**3.3 Second phases**

### **3.3.1 Attack around isolated IM particles**

IM particles may be either anodic or cathodic relative to the matrix under any particular set of solution conditions. As a result, two main types of pit morphologies are typically observed around isolated IM particles. Circumferential pits generally appear as trenches around a more or less intact particle and the corrosion attack is mainly in the matrix phase. In the case of clusters, the development of the trench around the cluster is a secondary corrosion reaction following extensive grain boundary attack. Trenching around isolated particles does not lead to more severe corrosion (Chen, Gao et al. 1996; Ilevbare, Schneider et al. 2004; Schneider, Ilevbare et al. 2004; Boag, Hughes et al. 2011). This can be understood in terms of a quantity called the pit stability product first reported by Galvele. This quantity is defined as *i.r* where *r* is the depth of the pit and *i* is the current density. This quantity must be greater than 10-2 Acm-1 for a nascent pit to be able to grow rapidly enough to establish a long enough diffusion path for an oxygen gradient to be established. This results in oxygen

Fig. 7. Dark field scanning transmission electron micrograph of constituent particle co-

Over the years there have been a number of studies that have assessed the effect of intermetallic particles on the corrosion susceptibility of specific aluminium alloys (Scully, Knight et al. 1993; Birbilis and Buchheit 2005),(Zamin 1981; Mazurkiewicz and Piotrowski 1983; Scully, Knight et al. 1993; Seri 1994). In the 1990s, Buchheit collected the corrosion potential values for intermetallic phases common to aluminium alloys mainly in chloride containing solutions (Buchheit 1995). More recently various groups have focussed on the electrochemical properties of Fe containing intermetallics (Pryor and Fister 1984; Afseth, Nordlien et al. 2002), and Cu containing intermetallics (Searles, Gouma et al. 2001; Birbilis, Cavanaugh et al. 2006; Birbilis and Buchheit 2008) which has been expanded into a comprehensive treatise covering a variety of common intermetallics present in aluminium alloys (Frankel 1998; Birbilis and Buchheit 2005; Birbilis and Buchheit 2008). A summary of

IM particles may be either anodic or cathodic relative to the matrix under any particular set of solution conditions. As a result, two main types of pit morphologies are typically observed around isolated IM particles. Circumferential pits generally appear as trenches around a more or less intact particle and the corrosion attack is mainly in the matrix phase. In the case of clusters, the development of the trench around the cluster is a secondary corrosion reaction following extensive grain boundary attack. Trenching around isolated particles does not lead to more severe corrosion (Chen, Gao et al. 1996; Ilevbare, Schneider et al. 2004; Schneider, Ilevbare et al. 2004; Boag, Hughes et al. 2011). This can be understood in terms of a quantity called the pit stability product first reported by Galvele. This quantity is defined as *i.r* where *r* is the depth of the pit and *i* is the current density. This quantity must be greater than 10-2 Acm-1 for a nascent pit to be able to grow rapidly enough to establish a long enough diffusion path for an oxygen gradient to be established. This results in oxygen

existing with S-phase (Al2CuMg) precipitate particles in AA2024-T3 sheet

the results of these studies is shown below in Table 7.

**3.3.1 Attack around isolated IM particles** 

**3.3 Second phases** 

reduction near the mouth of the pit at these early stages and an acidic salt solution at the active pit face. Studies have shown that this product is too low for trenching events to develop into stable pits around isolated cathodic particles even for S-phase dealloying (Schneider, Ilevbare et al. 2004).

The second type of pit morphology is due to the selective dissolution of the constituent particle. Pits of this type are often deep and may have remnants of the particle in them. Figure 8 shows a model of dealloying of an S-phase IM particle which leads to a Cu-enriched remnants as well as non-faradaic liberation of the Cu. Under neutral pH conditions magnesium and aluminium are preferentially dissolved from the Al2CuMg phase, leaving a Cu-enriched and high surface area remnant, which then exhibits solution potentials noble to the matrix (Buchheit, Grant et al. 1997). Ultimately, the form that copper takes on the surface is thought to be important in determining the corrosion-performance of alloys such as AA2024. The redistribution of copper has been demonstrated to enhance the kinetics of oxygen reduction processes and negatively affect corrosion. In dealloying from a bulk phase the physical structure of a surface has been predicted by percolation theory to be dependent upon the dissolution rate and concentration of the noble elements in the phase (Sieradzki 1993; Newman and Sieradzki 1994). Rapid dissolution rates lead to more porous network structures, where there is a possibility that unoxidized fragments enriched in the more noble metal will be released into solution, whereas slow dissolution allows surface diffusion and relaxation processes to maintain a stable surface structure. Also, if the noble metal content is sufficient, dealloying will not lead to an isolation of the percolation network. Theory suggests the copper concentration of 25 at% contained in the Al2CuMg phase allows it to dealloy and form both porous copper-rich networks and also to release clusters of both oxidized and unoxidized copper into an electrolyte. It is also noted that hydrodynamic forces may assist in the release of fragments (Buchheit, Martinez et al. 2000; Vukmirovic, Dimitrov et al. 2002; Muster 2009).


Table 7. Summary of corrosion potentials for intermetallic particles common to Al alloys

High Strength Al-Alloys: Microstructure, Corrosion and Principles of Protection 243

Localised Corrosion Hierarchy

<5 min 15 min 30 min 120 min Increasing time

Fig. 9. Hierarchy of localised corrosion attack in 0.1 M NaCl for AA2024-T3 from Boag et al.

information relating to microstructural features (from the work used to generate Table 7) was overlaid spatially above real micrographs of AA7075. The data represents the given anodic or cathodic current sustained (at a potential corresponding to the open circuit potential) for each microstructural feature. It is seen that, depending on environment, the currents realised are rather different; and this reconciles very well with 'ground truth'

An assumption behind the electrochemical work is that stable pits are on the extreme end (in an extreme value statistics sense) of the metastable pitting events as can be seen in Table 8 for total charge passed, pit stability product and the pit size. There have been other studies that suggest that the current transients are fundamentally different involving activation repassivation events within active pits (Sasaki, Levy et al. 2002). Some studies have also tried to connect phenomenological measurements to the current transients such as through measuring the volume of the trench around an IM particle to the total charge passed. While the total charge passed is similar for both cases, it can be seen (Figure 9) that S-phase dealloying is complete with 5 minutes which means that there are likely to be 75 trenching events occurring simultaneously/cm2 within the first 5 minutes. However, the frequency of metastable current transients on AA2024-T3 was close to 0.02 s which is nearly four orders of magnitude smaller. Transient currents associated with η-phase have been observed in

As described earlier, the reports of clustering of intermetallic particles in Al-alloys is an emerging area of research in aluminium corrosion. While severe pitting has for many years been attributed to clustering (Chen, Gao et al. 1996; Gao, Feng et al. 1998; Liao, Olive et al. 1998; Schneider, Ilevbare et al. 2004), it is only recently that detailed studies of the connections between clusters and severe pitting have emerged. Part of this work is based on electrochemical studies of current transients measured on corroding surfaces to try to

Trenching

De-alloy trench trench

S-phase

AlCuFeMn *e.g.* Al Cu Fe Al (Cu,Fe,Mn) 7 6 2

(Al,Cu) (Fe,Mn) Si x x

observations from SEM of exposed specimens.

AA7075 (Wloka and Virtanen 2008).

**3.3.2 Corrosion and particles clusters** 

Increasing activity

(Boag, Hughes et al. 2011)

Fig. 8. Schematic of dealloying of Al2CuMg phase contained within an aluminium alloy matrix. Anodic polarisation of the particle results in preferential loss of Al and Mg. A Curich network, which coarsens with age, and is susceptible to break-up during hydrodynamic flow, releases small Cu-rich particles (diameter approximately 10-100 nm). The electrically isolated particles are dissolved, making Cu ions available for replating as elemental Cu onto cathodic sites. The replated cathodic sites serve as efficient local cathodes that stimulate secondary pitting (after Buchheit et al., 2000)

Localised corrosion activity is, however, a complex phenomenon that is still under active research. Localised corrosion leads to local pH gradients as recently studied in detail by Ilevbare and Schneider (Ilevbare, Schneider et al. 2004; Schneider, Ilevbare et al. 2004). Enhanced oxygen reduction at cathodic sites generate hydroxyl ions promoting local pH increases, which can then modify the subsequent rate and morphology of corrosion propagation. Recent work by Boag et al. revealed the timescales on which various forms of attack occurred on isolated IM particles in AA2024-T3. In this study the IM particles in AA2024-T3 were divided into S-phase, AlCuFeMn phases and (Al,Cu)x(Fe,Mn)ySi since these compositions appeared to have separated timescales for attack (Figure 9). These compositions follow the electrochemical activity and provide one method for categorising the IM particles which brings together composition and electrochemistry. Whether this is sufficient for modelling purposes remains to be answered since it does not distinguish, for example, between compositional domains within the one particle that have different corrosion activities as seen in Figure 5. With respect to Figure 9, attack began at the S-phase particles which underwent dealloying indicating anodic behaviour, but then switched to trenching indicating active cathodic behaviour. This was followed by trenching around AlCuFeMn particles after 30 minutes and finally by trenching around (Al,Cu)xSi(Fe,Mn)y at 120 minutes.

The precise morphology of particle-induced pitting is important for emerging damage accumulation models. For these models to be predictive, it is necessary to develop a comprehensive, self-consistent accounting of this type of pitting. In cases where the electrochemical characteristics of constituent particles have been rigorously characterized, they have been found to have much more complicated behaviour than categorised by simple characterizations like "noble" or "active". In recent years, Buchheit and co-workers have attempted an accounting of such phenomena (Figure 10). In such work, electrochemical

Fig. 8. Schematic of dealloying of Al2CuMg phase contained within an aluminium alloy matrix. Anodic polarisation of the particle results in preferential loss of Al and Mg. A Curich network, which coarsens with age, and is susceptible to break-up during hydrodynamic flow, releases small Cu-rich particles (diameter approximately 10-100 nm). The electrically isolated particles are dissolved, making Cu ions available for replating as elemental Cu onto cathodic sites. The replated cathodic sites serve as efficient local cathodes that stimulate

Localised corrosion activity is, however, a complex phenomenon that is still under active research. Localised corrosion leads to local pH gradients as recently studied in detail by Ilevbare and Schneider (Ilevbare, Schneider et al. 2004; Schneider, Ilevbare et al. 2004). Enhanced oxygen reduction at cathodic sites generate hydroxyl ions promoting local pH increases, which can then modify the subsequent rate and morphology of corrosion propagation. Recent work by Boag et al. revealed the timescales on which various forms of attack occurred on isolated IM particles in AA2024-T3. In this study the IM particles in AA2024-T3 were divided into S-phase, AlCuFeMn phases and (Al,Cu)x(Fe,Mn)ySi since these compositions appeared to have separated timescales for attack (Figure 9). These compositions follow the electrochemical activity and provide one method for categorising the IM particles which brings together composition and electrochemistry. Whether this is sufficient for modelling purposes remains to be answered since it does not distinguish, for example, between compositional domains within the one particle that have different corrosion activities as seen in Figure 5. With respect to Figure 9, attack began at the S-phase particles which underwent dealloying indicating anodic behaviour, but then switched to trenching indicating active cathodic behaviour. This was followed by trenching around AlCuFeMn particles after 30 minutes and finally by trenching around (Al,Cu)xSi(Fe,Mn)y at

The precise morphology of particle-induced pitting is important for emerging damage accumulation models. For these models to be predictive, it is necessary to develop a comprehensive, self-consistent accounting of this type of pitting. In cases where the electrochemical characteristics of constituent particles have been rigorously characterized, they have been found to have much more complicated behaviour than categorised by simple characterizations like "noble" or "active". In recent years, Buchheit and co-workers have attempted an accounting of such phenomena (Figure 10). In such work, electrochemical

secondary pitting (after Buchheit et al., 2000)

120 minutes.

#### Localised Corrosion Hierarchy

Fig. 9. Hierarchy of localised corrosion attack in 0.1 M NaCl for AA2024-T3 from Boag et al. (Boag, Hughes et al. 2011)

Increasing time

information relating to microstructural features (from the work used to generate Table 7) was overlaid spatially above real micrographs of AA7075. The data represents the given anodic or cathodic current sustained (at a potential corresponding to the open circuit potential) for each microstructural feature. It is seen that, depending on environment, the currents realised are rather different; and this reconciles very well with 'ground truth' observations from SEM of exposed specimens.

An assumption behind the electrochemical work is that stable pits are on the extreme end (in an extreme value statistics sense) of the metastable pitting events as can be seen in Table 8 for total charge passed, pit stability product and the pit size. There have been other studies that suggest that the current transients are fundamentally different involving activation repassivation events within active pits (Sasaki, Levy et al. 2002). Some studies have also tried to connect phenomenological measurements to the current transients such as through measuring the volume of the trench around an IM particle to the total charge passed. While the total charge passed is similar for both cases, it can be seen (Figure 9) that S-phase dealloying is complete with 5 minutes which means that there are likely to be 75 trenching events occurring simultaneously/cm2 within the first 5 minutes. However, the frequency of metastable current transients on AA2024-T3 was close to 0.02 s which is nearly four orders of magnitude smaller. Transient currents associated with η-phase have been observed in AA7075 (Wloka and Virtanen 2008).

### **3.3.2 Corrosion and particles clusters**

As described earlier, the reports of clustering of intermetallic particles in Al-alloys is an emerging area of research in aluminium corrosion. While severe pitting has for many years been attributed to clustering (Chen, Gao et al. 1996; Gao, Feng et al. 1998; Liao, Olive et al. 1998; Schneider, Ilevbare et al. 2004), it is only recently that detailed studies of the connections between clusters and severe pitting have emerged. Part of this work is based on electrochemical studies of current transients measured on corroding surfaces to try to

High Strength Al-Alloys: Microstructure, Corrosion and Principles of Protection 245

H2 evolution None? H2 Evolution from within corrosion rings:

corrosion initiation within rings of corrosion product on AA2024-T3. Rings of corrosion product were observed a early as 1952 by Pearson et al. in 2S aluminium alloy and developed as early as 15 minutes after immersion. These rings were also observed in the work of Wei's group (Chen, Gao et al. 1996). Hughes et al. reported that the early stages of corrosion within these rings is characterised by H2 evolution and extensive grain boundary attack on the surface which happens prior to, or at the same time as trenching around IM particles (Figure 11). Boag et al. showed that only the most clustered sites on the surface of AA2024-T3 were associated with chloride signals drawing a strong link between clustering and stable pitting. The subsurface attack at these sites was almost exclusively intergranular, penetrating as much as 60 μm in 120 minutes exposure to 0.1M NaCl. The substantial intergranular attack extending deep into the AA2024-T3 and preceded the development of open pits, suggesting that intergranular attack proceeds prior to any substantial grain

These results suggest that trenching is either a secondary corrosion process or that there are different mechanisms of corrosion initiation which contribute to stable pitting. To further complicate the picture, Zhou, Thompson and co-workers observed that in some alloys there were large S/θ composite particles and that stable pits were established by direct penetration through the composite particle into the subsurface region of the alloy. This type of attack may be responsible for the widely accepted view that severe pitting can start at Sphase particles. Clearly there may be several avenues to the establishment of stable pits and

Severe pitting (Chen, Gao et al. 1996; Liao,

Co-operative corrosion (Hughes, Boag et al.

Coupling of different IM particle types Clustered trenching (Liao, Olive et al. 1998) Corrosion Rings (Liao, Olive et al. 1998)

Subsurface S-phase etchout

It can be continuous or irregular

> 13 (Trueman 2005) to 18 (Pride, Scully et

Grain etchout

al. 1994)

Olive et al. 1998)

2011)

Metastable Stable

Attack Type Individual IM particles Clusters

Propagation Virtually none GBA, IGA

*i.r* <10-2A/cm >10-2A/cm

Trenching/S-phase etch [(Kolics, Besing et al. 2001) (Liao, Olive et al. 1998; Schneider, Ilevbare et al. 2004) (Lacroix, Ressier et al. 2008), (Suter 2001; Blanc, Gastaud et al.

Attack descripition Pitting (Chen, Gao et al. 1996; Liao, Olive et al. 1998)

2003)

Attribution S-phase dealloying and etchout, Trenching

Total charge (μC) < 13 (Trueman 2005) to 18 (Pride, Scully et al. 1994)

Table 8. Corrosion Attack on A2024-T3

etchout, even at the surface.

this is still an area of active research.

Phenomenological

Electrochemical

Fig. 10. Left: 3-D representations of the spatial variation in electrochemical reaction rate associated with the 7075 alloy microstructure at a potential of -0.8Vsce in aerated 0.1M NaCl solution at 23°C and pH 2.5 (a), pH 6 (b), and pH 12 (c). Right: SE images of AA7075 after a 24 h exposure under free corrosion conditions in aerated 0.1M NaCl solution at 23°C and pH 2.5 (a), pH 6 (b), and pH 12 (c)

establish the boundary between the *i.r.* product for metastable and stable pits. Such current transients manifest in three ways (i) spontaneous nucleation/passivation events which have very short timescales, (ii) metastable pitting events which have lifetimes up to a several seconds on AA2024-T3 before decaying back to background currents and (iii) stable pitting events, which increase to a transient peak but then decay to a constant current indicating ongoing electrochemical activity. The literature in this area largely comes from a phenomenological or electrochemical perspective, using different terminologies, Table 8, summaries the activities from these different perspectives.

In terms of the relationship of clustering to stable or severe pitting, it can be seen clearly from Figure 10 that a higher density of IM particles, means a greater level of electrochemical activity. Excessive trenching is the most common explanation for the development of stable pitting from clusters of IM particles. However, in terms of developing models for corrosion initiation, it is important to understand all the pathways to the establishment of stable pits and the mechanisms for those pathways. Recently, Hughes and co-workers reported


Table 8. Corrosion Attack on A2024-T3

244 Recent Trends in Processing and Degradation of Aluminium Alloys

(a)

(b)

(c)

establish the boundary between the *i.r.* product for metastable and stable pits. Such current transients manifest in three ways (i) spontaneous nucleation/passivation events which have very short timescales, (ii) metastable pitting events which have lifetimes up to a several seconds on AA2024-T3 before decaying back to background currents and (iii) stable pitting events, which increase to a transient peak but then decay to a constant current indicating ongoing electrochemical activity. The literature in this area largely comes from a phenomenological or electrochemical perspective, using different terminologies, Table 8,

In terms of the relationship of clustering to stable or severe pitting, it can be seen clearly from Figure 10 that a higher density of IM particles, means a greater level of electrochemical activity. Excessive trenching is the most common explanation for the development of stable pitting from clusters of IM particles. However, in terms of developing models for corrosion initiation, it is important to understand all the pathways to the establishment of stable pits and the mechanisms for those pathways. Recently, Hughes and co-workers reported

Fig. 10. Left: 3-D representations of the spatial variation in electrochemical reaction rate associated with the 7075 alloy microstructure at a potential of -0.8Vsce in aerated 0.1M NaCl solution at 23°C and pH 2.5 (a), pH 6 (b), and pH 12 (c). Right: SE images of AA7075 after a 24 h exposure under free corrosion conditions in aerated 0.1M NaCl solution at 23°C and pH

Position (µm)

summaries the activities from these different perspectives.

Reaction Rate (A/cm2)

2.5 (a), pH 6 (b), and pH 12 (c)

corrosion initiation within rings of corrosion product on AA2024-T3. Rings of corrosion product were observed a early as 1952 by Pearson et al. in 2S aluminium alloy and developed as early as 15 minutes after immersion. These rings were also observed in the work of Wei's group (Chen, Gao et al. 1996). Hughes et al. reported that the early stages of corrosion within these rings is characterised by H2 evolution and extensive grain boundary attack on the surface which happens prior to, or at the same time as trenching around IM particles (Figure 11). Boag et al. showed that only the most clustered sites on the surface of AA2024-T3 were associated with chloride signals drawing a strong link between clustering and stable pitting. The subsurface attack at these sites was almost exclusively intergranular, penetrating as much as 60 μm in 120 minutes exposure to 0.1M NaCl. The substantial intergranular attack extending deep into the AA2024-T3 and preceded the development of open pits, suggesting that intergranular attack proceeds prior to any substantial grain etchout, even at the surface.

These results suggest that trenching is either a secondary corrosion process or that there are different mechanisms of corrosion initiation which contribute to stable pitting. To further complicate the picture, Zhou, Thompson and co-workers observed that in some alloys there were large S/θ composite particles and that stable pits were established by direct penetration through the composite particle into the subsurface region of the alloy. This type of attack may be responsible for the widely accepted view that severe pitting can start at Sphase particles. Clearly there may be several avenues to the establishment of stable pits and this is still an area of active research.

High Strength Al-Alloys: Microstructure, Corrosion and Principles of Protection 247

mainly to anodic dissolution of (i) the precipitation free zone where noble alloying elements such as copper are depleted, (ii) anodic second phase precipitates at the grain boundary, or (iii) grain boundaries with segregated alloying elements, such as magnesium, or impurity elements. In the case of the Al-Cu-Mg system, for example, numerous previous studies have suggested that in AA2024 aluminium alloy, when copper-rich precipitates (CuAl2) formed at grain boundaries, copper-depleted regions develop adjacent to the boundaries, which are anodic with respect to copper-rich grain boundaries and the grain matrix. A lot of IGC in AA2024 aluminium alloy can be explained as micro-galvanic corrosion of copper-depleted regions driven by cathodic areas of copper-rich grain boundaries and grain matrix. Further, when S phase (CuMgAl2) particles are preferentially precipitated at grain boundaries (typically <100 nm), their anodic nature with respect to the adjacent grain matrix, results in

Not all IGA is so well understood. In AA2024-T3, for example, IGC has been observed in the absence of second phase precipitates, and penetrates up to 60 µm within 120 min of immersion in 0.1 M NaCl at ambient temperature (Glenn, Muster et al. 2011; Hughes, Boag et al. 2011). While some of the attacked grain boundaries were decorated with CuAl2 or CuMgAl2 precipitates, many grain boundaries that were subject to intergranular attack were not associated with such precipitates, as shown in Figure 12. The TEM image of the corrosion front reveals two parts of a grain boundary, (i) the attacked grain boundary region behind the corrosion front and (ii) the intact grain boundary ahead of the corrosion front. Clearly, precipitation at the grain boundary is absent. Interestingly, although corrosion propagated in a confined region along the grain boundary and the corrosion front followed the grain boundary, corrosion development across the grain boundary is uneven with the interior of grain B being preferentially attacked. Corrosion has developed more than 70 nm

**200 nm**

Corrosion Front

Grain Boundary

their preferential dissolution.

into grain B, with comparably little attack on grain A.

Grain A

Grain B

Fig. 12. Transmission electron micrograph of the corrosion front, revealing two parts of a grain boundary: the attacked grain boundary region behind the corrosion front and the

Grain Boundary

Grain C

intact grain boundary ahead the corrosion front

Fig. 11. Stages to stable Pitting after Hughes et al. (Hughes, Boag et al. 2011)

### **3.4 Intergranular attack**

Intergranular corrosion is a phenomenon of which the precise mechanisms have been under debate for almost half a century (Hunter 1963). Whilst in a simple view intergranular corrosion can be considered as a special form of microstructurally influenced corrosion, intergranular corrosion can be summarised as a process whereby the grain boundary 'region' of the alloy is preferentially attacked, this is most often because it is anodic to the bulk or adjacent alloy microstructure. Intergranular corrosion can initiate at second phase IM particles in the surface, from pits and from grain boundaries at the surface. Intergranular corrosion penetrates more rapidly than pitting corrosion, and whilst both may have a deleterious effect on corrosion fatigue, the sharper tips produced by intergranular attack are drastic stress concentrators which may reduce the number of cycles to failure.

Corrosion activity may develop because of some heterogeneity in the grain boundary structure. In aluminium-copper alloys, precipitation of Al2Cu particles at the grain boundaries leaves the adjacent solid solution anodic and more prone to corrosion. With aluminium-magnesium alloys the opposite situation occurs, since the precipitated phase Mg2Al3 is less noble than the solid solution. Serious intergranular attack in these two alloys may however be avoided, provided that correct manufacturing and heat treatment conditions are observed.

The distribution of second phase material has a significant influence on the corrosion behaviour of high strength aluminium alloys. Thus, if second phases are located preferentially at grain boundaries, they may promote intergranular corrosion (IGC) due to their compositional, and, hence, electrochemical differences with respect to the adjacent alloy matrix. Further, as a result of precipitation at grain boundaries, the formation of a narrow band on either side of the grain boundary, the precipitation free zone (PFZ), also influences the corrosion behaviour of aluminium alloys since the PFZ is depleted of particular alloying elements. Thus, intergranular corrosion of high strength aluminium alloys is often attributed to compositionally different features at the grain boundary due

Time (mins) 2.5 5 10 20 30 120 >200

S-phase dealloy Trenching and Cu-enrichment

1st Appearance Max Depth

Rapid penetration

Continued deposition Well established

1st Appearance

Intermittent reactivation and new sites

1st Appearance Multiple Sites and larger

1st Appearance

1st Appearance

Fig. 11. Stages to stable Pitting after Hughes et al. (Hughes, Boag et al. 2011)

drastic stress concentrators which may reduce the number of cycles to failure.

Intergranular corrosion is a phenomenon of which the precise mechanisms have been under debate for almost half a century (Hunter 1963). Whilst in a simple view intergranular corrosion can be considered as a special form of microstructurally influenced corrosion, intergranular corrosion can be summarised as a process whereby the grain boundary 'region' of the alloy is preferentially attacked, this is most often because it is anodic to the bulk or adjacent alloy microstructure. Intergranular corrosion can initiate at second phase IM particles in the surface, from pits and from grain boundaries at the surface. Intergranular corrosion penetrates more rapidly than pitting corrosion, and whilst both may have a deleterious effect on corrosion fatigue, the sharper tips produced by intergranular attack are

Corrosion activity may develop because of some heterogeneity in the grain boundary structure. In aluminium-copper alloys, precipitation of Al2Cu particles at the grain boundaries leaves the adjacent solid solution anodic and more prone to corrosion. With aluminium-magnesium alloys the opposite situation occurs, since the precipitated phase Mg2Al3 is less noble than the solid solution. Serious intergranular attack in these two alloys may however be avoided, provided that correct manufacturing and heat treatment

The distribution of second phase material has a significant influence on the corrosion behaviour of high strength aluminium alloys. Thus, if second phases are located preferentially at grain boundaries, they may promote intergranular corrosion (IGC) due to their compositional, and, hence, electrochemical differences with respect to the adjacent alloy matrix. Further, as a result of precipitation at grain boundaries, the formation of a narrow band on either side of the grain boundary, the precipitation free zone (PFZ), also influences the corrosion behaviour of aluminium alloys since the PFZ is depleted of particular alloying elements. Thus, intergranular corrosion of high strength aluminium alloys is often attributed to compositionally different features at the grain boundary due

IMP

IGA

Domes

Grain Attack

**3.4 Intergranular attack** 

conditions are observed.

H2 Production

Corrosion rings

mainly to anodic dissolution of (i) the precipitation free zone where noble alloying elements such as copper are depleted, (ii) anodic second phase precipitates at the grain boundary, or (iii) grain boundaries with segregated alloying elements, such as magnesium, or impurity elements. In the case of the Al-Cu-Mg system, for example, numerous previous studies have suggested that in AA2024 aluminium alloy, when copper-rich precipitates (CuAl2) formed at grain boundaries, copper-depleted regions develop adjacent to the boundaries, which are anodic with respect to copper-rich grain boundaries and the grain matrix. A lot of IGC in AA2024 aluminium alloy can be explained as micro-galvanic corrosion of copper-depleted regions driven by cathodic areas of copper-rich grain boundaries and grain matrix. Further, when S phase (CuMgAl2) particles are preferentially precipitated at grain boundaries (typically <100 nm), their anodic nature with respect to the adjacent grain matrix, results in their preferential dissolution.

Not all IGA is so well understood. In AA2024-T3, for example, IGC has been observed in the absence of second phase precipitates, and penetrates up to 60 µm within 120 min of immersion in 0.1 M NaCl at ambient temperature (Glenn, Muster et al. 2011; Hughes, Boag et al. 2011). While some of the attacked grain boundaries were decorated with CuAl2 or CuMgAl2 precipitates, many grain boundaries that were subject to intergranular attack were not associated with such precipitates, as shown in Figure 12. The TEM image of the corrosion front reveals two parts of a grain boundary, (i) the attacked grain boundary region behind the corrosion front and (ii) the intact grain boundary ahead of the corrosion front. Clearly, precipitation at the grain boundary is absent. Interestingly, although corrosion propagated in a confined region along the grain boundary and the corrosion front followed the grain boundary, corrosion development across the grain boundary is uneven with the interior of grain B being preferentially attacked. Corrosion has developed more than 70 nm into grain B, with comparably little attack on grain A.

Fig. 12. Transmission electron micrograph of the corrosion front, revealing two parts of a grain boundary: the attacked grain boundary region behind the corrosion front and the intact grain boundary ahead the corrosion front

High Strength Al-Alloys: Microstructure, Corrosion and Principles of Protection 249

Fig. 13. a) Scanning electron micrograph of the alloy surface after 4 h of immersion in 0.5 M

boundaries; and b) grain stored energy map of the same area as a), obtained with threshold

Fig. 14. Scanning electron micrograph of the cross section of AA 2099-T8 aluminium alloy

In AA7xxx aluminium alloys, when anodic precipitates, such as η-MgZn2 phase, are formed at the grain boundaries, then these are relatively active with respect to the grain matrix. IGC occurs, with micro-galvanic coupling between the η-MgZn2 phase precipitates and the aluminium matrix adjacent to the particles providing the driving force (Wadeson, Zhou et al. 2006). Again however susceptibility to intergranular corrosion is strongly dependent on the heat treatment condition and its effect on grain boundary solute segregation and the morphology and composition of the grain boundary precipitate and the surrounding alloy matrix. (Knight 2003). The most resistant heat treatments are based on the use of over-aging to the T7 treatment or more complex heat treatments which involve retrogression and reaging to minimise the trade off between alloy strength and intergranular corrosion resistance (Polmear 1995; Davies 1999). The images in Figure 15 help rationalise the origins

**10 μm** 

after polarization to 0.824 V (SCE) in 0.5 M NaCl solution

of intergranular corrosion, in the Al-Zn-Mg system.

NaCl solution at ambient temperature, revealing attacked local regions along grain

value for misorientation set at 1.3

Further investigation of the relationship between grain boundary misorientation and its susceptibility to corrosion revealed that the distribution of grain boundary misorientation for the attacked grain boundaries was similar to that in the as-fabricated alloy (Glenn, Muster et al. 2011; Luo to be published). Thus, there must be other factor(s) that influence the corrosion susceptibility of grain boundaries and its zone of influence.

It was revealed using electron backscatter diffraction (EBSD) that grain stored energy plays a significant role in grain boundary attack. The stored energy of a pixel was determined from the misorientation of a pixel from its neighbours within grains. For each pair of pixels, with misorientation above a selected threshold value, a mean boundary energy was calculated using the Read-Shockley equation:

$$\chi\_{\mathbf{s}} = \chi\_0 \oplus (\mathbf{A} - \ln \theta)$$

where θ is the misorientation angle, γ0 and A are constants. The stored energy of an individual grain/subgrain is determined by averaging the sum of the mean boundary energies over the area of individual grains/subgrains. Then, the spatial distribution of the stored energy is represented as a map, which reflects the average population density of dislocations in individual grains/subgrains.

Figures 13(a) and (b) display a scanning electron micrograph of an AA2024-T3 aluminium alloy surface after immersion in 0.5 M NaCl solution for 4 h at ambient temperature and the stored energy map of the same area (Luo to be published). Brighter grains/subgrains have a higher level of stored energy. The corrosion product has been removed from the alloy surface which has been further cleaned using an argon plasma. The attacked local regions along grain boundaries are clearly evident. Comparing the SEM image with the stored energy map of Figure 13(b), it is evident that the regions of attack are located along the grain boundaries that surround grains of relatively high stored energy. Corrosion is not confined within the region immediately adjacent to the grain boundaries, but has developed 1-2 µm into the grains of relatively high stored energy, suggesting that grains with relatively high level of defects are more susceptible to corrosion. The cold work applied to the alloy to achieve the T3 temper resulted in relatively high population density of dislocations in the alloy. The population density of dislocations may vary from grain to grain since grains with different orientations to the rolling direction are subjected to higher levels of strain. Thus, more defects may be introduced in certain grains which have experienced more deformation than other grains. Consequently, some grains are more susceptible to corrosion than other.

Figure 14 displays a backscattered electron micrograph of the cross section of AA2099-T8 aluminium alloy after polarization to 0.824 V (SCE) in a 0.5 M NaCl solution. Grain boundary attack is evident to a depth up to ~55 μm. Individual subgrains were selectively attacked, as indicated by arrows. Some attacked subgrains are close to the surface region, with others being relatively deep into the bulk alloy. It is believed that the subgrains close to the surface region were exposed to the solution earlier than those far away from the surface region. Thus, the attack to certain subgrains within the inner regions suggested that the selective attack is determined by the intrinsic microstructure. EBSD indicates that the attacked sites tend to be the grains of relatively high stored energy compared with the intact grains that are surrounding the attacked grains, suggesting that the grains of high stored energy have relatively high corrosion susceptibility (Luo to be published).

Further investigation of the relationship between grain boundary misorientation and its susceptibility to corrosion revealed that the distribution of grain boundary misorientation for the attacked grain boundaries was similar to that in the as-fabricated alloy (Glenn, Muster et al. 2011; Luo to be published). Thus, there must be other factor(s) that influence

It was revealed using electron backscatter diffraction (EBSD) that grain stored energy plays a significant role in grain boundary attack. The stored energy of a pixel was determined from the misorientation of a pixel from its neighbours within grains. For each pair of pixels, with misorientation above a selected threshold value, a mean boundary energy was

γs = γ<sup>0</sup> θ (A – ln θ) where θ is the misorientation angle, γ0 and A are constants. The stored energy of an individual grain/subgrain is determined by averaging the sum of the mean boundary energies over the area of individual grains/subgrains. Then, the spatial distribution of the stored energy is represented as a map, which reflects the average population density of

Figures 13(a) and (b) display a scanning electron micrograph of an AA2024-T3 aluminium alloy surface after immersion in 0.5 M NaCl solution for 4 h at ambient temperature and the stored energy map of the same area (Luo to be published). Brighter grains/subgrains have a higher level of stored energy. The corrosion product has been removed from the alloy surface which has been further cleaned using an argon plasma. The attacked local regions along grain boundaries are clearly evident. Comparing the SEM image with the stored energy map of Figure 13(b), it is evident that the regions of attack are located along the grain boundaries that surround grains of relatively high stored energy. Corrosion is not confined within the region immediately adjacent to the grain boundaries, but has developed 1-2 µm into the grains of relatively high stored energy, suggesting that grains with relatively high level of defects are more susceptible to corrosion. The cold work applied to the alloy to achieve the T3 temper resulted in relatively high population density of dislocations in the alloy. The population density of dislocations may vary from grain to grain since grains with different orientations to the rolling direction are subjected to higher levels of strain. Thus, more defects may be introduced in certain grains which have experienced more deformation than other grains. Consequently, some grains are more

Figure 14 displays a backscattered electron micrograph of the cross section of AA2099-T8 aluminium alloy after polarization to 0.824 V (SCE) in a 0.5 M NaCl solution. Grain boundary attack is evident to a depth up to ~55 μm. Individual subgrains were selectively attacked, as indicated by arrows. Some attacked subgrains are close to the surface region, with others being relatively deep into the bulk alloy. It is believed that the subgrains close to the surface region were exposed to the solution earlier than those far away from the surface region. Thus, the attack to certain subgrains within the inner regions suggested that the selective attack is determined by the intrinsic microstructure. EBSD indicates that the attacked sites tend to be the grains of relatively high stored energy compared with the intact grains that are surrounding the attacked grains, suggesting that the grains of high stored

energy have relatively high corrosion susceptibility (Luo to be published).

the corrosion susceptibility of grain boundaries and its zone of influence.

calculated using the Read-Shockley equation:

dislocations in individual grains/subgrains.

susceptible to corrosion than other.

Fig. 13. a) Scanning electron micrograph of the alloy surface after 4 h of immersion in 0.5 M NaCl solution at ambient temperature, revealing attacked local regions along grain boundaries; and b) grain stored energy map of the same area as a), obtained with threshold value for misorientation set at 1.3

Fig. 14. Scanning electron micrograph of the cross section of AA 2099-T8 aluminium alloy after polarization to 0.824 V (SCE) in 0.5 M NaCl solution

In AA7xxx aluminium alloys, when anodic precipitates, such as η-MgZn2 phase, are formed at the grain boundaries, then these are relatively active with respect to the grain matrix. IGC occurs, with micro-galvanic coupling between the η-MgZn2 phase precipitates and the aluminium matrix adjacent to the particles providing the driving force (Wadeson, Zhou et al. 2006). Again however susceptibility to intergranular corrosion is strongly dependent on the heat treatment condition and its effect on grain boundary solute segregation and the morphology and composition of the grain boundary precipitate and the surrounding alloy matrix. (Knight 2003). The most resistant heat treatments are based on the use of over-aging to the T7 treatment or more complex heat treatments which involve retrogression and reaging to minimise the trade off between alloy strength and intergranular corrosion resistance (Polmear 1995; Davies 1999). The images in Figure 15 help rationalise the origins of intergranular corrosion, in the Al-Zn-Mg system.

High Strength Al-Alloys: Microstructure, Corrosion and Principles of Protection 251

alumina and therefore show extensive enrichment at the metal/oxide interface. In contrast, elements such as magnesium and lithium for example, have a lower ΔGº/n value than aluminium and, therefore, are more likely to appear in the oxide or electrolyte solution

In environments where aluminium alloys continually experience anodic dissolution it has been suggested that alloys with a wide range of copper concentrations (0.06 to 26 at%) can display copper enrichment at the metal-oxide interface (Blanc, Lavelle et al. 1997; Habazaki, Shimizu et al. 1997; Garcia-Vergara, Colin et al. 2004). Once a certain level of enrichment occurs, copper atoms (and most likely other noble alloying elements) are thought to arrange themselves into clusters through diffusion processes and eventually protrude from the alloy surface due to undermining of the surrounding aluminium matrix (Sieradzki 1993; Habazaki, Shimizu et al. 1997). These copper clusters may be released as elemental copper into the oxide layer by being undermined or copper ions may be oxidized directly from the protruding clusters. It has also been demonstrated that the level of copper enrichment is also influenced by grain orientation. In terms of general corrosion performance, the enrichment of copper at the alloy surface is also likely to increase the number of flaws that exist in the

This section covers general approaches to protection of aluminium alloys in view of recent advances in the understanding of alloy microstructure. It includes an overview of pretreatment processes such as anodising, conversion coating and organic coatings (barrier and inhibitor combinations). It will examine recent advances in inhibitor design such as building in multifunctionality and touch on self healing coating systems. Approaches using multifunctionality can target anodic and cathodic reactions more effectively than using

Standard metal finishing processes, which have been used for many years, are likely to continue to be used into the future unless they contain chemicals that are targeted for replacement such as chromium. The function of these coatings is primarily to provide better adhesion properties for paint coatings and a secondary role is to provide corrosion protection. The general approach for applying these coatings relies on metal finishing treatments (treatment prior to painting involving immersion in acidic and alkaline baths)) with the objective of reducing the heterogeneous nature of the metal surface such as removing the NSDL and second phase particles (Muster 2009). This is achieved in multistep treatment processes for metal protection (Twite and Bierwagen 1998; Buchheit 2003; Muster

2. deposition or growth of a manufactured oxide via electrochemical (anodising) or

3. use of an organic coating for specific applications, normally including a primer and a

On aluminium, most anodised coating processes produce an outer oxide with a cellular structure on top of a thin barrier layer that provides some protection against corrosion. Inhibitors can be incorporated into the outer porous layer of the anodized layer during formation or as a seal after formation to offer some extra protection upon damage. Chromic

1. selective deoxidation (IM particle removal and surface etching);

following corrosion processes (Muster 2009).

aluminium oxide.

2009) as for instance:

top-coat.

**4. Corrosion and protection** 

individual monofunctional inhibitors.

chemical (conversion coating) means;

Exfoliation corrosion (Davies 1999; Zhao and Frankel 2007) of aluminium alloys is also frequently due to intergranular corrosion. It generally occurs where the alloy microstructure has been heavily deformed (i.e. by rolling) and the grain structure has been flattened and extended in the direction of working. Intergranular corrosion attack from transverse edges and pits then run along grain boundaries parallel to the alloy surface. Exfoliation is characterised by leafing off of layers of relatively uncorroded metal caused by the swelling of corrosion product in the layers of intergranuar corrosion. Exfoliation is observed on aircraft components, for example, around riveted or bolted components.

Fig. 15. (a) Schematic of hypothetical grain boundary in an Al-Zn-Mg alloy. This schematic indicates the different chemistry that exists in the grain interior, solute depleted zone (precipitate free zone) and grain boundary precipitates - giving rise to electrochemical heterogeneity localised at the grain boundary region. (b) Conventional bright field TEM image of high angle grain boundary in AA7075-T651, revealing grain boundary precipitates (MgZn2) and a distinguishable precipitate free zone
