Corrosion Science 48 (2006) 22912303 www.elsevier.

com/locate/corsci

3D computer simulation of the inuence of microstructure on the cut edge corrosion behaviour of a zinc aluminium alloy galvanized steel
S.G.R. Brown
a

a,*

, N.C. Barnard

Materials Research Centre, School of Engineering, University of Wales, Singleton Park, Swansea SA2 8PP, UK b Engineering Doctorate Centre, Wales, University of Wales, Swansea SA2 8PP, UK Received 15 June 2005; accepted 16 July 2005 Available online 10 October 2005

Abstract Recently, experimental work has been reported that demonstrates the eects of microstructural variations within ZnAl Galfan type coatings on the corrosion behaviour of cut-edge material, i.e. those cases where both the underlying steel and the organic coated Galfan layer are simultaneously exposed to a corrosive environment [J. Elvins, J.A. Spittle, D.A. Worsley, Microstructural changes in zinc aluminium alloy galvanising as a function of processing parameters and their inuence on corrosion, Corros. Sci. 47 (11) (2005) 27402759]. In this paper a nite dierence numerical model of localized corrosion has been applied in an attempt to simulate this type of corrosion. Results from the model are compared to experimental observations. 2005 Elsevier Ltd. All rights reserved.
Keywords: A. Galfan coatings; B. Modelling; B. Finite dierence; C. Localized corrosion; C. Microstructure

1. Introduction Protection of steel used in the construction industry against the detrimental eects of corrosion is an area of signicant engineering importance. Corrosion eects have a
*

Corresponding author. Tel.: +44 1792 295284. E-mail address: S.G.R.Brown@swansea.ac.uk (S.G.R. Brown).

0010-938X/$ - see front matter 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.corsci.2005.08.015

2292

S.G.R. Brown, N.C. Barnard / Corrosion Science 48 (2006) 22912303

negative impact on material properties leading to unwanted economic consequences. Manufacturing methods that limit the deleterious inuence of physicochemical processes on the worlds durable goods and infrastructure often involve the use of protective coatings. Metallic coatings are used in the protection of iron and steel, traditionally applied by hot-dip methods. The capacity of zinc to aord barrier and cathodic protection as a steel coating has seen the wide-spread use of galvanized products [2]. In the construction industry there is an increasing demand for the improvement of corrosion resistance of organically coated steels, typically used in cladding and roong of buildings. Much interest concerns the development of galvanizing coatings containing $4.5 wt.% Al. Coating alloys of this type have been seen to exhibit greater ductility and superior corrosion resistance than conventional galvanizing alloy systems (galvanized steels suer larger Zn dissolution rates [3]) some of which include small aluminium additions in order to prevent large scale intermetallic formation during fabrication [15]. The 4.5 wt.% Al zinc alloy, often known as Galfan, is commonly applied to steel and used as the substrate for organically coated steel products, used as proled cladding in building solutions. In such applications the product is known to be susceptible to localized degradation at cut-edges leading to delamination of organic coating and degradation of the metallic coating remaining exposed. 2. Numerical model The numerical model comprises eld-based calculations for electrical potential and diffusion/migration coupled to standard electrochemical equations [6,7]. A brief overview of the model is provided below. The model uses an orthogonal structured computational grid to represent both the solid and liquid electrolyte in each simulation of dimensions nx ny nz. Dierential equations employed by the model are approximated via straightforward nite dierence methods. The model permits non-uniform nodal spacings in the x, y and z directions. Each cell in the grid possesses an integer value identifying the site as either electrolyte or solid. Electrolyte sites are then also dened in terms of concentration (c, mol dm3), electrical potential (/, V), current density vector (ix, iy, iz, A m2), volume fraction of the cell occupied by corrosion product (p) and the volume fraction of the cell that has already been corroded away. Solid sites contain information of composition and volume fraction corroded (f ). 2.1. A rst order model of corrosion Attempting to simulate the localized corrosion of zinc alloy steel coatings with a physically based numerical model inevitably requires simplifying assumptions to be made. A rst order model that deals with the evolution and movement of a single ionic concentration (increases of which are brought about by anodic dissolution) has been used. The model assumes that the Zn2+ rapidly undergoes hydrolysis to produce solid zinc hydroxide via Zn2 H2 Ol ! ZnOH H aq aq
ZnOHaq

H2 Ol ! ZnOH2s H

S.G.R. Brown, N.C. Barnard / Corrosion Science 48 (2006) 22912303

2293

In this way the model tracks the eective free proton concentration (c, mol dm3) arising from the dissolution of zinc and aluminium during the corrosion process. Direct estimation of the overall corrosion eects experienced considering only one electroactive species distinguishes the model from those that deal with individual species [8,9]. The distribution of Zn2+ and pH across a galvanic couple can be seen to support this approach due to the crucial role of corrosion products in the model [10,11]. 2.2. Determination of electrical potential Interface sites identied in the model as liquid sites adjacent to one or more solid sites are the subject of a calculation to determine the reversible potential, /0 (V) including perturbations in concentration, c ([H ]), deviating from the initial concentration, c0, using aq Eq. (3). H+ is frequently potential determining for metal surfaces in contact with aqueous solutions (due to the hydroxylated nature of the interface metal oxide lm). /0 E0   RT c ln nF c0 3

where R is the universal gas constant (J mol1 K1), F is the Faraday constant (C mol1), T is the temperature (K), E0 is the standard electrode potential (V) determined from a weighted average of all solid cells in the local Moore neighbourhood. These potentials are used as Dirichlet boundary conditions to determine the electrical potential in the remaining electrolyte cells using the Laplace equation r jr/ 0 4

where j is the electrical conductivity (X1 m1). The rate at which dissolution occurs at interface sites is related to the potentials experienced at those positions. Kinetics of dissolution for phases present at the solidliquid interface are approximated from experimental data obtained from the rotating disc electrode (RDE) technique providing the Tafel relationship for pure zinc and commercially pure aluminium samples. The rate of dissolution, a function of current density, is derived from the electrical potentials associated with the composition of a cell, and consequently from Eq. (3), thus in this model, current density, i / [H+]. 2.3. Determination of localized corrosion The amount by which each corroding metal/electrolyte interface cell corrodes is calculated from the local current density. The volume fraction of material dissolved in each cell is then updated. Once a cell is completely dissolved it is redened as an electrolyte cell. Therefore over a time increment the total increase in concentration in the evolving system is quantied via CA
ny nx nz XXX x1 y1 z1

dfx;y;z ca Dxx Dy y Dzz

2294

S.G.R. Brown, N.C. Barnard / Corrosion Science 48 (2006) 22912303

where nx, ny and nz dene the total number of sites on the mesh, fx,y,z is the volume fraction of a cell transforming, ca is a constant volume concentration contribution term and DxxDyyDzz is the volume of the cell. 2.4. Diusion and migration of electroactive species The movement of charged particles in solution, which are evolved during localized corrosion, is modelled by laws governing convection, diusion and the transport of ions through an electrical eld. Diusion through 5% (0.86 mol dm3) NaCl solution (ignoring the convection term) over a time increment due to ordinary diusion and migration eects is modelled by Eq. (6). oc r Drc zF r ucr/ ot 6

where D is the diusion coecient at 298 K (m2 s1), z is the unit charge and u is the ionic mobility (m2/(V s)). 2.5. The rate determining cathodic reaction and corrosion products Left unregulated the localized corrosion eects predicted by the model, as presented so far, would propagate unopposed. This would not be realistic since in this case it is known that the cathodic reaction is the rate determining process. To ensure electro-neutrality in the model there is a counterbalancing procedure to Eq. (5), whereby an equal and opposite concentration decrease, Cc, is forced at all metal/electrolyte interface sites (subject to satisfying the condition below), Cc CA V SE 7

where VSE is the total volume of all electrolyte cells at the solidelectrolyte interface that are able to accept a further decrease in concentration as part of the cathodic reaction. There is also a lower limit placed on cell concentration such that sites acting cathodically approaching this limit will cease to lower concentration. This has the eect of lowering the overall corrosion rate. Another mechanism regulating the localized corrosion eects in the model is the presence of corrosion products. Deposited at surface sites within pre-determined concentration and potential bands, these sites hinder the diusion of ionic species through them by reducing the diusion coecient according to the volume of any given cellx,y,z occupied by corrosion product (px,y,z). The simple quadratic relationship below has been used to calculate diusion coecients for all cells containing corrosion products, Dx;y;z D1 px;y;z 2 8

If px,y,z reaches unity then that site acts as a complete barrier to diusion. It has been assumed that at each time increment the volume of corrosion product deposited at the surface is 5% of the volume of the metal dissolved. Also, all sites containing corrosion product have their fraction, px,y,z, reduced by an arbitrary amount in each time increment to simulate gradual dissolution of corrosion product. It is recognised that this is a very simple model of corrosion product formation.

S.G.R. Brown, N.C. Barnard / Corrosion Science 48 (2006) 22912303

2295

2.6. Calculation of current density for comparison to reported experimental results using the scanning vibrating electrode technique (SVET) The current densities presented below are normalised vertical current densities calculated via numerical dierentiation of the electrical eld calculated via Eq. (9). This post-processing enables comparison of the predictions made by the computer model to experimentally derived data, e.g. SVET data, and the form of observed current density elds from the model has already been shown to match previously reported numerical/ analytical solutions [12,13]. iy 1 d/ b dy 9

where b is the electrical resistivity (X m) of 5% NaCl solution. Validation of a model attempting to simulate corrosion of more complex geometries would present challenges for comparison to SVET data [14]. 3. Results and discussion The discretized representation of the coating layer can be seen in Figs. 13 which show the 3D solid model of Zn4.5 wt.% Al steel coatings based on observations made by Elvins et al. [4]. Fig. 1 shows the presence of primary zinc crystals within a eutectic matrix (region D) and this is the setup which is used in simulating the cut-edge corrosion performance.

Fig. 1. 200 50 200 Element solid model of cut edge corrosion specimen viewed from below. Asteel, Binert resin layer, Celectrolyte and DZnAl coating layer comprising primary zinc dendrites (darker) in a eutectic matrix (lighter). Length, x = 0.77 mm, height, y = 0.54 mm and width, z = 250 lm.

2296

S.G.R. Brown, N.C. Barnard / Corrosion Science 48 (2006) 22912303

Fig. 2. 200 50 200 Solid model of cut edge corrosion specimen viewed from above with electrolyte not shown. Asteel, Binert resin layer, CZnAl coating layer comprising primary zinc dendrites (E) in a eutectic matrix (D). Length, x = 0.77 mm and width, z = 250 lm.

Fig. 3. 200 50 200 Solid model of cut edge corrosion specimen viewed from above with electrolyte not shown. Asteel, Binert resin layer, CZnAl coating layer comprising primary zinc dendrites in a eutectic matrix. Length, x = 2.5 mm and width, z = 0.77 mm.

Fig. 1 shows a view of the mesh from below. The presence of a steel substrate (region A) is included in addition to an inert polymeric resin material (region B) which prevents contact between the coating surface and electrolyte. Region C in Fig. 1 shows the electrolyte. Fig. 2 shows the same mesh from the top, but the electrolyte material has been removed. Regions D and E in Fig. 2 show eutectic and primary zinc dendrites respectively. In Figs. 1 and 2 the mesh used comprises 200 50 200 (x, y, z, respectively) elements and represents a cross-section of 250 lm 0.77 mm. This mesh was used to give good resolution of the eutectic/dendritic phases. In contrast, Fig. 3 shows another mesh of 200 50 200 elements, again viewed from the top with the electrolyte cells removed. This mesh represents an initial area of 2.5 mm 0.77 mm exposed to electrolyte and is closer to the geometry used in experiments [1,2,4] of 20 mm 2 mm. In all cases a non-uniform regular mesh has been used. This allows greater resolution in regions of interest, i.e. electrolyte elements away from the coating layer are relatively coarse (e.g. 1.5 32 0.5 lm) whereas elements near to and in the coating layer are ner (e.g. 1.5 8 0.5 lm) to allow a much more gradual simulation of the dissolution of metallic cells.

S.G.R. Brown, N.C. Barnard / Corrosion Science 48 (2006) 22912303

2297

In all cases zero ux boundary conditions are applied to the sides of the mesh. For the electrical potential calculation a zero ux boundary condition is applied to the top face of the electrolyte (i.e. the face furthest from the coating layer). In contrast, for the diusion calculations, a xed Dirichlet boundary condition (the initial electrolyte concentration) is applied to the top face of the electrolyte region to simulate the presence of excess electrolyte. Fig. 4 shows a view from below of the free proton concentration eld in the electrolyte for the mesh in Fig. 2 after 24 h exposure to electrolyte (solid elements not shown). The high concentration regions (lighter areas) indicate locations where localized corrosion of mainly primary zinc dendrites is occurring. Diusion/migration of this eld can be seen extending away form the corroding sites, although highest free proton concentrations remain close to the coating layer, reecting the involvement of H+ in the hydrolysis of evolving Zn2+. Fig. 5 shows a view from below of the corresponding calculated electrical potential eld in the electrolyte for the same simulation as Fig. 4 at the same time. Anodic areas (darker) can be seen above the two ZnAl layers with lighter cathodic regions over the central steel portion. The lighter cathodic areas tend to be most cathodic where less corrosion product has been deposited. The characteristic dome shape of these elds can be clearly seen. Figs. 6 and 7 show concentration and electrical potential elds respectively for the grid in Fig. 3 after 24 h exposure to electrolyte. Exactly the same form of distribution for these elds is obtained when compared to the smaller mesh results in Figs. 4 and 5. Fig. 8 shows the solid elements (but ignoring corrosion products) for the mesh in Fig. 2 after 24 h. Some primary zinc dendrites have been completely dissolved while other larger dendrites have been only partially removed. This may be compared to a published micrograph of a real specimen [1]. In all the above cases, the standard electrode potential of the primary zinc means that it corrodes in preference to the ZnAl eutectic and steel. In all cases the exposed steel acts as a large cathode although the rate of corrosion is controlled by the speed at which diusion/ migration can occur (mass transport control) and the diminishing eectiveness of the steel

Fig. 4. Concentration eld in NaCl electrolyte for 200 50 200 mesh after 24 h exposure (non-electrolyte elements not shown). Anodic regions can be seen as the lighter regions.

2298

S.G.R. Brown, N.C. Barnard / Corrosion Science 48 (2006) 22912303

Fig. 5. Electrical potential eld in NaCl electrolyte for 200 50 100 mesh after 24 h exposure (non-electrolyte elements not shown). Anodic regions can be seen as the darker regions (B) with cathodic zones (especially at A) developing over the central steel area.

Fig. 6. Concentration eld in NaCl electrolyte for 200 50 200 mesh after 24 h exposure (non-electrolyte elements not shown). Anodic regions can be seen as the lighter regions.

as a cathode due to corrosion product deposition. This is exactly the mechanism by which the coating layer protects the steel, preferential corrosion of the zinc phase and the associated covering of the steel by corrosion products. Eventually in the simulation, a sucient build up of corrosion product on the steel eectively prevents it acting as a cathode and the corrosion process eectively comes to a halt. Fig. 9 shows the simulated build up of corrosion product over time. Initially two parallel bands of product form that eventually link together to cover most of the steel surface. In this gure sites containing a volume fraction of P25% corrosion product are shown. Fig. 10 shows the calculated current density map 100 lm above the surface (to mimic published SVET results). Anodic regions can be seen as darker areas corresponding to primary zinc dendrite locations. The most active cathodic regions at this stage of the process can be seen at those areas of the steel least covered by corrosion products (cf. Fig. 9). While the

S.G.R. Brown, N.C. Barnard / Corrosion Science 48 (2006) 22912303

2299

Fig. 7. Electrical potential eld in NaCl electrolyte for 200 50 200 mesh after 24 h exposure (non-electrolyte elements not shown). Anodic regions can be seen as the darker regions (especially C) with cathodic zones (especially at A and B) developing in the central steel area.

Fig. 8. Magnied view of the 200 50 200 mesh after 24 h exposure. Primary zinc dendrites have completely dissolved in some positions on the coating layer (A) and are as yet only partially dissolved in other areas (B).

model is relatively simple at this stage the form of distribution of current density appears to match experimental observation although calculation of current density vectors via numerical dierentiation induces an inevitable numerical error. This comparison is also complicated by the fact that the SVET probe measures an averaged value over the surface area of the probe tip whereas the model performs current density calculations on an element by element basis. Finally, a useful comparison between model and experiment can be made when considering total zinc loss (tzl). Elvins et al. [1] compared zinc losses for three coatings. While each Galfan coating was the same nominal thickness they were solidied under dierent

2300

S.G.R. Brown, N.C. Barnard / Corrosion Science 48 (2006) 22912303

Fig. 9. The development of a corrosion product layer on the surface of the exposed steel. Top views of 200 50 200 element mesh from top to bottom at times 10, 16 and 24 h. The electrolyte is not shown. The inert resin layer is shown as A, the ZnAl coating layer (primary zinc dendrites and eutectic matrix) is shown as B. The exposed steel regions are C and the growing corrosion product layer is D.

Fig. 10. Predicted current density eld (A m2) 100 lm above the corroding cut-edge after 24 h exposure, length scales are in mm.

cooling rates. Their specimens were labelled as 55H, 80H and 100H in order of increasing cooling rate. Here the number refers to the power output of the cooling unit in percentage

S.G.R. Brown, N.C. Barnard / Corrosion Science 48 (2006) 22912303

2301

Fig. 11. Top view solid meshes used for lowest cooling rate (a) and highest cooling rate (b) cut edge specimens corresponding to 55H and 100H in Ref. [1]. From top to bottom in (a) and (b) the ve separate layers are: Inert coating, ZnAl Galfan layer, steel, Galfan layer, inert coating.

Fig. 12. Measured and simulated total zinc losses (tzl) for three cut edge specimens where 55H, 80H and 100H represent increased cooling applied to coating layer during cooling and solidication.

terms and H refers to heavy gauge material (i.e. 0.67 mm thick). Examination of the microstructures in the Galfan layer has been used to generate 3D solid meshes for the three

2302

S.G.R. Brown, N.C. Barnard / Corrosion Science 48 (2006) 22912303

dierent cases. Fig. 11 shows two such grids and the ner dendritic structure in the more rapidly cooled specimen can be seen in Fig. 11b. The model was then used to simulate corrosion for the three cases for 24 h immersion in 5% NaCl solution. A comparison of measured versus predicted total zinc losses is shown in Fig. 12. In both model and experiment the total zinc loss decreases with increasing cooling rate and the associated increased neness of the dendritic microstructure. While the predicted total zinc losses are of the right order it is clear that quantitatively there is a discrepancy. This may possibly be due to insuciently ne meshes in the model to represent the microstructure coupled with the simplicity of the corrosion deposition algorithm. However, the modelled result is considered to be good in that it appears able to capture the observed eect of metallurgical microstructure on corrosion behaviour. 4. Conclusions A simple rst order model of localized corrosion linked to nite dierence based eld calculations has been applied to the case of a ZnAl alloy coated steel strip immersed in electrolyte along its cut edge. The model is able to capture several features of the localized corrosion process and compares favourably to experimental measurements made for this particular case. In the system studied the role of corrosion products is important. Unfortunately, there is little quantitative information available for this aspect of the process and the model is therefore unavoidably approximate in this respect. However, this type of model would seem to oer potential for simulation and improved understanding of such corroding systems. With increased sophistication such an approach could provide a useful bridge between electrochemistry and metallurgy by being able to capture important features of both disciplines within the same computational framework. Acknowledgments This work was carried out under the Engineering Doctorate Scheme, the help and support of Corus Group and the funding of EPSRC is gratefully acknowledged. The authors are also grateful to Prof. H.N. McMurray at Swansea for many useful discussions concerning the work. References
[1] J. Elvins, J.A. Spittle, D.A. Worsley, Microstructural changes in zinc aluminium alloy galvanising as a function of processing parameters and their inuence on corrosion, Corros. Sci. 47 (11) (2005) 27402759. [2] H.N. McMurray, G. Parry, B.D. Jes, Corrosion resistance of ZnAl alloy coated steels investigated using electrochemical impedance spectroscopy, Iron Steelmaking 25 (3) (1998) 210215. [3] Y. Li, Corrosion behaviour of hot dip zinc and zincaluminium coatings on steel in seawater, Bull. Mater. Sci. 24 (4) (2001) 355360. [4] J. Elvins, J.A. Spittle, D.A. Worsley, Relationship between microstructure and corrosion resistance in ZnAl alloy coated galvanised steels, Corros. Eng. Sci. Technol. 38 (3) (2003) 197204. [5] H.C. Shih, J.W. Hsu, C.M. Sun, S.C. Chung, The lifetime assessment of hot-dip 5% AlZn coatings in chloride environments, Surf. Coat. Technol. 150 (1) (2002) 7075. [6] S.G.R. Brown, N.C. Barnard, H.N. McMurray, 3-dimensional modelling of localized corrosion eects in zinc and zinc alloy steel coatings, in: P. Vincenzini (Ed.), Proc. 3rd Int. Conf. on Computational Modelling and Simulation of Materials III (CIMTEC 2004), Acireale (CT), Italy, 2004, pp. 393403.

S.G.R. Brown, N.C. Barnard / Corrosion Science 48 (2006) 22912303

2303

[7] N.C. Barnard, S.G.R. Brown, H.N. McMurray, Modelling the localized corrosion eects experienced by zinc4.5 wt.% aluminium steel coatings in 5% NaCl solution, in: C.A. Brebbia, V.G. DeGiorgi, R.A. Adey (Eds.), Simulation of Electrochemical Processes, vol. 99, WIT Press, Southampton, 2005, ISBN 1-84564-0128. [8] J. Johnsen, A. Jssang, T. Jssang, P. Meakin, An experimental study of the quasi-two-dimensional corrosion of aluminium foils and a comparison with two-dimensional computer simulations, Physica A 242 (34) (1997) 356376. [9] P. Cordoba-Torres, R.P. Nogueira, L. de Miranda, L. Brenig, J. Wallenborn, V. Fairen, Cellular automaton simulation of a simple corrosion mechanism: mesoscopic heterogeneity versus macroscopic homogeneity, Electrochim. Acta 46 (19) (2001) 29752989. [10] E. Tada, S. Satomi, H. Kaneko, The spatial distribution of Zn2+ during galvanic corrosion of a Zn/steel couple, Electrochim. Acta 49 (14) (2004) 22792285. [11] E. Tada, K. Sugawara, H. Kaneko, Distribution of pH during galvanic corrosion of a Zn/Steel couple, Electrochim. Acta 49 (7) (2004) 10191026. [12] H.S. Isaacs, The eect of height on the current distribution measured with a vibrating probe, J. Electrochem. Soc. 138 (3) (1991) 722728. [13] D. Livingstone-Bridge, J.C. Myland, K.B. Oldham, A model of ionic current densities in the vicinity of a corroding disk-shaped region, Electrochem. Commun. 3 (8) (2001) 384389. [14] J.S. Lee, M.L. Reed, R.G. Kelly, Combining rigorously controlled crevice geometry and computational modeling for study of crevice corrosion scaling factors, J. Electrochem. Soc. 151 (7) (2004) 423433.