SITE-SPECIFIC MICROSTRUCTURE CHARACTERIZATION OF THE SURFACE LAYER IN A SAND-CAST Mg-Al ALLOY (AM50) CORRODED IN AN AQUEOUS

NACL SOLUTION  Mohsen Danaie1, Robert Matthew Asmussen2, Pellumb Jakupi2, David W. Shoesmith2, Gianluigi

A. Botton1

1 Department of Materials Science and Engineering, Brockhouse Institute for Materials Research and Canadian Centre for Electron Microscopy, McMaster University, Hamilton, Ontario, Canada 2 Department of Chemistry and Surface Science Western, Western University, London, Ontario, Canada Keywords: Magnesium Alloy, Sand Cast, Corrosion, Electron Microscopy

Abstract Site-specific analytical electron microscopy characterization was performed on representative locations within the microstructure of a corroded sand-cast AM50 alloy. Areas close to the eutectic microstructure were the regions with the least amount of corrosion damage. The corrosion layer in these areas consisted of a top columnar section, with a mixture of amorphous MgO and Mg(OH)2. Underneath this top corrosion film, an Al-rich layer with metallic character was identified. The corrosion film on the primary α-Mg grains showed a bi-layered morphology: a thin columnar film, similar to the superficial film in the eutectic region, and a much thicker and porous under-layer of MgO/Mg(OH)2.

1. Introduction The alloy AM50 (Mg-5 wt. %Al-0.3 wt.% Mn) is one of the magnesium-based AM-series alloys. This class of alloys was developed for the applications that require higher toughness and ductility, while maintaining the lightweight of the final components. Their development aimed to address the need for the applications for which the workhorse magnesium alloy, i.e. AZ91 (Mg-9 wt.% Al-1 wt.% Zn), is not suitable. With good castability and superior mechanical properties, AZ91 is the most widely used magnesium alloy [1]. With only 2-3% elongation, however, this alloy fails to perform in the applications that seek high impact strength and ductility. Alloys in the AM series were introduced in this context. With 10-15% elongation, AM-series alloys can be utilized in the automotive parts that need high absorption of impact energy e.g. seat frames, steering wheels, panels, brackets and fans [2]. The major drawback in the AM-group alloys is their poor corrosion resistant properties. The lower aluminum content in these alloys results in weaker corrosion resistance when compared to AZ91 [2]. The addition of manganese in these alloys is aimed at enhancing the corrosion resistance of these alloys by reducing the solubility of iron [3]. Addition of Mn also results in the formation of various intermetallics in the Al-Mn system, e.g. MnAl, MnAl6 or MnAl4. This helps both to withhold aluminum from forming the β-Mg17Al12 phase- detrimental from the ductility point of view-, and also to increase the overall strength via precipitation hardening. The mechanism of corrosion in AM-series alloys, especially in NaCl aqueous solution, is a matter of debate. It is acknowledged, however, that microgalvanic cells created by the presence

of various secondary phases in these alloys play a crucial role in determining the overall corrosion properties of these systems (See [4] and citations therein). Due to strong reactivity of magnesium, usually the secondary phases are cathodic towards the α-Mg matrix. This results in the preferential corrosion of the matrix. The extent of this behavior varies across different phases present in this system. For instance, it has been suggested that Al-Mn intermetallics with higher Mn content (AlMn and Al8Mn5) have higher cathodic current density and hence are more detrimental, causing preferential corrosion in the magnesium phase [5]. The role of β-Mg17Al12 phase is more ambiguous [4], since this phase forms a protective passive layer that impedes further corrosion. Microscopic characterization of the corrosion film on Mg alloys can lead to the identification of the weakest microstructural links and information on the active corrosion mechanisms. In pioneering studies by Nordlien, et al. [6, 7, 8], transmission electron microscopy (TEM) was used to examine the morphology of the oxide/hydroxide layer on both pure and alloyed Mg. The corrosion product on Mg-Al alloys exhibited a three-layer structure [8], with the layer closest to the alloy having a higher Al content, with respect to the bulk metal. X-ray photoelectron spectroscopy (XPS) demonstrated the accumulation of Al3+ species on the corroded surface of Mg-Al alloys [9, 10, 11]. This was interpreted as percolation of amorphous Al2O3 within the MgO/Mg(OH)2 corrosion layer [8], resulting in the observed enhanced corrosion resistance. It has also been claimed that Al-enrichment can be an artifact caused by sputter cleaning of the initial alloy surface, due to the higher sputtering rate of Mg [12]. In this study, our objective is to characterize the corrosion film formed on the Mg-Al alloy (AM50) at locations adjacent to the primary α-Mg grains and close to the eutectic microstructure in the alloy. Previous studies on the corrosion layer on Mg alloys have mainly focused on broad-beam techniques, such as XPS, that yield an averaged signal with low spatial resolution. Other studies employing electron microscopy and X-ray energy dispersive spectroscopy (XEDS) were performed predominantly on ultra-microtomed specimens, which do not allow selection of specific areas present in the alloy [6, 7, 8]. Using a focused ion beam for TEM sample preparation, our goal is to characterize the corrosion product layer in the vicinity of the main microstructural features in the alloy AM50. This allows us to examine the potentially corrosion resistant properties of the various constituents within the microstructure. The present contribution only includes a brief description of our work. A more thorough presentation of this research can be found elsewhere [13].

2. Experimental procedure  Sand-cast AM50 alloy samples were provided by General Motors in the form of a cylindrical ingot. The chemical composition is presented in Table 1. The details regarding sample surface preparation has been described elsewhere [13]. The microstructure of the as-cast alloy was studied at low magnification using a light microscope (Zeiss Axioplan 2 Imaging microscope, with a Hal 100 halogen lamp) and at higher magnification with a dual-beam scanning electron microscope / focused ion beam (SEM/FIB) (Zeiss NVision 40) equipped with an x-ray energy-dispersive spectrometer (XEDS - Oxford Inca, Silicon drift detector).

Table 1- Composition of the AM50 magnesium alloy used for this study

Corrosion studies were performed on polished specimens immersed in an aerated 1.6 wt.% NaCl (reagent grade, 99% assay) solution at room temperature. The sample was periodically removed from the solution, rinsed and sonicated in anhydrous ethanol and dried. Surface topography was analyzed using confocal laser scanning microscopy (CLSM) and XEDS. The CLSM results have been published elsewhere [14]. After 96 hr of total immersion, the alloy was analyzed in more detail. Three locations representative of the general microstructure were chosen for TEM analyses. Caution was taken not to alter the surface film structure during FIB preparation. For each region of interest, an ~1 µm-thick tungsten protective film was deposited, initially through electron-beam assisted deposition (known not to damage the surface) and subsequently through ion-beam assisted deposition. Thinning of the FIB lamella was performed in steps with an ion beam reduced in energy from 30 kV to 5 kV. Final polishing was performed at 1 kV accelerating voltage. The FIB samples were stored in a vacuum desiccator immediately after preparation until TEM analyses were performed. Microscopic characterization was performed on an FEI Titan 80-300 (scanning) transmission electron microscope, (S)TEM, equipped with an X-ray energy-dispersive spectrometer (XEDS- Oxford Inca, Si(Li) detector). The TEM accelerating voltage was set at 300 kV. Electron beam damage was minimized by conducting analyses on a cryogenically cooled sample holder (T = 95 K). Two complementary techniques were applied: (a) selected area diffraction (SAD) in parallel beam TEM mode along with bright-field (BF) imaging, and (b) analytical microscopy with elemental XEDS analysis in STEM mode. The first technique provided the local crystal structure of the phases present, and was performed first to monitor any artifacts caused by electron beam irradiation. Subsequently, in stage (b), STEM/XEDS maps were used to confirm the existence of various atomic species. A high-angle annular dark field (HAADF) detector was used in the STEM mode to form images from electrons scattered at high angles. The contrast observed in the HAADF images is mainly due to variation in the atomic number (Z), elements with a higher Z yielding a higher intensity. Simulations of electron diffraction patterns were carried out with the JEMS software. For XEDS data analysis, the commercial software INCA (Oxford Instruments) was used.

3. Results 3.1. Microstructure of the As-Cast State The representative microstructure of the sand-cast AM50 alloy in the as-polished state at low magnification is presented in Figure 1. The mild etch appearing in the light microscopy images is caused by the colloidal silica polishing. Both images in Figure 1 are acquired using the

differential interference contrast (DIC) technique [15]. At low magnification, Figure 1(A), a network of the secondary phases present in the microstructure of this alloy can be identified. At higher magnifications, Figure 1(B), the characteristic morphology of partially divorced eutectic (marked with the white arrow) and also the Al-Mn precipitates (an example marked with the red arrow) can be clearly seen.

Figure 1- Light microscopy images of the polished as-cast AM50 alloy (differential interference contrast imaging technique): (A) Low magnification image showing the general microstructure.

In (B) the β-phase in the partially divorced eutectic microstructure (white arrow) and also Al-Mn secondary phases (red arrow) can be observed.

The scanning electron microscopy (SEM) micrographs (Figure 2) provide a more detailed observation of the phases above. The secondary phases and precipitates are clearly visible in the backscattered SEM image, Figure 2(A). A partially divorced eutectic microstructure is present, with large eutectic β-Mg17Al12 phase and islands of the residue eutectic Mg scattered inside, Figure 2(B). This is shown in the XEDS maps of a similar area in Figure 3. A lamellar morphology is also occasionally present in the proximity of the partially divorced eutectic phase. The lamellar structure, neighboring the partially divorced eutectic, is indicative of a discontinuous precipitation reaction to form Mg and β-Mg17Al12 from eutectic Mg grains rich in Al, after completion of solidification [16]. This type of precipitation and morphology are observed at slow cooling rates, as is the case with the sand cast alloy.

Figure 2- Morphology of various phases present in the sand-cast AM50 alloy: (A) low

magnification SEM micrograph acquired using the backscattered electron detector. (B) A secondary electron image showing the partially divorced eutectic microstructure.

Beside the β-Mg17Al12 structures, there are precipitates with polygonal morphologies. The red arrow in Figure 1(B) indicates one of these precipitates. The major constituents of these precipitates are Al and Mn, with minor contributions of Si. Since there is a variety of

intermetallic phases in the Al-Mn system [17], we refer to the precipitates with similar morphology as Al-Mn precipitates. Area shown in Figure 4(A) contains two of these precipitates. Sections (B) through (E) of this figure present the XEDS elemental maps of Mg, Al, Mn, and Si from two selected regions.

Figure 3- A region of partially divorced eutectic microstructure (A) as observed in SEM. XEDS

elemental maps of a selected area in (A) are presented in (B) for Mg and (C) for Al.

Figure 4- (A) Area containing two Al-Mn precipitates (SEM micrograph). (B) through (E) XEDS

elemental maps of Mg, Al, Mn, and Si of the two selected regions of area shown in (A).

3.2. Characterization of the Corroded State Figure 5(A) presents a SEM micrograph of the corroded state of the AM50 alloy after 96 hours of immersion in 1.6 wt.% NaCl solution. The main features of the microstructure can still be identified based on the results in section 3.1 (For a more detailed description the reader is referred to [13, 14]). The α-Mg region exhibits significant corrosion damage and is covered with superficially cracked corrosion product. The Al-Mn precipitate, area marked as #3 in Figure 5(A), is surrounded by an apparently ruptured spherical section of corrosion products. Two areas, 1 and 2 in Figure 5(A), each containing a representative microstructural feature were chosen for TEM analyses. Area 1 is in the vicinity of the partially divorced eutectic and area 2 is within a primary α-Mg grain. Area 3 is chosen close to an Al-Mn precipitates. The evolution of the Al-Mn intermetallics during corrosion will be discussed elsewhere [18]. Figures 5(B) through (C) present the location where the FIB lamella was extracted for the further TEM characterization. The bright region in 5(B), spanning across the eutectic microstructure, is the protective tungsten layer deposited prior to the in-situ lift-out procedure. This protective layer, marking the location of sample from Area 2 can be observed in Figure 5(C). Here, due to a thicker corrosion product layer, some charging effects can also be detected. The micrograph in Figure 5(D) is showing the side of the milled region from area 3, neighboring an Al-Mn precipitate [18].  

Figure 5- (A) SEM micrograph showing the area of interest after 96 hours of immersion in 1.6

wt.% NaCl solution in water. The encircled areas mark the locations chosen for FIB sample preparation. (B) sample location for Area 1, close to the eutectic, (C) Sample location for Area 2,

on a primary α-Mg grain. (D) Sample from an area containing an Al-Mn precipitate. 3.2.1. Area 1- Close to a Partially Divorced Eutectic Microstructure Figure 6(A) shows the FIB sample cut from area 1, with the two windows that were thinned to electron transparency. This configuration of FIB sample allows us to both study larger areas of

the microstructure and also to provide a better mechanical stability of the prepared sample. A higher magnification view of the right side window is shown in Figure 6(B). The β-Mg17Al12 phase appears with higher intensity in the HAADF image due to the higher average atomic number of this phase compared to the Mg matrix.

Figure 6- (A) The FIB sample from Area 1, containing the partially divorced eutectic. The image

is an HAADF-STEM micrograph. The two thinned windows can be clearly observed. (B) STEM-HAADF image showing the eutectic at higher magnification. (C) Closer view of the top

columnar corrosion product layer. The columnar morphology of the outer layer of the film is clearly observed at higher magnification in Figure 6C and appears as a bright band in the HAADF image due to the deposition of W as a protective layer filling the porous structure during sample preparation. Figure 7(A) shows a bright field TEM micrograph of the columnar corrosion product layer chosen for SAD. The diffraction pattern, Figure 7(B), demonstrates that the only crystalline phase, beside the Mg base metal, is MgO. The dark-field imaging (not shown here) demonstrates that the MgO has an extremely fine grain size with the majority of grains located close to the base of the corrosion layer. This is consistent with recently published results on corroded commercially pure Mg [19].

Figure 7- (A) Bright field TEM image of the FIB section from area 1. (B) SAD from the same

area, along with the ring pattern simulation of MgO.

Figure 8- (A) STEM-HAADF image of the area 1 FIB sample. The XEDS elemental maps were acquired from the box shown in the micrograph in (A). XEDS maps of: (B) O, (C) C, (D) Mg,

and (E) Al. Figure 8 presents XEDS elemental maps acquired from the sample from area 1 (chosen from the left window of the FIB sample). The oxygen map, Figure 8(B), highlights the location of the surface film with respect to the substrate alloy. An enhanced Al signal is observed for a thin layer at the surface, Figure 8(E). The apparent slight enhancement of the Al signal on the outside of the oxide is an artifact arising from the increased background due to the presence of W in the protective layer. The presence of this Al-rich layer was also confirmed by a similar analysis within the right window and close to the β-phase region (data not shown here). The carbon can be caused either by contamination or presence of a carbonate phase. 3.2.2. Area 2- Primary Magnesium Grain Figures 9(A) and 9(B) present low and high magnification STEM-HAADF images of the FIB section cut from an α-Mg grain; area 2 in Figure 5(A). Figure 8(B) shows the columnar morphology of the oxide/hydroxide layer, which is similar to that present on the section cut from area 1 (Figures 7 and 8). From the XEDS maps of this area, Figure 10, no indication of an Al-rich layer at the alloy interface can be observed, Figure 10(B). A signal for Cl, uniformly distributed throughout the oxide is also observed. No Cl signal was observed for area 1. The oxygen signal is uniform across the corrosion layer.

Figure 9- (A) STEM-HAADF image of the FIB sample from area 2, on the primary α-Mg grain.

(B) Higher magnification view of the porous top corrosion product layer. The top columnar region of the corrosion layer is also visible.

To probe the crystal structure of the thick corrosion film present at this location, we acquired SAD patterns at 6 locations across this layer. Figure 11(A) shows the area analyzed, the initial beam location (1), and the direction over which the patterns shown in Figure 11(A) were recorded. Figure 11(B) shows the evolution in the SAD patterns across the corrosion layer. SAD #1 corresponds to the top columnar film of corrosion products. As the electron beam probes areas farther from the top surface, the diffraction ring pattern of MgO is always present with the gradual addition of diffraction spots from the Mg substrate. In SAD #6 the diffraction spot pattern is consistent with the 0001 zone axis of Mg (note the simulation). Farther away from the Mg substrate, the diffraction rings are consistent with the presence of only MgO (SAD #2 with a simulated MgO ring pattern).

Figure 10- (A) STEM-HAADF image of the FIB sample from area 2, primary α-Mg grain. The region marked in (A) is used for XEDS elemental analysis. XEDS elemental maps are shown in:

(B) Al, (C) Mg, (D) O, and (E) Cl.

Figure 11- (A) TEM-BF micrograph of area 2 sample. Region marked as #1 shows the selected

location of SAD #1. Consecutive SAD patterns were acquired from #1 to the base metal (#6). (B) SAD patterns and their evolution moving from the outer columnar layer to the base metal. The

ring pattern is consistently MgO (simulation shown on SAD #2).

4. Discussion  4.1. The Al-Rich Layer Observed in Area 1 (Close to the Eutectic) The experimental results presented in Figure 8, from the FIB sample prepared from an area close to the partially divorced eutectic, clearly demonstrate that an Al-rich layer is present between the base metal and the top corrosion film. The occurrence of this layer in corroded Mg-Al alloys has been reported in the literature, using either XEDS [8] or XPS [10]. The consensus in the literature, one study [9] being an exception, seems to indicate that this layer is causing the enhanced corrosion resistance in these alloys. The chemical state of this film has been suggested to be close to amorphous Al2O3 [10], creating a percolation layer within the MgO/Mg(OH)2 corrosion film [8]. Our electron energy-loss spectroscopy (EELS) measurements performed on this Al-rich layer indicate that this layer is close to metallic [13]. One plausible explanation for this observation is that the Al-rich layer is highly defective Al2O3. Song, et al. [20] speculated that, under corrosion conditions, Al2O3 would form as a semiconductor or electronic conductor rather than as a crystalline wide band gap insulator. This could potentially explain the observed sharp plasmon peak, characteristic of the metallic state, in EELS measurements that we have carried out [13]

(not shown). The formation of this Al-rich oxide layer can be linked to the higher Al alloy content in this region and the preferential dissolution of Mg during corrosion. The uniformity of the corrosion product layer on the eutectic Mg matrix and the β-Mg17Al12 eutectic phase suggests that the β-phase may not act as a highly active cathode. This observation is in agreement with the claim that the intermetallic β-phase is only slightly more noble than the α-Mg, with essentially the same corrosion rate [21]. The ability of the β-phase to act as a cathode would be expected to be related to the Al content of the Mg matrix phase: the higher the Al content in solid solution in the Mg grain, the smaller the electrochemical difference with the two phases [21] and the less likely galvanic coupling between a Mg anode and a β-phase cathode. 4.2. The Structure of the Corrosion Film in Area 2 (Primary α-Mg) At this location the corrosion film was substantially thicker (~1.5 to 2 µm) compared to that of location 1 (150-200 nm). A layer with a columnar morphology and a similar thickness to that at location 1 is present on the outside of the thicker sublayer of MgO and Mg(OH)2. Electron diffraction measurements, Figure 11, showed the Mg(OH)2 is present closer to the Mg substrate. This thick corrosion product was evenly permeated with chloride, strongly suggesting that it did not prevent continuous access of the chloride solution to the α-Mg substrate through the full corrosion exposure period. This contrast to the thin oxide at location 1, which contained no detectable chloride, confirms the protective effect of the oxide at locations where the thin Al-rich layer is present. The absence of the Al-rich layer can be attributed to the low Al content of the α-Mg at this location, which was measured to be around 2-3 wt.%. This is significantly lower than that of the eutectic Mg matrix at location 1 which, further away from the eutectic β-Mg17Al12, was close to 11-12 wt.% Al [13]. It is plausible that the mechanism through which the columnar sections of the corrosion film in Areas 1 and 2 are formed are similar, resulting from deposition or precipitation of corrosion products onto the surface. The same process was postulated by Nordlien, et al. [6] for the case of formation of the platelets in the top section of the corrosion layer of Mg in water.

5- Summary and Conclusion The detailed electron microscopic analysis of the corroded surface of the sand cast AM50 alloy has been performed. Two representative areas were analyzed: (i) a location close to a divorced eutectic microstructure where the extent of corrosion was minor; and (ii) an area located on an α-Mg grain where corrosion was extensive. At the lightly corroded area, the outer regions of the corrosion film were comprised of a columnar mixture of amorphous MgO and Mg(OH)2. Beneath this layer, at the alloy/oxide interface, a thin Al-rich layer was observed. This layer was metallic in character and is most likely a highly defective Al2O3 formed by the preferential dissolution of Mg from the alloy at this location which has a high Al content (11-12 wt.%). On the heavily corroded area, located on an α-Mg grain, the thick corrosion product layer is predominantly crystalline MgO. The absence of the Al-rich layer at this location allows corrosion to progress uninhibited. The inability of the surface to form the Al-rich layer can be attributed to the low Al content of the α-Mg grain (2-3 wt.%) at this location.

Acknowledgements Funding for this research was provided by General Motors Canada and the Natural Sciences and Engineering Research Council of Canada (NSERC). The microscopy work was carried out at the Canadian Centre for Electron Microscopy, a facility supported by NSERC and McMaster University.

References                                                                                                                [1] Westengen H, Aune TK. Magnesium Casting Alloys: in Friedrich HE, Mordike BL (Eds.) Magnesium Technology: Metallurgy, Design Data, Applications, Springer-Verlag, Berlin, 2006.    [2] Wei L-Y, Warren R. Microstructural characterization of several magnesium alloys in AM series. Materials Science and Technology 2007;23(6):745. [3] Mordike BL, Lukac P. Physical Metallurgy: in Friedrich HE, Mordike BL (Eds.) Magnesium Technology: Metallurgy, Design Data, Applications, Springer-Verlag, Berlin, 2006. [4] Song G, Atrens A. Understanding magnesium corrosion: A framework for improved alloy performance. Advanced Engineering Materials 2003;5(12):837. [5] Ma Y, Zhang J, Yang M. Research on microstructure and alloy phases of AM50 magnesium alloy. Journal of Alloys and Compounds. 2009;470:515. [6] Nordlien JH, Ono S, Masuko N, Nicancioglu K. J. Electrochem. Soc. 1995; 142(10): 3320-3322. [7] Nordlien JH, Ono S, Masuko N, Nisancioglu K. Corrosion Science 1997; 39(8): 1397-1414. [8] Nordlien JH,Nisancioglu K, Ono S, Masuko N. J. Electrochem. Soc. 1996; 143(8): 2564. [9] Splinter SJ, McIntyre NS. Surface Science 1994; 314: 157-171. [10] Wang L, Shinohara T, Zhang BP. Applied Surface Science 2010; 256: 5807-5812. [11] Feliu S Jr, Merino MC, Arrabal R, Coy AE, Matykina E. Surface and Interface Analysis 2009; 41: 143-150. [12] Jeurgens LPH, Vinodh MS, Mittemmeijer EJ. Acta Materialia 2008; 56: 4621-4634. [13]  Danaie M, Asmussen RM, Jakupi P, Shoesmith DW, Botton GA. Submitted to Corrosion Science. [14] Asmussen RM, Jakupi P, Danaie M, Botton GA, Shoesmith DW. Accepted for publication in Corrosion Science. [15] Oldfield RJ. “Differential Interference Contrast Light Microscopy”, Encyclopedia of Life Sciences, 2001, John Wiley & Sons, www.els.net. [16] Dahle AK, Lee YC, Nave MD, Schaffer PL, StJohn DH. Journal of Light Metals 2001; 1: 61-72. [17] Liu XJ, Ohnuma I, Kainuma R, Ishida K. Journal of Phase Equilibria 1999; 20(1): 45. [18]  Danaie M, Asmussen RM, Jakupi P, Shoesmith DW, Botton GA. In Preparation and To Be Submitted to Corrosion Science. [19] Taheri M, Phillips RC, Kish JR, Botton GA. Corrosion Science 2012; 59: 222-228. [20] Song G, Atrens A, Wu X, Zhang B. Corrosion Science 1998; 40(10): 1769-1791. [21] Mathieu S, Rapin C, Steinmetz J, Steinmetz P. Corrosion Science 2003; 45: 2741-2755.