Corrosion Science 163 (2020) 108296

Contents lists available at ScienceDirect

Corrosion Science
journal homepage: www.elsevier.com/locate/corsci

Moving boundary simulation and mechanistic studies of the electrochemical T

corrosion protection by a damaged zinc coating
Mohsen Saeedikhania, Sudesh Wijesinghea,b, Daniel John Blackwooda,*
a
Department of Materials Science and Engineering, National University of Singapore, 9 Engineering Drive 1, 117576, Singapore
b
Measurements and Characterization Unit, Singapore Institute of Manufacturing Technology, A⁎STAR, 2 Fusionopolis Way, 138634, Singapore

ARTICLE INFO ABSTRACT

Keywords: The mechanism of corrosion protection of steel substrate by a 96 wt% zinc primer that was damaged by scratch
Corrosion modeling was simulated. The primer was approximated to pure zinc layer due to high percentage of zinc. Results show that
Sacrificial protection while the middle of scratch is protected by sacrificial protection, at the corners of scratch a local cathodic
Barrier protection inhibition results in barrier protection due to variations in the dissolved oxygen concentration. Moving boundary
Zinc
simulation allowed to predict the changes to the geometry of the corroding electrodes. Finally, taking into
Moving boundary simulation
account zinc electrodes consumption, the period of sacrificial protection was predicted by simulations.

1. Introduction assembled onsite, such as above ground oil tanks and pipelines in re-
fineries. Note that the two prime components of ZRPs are the zinc dust
Corrosion as a spontaneous phenomenon has always been a major and a binder, which may be organic or inorganic.
expense to many industries [1]. Corrosion threatens industries’ safety The extensive literature available on zinc-based coatings states that
and is a prime factor in the failure of equipment. However, not all forms these provide two primary mechanisms of protection, namely sacrificial
of corrosion are harmful and in fact, industries can sometimes benefit protection and barrier protection [4–6,13]. Despite the large volume of
from corrosion phenomenon with the help of corrosion engineers. For work that has been dedicated to the galvanic corrosion of zinc-based
instance, the mechanism of galvanic corrosion, which may occur once coatings, quantification of the damage between the zinc coating and the
two dissimilar metals are electrically and electrolytically in contact steel substrate may not be easily obtained. There exist standard tech-
with each other [2,3], can be employed in order to protect certain niques that can assess the galvanic damage of a couple which are gal-
metallic structures by means of sacrificial protective primers and vanically in direct physical contact with each other. For instance, local
coatings [4, 5]. In such, a more active metal is chosen as the coating’s scanning probe techniques such as scanning vibrating electrode tech-
material in order to protect the underlying more noble substrate. Sa- nique (SVET), scanning kelvin probe (SKP), scanning electrochemical
crificial protective coatings are able to efficiently protect the metallic microscopy (SECM), and zero resistance ammeter (ZRA) are very useful
substrate even after mechanical damage to the coating [5]. to assess the galvanic damage. However, a drawback of such techniques
Of all the sacrificial coatings, zinc based coatings have always been could be the cost of the equipment and sometimes difficulties in sample
of most interest for the protection of steel substrates due to zinc’s preparation for certain specimens. In addition, although mixed poten-
abundance, high electrochemical activity and its environmentally be- tial theory [14,15] is used to estimate the galvanic corrosion rate, the
nign nature [6–12]. Hence, different varieties of zinc coatings, from hot individual components of the galvanic couple during these experiments
dip galvanized (HDG) coatings to zinc-rich primers, have been devel- are not in physical and electrical contact with each other, in contrast
oped to tackle the corrosion of steel structures with the help of sacri- with an actual galvanic couple like zinc-based coatings.
ficial protection. Hot dip galvanizing produces a layer of zinc metal on Therefore, the development of numerical models to simulate sys-
the steel base by immersing the steel in a kettle of molten zinc. Al- tems under galvanic corrosion is attractive. Modeling also has the ad-
though HDG coatings provide superior corrosion resistant for the steel, vantage of being faster and less expensive than experiments. Hence,
it is limited to structures that could be fitted and immersed in the kettle. several works have been dedicated to corrosion modeling [16–29].
Therefore, zinc rich primers (ZRPs) with one or two layer(s) on top are Some of the successful models have been developed using COMSOL
preferred for larger structures, especially those that need to be Multiphysics. For example, Amorim et al. [16] and Thebault et al.

⁎
Corresponding author.
E-mail address: [email protected] (D.J. Blackwood).

https://doi.org/10.1016/j.corsci.2019.108296
Received 4 July 2019; Received in revised form 22 September 2019; Accepted 17 October 2019
Available online 23 October 2019
0010-938X/ © 2019 Elsevier Ltd. All rights reserved.
M. Saeedikhani, et al. Corrosion Science 163 (2020) 108296

[17–21] developed 2D finite element method (FEM) models that si- a microscope. The backsides and edges of the samples were insulated by
mulate the cut edge corrosion of galvanized steels and Cross et al. [22] a sealing tape so that only the measured area was exposed.
developed a time-dependent FEM model to simulate the corrosion of The scratched samples were immersed in open to air beakers con-
zinc coatings galvanically coupled to a mild steel substrate in deaerated taining 3.5 wt% NaCl solution. A ca. 10 μm tip diameter tungsten pH
0.01 M H2SO4 electrolyte. However, Cross’ model diverged from the microelectrode (WPI™) was used to measure the local pH over the
experimental results after approximately 40 min, due to a decrease in scribed region of the 0 h, 2 h, 48 h, and 96 h immersed samples. The
the reactivity of the zinc surface. As such, there still appears to be no microelectrode pH measurements were carried out at ca. 150 μm above
successful model for long-term galvanic corrosion of zinc-based coat- the carbon steel surface in the scratched region, determined visually by
ings. a lateral camera. Further, to determine the pH distribution, the tungsten
Not only are ZRPs being used in conjunction with top layers, but pH microelectrode scanned the surface horizontally, with its movement
also galvanized coatings are nowadays being painted to increase their given by a micromanipulator.
lifetime/durability. However, simulation studies for galvanized coat- Samples were withdrawn after specific immersion period for up to
ings also do not consider the topcoat. This is due to the fact that the 96 h, rinsed with distilled water, and dried followed by an immediate
models wish to see how zinc could sacrificially protect the substrate, microRaman spectroscopy (Renishaw inVia Raman Microscope, 785 nm
regardless of the topcoat, in case of a damage such as a deep-through beamline laser, spot size ca. 50 μm × 300 μm). The Raman spectra were
scratch in our study or a cut-edge. Furthermore, there are actual cases obtained from only the top-view of the scratched region with the help of
that single layer ZRPs are used such as in environments with very low the optical microscope coupled with the Raman spectroscope. Finally,
corrosivity such as arid rural atmospheres. Paint fabricators normally the Raman spectra were smoothened and normalized using Origin
advice on the choice of the paint layer numbers (be it multiple layers of software.
the same primer or different grades of paint in each layer) based on the Additionally, optical microscopy (Olympus MX51) was conducted to
corrosion rate of that specific environment. study the zinc corrosion products through the cross section of the
Whilst most of the aforementioned simulations are focused on the scratchs. The samples were cut in desired dimensions by an ultra-slow
case of cut-edge corrosion with a constant size of zinc electrode during cutting speed precision cutting machine (Secotom 10). To minimize the
corrosion, this study simulates the galvanic corrosion within a scratch detachment of corroison products due to cut, a relatively thick layer of
geometry, which is a damage type that often happens to coated struc- cyanoacrylate adhesive was applied on the zinc corrosion products (on
tures. In the present work, the sacrificial protection provided by a the scratched region and up to 2 cm from both sides) to support the
scratched zinc-rich primer to a steel substrate has been simulated using interested regions and was let to fully cure before cutting. Note that
a moving boundary – time dependent model using COMSOL conventional mounting of the samples was impossible as the samples
Multiphysics, which allowed to visualize zinc consumption (corrosion) were relatively big in size (10 cm × 15 cm). After cutting, the samples
and to predict the sacrificial protection period. The amount of cathodic were grinded with SiC papers up to the 4000 grade and then polished.
and anodic current densities have been quantified over different cor- Finally, the applied adhesive was checked after polishing for each
rosion exposure periods. The moving boundary study in the model is sample and was yet well adhered to the sample.
able to track the deforming (corroding) interface(s) by including the The electrochemical measurements were performed in 3.5 wt%
concept of the Arbitrary Lagrangian–Eulerian (ALE) method, which is NaCl aqueous solution using a three-electrode configuration, the sample
able to capture greater deformation and resolution [23]. The ALE area, including the scratch at the center, was restricted to 14.6 cm2 by
method has previously been successfully used by Deshpande [24] to using a Gamry paint cell with a graphite rod counter electrode and
track the corroding interface of Mg-Fe and Al-Fe galvanic couples for saturated calomel electrode reference electrode (SCE). Potentiodynamic
72 h and by Sun et al. [25] to study crevice corrosion. Finally, the polarization measurements (PPM) for carbon steel was performed over
corrosion rates produced by the new model in this study are compared the potential range of carbon steel OCP (-0.68 V vs. SCE) to zinc OCP
with experimental data obtained from mixed potential theory. (-1.00 V vs. SCE) and the reverse range was chosen for zinc. Moreover,
The reason for choosing ZRP instead of a galvanized system is that PPM was conducted under the scan rate of 5 mV/S with the aid of a
our project’s aim is to simulate high thicknesses of zinc coats which are Gamry Interface 1000 potentiostat.
being used in reality for zinc primers. The thickness of galvanized
coating is in the range of ca. 5–20 microns whilst zinc primers have the 3. Modeling
thickness range of 50–150 microns. The 120 microns is the thickness
which is actually recommended by the paint/primer fabricators and 3.1. Geometry
being used in the industry. Last but not least, for (hot) galvanized steels,
there exists a metallic transition layer between the substrate and the Fig. 1(a) shows a schematic cross section of an iron substrate (as a
zinc coating which often is not considered in the modeling. Therefore, it surrogate for the carbon steel used in the physical experiments) with a
is likely that our single zinc layer model more resembles the protection 120 μm thick zinc primer (96 wt% Zn) layer covered by an electrolyte
behavior of a zinc coating than a hot-dip galvanized coating. (3.5 wt% NaCl) layer. The electrolyte’s thickness was set to 3000 μm as
this was successfully used by the previous modeling studies [17,19]. In
2. Experimental the case of a deep mechanical scratch through the primer, the elec-
trolyte will reach the iron surface and galvanic corrosion / protection
ASTM A36 carbon steel with the dimension of 10 cm × 15 cm was will come into account. To simplify the model, the primer/electrolyte
used. The surface of the carbon steel was ground by SiC papers up to interface is approximated to a zinc layer only (that is the 4 wt% of
grade 4000 and then polished. The zinc rich primer (ZRP) was coated as binder in the primer is neglected); this is reasonable as the primer dry
a single layer of 120 μm on steel surface by airless spray method. ZRP film contains 96 wt% zinc dust. Remark intercalation of zinc particles
was a commercial primer containing 96 wt% zinc dust on dry film. Due and loss of conductivity are indeed important parameters in zinc rich
to the very high percentage of the zinc dust, the primer approximately primers, but for lower percentages such as 40–70 wt% of zinc particles
models as a one-component zinc coat. on dry film. Intercalation was not the case for our studied ZRP which
An Elcometer scribing tool was used to introduce a deep-through- contains 96 wt% zinc particles on dry film which is remained with only
steel scratch to the primer. The scratch was a single line 500 μm wide 4 wt% binder, etc. Hence, the zinc particles in the primer are not split/
located in the middle of the sample and along the length of the sample. dispersed in the structure; rather they are compactly adhered together
The zinc-rich coating is far softer than steel, so the scribe is unlikely to due to the very high percentage of the zinc. Also, as there is only 4 wt%
damage the substrate, this was confirmed by visual observations under of binder in the primer, the ionic resistance of the binder could be

2
M. Saeedikhani, et al. Corrosion Science 163 (2020) 108296

Fig. 1. (a) The schematic cross section of a scratched paint on iron covered by an electrolyte layer, not drawn to scale for a better visualization of the scratch (b) The
simplified schematic geometry of the model.

neglected. Therefore, approximating it to a pure zinc layer was not far concentration, charge, mobility, Faraday constant, electric potential,
from reality. Note that it is always necessary to make simplifications to and solution-velocity, respectively. Since the electrolyte will be stag-
make the model work and to obtain practical models. Complex models nant in the model, the solution-velocity can be set to zero.
might not be attractive to the readers or even to industry. Therefore, the The mass balance of the species could be given by:
real surface of zinc would be equal to the boundaries in contact with
ci
electrolyte in the geometry as the primer was approximated to a zinc = . Ji + Ri
t (2)
layer due to the very high percentage of zinc dust.
Finally, the model’s boundaries are the surrounds of the electrolyte, with . Ji being the flux divergence and Ri being the production/
marked by black lines in Fig. 1(a) and shown in a simplified form in consumption rate of the species due to the homogeneous reactions in
Fig. 1(b). Note that the rightmost and leftmost boundaries in Fig. 1(b) the electrolyte.
are located sufficiently far away (1500 μm) from the edges of the iron Moreover, the bulk electrolyte is assumed electroneutral, that is:
cathode so that the cathodic electrochemical reactions (mainly oxygen n
reduction) occurring on the iron cathode are not restricted by the z i ci = 0
constrained geometry. i=1 (3)

Furthermore, the initial electrolyte solution contains 3.5 wt% NaCl

3.2. Theory, governing equations, and boundary conditions with pH = 6.5. At the electrolyte–air interface (boundary (ii) in Fig. 2a)
as well as the bulk electrolyte, oxygen initial concentration and carbon
Simulations were performed using the Corrosion Module of dioxide initial concentration was assumed to be distributed uniformly
COMSOL Multiphysics Software, which uses the finite element method and set to the constant values provided in the literature [19,26] and are
to solve the partial differential equations (PDE) problems. Fig. 2 shows reported in Table 1. The species that evolve the electrolyte’s chemistry
the geometry and two meshing sizes of the model, with corresponding with their respective charge number, diffusivity at infinite dilution, and
equations of each boundaries and domains that will be explained ac- initial equilibrium concentration are reported in Table 1. Eq. 3 was
cordingly. The Nernst-Planck equation represents the mass transport of applied to calculate for the sodium ion concentration.
the charge carriers in infinitely dilute aqueous electrolytes by asso- In this study, a semi-empirical model that was successfully used by
ciating the species’ fluxes with diffusion, migration and convection: Simillion et al. [27] is employed to find the species’ diffusion constants
Ji = Di c i zi Fui c i + civ (1) dependence on chloride ion concentration. That is:

where, Ji, Di, ci, zi, ui, F, φ, and v are the flux, diffusion coefficient, Di = Di0 . exp( ai . [Cl ]) (4)

Fig. 2. (a) Geometry of the model. For boundaries (i)-(iii): n = 0 . Laplace Equation 2 = 0 is only applied to the bulk electrolyte (b) Meshing of the model.
Smaller mesh size is related to oxygen diffusion and larger mesh size is related to the bulk electrolyte. Corresponding boundaries (iv) and (vii) in (a), move down

3
M. Saeedikhani, et al. Corrosion Science 163 (2020) 108296

Table 1 Table 3
The species in the electrolyte with their charge number, diffusivity at infinite Values of the parameters used for the finite element simulation.
dilution, fitting parameter of Eq. (4), and initial equilibrium concentration.
Parameter Description Value
Species Zi Di0×109 (m2.S1) ai Initial Concentration, ci (mol. m-3)
βFe Fe oxidation Tafel slope 0.150 V/dec.
Na+ +1 1.33 [20] 3.3 [27] 60 βZn Zn oxidation Tafel slope 0.021 V/dec.
Cl- −1 2.03 [20] 3.0 [27] 60 βO2 O2 reduction Tafel slope 50 mV/dec. [31]
Zn2+ +2 0.71 [20] 11.4 [27] 0 0 Open circuit potential of iron −681 mV vs. SCE
Fe
O2 0 2.40 [20] 1.0 [27] 0.26 [29] 0 Open circuit potential of zinc −1,004 mV vs. SCE
CO2 0 1.91 [30] 1.0 [27] 1.31 × 10−2 [30] Zn
0 Standard reduction potential of 570 mV vs. SCE
OH- −1 5.30 2.38 [27] 3 × 10−8 O2
oxygen
H+ +1 9.31 1.0 [27] 3 × 10−7
'
kFe Apparent rate constant for iron 1.45 × 10−6 mol. m-2.s-1
oxidation [19]
'
kZn Apparent rate constant for zinc 2.32 × 10−5 mol. m-2.s-1
Table 2 oxidation [19]
Homogenous reactions in the electrolyte with their equilibrium constants k 02 Interfacial rate constant for oxygen 10−5 m.s-1 [19]
(k) at 25 °C in the model. reduction
F Faraday constant 96,485 A.s. mol−1
Reaction log (K) at 25 °C

H2O = H++ OH− 14

Zn2+ + OH− = ZnOH+ 5.04 [17] concentration and other species are not uniform. The oxygen con-
ZnOH+ + OH− = Zn(OH)2(aq) 6.06 [17] centration, cO2, in the diffusion layer in Fig. 2(b) is calculated by the
Zn(OH)2(aq) + OH− = Zn(OH)3− 2.5 [17] simulation based on its rate of consumption (given by the magnitude of
CO2 + OH− = HCO3− 7.65 [17] the current density of ORR) and its rate of arrival, using the diffusion
HCO3− + OH− = CO3 2− + H2O 3.67 [17]
Zn2+ + CO32− = ZnCO3(aq) 5.3 [17]
and migration terms of Eq. (1). The thickness of the diffusion layer was
set to 500 μm as the same thickness was successfully used by the pre-
vious modeling studies [17]. The values for all the parameters indicated
where Di0 is the diffusion coefficient at infinite dilution and ai is a fit- in Eq.s (8)–(10) are reported in Table 3.
ting parameter reported in Table 1. The advantage of using Eq. (4) is Eq. (1) is solved on the electrolyte domain with the boundary
considering the effect of solution concentration on ion mobility. conditions indicated in Fig. 2. The current density across the anode’s
Therefore, ions would be less mobile in solutions containing multiple surface (boundaries (iv)-(vii)) is equal to the summation of the zinc
species. oxidation current density and oxygen reduction on zinc coating. Ac-
Additionally, seven homogenous chemical reactions were con- cordingly, the current density across the cathode’s surface (Boundary
sidered to occur in between the chemical species that present in the (viii)) is equal to summation of the iron oxidation current density and
electrolyte and reported in Table 2. All the reaction products were oxygen reduction on the exposed iron (scratch region). Moreover,
considered soluble. Note that the electrolyte’s conductivity Ohm’s law is applied to calculate potential gradient at the anode and
z2i F2Di ci cathode boundaries:
= RT
does exist as the migration term in Eq. (1), which Eq. (4)
( )
indirectly corrects the conductivity [27]. jloc
=
For the current study, three main electrochemical reactions on the
n
(11)
electrodes surfaces are taken into account; namely, iron oxidation, zinc where σ is the electrolyte conductivity and is the local current
( )
jloc
oxidation and oxygen reduction, with the latter reaction taken place density as a function of voltage across the anode’s and cathode’s sur-
both on the anode and cathode surfaces. Oxygen reduction reaction faces. jloc
( )
is replaced with Eq. (12) for the cathodic reaction and Eq.
(ORR) was considered to be under diffusion control. Note, similar to the (13) for the anodic reaction:
previous modeling studies [17,19–21,26,27], hydrogen evolution re-
( )
action (HER) on the electrodes was not considered as a contributing jloc , cathode = jloc, Fe + jloc, O2 (12)
reaction at this stage. Nevertheless, the electrochemical reactions are:
( )
jloc , anode = jloc, Zn + jloc , O2 (13)
Fe Fe 2 + + 2e (5)
and, for the insulated boundaries of (i) to (iii):
Zn Zn2 + + 2e (6)
n =0 (14)
O2 + 2H2 O + 4e 4(OH ) (7)
Moreover, Eq.s (8) to (10) express the local current densities at- 3.3. ALE method to track the moving (corroding) boundary
tributed to these electrochemical reactions:
0 ALE method is meant for moving mesh studies that uses the char-
'
jloc, Fe = 2FkFe exp Fe
acteristics of both Lagrangian and Eulerian methods. As mentioned in
Fe (8) the introduction, the ALE method is an advantageous technique that
0
simulates higher deformations (that can be corroded areas) [24, 25].
'
jloc, Zn = 2FkZn exp Zn According to Donea et al. [23], “in Lagrangian algorithms, each in-
Zn (9) dividual node of the computational mesh follows the associated mate-
rial particle during motion however its weakness is inability to follow
0
O2 large distortions of the computational domain without recourse to
jloc, O2 = 4Fk 02 cO2 exp
frequent re-meshing operations. Furthermore, the computational mesh
O2 (10)
is fixed, and the continuum moves with respect to the grid. In the Eu-
where jloc is the local current density, k’ is the apparent rate constant, k lerian description, large distortions in the continuum motion can be
is the interfacial rate constant, φ0 is the open circuit potential and β is handled with relative ease, but generally at the expense of precise in-
the Tafel slope. The subdomain below the bulk electrolyte, with smaller terface definition and the resolution of flow details. Because of the
mesh size, corresponds to the diffusion layer where the oxygen shortcomings of purely Lagrangian and purely Eulerian descriptions,

4
M. Saeedikhani, et al. Corrosion Science 163 (2020) 108296

ALE technique has been developed that succeeds, to a certain extent, in manner that keeps them equal, but of opposite sign such that net
combining the best features of both the Lagrangian and the Eulerian neutrality is maintained. However, as this moving boundary considers
approaches.” However, a drawback of ALE method is that it handles zinc consumption that leads to a reduction in the surface area ratio
relatively coarser meshes due to mesh entanglement. This drawback between the zinc and iron electrodes, so the anodic current density at
can be resolved by incorporating the adapting re-meshing feature of the the zinc surface increases with time, while the cathodic current density
COMSOL software in which the studied domain is re-drawn and re- at the iron electrode becomes larger over time.
meshed to restart the computation for time dependent studies in order Additionally, the average simulated cathodic current density on the
to obtain smooth meshes. Although cathodic to anodic surface area scratch surface was obtained as -22.5 μA.cm-2 after 96 h, which is the
ratio is very small, the combination of zinc self-corrosion and sacrificial point at which barrier protection by the primer outweighs sacrificial
corrosion lead to a considerable amount of deformation (corrosion) in protection. The -22.5 μA.cm-2 could be defined as the minimum
the anodic electrodes which draws the importance of simulating the cathodic current density criterion needed for sacrificial protection in
metal (zinc) loss. Therefore, ALE technique was employed in the current this study. Note the main criterion for sacrificial protection is potential
study to simulate the corrosion of the electrodes surface. and ability of anode to polarize the steel to cathodic potentials, how-
ALE method was employed to the meshed geometry shown in ever, the anode must be able to supply sufficient current in order to
Fig. 2b using COMSOL MultiPhysics. Moreover, in the ALE method achieve this. As per NACE RP-01-69 Standard, adequate sacrificial
mesh refinement is performed by moving nodes towards nodes of strong protection is achieved with a potential that is at least as negative as
solution gradients. It is noteworthy that COMSOL MultiPhysics solves -0.85 V (vs. Cu/CuSO4) which is equivalent to -0.78 V (vs. SCE).
Eq.s (15) and (16) to achieve the mesh displacement: Therefore, ZRP may not be sufficiently cathodically protecting the steel
substrate if the potential is positive of -0.78 V (vs. SCE). This criterion
x x
2
X + 2
Y =0 might be misleading if the surface area of anode to cathode ratio is very
t t (15)
high such as our case. Consider OCP of Zn is -1,004 mv vs. SCE and for
2 y 2 y Fe is -684 mv vs. SCE. In such a case, the galvanic potential of the
( )+ ( )=0 couple is likely always more negative than -0.78 V vs. SCE, which ob-
X
t
Y
t (16)
viously does not imply sacrificial protection at ALL times. Rather it is
In addition, the normal component n of velocity vector vn of the just because of the surface area of zinc is far larger than iron. Another
anode is calculated using Faraday law and expressed as: disadvantage of potential criterion is that it could be affected by IR drop
x (or y )
( )
M jloc and shows the potential more negative than -0.78 V vs. SCE. This matter
vn = n. = was previously studied extensively by the same authors [35]. Also,
t zF (17)
recently, McMahon et al. [36] reported that ZRP open circuit potential
Note that this velocity is zero for cathode as it is assumed not to be does not predict polarizability performance. Therefore, current density
corroding, i.e. a non-deforming boundary in the model. is a better criterion if only it could be measured such as by simulation.
Fig. 4a shows polarization curves of uncoated carbon steel and
4. Results and discussion unscratched zinc primer on a carbon steel substrate. Using the concept
of mixed potential theory, the intercept of these two curves yields an
4.1. Current density modeling estimate for jgalv of 5.8 μA.cm-2, but this would be for an anode to
cathode surface area ratio of 1:1. In the model, the ratio of the anode
Fig. 3(a) shows the evolution of the simulated current densities on (zinc primer) to the cathode (scratch) regions is 6:1. When this is fac-
the modeled surface of a scratched sample immersed in 3.5 wt% NaCl tored into the polarizations curves, Fig. 4b shows that jgalv increases to
solution from 0 h to 96 h. In all cases, the anodic current is related to 18 μA.cm-2 (per area of carbon steel). However, both these values for
the primer surface and the cathodic current is related to the iron surface jgalv neglect the influence of iR drop in the electrolyte that restricts the
exposed by the scratch. Starting from the edge of the scratch, which is throwing power of the system. That is to say, regions of the zinc coating
the anode/cathode interface, and moving outwards, it can be seen that that are far away from the scratch contribute less to the galvanic cur-
the magnitude of the local anodic current density decreases rapidly. rent density than those regions that are directly adjacent to the scratch;
This is due to the iR drop in the electrolyte [32–34] that restricts the this has been clearly demonstrated by scanning vibrating electrode
distance over which anodes and cathodes interact (throwing power). measurements [37]. The inability of the mixed potential theory to in-
However, over time the steep rise at the anode/cathode interface is corporate the influence of iR drop in the electrolyte is one of its major
depressed after 24 h until it is almost vanished after 96 h. At the same disadvantages in assessing galvanic corrosion, causing it to over-esti-
time, the average anodic current density increases from 4.2 μA.cm-2 at mate the value of jgalv. This is especially true for atmospheric corrosion,
0 h to 5.2 μA.cm-2 at 96 h, and the average cathodic current density where the thinness of the moisture film means results in a very high
decreases from -27.6 μA.cm-2 to -22.5 μA.cm-2. This phenomenon can be electrolyte resistance and thus short throwing distances [33]. Never-
explained as if during the first 24 h the primer contains high amount of theless, the two values of jgalv obtained from Fig. 4 likely represent the
free metallic zinc (zinc dust) that can readily corrode sacrificially to upper and lower bounds for the modelled scratched Zn coating; that is
protect iron. Over time the most of free zinc juxtaposed to the iron has 5.8 < jgalv < 18 μA.cm-2.
corroded such that this region is no longer able to provide cathodic Concerning the COMSOL model, which takes into account iR drop
current for the exposed iron as it did in the first 24 h; i.e. the zinc in this through the electrolyte, averaging the anodic current densities along
region cannot maintain the same corrosion rate. As a result, 96 h could the physical surface of freshly scratched sample (Fig. 3a) yields a value
be defined as a point at which the primer may not be able to provide for jgalv of 6.2 μA.cm-2. This is in excellent agreement with the range of
sacrificial protection as much as it was able before mainly due to the values determined by mixed potential theory and thus serves as vali-
lack of fresh zinc. dation of the model.
It is noteworthy that in all cases from 0 h to 96 h the anodic current Remark the short passivation range for the Zn in the presence of Cl–
densities are increasing, due to both zinc self-corrosion and zinc sacri- seen in Fig. 4 is unusual. However, it is not unique, as it was also seen in
ficial corrosion. At the same time, the cathodic current densities are other works studying zinc coatings such as the work of Li et al. [38].
decreasing as cathodic area is increasing by widening the scratch from Note that once the anode to cathode size ratio is taken into con-
500 μm at 0 h to 670 μm at 96 h due to the consumption of neighboring sideration (Fig. 4b), that this passivation range is positive of the gal-
zinc. Accordingly, zinc electrodes become smaller over 96 h. As ex- vanic corrosion potential and thus has no impact on the values fed into
pected, in all cases the total anodic and cathodic currents vary in a the model.

5
M. Saeedikhani, et al. Corrosion Science 163 (2020) 108296

Fig. 3. Simulated current densities on the anodes and cathodes physical surfaces for periods of (a) 0 h, (b) 2 h, (c) 6 h, (d) 24 h, (e) 48 h, (f) 96 h. The anodic current is
related to the painted surface and the cathodic current is related to the iron surface exposed by the scratch. Continuous change in current density for 96 h is provided
in supplementary data.

4.2. Oxygen transport modeling, micro-pH measurement and microscopy electrochemical reactions are about to take place. In addition, the
cathodic current density being less negative at the middle of scratch is
The ORR makes the major contribution to the total cathodic current due to this regions being the farthest from the zinc primer the cathodic
[30, 39, 40]. Therefore, to have a better understanding of the me- current density have been affected by iR drop [32–34], however, the
chanisms of sacrificial protection provided by the studied zinc rich amount of iR drop is not significant as the scratch is very narrow
primer it is necessary to determine the oxygen concentration profile (500 μm) and the electrolyte is highly conductive.
across the exposed iron surface with the scratched region. This was Assuming that the ORR is under diffusion control, then its rate will
obtained by modeling the oxygen transportation paths (Eq. (1)) through be directly proportional to the local oxygen concentration. Obviously,
the electrolyte at time zero, revealing that oxygen concentration is at its the oxygen concentration right on iron is always zero as it will be re-
maximum on the middle of the scratch and at its minimum on the duced immediately on the surface. Given that the oxygen is reduced to
corners (Fig. 5). Fig. 5 shows that oxygen concentration becomes lower hydroxide ions (Eq. (7)), the pH profile can be expected to match that of
by moving towards the bottom of the scratch (boundary (viii) in the oxygen concertation, i.e. the middle of the scratch region will be
Fig. 2a). Note one may conclude the oxygen concentration in Fig. 5 is more alkaline than the edges. Thebault et al. [20] confirmed this
less on the scratch corners compared to the middle of the scratch be- principle by measuring the pH across a cut edge of a zinc coated steel
cause more oxygen has been consumed (reduced) on the scratch corners sample coated with a pH microelectrode. They observed that pH was
and hence the cathodic current density is slightly more negative at rather acidic (pH 5–6) near to the zinc coating, while it increased to pH
scratch corners (Fig. 3). To the contrary, this conclusion is not correct as 10.5 at farthest distance away from the zinc coating, which was also
the simulation in Fig. 5 is related to initiation time that is when where most of the bulky zinc based corrosion products deposited.

6
M. Saeedikhani, et al. Corrosion Science 163 (2020) 108296

Fig. 4. Polarization behavior of carbon steel and zinc paint (a) 1:1 zinc paint to carbon steel surface area ratio and (b) 6:1 zinc paint to carbon steel surface area ratio.
The experiment is carried out in 3.5 wt% NaCl. The sweep rate is 2 mV.s−1.

Fig. 5. Simulated oxygen concentration at time zero at different cross sections

of 30 μm, 60 μm, and 150 μm away from the iron surface (boundary (viii) in
Fig. 2a). Fig. 6. pH distribution over the whole surface of the samples at 150 μm from
the iron surface (boundary (viii) in Fig. 2(a)) for different exposure hours.

Nevertheless, our micro-pH measurements depicted in Fig. 5 is in good

agreement with the simulated oxygen concentration profile (Figure (5)) and only the pH over the scratch region changes) implies that sacrificial
as well as the Thebault’s study [20]. and barrier protection mechanisms exist together along the surface of
The local pH in the electrolyte could provide information on the the scratch. The high alkalinity above the scratch’s center suggests this
local mechanisms [41]. For instance, the presence of zinc hydroxide region is under active sacrificial protection and the small areas near the
corresponds to the high pH regions, hence, could be an implication of scratch corners that are rather acidic, i.e. pH 5.5–6.5, are the areas
dominated sacrificial protection. Likewise, pH drop could imply that under barrier protection where the corrosion products layer is densely
sacrificial protection becomes less dominant. Fig. 6 depicts the local pH formed. As high alkalinity regions are far wider, sacrificial protection is
variation over the whole surface at ca. 150 μm above the carbon steel therefore more dominant within the scratch region.
surface recorded for immersed samples in 3.5 wt% NaCl solution for up Fig. 7 shows cross-section optical images of scratched zinc primer
to 96 h. For the 0 h sample, the local pH is almost uniform and equal to coated steel substrates that have been exposed to corrode for the set
bulk electrolyte’s initial pH (i.e., ca. pH 6.5). For the 2 h sample, the periods. It can be seen that the bulky corrosion products mainly form in
local pH around the corners of the scratch has fallen to ca. pH 5.5 whilst the middle of the scratch some distance away from the iron/zinc primer
the local pH above the scratch middle has increased to a maximum interfaces. Note that no corrosion products can be seen in Fig. 7c (24 h
value of ca. pH 9.5. This higher pH indicates that more ORR has oc- of exposure), but this is probably due to these being washed or eroded
curred on the middle of the scratch. Likewise, the pH fall at the corners away during the rinsing, cutting or grinding processes required to
of the scratch could be due to the reduction of OH– by Zn2+ (Table 2) prepare the samples for microscopy. The deposited zinc corrosion
which is expected at the corners of the scratch as these areas are likely products are expected to be only loosely adhered to the steel substrate
to have the highest Zn2+ concentrations since they are the closest area especially during the early stages of corrosion exposure [42]. The op-
to the zinc electrodes. tical images in Fig. 7 are thus consistent with the earlier work of The-
After 48 h, the alkalinity of the electrolyte above the middle of the bault et al. [19, 21].
scratch reaches a maximum of pH 11 followed by falling to pH 9.5 after The reason that the zinc corrosion products depositing in the high
96 h. Obviously, the pH profiles of the 2 h and 96 h samples are similar. pH regions in middle of the scratched region is due to the solubility of
The local pH variation over the scratch region between 2 h–96 h Zn2+ ions, which decreases with increasing pH with a minima at about
(note the pH over the zinc surfaces after 2 h is similar for all samples pH 10 after which the solubility of the Zn(OH)42−ions becomes

7
M. Saeedikhani, et al. Corrosion Science 163 (2020) 108296

Fig. 7. Optical microscopy images of the cross section of scratched samples exposed for (a) 2 h, (b) 6 h, (c) 24 h, (d) 48 h, (e) 96 h.

electrolyte from their source (the primer) to the cathodic region. Once
formed this thin film will act as a barrier layer inhibiting the oxygen
reduction reaction. Therefore, barrier action protection likely co-exist
with sacrificial protection within the scratched region from a very early
stage, even if it is outweighed by sacrificial protection later. This is
important as many researches did not recognize an early barrier [4, 5]
except for the work from Thebault et al. that noticed the early barrier
protection [19].

4.3. Micro-Raman spectroscopy

Fig. 9 shows Raman spectra collected from within the scratched

region of corrosion exposed samples for different periods. As the scratch
was very narrow, 500 μm, only a small amount of corrosion product
Fig. 8. SEM image of zinc corrosion products of the 2 h exposed sample ob- was available for analysis even after 96 h. As a result, some peaks in the
tained from the cross section at the right corner of the scratch in Figure a. Raman spectra were not sharp and thus hard to distinguish. Never-
theless, the Raman spectra obtained are still of sufficient quality to
allow some of the corrosion products to be identified by referencing the
important [43]. Nevertheless, Fig. 8 reveals that at high magnification a
peaks to characteristic shifts available in the literature (Table 4).
thin layer of zinc corrosion products does deposit at the corners of the
After 6 h Raman shifts of ε and β – Zn(OH)2, ZnO, simonkolleite [Zn
scratched region, even though this region is neutral or rather acidic
(Cl)2(Zn(OH)2)4] and hydrozincite [(ZnCO3)2(Zn(OH)2)3] were de-
(Fig. 6). This thin layer likely formed at the very early stages of the
tected across the scratch region, which remained the only detected zinc
corrosion exposure (less than 2 h); as these corners represent the
corrosion products until 96 h. ZnO and especially Zn(OH)2 are more
shortest distance that, the zinc ions have to migrate through the
stable at alkaline pH ranges and their formation could catalyze further

8
M. Saeedikhani, et al. Corrosion Science 163 (2020) 108296

Fig. 9. The Raman spectra of zinc corrosion products found within the scratched region for samples exposed to salt spray for different periods. Raman spectra were
obtained from top-view of the samples.

Table 4
The characteristic Raman shifts of the zinc corrosion products found within the scratched region.
Immersion time / h Raman Shift (s) / cm−1 Composition of corrosion product Name of corrosion product References

2 150 ε – Zn(OH)2 ε – Zinc Hydroxide [42,44,45]

220 ZnO Zinc Oxide
255, 390, 720, 900 Zn(Cl)2(Zn(OH)2)4 Simonkolleite
1000 ? ?
6 150 ε – Zn(OH)2 ε – Zinc Hydroxide
380, 740 β – Zn(OH)2 β – Zinc Hydroxide
220 ZnO Zinc Oxide
255, 390, 720, 900 Zn(Cl)2(Zn(OH)2)4 Simonkolleite
1070 (ZnCO3)2(Zn(OH)2)3 Hydrozincite
24 150, 1050 ε – Zn(OH)2 ε – Zinc Hydroxide
220, 440, 1140 ZnO Zinc Oxide
255, 390, 720, 900 Zn(Cl)2(Zn(OH)2)4 Simonkolleite
1070 (ZnCO3)2(Zn(OH)2)3 Hydrozincite
48 150, 367, 1050 ε – Zn(OH)2 ε – Zinc Hydroxide
220, 440 ZnO Zinc Oxide
390, 720, 900 Zn(Cl)2(Zn(OH)2)4 Simonkolleite
1070 (ZnCO3)2(Zn(OH)2)3 Hydrozincite
96 150, 760 ε – Zn(OH)2 ε – Zinc Hydroxide
220, 440 ZnO Zinc Oxide
390, 720 Zn(Cl)2(Zn(OH)2)4 Simonkolleite
1070 (ZnCO3)2(Zn(OH)2)3 Hydrozincite

9
M. Saeedikhani, et al. Corrosion Science 163 (2020) 108296

oxide growth as it has been proposed by Thomas et al. [43] that their
surfaces can support cathodic reactions. The presence of Zn(OH)2
within 96 h is in line with sacrificial protection to the exposed steel in
the scratched region during the first 96 h as was confirmed by micro-
electrode pH measurements. Nevertheless, the presence of si-
monkolleite and hydrozincite implies that barrier protection was also
present at this early stage; note the hydrozincite requires some hours to
form as it is being produced from Zn(OH)2 with CO2.
Altogether, it could be concluded that upon exposure to the elec-
trolyte ZnO and Zn(OH)2 immediately form in the scratch in smaller
amounts at the corners as the oxygen concentration is low in these areas
(Fig. 5). ZnO and Zn(OH)2 may contribute to the formation of si-
monkolleite and hydrozincite until a thin dense layer of these insulating
zinc corrosion products forms at the corners (Fig. 7), which electrically
disconnects these regions from the primer [46]. Therefore, within a
very short period, the corners of the scratch are likely to be protected by
a barrier action of the corrosion products rather than sacrificial action Fig. 10. Simulated electric potential distribution in the 3.5 wt% NaCl electro-
as these areas will not receive cathodic current density under the in- lyte domain at time zero. Color distribution: from bluish regions that are cor-
sulator barrier layer of corrosion products. Note that the present model respond to cathodic areas to reddish regions that correspond to anodic areas.
does not consider the precipitation of corrosion products at this point. Continuous change in electrolyte potential and geometry for 96 h is provided in
Furthermore, as was shown in Fig. 7, the majority of corrosion supplementary data.
products are formed at the middle of the scratch, where the con-
centration of oxygen (Fig. 5) and thus local pH (Fig. 6) are at maxima.
These middle-precipitated corrosion products are likely to be mostly Zn thickness is too thin the zinc dust near the scratch will be consumed
(OH)2 and ZnO which are stable in alkaline ranges [43]. Considering without providing sacrificial protection life that is long enough for the
oxygen continuously diffuses from boundary (ii) in Fig. 2, oxygen re- barrier layer to form. On the other hand, if the zinc primer is too thick
duction will continue to occur in the middle of the scratch region, its cost may outweigh its economic benefits, also the primer may lose.
which leads to further formation and accumulation of corrosion pro- Considering the importance of this matter, the corrosion behavior of
ducts over this region. Accumulation of the corrosion products will not zinc primer electrodes throughout the 96 h exposure period was simu-
lead to a full insulating barrier, as corrosion products do not form on lated (Fig. 11). Note for clarity, only the right half of simulations (which
some areas between the middle of the scratch and both corners which is are symmetrical around the center of the scratch) are shown in Fig. 11,
in agreement with the work of Thebault et al. [19]. Note the corners are along with corresponding optical images of real samples.
covered with dense insulating corrosion products (Fig. 8). This process As the wall of the scratch (boundary (vi) in Fig. 2) is the closet
continues until the free zinc within the throwing power distance of the source of zinc to the iron expose by the scratch, it was expected the
scratch is consumed, which is the point that the primer starts losing whole wall would corrode faster than at any other location boundary.
sacrificial protection action, i.e. 96 h in this study. Surprisingly, to the contrary, both the simulations and the optical
Furthermore, the homogenous reactions in Table 2, cause the elec- images of real samples in Fig. 10 show that zinc primer’ s highest
trolyte concentration to evolve, such that the local Zn2+ concentration corrosion rate occurs over an area comprised of the top half of
near the scratch walls becomes much higher than in the center of the boundary (vi) and leftmost part of boundary (vii), i.e. the circled areas
scratch, with the maximum Zn2+ being at the scratch corners. To in Fig. 10. There is good agreement between the real samples and the
maintain electroneutrality, chloride ions migrate towards the regions simulations in both cases the lower half of the wall of the scratch re-
rich in zinc ions, such that the total ionic strength of electrolyte is at its maining almost un-deformed (not corroded). Likewise, small indents
highest near the scratch walls, which impedes diffusion of oxygen. In occur in the primer structure just outside of the scratch in both the
our model we employed the approach of Simillion et al. [27] to find the models and the real coatings. The indent formation in the primer
species’ diffusion constants dependence on chloride ion concentration structure is due to formation of thin zinc corrosion products layer at the
(see Eq. 4). The net result of the reduced oxygen diffusion through the corners of the scratch (Fig. 7).
more concentrated electrolyte, is that oxygen consumed close to the With regards to the lower half of the scratch wall not being cor-
scratch corners is not replenished, so the local O2 concentration in this roded, despite the zinc in this region being the nearest to the exposed
region is lower than in the middle of the scratch. iron/steel, this phenomenon could be due to the formation of the thin
zinc-based corrosion layer that forms on the exposed iron near to the
4.4. Simulation of the electrodes deformation (corrosion) by ALE method primer wall (Fig. 8). Note as long as there is no turbulence or erosion
caused by the electrolyte, corrosion products can act as a barrier if they
Fig. 10 shows the initial electric potential distribution in the 3.5 wt are dense enough (such as Fig. 8) even if they are not well adhered. As
% NaCl electrolyte domain obtained by solving (Eq. (11)). As expected, this thin layer is insulating [43] it will impede dissolution of the zinc on
the electrolyte’s potential is dominated by the potential ranges that are the lower half of the scratch’s wall or, in other words, its barrier
close to the measured free corrosion potential of zinc (-1.004 V vs SCE), properties add resistance to current flow from the lower half of the
rather than that measured for iron (-0.681 V vs SCE), since the anode’s scratch wall. Hereafter, zinc dust located in the upper part of the scratch
surface is 6 times larger than the cathode’s surface. This is because zinc wall plus zinc dust from the top surface of the coating in the neigh-
is not polarized in presence of NaCl due to quick dissolution. However, borhood of the scratch start to corrode faster to compensate for the
steel can easily be polarized cathodically. The arrows in Fig. 9 indicate inactivated part of the scratch wall. This is the reason indents occur in
the direction of the electric field from more negative potentials (anode) the primer structure just outside of the scratch (Fig. 11c, d, & e). This
towards less negative potentials (cathodes). conclusion might be rather surprising that why would the upper part of
Recalling that Fig. 3 shows the anodic current density is highest at the scratch wall be able to corrode more if it is even further away from
the iron/zinc primer interface, meaning that is where the zinc dis- the middle of the scratch? To address this, Tada et al. [28] reported that
solution rate will be highest. This is an important factor for selecting the in their steel/zinc galvanic couple even though the nearest steel to zinc
right thickness for the coating. On the one hand, if the selected was covered by iron rust (after almost 3 h), zinc continued to protect

10
M. Saeedikhani, et al. Corrosion Science 163 (2020) 108296

Fig. 11. Simulated (left hand side) in 3.5 wt% NaCl electrolyte and optical microscopy (right hand side) images of corrosion of the zinc paint electrodes exposed to
salt fog for (a) 2 h, (b) 6 h, (c) 24 h, (d) 48 h, and (e) 96 h. Circles show the deformed (corroded) areas and arrows show the indent areas. Continuous change in
geometry for 96 h is provided in supplementary data.

the steel surface regions further away from the zinc and iron rust layer. corrosion rate were due to pH only, the lower and upper half of the
This could imply that the throwing power of the zinc coating allows scratch wall would have been corroded the same way. For our studied
protection at farther distances even though the nearest regions are not geometry, we would propose: although the lower half of the scratch
able to receive sacrificial protection. Also, if the reduction in the Zn wall is blocked/blinded by zinc corrosion products, the upper half and

11
M. Saeedikhani, et al. Corrosion Science 163 (2020) 108296

the left most and right most part of the primer are yet able to provide 4.5. Depiction of protection mechanism
sacrificial protection for a certain amount of time.
However, the present model considers the corrosion products are Overall, the results suggest that the studied mechanically damaged
soluble, yet it is still able to simulate the lower corrosion rate at the zinc-based coating protects the exposed steel substrate by the me-
base of the wall. To explain this, it is necessary to look at the root cause chanism depicted in Fig. 12. Almost immediately after the steel sub-
of the formation of these thin corrosion products, which is the low strate is exposed in the damaged region, zinc starts to corrode sacrifi-
oxygen concentration and consequently relatively low pH value in this cially to protect the exposed steel substrate as depicted in Fig. 12a.
vicinity. Therefore, considering oxygen transport in the electrolyte However, the rate of zinc’s corrosion may not be uniform due to loca-
makes the model close to the actual corrosion conditions, even though lized differences in the oxygen concentration. This is followed by the
the corrosion products were considered soluble in the model. deposition of a thin layer of zinc corrosion products that initiates in the
The observation that almost half of the scratch’s wall remained in- corners of the scratch, with the source of the zinc for this thin layer
tact throughout the full 96 h of corrosion exposure, after which time being mainly the walls of the scratch (Fig. 12b). Because the oxygen
barrier layer protection had developed, raises the question “would the concentration is lower at the corners (Fig. 5), the pH remains near
same primer but with half thickness (60 μm) provide the same level of neutral or possibly turns slightly acidic [37], which prevents the cor-
galvanic protection for the same width of scratch but at lower cost?” As rosion layer from thickening (i.e. the solubility of Zn2+ ions remain
such, further studies are in progress to simulate the selection of the high in this region). Nevertheless, as this thin layer evolves into si-
right thickness of the zinc primer depending on the width and size of monkolleite and hydrozincite, it becomes denser and more electrically
the scratch. insulating. This isolates the zinc in the base scratch wall from the ex-
Remark our current study is one step before delamination starts, posed iron and in turn, this results in a decrease in corrosion rate of the
which needs to be understood thoroughly. That is to say, zinc starts to wall. Hereafter, zinc dust located in the upper part of the scratch wall
corrode sacrificially first to protect the scratch region, and after this plus zinc dust from the top surface of the coating in the neighborhood of
step delamination may come to play. Based on our microscopy images the scratch start to corrode faster to compensate for the inactivated part
(see Fig. 7) delamination has not occurred throughout 96 h of corrosion of the scratch wall. This is the reason indents occur in the primer
exposure. Therefore, such simulation was not necessary at this step. In structure just outside of the scratch Fig. 11 & 12c). Finally, due to the
addition, delamination study, even though being very important, was high pH that develops in the middle of the scratch (resulting from the
not the focus of this work. higher oxygen concentration), zinc oxides and hydroxides precipitate
and accumulate as a thick barrier layer.

Fig. 12. The schematic of the evolution of the protection mechanism of steel from corrosion in the damaged region of a zinc rich paint. (a) early stages of corrosion
exposure mainly sacrificial protection, (b) pH change in the scratch, (c) zinc corrosion products precipitate to form a barrier layer.

12
M. Saeedikhani, et al. Corrosion Science 163 (2020) 108296

Our micro-pH in conjunction with the real optical images of the Appendix A. Supplementary data
samples clearly show that the majority (bulky) of zinc corrosion pro-
ducts are in the middle of the scratch where the pH is alkaline. This is in Supplementary material related to this article can be found, in the
line with the work of Tada et al. [28] where they found zinc corrosion online version, at doi:https://doi.org/10.1016/j.corsci.2019.108296.
products in the high pH region distant from the zinc layer, although
Prosek et al. [47] found zinc corrosion products in neutral regions. One References
possible reason for the difference between our study and Prosek et al.’s
study could be the geometry and zinc content difference as they have [1] J.R. Scully, Corrosion 75 (2019) 123.
studied on a cut edge with the zinc to iron ratio far smaller than our [2] M. Mouanga, M. Puiggali, B. Tribollet, V. Vivier, N. Pébère, O. Devos, Electrochim.
Acta 88 (2013) 6.
study. [3] M. Saeedikhani, S.L. Wijesinghe, D.J. Blackwood, Eng. J. 23 (2019) 223.
[4] H. Shi, F. Liu, E.H. Han, Surf. Coatings Technol. 205 (2011) 4532.
5. Conclusion [5] S. Shreepathi, P. Bajaj, B.P. Mallik, Electrochim. Acta 55 (2010) 5129.
[6] Z. Feng, G.S. Frankel, Corrosion 70 (2013) 95.
[7] S.M.A. Shibli, B.N. Meena, R. Remya, Surf. Coat. Technol. 262 (2015) 210.
The sacrificial protection provided by a scratched 96 wt% zinc-rich [8] Z. Mirzakhanzadeh, A. Kosari, M.H. Moayed, R. Naderi, P. Taheri, J.M.C. Mol,
primer to a carbon steel substrate has been simulated using a time Corros. Sci. 138 (2018) 372.
[9] P. Taheri, P. Laha, H. Terryn, J.M.C. Mol, Appl. Surf. Sci. 356 (2015) 837.
dependent model within the Corrosion Module of COMSOL
[10] L.I. Fockaert, P. Taheri, S.T. Abrahami, B. Boelen, H. Terryn, J.M.C. Mol, Appl. Surf.
Multiphysics. The corrosion rates produced by the new model were Sci. 423 (2017) 817.
validated by comparison with mixed potential theory. The main find- [11] P. Taheri, H. Terryn, J.M.C. Mol, Prog. Org. Coatings. 89 (2015) 323.
[12] A. Nazarov, E. Diler, D. Persson, D. Thierry, J. Electroanal. Chem. Lausanne 737
ings of the model are that while the middle of the scratch is protected
(2015) 129.
by sacrificial protection, at the corners of the scratch a local cathodic [13] R. Jain, R. Pitchumani, Surf. Coatings Technol. 337 (2018) 223.
inhibition results in barrier protection. This dual protection mechanism [14] M.T. Montañés, R. Sánchez-Tovar, J. García-Antón, V. Pérez-Herranz, Corros. Sci.
was concluded to be due to variations in the dissolved oxygen con- 51 (2009) 2733.
[15] R. Srinivasan, L.H. Hihara, ECS Trans. 45 (2013) 77.
centration over the surfaced of the scratched primer. By using Arbitrary [16] T.M. Amorim, C. Allély, J.P. Caire, Proc. COMSOL Conf. (2008) 1.
Lagrangian–Eulerian method, the model was able to simulate how the [17] F. Thébault, B. Vuillemin, R. Oltra, C. Allely, K. Ogle, Electrochim. Acta 82 (2012)
geometry of the corroding region would evolve over time and this was 349.
[18] F. Thébault, B. Vuillemin, R. Oltra, C. Allely, L. Dosdat, K. Ogle, ECS Trans. 3 (2007)
in agreement with observations made by optical microscopy. In addi- 343.
tion, the moving boundary simulation allowed to predict the sacrificial [19] F. Thébault, B. Vuillemin, R. Oltra, K. Ogle, C. Allely, Electrochim. Acta 53 (2008)
protection period as it takes into account the consumption of zinc. 5226.
[20] F. Thébault, B. Vuillemin, R. Oltra, C. Allely, K. Ogle, Electrochim. Acta 56 (2011)
Finally, the experimental and simulation results were combined to 8347.
produce a depiction of how the protection mechanism of mechanically [21] F. Thébault, B. Vuillemin, R. Oltra, C. Allely, K. Ogle, Corros. Sci. 53 (2011) 201.
damaged zinc rich primer s evolves over time. [22] S.R. Cross, S. Gollapudi, C.A. Schuh, Corros. Sci. 88 (2014) 226.
[23] J. Donea, A. Huerta, J.-Ph. Ponthot, A. Rodriguez-Ferran, Vol. 1, E. Stein, R. de
This work could open a horizon for future galvanic corrosion si-
Borst, and T.J.R. Hughes,, p.1 (Chapter 14), John Wiley and Sons Ltd., West Sussex
mulation researches to consider switching from simulation under con- (2004).
stant electrode size assumption to the more realistic case of deformed [24] K.B. Deshpande, Corros. Sci. 52 (2010) 3514.
[25] W. Sun, L. Wang, T. Wu, G. Liu, Corros. Sci. 78 (2014) 233.
(corroded) electrode over simulation (corrosion) time. As moving
[26] V. Topa, A.S. Demeter, L. Hotoiu, D. Deconinck, J. Deconinck, Electrochim. Acta 77
boundary simulations capture the zinc coating consumption, such (2012) 339.
models have the potential to be developed to become a tool for coating [27] H. Simillion, N. Van den Steen, H. Terryn, J. Deconinck, Electrochim. Acta 209
thickness selection based on the corrosivity of the environment, how- (2016) 149.
[28] E. Tada, K. Sugawara, H. Kaneko, Electrochim. Acta 49 (2004) 1019.
ever, corrosion modelers are fully aware that there exists a long way to [29] R.F. Weiss, Deep. Res. 17 (1970) 721.
go for such an achievement. [30] K. Ogle, V. Baudu, L. Garrigues, X. Philippe, J. Electrochem. Soc. 147 (2000) 3654.
[31] J.O.M. Bockris, S.U.M. Khan, A Molecular Level Approach, Springer Science, New
York, 1993, p. 319.
Data availability [32] J.T. Waber, B. Fagan, J. Electrochem. Soc. 103 (1956) 64.
[33] J.T. Waber, J. Electrochem. Soc. 103 (1956) 567.
The evolution of electrolyte potentials, electrode current densities, [34] J.W. Oldfield, Galvanic Corrosion, ASTM Special Technical Publication, 1988, p. 5.
[35] M. Saeedikhani, S. Wijesinghe, D. Blackwood, Corrosion 75 (2019) 756.
and electrodes deformation (corrosion) by time using the reported [36] M.E. McMahon, J.T. Burns, J.R. Scully, Prog. Org. Coatings 135 (2019) 392.
animations by COMSOL are provided in the PowerPoint slide as sup- [37] M. Yan, V.J. Gelling, B.R. Hinderliter, D. Battocchi, D.E. Tallman, G.P. Bierwagen,
plementary. One sample simulated current density data at 0 h in Corros. Sci. 52 (2010) 2636.
[38] R. Li, J. Liang, Y. Hou, Q. Chu, RSC Adv. 5 (2015) 60698.
Fig. 3(a) is also provided. The whole set of raw/processed data to re- [39] S. Thomas, I.S. Cole, Y. Gonzalez-Garcia, M. Chen, M. Musameh, J.M.C. Mol,
produce the exact findings cannot be shared at this stage as this re- H. Terryn, N. Birbilis, J. Appl. Electrochem. 44 (2014) 747.
search is ongoing. [40] A.P. Yadav, A. Nishikata, T. Tsuru, J. Electroanal. Chem. Lausanne 585 (2005) 142.
[41] M.E. McMahon, R.J. Santucci Jr., J.R. Scully, RSC Adv. 9 (2019) 19905.
The authors declare no conflict of interest.
[42] T. Ohtsuka, M. Matsuda, Corrosion 59 (2003) 407.
[43] S. Thomas, I.S. Cole, M. Sridhar, N. Birbilis, Electrochim. Acta 97 (2013) 192.
Acknowledgments [44] M.C. Bernard, A. Hugot-Le Goff, D. Massinon, N. Phillips, Corros. Sci. 35 (1993)
1339.
[45] A.H. Le Goff, S. Joiret, B. Saïdani, R. Wiart, J. Electroanal. Chem. Lausanne 263
The authors would like to acknowledge the NUS-SINGA Ph.D. (1989) 127.
Scholarship for funding to Mohsen Saeedikhani. A Part of the financial [46] A. Amirudin, D. Thierry, Prog. Org. Coatings. 28 (1996) 59.
support for this project was provided by the Singapore Institute of [47] T. Prosek, A. Nazarov, H.B. Xue, S. Lamaka, D. Thierry, Prog. Org. Coatings. 99
(2016) 356.
Manufacturing Technology (SIMTech) under project number U17-S-
016SU.

13