Influence of trace As content on the microstructure and corrosion behavior of the AZ91 alloy in different metallurgical conditions

Zhe Zhng, Guohu Wu,?, Andrej Atrens, Wenjing Ding

a National Engineering Research Center of Light Alloy Net Forming and State Key Laboratory of Metal Matrix Composites, School of Materials Science and Engineering, Shanghai Jiao Tong University, Shanghai 200240, China

b School of Mechanical and Mining Engineering, the University of Queensland, Brisbane, Qld 4072, Australia

Abstract The influence of trace As content (0.04, 0.05, 0.06wt%) on the microstructure and corrosion behavior of AZ91 alloy in different metallurgical conditions (as-cast, as-quenched, peak-aged) was firstly investigated. It is found that the corrosion resistance of AZ91 alloy in all conditions is significantly enhanced by As addition, mainly due to cathodic toxication, melt purification and incidence reduction of local corrosion. As alloying leads to the formation of Mg3As2 and makes the β-Mg17Al12 phase less continuous in the as-cast alloys. The alloy containing 0.06wt% As obtains a higher corrosion rate than those of the alloy containing 0.05wt% As owing to a more discontinuity of β phase. The β phase is dissolved into the matrix in the as-quenched alloys and reprecipitates along the grain boundaries after aging.The more continuous β phase distribution in the peak-aged alloys contributes to corrosion resistance. The corrosion rates are in the order of as-quenched>as-cast>peak-aged. The lowest corrosion rate (0.67mm/y) in 3.5wt% NaCl solution is obtained in the peak-aged AZ91-0.05As alloy, which is over 70% lower than that of as-cast AZ91 alloy (2.22mm/y).

Keywords: Magnesium alloy; AZ91; Microstructure; Corrosion behavior.

1. Introduction

Mg alloys are becoming more-widely used in various industries, such as industries related to automobiles, aerospace,computers and communications, etc., owing to their desirable properties of low density, high strength-to-weight ratio,low resistivity and good electromagnetic shielding[1,2].However, magnesium is an active metal in structural applications and is highly susceptible to corrosion especially in acidic and NaCl solution, which limits its more wide-spread application[3,4]. Although numerous efforts have been devoted to improving the corrosion resistance of magnesium alloys by different techniques such as surface coating technologies, surface treatment technologies, and alloying [5], there are only few alloys that have corrosion rates approaching the intrinsic corrosion rate of Mg. Furthermore, the surface coating technologies and surface treatment technologies may encounter environmental pressure, and typically the preparation process is complex.

More attention has been devoted to alloying technology,owing to its high economic efficiency and feasibility. The present work was undertaken to investigate the inflence of an alloying method on AZ91 alloy. AZ91 alloy is one of the most widely used magnesium alloys, due to its good casting ability and high yield strength [6]. It would be of significant technological importance to increase the corrosion resistance of AZ91 alloy.At present,the effects of the addition of Y,Ce,La, Ca, Si, Sb, Sm on the corrosion resistance of AZ91 alloy have been investigated [6–10]. Table 1 provides a compilation of corrosion rates for pure Mg, Mg–Al–Zn, Mg–Al–Mn and Mg–Al–RE alloys in recent years, in solutions like 3.5wt%NaCl [6–9,11–23]. Although a number of studies have been done,almost all of them reduced corrosion rate just by addingalloying elements to change the phase structure and reduce the influence of impurity elements on corrosion, contrary to few investigations on the restriction of anodic or cathodic kinetics in Mg alloys. Besides, the further improvement of corrosion behavior is urgent, in order to extend the application range of Mg alloys.

Table 1 Corrosion rates for pure Mg, Mg–Al–Zn, Mg–Al–Mn and Mg–Al–RE alloys for immersion tests at the open circuit potential (OCP) from recent years, by year of publication, first author, alloy (HP is high-purity), condition (wrought, W; as-cast, C; extruded, E; die-cast, DC; Friction stir processing, FSP), composition(wt%), test solution (0=0.1M NaCl, 1=0.5wt% NaCl, 2=1wt% NaCl, 3=0.6M NaCl, 4=3.5wt% NaCl saturated with Mg(OH)2, 5=5wt% NaCl, 6=salt spray, 7=Hanks’ solution (pH=8.2)) and measurement (weight loss, PW, hydrogen evolution, pH, electrochemical impedance spectroscopy, Pi,EIS), corrosion rate (mm y?1) and reference.

Table 1 (continued)

It was recently proposed that As could retard the cathodic kinetics on pure Mg [24,25]. Such an inhibiting effect on the cathodic reaction was claimed to reduce the corrosion rate of Mg below that of the rather impure Mg used, to 4mm/y.However, it is well-known that alloy compositions are complex, hence the addition of As in Mg alloy might change the interaction between different atoms to form different corrosion mechanisms compared with pure Mg. Therefore, it is necessary to study the specific role of As addition in the corrosion behavior of Mg alloy. At present, there are no studies of As being added to the AZ91 alloy. It is notable that pure As is non-toxic [26]. Less than 0.1wt% As addition is chosen in this work and our experiments including alloy melting and heat treatment are strictly protected. In addition, heat treatments also have great impact on the corrosion behavior due to the modification of the microstructure [21,27,28]. Investigations on the microstructural evolution of AZ91 alloys with As during heat treatment are essential to comprehensively understand the effect of As alloying on the corrosion behavior of AZ91 alloy.

The aim of the present study is to study the corrosion behavior of As alloyed into AZ91 alloy, in order to exploit a kind of high corrosion-resistant AZ91 alloy with trace As addition. The influence of the As content (x=0.04, 0.05, and 0.06wt%) on the microstructure and corrosion behavior of AZ91 alloy, in as-cast, as-quenched and peak-aged conditions is studied. The form of the As, the feasibility of As alloying to improve corrosion resistance, and the optimal As alloying are also discussed in detail.

2. Experimental

2.1. Materials

The cast Mg alloys AZ91-xAs were prepared from a commercial AZ91 and pure As, by melting in an electronic resistance furnace at 750 °C under the protection of a SF6–CO2gas mixture. 2wt% JDMJ refining agent (mixture of MgCl(45%), KCl (25%), NaCl (20%), others (10%)) [29] was added at 750 °C after the alloy was completely melted, followed by stirring vigorously, and isothermally holding at 730°C for about 20min to ensure the complete homogenization of the alloying elements. Then the melt was poured into a permanent mold pre-heated to 200 °C. The chemical compositions of the studied alloys, measured by Inductively Couple Plasma-Atomic Emission Spectroscopy (ICP-AES), are listed in Table 2. The alloys are hereafter referred to as the base,0.04As, 0.05As, and 0.06As alloy. The as-cast alloys were subjected to (i) solution treatment (410 °C×24h) in an electrical oven, and then quenched into cold water (as-quenched),followed by (ii) an aging treatment (200 °C×16h) in an oil bath (peak-aged) [21]. These as-quenched and peak-aged are designated as T4 and T6, respectively.

Table 2 Measured chemical compositions of the cast AZ91-xAs alloys.

2.2. Microstructural analysis

Optical metallography samples were prepared by mechanically grinding with SiC papers to 7000 grit, and polishing through standard procedures. Polished samples were etched with 4% (volume fraction) nital for 5s. The microstructure was observed with an optical microscope (OM, LEICA MEF4M) and a scanning electron microscope (SEM, Phenom XL). Energy dispersive X-ray spectroscopy (EDS) and X-ray diffraction (XRD, ADVANCE Da Vinci) were employed to identify the phases. The surface corrosion morphology of the specimens after immersion testing was observed using the SEM, and the surface corrosion products were analyzed using X-ray photoelectron spectroscopy (XPS, UltraDLD).

2.3. Immersion test

Specimens for immersion test were short cylinders (35mm×5mm). A hole, 0.35mm in diameter, was drilled through the specimens so that the specimens could be suspended in the solution by means of a nylon string. Specimens were ground with SiC papers to 7000 grit, polished through standard procedures, washed with distilled water, and cleaned by sonication in ethanol. Immersion testing was performed in quiescent 3.5wt% NaCl solution at 25 °C for 72h. The solution pH rapidly increased within the first few hours to a pH of 10.3 due to the low solubility of Mg(OH)2[4]. Thus, the testing solution was saturated with Mg(OH)2, so that the solution composition remained constant throughout the immersion test.A chromic acid solution (200g/L chromium trioxide, 10g/L silver nitrate and 20g/L barium nitrate) was used to remove from the surface the corrosion products prior to the determination of mass loss. The corrosion rate, vm(mg cm?2day?1),was evaluated using [30]:

where m0represents the weight prior to immersion, m1is the weight after immersion, S is the surface area of the sample,and t is the immersion time.The corrosion rate, PW(mm y?1)was determined from [30]:

2.4. Electrochemical measurement

Specimens for electrochemical measurement had dimensions of 20mm×20mm×5mm. One surface of the specimen was prepared by mechanically grinding with SiC papers to 7000 grit, polishing through standard procedures, and then cleaning by sonication in ethanol, while the other surfaces were sealed with insulating rubber. Quiescent 3.5wt% NaCl solution at 25 °C was employed as the test electrolyte for all electrochemical test using the PARSTAT 2273 system.A classical three electrode cell was used with graphite as the counter electrode,a saturated calomel electrode(SCE)as the reference electrode, and the sample, with an exposed area of 4cm2,as the working electrode. Before electrochemical impedance spectroscopy (EIS) measurements and potentiodynamic polarization (PDP), the working electrode was immersed into the corrosive solution at the open circuit potential (OCP) for 1h and 30h. EIS measurements were performed at the opencircuit potential with a frequency ranging from 100kHz to 100 mHz, with a 10mV amplitude voltage. Potentiodynamic polarization tests were carried out with a scan rate of 1mV/s,from ?0.25V against the open circuit potential,OCP,to about?1.9V.

3. Results

3.1. Microstructure

Table 2 shows that the Fe content decreased with the increase of the As content.

The phases in the as-cast AZ91-xAs (x=0, 0.04, 0.05,0.06) alloys are shown in Fig. 1. The base alloy consisted of α-Mg and the β-Mg17Al12phase, whereas additional diffraction peaks related to the Mg3As2phase were well-defined in the As-containing alloys(Fig.1(b)–(d)).The values of diffraction peak intensity of the Mg3As2phase were lower by comparison to the other peaks.

Fig. 2 shows the optical micrographs of the microstructures of the as-cast, as-quenched (410 °C×24h) (T4 temper)and after aging (200 °C×16h) (T6 temper) AZ91-xAs alloys. The as-cast AZ91 microstructure had typically α-Mg matrix and primary β-Mg17Al12phase. The skeleton-like β-Mg17Al12phase was distributed in the vicinity of the grain boundaries (Fig. 2(a)). The β-Mg17Al12phase of the as-cast AZ91-xAs (x=0.04, 0.05, 0.06) alloys was clearly less continuous and no Mg3As2phase was evident in Fig. 2(d), (g),and (j), which indicated that the trace amounts of As changed the morphology of the secondary phase of the as-cast alloys.Moreover, the Mg3As2content was too small to be visible with the optical microscope. There were fine lamellar eutectic microstructures, which consisted of the large β-Mg17Al12phase and the eutectic α-Mg phase (Fig. 2(g), (j)). After solution treatment, most of the β-Mg17Al12phase particles had been dissolved into the α-Mg matrix, and the average grain size increased to approximately 100μm. However, the average grain size was almost constant with increasing As content(Fig. 2(b), (e), (h), and (k)). The T6 aging treatment caused β-Mg17Al12phase precipitation along the grain boundaries,and within the grains of the α-Mg matrix (Fig. 2(c), (f), (i),and (l)).

Fig. 1. XRD patterns of as-cast AZ91 alloy with different As contents: (a) base alloy; (b) 0.04As alloy; (c) 0.05As alloy; (d) 0.06As alloy.

Fig. 3 shows the SEM microstructures and EDS results of as-cast AZ91 and AZ91-xAs (x=0.05, 0.06). The SEM results clearly show that the degree of discontinuity of the β-Mg17Al12phase improved with the increase of the As content in AZ91 alloys (Fig. 3(a)–(c)). The proportion of β-Mg17Al12phase was calculated by using Image-Pro Plus (IPP) software and showed in Table 3. It is obvious that the figures for the area ratio of β-Mg17Al12phase reduced with the increase of the As content in AZ91 alloys. This also demonstrated that as the content of As increased, β-Mg17Al12phase would be less continuous. Fig. 3(d) indicates the EDS point analysis results of as-cast AZ91-0.05As. According to the atomic ratio or weight ratio of Mg and As, Mg3As2phase was formed and located in the vicinity of grain boundaries. The size of these white particles was around 1μm. Typical SEM image mapping analyses of 0.05As alloy are showed in Fig. 4 which indicates that a portion of As was dissolved in α-Mg matrix.

3.2. Corrosion behavior

The corrosion rates of the alloys were evaluated by weight loss, EIS and polarization curves. Fig. 5 shows the corroded surfaces of the AZ91-xAs alloys after 72h immersion in the 3.5wt% NaCl solution.

For the as-cast alloys, the most of the surface of AZ91 alloy was covered with loose white corrosion products. In contrast, the surface of each AZ91 alloy with As was clean and bright, with only small areas that were corroded. This indicated that the As-containing AZ91 alloys had a better corrosion resistance in this test solution (Fig. 5(a)–(d)).

For the alloys in the T4 condition, white thick corrosion products covered a substantial part of the surface of the alloys. The corrosion of the base alloy was more severe than the other alloys. The coverage of the corrosion products increased in the order of AZ91-0.05As

A considerable amount of dark filiform corrosion was present on the surface of all the alloys with the T6 aging treatment (Fig. 5(i)–(l)). The surfaces of the AZ91-xAs (x=0.04,0.05, 0.06) alloys still maintained some metallic lustre with considerable filiform corrosion. The base alloy was covered with corrosion products.

The weight loss rate for the AZ91-xAs alloys is exhibited in Fig. 6 which indicates that there was a significant decrease in the corrosion rate of the AZ91 alloys containing As in comparison to that of the base AZ91 alloy. Overall T4 alloys had the highest corrosion rate. AZ91-0.05As alloy exhibited the largest decrease in the corrosion rate, and the lowest corrosion rate.

Fig. 2. Representative optical micrographs of AZ91-xAs alloys (x=0, 0.04, 0.05, 0.06) at as-cast state, as-quenched state (T4) and aged state (T6): (1) alloys:(a), (b), (c) base alloys; (d), (e), (f) 0.04As alloys; (g), (h), (i) 0.05As alloys; (j), (k), (l) 0.06As alloys. (2) treatment: (a), (d), (g), (j) as-cast; (b), (e), (h), (k)T4; (c), (f), (i), (l) T6.

The Nyquist and Bode spectra of the 12 alloys in the 3.5wt% NaCl solution after different immersion times are depicted in Figs. 7 and 8, respectively. The Nyquist spectra of these alloys all consisted of three loops, namely, one high frequency capacitance loop, one medium frequency capacitance loop, and one low frequency inductance loop. Their Nyquist spectra were similar except for the difference in the diameter of the loops. The impedance spectra indicated that the diameters for most alloys reduced in the order of T6>ascast>T4. The diameter of the capacitance loop of T6 AZ91-0.05As alloy was larger than the other alloys. The result of the impedance spectra was consistent with the weight loss rate in Fig. 6. Additionally, there were three-time constants in Fig. 8. A comparison of alloys after 1h and 30h immersion (Figs. 7 and 8), indicates that the diameter of capacitance of alloys in Nyquist spectra reduced as well as the values of phases and |Z| in Bode spectra dropped after 30h immersion.

The equivalent circuit shown in Fig. 9 was used to analyze the experimental EIS data. Rsis the solution resistance.Rtand the constant phase element (CPEdl) describe the high frequency capacitive loop. Rtis the charge transfer resistance,which illustrates the transferring of electrons in original substrate, and CPEdlstands for the electric double layer capacity.The CPEdlis used to compensate for the non-homogeneity in the system and is determined by CPEdl-Tand CPEdl-P. Rfand Cfcharacterize the medium capacitance loop, originated from the diffusion through a porous solid film on the surface of the alloy [31]. Rfis the film resistance and Cfrepresents the film capacitance. RLand L describe the low frequency inductive loop, corresponding to corrosion nucleation of localized corrosion. Besides the point where the spectrum first crosses the real axis at intermediate frequencies, the charge transfer resistance,Rt,is used to determine corrosion rate with the Stern–Geary equation which is not accurate, the low frequency impedance limit is the parameter of interest with respect to determination of the corrosion rate, which is definitively determined from on-line spectroelectrochemistry [31].EIS provides a method to determine the polarization resistance, Rp. This enables circuit simplification and the subsequent calculation of Rpto be estimated by Eq. (2) [31]:

The component-values of equivalent circuit are given in Table 4. The regularity of Rpvalues in as-cast and T6 AZ91-xAs alloys is consistent with that of weightlessness analysis, while Rpof T4 alloys is irregular.

Fig. 3. SEM and EDS results of as-cast AZ91-xAs alloys (x=0, 0.05): SEM backscattered micrograph of (a) Base alloy, (b) 0.05As alloy, (c) 0.06As alloy,(d) SEM backscattered micrograph of AZ91-0.05As alloy in a higher magnification and EDS results of point 1, point 2 and point 3.

Table 3 The area ratio proportion of β-Mg17Al12 phase in as-cast AZ91-xAs (x=0, 0.04, 0.05, 0.06) alloys (five parallel samples in each alloy).

Table 4 Fitting results of the EIS spectra for different conditions AZ91-xAs (x=0, 0.04, 0.05, 0.06) alloys for 1h and 30h immersion.

Fig. 4. (a) SEM backscattered micrograph of as-cast 0.05As alloy; elements map-scanning of: (b) Mg; (c) Al; (d) As.

Potentiodynamic polarization tests were also carried out.Representative polarization curves of the base alloy and AZ91-xAs (x=0.04, 0.05, 0.06) alloys are presented in Fig.10. The polarization response of AZ91 alloys with As was markedly different compared to that of AZ91 alloy. For 1h immersion, As-free AZ91 alloys had values of icorrthat were at least an order of magnitude lower than those of the As containing AZ91 alloys. However, after 30h immersion, the corrosion potential (Ecorr) of AZ91 alloys reduced and icorrincreased,while icorrof AZ91 alloys with As declined.Alloying additions of As generally led to a decrease of Ecorr. The cathodic reaction kinetics for AZ91-xAs alloys were comparable to those of AZ91 alloys, which indicated that the presence of alloyed As decelerated the cathodic processes. The other feature was that apparent pseudo‘passive window’trends existed for AZ91 alloys. Therefore, the anodic polarization curves were too complicated to allow accurate calculation of the corrosion current density.

3.3. Corrosion morphologies

Fig. 11(a)–(d) show the corrosion morphologies of as-cast AZ91-xAs alloys after 72h immersion in 3.5wt% NaCl. The SEM morphologies reveal that a denser corrosion product film was formed on the surface of as-cast AZ91 alloys with As than that on the base alloy. The corrosion morphology was almost the same, which indicated that As did not have an obvious effect on the corrosion products and the corrosion films. Moreover, because the AZ91-0.05As alloy had the best corrosion resistance among the as-cast alloys, the densities of surface corrosion products in the different states of the AZ91-0.05As alloys were compared in Fig. 11(c), (e), and (f). The corrosion morphologies of the T4 and T6 alloys were the same as that of as-cast alloy. The XPS compositions of the film of as-cast AZ91-0.05As alloy were 20 at% Mg, 41 at%O and 29 at% C, while the H element cannot be detected via XPS test (Fig. 11(c)). It is to be expected that some of the detected carbon was from the inevitable carbon contamination in the XPS chamber. The As did not exist in the corrosion product of the as-cast AZ91-0.05As alloy.Furthermore,representative corroded surface photographs in Figs. 5 and 11 indicate the similar corrosion product morphologies in AZ91 and the AZ91 alloys with As. The corrosion products of ascast AZ91 alloys with different As contents were determined by XRD as shown in Fig. 12. The XRD patterns indicate the same phases for the different alloys, including the alloy phases (α-Mg, β-Mg17Al12). Thus, the corrosion products were Mg(OH)2and MgCO3?3H2O.

Fig. 5. Representative corroded surface photograph of AZ91-xAs (x=0, 0.04, 0.05, 0.06) alloys at as-cast state, as-quenched state (T4) and aged state (T6)after 72h immersion in 3.5wt% NaCl.

Fig. 6. Weight loss rate for AZ91-xAs (x=0, 0.04, 0.05, 0.06) alloys at as-cast state, as-quenched state (T4) and aged state (T6) after 72h immersion in 3.5wt% NaCl.

Fig. 7. Impedance spectra of AZ91-xAs (x=0, 0.04, 0.05, 0.06) alloys at as-cast state, as-quenched state (T4) and aged state (T6) for immersion in 3.5wt%NaCl solution: (a) 1h, (b) 30h.

Typical cross-section morphologies of the base alloy and AZ91-0.05As alloy after 72h immersion in 3.5wt% NaCl are presented in Fig. 13. The deepest corrosion pit of the as-cast AZ91-0.05As alloy was about 157μm, which was somewhat deeper than that of the as-cast AZ91 alloy which was about 140μm (Fig. 13(a) and (b)). But the corrosion area of the ascast AZ91-0.05As alloy was smaller than that of as-cast AZ91 alloy (Fig. 5(a) and (c)), consistent with the corrosion rate of the AZ91 alloy with As being lower than that of the base alloy. Some β-Mg17Al12phases were in the forefront of the corrosion advance. Moreover, most of the corrosion pits were shallow but there were a few deeper pits when the alloys were heat treated (Fig. 13(c) and (d)), and the deepest corrosion pit of T6 AZ91-0.05As alloy was only about 136μm, much smaller than that of as-cast AZ91-0.05As alloy.

Fig. 8. Bode spectra of AZ91-xAs (x=0, 0.04, 0.05, 0.06) alloys at as-cast state, as-quenched state (T4) and aged state (T6) for 1h and 30h immersion in 3.5wt% NaCl solution.

Fig. 9. Equivalent circuits of the EIS spectra.

4. Discussion

4.1. Corrosion mechanism

In AZ91 alloys, α phase and β phase play an important role in the corrosion behavior [5,32]. The β phase is the galvanic cathode and accelerates the overall corrosion of the α matrix when its volume fraction is low. Nevertheless, if β phase has a high fraction and continuous morphology, it can be an anodic barrier against the overall corrosion of the alloy[33]. In addition, the aluminum concentration can vary from a few percent in the grain interior to 10% in the vicinity of the β phase, and the variation of the concentration of Al is in the range of about 35% in the β phase to about 6% in the α phase, and the region with less than 8% Al could be preferentially corroded. Thus, the lower Al content of the α matrix is the initiation site of corrosion. Moreover, the eutectic α phase also plays an important role in the corrosion behavior,because it corrodes faster than the α matrix if highly anodically polarized, whereas the eutectic α corrodes more slowly than the α matrix when not strongly polarized [33].

The XPS result indicates that the corrosion products mainly contained Mg, C and O (Fig. 11(c)). However, some researches [34] had shown that the corrosion products were Mg(OH)2, which indicated that a small amount of carbon dioxide was dissolved in the sodium chloride solution in the present experiment, and it was to be expected that some of the detected carbon was from the inevitable carbon contamination in the XPS chamber. According to the work of J?nsson [35], the corrosion products of AZ91 exposed to the atmosphere were the magnesium carbonates hydromagnesite[Mg5(CO3)4(OH)2?4H2O] and nesquehonite (MgCO3?3H2O)and long exposures resulted in the formation of pits containing brucite [Mg(OH)2] covered with hydromagnesite crusts in the presence of carbon dioxide gas. The primary reactions involved under polarization of α-Mg can be described by reactions (4)–(9) [35].

Fig. 10. Polarization curves of AZ91-xAs (x=0, 0.04, 0.05, 0.06) alloys at as-cast state, as-quenched state (T4) and aged state (T6) for 1h and 30h immersion in 3.5wt% NaCl solution: (a) as-cast alloys; (b) T4 alloys; (c) T6 alloys.

Reactions (4) and (5) are the overall anodic and cathodic half-cell reaction respectively during the dissolution process of Mg. The reaction (6) occurs due to the localized alkalization and the low solubility of Mg(OH)2; whilst the reaction(7) is the overall reaction of reaction (4)–(6). Owing to some CO2in the 3.5wt% NaCl solution, brucite can react directly with CO2to form magnesite by reaction (8). Magnesite can directly undergo the hydration reaction (9). The above explains why the products of the corrosion of AZ91 in these experiments were MgCO3?3H2O and Mg(OH)2.

In the case of the AZ91 alloy with As addition, some prior work suggested that the main action was to retard hydrogen evolution (HE) by poisoning the principal cathodic sites on the Mg surface [24], thereby to decrease the corrosion rate.McNulty et al. argued the corrosion behavior of any Mg alloy containing an appreciable Fe content was caused by the trace impurities of Fe-rich second-phase particles that constituted the principal cathodic hydrogen evolution reaction (HER) site on Mg[36].Yang et al.[15]studied the corrosion behavior of different contents of Fe in pure Mg, and found the corrosion rates of pure Mg with 15ppm Fe, 18ppm Fe and 25ppm Fe were about 2mm/y, 12.1mm/y and 106.3mm/y, respectively.Thus, Fe had a significant effect on the corrosion behavior of Mg. Song et al. [4,37] viewed two hypotheses for the tolerance limits of Fe that were phase precipitation and surface deposition. The phase precipitation hypothesis was that Fe-rich phases were precipitated inside Mg alloy and could accelerate the cathodic reaction. The surface deposition hypothesis was that Fe deposition on the Mg alloy surface was an effective cathode. As atoms can easily bond with the Fe deposition on the AZ91 alloy surface, which can possibly lead to effectively blocking the H atom recombination. And in the alloy, As was prone to combine with Fe-rich phase precipitation, leading to deactivation of Fe-rich phase, and thereby the suppression of cathode reaction. Moreover, Fig. 10 shows that the As alloying indeed had an appreciable influence on the hydrogen evolution reaction, and on the corrosion rate.

On the other hand, As did purify the Fe content of AZ91 alloys on the basis of Table 2. The maximum solid solubility of As in Fe is up to 10% at 835 °C, while 0.5% at 25 °C[38]. Thus, during the process of melting, As was easy to combine with Fe and the particles containing As and Fe were heavy and sank into the bottom of the melt and discharged into the slag when adding As to the melt during melting. This illustrates that As had purification effect during the melting process. Besides, the Fe content became less with increasing amount of As, which indicated that, as the As content increased, there was more of the purification effect. Thus, the purification effect indicates that the corrosion resistance of the AZ91 alloy is improved by the addition of As.

4.2. Influence of As content on microstructure of AZ91 alloys

For the as-cast AZ91 alloy, the As addition made the skeleton-like β-Mg17Al12phase less continuous, and there was Mg3As2phase in the as-cast AZ91-xAs alloys (Fig. 3).

Fig. 11. SEM images of corrosion morphologies of AZ91-xAs alloys after 72h immersion in 3.5wt% NaCl: (a) as-cast AZ91 alloy, (b) as-cast AZ91-0.04As alloy, (c) as-cast AZ91-0.05As alloy, (d) as-cast AZ91-0.06As alloy, (e) T4 AZ91-0.05As alloy, (f) T6 AZ91-0.05As alloy in a higher magnification.

Fig. 12. XRD patterns of corrosion products of as-cast AZ91 alloy with different As contents.

This might be largely because arsenic atoms acted as nucleation sites during solidification increasing nucleation sites,resulting in the discontinuity of the second phase. According to the binary phase diagram of magnesium and arsenic[39], As exists at the solid-liquid interface during solidification due to its eutectic reaction with magnesium, thereby it has an effect on the second phase formation. Moreover,the lattice misfit of As (a=b=0.869nm, c=0.636nm) and Mg17Al12(a=b=c=10.56nm) is 17.7%, which may form a semi-coherent interface, favorable for the second phase to take As atoms as its nucleation center. This explains the fact that Mg3As2phases were mainly in the vicinity of the grain boundary and As addition made the β-Mg17Al12phase less continuous. This might not block the corrosion process very well. Hence, this should be the reason why the excessive As could result in a lower corrosion resistant of AZ91 alloy. Meanwhile, As addition can also improve the corrosion resistance of AZ91 alloys due to the reduction of HER and Fe content, so AZ91-xAs alloys will have the most optimal As content in terms of corrosion resistance. The 0.05wt% As was the optimized alloy composition in this work.

After solution treatment, the β phase dissolved and the content of Al increased in the α matrix. T4 alloys suffered from general corrosion and thick corrosion products covered the surface of the alloys, so the corrosion rate of T4 alloys was the highest than that of as-cast and T6 alloys [40]. Zhou et al. [41] found there was a meta-stable, partially protective film on high aluminum content of the α matrix in the initial exposure but the dissolution rate of localized corrosion would be faster once the protective film was broken down due to weak localized sites of residual β phase and other Al–Mn or Al–Mn–Fe intermetallics. Therefore, the corrosion rate of T4 alloys significantly increased due to exposure in NaCl solution. Another reason is that after solution treatment, on account of the Mg–Fe phase diagram, the precipitation of the BCC Fe rich phases can accelerate the corrosion of the alloys[3,37]. Although the microstructures of As addition were not very different from the base alloy (Fig. 2(b), (e), (h), and(k)), adding As could increase corrosion resistance because it might prolong the time to protective film break down. In a word, the corrosion resistance of the whole alloys was worse after solution treatment but the alloys with As was slightly less corrosive in comparison.

For the T6 microstructure, a large number of fine second phases were precipitated at grain boundaries. The filiformlike corrosion morphology is of relevance to the evolution of local cathodes and anodes,during the corrosion process of Mg alloys [42,43]. The corrosion pits initiated at the anodic α-Mg matrix adjacent to the cathodic β-Mg17Al12precipitates.The microgalvanic action due to the β-Mg17Al12precipitates at grain boundary led to intergranular corrosion. Compared with samples of solution treatment, the β phases of T6 alloys blocked the corrosion process and the stable film on the β phase acted as barrier to inhibit corrosion [44]. As for the comparison with as-cast alloys, the depth of the pits of T6 alloys was relatively smaller,which would cause the corrosion resistance of T6 alloys to be better than that of as-cast alloy.And during the aging treatment, the Fe-rich phases of AZ91 alloys would be the nuclei to the precipitates formed and therefore these Fe-rich precipitates became deactivated [3,37].Moreover, As addition can reduce the corrosion rate and the reason is approximately the same as that of as-cast alloy.Therefore, the corrosion resistance of T6 alloys were better than as-cast and T4 alloys.

4.3. Influence of As content on electrochemical behavior of AZ91 alloys

Electrochemical methods are important and rapid tools for assessing the corrosion mechanisms of metals, but aspects of their interpretation have been shown to be non-trivial in the case of Mg alloys, since their corrosion behavior is more complex than most other metals [45]. The equivalent circuit of the EIS spectra (Fig. 7) demonstrates the same corrosion mechanism of AZ91-xAs alloys in cast, T4, T6 states. The capacitance loops in the impedance diagrams result from both charge transfer and film effect of the corrosion products and the low-frequency inductive loop can be ascribed to chlorideinduced pit formation or an adsorbed intermediate [45].

For the as-cast alloys (Table 4), about the same values of CPEdlindicate that the double-layer properties of these alloys were similar. Rtincreased with the increase of As content,which indicated that the As addition could restrain the electron transfer of the corrosion process consistent with both a cathodic poisoning mechanism, or the Fe-impurity mechanism. The intermediate frequency capacitor loop characteristics Rfand Cfrepresented the corrosion film of the alloys with As could better protect the substrate although As was not found in the corrosion products. This may be explained that As addition inhibited the HER with the decrease of the anode corrosion degree causing the corrosion film did not become porous structure. The RLvalues of the AZ91 alloys with As were much bigger than that of base alloy, which meaned As addition reduced the incidence of local corrosion due to the substantial reduction of amount of impurities that usually initiate pitting corrosion. Nevertheless, excessive As could make the β phase less continuous, which made the passivating film less protective, which could reduce the local corrosion resistance of the passivating film on the alloy surface, leading to the reduction of corrosion resistance.

The corrosion rates of the T4 alloys increased owing to easy transfer of electrons to form loose corrosion product films and easy pitting, which was consistent with the corrosion results of immersion test. The reason for the random of Rpvalues was that the corrosion rate relied on many factors such as solution, the shedding of corrosion products and the density of film when there was particularly high degree of corrosion. For T6 alloys, the double-layer properties were the same as those of as-cast alloys. However, there was little difference of RLvalues between T6 alloys and as-cast alloys,thus the pitting resistance of T6 alloys was not better than that of the as-cast alloys. For as much as the Rtvalues of T6 alloy were a lot more than those of as-cast alloys, the main reason for the good corrosion resistance of T6 alloys was the difficulty of eletron transfer. In a word, it should be noted that the corrosion behavior of AZ91-0.05As alloy with T6 state was best among these alloys. In addition, the reason for the decrease of Rtand RLvalues in different metallurgical conditions for 30h immersion was the increase of the heterogeneity of the surface in the corrosion process [46].

Dissolution of Mg-alloys under anodic polarization is accompanied by persistent hydrogen evolution (HE) that increase with the amount of anodic polarization, which is called the negative difference effect (NDE) [47,48]. And the existence of pseudo passivation tendency for AZ91 alloys made it harder to calculate the corrosion rate accurately(Fig.10).This phenomenon could stem from the presence of partially protective corrosion product films on the surface [23]. At these breakdown potentials, the corrosion product films broke down and the current density dramatically increased to a higher value, indicating rapid dissolution of the anodic Mg. Also,with the increase of the immersion time, the trend between As-containing and As-free AZ91 alloys had changed.The reason might be that electrochemical measurements which were made soon after specimen introduction into the solution might not be representative of steady state corrosion. The low corrosion rate of the As-free AZ91 alloys on first immersion in the solution accelerated, which was the opposite of the Ascontaining alloys.This might be attributed to the phenomenon that the existence of a discontinuous second phase in the Ascontaining AZ91 alloys led to faster corrosion, and then the corrosion rate gradually became slow due to a more protective corrosion film in comparison with As-free AZ91 alloys.In contrast, loose and porous corrosion film of As-free AZ91 alloys accelerated the corrosion rate.

It must be noted that the polarization curves provide a method by which the mechanistic effect (in an electrochemical sense) of alloying can be determined in a manner that provides discrimination between anodic and cathodic kinetics.

The polarization curves for all the samples were not symmetrical between their anodic and cathodic branches. There were much sharper changes in the anodic polarization branches than in the cathodic polarization branches. There might be two reasons for the complicated nature of the anodic polarization curves.One reason is the simultaneous combination of both anodic dissolution and anodic hydrogen evolution in the anodic region. Another reason is the occurrence of localized corrosion which could make the anodic process unstable. Furthermore, cathodic polarization curves of As-containing and As-free alloys indicate that an explanation based on As poisoning the HER is likely, therefore the explanation that As improved the corrosion resistance of AZ91 is that the As changed the cathodic kinetics and the amount and form of Fe-precipitation.

Additionally, compared with other literature values [6–9,11–22], the corrosion rate PW(0.67mm y?1) of T6 AZ91-0.05As alloy in this research was relatively low, which indicated As might have significant effect on reducing the corrosion rate, lower than the literature values for AZ91 of 0.9mm/y [49–51], 1.1mm/y Zhang [22].

5. Conclution

The present work provides the first fundamental study on the influence of trace As content on the microstructure and corrosion behavior of AZ91 alloys in different metallurgical conditions. The following conclusions could be drawn:

1. The as-cast AZ91 alloys containing As contain the β-Mg17Al12phase and Mg3As2phase.The degree of β phase discontinuity raises with the increment of As content. Further 0.06wt% As addition increases the corrosion rate,compared with 0.05wt% As. The β phase is dissolved into the α matrix in T4 alloys and reprecipitates along grain boundaries after aging, leading to the corrosion resistance of different metallurgical conditions in the order of T6>as-cast>T4.

2. It is notable that As can easily combine with Fe-rich phase to inhibit HER, thereby greatly decreasing cathodic kinetics. EIS spectraindicate that the As-containing alloys have lower electron transfer and reduce incidence of local corrosion.

3. A trace amount of As dramatically decreases the corrosion rate of all the AZ91 alloys by inhibiting HER and reducing the Fe content by melt purification. The lowest corrosion rate is AZ91-0.05As-T6 of 0.67mm/y,markedly lower than that of as-cast AZ91 alloy (2.2mm/y).

Acknowledgements

This work was supported by the National Natural Science Foundation of China (No. 51775334, 51821001), National Defense Science and Technology Innovation Special Zone Project (No. 002-002-01), the National Key R&D Program of China (Grant no. 2016YFB0701205)