Corrosion of magnesium alloys: the role of alloying

Transcription

1 International Materials Reviews ISSN: (Print) (Online) Journal homepage: K. Gusieva, C. H. J. Davies, J. R. Scully & N. Birbilis To cite this article: K. Gusieva, C. H. J. Davies, J. R. Scully & N. Birbilis (2015) Corrosion of magnesium alloys: the role of alloying, International Materials Reviews, 60:3, , DOI: / Y To link to this article: Published online: 22 Oct Submit your article to this journal Article views: 1990 View Crossmark data Citing articles: 102 View citing articles Full Terms & Conditions of access and use can be found at

2 Corrosion of magnesium alloys: the role of alloying K. Gusieva 1, C. H. J. Davies 2, J. R. Scully 3 and N. Birbilis *1,3 The demand for light-weighting in transport and consumer electronics has seen rapid growth in the commercial usage of magnesium (Mg). The major use of Mg is now in cast Mg products, as opposed to the use of Mg as an alloying element in other alloy systems and there is an emerging market of wrought Mg products and biomedical Mg components such that the past two decades have seen a significant number of new Mg-alloys reported. None-the-less, the corrosion of Mg alloys continues to be a challenge facing engineers seeking weight reductions by deployment of Mg. Herein, authors review the influence of alloying on the corrosion of Mg-alloys, with particular emphasis on the underlying electrochemical kinetics that dictate the ultimate corrosion rate. Such a review focusing on the chemistry corrosion link, both in depth and in a holistic approach, is lacking. As such the authors do not describe aspects such as high-temperature oxidation or cracking, but focus on delivering the state-of-the-art with regards to alloying influences on corrosion kinetics. It has been demonstrated that Mg itself will not be thermodynamically passive in environments of ph,11, regardless of the extent and type of alloying and hence corrosion kinetics require unique attention. Authors consolidate the presentation to include essentially all commercially available alloys and in excess of 350 custom alloys with wide variations in composition; in addition to reviewing the range of intermetallic compounds and impurities that form in such alloys systems. An update is also given regarding mechanistic advances and the role of grain size on corrosion of Mg. A wider understanding of the role of chemical effects upon corrosion of Mg is both timely and serves to highlight metallurgical approaches towards kinetically retarding the corrosion problem. The latter is of key relevance to next generation lightweight alloys and rational design of wrought Mg and bio-mg. Keywords: Magnesium, Magnesium alloys, Corrosion, Intermetallic, Polarisation Introduction Magnesium (Mg) has the lowest density (1?74 g cm 23 )of the engineering metals. 1 Mg consumption is growing in many industrial applications in the present era of lightweighting, 2,3 which is important for energy and fuel savings, reduced emissions, emerging biomedical applications, and user satisfaction of portable electronics. Mg and its alloys are machinable, die castable, non-magnetic and offer vibration and shock adsorption ability. However, the two persistent limitations that have restricted wider Mg use include limited room temperature ductility and high corrosion rates in aqueous environments. The former issue, ductility, is undergoing significant research in the mechanical metallurgy domain, and the latter is still an area of active present research. 1 Department of Materials Engineering, Monash University, VIC 3800, Australia 2 Department of Mechanical and Aerospace Engineering, Monash University, VIC 3800, Australia 3 Department of Materials Science and Engineering, The University of Virginia, VA 22904, USA *Corresponding author, nick.birbilis@monash.edu Mg and its alloys rapidly form a surface oxide/ hydroxide layer in moist conditions. 4 The high negative free energy of formation ensures such a layer forms rapidly, however, this surface layer does not offer suitable corrosion protection (as say, an oxide layer does to Al or Cr) on the basis of: N The oxide/hydroxide layer is soluble in most aqueous environments, and environments of high humidity. The primitive E ph diagram for Mg is seen in Fig. 1 (neglecting hydride formation 5 ). Mg readily dissolves to form Mg 2z, and the corresponding cathodic reaction is the hydrogen evolution reaction (this terminology is interchangeable with water reduction, because when water is reduced, hydrogen evolves). It is observed that Mg is prone to dissolution over a wide ph range extending from 22 to10?5 and only in highly alkaline conditions, magnesium hydroxide is insoluble and hence protective (i.e. a passive film). Mg completely loses its integrity in acidic conditions or in the presence of chloride, bromide, and sulphate ions, thereby placing some practical limits on widespread Mg use 6 to alkaline conditions or situations where suitable coating systems may be implemented. ß 2015 Institute of Materials, Minerals and Mining and ASM International Published by Maney for the Institute and ASM International Received 11 July 2014; accepted 2 October 2014 DOI / Y International Materials Reviews 2015 VOL 60 NO 3 169

3 Labels 0, 22, 24, and 26 are the log of soluble ion activity for the indicated lines, and labels a and b show conditions of stability for water and its decomposition products, hydrogen and oxygen respectively. Hydride formation has been neglected. 5 1 E ph diagram for the Mg in water (H 2 O) system at 25uC adapted from Ref. 8 N Mg surface layers only incompletely cover the metal surface and are highly defective arising from a Pilling Bedworth ratio,1; meaning the ratio of the volume of the surface oxide/hydroxide to the underlying (hexagonal) metal is not suited for complete coverage. 7 N The highly electronegative potential of Mg and its alloys renders Mg-alloys prone to galvanic and microgalvanic corrosion (i.e. internal corrosion between the micro-constituents of the alloy), along with the possibility of corrosion in deaerated environments; because oxygen is not required for the cathodic water reduction reaction that predominates at such negative potentials. As a result, thermodynamic stability of Mg is not considered to be the key issue in determination of Mg alloy corrosion, with the kinetics of corrosion determining the ultimate corrosion rate. This review will thus focus on the factors that determine, and the extent to which they dictate, the kinetics of corrosion. The introduction of various alloying elements into Mg can improve some vital mechanical properties and thus extend the range of potential engineering applications, however, this has an attendant influence on the corrosion of Mg. In regards to Mg-alloys, the main factors in controlling corrosion rate are chemical composition and microstructure, hence appropriate attention is given to the development of Mg microstructures in this review. The solid solubility of elemental additions in Mg will dictate the presence of any second phases or impurity particles. More so than in other commercial metal systems, the solid solubility of an alloying element is a key factor in determining its effect on the properties of Mg; as the majority of elements have either no solubility, or very limited solubility in Mg. This is an interesting scenario, as there are concomitantly a large number of elements that will not form intermetallics with Mg, such that when added to Mg a phase separation occurs to produce two pure metals (i.e. Mg and Fe in the case of the Mg Fe system). 9 The solid solubility depends on factors such as relative atomic size, valency and electronegativity as well as similarity in crystal structure. A certain liquid solubility is required to form a homogenous solution at the alloying temperature. Table 1 summarises maximum solid solubility values in Mg based on binary systems 9 13 and categorising elements according to their solubility. The elements shown in white cells represent elements with solid solubility (i.e. to levels.0?5 wc), the elements shaded in orange are defined as slightly soluble (0?05 0?5 wt-%) and those in grey are insoluble (i.e.,0?05 wt-%). Several elements also have an unknown solubility (green shading). The table also provides the values of the atomic radius of the elements for general information. A convenient way of thinking about the insolubility of elements in Mg is to define the following three scenarios: 1. Complete and full insolubility: Here no solid solution is formed and no Mg-intermetallic forms with the particular alloying element. This situation is typical of Fe, Mo and Nb. When such elements are present, they form second pure metal phases, with no mutual solubility. This situation causes dramatic corrosion No solubility, but the formation of a Mg-intermetallic: This is typically what is seen with Si (and Cu, Co, Ni). In such cases an Mg 2 X (where X5Si, Cu, Co, Ni) intermetallic will form. These compounds have no room temperature solubility and thus the retention of a homogenous (single phase) microstructure is not possible. Elements that form the Mg 2 X intermetallic are often very problematic from a corrosion perspective, as they enhance the cathodic reaction because of the large exchange 170 International Materials Reviews 2015 VOL 60 NO 3

4 2 Typical Mg-alloy microstructures, all compositions given in wt-% a Mg 0?4Zn 0?05Sr in the rolled condition. This alloy has alloying additions below the respective solubilitylimitsandisasolidsolutionalloywithuniformmatrix composition. The image is an optical micrograph following etching in glycol solution. 27 ; b AZ31 (Al 3Al 1Zn 0?4Mn) as sectioned from an extruded ingot. Backscattered electron imaging indicates that this is close to a solid solution alloy, however, as the Al content is approaching the solubility limit and solute enrichment is observed, together with fine second phase particles. This alloy also includes constituent Al x Mn particles, typical of scenarios with Mn additions; c Mg 6La as highpressure die cast. Backscattered electron imaging reveals that the addition of La well above the solubility limit results in a large volume fraction of Mg 12 La phase. 28 current density of metals such as Cu and Ni. 18 Special attention is nominally given to exclude Si from commercially produced pure Mg. 3. Some solubility with no Mg-intermetallic: It is possible for elements such as Zr. In such cases, the alloying element will enter the solid solution to a limited extent, after which any further alloying will result in a separate, pure phase of the element. Again, such phases are problematic from a corrosion perspective. 19,20 The limited solubility of most elements in crystalline Mg results in only minimal changes to the electrochemical potential of the Mg (a) phase. As such, the majority of Mg-alloys have a potential in the close vicinity of about 21?55 V SCE, and essentially all are below 21?4 V SCE. 21 During open circuit exposure, the difference in electrochemical characteristics between the Mg matrix (a) phase and any second phase (be it a precipitate or impurity/insoluble particle) can be defined as the local current density sustained between microstructural constituents at the iso-potential of the alloy. These locally sustained current densities dictate whether any microgalvanic coupling contributes to corrosion or not. To orient the reader with the typical microstructures observed in Mg alloys, a selection of alloy microstructures are seen in Fig. 2 for alloys with increasing alloy content. What is generally observed is that increasing alloy content leads to a rapid increase in the volume fraction of second phase. The extent and influence of the second phase will depend on the element added and its relative solid solubility. The variation in the volume fraction of second phase is evident in the examples seen in Fig. 2 with microgalvanic corrosion in Mg-alloys almost exclusively explained by the accelerated cathodic activity which arises from intermetallic particles (IMPs) able to support higher reduction reaction kinetics compared to pure Mg. 16 The fact that most elements have a greater efficiency at supporting the cathodic reaction stems from Mg having a low exchange current density, which is a physical characteristic of Mg itself. Consequently, IMPs with higher catalytic activity result in enhanced anodic dissolution of the Mg matrix. One well documented exception to this was reported by Kirkland, who indicated that Mg 2 Ca has a more negative electrode potential, such that it would be anodically polarised in the Mg matrix and undergoes extremely high dissolution rates at the expense of Mg. 22 The magnitude of the effect of IMPs on acceleration of anodic or cathodic kinetics depends on their respective chemistry (and hence electrochemistry) and volume fraction, so that the larger the volume fraction of IMPs in the alloy, the greater the impact on increasing the corrosion rate of the alloy. 23,24 The presence of impurities (which is the generic name given to insoluble elements) such as Fe, Ni, and Cu alters corrosion rates dramatically, even in the ppm range. 25,26 For example, islands of pure Fe or pure Cu in the matrix of Mg create a mixed potential that assures the Mg is polarised anodically (dissolves) and the pure insoluble elements can sustain the cathodic reaction at very high rates, as they are significantly polarised from their nominal corrosion potential values (often by.1 V). Arguably, the biggest single contribution to date with respect to moderating the corrosion rate of commercial Mg is the scavenging ability of Mn in conjunction with Al. 25,29 By virtue of Mn additions to the Mg Al systems, Fe is sequestered in an AlMnFe intermetallic, and hence the extent of microgalvanic coupling is dramatically reduced. 14 The moderation of corrosion/scavenging International Materials Reviews 2015 VOL 60 NO 3 171

5 3 A compilation of the typical potentiodynamic polarisation response of common Mg-alloys. The electrolyte used was 0?1M NaCl and a polarisation scan rate of 1 mv s ,33 35 Pure Mg was 99?9% purity effect in Al-free alloys remains largely unknown and is an area of current research 20 which will be significant in the development of creep resistant (i.e. Al-free) Mg-alloys. Corrosion kinetics of Mg-alloys Corrosion of Mg-alloys is a kinetic issue, which will depend on the alloy composition, together with the component design (i.e. if welded or joined) and environmental factors (such as ph, [Cl 2 ] and temperature). 30,31 The half-cell reactions for Mg corrosion are given in equations (1) and (2). Reaction (1) is Mg dissolution that produces electrons consumed at the cathode (2) resulting in generation of hydrogen gas. As will be described in the subsequent section, this process can be sustained during anodic polarisation of Mg 32 Mg?Mg 2z z2e { (1) 2H 2 Oz2e {?H 2 z2oh { (2) The relative rates of the above reactions, and their variation as a function of alloy potential, can differ significantly, as is shown in Fig. 3, which reveals the typical potentiodynamic polarisation response of the most common commercial alloys and pure Mg. The corresponding composition of the alloys shown in Fig. 3 is given in Table 2. Generally, the potentiodynamic polarisation response provides information about the relative rates of the anodic and cathodic kinetics for different alloys. The method can also provide an instantaneous corrosion rate (termed the corrosion current density i corr ) as approximated by a Tafel-type fit, usually from the cathodic polarisation data. This has been verified as a reasonable approximation to the corrosion rate as it relates to the rate obtained from other methods. 36 However, of key relevance here is that the polarisation technique, in spite of not being a long term predictor of corrosion rates owing to its instant nature, is an important tool in rationalising alloy behaviour as it is the most effective at providing information on the effect of a particular element on anodic or/and cathodic reaction rates. An archetypal demonstration of this was provided by Kirkland, who revealed that Mg 10Zn ( wt-%) and Mg 5Ca ( wt-%) show the same corrosion rates from mass loss tests, however, the origin of corrosion rates could only be ascertained from polarisation testing, which revealed that Zn enhanced cathodic kinetics (with little influence on anodic kinetics), while Ca enhanced anodic kinetics (with little influence on cathodic kinetics). 37 In Fig. 3, the representative polarisation curve for pure Mg is seen as a thick dark line. It is immediately obvious that there is a range of movements in the relative anodic and cathodic curves depending on the respective alloying additions. One striking feature, however, is that most of the commercial alloys have significantly more rapid cathodic kinetics than pure Mg, with the exception of WE54 and AE44, which both contain appreciable rare earth (RE) additions, yielding similar cathodic kinetics to pure Mg. Other features to note include that AM60, AZ31, AZ91 and ZE41 showed lower anodic kinetics than pure Mg, while AE44 and WE54 showed enhanced anodic kinetics relative to pure Mg. These simple descriptions are being made to orient the reader with the form and signatures in such data, while the effects are treated in a holistic treatise further below. The data in Fig. 3 provide the classic manifestations for Mg-corrosion kinetics, which includes 2 Elements that shift the anodic branch to lower rates nominally do so by solid solution doping, which is efficiently achieved by Al and Zn. However, this is countered by a corresponding increase in the cathodic kinetics. Both of these kinetic changes lead to ennobled E corr values. 2 Significant alloying additions (which in comparison to other metallurgical systems can in fact be modest) will lead to second phase development and hence, nominally more rapid cathodic kinetics (and hence corrosion rates). 2 It can be seen that Al-containing Mg-alloys tend to show the lowest anodic kinetics and the fastest cathodic kinetics. 2 Significant alloying with RE elements can reduce the E corr and enhance anodic kinetics on the basis that REs are themselves active elements. 2 Common alloying additions do not result in lower cathodic kinetics being imparted to Mg. Even REs 172 International Materials Reviews 2015 VOL 60 NO 3

6 4 Typical potentiodynamic polarisation response of selected experimental Mg-alloys. The alloy composition is denoted in the plot, while the pure Mg was 99?9% purity 50 tend to serve as local cathodes under free corrosion conditions Alloying additions nominally only alter E corr within a window of y100 mv. This, however, does not mean that corrosion rates cannot be spread over several orders of magnitude (as reported in the literature 39 ) since large changes in kinetics can be achieved on the basis that Mg is weakly polarisable. This means that only small changes in potential can result in large deviations of current (owing to a very low anodic Tafel slope in Mg systems, as opposed to say, passivating systems). In relation to the aforementioned alloys (Fig. 3), there is ample information in the literature concerning the microstructures developed, and hence our main focus herein is to present their corrosion kinetics in respect to the main microstructural constituents. Mechanical properties will not be covered herein, however, the authors note that the group of high(er) strength Mg-alloys (AE44, WE54, and ZK60) possess the presence of precipitates including Al 11 RE 3, Al 10 RE 2 Mn 7,Al 2 RE, 47 Mg x RE y, 44 and MgZn 2, 48 with improved mechanical properties at the expense of reduced corrosion performance. 49 Taking the finite commercial alloys as examples, the microstructural, and hence electrochemical, heterogeneity requires the quantification of microstructure as a requisite effort in the rationalisation of corrosion performance. Before discussing the unique effect of specific elements, a set of polarisation curves depicting the corrosion kinetics of selected experimental alloys is shown in Fig. 4. The selection of the data in Fig. 4 was deliberate, to indicate instances where anodic reaction rates can be rather markedly increased. This phenomenon occurs for two reasons. The first (Mg Ca) is when the alloying element is more reactive (in terms of dissolution kinetics) than Mg, which is the case when Ca is alloyed with Mg. 22 The second is when an alloying element triggers anodic activation. This is the case for Mg Sn. The anodic activation is a terminology used for elements that are able to disproportionally enhance anodic kinetics, when this phenomenon may not be expected on the basis that the alloying element is (in its pure form) less active than Mg. This phenomenon is common to Al alloys with heavy metals such as Pb and Sn, which are nominally low melting point elements that segregate to the surface disrupting the surface film and allowing dissolution en masse. The phenomenon is also observed in Mg (Fig. 4) and occurs with Sn (also with Pb, Sb, and Zr as discussed below). However, this phenomenon has not been studied in detail, perhaps because of the fact that it has no commercial relevance to date. Ironically, however, the low melting point elements that lead to anodic activation in Mg are also part of the family of (few) elements that have lower cathodic kinetics than Mg (Pb, Sn, and Sb have a low exchange current density, which is a physical property of each element 18 ). Unfortunately, the phenomenon of slowing cathodic reaction kinetics (evident in Fig. 5 for the Mg Sn alloy) is overwhelmed by the anodic activation for reported alloys to date. In the other examples provided that include ternary additions to Mg Al alloys, the overriding manifestation is the enhanced cathodic kinetics imparted by Al, while any beneficial effects in reduced anodic kinetics are not significant for the cases presented. In order to provide a synopsis of the electrochemical kinetics of constituents and precipitates which can populate Mg-alloys, a selection of potentiodynamic curves collected by Südholz 16 is presented for a finite number of Mg-based intermetallics (Fig. 5). Such data are not inclusive of the entire spectrum of possible Mgbased intermetallics, as the collection of such data necessitates the production of sufficiently large intermetallics for interrogation via micro-electrochemical testing. 23,51,52 What is observed from Fig. 5 is that the IMPs are nominally noble with respect to Mg, with the exception of Mg 2 Ca. This, therefore, indicates that such intermetallics, if in an Mg matrix, will be polarised cathodically, and will remain cathodically protected at the expense of matrix dissolution. Of most interest and importance, however, is that the rates of reaction on the International Materials Reviews 2015 VOL 60 NO 3 173

7 5 Potentiodynamic polarisation curves for intermetallic phase present in, and common to, the Mg-alloys systems. Date presented along with pure Mg for comparison (40 ppm by wt. Fe), in 0?1M NaCl. From Ref. 16 different intermetallics are distinctly different, and hence the chemical entity of the intermetallic is of great significance to the ultimate corrosion rate of the alloy, which they occupy. Authors emphasise the salient, but key point, that it is the kinetics of reaction on intermetallics that relate to their potency to support localised corrosion, and not the relative potential variation from Mg. Potential difference does not necessarily translate to kinetic effects, as seen in Fig. 5. Unfortunately, such a consolidated presentation of intermetallic electrochemistry does not exist for a range of environments (i.e. ph or [Cl 2 ]), with only reportage in 0?1M NaCl to date. In that vein, it is noted that the majority of reports for Mg corrosion exist in near-neutral chloride environments. This is considered to be reasonable on the basis that most atmospheric exposures can be approximated by dilute chloride solutions. The most common electrolytes appear to be 0?1 and 0?6M (i.e. 3?5 wt-%) NaCl. Some exceptions include works in more dilute electrolytes 53 and recent work in concentrated chloride electrolytes on the basis that drying aerosol droplets will lead to [Cl 2 ] between 1?0 and 5?0M. 54 Other environments in which testing has been conducted include sulphate solutions 53,55 and solutions saturated with Mg(OH) Additionally, in the field of Mg as candidates for bioresorbable implants, testing is executed in a variety of physiological environments and temperatures. 57 The ph dependence of pure Mg corrosion has been recently extensively covered by Ralston, 58 and readers are directed to that dedicated work, while the focus in this review remains on alloying effects. Focusing on the electrolyte for which the most data have been reported for a variety of alloys (0?1M NaCl), Fig. 6 represents the experimental data collected and reported over the past decade for a range of Mg-alloys, commercial and experimental, as subdivided by basic chemistry. This compilation includes data from 108 alloys and was limited to literature in which the combination of electrochemically determined corrosion current and gravimetrically determined corrosion rate was reported. Confining the representation to such 6 Corrosion rate expressed as corrosion current density (i corr ) v mass loss rate (mass loss determined from 1 day exposure in 0?1M NaCl). Inset shows same data, with logarithmic axes. Data from Refs. 16,17,19,27,28,33 35,59, International Materials Reviews 2015 VOL 60 NO 3

8 7 Upper plot shows the ICP-AES signal registered (i Mg2z ) together with current signal (j), along with the signal from different potential steps in the lower plot. The applied potential steps from left to right, are 21?4, 21?3, 21?2, 21?0, 20?8, 20?5, and 20?3 V (Ag/AgCl) applied each time for 240 s in 30mM NaCl electrolyte. From Ref. 53 alloys was deliberate so that the readers could ascertain (i) the relative corrosion rates seen in Mg alloys in units to which they can compare their own data, be it corrosion current or mass loss rate, (ii) to indicate that there is a correlation between tests, in that generally, alloys with a high corrosion current density will also have a high rate of mass loss, and (iii) to indicate the spread of corrosion rates which may arise as a function of alloying. This latter point merits some comment, as the authors can observe differences in over an order of magnitude in both the corrosion current density and mass loss rate. This spread is rather remarkable, and is perhaps even unique to Mg, in that alloying can have such a marked influence on corrosion kinetics for the same environment. A large spread of corrosion rates was also shown in selected alloys tested in physiological media, 39 however, the authors note that the present data (Fig. 6) only report results for 0?1M NaCl at room temperature. To orient readers, the authors note that low corrosion rates (typical of passivating systems, including Al-alloys) are nominally y1 ma cm 22 ; noting that the reported rates for the Mg alloys presented are much higher than this. Regarding the practical measurement and the measurement execution for quantifying of Mg corrosion electrochemically, gravimetrically or volumetrically (i.e. via hydrogen collection), a study by King et al. abridges and highlights recent developments. 54 Elaboration of the effects seen in Fig. 6 are rationalised from the descriptions of unique additions in Influence of alloy composition on corrosion of Mg-alloys section. Update on the mechanistic aspects of Mg corrosion The mechanism of Mg corrosion is a topic that has been studied throughout the past century. 53,61 68 While the corrosion of Mg is a process occurring at open circuit, it is of contextual relevance to mention that Mg displays a phenomenon known as the negative difference effect (NDE). The so-called NDE describes the phenomenon by which the amount of hydrogen evolved from an Mg electrode increases as the electrode potential is polarised anodically to more noble potentials. This is counterintuitive to conventional electrochemical wisdom on the basis that increasingly anodic polarisation should result in lower rates of the cathodic reaction (which is responsible for the evolution of hydrogen as per equation (2)). Somewhat controversially, in one study, the origin of this phenomenon was proposed by Petty 63 to be a result of the formation of unipositive Mg (Mg z ), which was later popularised by Atrens and co-workers, 65,66,69,70 who purported that superfluous hydrogen arises from a chemical reaction of Mg z with water at some unknown distance from the metal surface. The claims by Atrens and co-workers have led to somewhat of a nuanced understanding of Mg dissolution processes, largely as the proposed Mg z theory has no reasonable proof of its existence 66,70 74 and most recently the Petty experiments were refuted. 75 While the imagined existence of Mg z has captured attention and consumed scientific effort, it has no influence on the content of this review. At the time of writing this review, a large number of publications have emerged that provide significant detail regarding the mechanisms of Mg dissolution as well as physical descriptions of the NDE. Of those studies, the use of inductively coupled plasma atomic emission spectroelectrochemistry (AESEC) definitively demonstrated that Mg dissolution occurs with an n52 stoichiometry (i.e. Mg 2z ) in both sulphate and chloride solutions. 53 This observation was recently validated in an independent study using on-line mass spectroscopy. 76 The results of AESEC testing are seen in Fig. 7, which show excellent agreement between applied current and the Mg 2z dissolution current. This echoes the classic notions put forth by James and Straumanis, 64 albeit with modern International Materials Reviews 2015 VOL 60 NO 3 175

9 8 Schematic representation of the electrochemical impact of alloying elements studied (data extracted from the papers cited in Influence of alloy composition on corrosion of Mg-alloys section). Plot depicts the ability of alloying additions to modify anodic or cathodic kinetics (or both), leading to changes in the resultant corrosion rate (i corr ), along with changes in corrosion potential (E corr ). C S represents the solid solubility analytical support. With the Mg valency accounted for, the notion of superfluous hydrogen evolution during anodic polarisation (i.e. a parasitic cathodic reaction) was proposed by Frankel 67 to be the result of the cathodic reaction being catalysed by Mg dissolution, which is consistent with the classic works noting enhanced reducing ability 17 and has also recently shown by scanning vibrating electrode (SVET) measurements. 32 The atomistic origins of this enhanced catalytic activity are the goal of active studies. 77 It is, however, emphasised, that the alloying effects described herein are not dependent on the elucidation of the origins of enhanced catalytic activity in a first order sense, and hence this section will not be elaborated further in the present review. Influence of alloy composition on corrosion of Mg-alloys An attempt has been made to present an abridged summary of the influence of specific elements upon the corrosion kinetics of Mg. In many cases, the effect of an element is over and above that of an existing system (i.e. the effect of Mn, on say, the Mg Al system). However, such scenarios are described in words and the relevant references provided in order to yield a unique consolidation of information not available elsewhere. Aluminium (Al) AlisthemostcommonadditiontoMg,byvirtuethatitis (relatively) cheap, light, soluble, and considerably improves strength of Mg (i.e. from y70 to y250 MPa). 78 Additions of Al below the solubility limit tend to reduce theanodickineticsofmg.theadditionofaltomg ennobles the corrosion potential, with Mg Al alloys tending to be ennobled by y100 mv SCE compared to pure Mg, in Cl 2 environments. 65,66,79 81 The Mg Al alloys nominally have the lowest corrosion rates of the commercial Mg-alloys, 13 particularly AZ31 (Fig. 6), which gives an optimum balance between physical and corrosion properties for Mg-alloys to date. Although Al is soluble to y12 wt-% in Mg, this is dependent on temperature, and the room temperature solubility is much lower, with alloys richer in Al than AZ31 showing the presence of b-phase (Mg 17 Al 12 ). The extent of b-phase will depend on the time temperature history of the alloy, with the relative proportion of b-phase for the same given composition also varying rather widely depending on alloy cooling rate or subsequent heat treatment (i.e. whether sand-cast, die cast, or high-pressure die cast) Above y3 wt-%,al additions tend to enhance the cathodic reaction, which although further 93 ennobles E corr, is associated with an increase in corrosion rate. In open circuit conditions, b-phase serves as a local cathode. As such, cathodic kinetics of the following alloys increase in the order of AZ31,AZ61,AZ91#AM60. Al has been reported to increase the susceptibility to stress corrosion cracking in cases where b-phase exists in appreciable fractions. 78 It is also noted that the AZ alloy system undergoes solid-state phase changes at moderate temperatures, and hence its utility in more demanding modern applications is leading to the proliferation of Al-free Mg alloys. Silver (Ag) The effect of Ag additions on the corrosion of Mg was studied by Hanawalt more than 70 years ago, who suggested a tolerance limit for Ag in Mg to be y0?5 wt-%. 15 Above this concentration the mass loss rate increased monotonically from y1 mgcm 22 day 21 for 1 wt-% Ag, to y12 mg cm 22 day 21 for 5 wt-% Ag. In contrast, the effect of Ag on accelerating the corrosion of Mg is therefore as dramatic as that as Ca on Mg. 15 In more complex alloys, however, lower level quaternary additions of Ag (to,y2 wt-%) in the Mg RE Zr family of alloys are reported, because Ag improves the age hardening response and thus provides substantial increment in strength. 13 However, it is prudent to note at this point that although Ag has been added to the high strength Mg alloy family (which can be considered ultra 176 International Materials Reviews 2015 VOL 60 NO 3

10 Table 1 Maximum solubility of elements (at any temperature) in magnesium presented in wt-%: atomic radius of elements also given in picometeres 8 12 Atomic radius of elements also given in picometeres International Materials Reviews 2015 VOL 60 NO 3 177

11 9 The relationship between corrosion rate and grain size a pure Mg for current densities measured in 0?1M NaCl 225 and b for ZK60 processed by a combination of extrusion and equal channel angular pressing (ECAP) 219 high strength when considering specific strength), this has not been an alloy design with corrosion considerations in mind. None the less, addition of elements such as Ag that have the ability to modify nucleation/precipitation is an avenue that can be exploited. For example, trace additions of Ag (y0?1 wt-%) to AZ91 were recently shown to stimulate an interaction between Ag and the Mg 17 Al 12 phase, providing an increment in hardness without any significant loss of corrosion properties, i.e. a slight increase in cathodic kinetics is counterbalanced by decrease in anodic kinetics. 33 High concentrations of Ag led to increased corrosion rate because of co-formation Mg 4 Ag precipitates that can stimulate microgralvanic corrosion. 34,94 In addition, micro-additions of Ag (,0?5 wt-%) to Mg Zn Ca systems facilitate grain refinement and in cast and wrought alloys Back scattered SEM images (a, c, e) of polished specimens and the corresponding optical profilometry images following 330 min of exposure to 0?1M NaCl for ZK60: Each micrograph corresponds to different extents of processing and imaged on the ND TD plane for a and b initial, unprocessed condition; c and d intermediate, extrusion processed condition; and e and f the extrusionzecap processed condition International Materials Reviews 2015 VOL 60 NO 3

12 11 (left) Corrosion rates as a function of approximate crystallinity for two Mg-based bulk metallic glass (BMGs), Mg 65 Cu 25 Y 10 and Mg 70 Zn 25 Ca 5 measured from potentiodymamic polarisation testing in 0?1M NaCl. The 0% crystallinity designated the fully amorphous condition, while the increase in crystallinity to the fully devitrified state was achieved by heat treatment. (right) corresponding SEM micrographs from specimens denoted on the left plot following 12 h immersion in 0?1M NaCl. It is observed that Mg 65 Cu 25 Y 10 in the fully amorphous condition (i) undergoes filiform-like corrosion, while in the fully crystalline form, undergoes localised corrosion (ii) that is driven by the chemical entity of microstructural features (active particle dissolution). In the amorphous condition, Mg 70 Zn 25 Ca 5 is rather uniformly attacked (iii), forming blisters, while in the crystalline condition (iv) undergoes localised corrosion that is driven by the chemical entity of microstructural features (active particle dissolution). The figure is adapted from Ref. 107 Arsenic (As) The effect of As on Mg corrosion was explored on the basis of the potential to kinetically limit the cathodic reaction, with As known to be a cathodic poison preventing hydrogen recombination 96 and restricting the completion of the reaction in equation (2). The metallurgical influence of As has been recently reported by Birbilis et al., 97 whereby relatively small additions of sparingly soluble As (y0?37 wt-%) result in formation of Mg 3 As 2 phase. In spite of this second phase presence, a decrease in Mg corrosion rate (five times lower than that of pure Mg) was determined by simultaneous hydrogen collection and 7 day mass loss tests, as well as lower cathodic kinetics determined by potentiodynamic polarisation. The reduction in corrosion rates was associated with the ability of As to slow down the cathodic reaction rate of pure Mg. Bismuth (Bi) The addition of Bi to Mg Al alloys has the effect of refining Mg 17 Al 12 and is accompanied by co-formation of needle-shaped Mg 3 Bi 2 particles, even for Bi concentrations below the solubility limit. 98 When added to AZ91, Südholz reported that the presence of Mg 3 Bi 2 ennobles the corrosion potential of the alloy 34 and is shown to bring about the acceleration of both the anodic and cathodic reactions rates. 99 In spite of the negative influence on corrosion, Bi-containing particles seem to have a positive effect on enhancing tensile and creep properties via restricting grain boundary sliding. 100 Calcium (Ca) In the binary context, Ca additions in low levels (less than 0?35 wt-%) are essentially inert. 101 In general, however, Ca additions dramatically increase corrosion rates in Mg when added near to, or above the solubility limit (of y1?35 wt-%). Ca-containing Mg alloys result in exceptionally high corrosion rates. According to Hanawalt, the mass loss rate of Mg Ca binary alloys increases from y1 mgcm 22 day 21 for 0?5 wt-% Ca to about 6 mg cm 22 day 21 for 5 wt-% Ca. This is almost six times higher than the corrosion rate of Mg 5 wt-% Al. 15 In addition to possessing the highest corrosion rates of any candidates for structural alloys ever reported, Ca-containing Mg alloys dissolve to yield a voluminous corrosion product that is insoluble. 39 Owing to the biocompatibility of Ca, however (along with Zn), Mg Zn Ca alloys are being explored as potential Table 2 Precise chemical composition of commercial Mg-alloys for which polarisation data are given in Fig. 3, as determined by inductively coupled plasma-atomic emission spectroscopy (ICP-AES) 28,33 35 Mg Al Zn Mn Zr Fe Gd Y La Ce Nd Alloy wt-% AZ31 Bal AZ91E Bal AE44 Bal ,0. 005, y2 y1 y1 AM60 Bal ZE41 Bal ,0. 005, ZK60 Bal ,0. 005,0. 005,0. 005,0. 005, WE54 Bal International Materials Reviews 2015 VOL 60 NO 3 179

13 biomaterials. 22 More recently, very low levels of functional Ca additions are being explored in Al-free Mg-alloys, 102 in attempt to exploit the fact that Ca is one of the few soluble elements in Mg. Provided that Ca concentrations are kept below the solubility limit to avoid Mg 2 Ca formation, the corrosion rate will not sharply accelerate, while solid solutions of Mg Ca can display appreciable ductility Copper (Cu) The addition of Cu to Mg and Mg-alloys is generally avoided because of the insolubility of Cu, and the resultant formation of Cu islands in the microstructure. Cu is highly detrimental for corrosion, 78 rationalised on the basis that Cu has a high exchange current density, and hence an efficient cathode. The tolerance limit of 0?1 wt-% was established by Hanawalt, however, the limit is sharply reduced to 0?01 wt-% when Al and Mn are present in the alloy composition. 15 When added to AZ series alloys it forms an Mg Al Cu Zn phase, which is also a relatively strong local cathode resulting in local corrosion. 106 In general, Cu is avoided in the production of Mg-alloys, with measures in place to avoid Cu contamination and pick-up. Unfortunately, from a corrosion perspective, Cu is, however, a popular element in the formation of Mg metallic glasses, 107 where the influence of Cu on the resultant corrosion mechanism has been described as incongruent dissolution and discussed further below. Cerium (Ce) When added as a binary addition (i.e. Mg Ce) the formation of Mg 12 Ce occurs in increasing volume fraction with Ce additions, which can monotonically accelerate corrosion by enhancing cathodic kinetics. 24 In spite of the chemical reactivity of Ce, Mg 12 Ce is more noble than Mg, and Mg 12 Ce sustains the cathodic reaction at higher rates than Mg over the range of potentials typical of Mg-alloys. On the other hand, Ce additions to Al-containing Mgalloys can refine microstructure by formation of Al 4 Ce or Al 11 Ce 3 intermetallics. 108,109 These compounds are, however, more deleterious for corrosion than Mg 12 Ce. Rare earth elements are typically not combined with Alcontaining Mg alloys, with the exception of the commercial alloy AE44. In the more modern wrought series of Mg alloys, RE and Al are not combined, because of the ease of formation of Al x RE y compounds that reduce alloy ductility. Some reports suggest that Ce contributes to the surface film and stabilises Mg hydroxide. 110 Südholz, however, revealed via XPS analysis that RE elements were not seen to dope the surface oxide of Mg Al alloys. 28 The notion of oxide modification of Mg-alloys is an area of mixed reports, and one that is an open question. Irrespectively, elements with thermodynamic ability to replace Mg-oxides/hydroxide generally do not offer a wider window of passivity than Mg. 111 Erbium (Er) In favour of more cost effective RE elements, Er has not been well studied in Mg. In one study, however, Er, when added in 2 3 wt-%, was found to decrease corrosion rates in Mg 2 to 3Al (wt-%) alloys compared to commercial AM60 [nominal Mg 6Al 0?13Mn (wt- %)] when tested in borate buffer solution. 112 The assumption was made that Er facilitates an apparently enhanced protective effectiveness to the surface oxide giving rise to what was described as a pseudo-passivation effect by incorporation of Er into the Mg(OH) 2. However, there is still a lack of evidence to support this assumption. Furthermore, the testing of Mg in a buffer is likely to impose an unrealistic scenario of surface film evolution, which is nominally influenced by natural surface alkalinisation. Iron (Fe) Fe is the most common impurity in Mg-alloys. Fe contamination can be present because of the quality the starting pure Mg as well as because ofpickup from the casting process. Owing to its low solubility in Mg (y0?001 wt-%) Fe largely remains in its pure (body centred cubic) form. To date, the literature reports a number of tolerance limits for Fe, which is nominally restricted to as low a level as economically feasible. 13,113 The notion of an Fe tolerance limit dates back several decades, where the proposed limit of 150 ppm by weight was commonly accepted. 114,115 Recent works have focused more specifically on the role of Fe, as its ability to be scavenged by sequestration to AlMnFe particle formation is not possible in emerging Al-free Mg alloys. 17 The role of thermal history on the Fe tolerance limit has also been recently discussed on the basis of calculated phase diagrams. 116,117 The experimental validation and robustness of the related thermodynamic databases for Mg Fe remain open questions without further empirical work, particularly because it was seen that calculated phase predictions can vary from the structures characterised by electron microscopy. 17 Gadolinium (Gd) Of the elements that Mg may be alloyed with, Gd is unique on the basis that it has high solubility (.10 wt- %) in Mg, forming binary Mg Gd solid solutions over a wide range of compositions. Such solid solutions can be used as a basis for further loading of the matrix with solute elements in the case of wrought Mg alloys, or for carefully controlled precipitation processes in the family of Gd-containing precipitation-hardened Mgalloys In regards to this latter class of alloys, corrosion characterisation remains scarce. The addition of Gd in Mg Al alloys results in the precipitation of Al 2 Gd and Al Mn Gd phases that consumes Al and reduces the volume fraction of Mg 17 Al 12 phase. 122 These phases may contribute to a minor decrease in the total rate of cathodic reaction. 123 This tendency is commonly observed for Mg Al alloys modified with REs. However, the effect of Gd is more complex because it leads to more heterogeneous microstructures and diminished Al content in the Mg matrix. Therefore long term corrosion tests show a greater corrosion for Mgalloys containing Gd. 124 Other studies have also indicated that the influence of Gd on corrosion is generally defined as detrimental. 125 Mercury (Hg) Alloying with Hg is nominally only reserved for the production of high voltage anode materials for sea water batteries. 126 The Mg 3 Hg phase can form in binary Mg Hg alloy at concentrations of Hg at around 4 6 wt-%. 126,127 Thermomechanical processing of Mg Hg alloys can promote the redissolution of Mg 3 Hg into the Mg matrix, which subsequently retards microgalvanic corrosion and thus lowers corrosion rates. 128 Feng 180 International Materials Reviews 2015 VOL 60 NO 3

14 compared the electrochemical response of Mg 5Hg (wt- %) and Mg 6Hg (wt-%) in 3?5% NaCl solution, and concluded that Hg accelerates anodic reaction rates and leads to very negative corrosion potentials (E corr, Mg 5Hgy22?3 V SCE ), in fact representing the most negative corrosion potential reported of any Mg alloys. This was concomitant with high corrosion rates with i corr of Mg 5Hg being y27 ma cm 22, compared to y20 ma cm 22 for pure Mg. 127 The role of Hg in structural Mg alloys has not been explored. Holmium (Ho) The limited research to date regarding Ho has suggested that a Ho addition of 0?24 and 0?44 wt-% can decrease the rate of corrosion in an Mg Al alloy (AZ91D) by decreasing the volume fraction of Mg 17 Al 12 owing to the formation of Ho-containing intermetallic phases with Al. In one study, inspection of the reported data indicate a minor decrease in the cathodic reaction rate, along with as much as 10 times lower mass loss rates compared to that of the base Mg Al alloy. 129 This, however, is likely to lead to decrease in alloy strength and may be manifest as the same outcome as lowering of the alloy Al content. It has been posited that Ho presumably contributes to more uniform and compact layer of corrosion product because of the higher content of Al present in the surface film; however, more evidence is needed to support such assertions. Lanthanum (La) As a binary addition, La forms the Mg 12 La phase 24 when added above its solubility limit. This phase is a more efficient cathode than pure Mg leading to an overall increase in corrosion rates (from 10 to 30 ma cm 22 at concentrations varying from y0?5 to y3 wt-% La). 28 However, as an elemental addition to Mg Al alloys, La modifies the alloy microstructure (in a manner analogous to Ce), forming needle-like Al La compounds and alters Mg 17 Al 12 from discontinuous to continuous with a fine polygonal-shape. 130,131 Liu et al. reported a drop in corrosion rate (i corr ) of AZ91 from y603 to y3 ma cm 22 in 3?5% NaCl solution when 0?5 wt-% of La was added. This phenomenon was attributed to a refined microstructure and allegedly more protective corrosion product film. It should be noted, however, that the reported i corr of the base AZ91 was much higher than any other reports for AZ91; while the evidence for any La in the surface film was also not provided. In a separate study, Südholz also tested the effect of 0?3 wt-% of La on corrosion rate of AZ91E. In that study, however, the reduction in the i corr value was only by 0?3 ma cm 22 (from 7 ma cm 22 of AZ91E to 6?7 ma cm 22 of AZ91E 0?3La). 28 It is generally observed that any excess addition of La, above the solubility limit, reduces corrosion resistance because coarse Al La precipitation in Mg Al alloys. 130,132 In the context of Mg RE earth alloys, containing only Mg and REs, the presence of La in concert with Ce and Nd will lead to a different morphology and microchemistry of the second phase; transitioning from a lamellar eutectic to a divorced eutectic. 133 The relative proportions of La, Ce and Nd are important in the ultimate corrosion rate, as depicted in a combined electrochemical and microstructural study. 133 It was determined that Nd is less detrimental than La (or Ce) on the corrosion of Mg. Lithium (Li) Mg Li alloys because of their ultra-lightweight and superplastic behaviour remain promising materials for many critical engineering applications, 134 in spite of limited use to date. The microstructure of binary Mg Li alloys depending on the Li content may consist of either a single hexagonal a (HCP) phase, at 0 5?7 wt-% Li; two phase a and body-centred b (HCP and BCC) at 5?7 10?3 wt-% Li; or a single b (BCC) phase at Li concentrations greater than 10?3 wt-%. 135,136 Therefore, high additions of Li (.10?3 wt-%) can produce a uniform b phase that is cubic. Li is the only researched element able to impart a change in the crystal structure, which has major implications for ductility. Early work by Frost in 1955 and a later published report from Battelle in 1964 indicate superior corrosion resistance of binary Mg 11Li alloy compared to the alloys with lower Li content, for example, the mass loss rate of the alloys measured in 3% NaCl solution (at 35uC) over 8 days was Mg 2Li (4?92).Mg 4Li (4?44).Mg 9?3Li (0?77).Mg 11Li (0?57), mg cm 22 day ,137 The greater corrosion resistance of Mg 11Li alloy was attributed to the BCC structure. More recent studies have also validated aspects of the early research by reporting that Li additions below the bcc transition increase the corrosion rate of Mg (from 19 ma cm 22 for pure Mg to 45 ma cm 22 for Mg 8Li 138 ). Li was also seen to increase the corrosion rate of an Mg-alloy, from 12 mg cm 22 h 21 for AZ80 to about 200 mg cm 22 h 21 for AZ80 1?95Li. 139 In addition, over long exposure periods the corrosion rates of Mg Li alloys are more likely to remain constant or decrease much more slowly in contrast to AZ31 for instance, for which corrosion rate decreases with time. 135,138 Although some information on the effect of Li on bulk Mg corrosion exists, there is still a gap in regards to electrochemical effect of Li on corrosion kinetics of Mg and its alloys. Manganese (Mn) The influence of Mn additions on Mg has been studied in detail as far back as almost a century. 15,25 Its role is therefore well characterised and generally well understood. As a binary addition to Mg, Mn has been reported to show no significant effect on corrosion rate at concentrations up to y5 wt-%. 15 Mn is essentially always added to the Mg Al and Mg Al Zn systems (i.e. the commodity Mg-alloys). Compounds such as Al 8 Mn 5,Al 6 Mn or Al 4 Mn are formed, and the addition of Mn helps to reduce corrosion rate via the incorporation of lowly soluble metals in Al Mn intermetallic phases. The classic example of this behaviour is the incorporation of Fe into Al Mn Fe. 140 It has been reported that the amount of Mn to effectively mitigate the effect of Fe must be in the range that satisfies the maximum Fe/Mn ratio of 0?032, whereas greater ratios will sharply increase corrosion rates. 6 Although Mn is known to improve corrosion resistance, a low ratio between Al/Mn leads to higher cathodic potency. Therefore corrosion rate gradually increases with an increase in Mn content, 25,141 independent of the Fe sequestration aspect of Mn. More recently, the work of Gandel et al. reports the corrosion behaviour of Mg Mn binary alloys where the corrosion rate i corr reduces from 40 to about 22 ma cm 22 in 0?1M NaCl with Mn content ranging from 0?2 to 2 wt-%, International Materials Reviews 2015 VOL 60 NO 3 181