Fumin Xu, Ln Luo,b,?, Lei Xiong, Yong Liu,?

aKey Laboratory of Lightweight and High Strength Structural Materials of Jiangxi Province, Nanchang University, Nanchang 330031, China

b School of Materials Science and Engineering, Nanchang University, Nanchang 330001, China

Abstract There remains growing interest in magnesium(Mg)and its alloys,as they are the lightest structural metallic materials and potential metallic biomaterials.In spite of the greatest historical Mg usage at present, the wider use of Mg alloys remains restricted by the poor corrosion resistance.A nano amorphous film as the composition of Al2O3, had now been deposited on the AZ31 Mg alloy substrate by atomic layer deposition (ALD).Grazing incidence X-ray diffraction (GIXRD), X-ray reflect vity (XRR), X-ray photoelectron spectroscopy (XPS), atomic force microscope (AFM) and scanning electron microscopy (SEM) had been employed to identify the chemical compositions, microstructure and Al2O3/Mg interface of specimens firstl .Then corrosion behavior had been evaluated by neutral salt spray test and electrochemical measurement.The results showed that nano amorphous fil made a homogeneous cover on Mg alloy.The fil could improve the corrosion resistance of Mg alloy greatly, not only with a positive shift in Ecorr and a decrease in icorr, but also with a more uniform corroded mode.Furthermore, the roughness was found to be an important factor for corrosion resistant, in the way that rougher surface was corroded worse,and greater improvement would be in corrosion resistant after nano amorphous fil deposition.

Keywords: Mg alloys; Nano amorphous film Corrosion resistance; Surface roughness; Electrochemical measurement.

1.Introduction

Magnesium and its alloys are widely applied in the automotive, aerospace and electronics industries due to their outstanding physical and mechanical properties such as low density, high specifi strength, good vibration and cast-ability[1,2].In addition, utilization of Mg and its alloys as potential candidates for bone implants has drawn increasing attention, owing to their mechanical match with living bone and gradual degradation in the human physiological environment[3,4].However, the poor corrosion resistance in the working environment has become the main drawback, which restricts the applications of Mg and its alloys on a larger scale[5-7].Therefore, it is necessary to improve the corrosion resistance.

Lots of efforts have been done as alloying effects [5,8] or surface modificatio [2,9] while surface modificatio is considered being very effective, such as alkaline heat treatment[10,11], electrodeposition [12], conversion treatment [13,14],micro-arc oxidation [15,16], polymeric coating [17,18] and corrosion inhibiting molecules [19,20].It is commonly that the corrosion resistance would be improved by modifying barrier micrometer coatings.However, it is a great challenge to develop the nanometer protective film for the corrosion resistance improvement, which would lead to higher dimensional accuracy and lower modificatio cost for Mg components [21,22].Recent studies have confirme that nanometer amorphous film can also establishing an effective barrier against further oxidation for some metals [23,24].Is it still work on chemical active Mg alloys? Would is it also greatly sensitive by the surface roughness as the micrometer coatings[25,26]?

In this paper, the interplay between surface roughness and nanometer fil of the corrosion behavior was studied.A nanometer amorphous fil as Al2O3has been made on Mg alloy with different surface roughness to establish the effect on corrosion behavior.To ensure the fabrication quality, the deposition is carried out by atomic layer deposition[22,27] (ALD, based on sequential, self-limiting and selfsaturating surface chemical reactions, which is suitable for ultrathin fil fabrication).The fil on Mg alloy were characterized firstl (morphology, composition, phase, thickness,etc.), then corrosion behavior was evaluated by neutral salt spray test and electrochemical measurement.The corrosion mechanism for Mg alloy with nanometer fil and different surface roughness had been deduced finall .

2.Experimental procedure

2.1.Materials

AZ31 Mg alloy (wt%: 3.04% Al, 0.84% Zn, 0.3% Mn,0.014% Si, 0.002% Fe, 0.0049% Ca, 0.0012% Cu, and Mg balance) had been used as substrates in the size of 50×50×0.5mm3.For experiment, the specimens had been mechanically polished with emery papers of grade 1000 and diamond powder (diameter 0.05 μm) respectively.Therefore,the various surface roughness of the AZ31 Mg alloy substrates had been obtained for comparison.Then the rough substrate had been named as S1 and smooth substrate had been named as S2 respectively.After polishing, all the substrates had been cleaned with ultrasonic cleaner (PS-20) in isopropanol for 30min.After the cleaning,deoxygenation and dehydrogenation had been introduced on substrates in vacuum oven (DZF-6020) at 250 °C for 2h, then ALD process had been carried out.

2.2.Al2O3 fil fabrication

The Al2O3fil had been deposited by ALD (PEALD-100A) on AZ31 Mg alloy substrates.During the ALD process, the Al (CH3)3(trimethyl aluminum, TMA) and H2O (water vapor) had been the precursors for aluminum and oxygen, respectively.Deposition had been conducted in a fl w reactor at a pressure about 0.15Torr and a temperature of 100 °C, with N2as a carrier gas.The ALD process had consisted of identical cycles.Each cycle had contained the following sequence: H2O/0.02s→N2purge/60s→TMA/0.02s→N2purge/60s.The thickness of ALD fil can be precisely controlled by the number of growth cycles.When S1 and S2 had been deposited with Al2O3film they had been named as SF1 and SF2 respectively.

2.3.Microstructure characterization

Grazing incidence X-ray diffraction (GIXRD, PANalytical X’Pert PRO, incidence beam angle (α) is 1°, at 40kV and 40mA with the Cu-Kαradiation) had been employed to identify the fil phase constitutes.X-ray reflect vity (XRR)had investigated the fil thickness by measure the fil deposited on the glass substrate simultaneously.Atomic force microscope (AFM, Agilent 5500) had been used to measure the surface roughness by the average of ten measurements with a standard deviation of 0.1μm.Optical microscopy(OM, BX51M) had been used to observe the surface topography.Scanning electron microscopy (SEM, FEI Quanta 200F) with energy-dispersive spectroscopy (EDS, Oxford Xman,<130kV) had resolved the microstructure and composition in details.X-ray photoelectron spectroscopy (XPS,ESCALAB25OXI) with a monochromatic AlKα(1486.6eV)line of an X-ray source had been conducted to investigate the chemical bonding state and composition of samples.All binding energies calibrated by 284.8eV of adventitious carbon contamination (C 1s), except carbon materials.

The mechanical properties of the specimens were evaluated using nano-indentation tester (Fischerscope, HM2000).Microhardness tester (HV-1000A) had measured the microhardness and the load for the appearance of fil wrinkle had been determined with OM-examinations of Vickers hardness indentations.

2.4.Neutral salt spray test

Neutral salt spray test was performed to evaluate the longterm influenc function of the surface roughness and fil to bare substrates following the standard method ASTM B-117.The 3.5wt%NaCl solution spray at 35±2°C in the salt spray chamber and the salt spray test lasted for 48h.Reproducibility was confirme by three identical samples at the same conditions.The samples then were sent to SEM/ EDS without any further treatment.

2.5.Electrochemical measurements

Electrochemical impedance spectroscopy (EIS) and Potentio-dynamic polarization (PDP) had been performed using an Electrochemical Analyzer/Workstation (CHI650D)in 3.5wt% NaCl aqueous solution at room temperature.A three-electrode cell set-up had been used in which the prepared sample had been the working electrode and a platinum sheet and a saturated calomel electrode had been used as the counter and reference electrodes, respectively.

A stable open-circuit potential (OCP) had been established prior to PDP (?2 ～?1V/SCE at a scan rate of 1 mV·s?1)and EIS (at a disturbing potential of 10mV over a frequency range as 105～10?2Hz) measurements.Furthermore, ZView software (Version 3.0.0.22, UK) was used to get the equivalent circuits based on EIS data, as errors for the individual parameters<10% and chi-squared values<0.01.Reliability had been verifie by testing at least two specimens.

Fig.1.OM images of (a) S1, (b) SF1, (c) S2, (d) SF2; AFM images (inserts) of (e) S1, (f) S2 and XRR patterns (inserts) of (g) SF1, (h) SF2.

3.Results and discussion

3.1.Film analysis

Fig.1 illustrated the OM images of the AZ31 substrates with and without ALD-Al2O3film The insets gave the information on roughness and thickness.In Fig.1(a)(c), S1 and S2 displayed scrubbed traces, holes and some big inclusion protruding particles and the scrubbed traces in S1 were deeper than the one in S2.The AFM images of S1 and S2 (Fig.1(e)(f)) confirme that the surface of S1 was rougher than S2.Roughness values obtained by AFM for S1 and S2 were 14.00±0.1μm and 8.44±0.1μm, respectively.As expected, the surface defects (such as scratches)of specimens could be reduced after polishing with diamond powder which resulted in the decrease of local topographic heterogeneities.In Fig.1(b)(d), SF1 and SF2 almost kept the original roughness, as scrubbed traces and some big inclusion protruding particles could still be seen.Based on XRR results(Fig.1(g)(h)), the fil thickness could be calculated out by the following Eq.(1) [28,29]

In which D was the thickness of thin filmλwas the wavelength of X-ray,θcwas the critical angle of total reflectionθm,m+1was constructive interference diffraction peaks of incident angle.The fil thickness values for SF1 and SF2 were 18.44±0.01nm and 18.46±0.01nm, respectively.

The phase constituent of specimens had also been studied by using GIXRD (as shown in Fig.2).The GIXRD patterns illustrated thatα-Mg was the major phase for all the specimens and the crystal orientation ofα-Mg wasn’t influence significantl by the ALD surface modification Compared S1,S2 with SF1, SF2 respectively, there was just a small diffraction drum at (4 0 2) planes for Al2O3phase (the drums were more obvious in the amplifie figures Fig.2(e) and(f)), which indicated the fil was amorphous.It was worth noticing that amorphous structure might contribute to improve the corrosion resistance,which was similar to previous reports[23,30,31].

The second electronic images (Fig.3(b)(d)) gave another proof for that the surface roughness had not changed much after ALD deposition.As the big inclusion protruding particles and surface scratch still could be seen.However, the small precipitation and defects (Fig.3(e)(g)) were covered by the compact Al2O3film which leaded to the distribution of surface elements became more homogeneous.By the higher magnificatio images (Fig.3(f)(h)) of reign 4, and region 8,it could be noted Al2O3fil exhibited the similar condensed and compact microstructure even if the surface roughness was different, confirmin that surface roughness wouldn’t affect the surface morphology and structure of Al2O3film

EDS mapping images showed that all the elements (Mg,Al, Zn, Mn, O) were uniformly distributed (as shown in Fig.3(i), (j), (k) and (l)).Data of EDS analysis was listed in Table 1.The inclusion protruding particle and plane region were chosen as the typical area.After ALD deposition, O content increased apparently (5.77-4.71wt%) and Al content in the surface (5.74-4.57wt%) was higher than the one in AZ31 substrates (2.89-2.37wt%).It was found the compositions of point 3, area 4 (for SF1) and point 7, area 8 (for SF2) were similar to each other.It also indicated amorphous condensed Al2O3nano fil could lead to the distribution of surface elements become more uniform.

Fig.2.GIXRD patterns of (a) S1, (b) SF1, (c) S2, (d) SF2.The amplifie ones (inserts) of diffraction drums for amorphous Al2O3 in (e)SF1, (f)SF2.

Table 1Local EDS analysis of the specimens corresponding to Fig.3.

The chemical composition of the specimens was examined by XPS.Fig.4a and d exhibited XPS survey spectra of S1 and SF1 respectively, which consist from several peaks belongs to electronic states of Mg 1s, O1s and Al 2p.

The O 1s spectra (Fig.4b) consisted of two peaks located at 533.10eV and 531.12eV.The high binding energy (BE)peak came from chemisorbed OH?and CO32?, and low BE peak came from the MgO oxygen atom (O2?) [32].After ALD-deposited Al2O3film the O 1s spectra (Fig.4e) could be deconvoluted into three peaks.The additional component at a highest BE peak of O 1s could be ascribed to contamination,and the BE peak at 533.86 and 531.04eV ascribed to the chemisorbed OH?and Al2O3, respectively [23].Besides, the intensity of the component assigned to Al2O3increased and that related to chemisorbed oxygen decreased after coating,which indicated the efficien barrier of film

The Al 2p spectra (Fig.4c) could be deconvoluted into two bands located at 75.97eV and 71.06eV, respectively.The former band should be ascribed to the co-existence of Al oxide and/or Al hydroxide, the latter one should ascribe to the metallic Al [32,33].After ALD-deposited Al2O3film the spectra of Al 2p could be deconvoluted into two bands located at 76.04eV and 74.26eV, which were assigned to Al-OH bonds and Al2O3, respectively (Fig.4f).Besides, the BE difference between Al 2p and O 1s was 456.78eV for SF1,which was in a good agreement with reported values of fully oxidized amorphous Al2O3[34,35].

Fig.3.SEM images of the (a, e) S1, (b, f) SF1, (c, g) S2 and (d, h) SF2; the elemental mapping images of (i) S1, (j) SF1, (k) S2 and (l) SF2.

Fig.4.XPS survey spectra (a, d) and corresponding high-resolution XPS spectra of (b, e) O 1s, (c, f) Al 2p; (a-c) for S1, (d-f) for SF1.

Fig.5(a) showed the load-depth curves of instrumented indentation testing at a constant loading rate of 10 mN/s with the range of 0-100 mN.All the curves were smooth without any pop-in or pop-out behavior.The curves of SF1 and SF2 were fluen without any zigzag fluctuatio and disconnection,which indicated that Al2O3fil was smooth and no-cracking during the loading process.Furthermore, SF2 had the shallower depth under the same pressing load, which meant SF2 had the stronger resistance to external load, so it had the stronger resistance to plastic deformation.Fig.5(b) described the mean values for hardness, Young’s modulus and slenderness.After ALD deposittion, the hardness and slenderness increased while Young’s modulus decreased.In addition, the hardness and slenderness also increased while Young’s modulus decreased with the decrease of roughness.

Fig.5.(a) Indentation force versus depth curves, (b) the value of microhardness, Young’s modulus and slenderness obtained from (a), and (c) surface morphology after deformation by press indentation.

In addition, Fig.5(c) showed the surface morphology of loading sites in a Vickers indentation.The load for the appearance of fil wrinkle was determined with OM-examinations of Vickers hardness indentations and now served as an indirectly measure for adhesion [36].As shown in Fig.5(c),the adhesion of SF1 exhibited better than SF2 as less wrinkle exhibited.According to previous reports [37], the reason for the weaker adherence of the ALD alumina fil on S2 could be the excessive smoothness of the substrate surface.A rough surface roughness might be necessary for good fil adhesion and an initial roughness had been shown to increase the interface area to interlock the coating with the substrate [21].The results of Vickers indentation were coherent to Nanoindentation test (63.42 HV (S1) →67.14 HV (SF1), 73.51 HV (S2) →84.49 HV (SF2)).According to previous reports[23], the high microhardness of Al2O3fil could improve the hardness of substrates.In addition, the plastic deformation would take place when moved continuously over a surface(which related to surface hardening effect).The deeper polishing penetration, the more plastic deformation.As a result,the smoother surface would obtain a higher microhardness value [38].

Based on surface characterization (Section 3.1), following results could be gotten.With the decrease of surface roughness, the local topographic heterogeneities of AZ31 Mg substrate reduced.After the ALD deposited, an amorphous condensed Al2O3nano fil covered the surface with excellent conformity and uniformity.There was no obvious change in the surface roughness as the big inclusion protruding particles and surface scratch could still be seen.However, the distribution of surface elements became more uniform as the small precipitation and defects were covered.The microhardness of bare substrate would increase with the Al2O3fil and smooth surface.In addition, ALD-deposited Al2O3fil showed a good adhesion to the rough surface than smooth surface.

3.2.Corrosion behavior

3.2.1.Neutral salt spray test

Fig.6 showed the SEM morphologies of the specimens after neutral salt spray test.The surface morphology of S1 showed obvious corrosion pits and cracks (marked by yellow circle), indicating that the S1 was subject to severe nonuniform corrosion(Fig.6a).S2 showed the same surface morphology with S1 while the cracks were narrow(Fig.6c).After ALD-deposited Al2O3film the surface morphologies of SF1 and SF2 were fla and dense with fewer cracks and no pits(Fig.6b and d), showing uniform corrosion.So, the corrosion became more uniform after the deposition of Al2O3film Besides,the reducing of surface roughness would also contribute to the uniform corrosion slightly.

Fig.6.SEM images of the specimens of (a) S1, (b) SF1, (c) S2 and (d) SF2 after salt spray test.

Fig.7 showed the cross-sectional images and elemental mapping images of the specimens after salt spray test.Due to the penetration of H2O and Cl?, the corroded products magnesium hydroxide (Mg (OH)2) was produced (the elemental mapping images of Mg and O) [39].As could be seen in Fig.7a, the cross-sectional of S1 exhibited thick corroded layer (marked by yellow circle) with many cracks.With the decrease of surface roughness, the corrosion of S2 became gentle and the corroded layer was thinner than S1 (Fig.7c).In the case of the Al2O3film significan improvement of corrosion resistance was achieved, particularly for the SF2(Fig.7d).Only a shallow and uniform corrosion layer was observed on the SF2.Therefore, it could be speculated that the smooth surface and Al2O3fil improved significantl the corrosion resistance of AZ31 substrate.The elemental mapping showed oxygen distributed on both the corroded layers and the cross-sectional while a strong intensity occurred in the corroded layers.The chloride ion was concentrated in the corroded layer and also diffused into the substrate region surrounding the corroded layer.However, the chloride ion only concentrated with a uniform mode in the dense corroded layer after coating.This indicated not only the dense corroded layer could block the diffusion of Cl?, but also the dense corroded layer made a uniform distribution of Cl?on the surface (which would lead to a uniform corrosion) [40].Thus, the Al2O3fil would promote the formation of a dense corroded layer.

3.2.2.Potentio-dynamic polarization (PDP)

The corroded products of AZ31 substrates consisted of Mg(OH)2, MgAl2O4which were indicated by GIXRD patterns(Fig.8).The corroded products of AZ31 substrate with Al2O3fil were almost the same.However, the intensity of Mg(OH)2peaks for S1 increased apparently, demonstrating that the AZ31 Mg alloy undergone severer corrosion and a lot of new Mg (OH)2corroded products were formed.

The SEM surface morphology of the specimens after potentio-dynamic polarization test was shown in Fig.9.It could be observed that the corroded products and precipitations (white granular and bulk precipitates) covered the whole surface of specimens.In addition, many micro-cracked structures appeared and the surface of specimens exhibited bumps or potholes.Pits in specimens were circled out and the surface of the rough AZ31 substrate suffered from severer pitting corrosion.Cracks might be produced by the lower molar volume of corroded products resulting from dehydration during corrosion.Elemental mapping images showed that all the elements (Mg, O, Al, Zn, Na, Cl) were uniformly distributed(Fig.9(e), (f), (g) and (h)).The contents of aluminum in the precipitation layer of AZ31 Mg alloy with Al2O3fil were higher than those of bare substrates (Table 2).By comparison with Table 1 and Table 2, Mg element content increased apparently for AZ31 Mg alloy and Na, Cl elements also appeared after corrosion.Besides, it was also self-evident that the rougher the surface,the severer corrosion.The amounts of corroded products of SF1 was much more than that of SF2,and the products appeared in bumps or potholes were more than in those plane block.

Fig.7.SEM cross-sectional images of specimens (a) S1, (b) SF1, (c) S2 and (d) SF2 couple with the elemental mapping of Mg, O, C and Cl after salt spray test.

Fig.8.GIXRD patterns after potentio-dynamic polarization test: (a) S1, (b) SF1, (c) S2, (d) SF2.

Fig.9.SEM images after potentio-dynamic polarization test: (a) S1, (b) SF1, (c) S2 and (d) SF2; the elemental mapping images of (e) S1, (f) SF1, (g) S2 and (h) SF2.

Fig.10.(a) Potentio-dynamic polarization curves of the specimens and (b) Potentio-dynamic polarization parameters in 3.5wt% NaCl aqueous solution.

Table 2Local EDS analysis of the specimens corresponding to Fig.9 after potentiodynamic polarization test.

Fig.10 presented the potentio-dynamic polarization curves and main electrochemical parameters in 3.5wt.% NaCl aqueous solution.The data obtained from Tafel curves were shown in Table 3.icorrwas estimated by Tafel experimental method,which might be a little different from the results calculated by the Stern-Geary method [41].The corrosion rate was calculated based on the Faraday’s law [41] by Eq.(2) as:

In where A was the atomic weight of the metal,ρwas the density, n was the number of electrons exchanged in the dissolution reaction, and F was the Faraday constant(26.801 A h/mol).As shown in Fig.10 and Table 3, corrosion rates obtained from the method could be ranked in the increasing order: SF2 (0.01mm/y)

As shown in Table 3, with the decrease of surface roughness (comparing S1 with S2, and SF1 with SF2), there was a positive shift inEcorrvalues, a drop inicorrvalues and a rise in Rp.It indicated the tendency of corrosion initiation decreased due to thermodynamics and the corrosion rate got slow during corrosion,which meant the better corrosion resistance when the surface was smoother.The reason for the better corrosion resistance of the smooth substrate as compared to the rough substrate could be the reducing in local topographic heterogeneities.The defects (such as scrubbed traces,microparticles and holes) on the rough substrates remained as possible sites of corrosion since, in case of exposure to the aggressive environment via channel defects, which would result in trigger and accelerate locally the corrosion process.As shown in reaction (3) and (4), the main corroded products of magnesium degradation during electrochemical reaction in aqueous media were magnesium hydroxide (Mg (OH)2, solubility as 9×10?4g/100ml) and hydrogen gas (H2, solubility as 1.83ml/100ml) [42,43].Corroded products, Mg (OH)2would act as a protective layer and slow down the degradation of magnesium.

Table 3Electrochemical parameters calculated according to the polarization curves.

However, in the presence of chlorides, the hydroxide ions were converted to more soluble magnesium chloride (MgCl2,solubility as 59.6g/100ml) rather than magnesium hydroxide(Mg (OH)2).The Clˉions diffused through the holes and defects in the Mg (OH)2fil resulting in a reaction between Cl?and Mg (OH)2forming MgCl2.The MgCl2dissolved in aqueous solution thereby leaving the surface.

The big protruding particles, scratches and holes could be reduced or even eliminated by fin polishing treatment, and thereby smoothened the surface.Due to the reducing of the local topographic heterogeneities, the uniform surface of substrate would contribute to develop more uniform corrosion modules and form a continuous corroded products layer easily to protect substrate from corrosion.As for rough substrate,the interface area between specimen and corrosion medium would become larger while the surface roughness increased.Besides, the propensity of chloride ions to be preferentially adsorbed at certain non-uniformly distributed distinct defective locations meant that the chloride ions would be nonhomogeneously adsorbed on the specimen surface [40].The surface defects of rough substrate could also act as channel for corrosion medium making corrosion more serious.All of the mentioned above proved that the specimen with rough surface would undergo severer corrosion.The corrosion rate for S1 was about as much as 10.75 times of that for S2.After ALD deposition,while the roughness difference kept same but atomic number contrast difference decreased, the corrosion rate difference became smaller, as the corrosion rate of SF1 was as much as 4 times more than that for SF2.All these indicated that the surface roughness affected the corrosion resistance very much.

After ALD deposition (comparing S1 with SF1, and S2 with SF2), there was also a positive shift inEcorrvalues, a drop inicorrvalues and a rise in Rp.The positive shift ofEcorrmight be mainly ascribed to that the Al2O3fil (not soluble) exhibited excellent chemical inertness, and corrosion resistance than AZ31 Mg alloy [23,44-46].ALD-deposited Al2O3fil could cover the surface defects and make the distribution of surface elements more homogeneous,which could prevent the occurrence of pitting corrosion making corrosion more uniform.At the initial stage, the Al2O3fil effectively blocked the penetration of aggressive chloride ions and H2O molecules.However, the interface area between specimen and corrosion medium would become larger while the substrate surface roughness increased.Thickness of the Al2O3fil reduced as time went by.As for specimen with rough surface,the Al2O3fil would breakdown in some areas and the pitting corrosion would occur accelerating the corrosion rate.Moreover,the degradation of Al2O3fil for specimen with smooth surface would be more uniform.It was easier to form a continuous corroded products layer (Mg (OH)2) on the smooth surface than on the rough and the protective layer would contribute to delay the corrosion [47].Corrosion rate of S1 was about 27.2 times more than that for SF1, and the corrosion rate of S2 was about 11.0 times more than that for SF2,which verifie the Al2O3fil could protect the AZ31 Mg alloy well.

3.2.3.Electrochemical impedance spectroscopy (EIS)

To obtain information on the corrosion behavior of the Al2O3fil and surface roughness,electrochemical impedance spectroscopy (EIS) investigations were carried out for specimens in 3.5wt% NaCl aqueous solution.When an excitation signal with small amplitude was applied, the response depended on the electrode kinetics.It usually consisted of several different sub-processes.By analyzing these responses,the individual sub-process could be deduced.As shown in the Nyquist plots in Fig.11a,the diameter of the curves increased remarkably with the reducing of surface roughness and the deposition of Al2O3fil compared with the curves of the rough bare substrate, indicating that smooth surface and Al2O3fil would protect the magnesium alloy against corrosion [48].In addition, it was well known that materials with a higher Z modulus at lower frequencies exhibited better corrosion resistance for metal substrates [47].As shown in Fig.11(b), theZvalue of the S1, SF1, S2, SF2 were 59.60, 171.00, 71.65,387.21Ω·cm2, respectively.These results demonstrated that the smooth surface and Al2O3fil would enhance the corrosion resistance of the bare substrates.Thus, the evolution of the EIS data demonstrated that the Al2O3fil on the AZ31 Mg alloy could protect AZ31 Mg alloy from corrosion efficientl and the smooth surface would further enhance the corrosion resistance, which was good agreement with the results of the polarization curves.

Fig.11.EIS results of specimens (a) Nyquist plots (b) Bode plots of |Z| vs.frequency and (c) Bode plots of phase angle vs.frequency in 3.5wt% NaCl aqueous solution.

Table 4Electrochemical data obtained via equivalent circuit fittin of the EIS curves.

Fig.12.Equivalent circuit for corrosion modeling data: (a) Rs (CPE1Rct (RLL)) for S1 and S2, (b) Rs (CPE2[(CPE1Rp1) Rp2]) for SF1 and SF2.

The fitte results obtained from the equivalent circuit(ECs)are described by the black lines in Fig.11a, and the fittin parameters are listed in Table 4.It was reported in literature that the high frequency capacitive loop corresponds to the charge transfer and fil effect and the low frequency inductive loop indicated pitting of the alloy [15,49-51].The ECs developed for specimens were shown in Fig.12.In the ECs,Rs and Rct were the solution resistance and the charge transfer resistance, respectively.RLpresented the pitting corrosion with an inductance L.R1showed the resistance of the coating, and CPE was capacitive reactance of the protective layer(either Al2O3fil or the passive layer of hydroxyapatite) and n was the exponent in CPE [23,50].The equivalent circuit mode for both AZ31 Mg substrates was Rs (CPE1Rct (RLL))(Fig.12(a)) [52,53].The corroded products layer (such as Mg(OH)2) was characterized by CPE1 and Rct.The occurrence of pitting corrosion was described by RLand L.With the decrease of surface roughness, the Rct, RLand L increased from 118.20Ω·cm2, 41.40Ω·cm2, 98.60 H·cm?2to 152.40Ω·cm2, 142.71Ω·cm2, 564.16 H·cm?2respectively, indicating that the smooth surface contributed to the improvement of corrosion resistance.After ALD deposition, the equivalent circuit mode was changed from Rs (CPE1Rct (RLL)) to Rs(CPE2[(CPE1Rp1) Rp2]) (Fig.12b).From the fittin results shown in Table 4, the Rct of SF1 increased from 118.20 to 532.10Ω·cm2and the Rct of SF2 increased from 152.40 to 583.50Ω·cm2, confirmin that the compact Al2O3fil could effectively protect the AZ31 Mg alloy from corrosion [52].It also confirme that the smooth surface could also contribute to better corrosion resistance for the Al2O3film

Thus, the results of electrochemical measurements indicated that Al2O3fil could improve the corrosion resistance of Mg alloy substrate greatly and the decrease of surface roughness could further enhance the corrosion effectively.

4.Conclusion

A nano amorphous film as the composition of Al2O3, had been tried to deposit on the AZ31 Mg alloy substrate by atomic layer deposition (ALD).The fil on Mg alloy were characterized firstl (morphology, composition, phase, thickness, etc.), then corrosion behavior was evaluated by neutral salt spray test and electrochemical measurement.The conclusions had been deduced as follows:

(1) The compact Al2O3fil are obtained by ALD with a thickness of 18.44nm (SF1) and 18.46nm (SF2),respectively, and the fil exhibits uniform coverage and makes the element distribution of surface more homogeneous.

(2) It is verifie by the salt spray test that Al2O3fil induced a uniform corroded mode to Mg alloy.

(3) The electrochemical experiments indicate the Al2O3fil promote the corrosion resistance significantl , as theEcorrincreases from ?1.621(S1), ?1.401(S2) V to ?1.364(SF1), ?1.358(SF2) V respectively, and theicorrdecreases from 61.700(S1), 2.449(S2) μA/cm2to 2.054(SF1), 1.629(SF2) μA/cm2respectively.

(4) Surface roughness plays a critical role in the corrosion behavior of Mg alloy.By the reducing of surface roughness, theEcorrincreases by 0.220V and theicorrdecreases by 59.251 μA/cm2for S1 to S2(before coating),and theEcorrincreases by 0.006V and theicorrdecreases by 0.425 μA/cm2for SF1 to SF2 (after coating).Furthermore, the rougher bare substrate has a higher promotion in corrosion resistance after the nano amorphous fil deposition.

Acknowledgments

This work was supported by National Key Research and Development Program (Nos.2016YFB0701201,2016YFB0701203), National Natural Science Foundation of China (Nos.51671101), Domain Foundation of Equipment Advance Research of 13th Five-year Plan (No.61409220118), Natural Science Foundation of JiangXi Province (Nos.20171BCD40003) Key Research and Development Program of JiangXi Province (No GJJ150010)and Nanchang University Graduate Innovation Special Fund(No.CX2018038).