Corrosion Science 52 (2010) Contents lists available at ScienceDirect. Corrosion Science. journal homepage:

Transcription

1 Corrosion Science 52 (2010) Contents lists available at ScienceDirect Corrosion Science journal homepage: The influence of the microstructure on the atmospheric corrosion behaviour of magnesium alloys AZ91D and AM50 Martin Jönsson *, Dan Persson Swerea KIMAB, Drottning Kristinas väg 48, SE Stockholm, Sweden article info abstract Article history: Received 15 July 2009 Accepted 25 November 2009 Available online 3 December 2009 Keywords: A. Magnesium B. SEM C. Atmospheric corrosion Even though magnesium, as a structure metal, is most commonly used in an atmospheric environment, most investigations of magnesium are performed in solution. In the present work the atmospheric corrosion of two commonly used magnesium alloys, AZ91D and AM50, has been investigated from the initial stages up to the most severe forms of corrosion. A detailed investigation of the morphology of a corrosion attack and its development over time shows that the atmospheric corrosion mechanism is similar for the two alloys. Based on these findings a schematic model of the initial atmospheric corrosion attack on AZ91D is presented and discussed. Ó 2009 Elsevier Ltd. All rights reserved. 1. Introduction The low density of magnesium alloys has made them an interesting material for the automotive and aerospace industries, where weight is of importance. However, magnesium is a very un-noble metal, which makes it prone to corrosion. AZ91D and AM50 are two of the commercially most used magnesium alloys. The AZ91D alloy has an aluminium content of 9%, while that of AM50 is lower, 5%. Under atmospheric corrosion conditions [1], as well as in solution [2,3], the AZ91D magnesium alloy shows better corrosion resistance compared to the AM50 alloy. It is generally agreed that the presence of aluminium is beneficial for the corrosion protection of Mg alloys [2,4 6]. For optimum corrosion protection a lowest limit of aluminium content is required. Lunder et al. [5] reported this quantity to be about 2 4%, and a significant drop in corrosion rate was observed when the aluminium content in the alloy was increased above that level. The increasing corrosion resistance with increasing Al content is in part explained as due to the aluminium-containing microconstituents of the magnesium alloys, such as the presence of b-phase and eutectic a/b-constituent. Song et al. [7] investigated the effects on corrosion behaviour in 1 M NaCl solution of the b- phase (Mg 17 Al 12 ) and suggested that the b-phase works both as a barrier and as a site for the cathodic reaction. If the fraction (f) of the b-phase is low, it serves mainly as a galvanic cathode and accelerates the overall corrosion of the a-matrix. However, if the fraction (f) is high, then the b-phase acts as a barrier for the anodic * Corresponding author. Tel.: ; fax: addresses: martin.jonsson@swerea.se, martin.jonsson@kimab.com (M. Jönsson). reaction, retarding the overall corrosion attack. If the a-grains are fine, the b-phase fraction is nearly continuous. Therefore, it can be expected that f increases during corrosion and finally becomes high enough to make the b-phase act as a corrosion barrier. The corrosion rate should be low after this steady-state surface condition has been reached [7]. Lunder et al. [8] reported that the corrosion rate of the magnesium alloy AZ91D in solution is higher in the artificially aged (T6) condition compared to the solution heat-treated (T4) condition. In the T6 condition the precipitation of b-phase is almost continuous. Lunder et al. proposed that the b-phase here exhibits much better corrosion behaviour by forming a protecting network that prevents further corrosion of the underlying a-phase. However, these investigations were performed in solution at a high ph (10.5), and the ph has a strong influence on the activity of the Mg Al phases [6,8]. Mg is stable in alkaline solutions and exhibits an increasing corrosion rate at a lower ph, while aluminium is passive in neutral solutions but active in alkaline solutions [6,8]. The b-phase is stable in a wide ph range of 4 14 [8]. Mathieu et al. [6], on the other hand, measured the corrosion resistance of an Al-containing a-solid solution at a lower ph of 8.3 and found that the corrosion resistance of an Al-containing a-solid solution strongly depends on the Al content: E corr increases with Al, whereas i corr decreases. It is suggested that a higher corrosion resistance of semi-solid processed (thixomoulded) AZ91 can be explained by the higher aluminium content in the a-phase. The higher aluminium content gives a higher corrosion resistance and also lowers the galvanic potential difference between the two phases. However, the role of the aluminium content in the a-phase is somewhat disputed. Lunder et al. [8] measured current versus time curves for binary alloys containing magnesium and up to 8% aluminium. They showed that a small amount of aluminium content (up to 8%) in the a-phase X/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved. doi: /j.corsci

2 1078 M. Jönsson, D. Persson / Corrosion Science 52 (2010) leads to higher anodic activity. Similar results are reported by Song et al. [7] and Lee et al. [9]. Song et al. [7] made electrochemical measurements on binary Mg Al alloys containing Mg 2%Al and Mg 9%Al. They showed that if the a-grain is only weakly polarised anodically, the dissolution rate of Mg 2%Al is higher compared to Mg 9%Al, but vice versa if the anodic polarisation is strong. Hence, the eutectic a/b area would corrode to a greater extent in solution. The experiments referred to above were all performed in solution and there is a great difference between the corrosion of magnesium in solution and that under atmospheric conditions. As indicated by studies carried out by Tomashov [10], the dominating cathodic reaction during atmospheric corrosion is probably oxygen reduction. Furthermore, in solution the main corrosion product is Mg(OH) 2, which leads to a higher ph in the solution. Under atmospheric condition, on the other hand, the presence of CO 2 in the ambient air dissolves in the thin electrolyte layer and thereby lowers the ph in the electrolyte film. The corrosion product formed on the surface is initially magnesite (MgCO 3 ), and longer exposure results in a localised corrosion attack on the magnesium surface and the formation of shallow pits containing brucite (Mg(OH) 2 ) covered with crusts of corrosion products with hydromagnesite (Mg 5 (CO 3 ) 4 (OH) 2 4H 2 O) [11]. In solution, on the other hand, the main corrosion product is brucite (Mg(OH) 2 ) [12], which means that the corrosion process here differs from that under atmospheric conditions. In the present work the influence of the microstructure on the development of atmospheric corrosion of the magnesium alloys AZ91D and AM50 has been investigated, from the initial stages up to more severe forms of corrosion. The purpose has been to better understand the influence of the microstructure on the atmospheric corrosion properties of the AZ91D and the AM50 alloys. 2. Experimental Corrosion test panels of die-cast AZ91D and AM50 were supplied by Hydro Magnesium ASA Porsgrunn, Norway. The composition of the magnesium alloys is given in Table 1. The samples were machined to obtain small coupons (i.e mm3). The samples had a geometrical area of approximately 6.6 cm 2. The samples were sequentially polished and finally finished with 3 lm diamond paste. The morphology of a corrosion attack was examined, both after the initial stages and after longer exposure times Exposure conditions In the experiments with longer exposure times, AZ91D samples were contaminated with 70 lg/cm 2 NaCl. For the investigation of the initial stages of corrosion the AZ91D and AM50 alloys were exposed to a lower NaCl concentration of 1 lg/cm 2, and the exposure time of these samples was 6 days. The NaCl was applied by a pipette and care was taken to contaminate the whole surface equally. Both types of samples were exposed in 95% relative humidity (RH) in a closed vessel, and to achieve this RH a saturated solution of K 2 SO 4 was used [13]. After exposure both types of samples were pickled several times at ambient temperature in a solution of 200 g/l chromium trioxide (CrO 3 ) and 10 g/l silver nitrate (AgNO 3 ), then rinsed with water and ethanol, and finally dried. For the longer exposure samples the metal loss was calculated by weighing the samples after pickling, using three replicas for each exposure Methods of analysis Two Confocal microscopes were used in this work. The initial corrosion attack was investigated with a Confocal scanning laser microscope of model LEXT OLS3000 from Olympus Lext. In this instrument the light source is a blue laser with a wavelength of 408 nm. For investigation of samples exposed for longer times, with a higher chloride concentration, the Confocal microscope was instead of model lsurf manufactured by Nanofocus. This microscope uses a white xenon light source. The samples with 70 lg/cm 2 NaCl were exposed for 3, 6, 14, 21, 28, and 56 days. The surfaces were then pickled and the corrosion attack was examined using the Confocal microscopy. The measurement values obtained were then processed in MATLAB to obtain statistics on the development of the corrosion pits formed on the surface of the samples. The scanned area was 4.3 cm 2. Care was taken to include all the surface parts of the samples that were corroded. The scanned area included a total of measurement points. Further, an average of the 1000 deepest measurement values was calculated, and the maximum pit depth was also checked using a metallographic microscope SEM investigations Compositional analysis of the two magnesium alloys was performed using SEM model JSM-700F from JEOL, equipped with both Energy Dispersive Spectroscopy (EDS) and Wavelength Dispersive Spectroscopy (WDS). In the analysis performed with the EDS the acceleration voltage was set at 20 kev, and in the WDS measurement the acceleration voltage was set at 10 kev. The analysing software program connected to the SEM was of the model INCA supplied by Oxford Instruments. Both the EDS and the WDS capability of the SEM was utilized in these experiments. The WDS, which has a higher accuracy compared to the EDS, was used to identify the different phases that precipitate in the two alloys AZ91D and AM50. Using the EDS capability, on the other hand, gave fast measurements and the opportunity to do line scan, i.e. element composition measurements along a predetermined line. The EDS measurements show somewhat higher values (1 2 wt%) for aluminium, due to the overlap of Mg and Al peaks. Therefore, the result of EDS line analysis should not be viewed as an absolute value but as a relative value to be used when measuring differences within the microstructure in order to compare these with the morphology of the corrosion attack. It should also be noted that the SEM was used both in the backscattered mode (Figs. 1 and 2) and in the secondary electron mode (SEI) (Figs. 3, 4 and 7). 3. Results 3.1. Investigation of the microstructure of magnesium alloys AZ91D and AM50 In Figs. 1 and 2, a backscattered SEM image of the microstructure of AZ91D and AM50 can be seen. Further, the element compo- Table 1 Nominal composition of magnesium alloys AZ91D and AM50 (wt%). Al Zn Mn Si Ni Cu Fe Be (ppm) Mg AZ91D Remainder AM Remainder

3 M. Jönsson, D. Persson / Corrosion Science 52 (2010) Fig. 1. SEM backscattered image of the microstructure of the AZ91D magnesium alloy, the microstructure typically consisting of a matrix of primarily a-phase grains surrounded by eutectic a/b-constituent and b-phase (Mg 17 Al 12 ). The a/b-constituent and b-phase can be seen as the light gray and white areas, respectively. The composition of points 1 3 is given in Table 2. AZ91D, the b-phase (Mg 17 Al 12 ) is rarely observed [16]. But a number of white areas which, in the same way as in the case of AZ91D, correspond to the b-phase can be seen in Fig. 2 (marked (3)). The fraction of the b-phase is much lower in AM50 than in AZ91D. The primary a-phase grains in AM50 (marked (2) in Fig. 2) have an aluminium content of 2 wt%, which is also lower than that of AZ91D. In both of the alloys Al-Mn particles can be found (marked (3) in Figs. 1 and 2). These particles are often of the Al 8 Mn 5 type [16,17]. The influence of these particles on the atmospheric corrosion behaviour is further discussed elsewhere [18]. EDS line measurements show that in both AM50 and AZ91D a eutectic area can be found between the a-phase grains and the b-phase. The dendrite cores in both AZ91D and AM50 solidify first and contain a low concentration of Al [16,19]. The concentration increases, however, towards the dendrite periphery, and interdendritic areas become supersaturated with aluminium (and other solutes), forming eutectic a/b-constituent and b-phase. In Fig. 3, the aluminium content of a grain in the AZ91D and AM50 alloys is shown. The grain size has influence on the aluminium content, so care has been taken to investigate grains of similar sizes in the two alloys. Not surprisingly, the aluminium content of the grains is higher in AZ91D than in AM50. The diagrams in Fig. 3 show that the aluminium content of a larger grain (L1 and L2) is around 5% in AZ91D and around 3 4% in AM50, i.e. 2% lower. In a smaller grain (L3 and L4) the difference between the aluminium contents is higher. A smaller AZ91D grain (L3) has an aluminium content of around 8% in its middle, compared to 5 6% (L4) in the case of a similar AM50 grain. Towards the grain boundaries the aluminium content increases in both alloys Comparison of the initial corrosion behaviour of AZ91D and AM50 Fig. 2. SEM backscattered image of the microstructure of the AM50 magnesium alloy, a matrix of primarily a-phase grains surrounded by eutectic a/b-constituent and b-phase (Mg 17 Al 12 ). The a/b-constituent and b-phase can be seen as the light gray and white areas, respectively. The composition of points 1 3 is given in Table 3. sition, measured with WDS, of points 1 3 can be found in Tables 2 and 3. The microstructure of AZ91D typically consists of a matrix of primary a-phase grains (marked (2) in Fig. 1), the grains having an aluminium content of around 4 wt% in their middle. In the grain boundaries the b-phase (Mg 17 Al 12 ) can be found as white areas (marked (3) in Fig. 1). Even though the element composition of the b-phase (Mg 17 Al 12 ) in the WDS analysis (30 wt% Al) is somewhat lower than the aluminium content value given in the binary phase diagram for Al Mg (42 58 wt% Al) [14], these white areas are most probably b-phase. Between the b-phase and the a-phase a eutectic area of b-phase and a-phase can be found. The aluminium content in this phase varies and is discussed further down. The microstructure and the local nobility of the different intermetallic phases of AZ91D were more thoroughly investigated in an earlier study by the authors [15]. The microstructure of AM50 (Fig. 2) is similar to that of AZ91D. However, since the aluminium content is lower in AM50 than in In Figs. 4 and 7, SEM images of the corroded surfaces of AZ91D and AM50 can be seen. Both samples were exposed for 6 days to humid air (95% RH) and were contaminated by 1 lg/cm 2 NaCl. After that the corrosion products were removed by pickling. In Figs. 5 and 8, the corresponding confocal images show the morphology of the corrosion attack. A line scan over the surface shows the depth profile measured with the Confocal microscope as well as the aluminium content over the same line measured with the EDS capability of the SEM. In both AZ91D and AM50 trenches have formed in the a-phase in the vicinity of the eutectic a/b-constituent. The trenches in AM50 have a depth of around 1 lm and a width of around 2 3 lm. Similar trenches can be seen in AZ91D and, as shown in Figs. 6 and 9, the depth profile follows the aluminium content well in both alloys. In the case of AZ91D the trenches have formed in the a-phase, where the aluminium content is 5 6%. Just outside the trenches the aluminium content is over 10% and the area is not equally affected by corrosion. In AM50 the trenches are deeper than in AZ91D. An interesting feature can be observed here. The aluminium content in the trenches is 1 2% lower than that in the middle of the grain. Actually, the aluminium content should be constant within the grain, as observed before by EDS. Thus, the aluminium in the a-phase has undergone selective dissolution at the a-phase/eutectic a/b interfaces in these areas. Further, no corrosion attack can be seen in the vicinity of the intermetallic Al Mn particles either in AZ91D or in AM50. This has been explained as due to the position of the Al Mn particles within the microstructure [18]. Due to the solidification process the majority of the Al Mn particles can be found in the areas of a eutectic a/b-constituent. Under thin electrolyte layers in atmospheric environments galvanic coupling is only possible between the eutectic a/b-constituent and the intermetallic Al Mn particles, and due to the higher aluminium content in the eutectic a/b-constituent the driving force is not strong enough for the initiation of a corrosion attack at these locations.

4 1080 M. Jönsson, D. Persson / Corrosion Science 52 (2010) Fig. 3. (A) SEM secondary electron image (SEI) of the microstructure of the AZ91D magnesium alloy. (B) SEM backscattered image of the microstructure of the AM50 magnesium alloy. The aluminium content measured with EDS of lines 1 4 is given in the diagrams. Table 2 The element composition of points 1 3 in Fig. 1 measured with the WDS (wt%). Point Mg Al Mn Fe Zn 1 a a Mg varies in the analysis of the Al8Mn5 particles due to the surrounding magnesium. Table 3 The element composition of points 1 3 in Fig. 2 measured with the WDS (wt%). Point Mg Al Mn Fe 1 a a Mg varies in the analysis of the Al8Mn5 particles due to the surrounding magnesium. Fig. 4. SEM image of an AZ91D sample exposed for 6 days in 95% RH with 1 lg/cm 2 NaCl. The corrosion products have been removed through pickling The influence of the microstructure after longer exposure AZ91D samples exposed for times varying from 3 to 56 days were also investigated in the Confocal microscope. In Fig. 10, a typical corrosion attack on an AZ91D magnesium sample can be studied. The grains have corroded away leaving a structure consisting of the phases with a higher aluminium content, i.e. the eutectic a-phase and the b-phase. After the initial stages the corrosion is more severe, as can be seen in Fig. 11. The diagram shows the profile line of two pits with a depth of 20.0 and 15.3 lm, respectively. A higher magnification can be seen to the right in Fig. 11. The depth profile line of the corrosion attack shows a depth of 24.9 lm. Further, it seems that the eutectic a-phase/b-phase is unaffected by the corrosion. The measurement values obtained were processed to obtain an overview of the morphology of the corrosion attack. Fig. 12 shows the area of the sample covered with pits deeper than 5 and 10 lm, respectively. From the diagram it can be inferred that the area attacked by corrosion increases more or less linearly with

5 M. Jönsson, D. Persson / Corrosion Science 52 (2010) Fig. 5. Confocal image of the same area as seen in Fig. 4. Fig. 7. SEM image of an AM50 sample exposed for 6 days in 95% RH with 1 lg/cm 2 NaCl. The corrosion products have been removed through pickling. the time of exposure. After 3 days of exposure 1% of the surface is covered with pits with a depth of over 5 lm. The fraction of the total area attacked by corrosion increases with the time of exposure. After 56 days more than 7% of the surface is covered with pits with a depth of over 5 lm. Visual inspection of the sample surfaces shows that the corrosion attack is localised to a few clusters that grow laterally over the surface. In the same way the area covered with pits deeper than 10 lm increases with the time of exposure, from around 0.5% up to 3% after 56 days. Further, from the measurement an average of the 1000 deepest values of the corrosion attack after each exposure time was recorded. The maximum depth of the corrosion and the mass loss of the samples are given in Fig. 13. The corrosion of the samples was essentially proportional to the exposure time. The pit depth increases very rapidly during the first few days of exposure. After 3 days the corrosion attack on the surface is already in the order of 70 lm. After the initial very rapid growth of the corrosion attack, the corrosion depth increases at a slower rate and the deepest pits are in the order of 130 lm after 56 days of exposure. 4. Discussion In the previous work we have shown that the corrosion attack starts in the larger grains where the aluminium content is low [18]. Our present work shows that the corrosion, as indicated by the depth profile line in Fig. 6, follows the aluminium content very well at the beginning of the corrosion attack. Areas with a lower aluminium content corrode faster. Thus, under atmospheric condition, it is beneficial for the corrosion resistance to increase the aluminium content in the a-grains. This is contrary to the results of Lunder et al. [8] and Lee et al. [9], but in accordance with the results of Mathieu [6] discussed above. As seen in both Figs. 6 and Fig. 6. Depth profile and aluminium content of part of the surface in Figs. 4 and 5, as indicated by the lines in the Confocal and SEM images seen above.

6 1082 M. Jönsson, D. Persson / Corrosion Science 52 (2010) Fig. 8. Confocal image of the same area as seen in Fig. 7. Fig. 10. Confocal image of the corrosion attack on AZ91D magnesium exposed for 14 days in 95% RH contaminated with 70 lg/cm 2 NaCl. 9, areas with a high aluminium content tend to form a bi-metallic couple with areas low in aluminium. It was shown in previous work [18] that the Volta potential increased rapidly with aluminium contents up to 10 12% Al. Areas with a low aluminium content work as the anode in the formation of trenches, while areas with a higher aluminium content, >10%, can act as cathodes. In the case of AZ91D the corrosion process initiates in areas where the aluminium content is lower than around 6%. In AM50 the distinction between anodic and cathodic areas is not so pronounced. The corrosion appears to be more of a general type [20], but it can be seen from Fig. 9 that areas with an aluminium content lower than around 5% work as the anode in the initial stages. As the corrosion rate is dependent on the aluminium content in the different phases, the phases found in AM50 are more prone to corrosion compared to those of AZ91D due to the lower aluminium content both in the grain boundaries and in the grains. Further, in AM50 the lower aluminium content in both small and large grains makes these work as efficient anodes when coupled to the eutectic a/bconstituent with a higher aluminium content. Since the measurement has been performed in atmospheric conditions, the electrolyte layer is thin. This results in a large ohmic drop and a short separation between the anode and cathode on the surface. Hence the bi-metallic coupling is restricted to a small distance. The cathodic reaction on magnesium in atmosphere is thought to be due both to water reduction and to oxygen reduction [10]. However, the importance of the oxygen reduction increases with decreasing thickness of the electrolyte layer due to a small diffusion path of oxygen. Cathodic reaction : 2H 2 O þ 2e! 2H 2 ðgþþ2oh ðaqþ O 2 þ 2H 2 O þ 4e! 4OH ðaqþ Anodic reaction : Mg ðsþ!mg 2þ ðaqþþ2e ð1þ ð2þ ð3þ Fig. 9. Depth profile and aluminium content of part of the surface in Figs. 7 and 8, as indicated by the lines in the Confocal and SEM images seen above.

7 M. Jönsson, D. Persson / Corrosion Science 52 (2010) Fig. 11. Left: Severe corrosion attack on AZ91D magnesium exposed for 28 days in 95% RH contaminated with 70 lg/cm 2 NaCl. Right: A higher magnification of the white dotted square. The diagram shows the profile line of the corrosion attack corresponding to the white dotted lines. Fig. 13. The mass loss of AZ91D magnesium alloy versus time. The secondary y- axis: the average of the maximum pit formed on the AZ91D surface after different exposure times. Fig. 12. The fraction of the area covered by corrosion pits deeper than 5 and 10 lm, respectively, formed after different exposure times and measured with the Confocal microscopy. The dissolution of aluminium from the microstructural components is described by the following reactions in neutral and alkaline ph [21]: Al þ 3OH! AlðOHÞ 3 ðsþþ3e ð4þ AlðOHÞ 3 ðsþþoh ðaqþ!alðohþ 4 ðaqþ ð5þ The cathodic reaction produces OH (aq), Eqs. (1) and (2), which will lead to a local increase in the ph of the thin electrolyte in the vicinity of the cathodic sites. In these alkaline areas the aluminium suffers anodic dissolution, due to the formation of soluble species at higher ph, Eq. (5), and the a-phase grain can undergo selective dissolution of aluminium. Selective dissolution of aluminium in the Mg Al alloys in alkaline solutions has also been reported by other authors [8,9]. Fig. 14A C shows a schematic image of the atmospheric corrosion of AZ91D. The schematic image is based on the results of this work: A. During the initial steps of the corrosion attack, the large grains are attacked (Fig. 14A) due to the lower aluminium content in these grains. A microgalvanic element is formed between the a-phase and the eutectic a/b-constituent, and

8 1084 M. Jönsson, D. Persson / Corrosion Science 52 (2010) neutral solutions, but it has an increased corrosion rate in alkaline solutions [6,8]. Hence the aluminium-rich phases, i.e. the eutectic a/b-constituent and the b-phase, are to a larger extent affected by corrosion. Even more severe corrosion can be seen when these phases are corroded away. In our view the phases with a higher aluminium content, i.e. b- phase and eutectic a/b-constituent, do not form a protective barrier, as discussed in the literature [7,8]. After the initial rapid corrosion of the a-phase the corrosion rate exhibits a change and a slower corrosion process occurs with dissolution of phases with a higher aluminium content. 5. Conclusions Fig. 14. (A) Schematic image of the corrosion attack. In the initial stages of the corrosion attack large grains are attacked. A microgalvanic element is formed between the a-phase and the eutectic a/b-constituent. (B) After a few days the whole grain has corroded away. This results in the deep pits. The corrosion attack also starts to affect smaller grains in the vicinity, and a honeycomb structure is formed on the surface. (C) After longer exposure periods more severe corrosion occurs. The corrosion spreads in a hemispherical way, i.e. laterally and spatially, the result being a general type of corrosion attack. At this stage the eutectic a/bconstituent and the b-phase are also affected by the attack. trenches are formed in the a-phase, where the aluminium content is low, cf. Fig. 6. At this stage the main corrosion product is magnesite [11]. B. After a few days the whole grain is corroded away (Fig. 14B). After longer exposure times the corrosion also starts to affect grains in the vicinity, grains that have a smaller size and a high aluminium content and hence higher corrosion resistance. For AM50 the corrosion behaviour is similar, with anodic areas in the a-phase and the cathodic reactions taking place in the eutectic a/b-constituent. A structure consisting of the eutectic a/b-constituent and b-phase in AZ91D and AM50 is formed on the surface, similar to the case seen in Fig. 10. C. Even though the aluminium-containing phases decrease the overall corrosion rate due to their better corrosion resistance in near-neutral ph [6], these phases are not inert to a corrosion attack. Thus, after the initial stages the aluminium-containing phases are affected by corrosion and more severe corrosion can be observed when these phases are corroded away (Fig. 14C). Further, earlier investigations of the corrosion products [11] show that brucite Mg(OH) 2 is not transformed into magnesite MgCO 3 in the pits, which would be the case in ambient atmosphere. This shows that there is a low CO 2 content, indicating that the ph is higher in the pits compared to the surface. The higher ph is unfavourable for the aluminium-containing phases. Aluminium is passive in The atmospheric corrosion process is strongly related to the aluminium content in the different phases of the alloys. Phases low in aluminium are attacked to a greater extent, and phases with a higher aluminium content display higher corrosion resistance in atmospheric environments. Thus, the AM50 alloy, which has a lower aluminium content both in the grains and in the grain boundaries compared to the AZ91D alloy, will have a higher corrosion rate. A model of the atmospheric corrosion process of AZ91D is presented, where microgalvanic elements are formed with the a- phase as anode and where the eutectic a/b-constituent in the vicinity of the a-phase is the site of the cathodic reactions. Thus, the initial corrosion attack takes place in the a-phase, and while the dissolution of the a-phase progresses, pits are formed. A honeycomb-like corrosion pattern is formed on the surface with the remaining eutectic a/b-constituent and b-phase. After the initial rapid corrosion of the a-phase the corrosion slows down with dissolution of phases with a higher aluminium content, resulting in a corrosion attack of a more general nature. For the AM50 alloy the corrosion process is similar, with an initial attack on the a-phase in the grains and the cathodic reaction occurring in the aluminium-rich eutectic a/b-constituent. After longer exposure times aluminium-rich phases are attacked, and the corrosion develops laterally on the surface as well as in depth. Acknowledgments The authors thank Jan-Ingvar Skar at Norsk Hydro for supplying the materials and Jan Y. Jonsson and Fredrik Falkenberg at Outokumpu, Avesta, for all their help with the Confocal microscope. References [1] N. LeBozec, M. Jönsson, D. Thierry, Corrosion 60 (2004) 356. [2] R. Lindström, On the chemistry of atmospheric corrosion, a laboratory study on Zn and Mg/Mg alloys, Doctoral thesis, Department of Chemistry, Chalmers, Göteborg, [3] R.S. Hansen, in: Magnesium Alloys and their Applications: Symposium, 1992, p [4] G. Song, A. Atrens, Advanced Engineering Materials 5 (2003) 837. [5] O. Lunder, J.H. Nordlien, K. Nisangliou, Corrosion Reviews 15 (1997) 439. [6] S. Mathieu, C. Rapin, J. Steinmetz, P. Steinmetz, Corrosion Science 45 (2003) [7] G. Song, A. Atrens, M. Dargusch, Corrosion Science 41 (1999) 249. [8] O. Lunder, J.E. Lein, T.K. Aune, K. Nisancioglu, Corrosion 45 (1989) 741. [9] C.D. Lee, C.S. Kang, K.S. Shin, Metals and Materials 6 (2000) 441. [10] N.D. Tomashov, Theory and Protection of Metals: The Science of Corrosion, The Macmillan Company, New York, [11] M. Jönsson, D. Persson, D. Thierry, Corrosion Science 49 (2007) [12] R. Lindström, L.G. Johansson, J.E. Svensson, Materials and Corrosion 54 (2003) 587. [13] J.F. Young, Journal of Applied Chemistry 17 (1967) 241. [14] H. Baker, ASM Handbook, ASM International, Ohio, [15] M. Jönsson, D. Thierry, N. LeBozec, Corrosion Science 48 (2006) [16] V.Y. Gertsman, J. Li, S. Xu, J.P. Thomson, M. Sahoo, Metallurgical and Materials Transactions A: Physical Metallurgy and Materials Science 36 (2005) [17] E.M. Gutman, A. Eliezer, Y. Unigovski, E. Abramov, Materials Science and Engineering A 302 (2001) 63.

9 M. Jönsson, D. Persson / Corrosion Science 52 (2010) [18] M. Jönsson, D. Persson, R. Gubner, Journal of the Electrochemical Society 154 (2007) C684. [19] C.J. Simensen, B.C. Oberlander, J. Svalestuen, A. Thorvaldsen, Zeitschrift fûer Metallkunde/Materials Research and Advanced Techniques 79 (1988) 696. [20] R. Lindström, J.E. Svensson, L.G. Johansson, Journal of the Electrochemical Society 149 (2002) B103. [21] D.B. Blücher, J.E. Svensson, L.G. Johansson, M. Rohwerder, M. Stratmann, Journal of the Electrochemical Society 151 (2004) B621.