Nesic 2018

Download as pdf or txt
Download as pdf or txt
You are on page 1of 57

Implementation of a Comprehensive Mechanistic Prediction Model of Mild

Steel Corrosion in Multiphase Oil and Gas Pipelines

by Srdjan Nesic, Aria Kahyarian and Yoon Seok Choi


Institute for Corrosion and Multiphase Technology, Ohio University
342 W State Street, Athens, Ohio, 45701, USA

ABSTRACT

In the present study, the implementation of a comprehensive mechanistic predictive model for

corrosion of mild steel in the oil and gas transmission pipelines is described. The present model

simultaneously accounts for all major corrosion scenarios, including CO2 corrosion, H2S

corrosion, and corrosion in the presence of organic acid, and also incorporates the effect of

corrosion product layer formation, including iron carbonate and iron sulfide. With this approach,

the present model mechanistically reflects the mainstream understanding of corrosion in such

environments, and can be readily used to predict the corrosion rates in industrial applications.

The model was implemented by using a generic mathematical and programming approach that

has built in flexibility to add new chemical species and additional reactions in the future. The

model was designed to make it easy to extend and cover an even broader range of conditions

than it currently does, such as higher temperatures and pressures, non-ideal solutions, etc. The

mechanistic nature of the model allows it to be readily coupled with other applications such as

computational fluid dynamics (CFD) codes, multiphase flow simulators, process design

simulators, etc. In order to demonstrate the capabilities of this model, the calculated corrosion

rates were compared with the experimental corrosion rate data across a broad range of

environmental conditions and brine chemical compositions.


1. INTRODUCTION

The significance of the corrosion rate prediction in defining the design life of industrial

infrastructure with all the associated health, safety, environmental, and economic consequences

has been a strong driving force for developing better understanding of the corrosion phenomena

and advancements in the corrosion rate predictive tools. Corrosion rate predictive models for

internal pipeline corrosion in the oil and gas industry have undergone a long journey from the

first simplistic nomograms developed in 1970s1 to the comprehensive and elaborate mechanistic

mathematical models of today 2. The significant investment of resources into the development of

the ever more capable predictive models, was a response to the industrial demands for more

accurate corrosion rate predictions under increasingly more complex conditions. The new

generation of mechanistic models has also become a suitable platform on which it is possible to

implement the continuously advancing understanding of the underlying corrosion mechanisms

related to internal pipeline corrosion, and illustrate how all the “pieces of the puzzle” fit together

to describe the overall process.

The text below is focused on the implementation of the new generation of mathematical models

of internal corrosion of mild steel pipelines. Even though aqueous CO2 corrosion (aka. sweet

corrosion) is the most common mode of attack in wells and transmission pipelines, it is often

complicated by the presence of other corrosive species such as hydrogen sulfide (leading to so

called sour corrosion) and carboxylic (aka. organic) acids, as well as complex water chemistry,

formation of surface layers, multiphase flow, high pressure and temperature, etc. The

comprehensive modeling approach described below demonstrates how the modeling framework

is set up to reflect the current state of understanding of the physico-chemical phenomena


underlying sweet, sour and organic acid corrosion, but also to remain flexible and allow for

implementation of new knowledge as it emerges in future.

Looking back briefly, it is worth recalling that corrosion rate predictive models for internal

corrosion of mild steel pipelines developed in the past can be best classified depending on the

level of mathematical description of the fundamental thermodynamics and kinetic processes

underlying the corrosion process. That includes:

• empirical models employing arbitrary mathematical expressions with no true theoretical


3–5
underpinning, such as done for example in the so-called Norsok model or the model

proposed earlier by Dugstad et al. 6;

• semi-empirical models based on some rudimentary mechanistic considerations, such as

the series of models developed by de Waard and collaborators 1,7–11;

• elementary mechanistic models that use a simplified theoretical approach similar to what

was originally introduced by Gray et al. 12–15;

• comprehensive mechanistic models similar to what was introduced by Nesic et al. 2,16–18,

where majority of the processes are described based on the fundamental physiochemical

laws.

With the focus on the implementation of a most recent comprehensive mechanistic model, a brief

review of the key studies that had a significant impact on mathematical modeling of internal

pipeline corrosion is introduced first. The sections that follow describe the theory and

implementation of the corrosion model for a system containing the following corrosive species:

H+ ions, aqueous CO2, H2S, and carboxylic acids, however, the generic mathematical approach

allows for the inclusion of any number of additional species that may be present in a given
system, such as other weak acids, dissolved oxygen, etc. Furthermore, the present model is

developed to account for formation of porous corrosion product layers and their effect on the

corrosion process. Models for precipitation of solid iron carbonate and solid iron sulfide are

described below, while formation of additional solid species on the corroding steel surface such

as: magnetite, hematite, various types of scales such calcium carbonate, magnesium sulfate, etc.,

can be added with little effort. Finally, the verification of the model by using a broad

experimental database is outlined at the end of the paper in order to demonstrate the accuracy

and applicability of the model.

2. A BRIEF HISTORICAL REVIEW OF CORROSION MODELS

Being the most common corrosion scenario, so called sweet systems, with CO2 and its carbonate

derivatives as the main corrosive species, have been the first battleground for corrosion rate

prediction models and testing of various calculation approaches. The successful CO2 corrosion

models have subsequently been extended to cover corrosion rate predictions in other corrosion

scenarios, including sour systems and corrosion in presence of carboxylic acids. In this sense, a

historical view of developments in corrosion rate prediction models for oil and gas transmission

pipelines starts by looking back at some landmark studies in CO2 corrosion prediction. It is

helpful to start with the much simpler semi-empirical and mechanistic models, such as the

seminal semi-empirical models of de Waard et al., due to their significance in shaping the

understanding of CO2 corrosion as we know it today. Other similar variations of semi-empirical

models will not be discussed here, however, numerous reviews of such models are available in

the literature for further reference 3,19–24.


The initial study by de Waard and Milliams in 1975 has been considered the first mechanistic
10
attempt to describe and further, predict the CO2 corrosion of steel . Using a model developed

based on simplistic charge transfer relationships, de Waard and Milliams derived a simple

correlation between the corrosion curret and CO2 partial pressure. By considering the charge

balance at the corrosion potential (ia=ic) and using pH dependence expressions to relate the
10
potential to corrosion current , authors developed their well-known nomogram for corrosion

rate estimation1. The initial model developed by de Waard and Milliams did not include the

effect of other electroactive species, such as the hydrogen ions, the bicarbonate ions and water;

the pH was assumed to only be defined by CO2 equilibria; the effect of mass transfer, CO2

hydration reaction as well as other homogeneous chemical reactions associated with carbonate

species was also disregarded 10. That made the model simple, but drastically narrowed the range

of its applications. In a series of studies, extending over almost two decades, the initial model of
10
de Waard and Milliams was used as the foundation where the effect of various other relevant
1,7–9,11
parameters and environmental conditions was added . The effect of pH, flow rate, non-

ideal solutions, protective scales, glycol, top of line corrosion, and steel microstructure are
1,7–9,11
amongst those effects covered in the subsequent publications of de Waard et al. . These

new effects were accounted for by simply introducing additional empirical correction factors as

multipliers in the original de Waard and Milliams correlation. That transformed the original

mechanistic approach of de Waard and Milliams into a semi-empirical model with many obvious

disadvantages. However, it should be noted that the implementations of the original de Waard
10
and Milliams model and its subsequent derivatives are still found in the industry, where they

are still used (and abused) despite their known shortcomings and the fact that more advanced and

accurate models are now widely available, as described below.


The first elementary mechanistic model for CO2 corrosion of steel was introduced in 1989 by

Gray et al. 13. The authors developed the model with iron dissolution as the only anodic reaction

and hydrogen ion and carbonic acid reduction as the cathodic reactions. Their model accounted

for the mass transfer at a rotating disk electrode for hydrogen ion and carbonic acid reduction.

The effect of CO2 hydration reaction was also incorporated in the mass transfer calculation of

carbonic acid. They latter (over)extended their original model towards much higher temperatures

and pH values 12, suggesting that in the pH6 - pH10 range the reduction of bicarbonate ion also

becomes significant. The mechanistic approach to modeling of CO2 corrosion of steel proposed
12,13
by Gray et al. in these two studies , rapidly gained general acceptance, and was further

developed in the following years.

In 1996, an elementary mechanistic model developed by Nesic et al., mainly focused on

improving the estimated electrochemical rate constants and implementation of this mechanistic

approach to corrosion rate prediction for industrial applications 15. This model was developed by

considering mass transfer, the slow CO2 hydration reaction, and the kinetics of the
12,13
electrochemical reactions in a similar way as previously proposed by Gray et al. . Hydrogen

ion, carbonic acid, water and oxygen reduction were included in the model as the possible

cathodic reactions and iron dissolution as the only anodic reaction. In this model, Nesic et al.

assumed that the carbonic acid reduction was only limited by the CO2 hydration reaction, being

the preceding rate determining chemical reaction step, and the effect of mass transfer on the

chemical reaction limiting current of carbonic acid reduction was ignored. While such an

assumption is reasonable for stagnant conditions, it may lead to significant errors at high solution

velocities where the rate of mass transfer is comparable with the rate of the chemical reaction.
This issue was addressed by Nesic et al. in a later publication where the effect of mass transfer

was also included in chemical reaction limiting current calculations for turbulent flow regimes 14.

The elementary mechanistic models are now well-established for calculation of internal pipeline

corrosion rates. The scope of these models was expanded to incorporate more complex scenarios
25–27
such as the effect of corrosion product layer , multiphase flow 28, and the presence of other

corrosive species like oxygen, hydrogen sulfide, and organic acids 14,29–33.

An example of the elementary mechanistic models has been developed and published by Nesic et

al. 14 as an open source code freely available to the public, called FREECORP.

The elementary mechanistic models created a platform to implement the emerging understanding

of CO2 corrosion into corrosion rate predictions. With the mechanistic approach in the

calculations, these models also provided the opportunity for investigating the individual

underlying processes. However, the simple approach in implementation of physicochemical

theory in the elementary mechanistic models suffers from one fundamentally flawed assumption.

In these models, it is assumed that species are transferred from the bulk fluid toward the metal

surface and back independently from each other. In other words, the well-defined homogeneous

chemical reactions as well as the ionic interaction (electromigration) between species inside the

diffusion layer are ignored. Furthermore, the rates of all electrochemical reactions are typically

based on bulk solution concentrations, as this is much easier to do. These simplifications made

an accurate prediction of species concentrations at the corroding steel surface difficult, what then

rendered any mechanistic prediction of formation of protective corrosion product layers at the

surface problematic.
The main driving force for development of the new generation of more comprehensive

mechanistic models was the need to accurately predict the surface concentration of species,

which is often very different from those in the bulk. These are needed to properly calculate the

rates of various heterogeneous reactions occurring at the metal surface, such as the rate of

cathodic reactions and anodic dissolution of iron (corrosion), growth of corrosion product layers

(such as iron carbonate), etc. In an attempt to move beyond the “worst case scenario” corrosion

rate predictions, the comprehensive mechanistic models emerged that allowed for reasonable

prediction of the spontaneous protective corrosion product layers formation and the resulting low

corrosion rates that can make the use of mild steel without corrosion inhibition feasible in many

situations. To achieve these goals, accurate modeling of simultaneously occurring heterogeneous

electrochemical reactions, mass transfer and homogeneous chemical reactions in the aqueous

solution at the metal surface was needed.

While the first simplistic attempts to model CO2 corrosion using this kind of approach can be

traced back to early 1990 in the works by Turgoose et al.34 and Pots35, the comprehensive

mathematical models of CO2 corrosion of mild steel in its complete form was introduced in early
2,16–18,36
2000’s in series of publications by Nesic et al. . Besides the use of Nernst-Planck

equation to describe the concentration profiles of the chemical species in the solution, the

homogeneous chemical reactions and electrochemical reactions were treated with more details

than in the previous models 34,35. The scope of the model was further expanded by demonstrating

its ability to incorporate the corrosion product layer formation and by calculating the porosity

distribution throughout that layer.

The comprehensive mathematical models, with their in-depth analytical approach, have attracted

growing interest by researchers in the last two decades. In more recent years, similar models
have been developed and used to describe various corrosion scenarios. A few examples are the
37,38
studies of sour corrosion by Tribollet et al. , mechanistic study of CO2 corrosion and CO2
3940
corrosion under a thin water film by Remita et al. , pit propagation in CO2 and acetic acid

environment by Amri et al. 41, mechanistic study of acetic acid corrosion by Kahyarian et al. 42,

and top of the line corrosion by Zhang et al.43.

3. THEORETICAL BACKGROUND AND THE COMPREHENSIVE MODEL

The oil and gas transmission pipelines have been one of the main fronts for corrosion rate

prediction, mitigation and monitoring, due to the severe corrosivity of the environment and low

corrosion resistance of mild steel. The high corrosivity of pipeline internal environment is due to

presence of the co-produced aqueous phase with dissolved ionic species, organic acids, and acid

gases such as carbon dioxide and hydrogen sulfide. Upon dissolution in water these species form

a highly buffered acidic solution, where the metallic iron is thermodynamically unstable and it is

spontaneously converted to ferrous ions. The overall process may be expressed in term of net

redox reactions, as sown below in Reactions ( 1 ) to ( 3 ) for the case of carbon dioxide

corrosion, hydrogen sulfide corrosion, and acetic acid corrosion, respectively.

2+
𝐹𝐹𝐹𝐹(𝑠𝑠) + 𝐶𝐶𝑂𝑂2(𝑔𝑔) + 𝐻𝐻2 𝑂𝑂(𝑙𝑙) → 𝐹𝐹𝐹𝐹(𝑎𝑎𝑎𝑎) + 𝐶𝐶𝑂𝑂32−(𝑎𝑎𝑎𝑎) + 𝐻𝐻2(𝑔𝑔) (1)

2+ 2−
𝐹𝐹𝐹𝐹(𝑠𝑠) + 𝐻𝐻2 𝑆𝑆(𝑎𝑎𝑎𝑎) → 𝐹𝐹𝐹𝐹(𝑎𝑎𝑎𝑎) + 𝑆𝑆(𝑎𝑎𝑎𝑎) + 𝐻𝐻2(𝑔𝑔) (2)

2+ −
𝐹𝐹𝐹𝐹(𝑠𝑠) + 2𝐻𝐻𝐻𝐻𝐻𝐻(𝑎𝑎𝑎𝑎) → 𝐹𝐹𝐹𝐹(𝑎𝑎𝑎𝑎) + 2𝐴𝐴𝐴𝐴(𝑎𝑎𝑎𝑎) + 𝐻𝐻2(𝑔𝑔) (3)

In certain range of environmental conditions, the formation of solid ferrous deposits are

commonly observed in these systems. In order to describe this process, the overall reactions can

be restated as:
𝐹𝐹𝐹𝐹(𝑠𝑠) + 𝐶𝐶𝑂𝑂2(𝑔𝑔) + 𝐻𝐻2 𝑂𝑂(𝑙𝑙) → 𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹3(𝑠𝑠) + 𝐻𝐻2(𝑔𝑔) (4)

𝐹𝐹𝐹𝐹(𝑠𝑠) + 𝐻𝐻2 𝑆𝑆(𝑎𝑎𝑎𝑎) → 𝐹𝐹𝐹𝐹𝐹𝐹(𝑠𝑠) + 𝐻𝐻2(𝑔𝑔) (5)

The reactions shown above are condensed expressions of a large number of chemical and

electrochemical reactions that occur simultaneously. These are the subjects of the discussion in

the following sections, including the homogeneous chemical equilibria in the solution, the

heterogeneous chemical reactions at the metal surface, the formation of the solid deposits, and

the mass transfer of the involved chemical species from the bulk solution to the metal surface.

The mechanistic corrosion rate predictive models attempt to utilize the understandings of these

underlying physiochemical processes to develop a mathematical representation of such systems.

That allows these models to be used in a wide range of environmental conditions, and more

complex corrosion scenarios. The accuracy of such models is therefore tied to the validity of the

mechanistic understanding and inclusion of all key processes. Considering the overall

complexity, the current comprehensive models have moved well beyond the simplistic

nomograms of semi-empirical models9 and rudimentary calculations of simplistic mechanistic


12,13,15
models . As it becomes evident in the following text, the comprehensive mathematical

models take much longer to develop, they require much more extensive numerical

manipulations, and they are computationally more demanding. However, this is a small price to

pay when one considers the consequences of design and operational decisions relying on such

models. It is only logical that they harness the knowledge accumulated over decades of research

and development in this field of study and cast it in a form that can be used directly by the

corrosion engineers on the “front lines”.


3.1. AQUEOUS SOLUTION SPECIATION

The water chemistry calculation is the first computational step in any mechanistic model for

corrosion rate prediction purposes. This step is required to calculate the concentration of species

that are (electro)chemically significant in the corrosion process. In the case of aqueous corrosion

in oil and gas pipelines, numerous (potentially) corrosive species are typically present, including

CO2, H2S, carboxylic acids and their corresponding ions produced by dissociation. Despite

potentially large number of the involved species, the solution speciation can be obtained based

on the chemical equilibrium of the system, following some rather generic calculations. The

discussion bellow covers the conditions where all major species due to presence of CO2, H2S and

carboxylic acids are present in the solution. Following the same calculation approach, additional

species can be readily included.

Upon dissolution in the aqueous phase, CO2 undergoes a series of chemical reactions to form H+,

H2CO3, HCO3-, CO32-, as shown via Reactions ( 6 ) to ( 9 ).

𝐶𝐶𝐶𝐶2(𝑔𝑔) ⇌ 𝐶𝐶𝐶𝐶2(𝑎𝑎𝑎𝑎) (6)

𝐶𝐶𝐶𝐶2(𝑎𝑎𝑎𝑎) + 𝐻𝐻2 𝑂𝑂(𝑙𝑙) ⇌ 𝐻𝐻2 𝐶𝐶𝐶𝐶3 (7)


(𝑎𝑎𝑎𝑎)

𝐻𝐻2 𝐶𝐶𝐶𝐶3(𝑎𝑎𝑎𝑎) ⇌ 𝐻𝐻𝐻𝐻𝐻𝐻3− (𝑎𝑎𝑎𝑎) + 𝐻𝐻 + (𝑎𝑎𝑎𝑎) (8)

𝐻𝐻𝐻𝐻𝐻𝐻3− (𝑎𝑎𝑎𝑎) ⇌ 𝐶𝐶𝐶𝐶32− (𝑎𝑎𝑎𝑎) + 𝐻𝐻 + (𝑎𝑎𝑎𝑎) (9)

Similarly, the dissolved H2S (Reaction ( 10 )) as a diprotic weak acid, is involved in a

dissociation equilibrium according to Reactions ( 11 ) and ( 12 ).

𝐻𝐻2 𝑆𝑆(𝑔𝑔) ⇌ 𝐻𝐻2 𝑆𝑆(𝑎𝑎𝑎𝑎) ( 10 )


+ −
𝐻𝐻2 𝑆𝑆(𝑎𝑎𝑎𝑎) ⇌ 𝐻𝐻(𝑎𝑎𝑎𝑎) + 𝐻𝐻𝑆𝑆(𝑎𝑎𝑎𝑎) ( 11 )

− + 2−
𝐻𝐻𝑆𝑆(𝑎𝑎𝑎𝑎) ⇌ 𝐻𝐻(𝑎𝑎𝑎𝑎) + 𝑆𝑆(𝑎𝑎𝑎𝑎) ( 12 )

From Reactions ( 6 ) through ( 12 ) it can be seen that the fundamental difference between the

CO₂ and the H₂S water chemistry is that the aqueous CO₂ must undergo a hydration step to form

H₂CO₃ before dissociation while aqueous H₂S can directly dissociate after dissolution in the

aqueous phase. The concentration of aqueous H₂S concentration is about 1000 times higher than

aqueous H₂CO₃ at the same partial pressures of CO₂ and H₂S gases. This ratio is often used as

an argument to decide whether corrosion is sweet or sour, i.e. whether it is “dominated” by CO2

or H2S, although such designations are misleading.

Similar to aqueous H2CO3 and aqueous H2S, any other weak acid such as short chain carboxylic

acids (acetic acid in the present discussion) as well as water can partially dissociate to form H+

and their corresponding anions, as shown in Reactions ( 13 ) and ( 14 ).

𝐻𝐻𝐻𝐻𝐻𝐻(𝑎𝑎𝑎𝑎) ⇌ 𝐻𝐻 + (𝑎𝑎𝑎𝑎) + 𝐴𝐴𝐴𝐴 − (𝑎𝑎𝑎𝑎) ( 13 )

𝐻𝐻2 𝑂𝑂(𝑙𝑙) ⇌ 𝑂𝑂𝑂𝑂 − (𝑎𝑎𝑎𝑎) + 𝐻𝐻 + (𝑎𝑎𝑎𝑎) ( 14 )

In order to obtain the solution speciation, the composition of the gas phase is commonly used as

the initially known parameter, since it is easier to determine and control in industrial and/or

experimental systems. The gas/liquid equilibrium shown via Reactions ( 6 ) for CO2 and

Reactions ( 10 ) for H2S, can be expressed in the form of Henry’s law (or its modified forms) as

show via Equation ( 15 ):


𝐶𝐶𝑖𝑖 − 𝐻𝐻𝑖𝑖 𝑝𝑝𝑖𝑖 = 0 ( 15 )

where, pi is the partial pressure of CO2 or H2S (bar), Hi is Henry’s constant (M/bar), and Ci is the

concentration of dissolved species in liquid phase (M). The partial pressure of water in the gas

phase (saturation pressure), denoted by pws, can be calculated as shown in Table 1. It is common

in industrial applications for partial pressure of CO2 to be so high that a significant deviation

from ideal gas assumption of Equation ( 15 ) is observed. In such conditions the partial pressure

of CO2 in Equation ( 15 ) should be replaced with CO2 fugacity, 𝜑𝜑𝐶𝐶𝐶𝐶2 × 𝑝𝑝𝐶𝐶𝐶𝐶2 , for more accurate

calculations (shown in Table 1).

Using a generic formulation, any chemical equilibria, including the partial dissociation of

carbonic and sulfide species and those stemming from the presence of organic acids, can be

expressed in the form of Reaction ( 16 ). Thus, the chemical equilibrium of any given reaction j

can be mathematically expressed according to Equation ( 17 ), where Kj is the equilibrium

constant. The Kj values for H2CO3 dissociation (Kca), HCO3- dissociation (Kbi), and water

dissociation (Kw), can be found in Table 1. For H2S system and some other common weak acids

of significance in oil and gas pipeline corrosion the values of the equilibrium constant can be

found in Table 2.

𝑛𝑛𝑟𝑟 𝑛𝑛𝑝𝑝
( 16 )
� 𝑟𝑟 ⇌ � 𝑝𝑝
𝑟𝑟=1 𝑝𝑝=1

𝑛𝑛𝑝𝑝 𝑛𝑛𝑟𝑟 ( 17 )
� 𝐶𝐶𝑝𝑝 − 𝐾𝐾𝑗𝑗 � 𝐶𝐶𝑟𝑟 = 0
𝑝𝑝=1 𝑟𝑟=1

In addition to the abovementioned chemical equilibria, the concentration of ions in the aqueous

phase must also satisfy the electroneutrality constraint as shown by Equation ( 18 ).


� 𝑧𝑧𝑖𝑖 𝐶𝐶𝑖𝑖 = 0 ( 18 )
𝑖𝑖

Table 1. The equilibrium constants and fugacity coefficient for CO2/H2O system.*

𝐻𝐻𝐶𝐶𝐶𝐶2 ( I ) 44 𝜑𝜑𝐶𝐶𝐶𝐶2 ( II ) 45 𝐾𝐾𝑐𝑐𝑐𝑐 ( III ) 46 𝐾𝐾𝑏𝑏𝑏𝑏 ( III ) 46 𝐾𝐾𝑤𝑤 ( IV )47


𝑃𝑃𝑤𝑤𝑤𝑤 ( V ) 48
a1 1.3000 𝐸𝐸1 1.0000 233.51593 −151.1815 −4.098 0.1167 𝐸𝐸4

a2 −1.3341 𝐸𝐸 − 2 4.7587 𝐸𝐸 − 3 − −0.0887 −3245.2 −0.7242 𝐸𝐸6

a3 −5.5898 𝐸𝐸2 −3.3570 𝐸𝐸 − 6 −11974.3835 −1362.2591 2.2362 −0.1707 𝐸𝐸2

a4 −4.2258 𝐸𝐸5 − − − −3.984 𝐸𝐸7 0.1202 𝐸𝐸5

a5 − −1.3179 −36.5063 27.7980 13.957 −0.3233 𝐸𝐸7

a6 − −3.8389 𝐸𝐸 − 6 −450.8005 −29.5145 −1262.3 0.1492 𝐸𝐸2

a7 − − 21313.1885 1389.0154 8.5641 𝐸𝐸5 −0.4823 𝐸𝐸4

a8 − 2.2815 𝐸𝐸 − 3 67.1427 4.4196 − 0.4051 𝐸𝐸6

a9 − − 0.0084 0.0032 − −0.2386

a10 − − −0.4015 −0.1644 − 0.6502 𝐸𝐸3

a11 − − −0.0012 −0.0005 − −

* The ai values are rounded to four digits after the decimal.

∗ 𝑎𝑎3 𝑎𝑎4
(I) ln�𝐾𝐾𝐻𝐻,𝐶𝐶𝐶𝐶 ∗ � = 𝑎𝑎1 + 𝑎𝑎2 𝑇𝑇 + +
2 𝑇𝑇 𝑇𝑇 2
𝑎𝑎4 𝑎𝑎5 𝑎𝑎8
( II ) 𝜙𝜙𝐶𝐶𝐶𝐶2 = 𝑎𝑎1 + �𝑎𝑎2 + 𝑎𝑎3 𝑇𝑇 + + � 𝑃𝑃 + �𝑎𝑎6 + 𝑎𝑎7 𝑇𝑇 + � 𝑃𝑃2
𝑇𝑇 𝑇𝑇 − 150 𝑇𝑇
𝑎𝑎3 𝑎𝑎4 𝑎𝑎6 𝑎𝑎7 𝑎𝑎8 𝑎𝑎9 𝑎𝑎10 𝑎𝑎11
ln(𝑝𝑝𝑝𝑝𝑝𝑝. ) = 𝑎𝑎1 + 𝑎𝑎2 𝑇𝑇 + + + 𝑎𝑎5 𝑙𝑙𝑙𝑙(𝑇𝑇) + � + 2 + 𝑙𝑙𝑙𝑙𝑙𝑙 � + � + 2 + 𝑙𝑙𝑙𝑙𝑙𝑙 � (𝑝𝑝 − 𝑝𝑝𝑠𝑠 )2
( III) 𝑇𝑇 𝑇𝑇 2 𝑇𝑇 𝑇𝑇 𝑇𝑇 𝑇𝑇 𝑇𝑇 𝑇𝑇
Ps=1 if T<373.15, Ps=Pws if T>373.15.
𝑎𝑎2 𝑎𝑎3 𝑎𝑎4 𝑎𝑎6 𝑎𝑎7
( IV ) −log(𝐾𝐾𝑤𝑤 ) = 𝑎𝑎1 + + 2 + 3 + �𝑎𝑎5 + + 2 � log(10−3 𝜌𝜌𝑤𝑤 )
𝑇𝑇 𝑇𝑇 𝑇𝑇 𝑇𝑇 𝑇𝑇
4
2𝐶𝐶
𝑃𝑃𝑤𝑤𝑤𝑤 = 10 � �
( V) −𝐵𝐵 + (𝐵𝐵2 − 4𝐴𝐴𝐴𝐴)0.5
𝑎𝑎9
𝐴𝐴 = 𝜃𝜃 2 + 𝑎𝑎1 𝜃𝜃 + 𝑎𝑎2 ; 𝐵𝐵 = 𝑎𝑎3 𝜃𝜃 2 + 𝑎𝑎4 𝜃𝜃 + 𝑎𝑎5 ; 𝐶𝐶 = 𝑎𝑎6 𝜃𝜃 2 + 𝑎𝑎7 𝜃𝜃 + 𝑎𝑎8 ; 𝜃𝜃 = 𝑇𝑇 +
𝑇𝑇−𝑎𝑎10
The equilibrium speciation of the aqueous phase can therefore be obtained by solving the set of

mathematical equations presented above, covering liquid/gas equilibrium (Equation ( 15 )),

chemical equilibria in the aqueous phase (Equation ( 17 )), and the electroneutrality constraint

(Equation ( 18 )). When the solution pH is unknown in advance, the concentrations of the

involved chemical species are to be obtained using iterative solution schemes, due to the non-

linearity of the equilibrium expressions for weak acid dissociation reactions.

Table 2. The Equilibrium constants of H2S/H2O system and other common species.*

𝐻𝐻𝐻𝐻2𝑆𝑆 49
𝐾𝐾𝐻𝐻2𝑆𝑆 50
𝐾𝐾𝐻𝐻𝐻𝐻 − 51
𝐾𝐾𝐻𝐻𝐻𝐻𝐻𝐻 51

𝑎𝑎1 -6.3427 E2 7.8244 E2 -2.393 E1 -6.66104

𝑎𝑎2 -2.709 E-1 3.6126 E-1 3.0446 E-2 1.34916 E-2

𝑎𝑎3 1.1132 E-4 -1.6722 E-4 -2.4831 E-5 2.37856 E-5

𝑎𝑎4 1.6719 E4 -2.05657 E4 _ _

𝑎𝑎5 2.619 E2 -1.4274 E2 _ _


𝑎𝑎4
* log(𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝. ) = 𝑎𝑎1 + 𝑎𝑎2 𝑇𝑇𝐾𝐾 + 𝑎𝑎3 𝑇𝑇𝐾𝐾2 + + 𝑎𝑎5 log(𝑇𝑇𝐾𝐾 )
𝑇𝑇𝐾𝐾

The examples of such calculations are demonstrated below. Figure 1 shows the pH of the CO2

saturated aqueous solutions as a function of CO2 partial pressures up to 75 bar, assuming an

“open” system (i.e. constant composition in the gas phase). These results were obtained based on

the above-mentioned calculations using CO2 fugacity, and were found agreement well with the

experimental data of Meysammi et al. 52. Nevertheless, at higher pressures some deviations from

the experimental data is observed that could be associated with the departure from ideal aqueous

solution assumption.
4.00

3.90

3.80

3.70

3.60

3.50
pH
3.40

3.30

3.20

3.10

3.00
0 10 20 30 40 50 60 70
Pressure (pCO2+pws) (bar)

Figure 1. Calculated pH dependence of water saturated with CO2(g) at 305.15 oK, as a function of
total pressure (solid line) compared with experimental data (closed circles) taken from Meyssami
et al. 52.

Similar calculations can be used to determine the chemical speciation in an aqueous CO2

saturated solution. Figure 2 shows the variation in solution speciation of CO2 equilibria with pH,

under a 1 and 5 bar pure CO2 atmosphere, for an open system. An open system is defined as a

system where the key species (in our case CO2) resides predominantly in the gas phase and we

can assume that pCO2 is constant, what significantly influences the speciation in the aqueous

phase.
A) 1.0E+1

1.0E+0

1.0E-1
CO2 (aq)
Concentration / M

1.0E-2

HCO3- (aq) CO2-3 (aq)


1.0E-3

1.0E-4 H2CO3 (aq)

1.0E-5

1.0E-6
4 4.5 5 5.5 6 6.5 7 7.5 8
pH

B) 1.0E+1

1.0E+0
CO2 (aq)
1.0E-1
Concentration / M

1.0E-2 CO2-3 (aq)


HCO3- (aq)

1.0E-3
H2CO3 (aq)
1.0E-4

1.0E-5

1.0E-6
4 4.5 5 5.5 6 6.5 7 7.5 8
pH

Figure 2. Solution speciation in CO2/H2O equilibrium at various pH values at T=303.13 oK for an open system
(pws=0.042 bar). A) 1 bar total pressure. B) 5 bar total pressure.

In sour systems, where both CO2 and H2S are present in the gas phase, similar calculations can

be performed to obtain the mixed speciation of the solution. Figure 3 shows an example of
equilibrium concentrations as a function of pH for the H₂S/H2O. For an open system considered

here, which is most common in practical applications, the solution speciation from a mixed gas

composition is obtained by superposition of the speciation from both CO2/H2O and H2S/H2O

systems.

A)
1E+02

1E+00

H2S(aq)
Concentration / M

1E-02

1E-04
HS-(aq)

1E-06

1E-08 S2-(aq)

1E-10
3 4 5 6 7 8 9 10
pH

B)
1E+02

1E+00
H2S(aq)
Concentration / M

1E-02

1E-04 HS-(aq)

1E-06

1E-08 S2-(aq)

1E-10
3 4 5 6 7 8 9 10
pH

Figure 3. Equilibrium speciation of H2S/H2O system at 303.13oK. A) 0.1 bar H2S partial pressure. B) 1 bar H2S
partial pressure.
However, closed systems are sometimes more appropriate to represent water speciation, for

example in the case of CO2 corrosion in an autoclave or a confined reservoir or as is almost

always the case with carboxylic acids, as these species predominantly reside is the aqueous phase

and their volatility can be ignored. In the latter scenario, the known parameter is usually the total

concentration of the weak acid that is partially dissociated to its ionic form. In order to account

for this process, the water chemistry calculation in such cases include additional relationships

based on mass conservation law. Figure 4 demonstrates the results from such calculation for the

case of aqueous acetic acid solutions. In this example, the total concentration of acetic acid is

defined as shown by Equation ( 19 ). Since the dissociation equilibrium also involves hydrogen

ions (according to Reaction ( 13 )), the partitioning of acetic acid between its undissociated form

and acetate ion is strongly pH dependent, as shown in Figure 4.

𝐶𝐶𝑡𝑡,𝐻𝐻𝐻𝐻𝐻𝐻 = 𝐶𝐶𝐻𝐻𝐻𝐻𝐻𝐻 + 𝐶𝐶𝐴𝐴𝐴𝐴 − ( 19 )

Such calculation may be readily extended to include additional species commonly encountered in

oil and gas transmission pipelines, such as other carboxylic acids or acidic/alkaline salts, to

obtain the corresponding solution speciation in a very similar fashion.


30 °C 60 °C
1 0.985
0.9 0.983 0.866

0.8 0.852

0.7

C HAc / C t,HAc
0.6

0.5
0.392
0.4

0.3 0.365

0.2

0.1

0
2 3 4 5 6 7
pH

Figure 4. Acetic acid partitioning as a function of pH, shown as the fraction of undissociated acid to total acetate
concentration (the sum of undissociated form and acetate ion).

3.2. ELECTROCHEMICAL REACTIONS

Despite many decades of intense research, our understanding of the exact mechanism of

electrochemical reactions underlying mild steel corrosion in aqueous CO2 and H2S solutions
42,53–57
containing organic acids is still actively evolving . The model presented below was based

on the mainstream understanding of the electrochemical reaction mechanisms that have prevailed

in the past few decades.

It has been known for a long time that the spontaneous iron dissolution, causing the deterioration

of the steel infrastructure, is the key oxidation (anodic) process, while a family of hydrogen

evolving reactions are the main reduction (cathodic) reactions that provide the required electron

sink for the iron dissolution to progress. That includes the reduction of H+, carbonic species,

H2S, carboxylic acids, H2O, and other carboxylic acids, as shown in Table 3.
Table 3. Electrochemical reactions associated with aqueous acidic corrosion of mild steel in presence of various
acidic species.
Acidic solutions Iron dissolution Fe2+ (aq) + 2e− ⇋ Fe(s) ( 20 )
1
Hydrogen ion reduction H + (aq) + e− ⇋ H ( 21 )
2 2 (g)
Acidic solutions
1
Water reduction H2 O (l) + e− ⇋ H2 (g) + OH − (aq) ( 22 )
2
1
Carbonic acid reduction H2 CO3(aq) + e− ⇋ H2 (g) + HCO− 3 (aq)
( 23 )
CO2 containing 2
solutions 1
Bicarbonate ion reduction HCO3− (aq) + e− ⇋ H2 (g) + CO2− 3 (aq) ( 24 )
2
H2S containing 1 −
Hydrogen sulfide reduction 𝐻𝐻2 𝑆𝑆(𝑎𝑎𝑎𝑎) + 𝑒𝑒 − ⇋ 𝐻𝐻 + 𝐻𝐻𝑆𝑆(𝑎𝑎𝑎𝑎) ( 25 )
solutions 2 2 (𝑔𝑔)
Acetic acid 1 −
Acetic acid reduction 𝐻𝐻𝐻𝐻𝐻𝐻(𝑎𝑎𝑎𝑎) + 𝑒𝑒 − ⇋ 𝐻𝐻 + 𝐴𝐴𝐴𝐴(𝑎𝑎𝑎𝑎) ( 26 )
containing solutions 2 2 (𝑔𝑔)

It is worthwhile pointing out that these hydrogen-evolving reactions are thermodynamically

identical. That can be readily demonstrating by writing the Nernst equation for each of them,

which results in identical reversible potentials, when concentrations of the involved chemical

species are defined by chemical equilibria. Therefore, the main difference among these reactions

is in the charge transfer kinetics. Considering that H2 concentration in such systems is negligible,

the oxidation half-reactions can be disregarded. Therefore, in order to generalize the approach,

the cathodic current from reduction of a generic weak acid “HA” in the form of Reaction ( 27 ),

can be expressed via Equation ( 28 ).

1 ( 27 )
𝐻𝐻𝐻𝐻 + 𝑛𝑛𝑛𝑛 − → 𝐻𝐻2 + 𝐴𝐴−
2

Δ𝐻𝐻 1
�− 𝐻𝐻𝐻𝐻 � −
1 𝑠𝑠 𝑝𝑝 − (𝐸𝐸𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 −𝐸𝐸𝑟𝑟𝑟𝑟𝑟𝑟,𝐻𝐻𝐻𝐻 ) ( 28 )
𝑅𝑅 𝑇𝑇 𝑇𝑇𝑟𝑟𝑟𝑟𝑟𝑟
�� 𝐶𝐶𝐻𝐻𝐻𝐻 ( )
𝑖𝑖𝐻𝐻𝐻𝐻 = −𝑖𝑖0,𝐻𝐻𝐻𝐻 𝑒𝑒 � � 10 𝑏𝑏𝐻𝐻𝐻𝐻
𝑏𝑏
𝐶𝐶𝐻𝐻𝐻𝐻,𝑟𝑟𝑟𝑟𝑟𝑟
𝑠𝑠
where 𝑖𝑖0,𝐻𝐻𝐻𝐻 is the reference exchange current density, 𝐶𝐶𝐻𝐻𝐻𝐻 is the concentration of the reactant at

𝑏𝑏
the metal surface, 𝐶𝐶𝐻𝐻𝐻𝐻,𝑟𝑟𝑟𝑟𝑟𝑟 is the bulk concentration at reference condition, p is the apparent

reaction order, 𝐸𝐸𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠. is the electrical potential at the metal surface, 𝐸𝐸𝑟𝑟𝑟𝑟𝑟𝑟,𝐻𝐻𝐻𝐻 is the reference

potential, and other parameters have their common electrochemical meanings. The

electrochemical parameters used charge transfer rate calculations are listed in Table 4. In the

present model the contribution of water reduction reaction and bicarbonate ion reduction reaction

to the cathodic current were assumed to be negligible in acidic environments and were not

included in the calculations.

Table 4. Electrochemical parameters for cathodic hydrogen-evolving reactions used in the present model.

𝑏𝑏
𝑖𝑖0,𝑟𝑟𝑟𝑟𝑟𝑟 𝑐𝑐𝑟𝑟𝑟𝑟𝑟𝑟 𝐸𝐸 𝑟𝑟𝑟𝑟𝑟𝑟 ∆𝐻𝐻
𝑝𝑝 𝑏𝑏 𝑇𝑇𝑟𝑟𝑟𝑟𝑟𝑟 (𝐾𝐾)
(𝐴𝐴. 𝑚𝑚2 ) (𝑀𝑀) (𝑉𝑉) (𝑘𝑘𝑘𝑘. 𝑚𝑚𝑚𝑚𝑚𝑚 −1 )
𝐻𝐻 + 0.02 1.0 0.0001 -0.24 2.3𝑅𝑅𝑅𝑅/0.5𝐹𝐹 30 293.15
𝐻𝐻2 𝑆𝑆 0.0006 0.2 0.0001 -0.24 2.3𝑅𝑅𝑅𝑅/0.5𝐹𝐹 60 293.15
𝐻𝐻2 𝐶𝐶𝐶𝐶3 0.014 1.0 0.0001 -0.24 2.3𝑅𝑅𝑅𝑅/0.5𝐹𝐹 55 293.15
𝐻𝐻𝐻𝐻𝐻𝐻 0.04 0.5 0.0014 -0.24 2.3𝑅𝑅𝑅𝑅/0.5𝐹𝐹 55 293.15
𝐹𝐹𝐹𝐹 1 0 NA -0.488 2.3𝑅𝑅𝑅𝑅/1.5𝐹𝐹 37.5 293.15

When it comes to the dominating anodic reaction: iron dissolution (Reaction ( 20 )), the reverse

(cathodic) half-reaction: iron deposition is assumed to be negligible at the typical open circuit

potentials seen in CO2/H2S/HAc corrosion and can be ignored. The mechanism of iron oxidation

reaction in acidic media has been the subject of numerous studies over the last half a century 58–
69
, and has been proved difficult to explain. A detailed review of the literature is beyond the

scope of the present discussion, however, the interested readers may find a wealth of information

on this subject elsewhere 68,70.


There are two main classic mechanisms proposed for iron dissolution in acidic solutions: the

“catalytic mechanism” and the “consecutive mechanism”. These two mechanisms are associated

with two distinct electrochemical behaviors observed specifically in the active dissolution range.
71
The catalytic mechanism, first proposed by Heusler et al. , is based on the experimental Tafel

slope of 30 mV and second order dependence on hydroxide (OH-) ion concentration. On the

other hand, the consecutive mechanism proposed by Bockris et al. 66, was formulated to explain

the observed Tafel slope of 40 mV and a first order dependence on (OH-) ion concentration.

These two significantly different reaction kinetics are believed to be caused by the surface
59
activity of the iron electrode , i.e. the dissolution of cold-worked iron electrodes with high

internal stress occurs with a 30 mV Tafel slope, while a 40 mV Tafel slope was observed for

dissolution of annealed, recrystallized iron 59,67,69,70.

The anodic polarization curves obtained for mild steel dissolution in acidic CO2-saturated

environments have frequently been reported to have a 40 mV Tafel slope and a first order
12,13,15,72,73
dependence on hydroxide ion concentration , in accordance with the “consecutive

mechanism” proposed by Bockris et al. 66 . However, the mechanism proposed by Bockris et al.
66
is known to be only valid in acidic solutions of pH 4 or lower . It is well-known that the pH

dependence in this mechanism rapidly decreases and eventually vanishes at pH 5 and higher
62,66,67
. Considering that the great majority of the conditions in CO2/H2S/HAc corrosion fall

within this pH range, and due to a lack of a better understanding, the present model uses a

simplified rate expression. The rate of iron dissolution reaction is therefore calculated in the form

of Equation ( 28 ), with the constants shown in Table 4.


3.3. CORROSION PRODUCT LAYER FORMATION AND ITS

PROTECTIVENESS

Pipeline corrosion of mild steel is often accompanied by formation of a corrosion product layer

at the metal surface. The properties and composition of the corrosion product layer are affected

by numerous parameters such as water chemistry, temperature, fluid flow, steel composition and
74–82
microstructure, to name the most important ones . In the case of sweet corrosion, the

corrosion product layer is predominantly made of iron carbonate. The overall

precipitation/dissolution reaction for iron carbonate can be represented by a heterogeneous

chemical Reaction ( 29 ). The equilibrium for this reaction can be mathematically expressed via

Equation ( 30 ), where 𝐾𝐾𝑠𝑠𝑠𝑠,𝐹𝐹𝐹𝐹𝐹𝐹𝑂𝑂3 is the iron carbonate solubility (equilibrium) constant, shown in

Table 5.

𝐹𝐹𝐹𝐹 2+ (𝑎𝑎𝑎𝑎) + 𝐶𝐶𝐶𝐶32− (𝑎𝑎𝑎𝑎) ⇌ 𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹3(𝑠𝑠) ( 29 )

𝐾𝐾𝑠𝑠𝑠𝑠,𝐹𝐹𝐹𝐹𝐹𝐹𝑂𝑂3 = �𝐶𝐶𝐹𝐹𝐹𝐹 2+ 𝐶𝐶𝐶𝐶𝐶𝐶32− � ( 30 )


𝑠𝑠𝑠𝑠𝑠𝑠

Due to corrosion which releases ferrous ions, or due to increases in pH, temperature, or CO2

partial pressure, the solubility product of ferrous and carbonate ions concentrations in the

aqueous solution may exceed the saturation limit (𝐾𝐾𝑠𝑠𝑠𝑠,𝐹𝐹𝐹𝐹𝐹𝐹𝑂𝑂3 ) leading to net precipitation of iron

carbonate, i.e. Reaction ( 29 ) moves to the right. Given that the kinetics of iron carbonate

formation by precipitation is very slow, the resulting level of supersaturation can become quite

large. The extent of departure from equilibrium (as defined by Equation ( 30 )), is often

represented by the so-called “(super)saturation value” (𝑆𝑆𝐹𝐹𝐹𝐹𝐹𝐹𝑂𝑂3 ) defined as:


𝐶𝐶𝐹𝐹𝐹𝐹 2+ 𝐶𝐶𝐶𝐶𝐶𝐶32− ( 31 )
𝑆𝑆𝐹𝐹𝐹𝐹𝐹𝐹𝑂𝑂3 =
𝐾𝐾𝑠𝑠𝑠𝑠,𝐹𝐹𝐹𝐹𝐹𝐹𝑂𝑂3

The degree of supersaturation can be considered as a thermodynamic measure of the “driving

force” for iron carbonate precipitation 74. Therefore, the rate of iron carbonate precipitation is a

direct function of supersaturation but also strongly depends on temperature and surface to

volume ratio (i.e. available surface area for precipitation), as expressed by Equation ( 32 ) 83.

𝐴𝐴 ( 32 )
𝑃𝑃𝑃𝑃𝑖𝑖 = 𝑓𝑓(𝑇𝑇) 𝑔𝑔(𝑆𝑆𝑖𝑖 )
𝑉𝑉

In Equation ( 32 ), i is the precipitating species (in the discussion so far it is FeCO3), A/V is the

surface to volume ratio, and f(T) represents the temperature dependent precipitation rate constant

based on Arrhenius’ law. In the present model functions f and g for iron carbonate formation

were specified based on the findings of Sun and Nesic, as show in Table 5.

Table 5. Kinetic parameters of iron carbonate and iron sulfate precipitation rates.

parameter relationship reference

−64851.4
𝑓𝑓𝐹𝐹𝐹𝐹𝐹𝐹𝑂𝑂3 (𝑇𝑇) 84
𝑒𝑒 (28.2 − 𝑅𝑅𝑅𝑅
)

𝑔𝑔(𝑆𝑆𝐹𝐹𝐹𝐹𝐹𝐹𝑂𝑂3 ) 𝐾𝐾𝑠𝑠𝑠𝑠,𝐹𝐹𝐹𝐹𝐹𝐹𝑂𝑂3 (𝑆𝑆𝐹𝐹𝐹𝐹𝐹𝐹𝑂𝑂3 − 1) 84

𝑙𝑙𝑙𝑙𝑙𝑙𝐾𝐾𝑠𝑠𝑠𝑠,𝐹𝐹𝐹𝐹𝐹𝐹𝑂𝑂3 2.1963 77
−59.3498 − 0.041377𝑇𝑇 − + 24.5724𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 + 2.518𝐼𝐼0.5 − 0.657𝐼𝐼
𝑇𝑇

𝑓𝑓𝐹𝐹𝐹𝐹𝐹𝐹 (𝑇𝑇) 40,000 85


𝑒𝑒 (34.2− 𝑅𝑅𝑅𝑅
)

𝑔𝑔(𝑆𝑆𝐹𝐹𝐹𝐹𝐹𝐹 ) 𝐾𝐾𝑠𝑠𝑠𝑠,𝐹𝐹𝐹𝐹𝐹𝐹 (𝑆𝑆𝐹𝐹𝐹𝐹𝐹𝐹 − 1) 85

𝐾𝐾𝑠𝑠𝑠𝑠,𝐹𝐹𝐹𝐹𝐹𝐹 2848.779 86
� − 6.347�
10 𝑇𝑇 𝐾𝐾𝐻𝐻2 𝑆𝑆 𝐾𝐾𝐻𝐻𝐻𝐻 −

The porous iron carbonate precipitate on steel surface may affect the corrosion process in two

main ways:
• limiting the rate of mass transfer of the chemical species toward and away from the metal

surface, i.e. acting as a physical barrier.

• reducing the overall rate of electron transfer reactions (current density) by blocking

portions of the metal surface, making them unavailable as electrochemical reaction sites.

The degree of protection offered by a precipitated iron carbonate layer depends on its properties,

such as: porosity and adherence to the metal surface. Less porous (denser) iron carbonate layers

that are well adherent to the steel surface are more protective, and vice versa. These properties

are greatly affected by the kinetics of precipitation (faster precipitation leads to more protective

iron carbonate layers) but also by the corrosion rate of the mild steel “substrate” that undermines

the precipitating layer. The simplest way to quantify this process was proposed by van Hunnik et

al.74 as the so-called “scaling tendency” concept:

𝑃𝑃𝑃𝑃𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹3(𝑠𝑠) ( 33 )
𝑆𝑆𝑆𝑆 =
𝐶𝐶𝐶𝐶

where 𝑃𝑃𝑃𝑃𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹3(𝑠𝑠) is the precipitation rate and 𝐶𝐶𝐶𝐶 is the corrosion rate, both expressed in the same

volumetric units. This is a simple parameter that can serve as a practical indicator of the ability to
74,75,87
form a protective iron carbonate layer . The authors based their argument on the fact that

the formation of an iron carbonate layer never completely stops the corrosion process, what in

turn, causes the existing corrosion product layer to be undermined and sometimes detach from
74,88
the metal surface . The continuous process of precipitation, undermining and detachment

from the steel surface affects the adherence and porosity of the iron carbonate corrosion product

layer and ultimately its protectiveness.89 A scaling tendency of ST>>1 suggests that the

undermining by corrosion is overpowered by the rapidly forming iron carbonate precipitate,

creating a dense well attached and protective layer. On the other hand, a scaling tendency of
ST<<1 represents the case where the undermining by corrosion is much faster than the formation

of the corrosion product layer, therefore, only a porous, poorly attached, non-protective layer

may be formed, even at high (super)saturation values. 87,89.

Other factors may be important, such as the presence of an iron carbide (cementite) layer that can

serve as a suitable matrix for iron carbonate precipitation 74,81. Furthermore, the protectiveness of

an iron carbonate corrosion product layer can be compromised by various chemical and

mechanical removal processes 78,90,91.

The comprehensive mechanistic model of CO2 corrosion presented in this study accounts for the

formation iron carbonate corrosion product layer, its porosity and the protective effect.

Formation of iron carbonate layer by precipitation in the model is initiated when the local

supersaturation degree at the metal/solution interface exceeds unity (so that there is a driving

force), and when a solid (seed) surface is available for precipitation process to nucleate on.

Initially, this happens at the steel surface, which is a suitable substrate of iron carbonate

nucleation and growth. This is helped by the fact that the concentration of ferrous ions is highest

at the corroding steel surface due to electrochemical iron dissolution, and the concentration of

carbonate ions is also at its highest, due to a higher pH at the surface, all amounting to highest

local supersaturation. In the case of a mild steels with a robust iron carbide network that is

exposed by corrosion, this leads to even more favorable conditions for iron carbonate

precipitation. An exposed iron carbide network is a good nucleation site for iron carbonate, it

increases the level of supersaturation by presenting an additional diffusion barrier for species and

it diminishes/eliminates convective forces that may sweep away the corrosion products from the

steel surface.
As the iron carbonate corrosion product layer formed by precipitation presents a diffusion barrier

and blocks the steel surface, determining its porosity, 𝜀𝜀, distribution is the key to determining its

protectiveness. The other important property that governs the resistance to diffusion of dissolved

species through the porous layer is tortuosity, 𝜏𝜏, which is related to the shape of the pores. For

porous mineral structures it has been found that that the two are related, i.e. 𝜏𝜏 ≈ √𝜀𝜀. Generally,

highly porous corrosion product layers have a small effect on the corrosion rate, while layers

with low porosity (high density) are better barriers for transport of species and can reduce the

corrosion rate effectively. This is particularly true when dense iron carbonate layers form at the

steel surface, where in addition to forming a diffusion barrier, they lead to blocking of the

electrochemical reaction sites on the surface, thereby directly affecting the corrosion rate.

This background was used by Nesic et al. to propose a model for calculation of porosity
17,36
distribution in the iron carbonate layer by using a mass balance for solid iron carbonate

precipitate, which can be converted into an equation to calculate change of porosity over time

and in space, as:

𝜕𝜕𝜕𝜕 𝑀𝑀𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹3 𝜕𝜕𝜕𝜕 ( 34 )


=− 𝑃𝑃𝑃𝑃𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹3 − 𝐶𝐶𝐶𝐶
𝜕𝜕𝜕𝜕 𝜌𝜌𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹3 𝜕𝜕𝜕𝜕

The first term on right hand side is related to formation of the layer by precipitation (Equation (

32 )) and the second (convective-like) term arises from the undermining effect due to corrosion

of the steel substrate. This approach is broadly equivalent to using the concept of scaling

tendency as proposed by van Hunnik et al.74. The difference is that it is more physically realistic

and instead of obtaining a single parameter for characterizing the protectiveness of the corrosion

product layer (ST), the structure of the layer is predicted through calculating the distribution of

porosity in the layer over time. Also, the precipitation rate calculation is based on local
concentrations of species in the vicinity of the steel surface and in the porous corrosion product

layer, rather than basing it on bulk concentrations, as was originally done by van Hunnik et al.

(1996).

Using this approach, the porosity can be treated as an additional variable in the corrosion rate

calculations and its distribution through the diffusion layer can be obtained by solving Equation (

34 ) simultaneously with other relationship describing the potential and species concentration

distribution inside the boundary layer, as discussed in detail further below.

In the case of sour corrosion, various iron sulfides can form as corrosion products in H2S

corrosion of mild steel. These include amorphous ferrous sulfide (FeS), mackinawite (Fe1+xS),

cubic ferrous sulfide (FeS), troilite (FeS), pyrrhotites (Fe1-xS), smythite (Fe3+xS4), greigite

(Fe3S4), pyrite (FeS2) and marcasite (FeS2). Despite a large body of work being available on iron

sulfides, their role in corrosion remains unclear. When formed, iron sulfide corrosion product

layer acts as a diffusion barrier and block parts of the steel surface, just like iron carbonate does.

However, most iron sulfides found in corrosion of steel are electronic semi-conductors and allow

reduction of dissolved species on their surface. This produces galvanic effects due to enlarged

cathodic surface area. Therefore, the formation of iron sulfide corrosion product layers

sometimes does not readily lead to a marked decrease in corrosion rate and can even lead to

localized attack.

The first readily detectable iron sulfide that forms during corrosion of mild steel is mackinawite.

It is now believed that mackinawite forms by precipitation at the surface, although other theories

were put forward involving direct chemical reaction between iron in the steel and dissolved H2S

(a.k.a formation via a solid state reaction). It is true that mackinawite is ubiquitous and is almost
always found on the steel surface in experiments involving H2S, even when the bulk conditions

are far from favoring the precipitation (e.g. low pH). However, due to a much higher pH and

ferrous ion concentration at the corroding steel surface (particularly in low flow or stagnant

conditions), solubility of mackinawite is readily exceeded locally. Given the fast kinetics of

mackinawite precipitation (at least an order of magnitude faster than iron carbonate),

mackinawite forms first and then converts into other more stable forms such troilite, pyrrhotite

and pyrite.

As there is no clear understanding of whether one type of iron sulfide is any different from the

next, when it comes to corrosion product layers, the formation of a generic iron sulfide (FeS) is

implemented in the present model, much in the same way as was described above for carbonate,

via a precipitation reaction:

𝐹𝐹𝐹𝐹 2+ 𝑎𝑎𝑎𝑎 + 𝑆𝑆 2− (𝑎𝑎𝑎𝑎) ⇌ 𝐹𝐹𝐹𝐹𝐹𝐹(𝑠𝑠) ( 35 )

The solubility product for FeS can therefore be defined as shown in Equation ( 36 ), where Ksp,FeS

is the solubility constant of iron sulfide (Table 5).

𝐾𝐾𝑠𝑠𝑠𝑠,𝐹𝐹𝐹𝐹𝐹𝐹 = (𝐶𝐶𝐹𝐹𝐹𝐹 2+ 𝐶𝐶𝑆𝑆 2− )𝑠𝑠𝑠𝑠𝑠𝑠 ( 36 )

Similar to the case of iron carbonate layer formation, the main driving force for precipitation is

iron sulfide supersaturation 𝑆𝑆𝐹𝐹𝐹𝐹𝐹𝐹 , defined as:

𝐶𝐶𝐹𝐹𝐹𝐹 2+ 𝐶𝐶𝑆𝑆 2− ( 37 )
𝑆𝑆𝐹𝐹𝐹𝐹𝐹𝐹 =
𝐾𝐾𝑠𝑠𝑠𝑠,𝐹𝐹𝐹𝐹𝐹𝐹
As noted before, the precipitation rate of iron sulfide may also be expressed in term of the

generic formulation, shown via Equation ( 32 ). The corresponding parameters, f and g, for the

case of iron sulfide formation are listed in Table 5.

In mixed iron carbonate – iron sulfide precipitation, the buildup of the corrosion product layer is

affected by both processes. Therefore, Equation ( 34 ) can be expanded to accommodate for

simultaneous formation of both deposits as shown in Equation ( 38 ). Nevertheless, in most

cases of mixed precipitation, the layer is dominated by iron sulfide, due to its much faster

formation kinetics and the resulting competition for the same precursor cation (Fe2+).

𝜕𝜕𝜕𝜕 𝑀𝑀𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹3 𝑀𝑀FeS 𝜕𝜕𝜕𝜕 ( 38 )


=− 𝑃𝑃𝑃𝑃𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹3 − 𝑃𝑃𝑃𝑃𝐹𝐹𝑒𝑒𝑒𝑒 − 𝐶𝐶𝐶𝐶
𝜕𝜕𝜕𝜕 𝜌𝜌𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹3 𝜌𝜌FeS 𝜕𝜕𝜕𝜕

Iron acetate and other salts of small molecular weight organic acids are highly soluble in aqueous

solutions and therefore their formation is never seen in the pH range of interest for the present

application and does not need to be considered in the model. The approach described above for

precipitation of iron carbonate and iron sulfide can readily be extended to include other scale

forming species such as calcium carbonate, barium sulfate, etc., in order to account for their

effect on the corrosion rate.

3.4. MASS TRANSFER IN A CORRODING SYSTEM

During the corrosion process, the concentration of the aqueous species at the metal surface

deviate from those in the bulk solution, as a consequence of heterogeneous electrochemical

reactions occurring at the metal surface. Based on the known concentration of species in the

bulk, the comprehensive mathematical models are able to accurately calculate the surface

concentration of chemical species by considering the simultaneous mass transfer of multiple


species between the bulk and the surface and by accounting for chemical reactions amongst

them.

The mass transfer of species in corroding systems, or electrochemical systems in general, occurs

via three simultaneous mechanisms: convection – macroscopic movement of the bulk fluid

carrying the species; molecular diffusion – a result of the concentration gradient of the species;

electromigration – movement of the ions arising from the presence of an induced or a

spontaneously occurring electric field. Hence, the flux of any given species i can be described

through Equation ( 39 ) 92.

𝑁𝑁𝑖𝑖 = −𝑧𝑧𝑖𝑖 𝑢𝑢𝑖𝑖 𝐹𝐹𝐶𝐶𝑖𝑖 ∇𝜙𝜙 − 𝐷𝐷𝑖𝑖 ∇𝐶𝐶𝑖𝑖 + 𝑣𝑣𝐶𝐶𝑖𝑖 ( 39 )

The concentration change of any chemical species i in an elementary volume of the solution at

any location, can therefore be calculated through the balance of the fluxes of that species for that

volume (as it is going in and out) further corrected for the consumption/production of that

species through homogeneous chemical reactions. This is mathematically expressed via the
92
Nernst-Planck equation, which needs to be written for every species i in the system. This

constitutes a set of i vector (3-D) equations for i species. However, in most practical applications

we can ignore the lateral flux components and focus only on the direction perpendicular to the

metal surface – making the set of equations scalar (1-D). Furthermore, the electrical mobility of

ions can be estimated by using Nernst-Einstein relationship (ui=Di/RT). Therefore, for a one-

dimensional semi-infinite geometry in the direction x normal to the metal surface, Equation ( 39 )

can be simplified as following:

𝜕𝜕𝐶𝐶𝑖𝑖 𝑧𝑧𝑖𝑖 𝐷𝐷𝑖𝑖 𝐹𝐹𝐶𝐶𝑖𝑖 𝜕𝜕𝜙𝜙 ( 40 )


𝑁𝑁𝑖𝑖 = − 𝐷𝐷𝑖𝑖 − + 𝑣𝑣𝑥𝑥 𝐶𝐶𝑖𝑖
𝜕𝜕𝜕𝜕 𝑅𝑅𝑅𝑅 𝜕𝜕𝜕𝜕
The average bulk movement of the fluid in the direction normal to the surface is accounted for in

the convective flow term (vxC), where vx describes the velocity profile inside the diffusion layer.

However, in the case of corrosion in pipelines, the dominant mass transfer mechanism in the

bulk solution is in the form of turbulent mixing, which then decays as the solid wall is

approached – in the diffusion boundary layer. The turbulent mixing of the fluid can be grossly

simplified and expressed via a simple eddy diffusivity profile within the diffusion boundary

layer. The mathematical relationships for eddy diffusivity of turbulent flow through straight
93,94
tubes have been developed in a number of previous studies . A simple expression for eddy

diffusivity (Dt) at distance x (m) from the wall, and diffusion layer thickness (δ) is shown in
𝜇𝜇
Equation ( 41 ) and Equation ( 42 ), respectively 94. Here, 𝜐𝜐 = �𝜌𝜌 is the kinematic viscosity of

water in m2.s-1, which can be calculated based on parameters shown in Table 6, 𝑅𝑅𝑅𝑅 is the

Reynolds number, and d is the pipe diameter (m).

𝑥𝑥 ( 41 )
𝐷𝐷𝑡𝑡 = 0.18( )3 𝜐𝜐
𝛿𝛿

−7� ( 42 )
𝛿𝛿 = 25𝑅𝑅𝑅𝑅 8 𝑑𝑑

The eddy diffusivity (Dt) can then be lumped with molecular diffusion (Di) in Equation ( 40 ), in

order to account for the turbulent mixing, to replace the convective flow term (vxC):

𝜕𝜕𝐶𝐶𝑖𝑖 𝑧𝑧𝑖𝑖 𝐷𝐷𝑖𝑖 𝐹𝐹𝐶𝐶𝑖𝑖 𝜕𝜕𝜙𝜙 ( 43 )


𝑁𝑁𝑖𝑖 = − (𝐷𝐷𝑖𝑖 + 𝐷𝐷𝑡𝑡 ) − + 𝑣𝑣𝑥𝑥 𝐶𝐶𝑖𝑖
𝜕𝜕𝜕𝜕 𝑅𝑅𝑅𝑅 𝜕𝜕𝜕𝜕

The concentration distribution of the involved chemical species inside the boundary layer can be

defined based on mass conservation law using the flux Equation ( 43 ). When discussing the

corrosion in the presence of a corrosion produce layer with porosity 𝜀𝜀, the mass conservation of
any given species i can be expressed as Equation ( 44 ). Noting that the same one dimensional

assumption applied for the flux equation can be used to further simplify Equation ( 44 ), as well.

𝜕𝜕(𝜀𝜀𝜀𝜀𝑖𝑖 ) ( 44 )
= −∇. (𝜀𝜀 1.5 𝑁𝑁𝑖𝑖 ) + 𝜀𝜀𝜀𝜀𝑖𝑖
𝜕𝜕𝜕𝜕

The 𝜀𝜀 1.5 multiplier in the flux term is a result of accounting for both porosity and tortuosity

effects on molecular diffusion of species, and by assuming that the tortuosity is proportional to

√𝜀𝜀. In the absence of a porous medium, where 𝜀𝜀 = 1, Equation ( 44 ) is simplified to the well-

known Nernst-Planck equation. It is worthwhile to note that the porosity in the present

discussion is a separate variable, which is obtained by resolving Equation ( 38 ) and generally

varies with the distance from the metal surfaces.

Table 6. Temperature dependence of the physiochemical properties.


Parameter Relationship Reference
Water density (kg/m3) 𝜌𝜌𝑤𝑤 = 753.596 + 1.87748 𝑇𝑇 − 0.003562 𝑇𝑇 2 2
2
1.1709 �𝑇𝑇𝑟𝑟𝑟𝑟𝑟𝑟 −𝑇𝑇�−0.001827�𝑇𝑇𝑟𝑟𝑟𝑟𝑟𝑟 −𝑇𝑇�
( )
Water viscosity (cP) 𝜇𝜇 = 𝜇𝜇𝑟𝑟𝑟𝑟𝑟𝑟 10 (𝑇𝑇−273.15)+89.93 95

𝑇𝑇𝑟𝑟𝑟𝑟𝑟𝑟 = 293.15 𝐾𝐾, 𝜇𝜇𝑟𝑟𝑟𝑟𝑟𝑟 = 1.002 𝑐𝑐𝑐𝑐


𝑇𝑇 𝜇𝜇𝑟𝑟𝑟𝑟𝑟𝑟
Diffusion coefficient 𝐷𝐷𝑖𝑖 = 𝐷𝐷𝑖𝑖,𝑟𝑟𝑟𝑟𝑟𝑟
𝑇𝑇𝑟𝑟𝑟𝑟𝑟𝑟 𝜇𝜇

The Ri term in Equation ( 44 ) represent the effect of the homogeneous chemical reactions inside

the boundary layer. An accurate account of the homogeneous chemical reactions involved in the

complex water chemistry of CO2/H2S/HAc containing solutions is essential for calculating the

surface concentration of the chemical species. This is of significance, since the buffering system

of the solution containing weak acids such as carbonic acid, carboxylic acids, and hydrogen

sulfide may act as an additional source (or sink) for the chemical species as their concentrations

depart from the equilibrium seen in the bulk solution. Using the same generic format for

homogenous chemical reactions, as introduced by Reaction ( 16 ) above, the rate of any given
chemical reaction j can be calculated by Equation ( 45 ), where kf,j is the “forward” reaction rate

(rate of reaction j from left to right) and kb,j is the backward reaction rate (rate of reaction j from

right to left). Noting that at equilibrium, where Rj = 0, Equitation ( 45 ) simplifies to Equation (

17 ) to represent the reaction at chemical equilibrium.

𝑛𝑛𝑟𝑟 𝑛𝑛𝑝𝑝
( 45 )
𝑅𝑅𝑗𝑗 = 𝑘𝑘𝑓𝑓,𝑗𝑗 � 𝐶𝐶𝑟𝑟 − 𝑘𝑘𝑏𝑏,𝑗𝑗 � 𝐶𝐶𝑝𝑝
𝑟𝑟=1 𝑝𝑝=1

By simple mathematical manipulation, the rate of production (or consumption) of any given

species i via j chemical reactions (Ri,j), may be expressed in a matrix format. As an example, the

underlying chemical reactions of CO2/H2S/HAc corroding system (Reactions ( 6 ) to ( 14 )) can

be expressed as following:

𝑅𝑅𝐶𝐶𝑂𝑂2(𝑎𝑎𝑎𝑎)
⎤ ⎡−1 0 0 0 0 0 0 ( 46 )
⎡ ⎤
⎢ 𝑅𝑅𝐻𝐻(𝑎𝑎𝑎𝑎)
+ ⎥ ⎢0 1 1 1 1 1 1⎥

𝑅𝑅𝐻𝐻2 𝐶𝐶𝐶𝐶3(𝑎𝑎𝑎𝑎)
⎥ ⎢ ⎥ ⎡ 𝑅𝑅𝐶𝐶𝐶𝐶2 ,ℎ𝑦𝑦𝑦𝑦 ⎤
⎢ ⎥ ⎢ 1 −1 0 0 0 0 0 ⎥ ⎢ 𝑅𝑅
⎢ 𝑅𝑅𝐻𝐻𝐶𝐶𝐶𝐶− ⎥ ⎢ 0 𝑐𝑐𝑐𝑐 ⎥
3(𝑎𝑎𝑎𝑎) 1 −1 0 0 0 0⎥ ⎢ ⎥
⎢ ⎥ ⎢ ⎥ ⎢ 𝑅𝑅𝑏𝑏𝑏𝑏 ⎥
𝑅𝑅
⎢ 𝐶𝐶𝐶𝐶3(𝑎𝑎𝑎𝑎) ⎥ ⎢ 0
2− 0 1 0 0 0 0⎥
⎢ ⎥
⎢ 𝑅𝑅 ⎥=⎢ ⎥ × 𝑅𝑅𝐻𝐻2 𝑆𝑆,𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑
𝐻𝐻2 𝑆𝑆(𝑎𝑎𝑎𝑎) 0 0 0 −1 0 0 0 ⎢ ⎥
⎢ ⎥ ⎢ ⎥ 𝑅𝑅 −
⎢ 𝐻𝐻𝐻𝐻 ,𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑 ⎥
⎢ 𝑅𝑅𝐻𝐻𝐻𝐻(𝑎𝑎𝑎𝑎) ⎥ ⎢ 0

0 0 1 −1 0 0⎥ ⎢
⎢ ⎥ ⎢ ⎥ 𝑅𝑅𝐻𝐻𝐻𝐻𝐻𝐻,𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑 ⎥
𝑅𝑅 ⎢ ⎥
⎢ 𝐻𝐻𝐻𝐻𝐻𝐻(𝑎𝑎𝑎𝑎) ⎥ ⎢ 0 0 0 0 0 −1 0⎥
⎣ 𝑅𝑅 ⎦
⎢ 𝑅𝑅𝐴𝐴𝐴𝐴 − ⎥ ⎢0 0 0 0 0 1 0
⎥ 𝑤𝑤

⎥ ⎢ ⎥
(𝑎𝑎𝑎𝑎)

𝑅𝑅 −
⎣ 𝑂𝑂𝑂𝑂(𝑎𝑎𝑎𝑎) ⎦ ⎣ 0 0 0 0 0 0 1⎦

The kinetic rate constant of the homogeneous chemical reactions commonly encountered are

listed in Table 7.
Table 7. Kinetic rate constants of homogeneous chemical reactions. kf denotes the reaction progress
from left to right and K=kf/kb.
Reaction # Reaction rate constant Reference
11715
(7) 195.3−27.61𝑙𝑙𝑙𝑙𝑙𝑙𝑇𝑇𝐾𝐾 − 96
𝑘𝑘𝑓𝑓,ℎ𝑦𝑦𝑦𝑦 = 10 𝑇𝑇𝐾𝐾 (1/𝑠𝑠 )
−4 𝑇𝑇 2 +7.91×10−7 𝑇𝑇 3
(8) 𝑘𝑘𝑓𝑓,ca = 105.71+0.0526𝑇𝑇𝐶𝐶 −2.94×10 𝐶𝐶 𝐶𝐶 (1/𝑠𝑠) 97

(9) 𝑘𝑘𝑓𝑓,bi = 109 (1/𝑠𝑠) 2

( 11 ) 𝑘𝑘𝑓𝑓,𝐻𝐻2 𝑆𝑆 = 104 (1/𝑠𝑠) 2

( 12 ) 𝑘𝑘𝑓𝑓,𝐻𝐻𝐻𝐻 − = 1 (1/𝑠𝑠) 2

( 13 ) 𝑘𝑘𝑓𝑓,HAc = 3.2 × 105 (1/𝑠𝑠) 2

( 14 ) 𝑘𝑘𝑏𝑏,𝑤𝑤 = 7.85 × 1010 (1/𝑀𝑀. 𝑠𝑠) 98

Based on the discussion so far, the concentration of each chemical species involved in the

corrosion process can be determined from its corresponding mass conservation Equation ( 44 ).

The diffusion coefficients of the chemical species and their temperature dependence can be

found in Table 8 and Table 6, respectively.

Table 8. Reference diffusion coefficients at 25 oC (77˚F).


Diffusion coefficient in water
Species Reference
× 109 (m2/s)
𝐶𝐶𝐶𝐶2 1.92 99

𝐻𝐻2 𝐶𝐶𝐶𝐶3 2.00 2

𝐻𝐻𝐻𝐻𝐻𝐻3− 1.185 100

𝐶𝐶𝐶𝐶32− 0.923 100

𝐻𝐻 + 9.312 92

𝑂𝑂𝑂𝑂 − 5.273 100

𝑁𝑁𝑁𝑁+ 1.334 92

𝐶𝐶𝐶𝐶 − 2.032 92,100

𝐹𝐹𝐹𝐹 2+ 0.72 92

𝐻𝐻2 𝑆𝑆 1.93 101

𝐻𝐻𝐻𝐻 − 1.731 100

𝑆𝑆 2− 1.5 Estimated
𝐻𝐻𝐻𝐻𝐻𝐻 1.29 100

𝐴𝐴𝐴𝐴 − 1.089 100


However, for this system of equations to be complete, the electric potential appearing in the

electromigration term of Equation ( 43 ) must be specified. This can be done by using an

additional equation, which relates the electric potential in a medium with a uniform dielectric

constant to a given charge distribution, also known as the Poisson’s equation. For a medium with

porosity of :

𝐹𝐹 ( 47 )
∇. (𝜀𝜀 1.5 ∇𝜙𝜙) = −𝜀𝜀 � 𝑧𝑧𝑖𝑖 𝐶𝐶𝑖𝑖
𝜉𝜉
𝑖𝑖

where 𝜉𝜉 is the dielectric constant of the solution, F is the faraday’s constant, and other

parameters have their common electrochemical meaning.

Initial and Boundary conditions

Since all the transport Equations ( 44 ) and ( 47 ) are transient partial differential equation,

appropriate initial and boundary conditions need to be specified. At the initial time (t=0) it can be

assumed that a well-mixed solution comes into contact with the metal surface. Hence, the

concentrations of chemical species throughout the diffusion layer are initially constant, known

values, defined by the chemical equilibria of the solution as discussed above in the Section 3.1.

The boundary condition at the metal/solution interface is based on the defined fluxes and

includes all the electrochemical reaction rate calculations. For an electro-active chemical species,

the flux at the metal/solution boundary is equal to the rate of the corresponding electrochemical

reactions. For an electro-active species, i involved in j electrochemical reactions, the flux at the

metal surface can be described through equation ( 48 ).


𝑠𝑠𝑖𝑖𝑖𝑖 𝑖𝑖𝑗𝑗 ( 48 )
𝑁𝑁𝑖𝑖 |𝑥𝑥=0 = − �
𝑛𝑛𝑗𝑗 𝐹𝐹
𝑗𝑗

The negative sign in Equation ( 48 ) accounts for the sign conventions in current density, flux,

and stoichiometric coefficients. The current density of each electrochemical reaction can be

calculated based on the relationships shown in Section 3.2. Similarly to what is done with the

homogeneous chemical reactions, the set of Equations ( 48 ) can be transformed into a matrix

form in order to include all the electro-active species:

𝑁𝑁 2+ |𝑥𝑥=0 1 0 0 0 0 0 0 ( 49 )
⎡ 𝐹𝐹𝐹𝐹𝑎𝑎𝑎𝑎 ⎤ ⎡ ⎤
⎢ 𝑁𝑁𝐻𝐻𝑎𝑎𝑎𝑎+ |𝑥𝑥=0 ⎥ ⎢0 −1 0 0 0 0 0⎥ 𝑖𝑖𝐹𝐹𝐹𝐹�
⎢ ⎥ ⎢ ⎡ 2𝐹𝐹 ⎤
𝑁𝑁𝐻𝐻2 𝐶𝐶𝐶𝐶3,𝑎𝑎𝑎𝑎 |𝑥𝑥=0 ⎥
⎢ ⎥ ⎢ 0 0 −1 0 0 0 0 ⎥ ⎢ 𝑖𝑖𝐻𝐻 � ⎥
+

⎢ 𝑁𝑁𝐻𝐻𝐶𝐶𝐶𝐶− |𝑥𝑥=0 ⎥ ⎢0 0 ⎢ 𝐹𝐹 ⎥
3,𝑎𝑎𝑎𝑎 1 −1 0 0 0⎥ ⎢ 𝑖𝑖𝑐𝑐𝑐𝑐 ⎥
⎢ ⎥ ⎢ ⎥ �𝐹𝐹
⎢ 𝑁𝑁𝐶𝐶𝐶𝐶3,𝑎𝑎𝑎𝑎
2− |𝑥𝑥=0
⎥ ⎢0 0 0 1 0 0 0⎥ ⎢ 𝑖𝑖 ⎥
= × ⎢ 𝑏𝑏𝑏𝑏� ⎥
⎢ 𝑁𝑁 ⎥ ⎢ ⎥ 𝐹𝐹 ⎥
𝐻𝐻2 𝑆𝑆𝑎𝑎𝑎𝑎 |𝑥𝑥=0 0 −1 0 0⎥ ⎢
⎢ ⎥ ⎢0 0 0
⎢ 𝑖𝑖
⎢0 0
𝐻𝐻 2 𝑆𝑆� ⎥
⎢ 𝑁𝑁𝐻𝐻𝐻𝐻𝑎𝑎𝑎𝑎
− |𝑥𝑥=0 ⎥ 0 0 1 0 0⎥ ⎢ 𝐹𝐹 ⎥
⎢ ⎥ ⎢ ⎥ 𝑖𝑖𝐻𝐻𝐻𝐻𝐻𝐻�
⎢ 𝑁𝑁𝐻𝐻𝐻𝐻𝐻𝐻𝑎𝑎𝑎𝑎 |𝑥𝑥=0 ⎥ ⎢0 0 0 0 0 −1 0⎥ ⎢ 𝐹𝐹 ⎥
⎢ 𝑖𝑖 ⎥
⎢ 𝑁𝑁𝐴𝐴𝐴𝐴 − |𝑥𝑥=0 ⎥ ⎢0 0 ⎥

𝑤𝑤�
0 0 0 1 0 𝐹𝐹 ⎦
⎥ ⎢ ⎥
𝑎𝑎𝑎𝑎

⎣ 𝑁𝑁𝑂𝑂𝑂𝑂𝑎𝑎𝑎𝑎
− |𝑥𝑥=0
⎦ ⎣0 0 0 0 0 0 1⎦

For non-electroactive species the flux at the metal surface is zero:

𝑁𝑁𝑖𝑖 |𝑥𝑥=0 = 0 ( 50 )

Equation ( 48 ) and Equation ( 50 ) can be used to describe the mass transfer boundary condition

for all chemical species at the metal surface. The electric potential at the metal/solution is

defined by an arbitrary constant reference value (0 V).

Considering the governing equations, the initial conditions, and the boundary conditions

discussed above, this set of equations is fully specified if the potential at the metal surface (Esurf

in Equations ( 28 )) is known so that the rate of electrochemical reactions can be calculated. In


the present model this parameter, which is also known as the corrosion potential, is not known a

priori. Hence, an additional relationship is required to calculate it: the charge conservation at the

metal surface. At corrosion potential, all the cathodic (reduction) rates/currents are balanced by

the anodic (oxidation) rates/currents, meaning that the net current resulting from all j

electrochemical reactions is equal to zero. Therefore, the potential at the metal surface (Esur) is

found to satisfy this condition. The charge conservation can be mathematically expressed as

Equation ( 51 ):

� 𝑖𝑖𝑗𝑗 = 0 ( 51 )
𝑗𝑗

At the bulk solution boundary (x=δ) the concentration of chemical species remains unchanged at

all times (for t≥0). Therefore, the boundary condition can be defined for the bulk solution based

on the known concentration of species, and is identical to the initial condition. The solution

potential at the bulk boundary can be defined based on the Poisson’s equation. Considering that

the bulk solution is electrochemically neutral (Equation ( 18 )), the charge density in the right

hand side of Equation ( 47 ) is zero. Hence, the solution potential at the bulk boundary can be

specified as:

∇2 𝜙𝜙|𝑥𝑥=𝛿𝛿 = 0 ( 52 )

3.5. MODEL IMPLEMENTATION

The mathematical expressions describing the corrosion inside transmission pipelines based on

the above discussions are summarized in Table 9. For each chemical species present in the

solution, one mass conservation equation and its corresponding boundary and initial conditions

are included in the model. Additionally, the potential distribution within the boundary layer is
obtained based on the Poisson’s equation. These equations form a system of coupled, non-linear,

partial differential equations. Considering the one-dimensional geometry of equations, the

solution can be obtained using the finite difference method. This method has been widely used in

similar systems, and proven effective and efficient for such calculations 2,42,53,92,102.

Table 9. Summary of equations used in the comprehensive mathematical model.

Electrode surface boundary

𝑠𝑠𝑖𝑖𝑖𝑖 𝑖𝑖𝑗𝑗
𝑁𝑁𝑖𝑖 = − � For electroactive species
𝑛𝑛𝑗𝑗 𝐹𝐹
𝑗𝑗

𝑁𝑁𝑖𝑖 = 0 For non-active species

Φ=0

� 𝑖𝑖𝑗𝑗 = 0
𝑗𝑗
𝜕𝜕𝜕𝜕 𝑀𝑀𝐹𝐹𝐹𝐹𝐹𝐹3 𝑀𝑀FeS 𝜕𝜕𝜕𝜕
=− 𝑃𝑃𝑃𝑃𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹3 − 𝑃𝑃𝑃𝑃𝐹𝐹𝐹𝐹𝐹𝐹 − 𝐶𝐶𝐶𝐶
𝜕𝜕𝜕𝜕 𝜌𝜌𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹3 𝜌𝜌FeS 𝜕𝜕𝜕𝜕

Boundary layer
𝜕𝜕(𝜀𝜀𝜀𝜀𝑖𝑖 )
= −∇. (𝜀𝜀 1.5 𝑁𝑁𝑖𝑖 ) + 𝜀𝜀𝜀𝜀𝑖𝑖
𝜕𝜕𝜕𝜕 For all species

𝐹𝐹
∇. (𝜀𝜀 1.5 ∇𝜙𝜙) = −𝜀𝜀 � 𝑧𝑧𝑖𝑖 𝐶𝐶𝑖𝑖
𝜉𝜉
𝑖𝑖
𝜕𝜕𝜕𝜕 𝑀𝑀𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹3 𝑀𝑀FeS 𝜕𝜕𝜕𝜕
=− 𝑃𝑃𝑃𝑃𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹3 − 𝑃𝑃𝑃𝑃𝐹𝐹𝐹𝐹𝐹𝐹 − 𝐶𝐶𝐶𝐶
𝜕𝜕𝜕𝜕 𝜌𝜌𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹3 𝜌𝜌FeS 𝜕𝜕𝜕𝜕

Bulk boundary and Initial condition

𝐶𝐶𝑖𝑖 = 𝐶𝐶𝑖𝑖𝑏𝑏
For all species

∇2 𝜙𝜙|𝑥𝑥=𝛿𝛿 = 0

𝜀𝜀 = 1

92
The solution algorithm used in the present model is similar to that introduced by Newman ,

which can be used as an initial reference by an interested reader. The first step in implementation
is to discretize the partial differential equation. i.e. convert them into algebraic equations, using

Taylor’s series approximations to approximate the partial derivatives with respect to space and

time. In the present model, the spatial domain derivatives are discretized using second order

approximation, and the temporal derivatives are expressed using first order approximation.

The set of algebraic equations can be resolved using many different solution methods, generally

by converting them into a matrix form, i.e. by constructing a coefficient matrix that is multiplied

by the unknown variables (species concentrations and potential). The solution can be obtained

readily by inverting the coefficient matrix.

The presence of non-linear terms, such as those arising from chemical reactions or the

electromigration term introduces additional complexities. The presence of non-linear terms

means that some of the elements in coefficient matrix contain the unknown variables (i.e.

unknown species concentrations and potential). In a simple approach, the coefficient matrix

(hence, the inverse matrix) can be obtained by assuming (guessing) an initial value for the

unknown variables and then by iterating. This semi-implicit method, while valid, requires fine
42,53,102
temporal steps for the iterations to converge . However, the target application of the

present model is directed towards long simulation times in order to predicted the corrosion rates

over years and even decades, where fine temporal resolution is not needed nor practical, as it

would result in unacceptably long computational time. The alternative fully implicit method that

is more suitable and was used here is based on the so-called linearization of such non-linear

terms. An example of this linearization for a chemical reaction term, using Taylor’s series

expansion is shown below. The superscripts represent the different time steps with n being the

current step, Ri is the rate of production/consumption of species i through chemical reactions as


shown in Equation ( 46 ), c is the concentration of the chemical species, and m in the total

number of chemical species.

𝑚𝑚
𝜕𝜕𝑅𝑅𝑖𝑖 𝑛𝑛−1 𝑛𝑛 ( 53 )
𝑅𝑅𝑖𝑖𝑛𝑛 = 𝑅𝑅𝑖𝑖𝑛𝑛−1 +�� � �𝑐𝑐𝑘𝑘 − 𝑐𝑐𝑘𝑘𝑛𝑛−1 �
𝜕𝜕𝑐𝑐𝑘𝑘
𝑘𝑘=1

4. VERIFICATION OF RESULTS

To verify the performance of this comprehensive model, a large experimental database available

at the Institute for Corrosion and Multiphase Technology, Ohio University was used. Selected

comparisons between the predictions made by the model, as implemented in the software

package MULTICORP™ and the experimental results are presented, below to illustrate the

performance of the model, its strengths as well as areas where improvement is required. The

corrosion rate is presented as a function of the key parameters, such as the presence of CO2, H2S

and HAc, pH, pressure, temperature, velocity and time. The experimental data come from LPR

measurements, which have been done at least in duplicate and verified with weight loss

measurements. The exception are cases where the corrosion rate changed with time significantly

over the course of long experiments (due to protective corrosion product layer formation) and the

weight loss data did not offer a meaningful way to validate the LPR measurements.

For aqueous CO2 solutions, the change in corrosion rate with temperature and pH is shown in

Figure 5. In the experiments, no protective iron carbonate corrosion product layers formed even

at pH6 due to low Fe2+ concentration and relatively short exposures. Generally it can be observed

that the corrosion rate increased with temperature and decreased with pH. This is to be expected

as temperature accelerates all the physicochemical processes underlying corrosion and decreased

pH corresponds to a higher concentration of corrosive H+ ions. The performance of the model


can be deemed as reasonable and in most cases within the error of measurement. Some deviation

is seen at the “extremes”, i.e. for the combination of lowest and highest temperature/pH. Such

discrepancies were difficult to eliminate altogether, given that the model was not “tuned” to

match any particular set of experimental data, but was rather calibrated for optimal performance

over a wide range of operating parameters.

10

pH 4
7
Corrosion rate / (mm.yr-1)

4 pH 5

3 pH 6

0
10 20 30 40 50 60 70 80 90
Temperature / C

Figure 5. The CO2 corrosion rate as a function of temperature and pH; solid lines are generated by the model
(MULTICORP™ 5.5), points represent LPR experimental data taken from Nesic et al. 103: flow loop experiments,
v=2 m/s in a 0.0254 ID pipe, Fe2+< 1 ppm, [NaCl]=1 wt%, ptotal=1bar, note that pCO2 varies with temperature,
e.g. it is 0.98 bar at 20oC, 0.88 bar at 50oC and 0.53 bar at 80oC, due to the change in water vapor partial pressure,
pH2S=0 bar, [HAc]=0 ppm; experiments were repeated at least once and the error bars represent one standard
deviation.

In Figure 6, the effect of two other important parameters in aqueous CO2 corrosion of mild steel:

velocity and pCO2, is shown. Clearly, there is no major effect of velocity on the CO2 corrosion

rate, as the dominant electrochemical reactions, iron dissolution and reduction of carbonic acid

are not affected by flow. The slight flow dependence can be attributed to the reduction of H+ ions

whose limiting current is controlled by mass transfer, which is affected by turbulent pipe flow.
Formation of protective iron carbonate layers was avoided in these conditions due to a moderate

pH5 and low Fe2+ concentration. Therefore, the effect of increasing pCO2 is strong, with

extremely high corrosion rates obtained at higher pCO2, due to high concentration of carbonic

acid in the solution. The model captures both of these effects rather well, with some deviation at

the highest velocity and pCO2.

50

45

40

35 pCO2 = 20 bar
Corrosion rate / (mm.yr-1)

30

25
pCO2 = 10 bar
20

15

pCO2 = 3 bar
10

0
0 0.5 1 1.5 2
Velocity / (m.s-1)

Figure 6. The CO2 corrosion rate as a function of velocity and pCO2 in the absence of protective iron carbonate
corrosion product layers formation. Solid lines are generated by the model (MULTICORP™ 5.5), points
represent LPR experimental data taken from Wang et al.104. Conditions: flow loop experiments, 0.1 ID pipe,
T=60oC, pH5, Fe2+< 1 ppm, [NaCl]=1 wt%, pH2S=0 bar, [HAc]=0 ppm; experiments were repeated at least once
and the error bars represent minimum and maximum values.

When conditions are such that protective iron carbonate corrosion product layer does form, the

corrosion rate typically decreases with time, as shown in Figure 7. Due to high supersaturation of

the aqueous solution with iron carbonate at pH6.6, precipitation of solid iron carbonate leads to a

corrosion rate decrease by at least one order of magnitude over the course of a few days. The

porous iron carbonate corrosion product layer presents a diffusion barrier and blocks the
electrochemical reaction sites on the steel surface. This behavior is successfully captured by the

model, and both the trend and the magnitude of the corrosion rate change are successfully

predicted.

4.0

3.0
Corrosion rate / (mm.yr-1)

2.0

1.0

0.0
0 12 24 36 48 60 72
Time / hr

Figure 7. The decreasing CO2 corrosion rate with time due to formation of protective iron carbonate corrosion
product layers. Solid line is generated by the model (MULTICORP™ 5.5), points represent LPR data from five
repeated experiments, taken from Yang 105. Conditions: glass cell rotating cylinder experiments, equivalent
velocity v=0.63 m/s in a 0.1 ID pipe, T=80oC, pH6.6, Fe2+≈ 10 ppm, [NaCl]=1 wt%, pCO2=0.53 bar, pH2S=0 bar,
[HAc]=0 ppm.

The effect of H2S on corrosion rate, in the absence of formation of protective iron sulfide

corrosion product layers, is shown in Figure 8. The corrosion rate decreases with very small

amounts of aqueous H2S that adsorbs on the steel surface and interferes with the electrochemical

reactions. However, at partial pressure pH2S>0.1 mbar, this effect is overwhelmed by the

contribution of aqueous H2S to the overall reduction of corrosive species, thereby stimulating

more rapid dissolution of iron, increasing the corrosion rate. This behavior is captured by the

model successfully even if some discrepancies are seen at the very low H2S concentrations.
4.0

3.5

3.0

Corrosion rate / (mm.yr-1)


2.5

2.0

1.5

1.0

0.5

0.0
1.0E-07 1.0E-06 1.0E-05 1.0E-04 1.0E-03 1.0E-02 1.0E-01 1.0E+00
pH2S / bar

Figure 8. The H2S corrosion rate as a function of pH2S in the absence of protective iron sulfide corrosion product
layers formation. Solid line is generated by the model (MULTICORP™ 5.5), points represent LPR experimental
data taken from Zheng et al.31. Conditions: glass cell rotating cylinder experiments, equivalent velocity v=1.2 m/s
in a 0.0254 ID pipe, T=30oC, pH4, Fe2+< 1 ppm, [NaCl]=1 wt%, pCO2=0 bar, [HAc]=0 ppm; error bars represent
the standard deviation.

The effect of velocity on the H2S corrosion rate is presented in Figure 9. In contrast with CO2

corrosion (shown in Figure 6), a clear effect of velocity can be seen here, which is due to the fact

that aqueous H2S reduction limiting current is diffusion controlled and thereby affected by

turbulent flow. The model accounts for this, and therefore the simulations are in good agreement

with the experimental data.


3.0

2.5

Corrosion rate / (mm.yr-1)


2.0

1.5

1.0

0.5

0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Velocity /(m.s-1)

Figure 9. The H2S corrosion rate as a function of velocity in the absence of protective iron sulfide corrosion
product layers formation. Solid line is generated by the model (MULTICORP™ 5.5), points represent LPR
experimental data taken from Zheng et al.31. Conditions: glass cell rotating cylinder experiments (data show
equivalent velocity in a 0.0254 ID pipe), T=30oC, pH4, Fe2+< 1 ppm, [NaCl]=1 wt%, pH2S=0.00965 bar, pCO2=0
bar, [HAc]=0 ppm; error bars represent the standard deviation.

The pH effect in aqueous H2S corrosion of mild steel is shown in Figure 10. The marked

decrease in the corrosion rate between pH3 and pH4 by almost one order of magnitude

corresponds to the equivalent decrease of H+ ion concentration is solution, which is the main

cathodic species. This is not the case between pH4 and pH5, even if the H+ ion concentration

decreases another order of magnitude, as the main cathodic species becomes aqueous H2S, and

corrosion rate remains high. This is accounted for by the model, and the simulations agree rather

well with the experimental values.


14.0

12.0

10.0

Corrosion rate / (mm.yr-1)


8.0

6.0

4.0

2.0

0.0
2.5 3 3.5 4 4.5 5 5.5
pH

Figure 10. The H2S corrosion rate as a function of pH in the absence of protective iron sulfide corrosion product
layers formation. Solid line is generated by the model (MULTICORP™ 5.5), points represent LPR experimental
data taken from Zheng et al.31. Conditions: glass cell rotating cylinder experiments, equivalent velocity v=1.2 m/s
in a 0.0254 ID pipe, T=30oC, pH4, Fe2+< 1 ppm, [NaCl]=1 wt%, pH2S=0.0965 bar, pCO2=0 bar, [HAc]=0 ppm;
error bars represent the standard deviation.

In Figure 11, we can see the effect of protective iron sulfide corrosion product formation on the

corrosion rate over time. At 80oC, the layer forms rapidly and reduces the bare steel corrosion

rate by an order of magnitude to very low values, by presenting a diffusion barrier and blocking

the electroactive sites on the steel surface. At 25oC, the effect is much less pronounced due to the

much slower kinetics of iron sulfide precipitation; the bare steel corrosion rate starts out at a

lower rate, when compared to 80oC, and is reduced only slightly over the course of the exposure

as the iron sulfide corrosion product layer gradually layer builds up. This behavior is mimicked

by the simulations even if the very fast kinetics of the initial iron sulfide formation at 80oC is not

perfectly captured.
2.0

1.5

Corrosion rate / (mm.yr-1) 1.0

0.5

T = 25 oC

T = 80 oC
0.0
0 24 48 72 96 120
Time / hr

Figure 11. The H2S corrosion rate with time at different temperature; Solid lines are generated by the model
(MULTICORP™ 5.5), points represent LPR data from multiple experiments, taken from Zheng et al.85.
Conditions: glass cell rotating cylinder experiments, equivalent velocity v=0.4 m/s in a 0.1 ID pipe, pH6, Fe2+< 1
ppm, [NaCl]=1 wt%, pH2S = 0.1 bar at 25oC (red squares) and pH2S = 0.054 bar at 80oC (blue circles), pCO2=0
bar, [HAc]=0 ppm.

Finally, the effect of HAc presence on CO2 corrosion rate is illustrated in Figure 12. The

corrosion rate in the absence of HAc is already high due to the low pH, presence of CO2, flow

and moderately high temperature, however, addition of relatively small amounts of HAc

increases the corrosion rate significantly. Already at 8.5 ppm of undissociated HAc, the

corrosion rate almost doubles, while at higher concentrations, catastrophically high corrosion

rates are seen. This behavior is well represented by the simulations.


60

50

Corrosion rate / (mm.yr-1)


40

30

20

10

0
0 100 200 300 400 500 600 700 800 900
HAc concentration / ppm

Figure 12. The effect of undissociated HAc concentration on CO2 corrosion rate. Solid line is generated by the
model (MULTICORP™5.5), points represent LPR experimental data taken from George and Nesic 73.
Conditions: glass cell rotating cylinder experiments, equivalent velocity v=0.63 m/s in a 0.1 ID pipe, T=60oC,
pH4, Fe2+< 1 ppm, [NaCl]=3 wt%, pH2S=0 bar, pCO2=0.8 bar; error bars represent the standard deviation.

5. CONCLUSIONS

• A comprehensive mechanistic predictive model for corrosion of mild steel in the oil and

gas transmission pipelines can simultaneously accounts for CO2 corrosion, H2S corrosion,

and corrosion in the presence of organic acid, and account for the effect of corrosion

product layer formation: iron carbonate and iron sulfide.

• The model implementation was done by using a generic mathematical and programming

approach; this allows flexibility to add new chemical species and additional chemical and

electrochemical reactions in the future, it make it easy to extend the model to cover more

extreme conditions, such as higher temperatures and pressures, non-ideal solutions, etc.

• The model can be readily coupled with other applications such as computational fluid

dynamics (CFD) codes, multiphase flow simulators, process design simulators, etc.
• A large experimental data basewas used to successfully validate the model performance

and to highlight its strengths as well as areas where improvement is required.

6. THE CHALLENGES AHEAD

Uniform CO2 corrosion of mild steel can now be considered a mature topic in the context of

corrosion science and engineering. The understanding of the underlying physiochemical

processes enables construction of mechanistic models of varying complexity, which can be

successfully used to aid our understanding of the complex interplay between different parameters

and to predict the corrosion rate. Furthermore, these models serve as a repository of the current

knowledge on the topic, as well as a solid platform for building in new effects as they are

discovered and understood. While we have come a long way in the past few decades, plenty of

challenges lie ahead.

Modeling the effect of high pressure (close to and above the critical point for CO2) and high

temperature (above 100oC) is currently being addressed. Complexities arising from multiphase

flow affecting water wetting in oil transportation lines and water condensation in wet gas lines

are another major modeling challenge. The effect of non-ideal solutions (due to very high

concentrations of dissolved solids), scaling, under-deposit corrosion, erosion-corrosion and

corrosion inhibition are some of the new frontiers for this type of corrosion modeling. A number

of research groups around the world are currently working on many of these issues and as the

understanding matures, it will find its way into the mechanistic CO2 corrosion models of the

future.
Some recent advancements in understanding and modeling of H2S corrosion electrochemistry

and iron sulfide corrosion product layer formation have enabled smooth integration of sweet and

sour corrosion models, as described above. Yet, when it comes to understanding of the roles of

different iron sulfides on their protectiveness, there is a long way to go before we have sufficient

understanding that can lead to successful modeling.

Similarly, when it comes to organic acid corrosion the basic electrochemistry has been resolved

to the extent that it enables us to have accurate corrosion prediction models that can be integrated

with the CO2 and H2S corrosion models, as described in the present paper. However, the effect of

organic acids on integrity of iron carbonate and iron sulfide corrosion product layers still remains

a controversy and needs more research before the understanding can be implemented into the

models.

A special mention should be given to modeling of localized corrosion of steel in these

environments. This is the ultimately challenging topic lying ahead of us, as there is no single

cause or mechanism governing localized attack. However, research on this topic is ongoing and

some progress has been made. The solid foundation built in terms of comprehensive mechanistic

corrosion models, such as the one presented here, will serve as a platform for expanding these

models to address localized corrosion.

7. ACKNOLEDGMENTS

The model presented above has been developed over the past 20 years. Many individuals have

contributed significantly to building of the original model and the continuous improvements.

Over this long time, new and better physico-chemical models were implemented, the numerical

methods were enhanced to improve stability, accuracy and speed of calculations and finally – a
large effort was made to implement the model into the MULTICORP™ package and bring its

power to the fingertips of the corrosion engineers and scientists in the industry. The list is long

and in addition to the authors of this manuscript, it includes: Magnus Nordsveen, Hui Lee, John

K-L. Lee, Shihuai Wang, Jiyong Cai, Yang Yang, Ying Xiao, Hongbin Wang, Bert Pots, Yugui

Zheng, Wei Sun, Marc Singer, Bruce Brown, Dusan Sormaz, Arkopaul Sarkar, Zhengchao Tian

and numerous other graduate students and postdocs who helped with testing and verification of

the model. To them all we are grateful. This long term project was sponsored in part by the

following companies: Anadarko, Baker Hughes, BP, Champion Technologies, Chevron,

Clariant, CNPC, CNOOC, ConocoPhillips, DNV GL, ENI, ExxonMobil, Hess, Inpex, M-I

SWACO (Schlumberger), Multi-Chem (Halliburton), Nalco, Occidental Oil Company,

Petrobras, Petronas, PTT, Saudi Aramco, Shell, SINOPEC (China Petroleum), Tenaris,

TransCanada, Total, and Wood Group Kenny.

8. REFERENCES

1. de Waard, C., and D.E. Milliams, “Prediction of Carbonic Acid Corrosion in Natural Gas
Pipelines,” in Intern. Extern. Prot. Pipes (1975), pp. F1-1-F1-8.
2. Nordsveen, M., S. Nešić, R. Nyborg, and A. Stangeland, Corrosion 59 (2003): pp. 443–
456.
3. Olsen, S., “CO₂ Corrosion Prediction by Use of the Norsok M-506 Model - Guidelines
and Limitations,” in CORROSION (2003), p. Paper No. 623.
4. Olsen, S., A.M. Halvorsen, Per G. Lunde, and R. Nyborg, “CO₂ Corrosion Prediction
Model - Basic Principles,” in CORROSION (2005), p. Paper No. 551.
5. Halvorsen, A.M., and T. Sontvedt, “CO2 Corrosion Model for Carbon Steel Including
Wall Shear Stress Model for Multiphase Flow and Limits for Production Rate to Avoid
Mesa Attack,” in CORROSION (1999), p. Paper No. 42,
http://www.onepetro.org/mslib/app/Preview.do?paperNumber=NACE-
99042&societyCode=NACE.
6. Dugstad, A., L. Lunde, and K. Videm, “Parametric Study of CO₂ Corrosion of Carbon
Steel,” in CORROSION (1994), p. Paper No. 14.
7. de Waard, C., and U. Lotz, “Prediction of CO₂ Corrosion of Carbon Steel,” in
CORROSION (1993), p. Paper No. 069.
8. de Waard, C., U. Lotz, and A. Dugstad, “Influence of Liquid Flow Velocity on CO₂
Corrosion: A Semi-Empirical Model,” in CORROSION (1995), p. Paper No. 128.
9. de Waard, C., U. Lotz, and D.E. Milliams, Corrosion 47 (1991): pp. 976–985.
10. de Waard, C., and D.E. Milliams, Corrosion 31 (1975): pp. 177–181.
11. de Waard, C., L.M. Smith, and B.D. Craig, “Influence of Crude Oils on Well Tubing
Corrosion Rates,” in Corros. 2003 (2003), p. Paper No. 629.
12. Gray, L.G.S., B.G. Anderson, M.J. Danysh, and P.R. Tremaine, “Effect of pH and
Temperature on the Mechanism of Carbon Steel Corrosion by Aqueous Carbon Dioxide,”
in CORROSION (1990), p. Paper No. 40.
13. Gray, L.G.S., B.G. Anderson, M.J. Danysh, and P.R. Tremaine, “Mechanisms of Carbon
Steel Corrosion in Brines Containing Dissolved Carbon Dioxide At pH 4,” in
CORROSION (1989), p. Paper No. 464.
14. Nešić, S., H. Li, J. Huang, and D. Sormaz, “An Open Source Mechanistic Model for CO₂/
H₂S Corrosion of Carbon Steel,” in CORROSION (2009), p. paper No. 09572.
15. Nešić, S., J. Postlethwaite, and S. Olsen, Corrosion 52 (1996): pp. 280–294.
16. Nešić, S., M. Nordsveen, R. Nyborg, and A. Stangeland, “A Mechanistic Model for CO₂
Corrosion with Protective Iron Carbonate Films,” in CORROSION (2001), p. Paper No.
040.
17. Nešić, S., J. Lee, and V. Ruzic, “A Mechanistic Model of Iron Carbonate Film Growth
and the Effect on CO₂ Corrosion of Mild Steel,” in CORROSION (2002), p. Paper No.
237.
18. Nešic, S., and K. Lee, Corrosion 5 (2003): pp. 616–628.
19. Nyborg, R., “Overview of CO₂ Corrosion Models for Models for Wells and Pipelines,” in
CORROSION (2002), p. Paper No. 233.
20. Nyborg, R., P. Andersson, and M. Nordsveen, “Implementation of CO₂ Corrosion Models
in a Three-Phase Fluid Flow Model,” in CORROSION (2000), p. Paper No. 048.
21. Kapusta, S.D., B.F.M. Pots, and I.J. Rippon, “The Application of Corrosion Prediction
Models to the Design and Operation of Pipelines,” in CORROSION (2004), p. Paper No.
30.
22. Nešić, S., J. Postlethwaite, and M. Vrhovac, Corros. Rev. 15 (1997): pp. 211–240.
23. Nešić, S., Corros. Sci. 49 (2007): pp. 4308–4338.
24. Nyborg, R., “Field Data Collection, Evaluation and Use for Corrosivity Prediction and
Validation of Models,” in CORROSION (2006), p. paper no. 118.
25. Anderko, A., “Simulation of FeCO₃/FeS Scale Formation Using Thermodynamic and
Electrochemical Models,” in CORROSION (2000), p. No. 102.
26. Anderko, A., “Simulation of FeCO3/FeS Scale Formation Using Thermodynamic and
Electrochemical Models,” in Corros. 2000 (2000), p. Paper No. 102,
http://www.onepetro.org/mslib/servlet/onepetropreview?id=NACE-00629.
27. Dayalan, E., F.D. de Moraes, J.R. Shadley, S.A. Shirazi, and E.F. Rybicki, “CO₂
Corrosion Prediction in Pipe Flow under FeCO₃ Scale-Forming Conditions,” in
CORROSION (1998), p. Paper No. 51.
28. Zhang, R., M. Gopal, and W.P. Jepson, “Development of a Mechanistic Model for
Predicting Corrosion Rate in Multiphase Oil/water/gas Flows,” in CORROSION (1997),
p. Paper No. 601.
29. George, K., S. Nešić, and C. de Waard, “Electrochemical Investigation and Modeling of
Carbon Dioxide Corrosion of Carbon Steel in the Presence of Acetic Acid,” in
CORROSION (2004), p. Paper No. 379.
30. Han, J., J. Zhang, and J.W. Carey, Int. J. Greenh. Gas Control 5 (2011): pp. 1680–1683.
31. Zheng, Y., B. Brown, and S. Nešic, Corrosion 70 (2014): pp. 351–365.
32. Esmaeely, S.N., B. Brown, and S. Nešić, Corrosion 73 (2017): pp. 144–154.
33. Zheng, Y., J. Ning, B. Brown, and S. Nesic, Corros. 2014 71 (2014): pp. 316–325.
34. Turgoose, S., R.A. Cottis, and K. Lawson, “Modeling of Electrode Processes and Surface
Chemistry in Carbon Dioxide Containing Solutions,” in Comput. Model. Corros. ASTM
STP 1154 (1992), pp. 67–81.
35. Pots, B.F.M., “Mechanistic Models for the Prediction of CO₂ Corrosion Rates under
Multi-Phase Flow Conditions,” in CORROSION (1995), p. Paper No. 137.
36. Nešić, S., M. Nordsveen, R. Nyborg, and A. Stangeland, Corrosion 59 (2003): pp. 489–
497.
37. Tribollet, B., J. Kittel, A. Meroufel, F. Ropital, F. Grosjean, and E.M.M. Sutter,
Electrochim. Acta 124 (2014): pp. 46–51,
http://dx.doi.org/10.1016/j.electacta.2013.08.133.
38. Kittel, J., F. Ropital, F. Grosjean, E.M.M. Sutter, and B. Tribollet, Corros. Sci. 66 (2013):
pp. 324–329, http://dx.doi.org/10.1016/j.corsci.2012.09.036.
39. Remita, E., B. Tribollet, E. Sutter, V. Vivier, F. Ropital, and J. Kittel, Corros. Sci. 50
(2008): pp. 1433–1440.
40. Remita, E., B. Tribollet, E. Sutter, F. Ropital, X. Longaygue, J. Kittel, C. Taravel-Condat,
and N. Desamais, J. Electrochem. Soc. 155 (2008): p. C41,
http://jes.ecsdl.org/cgi/doi/10.1149/1.2801349.
41. Amri, J., E. Gulbrandsen, and R.P. Nogueira, Corros. Sci. 52 (2010): pp. 1728–1737,
http://dx.doi.org/10.1016/j.corsci.2010.01.010.
42. Kahyarian, A., A. Schumaker, B. Brown, and S. Nesic, Electrochim. Acta 258 (2017): pp.
639–652.
43. Zhang, Z., D. Hinkson, M. Singer, H. Wang, and S. Nešić, Corrosion 63 (2007): pp.
1051–1062.
44. Li, D., and Z. Duan, Chem. Geol. 244 (2007): pp. 730–751.
45. Duan, Z., R. Sun, C. Zhu, and I.-M. Chou, Mar. Chem. 98 (2006): pp. 131–139.
46. Duan, Z., and D. Li, Geochim. Cosmochim. Acta 72 (2008): pp. 5128–5145.
47. Marshall, W.L., and E.U. Franck, J. Phys. Chem. Ref. Data 10 (1983): pp. 295–304.
48. Cooper, J.R., “Revised Release on the IAPWS Industrial Formulation 1997 for the
Thermodynamic Properties of Water and Steam” (Lucerne, Switzerland: The International
Association for the Properties of Water and Steam, 2012), http://www.iapws.org.
49. Suleimenov, O.M., and R.E. Krupp, Geochim. Cosmochim. Acta 58 (1994): pp. 2433–
2444.
50. Suleimenov, O.M., and T.M. Seward, Geochim. Cosmochim. Acta 61 (1997): pp. 5187–
5198.
51. Kharaka, Y.K., W.D. Gunter, P.K. Aggarwal, E.H. Perkins, and DeBraal J. D.,
“Solmineq88 a Computer Program for Geochemical” (1988).
52. Meyssami, B., M.O. Balaban, and A.A. Teixeira, Biotechnol. Prog. 8 (1992): pp. 149–
154.
53. Kahyarian, A., M. Singer, and S. Nesic, J. Nat. Gas Sci. Eng. 29 (2016): pp. 530–549,
http://dx.doi.org/10.1016/j.jngse.2015.12.052.
54. Kahyarian, A., M. Achour, and S. Nesic, “Mathematical Modeling of Uniform CO₂
Corrosion,” in Trends Oil Gas Corros. Res. Technol., ed. A.M. El-Sherik (Elsevier, 2017),
pp. 805–849.
55. Kahyarian, A., M. Achour, and S. Nesic, “CO₂ Corrosion of Mild Steel,” in Trends Oil
Gas Corros. Res. Technol., ed. A. M. El-Sherik (Elsevier, 2017), pp. 149–190.
56. Kahyarian, A., B. Brown, S. Nesic, and S. Ne, Corrosion 74 (2018): pp. 851–859.
57. Kahyarian, A., and S. Nesic, Electrochim. Acta (2018),
https://linkinghub.elsevier.com/retrieve/pii/S0013468618327270.
58. Bockris, J.O., and D. Drazic, Electrochim. Acta 7 (1962): pp. 293–313.
59. Hibert, F., Y. Miyoshi, G. Eichkorn, and W.J. Lorenz, J. Electrochem. Soc. 118 (1971):
pp. 1919–1926.
60. Atkinson, A., and A. Marshall, Corros. Sci. 18 (1978): pp. 427–439.
61. El Miligy, A.A., D. Geana, and W.J. Lorenz, Electrochim. Acta 20 (1975): pp. 273–281.
62. Nešić, S., N. Thevenot, J.L. Crolet, and D. Drazic, “Electrochemical Properties of Iron
Dissolution in the Presence of CO₂ - Basics Revisited,” in CORROSION (1996), p. Paper
No. 03.
63. Keddam, M., O.R. Mattos, and H. Takenout, J. Electrochem. Soc. 128 (1981): pp. 257–
266.
64. Keddam, M., O.R. Mattos, and H. Takenout, J. Electrochem. Soc. 128 (1981): pp. 266–
274.
65. Felloni, L., Corros. Sci. 8 (1968): pp. 133–148.
66. Bockris, J.O., D. Drazic, and A. R. Despic, Electrochim. Acta 4 (1961): pp. 325–361.
67. Drazic, D., Mod. Asp. Electrochem. 19 (1989): pp. 62–192.
68. Dražić, D.M., and C.S. Hao, Electrochim. Acta 27 (1982): pp. 1409–1415.
69. Lorenz, W.J., G. Staikov, W. Schindler, and W. Wiesbeck, J. Electrochem. Soc. 149
(2002): pp. K47–K59.
70. Keddam, M., “Anodic Dissolution,” in Corros. Mech. Theory Pract. Third Ed. (CRC
Press, 2011), pp. 149–215, http://dx.doi.org/10.1201/b11020-4.
71. Heusler, K.E., Encyclopedia of Electrochemistry of the Elements. Vol. 9 (New York:
Marcel Dekker, 1982).
72. Ogundele, G.I., and W.E. White, Corrosion 42 (1986): pp. 71–78.
73. George, K.S., and S. Nešic, Corrosion (2007): pp. 178–186.
74. van Hunnik, E.W.J., B.F.M. Pots, and E.L.J.A. Hendriksen, “The Formation of Protective
FeCO₃ Corrosion Product Layers in CO₂ Corrosion,” in CORROSION (1996), p. Paper
No. 006.
75. Gulbrandsen, E., “Acetic Acid and Carbon Dioxide Corrosion of Carbon Steel Covered
with Iron Carbonate,” in CORROSION (2007), p. Paper No. 322.
76. Crolet, J.-L., N. Thevenot, and A. Dugstad, “Role Of Free Acetic Acid On The CO₂
Corrosion Of Steels,” in CORROSION (1999), p. Paper No. 24.
77. Sun, W., S. Nešić, and R.C. Woollam, Corros. Sci. 51 (2009): pp. 1273–1276.
78. Ruzic, V., M. Veidt, and S. Nešić, Corrosion 63 (2007): pp. 758–769.
79. Kermani, M.B., and A. Morshed, Corrosion 59 (2003): pp. 659–683.
80. Davies, D.H., and T. Burstein, Corrosion 36 (1980): pp. 416–422.
81. Dugstad, A., “Mechanism of Protective Film Formation during CO₂ Corrosion of Carbon
Steel,” in CORROSION (1998), p. Paper No. 31.
82. Johnson, M.L., and M.B. Tomson, “Ferrous Carbonate Precipitation Kinetics and Its
Impact on CO₂ Corrosion,” in CORROSION (1991), p. Paper No. 268.
83. Lasaga, A.C., Kinetic Theory in the Earth Sciences (Princeton University Press, 1998).
84. Sun, W., and S. Nešić, Corrosion 64 (2008): pp. 334–346.
85. Zheng, Y., J. Ning, B. Brown, and S. Nesic, Corrosion 72 (2015): pp. 679–691.
86. Benning, L.G., R.T. Wilkin, and H.L. Barnes, Chem. Geol. 167 (2000): pp. 25–51.
87. Sun, W., and S. Nesic, “Basics Revisited: Kinetics of Iron Carbonate Scale Precipitation
in CO₂ Corrosion,” in CORROSION (2006), p. Paper No. 365.
88. Pots, B.F.M., and E.L.J.A. Hendriksen, “CO₂ Corrosion under Scaling Conditions - the
Special Case of Top-of-Line Corrosion in Wet Gas Pipelines,” in CORROSION (2000), p.
Paper No. 031.
89. Nešić, S., and K.L.J. Lee, Corrosion 59 (2003): pp. 616–628.
90. Ruzic, V., M. Veidt, and S. Nešić, Corrosion 62 (2006): pp. 598–611.
91. Ruzic, V., M. Veidt, and S. Nešić, Corrosion 62 (2006): pp. 419–432.
92. Newman, J., and K.E. Thomas-Alyea, Electrochemical Systems, 3rd ed. (Wiley-
interscience, 2004).
93. Aravinth, S., Int. J. Heat Mass Transf. 43 (2000): pp. 1399–1408.
94. Davies, J.T., “Chapter 3: Eddy Transfer Near Solid Surfaces,” in Turbul. Phenom.
(Elserviere Inc., 1972), pp. 121–174.
95. Korson, L., W. Drost-Hansen, and F.J. Millero, J. Phys. Chem. 73 (1969): pp. 34–39.
96. Palmer, D.A., and R. Van Eldik, Chem. Rev. 83 (1983): pp. 651–731.
97. Green, N., ed., Comprehensive Chemical Kinetics, Volume 6 (Amsterdam, The
Netherlands: Elsevier Inc, 1972).
98. Delahay, P., J. Am. Chem. Soc. 74 (1952): pp. 3497–3500,
http://pubs.acs.org/doi/pdf/10.1021/ja01134a013%5Cnhttp://pubs.acs.org/doi/abs/10.1021
/ja01134a013.
99. Cussler, E., Diffusion: Mass Transfer in Fluid Systems, Engineering, Third Edit
(Cambridge University Press, 2009),
http://books.google.com/books?hl=en&amp;lr=&amp;id=TGRmfTrsPTQC&amp;oi=fnd
&amp;pg=PR17&amp;dq=Diffusion:+Mass+transfer+in+fluid+systems&amp;ots=7ERQ
wdA3wn&amp;sig=IVQzPbrfIg5nEsGU05pCp8BuI4w.
100. Haynes, W.M., ed., CRC Handbook of Chemistry and Physics, 84th ed. (CRC Press LLC,
2004).
101. Tamimi, A., E.B. Rinker, and O.C. Sandall, J. Chem. Eng. Data 39 (1994): pp. 330–332.
102. Kahyarian, A., B. Brown, and S. Nesic, J. Electrochem. Soc. 164 (2017): pp. H365–H374.
103. Nesic, S., G.T. Solvi, and J. Enerhaug, Corrosion 51 (1995): pp. 773–787.
104. Wang, S., K. George, and S. Nešić, “High Pressure CO₂ Corrosion Electrochemistry and
the Effect of Acid Acetic,” in CORROSION (2004), p. Paper No. 375.
105. Yang, Y., “Removal Mechanisms of Protective Iron Carbonate Layer in Flowing
Solutions,” Ohio University, PhD dissertation, 2012.

You might also like