Effect of TIG Welding and Rare Earth Elements Alloying on Corrosion Resistance of Magnesium Alloys

Magnesium alloys used in this study were obtained by gravity sand casting method as metal sheets with dimensions of 200 mm × 350 mm × 20 mm. The castings were milled and 10 mm-thick plates were used for welding.

Joints of magnesium alloys were made using a nonconsumable WT20 tungsten electrode of the diameter 3.2 mm (according to EN ISO 6848:2008) in the inert gas envelope (technical grade argon of the purity 99.995% at the flow rate of 12 dm³ min^–1), with the application of a Lincoln V 205AC/DC inverter welder and the alternating current. The upslope time was set to 2 s and the flash-off time was set to 4 s. The free flow-out time of argon was established to 3 s before beginning of the welding and to 4 s at the end of the process. Plates prepared for welding were chamfered for "Y" at the angle of 30°, leaving welding edge of 2 mm. Table I shows the chemical composition of the alloys used in the investigations.

Before welding, the plates were heated to the temperature of 100 °C. Welds had the form of string beads, the mode of the arrangement of the beads with the application of sealing run is presented in Fig. 1. The following parameters of the welding were applied: welding current 120 A, arch voltage 14 V, linear energy of the arch 3 kJ cm^–1. Filler wires of the diameter 2.4 mm, of the composition similar to the investigated materials (Table I), produced by Magnesium Elektron were used for welding. According to the recommendations of the producer (Magnesium Elektron), after welding the joints were subjected to heat treatment. Table II presents the parameters of the heat treatment.

Zoom In Zoom Out Reset image size

Measurements were conducted in the 0.5% NaCl (pure p.a., POCH S.A. Poland) solutions, additionally saturated with Mg(OH)₂, in a glass cell in a three-electrode setup at the free access of air. pH of the as-prepared solution was 9.8 ± 0.1. A platinum foil was used as a counter electrode and an Ag∣AgCl∣ saturated KCl solution electrode was used as a reference electrode. All potentials are given with respect to that electrode. Electrochemical measurements were performed using a Gamry 300 series potentiostat/galvanostat/ZRA. Samples of the cuboid shape were cut for each alloy (Fig. 2a). The HAZ samples were prepared separately in a manner depicted in Fig. 2b.

Zoom In Zoom Out Reset image size

Samples were embedded in epoxy resin in such a way that only one face of the sample cuboid was exposed to the solution. The electrical connection was made at the opposite side of the sample. Before measurements samples were wet (99.8% ethanol, POCH S.A) polished on SiC emery papers (Struers) of the gradation: 500, 1000, 2000, and finally 4000, next rinsed with ethanol and distilled water.

EIS spectra were registered in the frequency range from 3000 to 0.015 Hz, with the amplitude of the sinusoidal voltage signal 5 mV rms and with the sampling density of eight frequency points per decade in logarithmic steps. In selected cases EIS spectra were measured down to the frequency of 0.001 Hz. For the analysis of the impedance data the program Minuit⁴⁰ based on a complex least-square non-linear regression procedure was used. A more detailed description of the EIS data treatment procedure may be found elsewhere.⁴¹

LSV scans were performed with the potential sweep rate of 1 mV s^–1, starting from the potential of –0.3 V vs the open circuit potential (OCP). The scan was finished when the anodic current density attained approximately 2 mA cm^–2. SACV measurements were performed at the potential sweep rate of 1 mV s^–1 and the amplitude of ±10 mV, with the first cycle initiated in the cathodic direction. Three SACV cycles were registered during each measurement.

The measurements were performed according to two different time schedules. In the first schedule the investigated electrode was immersed in the solution and OCP was registered for two hours, then the potentiostat was set on to the lastly measured OCP and acquisition of the impedance spectrum was initiated. Two spectra at the same potential were consecutively registered to reveal the stationarity of the system, and then the LSV scan was performed.

In the second mode, the investigated electrode was immersed in the solution and OCP was registered for 8 min, then the potentiostat was set on to the lastly measured OCP and SACV measurement was started, then the potentiostat was set again to the OCP measurement. Such a procedure was repeated twelve times for 2 h. After the last SACV measurement, LSV was initiated. At least two repetitions of electrochemical measurements were carried out for each alloy (for BM, HAZ as well as for FZ material).

From the SACV voltammetry data, R_P was calculated as a slope of the E = f(I) curve. Similarly, R_P at the potential of current sign reversal (corrosion potential, E_corr) was calculated from the LSV data taking into account the points between –10 and +10 mV vs E_corr. In both cases, R_S (measured by EIS) was subtracted from the slope value to account for the Ohmic drop on the solution resistance. From the LSV data also the corrosion current density was calculated by extrapolation of the linear segment on the cathodic side of the Tafel plot to the zero-current potential. In that case the potential data were Ohmic-drop corrected before calculation of i_corr.

Visual inspection of the joints performed according to the EN ISO 17637:2017-02 showed that both faces of the weld did not contain any defects (Fig. 3a). No welding discrepancies according to EN ISO 10042:2018-09 were observed. No discontinuities on the surface of the joint were noticed in the liquid-penetrant inspection investigations performed according to the EN ISO 23277:2015-05. X-ray pattern analysis performed according to the EN ISO 17636-1:2013-06 standard also has not revealed any discrepancies (Fig. 3b). The observed macrostructure of the junction is characteristic of TIG welding. However, HAZ is not well discernible due to negligible structural changes. The shape factor of the welding (width-to-depth ratio) was approximately 0.8, meeting the requirements of EN ISO 10042:2008-09 for B-level quality welds (Fig. 3c).

Zoom In Zoom Out Reset image size

Significant refinement of the alloy structure of the weldments (Figs. 4b, 4d, 4f, 4h) in comparison to the base material (Figs. 4a, 4c, 4e, 4g) is clearly visible. No grain refinement but increased concentration of precipitations was observed in HAZ (Figs. 4b', 4d', 4f', 4h'). The chemical composition of the precipitations is the same in the whole volume of the joint. After the heat treatment (see Table II) the welded joint of the AZ91 alloy is composed of the grains of the solid solution (α phase) as well as plate-like precipitations of the Mg₁₇Al₁₂ phase (Figs. 4a, 4b, b'), the welded joint of the QE22 alloy is composed of the solid solution (α phase) as well as precipitations of (Mg,Ag)₁₂Nd phase (Figs. 4c, 4d, d'), the welded joint of the ZRE1 alloy is composed of the solid solution (α phase) as well as precipitations of (Mg,Zn)₁₂RE phase (Fig. 4e, 4f, f') whereas the WE43 alloy welded joint shows the grains of the solid solution (α phase) as well as precipitations of the β' phase (Mg₁₂Nd) and β₁ phase (Mg₃Nd) (Figs. 4g, 4h, h').⁴²

Zoom In Zoom Out Reset image size

The measured data are summarized in Table III. For all data calculated from LSV and OCP measurements the mean value (from at least five measurements) and its standard deviation (SD) are given. Only one or two SACV measurements were performed for each sample, so the data are given for one of them. Similarly in the case of EIS, the results for only one measurement, the most representative, were given, because in most cases the noises in the low frequency region were so high that no reliable fit was possible. Comparing the data in Table III with the data presented in Figs. 5–10 one should remember, that figures are traced for particular experiments, whereas most of the data in Table III are mean values.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The repeatability of OCP was good for all alloys except WE43 (see below). Except few cases SD of OCP measured after 2 h was below 15 mV (see Table III). However, in the case of WE43 alloy (for both FZ, HAZ and BM) the course of OCP in time group in two branches, removed from each other by approximately 55 mV in the case of FZ and 30 mV in the case of BM and HAZ. Within each branch, the scatter of the results did not exceed 5 mV (see Fig. 5a). All results presented in Fig. 5a were obtained using the same electrode, polished shortly before the measurement, and at virtually the same conditions. The most probable explanation for that division is the macroscopic inhomogeneity of the alloy. In the case of the AZ91 alloy, the course of OCP versus time was qualitatively the same as in the case of WE43 alloy, but good reproducibility was achieved (Fig. 5b). A quite different course of OCP versus time was observed in the case of ZRE1 alloy (see Fig. 5c). OCP first increased then after sudden drop stabilized. The time at which that sudden drop occurred was, however, different for each experiment. For QE22 alloy, the course of OCP in time was qualitatively the same as in the case of ZRE1 alloy (Fig. 5d).

In the case of all investigated alloys fluctuations of OCP were observed, however incomparably lower than those registered in the case of the NaCl solutions without Mg(OH)₂ (compare Figs. 5 and 6). The fluctuations of OCP in NaCl solutions (Fig. 6) are attributed to unstable pH near the electrode surface, the occurrence of local corrosion events, formation/liberation of small hydrogen bubbles, and formation of corrosion products.^43,44 Note that the saturation of the solution with Mg(OH)₂ stabilizes not only the Mg²⁺ concentration but also pH, both strongly influencing the course of magnesium alloy corrosion. Only in the case of WE43 alloy FZ, HAZ and BM materials showed a significant difference in OCP values (161 mV between BM and FZ and 45 mV between BM and HAZ), which suggests better corrosion resistance of the FZ material. In the case of other alloys, the difference was either negligible or even opposite.

Our preliminary experiments showed that prolonged cathodic polarization changes significantly the properties of the alloy and generates hydrogen bubbles on the surface of the electrode resulting in partially blocked metal surface. Therefore, LSV curves were traced with relatively high potential sweep rate (1 mV s^–1) to shorten the time of the measurement. E_corr covered a wide span of potentials, being the lowest for WE43 alloy and the highest for QE22 alloy (see Table III and Fig. 7).

In all cases, E_corr was significantly more anodic than OCP (see Table III). FZ material always showed bigger difference between OCP and E_corr than either BM or HAZ. In the case of QE22 and ZRE1 alloys that difference was much bigger (between 153 and 248 mV) than in the case of either WE43 or AZ91 alloys (between 60 and 89 mV). Only in the case of WE43 alloy E_corr for FZ material was significantly more anodic (by 161 mV) in comparison to BM. In other cases, the difference was much lower or even negligible (AZ91). Note, that the change of E_corr to more positive ("more noble") value is commonly assumed as a proof of better corrosion resistance.

The sudden increase of current density was observed on the anodic branch of polarization curves. This can be attributed to the breakdown of the passivating film on the surface.^43,45–48 No breakdown was observed in the case of BM and FZ samples of WE43 alloy and HAZ of ZRE1 alloy (Fig. 8a). In the case of other alloys breakdown potential (E_b) was more anodic for FZ material than for BM (see Table III and Fig. 8), which suggests that FZ material shows better corrosion resistance than BM. However, R_p measured at E_corr was always higher for BM in comparison to FZ material, which means opposite tendency, and in two cases (WE43 and AZ91) R_p measured at E_corr for HAZ was higher than either for BM or FZ. However, it must be said that in almost all cases the differences are statistically insignificant due to the high scatter of the results (see Table III). Similarly, i_corr measured at E_corr by extrapolation of the linear segment on the cathodic part of the polarization curve traced in the semi-logarithmic coordinates to E_corr was always lower in the case of BM in comparison to FZ material, however, the differences again were statistically insignificant. Note the self-consistence of the R_p and i_corr data, which were calculated independently from each other.

No matter what was the difference in E_corr between BM, HAZ and FZ materials of the same alloy, the current density in the anodic part of the LSV curve was always significantly higher for BM in comparison to either HAZ or FZ (see Fig. 8). It suggests the higher susceptibility to corrosion of BM in the conditions of anodic polarization. The most probable explanation of the decrease in anodic current in the case of FZ and HAZ materials in comparison to BM is the formation of thicker Mg(OH)₂ layer (remember also the increase in E_b). At the same time the current density in the cathodic part of the LSV curve was lowest for BM (except AZ91, for which the currents were close to each other). Evidently, welding increases the catalytic activity of the alloys in hydrogen evolution reaction (HER). The last effect is negligible for AZ91 alloy (Fig. 8c), which suggests that the reason for the enhanced catalytic activity in HER may be ascribed to the higher density of grain boundaries in the case of FZ material. Note, that RE containing precipitates are located mainly at the grain boundaries.

Note also the very good reproducibility of the cathodic Tafel slope (C_Tafel). In almost all cases (except AZ91 alloy, BM and HAZ) SD did not exceed 10% of the calculated value of C_Tafel (Table III).

In the case of all investigated alloys, the difference between OCP and E_corr was significant, and in the case of two of them (ZRE1 and QE22) very large. Although measurements of either R_p or i_corr at E_corr are commonly used to evaluate the corrosion resistance of magnesium and its alloys, it must be said that in practice magnesium alloys do not corrode at E_corr but at OCP, while OCP changes significantly in time. One may anticipate that when OCP changes, R_p changes too. Therefore, the course of R_p in time was studied using SACV. The results are presented in Fig. 9 and in Table III. The corrosion current density from the measured R_p was not extracted because, according to our opinion, no reliable data on anodic Tafel slope for all investigated alloys exist and even the cathodic Tafel slopes, measured at the potential significantly cathodic to OCP might not be applicable at OCP. Fortunately, for a particular alloy, a comparison of R_p values should be sufficient.

Figure 9 presents the course of R_p versus time for all investigated alloys in one of two series of such measurements. It is interesting to note that at the beginning of the experiment R_p of all alloys assumed values similar to each other, between 0.5 and 3.0 kΩ cm² (value extrapolated to t = 0). It may be postulated, that this value reflexes the corrosion resistance of the Mg α-phase, which constitutes most of the surface of all the alloys under study. Later on, R_p increases, most strongly in the case of AZ91 alloy. It may be ascribed to the forming of the composite oxide/hydroxide layer on the surface of that alloy composed mainly of Al₂O₃·xH₂O.^41,49 After two hours of corrosion test R_p of AZ91 alloy attain the value similar to the value registered in the LSV experiments (see Table III). Note that in the case of AZ91 alloy OCP and E_corr differ by less than 90 mV both in the case of BM, HAZ and FZ materials. WE43 alloy showed similar behavior like AZ91 alloy, but the final R_p after two hours was much lower than in the case of the latter (but still close to the value measured by LSV). Completely different behavior was observed in the case of both ZRE1 and QE22 alloys. After some time (different in particular experiments) sudden drop of R_p was observed. That drop correlated in time with the drop of OCP. After two hours, R_p attained the value of an order of magnitude lower than that measured by LSV at E_corr. The direction of OCP change during the sudden drop of R_p (towards more cathodic potential) together with the decrease of R_p suggests the increase in the anodic reaction rate (dissolution of the alloy) and constancy of the cathodic reaction rate (HER). Note the different behavior of HAZ material in the case of ZRE1 and QE22 alloys as compared to BM and FZ materials.

The primary condition to apply EIS is the stationarity of the measured system. In view of the fast changes of R_p and OCP observed at least in the initial period of time in the case of all investigated alloys (see Figs. 5 and 9) the time of the spectrum acquisition should be as short as possible. On the other hand, R_p is the low-frequency limit of the electrode impedance. Therefore, to obtain this low-frequency limit, acquisition should be continued to sufficiently low frequency. That seems to be an unresolved problem because acquisition to the low frequency requires long time.

The measured EIS spectra had a rather complex shape with at least two capacitive time constants in the high-to-medium frequency range and an inductive loop in the low-frequency range (Fig. 10).

Such EIS response of Mg alloys is typically observed in corrosion tests under similar exposure conditions.^50–54 The shape of the spectra in the low-frequency part suffered from significant scatter, which can be explained by a rather complex corrosion mechanism of Mg alloys in Cl⁻-containing media.^52,53 Mg alloys corrode non-uniformly, thus actively corroding and passivated regions exist simultaneously on the surface.⁵³ A porous layer of corrosion products covers weakly passivated regions. This layer is non-uniform and allows penetration of aggressive ions (in our case Cl^–), causing further corrosion attack.^52,55 It must be underlined here that there is no consensus on the exact corrosion mechanism of Mg alloys, nevertheless it is stated fairly precisely that inductive response in the low-frequency domain is due to the relaxation processes of adsorbed species.^52,55

Reliable qualitative and quantitative interpretation of the obtained EIS spectra is complicated and several different equivalent electrical circuits (EECs) are reported in the literature.^51–53 Less noisy spectra we were fitted using EEC presented in Fig. 10.

The equation describing that EEC has so many adjustable parameters, that the statistical validity of the parameters obtained from such a fit is questionable. However knowing all parameters of EEC one may calculate polarization resistance of the electrode process. For EEC presented in Fig. 10 R_P may be calculated from the following equation:

$$\begin{eqnarray}&&\displaystyle \frac{1}{{R}_{{\rm{P}}}}=\displaystyle \frac{1}{{R}_{{\rm{CT}}2}}+\displaystyle \frac{1}{{R}_{{\rm{CT}}1}+\displaystyle \frac{1}{\displaystyle \frac{1}{{R}_{{\rm{L}}}}+\displaystyle \frac{1}{{R}_{{\rm{C}}}}}}\end{eqnarray} \tag{ 1 }$$

Variables appearing in Eq. 1 are explained in the caption to Fig. 10. Selected, representative values of R_P are showed in Table III. In the case of all alloys under study, the results of EIS measurements agree (in the limits of measurement accuracy) with the results of SACV measurements, supporting eligibility of the interpretation approach used.