i
Diffusion in one dimension gives Ficks law of diffusion:

dci
dz
isf
J is
(87)
where z is the directional coordinate.
3.4.2
Fick diffusion expressed in terms of activities
Ficks law expressed in terms of activities can not just be derived by substituting
concentration in (87) by activities. It can be derived by combining (54) and (76) in
(82) which gives the following relation:
RT T , P ln aiD x T , P, n zi F I
RT
xi ct
B J
s
ij
s
j
(88)
j 1
Assume that all species are non-charged gives zero gradient of the potential, I=0.
Since d dxln y 1y dy
dx it reduces the above equation to:
RT
aiD x
aiD x
RT
xi ct
B J
s
ij
s
j
(89)
j 1
Assuming composition of the solutes to be negligible gives a relation similar to (84).

This reduces (89) even further to the following:
RT
aiD x
aiD x
RT J is
xi ct inf
(90)
Which rearranges to:

J is =inf
xi ct
aiD x
aiD x
(91)
The molar flux can be calculated by assuming the velocity of the solvent is zero and
by inserting (91) in (73):
N i =inf
xi ct
aiD x
aiD x
us
(92)
Comparing (86) and (91) shows that the result is different from the above expected
result. It is noteworthy that the thermodynamic factors have not been assumed to
follow ideality. They are therefore not given by *ii 1 and *ij ,i z j 0 .
- 22 -
34
3.4.3
The Nernst-Planck equation
The Nernst-Planck equation gives a more correct description of the flux in electrolyte
solutions. Still some significant assumptions are introduced which may be regarded as
rough approaches in moderately concentrated solutions having an ionic strength above
0.01 mol/kg H2O.
Equivalent to the Fick model the thermodynamic factor is assumed ideal, *ii 1 and
*ij ,i z j 0 . The driving force is then given by:
didriv
RT
xi zi F I
xi
(68)
The solvent concentration is assumed to be high compared to all other ions, xs 1 ,

xi z s | 0 . Since the solutes are assumed infinite dilute, the only diffusivity used is the
infinite dilute between ion and solvent, isf . Contributions to the B is given by
s
Biis
xs
isf
1
,
isf
Bijs
(93)
Taking (68) and (76) using B from (93) inserted in (82) results in
d i
RT
xi zi F I
xi
RT
J is
xi ct isf
(94)
Rearranging (94) gives the diffusion fluxes, J is

J is
ct
isf xi isf
xi zi F
I
RT
(95)
The molar fluxes are then given by the following expression by combining (73) and
(95) in:
Ni
ct
isf xi isf
xi zi F
I xi u s
RT
(96)
Equation (96) is known as the Nernst-Planck equation. Usually the potential gradient
I is unknown and needs to be either estimated or calculated separately from other
relations.
The following derivation considers a procedure which can be performed in order to
reach a relation where I is expressed from known variables.
The current carried by the solution is calculated by summing over the fluxes and
weighting with respect to ionic charge:
- 23 -
35

n
F z jN j
(97)
j 1
From (96) and (97) the resulting equation is

i
n
F2
I fjs x j z 2j
RT
j 1
ct F z j fjs x j ct
j 1
(98)
The solvent reference velocity is neglected due to electroneutrality:

n
c z x u
t
ct us z j x j
j s
j 1
(99)
j 1
To simplify (98) the following three definitions are used, the equivalent conductivity
of species i, # i :
F2
ct xi zi2 isf
RT
#i {
(100)
The equivalent total conductivity of the solution, # t :

n
#t { #i
i 1
(101)
F2 n
ct xi zi2 isf
RT i 1
and the transference number of species i which essentially is the current related to
species i, ti:
ti {
#i
#t
(102)
Applying the above definitions reduces (98) to

i
ct F z j fjs x j # t I
(103)
j 1
which can be rearranged to a relation of the gradient of the electrical potential:

I
#t
ct
#t
z
j
f
js
x j
(104)
j 1
Inserting (104) in (96) result in an equation of the molar flux without the potential
gradient:
- 24 -
36
Ni
ct
xi zi F i
cF
t
RT # t # t
isf xi xi u s isf
z
j
f
js
j 1
x j
(105)
By insertion of (102) reduces the above to

Ni
ct isf xi i
ti
ct
ti
Fzi zi
z
j
f
js
x j ct xi u s
(106)
j 1
For batteries the current is not zero, i z 0 , but in chromatographic processes or ion
exchange processes, there is no current flowing through the diffusive layer. For this
kind of processes it can be assumed that the current is zero:
i
(107)
Which simplifies (106) to

Ni
ct isf xi
ct ti
zi
z
j
f
js
x j ct xi u s , for i
(108)
j 1
This is the Nernst-Planck equation assuming that the solution does not transport
current over the diffusive layer which avoids the problems in calculating the potential
gradient. Comparing equation (108) to the Fickian diffusion model, a model which
looks like the Fickian model may be set up:
Ni
ct iseff xi ct xi u s , for i
(109)
where iseff is the observed effective diffusivity. The coefficient contains a term
related to normal transport and a term from migration due to a potential difference.
Comparing (108) to (109) it is clear that the effective diffusion is given by
iseff
t
isf i
zi
z
j
j 1
f
js
1
x j
xi , for i
ndim
(110)
The effective diffusivity is a function of the infinite dilution diffusivity and a term
related to the potential gradient. 1 xi is a vector of reciprocal gradients and ndim is
the number of dimensions in .
3.4.4
1
ndim
1 xi reduces to one by multiplication of xi .
Simple Extension of the Nernst-Planck equation
Instead of assuming the liquid to be ideal, we extend the expression to a diffusion

scheme, where part of the thermodynamic factor is retained in the equations. The
thermodynamic factor between component i and it self, *ii , is assumed to follow (62)
but other factors are assumed to be ideal *ij zi 0 . This scheme is chosen since the
driving force for component i then only contains the gradient of component i and not
the gradients of other species. It may be illustrated by an example: For the binary
system given by (85), *12=0 and d1 is therefore only related to x1 and not to x2.
- 25 -
37
The infinite dilution diffusivities between a component and the solvent, isf , are
applied since diffusivity rarely deviates at higher concentrations. This is discussed by
s
Wesselingh and Krishna6 p. 131. The B matrix is then given by:
xs
,
isf
Biis
Bijs
(111)
xs is the mol fraction of solvent and it is not assumed to be xs|1. However other
s
contributions to the B are neglected. The above assumptions result in the following
equation equivalent to the derivation of (94)

d i
RT
*ii xi zi F I
xi
RTxs s
Ji
xi ct isf
(112)
This can be rearranged to the diffusion flux as a function of the thermodynamic factor
and the potential gradient.
J is
ct
isf
xzF
1
*ii xi isf i i I
xs
xs RT
(113)
The thermodynamic factor may be calculated by an activity coefficient model. The

molar flux is, from (73):
Ni
ct
isf
xzF
1
*ii xi isf i i I xi u s
xs
xs RT
(114)
The current is now given by the following expression similar to (103) using (97)
i
ct F z j fjs
j 1
#
1
* ii x j t I
xs
xs
(115)
Rearranging to isolate the electrical potential gradient, I , and assuming the current
to be
i
(116)
Then equation (117) is obtained similar to (108)

Ni
ct
isf
t
1
*ii xi i
xs
zi
z
j
f
js
*ii x j xi u s , for i
(117)
j 1
where the ion-ion interaction is maintained. Comparing the result to (108) shows that
the two equations are more or less the same except the contribution from the
thermodynamic factor and the mol fraction of solvent, xs , is included.
- 26 -
38
3.4.5
Estimating missing diffusivities
The thermodynamic factors need to be calculated and Maxwell-Stefan diffusivities

need to be known in order to solve the full diffusion problem. Then the molar fluxes
can be calculated from (83). One bootstrap relation, which simplifies the calculation,
is zero solvent velocity. This is valid at low flow conditions. Additional bootstrap
relations must be supplied in order to solve the problem since only n-1 independent
flux equations may be written.
The thermodynamic factors are obtained from a thermodynamic model. The MaxwellStefan diffusivities are most likely not available in the open literature, remembering
that 12 n n 1 diffusivities are needed. Some of the diffusivities are tabulated, e.g. it
can be assumed that the infinite dilution diffusivities between solvent and other
species are valid in the concentration range of interest, is |isf . isf are usually
found in data collections. The missing inter-ionic diffusivities may be estimated from
is using the correlation from Wesselingh and Krishna6:
ij |
is js
2
i 0.55
z j zi
2.3
, i
- 27 -
39
0.5 zk2 xk
k 1
(118)
40
Theory of electrochemical kinetics and

corrosion
Corrosion sustained by a current flow is referred to as electrochemical corrosion. This

type of corrosion is the basis of many corrosion mechanisms. Crevice corrosion is one
of these types. It is caused by galvanic differences between metals or due to formation
of concentration cells in a cavity. Corrosion is therefore sustained due to
concentration difference between the bottom of the cavity compared to the outer
surface. This is similar to pitting or corrosion under insulation (CUI). De-alloying and
intergranular corrosion are also caused by a galvanic action where the individual
grains or crystals of the alloy are attacked. CO2 corrosion is also an electrochemical
corrosion process.
In this chapter the basic theory of electrochemical kinetics is presented. The current
understanding of the theory is extended in the sections below and the theory given in
section 4.3.2 to 4.3.4 has not been presented before.
4.1
Electrochemical kinetics
Electrochemical corrosion is described by two or more half cell reaction which

exchange electrons through a conducting medium. One half cell is typically the
dissolution of metal by oxidation at the anode. The general reaction is written by:
M U M z ze
(119)
The other half cell is the reduction of a compound in the liquid phase at the cathode.
This could be Y=H+ or Y=HCO3- or both at the same time.
Y m ze U Y n z
(120)
If one of the reactions stops, then the other also stops, since the production and
consumption of electrons is no longer sustained as shown by figure 4.
Figure 4: At the anode, metal is dissolved by oxidation of Fe; the electrons are transported trough the
metal to the cathode area where they go into reduction of other species.
- 29 -
41
Figure 5 shows the current density in A/m2 as function of the electrical potential of
one half cell reaction, (119). A similar plot could be shown for reaction (120). The
figure shows the current produced and consumed by reaction (119). The process
always runs forward and backward and it is reversible. The net reaction and current
density are dependent on the electrical potential in volts. Figure 5 shows an idealised
view of the electrochemical process. In reality at least two half cell reactions would
run simultaneously and exchange electrons. Figure 6 shows a more realistic scheme
where two half cells react and exchange current. It shows how both cells run forward
and backward. The net current is a sum of all anodic and cathodic currents.
5
4
current density (A/m )
inet
Equilibrium
0
-0.2
-0.1
0.1
0.2
-1
-2
-3
-4
-5
potential, S - I(V)
Figure 5: Current density produced and consumed by a half cell reaction.
Within electrochemistry two terms are typically used in order to describe the
electrochemical process, these are called polarization and overpotential.
A half cell is said to be polarized if the applied current has changed the potential
compared to the half cell equilibrium potential. The overpotential is the voltage by
which a half cell has polarized. The overpotential, K, is defined by:
K { S I S I
eq
(121)
(S-I) is the electrical potential difference between the metal and the liquid. Little
polarization is seen if a small current is allowed to flow between two half cells. In
corrosion there is no control of the current flow and the system is short-circuited. This
indicates that the current is allowed to flow freely.
The corrosion process is a multiple step mechanism. M is oxidized to Mz+ which
involves transport of Y-m to the electrode surface, combination of ions with electrons
at the metal surface and diffusion of Mz+ away from the surface. Ideally the process
- 30 -
42
would run instantly, but in reality the oxidation is finite, being limited by the slowest
step, also known as the rate determining step (rds).
The total resistance towards current flow may be viewed as a series of equivalent
resistances in series, related to the steps in the corrosion process.
10
8
current density (mA/cm)
6
4
2
inet
0
-0.2
-0.1
0.1
0.2
-2
-4
-6
-8
-10
potential, S - I (V)
Figure 6: Current density of two half cells.
The overpotential of a half cell reaction is related to these resistances. Four types of
mechanisms are typically taken into account:

Activation polarization
Concentration polarization
Reaction polarization
Resistance polarization (IR drop)
Activation polarization is by far the most important of the four. It is related to the
transfer of ions at the metal surface and through the surface during the half cell
reaction. The electrons must cross and activation energy barrier in order to proceed.
Concentration polarization is caused by the mass transport of ions to and from the
metal surface. If the charge transfer at the electrode is very fast, it may result in a lack
of reacting ions at the electrode surface which slows the overall corrosion process.
This is often related to slow diffusing compounds retarded by e.g. a film of corrosion
product.
Reaction polarization is related to the slow reaction in other parts of the systems not
directly related to the surface. This could for example be the dissolution process of
CO2(g) into the liquid phase or other electrolytic speciation reactions in the liquid.
Resistance polarization (IR drop) is observed when measuring current as function of
an applied electrical potential and it is related to the resistance of the liquid.
- 31 -
43
4.2
Activation polarization
Activation polarization was discovered by Julius Tafel in the beginning of the 20th
century. He showed experimentally that an applied potential is a logarithmic function
the current density. In the following sections a basic set of equations is given with the
purpose to show the nomenclature and relations of electrode kinetics.
Consider the electrochemical reaction:

ZZZ
X
Red YZZ
Z Ox ne
k
kf
(122)
Red is the reduced species and Ox is the oxidised species similar to reaction (119) and
(120). kf and kb are the forward and backward rate constants, and n is the number of
electrons produced. The reaction is said to be oxidation and anodic if the net process
is forward. Similar the reaction backward is a reduction and cathodic. In principle Red
and Ox are groups of species which are defined by:
Red
R,
Ox
(123)
Qi is the stochiometric coefficient of component i. R and X are the reduced and

oxidized compounds respectively.
It is assumed in the following derivation that the electron transfer mechanism is a
simple one-step reaction. This indicates that the mechanism only contains one
elementary reaction. Electrons are often not transferred this way but one by one
through numerous elementary reactions. The mechanisms can become complicated
for multi electron transfers especially mechanisms which include adsorption
phenomena. The theory for this kind of processes may be derived from the equations
presented below.
The net current density is the sum of the anodic and cathodic currents:
ia ic
inet
(124)
The anodic current, ia, is determined by the forward reaction of (122) which is
positive, and the cathodic current, ic, is the backward reaction which is negative. It is
important to note that the current density is a function of the electrode areas and they
are assumed not to change during the electrochemical process.
The current densities are proportional to the rate of reaction. They are defined by:
ia
nFvb
ic
nFv f ,
(125)
F is the Faradays constant and vf and vb are the forward and backward rates per area.
The rate of reaction is related to the concentrations of the species in reaction (122):
vf
*
k f aRed
,
vb
*
kb aOx
(126)
a are the activities, typically the molar unsymmetrical activities are used. Preferably
the rational unsymmetrical should be used. Here the asterisk, *, signifies surface
activities. This indicates that the activities are determined at the electrode surface. The
bulk concentrations are usually not applicable since the mass transport influences the
- 32 -
44
liquid composition at the surface. Therefore the surface and bulk concentrations are
*
*
and aOx
not the same. kf and kb are the forward and backward rate constant and aRed
are the activity product of reactants and products, which are consistent with definition
(123):
*
aRed
aQ
*
aOx
aQ
(127)
Inserting (125) and (126) in (124) reveals the net current density:
inet
*
*
nF k f aRed
kb aOx

(128)
forward and backward rate are equal at equilibrium and the net current density is
therefore zero, inet=0. From the theory of Koryta and Dvorak9 and Rubinstein10, it is
given that the following Arrhenius type equation is valid for the rate constant:
kb
'H 0f 1 D nF S I
Pf exp
(129)
RT
'H b0 D nF S I
Pb exp
, kf
RT
P is the pre-exponential factor and D is the charge transfer coefficient. The nominator
in the exponential terms of (129) is the surface electron transfer activation energy. P
and 'H 0 are independent of the electrode potential. D is 0<D<1 and typically D=0.5.
This is only true for reactions represented by (122) because only one elementary
reaction takes place.
In order to derive a rate constant which is independent of 'H 0 and liquid
composition, it is necessary to introduce the standard state potential, S I . In the

v
standard state the potential is equal to the standard state potential, S I

In this state the activities are one, a
*
Red
*
Ox
1 , as indicated by (16). At equilibrium
*
the current is zero, inet=0, and therefore (128) becomes k f aRed
kf
S I
*
kb aOx
which gives
kb . In combination with (129) this gives:
kv
kf
kb
'H b0 D nF S I v
Pb exp
RT
'H 0f 1 D nF S
Pf exp
RT
(130)
v
I
k v is the rate constant at equilibrium at the standard state. One could choose another
state as standard state, e.g. the equilibrium state. Then the potential would be equal to
eq
S I S I and inet=0. If this potential was used, then k becomes concentration
dependent, since S I is concentration dependent through the Nernst equation. It

is therefore important that literature values of k are used with care, in order to
eq
- 33 -
45
determine which standard state was used. It is of cause also important to recognise for
which systems it can be applied because some rate constants are obviously
concentration dependent. Dividing (129) by (130) reveals the following relations:
kb
D nF S I S I v
k exp
RT
, k
1 D nF S I S I v
k exp
RT
(131)
By defining 'S v { S I S I and inserting (131) in (128) gives:

v
inet
1 D nF 'S v *
D nF 'S v *
nFk v exp
aRed exp
aOx
RT
RT
(132)
The activities of the compounds at the surface, a*, can not be measured and are
typically not equal to the bulk activities. Instead we may rewrite (132) as a function of
the equilibrium potential. This potential is important in relation to corrosion. The
Nernst equation written for reaction (122) is a function of the standard state potential
and the surface activities:
S I S I
*
RT aOx
ln *
nF aRed
(133)
Liquid phase composition will distribute evenly up to the electrode surface at

equilibrium. The Nernst equation (133) at equilibrium is therefore a function of the
bulk concentration since the activities in the bulk and at the surface are identical. The
eq
equilibrium potential, S I , is therefore calculated by:
S I
eq
S I
RT aOx
ln
nF aRed
(134)
Substitution of S I in (132) by (134) reveals:

v
inet
1 D nF S I S I eq
nFk exp
RT
D nF S I S I eq
v
nFk exp
RT
Ox
aRed
Ox
aRed
1D
*
aRed
(135)
D
*
aOx
k v is still not a function of the liquid concentration. The net current density has
become a function of both surface and bulk activities. Equation (135) reduces to the
following equation by insertion of (121)
- 34 -
46
inet
*
a*
1 D nFK aOx
D nFK

i0 Red exp
exp
a
RT
RT
aRed
Red
(136)
Where i0 is the exchange current density defined by:

1D
i0 { nFk v aOx
aRed
(137)
i0 is the current density at equilibrium where the backward and forward reaction rate
of (122) are equal.
A high i0 correspond to a case with no activation control. This is equivalent to a
complete reversible behaviour where equilibrium is obtained instantly. It corresponds
physically to a surface where the activation energy to transport electrons is low and
the electron transport is high. This is also known as a Nernstian behaviour.
If i0 is low then electron transport is low and the reaction runs slowly.
One way to determine the charge transfer coefficient, D, is to plot i0 as a function of
aOx and aRed and D is determined through (137).
By assuming that bulk activities and surface activities are equal then a*=a. This
reduces equation (136) to the well known Volmer-Butler equation:
inet
1 D nFK
D nFK
i0 exp
exp
RT
RT
(138)
This equation gives an idealised representation of a system. In this equation it is

assumed that the electrode does not have a diffusion film up to the surface. The
Volmer-Butler equation is only valid at constant activity, since K and i0 are
concentration dependent through equations (134) and (137) respectively. The anodic,
ia, and cathodic, ic, current densities listed in equation (124) are given by the
following relation, obtained from (138):
ia
1 D nFK
i0 exp
,
RT
ic
D nFK
i0 exp
RT
(139)
the two currents are impossible to measure independently even though they are given
explicitly by (139). Plots of equations (138) and (139) are shown in figures 7 and 8.
It is seen how the net current density is the sum of the anodic and cathodic current
eq
densities. At equilibrium S I S I and inet=0, but ia and ic are not zero. The
difference between ia and ic is the exchange current, i0. Figure 8 shows a logarithmic
plot of the numerical current in figure 7. The dotted lines are the anodic and cathodic
currents in (139). The cathodic current approaches zero, ico0, at high overpotential
where K>>-0.05 and inet moves asymptotically towards ic. Similarly for low
overpotential, K<<-0.05, inetoia.
ia and ic meet in the encircled point which is ((S-I)eq, i0). i0 is typically determined by
intersection of the asymptotical values of inet at high and low K in figure 8.
- 35 -
47
5
4
inet
current density (P A/cm )
3
2
i0
ia
0
-0.2
-0.15
-0.1
-0.05
ic
K
0.05
0.1
0.15
0.2
-1
-2
-3
-4
-5
potential, S-I (V)

Figure 7: The net current density, inet, is a sum of the anode and cathode current densities, ia and ic.
Plotted for i0=1PA/cm2, D=0.3. Note how K=0 at inet=0PA/cm2 where (S-I)eq=-0.05V. The thick dashed
line shows the slope of inet at (S-I)o(S-I)eq also known as the polarization resistance, Rp.
absolute i [P A/cm ]
1.E+02
1.E+01
i0
ic
ia
1.E+00
inet
-0.20
-0.15
-0.10
-0.05
K
1.E-01
0.00
0.05
0.10
0.15
0.20
potential, S-I (V)
Figure 8: Showing the same as figure 7 just represented in a numerical log plot of the current density.
- 36 -
48
4.2.1
General activation controlled kinetics
In the above equations it was assumed, that one elementary reaction took place. It was
also assumed that it transferred all the electrons in one step.
In reality electrons are transferred in several steps. One of the steps is the rate
determining step and it has a much lower i0 value than all the other reactions. It is
therefore the bottleneck of electron transfer.
An empirical equation similar to (136) and (138), which is applicable to real system,
contains an anodic and a cathodic current density. The rds of the two are not the same
and the charge transfer coefficient, D, of the anode and cathode are therefore not the
same. Then D and (1-D) in equation (136) or (138) becomes decoupled and may
instead be represented by two separate empirical variables, Dc and Da, for the cathode
and anode. A general empirical Volmer-Butler equation is then
inet
D nFK
D nFK
i0 exp a
exp c
RT
RT
(140)
The Ds are no longer bound the same way as in (136) and (138). The Ds may
become higher then one but not less than zero. The new charge transfer coefficients
are found experimentally from voltamograms where a potential is applied while the
current density is measured. A more general equation can be given, if information are
known of the rds or the number of electrons transferred, see Rubinstein10.
A general set of equation for activation controlled reactions may be written. If all
reactions are similar to (122), then the total transferred current is the sum of all
currents transferred from the kth reaction:
inet
(141)
net , k
The general Volmer-Butler description follows by combining (138) and (141):

inet
0, k
1 D k nk FKk
exp
RT
D k nk FKk
exp
RT
(142)
There is a new exchange current density, i0,k, charge transfer coefficient, Dk, electron
number, nk, and overpotential, Kk, for every k redox reaction. Kk is defined relative to
the equilibrium potential of the kth reaction by:
K k { S I S I
4.2.2
eq , k
(143)
The Tafel law
The study by Tafel11 revealed experimentally that a measured current density is an

exponential function of the applied overpotential as shown by figure 8. It was shown
that for potentials far from equilibrium, that is K>>0 or K<<0, follow a relation of the
form:
a E log i
- 37 -
49
(144)
Where a and E are regressed parameters. This relation is equivalent to the following
equation if a=-Elog(i0):
i
i0
E log
(145)
It is evident that equations (138) and (139) reveals the same result at K>>0 and K<<0.
K E
K E
Because equation (138)o i010 a for Kof and equation (138)o i010 c for
Ko-f. Where the anodic and cathodic Tafel constants, Ea and Ec, are defined by:
Ea
ln 10 RT
,
nF 1 D
Ec
ln 10 RT
nFD
(146)
Figure 8 shows the logarithmic plot of the current density of figure 7. This type of
plot is known as a Tafel plot. The slope of the asymptotical values at K>>(S-I)eq and
K<<(S-I)eq are Ea and Ec. The two constant are usually determined experimentally
from a plot similar to figure 8 and then evaluate together with (146) in order to
determine what nD is.
4.2.3
Stern and Gearys theory
The basis of modern electrochemical corrosion theory was presented by Stern and
Geary12 (SG) and discussed in detail by Stern13,14. They assumed that two redox
reactions of (122) contribute to the net current density. It was additionally assumed
that the anode is only anodic and the cathode is only cathodic. This implies that the
anodic reaction only runs forward and the cathode reaction only runs backward. This
assumption is valid far from the equilibrium potential of the two reactions.
The equations are derived from (142) for a system of two redox reactions:
inet
1 D 2 n2 FK2
D1n1 FK1
i02 exp
i01 exp
RT
RT

(147)
inet ,1
inet ,2
The net current is a sum of the two redox reactions and subscript 1 identifies the first
redox and subscript 2 the second. It should be noted that reaction 1 is only cathodic
and reaction 2 is only anodic. The anodic part of reaction 1 and the cathodic part of
reaction 2 have been neglected. Therefore two different exchange current densities are
obtained, i01 and i02. But also two different Ds, Ks and ns since they are related to
each their redox reaction. The equilibrium potential of the two reactions are not the
same and therefore K1zK2 since (S-I)eq,1z (S-I)eq,2.
When the corrosion process runs freely, then the potential is equal to the corrosion
corr
potential S I S I . The net current density is zero, inet=0, since the cathode
consumes the exact same amount of electron produced by the anode. This is of cause
only true when electrons are not supplied from the outside. During corrosion, equation
(147) reduces to icorr = inet ,1 = inet ,2 since inet=0. icorr is similar to the exchange current
density, i0. icorr is used in the corroding system of multiple redox reactions. i0 is related
- 38 -
50
to a single redox reaction. icorr is therefore the current which flows during corrosion. It
is determined from (147) and inet=0 by
inet
icorr
0 at S I
inet ,2 inet ,1
S I
corr
1 D n F S I corr S I eq ,2
2
2
i02 exp
RT
D n F S I corr S I eq ,1
1 1
i01 exp
RT
(148)
On substituting i01 and i02 from (148) in (147) results in the net current as function of
icorr:
inet
1 D n F S I S I corr
2
2
icorr exp
RT
exp D n F S I S I
corr
1 1
1 D 2 n2 FK corr
D n FK corr
exp 1 1
icorr exp
RT
RT
RT
Where the corrosion overpotential, Kcorr, is defined by:
K corr
S I S I
corr
(150)
By rearranging (149) as a function of the 10-base logarithm and using the Tafel
constant from (146), gives the following expression of the net current density:
inet
icorr 10K
corr
Ea 2
10K
corr
E c1
(151)
Ea2 and Ec1 are not identical since D1zD2 and n1zn2. By expanding the above
logarithmic terms in (151) by a first order Taylor series results in the following
relation of the net current density:
inet
1
1
icorr ln 10 K corr

E
E
c1
a2
icorr ln 10 K
corr
E a 2 E c1
E a 2 E c1
(152)
This equation is only valid at low overpotentials due to the Taylor expansion. The
corrosion current density, icorr, is calculated by rearrangement of the above:
- 39 -
51
(149)
icorr
E a 2 E c1
inet
corr
ln 10 K E a 2 E c1
E a 2 E c1
1
ln 10 R p E a 2 E c1
(153)
which is known as the original Stern and Geary equation. The polarization resistance,
Rp, follows Ohms law at low polarization and it is defined by:
1
RP
inet
(154)
K corr
K corr is typically applied in a broad interval while inet is measured. This is done in
order to determine the E values through a Tafel plot as mentioned previously. RP is
determined online by polarizing the corrosion cell both anodic and cathodic. The
slope of a ( K corr , inet ) plot reveals RP. This is shown as an example in figure 7 as the
thick dashed line. Thereby icorr may be determined through equation (154) and it can
be converted directly to corrosion rate in mm/year which is illustrated below.
The SG equation is only valid if activation polarization is the only corrosion
mechanism. It is also assumed the net current may be described by (147). Therefore it
is assumed that the cathodic reaction of the anode and the anodic reaction of the
cathode do not contribute significantly to inet. A consequence is that the theory is only
valid if the corrosion potential is not close to the two redox equilibrium potentials. If
this is not the case, then the two ignored reactions become significant and the theory
fails.
4.2.4
Corrosion rate (mm/y) from current measurements
It is possible to derive an equation which may be used for converting a measured

corrosion current density to a corrosion rate in mm/year. Assume iron dissolution is
the only anodic reaction which is given by:
Fe s o Fe 2 aq 2e
(155)
The number of moles iron which dissolve, nM, may be determined by using Faradays
law written for reaction (155):
nM
QMQ
(156)
zF
M signifies an arbitrary metal. Q is the quantity of electrons produced, z=2 is the

number of electron per reaction, QM=1 is the stoichiometric coefficient of iron, and F
is the Faraday constant. The mass of metal dissolved, mM, is calculated by a
stoichiometric relation in combination with (156):
mM
M M nM mM
- 40 -
52
Q 0 It
zF
MM
(157)
because Q I t where t is the time in seconds and I the current in amperes. MM is the
molecular weight of metal (kg/mol). The corrosion rate of metal M in m/s, CRM, is
calculated by:
CRM
mM
At U M
(158)
Which states that the mass dissolved, mM, in kg divided by the area from which it is
dissolved, A; in m2 over the time, t, together with the metal density, UM, in kg/m3,
gives the corrosion rate in m/s. Equation (157) relates the dissolved mass to the
current which inserted in (158) gives:
CRM
Q 0 IM M
AU M nF
mM
At U M
Q 0 iM M
U M nF
(159)
Where i I A is the current density in A/m2. Equation (159) relates the measured or
calculated corrosion current, i, to the corrosion rate in m/s. The corrosion rate is then
calculated using the SG equation by insertion of (153) in (159). icorr could also be
determined by a direct experimental method in order to calculate what the corrosion
rate is in mm/y, by using (159). This is of cause under the assumption that corrosion is
uniform. The pre-exponential factor in (159), Q 0 M M U M nF , of iron is 1.16
mm/yearm2/A. The corrosion rate in mm/year is therefore 1.16 times greater than the
current density measured in A/m2 or PA/mm2.
It must be noted that equation (159) can be applied for calculation of corrosion rate in
a system where several metals dissolve simultaneously, for example a system where
two different metals dissolve simultaneously. This is done by separating the current
related to the dissolution of the specific metal. This may be difficult.
4.2.5
Extension of Stern and Geary theory for activation control
The SG theory can be extended to include a more general activation controlled

processed. In order to do this it must be understood how the SG theory is used in
practise. The principles of the theory is that the potential, (S-I), is varied around the
corrosion potential, (S-I)corr, in order to determined the slope of a ( K corr , inet ) plot so
icorr can be determined numerically. The corrosion rate may then be determined by
using icorr in equation (159). The slope of a ( inet , K corr ) plot is called the polarization
resistance, Rp. The reciprocal of Rp is therefore given by:
1
RP
dinet
d S I S I
(160)
S I
corr
By insertion of the net current defined by (141) results in the general equation of the
polarization resistance:
1
Rp
d
k
inet ,k
d
S
I

S I

S I corr
- 41 -
53
dinet ,k
S I
d S I
(161)
S I corr
It can be shown that the SG theory can be extended to a more general scheme by
using the following net current density instead of (147).
1 D1 n1 FK1
D1n1 FK1
i01 exp
exp
RT
RT
1 D 2 n2 FK2
D 2 n2 FK2
i02 exp
exp

RT
RT
inet
(162)
This equation contains both the forward and the backward reactions of two redox
reactions. One reaction will primarily run forward and the other will mostly run
backward. The effect of the opposite reactions can be introduced in the SG theory in
order to present a more exact version of the theory. Appendix B.2 shows that the
resulting polarization resistance can be expressed by:
1
Rp
dinet
d S I
S I
S I corr
corr
n FK S I
ln 10 n1 F
exp 1 1

icorr
RT
RT
E a1
ln 10
Ec 2
n FK S I
nF
2 exp 2 2
RT
RT
corr

1

1

1
1
(163)
Where Ea1 and Ec2 are defined by (146) for the two reactions and icorr is given in the
appendix as equation (333). The overpotential at the corrosion potential, K kS I
defined by:
K kS I
corr
{ S I
corr
S I
corr
eq , k
, is
(164)
It can be seen by comparing (152) and (163) that equation (163) reduces to (152)
when K1S I of and K 2S I o-f . This is the case when (S-I)eq,1 and (S-I)eq,2 are
far apart and (S-I)corr is far from both equilibrium potentials. This is expected. If
(S-I)corr is close to the equilibrium potentials, then the back- and forward reactions
become important in (148). These were neglected during the derivation of (152).
corr
corr
One of the difficult tasks of using equation (163) is to determine KkS I . It requires

the knowledge of the equilibrium potentials, which can be determined through the
Nernst equation, (134). This can be problematic since it requires the liquid phase
activities. It is therefore necessary to analyse the liquid phase composition in order to
use (163). Indeed equation (163) is an improvement of the SG theory, but the practical
application can be challenging compared to (152).
The problems can be justified, if the corrosion potential approaches one of the
equilibrium potentials.
corr
- 42 -
54
4.3
Concentration & activation polarization
The derivation of the SG theory, (152), required that the Volmer-Butler kinetics,
(138), was assumed to be valid. Sometimes this is acceptable, but there will be cases
where surface activities, a*, are not equal to the bulk activities. This is the case when
films cover the electrode surface. These films may either be porous corrosion product
or a liquid diffusion films. Equation (136) gives the exact electrode kinetic of a redox
reaction, and the surface activities are not known.
Transport of ions through the solution is governed by three mechanisms: Diffusion
due to gradients in the chemical potential, Pi, or migration due to an externally applied
potential or from attraction due to charged species. Assume migration is negligible.
This will be true at low potentials or high concentration of a supporting electrolyte.
Also assume component fluxes in mol/m2, jk, obeys Ficks law. Ficks law written as
function of activities are given by (92) which rewrites to:
jk
Dk
xk ct dak
ak dz
(165)
Where Dk is the diffusivity at infinite dilution and ak is the rational symmetrical

activity of component k. xk is the mol fraction, ct is the total concentration in mol/m3
and z is the directional coordinate. The flux is related to the net current density by:
nF
inet
QX
jX ,
inet
nF
QR
jR
(166)
It states that the current density is proportional to the number of electrons transferred.
Q k is the stoichiometric coefficient. The subscript R refers to the reduced compound
and X refers to the oxidised, defined by equation (123). Assuming a linear
concentration profile in the diffusive layer up to the electrode gives the following
from equation (165):
jk
Dk
xk ct ak ak*
ak
G
Dk
xk ct ak* ak
ak
G
(167)
G is the diffusion layer thickness and it is assumed constant. This constraint may be
applied experimentally by using a rotating disk electrode (RDE) setup5,15. G has a
typical magnitude of 0.001cm to 0.05cm. (166) and (167) are combined in order to
derive an expression for the surface activities:
a*X
aX aX
inetQ X G
,
nFDX x X ct
aR*
aR aR
inetQ RG
nFDR xR ct
(168)
The maximum current density is reached when the difference between activities are
maximum. This is obtained when ak* 0 . The phenomenon is called the limiting
current density.
The flux of compounds is at its maximum when this current is reached, since the
bottleneck is no longer the transfer of electrons through the surface but the supply of
ions to the surface. Even though the current transfer could run faster it is limited by
- 43 -
55
the diffusion process. The maximum current, limited by diffusion, is obtained by

setting ak* 0 in (168) which reveals the limiting current density, id:
iXd
nFDX
Q XG
iRd
x X ct ,
nFDR
Q RG
xR ct
(169)
The limiting current density of X is negative whereas the limiting current density of R
is positive. Back substituting (168) in (169) gives the following expression of the
surface activities as function of the net and limiting current densities:
a*X
i
,
a X 1 net
d
iX
aR*
i
aR 1 netd
iR
(170)
The general equation which contains the terms for both activation and diffusion
control kinetics are derived by inserting (170) into (136) using (127):
inet
QR
i Q X
1 D nFK
inet
D nFK

io 1 net
exp
1
exp
(171)
d
d
X iX
RT
iR
RT
The function gives the net current density as function of the overpotential and limiting
current densities. The above equation is not an explicit function of inet because the
multiplication symbol and the stoichiometric coefficients, Qi, prevents the isolation of
inet at the left hand side of the equation. It is apparent that if iRd >>inet and iXd <<inet,
then equation (171) reduces to a purely kinetic controlled relation, the Volmer-Butler
equation, (138).
Equation (171) can be rearranged in order to express inet explicitly. This is done by
including a suitable number of assumptions. Assume that only one reduced
compound, R, and one oxidised component, X, contribute to the limiting current. This
is usually an acceptable assumption since one compound is typically the limiting, and
not a collection of compounds. This removes the multiplication symbols in (171).
Assume also that the stoichiometric coefficients of reaction (122) are one, QR=QX=1.
Under these assumptions and isolating the net current from (171) gives
inet
1 D nFK
D nFK
i0 exp
exp
RT
RT
1
1 D nFK 1
D nFK
1 i0 d exp
d exp
RT
RT
iX
iR
(172)
The derivation if found in appendix B.1. The above equation incorporates both the
kinetics from resistance to current transfer, but also the limiting current due to
diffusion control. It is similar to the Volmer-Butler equation, (138), except that it
includes a denominator term which corrects for limiting currents. The anodic and
cathodic current densities are calculated equivalent to (139) by the following two
equations:
- 44 -
56
ia
1 D nFK
i0 exp
RT
1
1 D nFK 1
D nFK
1 i0 d exp
d exp
RT
RT
iX
iR
(173)
ic
D nFK
i0 exp
RT
1
1 D nFK 1
D nFK
1 i0 d exp
d exp
RT
RT
iX
iR
(174)
Figures 9 and 10 shows a plot of (172) to (174). The figures are plotted at the same
conditions as figures 7 and 8. Even though, figures 7 and 9 should be comparable, but
they are significantly different. This is also the case for figure 8 and 10. The reason is
that, in this example, inet no longer follow the dotted asymptotes though i0 instead it
goes asymptotically towards iRd =5PA/cm2 and iXd =-4PA/cm2. Figure 10 also shows
that ia and ic are no longer crossing at i0 where K=0 or (S-I)eq=-0.05V since the ids
are close to i0 and therefore affects ia and ic. The thick dotted lines in figure 9 shows
the slope of inet at K=0. The dashed lines are the previous slope obtained in figure 7
and the small-dotted line is the new slope which has been influenced by limiting
currents. This indicates that the SG theory can be improved in order to take care of
limiting currents.
5
inet
current density (PA/cm )
3
2
1
ia
i0
0
-0.2
-0.1
0.1
0.2
ic -1
K
-2
-3
d
X
-4
-5
potential, S-I(V)
d
Figure 9: Parameters are the same as figure 7, except that iR =5PA/cm2 and iX =-4PA/cm2. The dashed
line is the old line of figure 7, the small-dotted line obtained by the influence of id.
- 45 -
57
1.E+02
Asymptotes without i
d
X
absolute i [PA/cm ]
1.E+01
d
R
i0
1.E+00
inet
ia
-0.20
-0.15
-0.10
ic
1.E-01
-0.05
0.00
K
0.05
0.10
0.15
0.20
potential, S-I (V)
Figure 10: Obtained by the same parameters as figure 9. It shows the effect of limiting currents which
was not shown in figure 8.
4.3.1
General activation and concentration polarization
It was previously shown by (142), that a general equation may be written for inet
which is activation controlled. A general equation which is valid for both activation
and concentration controlled kinetics can be derived by combining (141) and (172).
The result for a k number of redox reactions is:
inet
1 D k nk FKk
D k nk FKk
exp
exp
RT
RT
k
1 D k nk FKk 1
1
D k nk FKk
1 i0 k d exp
d exp
i
RT
i
RT
Rk
Xk
(175)
There is a variable collection (K, D, i0 and ids) for every k redox reaction. The anodic
and cathodic currents of the kth redox is given by an expression similar to (173) and
(174).
4.3.2
Stern and Geary theory with concentration polarization
It is possible to include both activation and concentration controlled kinetics in the SG

theory. It requires that inet is no longer given by (147) or (162), instead it is assumed,
that two redox reactions are given by the following two reactions derived from (171):
- 46 -
58
i
D n FK
i01 1 netd ,1 exp 1 1 1
i1
RT
(176)
i
1 D 2 n2 FK 2
i02 1 netd,2 exp

i2
RT
(177)
inet ,1
inet ,2
It is assumed that one redox reaction will only react forward and the other will only
react backward. The two redox reactions are similar to (147), except that the limiting
current densities have been included. The total net current density is calculated by:
inet
inet ,1 inet ,2
Appendix B.3 shows that the polarisation resistance at S I
(178)
S I
corr
can be
expressed by the following equation as a function of the limiting current densities:

dinet
d S I S I
S I corr
D n F i 1 D 2 n2 F icorr
icorr 1 1 1 corr

1 d
i1d
RT
i2
RT
(179)
icorr is given in the appendix as equation (344). The above equation is applicable to
kinetics which has both activation and concentration controlled kinetics. i1d is the
cathodic limiting current which is negative and i2d is the anodic limiting current which
is positive. If there are negligible limiting current densities, i1d <<icorr and i2d >>icorr,
then the above equation reduces to the original SG equation, (152). It is important to
point out, that the above equation is 2nd order with respect to icorr, and has two
solutions. This can be seen by writing the above equation as:
1 D 2 n2 F 1 D1n1F 1 i D1n1F 1 D 2 n2 F dinet

2
0 icorr

corr

RT
i2d
RT i1d
RT
RT
d S I S I
4.3.3
S I corr
A general Stern and Geary theory
It is necessary to neglect the major assumptions, which were introduced in the SG

theory, in order to set up a general equation of the SG theory. Let the system consist
of k number of redox reactions and let these include limiting currents of anodic and
cathodic reactions. The net current is given by the general relation, (175). The
corrosion exchange current, icorr, is the current flowing while corrosion runs freely. It
is difficult to define in a general scheme. The reason is that there are several anodic
and cathodic reactions running at the same time. At (S-I)=(S-I)corr they all contribute
to inet.
The corrosion rate is no longer related to icorr, at (S-I)=(S-I)corr, as previously, but by
the sum of current densities which are related to the dissolution of the metal in focus.
This can be explained by visualizing corroding iron which dissolves by an anodic
process. If another anodic process is running in parallel, then only the current
- 47 -
59
transported by the iron redox reaction would be the actual corrosion current density. It
is therefore necessary to redefine what corrosion current is. It is possible to present
some of the theory. Solution of the problem would require more advanced computer
power, and the solution can not be given explicitly.
In the following derivations we seek an equation which can be used for calculating the
net current density. This is done in order to calculate the polarization resistance, RP, at
the corrosion potential, (S-I)=(S-I)corr. The theory can be set up differently, but this
approach is used, with the purpose of comparing the theory to the previously derived
SG equations.
Assume inet,k are defined by (172) for the kth reaction and the net current density is
calculated by (175). Differentiating (175) with respect to the potential reveals, RP, as
defined by (161). The derivative of the kth redox is given by the following relation as
shown in appendix B.4:
dinet ,k
d S I
1 D k nk FKk
inet , k
inet ,k
nk F
D k nk FKk
1 D k 1 d exp
D k 1 d exp
RT
iRk
RT
iXk
RT
D
K
1
n
F

k k k 1 exp D k nk FKk
1
1
d exp
d
i0 k iRk
RT
RT
iXk
This equation is not simple compared to the original SG theory, (154). On the
contrary this equation includes concentration and activation polarization of all redox
reactions and the SG equations represented by (154), (163) and (179) can be derived
from (180).
4.3.4
Comparison of Stern and Geary equations
Four different versions of the SG theory have been presented in the above sections:
The original SG theory, (154), the theory for advanced activation controlled kinetics,
(163), the original theory extended to concentration controlled kinetics, (179), and
finally a general formulation, (180).
The original SG theory includes the important activation controlled kinetics. The
extension given by (163), is beneficial for mechanism where the corrosion potential is
close to any of the equilibrium potentials. The drawback is that activities in the liquid
phase must be known.
The SG theory extended to concentration controlled kinetics, (179), is applicable to
diffusion controlled mechanisms. The activities of species have to be measured in
order to calculate the limiting currents. This may be avoided if limiting currents are
determined from measurements. The drawback is similar to the original SG theory.
The corrosion potential may not be too close to the equilibrium potentials.
The general SG theory given by (180) is of cause generally applicable, but much
information is needed in order to apply it.
Here two redox systems are used as a case study. They include limiting current
densities. This is with the intention to compare the various SG equations. Figure 11
shows the two redox reactions and the parameters used. The dotted lines are the
asymptotes of the anodic and cathodic reactions for pure activation polarization which
are used by the original SG theory. The circles indicate the exchange current
densities. The original SG theory predicts a corrosion exchange current density,
indicated by iStern-Geary, which is much higher than the real icorr. This is due to the
- 48 -
60
(180)
limiting current densities which are close icorr. This is also the reason why the
corrosion potential, (S-I)corr, is not predicted very well.
1.E+03
ia1
icorr
absolute i [P A/cm ]
i c2
iStern-Geary
1.E+02
icorr
1.E+01
i02
i 01
1.E+00
-0.30
-0.20
-0.10
corr
0.00
K =0
potential, S-I (V)
0.10
0.20
0.30
Figure 11: Shows a plot of the net anodic and cathodic currents densities (blue) and the net current
density (red). Equilibrium potential are r0.2, anodic and cathodic limiting currents are -80PA/cm2 and
200PA/cm2. Exchange currents are 1PA/cm2 and 2PA/cm2, Ds are 0.25 and 0.5.
Table 3 shows the polarization resistance, Rp, calculated by the various SG methods
(154), (163), (179), and (180) defined (161). An additional method has been used,
where Rp is determined numerically by the slope of ( inet , K corr ) at (S-I)=(S-I)corr. It is
obvious from the table, that the original SG theory, (154), performs not as good in a
system with limiting currents, (179) and (180). Equation (163) suffers the same
problems. Rp is under predicted an order of magnitude and therefore the corrosion rate
is over predicted by a factor two. The exact solution, (180), and the slope of the graph
give identical results.
Table 3: Comparison of the Stern and Geary equations applied for the case presented in figure 11.
Exact (180)
Graph slope
0.0006579
0.0006579
RP(Vcm2/PA)
iCorr =60.4095P A/cm 2 , S I corr -0.04727V
SG, Eq. (154)

0.000340080
Eq. (163)
0.00034081
These methods give almost the same results with minor variation. The original SG
theory is a conservative method since it will always tend to over-predict the corrosion
exchange current density and therefore over predict the corrosion rate. This may be
explained by the limiting currents which will tend to lower the corrosion exchange
current density compared to the purely activation controlled kinetics. From these
result it is clear that the original SG theory actually perform quite well compared to
the more exact methods. The theory becomes complicated when concentration
polarization is included in the equation scheme. It is a matter of weighing the benefit
from the more exact corrosion measurements using the advanced theory compared to
a less accurate result with the original SG theory. There is no reason why not to apply
- 49 -
61
Eq. (179)
0.0006581
the advanced theory since the advanced theory is consistent with the original SG
theory by setting the limiting current densities to infinite values. The result of the
advanced theory is that corrosion measurements can easily be improved if the limiting
current densities are known.
4.4
Reaction polarization and IR drop
Reaction polarization is due to slow reaction in the aqueous phase. It can become the
limiting variable in the electrode kinetics. The reaction is not related to the kinetics at
the metal surface and it is independent of the electrode. It can be related to the
limiting current density since it may influence the concentration of species in the bulk.
For example CO2 dissolution is slow, but the electrode reaction is fast as discussed by
Nesic et al.16 and Nesic and Postlewaite17,18. The used mechanism is:
CO2 g V CO2 aq
slow
(181)
CO2 aq e o 12 H 2 g HCO3 aq
fast
(182)
a limiting current may be observed, due to the slow production of CO2(aq). The
physical phenomena would enter into equation (172), but the derivation is not given
here. Nesic et al.16 and Nesic and Postlewaite17,18 uses a slightly different
representation of the compounds shown in reaction (181) and (182). The corrosion
literature often uses H2CO3 in the speciation scheme. This includes CO2(g) dissolution
to CO2(aq), which hydrates to H2CO3 and dissociates to HCO3- and CO32-. In this
speciation scheme CO2(aq) and H2CO3 are separate compounds which are present
simultaneously. This contradicts basic thermodynamic properties given by NIST19 and
CODATA20. NIST presents properties of H2CO3, but they are identical to H2O(l) +
CO2(aq). This indicate that either H2O(l) + CO2(aq) is used or H2CO3, not all three
compounds at the same time. This is supported by CODATA which does not include
H2CO3. The problem has not been addressed here, but needs to be solved in the future.
IR-drop is due to the resistance in the liquid phase. The potential drops and the
transported current is lower. This enters as a linear term in the net current density. IR
drop is not included in the derived equations above.
- 50 -
62
CO2 corrosion
CO2 corrosion
The amount of literature related to CO2 corrosion is extreme. A number of review

papers which gives an overview of the subject have been published. A brief historical
overview was given by Crolet21,22. Schmitt23 gave one of the first overviews of CO2
corrosion literature in the beginning of the 1980s. He stated that CO2 exhibited higher
corrosion rate compared to HCl at the same pH. This is one of the reasons why CO2
corrosion receives so much attention. Modelling is not possible using standard acid
corrosion models.
A number of books were published 24,25,26, they summarized the latest development
during the beginning of the 1980s. The focus on CO2 corrosion in the oil and gas
industry has increased exponentially since then. One of the main areas is cost. This
was discussed by Kermani and Harrop27. During the 1990s some of the more
advanced CO2 corrosion models were developed using computers. Dawson et al.28,29
summarized many of these models and Nesic30 gave a detailed overview of the
mathematical models which are the basis of most models today, ten years later.
The most comprehensive reviews of CO2 corrosion was presented by Kermani and
Morshed31. It gives a general overview and insight of the CO2 corrosion literature and
key corrosion parameters. Recently Schmitt and Hrstmeier32 presented an overview
of many of the discussed subjects within CO2 corrosion.
Flo M
w
lo o
p
Tra Fluid
nsp
pro
ort
per
tie
p
r
o
S
p
c
erti s
ale
Allo
e
s
ys
s
Ele
ctro
ch e
mi s
try
le
ca
es
typ
ll s
Fu
es
w
Flo
nc
rie
e
p
Ex
and experim
en
Ma
ie s
se
Ca
l
ies ode
d
stu
Resea
r ch
ts
Ind
ivities
t
c
a
try
rrosion literature
t
ivi
ct
us
CO2 co
ry
eo
Th
ti
ma
e
th
l
ca
ch
roa
p
ap
CO c o
2
rrosion literature
Figure 12: Overview of the topics discussed in the CO2 corrosion literature.
Figure 12 gives an overview of the topics discussed within the CO2 corrosion
literature. The industry and universities are the two parties which have different
approaches to corrosion. Industry related literature concerns mainly production
facilities issues. Field cases are often discussed in order to give a direct application of
the obtained solutions. Research institutions on the other hand often explore the
- 51 -
63
CO2 corrosion
theoretical or experimental details. This is often related to electrochemical kinetics of

CO2 corrosion. The industry and research meet in the middle. The common focus
areas are corrosion rate measurements, modelling and work related to improving the
understanding of the CO2 corrosion mechanism. Corrosion rate is a function of many
variables and the topics discussed over the years are
Corrosion rate measurements/models
o General CO2 corrosion mechanism
Alloys/steel types
o Effect of Fluid properties
Temperature, pressure, composition
pH, ionic strength
Additional components (H2S, acetic acid, O2)
Inhibitors
Transport properties
o Corrosion scales
Equilibrium/solubility
Kinetics
o Mass transport/flow
o Top of the line corrosion
o Effect of liquid wetting
The variables are investigated both experimentally and by models. The experimental
work is used in order to get a first hand impression and validation of theories and the
modelling is performed to get a general understanding of the corrosion mechanisms.
The CO2 corrosion mechanism is an electrochemical process and the corrosion
exchange current is a function of the liquid phase surrounding the alloy. The kinetic
parameters are affected by the composition of the alloy.
Temperature and pressure are two of the main variables, especially the partial
pressure of CO2. pH and ionic strength are also important and plays a key role in the
corrosion mechanism. Other species like H2S, oxygen, or acetic acid, contribute
similar to CO2 in the corrosion process. An overview of literature related to acetic
acid was presented by Dougherty33. The CO2 corrosion models are currently being
extended in the open literature, in order to take care of the above phenomena.
Corrosion scales has been known for a long time to play an important role in the
protection of the steel surfaces. The scales form a diffusion boundary which prevents
the corrosive species from diffusing to the surface and corroding it. The solubility of
the scales is closely related to the pH and recently the focus has been the solubility
and kinetics of scale precipitation. Iron carbonate, FeCO3, is typically formed as a
corrosion product. The transport properties of ions and the diffusive film layer
thicknesses has also become a focus area.
CO2 corrosion is normally observed in the liquid phase at the bottom of the pipeline.
Sometimes it corrodes under the roof of the pipeline. This is also known as top-ofline-corrosion (TLC), TOL corrosion, or TOP corrosion. TLC is CO2 corrosion, but it
is often related to the corrosion by acetic acid. It is connected to the condensation rate
of water and the physical transport of iron off the surface as shown by figure 13. The
increased dissolution is due to the acidification by acetic acid and the following
dissolution of FeCO3 as indicated by figure 2. The problem is a great concern to the
oil and gas industry since the rate of corrosion is higher in the top than in the bottom.
Corrosion inhibitors are usually injected in the pipelines. TLC is difficult to inhibit
- 52 -
64
CO2 corrosion
because the liquid ends up in the bottom of the pipes. Therefore inhibitors will never
reach the top of the line.
TLC (Top-of-the-Line Corrosion)
Buildup of water on pipe wall
Liquid phase from injected

inhibitor and condensation
H2O
Condensation of H2O
CO2
Dissolution of CO2
in water
Figure 13: Cross section of pipeline showing the condensation of water which influences the corrosion
process under the pipeline roof.
Volatile corrosion inhibitors (VCI) are sometimes used to prevent TLC. VCIs
condense in the top of the pipeline and protect against corrosion. VCIs are not
preferred due to the volatility and hazardous properties. Pipelines may also be
inspected regularly in order to inhibit TLC. This is done using special pigs which wet
the top using inhibitors from the bottom liquid phase.
The mechanism of CO2 corrosion in wet gas pipelines is generally the same in oil
pipelines. Therefore one of the new focus areas in corrosion literature has become
wetting of the surfaces since corrosion is only observed in water wet pipelines.
5.1
CO2 corrosion types
CO2 corrosion may cause four types of corrosion:

Pitting corrosion
Mesa attack
Flow induced corrosion
General uniform corrosion
Pitting or localized corrosion is a stochastic process which is difficult to predict. It is
observed as small areas of surface which are attacked heavily. Deep cavities are
formed and the corrosion rate is very high. The rate if often mm per year and failures
are most often related to pitting. Pitting is often observed at low fluid velocities. An
explanation is that pitting may be caused by a local electrochemical concentration cell
between the pit cavity and the pipe surface.
Pitting has been known since the beginning of CO2 corrosion research as discussed by
Baylis34,35. It was discussed by Xia et al.36 in relation to film formation and recently
by Sun et al.37,38,39 and Kvarekvl40. The stochastic phenomena was modelled by Xiao
and Nesic41,42. The oil and gas industry often use the empirical ASME B31G method43
to evaluate the remaining lifetime of pipelines attacked by pitting.
Mesa attack is also a stochastic pitting process. It is similar to normal pitting but the
mechanism is only observed at high flow rates. Figure 14 shows the mechanism of
mesa attack. It is caused by the dissolution of iron under the corrosion film. At a
certain point the film breaks and it is removed by the high flow rate. The dissolution
and removing of more film is a continuous process where more iron dissolves and
- 53 -
65
CO2 corrosion
film breaks off. Corrosion rate is very high and equivalent to normal pitting. Mesa
attack has been known since the first corrosion model was published44. The
phenomenon was discussed by Dunlop et al.45 and Videm and Dugstad46. It was not
until the late 1990s that Nyborg47,48,49 discovered the mechanism sketched in figure
14, using a video camera. A model which describes how to avoid mesa attack, is
available50. No models can predict mesa attack.
Steel
Steel
Steel
Figure 14: Illustrates mesa attack. A: High flow rate. B: Iron dissolves under corrosion product. C:
Corrosion product breaks off and more iron dissolves. Corrosion product breaks off continuously.
Flow induced corrosion typically contribute to mesa attack at high flow rates by
eating away the corrosion scales due to liquid vortices. A theoretical model was
discussed51.
General or uniform corrosion is defined by the dissolution of the corroding surface
which dissolves evenly. General corrosion is by far the most widely researched form
of corrosion in the literature. There are models to calculate electrochemical kinetics,
fluid properties, transport properties and a vast number of models to predict corrosion
rate. Three types of rate models are found in the literature:
Empirical or purely correlative models
Semi-empirical models
Mechanistic models
The number of models has grown considerably during the last two decades. The
performance, review, and comparison of the models have been published in a number
of studies. Kermani and Smith52 compared some of the semi-empirical models for the
European Federation of Corrosion. Similarly Nordsveen et al.53, Srinivasan54, and
Wang et al.55 gave brief overviews of the available software in the industry. Recently
Nesic et al.30, Woollam and Hernandez56, and Nyborg57,58,59 evaluated and described
the majority of empirical, semi-empirical, and mechanistic models. The history and
mathematical schemes were discussed. This was summarised by Kapusta et al.60.
5.2
Empirical or purely correlative models
Empirical models are correlation of measured corrosion rate in mm/year to a number

of input variables. The models have empirical parameters which are regressed to
corrosion rate data. The principles of the model have no physical meaning. These
models have won some acceptance in the industry since it is unimportant to the
companies how the predicted corrosion rate is obtained or at least unimportant which
scientific framework was used. Some models are purely data driven, where corrosion
rate is predicted by interpolating between previously observed cases to new unknown
cases61. Other models are fitted to databases using linear correlation62,63,64,65 or nonlinear correlations66. Some of the models are known by name e.g. the COPRA67,68 and
SWEETCOR69 correlations. Neural network models have become popular for
- 54 -
66
CO2 corrosion
correlating corrosion rate. Cottis et al.70 and Sinha and Pandey71 presented some of
the network theory. The principle of a model which is continuously being developed
was presented by Nesic et al.72,73,74,75. The drawback of these models is that they all
rely on huge databases, extrapolate very poorly, and can not be used for systems
which where not used in the parameter fitting process. The models require much new
data in order to extend the applicability to other systems. These models have not been
a focus of this study.
5.3
Semi-empirical models
These models are similar to the empirical models since they are fitted to corrosion rate
data. The semi-empirical models are set up to resemble some of the physical
principles of the real system and are not pure correlation. The unknown parameters in
the model are fitted in order to reproduce measured experimental corrosion rate data.
Semi-empirical models are some of the most frequently applied models. There are
two reasons for this. First of all they have been available for a long time and are
evaluated. They are simple and therefore easy to implement and use. They have been
fitted to measured corrosion rate data.
The drawback is the same as empirical models. The fitted parameters do not give
information of the mechanism of the system, they extrapolate poorly and it requires
much new data in order to extend the usage to other systems.
The amount of semi-empirical models is great and covers a subject in it self. Here the
significant works within the area are given. De Waard and Milliams presented the first
model in 197576,77,44, it has since then become very popular. The model was based on
a simple electrochemical model which depended on pH and temperature. From this
model the corrosion rate was related to an exponential function of pH. A simple
correlation of pH with partial pressure of CO2 was given. The two combined
equations were given and the corrosion could be calculated by:
log CR mm / year 7.96
2.32 103
5.55 103 t qC 0.67 log pCO2 bar
t qC 273
(183)
Where CR is the corrosion rate in mm/year. The model was fitted to corrosion rate
measured for grit blasted carbon steel. Protective layers of FeCO3 were not formed.
Some of the first experimental relations were presented by Townshend et al.78. The
success of the model lays in the simplicity, the experimental basis and the
nomogram/nomograph presented in the papers. A nomogram is a simple figure which
gives the user an easy tool to directly evaluate a complicated relation, using a ruler.
The model was updated approximately ten years later79. The empirical extension
could take care of protective FeCO3 scales and triethylene glycol or methanol. The
implementation was made as correction factors to equation (183). Which indicates
that corrosion rate was calculated using (183) and then a correction factor was
multiplied to obtain the final corrosion rate. It was discussed how to differentiate
between corrosion rates in the top and bottom of the pipeline. In the spirit of the
previous publication the model was again extended80,81 by empirical correction
factors. The scale related corrosion was improved and therefore the overall
temperature dependence was improved. A more accurate pH correlation was
introduced and the model could now also take care of condensation rate,
- 55 -
67
CO2 corrosion
hydrocarbons and effect of MEG and diethylene glycol (DEG). The inclusion of
inhibitor efficiency was discussed.
The whole principles of the model was revised in 199382,83. This was done since the
old models could not take flow velocity into account. The new model was a film
resistance model. Two contributions were considered, diffusion resistance and
reaction resistance. Parameters in the model were fitted to experimental corrosion rate
data. The new model was further improved84 in 1995 to take care of steel composition
and steel types. A correction factors for FeCO3 were again introduced and improved.
Recently a correction factor for oil wetting were fitted to experimental data85,86,87 and
addition of acetic acid has been discussed88. The model has been used as a sub-model
for evaluation of TLC as discussed by Vitse et al.89 and as a sub-model in multi flow
simulators53,90,91. Models which are similar to the pre-1993 model of de Waard were
presented by Dugstad et al.92 and Mishra et al.93.
Another type of model was presented by Efird94 and followed up by Jepson and coworkers95,96,97,98. The model relates the corrosion rate to the shear stress, Ww, by the
following relation:
CR
a W wb
(184)
Where a and b are variables which depend on the system.

Many models are denoted by name since they have been developed as computer
programs. Some authors give all or close to all model details. Other authors avoid
giving enough information and the model can be difficult to reproduce. The oil and
gas companies typically have each their CO2 corrosion model implementation.
The NORSOK M-506 model50,99,100,101,102 was developed by the Norwegian institute
of standards together with Statoil, Hydro, and IFE. It is the only documented and
publicly available program. Several other programs are used in the literature. They are
not described in detail and are only available to the companies. A lot of the models
are de Waard type, which indicates they have close resemblance to the models of de
Waard, e.g. HYDROCOR developed by Shell60,103,104, LIPUCOR developed by
Total105,106, Cassandra developed by BP107,108, CORPOS a position dependent model
developed by CorrOcean109,110, and PREDICT developed by
Intercor111,112,113,114,115,116. Other types of programs where the principles are not
revealed but the models are fitted to experimental corrosion rate data are given by
USL/ULL(University of Louisiana at Lafayette)117,118,119,120 developed for an
industrial consortium and CORMED developed by Elf121,122,123,124,125,126,127,128. There
are other semi-empirical models available, but the above have received some
attention. These models have not been the focus of this study since they do not
identify the corrosion mechanism but merely correlates corrosion rate measurements.
5.4
Mechanistic models
Mechanistic models are chemically mechanistic in the sense that they rely only on
chemical relations and the model has not been fitted to experimental corrosion rate
measurements.
- 56 -
68
CO2 corrosion
Gas Phase
H2O
CO2
Liquid Phase
CO2
CO32HCO3
HCO3-
H2O
H2
Fe2+
MEG
NaOH
Electrochemical
CO2 corrosion Model
CO2
Steel
Figure 15: The CO2 corrosion mechanism. CO2 dissolves in the condensed aqueous liquid phase
diffuses to the surface and dissolves iron by an electrochemical mechanism.
Figure 15 shows the principles of all available mechanistic models. The liquid phase
consists of water and additives. CO2(g) dissolves in the liquid phase, it hydrates,
dissociates to HCO3- and CO32- and makes the liquid acidic. CO2(aq) diffuses to the
pipe surface and reacts cathodically at the surface by using electrons and producing
HCO3- and H2(aq):
CO2 aq H 2O l e R 12 H 2 aq HCO3
(185)
H e R 12 H 2 aq
(186)
In corrosion literature CO2(aq) is often substituted by H2CO3(aq). There are slight

inconsistencies between thermodynamic modelling and corrosion literature. This is
pointed out in section 4.4. The above reactions proceed only if the electrons are
supplied by other reactions. The anodic process supplies the electrons by dissolution
of iron:
Fe s R Fe 2 2e
(187)
The produced species diffuses away from the surface. If the conditions are right,
FeCO3(s) precipitate on the pipe wall and forms a corrosion protective carbonate
layer. Figure 16 shows how FeCO3 becomes a diffusion barrier and the reaction rate
is lowered due to lack of CO2(aq) and similarly the electron transfer is smaller and
less iron dissolves.
- 57 -
69
CO2 corrosion
Gas Phase
H2O
CO2
Liquid Phase
CO2
CO32HCO3
CO2
Diffusion Control
FeCO3 precipitate
Diffusion control:
Lower corrosion rate
Fe2+
H2O
MEG
NaOH
Electrochemical
CO2 corrosion Model
FeCO3
Steel
Figure 16: The CO2 corrosion mechanism. If pH is high enough and if the temperature and liquid
composition is correct, then FeCO3 precipitate and forms a diffusion controlled mechanism which
inhibits the electrochemical corrosion process.
The core of all mechanistic corrosion models is the electrochemical kinetic model.
There are different variations of the mechanism and the above represented by (185) to
(187) is one of the suggested reaction mechanisms. The mathematical formulation of
the mechanism is commonly given by an equation similar to (142) or (175).
Not all models include FeCO3 scales. This was the case for the model by Turgoose et
al.129 from 1990. It is one of the first mechanistic CO2 corrosion models. The authors
showed the principles of solving an electrochemical model coupled to a diffusion
model. This was done in order to account for the Nernst diffusion layer, since the
concentration at the surface and in the bulk is not the same, as indicated by equation
(136). Much of the principles resembles the crevice corrosion model by Watson and
Postlethwaithe130 of the same year.
The Turgoose et al.129 model was a simulation of a rotating disk electrode. The
electrochemical model were therefore not given by a standard Volmer-Butler
mechanism similar to (136). Instead the current density was calculated by a
hydrodynamic equation which related the current density and Nernst diffusion layer
thickness to the diffusion coefficients and the rotation frequency.
Ficks law was assumed to be valid and the solution of the mathematical problem was
treated as a hyperbolic PDE (partial differential equation), where the diffusion model
was discretized in the diffusive layer in order to solve the system numerically. The
electrochemical model and the bulk composition were boundary conditions as shown
in figure 17.
- 58 -
70
CO2 corrosion
Figure 17: Principles of a discetized CO2 corrosion model. Green lines are boundaries. Blue lines are
discretization. Compartment numbered j, thickness 'x, and total thickness GN.
Turgoose et al.129 set up a mol balance for every discretized compartment, j, by use of
equation (43) written for compound i for a system where Ri=0, using (73) and (87)
and assumed a stagnant solvent, us=0:
2
dci
f d ci
i 2
dt j
dz j
(188)
This gives a set of n coupled PDEs which can be solved numerically on a computer.
The right hand side of (188) was calculated by the following second order Taylor
approximation.
c j 1 2c j c j 1
w 2 ci
2
wz j
'z
(189)
Combining (188) and (189) results in an equation which converts equation (188) to a
set of n coupled ODEs (ordinary differential equations). It rearrange to:
dci j
if dt
'z
j 1
2c j c j 1
The equation system was solved by using if dt
'z
(190)
=0.45 for H+, which had the
highest if . The concentration at the next time step was calculated assuming dci |'ci
and therefore ci k 1 = ci k 'ci k for time step k+1.
The initial composition at t=0 was assumed to be the equilibrium concentration at the
given CO2 pressure. Boundary conditions at the Nernst diffusion layer, z=GN, was
- 59 -
71
CO2 corrosion
assumed to be constant for tt0 and equal to the bulk concentration. Boundary
condition at z=0 for tt0 follows Faradays law. This indicates that the species
converted at the surface, produces or consumes current. The flux is related to the
current density by Faradays law:
Ni
Qi
ni F
ii
(for z=0)
(191)
Where n and Q are the stochiometric coefficients of the electron and the reacting
species. The flux of non-reacting compounds are zero. Some of the fluxes are for
example given by:
N Fe2
NH
N H 2O
iFe2
2F
iH
F
(for z=0)
(192)
(for z=0)
(193)
(for z=0)
(194)
The flux of iron has a negative sign since Fe2+ is produced at the surface. This is
opposite for H+ as shown by (186) and (187). The flux of water is zero since it does
not react on the surface. The model was used for evaluating the concentration profiles
up to the surface and was not used for prediction corrosion rate.
Crolet et al.131,132,133 presented a qualitative model which used some of the principles
similar to Turgoose et al.129.
Nesic134,135 gave a more physically correct model which included the carbonate
equilibria in the liquid phase. The boundary conditions at the Nernst diffusion layer,
z=GN, was similar to Turgoose et al.129. The boundary conditions at z=0 for tt0 was
set up to be a diffusion controlled scheme. Nesic made it clear that the model could
not predict corrosion rate but the model could, similar to Turgoose et al.129, predict
the concentration profiles in the diffusion layer. The model showed that pH increased
towards the surface by a value of 1.5. This effect was also shown by Turgoose et
al.129. They showed the increase was less than 0.01.
Pots136 presented the first mechanistic model in 1995 which could predict corrosion
rate. The boundary conditions were equivalent to Turgoose et al.129 except that the
kinetic model at the surface was calculated by a Volmer-Butler model similar to (138)
. The exchange current density, i0, was a function of the surface concentrations. These
were calculated from the diffusion model. The flux of species was assumed to follow
the more detailed Nernst-Planck equation, (96), instead of Ficks law. The potential
gradient, I, was evaluated as part of the model. The results showed that the model
could predict both activation and diffusion controlled corrosion schemes.
Dayalan et al.137 used a different approach. They modelled the mass transport in the
Nernst diffusion layer using the Chilton-Coulburn correlation. The boundary
conditions were similar to Pots136 and they also applied a Volmer-Butler model for the
surface kinetics. The model was solved as a non-linear problem of fifteen variables
and fifteen unknowns. Six surface concentrations, four current densities, four
equilibrium potentials and the corrosion potential.
- 60 -
72
CO2 corrosion
Sundaram et al.138 made an elaborate mathematical description of the corrosion

phenomena in 1996. Figure 18 shows the model principles.
Figure 18: Principles of Sundaram et al.s138 CO2 corrosion model which includes three diffusion
models. The distances to the interfaces are GP, GN, and Gb from the surface.
The core of the corrosion model was still the electrochemical kinetic model at the
surface. The diffusion model was extended compared to previous models. It contained
three compartments instead of one as the Turgoose et al.129 and Pots136 models in
figure 17. The turbulent layer was described by a mass transfer correlation in order to
take care of two-phase flow. Activity coefficients were mentioned in order to perform
an exact calculation of the phase composition. The diffusion layer was set up similar
to Pots136, with a Nernst-Planck description of the flux. The porous layer was assumed
to follow a Ficks law diffusion and it was assumed diffusion controlled. This indicate
that the surface concentration of species were assumed low compared to the interface
concentrations at z=GP. The mass balance was similar to (43) solved in steady state,
dci dt 0 , and Ri z 0 . Parameters used in the model were not given and the model is
not easy to reproduce. The model was further described and simplified during 2000 by
High et al.139.
A model similar to Nesic134,135 was discussed by Kvarekvl140 in 1997. The model
could only reproduce diffusion controlled corrosion mechanisms. It was extended in
order to account for slow reactions in the liquid phase by deriving a term for Ri as
given in equation (43). The liquid film thickness was calculated by using a correlation
of wall shear stress. The model showed that the carbonate concentration was
distinctively higher at the surface compared to the bulk concentration. This could
indicate formation of a protective porous FeCO3 layer which was not included in the
model.
Zhang et al.141 and Rajappa et al.142 discussed a mass transport equation similar to
Dayalan et al.137. They tested different correlation, one being the Chilton-Coulburn
model. Activity coefficients were included for the bulk composition calculations
similar to Sundaram et al.138. Many of the electrochemical rate constants were not
given. The same year Dayalan et al.143 extended their model to include a porous
diffusion layer and the set up is sketched in figure 19.
- 61 -
73
CO2 corrosion
Figure 19: Principles of Dayalan et al.s143 CO2 corrosion model which includes two diffusion models.
The distances to the interfaces are GP and GN from the surface.
The mass transport was modelled using transport correlations and the transport in the
porous layer was modelled by scaling the diffusion coefficients and the effective
diffusivity, eff . This was done by assuming the diffusion coefficients in the porous
media were related to the free diffusivity, , by:
eff
(195)
where W is the tortuosity. The equation is erroneous and should instead be given by:
eff
W2
(196)
as discussed in appendix C.1, where I is the porosity. The work was continued by
Wang et al.55 in order to test new mass transport correlations for slug flow. The model
was simplified and the porous layer and activity coefficients were neglected. Later
Anderko et al.144,145,146,147,148,149 (OLI) gave an electrochemical model coupled to an
activity coefficient model. The majority of surface activities were assumed to be equal
to the bulk activities. Some compounds were diffusion controlled by applying a mass
transport correlation equivalent to Dayalan et al.137.
A discretized two-dimensional PDE model was discussed by Wang and
Postlethwaite150,151,152. The model did not include a porous corrosion product film as
illustrated in figure 18 and 19. Sridhar et al.90 presented also a 2D model during 2000
similar to Wang and Postlethwaite150,151,152. The model was not intended for CO2
corrosion prediction. Instead a general equation scheme for mechanistic corrosion
models was presented and results applied for crevice corrosion was given. Few results
of CO2 effect were shown.
Recently Nesic et al.153,154,155 demonstrated an advanced mechanistic model known as
the KSC model. The setup was comparable to the work by Sundaram et al.138 and
Pots136. Figure 20 shows a sketch of their discretized diffusion model.
- 62 -
74
CO2 corrosion
Figure 20: Principles of Nesic et al.s154,156,155 CO2 corrosion model which includes one diffusion
model. The distances to the interfaces are GP and Gb from the surface. The diffusion coefficients are
position dependent. GP signifies a change in porosity and not a boundary.
The model was solved by applying a Nernst-Planck description of the flux in the
whole diffusive layer, from the surface to the bulk. The used mass balance was similar
to (43) except that porosity was empirically included by:
d H ci
dt
d H 2 Ni
dz
H R
(197)
Where H was the porosity. The initial condition was set equal to the equilibrium bulk
composition. The boundary condition was a kinetic electrochemical surface model
and a constant bulk composition The electrochemical model was linked to the
diffusion model by use of (191). A position dependent porosity and a position
dependent diffusion coefficient were used in order to account for the porous layer and
the turbulence. H and if were therefore functions of z. It was assumed, in their test
cases, that the porosity was H=0.1 from z=0 to z=GP=10Pm. The remaining free
diffusion from GP to Gb was modelled using H=0.6. An effective diffusion coefficient,
efff ,i , was used in the Nernst-Planck equation in order to account for turbulence. The
diffusion coefficient contained a term from free diffusion, if , and a contribution
from turbulence, if,t :
efff if i,tf
(198)
if,t was zero at zdGP and increased towards z=Gb. They concluded that the porosity
has a greater influence on the model result compared to the film layer thickness. The H
and GP were considered input parameters and Gb was calculated using correlation of
- 63 -
75
CO2 corrosion
the Reynolds number. The model was apparently simplified by Nesic et al.156,157,158 in
order to calculate H and GP. A Fick term was used instead of the Nernst-Planck
formalism. Nesic et al.159 also linked it to the stochastic model of Xiao and Nesic41,42
for estimation of pitting and it was linked to a two-phase flow model160,161,162 for
evaluation of wetting.
A model scheme for crevice corrosion was presented by Heppner163. They did not
focus on CO2 corrosion. Instead they presented an elaborate description of the
mechanistic equations which included a Pitzer model for bulk calculation and an ideal
Nernst-Planck diffusion model. The diffusion model was not linked to an activity
coefficient model. A new model was given by Song et al.164,165,167 during 2004. It
followed the principles of Pots136 using Ficks law instead of the Nernst-Planck
equation. The model did not include a porous diffusion layer.
It seems the most recent development will be a publicly available mechanistic
corrosion model made by Nesic and co-workers166. One of the benefits will be a
feature to plug-in 3rd party modules and thereby create a framework for further
corrosion model development. The model will be called FREECORP and will be
available during NACExpo 2008, but no results have been presented yet.
The focus of this study is the mechanistic corrosion models. It is clear from the above
list of empirical, semi-empirical, and mechanistic models, that a great amount of work
has already been performed for constructing a general CO2 corrosion model. The
above discussion shows the principles and assumptions for creating a mechanistic
corrosion model. Several crude assumptions are still used in the corrosion models;
these are discussed and evaluated in the following sections.
5.5
Electrochemical models
The core of all mechanistic CO2 corrosion models is an electrochemical kinetic

model. Two types of models are usually used either the Levich168 model used by
Turgoose et al.129 or a kinetic model similar to the Vomer-Butler description (142).
The Levich model is a hydrodynamic model used for RDE setups. The mechanism is
always diffusion controlled and the Nernst diffusion layer thickness is well defined
due to the rotation of the electrode. The Volmer-Butler model is normally used since
it is independent of the hydrodynamics and may be used for modelling activation
controlled mechanisms.
The steel composition is reflected in the kinetic model and the kinetic parameters are
directly related to the composition of the alloy as discussed by Lopez et al.169. The
mechanism used by most authors is represented by the steel dissolution (187) as the
anodic process. The cathodic reaction e.g. (186) is typically also observed in acid
corrosion. The reaction mechanism is not the same in all acids. CO2 reacts differently
compared to HCl as pointed out by Smitt23. CO2 corrosion is a combined mechanism
of (185) and (186). Therefore CO2 corrosion is faster compared to normal acidic
corrosion. The reason is that more current flows and the dissolution process is faster.
The reaction mechanism of CO2 corrosion has been discussed by many authors. An
overview of the most important works was recently given by Dugstad170 and was
briefly given Ogundele and White171, Dawson28, and Nesic et al.30.
Authors of electrochemical CO2 corrosion literature have dealt with three aspects of
the electrochemical mechanism: The cathodic process, the anodic process of iron
dissolution, or both.
The cathodic process is typically a reduction of H+ to H2. The total reaction contains
several elementary reactions and the total process is given by (186). Stern172 was one
- 64 -
76
CO2 corrosion
of the first authors who discussed the cathodic process principles. Later Smith and
Rothmann173,174,175 argued that the hydration of CO2 and direct reduction on the
surface, contributed to the current transport as described by (185). This was also
discussed by Wieckowski et al.176,177, Schmitt23, Eriksrud and Sntvedt204, and Gray
et al.179. Hurlen et al.180,181 suggested that direct reduction of HCO3 could have an
influence on the CO2 corrosion mechanism which was supported by Ogundele and
White171 and Gray et al.182. Nesic et al.183,184 presented the recently accepted
mechanism which was given by the two overall reaction represented by (185) and
(186).
The anodic process was discussed by Bockris et al.185,186 who gave one of the first
well described anodic mechanism of iron dissolution and Drazic187 summarised the
older significant works on the subject.
The mechanism by Bockris et al.185,186 was later used in many studies e.g. Gray et
al.179,182. Smith and Rothmann188 stated that the total anodic reaction was given by
(187). Davies and Burstein189 also discussed the dissolution process. Nesic et al.190,30
recently spread some doubts to the accuracy of the work of Bockris et al.185,186 at
pH>4. Nesic et al.190 stated that the previous anodic mechanism of iron dissolution
was erroneous and that the reactions were very dependent of pH. It was stated that the
mechanism changed completely between pH<4 and 4<pH<5 and again at pH>6.
Both cathodic and anodic currents were modelled by Rogers and Rowe191. The result
should be seen as a first attempt to estimate the electrochemical process. A complete
mechanism of anodic and cathodic current was presented by Nesic et al.154,155.
Recently Anderko et al.144,145,146,147,148,149 presented an extended electrochemical
model based of activities as given by (142) coupled to a Pitzer model. Unfortunately,
no parameters were published and the model can not be reproduced.
The number of works related to the electrochemical investigation of CO2 corrosion is
not elaborate. Many of the above studies give qualitative descriptions of the process
and there is still a need for a complete Volmer-Butler model which is valid in the
whole temperature, pressure, composition, and pH range. It is noteworthy that even
though an electrochemical model is the core of all mechanistic CO2 corrosion models,
then no one has published a precise description of the electrochemistry. Nesic et
al.154,155 give to date the most complete electrochemical model even though it can be
improved.
5.6
Improving CO2 corrosion models
5.6.1
Assumptions of mechanistic models
The above sections confirm that there is a noticeable amount of predictive CO2
corrosion models. The focus of this study is to improve the mechanistic models
instead of repeating the already known phenomena. The existing models apply a
number of assumptions. Two important phenomena are discussed in this study: The
non-ideality of the liquid phase and the properties of the porous FeCO3 corrosion
products which have not been thoroughly discussed before.
5.6.2
Thermodynamics of the fluid phases in CO2 corrosion
Table 2 shows a typical fluid composition in a wet gas pipeline. The ionic strength (I)
is very high, it is approximately 3 to 10 mol/kg H2O at the inlet and 0.1 mol/kg H2O at
- 65 -
77
CO2 corrosion
the outlet due to dilution by the condensed water. This indicates that the solutions
behave thermodynamically non-ideally. The well known Debye-Hckel limiting law
activity coefficient model is valid at Id0.01 mol/kg H2O and the extended DebyeHckel law is valid at Id0.1 mol/kg H2O. The working window can be extended
slightly by using the Davies rule192,193 which is valid at Id0.3 mol/kg H2O. An
advanced activity coefficient model is required in order to model the solutions
relevant for the work in this study. The Pitzer194,195 equation is typically valid up to
I=6 mol/kg H2O. None of these models can be used for multi-solvent systems.
It is assumed, in the majority of the above mentioned mechanistic CO2 corrosion
models, that the electrolytic solutions behave ideal. It is therefore assumed that
I<0.001 mol/kg H2O. It is noteworthy that an activity coefficient model was not
applied, since the condition in a wet gas pipelines are up to I=10 mol/kg H2O. It is not
only the bulk calculation which can be improved. The diffusion calculation was
shown by Newman5 and Wesselingh and Krishna6 to deviate from Ficks law at I>1
mol/kg H2O, and the electrochemical surface reactions are in principle not functions
of the surface concentrations, but functions of the activities as shown by (136).
Few of the mechanistic CO2 corrosion models have an integrated activity coefficient
model. This was done in order to account for the non-ideality of either the bulk phase
or the electrochemical surface calculations. None of the corrosion models were
improved by an activity coefficient in the diffusion calculation, and none were
improved in both electrochemical and bulk calculations. An activity coefficient model
should be implemented in order to account for non-ideality of the bulk calculation, the
diffusion calculation, and the surface kinetic models.
The authors who discussed activity coefficients in mechanistic CO2 corrosion models
were Sundaram et al.138 who used a Davies approximation and a Henrys law. This
was done in order to account for non-ideality of the bulk. They assumed ideality in
electrochemical and diffusion calculations. Zhang et al.141 and Rajappa et al.142
similarly used a Davies rule for their bulk calculations and assumed ideality in the
remaining parts of the model. A Pitzer model was applied in the electrochemical
surface model by Anderko et al.144,145,146,147,148,149. The model parameters are
unfortunately not available. None of the activity coefficient models gave a satisfactory
description of the bulk and surface calculation since they can not be applied to high
ionic strengths and mixed solvents.
An advanced thermodynamic model has still not been incorporated in a mechanistic
CO2 corrosion model. A mechanistic CO2 corrosion model would include a full bulk
calculation by applying (41) in a speciation routine. It would include the diffusion
scheme as presented in equation (83), or one of the simplification by (92) or (117),
and a kinetic model as given by (136). All equation should be expressed in terms of
activities.
The equation scheme for doing a more correct bulk calculation, diffusion calculation,
and kinetic surface reaction model is presented in this study. The extended
UNIQUAC model given by Thomsen et al.196,197,198 has shown to be able to predict
activities of multi-solvent electrolyte systems up high ionic strengths. The model will
be used for showing how the bulk, diffusion, and surface kinetics calculations can be
improved. Results of the thermodynamic modelling are shown in chapter 7 and 9.
5.6.3
Properties of FeCO3
It has been confirmed within the CO2 corrosion literature, that FeCO3 plays a central
role in the modelling of CO2 corrosion. FeCO3 is also known as siderite and chalybite.
- 66 -
78
CO2 corrosion
FeCO3 was discovered as a corrosion product very early by Baylis34,35 during the
1920s. The precipitated corrosion film of FeCO3 forms a layer of impermeable
corrosion product which retards the corrosion process and lowers the corrosion rate by
diffusion control. Few authors continued the research immediately after Baylis.
Kooijmans199 confirmed that FeCO3 formed in 1938. Later Riesenfeld and Blohm200
discovered that acid gas scrubbers corroded and FeCO3 precipitated due to a specific
temperature and saturation cycle. Hackerman and Glenn201 and Sowards and
Hackerman202 confirmed by X-ray diffraction that FeCO3 indeed precipitated and
formed a protective corrosion film. It was not until the late 1970s FeCO3 again
received attention. De Waard and Milliams76 published their well known semiempirical corrosion model. They were aware that FeCO3 could form, but it was not
observed in their studies.
A number of studies are available which discuss the thermodynamics of FeCO3.
There is some confusion to which properties are reliable.
An objective of this study has been to collect standard thermodynamic properties of
FeCO3 to compare and discuss the most reliable values. This was done since FeCO3
receives increasing attention in the oil and gas industry.
The amount of literature focused on FeCO3 in relation to corrosion has grown
considerably since the beginning of the 1980s. Sontheimer et al.203 discussed
protective FeCO3 corrosion scales and Schmitt23, Eriksrud and Sntvedt178, and
Dunlop et al.45 determined that FeCO3 is a significant factor in the oil and gas
production for corrosion protection. At the same time Wieckowski et al.176 used an
electrochemical approach to show that FeCO3 change the surface properties of
corroding iron.
One method is typically applied for preventing corrosion of mild carbon steel. The
method is called pH-stabilization. In this method FeCO3 is precipitated to form a
protective FeCO3 corrosion film. This is done by increasing the pH using NaOH,
NaHCO3, or a similar base since the solubility of FeCO3 is closely related to the pH
of a solution as shown by figure 2. The method has been known since the beginning
of the 20th century as discussed by Whitman et al.205, Leybold206,207, and Baylis34,35.
The use of pH-stabilization for industrial purposes has a long history. Crolet and
Samaran208 mentioned some of the first cases but Dunlop et al.45 published one of the
first newer discussions of the subject. pH-stabilization continues to be one of the
focus areas of the literature. Recent work was given by Haarberg et al.209, Olsen et
al.210, and Dugstad and Seiersten211 who described a corrosion control system running
continually in order to precipitate FeCO3 to protect against corrosion.
The discussion of FeCO3 properties is given in chapter 8.
- 67 -
79
80
Literature cited
Literature cited
[1]
CorrosionCost.com: http://www.corrosioncost.com/infrastructure/gasliquid/index.htm,
http://www.corrosioncost.com/utilities/gas/index.htm and
http://www.corrosioncost.com/prodmanu/oilgas/index.htm, July 2007.
[2]
Wang, P. M.; Anderko, A,; Young, R. D. A speciation-based model for mixed-solvent

electrolyte systems. Fluid Phase Equilibria 203(1-2), 141-176, 2002.
[3]
Taylor R.; Krishna R. Multicomponent Mass Transfer. John Wiley & sons, 1993.
[4]
Medvedev, Oleg. Diffusion Coefficient in Multicomponent Mixtures. PhD. Thesis, IVC-SEP,

Technical University of Denmark, 2005.
[5]
Newman, John. Electrochemical systems. John Wiley & Sons, 2004.
[6]
Wesselingh, J. A.; Krishna, R. Mass transfer in Multicomponent Mixtures. Delft University

Press, first edition, 2000.
[7]
Harned, H. S.; Owen, B. B. The physical chemistry of electrolytic solutions. Reinhold

Publishing Corporation, New York, third edition, 1964.
[8]
Robinson, R. A.; Stokes, R. H. Electrolyte solutions. Butteroworths Publications, London,

second edition, revised, 1965.
[9]
Koryta, J.; Dvorak, J. Principles of Electrochemistry. John Wiley & Sons, 1987.
[10]
Rubinstein, Israel. Physical Electrochemistry principles, methods, and applications. Marcel

Dekker, 1995.
[11]
Tafel, J.; Emmert, E. The cause of the spontaneous depression of cathode potential during the
electrolysis of dilute sulphuric acid. Zeitschrift fuer Physikalische Chemie, Stoechiometrie und
Verwandtschaftslehre 52, 349-73, 1905.
[12]
Stern, M.; Geary, A. L. Electrochemical Polarization I. A Theoretical Analysis of the Shape of

Polarization Curves. Journal of the Electrochemical Society 104(1), 56, 1957.
[13]
Stern, M. Electrochemical Polarization .2. Ferrous-Ferric Electrode Kinetics on Stainless Steel.

Journal of the Electrochemical Society 104(9), 559-563, 1957.
[14]
Stern, M. Electrochemical Polarization .3. Further Aspects of the Shape of Polarization

Curves. Journal of the Electrochemical Society 104(11), 645-650, 1957.
[15]
Levich, V. Physical Hydrodynamics. Prentic-Hall, New York and London, 1962.
[16]
Nesic, S.; Pots, B. F. M.; Postlethwaite, J.; Thevenot, N. Superposition of Diffusion and
Chemical Reaction Controlled Limiting Currents - Application to CO2 Corrosion. JCSE 1,
paper 3, 1995.
[17]
Nesic, S.; Postlethwaite, J. A predictive model of CO2 corrosion. NACE Canadian Region
Western Conference, p. 397-417, Calgary, 1994.
[18]
Nesic, S.; Postlethwaite, J. Modelling of CO2 corrosion mechanisms. In: Trethewey, K. R.;
Roberge, P. R. Modelling Aqueous Corrosion. NATO ASI series, Kluwer Academic
Publishers, p. 317, 1994.
[19]
D. D. Wagman, W. H. Evans, V. B. Parker, R. H. Schumm, I. Halow and S. M. Bailey. NIST

Chemical Thermodynamics Database Version 1.1. J. Phys. Chem. Ref. Data, 1982, 11
Supplement No. 2.
[20]
J. D. Cox, D. D. Wagman and V. A. Medvedev. CODATA Key values for thermodynamics.

Hemisphere Publishing Corp., New York, 1989.
[21]
Crolet, J.-L. Which CO2 Corrosion, Hence Which Prediction? Progress in the Understanding
and Prevention of Corrosion, Vol. 1, p. 473-497, London, UK, The Institute of Materials,
1993.
- 69 -
81
Literature cited
[22]
Crolet, J.-L. Which CO2 Corrosion, Hence Which Prediction? European Federation of
Corrosion (EFC) 13, p. 1, 1994.
[23]
Schmitt, G. Fundamental Aspects of CO2 Corrosion. CORROSION/83 paper no. 43, NACE,
1983.
[24]
Newton, L. E; Hausler, R. H. CO2 corrosion in oil and gas production - selected papers,
abstracts, and references. NACE, 1984.
[25]
Hausler, R. H.; Godard, H. P. Advances in CO2 Corrosion. NACE, vol. 1, 1984
[26]
Burke, P. A.; Asphahani, A. I.; Wright, B. S. Advances in CO2 corrosion. vol. 2, NACE, 1985.
[27]
Kermani, M. B.; Harrop, D. The impact of corrosion on the oil and gas industry. SPE
Production & Facilities 11(3), 186-190, 1996.
[28]
Dawson, J. L.; Shih, C. C.; Bartlett P. K. N. Models and prediction of CO2 corrosion Erosioncorrosion under flowing conditions. European Federation of Corrosion (EFC) 13, paper no. 8,
151, 1994.
[29]
Dawson, J. L.; Shih, C. C.; Bartlett, P. K. N. Models and prediction of CO2 corrosion Erosioncorrosion under flowing conditions. Progress in the Understanding and Prevention of
Corrosion 1, p. 513, 1993.
[30]
Nesic, S.; Postlewaite, J.; Vrhovac, M. CO2 Corrosion of carbon steel - from mechanistic to
empirical modelling. J. Corrosion Reviews 15, 211, 1997.
[31]
Kermani, M. B.; Morshed, A. Carbon dioxide corrosion in oil and gas production - A
compendium. Corrosion 59(8), 659-683, 2003.
[32]
Guenter, Schmitt; Michaela, Hrstemeier. FUNDAMENTAL ASPECTS OF CO2 METAL

LOSS CORROSION PART II: INFLUENCE OF DIFFERENT PARAMETERS ON CO2
CORROSION MECHANISMS. CORROSION/06 paper no. 112, NACE, 2006.
[33]
Dougherty, James A. A Review of the Effect of Organic Acids on CO2 Corrosion.

CORROSION/04 paper no. 376, NACE, 2004.
[34]
Baylis, John R. Factors other than dissolved oxygen influencing the corrosion of iron pipes.
Journal of Industrial and Engineering Chemistry 18(4), 370-80, 1926.
[35]
Baylis, John R. How to avoid loss by pipe corrosion. Water Works Engineering 83, 13-4 31-2
35-6, 1930.
[36]
Xia, Z.; Chou, K. C.; Szklarska-Smialowska, Z. Pitting corrosion of carbon steel in carbon
dioxide-containing sodium chloride brine. Corrosion 45(8), 636-42, 1989.
[37]
Yuhua, Sun; Nesic, Srdjan. A Parametric Study and Modeling on Localized CO2 Corrosion in
Horizontal Wet Gas Flow. CORROSION/04 paper no. 380, NACE, 2004.
[38]
Sun, Yuhua. Localized CO2 Corrosion in Horizontal Wet Gas Flow. PhD Thesis, Ohio
University, Chemical Engineering, 2003.
[39]
Sun, Yuhua; George, Keith; Nesic, Srdjan. The Effect of Cl- and Acetic Acid on Localized
CO2 Corrosion in Wet Gas Flow. CORROSION/03, paper no. 327, NACE, 2003.
[40]
Kvarekval, Jon. Morphology of localised corrosion attacks in sour environments.

[41]
Xiao, Ying; Nesic, Srdjan. A Stochastic Prediction Model of Localized CO2 Corrosion.
[42]
Xiao, Ying. A Two-Dimensional Stochastic Model for Prediction of Localized Corrosion.

Master of Science thesis, Ohio University, Chemical Engineering, 2004.
[43]
Feigel, R. E.; Hayden, L. E.; Gomez, G. J. Manual for Determining the Remaining Strength of
Corroded Pipe - A supplement for the ASME B31 pressure piping. ASME B31G, 1991.
[44]
de Waard, C.; Milliams, D.E. Prediction of Carbonic Acid Corrosion in Natural Gas Pipelines.
Ind. finishing and surface coatings 28(340), 24, 1976.
- 70 -
82
Literature cited
[45]
Dunlop, A. K.; Hassell, H.L; Rhodes, P. R. Fundamental considerations in sweet gas well
corrosion. CORROSION/83 paper No. 46, NACE, 1983.
[46]
Videm, K.; Dugstad, A. Effect of Flow Rate, pH, Fe2+ Concentration, and Steel Quality on the
CO2 Corrosion of Carbon Steels. CORROSION/87 paper no. 42, NACE, 1987.
[47]
Nyborg, Rolf. INITIATION AND GROWTH OF MESA CORROSION ATTACK DURING

CO2 CORROSION OF CARBON STEEL. CORROSION/98 Paper no. 48, NACE, 1998.
[48]
Nyborg, Rolf; Dugstad, Ame. MESA CORROSION ATTACK IN CARBON STEEL AND
0.5 % CHROMIUM STEEL. CORROSION/98 paper no. 29, NACE, 1998.
[49]
Nyborg, Rolf; Dugstad, Arne. Understanding and Prediction of Mesa Corrosion Attack.
[50]
Halvorsen, A. M. K.; Sndtvedt, T. CO2 Corrosion Model for Carbon Steel Including a Wall
Shear Stress Model for Multiphase Flow and Limits for Production Rate to Avoid Mesa
Attack. CORROSION/99 paper no. 42, NACE, 1999.
[51]
Schmitt, G.; Bosch, C.; Mueller, M.; Siegmund, G. A PROBABILISTIC MODEL FOR
FLOW INDUCED LOCALIZED CORROSION. CORROSION/00 paper no. 49, NACE,
2000.
[52]
Kermani, M. B.; Smith, L. M. CO2 corrosion control in oil and gas production - design
considerations. European Federation of Corrosion (EFC) 23, 1997.
[53]
Nordsveen, M.; Nyborg, R.; Hovden, L. Implementation of CO2 Corrosion Models in the
OLGA Three-Phase Flow Code. Multiphase '99 (Cranfield, UK: BHR Group), 279-292, 1999.
[54]
Srinivasan, Sridhar. REVIEW OF COMPUTER APPLICATIONS IN CORROSION

SCIENCE AND ENGINEERING. CORROSION/00 paper no. 473, NACE, 2000.
[55]
Wang, Hongwei; Cai, Ji-Yong; Jepson, William Paul. CO2 CORROSION MECHANISTIC
MODELING AND PREDICTION IN HORIZONTAL SLUG FLOW. CORROSION/02 paper
no. 238, NACE, 2002.
[56]
Woollam, R.C.; Hernandez, S.E. Assessment and Comparison of CO2 Corrosion Prediction
Models. SPE International Oilfield Corrosion Symposium, paper 100673-MS, 2006.
[57]
Nyborg, Rolf. Overview of CO2 Corrosion Models for Wells and Pipelines. CORROSION/02
paper no. 233, NACE, 2002.
[58]
Nyborg, Rolf; Dugstad, Arne. Flow assurance of wet gas pipelines from a corrosion viewpoint.
Proceedings of the International Conference on Offshore Mechanics and Arctic Engineering OMAE 4, 125-132, 2002.
[59]
Nyborg, Rolf. Field Data Collection, Evaluation and Use for Corrosivity Prediction and
Validation of Models. CORROSION/06, paper no. 118, NACE, 2006.
[60]
Kapusta, Sergio D.; Pots, Bernardus F.M.; Rippon, Ian J. The Application of Corrosion
Prediction Models to the Design and Operation of Pipelines. CORROSION/04 paper no. 633,
NACE, 2004.
[61]
Khajotia, Burzin; Sormaz, Dusan; Nesic, Srdjan. Case-based reasoning model CO2 corrosion
based on field data. CORROSION/07, paper no. 553, NACE, 2007.
[62]
Adams, C. D.; Garber, J. D.; Walter, F. H.; Singh, C. Verification of computer modelled
tubing life predictions by field data. CORROSION/93 paper no. 82, NACE, 1993.
[63]
Garber, J. D.; Walters, F. H.; Alapati, R. R.; Adams, C. D. Downhole parameters to predict
tubing life and mist flow in gas condensate wells. CORROSION/94 paper no. 25, NACE,
1994.
[64]
Walters, Frederick H. Uphole parameters indicate corrosion in gas-condensate wells. Oil and
Gas Journal 92(3), 59-60, 1994.
[65]
Lyle, F.F.; Schutt, H.U. CO2/H2S corrosion under wet gas pipeline conditions in the presence
of bicarbonate, chloride and oxygen. CORROSION/98 paper no. 11, NACE, 1998.
- 71 -
83
Literature cited
[66]
van Bodegom, L.; Gelder, K. V.; Paksa, M. K. F.; Raam, L. V. Effect of Glycol and Methanol
on CO2 Corrosion of Carbon Steel. CORROSION/87 paper no. 55, NACE, 1987.
[67]
Gatzke, L.K.; Hausler, R. H. The COPRA correlation. A Quantitative assessment of deep, hot
gas well and its control. CORROSION/83 Paper no. 48, NACE, 1983.
[68]
Hausler, R. H.; Garber, J. D. The Copra correlation revisited. CORROSION/90 paper no. 45,
NACE, 1990.
[69]
John, R. C.; Jordan, K. G.; Kapusta, S. D.; Young, A. L.; Thompson, W. T. SweetCor: An
Information System for the Analysis of Corrosion of Steels by Water and Carbon Dioxide.
[70]
Cottis, R. A.; Oing, Li; Owen, G.; Gartland, S. J.; Helliwell, I. A.; Turega, M. Neural network
methods for corrosion data reduction. Materials and Design 20(4), 169-178, 1999.
[71]
Sinha, S. K.; Pandey M. D. Probabilistic neural network for reliability assessment of oil and
gas pipelines. Comput.-Aided Civ. Inf. Eng. 17(5), 320-329, 2002.
[72]
Nesic, Srdjan; Vrhovac, Miran. A Neural Network Model for CO2 Corrosion of Carbon Steel.
JCSE 1, paper 6, 1999.
[73]
Nesic, S.; Nordsveen, M.; Maxwell, N.; Vrhovac, M. Probabilistic modelling of CO2 corrosion
laboratory data using neural networks. Corrosion Science 43(7), 1373-1392, 2001.
[74]
Hernndez, S.; Nesic, S.; Weckman, G.; Ghai, V. Use of Artificial Neural Networks for
Predicting Crude Oil Effect on CO2 Corrosion of Carbon Steels. CORROSION/05 paper no.
554, NACE, 2005.
[75]
Hernandez, S.; Nesic, S.; Weckman, G.; Ghai, V. Use of artificial neural networks for
predicting crude oil effect on carbon dioxide corrosion of carbon steels. Corrosion 62(6), 467482, 2006.
[76]
de Waard, C.; Milliams, D. E. Prediction of Carbonic Acid Corrosion in Natural Gas Pipelines.
First international conference on the internal and external protection of pipes, paper no. F1, 18, X85-X87, 1975.
[77]
de Waard, C.; Milliams, D. E. Carbonic-Acid Corrosion of Steel. Corrosion 31(5), 177-181,

1975.
[78]
Townshend, P. E.; Colegate, G. T.; van Waart, T. L. Carbon dioxide corrosion at low partial
pressure in major submarine pipelines. SPE European spring meeting, paper no. 741, 1972.
[79]
Simon Thomas, M. J. J.; de Waard, C.; Smith, L.M. Controlling factors in the rate of CO2
corrosion. UK corrosion '87, 147, Brighton, 1987.
[80]
de Waard, C.; Lotz, U.; Milliams, D.E. Predictive model for CO2 corrosion engineering in wet
natural gas pipelines. Corrosion 47(12), 976-985, 1991.
[81]
de Waard, C.; Lotz, U.; Milliams, D. E. Predictive Model for CO2 Corrosion Engineering.
[82]
de Waard, C.; Lotz, U. Prediction of CO2 Corrosion of Carbon Steel. CORROSION/93 paper
No. 69, NACE, 1993.
[83]
de Waard, C.; Lotz, U. Prediction of CO2 corrosion of Carbon Steel. European Federation of
Corrosion (EFC) 13, 30-49, 1994.
[84]
de Waard, C.; Lotz, U.; Dugstad, A. Influence of Liquid Flow Velocity on CO2 Corrosion, A
Semi-Empirical Model. CORROSION/95 paper No. 128, NACE, 1995.
[85]
de Waard, C.; Smith, L.; Craig, B. D. The influence of crude oil on well tubing corrosion rates.
EUROCORR 2001 paper no. 174, Riva del Garda, Italy, 2001.
[86]
de Waard, C.; Smith, L.: Bartlett, P.; Cunningham, H. Modelling Corrosion Rates in Oil
Production Tubing. EUROCORR 2001 paper No. 254, Riva del Garda, Italy, 2001.
[87]
de Waard, C.; Smith, L. M.; Craig, B. D. THE INFLUENCE OF CRUDE OILS ON WELL
TUBING CORROSION RATES. CORROSION/03 paper no. 629, NACE, 2003.
- 72 -
84
Literature cited
[88]
Smith, Liane; de Waard, Kees. Corrosion Prediction and Materials Selection for Oil and Gas
Producing Environments. CORROSION/05 paper no. 648, NACE, 2005.
[89]
Vitse, F.; Alam, K.; GunalTun, Y.; de Torreben, D. Larrey; Duchet-Suchaux, P. SemiEmpirical model for prediction of the Top-of-the-line Corrosion Risk. CORROSSION/02
[90]
Sridhar, N.; Dunn, D.S.; Seth, M. APPLICATION OF A GENERAL REACTIVE

TRANSPORT MODEL TO PREDICT ENVIRONMENT UNDER DISBONDED
COATINGS. CORROSION/00 paper no. 366, NACE, 2000.
[91]
Nyborg, Rolf; Andersson, Peter; Nordsveen, Magnus. Implementation of CO2 Corrosion

Models in a Three-Phase Fluid Flow Model. CORROSION/00 paper no. 48, NACE, 2000.
[92]
Dugstad, Arne; Lunde, Liv; Videm, Ketil. Parametric Study of CO2 Corrosion of Carbon Steel.
CORROSION/94 paper No. 14, NACE, 1994.
[93]
Mishra, B.; Al-Hassan, S.; Olson, D. L.; Salama, M. M. Development of a predictive model for
activation-controlled corrosion of steel in solutions containing carbon dioxide. Corrosion
53(11), 852-859, 1997.
[94]
Efird, K. D.; Wright, E. J.; Boros, J. A.; Hailey, T. G. Experimental Correlation of Steel
Corrosion in Pipe Flow with Jet Impingement and Rotating Cylinder Laboratory Tests.
Corrosion 49(12), 992, 1993.
[95]
Kanwar, D.; Jepson, W. P. A model to predict sweet corrosion of multiphase flow in horizontal
pipelines. CORROSION/94 paper no. 24, NACE, 1994.
[96]
Jepson, W. P.; Bhongale, S.; Gopal, M. PREDICTIVE MODEL FOR SWEET CORROSION
IN HORIZONTAL MULTIPHASE SLUG FLOW. CORROSION/96 paper no. 19, NACE,
1996.
[97]
Jepson, W. P.; Kang, C.; Gopal, M.; Stitzel, S. Model for Sweet Corrosion in Horizontal
Multiphase Slug Flow. CORROSION/97 paper no. 11, NACE, 1997.
[98]
Jepson, W. P.; Stitzel, S.; Kang, C.; Gopal, M. MODEL FOR SWEET CORROSION IN
HORIZONTAL MULTIPHASE SLUG FLOW. CORROSION/97 paper no. 602, NACE,
1997.
[99]
http://www.standard.no/imaker.exe?id=1369. NORSOK M-506 CO2 Corrosion Rate

Calculation Model (reversion 1, june 1998). Standardiseringen i Norge, 2007.
[100]
http://www.standard.no/imaker.exe?id=10405. NORSOK M-506 CO2 Corrosion Rate

Calculation Model (reversion 2, june 2005). Standardiseringen i Norge, 2007.
[101]
Olsen, Stein. CO2 Corrosion Prediction by use of the NORSOK M-506 Model Guidelines
and Limitations. CORROSION/03 paper no. 623, NACE, 2003.
[102]
Olsen, Stein; Halvorsen, Anne Marie; Lunde, Per G. CO2 Corrosion Prediction Model - Basic
Principles. CORROSION/05 paper no. 551, NACE, 2005.
[103]
Pots, B.F.M.; Hendriksen, E. L. J. A. CO2 CORROSION UNDER SCALING CONDITIONS THE SPECIAL CASE OF TOP-OF-LINE CORROSION IN WET GAS PIPELINES.
[104]
Pots, Bert.F.M.; John, Randy C.; Rippon, Ian J.; Simon-Thomas, Maarten J. J.; Kapusta,
Sergio D.; Girgis, Magdy M.; Whitham, Tim. IMPROVEMENTS ON DE WAARDMILLIAMS CORROSION PREDICTION AND APPLICATIONS TO CORROSION
MANAGEMENT. CORROSION/02 paper no. 235, NACE, 2002.
[105]
Gunaltun, Y.M. Combining research and field data for corrosion rate prediction.
[106]
Gunaltun, Y. M.; Larray, D. CORRELATION OF CASES OF TOP OF LINE CORROSION

WITH CALCULATED WATER CONDENSATION RATES. CORROSION/00, paper no. 71,
NACE, 2000.
- 73 -
85
Literature cited
[107]
Paisley, Dominic; Barrett, Nathan; Wilson, Owen. PIPELINE FAILURE: THE ROLES
PLAYED BY CORROSION, FLOW AND METALLURGY. CORROSION/99 paper no. 18,
NACE, 1999.
[108]
Hedges, Bill; McVeigh, Lorraine. THE ROLE OF ACETATE IN CO2 CORROSION: THE
DOUBLE WHAMMY. CORROSION/99 paper no. 21, NACE, 1999.
[109]
Gartland, Per Olav. CHOOSING THE RIGHT POSITIONS FOR CORROSION

MONITORING ON OIL AND GAS PIPELINES. CORROSION/98 paper no. 83, NACE,
1998.
[110]
Gartland, Per Olav; Salomonsen, Jan Erik. A PIPELINE INTEGRITY MANAGEMENT

STRATEGY BASED ON MULTIPHASE FLUID FLOW AND CORROSION
MODELLING. CORROSION/99 paper no. 622, NACE, 1999.
[111]
Srinivasan, Sridhar; Kane, Russell D. PREDICTION OF CORROSIVITY OF CO2/H2S

PRODUCTION ENVIRONMENTS. CORROSION/96 paper no. 11, NACE, 1996.
[112]
Jangama, V. R.; Srinivasan, S. A Computer Model for Prediction of Corrosion of Carbon

Steels. Corrosion/97 paper No. 318, NACE, 1997.
[113]
Srinivasan, Sridhar; Tebbal, Saadedine. CRITICAL FACTORS IN PREDICTING CO2/H2S

CORROSION IN MULTIPHASE SYSTEMS. CORROSION/98 paper no. 38, NACE, 1998.
[114]
Srinvasan, Sridhar; Kane, Russell D. Corrosion prediction models need to include field, lab
data - Using software system with full range of key parametric effects, a model can accurately
predict corrosion rates in carbon steel. Pipe Line and Gas Industry 82(6), 39-48, 1999.
[115]
Sangita, Kiran A.; Srinivasan, Sridhar. An Analytical Model to Experimentally Emulate Flow
Effects in Multiphase CO2/H2S Systems. CORROSION/00 paper no. 58, NACE, 2000.
[116]
Srinivasan, Sridhar; Kane, Russell D. CRITICAL ISSUES IN THE APPLICATION AND

EVALUATION OF A CORROSION PREDICTION MODEL FOR OIL AND GAS
SYSTEMS. CORROSION/03 paper no. 640, NACE, 2003.
[117]
Garber, James D.; Perkins, Richard S.; Jangama, Vamshidhar R.; Alapati, Rama R.
CALCULATION OF DOWNHOLE pH AND DELTA pH IN THE PRESENCE OF CO2
AND ORGANIC ACIDS. CORROSION/96 paper no. 176., NACE, 1996.
[118]
Fang, C. S.; Garber, J. D.; Perkins, R. S.; Reinhardt, J. R. Computer model of a gas condensate
well containing CO2 - phase II. CORROSION/89 paper no. 465, NACE, 1989.
[119]
Farshad, F. F.; Garber, J. D.; Polaki, Venugopal. Comprehensive model for predicting
corrosion rates in gas wells containing CO2. SPE Production & Facilities 15(3), 183-190, 2000.
[120]
Garber, James D.; Farshad, Fred; Reinhardt, James R.; Chen, Wei; Tadepally, Vamsee Priya;
Winters, Robert. INTERNAL CORROSION RATE PREDICTION IN PIPELINES AND
FLOWLINES USING A COMPUTER MODEL. CORROSION/04 paper no. 155, NACE,
2004.
[121]
Crolet, J. L.; Bonis, M. R. A tentative method for predicting the corrosivity of wells in new
CO2 fields. In: Burke, P. A.; Asphahani, A. I.; Wright, B. S. Advances in CO2 corrosion vol. 2,
23, 1985.
[122]
CO2 fields. CORROSION/85 paper no. 27, NACE, 1985.
[123]
CO2 fields. Materials performance 25(3), 41, 1986.
[124]
Crolet, J.L.; Bonis, M. R. An optimized procedure for corrosion testing under CO2 and H2S gas
pressure. Memoires et Etudes Scientifiques de la Revue de Metallurgie 85(7-8), 375-386,
1988.
[125]
Crolet, J. L.; Bonis, M. R. An Optimized Procedure for Corrosion Testing Under CO2 and H2S
Gas Pressure. CORROSION/89 paper no. 17, NACE, 1989.
[126]
Bonis, M. R.; Crolet, J. L. Basics of the Prediction of the Risks of CO2 Corrosion in Oil and
Gas Wells. CORROSION/89 paper no. 466, NACE, 1989.
- 74 -
86
Literature cited
[127]
Crolet, Jean-Louis; Bonis, Michel R. An Optimized procedure for corrosion testing under CO2
and H2S gas pressure. Materials Performance 29(7), 81-86, 1990.
[128]
Crolet, Jean-Louis; Bonis, M. R. Prediction of the risks of CO2 corrosion in oil and gas wells.
SPE Production Engineering 6(4), 449-453, 1991.
[129]
Turgoose, S.; Cottis, R. A.; Lawson, K. Modelling of Electrode Processes and Surface
Chemistry in Carbon Dioxide Containing Solutions. ASTM Symposium on computer
modelling of corrosion, STP 1154, San Antonio, Texas, 1990.
[130]
Watson, M. K.; Postlethwaite, J. Numerical simulation of crevice corrosion of stainless steels

and nickel alloys in chloride solutions. Corrosion 46, 522, 1990.
[131]
Crolet, J. L. Mechanisms of Uniform Corrosion Under Corrosion Deposits. Journal of

Materials Science 28(10), 2589-2606, 1993.
[132]
Crolet, J.L. The electrochemistry of corrosion beneath corrosion deposits. Journal of Materials
Science 28(10), 2577-2588, 1993.
[133]
Crolet, J. L.; Thevenot, N.; Nesic, S. Role of conductive corrosion products in the
protectiveness of corrosion layers. Corrosion 54, 194203, 1998.
[134]
Nesic, S. Prediction of Transport Processes in CO2 Corrosion. European Federation of

Corrosion (EFC) 13 paper 6, 120, 1994.
[135]
Nesic, S. Prediction of Transport Processes in CO2 Corrosion. Progress in the understanding

and prevention of corrosion 1, Institute of metals, London, 539, 1993.
[136]
Pots, B. F. M. Mechanistic Models for the Prediction of CO2 Corrosion Rates Under MultiPhase Flow Conditions. CORROSION/95 paper no. 137, NACE, 1995.
[137]
Dayalan, E.; Vani, G.; Shadley, J. R.; Shirazi, S. A.; Rybicki, E. F. MODELING CO2
CORROSION OF CARBON STEELS IN PIPE FLOW. CORROSION/95 paper no. 118,
NACE, 1995.
[138]
Sundaram, M.; Raman, V.; High, M. S.; Tree, D. A.; Wagner, J. Deterministic Modeling of
Corrosion in Downhole Environments. CORROSION/96 paper no. 30, NACE, 1996.
[139]
High, M. S.; Wagner, J.; Natarajan, S. Mechanistic Modelling of Mass Transfer in the Laminar
Sublayer in Downhole Systems. CORROSION/00 paper no. 62, NACE, 2000.
[140]
Kvarekvl, Jon. A KINETIC MODEL FOR CALCULATING CONCENTRATION

PROFILES AND FLUXES OF CO2-RELATED SPECIES ACROSS THE NERNST
DIFFUSION LAYER. CORROSION/97 paper no. 5, NACE, 1997.
[141]
Zhang, R.; Gopal, M.; Jepson, W. P. DEVELOPMENT OF A MECHANISTIC MODEL FOR

PREDICTING CORROSION RATE IN MULTIPHASE OIL/WATER/GAS FLOWS.
[142]
Rajappa, S.; Zhang, R.; Gopal, M. MODELING THE DIFFUSION EFFECTS THROUGH
THE IRON CARBONATE LAYER IN THE CARBON DIOXIDE CORROSION OF
CARBON STEEL. CORROSION/98 paper no. 26, NACE, 1998.
[143]
Dayalan, E.; de Moraes, F. D.; Shirazi, S. A.; Rybicki, E. F. CO2 CORROSION PREDICTION
IN PIPE FLOW UNDER FECO3 SCALE-FORMING CONDITIONS. CORROSION/98
[144]
Anderko, Andrzej; Young, Robert D. SIMULATION OF CO2/H2S CORROSION USING

THERMODYNAMIC AND ELECTROCHEMICAL MODELS. CORROSION/99 paper no.
31, NACE, 1999.
[145]
Anderko, Andrzej; McKenzie, Patrice ; Young, Robert D. COMPUTATION OF RATES OF

GENERAL CORROSION USING ELECTROCHEMICAL AND THERMODYNAMIC
MODELS. CORROSION/00 paper no. 479, NACE, 2000.
[146]
Anderko, A. Simulation of FeCO3/FeS Scale Formation Using Thermodynamic and

Electrochemical Models. CORROSION/2000 paper no. 102, NACE, 2000.
[147]
Anderko, A.; McKenzie, P.; Young, R. D. Computation of rates of general corrosion using
electrochemical and thermodynamic models. Corrosion 57(3), 202-213, 2001.
- 75 -
87
Literature cited
[148]
Anderko, Andrzej; Sridhar, Narasi. CORROSION SIMULATION FOR THE PROCESS

INDUSTRY. CORROSION/01 paper no. 348, NACE, 2001.
[149]
Anderko, Andrzej; Young, Robert D. A MODEL FOR CALCULATING RATES OF

GENERAL CORROSION OF CARBON STEEL AND 13% Cr STAINLESS STEEL IN
CO2/H2S ENVIRONMENTS. CORROSION/01 paper no. 86, NACE, 2001.
[150]
Postlethwaite, J.; Wang, F. Modelling mass transport in aqueous CO2 corrosion in turbulent
pipe flow. Materials Science forum 289-292(ELECTROCHEMICAL METHODS IN
CORROSION RESEARCH VI - part 2), 771-788, 1998.
[151]
Wang, Fangyou. Modelling of Aqueous Carbon Dioxide Corrosion in Turbulent Pipe Flow.
PhD Theses, University of Saskatchewan, Chemical Engineering, Canada, 1999.
[152]
Wang, F.; Postlethwaite, J. MODELLING OF AQUEOUS CO2 CORROSION OF IRON IN

TURBULENT PIPE FLOW. CORROSION/01 paper no. 41, NACE, 2001.
[153]
Nesic, Srdjan; Nordsveen, Magnus; Nyborg, Rolf; Stangeland, Aage. A Mechanistic Model for
CO2 Corrosion with Protective Iron Carbonate Films. CORROSION/01 paper No. 40, NACE,
2001.
[154]
Nesic, S.; Nordsveen, M.; Nyborg, R.; Stangeland, A. A diffusion-based model for uniform
corrosion. Diffusions In Materials, DIMAT2000 194-199(PTS 1 & 2 Defect And Diffusion
Forum), 1661-1674, 2001.
[155]
Nordsveen, M.; Nesic, S.; Nyborg, R.; Stangeland, A. A Mechanistic Model for Carbon
Dioxide Corrosion of Mild Steel in the Presence of Protective Iron Carbonate Films - Part 1:
Theory and Verification. Corrosion 59(5), 443-456, 2003.
[156]
Nesic, Srdjan; Lee, Kun-Lin John; Ruzic, Vukan. A MECHANISTIC MODEL OF IRON
CARBONATE FILM GROWTH AND THE EFFECT ON CO2 CORROSION OF MILD
STEEL. CORROSION/02 paper no. 237, NACE, 2002.
[157]
Nesic, S.; Nordsveen, M.; Nyborg, R.; Stangeland, A. A Mechanistic Model for Carbon
Dioxide Corrosion of Mild Steel in the Presence of Protective Iron Carbonate Films - Part 2: A
Numerical Experiment. Corrosion 59(6), 489-497, 2003.
[158]
Nesic, S; Lee, K. L. J. A Mechanistic Model for Carbon Dioxide Corrosion of Mild Steel in
the Presence of Protective Iron Carbonate Films - Part 3: Film Growth Model. Corrosion
59(7), 616-628, 2003.
[159]
Nesic, Srdjan; Xiao, Ying; Pots, Bernardus F. M. A Quasi 2-D Localized Corrosion Model.
[160]
Nesic, Srdjan; Wang, Shihuai; Xiao, Jiyong Ying. Integrated CO2 Corrosion - Multiphase flow
model, SPE paper 87555, Aberdeen, 2004.
[161]
Nesic, Srdjan; Cai, JiYong; Wang, Shihuai; Xiao, Ying. Integrated CO2 Corrosion Multiphase Flow Model. CORROSION/04 paper no. 626, NACE, 2004.
[162]
Neic, Srdjan; Cai, Jiyong; Lee, Kun-Lin John. A MULTIPHASE FLOW AND INTERNAL
CORROSION PREDICTION MODEL FOR MILD STEEL PIPELINES. CORROSION/05
[163]
Heppner, K. L.; Evitts, R. W.; Postlethwaite, J. Determining The Crevice Corrosion Incubation
Period of Passive Metals for Systems With Moderately High Electrolyte Concentrations
Application of Pitzers Ionic Interaction Model. CORROSION/03 paper no. 691, NACE, 2003.
[164]
Song, F. M.; Kirk, D. W.; Graydon, J. W.; Cormack, D. E. PREDICTION FOR CO2
CORROSION OF ACTIVE STEEL UNDER A PRECIPITATE. CORROSION/04 paper no.
382, NACE, 2004.
[165]
Song, F. M.; Kirk, D. W.; Cormack, D. E. A Comprehensive Model for Predicting CO2
Corrosion in Oil and Gas Systems. CORROSION/05 paper no. 180, NACE, 2005.
[166]
Nesic, Srdjan; Seagraves, Sam; Zhang1, Ziru; Li, Hui; Sormaz, Dusan. FREECORP OS: A
Free Open Source Model for Internal Corrosion of Mild Steel Oil and Gas Pipelines.
CORROSION/08, paper nr. not available yet, NACE, 2008.
- 76 -
88
Literature cited
[167]
Song, F. M.; Kirk, D. W.; Graydon, J. W.; Cormack, D. E. Predicting carbon dioxide corrosion
of bare steel under an aqueous boundary layer. Corrosion 60(8), 736-748, 2004.
[168]
Levich, V. G. Physicochemical Hydrodynamics. Prentice-Hall, Englewood Cliffs, New Jersey,

1962.
[169]
Lopez, D. A.; Perez, T.; Simison, S. N. The influence of microstructure and chemical
composition of carbon and low alloy steels in CO2 corrosion. A state-of-the-art appraisal.
Materials and design 24, 561-575, 2003.
[170]
Dugstad, Arne. Fundamental Aspects of CO2 Metal Loss Corrosion Part 1: Mechanism.
[171]
Ogundele, G. I.; White, W. E. Some Observations on Corrosion of Carbon-Steel in Aqueous

Environments Containing Carbon-Dioxide. Corrosion 42(2), 71-78, 1986.
[172]
M. Stern. The Electrochemical Behavior, including hydrogen overvoltage, of iron in acid

environments. J. Electrochemical Soc. 102, 609, 1955.
[173]
Schmitt, G.; Rothmann, B. Untersuchungen zum Korrosionsmechanismus von unlegiertem

Stahl in sauerstofffreien Kohlensurelsungen. I - Kinetik der Wasserstoffabscheidung.
Werkstoffe und Korrosion 28, 816, 1977.
[174]
Schmitt, G.; Rothmann, B. Zum Korrosionsverhalten von unlegierten und niedriglegierten

Sthlen in Kohlensurelsungen. Werkstoffe und Korrosion 29, 237, 1978.
[175]
Schmitt, G.; Rothmann, B. Studies on the corrosion mechanism of unalloyed steel etc. - I
kinetics of the liberation of hydrogen. Advances in CO2 corrosion vol. 1, 72, 1984.
[176]
Wieckowski, A.; Ghali, E.; Szklarczyk, M.; Sobokowski, J. The behaviour of iron electrode in
CO2 saturated neutral electrolyteI. Electrochemical study. Electrochimica Acta 28(11),
1619, 1983.
[177]
Wieckowski, A.; Ghali, E.; Szklarczyk, M.; Sobokowski, J. The behaviour of iron electrode in
CO2 saturated neutral electrolyteII. Radiotracer study and corrosion considerations.
Electrochimica Acta 28(11), 1627, 1983.
[178]
Eriksrud, E.; Sntvedt, T. Effect of Flow on CO2 Corrosion Rates in Real and Synthetic
Formation Waters. CORROSION/83 paper no. 44, NACE, 1984.
[179]
Gray, L. G. S.; Anderson, B. G.; Danysh, M. J.; Tremaine, P. R. Mechanisms of Carbon Steel
Corrosion in Brines Containing Dissolved Carbon Dioxide. CORROSION/89 paper no. 464,
NACE, 1989.
[180]
Hurlen, T.; Gunvaldsen, S.; Tunold, R.; Blaker, F.; Lunde, P. G. Effects of carbon dioxide on
reactions at iron electrodes in aqueous salt solutions. Journal of Electroanalytical Chemistry
and Interfacial Electrochemistry 180(1-2), 511-26, 1984.
[181]
Hurlen, T.; Gunvaldsen, S. Effects of buffers on hydrogen evolution at iron electrodes.

Electrochem. Acta 29, 1163, 1984.
[182]
Gray, L. G. S.; Anderson, B. G.; Danysh, M. J.; Tremaine, P. R. Effect of pH and Temperature
on the Mechanism of Carbon Steel Corrosion by Aqueous Carbon Dioxide. CORROSION/90
[183]
Nesic, S.; Postlethwaite, J.; Olsen, S. An Electrochemical Model for Prediction of CO2
Corrosion. CORROSION/95 paper no. 131, NACE, 1995.
[184]
Nesic, S; Postlethwaite, J; Olsen, S. An Electrochemical Model for Prediction of Corrosion of

Mild Steel in Aqueous Carbon Dioxide Solutions. Corrosion 52(4), 280-294, 1996.
[185]
Bockris, J. O'M.; Drazic, D.; Despic, A. R. The electrode kinetics of the deposition and
dissolution of iron. Electrochimica Acta 4(2-4), 325-361, 1961.
[186]
Bockris, J. O'M.; Reddy, A. K. N. Modern Electrochemistry volume 1 and 2, Plenum Press,

New York, 1970.
[187]
Drazic, D.M. Iron and its Electrochemistry in an Active State. In: Conway, B.E.; Bockris, J.
OM; White, R. E. Modern Aspects of Electrochemistry, vol. 19, Plenum Press, New York, 69,
1989.
- 77 -
89
Literature cited
[188]
Schmitt, G.; Rothmann, B. Untersuchungen zum Korrosionsmechanismus von unlegiertem

Stahl in sauerstofffreien Kohlensurelsungen. II - Kinetik der Eisenauflsung. Werkstoffe
und Korrosion 29, 98, 1978.
[189]
Davies, D. H.; Burstein, G. T. The Effects of Bicarbonate on the Corrosion and Passivation of
Iron. Corrosion 36(8), 416-422, 1980.
[190]
Nesic, S.; Thevenot, N.; Crolet, J.L.; Drazic, D. Electrochemical Properties of Iron Dissolution
in the Presence of CO2 Basics Revisited. CORROSION/96 paper no. 3, NACE, 1996.
[191]
Rogers, W. F.; Rowe, J. A. Corrosion effects of hydrogen sulphide and carbon dioxide in oil
production. Proc. 4th World Petroleum Congress, paper 3, 479-99, 1955.
[192]
Davies, C. W. The extent of dissociation of salts in water. VIII. An equation for the mean ionic
activity coefficient of an electrolyte in water, and a revision of the dissociation constants of
some sulfates. Journal of the Chemical Society 2, 2093-8, 1938.
[193]
Davies, C. W. Ion Association. Butterworths, London, 1962.
[194]
Zemaitis, F. D., Jr.; Clark, D. M.; Rafal, M.; Scrivner, N. C. Handbook of Aqueous Electrolyte
Thermodynamics. Theory & Application. DIPPR, AIChE: New York, 1986.
[195]
Pitzer, K. S. Activity coefficients in electrolyte solutions. CRC Press, 2nd ed., 1991.
[196]
Thomsen, K. Aqueous Electrolytes, model parameters and process simulation. Ph.D. Thesis,
IVC-SEP, Technical University of Denmark, 1997.
[197]
Iliuta, M.; Thomsen, K.; Rasmussen, P. Extended UNIQUAC model for correlation and
prediction of vapor-liquid-solid equilibria in aqueous salt systems containing non-electrolytes.
Part A. Metanol-water-salt systems. Chemical Engineering Science 55, 2673-2686, 2000.
[198]
Thomsen, K.; Rasmussen, P.; Gani, R. Correlation and prediction of thermal properties and
phase behaviour for a class of electrolyte systems. J. Chem. Eng. Science 51, 3675-83, 1996.
[199]
Kooijmans, Johannes. Corrosion and protective coating formation in water mains. GWF, das
Gas- und Wasserfach 81, 611-15 628-33, 1938.
[200]
Riesenfeld, F. C.; Blohm, C. L. Corrosion problems in gas-purification units employing MEA

(monoethanolamine) solutions. Petroleum Refiner 29(4), 141, 1950.
[201]
Hackerman, N.; Glenn Jr., E. E. Corrosion of steel by air-free, dilute, weak acids. Journal of
the Electrochemical Society 100, 339-44, 1953.
[202]
Sowards, D. M.; Hackerman, N. Kinetics of surface reactions of metals. I. Iron. Journal of the
Electrochemical Society 102, 297-303, 1955.
[203]
Sontheimer, H.; Klle, W.; Snoeyink, V. The siderite model of the formation of corrosion
resistant scales. American Water Works Association 73, 572-579, 1981.
[204]
Eriksrud, E.; Sntvedt, T. Effect of Flow on CO2 Corrosion Rates in Real and Synthetic
Formation Waters. CORROSION/83 paper no. 44, NACE, 1984.
[205]
Whitman, G. W.; Russell, R. P.; Altieri, V. J. Effect of Hydrogen-Ion Concentration on the

Submerged Corrosion of Steel. Industrial & Engineering Chemistry 17(7), 665, 1924.
[206]
Leybold, W. The solution of iron by carbonic acid. Angewandte Chemie 37, 190-1, 1924.
[207]
Leybold, W. Protection of wet gas meters against corrosion. GWF, das Gas- und Wasserfach
70, 294-5, 1927.
[208]
Crolet, J. L.; Samaran, J. P. The use of the anti-hydrate treatment for the prevention of CO2
corrosion in long crude gas pipelines. CORROSION/93 paper No. 102, NACE, 1993.
[209]
Haarberg, T.; Jakobsen, J. E.; Oestvold, T. The effect of ferrous iron on mineral scaling during
oil recovery. Acta Chemica Scandinavica 44(9), 907-15, 1990.
[210]
Olsen, S.; Lunde, O.; Dugstad, A. Stabilizing pH in Troll pipelines solves glycol-regen
problems. Oil & Gas Journal 97, 59-62, 1999.
[211]
Dugstad, A.; Seiersten, M. pH-stabilisation, a Reliable Method for Corrosion Control of Wet
Gas Pipelines. SPE paper 87560, Aberdeen, 2004.
- 78 -
90
Thermodynamic model and experiments in the mixed solvent electrolyte system CO2-Na2CO3NaHCO3-MEG-H2O
Thermodynamic model and experiments in

the mixed solvent electrolyte system CO2Na2CO3-NaHCO3-MEG-H2O
Experimental work and modelling has been studied for the CO2-NaHCO3-Na2CO3Monoethyle glycol(MEG)-water system. Measurements were conducted at 2 to 60 C
at atmospheric conditions. 212 new solid-liquid equilibrium (SLE) experimental data
points are presented here. The extended UNIQUAC model was used for correlating
SLE, vapour-liquid equilibrium (VLE), excess enthalpy, and excess heat capacity
based on 751 experimental data points. The determined parameters are valid between
-50 and 90 C. A method based on the principles of Schreinemakers et al.1,2 is
developed for interpreting the measured data. A simple density model of NaHCO3Na2CO3-NaCl-MEG-water is given.
The aim of this study is to investigate CO2 corrosion through experimental work and
modelling. The experimental work is focused on the CO2-NaOH-MEG-H2O system
found in wet gas pipelines. The system is a sub-system of CO2-Na2O-H2O-MEG as
shown by the grey line in figure 21. Similarly Na2CO3-NaHCO3-MEG-H2O is also a
sub-system of CO2-Na2O-MEG-H2O shown by the black line in figure 21. Therefore
solubility measurements may by conducted for the Na2CO3-NaHCO3-MEG-H2O
system in order to model the CO2-NaOH-MEG-H2O system.
CO2
100
10
90
90
10
80
H2CO3
70
30
20
80
wt%
40
70
60
NaHCO3
50
60
Na2CO33NaHCO3
50
50
Na2CO3
60
40
40
Na2CO3NaHCO32H2O
Na2CO3H2O
30
30
70
80
20
20
Na2CO37H2O
H2O+
90
MEG
Na2CO310H2O
10
10
100
0
10
20
30
40
50
60
70
80
NaOH
90
0
100
Na2O
Figure 21: An overview of the components in the CO2-Na2O-H2O system. Black line is the ternary
Na2CO3-NaHCO3-H2O system and the grey dotted line is the CO2-NaOH-H2O system.
- 79 -
91
An activity coefficient model based on the Pitzer methodology was previously

developed by Kaasa3 for the aqueous NaHCO3-Na2CO3-MEG-H2O system.
Sandengen4 recently extended the model to glycol solutions of Na2CO3-NaHCO3MEG-H2O.
Only a few experimental studies of the Na2CO3-NaHCO3-MEG-H2O system have
been published. This includes Oosterhof et al.5 who published solubility of Na2CO3 in
MEG-H2O at relatively high temperature. Grtner et al.6 continued their work with
MEG concentrations larger than 50 w % MEG. The work by Grtner et al.6 was
mainly focused on solubility in the quaternary Na2CO3-NaHCO3-MEG-H2O system.
Only few of these data concerned solubility in the ternary NaHCO3-MEG-H2O
system. Sandengen4 recently published solubility measurements of NaHCO3 in MEGH2O at low temperature and high MEG concentration. Their main objective was scale
prediction and not corrosion modelling.
The Na2CO3-NaHCO3-MEG-H2O system is modelled here, in order to describe the
activity coefficient through correlation of the experimental VLE, SLE and excess
enthalpy data using the extended UNIQUAC thermodynamic model for electrolytes
and non-electrolytes. It has previously been shown by Thomsen and co-workers7,8,9
that the model may be applied for mixed solvent electrolyte systems. Here the model,
method, and theory are used unchanged compared to previous works. A brief
overview of the thermodynamic model is given in appendix D.2 and D.3. The model
allows for estimation of pH, activity coefficients, and CO2 solubility in the liquid
phase which is important for corrosion estimation.
Two modelling approaches are used. One parameter set A is estimated based on
previously published parameters. In another parameter set B, new parameters for the
Na2CO3-NaHCO3-H2O sub-system is estimated using experimental values extracted
from the IVC-SEP electrolyte database10.
7.1
Thermodynamic modelling details
The extended UNIQUAC model is an activity coefficient model. It consists of the

original UNIQUAC model plus a Debye-Hckel long-range term. The gas phase
fugacities are calculated using the Soave-Redlich-Kwong (SRK) equation of state
employing a quadratic mixing rule for the a parameter and a linear mixing rule for
the b parameter. The used critical properties and the acentric factor, Z, is given in
table 4 The gas phases in this study is close to ideal and the accuracy of the properties
are not significant. No equation of state interaction parameters were used.
Table 4: Gas-phase properties used in the SRK equation.
H2O
CO2
Tc (K)
637.096
304.2
Pc (Bar)
220.64
72.83
Z
0.344
0.225
Generally the extended UNIQUAC parameter set is consistent and it allows for
simultaneous calculation of SLE, VLE, Liquid-Liquid equilibrium (LLE), excess
enthalpy, excess heat capacity, heat of solution, heat of dilution, activity, and osmotic
coefficient with the same set of parameters. The extended UNIQUAC model reduces
to the original UNIQUAC model if no ions are present and the model may also be
used for calculation of properties in the non-electrolyte system MEG-H2O.
- 80 -
92
A Gibbs energy minimization routine supports the calculation of the properties of the
electrolytic fluid phase. This is done in order to perform speciation equilibrium
calculations and phase equilibrium calculations. The method of calculating the SLE,
VLE, excess enthalpy, heat of solution and excess heat capacity was briefly
mentioned by Iliuta et al.9 and in detail by Thomsen7.
The following equilibria are considered in this work:
Speciation equilibria:
H 2O l R H OH
(199)
CO2 aq H 2O l R H HCO3
(200)
HCO3 R H CO32
(201)
Vapour-Liquid Equilibria (VLE):

H 2O g R H 2O l
(202)
CO2 g R CO2 aq
(203)
H 2O s R H 2O l
(204)
MEG s R MEG aq
(205)
MEGH 2O s R MEG aq H 2O l
(206)
NaHCO3 s R Na HCO3
(207)
Na2CO3 s R 2 Na CO32
(208)
Na2CO3 H 2O s R 2 Na CO32 H 2O
(209)
Na2CO3 7 H 2O s R 2 Na CO32 7 H 2O
(210)
Na2CO3 10 H 2O s R 2 Na CO32 10 H 2O
(211)
Na2CO3 NaHCO3 2 H 2O s R 3 Na CO32 HCO3 2 H 2O
(212)
Na2CO3 3NaHCO3 s R 5 Na CO32 3HCO3
(213)
Solid-Liquid Equilibria (SLE):
- 81 -
93
The equilibria (202) and (203) are used in the VLE calculations. It is assumed that
MEG is not present in the gas phase since we restrict VLE data to 90 C. Glycol
contributes 1.5 % to the total pressure at 90 C and less than 1 % at most
temperatures11. The equilibria (199), (200), and (201) describe the speciation.
Reaction (201) results in an increased challenge in the experimental work and
modelling, since it shows how carbonate and bicarbonate salts may convert to the
other. The equilibria (204) through (213) shows the formation of possible solid
phases. Four hydrates may form in addition to the two double salts, trona
(Na2CO3NaHCO32H2O) and wegscheiderite (Na2CO33NaHCO3). The molal
equilibrium constants, KT , Po ,k , of reactions k at temperature T and pressure Po is
calculated from the standard state properties and the thermodynamic relations:
d ln KT , Po , k
' k H To, Po
dT
RT 2
d ' k H To, Po
dT
' k C po T , Po
(214)
' k signifies the change in enthalpy(H) or heat capacity(Cp) of reaction (202) to (213).
Superscript o indicates standard state values. NIST12 standard state properties were
used at reference conditions To=298.15 K and Po=1 bar. The correlation of the heat
capacity is given by the following equation, defined by Thomsen7:
C po ,i
ai bT
i
ci
T T4
(215)
By integrating equation (214) using (215) results in the following temperature

function of the equilibrium constants expressed in terms of the standard state
thermodynamic properties, Gibbs energy (G) and enthalpy:
R ln KT , Po ,k
1 1
T T
' k GToo , Po To ' k H Too , Po 'a ln o 1

To T
To T
2
T To 'c T T
T T4
T
4
ln
0.5'b
ln o

T
T
T
To T4
T
4
(216)
where T4 =200K and 'a , 'b and 'c are calculated using the stoichiometric
coefficients of component i in reaction k, Q ik , by 'u
u ,u
ik i
a, b, c .
i 1
Coefficient bi and ci are assumed to be zero for all solid and gaseous species. Lists of
used a, b and cs are found in the tables page 116 and 117. A constant heat capacity is
therefore used for these species. Heat capacities used in set A were either those from
NIST12 or were determined in a previous study by Thomsen et al. 7,8,13. The heat
capacities for solids were updated in parameter set B by using a modified Kopps rule14
for the values that were not listed in NIST12.
- 82 -
94
7.2
Experimental methods and apparatus
7.2.1
Chemicals
The purity of the chemicals is listed in table 5. The solid chemicals were additionally
analysed by powder X-ray diffraction and titration and showed no impurities. The
used demineralised water was produced onsite and had a conductivity of less than 0.2
S.
Table 5: Purity of the used chemicals.
Chemical
MEG
NaHCO3
CAS
107-21-1
144-55-8
Anhydrous Na2CO3
487-19-8
HCl
Water
7647-01-00
7732-18-5
7.2.2
Purity
Puriss p.a. ACS >99.5 %
Puriss >99.5 %
dried at 300 C
36 %-38 %
Demineralised, < 0.2 PS
Supplier
Riedel-deHan
Fluka
ACROS
Riedel-deHan
J. T. Baker
Technical University
of Denmark
Experimental setup
The experimental setup consisted of two units, an equilibrium unit and an analysis
unit. The equilibrium unit consisted of seven cells connected in parallel to a Julabo
F25 heating/cooling circulation bath which controlled temperature within r0.01K.
Figure 22: The seven equilibrium cells connected in parallel to the heating/cooling unit.
- 83 -
95
Figure 22 shows the equilibrium cells. The first three on the left consisted of
equipment assembled from spare parts in the laboratories. The last four on the right
was custom made for this setup. Each cell consisted of a glass lid and a heat jacket
surrounding an equilibrium chamber. A cell is sketched in figure 23. The equilibrium
chamber was a closed 50 mL poly ethylene container with screw cap and a magnetic
stirrer. A dead weight was placed on top of the containers to keep them submerged.
Water was used for transferring heat from the heat jacket to the equilibrium chamber.
Heat jacket lid

Heat outlet
Dead weight
Chamber lid
Equilibrium chamber
Heat transfer liquid
Magnetic stirrer
Heat Jacket
Heat inlet
Figure 23: Sketch of a SLE equilibrium cell.
The saturated solutions were analyzed by potentiometric titration using an autoburette,

ABU93 triburette, from Radiometer. The pH electrodes were an Orion 617500 and a
Metrohm 6.0228.000 connected to a ThermoOrion 635 unit and a Metrohm 694 pH
meter respectively.
Figure 24: The analysis unit consisting of to titration cells in the left part of the picture and the
computer on the right for data acquisition and burette control. The ABU unit is seen behind the titration
cells.
- 84 -
96
The two electrodes are seen in the titration process in figure 24, the Orion unit at the
left and the Metrohm unit at the right of the black ABU unit in the back. The
computer for data acquisition and burette control is seen in the right part of the
picture. The ABU unit allows up to three simultaneous titrations but only two were
used. Software to control the ABU unit with respect to the measured pH was
developed. The software allowed for high precision titrations to be performed close to
the equivalence points.
7.3
Experimental procedure
The chemicals were mixed from the pure chemicals listed in table 5 and a magnet was
added to the equilibrium chamber. Containers were kept closed until sampling. The
size of the vapour phase was kept low and assumed to consist mainly of air. Solutions
were stirred at approximately 500 rpm and set to equilibrate over a period of 16 h to
78 h. The time necessary for equilibration was determined to be a few hours.
The pressure was maintained at 1 atm r 0.05 atm and the temperature was measured
in the heat transferring liquid just before sampling at an accuracy of r0.05K. The CO2
pressure was not measured but assumed to correspond to the atmospheric background
of approximately 0.0004 bar.
The liquid was filtered before sampling by one of four methods. Preferably a cone
shaped filter paper was put into the equilibrium chamber and the liquid was allowed
to flow inside the filter. If this method did not produce a sample large enough for
analysis, the content of the equilibrium chamber was filtered through a conically
shaped paper filter on top of small beaker. If these two procedures failed to produce
enough liquid for analysis, saturated liquid was absorbed with a known amount of
paper towel inside a cone shaped filter, similar to method one. Finally, liquids of high
glycol concentrations were vacuum filtered using ceramic filters because they only
wet the surface of the paper filters poorly.
Samples were obtained by micropipette when possible at a volume of 100 to 1000 PL
and weighed. Samples were between 0.1 g and 1.3 g. Samples were taken in triple
from each equilibrium chamber. The saturated liquid was allowed to stand no more
than 5 minutes before being weighed. The densities of the samples were determined
as weight per volume sample at system temperature.
Samples were stored up to 3 days after sampling in closed weighing containers similar
to the equilibrium chambers and titrated by 0.1 M HCl. Approximately 15 g of water
was added to the samples prior to the automatic titration which took 30 minutes on
average and samples were stirred at 400 rpm while titrating. The titration was
configured so that a stable pH was obtained over a period of 10 seconds before the
next addition of HCl. Titrations required between 1.5 mL and 25 mL HCl. The diluted
HCl was produced from concentrated HCl and standardized by Na2CO3.
The solid phases were analysed when in doubt by powder X-ray diffraction to
determine the type of solid phase. This was done by decanting the liquid and
analysing the solid within 30 minutes after separation.
7.3.1
Previous work on analysis for sodium and carbonate
Solubility in the CO2-Na2O-H2O system has been thoroughly studied in the open
literature. Table 6 gives a summary of the used experimental methods in the literature.
- 85 -
97
The methods were used for solubility determination in aqueous systems and may
similarly be used for the solubility determination in the mixed solvent Na2CO3NaHCO3-MEG-H2O system.
The techniques fall into two categories, one for determining carbon species, another
for determining sodium content. The measurements are always converted to solubility
of solid Na2CO3, NaHCO3 or NaOH, and rarely tabulated as ionic species.
Table 6: Overview of analytical methods used in the literature for analysing the CO2-Na2O-H2O
system.
#
1
Method
Acid titration
2
3
Acid + absorbed CO2

Winklers method
4
5
6
Gravimetrical
precipitation by BaCl2
Triple Acetate method
Drying
ICP-AES
Determines
Total alkalinity (sodium)
or total carbon or
bicarbonate or carbonate.
Total carbon content

Carbonate and
Bicarbonate or
Carbonate and hydroxide
Total carbon (as BaCO3)
Total Sodium
Total Sodium (as NaCl
or Na2CO3)
Total Sodium
Used by reference
Alkali: 15,16,17,18,19,20,21
HCO3 and CO32 : 22,23,6,24
Only HCO3 : 25,26,27,28,29
Only CO32 : 30, 31,32
16,17,18,33,19,34,29,20
35,36,20,29
35,29
37, 38, 39, 29, 24, 20
NaCl: 33,40,41.
Na2CO3: 21,19
5
The solubility of a salt in the Na2CO3-MEG-H2O or in the NaHCO3-MEG-H2O

system may be determined by using one or more of the methods mentioned in table 6.
A second or third analysis only serves to check the result of the first. If the
precipitating salt is a hydrate, the water-MEG ratio is changed from its initial value.
This ratio needs to be determined separately.
The determination of solubility of salts in a Na2CO3-NaHCO3-MEG-H2O solution
requires the use of at least two different methods from table 6, one for determining
sodium content and another for determining carbon content. From these results, the
composition of the solution can be determined. Again, if a salt hydrate precipitates
from the initial solution, the water-MEG ratio is changed from its initial value.
The first method mentioned in table 6 is titration. It is a well-known method for
analysing carbonate mixtures which is also applied in this study. The solution is
titrated using a strong acid. Interpreting the equivalence points however is not a trivial
task.
In the second approach, total carbon content is determined by adding an excess of
strong acid to a sample to convert carbonates and bicarbonates to CO2. It is stripped
off with nitrogen. The vapour must be dried and the total carbon is determined by
adsorption of CO2 or by the weight loss of the solution. This method requires
relatively big sample sizes and very accurate weights. It has been used by many
authors over the years despite the difficulties16,17,18,29,20,33,19,34. This method was also
tested in this study, but consistent results were not obtained.
Winklers method42 may be applied in two different ways; either for analysing
mixtures of carbonate and hydroxide or mixtures of carbonate and bicarbonate. In
both cases the total sodium is determined by a preliminary titration. Hydroxide in a
- 86 -
98
carbonate/hydroxide solution is determined by precipitating carbonate as BaCO3 and

titration of hydroxide to the equivalence point35. The bicarbonate in a
carbonate/bicarbonate solution is determined by adding a known excess of NaOH,
precipitating the existing and formed carbonate as BaCO3 and back-titrating the
excess hydroxide. The amount of bicarbonate is calculated from the difference of the
added and back-titrated NaOH36,29,20.
In a fourth method, aqueous BaCl2 is added to precipitate carbonates as BaCO3. The
total carbon is determined gravimetrically or by titration of the precipitated
BaCO335,29. The gravimetrical determination may require big samples and careful
filtration. Barium requires special handling in the laboratory.
The total amount of sodium may be determined by either of the three following
methods or by titration. Titration is used in this work.
One method is the triple acetate method, in which sodium is precipitated as the triple
uranyl acetate salt and weighed gravimetrically as UO2 3 ZnNa CH 3COO 9 6 H 2O .
Small amounts of sodium can be determined, due to the large molecular mass of triple
uranyl acetate. A description is given by Barber and Kolthoff38 and Salit37, but the
method has limitations39 and the uranium makes the solutions hazardous.
Another method in which total sodium is determined is drying of the liquid phase. By
adding excess amounts of HCl, CO2 vaporizes and sodium is determined as
precipitated NaCl by drying33,40,41. If carbonate is the only anion it may be determined
by drying and weighing as Na2CO321,19. Drying is problematic in solutions containing
bicarbonate which decomposes at temperatures above 60 C by
2NaHCO3 o Na2CO3 CO2 g H 2O g
(217)
This method is also problematic when used for Na2CO3-NaHCO3-MEG-H2O solutions

because of the low vapour pressure of MEG.
The third method of determining total sodium is ICP-AES (Inductively Coupled
Plasma - Atomic Emission Spectrometry). This method was used by Oosterhof et al.5
7.3.2
Interpreting the titration results
Very few authors have reported the complete compositions of the solvents of the
saturated solutions. The solvent composition may change significantly if solvent
molecules precipitate together with the salts. According to equations (209) to (212)
Na2CO3, can precipitate with various amounts of hydrate water. The amount of solid
phase precipitated must therefore be known in order to determine the new solvent
composition. This is a general problem in mixed solvent electrolyte systems.
Oosterhof et al.5 solved this problem by applying a total organic carbon method
(TOC) for their poly ethylene glycol mixtures. Unfortunately they did not apply this
method for their MEG mixtures. Instead they assumed the amount of precipitated
solvent to be negligible. The solvent compositions they reported are therefore not
accurate. In this work a method is presented to solve this problem. It may also be
applied to other similar mixed solvent electrolyte systems.
- 87 -
99
10
9
8
7
6
5
4
3
2
1
0
-5
-10
-15
-20
pH
-25
d(pH)/dV
VA
0.5
d(pH)/dV
pH
VB
1.5
-30
2
mL 0.1M HCl
Figure 25: A regular titration curve of a 0.3 g sample saturated with NaHCO3 in a 63 wt% MEG
solution at 25 C.
Figure 25 shows a typical titration curve of a saturated solution from the Na2CO3NaHCO3-MEG-H2O mixture obtained in this work. Volumes of titrant at the two
equivalence points are VA and VB. The interpretation of the titration curves is not
straight forward. Three different methods for interpreting this type of measurements
are found in published scientific papers giving significantly different results.
The methods depend on the concentration of the titrant, CHCl, and the sample size in
kilo grams, msample. Equation (218) to (220) represent the three methods and give the
liquid phase composition in mol/kg solution of component i, Si. Method I provide both
the composition of carbonate and bicarbonate, whereas method II and III only give the
total carbon, TC, or total sodium, TNa:
Method I:
SCO2
3
Method II:
Method III:
CHClVA msample ,
STC
SCO2 S HCO
STNa
CHClVB msample
S HCO
3
CHCl VB 2VA msample
CHCl VB VA msample
(218)
(219)
(220)
Method I and method II both depend on VA. Our results show that VA can not be
accurately determined. The reasons are due to the evaporation of an unknown amount
of carbon dioxide, difference in titration and equilibrium temperatures, and the degree
of dilution of the sample during titration. A typical accuracy of 5 % to 10 % is
obtained, which is unacceptable, see table 7 column SCO2 .
3
In method I it is assumed that the moles of carbonate and bicarbonate are determined
from the titration volumes at the two equivalence points. This method apparently
reveals HCO3 and CO32 content and no more analysis is needed. This is a
- 88 -
100
misinterpretation. Figure 25 illustrate the problem, it shows a titration curve and its
gradient for the titration of a saturated solution of NaHCO3. It is determined that VA >
0 which indicates that the solution contains a considerable amount of carbonate. The
NaHCO3 salt was checked for NaOH and Na2CO3 impurities by X-ray diffraction and
titration and showed no impurities. The carbonate must originate from the hydrolysis
of bicarbonate to carbon dioxide and carbonate. Apparently carbon dioxide is
evaporating even though the cell was filled and closed during the experiments.
Interpreting the titration curve using method I results in an amount of Na2CO3, even
though nothing was present. If the same practice is carried out for a solution with an
infinitesimal amount of Na2CO3 added, an erroneous high amount of carbonate is
determined and equivalently an erroneous low amount of NaHCO3. The procedure has
unfortunately been used by some authors6,22,23,24 and should be avoided.
Method II gives the amount of carbon as HCO3 + CO32 . An additional analysis is
required in order to fully determine the liquid composition. Method III is similar to
method II, but method III is more accurate since it is independent of VA.
Method III is used for determining the total sodium or total alkali. The acid titration
gives the amount of alkali, OH-+ HCO3 +2 CO32 -H+. The H+ concentration is
negligible in the pH range used here. Na+ is therefore the only counter ion. Total
sodium can therefore be calculated from (220). Some authors state that the amount of
CO32 30,31,32 in a solution of carbonates or HCO3 6,25,26,27,28,29 in a solution of
bicarbonate is determined this way. In reality they determined the amount of sodium
ion.
Typically, method III is used for analysing solutions of a single salt. For analyzing
solutions containing two salts, method I is used and it is assumed that this method
gives the amounts of carbonate and bicarbonate. As described above, this is not
correct and a discontinuity in solubility is observed due to the switching between the
two analysis methods. Switching to method I at the first addition of Na2CO3 after
measuring NaHCO3 solubility creates a non-continuous solubility curve in the limit of
nNa CO o 0 . The discontinuity amounts to several wt%. Some authors are aware of
2
this and they support their analysis by the CO2 absorption method to get a fully
determined liquid phase15,16,17,18,19,20,21.
In this work we determined total sodium by titration. Afterwards, we calculated the
liquid composition using the reverse Schreinemakers method outlined below.
7.3.3
The reverse Schreinemakers method
Schreinemakers method1 was originally developed as a graphical method for

determining a solid phase from the compositions of the initial and the saturated
solutions. The method was also applied to multi solvent systems through a more
complicated graphical method2. The method requires two different initial
compositions from which the same solid precipitates. By drawing levers through the
points marking the initial composition and the composition of the saturated liquid the
solid phase can be identified at the intersection between the levers.
- 89 -
101
Total Solvents
0
100
10
20
90
30
Na2CO310H2O
40
80
70
60
50
50
60
40
70
30
o
20 C
80
20
90
100
0
Na2CO3
10
20
30
40
50
60
70
80
10
0
90 100
NaHCO3
Figure 26: Phase diagram and the lever rule used in the reverse Schreinemakers method. L(saturated
liquid), P(solid precipitate), and I(initial solution). The diagram shows the area marked by black lines
in figure 21.
Here a method is developed in which Schreinemakers method is used in reverse.

Figure 26 shows the principle. From the guessed or known precipitating solid (P) and
initial composition (I) a lever may be drawn through the two points to determine the
saturated solution (L). Only one lever is drawn but the length of the lever is
determined by the amount of mol sodium/kg total determined by titration. The method
is applicable for 2 salts and n solvents.
One solvent may precipitate from the solution together with a salt. For example
Na2CO310H2O may precipitate from the quaternary Na2CO3-NaHCO3-MEG-H2O
system. The amount of precipitated hydrate is usually unknown and the MEG/H2O
ratio is therefore different from the ratio in the initial solution. The method requires
four criteria to be fulfilled for a quaternary solution:
Initial composition must be known
Only one solid precipitates
The identity of the solid is known or guessed
One common ion exists
The weight fraction of solids in the saturated liquid phase is determined by the lever
rules formulated in equations (221) and (222) and the sodium mole balance (223).
- 90 -
102
This requires the solution of three equations of three unknowns. The equations are
generally applicable and may be used in an n solvent system.
Salt 1: fw1L 1 f w1P
w1I
(221)
Salt 2: fw2 L 1 f w2 P
w2 I
(222)
Sodium balance: nNa , L Q 1n1L Q 2 n2 L
(223)
wim is the weight fraction of compound i in point m. Point m can be one of the
following points marked in figure 26: L(Liquid solution) ,P(solid precipitate), or
I(initial solution). Compound i can be one of the following four, 1: NaHCO3, 2:
Na2CO3, 3: H2O and 4: MEG. wiP is assumed known or guessed and wiI is known from
weighing. Qi is the stoichiometric coefficient of the sodium ion in the solid phase. nNa,L
is the total amount of sodium in the saturated solution, L, and f is the lever factor. The
unknowns are f and wiL.
The titration result given in this work is calculated by equation (220) as STNa in mol
sodium/kg sample. STNa may be expressed in terms of the mass of a liquid sample,
msample , in kilo grams, and nNa,sample, the number of moles of sodium ions in the liquid
sample using equation (223):
STNa
nNa , sample
msample
Q
w1L
w
Q 2 L
M1
M2
(224)
Mi (kg/mol) is the molar mass of compound i. The mass fraction of NaHCO3 in the
liquid sample solution, w1L, is derived from equation (224) and w2L is the only
unknown:
w1L
STNa M 1
Q
Q w2 L
M1
M2
(225)
Equation (221) and (222) can be rearranged by elimination of f.
w1I w1P w1P
w1L w2 I w2 P w1I w2 P w1P w2 I
(226)
After inserting (225) in (226) and rearranging, the weight fraction of Na2CO3 in the
mixed solvent system, w2L, can be expressed as:
w2 L
STNa M 1 w2 I w2 P Q 1 w1I w2 P w1P w2 I

M
w1I w1P Q 1 w2 I w2 P
M2
(227)
The solubility of NaHCO3 in weight fraction, w1L, is calculated by equation (225) and
the solubility of Na2CO3, w2L, by equation (227).
The initial mass of water, m3I, is distributed between the saturated liquid (L) and the
solid phase (P). The water mass balance is:
m3 I
m3 P m3 L
- 91 -
103
(228)
The mass of water in the solid, m3P, is determined from the moles of water in the
solid, m3P/M3, which is related to the ratio QWNa,P of moles water to the moles of
sodium, nNa,P, in the solid:
m3 P Q WNa , P nNa , P M 3
(229)
QWNa,P is known from X-ray analysis of the solid or guessed. It is obtained from the
stoichiometric composition of the solid, examples are QWNa,P=0 for NaHCO3, QWNa,P=5
for Na2CO310H2O and QWNa,P= 23 for trona. QWNa,P may be guessed and a correct value
of QWNa,P gives a continues saturation line by repeated samples. An incorrect value
results in a discontinuous saturation line. nNa,P is determined from a sodium mol
balance:
nNa , L nNa , P
nNa , I
(230)
where nNa,I is the initial total amount of sodium which can be expressed from the
initial kilo grams of NaHCO3, m1I, and Na2CO3, m2I:
nNa , I
Q
m1I
m
Q 2 I
M1
M2
(231)
The stoichiometric coefficients of sodium in NaHCO3 and Na2CO3, Q1=1 and Q2=2,
are the same as those used in (223). The moles of sodium in the sample, nNa,L, is
calculated from the mass of solvents:
msolvents , L J
nNa , L
(232)
where msolvents,L is the total kilo grams of solvent in the saturated liquid, H2O + MEG.
J is the number of moles of sodium per kilo gram solvents defined by:
J
nNa , sample
msolvents , sample
(233)
J is determined by sampling the liquid L and determining the moles of sodium in the
sample, nNa,sample, by titration. It is divided by the mass of solvents in the sample,
msolvents,sample, which is calculated by:
wsolvents , L msample
msolvents , sample
(234)
The sample mass, msample, is known from weighing of the sample. The total solvents
mass fraction, wsolvents,L, is calculated from equations (225) and (227):
1 w1L w2 L
wsolvents , L
(235)
The total mass of solvent in the saturated liquid (L), msolvents,L, is calculated by the
following relation
msolvents , L
m3 L m4 L
- 92 -
104
(236)
For a n solvent system the above equation is given as a sum over all solvents. It is
assumed that MEG does not precipitate and stays in solution and m4L is therefore
determined by:
m4 L
m4 I
(237)
where m4I is the initial amount of glycol, which is known. For the n solvent system the
initial amount of all solvent are given similar to (237).
The only unknown in the above equations is the amount of water in the saturated
liquid, m3L. Contracting (228) to (236) to one equation by inserting (229) in (228),
(230) in (229), (231) in (230), (232) in (230), (237) in (236), and (236) in (232) gives
the final equation of water in the saturated liquid phase:
m3 L
m3 I M 3Q Na , P Q mM11I Q
m2 I
M2
1 M 3Q WNa , P J
Jm4 I
(238)
For the n solvent system m4I is the initial sum of all solvent in the system except
water. All variables are known from the initial composition or the titration data since J
is related to the solubility, STNa, by J=STNa/(1-w1L-w2L), derived from equations (224),
(233), (234), and (235). Values of the initial compositions and STNa are found in table
7.
The total mass of the saturated liquid, mL, is calculated using equation (225), (227),
(235), (237), and (238) by
mL
m3 L m4 L
wsolvents
(239)
For the n solvent system m4L is the sum of all solvents except water. The weight
fraction of solvents, H2O(3) and MEG(4), in the liquid phase (L) may now be
calculated by:
w4 L
m4 L
,
mL
w3 L
m3 L
mL
(240)
The mass of solid, mP, is determined from a total mass balance:

mI
mL mP
(241)
By using the equations presented here, the identity of the solid phase and the
composition of the liquid phase can be determined from the titration results. This
method may be called a reverse Schreinemakers method. This is a general method not
only applicable to the Na2CO3-Na2CO3-MEG-H2O system. It is valid for an n solvent
system and allows for the full determination of the identity of the solid phase and the
composition of the saturated liquid phase from a small amount of information. The
method is restricted to single salt precipitation and requires a common ion (in this
case the sodium ion) to calculate the composition. The method has not been extended
to systems in which more than one solid phase is in equilibrium with the same liquid.
- 93 -
105
7.4
Experimental results
The solubility of Na2CO3 and NaHCO3 has been determined in the mixed solvent
MEG-H2O at atmospheric pressure between 2 and 60 C. Table 7 gives the
experimental initial compositions, titration results, and densities, together with the
compositions of the saturated solutions, and the identity of the solid phases
determined by the reverse Schreinemakers method.
The initial composition is listed as amounts of pure compounds in grams. The titration
results are listed in mol/kg total as described above. Every measurement was repeated
three times and the deviation is given. The SCO2 is listed in table 7. It may be used
3
together with the intial composition and Schreinemakers method to get an idea of the
identity of the solid phase. The standard deviation is high and reflects the difficulties
in determining VB accurately. Some solution could only be measured once and the
deviation is missing.
The density measurements were difficult to reproduce and the deviations are high. In
some cases, the determination of density failed completely due to one or more
reasons. Either the micropipette took in air, it was not completely emptied, or too
small samples were used. In a few cases the density could not be measured, in other
cases only one measurement could be made, and the standard deviation could not be
determined. The sampling technique was refined over time and the measurements of
the density of solutions saturated with Na2CO3, which were measured last, are
therefore more accurate than the corresponding data for solutions saturated with
NaHCO3.
The titration data were interpreted using the reverse Schreinemakers method. The
results are given in table 7 as wt% Na2CO3, NaHCO3 and MEG. The standard
deviation is very low and shows the high reproducibility obtained by this method.
The reverse Schreinemakers method developed in this work can not be used for
interpreting titration results for solutions with two solid phases in equilibrium with the
same liquid. The titration results for such solutions were therefore not converted to
wt% in table 7.
The identities of selected solid phases were determined by X-ray diffraction to verify
the results of the reverse Schreinemakers method.
- 94 -
106
107
2.45 15.60
2.55 42.22
2.35 63.21
10.20
0.00
10.55
0.00
10.35
3.08
10.40
3.15
10.05 14.96
10.20 15.32
10.15 28.05
10.20 28.12
10.35 42.22
10.20 53.05
10.10 53.13
10.25 63.21
10.20 77.89
10.45 78.10
10.30 100.00
20.15
0.00
20.00
0.00
20.10
2.70
20.00
2.93
20.10
3.21
20.00 15.09
20.10 15.12
20.00 15.43
20.10 27.89
20.00 28.42
19.60 42.22
20.10 52.82
20.00 52.99
20.10 60.02
19.60 63.21
20.10 67.44
26.1109
17.2869
9.9027
10.0318
10.0009
9.7773
9.6940
8.4850
8.5744
7.2142
7.2232
20.3200
4.7262
4.6877
14.0064
2.2212
2.1872
0.0000
10.0660
9.9966
9.6800
9.7341
9.6970
8.5923
8.5210
8.4692
7.2172
7.1993
20.3200
4.7172
4.7049
4.0003
14.0064
3.2574
4.8273
12.6312
17.0171
0.0000
0.0000
0.3109
0.3156
1.4924
1.5507
2.8118
2.8261
14.8473
5.3404
5.3131
24.0690
7.8272
7.7994
9.9890
0.0000
0.0000
0.2684
0.2942
0.3215
1.5275
1.5173
1.5448
2.7908
2.8586
14.8473
5.2816
5.3028
6.0058
24.0690
6.7474
Initialc
4.6238
3.3322
2.4722
1.0104
1.9083
1.0030
1.9070
1.6265
0.9111
1.3677
0.7926
3.5293
0.6002
1.1178
2.5329
0.5056
1.1120
1.3650
1.8708
1.9238
1.8776
1.1232
1.8969
2.0342
1.5716
1.1100
1.3003
1.3270
3.5293
1.1088
1.0919
0.9790
2.5329
1.0456
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.733
0.475
0.375
0.962
1.009
0.910
0.963
0.792
0.747
0.634
0.607
0.477
0.441
0.447
0.382
0.308
0.340
0.647
1.088
1.117
1.049
1.036
1.072
0.919
0.867
0.847
0.697
0.701
0.568
0.478
0.480
0.444
0.435
0.411
0.38
1.24
0.31
1.33
0.49
1.02
1.12
0.07
0.88
0.55
5.33
0.97
0.32
0.20
1.09
1.58
0.86
0.08
2.15
0.12
0.04
0.21
0.17
0.31
1.38
0.23
0.02
1.24
0.27
0.43
4.81
0.94
0.115
0.085
0.095
0.054
0.104
0.071
0.118
0.112
0.074
0.118
0.063
0.100
0.037
0.079
0.088
0.024
0.053
0.083
0.163
0.217
0.153
0.190
0.206
0.170
0.129
0.169
0.089
0.126
0.141
0.045
0.077
0.057
0.093
0.060
11.57
24.69
16.08
12.62
11.22
11.16
6.02
3.42
11.41
2.57
6.63
37.49
2.40
8.98
15.75
2.37
0.14
0.30
1.59
0.79
0.65
2.33
9.53
10.52
26.93
0.63
2.32
2.32
35.50
9.29
Equilibriumd
1.057
1.049
1.044
1.048
1.045
1.047
1.055
1.066
1.069
1.091
1.072
1.117
1.065
1.072
1.042
1.048
1.065
1.083
1.064
1.068
1.079
1.080
1.094
1.097
1.101
1.098
0.09
0.24
0.27
2.24
0.21
1.17
0.05
0.09
1.15
0.34
0.28
0.42
0.61
0.39
0.04
0.67
0.41
0.21
0.32
0.59
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
6.16
0.024 26.1109
3.992
- 17.2869
3.15
0.039 9.9027
8.08
0.025 10.0318
8.47
0.11 10.0009
7.64
0.037 9.7773
8.09
0.082 9.6940
6.65
0.075 8.4850
6.278 0.0046 8.5744
5.33
0.047 7.2142
5.10
0.028 7.2232
4.01
0.21 20.3200
3.702
- 4.7262
3.75
0.036 4.6877
3.21
0.010 14.0064
2.589 0.0052 2.2212
2.86
0.031 2.1872
5.44
0.086 0.0000
9.14
0.079 10.0660
9.383 0.0076 9.9966
8.81
0.19 9.6850
8.71
0.010 9.7341
9.003 0.0037 9.6970
7.72
0.017 8.5923
7.28
0.012 8.5210
7.12
0.022 8.4692
5.86
0.081 7.2172
5.89
0.013 7.1993
4.7747 0.00093 20.3200
4.02
0.050 4.7172
4.03
0.011 4.7049
3.73
0.016 4.0003
3.65
0.18 14.0064
3.46
0.032 3.2574
Reverse Schreinemakerse
14.642
40.534
61.223
0.000
0.000
2.846
2.898
13.963
14.354
26.551
26.688
40.526
51.087
51.133
61.183
75.879
75.868
94.564
0.000
0.000
2.459
2.678
2.920
13.928
14.014
14.328
26.253
26.748
40.203
50.701
50.853
57.782
60.905
65.111
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
Table 7: Experimental initial composition, equilibrium composition, results of the reverse Schreinemakers method and solid phase information from the Na2CO3-Na2CO3-MEG-H2O system at 1
bar and 2 C to 60 C.
wNaHCO3 VNaHCO
mH2O
Ta
mMEG
mNaHCO3 mNa2CO3
STNa
dev SCO 2wNa2CO3
wMEG
P1f P2f
dev Usat. iquid dev
mH2O
VNa2CO3
w MEGb
3
3
C
%
g
g
g
g
mol/kg
%
mol/kg
%
g/cm3
%
%
%
%
%
g
%
108
w MEGb
75.01
78.07
82.48
90.02
97.42
100.00
100.00
42.22
63.21
0.00
2.87
2.93
15.01
15.43
28.11
52.81
77.33
100.00
0.00
3.36
14.96
15.60
28.01
42.22
52.88
63.21
78.17
100.00
0.00
2.97
3.04
15.13
15.34
15.60
27.85
42.22
53.03
63.21
Ta
20.10
20.00
20.10
20.10
20.10
20.10
20.00
24.40
24.60
29.80
29.90
29.90
30.00
29.90
30.00
30.00
30.00
30.00
34.90
34.70
34.90
32.70
34.90
32.70
34.90
33.00
34.90
34.90
39.90
39.40
39.80
39.95
39.60
40.50
39.90
40.00
39.90
40.60
2.4977
2.2024
1.7574
0.9995
0.2592
0.0000
0.0000
20.3200
14.0064
10.0320
9.6877
9.7167
8.5444
8.4689
7.2252
4.7268
2.2914
0.0000
10.0139
9.7142
8.5202
26.1109
7.1995
17.2869
4.7139
9.9027
2.1880
0.0000
10.0165
9.7180
9.7116
8.4968
8.3722
26.1109
7.2561
17.2869
4.7096
9.9027
mH2O
7.4957
7.8396
8.2719
9.0123
9.7689
10.0005
9.9970
14.8473
24.0690
0.0000
0.2858
0.2936
1.5094
1.5447
2.8258
5.2907
7.8158
10.0247
0.0000
0.3375
1.4989
4.8273
2.8007
12.6312
5.2903
17.0171
7.8359
10.0162
0.0000
0.2975
0.3048
1.5145
1.5168
4.8273
2.8009
12.6312
5.3171
17.0171
mMEG
0.9924
1.1731
0.9970
1.0003
0.9833
1.2820
1.3853
3.5293
2.5329
1.3015
1.3034
1.5166
1.1944
1.4930
1.0006
0.6992
0.6260
0.9025
1.8971
1.9009
1.6346
4.6238
1.3540
3.3322
1.1031
2.4722
1.1269
1.4230
1.4127
1.3060
1.9159
1.1928
1.9049
4.6238
1.0128
3.3322
0.7055
2.4722
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
mNaHCO3 mNa2CO3
0.388
0.383
0.373
0.401
0.572
0.744
0.772
0.594
0.435
1.262
1.181
1.196
1.030
0.986
0.796
0.517
0.367
0.810
1.346
1.261
1.100
1.077
0.863
0.658
0.570
0.509
0.420
0.871
1.414
1.343
1.378
1.147
1.139
1.107
0.890
0.677
0.579
0.516
STNa
1.34
1.43
0.30
0.28
0.16
0.10
0.08
0.34
1.03
1.84
0.34
0.04
0.25
0.17
0.49
0.98
0.31
0.09
0.08
0.11
0.12
0.37
0.09
0.49
0.18
0.35
0.06
1.12
0.23
0.10
0.04
0.11
1.60
0.23
1.48
3.81
0.19
0.31
dev
SCO320.055
0.057
0.047
0.046
0.051
0.045
0.086
0.095
0.076
0.105
0.116
0.149
0.139
0.135
0.135
0.065
0.031
0.059
0.262
0.261
0.238
0.314
0.174
0.212
0.089
0.148
0.081
0.140
0.120
0.152
0.238
0.162
0.236
0.083
0.112
0.113
0.070
0.078
8.48
2.70
0.06
5.48
6.21
3.58
11.11
8.22
2.49
13.05
9.97
2.44
4.99
6.93
6.24
9.79
5.77
2.41
2.48
3.52
18.79
23.11
17.48
7.09
3.46
19.86
43.44
12.37
8.71
3.03
3.70
4.47
1.77
31.01
2.58
1.34
dev
1.110
1.117
1.117
1.128
1.146
1.160
1.151
1.084
1.071
1.070
1.074
1.085
1.080
1.082
1.110
1.116
1.161
1.110
1.108
1.100
1.114
1.109
1.123
1.149
1.093
1.084
1.115
1.122
1.071
1.103
1.090
1.111
-
Usat. iquid
0.64
0.98
0.53
0.14
0.02
0.04
0.03
0.04
0.16
2.10
0.57
1.42
0.21
0.11
0.34
0.74
0.31
1.21
0.17
0.77
2.61
0.98
0.24
1.47
1.00
-
dev
wNa2CO3
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
VNa2CO3
3.26
3.22
3.137
3.372
4.804
6.249
6.482
4.99
3.66
10.60
9.92
10.043
8.65
8.28
6.69
4.34
3.087
6.806
11.311
10.60
9.24
9.05
7.248
5.53
4.787
4.27
3.532
7.32
11.88
11.28
11.579
9.63
9.57
9.30
7.47
5.69
4.866
4.33
wNaHCO3
0.044
0.046
0.0094
0.0094
0.0077
0.0065
0.0051
0.017
0.038
0.20
0.034
0.0042
0.022
0.014
0.033
0.043
0.0096
0.0060
0.0086
0.012
0.011
0.034
0.0063
0.027
0.0085
0.015
0.0020
0.082
0.028
0.011
0.0051
0.011
0.15
0.021
0.11
0.22
0.0093
0.014
VNaHCO3
2.4977
2.2024
1.7574
0.9995
0.2592
0.0000
0.0000
20.3200
14.0064
10.0320
9.6877
9.7167
8.5444
8.4689
7.2252
4.7268
2.2914
0.0000
10.0139
9.7142
8.5202
26.1109
7.1995
17.2869
4.7139
9.9027
2.1880
0.0000
10.0165
9.7180
9.7116
8.4968
8.3722
26.1109
7.2561
17.2869
4.7096
9.9027
mH2O
72.563
75.554
79.890
86.981
92.735
93.751
93.518
40.113
60.903
0.000
2.581
2.639
13.715
14.148
26.235
50.523
74.942
93.194
0.000
3.002
13.578
14.191
25.977
39.884
50.349
60.513
75.411
92.680
0.000
2.635
2.691
13.670
13.871
14.152
25.769
39.818
50.449
60.476
wMEG
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
P1f P2f
109
w MEGb
70.01
75.09
78.10
82.44
89.81
97.51
100.00
0.00
2.89
3.04
14.74
15.34
27.86
53.02
78.05
100.00
3.16
0.00
14.93
27.93
53.05
77.91
100.00
0.00
3.04
3.94
14.93
27.93
54.34
77.95
100.00
0.00
3.10
9.10
14.97
36.50
44.87
78.11
Ta
39.90
39.60
39.80
39.90
39.90
39.90
39.85
49.90
48.65
48.70
49.85
49.00
49.80
49.80
49.75
49.70
56.85
59.65
59.75
59.60
59.60
59.50
59.40
10.20
10.15
10.40
10.20
10.20
10.20
10.20
10.20
15.20
15.20
15.20
15.10
15.20
15.20
15.20
3.2669
2.4916
2.1983
1.7665
1.0243
0.2487
0.0000
10.0034
9.7089
9.6849
8.5482
8.4856
7.1944
4.6912
2.1902
0.0000
9.7463
10.0668
8.5408
7.2290
4.7048
2.2126
0.0000
10.0134
9.6966
9.7279
8.5053
7.2202
4.7315
2.2120
0.0000
9.9952
9.7155
9.1058
8.4855
6.3496
5.5182
2.1900
mH2O
7.6250
7.5101
7.8386
8.2936
9.0243
9.7557
10.0544
0.0000
0.2885
0.3040
1.4778
1.5374
2.7780
5.2946
7.7858
10.0219
0.3179
0.0000
1.4984
2.8021
5.3158
7.8048
10.0630
0.0000
0.3043
0.3990
1.4925
2.7978
5.6305
7.8197
10.0181
0.0000
0.3112
0.9111
1.4940
3.6500
4.4917
7.8164
mMEG
0.5313
0.5415
0.6093
0.5126
0.5135
0.6277
0.9057
1.5841
1.6013
1.9179
1.2886
1.9052
1.2150
0.8969
0.7096
0.8838
0.5975
1.7070
1.4310
1.3035
0.9456
0.7067
0.9045
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
2.5072
4.2919
2.7720
2.2930
2.1673
1.7704
1.7773
1.7743
2.8448
2.8648
2.0813
2.1267
2.0555
2.5300
0.8600
mNaHCO3 mNa2CO3
0.456
0.413
0.410
0.422
0.426
0.613
0.856
1.603
1.479
1.580
1.317
1.323
1.036
0.660
0.458
0.832
1.701
1.808
1.524
1.186
0.744
0.510
0.874
2.236
2.298
2.083
1.772
1.643
2.223
1.131
0.974
2.807
2.641
2.438
2.241
2.291
2.820
1.118
STNa
0.38
0.22
0.11
0.73
0.22
0.44
0.62
0.14
0.90
0.30
0.10
6.11
0.21
0.18
0.34
0.57
0.26
0.14
0.30
0.09
1.27
0.21
1.51
0.95
0.00
2.08
3.63
0.33
0.22
0.17
0.31
0.69
0.14
0.44
1.52
0.36
0.15
dev
SCO320.041
0.032
0.037
0.043
0.032
0.061
0.036
0.207
0.187
0.294
0.197
0.304
0.151
0.072
0.042
0.055
0.216
0.214
0.239
0.181
0.065
0.050
0.075
1.109
0.942
1.030
0.876
0.796
1.104
0.557
0.427
1.399
1.311
1.213
1.109
1.131
1.398
0.549
19.90
16.30
7.39
0.74
7.14
45.41
12.61
1.55
10.34
9.41
5.86
21.19
1.30
1.40
4.37
5.57
3.12
8.23
0.64
5.77
85.38
4.27
3.05
0.88
4.41
2.86
3.64
0.61
1.61
0.60
0.47
0.41
0.47
1.17
1.53
0.88
0.15
dev
1.117
1.125
1.129
1.136
1.149
1.164
1.140
1.139
1.126
1.129
1.114
1.113
1.131
1.166
1.105
1.152
1.121
1.122
1.125
1.132
1.155
1.123
1.105
1.142
1.144
1.191
1.142
1.118
1.153
1.143
1.128
1.131
1.164
1.210
1.145
Usat. iquid
wNa2CO3
VNa2CO3
0
0.51
0
0.25
0
0.27
0
0.19
0
0.13
0
0.23
0
0
2.39
0
0.60
0
1.98
0
0
0.38
0
0.93
0
0.29
0
0.68
0
0.46
0
0
0.92
0
8.01
0
2.78
0
0.24
0
0.25
0
0.32
11.85
0.11
- 12.17933 0.000020
0.84
11.04
0.23
1.14
9.39
0.34
2.10
8.708
0.36
11.78
0.038
1.16
5.99
0.013
2.22
5.160
0.0085
0.09
14.87
0.046
0.46
13.99
0.097
0.59
12.92
0.019
0.31
11.88
0.052
0.62
12.14
0.18
1.47
14.95
0.054
0.55
5.924
0.0090
dev
3.83
3.472
3.447
3.54
3.578
5.15
7.19
13.47
12.42
13.27
11.06
11.12
8.70
5.544
3.85
6.99
14.29
15.19
12.81
9.959
6.25
4.283
7.34
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
wNaHCO3
mH2O
0.014 3.2669
0.0076 2.4916
0.0037 2.1983
0.026 1.7665
0.0077 1.0243
0.023 0.2487
0.045 0.0000
0.019 10.0034
0.11 9.7089
0.039 9.6849
0.011 8.5482
0.68 8.4856
0.019 7.1944
0.0097 4.6912
0.013 2.1902
0.040 0.0000
0.037 9.7463
0.021 10.0668
0.038 8.5408
0.0094 7.2290
0.080 4.7048
0.0090 2.2126
0.11 0.0000
7.4558
3.2361
6.4640
5.9132
4.7621
4.6642
2.0165
0.0000
7.3395
6.8177
7.7536
6.7621
4.8530
5.3831
2.1505
VNaHCO3
67.323
72.481
75.406
79.521
86.593
92.493
92.806
0.000
2.527
2.639
13.110
13.633
25.433
50.082
75.041
93.015
2.707
0.000
13.014
25.152
49.731
74.575
92.655
0.000
7.549
5.172
18.262
33.786
48.251
74.733
94.840
0.000
3.754
9.156
15.947
37.716
38.688
73.778
wMEG
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
D
D
D
D
D
M
M
C
D
D
D
D
D
M
M
P1f P2f
110
w MEGb
100.00
0.00
3.07
8.98
15.26
21.02
28.13
39.99
15.43
0.00
3.04
3.06
15.11
15.34
27.94
52.90
78.45
100.00
0.00
2.74
15.04
15.34
27.76
52.72
77.30
100.00
0.00
2.97
15.16
27.92
53.13
77.53
100.00
3.04
3.04
3.04
3.04
3.04
Ta
15.20
20.10
20.10
20.10
20.10
20.10
20.10
20.10
29.90
39.80
39.80
39.50
40.00
39.50
40.00
40.00
39.90
39.90
49.60
48.85
49.80
49.60
49.70
49.70
49.70
49.70
59.55
58.20
59.75
59.65
59.55
59.55
59.45
10.15
10.50
10.30
10.30
10.25
0.0000
10.0175
9.7043
9.1021
8.5166
7.9181
7.2056
6.0083
8.4562
9.9823
9.6958
9.7113
8.4945
8.4566
7.2336
4.7177
2.2158
0.0000
9.9849
9.6756
8.5359
8.4741
7.2678
4.7511
2.2837
0.0000
10.0055
9.6703
8.5310
7.1741
4.6970
2.2641
0.0000
9.7296
9.7031
9.6998
9.7057
9.6945
mH2O
10.0145
0.0000
0.3075
0.8975
1.5333
2.1070
2.8202
4.0036
1.5425
0.0000
0.3043
0.3068
1.5115
1.5322
2.8046
5.2990
8.0670
10.0301
0.0000
0.2729
1.5107
1.5353
2.7930
5.2973
7.7780
10.0678
0.0000
0.2955
1.5248
2.7784
5.3252
7.8099
9.7490
0.3054
0.3046
0.3044
0.3046
0.3043
mMEG
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
1.0819
1.7053
1.0567
1.2318
0.1307
0.8703
2.5408
2.5149
2.5114
2.5043
2.5518
2.5101
2.5520
4.5921
5.7611
4.6687
5.7592
4.9300
4.6623
4.2938
1.7583
1.1124
1.1395
5.5709
5.8053
4.1711
4.6669
3.3213
1.5709
0.7996
0.8105
5.8730
5.7978
4.2020
3.2790
1.5942
0.7861
0.8038
0.1704
0.4846
0.7047
1.1119
2.3848
mNaHCO3 mNa2CO3
0.869
3.497
3.357
3.109
2.919
2.828
2.876
3.144
4.917
6.537
5.813
6.087
5.018
4.826
3.886
1.950
0.935
1.128
6.273
5.935
4.822
4.701
3.687
1.819
0.972
0.705
6.054
5.817
4.675
3.527
1.756
0.802
0.454
1.152
1.579
1.889
2.375
2.299
STNa
0.10
0.26
0.08
0.09
0.17
0.14
0.10
0.16
0.20
0.40
0.10
0.18
0.15
0.32
0.26
0.19
2.65
1.92
0.59
0.14
0.44
1.72
0.73
0.79
1.78
2.68
1.61
0.49
0.48
0.70
0.51
0.38
4.60
2.61
0.40
0.43
1.01
0.98
dev
0.410
1.735
1.663
1.541
1.441
1.397
1.417
1.558
2.446
3.283
2.902
3.031
2.514
2.391
1.931
0.966
0.453
0.549
3.127
2.963
2.409
2.289
1.832
0.862
0.479
0.317
3.038
2.907
2.320
1.747
0.853
0.389
0.215
0.303
0.523
0.716
0.946
0.996
SCO320.85
0.32
0.08
0.06
0.12
0.07
1.06
0.11
0.73
0.61
0.59
0.16
0.47
0.54
0.46
0.44
2.36
2.66
1.23
0.46
0.97
3.38
1.13
2.98
2.32
3.99
2.97
1.35
1.05
0.36
3.63
0.19
4.76
1.87
8.04
6.42
2.76
1.68
dev
1.159
1.176
1.170
1.163
1.163
1.170
1.189
1.166
1.283
1.330
1.367
1.307
1.232
1.246
1.225
1.160
1.112
1.143
1.303
1.258
1.321
1.306
1.238
1.178
1.147
1.151
1.367
1.351
1.340
1.267
1.149
1.166
1.122
1.083
1.068
1.104
1.147
-
Usat. iquid
0.90
0.57
0.43
0.59
0.76
0.36
0.26
0.22
1.15
1.34
1.93
1.83
0.92
2.28
0.81
1.70
1.48
3.67
2.96
0.87
1.47
1.57
2.30
14.19
0.46
1.66
3.87
1.98
0.81
0.40
2.22
0.05
0.49
1.20
3.19
-
dev
4.603
18.53
17.79
16.47
15.47
14.99
15.24
16.66
26.06
34.64
30.81
32.26
26.59
25.57
20.60
10.34
4.96
5.98
33.24
31.45
25.55
24.92
19.54
9.64
5.15
3.74
32.08
30.83
24.77
18.69
9.30
4.25
2.41
1.549
4.322
6.181
9.51
11.24
wNa2CO3
0.0045
0.049
0.014
0.014
0.027
0.022
0.014
0.026
0.052
0.14
0.030
0.057
0.039
0.082
0.053
0.020
0.13
0.12
0.20
0.045
0.11
0.43
0.14
0.076
0.092
0.10
0.52
0.15
0.12
0.13
0.048
0.016
0.11
0.0043
0.0027
0.0050
0.024
0.12
VNa2CO3
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
7.22
6.41
6.07
4.88
1.499
wNaHCO3
0.26
0.057
0.076
0.24
0.0072
VNaHCO3
0.0000
9.2920
8.7673
7.7236
6.8753
6.0142
6.1353
5.9115
8.2630
9.8945
9.6560
9.5285
8.2582
8.2346
6.9331
4.6131
2.1170
0.0000
9.8738
9.4468
8.4055
8.2318
7.1121
4.6647
2.2403
0.0000
9.7938
9.4209
8.3707
6.9988
4.5991
2.2060
0.0000
9.7296
9.7031
9.6998
9.7057
7.3067
mH2O
95.397
0.000
2.786
8.695
15.414
22.056
26.691
33.650
11.632
0.000
2.114
2.113
11.357
11.675
22.870
47.934
75.287
94.021
0.000
1.925
11.342
11.803
22.689
48.049
73.640
96.264
0.000
2.104
11.592
23.106
48.666
74.662
97.594
2.776
2.716
2.670
2.605
3.489
wMEG
C
D
D
D
D
D
H
M
M
M
M
M
M
M
M
M
M
C
M
M
M
M
M
M
M
C
M
M
M
M
M
M
C
B
B
B
B
D
P1f P2f
111
w MEGb
3.04
3.04
15.34
15.34
15.34
15.34
15.34
15.34
3.19
3.05
15.46
14.98
2.93
2.93
2.93
2.93
2.93
3.01
2.99
2.93
2.93
2.93
15.43
15.43
15.43
15.43
15.43
15.43
15.21
14.95
15.43
15.43
2.93
2.93
2.93
2.93
15.43
15.43
Ta
10.10
10.30
10.20
10.15
10.30
10.15
10.20
10.15
15.10
15.20
15.20
15.20
20.10
20.00
20.00
20.00
20.00
20.10
20.10
20.00
20.00
20.10
20.10
20.00
20.00
20.00
20.00
20.00
20.10
20.10
20.00
20.10
29.80
30.00
30.00
29.90
29.80
30.00
9.7080
9.6841
8.4994
8.4642
8.4954
8.4546
8.4820
8.4844
9.7140
9.7010
8.5047
8.4929
9.7292
9.6991
9.7237
9.7171
9.7268
9.7138
9.6938
9.7010
9.7060
9.7172
8.4512
8.4543
8.4606
8.4529
8.4598
8.4677
8.5024
8.5242
8.4715
8.4610
9.7021
9.7264
9.7254
9.7080
8.4709
8.4575
mH2O
0.3047
0.3040
1.5399
1.5335
1.5392
1.5318
1.5368
1.5372
0.3205
0.3048
1.5553
1.4963
0.2940
0.2931
0.2939
0.2937
0.2940
0.3013
0.2986
0.2932
0.2933
0.2937
1.5415
1.5421
1.5432
1.5418
1.5431
1.5445
1.5257
1.4983
1.5452
1.5433
0.2932
0.2939
0.2939
0.2934
1.5451
1.5427
mMEG
0.0614
1.4853
1.0136
1.1697
1.0409
1.2177
0.0619
1.4805
0.3268
0.1588
0.1837
0.0961
1.1381
1.1647
1.2183
1.2884
1.3226
0.3269
0.1626
1.4289
1.4271
1.3712
1.1279
1.1652
1.2227
1.2485
1.4300
1.4939
0.1515
0.3177
1.2890
1.3521
1.1946
1.5389
1.6170
1.3653
1.1963
1.4907
2.3601
3.3331
0.2222
0.4706
0.5814
0.9746
2.3604
3.3366
2.1216
2.0646
2.0719
2.0797
0.1453
0.5314
0.9686
1.5513
1.9235
2.5787
2.5346
2.8880
2.8647
2.3268
0.1213
0.4610
0.9750
1.2541
2.8600
3.3353
2.1462
2.1697
1.5319
2.1621
0.7175
3.8482
4.5138
2.2428
0.7237
3.3510
mNaHCO3 mNa2CO3
2.042
2.491
1.038
1.374
1.517
2.064
1.741
1.949
2.879
2.756
2.402
2.341
1.186
1.661
2.207
2.912
3.358
3.627
3.488
3.470
3.451
3.483
0.987
1.413
2.086
2.443
2.741
3.067
3.098
3.158
2.550
2.510
2.033
3.956
4.408
3.467
1.850
3.148
STNa
0.24
1.73
0.16
0.51
0.11
0.38
1.84
0.29
0.21
0.03
0.37
0.18
0.12
0.32
0.11
0.25
0.18
0.14
0.14
0.19
0.23
0.18
0.22
0.18
0.23
0.23
0.02
0.04
0.10
0.11
0.10
0.17
0.13
0.28
0.35
0.12
0.26
0.04
dev
SCO320.960
0.928
0.305
0.469
0.558
0.854
0.794
0.747
1.267
1.288
1.088
1.106
0.263
0.521
0.822
1.219
1.467
1.641
1.660
1.537
1.532
1.532
0.232
0.464
0.833
1.026
1.242
1.443
1.467
1.477
1.083
1.065
0.657
1.832
2.115
1.490
0.648
1.478
0.54
1.71
0.22
2.40
0.83
0.21
0.95
0.46
0.46
0.91
1.45
0.09
0.93
0.13
0.49
0.30
0.21
0.35
0.54
6.52
0.98
0.75
0.65
0.27
0.46
0.68
0.55
0.35
0.24
0.41
0.60
0.32
0.44
0.30
dev
1.093
1.057
1.084
1.064
1.118
1.156
1.145
1.154
1.134
1.066
1.088
1.120
1.166
1.180
1.186
1.169
1.197
1.193
1.195
1.072
1.096
1.128
1.156
1.169
1.172
1.166
1.173
1.152
1.149
1.107
1.203
1.245
1.192
1.110
1.203
Usat. iquid
2.77
2.14
2.19
2.87
0.97
0.15
0.51
0.76
0.35
0.40
0.43
0.53
0.81
0.19
0.16
0.56
0.40
0.29
0.31
0.15
0.09
0.80
0.50
0.65
0.53
0.31
0.10
0.66
0.90
1.50
0.34
1.59
0.34
0.37
dev
10.36
2.0468
4.282
5.2330
8.555
8.73
13.30
13.657
11.54
11.78
1.3164
4.723
8.349
12.87
15.52
17.38
17.56
1.1212
4.177
8.525
10.763
12.807
15.079
15.57
6.214
19.19
22.22
6.372
15.400
wNa2CO3
0.027
0.00032
0.0029
0.00081
0.0068
0.17
0.034
0.0044
0.049
0.023
0.00018
0.0025
0.0020
0.010
0.011
0.029
0.026
0.00023
0.0010
0.0041
0.0063
0.0019
0.0034
0.017
0.0016
0.033
0.043
0.0030
0.0035
VNa2CO3
VNaHCO3
0.7295 0.00073
5.48
0.015
4.76
0.064
4.45
0.015
3.78
0.076
0.780 0.0047
3.101 0.0045
1.5063 0.00028
1.881 0.0036
1.0007 0.00089
7.88
0.013
6.47
0.049
5.30
0.023
4.05
0.078
3.60
0.069
2.911 0.0043
1.463 0.0020
6.51
0.019
5.24
0.023
4.01
0.046
3.46
0.057
2.727 0.0019
1.860 0.0038
1.344 0.0011
7.23
0.024
2.81
0.041
1.80
0.062
5.44
0.045
2.037 0.0039
wNaHCO3
7.1789
8.4994
8.4642
8.4954
8.4546
5.6488
8.4903
8.6391
6.8987
6.8802
9.7292
9.6991
9.7237
9.7171
9.7268
8.6482
8.7037
8.4512
8.4543
8.4606
8.4529
7.9783
7.9178
7.8365
9.7021
9.2133
9.1278
8.4709
7.9183
mH2O
3.620
14.185
13.952
13.854
13.447
19.352
3.041
2.891
15.928
15.581
2.664
2.605
2.533
2.437
2.373
2.683
2.686
14.249
13.973
13.493
13.232
13.689
13.558
13.540
2.539
2.412
2.370
13.604
13.463
wMEG
D
B
B
B
B
B
D
B
D
D
D
D
B
B
B
B
B
D
D
B
T
B
B
B
B
B
T
T
D
D
B
B
B
T
T
B
B
T
T
T
T
T
T
B
T
P1f P2f
112
8.4640
8.4631
8.4584
10.0081
9.7036
9.6833
9.7284
9.6826
9.7064
9.7241
8.4632
8.4733
8.4627
8.5264
8.5100
7.2394
4.7356
2.2556
9.7088
9.7056
9.6974
8.4761
8.4610
8.4607
8.4660
8.4716
1.5439
1.5437
1.5429
0.0000
0.3046
0.3039
0.3053
0.3039
0.3047
0.3905
1.5333
1.5352
1.5332
1.5086
1.4961
2.8331
5.6010
7.8178
0.3047
0.3046
0.3044
1.5357
1.5330
1.5329
1.5338
1.5349
mMEG
1.6168
1.3154
0.4061
0.2910
1.2050
1.5510
1.6437
1.3556
0.4710
0.2979
1.1500
1.2068
1.4300
0.2587
0.2762
0.3182
0.2300
0.2096
1.2223
1.4822
1.5275
1.0151
1.1913
1.4304
1.5161
1.5757
4.5217
1.6136
5.5990
4.4064
0.8367
3.8741
4.7442
2.3409
5.0355
3.5821
0.3445
0.8436
2.8565
2.8435
2.8792
1.7991
0.7296
0.3681
0.9949
3.3377
3.6444
0.3467
0.7143
2.8570
3.6380
4.1760
mNaHCO3 mNa2CO3
4.168
2.514
4.900
5.566
2.253
4.074
4.600
3.527
5.802
4.889
1.515
2.178
2.943
4.081
4.075
2.696
1.211
0.675
2.625
4.002
4.163
1.607
2.093
3.804
4.540
4.519
STNa
0.29
2.26
0.26
0.93
0.55
0.26
0.05
0.14
0.10
0.66
0.16
0.21
0.25
0.06
0.18
0.95
1.68
0.79
0.11
0.06
0.14
0.44
0.65
0.26
0.18
0.08
dev
SCO322.056
1.029
2.441
2.771
0.750
1.873
2.221
1.459
2.881
2.404
0.418
0.783
1.293
1.984
1.972
1.278
0.515
0.255
0.907
1.766
1.876
0.407
0.659
1.800
2.207
2.099
0.48
1.97
0.91
1.51
0.62
0.47
0.10
0.53
0.45
1.15
1.31
0.88
0.43
0.25
0.04
0.84
1.51
2.11
5.07
0.48
0.17
1.31
1.74
0.24
0.37
9.72
dev
1.215
1.176
1.268
1.250
1.109
1.226
1.233
1.236
1.324
1.289
1.123
1.186
1.204
1.250
1.222
1.165
1.119
1.138
1.164
1.240
1.260
1.137
1.167
1.108
1.228
1.291
Usat. iquid
3.78
1.39
1.44
2.54
0.62
2.24
0.37
1.97
0.79
2.27
0.65
1.86
0.68
1.05
1.16
1.76
2.61
0.30
1.48
1.25
0.31
0.96
2.20
0.93
6.89
1.75
dev
21.59
29.22
7.127
19.60
23.343
25.09
3.0696
7.244
13.45
20.975
21.02
13.50
5.60
2.71
8.223
18.043
19.26
3.057
6.148
18.76
23.205
-
wNa2CO3
0.034
0.12
0.0092
0.032
0.0062
0.084
0.00071
0.0034
0.024
0.0072
0.021
0.081
0.070
0.019
0.0025
0.0073
0.017
0.0021
0.0086
0.010
0.0055
-
VNa2CO3
0.79
0.43
7.63
3.16
1.641
1.30
7.86
6.81
3.40
1.04
0.91
1.25
1.29
1.37
9.01
5.021
4.43
8.66
7.83
2.21
1.35
-
wNaHCO3
0.049
0.25
0.12
0.040
0.0094
0.14
0.021
0.044
0.024
0.010
0.029
0.087
0.061
0.015
0.028
0.0081
0.020
0.062
0.13
0.065
0.059
-
VNaHCO3
7.8114
9.9096
9.7036
9.1846
9.1117
9.6724
8.4632
8.4733
8.0168
8.4723
8.4410
7.1660
4.6982
2.2271
9.7088
9.3398
9.2812
8.4761
8.4610
8.4607
8.4660
-
mH2O
T
M
T
M
P1f P2f
12.810 T
- B
- T
0.000 T
2.594 B
2.474 T
2.432 T
- B
- T
2.857 T
13.663 B
13.183 B
13.350 T
11.788 T
11.754 T
24.154 T
50.634 T
74.652 T
2.519 B
2.430 T
2.423 T
13.542 B
13.194 B
12.121 W
11.572 W
- W
wMEG
MEG , salt-free weight percent of MEG in the initial composition, mMEG/(mMEG+mH2O). c The weight of the initial composition in grams, mi, of component i. d
equilibrium temperature in C. b w
15.43
15.43
15.43
0.00
3.04
3.04
3.04
3.04
3.04
3.86
15.34
15.34
15.34
15.03
14.95
28.13
54.19
77.61
3.04
3.04
3.04
15.34
15.34
15.34
15.34
15.34
30.00
29.90
30.00
39.90
39.60
39.80
39.80
39.70
39.80
39.90
39.20
39.80
39.60
39.90
39.90
39.90
39.80
39.90
49.70
49.75
49.70
48.80
49.60
49.75
49.70
49.40
mH2O
saturated liquid at equilibrium temperature in g/cm3. The deviation of STNa, SCO32-, and Usolution (dev) is calculated as the standard deviation, V, divided by the mean of the samples times hundred. e
The result of the reverse Schreinemakers method. wi is the weight percent of component i per kilogram of total solution calculated by equation (225), (227) and (240). mH2O is the grams of water in
the saturated liquid phase at equilibrium calculated by equation (238). Vi, is the standard deviation of the weight fraction i. f P1 and P2 is the precipitated solid phases. B: NaHCO3, C: Na2CO3, M:
Na2CO3.H2O, H: Na2CO3.7H2O, D: Na2CO3.10H2O, T: trona, and W: wegscheiderite.
The equilibrium composition determined by titration. STNa is calculated by equation (220) as mol sodium/kg total and SCO32- by equation (218) in mol CO32-/kg total. Usat. liquid is the density of the
w MEGb
Ta
7.4.1
Meta stable solutions
A challenge when measuring solubility in the Na2CO3-NaHCO3-MEG-H2O system at

lower temperature is the formation of meta-stable solutions. Several different metastable phases may form depending on the initial concentration and the added seed
crystals.
10
Stable 3 wt % MEG
20C
Meta-Stable 3 wt % MEG
wt % NaHCO3
Stable 15 wt % MEG
Stable
NaHCO3
Meta-Stable 15 wt % MEG
Stable
Na2CO3NaHCO32H2O
Meta-Stable
Na2CO3NaHCO32H2O
Meta-Stable
Na2CO3H2O
Stable
Na 2CO310H 2O
0
0
10
15
20
25
30
wt % Na2 CO 3
Figure 27: Stable and meta-stable data at 20 C in the Na2CO3-NaHCO3-MEG-H2O system. , :
NaHCO3. , , : Trona. , : Na2CO310H2O. : Na2CO3H2O. Dotted line: meta-stable trend.
In this study meta-stable solutions have a tendency to form below 35 C from initial
solutions from which salt hydrates were expected to precipitate. For example Na2CO3
solutions may stand for several days at very high super-saturation at 10 C to 25 C
without solids precipitating. By seeding, precipitation occurs instantly. Seeding an
aqueous Na2CO3 solution at 10 C by Na2CO3 will make Na2CO3H2O precipitate, but
seeding using Na2CO310H2O initiates Na2CO310H2O precipitation. The equilibrium
experiments related to Na2CO3 were carried out by supersaturating a solution by a few
wt% Na2CO3, seeding, and determining the composition by the reverse
Schreinemakers method. Na2CO37H2O was not used as a seeding crystal, but the
reverse Schreinemakers method revealed that one solution precipitated Na2CO37H2O.
Table 8 shows the meta-stable data determined in this study. The columns correspond
to those of table 7. Figure 27 shows a selection of meta-stable data from table 8
determined in this study. The stable branches of the isotherms consist of a NaHCO3
branch, a trona branch and a Na2CO310H2O branch. The trona branch at 3 wt% MEG
is very short. The meta-stable isotherms have longer trona branches and the
- 101 -
113
Na2CO310H2O branch is replaced by a Na2CO3H2O branch. The meta-stable

solutions in equilibrium with trona can easily be reproduced and it is difficult to get
the stable phase of Na2CO310H2O without using seed crystals.
The long meta-stable trona branch in 3 wt% MEG indicates that trona can be produced
consistently at 20 C even though Na2CO310H2O is expected to precipitate. On the
other hand if impurities of Na2CO310H2O are present in the crystallizer, then
Na2CO310H2O precipitates immediately.
Seed crystals of Na2CO310H2O can be stored at room temperature in a saturated
Na2CO3 solution. The crystals grow to big hydrates over a few days. Na2CO310H2O
may not be stored in open air since it quickly loses water. A test was made where an
amorphous single 0.25cm3 crystal was stored in air. It decomposes in a few hours due
to evaporation of water. It forms crystals of lower Na2CO3 hydrates and becomes
white.
- 102 -
114
115
Initialc
mol/kg
g/cm3
Meta-stable solutiond
mol/kg
30.00
2.93 9.7162 0.2936 0.0000 4.9251 4.429 0.01
2.207
0.11
1.259 0.28 23.472
29.90
2.93 9.7317 0.2941 0.0000 4.5903 5.421 0.31
2.696
0.60
1.301 1.22
28.73
30.00
15.17 8.5091 1.5221 0.0000 2.8568 3.403 0.90
1.692
1.16
1.198 0.29
18.04
29.90
15.43 8.4678 1.5445 0.0000 4.7036 4.935 3.14
2.484
2.86
1.303 1.27
26.15
29.80
15.43 8.4507 1.5414 0.0000 4.5941 5.101 0.12
2.531
0.24
1.306 0.63
27.03
29.90
27.93 7.2273 2.8006 0.0000 2.5354 2.704 0.75
1.337
1.07
1.173 0.45
14.33
20.00
2.93 9.7135 0.2936 0.9606 2.7433 3.563 0.15
1.594
0.36
1.203 0.79
16.76
20.00
2.93 9.7023 0.2932 0.7427 2.6877 3.685 0.07
1.675
0.60
1.207 0.49 17.735
20.00
2.93 9.7021 0.2932 0.5215 2.6353 3.804 0.22
1.767
0.65
1.203 0.24
18.7
20.10
2.93 9.7144 0.2936 0.3914 2.5992 3.871 0.24
1.804
0.10
1.227 0.46
19.23
20.00
2.93 9.7253 0.2939 0.4096 2.7680 3.993 0.30
1.874
0.33
1.232 0.05
20.01
20.00
2.93 9.7234 0.2939 1.6334 4.5144 4.312 0.03
2.075
0.52
1.255 2.23 21.901
20.00
2.93 9.7055 0.2933 0.4132 2.9801 4.148 0.14
1.982
0.01
1.236 0.63
21.07
20.10
2.93 9.7235 0.2939 0.7449 4.1752 4.832 0.11
2.387
0.13
1.258 0.46
25.3
20.00
2.93 9.7249 0.2939 0.4399 4.0581 4.530 0.17
2.201
0.23
1.253 0.47
20.00
15.43 8.4662 1.5443 1.6138 4.5150 4.334 0.24
2.138
0.21
1.267 0.20
22.04
20.10
15.43 8.4573 1.5427 1.5409 3.8445 3.509 2.68
1.703
2.63
1.187 1.64
17.86
20.00
15.43 8.4614 1.5434 0.0603 3.3499 4.556 0.19
2.246
0.01
1.277 0.41
23.85
20.00
15.43 8.4606 1.5433 0.5692 3.5209 4.319 0.50
2.117
1.269 0.48
22.61
20.00
15.43 8.4653 1.5441 0.1216 3.3729 4.511 0.21
2.230
0.47
1.273 0.99
20.00
15.43 8.4678 1.5446 0.1742 3.3933 4.537 0.50
2.232
0.68
1.278 0.37
20.10
15.43 8.4698 1.5449 0.2601 3.4191 4.517 0.75
2.228
0.70
1.280 0.76
20.00
15.43 8.4623 1.5435 0.4161 3.4759 4.353 0.19
2.144
0.23
1.272 0.25
20.00
15.43 8.4635 1.5438 0.7440 4.1814 4.723 0.39
2.355
1.268 0.41
30.00
2.93 9.7058 0.2933 0.0275 4.6715 5.997 0.13
2.981
0.51
1.355 1.03
31.66
29.80
2.93 9.7129 0.2935 0.5414 4.9622 5.423 0.49
2.682
0.62
1.287 0.93
29.90
2.93 9.7225 0.2938 0.1836 4.6845 4.488 0.04
2.183
0.34
1.265 0.51
30.00
15.43 8.4640 1.5439 0.5630 3.0946 3.811 0.08
1.856
0.32
1.227 0.02
19.93
29.90
15.43 8.4597 1.5431 0.0280 4.6156 4.745 0.35
2.359
0.15
1.270 1.59
25.01
30.00
15.43 8.4630 1.5437 0.1868 4.6814 3.330 0.90
1.631
0.85
1.215 0.06
a
system temperature in C. b c e f see table 7. d The meta-stable composition determined by titration. Symbols defined in table 7.
0.0033
0.088
0.16
0.82
0.031
0.11
0.016
0.0078
0.025
0.027
0.034
0.0034
0.017
0.013
0.029
0.28
0.045
0.059
0.042
0.0088
0.088
-
0
0
0
0
0
0
3.36
2.843
2.31
2.04
1.83
1.506
1.45
0.49
1.48
1.17
0.4578
0.45
0.1875
0.42
0.2142
0.018
0.0092
0.031
0.034
0.045
0.0049
0.023
0.023
0.041
0.35
0.00034
0.089
0.00015
0.012
0.00031
-
9.3835
9.6315
8.3944
8.2586
8.2888
7.0772
9.4781
9.5346
9.6027
9.6570
9.6495
9.1021
9.6077
9.4314
7.8518
7.8550
8.4261
8.2410
9.7010
8.2446
8.2309
-
Reverse Schreinemakerse
%
2.322
2.112
12.581
11.635
11.442
24.289
2.400
2.370
2.340
2.323
2.310
2.395
2.295
2.242
12.571
13.292
11.717
12.136
2.000
12.563
11.805
-
%
M
M
M
M
M
M
T
T
T
T
T
T
T
T
M
T
T
M
T
T
T
T
T
T
M
T
M
T
M
D
M
T
M
M
M
M
M
Table 8: Experimental initial composition, meta-stable composition, results of the reverse Schreinemakers method and solid phase information from the Na2CO3-Na2CO3-MEG-H2O system at 1
bar and 20 C to 30 C.
wNaHCO3 VNaHCO
mH2O
Ta
dev
mMEG
mNaHCO3 mNa2CO3
STNa
dev
wMEG
P1f P2f
Usat. liquid dev wNa2CO3 VNa CO
w MEGb mH2O
SCO 2-
7.5
Parameter fitting
The parameter estimation routines used in this study consist of four parts. A
thermodynamic model, a speciation routine for calculating the equilibrium
composition in the liquid and the gas phase, a set of routines for calculating specific
properties and a parameter optimisation routine. The modelling extends the previous
work by Thomsen and co-workers7,8,9,43.
An object function is calculated for the optimisation routine in which experimental
and calculated properties are compared. The value of the object function, and thereby
the difference between calculated and experimental properties, is minimized by
varying the parameters of the extended UNIQUAC model using a combination of
Nelder-Mead and Marquardt-Levenberg minimization routines. The extended
UNIQUAC parameters are constrained by a penalty function in the objective function.
The volume and surface parameters, ri and qi, are constrained to the interval from 0.05
to 15, the interaction parameters, uijo , to the interval from -2500 to 2500 and the
temperature gradients of the interaction parameters, uijt , to the interval from -20 to 20.
If parameters were determined to have a value corresponding to one of these
boundaries, it was taken as an indication that either the parameter was unnecessary,
some data were wrong, or that a parameter was given a bad initial value. Maximum
step sizes were used for the parameters to avoid extreme values. The objective
function is given by
Pcalc Pexp
0.05 Ptotal,exp 0.01
Ex
Ex
H calc
H exp
14.4 R
VLE data
H Ex data
' sol H data
' sol H calc ' sol H exp
120.3R
Lit . SLE data
(242)
' r G o RT i Q ir ln ai
0.05
Our SLE data
5 STNa ,exp STNa ,calc
The first term is related to the binary and ternary VLE data. The bubble point pressure
is calculated for a given solution. The weighting factor of 0.05 is used to make a
relative deviation of 5 % contribute by 1 in the objective function. The residual at low
pressures may become very high when dividing by Ptotal,exp. This is overcome by
adding 0.01 bar, and therefore pressures below 0.01 bar receive little weight. The
bubble point pressures of the solutions considered here were between 0.0002 and 5
bar. The weighted residuals calculated with the optimized parameters are in the range
from -3 to 3.
Absolute deviations are used for calculating the contribution of excess enthalpy data
to the object function. A difference of 14.4R =120 J/mol between calculated and
- 104 -
116
Ex
Ex
experimental excess enthalpy corresponds to H calc
H exp
(14.4 R) 1 and therefore
contributes by 1 to the object function. The weighted residuals for excess enthalpy
data calculated by the optimized parameters are in the range from -1 to 1. The
contribution of heat of solution data to the objective function are also calculated from
absolute differences. A difference of 120.3R=1000J/mol between calculated and
measured heats of solution gives ' sol H calc ' sol H exp 120.3R 1 and therefore
contributes by 1 to the objective function. The weighted residuals of this type
calculated with the optimized parameters are in the range from -0.2 to 0.2.
The deviation of SLE data was added to the objective function by determining the
logarithm of the saturation index, ln(SIk) = ln
Q ik
i in k
Q
i
ik
KT , Po ,k = ' k GTo, P RT +
o
ln ai for reaction k with equilibrium constant Kk and stoichiometric coefficients
Qik of the ith component with activity ai. The experimental data define which salts
saturate the liquid and ln(SI) of those salts are returned in the objective function and
should be 0 when parameters are optimized. The weighting factor of 0.05 scales the
contribution of SLE data to be 1 if the calculated ln(SI) is 0.05. A penalty is added if a
wrong salt precipitates compared to salts determined by the dataset. SLE data of ice
formation are weighted 100 times higher because this type of data are usually very
accurate. The weighted residuals of SLE data calculated with the optimized
parameters are in the range from -10 to 10.
The equilibrium data measured in this work are represented by the last term of the
objective function, (242). The equilibrium routine was set up to use the initial
composition, the identity of the solid phase, and the temperature to calculate mol
sodium/kg total, STNa,calc, The calculated value was compared to the measured value.
The weighting was set to give a contribution of 1 when the deviation is 0.2 mol
sodium/kg total. The weighted residuals calculated with the optimized parameters are
in the range from -2 to 2. Results of the reverse Schreinemakers method were only
used for plotting and were not used in the fitting process. Therefore the two-salt
saturation points are not represented in the figures even though they were used in the
fitting process.
350 VLE data, 192 thermal property data and 209 SLE data were used in the fitting
process. All literature values used were taken from original articles. Data collections
were deliberately not used in order to avoid printed errors and misinterpretations.
Two MEG specific parameters and 12 MEG related interaction parameters were
optimized by minimizing the objective function in equation (242).
An average relative deviation, ARD, was calculated in order to compare experimental
and calculated values. The calculated ARD is defined by
ARD
1
nexp
nexp
uexp ucalc
i 1
uexp
u is an arbitrary calculated or experimental value and nexp is the number of

experimental data.
- 105 -
117
(243)
7.5.1
Parameter estimation procedure
Binary, ternary, and quaternary data were used in the fitting process. In the parameter
estimation procedure the initial UNIQUAC volume and surface area parameters for
MEG, rMEG and qMEG, were given guesses of 5. Interaction parameters were given
initial values of 500 with a temperature gradient of 0.5.
The fitting process was divided into five steps. First the MEG-MEG and MEG-H2O
interaction parameters were fitted to the binary VLE, HEx, and freezing point
depression data for the MEG-H2O system. Freezing point depression data were only
used for ice precipitation up to 55 wt% MEG since thermodynamic properties of
MEGH2O(s) and MEG(s) were unknown. Second the CO2-MEG interaction
parameters were fitted to the CO2-MEG-H2O data. In the third step r, q, and MEGNa+, MEG- HCO3 and MEG- CO32 interaction parameters were fitted to SLE data for
the NaHCO3-MEG-H2O and Na2CO3-MEG-H2O systems. In the last step, all these
parameters were simultaneously fitted to all the binary, ternary, and quaternary
Na2CO3-NaHCO3-MEG-H2O data. The parameters were gradually refined by
repeating this procedure a few times until a low objective function value was
obtained. Finally the standard state property for MEGH2O(s) and MEG(s) were fitted
to SLE data for the MEG-H2O system in the whole composition range.
The MEG-OH- interaction was not fitted in this study due to the lack of data. It was
o
t
therefore given the fixed value uOH
0 equivalent to no
2500 and uOH
MEG
MEG
interaction. The OH concentrations in the solutions studied are up to 10-3 molal.
Therefore interaction has little or no effect on the thermodynamics of the solutions
studied here. The MEG-H+ interaction parameter is defined in accordance with the
systematic approach defined by Thomsen7, see appendix D.3.
7.6
Results and discussion
7.6.1
Correlation of density
The densities measured in this study are reported in table 7. It was found that the
density data for solutions containing H2O, MEG, NaHCO3, Na2CO3, and NaCl could
be correlated using the simple, empirical equation:
U sol
U H O 'U MEG 'U NaHCO 'U Na CO 'U NaCl

2
(244)
The density of water, U H 2O (g/cm3), was calculated by the Kell44 correlation:
UH O
2
0.99983952 16.945176 10
3
t 7.9870401 106 t 2
46.170461 109 t 3 105.56302 1012 t 4

280.54253 10
15 5
(245)
1 16.87985 10 t
3
where t is the temperature in C. The contribution from glycol, 'U MEG (g/cm3), was
fitted to the following equation:
- 106 -
118
'U MEG
w MEG d1 w MEG d 2 d3 t 25qC d 4 t 25qC

2
(246)
w MEG is the salt-free weight fraction of MEG, tilde indicate salt-free solution. d1 and
d2 were fitted to data at 25 C of Hayduk and Malik45 and Huot et al.46 as shown in
figure 28. d1 and d2 are found in table 9.
0.12
25 C
0.10
'UMEG
0.08
0.06
0.04
Model
Hayduk and Malik
0.02
Huot et al.
0.00
0
0.2
0.4
0.6
0.8
salt free MEG wt. frac.

Figure 28: Density of MEG-H2O solutions. 'UMEG as function of salt-free MEG wt. fraction.
Mol fraction of MEG was not used in this study since fitting 'U MEG as a polynomial
function of MEG mol fraction requires more parameters. A small refinement of the
model was made by fitting the temperature dependence of 'U MEG , d3 and d4, to data
from Kapadi et al.47, Sakurai48, George and Sastry49, Cocchi et al.50, and Douheret
and Hoiland51. The calculated density was compared to Gallants graph52 and the
agreement is good as shown in figure 29. The density measured by Yang et al.53 were
deliberately not used since they diverge compared to the remaining data. d3 and d4 are
found in table 9.
The densities of solutions containing 0 to 15 wt% MEG, saturated in one or both salts
of Na2CO3 and NaHCO3, are shown in figure 30. The plot shows a linear trend of the
density as function of STNa. The trend is also observed by looking at the two salts
separately, which is not shown here. Densities of solutions containing higher MEG
concentrations are also linear, but have a higher density compared to densities in
figure 30.
- 107 -
119
1.14
Model
Gallant
Sakurai
George & Sastry
Cocchi et al.
Kapadi et al.
Douheret et al.
Density of MEG-water (g/cm 3)
1.12
1.10
1.08
1.06
100 wt.% MEG
1.04
80 wt.% MEG
60 wt.% MEG
1.02
40 wt.% MEG
1.00
20 wt.% MEG
0.98
0 wt.% MEG
0.96
-1
19
39
59
79
99
119
Temperature (C)
Figure 29: Density of MEG-H2O solution as function of temperature and salt-free MEG wt. fraction.
From these results it is assumed that contributions from salts, 'U NaHCO3 , 'U Na2CO3 , and
'U NaCl , in g/cm3 in (244) follows a linear dependence by:
'U NaHCO3
d5 wNaHCO3
'U Na2CO3
d 6 wNa2CO3
'U NaCl
(247)
d 7 wNaCl
The parameters d5 and d6 were fitted to the experimental data from table 7 calculated
using the reverse Schreinemakers method. This indicate that the experimental
densities were fitted using the measured experimental compositions of the liquid
phase. The result is shown in figure 31 and 32. The calculated density shown in figure
31 is lower than the experimental density at high MEG concentrations. The deviation
is related to the accuracy of the extended UNIQUAC model.
Figure 32 shows a curvature below 40 wt% MEG at low temperature. This is due to
the precipitation of Na2CO310H2O also shown by figure 45. The density has a very
distinct temperature dependence which is reproduced by the model.
- 108 -
120
1.40
Single and double salt

experimental data at
0 to 15 wt% MEG.
2 to 60 C.
1.35
density (g/cm )
1.30
1.25
1.20
1.15
1.10
1.05
1.00
0
STNa (mol Na/kg)

Figure 30: Density of Na2CO3-NaHCO3-MEG-H2O solution as function of sodium content.
density (g/cm ) at eq. temperature
1.18
1.16
1.14
1.12
1.10
1.08
1.06
10C
20C
30C
35C
40C
50C
1.04
1.02
10C
20C
30C
35C
40C
50C
1.00
0
0.2
0.4
0.6
0.8
salt free MEG wt. fraction

Figure 31: Density of saturated NaHCO3-MEG-H2O solution determined using equation (244).
Composition of the saturated solution determined using the extended UNIQUAC model.
- 109 -
121
d7 was fitted to the un-saturated NaCl-MEG-H2O data of Sandengen and Kaasa54. A

comparison is shown in figure 33. The linear trend with respect to wt. fraction of
NaCl is obvious. The parameters d5 to d7 are found in table 9.
No temperature dependence of 'U NaHCO3 , 'U Na2CO3 , and 'U NaCl was included since the
above correlations reproduce the experimental results satisfactory in the used

temperature range. This is for example verified by looking at figure 33. The result is
predicted equally well at 25 and 50 C with no temperature correction.
density (g/cm3) at eq. temperature
1.40
10C
15C
20C
30C
40C
50C
1.35
1.30
1.25
10C
15C
20C
30C
40C
50C
1.20
1.15
1.10
1.05
1.00
0.0
0.2
0.4
0.6
0.8
1.0
MEG wt. fraction

Figure 32: Density of saturated Na2CO3-MEG-H2O solution determined using equation (244).
Composition of the saturated solution determined using the extended UNIQUAC model.
The average relative deviation (ARD) between experimental densities and densities
calculated with equation (244) was used as objective function. The ARD obtained was
0.79 % which is within the experimental error. The data are scattered but the model
reproduces the temperature dependence even though there is no temperature
dependence of the salts in the model.
Equation (244) is also valid for salt-free systems and may be used for calculating
water, MEG, or MEG-H2O density between 0 and 100 C. It must be emphasized that
the model gives an engineering approach to calculating the density and is only
accurate down to 0.01 g/cm3. Figure 34 shows a comparison of experimental and
calculated densities in the salt solutions. There is no tendency to either over- or under
predict. The plot shows that outliers are approximately 5% which is related to the
quality of the density measurement.
- 110 -
122
1.20
1.20
100% MEG
1.15
0% MEG
1.15
100% MEG
0% MEG
density (g/cm )
1.10
1.05
77.4% MEG
1.10
1.05
25 C
50 C
1.00
1.00
model
model
Sandengen and Kaasa

0.95
0.00
0.05
0.10
0.15
0.20
0.25
0.95
0.00
0.30
Sandengen and Kaasa

0.05
0.10
NaCl Wt. fraction
0.15
0.20
0.25
NaCl Wt. fraction
Figure 33: Density of NaCl-MEG-H2O solutions determined by equation (244) at 25 and 50 C.
1.40
5%
1.35
3
calc. density(g/cm )
density (g/cm )
77.4% MEG
2.5%
2.5%
1.30
5%
1.25
1.20
1.15
1.10
1.05
1.00
1
1.1
1.2
1.3
1.4
exp. density (g/cm 3 )

Figure 34: Comparison of calculated and experimental density of NaCl-Na2CO3-NaHCO3-MEG-H2O
solutions between 2 to 60 C.
The density model presented by (244) has certain similarities to the model of
Sandengen and Kaasa54. Their model uses three parameters per salt, and is only valid
at 15 and 20 C. (244) uses one parameter per salt and is valid in a broader
temperature range. The model by Sandengen and Kaasa54 was fitted to the NaHCO3NaCl-MEG-H2O system and not for systems with Na2CO3. In our model Na2CO3 and
- 111 -
123
0.30
NaHCO3 was only fitted to data for saturated solutions but works satisfactory for unsaturated solutions. This was evaluated by comparing our correlation to the dilute
NaHCO3 solutions of Sandengen and Kaasa54. At 20 C it gives an ARD of 0.14 %.
The model by Sandengen and Kaasa54 gives an ARD of 0.12 % for the same data and
the agreement is good. Comparing the models to our data at 20 C for solutions of
NaHCO3 give an ARD of 0.44 % for equation (244) and 0.47 % for the model by
Sandengen and Kaasa54. This is understandable since their data were not used in our
fitting process. Equation (244) gives an ARD of 0.19 % for the NaCl related data at
20 C and the model by Sandengen and Kaasa54 gives an ARD of 0.02 %. It shows
that their model performs better for the NaCl-MEG-H2O system at 20 C. Our model
performs equally well at 50 C which the correlation by Sandengen and Kaasa54 does
not.
Table 9: Correlation parameters for equation (246) and (247).
Parameter
d1
(g/cm3)
-0.0358
d2
d3
d4
d5
(g/cm3) (g/cm3K2) (g/cm3K) (g/cm3)
0.149
3.4110-6 -495.910-6 0.921
d6
(g/cm3)
0.714
d7
(g/cm3)
0.714
The density model represented in equation (244) uses few parameters and is valid at
concentrations up to saturation of the two salts at temperatures between 2 and 60 C.
It is expected that the model can be used in a wider temperature range with good
accuracy. The parameters of the correlation are found in table 9.
Salt contribution for Na2CO3 and NaHCO3 can alternative be calculated by
'U NaHCO3 + 'U Na2CO3 = STNa 0.0509 kg mol . This is a simplified approach and gives an
ARD of 1.04%, where the contribution to NaHCO3 is not as accurately calculated.
7.6.2
Model of the Na2CO3-NaHCO3-MEG-H2O system
Eight species are assumed to be present in the liquid phase as shown in table 10. Two
parameters sets were determined in this work, denoted set A and B. Table 10, 11, 12,
and 13 give the used standard state properties and the regressed interaction
parameters.
Parameter set A contains mainly standard state properties from NIST12 and a few of
such properties regressed in this and in a previous study8 as shown in table 10 and 11.
The interaction parameters were determined by Thomsen et al.13,8 and Garcia et al.55
except for the MEG parameters determined in this study.
Set B include most of the liquid phase parameters used in set A, but part of the
standard state properties of the solid species were optimised, as seen in table 11. The
interaction parameters Na CO32 , Na HCO3 , and OH CO32 were also
optimised to get a better representation of the properties of the aqueous Na2CO3NaHCO3-H2O system as well as of the mixed solvent system, Na2CO3-NaHCO3MEG-H2O.
Figure 35 shows the solubility of Na2CO3 in pure water. The A parameters give an
offset at T>35 C. The same is the case for the NaHCO3 solubility, not shown here.
The improvement is related to better values of the standard state enthalpies of
Na2CO3H2O and Na2CO3(s), and better interaction parameters.
- 112 -
124
Na2CO37H2O
35
Na2CO3H2O
Na2CO3
Na2 CO 3 10H2 O
30
Wt % Na 2 CO3
25
20
15
A parameters
10
B parameters
Experimental
Ice
0
-10
10
30
50
70
90
110
130
Temperature C
Figure 35: The temperature dependence of Na2CO3 solubility in water. Lines calculated with the A and
the B parameters. Experimental data points are from the IVC-SEP database10.
Figure 36 shows the temperature dependence of the two salt saturation lines in the
Na2CO3-NaHCO3-H2O system, they also represent ternary univariant compositions.
The y-axis corresponds to the salt fraction nNa 2CO3 nNa 2CO3 nNaHCO3 in the saturated
solution. The ternary invariant points are shown and represent the points where three
solid phases are in equilibrium with a liquid and a gas phase. The binary invariant
points are shown on the top and bottom axes where salt fraction is zero or one, and
two solid phases are in equilibrium. It is shown how set B gives a better representation
of the two salt lines, especially the NaHCO3-trona (Na2CO3NaHCO32H2O) line, but
also a better transition temperature between Na2CO3H2O(s) and Na2CO3(s).
- 113 -
125
Na2CO37H2O
Na2CO3H2O
1
Na2CO3 10H2O
Na 2 CO 3
0.9
0.8
Na2CO3NaHCO32H2O
0.6
Na2CO33NaHCO3
0.5
NaHCO3
0.4
Ice
salt fraction
0.7
Na2CO3
0.3
A parameters
0.2
NaHCO 3
B parameters
0.1
Experimental
0
-10
10
30
50
70
90
110
130
Temperature, C
Figure 36: Temperature dependence of the two-salt lines in the aqueous Na2CO3 NaHCO3 system.
Phase lines calculated with the A and the B parameters. Experimental data points are from the IVC-SEP
database10.
A comparison of the standard state properties used in set A and B, reveals that a
change of less than 1 % in these properties gives a better result as shown in table 11.
o
The significant difference, is the change in ' f H Na
. Parameter set B is not
2 CO3 s
compatible with the previously determined parameters, such as those of Iliuta et al.9
and Thomsen et al.43.
The standard state properties of MEG(s) and MEGH2O(s) were determined in this
study. The values are almost identical between set A and B and the differences are
within the confidence intervals of the properties as shown by table 11.
o
o
The rational symmetrical standard state properties ' f GMEG
l and ' f H MEG l , were
obtained from the NIST tables12 as listed in table 11 and the extended UNIQUAC
model were used in the conversion of MEG(l) to MEG(aq). MEG(aq) is assumed to
be the only liquid glycol compound in solution. The calculated molal unsymmetrical
*
*
standard state properties ' f GMEG
aq and ' f H MEG aq , are listed in table 10. These
values were calculated by the method given by Iliuta et al.9 . The values are
*
*
' f GMEG
aq =-335.1r0.3kJ/mol and ' f H MEG aq =-461.6r1.1kJ/mol, calculated using
parameter set A. The heat capacity of MEG(aq), C *p , MEG aq , is calculated by using the
- 114 -
126
DIPPR heat capacity correlation of MEG(l), C po , MEG l , and the extended UNIQUAC
model. The conversion is performed by a linear interpolation of C *p , MEG aq between 0
and 110 C giving the correlation parameters for MEG(aq) shown in table 10 for
equation (215). Heat capacity correlation parameters of other compounds are given in
table 10 and 11. The UNIQUAC volume and surface area parameters for MEG,
rMEG(aq) and qMEG(aq), were determined in this work by fitting to data. The values are
shown in table 10. They are different from those obtained by the calculation method
of Abrams and Prausnitz56 and Bondi57, rMEG(l)=2.4088 and qMEG(l)=2.248, as given by
Skjold-Jrgensen et al.58 and later by Chiavone-Filho et al.59,60, Lancia et al.61, and
Horstmann et al.62. Our regressed values are considerably higher, but they are
comparable to values determined by Iliuta et al.9 who obtained r=3.3 and q=3.7 for
methanol, and Thomsen et al.43 who got r=q=5.9 for ethanol and r=10.3, q=9.5 for 1butanol. Since the MEG molecule is larger than methanol it should have higher r, and
q values than methanol and comparable or higher values than ethanol but less than 1butanol, which it also has.
The extended UNIQUAC interaction parameters are summarized in table 12 and 13.
The most noteworthy difference between set A and B is found in table 13. The
temperature gradient of the MEG-MEG, MEG-Na+, and MEG- HCO3 interaction
energy parameters are similar in numerical size but have opposite sign. The
temperature gradient of the Na+- HCO3 interaction parameter is noticeable different
in the two parameter sets. The improved temperature dependence of the bicarbonate
solubility might be related to this difference.
- 115 -
127
128
A and B
A
B
A and B
A and B
A and B
A and B
A and B
A and B
M
ri
qi
g/mol
18.01528 0.9200 1.4000
62.06844 5.4996 7.0797
62.06844 4.0060 5.4610
44.00980 5.7410 6.0806
22.98977 1.4034 1.1990
1.007940 0.13779 110-16
17.00734 9.3973 8.8171
60.00920 10.828 10.769
61.01714 8.0756 8.6806
kJ/mol
-237.129
-334.9996d
-334.7257d
-385.98
-261.905
0
-157.2481e
-527.81
-586.77
' f G*
f
kJ/mol
-285.83
-461.7981d
-461.3862d
-413.8
-240.12
0
-230.2433e
-677.14
-691.99
'f H*
ai
J/(molK)
58.36952
55.90368d
72.35423d
243.0000
600.6158
0
1418.157
894.6877
-0.6770971
bi
J/(molK2)
0.03896110
0.3134489d
0.2843604d
0
-1.100623
0
-3.445769
-2.827237
0.2737451
ci
J/mol
523.8794
0
0
0
-23231.92
0
-51473.13
-21149.44
-10089.51
ri, qi, ai, bi, and ci determined/defined by Thomsen et al. 7,13. b ri, qi, ai, bi, and ci Determined by Thomsen and Rasmussen8. c Parameters determine in this study. d Calculated using
the extended UNIQUAC model from MEG(l) properties. e Determined by Kaj Thomsen, previously unpublished. f Obtained from NIST12 if not stated otherwise.
CO2(aq)b
Na+(aq)a
H+(aq)a
OH-(aq)a
CO32-(aq)b
HCO3-(aq)b
Unit
H2O(l)a
MEG(aq)c
Set
Table 10: Molar mass, M, extended UNIQUAC parameters of set A and B, standard state formation properties and correlation parameter for the heat capacity equation (215) for
liquid species.
129
B
-1045.387b
-3428.800b
-1287.599b
-2715.602b
-236.538f
-851
-851.352b
c
-2379.85
-2382.157b
c
-3599.73
-3606.270b
-228.572
-394.359
-321.4179r0.8b -321.4198r0.6b
-558.298r1.1b
-557.9913r0.8b
-323.08
B
-1121.155b
-4078.249b
-1426.673b
-3197.269b
-292.624f
-950.81
-947.6426b
-2684.9
-2676.566b
-3982.7
-3979.125b
-241.818
-393.509
-467.4966r3.1b -467.4533r2.5b
-764.9524r3.7b -766.1036r2.7b
-454.8
A
-1130.68
-4081.32
-1431.26
-3199.97
kJ/mol
kJ/mol
A
-1044.44
-3427.66
-1285.31
-2714.2
' f H o (25 C, 1 bar)
' f G o (25 C, 1 bar)
ai, bi=ci=0
J/(molK)
A
B
112.3
550.32
145.6
300c
303.31a
47.89955 f
87.61
349.7375c 245.51a
781.5798 358.23a
33.577
37.11
93.98a
122.52a
molK
4
35.53 0.43678T 1.8486 10 T (T in Kelvin). f Estimated by Kaj Thomsen. Not previously published.
Value calculated using Kopps modified rule14. b Estimated in this work. c Estimated in a previous study by Thomsen and Rasmussen8. d Properties of MEG(l) is used for estimating
105.9887
286.1415
124.004
232.0957
18.01528
84.00691
226.0262
358.0095
18.01528
44.0098
62.06844
98.099
62.06844
MEG(aq). e DIPPR correlation: C p , MEG J
Set
Na2CO3(s)
Na2CO310H2O(s)
Na2CO3H2O(s)
Na2CO37H2O(s)
Ice
NaHCO3(s)
Na2CO3NaHCO32H2O(s)
Na2CO33NaHCO3(s)
H2O(g)
CO2(g)
MEG(s)
MEGH2O(s)
MEG(l)
M
g/mol
Table 11: Standard state properties of set A and B. Correlation parameter for the heat capacity equation (215).Values not marked are from NIST12.
130
H2O
H2O
0a
0a
4.2392c
0.65204c
-0.21916c
4.4442c
-3.5685c
CO2(aq)
A and B
13.629b
-0.22080b
0a
0a
0a
2.7730b
B
0a
0a
20.278b
3.7820b
2.7887c
b
1.8290 -0.040973c
Na+(aq)
H+(aq)
A and B
0a
0a
0a
0a
B
5.6169b
2.7496d
3.6426c
a
0
OH-(aq)
CO32-(aq) HCO3-(aq)
A and B
A and B
-1.3448b
1.7241b -0.019813b
Defined as a reference or no data available to fit this parameter. b Determined by Thomsen and Rasmussen8. c Determined in this study. d Determined by Garcia et al.55
2.1482c
-0.50597c
0.56247c
5.0355c
3.9323c
MEG(aq)
u tji in K2, extended UNIQUAC interaction energy parameters for parameter set A and B
set
A
B
0a
H2O
0.71268c
0.18095c
MEG(aq)
7.5184b
CO2(aq)
+
0.48719b
Na (aq)
+
0a
H (aq)
8.5455b
OH (aq)
23.3516b
CO3 (aq)

HCO3 (aq)
-0.38795b
CO2(aq)
Na+(aq)
H+(aq)
OH-(aq)
CO32-(aq) HCO3-(aq)
A and B
A and B
A and B
A and B
A
B
A
B
40.51756b
1788.825b
0a
9a
110
1109a
0a
2500a
1398.137b
1109a
1562.881b
2500a
547.9543b 454.124c 1109a 1588.025b 1423.716c 1458.344b
651.045b 1101.92b 979.940c 1109a
2500a
800.0081b 771.0377b
Defined as a reference or no data available to fit this parameter. b Determined by Thomsen and Rasmussen8. c Determined in this study.
Table 13: uij
339.9663c 350.4045c
174.5583c 173.2765c
292.9922c 496.3903c
1109a
2500a
671.8906c 566.6611c
676.5772c 616.8792c
MEG(aq)
u oji in K, extended UNIQUAC parameters for parameter set A and B
set
A
B
0a
H2O
249.7378c 241.2451c
MEG(aq)
41.07171b
CO2(aq)
+
733.2863b
Na (aq)
+
1104a
H (aq)
600.4952b
OH (aq)
2361.3877b
CO3 (aq)
HCO3 (aq)
577.0502b
Table 12: uij
The result of the MEG-H2O VLE data correlation is illustrated in figure 37. It shows
the bubble point pressure as a function of liquid composition. The trend indicates that
MEG behaves ideal since the pressure is linear with composition. This is only the case
for the MEG-H2O system and not for the ternary and quaternary salt solubility.
0.8
Model, 0C
Model, 25C
Model, 50C
Model, 60C
Model, 65C
Model, 70C
Model, 80C
Model, 90C
0.7
total pressure (bar)
0.6
0.5
Exp, 0C
Exp, 25C
Exp, 50C
Exp, 60C
Exp, 65C
Exp, 70C
Exp, 80C
Exp, 90C
0.4
0.3
0.2
0.1
0
0.0
0.2
0.4
0.6
0.8
1.0
mol fraction MEG

Figure 37: Total vapor pressure of the mixed solvent MEG-H2O system, lines calculated using the
model and parameter set B. References for the used experimental data points are listed in table 14.
Table 14 summarizes the experimental VLE data used for parameter estimation in this
study. Column T, P gives the temperature and pressure interval of the data. N signifies
the number of data used and the number of data points found in each paper. The
temperature interval 0 C to 90 C is the focus of this study. In this temperature range
the vapour pressure of MEG is insignificant compared to the vapour pressure of water
and CO2. Therefore we did not use vapour-liquid equilibrium data for pure MEG.
Trimble and Potts63 and Sokolov et al.64 made measurements in a broad temperature
interval but only data from the lower temperatures were included here. Villamanan65
published a collection of data, but they were not used in this study since they had
previously been published by Gonzalez et al.66 and Villamanan et al.67. The data by
Skripach and Temkin68 were not used since they were too scattered as noted by
Kumar et al.69. The data from a few authors were not included because they only
measured a few data at high temperatures70,71,72,73,74,75,69,61. Liu et al. 76 did not report
the equilibrium pressure and their data are not suited for thermodynamic modelling.
The thorough study of Gallant52 shows useful graphs but no tabulated experimental
values. The study was therefore only used for comparison.
The VLE performance of parameter set A and B is shown in table 14. The average
relative deviation in the total pressure of set B is slightly better than A. Figure 38.1
and 39.1 shows the predictions by the model and the two set perform very similar.
- 119 -
131
Table 14: Used experimental MEG-H2O and CO2-MEG-H2O VLE data.
Author
MEG-H2O data
Villamanan et al.67
Gonzalez et al.66
Horstmann et al.62
Chiavone-Filho et al.59
Trimble and Potts63
Nath and Bender77
Sokolov et al.64
P interval
bar
T interval
C
# data used
(actual # data)
P Deviation
% (ARD)
Parameter set
A
B
4.4
3.7
4.3
3.6
8.1
7.4
4.0
3.5
1.7
1.8
5.2
4.9
10.0
10.0
60
50
60, 80
70, 90
72 - 89
65, 77, 90
18 - 87
21(24)
20(21)
60(84)
38(40)
15(79)
36(42)
41(51)
Mokbel et al.11
Villamanan65
Skripach and Temkin68
Efremova et al.70
Frolova et al.71
Ramanujam and Laddha72
Ogorodnikov et al.73
Liu74
Kireev and Popov75
Kumar et al.69
Lancia at al.61
Liu et al.76
Gallant52
0.01
- 0.19
0.006 - 0.12
0.003 - 0.47
0.06
- 0.66
0.30, 0.57
0.02
- 0.67
0.01, 0.03, 0.06,
0.13
0.0002 - .62
0.0001 - 0.2
0.1
1.01325
1.01325
1.01325
1.01325
1.0138
1.01325
0.97
0.4
- 1.5
Not given
1
- 35
-15 - 90
50,60
47 - 136
103 - 170
102 - 183
107, 118, 136
103 - 197
103 - 195
100 - 189
99 - 197
98, 110, 122
103 - 195
0
- 240
97(106)
0(48)
0(11)
0(9)
0(11)
0(3)
0(10)
0(10)
0(18)
0(7)
0(42)
0(10)
0(0)
5.6
CO2-MEG-H2O data
Hayduk et al.45
Davis et al.78
Won et al.79
Kobe and Mason80
Sandengen4 a
1.01325
5
1.01325
1.01325
1
25
0, 25, 60
25
25
4, 22
10(13)
7(18)
3(3)
0(3)
0(3)
23.5
31.3
25.1
The data of Sandengen were measured in the system CO2-NaHCO3-MEG-H2O. pH was the only
dependent variable measured.
- 120 -
132
4.9
-
15.0
29.5
12.3
-
exp. Heat excess/R
-100
0.7
-80
-60
-40
-20
0
VLE data
Heat Excess data
-10
0.6
-20
calc. Heat excess/R
calc. total P (bar)
0.5
-30
-40
0.4
-50
0.3
-60
-70
0.2
-80
0.1
0
0
0.2
0.4
exp. Total P (bar)
-90
-100
0.6
2.5
Literature SLE data
6
2
5
1.5
calc. SI
calc. Na in sat. solution (mol/kg total)
SLE data from this study
0.5
0
0
exp. Na in sat. solution (mol/kg total)
50
100
150
data number
Figure 38: Comparison of experimental and calculated properties using parameter set A. 1:
Comparison of VLE data. 2: Comparison of excess enthalpy. 3: Our SLE data compared to predictions
by the model. 4: Model prediction of literature data.
The extended UNIQUAC model is also capable of calculating the excess enthalpy
within reasonable accuracy. The excess enthalpy in the binary MEG-H2O system is
related to the surface area parameter qMEG and the MEG-H2O and MEG-MEG
interaction parameters. The volume parameter rMEG does not influence the excess
enthalpy calculation. It is therefore crucial that qMEG is used as a fitting parameter to
model the excess enthalpy.
- 121 -
133
exp. Heat excess/R
-100
0.7
-80
-60
-40
-20
0
VLE data
Heat Excess data
-10
0.6
-20
calc. Heat excess/R
calc. total P (bar)
0.5
-30
-40
0.4
-50
0.3
-60
-70
0.2
-80
0.1
0
0
0.2
0.4
exp. Total P (bar)
-90
-100
0.6
2.5
Literature SLE data
6
2
5
1.5
calc. SI
calc. Na in sat. solution (mol/kg total)
SLE data from this study
2
0.5
1
0
0
0
0
exp. Na in sat. solution (mol/kg total)
50
100
150
data number
Figure 39: Comparison of experimental and calculated properties using parameter set B 1: Comparison
of VLE data. 2: Comparison of excess enthalpy. 3: Our SLE data compared to predictions by the
model. 4: Model prediction of literature data.
Table 15 outlines the used literature sources for excess enthalpy data. Kracht et al.81
are clearly the most valuable contributors. The data from Knnecke et al.82 were
excluded as these data are even more scattered than those in figure 40. Wang et al. 83
made an experimental study but did not publish the experimental values. Figure 40
shows the calculated and experimental excess enthalpy as a function of solvent
composition. The data at 25 C are very scattered and the model represents the data
within the experimental error. The model also reproduces the temperature dependence
of the excess enthalpy, which increase with increasing temperature in accordance with
the experimental data.
- 122 -
134
mol fraction MEG

0.00
0.20
0.40
0.60
0.80
1.00
0
Model, 25C
-10
Model, 50C
Model, 65C
-20
Exp, 25C
Heat Excess/R
-30
Exp, 50C
Exp, 65C
-40
-50
-60
-70
-80
-90
-100
Figure 40: Excess enthalpy for the mixed solvent MEG-H2O system, lines calculated using the model
and parameter set B. The shown experimental data points are obtained from references in table 15.
Table 15: Used experimental MEG-H2O data on heat excess and heat of solution.
Author
Heat excess (HEx)
Villamanan et al.67
Kracht et al.81
Matsumoto at al.84
Dohnal et al.85
Gallego86
Biros et al.87
Rehm and
Bittrich88
Knnecke et al.82
Wang et al.83
Heat of solution
Manin et al.89
Batov et al.90
Nakayama91
Nichols et al.92
MEG interval
wt%
T interval
C
# data used
(actual # data)
18(20)
90(93)
96.6 50
98.5 12.5, 25, 35,
50, 65
97.3 25
3.5 25
96.9 25
92.0 25
63.6 25
Deviation
% (ARD)
Parameter set
A
B
7.6
7.2
5.8
7.1
31(33)
5(5)
18(18)
22(22)
3(3)
9.7
3.0
18.5
5.0
2.4
95.1 25, 35, 45

-
0(21)
0(0)
7.0
15.3
2.3
0.6
9.9
12.6
62.0
18.0
0.3
- 0.6
0.2
Not given
Not given
25
25
25, 40
20, 25, 30
- 123 -
135
2(2)
3(3)
0(2)
0(3)
10.6
2.7
19.7
4.0
5.3
-
1.9
3.5
6.5
2.6
-
Table 15 shows the capability of the model to reproduce the experimental data. The
performance is good and parameter set B reproduces the data better than A. The
difference between the two parameter sets is not obvious when comparing figure 38.2
and 39.2, but the temperature dependence of set B is significantly better than A. The
data by Gallego86 gives a very high deviation and the work may be erroneous, but the
data were still included in the fitting process.
175
Model, 5C
Model, 25C
Model, 45C
Model, 90C
165
155
Exp, 5C
Exp, 25C
Exp, 45C
Exp, 90C
Cp J/(molK)
145
135
125
115
105
95
85
75
0.0
0.2
0.4
0.6
0.8
1.0
mol fraction MEG

Figure 41: Model predictions (lines) and experimental heat capacity data points for the mixed solvent
MEG-H2O system. Data were obtained from various authors93,94,46,53,51. Table 11 shows the parameters
of the used heat capacity correlation (215). Mixture properties calculated using parameter set B.
A few data of heat of solution were included in the objective function. Nakayama91
and Nichols et al.92 published studies on the subject but did not report the
composition. Their work can therefore not be used for determining model parameters.
Table 15 shows how well the heat of solution is predicted. Set A performs better but
this may be due to the few number of data and the relative low weight they obtained
in the objective function.
The parameters of the extended UNIQUAC model are consistent. This is proved by
the result in figure 41 which shows the predicted MEG-H2O heat capacity. It was not
used in the parameter fitting process but the model resembles the measured
values93,94,46,53,51 closely. The ARD is 1.0 %, which is acceptable. The model overpredicts slightly at 5 C, but it is still satisfactory. The apparent molal heat capacity
below 30 mol% MEG is not shown here, but it is predicted less accurately, and
deviation up to 20 % can be expected at infinite dilution. This is acceptable since
- 124 -
136
experimental determination of apparent molal heat capacity is very challenging and

requires high precision equipment. The deviation may be related to the experimental
data and not to the model.
The freezing point depression of MEG in water is shown in figure 42. The result is
very satisfactory. Only the ice precipitation up to 55 wt% MEG data was used for
determining model parameters of MEG(aq). The data for the precipitation of solid
MEG hydrate and pure MEG were used to determine the standard state properties of
MEGH2O(s) and MEG(s). The experimental data used for parameter estimation are
listed in table 16. The few data points not used, are the freezing point of pure water
and solutions in meta-stable equilibrium with MEGH2O(s). Zinchenko and
Zinchenko95 did a very thorough investigation of the SLE behaviour, but
unfortunately they did not tabulate their data, and their work was not included in this
study. Table 16 reveals that parameter set A and B performs equally well for solid
liquid equilibrium calculation in this system and a very low deviation in the saturation
index (SI) is observed. The calculated deviation of SI may be examined by looking
data point 23-88 in figure 38.4 and 39.4. There is an amount of scatter, but the
correlation is generally good. The freezing point data are less scattered as shown by
data point 1-22. The remaining freezing point data above 55 % MEG are represented
by data point 89-129. These two last series show low scatter.
Model
-5
Experimental
MEG(s)
-25
Ice
Liquid
-35
-42.93
-45
MEGH2O(s)
Temperature, C
-15
-49.03
-55
-65
0
20
40
60
80
100
wt % MEG
Figure 42: SLE curve for the mixed solvent MEG-H2O system. Circles represent experimental values
obtained from references in table 16 and line calculated using the model and parameter set B.
Only scarce amount of equilibrium data are available for the ternary CO2-MEG-H2O
system as shown by table 14. Hayduk et al.45, Won et al.79, and Kobe and Mason80 all
tabulated the solubility in mol/L. The individual density measurements were used in
- 125 -
137
order to eliminate the volume dependence if possible. Kobe and Mason80 did not
report density and their data were not used since our density correlation was not
available at the time of fitting the data. One data point of Hayduk et al.45 seemed to be
an outlier and two other data points at high glycol concentration were not used. Most
of the scattered data at low temperature by Davis et al.78 were not used. Sandengens
data4 were not included as they were not available to us during the parameter
estimation. Binary CO2-MEG data96,97 were not included in the fitting process. Result
of the fitting is also shown in table 14. Parameter set B performs somewhat better than
set A. But still a relatively high deviation is observed for this kind of data. Figure 43
shows that the curvature corresponding to the experimental data is not reproduced by
the model. The overall temperature and pressure dependence is reproduced
satisfactory. The deviation may be related to the standard data of the carbonate
species or the volume and surface parameter of the CO2 compound. It was not
attempted to improve the CO2 parameters.
0.012
Model, 5 bar, 25 C
Model, 5 bar, 60 C
Model, 1 atm, 25 C
Exp, 5 bar, 25 C
Exp, 5 bar, 60 C
Exp, 1 atm, 25 C
mol CO2 /mol total
0.01
0.008
0.006
0.004
0.002
0
0.0
0.2
0.4
0.6
0.8
1.0
mol fraction MEG

Figure 43: CO2 solubility in the mixed solvent MEG-H2O system. The points represent experimental
data from references in table 14 and lines were calculated using the model and parameter set B.
Data for salt solubility in MEG are needed in order to fit the MEG- HCO3 and MEGCO32 interaction parameters. As the solubility is relatively low in these systems, data
for salt solubility in the ternary salt-MEG-H2O system are used instead. Quaternary
salt1-salt2-MEG-H2O solubility data were also included. As the model only contains
binary interaction parameters, only ternary data are strictly needed. Quaternary data
are needed if a solid precipitate which consist of salt1-salt2-MEG-H2O. It is not the
case for the systems in this work.
- 126 -
138
Oosterhof et al.5, Grtner et al.6, Dugstad98, Sandengen4, and Atakhodzhaev and

Dzhuraev99 have all published SLE data related to the Na2CO3-NaHCO3-MEG-H2O
system, see table 17. Atakhodzhaev and Dzhuraev 99 only show their experimental
NaHCO3-MEG-H2O results in graphs and their results were therefore not used in this
study. Dugstad98 lists a volume dependent solubility. Density is not given, and our
density model was not available at the time of fitting. The data were therefore not
usable for modelling. Sandengen4 reports a few data which were not used since they
were unavailable at the time of fitting. Oosterhof et al.5 did a very useful study on
Na2CO3 solubility in MEG-H2O. They made a very consistent analysis of the liquid
phase but did not analyse the solid phase. The missing information is of key
importance when reporting solubility data. This is especially true in the carbonate
system where several different hydrates may precipitate. Their study was fortunately
carried out at relatively high temperature where Na2CO3 and Na2CO3H2O are the
only possible solid phases. Grtner et al.6 did a comprehensive study on the Na2CO3NaHCO3 solubility in the mixed solvent. They briefly analysed the solubility of
NaHCO3 in the mixed solvent, but the main part of their data are quaternary solubility
data. The measurements were done in solutions with a high glycol content, >50 wt%
MEG (22 mol%).
The ternary data of Grtner et al.6 are comparable to our results and were used in this
study. None of their quaternary data were included, since Grtner et al.6 determined
solubility by a method that may be based on an inaccurate interpretation of the
titration experiments. The inaccuracy is in our experience in the order of 0.5 to 2.0
wt%. Besides, they list only the salt-free glycol concentration in the initial solutions
and assumed that the solvent composition did not change due to precipitation of salt
hydrates or evaporation of water at the high temperatures.
18
16
wt % NaHCO3
14
Model, 20 C
Exp, 20 C
Model, 35 C
Exp, 35 C
Model, 50 C
Exp, 50 C
Model, 80 C
Exp, 80 C
12
NaHCO3
10
8
6
4
2
0
0
20
40
60
80
100
MEG wt %, saltfree
Figure 44: Solubility of NaHCO3 in the mixed solvent MEG-H2O system. The points represent
experimental data from references in table 17 and model lines calculated using parameter set B.
- 127 -
139
The result of modelling the ternary SLE system NaHCO3-MEG-H2O is shown in

figure 44. The experimental data points are mainly from this study, but some are from
Grtner et al.6. The data for solubility of NaHCO3 in pure water were obtained from
the IVC-SEP electrolyte data bank and were only used in set B for fitting the standard
state properties. NaHCO3 is the only solid phase that precipitates in this system above
the freezing point. An unusual increase in the solubility of NaHCO3 is observed in
solutions with more than 85 wt% MEG. The phenomenon is unexpected and was
validated several times in this study. It was also confirmed by powder X-ray
diffraction, that the solid phase is pure NaHCO3 and not a complex or solid solution
formed by MEG and NaHCO3. Grtner et al.6 suggested that the solubility increase
may be explained by the presence of a liquid complex. It may rather be explained by
the increase in CO2 solubility at high MEG concentrations, shown in figure 43, and
the dissociation of bicarbonate to CO2. Comparing figure 43 and 44 shows that the
model predicts a too low CO2 solubility at high MEG concentrations and equivalently
a too low solubility of NaHCO3.
35
Model, 10C
Model, 15C
Model, 20C
Model, 25C
Model, 40C
Model, 80C
30
wt % Na2CO3
25
Exp, 10C
Exp, 15C
Exp, 20C
Exp, 40C
Exp, 80C
Na2CO37H2O
20
15
Na2CO3H2O
10
Na2CO310H2O
Na2CO3
0
0
20
40
60
80
100
MEG wt%, saltfree

Figure 45: Solubility of Na2CO3 in the mixed solvent MEG-H2O system. The points represent
experimental data from references in table 17 and model lines calculated using parameter set B. Thin
black: Na2CO310H2O. Thick grey: Na2CO37H2O. Thin grey: Na2CO3H2O. Thick black: Na2CO3.
The solubility of Na2CO3 in the mixed solvent system is shown in figure 45. Most of
the shown experimental values were produced in this work. Some of the data at 40 C
and 80 C are from Oosterhof et al.5. Na2CO3H2O and Na2CO3 precipitate at these
temperatures and the solubility decrease linearly with MEG wt%. Figure 45 shows
that the model predicts a high solubility compared to their experimental values at
- 128 -
140
these temperatures. Apparently Oosterhof et al.5 did not correct for solvent
composition change even though water precipitated in Na2CO3H2O. Therefore MEG
concentration is higher at equilibrium then at initial conditions. Their solubility data
of Na2CO3H2O is lower compared to our values since their MEG concentration in
reality was higher than expected. Figure 45 shows the solubility profile plotted as a
function of the initial MEG wt% on a salt-free basis. Na2CO310H2O precipitates at
low temperature and low MEG concentration followed by a small Na2CO37H2O
branch, a Na2CO3H2O branch and finally a Na2CO3 branch. The model predicts the
Na2CO37H2O precipitation and it was only supported by one experimental data point.
14
w/w total % NaHCO3
12
10
Model, 10C
Exp, 10C
Model, 20C
Exp, 20C
Model, 30C
Exp, 30C
Model, 40C
Exp, 40C
Model, 50C
Exp, 50C
8
6
3 wt % MEG
4
2
0
0
10
20
30
40
w/w total % Na2CO3

Figure 46: Experimental and calculated solubility of Na2CO3 and NaHCO3 in the mixed solvent
system MEG-H2O with 3 wt% MEG. Experimental data points were obtained in this study. Lines were
calculated using the model and parameter set B. Thin black lines: NaHCO3. Thick black:
Na2CO310H2O. Thin grey: trona. Thick grey: Na2CO3H2O.
Table 16: Used experimental freezing point depression data of MEG-H2O.
MEG interval
wt%
Author
Parameter set
Cordray et al.100
Ott et al.101
Baudot and
Odagescu102
Zinchenko and
Zinchenko95
T interval
C
# data used
(actual # data)
10.0 - 90.9
15.6 - 97.2
40.0 - 50.0
-49.9 - -3.6
-49.0 - -5.7
-36.2 - -22.4
33(34)
25(27)
5(5)
-50
0(0)
- 100
- 0
SI Deviation
%c
A
B
0.06a, 1.6 b
0.4a, 1.6 b
-0.7a, -0.6 b -0.3a, -0.6 b
-0.2
0.2
-
<55 wt% MEG. b >55 wt% MEG. c % deviation in the saturation index, SI, calculated as dev. % =
100 SI 1 nexp
- 129 -
141
Not all the data of Oosterhof et al.s5 were used in this study, since it is unclear what
the solid phases were. Most of the solid phases were determined by Grtner et al.6 but
still some phases are ambiguous at higher MEG concentration.
Parameter set A and B does not perform equally well for calculating Na2CO3
solubility. Figure 45 shows the predictions using parameter set B. Set A, not shown
here, does not perform as well at MEG concentrations >80 wt%. This is mainly
related to the improved thermodynamic properties of Na2CO3H2O in set B.
14
w/w total % NaHCO3
12
10
Model, 10C
Exp, 10C
Model, 20C
Exp, 20C
Model, 30C
Exp, 30C
Model, 40C
Exp, 40C
Model, 50C
Exp, 50C
8
6
15 wt % MEG
4
2
0
0
10
20
30
40
w/w total % Na2CO 3

Figure 47: Experimental and calculated solubility of Na2CO3 and NaHCO3 in the mixed solvent
system MEG-H2O with 15 wt% MEG. Experimental data points were obtained in this study. Lines were
calculated using the model and parameter set B. Thin black lines: NaHCO3. Thick black:
Na2CO310H2O. Thin grey: trona. Thick grey: Na2CO3H2O.
Table 17: Used experimental salt solubility data in Na2CO3-NaHCO3-MEG-H2O
T interval
C
Author
This study
Oosterhof et al.5
Grtner et al.6
Dugstad98
Sandengen4
Atakhodzhaev and Dzhuraev99
a
2
40
50
- 50
- 90
- 90
20, 124
4
18
# data used
(actual # data)
133(212)
46(58)
20(159)
0(3)
0(2)
0(0)
Deviation
%
A
B
-0.04a
-0.04a
-6.1b
-4.7b
b
0.6
-0.05b
-
deviation in ARD of calculated using mol Na/kg total. b % deviation in the saturation index calculated
as dev %
100 SI 1 nexp .
- 130 -
142
The predicted solubility in the quaternary Na2CO3-NaHCO3-MEG-H2O system is

depicted in figure 46, 47, and 48. Figure 46 gives the solubility in a solvent consisting
of 3 wt% (1 mol%) MEG on a salt-free basis. Figure 47 shows solubility in a solvent
consisting of 15 wt% (5 mol%) MEG, salt-free basis. Figure 48 shows the change in
the composition by addition of MEG for solutions saturated with two salts. Adding
MEG makes it a quaternary system and the ternary invariant points shown in figure 36
become univariant and therefore dependent on MEG concentration. Invariant points in
the quaternary system exist where four solid phases are in equilibrium, none are
shown for the specific MEG concentration in figure 48.
1
0.9
0.8
0.6
0.5
0.4
MEGH2O(s)
salt fraction
0.7
0.3
0.2
NaHCO 3
0.1
Model at 0% MEG
Model at 75% MEG
0
-50 -30 -10
10
30
50
70
90
110 130
Temperature, C
Figure 48: The influence of temperature and MEG concentration on the precipitation fields in the
Na2CO3-NaHCO3-MEG-H2O system. Description of areas given in figure 36. The ternary invariant
points move to lower temperatures by addition of MEG.
The deviation observed between experimental and calculated values in figure 46 and
47 is due to two reasons. The MEG concentration and temperature used in model
calculations are constant. But the experiments were performed at slightly different
temperatures, and the water MEG ratio was not constant due to precipitation of
hydrated salts. Table 7 shows the exact experimental temperatures and MEGconcentrations plotted in figure 46 and 47. Comparing the 20 C isotherm in figures
46 and 47, shows how trona (Na2CO3NaHCO32H2O) breaks through by the addition
of MEG. This is similar to the breakthrough of wegscheiderite at 50 wt% and 50 C to
90 C measured by Grtner et al.6. Additionally it can bee seen how the trona area
widens by increased temperature. The residuals of the SLE data are summarised in
figure 38.3, 38.4, 39.3, and 39.4. The data by Oosterhof et al.5 and Grtner et al.6 are
data number 23 to 88 and a lower scatter is seen when using parameter set B. The
model predicts the Na2CO3-NaHCO3-MEG-H2O solubility satisfactory. The results of
- 131 -
143
set A and B are similar at 3 wt% MEG but the result at 15 wt% MEG is slightly better
described by parameter set B. Table 17 compares the correlation of SLE data using
parameter set A and B. The performance of set B is better. Table 17 shows that not all
our experiments were used for modelling. Data in pure water were not used and a
collection of data was produced after the modelling to verify part of the system.
Figure 49: The calculated infinite dilution activity coefficient of MEG in water compared to literature
values of Suleiman and Eckert103.
Figure 48 shows a plot similar to figure 36, but for the mixed solvent system. The
two-salt invariant lines move towards lower temperatures by adding 75 wt% MEG. It
shows how glycol changes the water activity and displaces the precipitation fields of
this system The trona area has moved to the temperature range from -40 C to 70 C
instead of 20 C to 115 C. The wegscheiderite area has similarly moved to lower
temperatures and wegscheiderite may precipitate at 0 C. The trend has changed
above 80 C. Instead of a broad area of NaHCO3, a broad area of Na2CO3 is observed.
All the binary and ternary invariant points have moved 30 to 50 C down and the
solid phases Na2CO37H2O and Na2CO310H2O are no longer present. The same
phenomenon was observed in figure 44 and 45. At low temperature MEGH2O is
precipitating at approximately -44 C instead of ice at approximately 3 C. These
temperatures are lower than the freezing point of the MEG-H2O system, shown in
figure 42, since the solutions are saturated with Na2CO3 and NaHCO3. Comparing the
results of figure 46 and 47 to 48 shows that MEG has the tendency to widen the trona,
wegscheiderite, Na2CO3H2O, and the Na2CO3 fields, but to narrow the NaHCO3 field
at isothermal conditions. For example trona may precipitate at salt fractions 0.7 to
0.97 at 40 C in pure water whereas it has widened to 0.5 to 0.97 at 0 C in 75 wt%
MEG. Table 34 and 35 in appendix D.1 show the correlation matrix and the
confidence interval of the parameters fitted in parameter set A. Table 36 and 37 give
- 132 -
144
the same, for parameter set B. The tables show indications of correlation between
o
o
t
t
uMEG
, MEG ~ u MEG , H 2O and u MEG , MEG ~ u MEG , H 2O . These parameters are related to the VLE
data shown in figure 37 and freezing point depression data shown in figure 42. The
correlation indicates that the system may be slightly over parameterized. Table 35 and
o
o
37 shows on the other hand that uMEG
, MEG and u MEG , H 2O have a low confidence
t
t
interval. uMEG
, MEG and u MEG , H 2O could probably be fixed but were nonetheless fitted.
o
o
There is also a correlation between uMEG
~ uMEG
which is expected due to the
,CO 2
, HCO
3
speciation equilibria given by (201).

o
o
The uMEG
, MEG and u MEG ,CO2 parameters are related to the CO2-MEG-H2O equilibria
shown in figure 43. The two parameters show a minor correlation, possibly due to the
o
low amount of available data. The low confidence interval of uMEG
,CO2 shown in table
o
35 and 37 indicates that uMEG
,CO2 is necessary in the fitting process.
The infinite dilution activity coefficient of MEG in the binary MEG-H2O system is
shown in figure 49. The calculation is shown for both parameter set A and B. The sets
predict almost the same values. It is a coincidence that the two follow the same trend.
The calculations are carried out using the extended UNIQUAC model applying the
surface and volume parameters r and q, of MEG (M) and water (W):
ln J Mf
r q
r q
rM rM
1 5qM ln M w M w 1
rw rw
rw qM rw qM
u u
u uMM
qM 1 wM
exp ww Mw
T
T
ln
(248)
r and q are obtained from table 10. The interaction parameters of the MEG-H2O, are
calculated by the linear temperature relation:
uij
uijo uijt T 298.15
(249)
where uijo and uijt are given in table 12 and 13. Figure 49 shows the comparison of the
model to the work by Suleiman and Eckert103. Studying their work shows that the
experimental value of the infinite dilution coefficient were determined at a very low
precision. Therefore the calculated values are significantly different compared to the
experimental values.
Comparison of predicted and experimental solubility data of Grtner et al.6 are shown
in Figure 50 for 50 wt% MEG from 50 to 90 C. The solid lines represent the
calculated solubility using parameter set B. The coloured areas indicate the solubility
of a single salt. The interfaces between two areas correspond to the 2-salt solubility
points. The experimental data are offset compared to the predicted values. For
example the ternary solubility of NaHCO3 and Na2CO3H2O is experimentally higher
than the calculated values as shown by A which is also observed for the quaternary
solubility of trona as shown by B. The 2-salt interface between NaHCO3 and
Na2CO33NaHCO3 is offset as shown by C and similarly the Na2CO3NaHCO32H2ONa2CO3H2O interface, indicated by D.
- 133 -
145
7.7
Conclusion
An experimental study was conducted for the NaHCO3-Na2CO3-MEG-H2O system in

order to improve the experimental description of this system. The experimental work
showed that trona and the hydrates of Na2CO3 easily form meta-stable phases.
Ternary mixtures of Na2CO3-MEG-H2O and NaHCO3-MEG-H2O both show unusual
solubility profiles. The solubilities of the salts seem to go through local minima and
increase toward higher MEG concentrations.
A literature study has shown that titration alone may not be used for carbonate and
bicarbonate determination. It may only be used for determination of alkalinity and a
second method may support the titration. For this purpose the reverse Schreinemakers
method was derived in order to determine the composition of the saturated liquid from
little experimental information. The method has not yet been extended to solutions
saturated with more than one salt.
Based in part on the new experimental data, parameters in an activity coefficient
model were determined with the purpose to improve CO2 corrosion models, scale
prediction programs, or gas drying simulators in the mixed solvent electrolyte system
of CO2-NaHCO3-Na2CO3-MEG-H2O.
The extended UNIQUAC model was shown to be a suitable model for reproducing
and predicting a number of properties of the CO2-NaHCO3-Na2CO3-MEG-H2O and its
binary, and ternary sub-systems. This is done using only binary interaction parameters
with linear temperature dependence. For the particular system 751 SLE, VLE, and
excess enthalpy data points were fitted satisfactory using 12 ion specific interaction
parameters and two UNIQUAC volume and surface parameters for MEG. The model
predicts the heat capacity with good accuracy even though it was not used in the
parameter regression. The model predicts solubility in the quaternary system based on
parameters fitted to data for binary and ternary systems. It is expected that the model
predicts activity coefficients equally well in saturated and un-saturated solutions since
both types of data were used in the fitting process.
The activity coefficient model with the parameters determined in this work is not only
valid for modelling corrosion or similar processes, but may as well be used for CO2
capture, scale prediction, gas drying or similar processes involving CO2 and MEG.
A simple engineering correlation which reproduces the density of NaHCO3-Na2CO3MEG-H2O solutions was presented. The correlation reproduces the density data
within the experimental error using only one salt specific parameter for each salt in
the whole concentration and temperature interval studied.
There is a general lack of experimental data for CO2 solubility in MEG-H2O,
NaHCO3-H2O, and Na2CO3-H2O systems. These could improve the current modelling
of the system.
- 135 -
147
7.8
Literature cited
[1]
Schreinemakers, F.A.H. Graphical deductions from the solution isotherms of a double salt and
its components. Z. Physik. Chem., 1893, 11, 75-109.
[2]
Schreinemakers, F.A.H.; van Dorp, W.A.; Cocheret, D.H.; Filippo, H.; Waal, A.J.C.
Quaternary equilibrium systems. Z. Physik. Chem., 1907, 59, 641-669.
[3]
Kaasa, B. Prediction of pH, mineral precipitation and multiphase equilibria during oil
recovery. Ph.D. Thesis no. IUK 103, NTNU, 1998.
[4]
Sandengen, Kristian. Prediction of mineral scale formation in wet gas condensate pipelines and
in MEG (mono ethylene glycol) regeneration plants. PhD thesis, NTNU, Norway, 2006.
[5]
Oosterhof H.; Witkamp, G.J.; van Rosmalen, G.M. Some antisolvents for crystallisation of
sodium carbonate. Fluid Phase Equilibria, 1999, 155, 219-227.
[6]
Grtner, R.S.; Seckler, M.M.; Witkamp, G.-J. Solid Phases and Their Solubilities in the
System Na2CO3 + NaHCO3 + Ethylene Glycol + Water from (50 to 90)C. J. Chem. Eng.
Data, 2004, 49, 116-125.
[7]
Thomsen, K. Aqueous Electrolytes, model parameters and process simulation. Ph.D. Thesis,
IVC-SEP, Technical University of Denmark, 1997.
[8]
Thomsen, K; Rasmussen, P. Modeling of Vapor-liquid-solid equilibrium in gas-aqueous

electrolyte systems. Chemical Engineering Science, 1999, 54, 1787-1802.
[9]
Iliuta, M.; Thomsen, K.; Rasmussen, P. Extended UNIQUAC model for correlation and
prediction of vapor-liquid-solid equilibria in aqueous salt systems. Chemical Engineering
Science, 2000, 55, 2673-2686.
[10]
http://www.ivc-sep.kt.dtu.dk. IVC-SEP, Center for phase equilibria and separation processes,

Electrolyte database, 2007.
[11]
Mokbel, I.; Porcedda, S.; Guetachew, T.; Marongiu, B.; Jose, J. Static measurements of the
total vapor pressure of ethane-1,2-diol + water mixtures at temperatures from 258 K to 363 K.
The International Electronic Journal of Physico-Chemical Data, Eldata, 1999, 5(2), 79-84.
[12]
Wagman, D.D.; Evans, W.H.; Parker, V.B.; Schumm, R.H.; Halow, I.; Bailey, S.M. Journal of
Physical and Chemical Reference Data, Vol. 11, Suppl. No. 2: The NBS Tables of Chemical
Thermodynamic Properties. Selected Values for Inorganic and C1 and C2 Organic Substances
in SI Units. J. Phys. Chem. Ref. Data, 1982, 11 Supplement No. 2.
[13]
Thomsen, K.; Rasmussen, P.; Gani, R. Correlation and prediction of thermal properties and
phase behaviour for a class of electrolyte systems. J. Chem. Eng. Science, 1996, 51, 3675-83.
[14]
Hurst, J.E.; Harrison, B.K. Estimation Of Liquid And Solid Heat Capacities Using A Modified
Kopp's Rule. Chemical Engineering Communication, 1992, 112, p 21-30.
[15]
Oglesby, N.E. A Study Of The System Sodium Bicarbonate-Potassium Bicarbonate-Water. J.

Am. Chem. Soc., 1929, 51, 2352-2362.
[16]
Waldeck, W.F.; Lynn, G.; Hill, A.E. Aqueous solubility of salts at high temperatures II. The
ternary system Na2CO3-NaHCO3-H2O from 100 200. J. Am. Chem. Soc., 1934, 56, 43-47.
[17]
Hill, A.E.; Bacon, L.R. Ternary Systems VI. Sodium Carbonate, Sodium Bicarbonate and
Water. J. Am. Chem. Soc., 1927, 49, 2487-95.
[18]
Hill, A.; Smith, S. Equilibrium between the carbonates and bicarbonates of sodium and
potassium in aqueous solution at 25C. J. Am. Chem. Soc., 1929, 51, 1626-1636.
[19]
Hill, A. Double salt formation among the carbonates and bicarbonates of sodium and
potassium. J. Am. Chem. Soc., 1930, 52, 3813-17.
[20]
Green, S.J.; Frattali, F.J. The system Na2CO3-Na2SO4-NaOH-H2O at 100C. J. Am. Chem.
Soc., 1946, 68, 1789-1794.
- 136 -
148
[21]
Itkina, L.S.; Chaplygina, N.M. The solubility isotherm in the 2Li+, 2Na+. CO32-, 2OH- + H2O
system at 50C. Russ. J. Inorg. Chem., 1963, 8(6), 768-772.
[22]
Morozova, V. A.; Rzhechitskii, E. P. Solubility in The sodium fluoride-sodium bicarbonatewater, sodium fluoride-sodium sulfate-water, and sodium fluo. Russ. J. Inorg. Chem., 1977,
22, 485-486.
[23]
Ellingboe, J. L.; Runnels, J. H. Solubilities of Sodium Carbonate and Sodium Bicarbonate in

Acetone-Water and Methanol-Water Mixtures. J. Chem. Eng. Data, 1966, 11, 323-324.
[24]
Sedel'nikov, G.S.; Trofimovich, A.A. The reciprocal system 2K+, 2Na+, 2HCO3-, CO32-, H2O at
75C. Russ. J. Inorg. Chem., 1959, 4(6), 649-652.
[25]
Fedotieff, P.P.; Koltunoff, J. Another form of ammonia-soda process. Z. anorg. allg. Chemie,
1914, 85, 247-260.
[26]
Bogoyavlensky, P.S.; Manannikova, A.S. The NaCl-NaHCO3-H2O system (Carlsbad salt) at

25 and 38C. J. Applied Chemistry, USSR, 1955, 28, 225-28.
[27]
Ponizovskii, A.M.; Vladimirova, N.M.; Gordon-Yanovskii, F.A. Solubility in the Na Mg Cl

HCO3 - H2O system at 0C with carbon dioxide at 4 and 10 kg/cm2. Russ. J. Inorg. Chem.,
1960, 5, 1250-2.
[28]
Sukmanskaya, G.V.; Bogoyavlenskii, P.S. Solubility in the Sodium Sulfate - Sodium

Bicarbonate - Sodium Chloride - Water system. Russ. J. Inorg. Chem., 1960, 5, 468-71.
[29]
Trypuc, M.; Kielkowska, U. Solubility in the NH4HCO3 + NaHCO3 + H2O System. J. Chem.
Eng. Data, 1998, 43, 201-204.
[30]
Plyushchev, V.E.; Kurtova, L.V. The Li+,Na+//CO32-, NO3--H2O system. Russ. J. Inorg. Chem.,
1965, 10, 800-803.
[31]
Waldeck, W.F.; Lynn, G.; Hill, A.E. Aqueous solubility of salts at high temperatures. I
Solubility of Sodium Carbonate from 50 to 348C. J. Am. Chem. Soc., 1932, 54, 928-936.
[32]
Wells, R.C.; McAdam, D.J. Phase Relations Of The System: Sodium Carbonate And Water. J.
Am. Chem. Soc., 1907, 29, 721-727.
[33]
Morey, G.W.; Burlew, J.S. Studies of solubility in systems containing alkali and water IV. J.
Phys. Chem., 1964, 68, 1706-1712.
[34]
Freeth, F.A. The system Na2O-CO2-NaCl-H2O considered as two four component systems.
Phil.Trans.Roy.Soc.(London), ser A., 1922, 223, 35-87.
[35]
Itkina, L.S.; Kokhova, V.F. Solubility and composition of solid phases in the Na2SO4-Na2CO3NaOH - H2O system between 25 and 150C. Russ. J. Inorg. Chem., 1960, 5, 622-626.
[36]
Wegscheider, R.; Mehl, J. The system sodium carbonate-sodium bicarbonate-water and the
region of existence of trona, Na2CO3.NaHCO3.2H2O. Monatshefte Chem. sterreichische
Akademi der Wissenschaften, 1928, 49, 283-315.
[37]
Salit, P.W. A New Triple Acetate Method For Sodium Determinations In Biological Materials.
J. Biol. Chem., 1932, 96(3), 659.
[38]
Barber, H. H.; Kolthoff, I. M. A Specific Reagent For The Rapid Gravimetric Determination
of Sodium. J. Am. Chem. Soc., 1928, 50, 1625.
[39]
Ervin, G.J.; Giorgi, A.L.; McCarthy, C.E. The system Potassium Carbonate-Sodium
Carbonate-Water at 100 and 150C. J. Am. Chem. Soc., 1944, 66, 384-7.
[40]
Kremann, K.; Zitek, A. The Formation of Saltpeter from Sodium Nitrate and Potash from the
Standpoint of the Phase Rule. Monatshefte fr Chemie, 1909, 30, 311-340.
[41]
Osaka, J. On Sodium Potassium Carbonates. Mem. Coll. Sci. Eng. (Kyoto), 1911, 3, 55-61.
[42]
Skoog, D.A.; West, D. M.; Holler, F. J. Fundamentals of Analytical Chemistry, 7th edition,
Sounders college publishing, 1996.
[43]
Thomsen, K.; Iliuta, M.; Rasmussen, P. Extended UNIQUAC model for correlation and
prediction of vapor-liquid-liquid-solid equilibria in aqueous salt systems containing non-
- 137 -
149
electrolytes. Part B. Alcohol (Ethanol, Propanols, Butanols) - water - salt systems. Chemical
Engineering Science, 2004, 59, 3631-3647.
[44]
George, S. K. Density, thermal expansivity, and compressibility of liquid water from 0 to

150. Correlations and tables for atmospheric pressure and saturation reviewed and expressed
on 1968 temperature scale. J. Chem. Eng. Data, 1975, 20, 97-105.
[45]
Hayduk, W.; Malik, V. K. Density, Viscosity, and Carbon Dioxide solubility and diffusivity in
aqueous ethylene glycol solutions. J. Chem. Eng. Data, 1971, 16(2), 143-146.
[46]
Huot, J. Y.; Battistel, E.; Lumry, R.; Villeneuve, G.; Lavallee, J. F.; Anusiem, A.; Jolicoeur, C.
A Comprehensive Thermodynamic Investigation Of Water-Ethylene Glycol Mixtures At 5,25,
And 45C. Journal Of Solution Chemistrym, 1988, 17(7), 601-636.
[47]
Kapadi, U. R.; Hundiwale, D. G.; Patil, N. B.; Patil, P. R.; Lande, M. K. Densities, excess
molar volumes, viscosities of binary mixtures of ethanediol with water at various temperatures.
Journal Of The Indian Chemical Society, 2000, 77(7), 319-321.
[48]
Sakurai, M. Partial Molar Volumes Of Ethylene-Glycol And Water In Their Mixtures. Journal
Of Chemical And Engineering Data, 1991, 36(4), 424-427.
[49]
George, J.; Sastry, N. V. Partial excess molar volumes, partial excess isentropic
compressibilities and relative permittivities of water + ethane-1,2-diol derivative and water +
1,2-dimethoxyethane at different temperatures. Fluid Phase Equilibria, 2004, 216(2), 307-321.
[50]
Cocchi, M.; Marchetti, A.; Pigani, L.; Sanna, G.; Tassi, L.; Ulrici, A.; Vaccari, G.; Zanardi, C.
Density and volumetric properties of ethane-1,2-diol plus di-ethylen-glycol mixtures at
different temperatures. Fluid Phase Equilibria, 2000, 172(1), 93-104.
[51]
Douheret, G.; Pal, A.; Hoeiland, H.; Anowi, O.; Davis, M. I. Thermodynamic Properties Of
(Ethan-1,2-Diol + Water) At The Temperature 298.15-K .1. Molar Volumes, Isobaric Molar
Heat-Capacities, Ultrasonic Speeds, And Isentropic Functions. Journal Of Chemical
Thermodynamics, 1991, 23(6), 569-580.
[52]
Gallant, R. W. Physical properties of hydrocarbons. XIII. Ethylene glycols. Hydrocarbon

Processing (1966-2001), 1967, 46(4), 183-96.
[53]
Yang, C.; Ma, P.; Jing, F.; Tang, D. Excess Molar Volumes, Viscosities, and Heat Capacities
for the Mixtures of Ethylene Glycol + Water from 273.15 K to 353.15 K. Journal of Chemical
and Engineering Data, 2003, 48(4), 836-840.
[54]
Sandengen, K.; Kaasa, B. Estimation of Monoethylene Glycol (MEG) Content in Water +

MEG + NaCl + NaHCO3 Solutions. J. Chem. Eng. Data, 2006, 51 (2), 443 -447.
[55]
Garcia, A.V.; Thomsen, K.; Stenby, E.H. Prediction of mineral scale formation in geothermal
and oilfield operations using the extended UNIQUAC model Part II. Carbonate scaling
minerals. Geothermics, 2006, 35, 239-284.
[56]
Abrams, D.S.; Prausnitz, J.M. Statistical thermodynamics of liquid mixtures: A new

expression for the Gibbs excess energy of partly or complete miscible systems. AIChE Journal,
1975, 21, 116-128.
[57]
Bondi, A. Physical Properties of Molecular Crystals, Liquids and Glasses, 1968, Wiley. New
York,
[58]
Skjold-Jrgensen, S.; Kolbe, B., Gmehling, J., Rasmussen, P. Vapor-Liquid-Equilibria By

Unifac Group Contribution - Revision And Extension. Industrial & Engineering Chemistry
Process Design And Development, 1979, 18(4), 714-722.
[59]
Chiavone-Filho, O.; Proust, P.; Rasmussen, P. Vapor-Liquid-Equilibria For Glycol Ether Plus
Water-Systems. Journal Of Chemical And Engineering Data, 1993, 38(1), 128-131.
[60]
Chiavone-Filho O.; Rasmussen P. Modeling of salt solubilities in mixed solvents, Brazilian

Journal Of Chemical Engineering, 2000, 17(2), 117-131
[61]
Lancia, A.; Musmarra, D.; Pepe, F. Vapour liquid equilibria for mixtures ethylene glycol,
propylene glycol, and water between 98C and 122C. Journal of Chemical Engineering of
Japan, 1996, 29(3), 449-455.
- 138 -
150
[62]
Horstmann, S.; Gardeler, H.; Wilken, M.; Fischer, K.; Gmehling, J. Isothermal Vapor-Liquid
Equilibrium and Excess Enthalpy Data for the Binary Systems Water + 1,2-Ethanediol and
Propene + Acetophenone. Journal of Chemical and Engineering Data, 2004, 49(6), 1508-1511.
[63]
Trimble, H. M.; Potts, W. Glycol-water mixtures. Vapor pressure-boiling point-composition

relations. Ind. Eng. Chem., 1935, 27(1), 66-68.
[64]
Sokolov, N. M.; Tsygankova, L. N.; Zhavoronkov, N. M. Liquid-vapor phase equilibrium in

water-ethylene glycol and water-1,2-propylene glycol systems at various pressures. Teor. Osn.
Khim. Tekhnol., 1971, 5, 900-904.
[65]
Villamanan, M. A. Liquid-vapor equilibrium. 1,2-Ethanediol - water system. International

DATA Series, Selected Data on Mixtures, Series A: Thermodynamic Properties of NonReacting Binary Systems of Organic Substances, 2004, 32(3), 189.
[66]
Gonzalez, C.; Van Ness, H. C. Excess thermodynamic functions for ternary systems. 9. Totalpressure data and GE for water/ethylene glycol/ethanol at 50C. J. Chem. Eng. Data, 1983, 28,
410.
[67]
Villamanan, M.A.; Gonzalez, C.; van Ness, H.C. Excess thermodynamic properties for
water/ethylene glycol. J. Chem. Eng. Data, 1984, 29(4), 427-429.
[68]
Skripach, T. K.; Temkin, M. I. Studies of the equilibrium liquid-vapor in the systems: waterethylene glycol, water-diethylene glycol, ethylene glycol-diethylene glycol, diethylene glycoltriethylene glycol. Zhurnal Prikladnoi Khimii (Sankt-Peterburg, Russian Federation), 1946, 19,
180-6.
[69]
Kumar, P.; Sudhakar, K.; Rao, D. Phaneswara. A study of vapor-liquid equilibrium from
isobaric boiling-point data for the monoethylene glycol-diethylene glycol-water system. Indian
Chemical Engineer (1959-1993), 1982, 24(2), 14-16.
[70]
Efremova, S. A.; Komarova, L. F.; Garber, Yu. N.; Tikhanovich, E. V. Study of phase
equilibria in binary subsystems of the acetone-ethanol-water-methylcellosolve-ethylene glycol
system. Zhurnal Prikladnoi Khimii (Sankt-Peterburg, Russian Federation), 1988, 61(11), 257981.
[71]
Frolova, E. A.; Ershova, T. P.; Stepanova, V. A.; Pavlov, S. Yu. Separation of mixtures from
the synthesis of ethylene glycol dimethacrylate. Osnov. Organ. Sintez i Neftekhimiya,
(Leningrad), 1979, 11, 90-5.
[72]
Ramanujam, M.; Laddha, G. S. Vapour liquid equilibria for ethanol-water-ethylene glycol

system. Chemical Engineering Science, 12(1), 1960, 65-68.
[73]
Ogorodnikov, S. K.; Kogan, V. B.; Morozova, A. I. Liquid-vapor equilibrium in the system

ethylene glycol-water. Zh. Prikl. Khim., 1962, 35, 685.
[74]
Liu, F. Mathematical modelling and indentification in extractive distillation of alcohol-water

with glycol and diethylene glycol. J. east china. inst. chem. techn., 1993, 19(3), 254.
[75]
Kireev, V. A.; Popov, A. A. Equilibria in mixtures of liquids and solutions. I. Boiling points
and the composition of aqueous solutions of ethylene glycol and ethylene oxide. Zhurnal
Prikladnoi Khimii (Sankt-Peterburg, Russian Federation), 1934, 7, 489-94.
[76]
Liu, F.; Zhang, C.; Huang, F.; Zhang, C. Studies on separation of alcohols and water by
extractive distillation. Fuel Science & Technology International, 1993, 11(11), 1537-50.
[77]
Nath, A.; Bender, E. Isothermal Vapor Liquid Equilibria Of Binary And Ternary Mixtures
Containing Alcohol, Alkanolamine, And Water With A New Static Device. Journal Of
Chemical And Engineering Data, 1983, 28(4), 370-375.
[78]
Davis, P.M.; Clark, P.D.; Fitzpatrick, E.; Lesage, K.L.,; Svrcek, W.Y.; Satyro, M. The impact
of sulfur species on glycol dehydration - a study of the solubility of certain gases and gas
mixtures in glycol solutions at elevated pressures and temperatures. Gas Processors
Association, Research report 183, 2002.
[79]
Won, Y. S.; Chung, D. K.; Mills, A. F. Density, viscosity, surface tension, and carbon dioxide
solubility and diffusivity of methanol, ethanol, aqueous propanol, and aqueous ethylene glycol
at 25 C. Journal of Chemical and Engineering Data, 1981, 26(2), 140-1.
- 139 -
151
[80]
Kobe, K.A.; Mason, G.E. Aqueous Solutions of Alcohols as Confining Liquids for Gas
Analysis. Industrial and Engineering Chemistry, Analytical Edition, 1946, 18, 78 79.
[81]
Kracht, C.; Ulbig, P.; Schulz, S. Measurement and correlation of excess molar enthalpies for
(ethanediol, or 1,2-propanediol, or 1,2-butanediol plus water) at the temperatures (285.65,
298.15, 308.15, 323.15, and 338.15) K. Journal Of Chemical Thermodynamics, 1999, 31(9),
1113-1127.
[82]
Knnecke, H. G.; Steinert, H.; Leibnitz, E. Heats of mixing of binary systems on the example
benzene/water and water/glycols. Z. Phys. Chem. (leipzig), 1958, 208, 147-156.
[83]
Wang, S.; Luo, B.; Zheng, D. Study of the relationship between heat of mixing and vaporliquid equilibrium for binary aqueous solutions of glycols. Beijing Huagong Xueyuan Xuebao,
Ziran Kexueban, 1992, 19(3), 1-9.
[84]
Matsumoto, Y.; Touhara, H.; Nakanishi, K.; Watanabe, N. Molar excess enthalpies for water +
ethanediol, + 1,2-propanediol, and + 1,3-propanediol at 298.15 K. Journal of Chemical
Thermodynamics, 1977, 9(8), 801-5.
[85]
Dohnal, V.; Roux, A. H.; Hynek, V. Limiting partial molar excess enthalpies by flow
calorimetry: some organic solvents in water. Journal of Solution Chemistry, 1994, 23(8), 889900.
[86]
Arroyo, G. A. Excess enthalpy of binary mixtures of water with mono-, di-, tri-, and
tetraethylene glycols. Anales de Quimica, Serie A: Quimica Fisica e Ingenieria Quimica, 1988,
84(2), 179-82.
[87]
Biros, J.; Pouchly, J.; Zivny, A. A calorimetric investigation of interactions in aqueous

solutions of poly(oxyethylene). 1. Heats of mixing of oligomeric models. Makromolekulare
Chemie, 1987, 188(2), 379-94.
[88]
Rehm, K.; Bittrich, H.-J. Heats Of Mixing Of Binary Liquid Systems Using Thermometric
Titration. Z. Phys. Chem (leipzig), 1972, 251, 104.
[89]
Manin, N.G.; Antonova, O.A.; Korolev, V.P. Thermochemical study of the solvation of
ethylene glycol, glycerine and formamide in water-tert-butanol mixed solvent. Russian journal
of general chemistry, 1996, 66(2), 204-211.
[90]
Batov, D. V.; Antonova, O. A.; Egorova, N. G.; Korolev, V. P. Enthalpies of ethyleneglycol,

1,4-butyleneglycol, 2,3-butyleneglycol, glycerine and 2-ethoxyethanol solution in watermethanol mixed solvent. Russian journal of general chemistry, 1996, 66(8), 1331-1234.
[91]
Nakayama, H. Temperature dependence of the of solution of poly(ethylene glycol) and of

related compounds. Bull. Chem. Soc. jpn., 1970, 43(6), 1683.
[92]
Nichols, N.; Skjold, R.; Sprink, C.; Wadso, I. Thermochemistry of solutions of biochemical
model compounds 6. alpha,omega-dicarboxylic acids, -diamines, and -diols in aqueous
solution. J. Chem. Thermodyn., 1976, 8(10), 993-999.
[93]
Zhaodong, N.; Liu, B.P.; Zhicheng T. Calorimetric investigation of excess molar heat
capacities for water plus ethylene glycol from T=273.15 to T=373.15 K. Journal Of Chemical
Thermodynamics, 2002, 34 (6), 915-926.
[94]
Origlia-Luster, M.L.; Patterson, B.A.; Woolley, E.M. Apparent molar volumes and apparent
molar heat capacities of aqueous ethane-1,2-diol, propane-1,2-diol, and propane-1,3-diol at
temperatures from 278.15 K to 393.15 K and at the pressure 0.35 MPa. Journal Of Chemical
Thermodynamics, 2002, 34(4), 511-526.
[95]
Zinchenko, A. V.; Zinchenko, V. D. On phase transitions in the water-ethylene glycol system

at subzero temperatures under non-isothermal conditions. Cryo-Letters, 2001, 22(3), 191-198.
[96]
Jou, F.Y.; Deshmukh, R.D.; Otto, F.D.; Mather, A.E. Vapor-Liquid-Equilibria of H2S and CO2
and ethylene-glycol at elevated pressures. Chemical Engineering Communications, 1990, 87,
223-231.
[97]
Da-Qing, Z.; Wen-Dong, M.; Rei, W.; Tian-Min, G. Solubility study of methane, carbon
dioxide and nitrogen in ethylene glycol at elevated temperatures and pressures. Fluid Phase
Equilibria, Volume, 1999, 155(2), 277-286.
- 140 -
152
[98]
Dugstad, A.; Seiersten, M. pH-stabilisation, a Reliable Method for Corrosion Control of Wet
Gas Pipelines. SPE, 2004, paper 87560.
[99]
Atakhodzhaev, A. K.; Dzhuraev, B. Solubility sodium bicarbonate in mixed alcohol-water

solvents. Doklady Akademii Nauk USSR, 1988, (7), 39-40.
[100]
Cordray, D.R.; Kaplan, L.R.; Woyciesjes, P.M.; Kozak, T.F. Solid-liquid phase diagram for
ethylene glycol + water. Fluid Phase Equilibria, 1996, 117(1-2), 146-152.
[101]
Ott, J.B.; Goates, J.R.; Lamb, J.D. Solid-liquid phase equilibria in water + ethylene glycol.
J.Chem.Thermodynamics, 1972, 4, 123-126.
[102]
Baudot, A.; Odagescu, V. Thermal properties of ethylene glycol aqueous solutions.

Cryobiology, 2004, 48(3), 283-294.
[103]
Suleiman, D.; Eckert, C.A. Limiting Activity-Coefficients of Diols on Water by a Dew-Point

Technique. Journal Of Chemical And Engineering Data, 1994, 39(4), 692-696.
- 141 -
153
154
Review and recommended thermodynamic properties of FeCO3 corrosion films
8
8.1
Review and recommended thermodynamic

properties of FeCO3 corrosion films
Introduction
CO2 corrosion or sweet corrosion has been known for a long time and continues to be
a problem in the process industry. Properties of FeCO3 are important in order do
understand FeCO3 stability. The solubility constant of FeCO3 may for example be
used for calculating the solubility of FeCO3 and to show how FeCO3 precipitates with
respect to pH or temperature. It may also be used in kinetic models for calculating the
precipitation rate and thereby the scale build-up. These calculations require high
precision of the equilibrium constant.
FeCO3 is not only observed in CO2 corrosion systems, it is also found in other
systems which contain iron: The availability of aqueous Fe2+ is frequently controlled
by the solubility of FeCO3 and the iron concentration follows the saturation by
FeCO3. The calculated iron concentration is dependent on the thermodynamic
properties of FeCO3 and inaccurate properties will influence the calculated
concentrations. Using an erroneous solubility constant of FeCO3 may lead to false
conclusions. This is especially important in geological systems as shown by Jensen et
al.1,2. Iron also influences the production of crops and Schwab and Lindsay3 has
shown how FeCO3 control the chemistry in agriculture. Previously Flgge4 discussed
how to produce FeCO3 for human consumption. It is also related to ground water5,6,
aeration of ground water7,8 and natural waters9,10. The water transported in pipes
receives an amount of iron from the dissolved transport equipment which gives a
contribution to iron in tap water as examined by Tillmans and Klarmann11. FeCO3 has
an economical aspect since it occurs naturally as technical pure crystals and is used
for iron and steel production12,13. Consequently iron has leaked into ground waters
and it has become an environmental problem which also has been addressed14. The
prehistoric high temperature and pressure ore formation has also been examined by
French15, Weidner16, and Rosenberg17.
Recently authors have studied the crystals for a Mars evolution theory18, but also in
scenarios of fuel cells19. As an alternative to existing CO2 capture processes Palandri
and Kharaka20 have discussed FeCO3 in a process for underground sequestration of
CO2 from flue gas.
Since then many contributors have published and discussed the FeCO3 properties
used for corrosion modelling. Properties have been used for Pourbaix stability plots
by several authors21,22,7,23,14,24,25,26. Properties have also been used in kinetic models to
predict the rate of FeCO3 precipitation and dissolution. Johnson and Tomson27,28,29
and Hunnik et al.30 published two of the most cited models, other models have been
published11,31,32,33,34,14,35,36 even recently by Golubev et al. 37. The kinetic models
continue to be the focus of the corrosion literature. Some of the newer applications
were made by Dugstad38, Nesic et al.39,40, Sun et al.41, Sun and Nesic42, and
Gulbrandsen43.
Two reviews have previously been published by Grauer44 and Preis and Gamsjger45
which focused mainly on the solubility constant and the enthalpy of formation of
FeCO3. In this study a comprehensive collection of FeCO3 properties are reviewed.
This includes entropy, enthalpy of formation and Gibbs energy of formation, heat
- 143 -
155
capacity, aqueous solubility and solubility constants. Properties of Fe2+ have also
been collected since they are closely related to FeCO3. Works on thermal
decomposition of FeCO3 are presented and an overview of synthesis of FeCO3 is
given.
The focus of this study is mainly related to aqueous systems between 0 and 100 C
and pressures up to 50 bar CO2. A few high temperature salt systems are discussed in
order to show how low temperature properties may be determined from high
temperature measurements. Much literature has been published on using FeCO3
properties to predict other properties. In this study only works related to measured
FeCO3 properties has been included. An extensive review of entropy, enthalpy of
formation and Gibbs energy of formation, heat capacity, aqueous solubility and
solubility constant of FeCO3 is given. A consistent set of thermodynamic properties
for FeCO3 and relevant aqueous species is selected and recommended for use.
Speciation schemes for aqueous FeCO3 are reviewed and evaluated. Issues related to
supersaturation of FeCO3 are discussed. Works on the thermal decomposition of
FeCO3 are presented and an overview of measured solubility and synthesis of FeCO3
is given. Measurements of non-aqueous systems are not discussed here.
- 144 -
156
8.2
Standard state properties of FeCO3
8.2.1
Databases of standard state formation properties
Standard state properties of FeCO3 have been collected in several databases.

Table 18, 19 and 20 give an overview of the available properties of CO32 , Fe2 , and
FeCO3 respectively which will be discussed in the following subsections. The
database collections of thermodynamic properties are also discussed and require
special attention since they have been used in many studies. The databases by the
United States Geological Survey (USGS)46,47,48 and Nordstrom et al.49 were used in a
range of ground water and geological studies. The book by Latimer50 was used in the
past as a source of properties and was therefore included in several works. Also the
smaller iron database by Langmuir51 was typically used since it covers many iron
species. The collections by Barin et al.52 and Martell and Smith53 were used less and
the collections by Wagman et al.54, NIST55, and Stumm22 have rarely been used.
The first carbonate property database was compiled by Kelley and Anderson56. The
properties of FeCO3 were estimated but not measured. The result of their work has
been used by almost all later data collections and their assumptions have spread into
many of the later data collections, which is discussed below. This assessment is based
on all references to databases with standard thermodynamic properties for FeCO3
encountered in this study.
The data collections by CODATA57,58 related to Fe(II) properties and the book by
Baes and Mesmer59 related to hydroxide properties, are also used in the discussion of
FeCO3 properties below.
8.2.2
Standard state heat capacity of FeCO3
The standard state heat capacity, C p , was measured and tabulated by Anderson60 and
Robie et al.61. Kelley and Anderson56 gave an approximate linear correlation which
was estimated using the heat capacity of iron, oxygen and carbon valid at room
temperature. The correlation has apparently been adopted by many authors through
time. Some authors refer to Kelley62 as the origin, but the work originates from Kelley
and Anderson56 even though Kelley62 states it was derived by Anderson60. FeCO3 has
a Nel transition temperature between 30 K and 40 K which was not recognised by
Anderson60. The Nel transition temperature is a second order magnetic phase
transition which is observed as a spike in the heat capacity. This was determined by
Kalinkina63, Kostryukov and Kalinkina64, Kalinkina and Kostryukov65, and Robie et
al.61. Robie et al.61 measured the low temperature heat capacity more accurately and
refined the correlation. Figure 51 shows the two correlations compared to the
measured heat capacities. The values by Anderson60 were measured in the interval 54
K to 296 K and are therefore only plotted at low temperature. Kalinkina63 and
Anderson60 gave a Debye and Einstein correlation of the heat capacity, which may be
used for calculating the heat capacity at low temperature.
Table 20 shows the literature values of the standard state heat capacity of FeCO3 at 25
C. Only small variations are observed through time and a good agreement between
authors is found. A recommended value of the heat capacity at 298.15 K is
82.25 kJ molK obtained from the correlation given by Robie et al.61.
- 145 -
157
110
Kelley and Anderson
Robie
Robie (exp)
Anderson
105
Cp (J/(molK))
100
95
90
85
80
75
70
-50
50
100
150
200
250
Temperature (C)
Figure 51: The standard state heat capacity of FeCO3(s) as function of temperature.
8.2.3
Standard state entropy of FeCO3
Table 20 lists the standard state entropy of FeCO3, S. The entropy follows the third
law of thermodynamics which states that S o (T=0 C, 1bar)=0 J/(molK). It is
calculated by integrating the standard state heat capacity, C po , and the sum of heat of
phase changes, ' trs H o , by:
o
S 298.15
K
298.15 K
0K
C po T ,1bar
T
dT
trs
' trs H o
Ttrs
(250)
Anderson60 and Kostryukov and Kalinkina64 estimated the entropy from the measured
heat capacity. The value by Anderson60 has been used by many authors and the value
is still listed in the NIST55 database, see table 20. The value was revised by Robie66,
re-evaluating the work by Anderson60 using a rough correction. It is not likely that the
new value is more accurate then the Anderson60, but the value was used in the data
collections by Robie et al.47, Robie and Waldbaum46, and Helgeson et al.67, see table
20. French15 determined a new value by extrapolation from high temperature using
the heat capacity correlation by Kelley and Anderson56 even though the correlation is
not valid in the applied temperature interval. This explains the deviation as indicated
by French15. Robie et al.61 improved the heat capacity measurements as explained
above and estimated a more reliable entropy. Still it may be improved since the
measured FeCO3 crystals were not pure. A recommended value of the standard state
- 146 -
158

o
entropy of FeCO3 is S FeCO
=95.47J/(molK), obtained by Robie et al.61, which is
3
consistent with their measured value of the heat capacity.
o
The entropy listed in table 20 is the standard state entropy of FeCO3, S FeCO
is
3
o
o
significantly different from the entropy of formation, ' f S FeCO
. ' f S FeCO
is calculated
3
3
from the entropy of iron carbonate, crystalline iron, graphite, and gaseous oxygen. For
example, using the NIST55 and Robie et al.61 standard state entropies of:
Fe D s C s
27.28
5.74
3
2
3
2
O2 g
R FeCO3 s
205.138
95.47
J molK
(251)
gives ' f S o =-245.26J/(molK). Using the FeCO3 entropy of Anderson60 gives

o
' f S FeCO
=-247.8J/(molK), a 1 % difference. Kelley and Anderson56 actually used the
3
o
=-247.7J/(molK). This value was
value of Anderson60, but listed a value of ' f S FeCO
3
obtained by using other values of the standard state entropies of iron, graphite, and
o
oxygen. ' f S FeCO
thus depends not only on properties of FeCO3.
3
8.2.4
Fe2+ and CO32- standard state properties
The standard state properties of Fe2 and CO32 are often used in the calculation of
the FeCO3 properties. CO32 properties have remained constant over the years as
shown by table 18, but Fe2+ properties have varied as shown in table 19.
Table 18: Standard state properties of CO32- listed in databases related to FeCO3 at 298.15K and 1 bar.
Authors
Year
Robie and Waldbaum46

Wagman et al.68
Robie et al.47
NIST55
CODATA58
Robie and Hemingway48
1968
1969
1978
1982
1989
1995
'fH
(kJ/mol)
-677.14
-677.14
-677.14
-675.23
-675.2
'fG
(kJ/mol)
-527.9
-527.90
-527.9
-527.81
-527
S
(J/(molK))
-56.9
-56.9
-56.9
-50.0
-50
The controversy was discussed by Tremaine et al.69, Tremaine and LeBlanc70, and
Parker and Khodakovskii57. They showed that the discrepancies between the
o
determined ' f GFe
2 are related to the contamination by oxygen and hydrogen. The
value by Randall and Frandsen71, see table 19, is often used and was obtained in an
atmosphere contaminated by oxygen as discussed by Patrick and Thompson72.
Randall and Frandsen71 measured the standard state potential of iron, 'E o 0.4402V ,
o
o
and calculated the Gibbs energy of formation, ' f GFe
2 = nF 'E =-84.9kJ/mol.
o
In the oxygen free environment, two different ' f GFe
2 values have been obtained.
o
Patrick and Thompson72 obtained a higher value in vacuum, ' f GFe
2 =-78.9 kJ/mol,
o
and Larson et al.73 obtained a lower value in a hydrogen atmosphere, ' f GFe
2 =-91.2
55
kJ/mol, see table 19. The high value was accepted by NIST . The lower value was
- 147 -
159
accepted by Langmuir51 and has therefore been used in many studies. Johnson and
Bauman74 support the NIST55 value, but the value by Parker and Khodakovskii57 was
recently evaluated and accepted by Preis and Gamsjger45,75. Langmuir51, Preis and
Gamsjger45, and Gamsjger et al.76 indicated also that the value by NIST55 could be
improved. They showed that it could be inconsistent with the enthalpy of FeCO3,
even though it is physically more correct. This is presumed since it was measured in
vacuum and not contaminated by hydrogen of oxygen.
Table 19: Frequently used standard state properties of Fe2+ at 298.15K and 1 bar.
Author
Year
Randall and Frandsen71

Latimer50
Patrick and Thompson72
Larson et al.73
Langmuir51
Wagman et al.54
Robie et al.47
Tremaine and LeBlanc70
NIST55
Parker and Khodakovskii57
1932
1952
1953
1968
1968
1969
1969
1978
1980
1982
1995
1995
'fH
(kJ/mol)
-87.86
-92.5
-89.1
-89.1
-89.1
-90.0
-91.1
'fG
(kJ/mol)
-84.98
-84.94
-78.9
-84.94
-91.2
-91.2 a
-78.9
-78.87
-88.92
-78.9
-90.53
-90.0
S
(J/(molK))
-113.4
-107.1
-137.7
-138
-107.1 a
-137.7
-101.6
-107.1 a
Obtained without refitting from Larson et al.73
o
' f H Fe
2 remained reasonably constant between studies, see table 19, but it has been
o
determined in many different ways. For example, Larson et al.73 estimated S Fe
2 from
55
o
used the
the entropy of iron sulphate and from this calculated ' f H Fe
2 . NIST
o
o
o
and ' f GFe
opposite approach where S Fe
2 was calculated ' f H
2 . A relatively low
Fe 2
o
o
value of S Fe
determined by Larson et al.73 was recently accepted
2 was found. S
Fe2
and included in the work by Robie and Hemingway48.
o
o
has remained relatively constant, the
Since ' f GFe
2 is scattered and ' f H
Fe 2
o
calculated values of S Fe
2 are also scattered as shown in table 19.
It is not possible to recommend a reliable set of thermodynamic properties of Fe2+,

due to the observed scatter. The NIST55 or Parker and Khodakovskii57 values are
equally good from a physical point of view even though they are significantly
different.
8.2.5
Gibbs energy of formation of FeCO3
o
The standard state Gibbs energy of formation of FeCO3, ' f GFeCO
, has been
3
determined by several authors. Table 20 lists the available values and a noteworthy
interval between -697kJ/mol and -666kJ/mol is observed. We have found that all
o
from one of the
except Robie and Hemingway48 and Barin et al.52 calculated ' f GFeCO
3
following three reactions named sp (solubility product), psp (pressure solubility

product), and therm (thermal decomposition):
- 148 -
160
sp:
FeCO3 s R Fe 2 CO32
(252)
psp: FeCO3 s 2 H R Fe 2 CO2 g H 2O l
(253)
therm: 3FeCO3 s R Fe3O4 2CO2 g CO g
(254)
The equilibrium constant of reaction k is symbolized by Kk. It is related to the Gibbs

energy of reaction, ' r , k G o , the stochiometric coefficient, Q i , and the Gibbs energy of
formation, ' f Gio , of component i in reaction k by:
' r ,k G o
Q '
i
Gio
RT ln K k
(255)
i in k
'r,k is the change of any property related to reaction k. 'r,spH is for example the
enthalpy of reaction (252). The negative base-ten logarithm to the equilibrium
constant is -logKk=pKk, and written for reaction (252) is denoted by pKsp. Reaction
(252) is most frequently used in the open literature, but both of the above reactions
o
have been used for determining ' f GFeCO
. This is done using equation (255) which
3
written for reaction (252), (253), and (254) gives:
o
' f GFeCO
3
o
o
RT ln K sp ' f GCO
2 ' f G 2
Fe
(256)
o
' f GFeCO
3
o
o
o
RT ln K psp ' f GHo 2O ' f GCO
' f GFe
2 2 ' f G
H
2g
(257)
o
3' f GFeCO
3
o
o
o
RT ln K therm ' f GFe
2' f GCO
' f GCO
g
3O4
2g
(258)
The equilibrium constant of reaction (256) and (257) are determined at low
temperature in aqueous media while (258) is determined through decomposition of
FeCO3 at high temperature.
o
o
is a linear function of ' f GFe
Equation (256) and (257) demonstrates how ' f GFeCO
2 .
3
o
This is problematic, since ' f GFe
2 varies significantly between databases as shown by
o
o
which depends on the S Fe
section 8.2.4. The same is true for S FeCO
2 . It illustrates the
3
2+
importance of knowing which Fe properties were used in the calculation. This
problem was also noted by Preis and Gamsjger45.
The issue may be avoided by tabulating the equilibrium constant of reaction (252) or
o
, since they are independent of the Fe2+ properties.
(253) instead of ' f GFeCO
3
o
' f GFeCO
determined at high temperature is independent of the Fe2+ properties. It is
3
determined using equation (258) at approximately 1000K and extrapolated to

298.15K using the heat capacity of FeCO3. This is a drawback since the extrapolation
o
involves an amount of uncertainty. To demonstrate this, consider ' f H FeCO
3
determined by Holland77, see table 20. He used the heat capacity correlation of Robie
et al.61. The obtained enthalpy is low compared to other works, see table 20. It can be
o
by:
shown that this value is too low by calculating ' f GFeCO
3
- 149 -
161

o
' f GFeCO
3
o
o
' f H FeCO
T ' f S FeCO
3
3
(259)
Table 20: Standard state formation properties of FeCO3 at 298.15K and 1 bar.
'fH
(kJ/mol)
Authors
Year
Anderson60
Kelley and Anderson56
Latimer50
Wilcox and Bromley78
Kostryukov and
Kalinkina64
Robie66
Langmuir51
Wagman et al.54
Singer and Stumm6
French15
Barin et al.52
Helgeson et al.67
Robie et al.47
Reiterer et al.79,80
NIST55
Robie et al.61
1934
1935 -747.60(T)
1952 -744.8
1963 -748.9(xc)
1964
1965
1968
1969
1969
1970
1971
1973
1978
1978
1980
1982
1984
Stubina and Toguri81

Holland77
Chai and Navrotsky82
Preis and Gamsjger45
1989
1990
1994
1995
2002
'fG
(kJ/mol)
-673.75(T,xa)
-673.88
S
(J/(molK))
92.9(D)
92.9 b
92.9
Cp
(J/(molK))
83.35(x,D,M)
82.09(T)
96.1(M)
-743.965
-740.57
-775.71*f
-740.57(M)
-749.656(xk)
-736.985
-740.57
-753.22(xk)
-749(xf)
-761.18(xf)
-750.6(xf)
-755.9
-752.0
105.0(*,xd)
105.0
-673.749 e
-679.44(xa)
-666.72
-671.53
-697.05*
-768.26(xg,M)
-679.440h
-666.698(xa)
-669.02(xi,M,D)
-666.67(xa)
-680.03(xj)
92.9
95.47
82.13
82.44(M,D)
82.25(x,T)
-688.04(xl)
95.50
82.27(T)
-682.8m
-678.9
95.47
95.47
82.27(*,T,M)
92.9
82.13
74.1*
92.9(M)
105.0
105.0
82.08(T)
82.09(T)
* ~ Rough estimate due to scatter in data, extrapolation or guess, x ~ calculated, T ~ temperature

correlation given, D ~ experimental data given, M ~ properties at several temperatures given. a
2
calculated from equilibrium constant of FeCO3 and properties of Fe2+ and CO3 . b they calculated
D
' f S o =-247.7. c estimated by a group contribution method. d a corrected value of the S FeCO
by
3
Anderson60. e obtained directly from Kelley and Anderson56 without refitting. f determined from the
high temperature measurements. g the listed GFeCO3 value is not identical to ' f GFeCO3 . It is not
o
comparable to other ' f GFeCO3 values. h obtained directly from Langmuir51 without refitting. i
o
corrected to 25 C from 50 C. j calculated from equilibrium constant of FeCO3 and properties of

o
o
CO32 using Larson et al.73 properties of Fe2+. k calculated from ' f GFeCO
and ' f S FeCO3 . l
3
calculated using ' f H FeCO3 and ' f S FeCO3 . ' f S FeCO3 was calculated using Hollands77 own
o
properties, see text. The value is erroneous which is identified by calculating Ksp, see text.
- 150 -
162

o
Which is carried out using an entropy of formation of ' f S FeCO
=-245.32J/(molK),
3
77
calculated from Hollands own data. The obtained Gibbs energy is very low, see
table 20, which indicates that the measured enthalpy is similarly too low. This can be
o
to calculate the equilibrium constant of reaction
further illustrated by using ' f GFeCO
3
(252). Using the NIST55 properties of Fe2+ and CO32 , reveals pK sp =12.45 and
similarly using Robie and Hemingway48 ion properties reveals pK sp =14.25. Both
equilibrium constants, Ksp, are several orders of magnitude too low compared to
o
literature values, see table 21 and figure 52. French15 likewise determined ' f GFeCO
3
which was even lower than the value by Holland77, see table 20. This value is
inaccurate and the calculated Ksp would be four orders of magnitude different
compared to literature.
o
Robie and Hemingway48 also applied equation (259) while determining ' f GFeCO
.
3
They used the reaction enthalpy of FeCO3 dissolution of Nordstrom et al.49,
o
' r , sp H o =-10.4 kJ/mol, and the entropy of Robie et al.61 and obtained a ' f GFeCO
3
comparable to other works, see table 20. Unfortunately they obtained pK sp =11.53
which is very high compared to other works, see table 21 and figure 52.
o
o
o
Barin et al.52 determined a GFeCO
. It was calculated from ' f H FeCO
and S FeCO
using
3
3
3
o
o
o
was used instead of ' f S FeCO
and the GFeCO
value obtained is
equation (259). S FeCO
3
3
3
o
.
not identical to ' f GFeCO
3
o
is given during the final discussion.
A consistent value of ' f GFeCO
3
8.2.6
Enthalpy of formation of FeCO3
The standard state enthalpy of formation of FeCO3 is given in the literature and
determined by two different methods. Either calculated at 298.15K by
o
' f H FeCO
3
o
o
' f GFeCO
T ' f S FeCO
3
3
(260)
or extrapolated from high temperature to 298.15K. Barin et al.52 and Robie and
Hemingway48 are exceptions. They used other approaches as described above. The
variation in the enthalpy is shown in table 20, the value span -775.7 kJ/mol to -737.0
kJ/mol.
o
o
is calculated through equation (260) using ' f GFeCO
.
In the first approach ' f H FeCO
3
3
o
' f GFeCO
is determined from the equilibrium constant of reaction (252) or (253) using
3
o
o
' f GFe
2 through equation (256) or (257). ' f G 2 is uncertain due to the variation in
Fe
o
o
o
' f GFe
is uncertain. Approximately ' f H FeCO
=-740 kJ/mol
2 and therefore ' f H FeCO
3
3
72
o
as NIST55 does or -750
is obtained by using the ' f GFe
2 of Patrick and Thompson
kJ/mol if the value by Parker and Khodakovskii57 is used. Accordingly -745 kJ/mol is
71
o
. Small variations are related
obtained by using the ' f GFe
2 of Randall and Frandsen
to differences in the used entropy of FeCO3, given by either Anderson60, Robie66, or
Robie et al.61.
- 151 -
163
In the second approach French15, Stubina and Toguri81, Holland77, and Chai and
o
from high temperature down to 298.15K. The
Navrotsky 82 all extrapolated ' f H FeCO
3
variation is notable as shown in table 20, which indicates that the procedure is
difficult. It is valuable that the extrapolation procedure is independent of the variation
o
in ' f GFe
2 . A value in the lower range of the enthalpy interval is typically obtained.
Stubina and Toguri81 and Chai and Navrotsky 82 obtained an enthalpy of
o
approximately -750 kJ/mol similar to studies using the ' f GFe
2 of Parker and
57
Khodakovskii shown above. Either their value is more correct or the high
temperature experiments were polluted by hydrogen or oxygen as indicated by Patrick
and Thompson72. Preis and Gamsjger45 made a critical evaluation of the enthalpy and
obtained a similar value pointing out the problems shown here. They argued that the
enthalpy of reaction (253) is only consistent with the Parker and Khodakovskii57
enthalpy of Fe2+.
o
is difficult to determine since it varies with the
A recommended value of ' f H FeCO
3
o
. It must be pointed out
values used for the properties of Fe2+, similar to ' f GFeCO
3
o
to obtain
which properties of Fe2+ and CO32 were used in the calculation of ' f H FeCO
3
consistency.
8.2.7
Solution enthalpy of FeCO3, reaction (252)
o
o
The enthalpy of reaction (252) or (253) are linear function of ' f H FeCO
, ' f H CO
,
3
2
o
o
o
o
. ' f H Fe
vary only slightly between studies as
' f H Fe
2 and ' f H
2 and ' f H
CO 2
CO 2
3
o
as discussed. The enthalpy of solution
shown in table 18 and 19, contrary to ' f H FeCO
3
written for reaction (252) is:
' r , sp H o
o
o
o
' f H Fe
' f H FeCO
2 ' f H
CO 2
3
(261)
also ' r , sp H o varies between studies as shown in table 21 with values in the range
o
between -30.14 kJ/mol and -9.46 kJ/mol. It is important to point out that both ' f GFe
2
o
o
o
even though ' f GFe
and ' f H Fe
2 are used in the calculation of ' r , sp H
2 is not found
o
explicitly in equation (261). This is due to ' f GFe
2 is used in the calculation of
o
' f H FeCO
as explained previously.
3
o
o
and it seems that the
Most authors used consistent values of ' f GFe
2 and ' f H
Fe 2
o
variation in ' r , sp H o is not connected to the variation observed in ' f GFe
2 .
o
Some irregularities were introduced by for example Langmuir51 who used ' f H FeCO
3
from Robie and Waldbaum46 calculated on the basis of Randall and Frandsens71
51
73
o
o
' f GFe
used the ' f H Fe
. Table 19 shows that Randall
2 . Langmuir
2 of Larson et al.
and Frandsen71 and Larson et al.73 are not consistent.
- 152 -
164

o
A similar combination was used by Nordstrom et al.49 who used ' f H FeCO
of Robie et
3
73
o
, while Nordstrom et al.49 used the
al.61 who again used ' f GFe
2 from Larson et al.
55
o
' f H Fe
. Table 19 shows that the Larson et al.73 and NIST55 properties
2 from NIST
of Fe2+ are inconsistent. ' r , sp H o given by Nordstrom et al.49 has unfortunately been
implemented in some later studies e.g. Ptacek and Blowes12, Robie and Hemingway48,
and Marion et al.18.
o
o
of Robie and Waldbaum46 who used ' f GFe
Helgeson83 also used ' f H FeCO
2 of
3
50
o
and therefore
Randall and Frandsen71. Helgeson83 used the ' f H Fe
2 from Latimer
obtained a more consistent value since the Latimer50 value is compatible with Randall
and Frandsen71.
Greenberg and Tomson33, Johnson and Tomson27,28,29, and Preis and Gamsjger45 all
fitted the enthalpy and obtained this way internal consistency of their data.
Preis and Gamsjger45 showed the temperature dependence of reaction (253) and
concluded that the Fe2+ properties of Parker and Khodakovskii57 are better than those
given by NIST55. Preis and Gamsjger45 determined ' r , psp H o which correspond to
' r , sp H o =-13.2 kJ/mol. They also showed that using the NIST55 properties for Fe2+
results in a biased temperature dependence of FeCO3. This indicates that the solubility
experiments were either polluted by oxygen or that the Parker and Khodakovskii57
Fe2+ properties are more consistent than the NIST55 properties.
o
to obtain a consistent property
It is recommended to either fit ' r , sp H o or ' f H FeCO
3
o
o
and ' f H Fe
set, or to use properties of ' f H FeCO
2 from the same data collection
3
which are consistent.
8.2.8
Solubility constants of FeCO3
Reaction (252) and (253) define the two solubility constants of FeCO3 discussed in
the open literature. Table 21 and figure 52 gives an overview of the published values
and correlations of pK sp of reaction (252) while table 22 gives an overview of pK psp
of reaction (253). pK sp has primarily been examined in the literature. The 25 C
value range between 10.24 and 11.53 as observed in figure 52. The lower value gives
a solubility constant, Ksp, which is twenty times higher than Ksp calculated from the
upper value.
The equilibrium constants are much easier to compare between works compared to
standard state properties. The reason is that pK sp is independent of Fe2+ properties
which vary between studies. This gives a direct way to identify why some authors
obtained their standard state property values.
The equilibrium constant has still not been determined accurately due to few
misinterpretations which have accumulated in data collections through time as
discussed in the following sections.
- 153 -
165

Table 21: Literature, 'r,spH, speciation scheme, and pKsp, of reaction (252) at 25 C and 1 bar.
Authors
Year
Smith84
Tillmans and Klarmann11
Kelley and Anderson56
1918
1924
1935
Latimer50
Langmuir51
1952
1968
1969
'r,spH
(kJ/mol)
Speciation scheme
pKsp
No extra species
No extra species
No extra species
10.46 (30 C)
9.57 (18 C)
10.67 (xa),
10.50 (30 C)
10.68 b
10.67 b
10.55 (T, M)
-19.37
No extra species
No extra species
-25.20
FeOH , Fe OH 2 aq ,
Fe OH 3 , Fe OH 4

Helgeson83
Singer and Stumm6
Martell and Smith53
Bardy and Pr85
Serebrennikov86
1969
1970
1976
1976
1977
-21.05 c
-19.37 d
Robie et al.47
NIST55
Robie et al.61
Greenberg32
1978
1980
1982
1984
1986
-29.26(x)
2
10.69 (M,x) b
10.24 (M)
10.68 b
10.46 (M)
10.54 (x)
No extra species
No extra species
No extra species
FeHCO3 , FeCO3 aq
-25.67
No extra species
No extra species
FeOH
-22.8
FeHCO3 , FeCO3 aq ,
10.50 b
10.91 (xe)
10.50 (x) b
10.60
10.56 (xf)
FeOH , Fe OH 2 aq
Greenberg and Tomson33
1986
-22.9
FeHCO3 , FeCO3 aq ,
10.605(T,D,M),
10.58 (x)
FeOH , Fe OH 2 aq
Nordstrom et al.
49
1990
-10.38
Haarberg et al.87
Johnson and Tomson27,28,29
1990
1991
-30.14
Braun35
Ptacek and Reardon88
1991
1992
Ptacek13
1992
36
Greenberg and Tomson

Bruno et al.
24
1992
48
Robie and Hemingway

Garber et al.89
Grauer44
10.89, 10.45 g
No extra species
FeHCO3 , FeOH
No extra species
FeHCO3 , FeCO3 aq , FeOH
10.37 (x,T)
10.45 (T)
-9.46
FeHCO , FeCO3 aq , FeOH

3
FeCO3 aq , Fe CO3
1992
Ptacek and Blowes12
1994
-10.38 i
1995
1996
1999
i,j
-10.4
Silva et al.90
2002
aq
2
2
aq
FeHCO , FeCO3 aq , maybe

3
FeOH
2002

Unknown
Unknown
FeCO3 aq , Fe CO3
2002
2
-13.23(x)
FeCO3 aq , Fe CO3
2
2
FeCO3 aq maybe more
- 154 -
166
aq
10.99(x *,h,M,T)
11.06
11.0-11.2 (D)
10.78
10.8
11.03 (M)
11.53 (xk)
10.66 (x,T)
10.8 b,l
11.03,
10.43 (D)m
10.59 (x)
10.9 (*)
Authors
Year
Marion et al.18
Sun and Nesic91
'r,spH
(kJ/mol)
Speciation scheme
pKsp
2003
FeCO3 aq , FeOH
11.06 (Tn)
2004
Unknown
10.59 (x)
* ~ Rough estimate due to scatter in data, extrapolation or guess, x ~ calculated by regressed

correlation or similar, T ~ temperature correlation given, D ~ experimental data given, M ~ properties
at several temperatures given. a calculated from ' r , sp G
60.92 kJ/mol given in their work. b the
value was obtained from Kelley and Anderson56. c a value of ' r , sp S FeCO3 =-275.3 J/(molK) was given.
o
2
obtained from Latimer50. e calculated using NIST55 properties of Fe2+ and CO3 . f calculated by their
given correlation. g 10.89 is calculated consistently with the remaining speciation. 10.45 is an
evaluation of Smith84 measurements. h The value is calculated at a ionic strength of 0.1. i obtained from
Nordstrom et al.49. j a value of ' f S FeCO3 =-252.6 J/(molK) was given. k The value is calculated from
o
the given thermodynamic properties. l obtained from Bruno et al.24. m 11.53 of dry crystals and 10.43 of
wet crystals. n Temperature dependence obtained from ' sp H of Nordstrom et al.49.
o
- 155 -
167
Speciation Schemes
A speciation scheme refers to the number of liquid compounds assumed to be present

in a liquid phase. It is usually assumed that H2O(l), H+, OH-, Fe2+, CO2(aq), HCO3-,
and CO32- are present in a system saturated with FeCO3. In the following sections this
is classified as a basic speciation scheme. It could also be assumed that a liquid iron
complex is formed and the total iron is distributed between several liquid iron species
for example FeHCO3+ in addition to Fe2+. This will be known as an extended
speciation scheme.
Table 21 gives an overview of the various liquid iron species used in the calculation
of pK sp . Fe 2 is a compound in all calculation of pK sp and therefore not mentioned
in table 21. The difference between using a basic and an extended speciation scheme,
is the calculated amount of free Fe 2 ions. To illustrate this, take two similar
solutions of iron. Assume all iron is Fe 2 in one solution. In the second solution the
same amount of iron is distributed between Fe2 and FeCO3 aq . Evidently the
amount of free Fe 2 is reduced if FeCO3 aq is present. Accordingly the value of
K sp will decrease since the concentration of Fe 2 is reduced. The consequence is an
observed increase in the pK sp value. It is therefore important that the speciation

scheme is carefully chosen equivalent to the scheme for which pK sp was calculated.
In the following it is illustrated, how the choice of speciation schemes influence the
resulting value of pK sp .
Solubility constant of Reaction (252)
Smith84 was the first to determine the equilibrium constant of FeCO3. This was done
by measuring the total iron and CO2 content in the liquid phase at anaerobe conditions
and at various CO2 pressures. The value was determined by setting up the mass action
laws for the CO2-H2O system and reaction (252) assuming the solid phase is in its
standard state:
aH aHCO
K CO2 aq
aCO2 aH 2O
aH aCO2
K HCO
(262)
(263)
aFe2 aCO2
(264)
aHCO
K sp
Fe2 , HCO3 , and Fe HCO3 2 aq were assumed to be the species in the liquid
phase. Smith84 tabulated the total concentration of iron as Fe HCO3 2 aq . The

molality of Fe 2 , bFe2 , was calculated by:
- 156 -
168
D bFe HCO aq
bFe2
(265)
3 2
D is the degree of ionization. The above relation is an uncommon approach since the
total iron is expected to be bT , Fe = bFe bFe HCO aq = 1 D bFe HCO aq . This way
2
3 2
3 2
Smith assumed that the total amount of iron is (1+D) times higher than the measured
value bFe HCO3 aq . This indicates that some inconsistency was introduced.
84
Smith84 wrote an electroneutrality balance in order to calculate the bicarbonate

concentration. It was assumed that only Fe2 and HCO3 were present and therefore
2bFe2
bHCO
(266)
K sp was determined by inserting (262) and (263) in (264) to remove aH and aCO2
3
from the equation:

2
aHCO
a
Fe2
K HCO
K CO2 aq aCO2 aq aH 2O
K sp
(267)
Rearranging and substituting (265) and (266) in the above, while assuming solute
activities equal to molalities and a water activity of aH 2O 1 , gives:
D b
K{
Fe HCO3 2
K CO2 aq K sp
bCO2 aq
(268)
4 K HCO
3
K is defined by the left hand side of the equation. Smith84 assumed D | 0.9 and
determined K from experimental measurements of iron ( bFe HCO3 ) and CO2 ( bCO2 aq ).
2
He obtained
4.04 103 and K sp was determined from (268) by:

K sp
K HCO K
(269)
4 K CO2 aq
using a set of approximated K HCO and K CO2 aq values. Smith84 obtained

3
3.453 1011 at 30 C which is equivalent to pK sp 30o C =10.46. He stated that
K sp
the solubility at zero CO2 pressure would be CFe2 | K sp | 5.8 106 mol L . Taking
Smiths84 data, setting D
1 and bT , Fe = bFe HCO3
aq
results in pK sp =10.32 at 30 C.
Tillmans and Klarmann used the same approach using D 1 and arrived at
3
K 7.05 103 for their data at 18 C. They also used approximate values of K HCO
11
and K CO2 aq and calculated pK sp 18o C 9.57 . This value is very low compared to
other later works as shown in figure 52 which may be related to the applied values of
K HCO and K CO2 aq . The work by Tillmans and Klarmann11 has not been used in any
3
- 157 -
169
later works, but the experimental data by Smith84 has been reinterpreted numerous
times.
13
12.5
H
H
pKsp of FeCO3
12
11.5
F
H
11
10.5
A
F
D
G
I
C
F
E
B
A
G
A
B
10
C
D
E
F
G
9.5
0
10
20
30
40
50
Smith 18
Tillmans 24
Kelley 35
Latime 52
Langmuir 69
Langmuir 69
Helgeson 69
Singer 70
Bardy 76
Serebrennikov 77
Martell 81
Reiterer 81
Robie 84
Greenberg 86
Greenberg thesis 86
Greenberg thesis 86
Nordstrom 90
60
70
80
Haarberg 90
Braun 91
Braun 91
Johnson 91
Bruno 92
Greenberg 92
Greenberg 92
Ptacek 92
Robie 95
Garber 96
Grauer 99
Jensen 02
Preis 02
Silva 02
Marion 03
Sun 04
Gulbrandsen 07
90
100
Temperature (C)
Figure 52: Negative base-ten logarithm to the equilibrium constant, pKsp, of reaction (252) as function
of temperature determined by various authors in the literature. Lines are correlations, symbols are
single values.
Kelley and Anderson56 made an evaluation of the work by Smith84. This evaluation is
central in the literature, since it was and is still included in many data collections.
Kelley and Anderson56 used the measured 3 K 4.04 103 by Smith84 at 30 C using
a different set of K HCO and K CO2 aq and obtained pK sp 30o C =10.50 which
3
correspond to a Gibbs energy of reaction of ' r , sp G 30o C = 60.92 kJ mol . They
corrected Ksp to the temperature 25 C; even though the method is not described as
noted by Singer and Stumm6. Their conversion reveals a value of
pK sp 25o C 10.67 . The result may only be obtained by assuming
' r , sp G 30o C = ' sp G 25o C which is a rough approach. pK sp 25o C 10.67 has
unfortunately been adopted by many data collections: Latimer50, Helgeson83, Martell

and Smith53, Garber et al.89, Palmer and Eldik92, Hgfeldt93, Zhuk94, and Stumm and
Lee22. Even the pK sp 30o C 10.50 at 30 C was adopted by NIST55 and Robie et
al.47 as the value at 25 C. The correlation by Sun and Nesic91 seems to originate from
- 158 -
170
the values given by Helgeson83. This may be recognised by comparing the two works
in figure 52.
Singer and Stumm6 were aware of the discrepancies between works. They showed
how the biased pK sp has led to numerous wrong interpretations of the FeCO3
solubility and wrong conclusion of supersaturation also noted by Ghosh95. Singer and
Stumm6 conducted solubility measurements at four different temperatures between 17
and 30 C and the solubility constant was re-evaluated. Unfortunately no experimental
data were given. A temperature correction was employed by using the heat of solution
from Latimer50 and a value of pK sp 10.24 at 25 C was obtained. The equilibrium
constant is three times higher than the previous pK sp
10.67 by Kelley and
Anderson56.
Bardy and Pr85 conducted experiments in order to resolve some of the
disagreements. They performed solubility measurements of the pure Fe-CO2-H2O
system, but also in systems of additional ions. pK sp was estimated at zero ionic
strength similar to the work by Singer and Stumm6 using the activity coefficient
correlation of Davies96. They determined pK sp experimentally at 20 C and
calculated it at T=25 C using the temperature correction of Singer and Stumm6 and
found pK sp 10.46 . It is uncertain how high a precision they obtained from the
assumptions they applied.
Many additional authors have determined the equilibrium constant of FeCO3. There
are three trends in the literature. One group uses an extended speciation scheme.
Another group uses a basic speciation scheme similar to the works mentioned above.
Finally a group uses any speciation scheme but base the calculations on scattered
measurements and obtains an inaccurate high pKsp value. The groups are discussed
here.
Langmuir51 collected properties of liquid iron species from different FeCO3 related
works. An evaluation of Smith84 data was given even though the results were based
on Langmuirs51 own measurements. He found a relatively high, pK sp 10.55 , and
gave a temperature correlation using the vant Hoff law. The high value is probably
(2 n )
due to the included Fe OH n
species. He stated that the properties of
Fe OH 2 aq should be considered as tentative and the existence not adequately
documented.
Wersin et al.97 took another approach. They used the speciation scheme by Fouillac
and Criaud98, and added FeOH without refitting. The same approach was used by
Nordstrom et al.49 who included the Fouillac and Criauds98 properties of FeHCO3 ,
the Langmuir99 properties of FeCO3 aq , and FeOH properties of Baes and
Mesmer59. Nordstrom et al.49 chose to give two different values of pK sp . An
evaluation of Smith84 data, pK sp =10.45 , and a recalculation which was consistent
with the given speciation scheme, pK sp =10.89. This shows clearly the influence of
using a different speciation scheme and a significantly higher pK sp is obtained.
The work by Nordstrom et al.49 was later included in the computer models
WATEQ4F, MINTEQA2, and PHREEQC, and in the collection by Stumm and
- 159 -
171
Morgan100 which has since made their equilibrium constants widely used. Still, the
database has room for improvements.
Greenberg and Tomson36 applied the Nordstrom et al.49 speciation scheme using
activity coefficients correlation of Davies96 and a relatively high equilibrium constant
was found, pK sp 10.78 . Johnson and Tomson27,28,29 made a similar study on FeCO3.
The method is not described in detail but they determined a temperature correlation of
the equilibrium constant using less speciation compared to Greenberg and Tomson36.
Even with the amount of species included they found a low pK sp 10.45 .
Previously Greenberg and Thomson32,33 used another scheme and arrived at
approximately pK sp 10.6 . Their calculations are not clear and they used CaCO3
properties in their FeCO3 calculations. Ptacek et al.88,12,13 adopted the speciation
scheme from Nordstrom et al.49 and refitted the equilibrium constant of FeCO3 using
a Pitzer model and obtained approximately pK sp | 11 . Marion et al.18 reinterpreted
the data by Ptacek13. They arrived at more or less the same Ksp since they used a
similar speciation scheme and many of the Nordstrom et al.49 properties. It is
unfortunate that they are missing crucial parameters of the liquid iron species and
used parameters from MgCO3 assuming they were valid for FeCO3. Figure 52 shows
pKsp used by Marion et al.18 plotted together with pKsp of other authors. The
temperature dependence of the correlation by Marion et al.18 is significantly different
from other works. It should be seen as uncertain.
Jensen et al.1 made a critical evaluation of pK sp . They were aware of some of the
problems described here, even though parts of the Nordstrom et al.49 speciation
scheme was used. Two different equilibrium constants of FeCO3 were determined for
two types of measurements, a pK sp 10.32 and pK sp 10.68 were determined. The
Fouillac and Criaud98 equilibrium constant of FeCO3 aq and FeHCO3 was also
used and two new equilibrium constants, pK sp
10.43 and pK sp
11.03 , were
found. The two scenarios clearly show the influence of using a different speciation
scheme. The Fouillac and Criaud98 scheme produces more FeCO3 aq and FeHCO3
and resulted in a lower equilibrium constant of FeCO3 s . If a simple speciation
scheme had been used then evidently a smaller pK sp was found.

An alternative speciation scheme was suggested by Davies and Burstein101 in which
Fe CO3
2
2
aq was presumed to exist. They referred to the work by Sidgwick102 but
Sidgwick102 only mentioned Fe HCO3 2 aq . Fe CO3

Wieckowski et al.
103
24
2
2
aq was also taken up by

9
and later by Bruno et al. and Wersin who determined the
equilibrium constant of Fe CO3
2
2
aq . Bruno et al.24 gave a value of
pK sp
10.8 ,
which is relatively high due to the unusual speciation.

Grauer et al.44 made a summary of some of the FeCO3 related works and chose to
recommend the value of Bruno et al.24. The discussion is not elaborate and the value
is recommended from a rough weighing of available works.
Recently Fe CO3
2
2
aq was included in the work by Preis and Gamsjger45 who
made a critical study of FeCO3 properties and re-evaluated the data from Smith84,
Bruno et al.24, Reiterer et al.79,80, and Silva et al.90. They included a set of recalculated
- 160 -
172
equilibrium constants from Greenberg and Tomson36 and included almost the same
speciation scheme of Bruno et al.24 and arrived at pK sp 10.59 . This value is
somewhat intermediate but indicates that the pK sp may be lower if the Bruno et al.24
speciation was not used. Even though the work by Preis and Gamsjger45 should be
seen as consistent within the framework of their speciation scheme.
It is apparent from the overview given here that values of the equilibrium constants
depend on the speciation considered. It must be emphasized that equilibrium constants
from different studies should not be combined without a critical examination of the
assumptions.
Some authors have chosen to keep a basic speciation scheme applying a low amount
of iron species. One of the works was made by Haarberg et al.87 who reinterpreted the
experimental studies by Smith84, Bardy and Pr85, and Greenberg32 using a basic
speciation scheme and a Pitzer equation. Their result is consistent and they present a
correlation of the pK sp as function of temperature up to 200 C. The calculated value
at 25 C pK sp ,25qC
10.37 which is comparable to the equilibrium constant of Smith84
and Singer and Stumm6.

Robie et al.61 also re-evaluated the data by Smith84 together with a set of unpublished
data. The speciation by Langmuir51 was used and they obtained a pK sp 10.60 ,
which is similar to the work of Langmuir51, pK sp
10.55 .
Braun35 used a basic speciation scheme analogous to Haarberg et al.87. The obtained
equilibrium constant was determined at an ionic strength of 0.1 mol/kg H2O and they
found a pK sp 10.99 . This is a high value and may be explained by the applied ionic
strength but also by the high scatter in their data. Their pK sp was determined in the
temperature interval 30 to 80 C, but if the measurements are constrained between 40
and 60 C, then pK sp ,25C 10.4 is obtained. The same problems were introduced by
Silva et al.90. They conducted an experimental study at 25 C in NaCl solutions. Silva
et al.90 applied the equilibrium constants of Thurmond and Millero104 using a Pitzer
model by Millero et al.105,10. The Pitzer model was used with a speciation scheme
similar to Bruno et al.24, see table 21. The scheme is inconsistent with the original
Bruno et al.24 scheme since FeCO3 aq was included. Their equations, figures and
tables are in disagreement and the modelling should be seen as tentative. They arrive
at an equilibrium constant of pK sp 10.9 due to the speciation used, but also due to
the high scatter of their result. By removing two out of their nine data points used for
their linear pK sp correlation gives a pK sp 10.4 . Their result is uncertain and may be
improved.
Solubility constant of Reaction (253)
The solubility constant of FeCO3 represented by reaction (253) can be determined

explicitly, since the equilibrium constants of the CO2-H2O system are not required.
Instead it is obtained directly by the mass-action law:
K psp
aFe2 fCO2 aH 2O
aH2
- 161 -
173
(270)
This is done by measuring pH, iron content and CO2 pressure. Water is usually
assumed ideal and the partial pressure of CO2 is substituted for the fugacity of CO2.
Table 22 gives an overview of the authors who used this method. Reiterer et al.79,80
o
was determined and
were the first to use this approach for FeCO3. ' f GFeCO
3 ,50q C
o
converted to ' f GFeCO
=-669.02 kJ/mol. Jensen et al.1, Silva et al.90 and Bruno et
3 ,25q C
al.24 stated that Reiterer et al.79 found a value of pK sp 11.20 , but it is unclear how
they obtained this value. Reiterer et al.79 did not publish this value but gave a
pK psp 7.61 at 50 C and an ionic strength of one. Using the NIST55 standard
thermodynamic properties of Fe 2 aq and CO32 aq results in a pKpsp converted to
pK sp
10.9 . This value is high, which may be related to the high ionic strength and
the difference in the used properties of Fe 2 aq . Reiterer et al.79 did not correct the
calculated pKpsp using an activity coefficient model even though they stated so.
Table 22: Literature, 'r,pspH, speciation scheme, and pKpsp, of reaction (253) at 25 C and 1 bar.
Authors
Year
Wersin9
1980
1990
Bruno et al.
24
Silva et al.
45
2002
2002
Speciation scheme
pKpsp
No extra species
1992
Preis and Gamsjger

90
'r,pspH
(kJ/mol)
-17.3
FeCO3 aq , Fe CO3
2
FeCO3 aq , Fe CO3
2
FeCO3 aq , Fe CO3
2
aq
aq
aq
FeCO3 aq maybe more
-7.61 (50 C, I=1molal)

-7.69 (I=1M),
-7.35 (I=0M)
-7.59 (I=1M),
-7.65 (I=?)
-7.56
-6.95(*)
* ~ Rough estimate due to scatter in data, extrapolation or guess. I ~ Ionic strength.
Reiterer et al.79 used a basic speciation scheme, opposite Wersin9, Bruno et al.24, and
Preis and Gamsjger45 who all introduced Fe CO3
2
2
aq , see table 22. Wersin9 and
Bruno et al.24 determined pKpsp at an ionic strength of one and calculated the value at
zero ionic strength using the SIT106 model. Both authors obtained high values which
are comparable to the value of Reiterer et al.79. Bruno et al.24 obtained an unusual
result; the pKpsp is lower at an ionic strength of one compared to an ionic strength zero
which is opposite all other studies so far.
The work by Silva et al.90 must be considered as uncertain since their method and
speciation scheme is questionable as discussed previously.
8.2.9
Equilibrium constant of FeHCO3+, FeCO3(aq), and Fe(CO3)22-
The existence of the three species, FeHCO3 , FeCO3 aq , and Fe CO3
2
2
aq , has
been widely debated in the literature. Some authors decided to include them; others
have deliberately left them out of their speciation. Table 23, 24, and 25 give an
overview of the authors who determined the related properties of the following
reactions respectively:
FeCO3 aq R Fe2 aq CO32 aq
- 162 -
174
(271)
FeHCO3 aq R Fe 2 aq HCO3 aq
(272)
Fe CO3 2 aq R Fe 2 aq 2CO32 aq
(273)
2
Table 21 to 25 gives an overview of all speciation schemes used. All three species are
rarely included in combination. Usually only FeCO3 aq and FeHCO3 are used or
FeCO3 aq and Fe CO3
2
2
aq . Very few of their properties were determined
consistently, meaning that they were not found by the same authors by fitting to the
same data simultaneously.
It has previously been shown that CaHCO3 and MnHCO3 could exist as
compounds6,107. Larson108 was the first to discuss how FeHCO3 could explain
supersaturation. Later Serebrennikov86, Mattigod and Sposito109, and Fouillac and
Criaud98 based their FeHCO3 and FeCO3 aq speciation on correlations of other
works and extrapolated to the non investigated iron systems. Their results should be
seen as approximate.
Table 23: Properties related to FeCO3(aq), reaction (271).
Authors
Serebrennikov86
Year Speciation scheme

1977 FeHCO3 , FeCO3 aq
Mattigod and
Sposito109
1977
Turner et al.112
1981
pK
-4*
FeHCO3 , FeCO3 aq , Fe CO3
2
2
aq ,
-6.57(*,x)
FeOH , Fe OH 2 aq , Fe OH 3 , Fe OH 4

2
FeCO3 aq , FeOH , Fe OH 2 aq ,
-4.73
Fe OH 3 , Fe OH 4

2
FeHCO3 , FeCO3 aq
Fouillac and Criaud98
1984
Nordstrom et al.49
1990
-4.38 b
Bruno et al.24
1992
FeCO3 aq , Fe CO3
-5.5
Millero and Hawke105 1992
-4.73 a
2
2
aq
2
2
-5.45 c
aq ,
FeOH , Fe OH 2 aq
90
Silva et al.
2002
FeCO3 aq maybe more
Preis and
Gamsjger45
2002
FeCO3 aq , Fe CO3
2
2
-6.3(*)
aq
-5.3
~ Rough estimate due to scatter in data, extrapolation or guess. x ~ calculated. a ' f H FeCO aq =0.29
o
kJ/mol. b calculated by the oxalate correlation of Langmuir99. c obtained from Bruno et al.24.
Nordstrom et al.49 collected some of these properties. FeHCO3 properties originated

from Fouillac and Criaud98 who did not include FeOH . This is noteworthy, since
Nordstrom et al.49 included it. The equilibrium constant of FeOH was based on the
work by Baes and Mesmer59,110 and Sweeton and Baes111. They justified the existence

of FeOH , Fe OH 2 aq , and Fe OH 3 by the least square fitting of equilibrium
- 163 -
175
constants to the solubility data of Fe(OH)2 and Fe3O4. Nordstrom et al.49 included
only FeOH . A consistent model would include all three complexes of iron.
Nordstrom et al.49 also included the FeCO3 aq properties; they were not taken from
the work of Fouillac and Criaud98 as expected, similar to FeHCO3 , but from the
oxalate correlation of Langmuir99. Nordstrom et al.49 obtained some consistency by
fitting one of the equilibrium constants similar to Silva et al.90. Bruno et al.24 fitted all
equilibrium constants to their data. Turner et al.112 and Millero and Hawke105
collected properties without refitting. Millero and Hawke105 used the equilibrium
constants of Bruno et al.24 and extended the speciation scheme by FeHCO3 . This is
noteworthy since Bruno et al.24 and Wersin9 determined that FeHCO3 does not exist
as species, even though Wersin et al.97 used a speciation scheme which included this
compound.
Marion et al.18 stated that they re-estimated much of the thermodynamic properties of
FeCO3 aq and FeOH even though they originate from Nordstrom et al.49.
Table 24: Properties related to FeHCO3+(aq), reaction (272).
Authors
Serebrennikov86
Mattigod and
Sposito109

1977 FeCO3 aq , FeHCO3
1977 FeHCO , FeCO aq , Fe CO 2 aq ,
3
3
3
Fouillac and Criaud98
1984
Nordstrom et al.49
1990
Millero and Hawke105
1992
pK
-2*
-2.05(*,x)

2
FeCO3 aq , FeHCO3
-2.17 a
-2 b
2
2
-1.47(*,xc)
aq ,
FeOH , Fe OH 2 aq
*
~ Rough estimate due to scatter in data, extrapolation or guess. x ~ calculated. a
o
' f H FeHCO
=4.35
3
kJ/mol. b obtained from the rough estimate by Fouillac and Criaud98. c extrapolated from other system
from correlation of Ksp.
The question is whether FeHCO3 , FeCO3 aq , and Fe CO3
2
2
aq exist at all.
Singer and Stumm discussed this subject. They concluded that FeHCO3 does not
6
form and most likely not FeCO3 aq . The same was concluded by Johnson and
Bauman74. Davies and Burstein101 indicated on the other hand, that Fe CO3

2
2
86
exist, but not FeOH . This also contradicts Serebrennikov who stated that
FeHCO3 could account for up to 50% of the total iron and should therefore be
equivalent to the concentration of Fe 2 . In the speciation by Jensen2, FeHCO3
account for 26-41% and FeCO3 aq for 1-5%.
- 164 -
176
aq

Table 25: Properties related to Fe(CO3)22-(aq), reaction (273).
Authors
Mattigod and
Sposito109

1977 FeHCO , FeCO aq , Fe CO 2 aq ,
3
3
3
Bruno et al.24
1992
pK
-9.51(*,x)

Millero and Hawke105
1992
FeCO3 aq , Fe CO3
2
2
2
aq
-7.1
2
2
-7.17 a
aq ,
FeOH , Fe OH 2 aq
2002
FeCO3 aq , Fe CO3
2
2
-7.1 a
aq
~ Rough estimate due to scatter in data, extrapolation or guess. x ~ calculated. a obtained from Bruno
et al.24.
It is unfortunate that species which have low experimental basis have been included in
the highly cited works of for example Nordstrom et al.49. It is also notable that the
critical evaluation by Preis and Gamsjger45 chose to include Fe CO3
2
2
aq even
though it has a low experimental basis.
8.2.10 Supersaturation
The possible supersaturation of FeCO3 solutions has been widely debated in the
literature and continues to be so. There are two groups which discuss this. One group
used thermodynamic properties of FeCO3 where the calculated and the measured
solubilities were different. They concluded that the solutions were supersaturated.
Another group observed supersaturation due to improper handling of the liquid
samples.
The reason why the first group observed supersaturation is basically due to
inconsistency between equilibrium constants, speciation scheme and the measured
solubilities. A group of studies were unaware of this problem. Bruno et al.24 for
example stated that supersaturation was observed and refer to the existence of
Fe CO3
2
2
aq . Similarly Larson108 stated that supersaturation is related to the
formation of FeHCO3 . Emerson113 indicated that using an erroneous equilibrium

constant resulted in a wrong calculated solubility. Still it was discussed that
supersaturation could be related to FeHCO3 formation. Ghosh95 showed that using
the correct equilibrium constant gives saturation and not supersaturation. An apparent
supersaturation can be observed if an erroneous equilibrium constant of FeCO3 s is
used or if a constant which is inconsistent with the speciation scheme is used. The
authors tend to include new unusual compounds in order to explain an observed
supersaturation. This is also why the existence of FeHCO3 , FeCO3 aq , and
Fe CO3
2
2
aq have been suggested. The reason is often that their equilibrium
constants are fitted to experimental data in combination with other equilibrium

constants of lower quality.
The other group which has discussed supersaturation were e.g. Hem5 and Stumm and
Singer8. Their discussions were related to ground-waters, but the issues are also
- 165 -
177
applicable to other systems. The problems were mainly related to improper or poor
handling of the samples which could for example explain the supersaturation
observed by Postma114. Problems were basically due to measurements at atmospheric
laboratory condition, which did not resemble the reservoir conditions. Other problems
could be related to the used equipment. The equilibrium measurements discussed by
Gamsjger115 and Gamsjger and Knigsberger116 showed that vigorous stirring could
result in supersaturation due to carryover of solid to the measuring cells.
Ghosh et al.7 observed a 20 to 30 time supersaturation in their measurements, which
may be explained by the pK sp of FeCO3 at 25 C, used at 10 C. Jensen et al.1,2
observed continuous supersaturation, which could be related to leaking containers or
due to the added MnCl2 and CaCl2 salt.
Other authors were unaware of supersaturation and have inadvertently published
supersaturated results as equilibrium data. Tillmans and Klarmann11 and Klarmann117
state that a group of old works before 1900 did not recognise how FeCO3 may
supersaturate before equilibrium. The time to reach equilibrium was shown
experimentally by Mller et al.118,119. Equilibrium was attained within few hours at 50
atm CO2. Dugstad120 stated it could take 40 hours and Leybold121 stated that
equilibrium was not obtained within ten weeks at 1 atm CO2. Jensen et al.1 observed
even longer equilibrium times. Wajon et al.31 observed similar kinetics in CaCO3
solutions where equilibrium was not reached within six weeks.
In recent studies a soluble iron salt has been used to deliberately study the effect of
supersaturation e.g. Dugstad120 showed the importance of iron supersaturation on CO2
corrosion and the resulting corrosion protection and Dugstad and Drnen122 discussed
the influence of supersaturation on the morphology and structure of the corrosion
scales. Their studies are ongoing and recently Gulbrandsen43 discussed
supersaturation to prevent corrosion from acetic acid in iron solutions.
Often the observed Fe2+ supersaturation is not appreciated and usually it is sought
avoided. The above shows that supersaturation may be related to many different
issues during measurements and calculation of results. The experimental equipment
should be carefully setup in order to avoid erroneous measurements and the
equilibrium constants should only be used in the speciation scheme that they were
originally determined.
Other speciation schemes in the Fe-CO2-H2O system
A number of authors have applied speciation schemes for the Fe-CO2-H2O system
which are not typical. Table 26 gives an overview of some of the species used in the
various studies. Stumm and Lee22 gave a review of iron chemistry. They used the Ksp
of Kelley and Anderson56 in combination with hydroxide properties. The resulting
scheme is not accurate since hydroxides were not used by Kelley and Anderson56.
Palmer and Eldik92 followed the same approach but included the FeHCO3 complex
instead of hydroxides.
- 166 -
178
Table 26: Alternative speciation used for FeCO3 modelling.
Authors
Stumm and Lee22
Year
1960
Speciation scheme
Langmuir99
1979
FeHCO3 , FeCO3 aq
Palmer and Eldik92
1983
FeHCO3
Wersin et al.97
1989
Millero et al123
1995
2
2
King124
1998
FeOH , Fe OH 2 aq , Fe OH 3
aq ,
FeOH , Fe OH 2 aq
aq ,
FeCO3 OH , FeOH ,
aq ,
FeOH
Fe OH 2 aq
Jimenez-Lopez and
Romanek125
2004
2
2
A correlation between oxalate equilibrium constants were given by Langmuir99. The

data were not fitted consistently but the study is important since it was used in a
significant amount of speciation schemes, for example the one by Nordstrom et al.49.
Wersin et al.97 applied a speciation scheme almost equivalent to the later work by
Nordstrom et al.49. They decided to change this scheme in the study of Bruno et al.24.
Millero et al123, King124, and Jimenez-Lopez and Romanek125 used an extended
speciation scheme which included FeHCO3 , FeCO3 aq , and Fe CO3
2
2
aq . All
24
authors stated that the work was based on the speciation of Bruno et al. . This is
notable since Bruno et al.24 stated explicitly that the existence of FeHCO3 could not
be justified.
Other questionable assumptions have been made. For example Millero et al123
included a number of interaction parameters in their FeCO3 model which were fitted
to the copper system. King124 used the Pitzer model of Milllero and Hawke105
unmodified. He added more species without refitting the equilibrium constants and

the Pitzer parameters. It is stated that FeCO3 OH could be justified by the
modelling given by the following reaction:
Fe CO3 OH aq R Fe 2 OH CO32

He stated that the equilibrium constant, pK Fe CO

OH
(274)
, of the above reaction was
-9.97. The effect of the new species is very small and less than the scatter in the
experimental data. It is doubtful whether this compound exists. It is up for discussion
whether the model is consistent, since constants and parameters of different models
were used without refitting.
8.2.11 Comment on equilibrium constants of iron hydroxides

The iron hydroxide species, Fe OH n
2 n
FeHCO3 , FeCO3 aq , and Fe CO3
2
2
n=0..4, have been discussed equivalent to
aq .
- 167 -
179
Sweeton and Baes111 studied the hydrolysis of iron which has been re-evaluated by
Baes and Mesmer59,110. The properties were justified by fitting equilibrium constants
to solubility measurements, and activity models were not applied. Table 27 gives an
overview of the speciation used by the various authors. It shows how Baes and
Mesmer changed the speciation scheme between studies. Baes and Mesmer110 stated
2
specifically that Fe OH 4 could not be justified. The studies gave significantly
different results. Equilibrium constants are 10 orders of magnitude different, and the
speciation schemes and thermal properties vary significantly. The constants by Baes
and Mesmer59 are widely used since one was included in the work by Nordstrom et
al.49 and NIST55. But these works only included the properties of FeOH and not the
remaining three species.
The properties determined by Leussing and Kolthoff126 were used in the collection by
Langmuir51 and have therefore also been used by a considerable number of authors.
More recently Shock et al.127 gave an alternative speciation scheme which included
FeO aq and HFeO2 . It has been shown that FeO is not stable at lower
temperatures128 and there are some disagreements between works. The same was
suggested previously by Gayer and Woontner129, who determined the solubility of

2
Fe(OH)2(s) in NaOH solutions. Their data indicate that Fe OH 3 and Fe OH 4 are
more likely to be HFeO2 and FeO22 which was also mentioned by Tremaine et al.69
Table 27: Speciation used by a selection of ferrous hydroxide studies.
Authors
Gayer and Woontner129
Year
1956
Speciation scheme
Sweeton and Baes111
1970
Baes and Mesmer59
1976
Tremaine et al.69
1977
FeOH , HFeO2 , FeO22
Tremaine and LeBlanc70
1980
Baes and Mesmer110
1981
Shock et al.
127
1997
FeOH , Fe OH 2 aq , HFeO2 , FeO22

2

FeOH , FeO aq , HFeO2
8.3
Experimental literature data on FeCO3
8.3.1
Aqueous solubility
The solubility of FeCO3 has been determined a number of times. Table 28

summarises the works and outlines the used temperature, pressure, and ions. The ionic
strength of the solution is often set deliberately. This is done in order to get a low
liquid junction potential while measuring pH or to simplify the equilibrium constant
calculations by fixing the ionic strength for the activity coefficient models.
Frequently authors6,85,79,80,24 assume the ions to be independent of the solvent medium.
This indicates that there is no distinction between FeCO3 solubility in for example the
NaCl-H2O medium or in the KNO3-H2O medium. In reality activity coefficient are
- 168 -
180
different between the systems and ionic interaction are present. It is not advisable to
set the ionic strength of the system, since the error related to the activity coefficient
calculation can be significantly lowered by looking at the solubilities in the pure
systems instead.
Recently some authors24,1,90 applied more advanced methods to calculate activity
coefficients which included the interaction between ions. Usually interaction
parameters are not available, and occasionally these are used across systems13. This
method is inconsistent and should be avoided.
Oxygen contamination was carefully eliminated by most authors and it should be,
since oxygen rapidly oxidises Fe2+ to Fe3+. Often this is done by purging the solutions
using N2 or CO2.
The works shown in table 28 may be divided into two groups: Data that are either
suitable or not suitable for thermodynamic modelling. Data that are not suitable for
such use may be due to missing composition, missing temperature, missing
information on CO2 pressure, or ions were added to set the ionic strength.
Incomplete datasets were given by Langmuir51, Singer and Stumm6, Davies and
Burstein101, Braun35, and Hunnik et al.30. Most of the data were shown in figures, but
measured values were not tabulated. Similarly Robie et al.61 used data which were
never published.
In the works by Ehlert and Hempel130, Leybold121, Baylis131, and Davies and
Burstein101 the equilibrium temperature was not measured. Lan and Jia132 and Jensen
et al.1 did not focus on temperature control and their data are given with an accuracy
of r2.5 C.
The works by Gould133, Leybold121, and Haehnel134 are subject to some uncertainty
and it is unclear if equilibrium was obtained. A kinetic study was performed by
Johnson and Tomson27,28,29. The solubility was measured while changing the
temperature and the temperature was kept constant 48h before and a few hours after
the experiment. It is unclear if equilibrium was obtained within this time. Similarly
Klarmann117, discusses some of the previous works before 1900, which have not been
included here, since it is also unclear if equilibrium was obtained.
Haehnel134 and Dugstad120 gave small datasets on FeCO3 solubility but unfortunately
listed the concentration in an undefined unit. The works by Brauns135, Lyden136 and
Teodorovich137 listed solubility of minerals but the salts were taken from nature and
were not pure.
The CO2 pressure has a significant influence on the FeCO3 solubility. It affects the pH
and the carbonate content in the liquid phase. Gould133 and Leybold121 used a high
CO2 pressure but neglected to measure the value, which is problematic. The same
problem was introduced by Bardy and Pr85. They equilibrated CO2 in oxygen free
water at 20 C and used the enriched CO2 water for their FeCO3 solubility
experiments. The pressure was not measured, but the total carbon content was
determined. Their data table lists the carbon content, but the unit is unfortunately
missing.
Wells138, Konz139, and Lan and Jia132 did not report the CO2 content or pressure. The
same was the case for Braun35 who flushed the solutions by nitrogen; it is unclear if
CO2 evaporated. Johnson and Tomson27,28,29 and Jensen et al.1 used containers
completely filled with liquid and used small gas phases. The CO2 content was low and
equal to the CO2 pressure of aqueous FeCO3 solutions.
Ions were used in some works to set the ionic strength of the solutions. Singer and
Stumm6, Reiterer et al.79,80, and Wersin9 used perchlorate to set the ionic strength
- 169 -
181
since it was believed not to form aqueous iron complexes. Bruno et al.24 published the
works of Wersin9 and the two datasets are identical. Similarly Bardy and Pr85 and
Silva et al.90 used other ions to set the ionic strength. Ehlert and Hempel130, Konz139,
Babcan140, Ptacek13, and Jensen et al.1 determined the solubility to show the effect of
different ions. Braun35 and Lan and Jia132 added various pH buffers to set the pH. This
complicates matters significantly, since the buffers contain several salts and the
interactions are unknown. The work by Silva et al.90 show the effect of pH, from their
description it is unclear how they changed pH and exactly which ions were added to
their mixtures.
Table 28: List of experimental literature data on aqueous FeCO3 solubility.
Authors
Gnzburg 141
Ehlert and Hempel130
Year
1911
1912
Wells138
Smith84
Haehnel134 b,c
Leybold121 b
Tillmans and Klarmann11
Baylis131
Mller and Henecka119
Mller and Harant118
Konz139
Gould133 b
Langmuir51 d
Singer and Stumm6 d
Babcan140
Bardy and Pr85 c
1915
1918
1924
1924
1924
1926
1929
1935
1966
1968
1969
1970
1974
1976
Reiterer and Gamsjger79,80

Davies and Burstein101
Johnson and Tomson27,28,29 b,d
Braun35 d
Dugstad120 c
Bruno et al.24 and Wersin9
Ptacek13
1980
1980
1991
1991
1992
1992
1992
Greenberg and Tomson36

Andersen and Valle142
Hunnik et al.30
Lan and Jia132
Jensen et al.1
Silva et al.90
1992
1993
1996
1996
2002
2002
Added salts
None
None or Na2SO4, NaCl,
MgCl2, MgSO4
None, sulphates
None
None
Tab water
None
None or NaOH and Ca(OH)2
None
None
K2CO3, KHCO3
None
NA
NaHCO3 and HClO4
FeCl2, MgCO3, CaCO3, NaCl
None or NaClO4, NaCl,
KNO3, Na2SO4, MgSO4,
CaSO4, CaCl2, MgCl2
NaClO4 + HClO4
KHCO3
None
pH Buffers (I=0.1)
None
NaClO4
None or NaCl, NaHCO3 or
Na2SO4 or FeSO4
None
NA
NA
pH Buffers
CaCl2, MnCl2, NaHCO3
NaCl, NaHCO3
CO2 Pressure
1 atm
none and 2 atm
Temperature
25 C
NA
NA
15 C
30 C
18 C
NA
18 C
NA
20 C
40,60,80 C
110 C
23.9 C
NA
22.5 C
25 to 200 C
20 C
1 and 56 atm
NA (high)
a
c
50 atm
50 atm
NA
NA (high)
NA
a
2 to 22 atm
NA
(higher than 0)
0.084, 0.84 bar
NA
None (low or 0)
None (low or 0)
1bar
0.01, 0.05 atm
0.0001 to 0.003
atma
2.5-3.0 bar
0.2-3.0 bar
1-5.6 bar
NA
None (low or 0)
0.05 bar CO2
NA ~ Not stated. a CO2 content in liquid phase determined. b Unclear if equilibrium was attained. c
missing unit on a measured value. d no experimental data tabulated.
Another group of authors gave complete dataset and chose not to adjust the ionic
strength. Smith84 and Tillmans and Klarmann11 measured the FeCO3 solubility and
- 170 -
182
50 C
NA
35 C, 79 C
30 to 80 C
15 to 72 C
25 C
25 C
25 to 94 C
20 to 80 C
40 to 106 C
22.5Cr2.5C
15 Cr1.5 C
25 C
determined the total carbon content in the liquid phase. Gnzburg 141, Haehnel134,
Mller and Henecka119, Mller and Harant118, Dugstad120, Greenberg and Tomson36,
and Andersen and Valle142 measured the solubility and gave the CO2 pressure. The
measurements by Andersen and Valle142 were briefly described and no experimental
details were revealed. Ptacek13 is the only author who determined both CO2 pressure
and the total carbon content in the liquid phase. The main part of Ptaceks13 work is
focused on mixtures of many salts, but seven data points were measured in pure water
or the mixture of NaHCO3-H2O. A larger number of her data points were measured in
the NaHCO3-NaCl-H2O salt mixture.
Mller and Henecka119 performed their experiments in a broad temperature and
pressure interval, but chose only to give the solubility at 20 C and 50 atm. The work
was continued by Mller and Harant118. The two works are comparable. Both works
show a maximum in the solubility at approximately 60 C, at 50 atm CO2. Mller and
Henecka119 show how the solubility increases a factor 2 over the pressure range 20 to
50 atm CO2. Mller and Henecka119 made also a number of interesting experiments in
the CO2-FeCO3-NaOH-H2O system. Which show how FeCO3 is partly converted to
Fe(OH)2.
8.3.2
Solid phases
The majority of the authors listed in table 28 determined the purity of FeCO3 by Xray diffraction before the solubility experiments. Only few authors took the effort to
determine the precipitated solid phase after the solubility experiment. The two
measurement do not necessarily give the same result and FeCO3 may precipitate as
Fe(HCO3)2 or as a hydrate of FeCO3.
Singer and Stumm6 and Reiterer80 determined that crystalline FeCO3 precipitated
during their experiment. Jensen et al.1 determined that several carbonate phases were
in equilibrium with their solutions. This indicates the measurements by Jensen et al.1
are in fact the 2-salt solubility of co-precipitated FeCO3 and CaCO3 and not the
solubility of FeCO3. This is important, since the solubility phenomena are different in
the two cases. Berg and Buzdov143 show that FeCO3 does not form hydrates at 25 C,
but small inclusion of water may be expected. Reynolds144 findings show that
FeCO3K2CO34H2O may be formed. This indicates that maybe other double
carbonate salts of iron may form, for example double salts of Na2CO3 or similar.
Rosenberg17, Rosenberg and Foit145, and Chai and Navrotsky146 states that the double
salt ankerite, CaCO3FeCO3, may not form at low temperature but it will form solid
solutions at high temperatures. Some indications show that it may form at lower
temperatures147. Ghosh et al.7 states that Fe(OH)2FeCO3 may precipitate even though
it is not experimentally proved. Other authors have recently made thorough
investigations and determined experimentally that it exists148,149,26,150,151,152,153. AlHassan et al.149 even stated that the precipitation of this solid phase could explain the
unexpected CO2 corrosion behaviour above 60 C. Xia et al.154 have also shown
experimentally that Fe(HCO3)2(s) may form as a intermediate meta-stable phase
before it converts to the final FeCO3 corrosion product.
8.3.3
Specific heat capacity of FeCO3
The heat capacity of FeCO3 was measured by Anderson60 between 54K and 296K.
The purity of the used FeCO3 was 88% and the impurities were CaCO3, MnCO3, and
- 171 -
183
MgCO3. Anderson60 corrected the result for impurity, but the method was not
described.
Robie et al.61 measured the value more accurately between 5 K and 373 K; still the
crystal had 4.4% impurities of MnCO3. The measurements were improved at low
temperature compared to the work by Anderson60.
The heat capacity was also measured by Kalinkina63, Kostryukov and Kalinkina64,
and Kalinkina and Kostryukov65. Kalinkina64 stated that the used FeCO3 were 97%
pure, which is an improvement compared to Robie et al.61, but the measured values
were only discussed and shown in figures and not tabulated.
These measurements may still be improved since the used crystals have not been pure
even though the result is acceptable.
8.3.4
Thermal decomposition of FeCO3
High temperature analysis of FeCO3 is not the focus of this study, but a number of
groups have investigated the decomposition temperature and pressure of FeCO3 and
the properties are given here. French15 approached the subject and gave an outline of
the previous work.
Berg and Buzdov143,155 and Stubina and Toguri 81 found that FeCO3 decompose
according to reaction (275) in low CO2 pressure atmosphere, close to vacuum at
approximately 300 C, to form magnetite, Fe3O4 or FeOFe2O3156:
3FeCO3 o Fe3O4 2CO2 CO
(275)
Powell128 showed that the CO:CO2 ratio is not 1:2, in the initial part of the reaction.
He suggests that FeCO3 decompose to wstite, FeO, and CO2 is later reduced to CO
by FeO. Berg and Buzdov155 show that the CO2 + CO pressure of FeCO3 is 10-18 atm
at 25 C. Reiterer80 assumed that FeO was the reaction product and estimated the
pressure to be 0.0003 atm.
The decomposition of FeCO3 is considerably more complicated in an oxygen
atmosphere143. Berg and Buzdov143 shows that oxygen oxidises magnetite to
maghemite, D Fe2O3 , with an intermediate form, J Fe2O3 . Trace amount of
oxygen and high CO2 pressure makes FeCO3 decompose at approximately 360 C by
the following reaction as studied by French15:
3FeCO3 12 O2 o Fe3O4 3CO2
(276)
He also showed that hematite, Fe2O3, may be produced as a side product. French15
and Weidner45 found that above 450 to 550 C pure magnetite and graphite is always
produced. The subject was also discussed by Chai and Navrotsky82.
8.4
Synthesis of FeCO3
FeCO3 has been synthesised by different reaction paths. Nine patents have been
obtained on FeCO3 synthesis, as shown in table 29. Table 30 gives an overview of the
used reaction species and the colour of the produced FeCO3. Babcan157 gives an
overview of the old literature on FeCO3 production and Ptacek13 tested and compared
some of the methods. The difference between reaction paths depends on the iron
source, the carbonate source, and process condition of temperature and pressure.
- 172 -
184
A typical source of iron is iron powder, iron wire, or an iron salt of sulphate, chloride,
chlorate or oxalate. The source of carbonate may either be gaseous CO2 or a salt of
bicarbonate, carbonate, oxalate or urea. Three types of process condition are used,
room temperature and pressure, mid temperature and pressure and very high
temperature and pressure.
The above conditions fall in four groups. They are all related to the reaction path for
the carbonate formation.
One group uses very high temperature and pressure15,16,82,158,146. Iron oxalate is always
used in these methods since it decomposes to CO2 and carbonate. Ehrhardt et al.158
show that a very high CO2 pressure and temperature is needed to synthesise the
desired quality. The method requires specialised high pressure equipment and only
small samples may be produced.
Another approach uses pure gaseous CO2 to form carbonate. CO2 dissolves in water
and dissociates to carbonate. Iron in the form of wire or powder dissolves and FeCO3
precipitates31,133,11. The method is suited for producing high purity FeCO3 since it
does not require the interference from other ions as discussed by Gould133. The
method is slow since it requires dissolution of the iron. Gaseous CO2 has also been
used with aqueous iron salt mixtures to precipitate FeCO380,159. This process requires
high pressure equipment and a thorough deaeration of the setup. The same is true
when using urea ((NH2)2CO)115,160. Here ammonia is produced as a side product and
the product may be impure. The work by Huxley and Fetchin161 state that FeCO3 may
be produced in a CO2-ammonia process, in reality they produce iron carbamate,
Fe NH 2CO2 2 .
In a third process FeCO3 may be precipitated from solutions of carbonate. This
involves higher pH and the precipitate may be polluted by co-precipitated Fe(OH)2.
The precipitate is typically of lower quality and has been used in processes to remove
iron as shown by Konz139. Sodium or potassium carbonate is typically used in this
process141,143,162,139,163,164,165,166 or calcium carbonate168,167,128,169,157,140,81. Ptacek13
showed that SrCO3 could alternatively be used as carbonate source. Reynolds144 has
shown on the other hand that the precipitated solid is most likely not iron carbonate.
The key is the salt concentration. He explained that stable FeCO3 only precipitates
from the dilute solution. Concentrated solution also precipitate FeCO3 but within
minutes it reprecipitates as FeCO3K2CO34H2O. Another issue is that CaCO3 may
easily incorporate in the FeCO3 crystal as discussed by Rosenberg and Foit145.
The final method used by the majority of authors employs bicarbonate to precipitate
FeCO3. Iron sulphate is typically used as iron source84,85,24,9,170,162,171,101, a few authors
used perchlorate6,125 or Mohrs salt (FeSO4(NH4)2SO46H2O) 36,32,28,90,172,173. Lyon174
used iron chloride and the method is equivalent to Sharp170. Allison et al.175 and
Ptacek13 are the only authors who suggest using a mixture of carbonate and
bicarbonate. Allison et al.175 even states that carbamate may be employed. Any of the
setups applying a carbonate or bicarbonate salt were often supported by an
atmosphere of CO2 often at atmospheric conditions and room temperature. The
method is popular since it is simple and requires less sophisticated equipment. The
produced FeCO3 were sometimes post treated in an anaerobic oven between 100 and
200 C for drying.
- 173 -
185
Table 29: Patents on synthesis of FeCO3.
Authors
Hoy162
Year
1899
Parker164
1903
Flgge171,176,4,177 1905
Lilly163
1908
Gill168
1911
Allison et al.175
1958
Konz139
1966
Gould133
Lyon174
1968
1987
Short description of method

FeSO4 with K2CO3 or NaHCO3 at anaerobic conditions
in glove box. Washed as same conditions.
NaHSO4 and Iron to get FeSO4 precipitated as FeCO3,
Fe(OH)2 with carbonate or hydroxide.
FeSO4 and Na/KHCO3 to give FeCO3 which is washed
in CO2-water at anaerobic conditions.
FeSO4 and Na2/K2CO3 in glycerine at anaerobe
conditions under oil. States Na2/K2CO3 may be
substituted by CO2+NaOH.
As part of their Fe2O3 production, CaCO3 and CO2 are
mixed with FeCl2 to get FeCO3.
Ammonium bicarbonate or carbonate or carbamate
with FeSO4.
K2CO3 + CO2 precipitate FeCO3 by iron from the
process equipment at anaerobe conditions.
Iron wire and CO2 at anaerobic conditions.
FeCl2 and NaHCO3 (1:2) in a closed vessel at 200 C
(probably high CO2 pressure due to decomposition of
NaHCO3). Precipitate filtered in glove box.
Colour
NA
NA
Greenish-white a
NA
NA
NA
NA
Bluish grey,
White or pale
cream
Ghosh et al.7 states this may be due to oxidation
The colour of FeCO3 was often white, but beige is also typical. Krustinsons167
mentions how FeCO3 precipitate as grey-white at 120 C, which change to brow
above 130 C. Problems while performing the precipitation reveals green, blue or
black precipitates. FeCO3 oxidises to red-brown magnetite and hematite. Baudisch
and Welo166 and Gayer and Woontner178 show some of the difficulties observed when
precipitating ferrous iron. The observed precipitate has various colours and they are
not related to ferric iron. The colour scheme is similar to precipitation by mixing
aqueous FeCl2 and Na2CO3: First dirty green, then black and finally red-brown. Gayer
and Woontner178 suggest the different colour are due to production of different
hydrates of Fe(OH)2 and finally production of ferric iron. It must be underlined, that
iron carbonate may only be produced at anaerobic condition where all oxygen has
been strictly removed.
- 174 -
186

Table 30: Methods for synthesising FeCO3 as published in the literature.
Authors
Gnzburg 141
Smith84
Year
1911
1918
Tillmans and
Klarmann11
Baudisch and Welo166
Lemke and Biltz160
Krustinsons167
Sharp170
Graf165
Berg and Buzdov143
Powell128
Johannes169
Singer and Stumm6
Babcan157,140
1924
Weidner16, French15,
Bardy and Pr85
1925
1934
1935
1960
1961
1961
1965
1968
1970
1970,
1974
1971,
1994
1976
Short description of method

FeSO4(NH4)2SO4?H2O and Na2CO3. O2 avoided.
FeSO4 and NaHCO3 in saturated CO2 solution at 100
C and 400psi (27.6bar) CO2.
CO2 and iron particles. O2 avoided by vacuum and
boiling.
FeSO4 and Na2CO3 at anaerobic condition and room T.
FeCl2 and Urea at 200 C. Purity is low, 15%.
FeCl2 and CaCO3 in melted glass pipe for 20h, 120 C.
FeSO47H2O and NaHCO3 with CO2(g).
FeSO4 and Na2CO3 with CO2(g) at 143 C or 300 C.
FeSO4 and Na2CO3 at 25 C at anaerobic conditions.
Excess FeCl2 and CaCO3, 170C for 24h. O2 avoided.
FeCl2 and CaCO3 at 150 to 300 C at 1000 bar.
FeClO4 and NaHCO3 with HClO4 in N2 flush.
FeCl2 and CaCO3 at 140 C at low pressure. O2 content
unknown.
FeC2O42H2O heated in 97% CO2/CO at 380 C and
3kbars for 48h.
FeSO4 and NaHCO3 in argon atmosphere at 80 C, post
washed.
FeCl2 with HCl, urea and CO2 atmosphere of 40-80atm
at 140-160 C in autoclave, washed and dried at 80 C.
1: dry ice and FeCl24H2O in autoclave at 150 C
2: CO2 from bottle in FeCl2 solution at 220 C.
FeC2O42H2O heated at 470 C and 3500bar CO2.
FeSO47H2O and CO2 at 1200atm at 180 C
Fe(OH)2 and CO2 at 2500atm at 150 C.
FeSO4 and KHCO3 in N2 atmosphere.
Fe wire and CO2.
Mohrs salt and NaHCO3 with CO2 at room T, heated at
70-90 C. Anaerobic conditions.
Excess FeCl24H2O and CaCO3 at 170 C dried at 100
C. Unclear if anaerobic conditions were obtained.
FeSO47H2O with HCl and NaHCO3 in autoclave at
100 C and 19bar CO2.
Gamsjger and
Reiterer115
Reiterer80
1979
Ehrhardt et al.158
Ehrhardt et al.159
1980
1980
Davies and Burstein101

Wajon et al.31
Greenberg and
Tomson36,32,33
Stubina and Toguri81
1980
1985
1986
Wersin et al.97 and

Wersin9 used by
Bruno et al.24
Johnson and
Tomson27,28,29
Ptacek13
1990,
1992,
1989
1991 Mohrs salt and NaHCO3 at anaerobic condition.
Temperature unknown.
1992 FeCl24H2O and 1:4 Na2CO3/NaHCO3 at anaerobic
conditions in glove box at room temperature. Removed
Fe(III) and O2 by precipitation. Heated at 160 C.
1996 FeC2O42H2O and CO2 at 850 C and 18kbars for 24h.
1999 Method by Johnson and Tomson28 at room temperature
post heated at 75 C.
2002 Equivalent to Greenberg and Tomson36,32,33 and details
from Bruno et al.24.
2004 Fe(ClO4)2 and NaHCO3 at 25 C and 1atm, 10% CO2,
at anaerobic conditions.
2006 FeSO4(NH4)2SO46H2O and KHCO3. O2 avoided.

Heuer and Stubbins173
Silva et al.90
Jimenez-Lopez And
Romanek125
Gogolev172
1980
1989
Colour
White
White
White
NA
Grey-white a
Yellow-brown b
NA
White
Grey
NA
NA
NA
Tan (beige)
NA
NA
White
Light-brown
Colour-less
White, pale-greenc
Green
Light grey, white
NA
Beige
White
White
White
White
NA
NA
White
Changed to brown while heated to 130-180C. b The colour changed to Yellow-brown while drying. c
Low concentration of KHCO3: white, high KHCO3: pale-green.
- 175 -
187
8.5
Discussion and conclusion
This study gives a review of available thermodynamic properties of FeCO3 found in

the open literature. The heat capacity has been determined experimentally by a few
authors. Until date the work by Robie et al.61 is the most accurate. Their temperature
correlation of heat capacity is also the most reliable. The standard state entropy, which
obeys the third law of thermodynamics, was determined by Robie et al.61 and is
consistent with their measurements of heat capacity.
Based on the study performed in this work we recommend the properties given in
table 31, 32 and 33. Two different alternatives are given. Table 31 contains a set of
properties which are consistent with the NIST55 thermodynamic properties. Table 32
presents properties which are consistent with the CODATA57,58 thermodynamic
properties. We consider these two alternatives to be of equal quality and consistent.
Table 31: Recommended standard state properties consistent with NIST55 at 298.15K at 1 bar.
Properties are from NIST if not stated otherwise.
Fe2+
CO32FeCO3
a
' f Gq
'f Hq
' f Sq
kJ/mol
-78.9
-527.81
-665.16 b
kJ/mol
-89.1
-677.14
-738.28 c
J/(molK)
-137.7
-56.9
95.47 d
J/(molK)
-34.30
-501.03
-245.26
Values were calculated using the NIST55 standard entropies S Fe s =27.28 J/(molK), SC s =5.74
D
J/(molK), SO g =205.138 J/(molK), S H g =130.684 J/(molK), and S H aq =0 J/(molK). b Values

2
2
D
were calculated consistently with ' f GFe2 and ' f GCO 2 using pKsp=10.24 of Singer and Stumm6. c
3
calculated using equation (260). d obtained from Robie et al.61.

Table 32: Recommended standard state properties consistent with CODATA57,58 at 298.15K at 1 bar.
' f Gq
'f Hq
kJ/(molK)
-90.53
-527.90 c
-676.88 e
kJ/(molK)
-90
-675.23 d
-750.01 f
' f Sq
J/(molK)
J/(molK)
-101.6
1.76
-50 d
-494.15
95.47 g
-245.30
D
a
57,58
Values were calculated using the CODATA
standard entropies S Fe s =27.319 J/(molK),
Fe2+ b
CO32FeCO3
SCD s =5.74 J/(molK), SOD2 g =205.138 J/(molK), S HD 2 g =130.68 J/(molK), S HD aq =0 J/(molK). b

D
Obtained from Parker and Khodakovskii57. c Calculated from ' f H CO 2 and ' f SCO 2 . d Obtained
3
from CODATA58. e Values were calculated consistently with ' f GFe2 and ' f GCO 2 using
3
pKsp=10.24 of Singer and Stumm6. f calculated using equation (260). g obtained from Robie et al.61.
The two alternatives are different since there is a controversy regarding the
thermodynamic properties of Fe2+. There are two sides in the literature, both being
equally correct, but they give different results. The two opposing groups believe that
the potential of Fe2+ should be measured in vacuum or in hydrogen atmosphere. The
discrepancies influence the standard state formation properties of FeCO3 which are
shown in table 31 and 32. It can be concluded that the use of any Fe2+ properties
- 176 -
188
should be done with care and properties must be evaluated before use. It is
recommended never to combine properties from different data collections in the same
work. It will eventually lead to inconsistencies between thermodynamic properties of
Fe2+ and FeCO3.
D
D
S FeCO
was obtained from Robie et al.61 and shown both in table 31 and 32. ' f S FeCO
3
3
is comparable in the two tables, since the standard state properties of the reference
compounds are similar in NIST and CODATA.
The solubility constant of FeCO3 can be determined experimentally and
independently of the Fe2+ properties. At the same time it may be used to consistently
determining the standard state formation properties of FeCO3 from the Fe2+
properties. It has been determined by a considerable number of authors, but the work
by Kelley and Anderson56 is inaccurate and should not be used even though it is found
in most property collections. The equilibrium constants determined by Singer and
Stumm6 or Haarberg et al.87 on the other hand, are more consistent. These two works
also apply a basic speciation scheme, opposite many other works. There is an ongoing
debate on whether FeHCO3 , FeCO3 aq , Fe CO3
2
2
aq , and
FeCO3 OH
exist. The properties of these compounds are uncertain and they have unfortunately
been included in many works recently. If these compounds are used, a relatively high
solubility constant, pK sp , of FeCO3 is found. This is due to the lower observed free
Fe2+ concentration in favour of the other iron species. If any of the species are
included, then it is suggested to refit all equilibrium constants in order to obtain a
consistent speciation scheme. Similarly the solubility constant of FeCO3 must always
be used in the speciation scheme that it was fitted to. It becomes inconsistent if
properties from different data collections are combined.
Table 33 shows recommended standard state properties of FeCO3 dissolution of
reaction (252) and (253). Properties consistent with both NIST and CODATA are
shown. The equilibrium constant was assumed to follow the calculated value by
Singer and Stumm6 of pKsp=10.24.
Table 33: Recommended properties of reaction (252) and (253), consistent with NIST55 or
CODATA57,58 at 298.15 K and 1 bar.
Reference
Reaction (k)
NIST55
(252) (sp)
CODATA57,58 (252) (sp)
NIST55
(253) (psp)
CODATA57,58 (253) (psp)
a
pKk
' r ,k H D
' r ,k S D
10.24 a
10.24 a
-7.92 c
-7.91 c
kJ/(molK)
-27.96 b
-15.21 b
-30.16 d
-19.32 e
J/(molK)
-289.81 b
-247.05 b
-50.55 d
-86.68 e
Obtained from Singer and Stumm6. b Calculated from properties in table 31 and 32. c Calculated from
D
' psp H D and ' psp S D d Calculated from properties in table 31 and NIST: ' f H CO
=-393.509
2g
D
kJ/mol, ' f H H O l =-285.83 kJ/mol, SCO g =213.74 J/(molK), and S H O l =69.91 J/(molK) e
2
2
2
D
Calculated from properties in table 32 and CODATA: ' f H CO g =-393.51 kJ/mol,

2
D
D
' f H HD 2O l =-285.83 kJ/mol, SCO
=213.785 J/(molK), and S H O l =69.95 J/(molK).
2g
2
' r , sp H D was calculated from the enthalpies of formation of FeCO3, CO32-, and Fe2+
shown in table 31 and 32. The NIST value is an order of magnitude lower than the
CODATA value. The value is uncertain but consistent with the remaining database
- 177 -
189

D
properties. The difference is related to the inequality of ' f GFe
2 between NIST and
CODATA. Preis and Gamsjger45 discussed this problem and recommend the
CODATA properties by evaluating ' r , sp H D . Unfortunately they included an extended
speciation scheme in their calculations, using FeCO3 aq and Fe CO3
2
2
aq
despite the uncertainties outlined in the sections above. The discrepancy in ' r , sp S spD
D
and ' r , psp S D between NIST and CODATA is also related to the ' f GFe
2 . The pKpsp is
calculated from the pKsp value and the small difference between NIST and CODATA
is related to the difference in the S properties of the reference compounds.
An overview of experimental literature on FeCO3 solubility measurements is given in
this study. The majority of literature studies fix the ionic strength by adding another
salt. The added ions will influence the solubility significantly. The best way to avoid
errors is to keep the amount of unnecessary ions low. Some studies neglect to measure
temperature or apply vague temperature control. CO2 content in the system is an
important parameter for FeCO3 solubility and measurements should always include a
CO2 determination. Often measurements are misinterpreted as supersaturated due to
improper use of equilibrium constants. In reality it has been shown, that FeCO3 may
form supersaturated solutions. It has also been shown that equilibrium is eventually
attained.
FeCO3 decompose at high temperature and a brief overview of the related literature is
given in this study. A section on producing FeCO3 is also given and many different
methods may be used. Hydroxide concentration and oxygen contamination are two
key parameters to control in order to prevent precipitation of Fe(OH)2 and to avoid
oxidation of Fe2+ to Fe3+. Undesired reaction products may be obtained if oxygen is
not carefully removed.
- 178 -
190
8.6
Literature cited
[1]
D.L. Jensen, J. K. Boddum, J. C. Tjell and T. H. Christensen: The solubility of rhodochrosite

(MnCO3) and siderite (FeCO3) in anaerobic aquatic environments, Applied Geochemistry,
2002, 17(4), 503-511.
[2]
D. L. Jensen: Speciation of dissolved iron(II) and manganese(II) in a groundwater pollution

plume, Environmental Science and Technology, 1998, 32(18), 2657-2664.
[3]
A. P. Schwab and W. L. Lindsay: Effect of redox on the solubility and availability of iron,
Soil Science Society of America Journal, 1983, 47(2), 201-5.
[4]
A. Flgge: Process of making Ferrous Carbonate, US patent 842452, 1907.
[5]
J. D. Hem: Restraints on Dissolved Ferrous Iron Imposed by Bicarbonate, Redox Potential,

and pH, U. S. Geol. Surv. Water Supply, 1960, paper 1459-B.
[6]
P. C. Singer and W. Stumm: The solubility of ferrous iron in carbonate-bearing waters,

Journal - American Water Works Association, 1970, 62, 198202.
[7]
M. M. Ghosh, J. T. O'connor and R. S. Engelbrecht: Rate of precipitation of iron in aerated

groundwaters, J. Sanit. Eng. Div. Am. Soc. Civil Engrs., 1996, 92(1), 199-213.
[8]
W. Stumm and P. C. Singer: Precipitation of Iron in Aerated Groundwaters, discussion, J.

San. Eng., ASCE, 1966, 92, 120.
[9]
P. Wersin: The Fe(II)-CO2-H2O system in anoxic natural waters: Equilibria and surface
chemistry, PhD dissertation, Swiss Fed. Inst. of Technol., Zrich, 1990.
[10]
F. J. Millero and D. Pierrot: A chemical equilibrium model for natural waters, Aquatic
Geochemistry 1998, 4(1), 153-199.
[11]
J. Tillmans and B. Klarmann: The solution of iron by oxygen-free natural waters,

Angewandte Chemie, 1924, 36, 94-7, 103-4, 111-5.
[12]
C. J. Ptacek and D. W. Blowes: Influence of Siderite on the Pore-Water Chemistry of Inactive

Mine-Tailings Impoundments, ACS Symposium Series, 1994, 550, 172-189.
[13]
C. J. Ptacek: Experimental determination of siderite solubility in high ionic-strength aqueous

solutions, Ph.D. thesis, University of Waterloo, Ontario, Canada, 1992.
[14]
P. E. Dresel: The dissolution kinetics of siderite and its effect on acid mine drainage, Ph.D.
thesis. Pennsylvania State University, 1989.
[15]
B. M. French: Stability Relations of Siderite (FeCO3) In System Fe-C-O, American Journal

of Science, 1971, 271(1), 37-78.
[16]
J. R. Weidner: Equilibria In System Fe-C-O .1. Siderite-Magnetite-Carbon-Vapor

Equilibrium From 500 to 10,000 Bars, American Journal Of Science, 1972, 272(8), 735-751.
[17]
P. E. Rosenberg: Subsolidus relations in the system CaCO3-FeCO3, Am. J. Sci., 1963, 261,
683-690.
[18]
G. M. Marion, D. C. Catling and J. S. Kargel: Modeling aqueous ferrous iron chemistry at

low temperatures with application to Mars, Geochimica Et Cosmochimica Acta, 2003,
67(22), 4251-4266.
[19]
G. H. Rau: Possible use of Fe/CO2 fuel cells for CO2 mitigation plus H2 and electricity
production Energy Conversion and Management, 2004, 45(13-14), 2143-2152.
[20]
J. L. Palandri and Y. K. Kharaka: Sequestration of CO2 in redbeds: siderite precipitation from

CO2 and SO2 waste gas. Water-Rock Interaction, Proceedings of the International Symposium
on Water-Rock Interaction, 11th, Saratoga Springs, NY, United States, 2004, 1, 567-571.
[21]
E. Deltombe and M. Pourbaix: Equilibrium potential-pH diagram for the system Fe-CO2-H2O
at 25, Proc. 6th Meeting Intern. Comm. Electrochem. Thermodynam. and Kinet., 1955, 12432.
- 179 -
191

[22]
W. Stumm and G. F. Lee: The chemistry of aqueous iron, Schweizerische zeitschrift fr

hydrologie, 1960, 22, 295-319.
[23]
J. D. Hem: Equilibrium chemistry of iron in ground water, In Principles and application of

water chemistry , (Ed. S.D. Faust and J.V. Hunter), 625; 1967, John Wiley and sons, inc.,
New York.
[24]
J. Bruno, P. Wersin, W. Stumm and F. Brandberg: On the influence of carbonate in mineral

dissolution: II. The solubility of FeCO3(s) at 25C and 1 atm total pressure, Geochim.
Cosmochim. Acta., 1992, 56(3), 11491155.
[25]
B. Mishra, S. Al-Hassan, D. L. Olson and M. M. Salama: Development of a predictive model

for activation-controlled corrosion of steel in solutions containing carbon dioxide, Corrosion,
1997, 53(11), 852-859.
[26]
J. Dong, T. Nishimura and T. Kodama: Corrosion behavior of carbon steel in bicarbonate

(HCO3-) solutions, Materials Research Society Symposium Proceedings, 2002, 713(Scientific
Basis for Nuclear Waste Management XXV), 105-112.
[27]
M. L. Johnson and M. B. Tomson: Ferrous carbonate precipitation kinetics and its impact
CO2 corrosion, CORROSION/91, NACE, 1991, paper 268.
[28]
M. B. Tomson and M. L. Johnson: How Ferrous Carbonate Kinetics Impacts Oilfield

Corrosion, Proceedings, International Symposium on Oilfield Chemistry, Anaheim, CA, SPE,
1991, Paper No. 21025, 257-264.
[29]
M. L. Johnson: Ferrous Carbonate Precipitation Kinetics - A Temperature Ramped

Approach, Ph. D. Thesis, Rice University, Houston, Texas, 1991.
[30]
E. Van Hunnik, B. F. M. Pots and E. L. J. A. Hendriksen: The Formation of Protective FeCO3

Corrosion Product Layers in CO2 Corrosion, CORROSION/96, NACE, 1996, paper no. 6.
[31]
J. E. Wajon, G.-E. Ho and P. J. Murphy: Rate of precipitation of ferrous iron and formation of
mixed iron-calcium carbonates by naturally occurring carbonate materials, Water Research,
1985, 19(7), 831-7.
[32]
J. L. Greenberg: High Temperature Kinetics of Precipitation and Dissolution of Ferrous

Carbonate, M.S. Thesis. Rice University, Houston, Texas, 1986.
[33]
J.L. Greenberg and M. B. Tomson: Kinetics of Precipitation and Dissolution of Ferrous

Carbonate, International Symposium on Metals Speciation, Separation, and Recovery, (ed., J.
W. Patterson), 1986, Lewis Publication, Inc., Chelsea, MI, pp. I-1/17.
[34]
G. I. Ogundele and W. E. White: Some Observations on Corrosion of Carbon-Steel in

Aqueous Environments Containing Carbon-Dioxide, Corrosion, 1986, 42(2), 71-78.
[35]
D. B. Robert: Solubility of iron(II) carbonate at temperatures between 30 and 80C, Talanta,

1991, 38(2), 205211.
[36]
J. L. Greenberg and M. B. Tomson: Precipitation and dissolution kinetics and equilibria of

aqueous ferrous carbonate vs. temperature, Appl. Geochem., 1992, 7(2), 185190.
[37]
S. V. Golubev, P. Bnzeth and J. Schott: Siderite dissolution kinetics in acidic aqueous

solutions from 60 to 100C and 0 to 50 atm pCO2, Geophysical Research Abstracts, 2007, 9,
04038.
[38]
A. Dugstad: Mechanism of Protective Film Formation During CO2 Corrosion of Carbon

Steel, CORROSION/98, NACE, 1998, paper no. 31.
[39]
S. Nesic, M. Nordsveen, R. Nyborg and A. Stangeland: A Mechanistic Model for CO2

Corrosion with Protective Iron Carbonate Films, CORROSION/01, NACE, 2001, Paper No.
40.
[40]
M. Nordsveen, S. Nesic, R. Nyborg and A. Stangeland: A mechanistic model for carbon

dioxide corrosion of mild steel in the presence of protective iron carbonate films - Part 1:
Theory and verification, Corrosion, 2003, 59(5), 443-456.
[41]
W. Sun, K. Chokshi and S. Nesic: Iron Carbonate Scale Growth and the Effect of Inhibition
in CO2 Corrosion of Mild Steel, CORROSION/05, NACE, 2005, paper no. 285.
- 180 -
192

[42]
W. Sun, S. Nesic: Basics Revisited: Kinetics of Iron Carbonate Scale Precipitation in CO2
Corrosion, CORROSION/06, NACE, 2006, paper no. 365.
[43]
E. Gulbrandsen: Acetic acid and carbon dioxide corrosion of carbon steel covered with iron
carbonate, CORROSION/07, NACE, 2007, paper no. 322.
[44]
R. Grauer, Translation by: Urs Berner: Solubility products of M(II) Carbonates, PSI
Bericht 99-04, Paul Scherrer Institute, Villigen, 1999.
[45]
W. Preis and H. Gamsjager: Critical evaluation of solubility data: enthalpy of formation of

siderite, Physical Chemistry Chemical Physics, 2002, 4(16), 4014-4019.
[46]
R. A. Robie and D. R. Waldbaum: Thermodynamic properties of minerals and related

substances at 298.15K (25C) and one atmosphere (1.013 bars) pressure and at higher
temperatures, U.S. Geol. Survey Bull. 1259, 256 pages, 1968.
[47]
R. A. Robie, B. S. Hemingway and J. R. Fisher: Thermodynamic properties of minerals and

related substances at 298.15 K and 1 bar (105 pascals) pressure and at higher temperatures,
U.S. Geol. Survey Bull. 1452, 456 pages, 1978.
[48]
R. A. Robie and B. S. Hemingway: Thermodynamic properties of minerals and related

substances at 298.15 K and 1 bar (105 Pascals) pressure and at higher temperatures, U.S.
Geological Survey Bulletin 2131, Washington, 1995.
[49]
D. K. Nordstrom, L. N. Plummer, D. Langmuir, E. Busenberg, H. M. May, B. F: Jones and D.

L. Parkhurst: Revised chemical equilibrium data for major water-mineral reactions and their
limitations, Chemical Modelling of Aqueous Systems II, ACS symposium series, 1990, 416,
398-413.
[50]
W. E. Latimer: The Oxidation States of the Elements and Their Potentials in Aqueous
Solutions, 2nd Ed., PrenticeHall Inc. Englewood Cliffs, N. J., 1952.
[51]
D. Langmuir: The Gibbs free energies of substances in the system Fe-O2-H2O-CO2 af 25 C,

U.S. Geol. Surv. Prof. Pap., 1969, 650B, p. 180-184.
[52]
I. Barin, O. Knacke and O. Kubaschewski: Thermochemical Properties of Inorganic

Substances, Berlin: Springer-Verlag. 1973 + Supplement 1977.
[53]
A. E: Martell and. R. M. Smith: Critical stability constants. 4: Inorganic complexes, New

York,N.Y., Plenum press., 1976 repr. 1981.
[54]
D. D. Wagman, W. H. Evans, V. B. Parker, I. Halow, S. M. Bailey and R. H. Schumm:

Selected values of chemical thermodynamic properties. Tables for elements 35 through 53 in
the standard order of arrangement, Natl. Bur. standards Tech. Note 270-4, 154 pages, 1969.
[55]
D. D. Wagman, W. H. Evans, V. B. Parker, R. H. Schumm, I. Halow and S. M. Bailey: NIST

Chemical Thermodynamics Database Version 1.1, J. Phys. Chem. Ref. Data, 1982, 11
Supplement No. 2.
[56]
K. K. Kelley and C. T. Anderson: Contributions to the data on theoretical metallurgy. IV.

Metal carbonates: Correlations and applications of thermodynamic properties, U.S.
Department of the Interior, Bureau of Mines, 1935, 384, 1-73.
[57]
V. B. Parker and I. L. Khodakovskii: Thermodynamic properties of the aqueous ions (2+ and
3+) of iron and the key compounds of iron, Journal of Physical and Chemical Reference Data,
1995, 24(5), 1699-1745.
[58]
J. D. Cox, D. D. Wagman and V. A. Medvedev: CODATA Key values for thermodynamics,

Hemisphere Publishing Corp., New York, 1989.
[59]
C. F. Baes and R. E. Mesmer: The Hydrolysis of Cations. Iron., Wiley, New York, 226-237,
1976 repr. 1986.
[60]
C. T. Anderson: The heat capacities of Magnesium, Zinc, Lead, Manganese and iron
carbonate at low temperature, J. Am. Chem. Soc., 1934, 56, 849.
[61]
R. A. Robie, H. T. Haselton and B. S. Hemingway: Heat-Capacities and Entropies of

Rhodochrosite (MnCO3) and Siderite (FeCO3) Between 5K and 600K, American
Mineralogist, 1984, 69(3-4), 349-357.
- 181 -
193

[62]
K. K. Kelley: Contribution to the data on theoretical metallurgy. XIII High temperature heat
content, heat capacity and entropy data for the elements and inorganic compounds, U.S. Bur.
Mines Bull. 584, 1960.
[63]
I. N. Kalinkina: Magnetic specific heat of antiferromagnetic Co, Ni, Mn, and Fe carbonates,
Soviet Physics JETP, 1963, 16(6), 1432.
[64]
V. N. Kostryukov and I. N. Kalinkina: Heat capacity and entropy for manganese, iron, cobalt,
and nickel carbonates at low temperatures, Zhurnal Fizicheskoi Khimii, 1964, 38(3), 780-1.
[65]
I N. Kalinkina and V. N. Kostryukov: Discontinuities of heat capacity in antiferromagnetic

carbonates, Fizika Tverdogo Tela (Sankt-Peterburg), 1966, 8(1), 176-80.
[66]
R. A. Robie: Heats and free energies of formation of troilite, herzenbergite, magnesite, and
rhodochrosite calculated from equilibrium data, U.S. Geol. Survey Prof., 1965, paper 525-D,
65-72.
[67]
D. C. Helgeson, J. M. Delany, H. W. Nesbitt and D. K. Bird: Summary and critique of the

thermodynamic properties of rock-forming minerals, Amarican journal of science, 1978,
278A, 1-229.
[68]
D. D. Wagman, W. H. Evans, V. B. Parker, I. Halow, S. M. Bailey and R. H. Schumm:

Selected values of Chemical Thermodynamic Properties. Tables of the first 34 elements in the
standard order of arrangement, Natl. Bur. standards Tech. Note 270-3, 268 pages, 1968.
[69]
P. R. Tremaine, R. Vonmassow and G. R. Shierman: A Calculation of Gibbs Free-Energies

for Ferrous-Ions - Solubility of Magnetite in H2O and D2O to 300 C, Thermochim Acta,
1977, 19(3), 287-300.
[70]
P. R. Tremaine and J. C. LeBlanc: The Solubility of Magnetite and the Hydrolysis and
Oxidation of Fe2+ in Water to 300C, Journal of Solution Chemistry, 1980, 9(6), 415-442.
[71]
M. Randall and M. Frandsen: The standard electrode potential of iron and the activity
coefficient of ferrous chloride, J. Am. Chem. Soc., 1932, 54, 47-54.
[72]
W. A. Patrick and W. E. Thompson: Standard electrode potential of the iron-ferrous ion

couple at 25, Journal of the American Chemical Society, 1953, 75, 1184-7.
[73]
J. M. Larson, P. Cerutti, H. K. Garber and L. G. Hepler: Electrode potential and

thermodynamic data for aqueous ions. Copper, zinc, cadmium, iron, cobalt, and nickel, J.
Phys. Chem., 1968, 72, 2902.
[74]
G. K. Johnson and J. E. Bauman: Equilibrium-Constants for Aquated Iron(II) Cation, Inorg

Chem, 1978, 17(10), 2774-2779.
[75]
W. Preis and H. Gamsjager: Thermodynamic investigation of phase equilibria in metal

carbonate-water-carbon dioxide systems, Monatshefte Fur Chemie, 2001, 132(11), 13271346.
[76]
H. Gamsjager, E. Konigsberger and W. Preis: Solubilities of metal carbonates, Pure And

Applied Chemistry, 1998, 70(10), 1913-1920.
[77]
T. J. B. Holland and R. Powell: An enlarged and updated internally consistent thermodynamic

dataset with uncertainties and correlations: the system K2O-Na2O-CaO-MgO-MnO-FeO-Fe2OAl2O3-TiO2-SiO2-C-H2-O2, J. of Metamorphic Geology, 1990, 8, 89-124.
[78]
D. E. Wilcox and L. A. Bromley: Computer estimation of heat and free energy of formation
for simple inorganic compounds, Journal of Industrial and Engineering Chemistry
(Washington, D. C.), 1963, 55(7), 32-9.
[79]
F. Reiterer, W. Johannes, H. Gamjger: Semimicro determination of solubility constants:

copper(II) carbonate and iron(II) carbonate, Mikrochimica Acta, 1981, 1, 6372.
[80]
F. Reiterer: Lslichkeitskonstanten und Freie Bildungsenthalpien neutraler

bergangsmetallcarbonate, Thesis, Montanuniversitt Leoben, 1980.
[81]
N. M. Stubina and J. M. Toguri: The decomposition pressure of synthetic siderite (FeCO3),

Trans. Iron Steel Soc., 1989, 10, 87-90.
- 182 -
194

[82]
L. Chai and A. Navrotsky: Enthalpy of formation of siderite and its application in phase
equilibrium calculation, American Mineralogist, 1994, 79(9), 921-929.
[83]
H.C. Helgeson: Thermodynamics of Hydrothermal Systems at Elevated Temperatures and

Pressures, American journal of science, 1969, 267, 729.
[84]
H. J. Smith: On Equilibrium in the system: ferrous carbonate, carbondioxide and water,

Journal of the American Chemical Society, 1918, 40, 879883.
[85]
J. Bardy and C. Pr: Determination experimentale du coefficient de solubilite du carbonate

ferreux en milieu aqueux, Tribune du CEBEDEAU, 1976, 387, 75.
[86]
V.S. Serebrennikov: Redox conditions in the low caucasus carbonated mineral springs,
Geochem. Int., 1977, 14, 925-945.
[87]
T. Haarberg, J. E. Jakobsen and T. Oestvold: The effect of ferrous iron on mineral scaling
during oil recovery, Acta Chemica Scandinavica, 1990, 44(9), 907-15.
[88]
C. J. Ptacek, E. J. Reardon: Solubility of siderite (FeCO3) in concentrated NaCl and Na2SO4

solutions at 25C, Water-Rock Interaction, Proc. Int. Symp., 7th, 181183, 1992.
[89]
J. D. Garber, R. S. Perkins, R. J. Vamshidhar and R. A. Rama: Calculation of Downhole pH

and Delta pH in the Presence of CO2 and Organic Acids, CORROSION/96, NACE; 1996,
paper no. 176.
[90]
C. A. R. Silva, X. W. Liu and F. J. Millero: Solubility of siderite (FeCO3) in NaCl solutions,

Journal of Solution Chemistry, 2002, 31(2), 97-108.
[91]
Y. Sun, S. Nesic: Parametric Study and Modeling on Localized CO2 Corrosion in Horizontal
Wet Gas Flow, CORROSION/04, NACE, 2004, paper no. 380.
[92]
D. A. Palmer and R. van Eldik: The chemistry of metal carbonato and carbon dioxide
complexes, Chemical Reviews, 1983, 83(6), 651731.
[93]
E. Hgfeldt: Stability constants of metal-ion complexes - part A: inorganic ligands, IUPAC

chemical data series volume 21, 1982.
[94]
N. P. Zhuk: Thermodynamic constants of sulfates, carbonates, chromates, bromates, iodates,

oxalates, and other salts slightly soluble in water, Zhurnal Fizicheskoi Khimii, 1954, 28,
1690-7.
[95]
M. M. Ghosh: Oxygenation of ferrous iron(II) in highly buffered waters, In: AqueousEnvironmental Chemistry of Metals , (Ed. Rubin, A. J), Ann Arbor Science: Ann Arbor,
1974; 193-217.
[96]
C. W. Davies: The Extent of dissociation of salts in water, Journal of the Chemical Society,
1938, 2, 2093-2100.
[97]
P. Wersin, L. Charlet, R. Karthein and W. Stumm: From adsorption to precipitation: Sorption

of Mn2+ on FeCO3(s), Geochim. Cosmochim. Acta., 1989, 53, 27872796.
[98]
C. Fouillac and A. Criaud: Carbonate and bicarbonate trace metal complexes: critical
reevaluation of stability constants, Geochem. J., 1984, 18, 297303.
[99]
D. Langmuir: Techniques of estimating thermodynamic properties for some aqueous

complexes of geochemical interest, ACS Symposium Series, 1979, 93, 353-87.
[100]
W. Stumm and J. J. Morgan: Aquatic chemistry : Chemical equilibria and rates in natural
waters, 3.ed. New York,N.Y., Wiley., 1996.
[101]
D. H. Davies and G. T. Burstein: The Effects of Bicarbonate on the Corrosion and Passivation
of Iron, Corrosion, 1980, 36(8), 416-422.
[102]
N. W. Sidgwick: The Chemical Elements and Their Compounds. Vols. II., University Press,
Oxford, 1952.
[103]
A. Wieckowski, E. Ghali, M. Szklarczyk and J. Sobkowski: The behaviour of iron electrode

in CO2 saturated neutral electrolyteI. Electrochemical study, Electrochimica Acta, 1983,
28(11), 1619.
- 183 -
195

[104]
V. Thurmond and F. J. Millero: Ionization of carbonic acid in sodium chloride solutions at

25C, J. Solution Chem., 1982, 11(7), 447.
[105]
F. J. Millero and D. J. Hawke: Ionic Interactions of Divalent Metals in Natural-Waters,

Marine Chemistry, 1992, 40(1-2), 19-48.
[106]
I. Grethe, A. V. Plyasunov and K. Spahiu: Estimations of Medium Effects on

Thermodynamic Data, in: Modelling in Aqautic Chemistry, OECD NEA, Paris, 1997, 325426.
[107]
L. N. Plummer and E. Busenberg: The Solubilities of Calcite, Aragonite and Vaterite in CO2H2O Solutions Between 0-C and 90 C, and an Evaluation of the Aqueous Model for the
System CaCO3-CO2-H2O, Geochim. Cosmochim. Acta, 1982, 46, 1011-1040.
[108]
T. E. Larson, Oxidation of Metals and Ions in Solution, In Principles and application of

water chemistry, (Ed. S.D. Faust and J.V. Hunter), John Wiley and sons, inc., New York,
1967, p. 433.
[109]
S. V. Mattigod nd G. Sposito: Estimated Association Constants for Some Complexes of

Trace-Metals With Inorganic Ligands, Soil Science Society of America Journal, 1977, 41(6),
1092-1097.
[110]
C. F. Baes and R. E. Mesmer: The Hydrolysis of Cations. Iron, American journal of science,
1981, 281, 935-962.
[111]
F. H. Sweeton and C. F. Baes, Jr.: The solubility of magnetite and hydrolysis of ferrous ion in
aqueous solutions at elevated temperatures, Journal of Chemical Thermodynamics, 1970,
2(4), 479.
[112]
D.R. Turner, M. Whitfield and A.G. Dickson: The equilibrium speciation of dissolved
components in freshwater and seawater at 25C and 1atm pressure, Geochim. Cosmochim.
Acta., 1981, 45, 855881.
[113]
S. Emerson: Early diagenesis in anaerobic lake sediments: chemical equilibriums in

interstitial waters, Geochimica et Cosmochimica Acta, 1976, 40(8), 925-34.
[114]
D. Postma: Pyrite and Siderite Formation in Brackish and Fresh-Water Swamp Sediments,
American Journal of Science, 1982, 282(8), 1151-1183.
[115]
H. Gamsjger and F. Reiterer: Investigation of equilibria involving carbon dioxide,

carbonates and water, Environ. Internatl., 1979, 2, 419-424.
[116]
H. Gamsjger and E. Knigsberger: Solubility of sparingly soluble ionic solids in liquids, In

The experimental Determination of Solubilities, (Ed. Hefter and R.P.T. Tomkins), Wiley,
2003, 315.
[117]
B. Klarmann: Die Lsung von Eisen durch Kohlensure, Zeitschrift fr Angewandte

Chemie, 1924, 37(34), 645-646.
[118]
R. Mller and L. Harant: Investigating the solubility of siderite and the electro-deposition of
iron from these solutions, Berg und Huttenmannisches Jahrbuch, 1935, 83, 93-102.
[119]
E. Mller and H. Henecka: The action of carbon dioxide under high pressure on metallic
iron, Z. anorg. allgem. Chem., 1929, 181, 159-71.
[120]
A. Dugstad: The Importance of FeCO3 Supersaturation on the CO2 Corrosion of Carbon

Steels, CORROSION/92, NACE, 1992, paper no. 14.
[121]
W. Leybold: The solution of iron by carbonic acid, Angewandte Chemie, 1924, 37, 190-1.
[122]
A. Dugstad and P.-E. Drnen: Efficient Corrosion Control of Gas Condensate Pipelines by
pH-Stabilisation, CORROSION/99, NACE, 1999, paper no. 20.
[123]
F. J. Millero, Y. Wensheng, J. Aicher: The speciation of Fe(II) and Fe(III) in natural waters,
Marine Chem., 1995, 50(1-4), 21.
[124]
D. W. King: Role of carbonate speciation on the oxidation rate of Fe(II) in aquatic systems,
Environmental Science & Technology, 1998, 32(19), 2997-3003.
- 184 -
196

[125]
C. Jimenez-Lopez and C. S. Romanek: Precipitation kinetics and carbon isotope partitioning

of inorganic siderite at 25C and 1 atm, Geochimica et Cosmochimica Acta, 2004, 68(3),
557571.
[126]
D. L. Leussing and I. M, Kolthoff: The solubility of ferrous hydroxides and the ionisation of
aquo-ferrous ion, J. am. chem. soc., 1953, 75, 2476.
[127]
E. L. Shock, D. C. Sassani, M. Willis, S. A. Sverjensky: Inorganic species in geologic fluids:

correlations among standard molal thermodynamic properties of aqueous ions and hydroxide
complexes, Geochimica et Cosmochimica Acta, 1997, 61(5), 907-950.
[128]
H. E. Powell: Thermal decomposition of siderite and consequent reactions, U.S. Bur. Mines
Rept. Inv. 6643,44 p, 1965.
[129]
K. H. Gayer and L. Woontner: The solubility of ferrous hydroxide and ferric hydroxide in
acidic and basic media at 25, Journal of Physical Chemistry, 1956, 60, 1569-71.
[130]
H. Ehlert and W. Hempel: ber die lslichkeit einiger salze, Z. Elektrochem, 1912, 18, 727.
[131]
J. R. Baylis: Factors other than dissolved oxygen influencing the corrosion of iron pipes,
Journal of Industrial and Engineering Chemistry, 1926, 18(4) 370-80.
[132]
B. Lan and J. Jia: Carbonate solubility in water at different pH values, Xi'an Dizhi Xueyuan
Xuebao, 1996, 18(2), 80-86.
[133]
F. T. Gould: Ferrous carbonate, US Patent 3416883, 1968.
[134]
O. Haehnel: Solubilities and properties of the carbonates of strontium, barium, and heavy
metals in water containing carbon dioxide under pressure, Journal fuer Praktische Chemie
(Leipzig), 1924, 108, 187-93.
[135]
R. Brauns: Chemische Mineralogie, Leipzig p. 398, 1896.
[136]
R. Lyden: Studies on the solubilities of carbonate minerals and carbonates in carbonic acid
solutions, Suomen Kemistiseuran Tiedonantoja, 1925, 34, 72-9.
[137]
G. I. Teodorovich and T. M. Shvedova: Solubility of ferrous and other carbonate minerals in

hydrochloric acid, Biull. Moskov. Obshchestva Ispytatelei Prirody, Otdel. Geol., 1960, 35(4),
100-4.
[138]
R. C. Wells: The fractional precipitation of som ore-forming compounds at moderate

temperatures, Geol. Surv. Bull. (U. S.) 609, 1915.
[139]
P.R. Konz: Method of gas purification and removal of ferrous carbonate from the absorption
solution, US patent 3264056, 1966.
[140]
J. Babcan: Modelling of the low-temperature metasomatism of carbonates by siderite,

Geologicky Zbornik--Geologica Carpathica, 1974, 25(1), 53-63.
[141]
J. Gnzburg: ber die Darstellung und die Eigenschaften von Ferrobikarbonatlsungen,

Inaugural dissertation,Karlsruhe, 1911.
[142]
T. R. Andersen and A. Valle: Prediction of top of the line corrosion in pipelines carrying wet
CO2 containing gas, CORROSION/93, NACE; 1993, paper no. 73.
[143]
L. G. Berg, and K. A. Buzdov: Synthesis and thermal dissociation of iron(II) carbonate,

Zhurnal Neorganicheskoi Khimii, 1961, 6, 2003-8.
[144]
W. C. Reynolds: Chemical properties of concentrated solutions of certain salts. Part I. Double

potassium carbonates, Journal of the Chemical Society, Transactions, 1898, 73, 262-267.
[145]
P. E: Rosenberg and F. F. Foit Jr.: The stability of transition metal dolomites in carbonate
systems: a discussion, Geochim. Cosmochim. Acta, 1979, 43, 951.
[146]
L. Chai and A. Navrotsky: Synthesis, characterization, and energetics of solid solution along
the dolomite-ankerite join, and implications for the stability of ordered CaFe(CO3)2,
American Mineralogist, 1996, 81, 1141-1147.
[147]
P. M. Davidson: Ternary iron, magnesium, calcium carbonates: a thermodynamic model for

dolomite as an ordered derivative of calcite-structure solutions, American Mineralogist, 1994,
79(3-4), 332-9.
- 185 -
197

[148]
E. Erdoe and H. Altorfer: A basic iron carbonate similar to malachite as a corrosion product
of steel, Werkstoffe und Korrosion, 1976, 27(5), 304-12.
[149]
S. Al-Hassan, B. Mishra, D. L. Olson and M. M. Salama: Effect of microstructure on

corrosion of steels in aqueous solutions containing carbon dioxide, Corrosion, 1998, 54(6),
480-493.
[150]
S. Music, I. Nowik, M. Ristic, Z. Orehovec and S. Popovic: The effect of

bicarbonate/carbonate ions on the formation of iron rust, Croatica Chemica Acta, 2004, 77(12), 141-151.
[151]
W. Kamolpornwijit, L. Liang, G. R: Moline, T. Hart and O. R. West: Identification and

quantification of mineral precipitation in Fe0 filings from a column study, Environmental
Science and Technology, 2004, 38(21), 5757-5765.
[152]
R. de Marco, Z.-T. Jiang, B. Pejcic and E. Pionen: An In Situ Synchrotron Radiation Grazing
Incidence X-Ray Diffraction Study of Carbon Dioxide Corrosion, Journal of the
Electrochemical Society, 2005, 152(10), B389-B392.
[153]
R. de Marco, Z.-T. Jiang, D. John, M. Sercombe and B. Kinsella: An in situ electrochemical

impedance spectroscopy/synchrotron radiation grazing incidence X-ray diffraction study of the
influence of acetate on the carbon dioxide corrosion of mild steel, Electrochimica Acta, 2007,
52(11), 3746-3750.
[154]
Z. Xia, K. C. Chou and Z. Szklarska-Smialowska: Pitting corrosion of carbon steel in carbon

dioxide-containing sodium chloride brine, Corrosion, 1989, 45(8), 636-42.
[155]
L. G: Berg, and K. A. Buzdov: Determination of heat effects in the dissociation of FeCO3,

Zhurnal Neorganicheskoi Khimii, 1962, 7, 1773-8.
[156]
H. Sontheimer, W. Kolle and V. Snoeyink: The siderite model of the formation of corrosion
resistant scales, American Water Works Association, 1981, 73, 572-579.
[157]
J. Babcan: Low-temperature synthesis of siderite, Geologicky Zbornik--Geologica

Carpathica, 1970, 21(1), 89-97.
[158]
H. Ehrhardt, H. C. Schober and H. Seidel: Hochdrucksynthesen von Carbonaten. III.

Darstellung von Carbonaten zweiwertiger Metalle durch Zersetzung ihrer Oxalate unter hohem
CO2-Druck, Z. Anorg. Allg. Chem., 1980, 465, 83.
[159]
H. Ehrhardt, H. Schweer and H. Seidel: Syntheses of some carbonates with carbon dioxide
under high pressures, Zeitschrift fuer Anorganische und Allgemeine Chemie, 1980, 462, 18598.
[160]
A. Lemke and W. Biltz: Zur Hydrothermalsynthese einiger Carbonate, Z. Anorg. all. Chem.,
1934, 220, 312.
[161]
E. E. Huxley and J. Fetchin: Process for producing ferrous carbonate and removing oil from
ammonia therewith, US patent, US3058810, 1962.
[162]
C. Hoy: Improvements in the manufacture of Iron carbonate, Great British patent, 7465,
1899.
[163]
J. K. Lilly: Production of ferrous carbonate, US patent 876366, 1908.
[164]
C. L. Parker: Improvements in & connected with the Production of Glauber Salts, Ferrouscarbonate, Ferrous-hydrate & Ammonia from Nitre-cake, Great British patent, 24639, 1903.
[165]
D. L. Graf: Crystallographic tables for the rhombohedral carbonates, Am. Mineralogist,

1961, 46, 1283-1316.
[166]
O. Baudisch and L. A. Welo: The aging of ferrous hydroxide and ferrous carbonate, Journal
of Biological Chemistry, 1925, 64, 753-70.
[167]
Von J. Krustinsons: Dissociation of ferrous carbonate, Z. anorg. allgem. Chem. 1935, 225,
93-96.
[168]
J. Gill: Manufacture of Ferric oxide, US patent 842452, 1911.
- 186 -
198

[169]
W. Johannes: Experimentelle Sideritbildung aus Calcit + FeCl2, Contrib. Mineral, Petrology

(Berlin), 1968, 17(2), 155-64.
[170]
W. E. Sharp: The cell constants of artificial siderite, American Mineralogist, 1960, 45, 2413.
[171]
A. Flgge: Pure ferrous carbonate, Germany patent 178878, 1905.
[172]
A. V. Gogolev, E. V. Zakharova, N. I. Rodygina, A. M. Fedoseev, and V. P. Shilov:

Reduction of neptunium(V) and uranium(VI) with iron(II) in bicarbonate solutions,
Radiochemistry, 2006, 48(3), 249-253.
[173]
J. K. Heuer and J. F. Stubbins: An XPS characterization of FeCO3 films from CO2 corrosion,
Corrosion Science, 1999, 41(7), 1231-1243.
[174]
W. G: Lyon: Process for preparing ferrous carbonate, US patent 4657752, 1987.
[175]
C.F. Allison, GM. Dreher and C. A. Trathowen: Production of ammonium sulfate and ferrous
carbonate from ferrous sulfate-containing solutions, US patent 2845332, 1958.
[176]
A. Flgge: Iron carbonate, Patent of France FR 371876, 1906.
[177]
A. Flgge: An improvement of manufacture of Ferrous Carbonate in the cold, Great British

Patent 27316, 1907.
[178]
K . H. Gayer, and L. Woontner: The stability of iron(II) hydroxide, Journal of Chemical

Education, 1957, 34, 178-9.
- 187 -
199
200
The results of the thermodynamic modelling of the bulk phase are shown in section 7.
This model may also be used for evaluating the thermodynamic factor, *ii, used in the
extended Nernst-Planck formulation defined by (62) and used in (114). The remaining
thermodynamic factors *ij for izj are not shown in this study since they are assumed
to be negligible. In the result shown here only *ii is illustrated. The reason is that
including more thermodynamic factors complicates the equation scheme and the
result can become difficult to interpret. This is illustrated by the following example:
Equation (83) may be written for a binary system:
d1
RT
*11x1 *1s xs z1F I
x1
RTxs s
J1
x1ct 1fs
(277)
Component 1 is the solute and component s is the solvent. The gradient of the solvent
xs is often close to zero since the solvent composition is almost constant in the
diffusion layer. Equation (277) shows that *1s is related to the gradient of xs. The
diffusion problem has not been solved in the figures shown below and therefore xs
has not been calculated. If *1s is assumed to be zero, then *11 is the correction factor
for the diffusivity at infinite dilution since the above rearranges to
J1s
ct
xzF
1,eff x1 1 1 1,eff I
xs
RTxs
(278)
Where the effective diffusivity is given by:

1,eff =1fs *11
(279)
*ii is therefore the correction factor of the diffusivity in order to account for nonideality of the liquid phase.
The calculation of *ii requires a thermodynamic activity coefficient model and a
speciation routine. The thermodynamic model gives the activities and the derivatives
with respect to T, P, and composition. The speciation routine calculates the
equilibrium composition from the given T, P, and salt amount. In this study,
parameter set B is used as determined in section 7. The rational symmetrical standard
state is used for water and the rational unsymmetrical standard state for the other
compounds.
- 189 -
201
7.8
1.3
7.8
1.2
7.3
1.2
7.3
1.1
6.8
1.1
6.8
0.8
0.5
5.3
4.8
Imax: 1.22
P CO2: 1bar
4.3
Temperature: 298.15K
0.2
0.4
0.9
0.8
0.7
5.8
5.3
4.8
Imax: 1.22
P CO2: 1bar
0.6
4.3
MEG Fraction: 0
0
6.3
H2O
CO2
MEG
+
Na
HCO
3
+
H
OH 2CO3
pH
0.6
0.8
ii
NaHCO3 mol/kg total
3.8
1.2
Saturation
by NaHCO3
MEG Fraction: 0
0.5
0.1
0.2
0.3
0.4
Na2CO3 mol/kg total
1.3
7.8
1.2
7.3
1.1
6.8
0.5
3.8
0.6
Saturation
by NaHCO3
6.3
H2O
CO2
MEG
Na+
HCO3H+
OHCO32pH
0.9
0.8
0.7
5.8
pH
0.7
5.8
pH
0.9
0.6
Thermodynamic factor,
6.3
H2O
CO2
MEG
Na+
HCO3H+
OHCO32pH
pH
ii
1.3
ii
5.3
4.8
Imax: 1.22
0.6
0.5
P CO2: 1bar
4.3
MEG Fraction: 0
0
0.2
0.4
0.6
0.8
NaOH mol/kg total
3.8
1.2
Saturation
by NaHCO3
Figure 53: Thermodynamic factors, *ii, as function of added pH-stabilizing salt, NaOH, NaHCO3, and
Na2CO3. The results of the three salts are equivalent since CO2 reacts in the liquid phase. Saturation by
NaHCO3 is reached in all three cases when salt concentration is high. Imax indicates the maximum ionic
strength obtained at saturation.
*ii will always approach one as the concentration goes to infinite dilution. Figure 53
shows *ii and pH of the Na2CO3-NaHCO3-MEG-H2O system plotted for 1 bar partial
pressure of CO2 from zero concentration of the three salts NaOH, NaHCO3, and
Na2CO3 at T=25C to precipitation of a salt. MEG concentration is zero. pH is
calculated using the molality of H+, bH , by:
pH
log bH
(280)
CO2 dissolves into the liquid phase and dissociates to HCO3 and CO32 by the
equilibria given by (201). The speciation routine is set up in order to calculate the
liquid phase composition at a given temperature and CO2 pressure at a specified
amount of salt added. Comparing the three plots in figure 53 demonstrates that there
is no difference between adding NaOH, NaHCO3 or Na2CO3.
- 190 -
202
7.8
1.3
8.3
1.2
7.3
1.2
7.8
1.1
6.8
1.1
6.3
0.9
5.8
0.7
Imax: 1.02
0.6
0.5
P CO2: 1bar
MEG Fraction: 0.2
0
0.2
5.3
H2O
CO2
MEG
Na+
HCO3H+
OH- 2CO3
pH
0.4
4.8
0.6
7.3
6.8
pH
6.3
0.9
5.8
0.8
H2O
CO2
MEG
Na+
HCO3H+
OH- 2CO3
pH
0.7
4.3
0.6
3.8
0.5
0.8
Saturation
by NaHCO3
Imax: 0.98
P CO2: 1bar
MEG Fraction: 0.4
0
0.2
0.4
NaHCO3 mol/kg total
5.3
4.8
4.3
3.8
0.6
0.8
Saturation
by NaHCO3
1.3
8.3
1.3
8.3
1.2
7.8
1.2
7.8
6.8
1
6.3
0.9
5.8
0.8
0.7
0.6
0.5
H2O
CO2
MEG
Na+
HCO
3
H+
OHCO32pH
Imax: 1.12
P CO2: 1bar
MEG Fraction: 0.6
0
0.2
5.3
ii
1.1
4.8
NaHCO3 mol/kg total
0.6
Saturation
by NaHCO3
6.8
1
6.3
0.9
5.8
0.8
H2O
CO2
MEG
Na+
HCO
3
H+
OHCO32pH
0.7
Imax: 1.82
0.6
4.3
3.8
0.4
7.3
1.1
pH
7.3
pH
ii
NaHCO3 mol/kg total
0.8
ii
1.3
pH
ii
0.5
0.8
P CO2: 1bar
MEG Fraction: 0.8
0
0.2
5.3
4.8
4.3
3.8
0.4
NaHCO3 mol/kg total
0.6
0.8
Saturation
by NaHCO3
Figure 54: Setup is similar to figure 53 for increasing amount of wt% MEG on a salt-free basis.
The outcome of the calculation is the exact same result since the routine adds CO2
until the defined pressure has been reached. This is similar to the real scenario in wet
gas pipelines. The liquid phase is very small compared to the gas phase. CO2 will
dissolve the first few hours of production. New CO2 is pumped through the pipeline
and the partial pressure will quickly reach steady state and remain constant. The liquid
phase becomes saturates with CO2. The amount of CO2 dissolved in the liquid is
directly related to the added amount of pH-stabilizing salt. Figure 53 also shows that
NaHCO3 will precipitate if the concentration of pH-stabilizer is too high.
The thermodynamic factors, *ii, shown in figure 53 of i= HCO3 is approximately one
below 0.3 mol NaHCO3/kg total. It illustrates that the effective diffusion coefficient is
considerable different from the infinite dilution activity coefficient at higher
concentration. It is 20% higher at saturation and the *ii of i=Na+ is 0.55 which
indicates the effective diffusivity is 55% of the value at infinite dilution. The
concentration of CO32 is low in this solution, 0.01molal, and *ii of i= CO32 is
consequently also close to one, here 1.02. The concentration of the remaining
components are close to infinite dilution and *ii of these compounds are close to
*ii=1. The ionic strength, Imax, of the systems shown in figure 53 is approximately 1.2
mol/kg H2O at saturation. This is an indication that the solutions behaves non-ideal.
- 191 -
203
Figure 54 demonstrates a calculation similar to figure 53. The effect of increasing

amount of MEG is shown. Concentration of 20 wt% to 80 wt% MEG on a salt-free
basis is given where figure 53 shows for 0 wt% MEG. It illustrates how *ii of
i= HCO3 changes from 1.20 to 0.9 at saturation by addition of MEG. It also show that
*ii of i=Na+ changes from 0.55 towards 0.9. This indicates the diffusion process is
more ideal in the pure MEG. MEG lowers the activity of water and in addition it
decreases the diffusivity of HCO3 , as figure 54 shows. This could explain why MEG
reduces corrosion since the diffusion process of HCO3 is brought noticeably down.
The figure also shows that the effective diffusion coefficient of MEG it self in a 20
wt% MEG solution is 1.2, decreasing slightly with salt concentration. It is noteworthy
that the ionic strength is 1.8 mol/kg H2O at high MEG concentration due to the low
amount of water present.
1.4
8.3
1.4
8.8
1.3
7.8
1.3
8.3
7.3
1.2
7.8
1.1
7.3
ii
0.8
0.7
0.6
6.3
H2O
CO2
MEG
+
Na
HCO3+
H
OHCO32pH
0.9
Imax: 1.86
5.8
5.3
4.8
0.8
0.7
0.6
P CO2: 1bar
0.5
0.4
MEG Fraction: 0
0.5
Imax: 3.27
6.3
5.8
5.3
4.8
P CO2: 1bar
6.8
H2O
CO2
MEG
Na+
HCO
3
+
H
OHCO32pH
0.9
1.5
NaHCO3 mol/kg total
2
Saturation
by NaHCO3
4.3
0.5
3.8
0.4
2.5
4.3
MEG Fraction: 0
0
0.5
3.8
1
1.5
NaHCO3 mol/kg total
2.5
Saturation
by NaHCO3
Figure 55: The effect of temperature on the thermodynamic factors, *ii. Shown for 50 and 75C.
Solubility increases but profiles are almost unchanged.
Figure 55 reveals the effect of increasing temperature. The profiles are similar except
that the solubility increases. By comparing figure 53 and the plots in 55 shows that the
thermodynamic factor *ii of i= HCO3 and i=Na+ are almost unchanged, but *ii of
i= CO32 increases with temperature. Figure 55 illustrates that the effective diffusivity
of CO32 at 75 is 1.4 at saturation. This indicates that carbonate diffuses much faster
close to saturation at high temperature. The high transport of carbonate could have an
unexpected influence on the build-up of protective FeCO3 layers.
- 192 -
204
pH
6.8
1
1.1
pH
ii
1.2
8.8
1.3
1.2
1.2
7.8
7.8
1.1
H2O
CO2
MEG
Na+
HCO
3
+
H
OHCO32pH
0.8
0.7
0.6
pH
0.9
6.8
5.8
Imax: 1.72
1
6.8
0.9
0.8
0.7
0.6
4.8
Imax: 1.16
P CO2: 0.1bar
0.5
0.4
5.8
4.8
P CO2: 5bar
0.5
MEG Fraction: 0
0
H2O
CO2
MEG
Na+
HCO
3
+
H
OHCO32pH
pH
ii
ii
1.1
3.8
0.5
MEG Fraction: 0
0.4
1.5
Saturation
by NaHCO3
NaHCO3 mol/kg total
3.8
0.5
NaHCO3 mol/kg total
Saturation
by NaHCO3
1.5
8.8
1.3
1.2
7.8
ii
1.1
1
6.8
0.9
H2O
CO2
MEG
Na+
HCO3H+
OH- 2CO3
pH
0.8
0.7
0.6
Imax: 1.05
pH
8.8
1.3
5.8
4.8
P CO2: 30bar
0.5
0.4
MEG Fraction: 0
0
3.8
0.5
NaHCO3 mol/kg total
Saturation
by NaHCO3
1.5
Figure 56: The effect of partial pressure of CO2 on the thermodynamic factors, *ii. High pressure
lowers the thermodynamic factor of carbonate.
Figure 56 show the sensitivity to pressure. It compares to figure 53 at 1 bar CO2. The
high CO2 pressure sours the liquid phase as observed on the pH axis. The pH is below
7 at pCO2 =5 and 30 bar and therefore CO32 has dissociates to HCO3 . Therefore *ii of
i= CO32 decreases with increasing pCO2 . *ii of i= HCO3 and i=Na+ remain almost
unchanged. The high CO2 pressure lowers the solubility limit since the activity of
HCO3 has increased. The figure also shows that CO2 pressure has a relatively high
effect on *ii at low pressures but limited effect at high pressure. At high pressure the
change in the thermodynamic factor of CO2 becomes noticeable.
The results shown here proves that a thermodynamic activity coefficient model can be
used for evaluating the thermodynamic factors and thereby improving the results of a
diffusion model. The correction factor of the diffusion coefficient is in the order of 0.4
to 1.6. It may be agued that this is within the experimental accuracy of the determined
diffusion coefficients. If an activity coefficient is used for the bulk calculations, there
is no reason not to include it in the diffusion model. One of the reasons why this is not
done is probably because of the lack of theory. In this work a comprehensive set of
- 193 -
205
equations are presented which may be used for setting up the diffusion problem as
function of the activity coefficient model.
- 194 -
206
Conclusion
10 Conclusion
The main objective of this work has been to improve the understanding of the CO2
corrosion mechanism. The focus has mainly been wet gas pipelines for transportation
of natural gas offshore.
An extensive mathematic theory is presented in order to improve existing mechanistic
CO2 corrosion models.
It is shown in brief how to perform a full thermodynamic bulk calculation.
A detailed scheme of the multi component electrolyte diffusion process is given.
Several setups for implementation of the diffusion and thermodynamic models are
given. The theory of extending the Nernst-Planck theory is given and expressed in
terms of thermodynamic factors which have not been presented previously.
A precise representation of the electrochemical kinetics is given. Suggestions for
enhancing the Stern and Geary theory is presented and thereby enhancement the
theory behind linear polarization resistance and corrosion measurements. The
alternative extensions of the Stern and Geary theory are compared. The result show
that the original Stern and Geary theory is suited for normal activation controlled
kinetics, but it over predicts considerably if the corrosion process is affected by
limiting current densities.
An extensive review of CO2 literature is presented and a review of existing
mechanistic CO2 corrosion models is given.
Literature has shown that FeCO3 plays an important part in the corrosion process
especially the solubility if FeCO3. There is a lack of accuracy on existing
thermodynamic properties of FeCO3. This indicates that the calculated solubility and
temperature dependence of the solubility has been imprecise. This is due to the
properties of Fe2+. These vary significantly depending of the atmosphere of either H2
or vacuum. Existing properties are validated and a consistent set of FeCO3 properties
are given.
There are a great number of mechanistic CO2 corrosion models in the open literature.
The majority of models assume ideal thermodynamic condition. This is a crude
assumption since the ionic strength is up to 10 mol/kg H2O in certain pipelines. A
thermodynamic model is build in order to improve the thermodynamic description.
Parameters for the extended UNIQUAC model are presented for the extension to
MEG calculations in the multi solvent electrolyte system: CO2-NaHCO3-Na2CO3MEG-water. Parameters are regressed to literature values of SLE, VLE, heat excess,
and validated for heat capacity data. 212 experimental SLE data points were measured
in this work and used in the thermodynamic modelling. The determined parameters
are valid between -50 and 90 C. Measurements were conducted at 2 to 60 C at
atmospheric conditions. A method called the reverse Schreinemakers method was
developed during the experimental work. It improves the accuracy of the solubility
measurements and reduces the amount of experimental work needed since it gives
almost the same information as a new experiment. The method may also be used for
obtaining information on the unknown solid phase, just like the original
Schreinemakers method.
A model of the density in the NaCl-NaHCO3-Na2CO3-MEG-water system is given
from literature values and experimental work measured in this study. The model is
valid between 2 and 60C.
- 195 -
207
Conclusion
The thermodynamic model is a general model and it is not restricted to the corrosion
system. It may be used for calculating the activity coefficient in the CO2-NaHCO3Na2CO3-MEG-water system. The system is for example found in some gas drying
facilities and other closed CO2 loops. The model may additionally be used for scale
prediction CO2 capture calculations. The representation of the phase equilibria by this
model is good.
A database of approximately 3500 data points on phase equilibrium measurements in
systems containing CO2, CH4, MEG, DEG, TEG, and water was collected and
evaluated during the modelling.
The effect of applying the above thermodynamic model in a diffusion model is
shown. The thermodynamic factors have been calculated in the system CO2-NaOHNaHCO3-Na2CO3-MEG-water. The results show that the diffusivities of Na+, HCO3 ,
CO32 , MEG, and CO2(aq) may vary 60% up or down compared to the infinite dilute
diffusivities. The outcome is independent of whether NaOH, NaHCO3, or Na2CO3 is
used. The effect on Na+ diffusivity is high in all cases and MEG is affected at low
MEG concentrations. Diffusivity of HCO3 is only affected by the MEG
concentration. The system seems to perform more ideally in term of diffusion at high
MEG concentrations. CO32 diffusivity is a strong function of temperature and partial
pressure of CO2. Diffusivity of CO2(aq) is also effected at high CO2 pressure.
The work done in this thesis has shown that all mechanistic CO2 corrosion models
may be improved by applying a thermodynamic model. It will involve some
difficulties in the mathematical description. This will improve the capabilities of the
models to predict corrosion and to extrapolate to unknown system at high ionic
strength.
- 196 -
208
Future, short term long term
11 Future, short term long term

The future work would be to include FeCO3 in the extended UNIQUAC model. Fe2+
is already in the aqueous model, but the interaction between Fe2+-CO2, Fe2+- HCO3 ,
Fe2+- CO32 , and Fe2+-MEG needs to be fitted. Results for the effect of MEG on
FeCO3 would be interesting to investigate. This would require minimum work since
the solubility of FeCO3 was measured in this work both in water and in MEG-water,
though not given in this thesis. This may show that the solubility of FeCO3 is lower in
MEG and MEG therefore enhances the corrosion protection.
It could be interesting to see the effect of DEG compared to MEG. Similarly the effect
of methanol would be obvious to investigate since some of the industry is using the
cheaper methanol. The extended UNIQUAC model has already been regressed for
methanol. Showing the thermodynamic factors in the water system would be an
interesting issue and to compare to MEG. It may involve a new regression of the
methanol parameters since CO2 has never been fitted to methanol.
Natural gas has been considered not to influence the equilibria in this study. It may be
investigated by including CH4 in the model. It would be an easy task since
experimental data has already been collected in this work. The procedure is only
interesting from a scientific point of view, since the effect of CH4 is expected to be
very low.
An obvious task would be to implement a full diffusion model which includes the
extended UNIQUAC model. The objective would be to show the effect of the
extended UNIQUAC model on the predicted corrosion rates. It would be interesting
to see the effect of the thermodynamic model compared to the same corrosion model
at ideal conditions. The OLI thermodynamic model has been used in a public study to
create an advanced electrochemical model. It would be interesting to see the effect of
implementing their model in a full diffusion scheme which include a
thermodynamically correct bulk calculation and the thermodynamic factors in the
diffusion model. This may be very difficult to obtain since OLI is very restrictive
towards their internal computer code.
The electrochemical kinetics of CO2 corrosion is understood to some extended, but
there are still improvements to be made, especially the CO2 corrosion mechanism in
combination with H2S, O2 and acetic acid. The accuracy of the corrosion models
always rely on the precision of the core electrochemical model. A mathematically
model has still not been presented.
The corrosion literature and the thermodynamic property collections have reached a
slight inconsistency level. The corrosion literature has accepted the existence of
H2CO3, but it has been rejected by thermodynamic databases. This issue needs to be
addressed in order to build consistent corrosion models.
The predictive CO2 corrosion models are currently modelled as 1D discretized PDEs.
The corrosion models of the future will and should be implemented in more general
flow simulation scheme. One of the difficult tasks in the future will be to reliably and
easily predict pitting and crevice corrosion. It would be a task to implement corrosion
models directly in process simulators in order the quickly determine whether the
equipment could withstand corrosion or which alloy would be optimal for the specific
operating conditions.
- 197 -
209
210
12 List of figures and tables

12.1 Figures
Figure 1: An overview of the gas transportation systems.............................................1
Figure 2: Ideal calculation of FeCO3 solubility in water. The liquid composition
at pH >8 is a hypothetical since NaHCO3 is most likely to precipitate. It
was not included in the above calculations....................................................2
Figure 3: Mol balance of component i over time........................................................13
Figure 4: At the anode, metal is dissolved by oxidation of Fe; the electrons are
transported trough the metal to the cathode area where they go into
reduction of other species. ...........................................................................29
Figure 5: Current density produced and consumed by a half cell reaction.................30
Figure 6: Current density of two half cells. ................................................................31
Figure 7: The net current density, inet, is a sum of the anode and cathode current
densities, ia and ic. Plotted for i0=1PA/cm2, D=0.3. Note how K=0 at
inet=0PA/cm2 where (S-I)eq=-0.05V. The thick dashed line shows the
slope of inet at (S-I)o(S-I)eq also known as the polarization resistance,
Rp..................................................................................................................36
Figure 8: Showing the same as figure 7 just represented in a numerical log plot of
the current density........................................................................................36
Figure 9: Parameters are the same as figure 7, except that iRd =5PA/cm2 and
iXd =-4PA/cm2. The dashed line is the old line of figure 7, the smalldotted line obtained by the influence of id. ..................................................45
Figure 10: Obtained by the same parameters as figure 9. It shows the effect of
limiting currents which was not shown in figure 8......................................46
Figure 11: Shows a plot of the net anodic and cathodic currents densities (blue)
and the net current density (red). Equilibrium potential are r0.2, anodic
and cathodic limiting currents are -80PA/cm2 and 200PA/cm2. Exchange
currents are 1PA/cm2 and 2PA/cm2, Ds are 0.25 and 0.5...........................49
Figure 12: Overview of the topics discussed in the CO2 corrosion literature.............51
Figure 13: Cross section of pipeline showing the condensation of water which
influences the corrosion process under the pipeline roof. ........................53
Figure 14: Illustrates mesa attack. A: High flow rate. B: Iron dissolves under
corrosion product. C: Corrosion product breaks off and more iron
dissolves. Corrosion product breaks off continuously.................................54
Figure 15: The CO2 corrosion mechanism. CO2 dissolves in the condensed
aqueous liquid phase diffuses to the surface and dissolves iron by an
electrochemical mechanism. ........................................................................57
Figure 16: The CO2 corrosion mechanism. If pH is high enough and if the
temperature and liquid composition is correct, then FeCO3 precipitate
and forms a diffusion controlled mechanism which inhibits the
electrochemical corrosion process. ..............................................................58
Figure 17: Principles of a discetized CO2 corrosion model. Green lines are
boundaries. Blue lines are discretization. Compartment numbered j,
thickness 'x, and total thickness GN.............................................................59
- 199 -
211
Figure 18: Principles of Sundaram et al.s138 CO2 corrosion model which includes
three diffusion models. The distances to the interfaces are GP, GN, and Gb
from the surface. ..........................................................................................61
Figure 19: Principles of Dayalan et al.s143 CO2 corrosion model which includes
two diffusion models. The distances to the interfaces are GP and GN from
the surface. ...................................................................................................62
Figure 20: Principles of Nesic et al.s154,156,155 CO2 corrosion model which
includes one diffusion model. The distances to the interfaces are GP and
Gb from the surface. The diffusion coefficients are position dependent. GP
signifies a change in porosity and not a boundary.......................................63
Figure 21: An overview of the components in the CO2-Na2O-H2O system. Black
line is the ternary Na2CO3-NaHCO3-H2O system and the grey dotted
line is the CO2-NaOH-H2O system. .............................................................79
Figure 22: The seven equilibrium cells connected in parallel to the
heating/cooling unit. ....................................................................................83
Figure 23: Sketch of a SLE equilibrium cell. .............................................................84
Figure 24: The analysis unit consisting of to titration cells in the left part of the
picture and the computer on the right for data acquisition and burette
control. The ABU unit is seen behind the titration cells..............................84
Figure 25: A regular titration curve of a 0.3 g sample saturated with NaHCO3 in a
63 wt % MEG solution at 25 C. .................................................................88
Figure 26: Phase diagram and the lever rule used in the reverse Schreinemakers
method. L(saturated liquid), P(solid precipitate), and I(initial solution).
The diagram shows the area marked by black lines in figure 21.................90
Figure 27: Stable and meta-stable data at 20 C in the Na2CO3-NaHCO3-MEGH2O system. , : NaHCO3. , , : Trona. , : Na2CO310H2O. :
Na2CO3H2O. Dotted line: meta-stable trend.............................................101
Figure 28: Density of MEG-H2O solutions. 'UMEG as function of salt free MEG
wt. fraction. ................................................................................................107
Figure 29: Density of MEG-H2O solution as function of temperature and salt free
MEG wt. fraction. ......................................................................................108
Figure 30: Density of Na2CO3-NaHCO3-MEG-H2O solution as function of
sodium content...........................................................................................109
Figure 31: Density of saturated NaHCO3-MEG-H2O solution determined using
equation (244). Composition of the saturated solution determined using
the extended UNIQUAC model.................................................................109
Figure 32: Density of saturated Na2CO3-MEG-H2O solution determined using
equation (244). Composition of the saturated solution determined using
the extended UNIQUAC model.................................................................110
Figure 33: Density of NaCl-MEG-H2O solutions determined by equation (244) at
25 and 50 C. .............................................................................................111
Figure 34: Comparison of calculated and experimental density of NaCl-Na2CO3NaHCO3-MEG-H2O solutions between 2 to 60 C. ..................................111
Figure 35: The temperature dependence of Na2CO3 solubility in water. Lines
calculated with the A and the B parameters. Experimental data are from
the IVC-SEP database10. ............................................................................113
Figure 36: Temperature dependence of the two-salt lines in the aqueous Na2CO3
NaHCO3 system. Phase lines calculated with the A and the B
parameters. Experimental data are from the IVC-SEP database10.............114
- 200 -
212
Figure 37: Total vapor pressure of the mixed solvent MEG-H2O system using
parameter set B. The sources of the experimental data are listed in table
14. ..............................................................................................................119
Figure 38: Comparison of experimental and calculated properties using parameter
set A. 1: Comparison of VLE data. 2: Comparison of excess enthalpy. 3:
Our SLE data compared to predictions by the model. 4: Model
prediction of literature data........................................................................121
Figure 39: Comparison of experimental and calculated properties using parameter
set B 1: Comparison of VLE data. 2: Comparison of excess enthalpy. 3:
Our SLE data compared to predictions by the model. 4: Model
prediction of literature data........................................................................122
Figure 40: Excess enthalpy data of the mixed solvent MEG-H2O system using
parameter set B. The sources of the experimental values are listed in
table 15.......................................................................................................123
Figure 41: Model predictions and experimental heat capacity for the mixed
solvent MEG-H2O system. Experimental data obtained from various
authors93,94,46,53,51. Table 11 shows the parameters of the used heat
capacity correlation (215). Mixture propertied calculated using
parameter set B...........................................................................................124
Figure 42: SLE curve for the mixed solvent MEG-H2O system. The sources of
experimental values are listed in table 16. Model calculated using
parameter set B...........................................................................................125
Figure 43: CO2 solubility in the mixed solvent MEG-H2O system. The sources of
the experimental values are listed in table 14. Model calculated using
parameter set B...........................................................................................126
Figure 44: Solubility of NaHCO3 in the mixed solvent MEG-H2O system. The
sources of experimental data are given in table 17. Model calculated
using parameter set B.................................................................................127
Figure 45: Solubility of Na2CO3 in the mixed solvent MEG-H2O system. The
sources of experimental data are given in table 17. Model calculated
using parameter set B.................................................................................128
Figure 46: Experimental and calculated solubility of Na2CO3 and NaHCO3 in the
mixed solvent system MEG-H2O with 3 wt % MEG. Experimental data
were obtained in this study. Model calculated using parameter set B.
Thin black lines: NaHCO3. Thick black: NaHCO310H2O. Thin grey:
trona. Thick grey: NaHCO3H2O. ..............................................................129
Figure 47: Experimental and calculated solubility of Na2CO3 and NaHCO3 in the
mixed solvent system MEG-H2O with 15 wt % MEG. Experimental data
were obtained in this study. Model calculated using parameter set B.
Thin black lines: NaHCO3. Thick black: NaHCO310H2O. Thin grey:
trona. Thick grey: NaHCO3H2O. ..............................................................130
Figure 48: The influence of temperature and MEG concentration on the
precipitation fields in the Na2CO3-NaHCO3-MEG-H2O system.
Description of areas given in figure 36......................................................131
Figure 49: The calculated infinite dilution activity coefficient of MEG in water
compared to literature values of Suleiman and Eckert103. .........................132
Figure 50: Comparison of predicted and experimental solubility of Grtner et al.6
between 50 and 90 C in 50 wt% MEG. Deviation indicated by ABCD.
: Exp. NaHCO3 solubility. : Exp. Na2CO3H2O solubility. : Exp. 2salt solubility of NaHCO3-wegscheiderite. : Exp. single salt trona
- 201 -
213
solubility. : Exp. 2-salt solubility of trona-Na2CO3H2O. Red circle:

Exp. Trona-Na2CO3 solubility. ..................................................................134
Figure 51: The standard state heat capacity of FeCO3(s) as function of
temperature. ...............................................................................................146
Figure 52: Negative base-ten logarithm to the equilibrium constant, pKsp, of
reaction (252) as function of temperature determined by various authors
in the literature. Lines are correlations, symbols are single values. ..........158
Figure 53: Thermodynamic factors, *ii, as function of added pH-stabilizing salt,
NaOH, NaHCO3, and Na2CO3. The results of the three salts are
equivalent since CO2 reacts in the liquid phase. Saturation by NaHCO3
is reached in all three cases when salt concentration is high. Imax
indicates the maximum ionic strength obtained at saturation....................190
Figure 54: Setup is similar to figure 53 for increasing amount of wt% MEG on a
saltfree basis...............................................................................................191
Figure 55: The effect of temperature on the thermodynamic factors, *ii. Shown
for 50 and 75C. Solubility increases but profiles are almost unchanged. 192
Figure 56: The effect of partial pressure of CO2 on the thermodynamic factors,
*ii. High pressure lowers the thermodynamic factor of carbonate. ...........193
Figure 57: Figure showing the principle of the tortuosity. .......................................223
- 202 -
214
12.2 Tables
Table 1: Composition of mild carbon steel, X65, used for pipelines............................1
Table 2: Composition of a typical natural gas transported in pipelines in focus of
this study. .......................................................................................................2
Table 3: Comparison of the Stern and Geary equations applied for the case
presented in figure 11. .................................................................................49
Table 4: Gas-phase properties used in the SRK equation...........................................80
Table 5: Purity of the used chemicals. ........................................................................83
Table 6: Overview of analytical methods used in the literature for analysing the
CO2-Na2O-H2O system. ...............................................................................86
Table 7: Experimental initial composition, equilibrium composition, results of the
reverse Schreinemakers method and solid phase information from the
Na2CO3-Na2CO3-MEG-H2O system at 1 bar and 2 C to 60 C. ................95
Table 8: Experimental initial composition, meta-stable composition, results of the
reverse Schreinemakers method and solid phase information from the
Na2CO3-Na2CO3-MEG-H2O system at 1 bar and 20 C to 30 C. ............103
Table 9: Correlation parameters for equation (246) and (247). ................................112
Table 10: Molar mass, M, extended UNIQUAC parameters of set A and B,
standard state formation properties and correlation parameter for the
heat capacity equation (215) for liquid species. ........................................116
Table 11: Standard state properties of set A and B. Correlation parameter for the
heat capacity equation (215).Values not marked are from NIST12............117
Table 12: uijo u oji in K, extended UNIQUAC parameters for parameter set A and
B .................................................................................................................118
Table 13: uijt u tji in K2, extended UNIQUAC interaction energy parameters for
parameter set A and B ................................................................................118
Table 14: Used experimental MEG-H2O and CO2-MEG-H2O VLE data.................120
Table 15: Used experimental MEG-H2O data on heat excess and heat of solution..123
Table 16: Used experimental freezing point depression data of MEG-H2O. ............129
Table 17: Used experimental salt solubility data in Na2CO3-NaHCO3-MEG-H2O ..130
Table 18: Standard state properties of CO32- listed in databases related to FeCO3
at 298.15K and 1 bar..................................................................................147
Table 19: Frequently used standard state properties of Fe2+ at 298.15K and 1 bar..148
Table 20: Standard state formation properties of FeCO3 at 298.15K and 1 bar. ......150
Table 21: Literature, 'r,spH, speciation scheme, and pKsp, of reaction (252) at 25
C and 1 bar. ..............................................................................................154
Table 22: Literature, 'r,pspH, speciation scheme, and pKpsp, of reaction (253) at
25 C and 1 bar. .........................................................................................162
Table 23: Properties related to FeCO3(aq), reaction (271).......................................163
Table 24: Properties related to FeHCO3+(aq), reaction (272). .................................164
Table 25: Properties related to Fe(CO3)22-(aq), reaction (273).................................165
Table 26: Alternative speciation used for FeCO3 modelling....................................167
Table 27: Speciation used by a selection of ferrous hydroxide studies. ...................168
Table 28: List of experimental literature data on aqueous FeCO3 solubility............170
Table 29: Patents on synthesis of FeCO3..................................................................174
Table 30: Methods for synthesising FeCO3 as published in the literature................175
- 203 -
215
Table 31: Recommended standard state properties consistent with NIST55 at

298.15K at 1 bar. Properties are from NIST if not stated otherwise. ........176
Table 32: Recommended standard state properties consistent with CODATA57,58
at 298.15K at 1 bar.....................................................................................176
Table 33: Recommended properties of reaction (252) and (253), consistent with
NIST55 or CODATA57,58 at 298.15 K and 1 bar. .......................................177
Table 34: Correlation matrix of parameters fitted in parameter set A. .....................225
Table 35: Confidence intervals of parameters fitted in parameter set A...................225
Table 36: Correlation matrix of parameters fitted in parameter set B. .....................226
Table 37: Confidence intervals of parameters fitted in parameter set B...................226
Table 38: Number of interaction parameters for example systems. .........................231
- 204 -
216
Appendix
13 Appendix
A. Theory of electrolyte thermodynamics
A.1
Converting between activity coefficient types
The following gives a general relation for the chemical potential from a combination
of (23) and (13):
Pi
Pi4s RT ln siJ i4s si4
(281)
Piv x RT ln xiJ iv x xiv
(282)
Pi* x RT ln xiJ i* x xi*
(283)
Pi*m RT ln biJ i*m bi*
(284)
Pi*c RT ln ciJ i*c ci*
(285)
This section has its basis in eq. (282) to (285), here equations which relate J iv x , J i*c ,
J i*m and J i* x are derived.

The standard state concentration, si4 , is assumed 1 and therefore neglected in the
following equations.
Only three relations are needed to convert between the four different activities. Four
relations are given just to give a wider relation to other explicit variables. Equation
(306) and (311) are in principal related and should be taken as two different
representations of the same equation.
The equations derived in this appendix are listed here:
J i* x
J iv x
J ifx
(290)
J i*m
xsJ i* x
(303)
Us M t
M s Ut
(306)
U s mt
Ut ms
(311)
J i*c
J i* x
J i*c
J i*m
Nomenclature is found at the derived equations in the subsections below.

The above written as one gives:
- 205 -
217
Appendix
J i*m
xsJ i* x
xs
J iv x
J ifx
xsJ i*c
Ut M s
Us M t
J i*c
Ut ms
U s mt
(286)
Remember that if a special standard state is applied which does not use a standard
state concentration of si4 1[unit ] , then si4 will become part of the equations above.
A.1.1 Converting between rational symmetrical and rational

unsymmetrical activity coefficients
From (282) and (283) by equalizing gives:
Pi
Piv x Pi* x
RT
Piv x Pi* x
RT
Piv x RT ln xiJ iv x Pi* x RT ln xiJ i* x

x J *x
ln i iv x
xiJ i
J *x
ln iv x
Ji
(287)
At infinite dilution the rational unsymmetric activity coefficient approach one,

J i* x o 1 for xi o 0 . The rational symmetric activity coefficient is therefore very far
from its ideal situation of J iD x o 1 for xi o 1 . At infinite dilution it has a finite value,
defined by:
lim J iv x { J ifx
(288)
xi o0
J ifx is the rational symmetric activity coefficient at infinite dilution in the solvent.
At infinite dilution (287) is rewritten with (288) and J i* x o 1 to give:
Piv x Pi* x
RT
J *x
ln iv x
Ji
1
o ln fx
Ji
for
xi o 0
(289)
Because (287) must be valid also at infinite dilution, (289) and (287) is combined to
give:
J i* x
ln J
*x
i
J iv x
J ifx
ln J
vx
i
ln J
- 206 -
218
fx
i
(290)
Appendix
A.1.2 Converting between rational symmetrical and molal

From eq. (282) and (284):
Piv x RT ln xiJ iv x Pi*m RT ln biJ i*m
Pi
(291)
b J *m
ln i iv x
xiJ i
Piv x Pi*m
RT
The quantity bi xi is found from:

bi
xi
ni ns M s
ni nt
nt 1
ns M s
1
xs M s
(292)
where ni is the number of moles of i, ns and Ms is the number moles and mol weight
(kg/mol) of the solvent s. nt is the total number of moles in the phase. At infinite
dilution xs | 1 which from (292) gives:
bi
xi
1
1
o
xs M s
Ms
for
xi o 0
(293)
at infinite dilution (291) and (293) gives a relation equivalent to (288) which is valid
because J i* x o 1 for xi o 0 :
Piv x Pi*m
RT
b J *m
1
ln i iv x o ln
fx
x
J
i i
M sJ i
for
xi o 0
(294)
From (291) and (294) the relation between unsymmetric molal and rational
symmetrical activity coefficient is found because (291) must be valid also at infinite
dilution, using (292) gives:
biJ i*m
xiJ iv x
J i*m
1
M sJ ifx
xs
J iv x
J ifx
the above could easily be used in combination of (290) to give (303).
A.1.3 Converting between rational symmetrical and molar

From (282) and (285):
- 207 -
219
(295)
Appendix
Pi
Piv x Pi*c
RT
Piv x RT ln xiJ iv x Pi*c RT ln ciJ i*c

(296)
c J *c
ln i iv x
xiJ i
The quantity ci xi is found from:

ci
xi
ni Vt
ni nt
nt Ut
mt
Ut
(297)
Mt
ci is the concentration in(mol/L solution). Vt is the total volume (L) of the phase and
Ut is the total density in (kg/L solution) of the phase. mt is the total mass of matter
and M t is the number average molecular weight of solution (kg/mol) calculated as
Mt
xM
i
summed over all species in the phase.
At infinite dilution, the above reduces to the following as Vt o Vs for xi o 0 :

ci
n V
o i s
xi
ni ns
ns U s
ms
Us
Ms
for
xi o 0
(298)
U s is the density of pure solvent. An equivalent to (288) in combination with the

above gives the following at infinite dilution:
Piv x Pi*c
RT
c J *c
Us
ln i iv x o ln
fx
xiJ i
M sJ i
for
xi o 0
(299)
By (296), (297) and (299) the relation between the molar unsymmetrical and rational
unsymmetrical activity coefficient is found.
ciJ i*c
xiJ iv x
*c
i
Us
M sJ ifx
J iv x U s M t
J ifx Ut M s
(300)
because (296) must be valid at infinite dilution. The above should be seen in relation
to (290) and (306).
A.1.4 Converting between rational unsymmetrical and molal

From (283) and (284):
- 208 -
220
Appendix
Pi
Pi* x Pi*m
RT
Pi* x RT ln xiJ i* x Pi*m RT ln biJ i*m

(301)
b J *m
ln i i* x
xiJ i
at infinite dilution by the use of (293):
Pi* x Pi*m
RT
b J *m
1
ln i i* x o ln
xiJ i
Ms
for
xi o 0
(302)
From the combination of (292), (301) and (302) the relation between the molal
unsymmetrical and rational unsymmetric activity coefficient is found by:
biJ i*m
xiJ i* x
Ms
(303)
xJ
*m
i
*x
s i
because (301) must be valid at infinite dilution.

By taking a look at (290) it is obvious that (295) is the same as (303).
A.1.5 Converting between rational unsymmetrical and molar

From (283) and (285):
Pi
Pi* x Pi*c
RT
Pi* x RT ln xiJ i* x Pi*c RT ln ciJ i*c

(304)
c J *c
ln i i* x
xiJ i
which reduces to the following by the combination of (298) at infinite dilution:
Pi* x Pi*c
RT
c J *c
U
ln i i* x o ln s
xiJ i
Ms
for
xi o 0
(305)
(304) and (305) gives the relation between the molar unsymmetric and rational
unsymmetric activity coefficient using (297):
J i*c
J i* x
Us M t
M s Ut
because (304) must hold at infinite dilution.

By taking a look at (290) it is obvious that (300) is the same as (306).
- 209 -
221
(306)
Appendix
A.1.6 Converting between molal unsymmetrical and molar

From (284) and (285):
Pi*m RT ln biJ i*m Pi*c RT ln ciJ i*c
Pi
Pi*m Pi*c
RT
(307)
c J *c
ln i *i m
biJ i
the ratio ci bi is calculated from the relation, where ms is the mass of solvent and mt
is the total mass in the phase. Ut is the overall density of the phase:
ci
bi
ms Ut
mt
ni Vt
ni ms
(308)
At infinite dilution the volume of the phase is Vt o Vs as xi o 0 , which reduces the

above to:
ci
n V
o i s
bi
ni ms
Us
for
xi o 0
(309)
in combination with (307) at infinite dilution, this gives:
Pi*m Pi*c
RT
c J *c
ln i *i m o ln U s
biJ i
for
xi o 0
(310)
From equalization of (307) and (310) the relation between the molar unsymmetrical
and molal unsymmetrical activity coefficient is obtained:
ciJ i*c
biJ i*m
*c
i
Us
J
*m
i
U s mt
Ut ms
(311)
because (307) has to be true at infinite dilution.

m are the masses given in equation (308), which are related to the mole number and
molecular weights by ms ns M s and mt nt M t .
A.2
Relations between standard state fugacities
A relation between the standard state fugacities fi *m , f i * x , fi *c and fi D x is derived.

Equation (282) to (285) and links to the original relation of chemical potential and
- 210 -
222
Appendix
standard properties in (13) is used. The standard state concentration, siD , is assumed 1
and therefore neglected in the following equations.
The equations derived in this appendix are:
fi *x
fi *m
fi D x
f i D xJ ifx
(315)
fi D x M sJ ifx
(318)
U s f i *c
M sJ ifx
(321)
Equation (324), (327) and (330) are linear combinations of the above three and may
be derived from the above.
This section was originally written, in order to show the relation between Henrys law
constants at different scales. The above equations allow for the use of Henrys law to
determine the symmetrical standard state fugacity. The reader should remember that
all the above equations are functions of (T,P) and there are a new standard state for
every (T,P).
To summarize: The above equations written as one:
f i *m
U s f i *c
M s fi* x
M s fi D xJ ifx
(312)
Remember that if a special standard state is applied which does not use a standard
state concentration of siD 1[unit ] , then siD will become part of the equations above.
A.2.1 Relation between rational symmetrical and rational

unsymmetrical standard state fugacity
The chemical potential in the liquid phase can be expressed by the rational
symmetrical and rational unsymmetrical standard state. The chemical potential has to
be the same independent of the standard state. Using (13) for the same phase using
different standard and substituting the solution fugacities in (23) by activity
coefficients shows
fi
Dx
fi
PL
PiD x RT ln
fi
*x
fi
Pi* x RT ln
(313)
PiD x RT ln xiJ iD x xiD Pi* x RT ln xiJ i* x xiD

Rearranging (313) gives:
PiD x Pi* x
RT
J i* x
J iD x
RT ln xiJ i* x RT ln xiJ iD x
fi D x
fi* x
- 211 -
223
f
f
RT ln *i x RT ln Di x
f
f
i
i
(314)
Appendix
From (290) J i*x J iD x is substituted for J ifx

f i D xJ ifx
fi *x
(315)
A.2.2 Relation between rational symmetrical and molal

Using (13) for the same phase using different standard states and substituting the
solution fugacities in (23) by activity coefficients shows
PL
fi
Dx
fi
PiD x RT ln
P RT ln x J
Dx
i
fi
*m
fi
Pi*m RT ln
Dx
i i
*m
i
RT ln b J
*m
i i
(316)
which rearranges to:
PiD x Pi*m
RT
biJ i*m
xiJ iD x
RT ln biJ i*m RT ln xiJ iD x
f
f
RT ln *im RT ln Di x
f
f
i
i
fi D x
f i *m
(317)
From (295) the above rearranges to

fi D x M sJ ifx
fi *m
(318)
M s is the molar mass of the solvent in kg/mol.
A.2.3 Relation between rational symmetrical and molar

PL
fi
Dx
fi
PiD x RT ln
P RT ln x J
Dx
i
Dx
i i
this gives:
- 212 -
224
fi
*c
fi
Pi*c RT ln
P RT ln c J
*c
i
*c
i i
(319)
Appendix
PiD x Pi*c
RT
ciJ i*c
xiJ iD x
f
f
RT ln *i c RT ln Di x
f
f
i
i
RT ln ciJ i*c RT ln xiJ iD x

fi D x
f i *c
(320)
Using (300) the above results in

ciJ i*c
xiJ iD x
fi
Us
M sJ ifx
fi D x
f i *c
(321)
U s f i *c
M sJ ifx
Dx
M s is the molar mass of the solvent in kg/mol and U s is the density of the pure
solvent at (T,P) equal to the (T,P) at which the f i *c , J ifx and fi D x are found.
A.2.4 Relation between rational unsymmetrical and molal

PL
fi
*x
fi
Pi* x RT ln
P RT ln x J
*x
i
*x
i i
fi
*m
fi
Pi*m RT ln
*m
i
RT ln b J
*m
i i
(322)
Isolating Pi* x Pi*m RT gives
Pi* x Pi*m
RT
biJ i*m
xiJ i* x
RT ln biJ i*m RT ln xiJ i* x
f
f
RT ln *im RT ln *i x
f
f
i
i
(323)
fi* x
f i *m
From (303) this gives:

biJ i*m
xiJ i* x
fi
*m
fi*x
f i *m
1
Ms
M s fi
*x
M s is the molar mass of the solvent in kg/mol.
- 213 -
225
(324)
Appendix
A.2.5 Relation between rational unsymmetrical and molar

PL
fi
*x
fi
fi
*c
fi
Pi* x RT ln
P RT ln x J
*x
i
*x
i i
Pi*c RT ln
P RT ln c J
*c
i
(325)
*c
i i
Finding Pi* x Pi*c RT rearranges the above to:
Pi* x Pi*c
RT
ciJ i*c
xiJ i* x
f
f
RT ln *i c RT ln *i x
f
f
i
i
RT ln ciJ i*c RT ln xiJ i* x
(326)
fi*x
f i *c
Some reference to how Us/Ms came about (see my master thesis)

ciJ i*c
xiJ i* x
Us
fi* x
f i *c
Ms
Us
fi*x
(327)
f i *c
Ms
M s is the molar mass of the solvent in kg/mol and U s is the density of the pure
solvent at (T,P) equal to the (T,P) at which the fi *c is found.
A.2.6 Relation between molal unsymmetrical and molar

PL
fi
*m
fi
Pi*m RT ln
P
*m
i
RT ln b J
*m
i i
This results in the following
- 214 -
226
fi
*c
fi
Pi*c RT ln
P RT ln c J
*c
i
*c
i i
(328)
Appendix
Pi*m Pi*c
RT
ciJ i*c
biJ i*m
RT ln ciJ i*c RT ln biJ i*m
f
f
RT ln *i c RT ln *im
f
f
i
i
(329)
fi*x
f i *m
From (311) this give

ciJ i*c
biJ i*m
fi
f i *m
f i *c
Us
U s fi
*m
(330)
*c
U s is the density of the pure solvent at (T,P) equal to the (T,P) at which the f i *c is
found.
A.3
Converting standard properties between standard states
Converting between standard states is performed using (14):
Pi4 s Pi4 s
2
RT ln
fi 42 s
f 41s
(14)
The relations between standard state fugacities, fi 4i s , is given by (315), (318) and
(321) and the fractions f 42 s f 41s are basically a small rearrangement of the standard
i
state fugacities. :
fi*x
fi D x
f i *m
fi D x
fi D x
f i *c
J ifx
(315)
M sJ ifx
(318)
Us
(321)
M sJ ifx
If we for example had a huge database of standard state data in molal unsymmetrical
standard state, fi *m , at a reference T D , P D , like the NIST database. Take that we want
to convert it to rational symmetrical standard state, fi D x , at the same T D , P D we
would insert (318) in (14)
PiD x
Pi*m RT ln
P
*m
i
fi D x
f i *m
RT ln M J
fx
s i
- 215 -
227
(331)
Appendix
This indicates that by subtract RT ln M sJ ifx from all our molal unsymmetrical
standard state properties will convert those into rational symmetrical standard state. A
drawback to this approach is that a model of J ifx at T D , P D is needed.
To convert standard state data to other reference temperatures would require standard
state properties.
- 216 -
228
Appendix
B. Theory of electrochemistry
B.1
Isolation of net current in equation (171)
Assume:
QX
QR 1
Assume also that one reactant and one product is found in (123). This lead to the
following net current density of the involved redox reaction from (171):
i
1 D nFK
i
i
D nFK
i0 1 d exp
1 d exp
i
RT
i
RT
R
X
Isolating of the exchange current density and the limiting current densities gives:
i
i0
1 D nFK i
1 D nFK
D nFK i
D nFK
exp
d exp
exp
d exp
RT
i
RT
RT
i
RT
X
Moving the current densities to the left side gives:

1 1
1 D nFK 1
1 D nFK
D nFK
D nFK
i d exp
d exp
exp
exp
i i
RT
i
RT
RT
RT
0 R
The final equation of the net current density as function of the limiting currents and
the overpotential:
B.2
1 D nFK
D nFK
i0 exp
exp
RT
RT
1
1 D nFK 1
D nFK
1 i0 d exp
d exp
RT
RT
iX
iR
(172)
Derivation of equation (163)
It can be shown that the SG theory can be extended to a more general scheme by
using the net following net current density instead of (147).
inet
1 D1 n1 FK1
D1n1 FK1
i01 exp
exp
RT
RT
1 D 2 n2 FK2
D 2 n2 FK2
i02 exp
exp

RT
RT
- 217 -
229
(162)
Appendix
which rearranges to:

inet
D n FK
n FK
i01 exp 1 1 1 exp 1 1 1
RT
RT
(332)
D n FK
n FK
i02 exp 2 2 2 exp 2 2 1
RT
RT
at S I
KkS I
corr
S I where inet=0 and by using definition (164) where

corr
eq , k
S I S I result in:
corr
inet
icorr
inet ,1 inet ,2
0 at S I
D n FK S I
{ i01 exp 1 1 1
RT
corr
D n FK S I
i02 exp 2 2 2
RT
S I
corr
n FK S I
exp 1 1
RT
corr
corr
n FK S I
exp 2 2
RT

1

corr
(333)

1

This rearranges to the following to relations:

D n FK S I
i01 exp 1 1 1
RT
corr
i
corr
corr
n FK S I
exp 1 1
RT

1

D n FK S I
i02 exp 2 2 2
RT
corr
icorr
corr
n FK S I
exp 2 2
RT

1

(334)
Now differentiating (162) with respect to the potential gives:

1 D1 n1 F
dinet
1 D1 n1 FK1 D1n1 F
D n FK
exp
exp 1 1 1
i01

S
I
d
RT
RT
RT

RT

1 D 2 n2 F
1 D 2 n2 FK2 D 2 n2 F
D n FK
exp
exp 2 2 2
i02

RT
RT
RT
RT
(335)
After some rearrangement, this gives:

dinet

D1n1 F
D n FK
n FK 1
exp 1 1 1 exp 1 1 1 1
i01

d
RT
RT
RT
S
I
D

1

i02

D n FK
n FK 1
exp 2 2 2 exp 2 2 1 1
RT
RT
RT D 2
D 2 n2 F
At the corrosion potential, the above is:
- 218 -
230
(336)
Appendix
dinet
d S I S I
i01
D n FK S I
exp 1 1 1
RT
RT
D1n1 F
S I corr
i02
corr
D n FK S I
exp 2 2 2
RT
RT
D 2 n2 F
n FK S I
exp 1 1
RT
corr
n FK S I
exp 2 2
RT
corr
1

1 1
D1
corr
1

1 1
D2
(337)
And substituting (334) into the above gives:

dinet
d S I S I
n FK S I
DnF 1
icorr 1 1 exp 1 1
RT
RT D1
corr
S I corr
icorr
n FK S I
D 2 n2 F 1
exp 2 2
RT
RT D 2
corr

1 1

corr
n FK S I
exp 1 1
RT
corr
n FK S I
exp 2 2
RT

1 1

which is the same as:

dinet
d S I S I
S I corr
corr
n FK S I
DnF nF
icorr 1 1 1 1 exp 1 1
RT
RT
RT
corr
n FK S I
D nF nF
icorr 2 2 2 1 exp 2 2
RT
RT
RT
1
(338)
1
Which rearrange to:

dinet
d S I
S I
S I corr
1
corr
n FK S I
1 D1 n1 F n1 F

1
1
1 exp
1
icorr

RT
RT
RT

1

corr
n FK S I
D nF nF

icorr 2 2 2 1 exp 2 2
RT

RT
RT
Which rearrange to:

dinet
d S I S I
S I corr
corr
1 D1 n1F n1 F exp n1FK1S I

icorr
RT
RT
RT
corr
n FK S I
D 2 n2 F n2 F
2
2
exp
icorr

RT
RT
RT

1

1

1
1
by looking at (152) it is essentially the same plus two correction factors, for the
redox:
- 219 -
231
(340)
(339)
Appendix
corr
n FK S I
n1 F
1
1
exp
RT
RT
B.3

1

1
corr
n FK S I
n2 F
2
2
exp
and
RT
RT

1

1
(341)
Assume two redox reactions. Reaction 1 includes only cathodic reaction and reaction
2 only anodic. This way and equation is obtained, which is more precise then (152).
The net current of each of the redox reactions are given by (176) and (177). Isolating
the net currents, gives:
inet ,1
inet ,2
at S I
S I
becomes icorr
corr
inet ,1
D n FK
i01 exp 1 1 1
RT
i01
D n FK
1 d exp 1 1 1
i1
RT
(342)
1 D 2 n2 FK2
i02 exp
RT
1 D 2 n2 FK2
i02
1 d exp
i2
RT
(343)
the net current is zero and defined, inet
0 . Therefore icorr
inet ,2 because only one cathodic and only one anodic reaction
occur. From (342) and (343) this gives:

inet
inet ,1 inet ,2
icorr
D n F S I corr S I eq1
1 1
i01 exp
RT
D n F S I corr S I eq1
1 1
i
1 01d exp
i1
RT
1 D n F S I corr S I eq 2
2
2
i02 exp
RT
1 D n F S I corr S I eq 2
2
2
i
1 02d exp
i2
RT
By using (160) and differentiating the net current, (178), gives:
- 220 -
232
(344)
Appendix
dinet
d S I
i02
i01
D1n1 F
RT
D n FK i D n F
D n FK
exp 1 1 1 01d 1 1 exp 1 1 1 inet ,1
RT i1 RT
RT
i01
D1n1 FK1
1 d exp
i1
RT
1 D 2 n2 F exp 1 D 2 n2 FK2 i02 1 D 2 n2 F exp 1 D 2 n2 FK2 i
RT
d
RT
i2
1 D 2 n2 FK 2
i
1 02d exp
i2
RT
RT
RT
net ,2
(345)
inet,k is given by (342) and (343) and they can easily be substituted into the above
equation. At S I
S I
S I
corr
S I
dinet
d S I
S I
eq , k
corr
the above gives, where K kS I
D n FK S I
exp 1 1 1
RT
RT
D1n1 F
corr
inet , k S I
corr
S I corr
i DnF
D n FK S I
01d 1 1 exp 1 1 1
i1 RT
RT
corr
D n FK S I
i
1 01d exp 1 1 1
i1
RT
corr
1 D n FK S I
1 D 2 n2 F

2
2
2
i02
exp
RT
RT

corr
S I
1 D 2 n2 FK2
i
1 02d exp
i2
RT
S I corr
i02 1 D 2 n2 F
1 D 2 n2 FK2
corr
inet ,2 S I
exp
d
i2
RT
RT

S I corr
i
1 D 2 n2 FK2
1 02d exp
i2
RT
i01
corr
corr
corr
i
S I
net ,1
indicates that (342) and (343) has to be evaluated at
S I S I . Essentially this can easily be done by the use of (344). One can
corr
corr
see that inet ,1 S I icorr and inet ,1 S I icorr . By substituting (344) into
corr
the above, a very simple equation appears:
- 221 -
233
Appendix
dinet
d S I S I
B.4
S I corr
D n F i 1 D 2 n2 F icorr
icorr 1 1 1 corr

1 d
i1d
RT
i2
RT
(179)
Assume the net current density is given by (141) and the kth net current density is
given by (172). The derivative of the net current density with respect to potential is:
1 D k nk F exp 1 D k nk FKk D k nk F exp D k nk FKk
RT
RT
RT
1
n
F
D
K

k k k 1 exp D k nk FKk
1
1 i0 k d exp
d
iRk
RT
RT
iXk
1 D k nk F exp 1 D k nk FKk D k nk F exp D k nk FKk
d
d
RTiRk
RT
RT
RTiXk
inet ,k

1 D k nk FK k 1
1
D k nk FK k
1 i0 k d exp

exp
d
iRk
RT
RT
iXk
dinet ,k
d S I
RT
Which rearrange to:
dinet ,k
d S I
1 D k nk FK k
inet ,k
inet ,k
nk F
D k nk FK k
1 D k 1 d exp
D k 1 d exp
RT
i
RT
i
RT
Rk
Xk
1 D k nk FKk 1
1
1
D k nk FK k
d exp
d exp
i0 k iRk
RT
i
RT
Xk
- 222 -
234
(180)
Appendix
C. Theory of diffusion
C.1
Porous media
Diffusion in porous media is like free diffusion. The basic difference is the travel
distance from a to b as shown in figure 57. The difference in length is due to the
porous structure. In free diffusion the travel distance on a macroscopic level is L. The
travel distance in a porous media is higher due to the solid matrix through which the
molecules have to travel. This distance is designated Le, Le>L. The ratio Le L is
called the tortuosity, W:
W{
Le
!1
L
(346)
The porosity, I, is the fraction of the total volume of a samples which are voids or
occupied by liquid. The porosity is defined by:
Vvoids
Vtotal
(347)
A porosity of 0 is a solid sample and a porosity of 1 is a sample without a solid

structure, like in free diffusion. The tortuosity is a relative length and the porosity is a
relative volume.
Figure 57: Figure showing the principle of the tortuosity.
The paper by Epstein1 gives a thorough derivation of the relation between eff and
:
N. Epstein. On tortuosity and the tortuosity factor and diffusion through porous media. Chem Eng.
Sci. 44(3), 777, 1989.
- 223 -
235
Appendix
eff
W2
(196)
By rearranging this gives:
H{
eff
Where H is the porosity-tortuosity-factor.
- 224 -
236
I
W2
(348)
237
0.1861
0.1487
0.5514
0.4756
-0.0042
-0.8017
0.522
0.5713
0.0757
0.1204
0.3059
0.1866
0.0501
0.0552
-0.3396
-0.8301
0.556
0.3399
0.7312
0.1972
-0.0867
0.2831
0.3495
0.6746
0.0493
1
q
MEG
1
0.0493
r
MEG
0.3459
0.0567
-0.3162
-0.0562
0.3722
0.416
0.5564
0.5038
0.7896
0.0775
1
0.9538
0.2831
0.1972
uij0
MEG
H2O
0.1909
-0.0853
-0.271
0.1882
0.2302
0.3014
0.5716
0.5055
0.9253
0.341
0.9538
1
0.3495
-0.0867
MEG
MEG
j
Par. value
conf. %
5.4996
7.66
r
i MEG
7.0797
7.95
q
MEG
249.738
10.65
uij0
MEG
H2O
339.966
12.5
MEG
MEG
174.558
19.59
MEG
CO2
MEG
Na+
671.891
5.67
0.0162
-0.186
0.0974
0.0786
0.2375
0.2735
0.9496
0.5461
-0.0991
0.5038
0.5055
0.6746
0.1861
MEG
CO32-
676.577
3.58
0.71268
37.6
0.56247
74.32
5.0355
5.18
MEG
CO2
0.6146
0.3605
0.5199
-0.3567
1
0.9929
0.2019
0.2375
0.0354
-0.4058
0.3722
0.2302
0.0757
0.5514
uijt
MEG
H2O
MEG
MEG
-0.0302
-0.1102
0.0658
0.0863
0.2019
0.2445
0.9496
0.6268
0.0112
0.5564
0.5716
0.7312
0.1487
MEG
HCO3-
uijt
MEG
MEG
HCO3H2O
-0.2909
-0.2391
0.0534
0.6929
-0.4058
-0.3181
0.0112
-0.0991
0.6081
1
0.0775
0.341
0.3399
-0.8301
MEG
Na+
MEG
CO32-
0.0004
-0.2045
-0.1406
0.432
0.0354
0.1261
0.6268
0.5461
1
0.6081
0.7896
0.9253
0.556
-0.3396
292.992
46.08
Table 35: Confidence intervals of parameters fitted in parameter set A.
j
r
MEG
q
MEG
uij0 MEG H2O
MEG MEG
MEG CO2
MEG Na+
MEG CO32MEG HCO3uijt MEG H2O
MEG MEG
MEG CO2
MEG Na+
MEG CO32MEG HCO3-
MEG
CO2
Correlation matrix and confidence intervals
Table 34: Correlation matrix of parameters fitted in parameter set A.
D.1
D. Model details of Na2CO3-NaHCO3-MEG-water
3.9323
61.8
MEG
Na+
0.5936
0.3232
0.5354
-0.2821
0.9929
1
0.2445
0.2735
0.1261
-0.3181
0.416
0.3014
0.1204
0.4756
MEG
MEG
2.1482
24.67
MEG
CO32-
0.1361
0.1152
1
0.1176
0.5199
0.5354
0.0658
0.0974
-0.1406
0.0534
-0.3162
-0.271
0.3059
-0.0042
MEG
CO2
-0.50597
-69.17
MEG
HCO3-
-0.7423
-0.8303
0.1176
1
-0.3567
-0.2821
0.0863
0.0786
0.432
0.6929
-0.0562
0.1882
0.1866
-0.8017
MEG
Na+
0.8332
0.1152
-0.8303
0.3605
0.3232
-0.1102
-0.186
-0.2045
-0.2391
0.0567
-0.0853
0.0501
0.522
MEG
CO32-
0.8332
0.1361
-0.7423
0.6146
0.5936
-0.0302
0.0162
0.0004
-0.2909
0.3459
0.1909
0.0552
0.5713
MEG
HCO3-
238
MEG
MEG
MEG
MEG
MEG
MEG
MEG
MEG
MEG
MEG
MEG
MEG
MEG
MEG
H2O
MEG
CO2
Na+
CO32HCO3H2O
MEG
CO2
Na+
CO32HCO3-
0.1112
-0.0639
0.4986
0.4338
0.5837
0.3796
-0.1338
0.3309
0.3356
-0.0964
-0.2383
0.2474
0.534
0.4922
-0.1339
-0.6485
0.0778
-0.3029
0.3753
0.3411
0.1869
0.0361
0.66
1
0.0953
-0.0108
1
0.66
0.2527
0.0078
-0.5143
0.1376
0.2784
0.36
0.4692
0.4548
0.9127
-0.0782
1
0.986
0.0953
0.1869
q
uij0
MEG
MEG
H2O
0.2
-0.0631
-0.5261
0.1501
0.2182
0.3099
0.4933
0.4958
0.9475
0.0181
0.986
1
-0.0108
0.0361
MEG
MEG
Par. value
conf. %
4.006
5.77
5.461
6.76
241.245
11.03
r
q
uij0
MEG
MEG
MEG
i
H2O
j
350.404
11.62
MEG
MEG
173.276
15.7
MEG
CO2
496.39
17.47
MEG
Na+
Table 37: Confidence intervals of parameters fitted in parameter set B.
uijt
r
q
uij0
r
MEG
Table 36: Correlation matrix of parameters fitted in parameter set B.
566.661
5.68
MEG
CO32-
0.1443
-0.0612
-0.4505
0.1355
0.1108
0.2036
0.648
0.6623
1
0.1716
0.9127
0.9475
0.0778
-0.1339
MEG
CO2
616.879
3.71
MEG
HCO3-
0.1273
0.078
-0.042
-0.5934
-0.3333
-0.3041
-0.2182
-0.1722
0.1716
1
-0.0782
0.0181
-0.3029
-0.6485
MEG
Na+
0.18095
133.74
uijt
MEG
H2O
-0.2635
-0.2974
-0.1109
0.5878
-0.0036
0.0512
0.9545
0.6623
-0.1722
0.4548
0.4958
0.2474
-0.2383
MEG
CO32-
-0.21916
-176.58
MEG
MEG
-0.3119
-0.2164
-0.1082
0.5559
0.0299
0.0796
0.9545
0.648
-0.2182
0.4692
0.4933
0.3309
-0.1338
4.4442
6.14
MEG
CO2
0.5654
0.4525
0.5509
0.0595
1
0.9932
0.0299
-0.0036
0.1108
-0.3333
0.2784
0.2182
0.3753
0.534
uijt
MEG
MEG
HCO3H2O
-3.5685
-28.58
MEG
Na+
0.5647
0.4244
0.4956
0.0757
0.9932
1
0.0796
0.0512
0.2036
-0.3041
0.36
0.3099
0.3411
0.4922
MEG
MEG
4.2392
10.9
MEG
CO32-
0.2465
0.375
1
-0.0538
0.5509
0.4956
-0.1082
-0.1109
-0.4505
-0.042
-0.5143
-0.5261
0.3356
0.1112
MEG
CO2
0.65204
49.23
MEG
HCO3-
-0.6377
-0.7296
-0.0538
1
0.0595
0.0757
0.5559
0.5878
0.1355
-0.5934
0.1376
0.1501
-0.0964
-0.0639
MEG
Na+
0.8308
0.375
-0.7296
0.4525
0.4244
-0.2164
-0.2974
-0.0612
0.078
0.0078
-0.0631
0.5837
0.4986
MEG
CO32-
0.8308
0.2465
-0.6377
0.5654
0.5647
-0.3119
-0.2635
0.1443
0.1273
0.2527
0.2
0.3796
0.4338
MEG
HCO3-
Appendix
D.2
The extended UNIQUAC model
The gibbs excess, GE, is defined by difference between the total gibbs energy in the
standard state, Gt4s,id, and the total Gibbs energy of the system, Gt:
G t4s , E { G t G t4s ,id
(349)
Gt4s,E is therefore a function of the chosen standard state and concentration scale, 4s.
It can be shown that
w G t4s , E
wni
P ,T , n j
RT ln J i4s
(350)
the log. of the activity coefficient is a partial derivative of the total Gibbs excess
energy of compound i with respect to constant T,P, and the jth compound.
The extended UNIQUAC model describes the Gibbs excess property by the following
contributions to a Gibbs-excess energy:
GtD x,E
G D x ,C G D x , R G D x , DH
(351)
Gx,C, Gx,R, and Gx,DH are the Gibbs combinatorial, residual and Debye-Hckel
contributions respectively. Which are given by:
G D x ,C
RT
G D x , DH
RT
Ii
Ii
x ln x 2 q x ln T
i
i i
(352)
G D x, R
RT
qi xi ln T k\ k
i
k
(353)
xw M w
4A
b2 I
ln 1 b I b I
3
b
2
(354)
The variables are explained in the text below. The rational symmetrical activity
coefficient of water, J wv x , denoted by subscript w is calculated using (350):
ln J wv x ln J wv x ,C ln J wv x , R ln J wv x , DH
(355)
J wv x ,C , J wv x , R and J wv x , DH are the combinatorial, residual and Debey-Hckel rational

symmetrical activity coefficients of water respectively. The rational unsymmetrical
activity coefficient of the ions, J i* x , is:
ln J i* x
ln J i* x ,C ln J i* x , R ln J i* x , DH
ln J iv x ,C ln J ifx ,C ln J iv x , R ln J ifx , R ln J i* x , DH
- 227 -
239
(356)
Appendix
Which is calculated from the rational symmetric activity coefficient using the rational
symmetrical activity coefficients at infinite dilution, J wfx ,C , J wfx , R . J i* x , DH is the
rational unsymmetrical Debye-Hckel activity coefficient of ion i. The rational
symmetrical combinatorial activity coefficient is expressed from (352) using (350):
ln J iv x ,C
I
ln i
xi
Ii z Ii
1 qi ln
xi 2 T i
Ii
1
T
i
N
N
ri xk qk
ri xk qk (357)
ri
ri
z
1 N
1 kN
qi ln kN
ln N
2
qi xk rk
k xk rk
xk rk
qi xk rk
k
k
xi is the mole fraction of component i, z is the combinatorial number, z=10.
Ii and T i are defined as:

Ii
xi ri
(358)
x q
k
Ti
xi qi
(359)
xk qk
k
where ri is the volume parameter and qi is the surface parameter of component i with
a total of N liquid components. The rational symmetrical residual activity coefficient
is expressed from (353) using (350):
ln J
v x,R
i
N
N T\
qi 1 ln T k\ ki N k ik
k
k
l T l\ lk
(360)
where the energy interaction parameters, ukl , are used in the \ i as:
\ kl
u u
u
exp kl ll exp kl
T
W kl
u
exp kl
T
ull W kl
exp
T W ll
(361)
(362)
T is the temperature in Kelvin. There are two interaction parameters, ukl0 and uklt , for
every component pair since ukl is expressed by the linear temperature relationship as:
- 228 -
240
Appendix
ukl
ukl0 uklt T 298.15 K
(363)
The binary interaction between component A and B are assumed to be the same as the
interaction between B and A therefore:
ukl0
ulk0
(364)
uklt
ulkt
(365)
The rational symmetrical combinatorial and residual activity coefficient are found by
setting xw 1 , and letting composition of others species be zero.
ln J ifx ,C
ri
r z rq
rq
1 i qi ln i w 1 i w
rw
rw 2 rw qi
rw qi
ln
ln J ifx , R
(366)
qi 1 ln\ wi \ iw
(367)
the rational symmetrical Debye-Hckel activity coefficient of water is:

ln J wv x , DH
Mw
2A
1
2 ln 1 b I
1 b I
b3
1 b I
(368)
and the rational unsymmetrical Debye-Hckel activity coefficient for ions are:
ln J i* x , DH
zi2
A I
1 b I
(369)
A is assumed temperature dependent by:

A
1.131 1.335 10 T 273.15 1.164 10 T 273.15

3
5
kg / mol (370)
A is valid in the temperature range: 273.15K<T<383.15K and typically also used

down to 230K. b is b 1,5 kg / mol . M w is the molecular weight of water in kg/mol
and zi is the charge of ion i including sign. I is the molal ionic strength of the
compounds in mol/kg H2O:
1 N
bi zi2
2 1
1 N
xi zi2
2 1
xw M w
bi is the molal concentration in mol compound/kg water of component i.
- 229 -
241
(371)
Appendix
D.3
The extended UNIQUAC parameters
The core parameters of the UNIQUAC model are ri , qi , ukl0 and uklt . ri and qi of nonionic species are typically calculated from their geometry as given by the theory of
Abrams and Prausnitz2 and Bondi3. The ri and qi of water were originally fitted4 to
experimental VLE data of water and should be regarded as constants. In the extended
UNIQUAC model other ri and qi are fitted as parameters in the model. The ri , qi
and interaction parameters of the solids are not known, since they are not present in
the liquid phase.
It is not possible to measure the thermodynamic properties of an ion independently of
other ions, because electro neutrality has to be fulfilled. By convention the ions are
measured relative to the hydrogen ion. The interaction parameters are defined by
(363) to (365). ukl0 for H+ and ion are defined as 109 and uklt for H+ and ions are
defined as 0. H2O-H+ interaction is 104. ukl0 and uklt of H2O-H2O are set to zero. The
equivalent parameter for other solvents is not set to zero but fitted.
The Gibbs energy of H+ of formation, enthalpy of formation and heat capacity is set to
0 in the molal unsymmetrical standard state at all T,P as defined by NIST:
P H*m T , P { 0
(372)
By assuming that the contribution of H+ to the excess enthalpy is zero, then the
surface parameter becomes:
qH | 0
This result in ln J Hfx, R
(373)
0 by insertion of (373) in (367). Assuming that the rational
symmetrical activity coefficient of H+ at infinite dilution is 1, then ln J Hfx

assumptions results in ln J
fx , C
H
0 from (356) since ln J
v x ,C
i
ln J
v x,R
i
ln J
0 . These
* x , DH
i
0 at
infinite dilution. rH is then calculated from (366) by:

ln J ifx ,C
rH
ln
rH
rw
1
rH
rw
(374)
0.13779
Therefore rH and qH are known by convention. For numerical reasons rH
1014 .
The reason why the above assumptions are made, is to anchor the ion parameters to
the H+ ion or else an infinite number of interaction parameters could be found. The
Abrams, D. S; Prausnitz J. M. Statistical Thermodynamics of liquid mixtures: a new expression for

the Gibbs energy of partly of completely miscible system; AIChE J. 21(1), 116, 1975.
3
Bondi, A. Physical Properties of Molecular Crystals, Liquids and Glasses. Wiley, New York, 1968.
4
Prausnitz, J. M; Anderson, T. F.; Grens, E. A.; Eckert, C. A.; Hsieh, R.; OConnel, J. P. Computer
calculations for multicomponent vapour-liquid and liquid-liquid equilibria. Prentice-Hall, Inc., New
Jersey, p. 42, 1980.
- 230 -
242
Appendix
above anchor-scheme is used in the extended UNIQUAC model. Other schemes

may be used, as presented by Thomsen5.
Anion-anion interaction is determined from the equivalent acid. For example Cl--Clinteraction is fitted to the properties of HCl. The cation-cation interaction is set to
zero, which is equivalent to fitting the interaction parameters relative to the anions.
For example in the NaCl-H2O system the Na+-Na+ interaction is zero and the Na+-Clinteraction is relative to the Cl--Cl- interaction.
The total number of interaction parameters are 2 N 2 for the N liquid component
system.
From the rules given above the number of fitted interaction parameters, Nu, are :
Nu
1
2N 2 N N N 1 N netutral 1 N anions

# tables
neutral y water
H
table ydiagonal
(375)
N is the number of all liquid species in the system, water, solvent, and ions. Nneutral is
the number of neutral species, including water. Nanions is the number of anions. Nu is
calculated from (375). The first term is obtained from N2 interaction parameters minus
the diagonal, N2-N, since the diagonal is described by the remaining terms. The factor
is observed, since uij=uji. The second term, N-1, are the number of interaction
parameters predefined in relation to H+. The -1 is observed since the diagonal has
already been subtracted. The third term, Nneutral-1, is the number of parameters for
neutral species, except water which is predefined. The fourth term, Nanions, refers to
the number of anion-anion interaction parameters. The above rearranges to:
Nu
N 2 3N 2 N neutral 2 N anions
(376)
Typical values are given in the table 38.

Table 38: Number of interaction parameters for example systems.
N
4
5
6
8
Example
NaOH-H2O
NaCl-H2O
Na-K-Cl-H2O
The system in section 7
Nneutral
1 (water)
1 (water)
1 (water)
3 (water, MEG, CO2)
Nanions
1 (i.e. OH-)
2 (i.e. OH-, Cl-)
2 (i.e. OH-, Cl-)
3 (i.e. OH-, CO32-, HCO3-)
Nu
8
16
24
52
The number of interaction parameters shown in table 38 is of cause the maximum

number of parameters which will be fitted. Some of the parameters are arbitrarily
defined to a low interaction. For example it is very unlikely that CO2(aq) and CO32will ever meet each other in the same carbon dioxide system. One of the
compounds will be extremely low in concentration at a given pH due to dissociation
to HCO3-. This kind of interaction is set arbitrarily to no interaction, which is
ukl0 =2500 and uklt =0.
Thomsen, K. Aqueous Electrolytes, model parameters and process simulation. Ph.D. Thesis, IVCSEP, Technical University of Denmark, 1997.
- 231 -
243
244
Appendix
14 Index
Acentric factor ..................................80
Activation
Controlled kinetics .....35, 37, 41, 44,
46, 48, 49, 60, 64, 195
Energy .........................14, 31, 33, 35
Polarization .................31, 32, 40, 48
Activity ......7, 33, 35, 43, 80, 132, 157,
168, 192
Activity coefficient ....8, 17, 61, 66, 80,
133, 135, 159, 168, 189, 196, 205,
227
Symmetrical ........ See Standard state
Unymmetrical ..... See Standard state
Alloy ...................................52, 64, 197
Anode............29, 32, 35, 38, 40, 57, 65
Average relative deviation ......105, 110
Boundary conditions .14, 58, 59, 60, 63
Breakdown of pipeline........................1
Carbamate .......................................173
Cathode .....................29, 32, 35, 38, 64
Centrifugal force ...............................15
Charge transfer..................................31
Charge transfer coefficient....33, 35, 37
Chemical potential .....7, 11, 12, 14, 43,
205, 211
Concentration controlled kinetics .....31
Concentration scale...........7, 8, 17, 227
Conductivity................................24, 83
Confidence interval.........114, 132, 225
Control volume .................................13
Correlation matrix...........................225
Corrosion
Crevice ................29, 58, 62, 64, 197
De-alloying ...................................29
Electrochemical.............3, 29, 38, 64
Exchange current ........47, 48, 49, 52
Film .................................53, 67, 143
Flow induced.................................54
Galvanic ........................................29
General and uniform .....................54
Intergranular..................................29
Mesa attack ...................................53
Models............................ See Models
Pitting........................29, 53, 64, 197
Prediction ..................................3, 54
Rate .....3, 40, 51, 53, 55, 60, 67, 197
Scale..........................52, 54, 67, 166
Top of the line...............................52
Under insulation............................29
Critical properties .............................80
Current density....30, 32, 33, 37, 58, 60
Davies rule ........................66, 159, 160
Debye-Hckel .....................66, 80, 227
DEG ..................................56, 196, 197
Del operator ......................................14
Density......................94, 103, 106, 135
Diffusion
Coefficient ..........18, 21, 62, 63, 191
Flux ...................................19, 23, 26
Diffusivity...18, 20, 21, 27, 43, 62, 189
Dissolution...3, 29, 40, 53, 57, 64, 143,
173
Driving force...............................14, 20
Economy .....................................1, 143
Electrical potential ................15, 21, 30
Electrochemistry .......................29, 217
Electroneutrality .......................24, 157
Electrophoretic effect........................21
Elementary Reaction.............32, 33, 64
Enthalpy........6, 82, 143, 151, 152, 176
Entropy ...............................6, 146, 176
Equilibrium
Constant .12, 82, 149, 156, 161, 165,
167, 176
Criteria ..........................................11
Potential ............................30, 34, 37
Excess
Enthalpy................79, 104, 121, 123
Gibbs.....................................11, 227
Heat capacity ................................80
Exchange current ........................35, 60
Experimental setup ...........................84
extended UNIQUAC ...66, 79, 80, 104,
115, 116, 118, 135, 195, 227, 230
Faraday's
Constant ............................15, 32, 40
Law .........................................40, 60
FeCO3 .............2, 55, 66, 143, 176, 192
Fick
Diffusion coefficient...............18, 21
Law .................21, 22, 43, 58, 60, 64
Flux .......................................13, 19, 43
Freezing point depression ...............106
Friction force ....................................18
Fugacity ..............................6, 162, 211
Coefficient ......................................9
- 233 -
245
Index
Gas ....................................1, 51, 65, 79

Gibbs energy ..6, 12, 81, 143, 148, 176,
227
Gibbs phase rule................................12
Gradient.............................................14
Half cell.......................................29, 57
HCl..............51, 64, 83, 85, 87, 88, 231
Ideal
Activity ...........................................8
Activity coefficient .........................8
Solubility.........................................2
Improving diffusion models............189
Infinite dilute
Activity coefficient ..8, 21, 133, 191,
206, 228
Diffusivity .............................23, 196
Inhibitor.........................................2, 52
Interaction parameters......80, 104, 115,
118, 126, 133, 135, 167, 228, 230
Invariant point.........................113, 132
Ionic strength ...3, 23, 52, 65, 155, 159,
161, 162, 168, 191, 229
Kopps rule.........................................82
Kronecker delta function...................17
Laplace operator................................14
Mass transport.................13, 32, 52, 61
Maxwell-Stefan theory......................18
MEG................................2, 56, 79, 190
Meta-stability ..........101, 125, 135, 171
Minimization routine ................81, 104
Mixed-solvent .........................8, 79, 91
Models
Electrochemical.........29, 58, 64, 197
Empirical...............................54, 106
Semi-empirical..............................55
Thermodynamic ....................79, 227
Mohr's salt.......................................173
Mol Balance ..........................13, 59, 92
Mol fraction scale ...............................7
Molal scale ..........................................7
Molar scale..........................................7
Motivation...........................................3
Na2CO3 ......................79, 171, 174, 190
Nabla operator...................................14
NaHCO3 ..........2, 67, 79, 170, 174, 190
NaOH ..............2, 67, 79, 168, 171, 190
Natural gas ...............................See Gas
Nernst
Diffusion layer ..............................58
Equation ........................................34
Nernstian behaviour..........................35
Nernst-Planck .....................23, 60, 189
Extension ......................................25
Nomogram/Nomograph....................55
Non-ideality ......2, 5, 6, 10, 65, 66, 189
Objective function ..................104, 110
Oil ...............................1, 51, 53, 56, 67
Optimization routine.......................104
Over potential ...................................30
Corrosion ......................................39
Overview...........1, 51, 54, 64, 144, 153
Oxygen.2, 52, 147, 152, 169, 172, 174,
178
Parameters...............112, 116, 117, 118
Partial derivative.......................16, 227
Partial differential equation 14, 58, 197
pH ...................................2, 55, 84, 190
Phase rule..........................................12
pH-stabilization.....................2, 67, 191
Pipeline .......1, 52, 55, 65, 79, 191, 195
Pitzer .............64, 65, 80, 160, 161, 167
Platform ..............................................1
Polarization .......................................30
Activation .....................................32
Concentration & Activation..........43
IR drop ..........................................50
Reaction ........................................50
Porosity .....................................62, 223
Porous diffusion layer.........43, 61, 223
Potential ...........See Electrical potential
Potential gradient ........................23, 60
Rate determining sted .................31, 37
Rational scale......................................7
RDE .......... See Rotating disk electrode
Redox................................................32
Reference (T,P)................................5
Review ........................... See Overview
Rotating disk electrode .........43, 58, 64
Sampling ...............................85, 92, 94
Saturated solution ...84, 87, 90, 94, 112
Saturation index ..............105, 125, 129
Schreinemakers
Method..........................................79
Reverse method ..............89, 94, 105
Seeding ...........................................101
Separation .....................................1, 85
Siderite......................................66, 143
SLE ...............................12, 79, 81, 105
Sodium hydroxide................See NaOH
Solubility...................................94, 168
- 234 -
246
Appendix
Speciation
Routine............11, 66, 104, 189, 190
Schemes ................................81, 156
SRK...................................................80
Standard state ......................................5
Chemical potential ....................7, 15
Fugacity.......................7, 8, 210, 215
Mixed-Solvent.................................8
Symmetrical ............................8, 205
Unsymmetrical........9, 114, 189, 205
Steady state .........................14, 61, 191
Step size ..........................................104
Stern and Geary.................................38
Comparison ...................................48
Extension.......................................46
Supersaturation101, 144, 159, 163, 165
Symmetrical ............ See Standard state
Tafel
Constants.......................................38
Law ...............................................37
Thermodynamic factor..............17, 189
Titration .....83, 84, 85, 88, 90, 94, 127,
134, 135
Transference number ........................24
Transportation.............................1, 195
Trona.................................82, 131, 133
UNIQUAC... See extended UNIQUAC
Unsymmetrical.........See Standard state
VLE.......12, 79, 81, 104, 105, 119, 230
Volatile corrosion inhibitors .............53
Volmer-Butler.................35, 58, 60, 64
Wegscheiderite .................82, 131, 134
Wetting .................................53, 56, 64
Winkler's method..............................86
- 235 -
247
248
the problem is presented. The chemical system in the pipelines consists of

CO2-Na2CO3-NaHCO3-MEG-H2O. Sodium is injected as aqueous NaOH or
NaHCO3 in order to pH-stabilize and avoid corrosion and MEG is injected in
order to prevent gas hydrates.
There are a great number of models available in the literature which may
predict CO2 corrosion. These models are not accurate and assume ideality in
the main part of the equations. This thesis deals with aspect of improving the
models to account for the non-ideality.
Theories on electrolyte thermodynamics, electrolyte mass transport, and
electrochemical kinetics are presented. A literature overview of CO2 corrosion
is given and possible extensions of existing models are discussed.
A literature review presented in this work shows that FeCO3 plays a main part
in the protection of steel. Especially the solubility of FeCO3 is an important
factor. The thermodynamic properties of FeCO3 are discussed and validated.
A consistent set of FeCO3 properties are given.
A mixed solvent electrolyte model is regressed for the CO2-Na2CO3-NaHCO3MEG-H2O system. Parameters of the extended UNIQUAC model is fitted to
literature data of VLE, SLE, heat excess, and validated against heat capacity
data. The model is also fitted to experimental data produced in this work for
SLE in the Na2CO3-NaHCO3-MEG-H2O system.
The application of the above model is shown. It demonstrates the calculated
thermodynamic correction factors. These show how the diffusion process in
CO2 corrosion models deviate from the ideal case.
Department of Chemical
and Biochemical Engineering
DTU Building 229
Sltofts Plads
DK-2800 Kgs. Lyngby
www.kt.dtu.dk
Carbon Dioxide Corrosion: Modelling and Experimental Work Applied to Natural Gas Pipelines 2008
The focus of this study is carbon dioxide corrosion. CO2 corrosion is observed in offshore natural gas transportation pipelines. A general overview of
Carbon Dioxide Corrosion

Modelling and Experimental Work Applied to Natural Gas Pipelines

2008

Phillip Loldrup Fosbøl - 87-91435-89-7

Uploaded by

Copyright:

Available Formats

Phillip Loldrup Fosbøl - 87-91435-89-7

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Phillip Loldrup Fosbøl - 87-91435-89-7

Uploaded by

Copyright:

Available Formats

Philip Loldrup Fosbl

the problem is presented. The chemical system in the pipelines consists of

Carbon Dioxide Corrosion

Philip Loldrup Fosbl

Carbon Dioxide Corrosion:

Philip Loldrup Fosbl

Technical University of Denmark (DTU)

Front cover by Philip L. Fosbl: Severe CO2 attacked pipeline.

Theory of electrolyte thermodynamics

Theory of electrolyte mass transport

Fick diffusion .................................................................................................... 21

Theory of electrochemical kinetics and corrosion

General activation controlled kinetics ............................................................ 37

General activation and concentration polarization........................................ 46

CO2 corrosion types............................................................................................... 53

Assumptions of mechanistic models................................................................ 65

Previous work on analysis for sodium and carbonate.................................... 85

Meta stable solutions ..................................................................................... 101

Parameter estimation procedure................................................................... 106

Correlation of density..................................................................................... 106

Review and recommended thermodynamic properties of FeCO3 corrosion films 143

Experimental literature data on FeCO3 ............................................................ 168

Databases of standard state formation properties....................................... 145

Synthesis of FeCO3 ............................................................................................. 172

Improving diffusion models

Future, short term long term

List of figures and tables

Theory of electrolyte thermodynamics

Converting between activity coefficient types .................................................. 205

Model details of Na2CO3-NaHCO3-MEG-water

Correlation matrix and confidence intervals .................................................... 225

Process equipment is exposed to corrosion on a daily basis. Maintenance and repair

Wet GAS Line

Figure 1: An overview of the gas transportation systems.

A pipeline transports on average 4 million m3 natural gas/day. A breakdown will be

Optimal Operating Window

pCO 2(bar),FeCO 3(g/kg H 2O)

When small amounts of NaOH is added, pH increases fast and at an addition of

Motivation and method

Theory of electrolyte thermodynamics

Theory of electrolyte thermodynamics

Thermodynamic properties like G, H, and S, etc. (Gibbs energy, enthalpy, and

Theory of electrolyte thermodynamics

The thermodynamic framework

differentiation of the above gives the following relations:

Equation (6) reduces to the following, at constant T:

Theory of electrolyte thermodynamics

Theory of electrolyte thermodynamics

Standard states and concentration scales

Theory of electrolyte thermodynamics

denoted by superscript *. The unsymmetrical standard state fugacity, fi *s , is defined

Liquid phase fugacities, J approach

The fugacity of component i in the liquid phase at (T,P,n), fi L , where n and s

The calculated value of fi L is independent of the chosen standard state and

Vapour phase fugacities, I approach

Theory of electrolyte thermodynamics

Which inserted in (24) is:

Liquid phase fugacity, I approach

Which inserted in (27) becomes:

denoted by superscript . The unsymmetrical standard state fugacity, fi s , is defined

't A Ni z 't Ri V 't A Ni z 'z ci t 't ci t V

xi Pi RT , this is principally wrong since it has a dimension

to the velocity difference, ui u j , and the drag coefficients, RT ij . ij is therefore

n n 1 diffusivities are required for a n component system. For multi component

Fick diffusion coefficient are not symmetrical and requires n 1 n 1 coefficients