Acta Materialia 133 (2017) 90e99

Contents lists available at ScienceDirect

Acta Materialia
journal homepage: www.elsevier.com/locate/actamat

Full length article

Crystallographic effects on the corrosion of twin roll cast AZ31 Mg

alloy sheet
S. Pawar a, T.J.A. Slater a, T.L. Burnett a, X. Zhou a, G.M. Scamans b, c, Z. Fan b,
G.E. Thompson a, P.J. Withers a, *
a
School of Materials, The University of Manchester, Manchester, M13 9PL, UK
b
The EPSRC Centre e LiME, BCAST, Brunel University, Uxbridge, London, UB8 3PL, UK
c
Innoval Technology Ltd., Beaumont Close, Banbury, Oxon, OX16 1TQ, UK

a r t i c l e i n f o a b s t r a c t

Article history: The inuence of crystal orientation on the corrosion behaviour of twin roll cast (TRC) AZ31 Mg alloy
Received 21 February 2017 sheet has been investigated using correlative microscopy and immersion testing in 3.5% NaCl. Corrosion
Received in revised form initiates as a consequence of the microgalvanic coupling between cathodic intermetallic phases,
12 May 2017
including sub-micron sized, rosette-shaped Al8Mn5 particles, b-Mg17(Al,Zn)12 phase and the Mg matrix.
Accepted 15 May 2017
Subsequent development of lateral liform-like corrosion fronts are observed running along the dendrite
arms whilst being constrained by the Al- and Zn-rich interdendritic boundaries. Prolonged immersion
showed variable corrosion depths across the surface. By correlating with EBSD analysis, it was conrmed
Keywords:
Automotive sheet
that the corrosion depth is considerably higher in a-Mg grains with non-basal planes on the surface. The
Mg-Al alloy severity of attack suffered by a-Mg grains with non-basal planes on the surface is on average twice that
Micro-galvanic corrosion for grains with basal or near-basal with planes on the surface. At the same time, corroded a-Mg grains
Mg dissolution with basal planes on the surface display nanoscale, canyon-like corrosion features developing into the
alloy perpendicularly. A mechanism has been proposed to explain the sequential corrosion events.
2017 Acta Materialia Inc. Published by Elsevier Ltd. This is an open access article under the CC BY
license (http://creativecommons.org/licenses/by/4.0/).

1. Introduction methods for production of Mg-based alloys with improved prop-

erties. Conventional sheet production involves several processing
Magnesium alloys are attractive candidates for lightweight steps including melting, ingot formation, homogenization, milling,
structural components due to their low density, high specic heating, hot rolling and annealing. As the cost of such a multi-step,
stiffness and strength [1]. However, Mg-based alloys generally labour intensive, Mg sheet process is relatively high, an efcient
suffer from poor corrosion resistance. This has been largely attrib- and cost-effective production route is required. The twin roll cast-
uted to the micro-galvanic inuence of impurities [2e5], the ing (TRC) process offers several advantages, potentially resulting in
microstructure and surface modication [6e8]. Specically, in the signicant cost savings [15,16]. The TRC process involves relatively
case of Mg-Al based alloys, the alloying additions of Al, Zn, Mn, etc., high cooling rates (102e103 K/s) compared to conventional ingot
result in the formation of discrete cathodic intermetallics such as casting, which results in rapid solidication [17,18]. The typical
AlxMny and b-Mg17Al12 phases, etc., which lead to localized micro- coarse and non-uniform grain sizes observed in TRC AZ31 Mg alloy
galvanic corrosion [8e11]. are attributed to the formation of a-Mg grains on coarse and
Mg-Al based alloys have been used for automotive sheet ap- randomly distributed MgO particles, which act as nucleating sites
plications, for example, seat frames, instrument panels, etc., with [19]. Grain size variations have been reported for Mg alloy sheets of
AZ31 being one of the most promising candidates [12e14]. The different sheet thicknesses [15]. TRC Mg sheet surfaces have also
production routes of metallic alloys affect their properties and the been reported to exhibit a strong (0001) basal texture, which is also
past decades have shown a substantial progress in different known to adversely affect the tensile properties, especially the
formability [16,20,21].
Crystallographic orientation has also been reported to have a
profound effect on the corrosion rate and path of Mg and its alloys.
* Corresponding author. EBSD studies on several Mg alloys have shown strong basal textures
E-mail address: [email protected] (P.J. Withers).

http://dx.doi.org/10.1016/j.actamat.2017.05.027
1359-6454/ 2017 Acta Materialia Inc. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
 S. Pawar et al. / Acta Materialia 133 (2017) 90e99 91

arising from the preferred [0001] crystal orientation of Mg grains, using successive grades of SiC abrasive papers from 240 to 4000
which signicantly inuence the corrosion performance [22e24]. grits, followed by polishing using 6 mm to mm diamond pastes
The strong basal texture and the consequent electrochemical with a non-aqueous lubricant and subsequent ultrasonic cleaning
anisotropy of Mg surfaces result in increased corrosion resistance in an acetone bath for 15e20 min. The metallographic specimens
[23e26]. The relationship between the crystal orientation and were etched using an acetic-picral etchant, comprising 5 ml acetic
corrosion in commercially pure Mg was demonstrated to be a acid 4.2 g picric acid 10 ml distilled water 100 ml (95%)
consequence of the difference in surface energies between the ethanol, for up to 10 s to reveal the microstructural features.
basal (0001), prismatic (1010) and (1120) crystal planes. Theoreti-
cally calculated surface energies for the crystal planes of Mg (0001), 2.3. Microstructure and texture characterization
(1010) and (1120) were reported to be 1.54
104, 3.04
104 and
2.99
104 J mol1 respectively, where the theoretical dissolution Optical microscopy was conducted using a Carl Zeiss light mi-
rates of (1010) and (1120) crystal planes were suggested to be about croscope tted with a polarizer and a digital camera. Scanning
20 times higher than that of the basal (0001) Mg crystal planes [27]. electron microscopy (SEM) was undertaken using a Carl Zeiss Ultra
Further, Song et al. showed that Mg grains with a basal orientation 55 FEG-SEM, equipped with an energy dispersive X-ray spectros-
in commercially pure Mg are more stable and corrosion resistant, copy (EDS) detector, operating at accelerating voltages in the range
exhibiting a more positive corrosion potential, lower anodic po- of 3e20 kV, using secondary (SE) and backscattered electron (BSE)
larization current density, larger impedance, and thinner surface detectors. Additionally, an FEI Magellan FEG-SEM with MAPS
lm than grains with a non-basal orientation [28]. On the other software was used to capture images from macroscale corroded
hand, formation of liform-like corrosion morphologies were surfaces at ultra-high resolution, operating at an accelerating
observed on Mg surfaces when exposed to chromate and chloride voltage of 5 kV and a beam current of 0.4 nA. Secondary electrons
solutions [23,29]. McCall et al. [30], showed that Mg (0001) surfaces were collected using the Everhart Thornley detector (ETD) detector
exhibited pitting corrosion susceptibility and that the (1010) and to create the image, and an array of small tiles collected and
(1120) crystal surfaces were passive at open circuit potentials. Also, automatically stitched together using the MAPS software. The array
pitting initiated only at potentials slightly anodic to their open of 750 (25
30) tiles, collected with a step size of 500 nm, provided
circuit potentials, the ndings were supported later by Lillard an ability to investigate the development of the corrosion fronts
et al.s simulation analysis [31]. Considering the results reported in across the entire (7.5 mm2) surface, without loss of resolution.
the literature, it is clear that the crystal orientation inuences the Electron backscattered diffraction (EBSD) was conducted on an FEI
corrosion properties of Mg surfaces. Magellan FEG-SEM, operating at an accelerating voltage of 20 kV
Most of the observations have been largely conducted on and the EBSD maps were processed using HKL Channel 5 software.
commercially pure Mg and wrought AZ31B Mg specimens, whereas Specimens were subjected to EBSD analyses both prior to and after
other methods of Mg alloy production can also have crucial rami- immersion tests. A total of 234 individual EBSD maps
cations on the resultant Mg alloy surfaces and their properties. (300
300 mm) were obtained, with a step size of 1.5 mm, which
AZ31 Mg alloy sheet [16], produced by the conventional TRC pro- were stitched together using Aztec software to create a single
cess, presents a genuine opportunity to investigate the inuence of snapshot of the corroded surface.
dendritic and textured microstructures on corrosion properties. In- Quantitative texture measurements were performed using X-
depth microscopic characterization of corroded surfaces is ray diffraction (XRD), using a Philips X'pert Modular Powder
extremely important in order to determine the structural integrity Diffractometer (CuKa) with a scanning range (2q) of 5 e85 . The
of such Mg alloy surfaces. Here, a correlative approach bringing resulting peaks were tted by the X'Pert HighScore software pro-
together various optical and electron microscopy techniques is gram and the data were used to calculate (0002), (1010) and (1120)
applied to study the inuence of the crystal orientation of a-Mg pole gures. Thin foils required for imaging in the scanning trans-
grains on the corrosion propagation mechanism. The role of inter- mission electron microscope (STEM) were mechanically ground
metallic phases in the corrosion process for AZ31 has also been and punched into 3 mm discs with an average thickness of less than
carefully considered. 100 mm. The discs were ion beam thinned using a Gatan precision
ion polishing system (PIPS) at 5.0 kV and at an incident angle of
2. Experimental procedure 4e6 . STEM-EDS elemental mapping was conducted on the FEI
Talos instrument, equipped with the Super-XEDS detectors and
2.1. Material operated at an accelerating voltage of 200 kV.

AZ31 magnesium alloy sheets of 6 mm thickness were produced 2.4. Immersion testing in 3.5% NaCl solution
by the twin roll casting (TRC) process [16]. Compositional analysis
of the TRC AZ31 Mg alloy, shown in Table 1, was carried out using a The immersion testing was conducted at ambient temperature
Perkin-Elmer Optima 5300 dual view inductively coupled plasma in naturally aerated, near-neutral 3.5% NaCl solution, according to
atomic emission spectroscope (ICP-AES). ASTM-G31-72 [32]. The specimens were analysed by SEM-EDS
before and after chemical cleaning for 1 min in a boiling solution
containing 20% CrO3 1% AgCrO4 [33]. After removal of corrosion
2.2. Specimen preparation products, the specimen surfaces were rinsed in deionized water
and acetone, followed by drying in a cool air stream. Measurement
Sample preparation comprises sequential mechanical grinding of corrosion depths from individual Mg grains after immersion and
removal of corrosion products was performed using a 3D Bruker
Table 1
prolometer. The complex dendritic TRC microstructure (Fig. 1a)
Compositional analyses (wt.%) of the twin roll cast AZ31 magnesium alloy obtained posed a challenge as the corrosion depths varied across dendrite
by ICP-AES analyses. arms; hence, a series of individual line proles were taken across
Wt.% Al Zn Mn Si Fe Cu Ni
the corrosion features in each Mg grain and an average of at least 5
such proles per Mg grain are plotted in a bar graph, revealing the
AZ31 2.9 0.88 0.34 0.02 0.007 0.009 0.004
corrosion depths across several Mg grains.
92 S. Pawar et al. / Acta Materialia 133 (2017) 90e99

3. Results 3.2. Initiation of corrosion

3.1. Microstructural characterization This section highlights the initial events occurring on the TRC
AZ31 Mg alloy sheet after immersion in 3.5% NaCl solution. It is
The etched surface of TRC AZ31 Mg alloy in the optical micro- important to examine the immediate response of the sheet upon
graph of Fig. 1a, shows a dendritic structure of a-Mg grains (HCP, electrolyte exposure in order to develop a basic understanding of
space group P63/mmc, a 0.32 nm, c 0.521 nm). Fig. 1b reveals the corrosion initiation mechanism.
the interior of a single a-Mg grain, displaying the dendrite arms
comprising the Mg matrix and the interdendritic regions decorated 3.2.1. Localized micro-galvanic corrosion
with a complex distribution of intermetallics, evident in the back- Formation of hydrogen gas bubbles on the alloy surface is
scattered electron micrograph. The alloy surface shown in Fig. 1b evident from the visual observations within several minutes of
represents the RD-ND plane of the Mg sheet (RD: rolling direction, immersion in 3.5% NaCl [35]. Unreported visual observation after
ND: normal direction). The sub-micron particle size and the pref- 30 min showed initiation of localized corrosion at the intermetallic
erential distribution of intermetallics (bright features) on the particles with consequent hydrogen evolution, which is evident
interdendritic regions (between the dotted lines) are clearly from the dark, spherical, pits arrowed in Fig. 3a. After removal of
evident in the high magnication SEM image (Fig. 1c). Further, the the corrosion products, cathodic phases such as b-Mg17(Al,Zn)12
high angle annular dark eld scanning transmission electron and Al8Mn5 are revealed within the corrosion sites. For example,
micrograph (HAADF-STEM) from a typical interdendritic region Fig. 3b and c shows the location and morphology of the interme-
(Fig. 1d) and associated EDS maps conrm that the intermetallic tallic particle, where the EDS analyses conrmed to be Mn- and Al-
particles are mainly b-Mg17(Al,Zn)12 phase (Fig. 1e, f, & g) alongside rich and based on previous work, it is presumed to be an Al8Mn5
the ower-shaped Al8Mn5 intermetallic (Fig. 1g and h). The intermetallic particle. There is a difference in the contrast of the
microscopic analyses presented here highlight the general micro- corroded surface which can be seen between Fig. 3a and b; this is
structure of the alloy, consistent with the work reported previously the same region but in Fig. 3b the corrosion product formed during
[10,34]. immersion has been removed as described in the experimental
EBSD investigation was conducted prior to immersion, as shown section. This was to reveal the underlying microstructure for
in Fig. 2. The SE image in Fig. 2a reveals the distribution of coarse, as analysis after corrosion.
well as ne (60e600 mm) dendritic a-Mg grains possessing irreg-
ular shapes, which can be attributed to the nucleation kinetics
3.2.2. Lateral liform-like corrosion
associated with the TRC process [16]. The texture was measured on
Gradual development of lateral liform-like corrosion is
the RD-ND plane of the Mg sheet. The inverse pole gure map in
observed, which is associated with the localized corrosion pits
Fig. 2b highlights the characteristic texture of the TRC AZ31 Mg
formed initially at the cathodic intermetallic sites and the sur-
alloy sheet. It is evident from the pole gures that the sheet surface
rounding Mg matrix. After 2 h of immersion, it was observed that
possesses a relatively strong basal texture, with a majority of grains
the corrosion fronts preferentially follow the dendrite arms, as
oriented with the basal (0002) pole approximately parallel to the
shown in Fig. 4a. The dendrite arms comprising the Mg matrix
ND (red grains in Fig. 2b) in accordance with previous observations
preferentially corrode; this is anodic and highly corrosion suscep-
[16]. The distribution of a high fraction of a-Mg grains in basal
tible. These liform-like corrosion fronts are inhibited at the
orientation on the TRC surface, evident from the pole gures
interdendritic and grain boundaries. As conrmed from the SEM-
(Fig. 2c), representing an alloy surface area of 10 mm2 can be
EDS analyses, these interdendritic and grain boundaries showed
attributed to the inuence of rolling conducted during the TRC,
relatively high Al and Zn contents compared to the dendrite arms,
which orients the basal crystal planes of the a-Mg grains prefer-
which makes them relatively resistant and act as corrosion barriers.
entially towards the roll surface as observed for most as-rolled Mg
However, lateral advancement of corrosion fronts into adjacent Mg
sheet surfaces [20,21,23,24].
grains is also observed at some sites. The two individual corrosion

Fig. 1. (a) Optical micrograph under polarized light of the etched surface illustrating the dendritic structure, (b) backscattered electron micrograph highlighting the dendrite arms
and distribution of intermetallics within a single a-Mg grain, (c) high magnication SEM image showing the b-phase and Mn-containing intermetallic distribution within the
interdendritic (ID) regions (shown within the dotted lines), (d) HAADF-STEM showing discrete intermetallic particles in the ID region along with TEM-EDX maps (eeh) conrming
the Mg, Zn, Al and Mn elemental distributions respectively.
 S. Pawar et al. / Acta Materialia 133 (2017) 90e99 93

Fig. 2. EBSD analyses of the roll surface, where the texture was measured on the RD-ND plane (also casting direction) (a) SE image, (b) Inverse pole gure map, where the colours of
the grains correspond to the crystallographic axes in the stereographic triangle or inverse pole gures, as shown in the inset. (c) {0002}, {1010} and {1120} pole gures, referring to
the RD plane plotted from the XRD texture measurement conducted on a (10 mm2) area.

i.e. 2 h immersion, which is shown in Fig. 4b. These were chosen as

representative regions as observed from the top surface. Type A
grains showed a corrosion depth of ~35 mm while Type B grains
showed a corrosion depth of ~55 mm. Further, immersion tests in
3.5% NaCl for 24 h revealed that the corrosion depths varied
signicantly. An example of the corrosion depths measured for
grains 1, 2 and 3 shown in Fig. 4b were 115 mm, 97 mm and 196 mm
respectively. It is interesting to note that the corrosion morphol-
ogies in individual grains are signicantly different, where grain 1
and grain 3 show deeper corrosion penetration, while grain 2
showed relatively uniform and laterally spread corrosion
morphology with reduced corrosion depth. This event has been
referred to as Stage II corrosion.
Therefore, it is considered that Stage I corrosion is mainly a
consequence of the inuence of the cathodic intermetallic phases
on the anodic Mg microstructure. The corroded surface shows ev-
idence of local micro-galvanic corrosion, which subsequently
shows development of liform-like corrosion morphologies across
the alloy surface. As the non-uniform corrosion depths in individual
Mg grains (Fig. 4b) observed in Stage II corrosion indicate a dif-
Fig. 3. (a) Backscattered electron micrograph of the alloy surface along with the ference in the corrosion attack in Mg grains with different crystal
corrosion products, after immersion in 3.5% NaCl solution for 30 min. The arrows orientations, the reason for the difference in corrosion propagation
indicate corrosion initiation sites (arrowed), (b) a typical corrosion pit after removal of
rates of individual Mg grains will be explored in the next section.
corrosion products, and (c) a high magnication image presumably showing a ower-
shaped Al8Mn5 intermetallic in the site interior, evidenced by the Al-Mn-content from
the EDS spectrum, shown in (d). The characterization highlights the inception of Stage 3.3. The crystallographic inuence on corrosion
I corrosion.

As evident from the corrosion depths in Fig. 4b corrosion

events, including the micro-galvanic corrosion occurring at local- propagates unevenly from grain to grain. Here this is investigated
ized sites on the alloy surface, followed by the development of using electron backscattered diffraction in order to identify the
liform-like corrosion morphologies from these corroded sites relationship between the crystal orientation and corrosion rate in
along the dendrite arms can be summarized together as Stage I individual grains in Stage I corrosion.
corrosion. The gradual progression of such corrosion fronts is
eventually observed to cover the alloy surface, with a relatively 3.3.1. Corrosion initiation
shallow corrosion attack. Fig. 5a shows a backscattered electron image of the corroded
Although it is evident that the corrosion attack progresses surface after 1 h of immersion, along with the crystal orientations
laterally across the surface, the SEM images from the transverse of the a-Mg grains. The measurement of corrosion depths from
sections revealed non-uniform corrosion attack occurring in Stage I, each Mg grain has been undertaken using a 3D Bruker prolometer.
94 S. Pawar et al. / Acta Materialia 133 (2017) 90e99

Fig. 4. SEM images of the corroded alloy specimen after immersion in 3.5% NaCl. (a) the top view after 2 h reveals preferential corrosion in the alloy matrix following the dendrite
arms in the form of lateral liform-like corrosion fronts (Stage I), (b & c) the transverse sections which reveal variable corrosion depths, suggesting non-uniform corrosion attack in
individual Mg grains after 2 h (Stage I) and 24 h (Stage II). These were chosen as representative regions as observed from the top surface.

The threshold for the measurement of corrosion depths was the corroded specimen was characterized, where the Stage II
considered with respect to the unexposed specimen edge. The a- corrosion had occurred after 24 h immersion testing. The liform-
Mg grains are assigned to two sets, where set (A1-A8) comprises like corrosion morphologies were clearly evident from the
grains having non-basal crystal orientations and set (B1-B8) those corroded surface, as shown in Fig. 6a. The severely attacked regions
grains with basal crystal orientations. From the graph of the show an average depth of ~160e180 mm, while the surrounding
corrosion depths, shown in Fig. 5b, it is clear that on average the surface shows shallow corrosion with an average depth of ~75 mm
grains in set (A1-A8) show corrosion to greater depths (~52 mm), (measured using a 3D Bruker prolometer). The lateral propagation
compared to those in set (B1-B8) (~26 mm). The a-Mg grains rep- of corrosion fronts from the severely corroded areas appears to
resenting non-basal crystal planes are referred as G(NB) while those progress via the dendrite arms (arrowed). When observed under
with basal or near-basal crystal orientations are referred as G(B) in polarized light in the optical microscope (Fig. 6b), this region
the following text. This suggests that the severity of attack suffered comprises several Mg grains (some are labelled in Fig. 6a), which
by G(NB) is on average twice that for G(B). were severely attacked. Hence, it is crucial to determine the typical
The extent of corrosion from grain to grain varies widely, as is crystal orientations of preferentially corroded Mg grains. At the
clear from Fig. 5c, where regions in red indicate shallow and same time, the inuence of etching can be observed at some grain
reduced corrosion while regions in blue indicate increased corro- boundaries with corrosion cracks. It is important to note that the
sion depths. G(B) (1) shows relatively uniform corrosion, with an optical characterization of the corroded specimen was performed
average corrosion depth of ~2 mm (Fig. 5d), whereas G(NB) (2) suffers after storing in the desiccator, hence additional corrosion features
from severe corrosion attack, with an average depth of ~31 mm are apparent that can be ignored for the moment. From a structural
(Fig. 5e). At the same time, it is interesting to note from the 3D integrity perspective these demand further investigation going
images (Fig. 5d and e) that the resulting surface proles of the two forward.
grains appear to be substantially different. This suggests that the The band contrast map (Fig. 7a) shows dark features on the
Mg dissolution/corrosion in G(B) (1) is relatively uniform compared corroded surface which highlight the preferentially corroded areas
to G(NB) (2), which shows a deep, liform-like corrosion attack and during Stage II, while the inverse pole gure map (Fig. 7b) high-
hence increased Mg dissolution (Fig. 5e). Clearly, the corrosion lights the crystal orientations of respective Mg grains. The crystal
depth measurements coupled with the EBSD analyses after Stage I orientations of the corroded Mg grains, evident in the EBSD map,
indicate that the crystal orientation signicantly affects the Mg have been considered to discuss the correlation between the crystal
dissolution in individual Mg grains. To investigate the inuence of orientation and the corrosion propagation mechanism. The indexed
crystal orientation after Stage II, a detailed microscopic analyses EBSD patterns and crystal orientations for some corroded Mg grains
using a similar approach was conducted. have been highlighted in Fig. 7 (c1-c3), (d1-d3) and (e1-e3), which
conrms that preferentially corroded grains 1e3, 5, 6 & 8 show
3.3.2. Corrosion propagation non-basal crystal planes while grains 4 & 7 exhibit a basal or near-
Subsequently, the alloy surface shows the formation of corro- basal crystal planes. From the statistical analyses performed on the
sion products, usually Mg(OH)2, post-Stage I. The breakdown of corroded areas, it was evident that the population of corroded non-
Mg(OH)2 layer was observed [39], which was accompanied by basal Mg grains was signicantly higher compared to the Mg grains
spontaneous and vigorous H2 evolution, an indication of the trig- with the basal planes parallel to the surface. Secondly, the corroded
gering of Stage II corrosion. This event was visually observed and non-basal Mg grains typically showed increased corrosion depths
 S. Pawar et al. / Acta Materialia 133 (2017) 90e99 95

Fig. 5. (a) Characterization of the corroded surface after 1 h immersion (Stage I) and removal of corrosion products, using EBSD. The Mg grains are represented as two sets, (A1-A8)
comprises Mg grains in non-basal crystal orientations and (B1-B8) those having basal/near-basal orientations, (b) graph shows average corrosion depths measured by optical
prolometer for each Mg grain, (c) prole of the corroded surface after removal of corrosion products highlighting the variation in corrosion depth in individual Mg grains. (d) grain
(1) shows shallow and uniform corrosion attack whereas (e) grain (2) exhibits liform-like corrosion with increased corrosion depth (note different colour scales).

Fig. 6. (a) Back scattered electron micrograph of corroded surface after 24 h immersion (Stage II) reveals severely corroded regions (dark areas) and lateral liform-like corrosion
fronts, emanating from severely corroded sites, (the grains marked here are identied in Fig. 7). The surrounding surface is also affected from the corrosion attack; although not
evident in low magnication and (b) plane polarized light image of the same region shows Mg grains with a wide range of crystal orientations. Some grain boundaries, highlighted
in the image seem to be preferentially attacked, which can be attributed to etching (artifact).
96 S. Pawar et al. / Acta Materialia 133 (2017) 90e99

Fig. 7. EBSD characterization of the corroded surface shown in Fig. 6, after 24 h immersion (Stage II). (a) Band contrast map reveals the dark, severely corroded areas and (b) inverse
pole gure (IPF) map, which shows the crystal orientations of the Mg grains, (c1-c5) electron backscattered diffraction patterns, (d1-d5) indexed patterns and (e1-e5) crystal
orientations, of some of the severely corroded Mg grains.

compared to the basal Mg grains, also observed in Stage I (Fig. 5). from Stage I to Stage II and is expected to continue upon further
This observation therefore conrms that the crystallographic in- exposure. The in-depth microscopic characterization of corroded
uence on the sequential development of corrosion is extended areas from Stage II, shown in the micrograph, will be considered in

Fig. 8. Multiscale image comprising (a) low resolution composite backscattered electron micrograph of the corroded surface after 24 h immersion (Stage II) showing a region of
severe corrosion one having negligible corrosion, (b) high magnication SEM image from the highly corroded region of interest (RoI1) showing the boundary between two Mg
grains. EBSD conrms Grain 1 in (0002) orientation and Grain 2 in (0110) orientation (c) exploded image from RoI2, exhibiting very shallow and uniform corrosion attack, with
valley-like corrosion features across the interdendritic regions and remnants of Al8Mn5 intermetallics and b-phase and (d) higher magnication image of RoI2 displaying the
canyon-like corrosion features, preferentially propagating along the prismatic plane.
 S. Pawar et al. / Acta Materialia 133 (2017) 90e99 97

the next section. macroscopically uniform corrosion attack. However, the corrosion
morphology is signicantly different from that observed in the
3.4. Characterization of corrosion features in individual Mg grains corroded Mg grain of non-basal orientation (Fig. 8b). It is inter-
on the surface esting to note the nanoscopically deep, crystallographic, canyon-
like corrosion features within the remnant Mg matrix at high
In order to obtain further evidence related to the corrosion magnication (Fig. 8d). These canyons propagate into the grain
propagation in individual Mg grains with basal and non-basal interior, along the prismatic planes, beneath the corrosion prod-
crystal orientations after Stage II, the corrosion features have ucts. The parallel sided, canyon-like corrosion features are located
been probed by HR-SEM. Prior to the microscopic analyses, the at ~90 to the interdendritic boundaries, which are highlighted by
corrosion products were removed from the corroded sample dotted lines in Fig. 8c [36]. The canyon-like corrosion features are
surface. observed across the interiors of dendritic cells in several corroded
G(B)s. Therefore, this result suggests that the G(B)s exhibiting
higher corrosion resistance [24e28], evident from the shallow
3.4.1. Mg grains in non-basal orientations
corrosion depth (~75 mm) show preferential corrosion propagation
The microstructural analyses from a region of interest (RoI1) in
along the prismatic crystal planes (Fig. 8d). At the same time, there
Fig. 8a is shown in the high magnication image (Fig. 8b), which
is no evidence of corrosion propagation surrounding the cathodic
reveals the grain boundary interface between the adjacent a-Mg
phases like Al8Mn5 and b-phase (Fig. 8c).
grains. In the image, Grain 1 displays shallow corrosion, whereas
Grain 2 shows a severe corrosion attack along the dendrite arms,
with the interdendritic boundaries largely unaffected. These act as 4. Discussion
a corrosion barrier to the corrosion taking place within the grain
interior. At the same time, the grain boundary, highlighted with The current study typically highlights the inuence of TRC
dotted lines in Fig. 8b, shows a signicant amount of b-phase, microstructure on the corrosion properties of AZ31 Mg alloy sheet.
which appears to confer additional resistance to the progression of Electron microscopy of the alloy surface shows cathodic phases
the corrosion fronts at the grain boundary. EBSD reveals that G(B) namely, Al8Mn5, b-Mg17(Al,Zn)12, along with isolated Fe-particles
(1) exhibits an [0002] crystal orientation while G(NB) (2) shows a (Fig. 1). The inuence of the relatively high cooling rates associ-
[0110] crystal orientation conrming that higher Mg dissolution in ated with the TRC process [28], is clearly evident from the ne
Grain 2 can be associated with the typical non-basal crystal particle sizes of Al8Mn5 and b-phases (Fig. 1). Secondly, rolling,
orientation [23e26]. which is an integral part of the TRC process also dominates the
distribution of the crystal planes of the Mg grains in the solidied
3.4.2. Mg grains in basal orientations microstructure (Fig. 2). From the EBSD results, it was conrmed
Fig. 8c shows higher resolution analysis of G(B) in [0002] that the surface exhibits a higher fraction of Mg grains of basal
orientation, which suffered from a shallow corrosion attack (region orientation (Fig. 2). Therefore, the discussion is based on the in-
of interest RoI2 in Fig. 8a). It is evident from the relatively at uence of local chemistries of microconstituent phases and crystal
corrosion morphology that the grain interior suffered orientation of Mg grains which directly affects the corrosion at

Fig. 9. Schematic diagrams showing the corrosion mechanism in TRC AZ31 Mg alloy in 3.5% NaCl solution. The surface indicates the 3D crystal orientations w.r.t. RD, whereas the
transverse section microstructure reveals corrosion depths. (a) image reveals G1 (near-basal), G2 (basal) and G3 (non-basal) crystal orientations, along with the micro-constituent
features, (b) corrosion initiation occurs at local sites constituting Al8Mn5/b/Fe particles (cathodes), (c) liform-like corrosion fronts develop from the initiation sites and follow the
dendrite arms. Both (b) and (c) constitute Stage I corrosion, (d) subsequent immersion shows corrosion propagation with increase in corrosion depth as a function of immersion
time, which constitutes Stage II corrosion. The liform-like corrosion fronts are observed in both G1 and G3 which indicate increased and preferential corrosion. G3 with non-basal
crystal orientation shows increased corrosion depth followed by G1 with near-basal crystal orientation, while G2 with basal crystal orientation reveals a relatively uniform and
reduced corrosion depth and (e) the exploded view of the corrosion occurring beneath the corrosion product in G2 (basal) grain which shows nano-sized, canyon-like corrosion
features.
98 S. Pawar et al. / Acta Materialia 133 (2017) 90e99

different stages. A schematic diagram (Fig. 9) is displayed in order surfaces (3D prolometric analyses), it is clear that G(NB)s suffer a
to explain the sequential corrosion events. non-uniform and localized corrosion attack (Fig. 5e) compared to
G(B)s which exhibit a relatively uniform corrosion attack (Fig. 5d).
4.1. Schematic representation of the corrosion events Therefore, the difference in the resultant corrosion morphologies
observed in both G(B)s and G(NB)s shed valuable insight on the
Fig. 9a shows the TRC AZ31 Mg alloy microstructure, with Mg mode of corrosion attack and the anodic Mg dissolution mecha-
grains with crystal orientations, namely G1 (near-basal), G2 (basal) nism of Mg grains with a difference in the crystal planes. As
and G3 (non-basal), relative to RD (surface). On immersion, the mentioned in the introduction, for the case of pure polycrystalline
corrosion initiates at localized cathodic sites, namely Fe-, Al8Mn5 Mg, the Mg dissolution rates in G(B)s were considerably slower
and b-Mg17(Al,Zn)12, resulting in H2 evolution and formation of compared to G(NB)s. This effect was attributed to the closely packed
corrosion products on the TRC surface, mainly, Mg(OH)2, as shown crystallographic plane (0001) and its lower surface energy, result-
in Fig. 9b. Further, lateral development of liform-like corrosion ing in reduced Mg dissolution [23]. Saradha et al. [37] had reported
fronts, which initiate from the local corrosion sites and follow the that the theoretical work function value of a closely packed crys-
dendrite arms (Fig. 4a), preferentially consuming the Mg matrix. tallographic plane was signicantly more positive than that of a less
These corrosion fronts are restricted at the Al- and Zn-rich inter- closely packed plane, which was further experimentally conrmed
dendritic/grain boundaries, highlighted in Fig. 9c. The localized by Song et al. [23]. Although the work function values have not
micro-galvanic corrosion shown in Fig. 9b and the liform-like been conrmed in the current study, the experimental ndings
corrosion morphologies in Fig. 9c together constitute Stage I here strongly support the hypothesis proposed by Saradha et al. The
corrosion event. Prolonged immersion shows propagation of OCP of the metal is linearly correlated to the work function, which
corrosion fronts in individual Mg grains (Fig. 9d), evident in G3 & has been theoretically deduced and experimentally veried by
G1, which is signicantly higher compared to G2, refers to Stage II authors elsewhere [23,29,38,39]. The non-uniform corrosion attack
corrosion. The difference in the corrosion depths is highlighted observed in the current study can be attributed to the inuence of
with the colour contrast of the Mg(OH)2 corrosion product in in- the TRC process on the surface, which constitutes a high fraction of
dividual dendrite arms, the Mg grains, and also being observed G(B)s along with a random distribution of G(NB)s (Fig. 2). The
from the extent of corrosion attack on the cross section. Following consequent effect of crystal orientation of the Mg grains with
with the high resolution microscopy shown in Fig. 9e, nanoscopi- different crystal planes is evident from the corrosion morphologies
cally deep, crystallographic corrosion features, develop beneath the (Fig. 5) is similar the reported literature [23,24]. Prolonged expo-
corrosion products, indicating the corrosion propagation in Mg sure showed increased corrosion depths in Mg grains with crystal
grains with basal crystal planes. planes other than (0002) observed in the current study (Fig. 8).
Hagihara et al. [26] recently examined the orientation-dependent
4.2. Micro-galvanic corrosion corrosion behaviour of pure-Mg single crystals. The corrosion
rates increased in the order (0001) < 1120 < 1010 < 1123 <
The role of intermetallics on the initiation of corrosion in TRC 1012. The dynamic polarization curves showed that the
AZ31 Mg alloy has been reported earlier by the authors [10], and orientation-dependence of the current density was signicant in
derives from the localized nobility of micro-constituent phases. the anodic region, while it was almost negligible in the cathodic
Scanning Kelvin probe force microscopy (SKPFM) analyses on the region. Therefore, it was reported that the anodic reaction governs
TRC AZ31 Mg alloy sheet conrmed that the local Volta potential the orientation-dependent corrosion behaviour and that the
difference was capable of generating micro-galvanic coupling be- charge-transfer and lm resistances were strongly orientation-
tween the micro-constituents including Fe-particles, Al8Mn5 in- dependent.
termetallics and b-Mg17(Al,Zn)12 phase and the Mg matrix, which Importantly, striated corrosion features were specically
leads to the corrosion initiation (Fig. 3). The inuence of interme- observed on (0001) crystal planes in the case of pure Mg, when
tallic phases, mainly AlxMny and the b-phase, on the corrosion exposed to 3.5% NaCl [25,36]. However, it needs to be considered
behaviour of Mg-Al based alloys has also been widely reported; that the sample surfaces were examined along with the corrosion
such phases act as cathodes, and, are responsible for micro-galvanic products; hence, it is quite possible that the corrosion morphol-
corrosion [9e11]. The results from short exposure times during the ogies beneath the corrosion products were not clearly evident. It
immersion experiments in the current study (Fig. 3), revealing was suggested that there was an underlying crystallographic in-
localized corrosion are in good agreement with the literature. uence on the development of surface structures. The evidence of
Following the localized corrosion initiation on the surface, conse- preferential Mg dissolution along prismatic crystal planes in G(B)s
quent liform-like corrosion fronts emerge, connecting the local was clearly evident from the current study (Fig. 8ced) which sug-
corrosion sites within the dendrite arms, as a result of the anodic gests that the striated corrosion features must be connected and
Mg dissolution. Lateral progression of the corrosion paths, which resemble to the canyon-like features observed in the current study
are restricted by the interdendritic boundaries, relative to the Mg and drive the corrosion propagation in G(B)s [25].
matrix, can be attributed to the corrosion resistant Al-and Zn-rich
coring. Such corrosion morphologies are a commonly observed
phenomena in Mg-Al based Mg alloys [9e11]. 5. Conclusions

4.3. Crystallographic orientation dependence 1. Corrosion initiation occurred preferentially at local sites, due to
micro-galvanic coupling between the cathodic phases, namely:
Although the corrosion fronts were observed to laterally cover Al8Mn5 intermetallic particles, b-Mg17(Al,Zn)12 particles and the
the alloy surface, the cross sections showed variation in the anodic Mg matrix in the dendrite arms. Subsequently laterally
corrosion depths (Fig. 4b). Referring to the initial corrosion depths spread, liform-like corrosion morphologies were observed
(Fig. 5b), coupled with the crystal orientation of the Mg grains across the alloy surface, consuming the Mg matrix along the
(Fig. 5a), it is clear that G(NB)s showed increased rates of corrosion dendrite arms, and the corrosion fronts being restricted at the
propagation which in turn indicate faster Mg dissolution (Fig. 5c); Al- and Zn-rich interdendritic and grain boundaries. After im-
compared to G(B)s (Fig. 5d). From the appearance of remnant mersion in 3.5% NaCl for 1 h, a-Mg grains with non-basal plane
 S. Pawar et al. / Acta Materialia 133 (2017) 90e99 99

on the surface showed an average depth of corrosion twice that [10] S. Pawar, X. Zhou, G.E. Thompson, G. Scamans, Z. Fan, J. Electrochem. Soc. 162
of the a-Mg grains with {0002} planes on the surface. (9) (2015) C442eC448.
[11] S. Pawar, X. Zhou, G.E. Thompson, J.D. Robson, I. Bayandorian, G. Scamans,
2. Prolonged immersion in 3.5% NaCl showed non-uniform Z. Fan, Corrosion 68 (6) (2012) 548.
corrosion depths in individual Mg grains. EBSD characteriza- [12] J.T. Carter, R. Verma, P.E. Krajewski, Magnes. Technol. (2008) 69e74.
tion of the TRC surface conrmed that the corrosion propagation [13] R. Verma, J. Carter. SAE Paper 2006-01-0525.
[14] P.E. Krajewski, S. Kim, J.T. Carter, R. Verma, Magnesium sheet: automotive
in non-basal Mg grains is pronounced compared to Mg grains of applications and future opportunities, Korean Inst. Metals Mater. Trends
[0002] orientation. Metals Mater. Eng. 20 (5) (2007) 60e68.
3. For the Mg grains with {0002} planes on the surface, nanoscale, [15] S. Das, N. Barekar, O.E. Fakir, X. Yang, J.P. Dear, Z. Fan, Mater. Lett. 144 (2015)
54e57.
deep, crystallographic, canyon-like corrosion features had [16] I. Bayandorian, Y. Huang, Z. Fan, S. Pawar, X. Zhou, G.E. Thompson, Metall.
developed into the alloy perpendicularly, conrming that Mater. Trans. A 43 (2012) 1035e1047.
corrosion in such grains propagates preferentially along the [17] Y.S. Park, S.B. Lee, N.J. Kim, Mater. Trans. 44 (2003) 2617.
[18] S. Das, N.S. Barekar, O.E. Fakir, L. Wang, A.K. Prasada Rao, J.B. Patel,
prismatic planes. H.R. Kotadia, A. Bhagurkar, J.P. Dear, Z. Fan, Mater. Sci. Eng. A 620 (2015)
223e232.
Acknowledgements [19] H. Men, B. Jiang, Z. Fan, Acta Mater. 58 (2010) 6526e6534.
[20] M. Masoumi, F. Zarandi, M. Pekguleryuz, Mater. Sci. Eng. A 528 (2011)
1268e1279.
The authors are grateful to the TARF-LCV Grant (EP/I038616/1), [21] N. Tang, M.P. Wang, H.F. Lou, Y.Y. Zhao, Z. Li, Mater. Chem. Phys. 116 (2009)
LiME Hub (EP/N007638/1) and the LATEST2 Programme Grant (EP/ 11e15.
[22] F. Kaiser, J. Bohlen, D. Letzig, K.-U. Kainer, A. Sfyczynski, C. Hartig, Adv. Eng.
H020047/1). PJW is grateful for funding through an European
Mater. 5 (12) (2003) 891e896.
Research Council Advanced Grant 695638 (CORREL-CT). The au- [23] G. Song, R. Mishra, Z. Xu, Electrochem. Commun. 12 (2010) 1009.
thors would like to acknowledge EPSRC funding for the Henry [24] G. Song, Z. Xu, Corr. Sci. 54 (2012) 97e105.
Moseley X-ray Imaging Facility under grants EP/M010619, EP/ [25] M. Liu, D. Qiu, M.C. Zhao, G. Song, A. Atrens, Scr. Mater 58 (2008) 421.
[26] K. Hagihara, M. Okubo, M. Yamasaki, T. Nakano, Corr. Sci. 109 (2016) 68e85.
K004530, EP/F007906, EP/F028431. [27] B.-Q. Fu, W. Liu, Z.-L. Li, Appl. Surf. Sci. 255 (2009) 9348e9357.
[28] G.-L. Song, Z. Xu, Corr. Sci. 63 (2012) 100.
References [29] P. Schmutz, V. Guillaumin, G.S. Frankel, R.S. Lillard, J.A. Lillard, J. Electrochem.
Soc. 150 (2003) B99eB110.
[30] C.R. McCall, M.A. Hill, R.S. Lillard, Corros. Eng. Sci. Technol. 40 (2005) 337.
[1] G.L. Makar, J. Kruger, Int. Mater. Rev. 38 (1993) 138e153.
[31] R.S. Lillard, G.F. Wang, M.I. Baskes, J. Electrochem. Soc. 153 (2006) B358eB364.
[2] G. Song, A. Atrens, Adv. Eng. Mater. 5 (2003) 837e858.
[32] ASTM-G31-12a. 2012; http://dx.doi.org/10.1520/G0031-12A.
[3] J.D. Hanawalt, C.E. Nelson, J.A. Peloubet, Trans. AIME 147 (1942) 273.
[33] ASTM-G1. 2011; http://dx.doi.org/10.1520/G0001-03R11.
[4] G.S. Frankel, A. Samaniego, N. Birbilis, Corr. Sci. 70 (2013) 104e111.
[34] S. Pawar, X. Zhou, T. Hashimoto, G.E. Thompson, G. Scamans, Z. Fan, J. Alloys
[5] S. Feliu Jr., A. Samaniego, A.A. El-Hadad, I. Llorente, Corr. Sci. 67 (2013)
Compd. 628 (2015) 195.
204e216.
[35] Y. Yang, F. Scenini, M. Curioni, Electrochim. Acta 198 (2016) 174e184.
[6] H. Matsubara, Y. Ichige, K. Fujita, H. Nishiyama, K. Hodouchi, Corr. Sci. 66
[36] L. Yang, X. Zhou, M. Curioni, S. Pawar, H. Liu, Z. Fan, G. Scamans, G. Thompson,
(2013) 203e210.
J. Electrochem. Soc. 162 (2015) C362eC368.
[7] F. Cao, Z. Shi, J. Hofstetter, P.J. Uggowitzer, G. Song, M. Liu, A. Atrens, Corr. Sci.
[37] R. Saradha, M.V. Sangaranarayana, J. Phys. Chem. B 102 (1998) 5099e5106.
75 (2013) 78e99.
[38] M. Stratman, H. Streckel, Corr. Sci. 30 (1990) 681e696.
[8] J. Liao, M. Hotta, S. Motoda, T. Shinoharra, Corr. Sci. 71 (2013) 53e61.
[39] P. Schmutz, G.S. Frankel, J. Electrochem. Soc. 145 (1998) 2285e2295.
[9] A. Pardo, M.C. Merino, A.E. Coy, R. Arrabal, F. Viejo, E. Matykina, Corr. Sci. 50
(2008) 823e834.