Performance of AZ31 Alloy as Anodes for Primary Magnesium-Air Batteries under High Current Discharge

Abstract

The discharge behavior of Mg-air battery consisting of AZ31 Mg alloy anode and a graphite-based air cathode was investigated under constant discharge current densities of up to 40 mA cm⁻². Discharge with higher current density increased the power density, which reached its maximum value of 25 mW cm⁻² at 30 mA cm⁻². Further discharge caused the cell voltage to drop and fluctuate. Electrochemical impedance spectra of the Mg anode revealed specific trajectories according to the discharge current. A new electrical equivalent circuit modeling the dissolution at the Mg anode was proposed. After checking its validity, the dependence of the circuit elements on the discharge current was analyzed. A discharge product in the form of hydrotalcite was preferentially formed under high current discharge, which was gradually transformed to Mg(OH)₂ and Al₂O₃. Higher discharge current promoted uniform dissolution of the Mg anode and improved the anodic efficiency. The removal of Al–Mn intermetallic particles was considered as the cause, which was verified by SEM/EDS analysis. Considering the anode kinetics and the performance of the air cathode, the optimal operation condition of 30 mA cm⁻² discharge was proposed for the present Mg-air battery configuration.

Nyquist diagram of Mg-air battery with 10 mA cm⁻² discharge, showing two inductive loops in the lower frequencies.

1. Introduction

Magnesium (Mg) has a negative standard electrode potential of −2.37 V and a high Faradaic capacity of 2.205 Ah·g⁻¹, which are attractive features for battery anodes. Magnesium primary batteries,¹⁾ sea water-activated Mg batteries^2,3) have been developed and widely used. The research and development of Mg-ion secondary batteries is currently underway.^4,5)

Mg-air batteries (MABs) equipped with air cathodes, have attracted attention as small-to-middle sized power supplies for emergency or outdoor purposes.^6,7) While Mg anodes are stable in dry atmosphere by forming an oxide film, they are quickly activated in contact with a neutral aqueous electrolyte such as NaCl. As atmospheric oxygen acts as a reactant, MABs can generate electricity until the Mg anode is exhausted. This ensures long storage life, easy operation, and safe disposal of used MABs.

There are reviews⁸⁾ and articles discussing the discharge performance of MAB anodes^9–13) air cathodes,^14,15) and the electrolyte.¹⁶⁾ Studies of Mg anodes have been well conducted, since they are consumable and critical for ensuring cell voltage and anodic efficiency. Commercial Mg–Al alloys such as AZ31 (Mg–3%Al–1%Zn–0.3%Mn, % in mass) and AZ91 (Mg–9%Al–1%Zn–0.3%Mn) have been widely utilized for their low cost and good corrosion resistance.^9,10) Special Mg alloys were developed for obtaining high discharge capacities.^11–13) Most of MAB studies are conducted under moderate current around 10–20 mA cm^{−2 9–11)} or using half-cell configurations¹²⁾ to provide stable discharge conditions. Aside from the material aspects, the discharge performance is influenced by the operation conditions as discharge current densities,^11,13) ambient temperature,¹³⁾ and the full cell configurations that limit the electrolyte volume.¹¹⁾ Insoluble discharge products are gradually accumulated on the anode surface at the bottom of the cell. It would be useful if one can monitor how such conditions influence the discharge behavior of MAB. Electrochemical impedance spectroscopy (EIS) is widely used for analyzing electrochemical phenomena as corrosion, surface treatment and batteries.¹⁷⁾ There seems to be few EIS studies to monitor the discharge performance of MAB, or Mg alloys in anodically polarized condition.^18,19)

The present work investigated the effect of discharge current on a commercial MAB. The electrode potential and electrochemical impedance of AZ31 Mg anode were simultaneously monitored under the constant current discharge up to 40 mA cm⁻². The impact of discharge current on the discharge products, surface roughness and the anodic efficiency of Mg electrode was also discussed.

2. Experimental Procedure

Rolled AZ31 Mg alloy plates (supplied by Osaka-Fuji Corp.) were cut to 25 mm × 50 mm × 1 mm and used as the Mg anode. The chemical composition is listed in Table 1.

Table 1 Chemical composition of Mg anode in mass percent.

The surface was polished with SiC abrasive paper up to 1000 grit and cleaned with ethanol and pure water. A commercial air cathode used for MAB (supplied by ARV Corp., MG201) was employed and cut to the same size as the Mg anode. Main composition was graphite, metal oxide catalyst, and binder.²⁰⁾ A test cell enclosure for MAB (supplied by EC frontier, model EC-1) was employed. The anode and cathode were facing each other across 10 mm distance, which was similar to MG201. The surface area was regulated as 3.14 cm² by using a Viton ring. Saturated Ag/AgCl electrode was used as the reference electrode and placed between the above-mentioned electrodes. 10 mass% NaCl solution was used as the electrolyte by pouring into the cell.

The constant current discharge test was performed under current densities (i_d) of 5, 10, 20, 30, and 40 mA cm⁻². The test period (t) was chosen such that the total discharge (q) was 324 C cm⁻². The potential of the anode (E_a), cathode (E_c), and cell voltage (V_cell) were monitored. EIS of the anode was performed in galvanostatic mode using Solartron Analytical ModuLab XM ECS. The amplitude of current input was set to 1% of the discharging current, and the frequency response from 10 kHz to 1 Hz was acquired. The spectrum was analyzed with a curve fitting software (supplied by Scribner Associates, ZView ver. 3.5). The discharge test was carried out at room temperature and repeated at least three times to verify the reproducibility.

After the test, the discharge product was collected from the anode surface and analyzed by XRD using Rigaku Ultima III. The anode was further cleaned in 10 mass% boiling CrO₃ solution to measure the weight loss. Anodic efficiency (η) was calculated by the following equation:   

\begin{equation} \eta = i_{d}\cdot A\cdot t\cdot M/2F (W_{i} - W_{f})\times 100 \end{equation} (1)

where A the surface area (3.14 cm²), M the atomic mass of Mg (24.3 g mol⁻¹), F Faraday constant (96485 C mol⁻¹), W_i and W_f the weights of the anode before and after the discharge test (mg), respectively.

The surface roughness of the anode after cleaning was measured by Mitsutoyo SJ-500. Linear scanning was conducted three times and the average value (R_a) was calculated. The surface was further characterized with SEM/EDS using JEOL JSM-IT500.

3. Results and Discussions

3.1 Electrode potentials and cell voltage

Figure 1 shows the time evolution of E_a, E_c, and V_cell for i_d at 10, 30, and 40 mA cm⁻². At 10 mA cm⁻², E_a and E_c reached their steady state values quickly and provided a constant V_cell of around 1.1 V. V_cell dropped at higher i_d due to the polarizations of the anode and cathode. In addition, a certain period was needed to attain the steady state values. At 40 mA cm⁻², V_cell fluctuated intermittently. While the large fluctuations were mostly attributed to E_c, E_a contributed to some spike-like changes that appeared in V_cell. Figure 2 shows the E_a, E_c, and power density (p) at the end of discharge. The value p reached its maximum of 25 mW cm⁻² at 30 mA cm⁻², which corresponds to V_cell of 0.8 V.

Fig. 1

Discharge curves of anode/cathode potential and cell voltage at discharge current of (a) 10, (b) 30, and (c) 40 mA cm⁻².

Fig. 2

Discharge characteristics of the present MAB.

3.2 Electrochemical impedance spectroscopy of Mg anode

Figure 3 shows typical Nyquist and Bode plots of the Mg anode at the end of discharge. The trajectory revealed both capacitive and inductive behaviors, depending on the frequency region. There are several EIS studies of pure Mg or Mg alloys corroding in NaCl solution.^19,21–24) King et al.¹⁹⁾ proposed an electrical equivalent circuit of pure Mg corroding in NaCl solution, shown in Fig. 4(a), that simulated the spectrum using two capacitive loops and one inductive loop. The circuit consists of series of resistances (R_s, R₁, R₂, R₃), capacitances (C₁, C₂), and an inductance (L₁), where R_s is the solution resistance. Curioni et al.²¹⁾ adopted the same circuit and interpreted the physical meaning of each element. The validity of the circuit was experimentally confirmed with AZ31 alloy corroding in NaCl solution.^22,23) The same circuit was applied to the present study, but no convincing results were obtained, especially for simulating inductive behaviors observed in the low-frequency region. Then, referring to King’s model, a circuit was newly devised by adding one R-L element (R₄ and L₂) in parallel, as shown in Fig. 4(b). Curve fitting was carried out using a simulation software and the result was superimposed in Fig. 3. To judge the validity of the proposed circuit, the residuals of the impedance ΔRe(ω) and ΔIm(ω)²⁴⁾ were calculated as;   

\begin{equation} \varDelta \mathit{Re}(\omega_{\iota}) = (Z_{\textit{Re}}(\omega_{\iota})-\skew3\check{Z}_{\textit{Re}}(\omega_{\iota}))/|Z(\omega_{\iota})| \end{equation} (2)

and   

\begin{equation} \varDelta \mathit{Im}(\omega_{\iota}) = (Z_{\textit{Im}}(\omega_{\iota})-\skew3\check{Z}_{\textit{Im}}(\omega_{\iota}))/|Z(\omega_{\iota})| \end{equation} (3)

Here, Z_Re(ω_ι) and Z_Im(ω_ι) are the real and imaginary parts of the measured impedance at the angular frequency ω_ι (= 2πf_i), $\skew3\check{Z}_{Re}(\omega _{\iota })$ and $\skew3\check{Z}_{Im}(\omega _{\iota })$ are the real and imaginary parts of the fitted impedance, and f_i is the ith frequency (Hz). The cumulative error X_Im² between the actual and modeled imaginary parts²⁴⁾ is given as;   

\begin{equation} X_{\textit{Im}}{}^{2} = \frac{1}{N}\sum\nolimits_{i = 1}^{N}\Delta \mathit{Im}(\omega_{i})^{2} \end{equation} (4)

where N is the number of frequency points of the impedance spectrum.

Fig. 3

EIS of MAB anode for 10 mA cm⁻² at the end of discharge: (a) Nyquist plot, (b) Bode plot. The result of curve fitting was also indicated.

Fig. 4

Electrical equivalent circuit of Mg anode: (a) at open circuit, (b) during discharge.

Figure 5 shows the frequency dependence of the residuals obtained from the spectra of Fig. 3. The values were fit within ±3% throughout the frequency range, and exhibited good agreement with the proposed model.

Fig. 5

Dependence of the residuals of the real and imaginary parts of the Nyquist plot shown in Fig. 3.

The dependence of circuit elements on i_d are illustrated in Fig. 6. X_Im² was around 10⁻⁴ regardless of i_d, suggesting that the prediction based on the set of modeled elements resulted in a similar degree of fitting. Except C₁ and R₂, the parameters vary in proportion or inverse proportion with i_d. The value of C₁ is around 30 µF cm⁻², which corresponds to electrochemical double layer. In contrast, C₂ increases with i_d, which could be attributed to the capacitive behavior of oxide/hydroxide film. The resistive elements of the equivalent circuit have a close relationship with the actual corrosion rate.¹⁷⁾ Polarization resistance (R_p) can be evaluated from the zero-frequency impedance (f → 0) at −Z_Im = 0. On the other hand, charge transfer resistance (R_t) is obtained from the impedance maximum. In both cases, the solution resistance R_s is subtracted. It is reported that the reciprocals of both parameters (R_p⁻¹ and R_t⁻¹), naming conductance, are in good correlation with the corrosion rate.^19,21,23) In the proposed circuit of Fig. 4(b), the values are derived as   

\begin{equation} R_{p}{}^{-1} = 1/(R_{1} + R_{2}) + 1/(R_{3} + R_{4}) \end{equation} (5)

and   

\begin{equation} R_{t}{}^{-1} = 1/(R_{1} + R_{2}) \end{equation} (6)

Fig. 6

Dependence of circuit elements on discharge current; (a) resistance, (b) capacitance, (c) inductance.

Figure 7 exhibits the relationships of conductance values with i_d. They show good linearities up to 20 mA cm⁻², but start to deviate with further increase in i_d. The origin of the deviation is related to the increase of R₂ with i_d higher than 30 mA cm⁻², which will be discussed in the later section.

Fig. 7

Relation between discharge current and the reciprocals of R_p and R_t.

Inductive behaviors are often observed in Mg corrosion. The relaxation of adsorbed species such as Mg(OH)_ads⁺ or Mg(OH)₂,^25,26) the occurrence of pitting,²⁷⁾ or the increase/decrease in the area of anode sites²¹⁾ are considered to be the origins of such behaviors. King et al.¹⁹⁾ reported an enhancement in the inductive behavior when performing EIS on Mg above the free corrosion potential, which supports our results shown in Fig. 6(c). In addition, our analysis clearly indicates the presence of two inductive elements. Since such behavior has not been reported in corrosion studies, it would be a phenomenon specific to the anodic dissolution of Mg under high current densities.

3.3 Analysis of discharged product

During the discharge test, the cell enclosure was gradually filled with discharged product. Figure 8 shows the side view of the cell and the anode surface at the end of test. At 5 mA cm⁻², a white product appeared above a gray product; the former appears as a crust with a few mm thickness, the latter deposited as sediment at the bottom of the cell. In case of 40 mA cm⁻² discharge, only a white and flocculent precipitation was visible. The anode surface at 5 mA cm⁻² discharge was covered by a gray product. The product layer was thin which was easily removed by running water. In case of 40 mA cm⁻² discharge, a solid gray crust was observed. After removing the crust, a thick and dense gray product laid behind. XRD was used to characterize the product found at 40 mA cm⁻², and the results were shown in Fig. 9. From peak analysis, Mg(OH)₂, Al₂O₃, and Mg₆Al₂(OH)₁₆CO₃ 4H₂O (hydrotalcite) were identified. It is interesting to note that hydrotalcite was only found in the gray product. In the white product, instead, peaks relating to Al₂O₃ and Mg(OH)₂ were detected. Furthermore, it was found that the color of the gray product changed to white if the cell was left in ambient atmosphere for 2–3 days after the discharge. Therefore, it was supposed that hydrotalcite was not chemically stable and gradually transformed to white Mg(OH)₂ and Al₂O₃ in the electrolyte.

Fig. 8

Side view of the cell enclosure and the Mg anode surface after discharge test. C: side at air cathode, A: side at Mg anode.

Fig. 9

XRD chart of the white and gray colored discharge products found in the Mg anode surface after the discharge current of 40 mA cm⁻².

3.4 Surface characterization of Mg anode after discharge

Table 2 shows the anodic efficiency η and the surface roughness R_a of the Mg anode after cleaning in CrO₃ solution. The value of η below 100% suggests the self-discharge of Mg anode.¹³⁾ The present result of η between 50 and 70% is within the range observed in previous studies.¹³⁾ The value of η tends to increase with higher i_d. Similar trend was observed for R_a.

Table 2 Surface roughness and anodic efficiency for different discharge current.

Figure 10 shows the SEM images and element mapping analyses of Mg, Al, and Mn of the Mg anodes with different i_d. The dimpled surface caused by the anodic dissolution was less prominent at higher i_d. In the Mg mapping, color loss sections with the size of 10 µm or less were observed. Instead, Al and Mn signals were detected at these sections, suggesting the presence of Al–Mn intermetallic particles (IMPs) as Al₄Mn or Al₆Mn.²⁸⁾ It is seen that the number of large-sized IMPs were selectively reduced on the surface at higher i_d.

Fig. 10

SEM image and element maps obtained by SEM-EDS results of Mg anode surface after cleaning with CrO₃ solution. Discharge test was carried out with: (a) 5 mA cm⁻², (b) 40 mA cm⁻². The scale length is 50 µm.

3.5 Effect of discharge current on the performance of Mg anode

The above studies indicate that i_d influences the discharge performance of MAB in various aspects. When i_d is low, E_a presents a constant value during discharge. As i_d increases, E_a shifts to more positive values and starts fluctuating. Srinivasan et al.¹⁸⁾ reported the phenomenon of rapid change in current (1–5 mA cm⁻²) for anodically potential controlled AZ31 alloy in a simulated body fluid. They interpreted the change in terms of the breakdown and repair of the passive film formed on the surface. In our study, corrosive 10 mass% NaCl was chosen as the electrolyte, in which AZ31 corrodes actively even in open circuit condition. Rather, the fluctuation in E_a is associated with the accumulation of discharged products shown in Fig. 8. This phenomenon was also appeared in the increase in R₂ in Fig. 6(a) which leads the deviations from the linearity of R_t⁻¹ and R_p⁻¹ in Fig. 7 at i_d higher than 30 mA cm⁻². According to Curioni et al.,²¹⁾ R₂ accounts for the process of electron transfer. The increase of R₂ regardless of higher i_d suggests the process of anode kinetics seems altered under the present MAB configuration.

The improvement of anodic efficiency and the smaller surface roughness with higher i_d are also noticeable. These phenomena seem to have a close relation with the disappearance of large-sized Al–Mn IMPs in Fig. 10(b). It is known that they are electrochemically noble with respect to the Mg matrix. As a result, they tend to act as micro-cathodes and accelerate the corrosion of the Mg matrix surrounding them, while remaining on the surface.²⁸⁾ Application of higher i_d shifts E_a in the noble direction so that the cathodic activity of IMPs was reduced. In addition, a concomitant flow of hydrogen gas which occurs as the result of self-discharge reaction of Mg anode, got more active and detached IMPs from the surface. As a result, smoother surface was observed at higher i_d. The removal of IMPs also suppresses the self-discharge reaction of Mg anode and improves the anodic efficiency. On the other hand, higher i_d generates the solid crust and dense discharge products on Mg anode that influence E_a and R₂. The formation of discharge products with different crystal structure remains uncertain. Still, the flocculent nature of precipitations would interfere the removal of discharged products for continuous operation of MAB.¹⁵⁾

As shown in Fig. 2, the maximum power density was obtained at 30 mA cm⁻². All these facts can lead to conclude that the optimum operation condition of the present MAB is discharging with i_d of 30 mA cm⁻² or less.

3.6 Effect of discharge current on the performance of air cathode

As seen in Figs. 1 and 2, the increase in i_d also resulted in a negative impact on the air cathode. A close look at Fig. 1 revealed that the cathode exhibited a negative value at the start that gradually recovered to a more positive value. Discharge under higher i_d made this phenomenon more prominent. In addition, the voltage did not attain a constant value, but rather varied irregularly. It is likely that this behavior presents the limit of its discharge performance. Thus, a discharge current of 30 mA cm⁻² or less is also preferable from a viewpoint of air cathode. Compared to alkaline Zn–MnO₂ batteries, MABs are no good at high current discharge.⁸⁾ This is attributed to the sluggish kinetics of oxygen reduction reaction in neutral NaCl solution. The development of an efficient oxygen catalyst is required to improve the performance of existing MAB.

4. Conclusion

The discharge behavior of MAB with AZ31 alloy anode and air cathode was investigated by constant discharge tests up to current densities of 40 mA cm⁻² in 10 mass% NaCl solution. The results are summarized as follows:

1. (1)    Under 10 mA cm⁻² discharge, the anode and cathode potentials quickly reached the constant value of 1.1 V. An increase in the discharge current induced polarization of both the electrodes and lowered the cell voltage, accompanied by some fluctuation. The power density reached its maximum value of 25 mW cm⁻² at 30 mA cm⁻², with the cell voltage being 0.8 V.
2. (2)    EIS spectra of the Mg anode revealed trajectories with two capacitive and two inductive loops. A new electrical equivalent circuit of the Mg anode dissolving in NaCl electrolyte was proposed. The features of each circuit element were analyzed. The reciprocal of the resistive element in the discharge kinetics showed good linearity with i_d up to 20 mA cm⁻².
3. (3)    White- and gray-colored discharge products were gradually accumulated in the cell. The gray product was densely accumulated at higher discharge current. XRD analysis identified the gray product as Mg(OH)₂ and Mg₆Al₂(OH)₁₆CO₃ 4H₂O, whereas the white product consisted of Mg(OH)₂ and Al₂O₃.
4. (4)    Surface roughness and anode efficiency were improved by higher discharge currents. The SEM/EDS analysis revealed the removal of Al–Mn IMPs on the discharged electrode surface. It is likely that the application of higher discharge current diminished the micro-galvanic corrosion between Al–Mn IMPs and the Mg matrix.
5. (5)    Discharge below 30 mA cm⁻² was found to provide the optimum performance of the Mg anode, which was also preferable for the stable operation of the paired air cathode.

REFERENCES

• 1)  P.J. Spellman: Linden’s Handbook of Batteries, ed. by T.B. Reddy, (McGraw-Hills, New York, 2011) 10.1–10.13.
• 2)  R.D. Lucero and A.P. Karpinski: Linden’s Handbook of Batteries, ed. by T.B. Reddy, (McGraw-Hills, New York, 2011) 34.1–34.7.
• 3)   Q.  Liu,  Z.  Yan,  E.  Wang,  S.  Wang and  G.  Sun: Int. J. Hydrogen Energ. 42 (2017) 23045–23053.
• 4)   P.  Saha,  M.K.  Datta,  O.I.  Velikokhatnyi,  A.  Manivannan,  D.  Alman and  P.R.  Kumta: Prog. Mater. Sci. 66 (2014) 1–86.
• 5)   H.D.  Yoo,  I.  Shterenberg,  Y.  Gofer,  G.  Gershinsky,  N.  Pour and  D.  Aurbach: Energy Environ. Sci. 6 (2013) 2265–2279.
• 6)   A.  Ito,  H.  Abe,  M.  Kubota,  Y.  Taira,  H.  Inazuka,  K.  Saita,  T.  Takahara and  Y.  Ono: FB Technical News 70 (2014) 23–27.
• 7)  K. Yamazaki: Proc. 73rd Ann. Mg World Conference, (Intl. Mg Assoc., 2016) pp. 83–88.
• 8)   T.  Zhang,  Z.  Tao and  J.  Chen: Mater. Horiz. 1 (2014) 196–206.
• 9)   G.  Huang,  Y.  Zhao,  Y.  Wang,  H.  Zhang and  F.  Pan: Mater. Lett. 113 (2013) 46–49.
• 10)   M.  Yuasa,  X.  Huang,  K.  Suzuki,  M.  Mabuchi and  Y.  Chino: Mater. Trans. 55 (2014) 1202–1207.
• 11)   M.  Yuasa,  X.  Huang,  K.  Suzuki,  M.  Mabuchi and  Y.  Chino: J. Power Sources 297 (2015) 449–456.
• 12)   N.  Wang,  R.  Wang,  C.  Peng,  B.  Peng,  Y.  Feng and  C.  Hu: Electrochim. Acta 149 (2014) 193–205.
• 13)   N.  Wang,  R.  Wang,  C.  Peng,  C.  Hu,  Y.  Feng and  B.  Peng: Trans. Nonferrous Met. Soc. China 24 (2014) 2427–2439.
• 14)   C.  Shu,  E.  Wang,  L.  Jiang and  G.  Sun: Int. J. Hydrogen Energ. 38 (2013) 5885–5893.
• 15)   Y.  Milusheva,  R.I.  Boukoureshtlieva,  S.M.  Hristov and  A.R.  Kaisheva: Bulg. Chem. Commun. 43 (2011) 42–47.
• 16)   M.  Mayilvel Dinesh,  K.  Saminathan,  M.  Selvam,  S.R.  Srither,  V.  Rajendran and  K.V.I.S.  Kaler: J. Power Sources 276 (2015) 32–38.
• 17)  M.E. Orazem and B. Tribollet: Electrochemical Impedance Spectroscopy, 2nd ed., (Wiley, Hoboken, 2017).
• 18)   A.  Srinivasan,  K.S.  Shin and  N.  Rajendran: RSC Advances 4 (2014) 27791–27795.
• 19)   A.D.  King,  N.  Birbilis and  J.R.  Scully: Electrochim. Acta 121 (2014) 394–406.
• 20)  S. Natsume and Y. Ando: Japanese Patent, 2012.9.27.
• 21)   M.  Curioni,  F.  Scenini,  T.  Monetta and  F.  Bellucci: Electrochim. Acta 166 (2015) 372–384.
• 22)   M.C.  Delgado,  F.R.  García-Galvan,  I.  Llorente,  P.  Pérez,  P.  Adeva and  S.  Feliu, Jr.: Corros. Sci. 123 (2017) 182–196.
• 23)  M. Ohara, I. Nakatsugawa, Y. Matsumoto, F. Kido, T. Matsumoto and S. Saikawa: Collected Abstracts of the 2018 Autumn Meeting of the Japan Institute of Light Metals, (2018) pp. 113–114.
• 24)   M.  Schönleber,  D.  Klotz and  E.  Ivers-Tiffée: Electrochim. Acta 131 (2014) 20–27.
• 25)   G.  Baril and  N.  Pébère: Corros. Sci. 43 (2001) 471–484.
• 26)   G.  Galicia,  N.  Pébère,  B.  Tribollet and  V.  Vivier: Corros. Sci. 51 (2009) 1789–1794.
• 27)   Y.  Song,  D.  Shan,  R.  Chen and  E.-H.  Han: Corros. Sci. 51 (2009) 1087–1094.
• 28)   L.G.  Bland,  N.  Birbilis and  J.R.  Scully: J. Electrochem. Soc. 163 (2016) C895–C906.