Applied Surface Science 497 (2019) 143772

Contents lists available at ScienceDirect

Applied Surface Science

journal homepage: www.elsevier.com/locate/apsusc

Full length article

Formation mechanism of oxide layer on AZ31 Mg alloy subjected to micro- T

arc oxidation considering surface roughness
⁎
N. Nashraha, M.P. Kamila, D.K. Yoona, Y.G. Kimb, Y.G. Koa,
a
Materials Electrochemistry Group, School of Materials Science and Engineering, Yeungnam University, Gyeongsan 38541, Republic of Korea
b
Extreme Fabrication Technology Group, Korea Institute of Industrial Technology, Daegu 42994, Republic of Korea

A R T I C LE I N FO A B S T R A C T

Keywords: This study investigated the formation mechanism of the oxide layer generated by micro-arc oxidation (MAO)
Mg alloy considering surface roughness of AZ31 Mg alloy. For this purpose, a series of MAO treatments under alternating
Micro-arc oxidation current in a silicate-based electrolyte were performed on two present samples with grooved and flat surfaces. The
Surface roughness time-voltage response presented that the grooved sample reached the breakdown voltage earlier than the flat
Nucleation
sample. This indicated the appearance of micro-discharges on the grooved sample occurred earlier than that of
Growth
the flat sample since the oxide fragments nucleated preferentially at the ridge areas where electrons were
concentrated severely under high electrical field. The oxide layer of the grooved sample was relatively less dense
than that of the flat sample. This was attributed to the fact that the oxide layer of the grooved sample was likely
to grow with ease on pre-existing coating layer whereas the oxide layer of the flat sample developed conformably
throughout the whole areas. Thus, the growth rate of the grooved sample was higher than that of the flat sample,
which agreed well with micro-discharge activities. Such formation mechanism was proposed in relation to
nucleation and growth of the oxide layer formed on different surface structures.

1. Introduction parameters of rectifier [10,11], and by controlling substrate features,

such as grain size and surface topography [12,13]. Among processing
In recent years, magnesium (Mg) and its alloys have been in- variables, the surface morphologies of the substrate would be known to
vestigated extensively as one of the most promising metallic candidates affect the formation of the oxide layer during MAO coating. Moreover,
suitable for a variety of commercial applications within the scopes of the influences of surface roughness on corrosion, wear, and fatigue
electronic and automobile components due to high electromagnetic properties were significant [14–16]. For instance, Yoo et al. [17] pos-
shielding properties and good specific strength at room temperature. tulated from direct current PEO coating on AZ91 Mg alloy samples
Despite their attractive properties, these applications were un- whose surface roughness values were measured to be 0.5, 1, and 2.5 μm
fortunately obstructed by inherent poor corrosion resistance of Mg alloy that, with an increment in surface roughness, corrosion resistance in
against corrosive environment [1–3]. Thus, a variety of surface mod- 3.5 wt% NaCl solution decreased to some extent whereas the transfer of
ification methods, such as metal-based plating, chemical-conversion current across the oxide layer became easier due to the fact that the flat
coating, and micro-arc oxidation (MAO, also known as plasma elec- surface might cause crystalline phase in the oxide layer during growth
trolytic oxidation), have been applied [4–6]. MAO, where the electro- [13,17]. Another study reported by Wang et al. [18] suggested that the
chemical reactions accompanying a myriad of micro-arcs over dielectric regular micro-grooves with a trapezoidal shape formed on Al alloy
ignition triggered the nucleation of the oxide layer in an eco-friendly would lead to the high fractions of α-Al2O3 and 3Al2O3·2SiO2 working
electrolyte, was considered to be desirable to fabricate the protective as hard constituents in the oxide layer due to high temperature and
oxide layer with excellent adhesion to the metal substrate. This tech- pressure in the sharp interface. This would give rise to superior wear
nique differed from the conventional hard-coatings [7]. resistance to the flat counterpart [18]. According to the fatigue results
Earlier studies reported the effects of several parameters on the of 2024 Al alloy samples processed by MAO, the fatigue life was seen to
improvement in the electrochemical properties as well as on the for- decrease with increasing initial surface roughness in which hill and
mation of the oxide layer by modifying electrolytes with the additions valley were identified [16].
of organic and/or inorganic compounds [8,9], by changing electrical While it has been well accepted that surface roughness was a crucial

⁎
Corresponding author.
E-mail address: younggun@ynu.ac.kr (Y.G. Ko).

https://doi.org/10.1016/j.apsusc.2019.143772
Received 28 May 2019; Received in revised form 19 August 2019; Accepted 22 August 2019
Available online 23 August 2019
0169-4332/ © 2019 Elsevier B.V. All rights reserved.
N. Nashrah, et al. Applied Surface Science 497 (2019) 143772

factor affecting the inherent performances of the oxide layer, the fun-
damental mechanism that would describe nucleation and growth of the
oxide layer during plasma electrolysis remained less understood by
considering a role of surface roughness.
Therefore, the present study attempts to scrutinize the nucleation
and growth influenced by surface roughness of the initial sample. The
nucleation phenomena of the oxide fragments formed at the ridge and
valley of the initial substrate are compared at the onset of breakdown
where the plasma discharges start to appear. Subsequently, both growth
pattern and rate of the oxide layer are studied to propose formation
mechanism in relation to microstructural interpretation. Two different
corrosion assessments are used to describe the difference in electro-
chemical performances related to surface roughness. In addition, the
stress state of the oxide layer is discussed based on high-energy plasma
discharges formed on the surface during MAO.

2. Experimental procedures

A commercial-graded AZ31 Mg alloy plate was cut into a dimension

of 15 (L) × 25 (W) × 4 (T) mm3. Prior to MAO coating, two samples
were polished mechanically with SiC paper up to #2400 grit followed
by diamond suspension to obtain a nearly flat surface (FS). Then, one of
the samples was ground again using SiC paper with #100 grit in order
to produce a grooved surface (GS). Both samples were rinsed with
distilled water, cleaned ultrasonically with pure ethanol, and dried in a
stream of warm air. Surface roughness of the present samples was
measured by atomic force microscopy (AFM, Park systems NX-10). To
quantify surface roughness, the value of Ra that expressed an average
deviation of the oscillations along a regression line drawn through
particular roughness profile between the ridge and valley areas in the
samples was used. Such values for both FS and GS samples were mea-
sured to be 0.025 and 1 μm as shown in Fig. 1.
A series of MAO treatments were performed in an aqueous elec- Fig. 1. Representative AFM images showing the surface morphologies of the
trolyte with 0.1 M KOH and 0.05 M Na2SiO3 (pH: 11.83, conductivity: present samples with (a) flat and (b) grooved surfaces prior to MAO coating.
12.39 mS·cm−1) at current density and frequency of 50 mA·cm−2 and These structures are uniform throughout the surfaces after mechanical pre-
paration.
60 Hz, respectively. The present samples worked as the anode while the
stainless steel net was set as the cathode. A glass vessel was used as the
electrolyte container, equipped with magnetic stirrer and water cooling sample via MAO for 20 s is not provided due to less noticeable micro-
system to maintain the electrolyte temperature around 288 K for the discharges on the anode surface. From the increasing tendency of re-
stabilization of electrochemical reactions throughout the process. sponding voltage shown in Fig. 2a, two stages were divided clearly by
Details of the experimental setup were found elsewhere [19]. During their own slopes. In stage I, the responding voltage would tend to in-
MAO, the real-time images of micro-discharges formed on the surface crease linearly, obeying Ohm's law due to an increase in electrical re-
were taken with respect to processing time. The morphologies of the sistance of the thin passive film that formed on the substrate by che-
surface and cross section in the oxide layer were observed by field- mical reactions [9,12,19]. Unless the thin passive film would withstand
emission scanning electron microscope (FE-SEM, HITACHI S-4800) the intense electric field any longer, breakdown phenomenon began to
coupled with energy-dispersive spectrometer. appear together with a number of micro-discharges as the lighting dots.
The electrochemical response of the oxide layer was evaluated Then, stage II was started. The dotted red-circle in both curves in-
through potentiodynamic polarization and electrochemical impedance dicated the breakdown voltage whose values were measured to be
spectroscopy (EIS) in 3.5 wt% NaCl solution near to sea water. The ~220 V for both cases. Interestingly, times to reach breakdown were
polarization tests were scanned from −0.25 to +0.4 V with respect to different from each other in this study. FS and GS samples reached such
open circuit potential at a scan rate of 1 mV·s−1. EIS tests were per- ignition in 30 and 20 s respectively, suggesting that surface roughness
formed at frequency ranging 106 to 10 Hz with 10 mV rms perturbation. would influence the initial stability and concurrent nucleation of micro-
For each sample, electrochemical tests were repeated at least three discharges on the sample. It is inferred that the higher surface rough-
times to ensure that the data obtained would be feasible. ness, the earlier breakdown took place. After breakdown, the voltage in
stage II increased with slope value lower than that in stage I as the
3. Results and discussion electrical resistance increased by the growth of the oxide layer covering
the anode surface. After MAO for 300 s, the final voltages of FS and GS
3.1. Voltage-time response with micro-discharge samples reached ~290 and ~320 V, respectively. Such a difference in
the final voltage would be associated with the thickness and morphol-
Fig. 2 presents the voltage-time curves and histograms of FS and GS ogies of the oxide layer, which would be discussed microstructurally in
samples coated by MAO at 50 mA·cm−2 as a function of coating time the following section.
ranging from 20 to 300 s. The changes in the responding voltages to- The inset images in Fig. 2a evidenced that micro-discharges on FS
gether with couples of inset images showing the characteristics of and GS samples were evolved in different manners that allowed their
micro-discharges for 20, 30, 120, and 300 s are displayed in Fig. 2a, and size and fraction to vary moderately with increasing coating time.
the histograms of both size distribution and fraction of micro-discharges These results were analyzed in Fig. 2b. After MAO for 20 s, a number of
with respect to coating time are provided in Fig. 2b. A histogram of the fine micro-discharges with a dim color started to appear on GS sample

2
N. Nashrah, et al. Applied Surface Science 497 (2019) 143772

(caption on next page)

3
N. Nashrah, et al. Applied Surface Science 497 (2019) 143772

Fig. 2. (a) Responding voltage vs. coating time curves of FS and GS samples during MAO at 50 mA·cm−2 for 300 s. The insets show the evolution of micro-discharges
formed on surfaces with respect to coating time, and (b) the size distribution and fraction of micro-discharges taken from the insets.

Fig. 3. Representative SEM images showing the surface and cross section structures after MAO for 30 and 20 s when breakdown took place for the first time: (a), (b)
FS and (c), (d) GS samples. Points A and B are used for compositional analyses in EDS.

Table 1 MAO at breakdown point corresponding to FS and GS samples for 30

EDS results of GS sample after MAO for 30 s when breakdown took place for the and 20 s, respectively. As presented in Figs. 3a and b, a thin oxide layer
first time. The measurements are taken from the representative points at the with a fairly uniform thickness of 1.2 μm was observed clearly on the
ridge and valley areas, which are denoted as points A and B in Fig. 3c. whole surface of FS sample. Several micro-pores with an average size of
Area Point Mg (wt%) O (wt%) Si (wt%) 0.5 μm were also detected due to micro-discharges with high energy.
Those structures were typical characteristic of metal-oxide layer formed
Ridge A 48.8 ± 0.8 41.0 ± 1.3 10.2 ± 0.6
by MAO [20,21]. In case of GS sample, the formation of the oxide layer
Valley B 63.6 ± 2.7 29.1 ± 0.5 7.30 ± 0.7
seemed to be premature. From Fig. 3c, both ridge and valley were still
visible even after breakdown. To confirm the elemental composition,
while no micro-discharge was detected on FS sample. This fact was EDS point measurements were carried out at least five times to show the
confirmed again by the different times to reach breakdown in the vol- element difference on the ridge and valley areas as indicated by the
tage-time data. In particular after 300 s, the size distribution of micro- black (A) and white (B) dots in Fig. 3c. The amount of O and Si in wt%
discharges in FS sample was more homogenous than that in GS sample on the ridge was relatively higher than that on the valley as listed in
which showed a bimodal distribution where coarse and fine sparks Table 1, confirming the un-uniform formation of oxides on the ridge at
coexisted. Thus, the fraction of micro-discharges in GS sample was the initial stage of MAO coating. From the observations on the cross
lower than that in FS counterpart, which was attributed mainly to the section, it is important to note that some oxide fragments were also
cluster of micro-discharges in local areas of GS sample. This finding found to nucleate locally around the ridge to a greater extent than that
suggested that growth behavior of the oxide layers was expected to be near to the valley as shown in Fig. 3d. This fact implied that breakdown
different in the samples with different surface topographies. phenomenon was initiated around the ridge earlier than around the
adjacent valley.
It is found that both chemical decay of metal and formation of
3.2. Nucleation behavior upon breakdown passive film were processed in the same way that allowed the move-
ment of the anion species in the wet-process and their interaction with
Fig. 3 shows an initial occurrence of the oxide layers produced by

4
N. Nashrah, et al. Applied Surface Science 497 (2019) 143772

metal anode to be attained during natural chemical reactions. This sample, intense plasma arcs would result in the formation of discharge
process was influenced significantly by how the surface features after channel present in the oxide layer. On the other hand, the size and
mechanical finish would be formed, such as artificial grid in the whole distribution of the micro-pores were relatively uniform in FS sample
surface and the related surface roughness in local area per se. According after MAO for 300 s.
to the recent investigation in the correlation between surface groove This phenomenon was explained by continuous monitoring on
and chemical reactions reported by Zuo et al. [22] using 316 L stainless short-duration PEO treatment that the serial formation of micro-dis-
steel in 0.01 M NaCl solution, the aspect ratio of the surface groove charge was expected on the local region with low electrical resistivity as
defined as w/d (openness parameter), in which w and d were the width discussed by Clyne research group [24]. The oxide layer with either
and depth of one groove at the opening, would control the rate of local deep micro-pores or thin thickness would be the plausible candidate
chemical reactions. They suggested that, depending upon surface finish responsible for electron transfer across the insulation barrier during the
method, the value of w/d higher than threshold would lead to the nu- wet-based oxidation coating. It should be noted that this tendency was
cleation of localized decay by the accumulation of anions. In this study, understood on the same system in a ‘relative’ manner. In this study, the
the present value of openness parameter in any ridge-valley region was valley area with lower volume of the oxide fragments would be sus-
measured to be ~2.5 (w: ~10 and d: ~4). It is certain that the formation ceptible to next breakdown accompanying intense cascades because
of the thin passive film during anodization prior to ignition accom- nucleation of the oxide layer appeared first at the ridge area as evi-
panying plasma arcs was prone to be initiated primarily in the vicinity denced in Fig. 3d. Thus, micro-discharges concentrated at the valley
of valley area with a high surface area. Thus, the passive film at the area would play an important role in the formation of type-B discharge
valley was relatively thicker than that at the ridge. This fact would channels (defined by Hussein et al. [25]) as well as relatively fast
affect the nucleation of the oxide fragments on the onset of breakdown. growth of the oxide layer in the local area. Both XRD data of two
The thin passive film with low electrical resistivity would possess a samples coated by MAO for 300 s were composed of peaks associated
number of electrons to move across the interfaces easier than what the with Mg, MgO, and MgSiO3 as shown in the Fig. 4g. A significant dif-
thick passive film acted unless the thickness of the passive film was ference was detected rarely due to the same substrate and electrolyte
uniform entirely [23]. As a result, myriads of electrons at the ridge area used in this study. The thickness values of the final oxide layers in FS
in PE field would trigger the initial appearance of micro-discharge with and GS samples were measured to be ~11 and ~14 μm, respectively
ease. This was consistent with Fig. 2a that the time to undergo break- after MAO for 300 s as shown in Fig. 4h. Hence, the growth rate of GS
down in GS sample was shorter than that in FS counterpart. It is con- sample was higher than that of FS sample. This finding would be in line
cluded that the preferential nucleation of the oxide fragments was fa- with final responding voltage of GS sample higher than that of FS
cilitated first at the ridge area whereas a relatively uniform the oxide counterpart. All results were also reinforced indirectly by AFM ob-
layer with a thin thickness of 1.2 μm was formed on FS sample, prob- servations as shown in Fig. 5. For FS sample with initial substrate
ably owing to the occurrence of homogeneous nucleation. As the nu- roughness was measured to be ~0.025 μm, the values of Ra would tend
cleation of the oxide fragments related closely to surface roughness to increase with respect to coating time, reaching ~0.05, however, the
might have an influence on subsequent process to develop the oxide initial Ra value of GS sample measured to be ~1 μm would tend to
layer, it is necessary to look into the growth behavior of the oxide layer decrease, reaching ~0.1 μm. Thus, it is inferred that the growth rate of
in the next section. the oxide layer at the valley area was higher than that at the ridge area
in the grooved condition.
3.3. Growth behavior of oxide layer In general, the stress would be considered during MAO coating. As
discussed by Clyne and Troughton [26] in the area of plasma electro-
Fig. 4 displays the growth behavior of the present oxide layer lysis, the generation of stress was known to be related to discharge
through SEM micrographs of the cross-sectional morphologies in FS and characteristics of high-energy plasma arcs that were affected by several
GS samples subjected to MAO for 120 and 300 s together with XRD data variables, such as electrical conditions (current density, frequency, etc.)
of both samples after MAO for 300 s. A graph of the thickness of the and substrate states (crystallinity, surface roughness, etc.). Thus, the
oxide layer with respect to coating time up to 300 s is provided to figure formation of stress might depend on surface roughness. More im-
out the different growth rate of the oxide layer. As seen from Figs. 4a portantly, however, stress would be relaxed by thermal contribution of
and b, the oxide layers with not only fairly uniform thickness but also plasma discharges. During MAO where plasma-assisted electrochemical
relatively dense structure were detected clearly on FS sample. If the reactions governed, the local temperature of transient plasma arcs was
nucleation of the oxide layer occurred uniformly on the substrate reported to be as high as ~2000 K, exceeding the melting temperature
during ignition, the occurrence of micro-discharges would take place of the newly-formed oxide layer. For example, Hussein et al. [25] es-
evenly throughout the surface. This finding was confirmed by surface timated the plasma electron temperature ranging from 4500 to
morphologies of the oxide layer that showed micro-pores which are 10,000 K linked with collective discharge behavior by means of optical
fairly uniform in size (Fig. 4c). In case of GS sample, the contour of the emission spectroscopy. This result was obtained from 1100 Al alloy.
interface between the oxide layer and substrate that defined the ridge Indeed, the present group reported the temperature of plasma arcs on
and valley areas was still visible as shown in Figs. 4d and e. In addition, AZ91 Mg alloy in the range of 2116–2643 K [27]. Thank to high-tem-
the oxide layer in GS sample exhibited more porous structure than that perature, the spontaneous phenomena including solidification and (re-)
in FS sample due to the occurrence of discharge channels that would melting of Mg oxide would take place. Although the evident expansion
provide a short path for harmful ions from corrosive environment. The of plasma arcs after critical voltage affected the local surface area,
present work confirmed the effect of surface roughness on controlling thermal relaxation resulting from the concurrence of high-energy
the overall characteristics of the micro-defects present in the oxide plasma arcs was dominant. As a result, stress generated would tend to
layer. Surface roughness affected directly discharge characteristics be relieved. Thus, the present process differed from the stress devel-
during growth in which GS sample exhibited plasma arcs with various opment during conventional depositions [28]. In this study, the
size and distribution as shown in the inset of Fig. 2a. These plasma morphologies of the oxide layer as well as defect structure were related
activities would modify the defect-structure of the oxide layer in terms closely to plasma characteristics.
of density and size on the basis of the morphological comparison be-
tween Figs. 4c and f. The pores density on the surface in GS and FS 3.4. Electrochemical response of oxide layer
sample treated by MAO for 300 s showed some differences in different
locations as shown in Figs. 4c and f. Considering the bimodal size dis- Electrochemical responses of the oxide layers on GS and FS samples
tribution of plasma arcs shown in the inset of Fig. 2a in case of GS evaluated by potentiodynamic polarization and electrochemical

5
N. Nashrah, et al. Applied Surface Science 497 (2019) 143772

Fig. 4. Representative SEM images showing the cross section structures after MAO of (a) FS and (d) GS samples for 120 s, and of (b) FS and (e) GS samples for 300 s,
the surface structures of (c) FS and (f) GS samples for 300 s. (g) XRD spectra of FS and GS samples. The scan ranges are from 10 to 90° with Cu Kα radiation source and
(h) the thickness variation of the oxide layers formed on FS and GS samples with respect to coating time. The growth rate is calculated based on a linear ap-
proximation at the point of 120 s.

impedance tests are presented in Fig. 6. The values of corrosion para- FS sample showed higher impedance value than that on GS sample.
meters, such as corrosion current density (Icorr), corrosion potential Based on equivalent circuit model, it is of importance to note that a
(Ecorr), and Tafel slopes of anodic and cathodic branches (βa and βc) significant difference in Ro value was found between FS and GS samples
were tabulated in Table 2. Polarization resistance (Rp) was calculated whereas the value of Ri remained unchanged each other. On the con-
using representative equation proposed by Stern-Geary [29]. The cor- trary, the value of Ro in FS sample was approximately one-order higher
rosion analysis based on Tafel iteration revealed that the Rp values of than that in GS sample. This was attributed to high porosity in the oxide
GS sample was approximately two-order lower than that of FS sample layer formed on GS sample, which was consistent with microstructural
due to the porous structure of the oxide layer formed on GS sample. observations shown in Fig. 4. In contrast to FS sample, the existence of
EIS analysis was used to describe the protective capability by con- appreciable discharge channels in GS sample would provide the short
sidering microstructural characteristics of the oxide layer of FS and GS paths for Cl− ions that penetrated toward the substrate (Table 3).
samples in a quantitative manner. As reported earlier [30], the present Although the micro-defects exhibited negative effect on corrosion
oxide layer formed by MAO would comprise two different layers where resistance aforementioned, the enlarged surface area by the micro-
the inner layer were thin and compact while the outer layer were thick pores would be beneficial for interface anchoring and heterogeneous
and porous in nature. From Nyquist plot shown in Fig. 6b, the im- compositing. For instance, the ‘porous’ Mg-phosphate layer after MAO
pedance features of the samples were proportional directly to the whole provided preferential sites for secondary substances via chemical in-
diameters of their semicircles, suggesting that the oxide layer formed on terlocking, such as TiO2 and chitosan on OH-terminated Mg-phosphate

6
N. Nashrah, et al. Applied Surface Science 497 (2019) 143772

Fig. 5. Representative AFM images showing the surface roughness after MAO of (a) FS and (c) GS samples for 120 s, and of (b) FS and (d) GS samples for 300 s.

Fig. 6. (a) Potentiodynamic polarization curves of FS and GS samples treated by MAO for 300 s, ranging from −0.25 to 0.4 V vs. open circuit potential at a scan rate
of 1 mV·s−1 and (b) Nyquist plots taken from 106 to 10 Hz with 10 mV rms perturbation. The inset in Fig. 6b shows the equivalent circuit model that describes
electrochemical response. Both testing are performed in 3.5 wt% NaCl solution.

Table 2
Potentiodynamic polarization results of FS and GS samples treated by MAO for 300 s, which are measured from −0.25 to 0.4 V vs. open circuit potential at a scan rate
of 1 mV·s−1 in 3.5 wt% NaCl solution.
Sample Ecorr (V) Icorr (A·cm−2) βa (V·decade−1) βc (V·decade−1) Rp (Ω·cm2)

FS −0.64 7.38 × 10−9 0.78 0.31 3.09 × 107

GS −0.94 2.25 × 10−7 0.93 0.24 6.37 × 105

7
N. Nashrah, et al. Applied Surface Science 497 (2019) 143772

Table 3
Electrochemical impedance results of FS and GS samples treated by MAO for 300 s, which are measured from 106 to 10 Hz with 10 mV rms perturbation in 3.5 wt%
NaCl solution. All values are obtained based on equivalent circuit model in inset of Fig. 6b. In this model, Rs, Ro, and Ri refer to the resistance of the electrolyte, outer
layer, and inner layer, respectively. CPEo and CPEi are the constant phase elements related to the outer and inner layers, respectively.
Sample Rs (Ω·cm2) Ro (Ω·cm2) CPEo (S·sn·cm−2) Ri (Ω·cm2) CPEi (S·sn·cm−2)

FS 24.3 4.04 × 105 1.15 × 10−13 5.38 × 106 1.74 × 10−11

GS 23.6 1.83 × 104 2.15 × 10−12 1.91 × 106 9.69 × 10−10

Fig. 7. A schematic illustration on formation mechanism of the oxide layer formed on grooved surface of AZ31 Mg alloy: (a) initial condition with movement of
anions, (b) the onset of breakdown, (c) nucleation of the oxide fragments formed on ridge area, and (d) growth of the oxide layer as a final stage.

surface [31–33]. Such anchoring, in turn, sealed the micro-pores ef- was in contrast to FS sample where any flaws were removed on surface.
fectively to improve corrosion resistance of Mg alloy, enhancing the In GS sample with relatively large ratio of w/d, the valleys with higher
potential of MAO. surface area than the ridges would provide a preferable site for the
accumulation of anion species, such as OH– and SiO32− in the present
3.5. Formation mechanism of oxide layer on grooved surface electrolyte as shown in Fig. 7a. This promoted the chemical reactions at
the valleys to a greater extent than that at the ridges [22,35]. Hence,
All microstructural evidences aforementioned supported that the the thickness of the passive film formed on the valley was presumably
different surface roughness of the substrate played a critical role in the thicker than that on the ridges, resulting in non-uniform resistivity on
nucleation and growth behavior of the oxide layer formed on AZ31 Mg surface. The formation of the passive film exhibited a linear increase in
alloy via MAO coating. The schematic illustration shown in Fig. 7 de- responding voltage due to the steady growth of the passive film as
scribes the formation mechanism of the oxide layer in a chronological shown in Fig. 2a. Once the passive film would no longer withstand the
order by considering meaningful variation in surface topographies. high electric field by MAO, the number of electrons from the electrolyte
Prior to MAO, the interaction of ions in the present electrolyte with the would be injected into the passive film, generating the emission of
metal substrate would result in the formation of the fragile passive film plasma sparks on the surface as depicted in Fig. 7b. Since the dielectric
through chemical reactions similar to those by anodization [34]. These breakdown would take place first at the ridge area due to lower elec-
chemical reactions to form the passive film were affected locally by the trical resistance as compared to that at the valley area, it was clear that
ridge and valley in GS sample with a uniform period of ~10 μm, which the oxide fragments were nucleated preferentially on the ridge area.

8
N. Nashrah, et al. Applied Surface Science 497 (2019) 143772

After the nucleation occurred at the ridge area, the concurrent J. Li, E. Ma, Z. Shan, Turning a native or corroded mg alloy surface into an anti-
transition to develop next breakdown from the ridge to the valley corrosion coating in excited CO2, Nat. Commun. 9 (2018) 4058.
[3] H.F. Guo, M.Z. An, Growth of ceramic coatings on AZ91D magnesium alloys by
would be favored by the reason that the presence of premature surface micro-arc oxidation in aluminate-fluoride solutions and evaluation of corrosion
that has not fully covered by the oxide fragments would create a resistance, Appl. Surf. Sci. 246 (2005) 229–238.
pathway with possessing relatively low electrical resistance. This was [4] Y.W. Song, D.Y. Shan, E.H. Han, High corrosion resistance of electroless composite
plating coatings on AZ91D magnesium alloys, Electrochim. Acta 53 (2008)
described in Fig. 7c. Thus, consecutive breakdown phenomena of GS 2135–2143.
sample would differ from FS sample, which was confirmed by both [5] A. Francis, Y. Yang, A.R. Boccaccini, A new strategy for developing chitosan con-
population and size distribution of micro-discharges shown in Fig. 2b. version coating on magnesium substrates for orthopedic implants, Appl. Surf. Sci.
466 (2019) 854–862.
Such distribution of micro-discharges was observed to be bimodal in GS [6] M. Kaseem, Y.G. Ko, On the compactness of the oxide layer induced by utilizing a
sample after MAO coating for 300 s. These micro-discharges con- porosification agent, Appl. Surf. Sci. 473 (2019) 715–725.
centrated at the valley region would facilitate plasma-enhanced elec- [7] A. Nominé, J. Martin, G. Henrion, T. Belmonte, Effect of cathodic micro-discharges
on oxide growth during plasma electrolytic oxidation (PEO), Surf. Coat. Technol.
trochemical reactions to thicken the oxide layer with an aid of in-
269 (2015) 131–137.
creasing surrounding temperature which would induce local [8] W. Al Zoubi, Y.G. Ko, Flowerlike organic-inorganic coating responsible for extra-
overheating [36]. This fact allowed the oxide layer at the valley area to ordinary corrosion resistance via self-assembly of an organic compound, ACS
grow at a relatively higher rate than that at the ridge area. On the other Sustain. Chem. Eng. 6 (2018) 3546–3555.
[9] M.P. Kamil, M. Kaseem, Y.G. Ko, Soft plasma electrolysis with complex ions for
hand, the group of intense micro-discharges at the valley led to the optimizing electrochemical performance, Sci. Rep. 7 (2017) 44458.
formation of appreciable discharge channels as short-circuit path as [10] K.R. Shin, Y.S. Kim, H.W. Yang, Y.G. Ko, D.H. Shin, In vitro biological response to
shown in Fig. 7d. The final morphologies of the oxide layer in GS the oxide layer in pure titanium formed at different current densities by plasma
electrolytic oxidation, Appl. Surf. Sci. 314 (2014) 221–227.
sample were found to be more porous than those in FS sample. [11] S.C. Troughton, T.W. Clyne, Cathodic discharges during high frequency plasma
Although the porous structure would suffer from the facile pene- electrolytic oxidation, Surf. Coat. Technol. 352 (2018) 591–599.
tration of various corrosive ions onto the metal substrate, the present [12] S. Fatimah, H.W. Yang, M.P. Kamil, Y.G. Ko, Control of surface plasma discharge
considering the crystalline size of Al substrate, Appl. Surf. Sci. 477 (2019) 60–70.
study would offer a new insight into the significant role of surface [13] L. Zhu, R.S. Petrova, J.P. Gashinski, Z. Yang, The effect of surface roughness on
roughness on the modification of micro-discharge characteristics. The PEO-treated Ti-6Al-4V alloy and corrosion resistance, Surf. Coat. Technol. 325
plasma activities determined by surface roughness of the substrate (2017) 22–29.
[14] M. Zhang, L. Deng, D. Yao, J. Jin, X. Wang, Effect of surface morphology on the
would modify the defect-structure of the oxide layer in terms of density oxidation behavior of bulk metallic glass, Corros. Sci. 98 (2015) 626–633.
and size. As a result, the surface roughness of the substrate that would [15] H. Hanjun, C. Zhen, L. Xingguang, F. Xingguo, Z. Yugang, Z. Kaifeng, Z. Hui, Effects
facilitate the distinctive discharge characteristics during MAO would of substrate roughness on the vacuum tribological properties of duplex PEO/
bonded-MoS2 coatings on Ti6Al4V, Surf. Coat. Technol. 349 (2018) 593–601.
affect the structure of the oxide layer in a direction toward the enlarged
[16] W.B. Dai, L.X. Yuan, C.Y. Li, D. He, D.W. Jia, Y.M. Zhang, The effect of surface
surface area that would be anticipated to be beneficial for inducing roughness of the substrate on fatigue life of coated aluminum alloy by micro-arc
emerging properties like bio-implanting. oxidation, J. Alloys Compd. 765 (2018) 1018–1025.
[17] B. Yoo, K.R. Shin, D.Y. Hwang, D.H. Lee, D.H. Shin, Effect of surface roughness on
leakage current and corrosion resistance of the oxide layer on AZ91 Mg alloy pre-
4. Summary pared by plasma electrolytic oxidation, Appl. Surf. Sci. 256 (2010) 6667–6672.
[18] J. Wang, S. Huang, H. Huang, M. He, P. Wangyang, L. Gu, Effect of micro-groove on
The formation mechanism of the oxide layer in AZ31 Mg alloy microstructure and performance of MAO ceramic coating fabricated on the surface
of aluminum alloy, J. Alloys Compd. 777 (2019) 94–101.
processed by MAO coating was investigated by taking the effect of [19] Y.G. Ko, S. Namgung, D.H. Shin, Correlation between KOH concentration and
surface roughness into account. Irrespective of surface roughness used surface properties of AZ91 magnesium alloy coated by plasma electrolytic oxida-
in this study, the time to reach breakdown in GS sample was earlier tion, Surf. Coat. Technol. 205 (2010) 2525–2531.
[20] W. Al Zoubi, M.P. Kamil, H.W. Yang, Y.G. Ko, Electron-donor and -acceptor agents
than that in FS sample due to the fact that the nucleation of the oxide responsible for surface modification optimizing electrochemical performance, ACS
fragments was prone to occur preferentially at the ridge area. This re- Appl. Mater. Inter. 9 (2017) 28967–28979.
sult suggested that the nucleation kinetics would be triggered with in- [21] M.P. Kamil, H.W. Yang, Y.G. Ko, Impurity-induced discharges in plasma electro-
lysis, Electrochem. Commun. 91 (2018) 54–59.
creasing surface roughness. Once the micro-discharge became in-
[22] Y. Zuo, H. Wang, J. Xiong, The aspect ratio of surface grooves and metastable
tensified subsequent to breakdown, the oxide layer of GS sample grew pitting of stainless steel, Corros. Sci. 44 (2002) 25–35.
locally on the pre-existing oxide fragments whereas the oxide layer of [23] J.W. Schultze, M.M. Lohrengel, Stability, reactivity and breakdown of the passive
films. Problems of recent and future research, Electrochim. Acta 45 (2000)
FS sample grew uniformly throughout the entire surface which implied
2499–2513.
that the growth rate of GS sample was higher than that of FS counter- [24] S.C. Troughton, A. Nominé, J. Dean, T.W. Clyne, Effect of individual discharge
part. Although GS sample possessed the thicker oxide layer of ~14 μm cascades on the microstructure of plasma electrolytic oxidation coatings, Appl. Surf.
than FS sample after MAO for 300 s, however, the oxide layer contained Sci. 389 (2016) 260–269.
[25] R.O. Hussein, X. Nie, D.O. Northwood, A. Yerokhin, A. Matthews, Spectroscopic
somewhat noticeable discharge channels with an aid of intense micro- study of electrolytic plasma and discharging behavior during the plasma electrolytic
discharges that were favored at the valley area during growth of the oxidation (PEO) process, J. Phys. D. Appl. Phys. 43 (2010) 105203.
oxide layer. Controlled by grooved surface, such a porous structure of [26] T.W. Clyne, S.C. Troughton, A review of recent work on discharge characteristics
during plasma electrolytic oxidation of various metals, Int. Mater. Rev. 64 (2019)
the oxide layer that might sacrifice long-term electrochemical stability 127–162.
would be desirable to achieve the enlarged surface area responsible for [27] K.M. Lee, B.U. Lee, S. Il Yoon, E.S. Lee, B. Yoo, D.H. Shin, Evaluation of plasma
functional properties, such as medical implant, photo catalysis, dye- temperature during plasma oxidation processing of AZ91 Mg alloy through analysis
of the melting behavior of incorporated particles, Electrochim. Acta 67 (2012)
sensitized solar cell, etc. 6–11.
[28] G. Cao, Y. Yun, H. Xu, G. Yuan, J. Hu, G. Shao, A mechanism assessment for the
Acknowledgment anti-corrosion of zirconia coating under the condition of subcritical water corrosion,
Corros. Sci. 152 (2019) 54–59.
[29] M. Stern, A.L. Geary, A theoretical analysis of the shape of polarization curves, J.
This work was supported by the Fundamental Research Project Electrochem. Soc. 104 (1957) 56–63.
funded by the National Research Foundation of Korea, Republic of [30] K.R. Shin, Y.S. Kim, G.W. Kim, Y.G. Ko, D.H. Shin, Development of titanium oxide
layer containing nanocrystalline zirconia particles with tetragonal structure:
Korea (NRF-2017R1D1A1A09000921).
structural and biological characteristics, Colloids Surfaces B Biointerfaces 131
(2015) 47–53.
References [31] Y. Zhang, K. Bai, Z. Fu, C. Zhang, H. Zhou, L. Wang, S. Zhu, S. Guan, D. Li, J. Hu,
Composite coating prepared by micro-arc oxidation followed by sol-gel process and
in vitro degradation properties, Appl. Surf. Sci. 258 (2012) 2939–2943.
[1] A.L. Yerokhin, X. Nie, A. Leyland, A. Matthews, S.J. Dowey, Plasma electrolysis for [32] G. Cao, L. Wang, Z. Fu, J. Hu, S. Guan, C. Zhang, L. Wang, S. Zhu, Chemically
surface engineering, Surf. Coat. Technol. 122 (1999) 73–93. anchoring of TiO2 coating on OH-terminated Mg3(PO3)2 surface and its influence
[2] Y. Wang, B. Liu, X. Zhao, X. Zhang, Y. Miao, N. Yang, B. Yang, L. Zhang, W. Kuang, on the in vitro degradation resistance of Mg-Zn-Ca alloy, Appl. Surf. Sci. 308 (2014)

9
N. Nashrah, et al. Applied Surface Science 497 (2019) 143772

38–42. Surf. Sci. 425 (2017) 83–94.

[33] K. Bai, Y. Zhang, Z. Fu, C. Zhang, X. Cui, E. Meng, S. Guan, J. Hu, Fabrication of [35] R. Walter, M.B. Kannan, Y. He, A. Sandham, Effect of surface roughness on the in
chitosan/magnesium phosphate composite coating and the in vitro degradation vitro degradation behaviour of a biodegradable magnesium-based alloy, Appl. Surf.
properties of coated magnesium alloy, Mater. Lett. 73 (2012) 59–61. Sci. 279 (2013) 343–348.
[34] Y. Long, D.G. Li, D.R. Chen, Influence of square wave anodization on the electronic [36] A.L. Yerokhin, L.O. Snizhko, N.L. Gurevina, A. Leyland, A. Pilkington, A. Matthews,
properties and structures of the passive films on Ti in sulfuric acid solution, Appl. Discharge characterization in plasma, J. Phys. D. Appl. Phys. 36 (2003) 2110–2120.

10