Evidence of the Enrichment of Transition Metal Elements on Corroding Magnesium Surfaces Using Rutherford Backscattering Spectrometry

99.9% pure Mg was used in this investigation as supplied by Alfa Aesar (Table I). Each sample was cold mounted in epoxy resin and mechanically ground to a 1200 grit SiC finish using ethanol (C₂H₆O) as lubricant. Sample surfaces were cleaned with ethanol and dried using compressed air after each grinding step.

Electrochemical testing was performed using a three electrode vertical cell with a platinum mesh counter electrode and saturated calomel reference electrode (SCE). An electrolyte filled burette with an inverted funnel attached was centered over the sample to aid in collection of H₂ as described elsewhere⁴² emphasizing the important aspect that the electrolytes were pre-saturated with hydrogen prior to commencement of testing. This setup provided an exposed area of 0.5 cm² for each test specimen. Potentiostatic polarization was performed at various net anodic and cathodic potentials ranging from −1.4 V_SCE to −1.8 V_SCE for 24 hours in quiescent 0.6 M NaCl solution using 18.2 MΩ deionized water at 25°C; the initial pH of each solution was in the range 8–9 as a result of water reduction when pre-saturating the solution with hydrogen. Following testing, samples were cleaned using 200 g/L CrO₃ according to ASTM G1⁴³ to remove all corrosion products for determination of mass loss and evaluation using optical microscopy. Negligible mass loss was recorded on as-polished samples as seen elsewhere verifying this method.⁴⁴

Under anodic polarization the total charge consumed is described by:

following fundamental mixed potential theory⁴⁵ where Q_Δm is the anodic charge consumed from metal dissolution which can be accessed from gravimetric mass loss after removing corrosion products, is the cathodic charge consumed on the anode due to hydrogen evolution, and Q_{I net anodic} is the charge measured by the potentiostat as the applied anodic current over time. The gravimetric mass loss for each sample with ± 0.1 mg resolution was converted to consumed anodic charge (Q_Δm) via Equation 2:

where z is the equivalent electrons per mole of Mg oxidized, Δm is the change in mass, F is Faraday's constant, and a is the molar mass of Mg. In this study, z is assumed to be 2 following the half-cell reaction Mg → Mg²⁺ + 2e⁻. The hydrogen captured during testing of pre-saturated solutions can be converted to a corresponding cathodic charge using the ideal gas law in combination with Faraday's Law:⁴²

where z = 2 (i.e., 2e⁻ + 2H⁺ → H₂), P is the pressure inside the burette (1 atm), n is the number of moles of H₂ gas given by the ideal gas law, V is the volume of gas in the burette, F is Faraday's constant, R is the ideal gas constant, and T is temperature. The net charge from the measured or applied current was determined from the integrated signal of current vs. time transient at each potential:

In additional electrochemical testing, cathodic potentiodynamic polarization scans were performed after 24 hour anodic and cathodic potentiostatic holds. These samples were not cleaned or removed from the electrochemical cell. Immediately following the potentiostatic hold, samples were allowed to stabilize at their open circuit potential (OCP) for 10 minutes before scanning from +0.05 V above OCP to a potential of −2.3 V_SCE at a rate of 1 mV/s. A Bio-Logic VMP multi-channel potentiostat was used for all electrochemical testing. Each potential was tested a minimum of three times showing limited scatter. Additionally, control samples were polished, cleaned in CrO₃, and tested as is or subjected to the same cathodic polarization test after 10 min hold at OCP. A second control sample was subjected to the anodic potentiostatic hold, cleaned in CrO₃, and tested in a film free condition but with the possible surface enrichment of TM elements.

All solutions were retrieved and analyzed using ICP-OES. A Thermo Scientific iCAP 7200 ICP-OES was used for analysis. Samples were acidified by adding 1 M HCl into each solution after electrochemical testing in order to dissolve any insoluble corrosion products suspended in solution. The acidified solutions were then subjected to a 10 part dilution prior to testing to minimize interferences and the background signal as a result of the NaCl content in solution. Corrosion products remaining on the surface could not be analyzed because of Cr interferences with Fe emission signals due to the CrO₃ cleaning. In this work, the emission intensity of Mg (279.553 nm), Mg (280.270 nm), Al (226.910 nm), Al (308.215 nm), Al (396.152 nm), Fe (238.204 nm), Fe (239.562 nm), Mn (257.610 nm), Mn (259.373 nm), Zn (206.200 nm), Zn (213.856 nm), Zr (339.98 nm) and Zr (343.823 nm) were recorded and used for the calculation of charge consumed and percent of elements in solution using solution analysis. The approximate detection limit for each element is shown in Table II.

RBS was carried out at the University of Western Ontario, using a 1.7 MeV Tandetron, solid state, gas insulated, high frequency accelerator. The incident ion used for RBS was ⁴He (referred to as an α particle hereafter) accelerated to an energy of 3050 keV. This energy was used in order to take advantage of the resonance of the elastic ¹⁶O(α,α)¹⁶O scattering cross-section near 3040 keV which results in enhancement of the oxygen signal up to 29 times greater than that of the Rutherford cross-section depending on the scattering angle.⁴⁶ The energy of scattered particles without energy loss is defined by Equation 5:³⁶

where E is the energy of the ⁴He projectile after collision with the substrate, K is the kinematic factor, and E_o is the incident projectile energy. The kinematic factor is defined by Equation 6:³⁶

where M₁ is the mass of the projectile, M₂ is the target mass, and θ is the scattering angle. The energy resolution, ΔE, between two masses on the same target that differ by a small amount ΔM₂ is therefore:

The values of K and E used for the elements detected in this study are given in Table III. The energy loss of an alpha particle after passing through a layer with thickness x (atoms/cm²) is given by the equation:

where x is the depth into the sample layer and S(E) is the specific energy dependent stopping cross section of an alpha particle for the elements considered in the α particle-substrate collision.

The geometry and design of the RBS experiments was as follows: the incident angle was 0 degrees with an exit angle of 20 degrees resulting in a scattering angle of 160 degrees. The beam spot size was approximately 1.5 mm × 0.5 mm whilst the detector was a solid state device with an energy resolution of 15 keV. The beamline contained a quadrupole magnet for focusing, an analyzer magnet, a raster-scanner, and a steerer. The chamber pressure was in the range of 1.3 × 10⁻⁵ Pa. Data analysis was performed with the program SIMNRA v6.06.⁴⁷ In this analysis the composition depth profile was established by iterative simulation of both elements in each layer and layer thicknesses until the experimental RBS data was matched for a given beam energy. The match was achieved accounting for energy loss and element concentration in a Mg matrix. Three zones were assumed: oxide/hydroxide outer layer, metallic enriched zone with the Mg below the oxide/hydroxide, and bulk Mg. This layered situation was modeled based on the ⁴He stopping cross-sections for a given layer thickness and composition. To achieve a match, the layer thickness, atomic numbers included in each zone, and thickness assumed needed to accurately reproduce the experimental RBS spectrum. Hence the atomic mass resolution produces the only uncertainty. A composition versus depth plot was constructed only after a match of the experimental data was obtained.

The typical potentiostatic polarization data for 99.9% pure Mg over a 24 hour period are shown in Figure 1. Figure 1a indicates the net current density vs. time for potentials that accumulated a net anodic charge whilst Figure 1b shows the net current densities vs. time for net cathodic potentials; the OCP is approximately −1.6 V_SCE. All net anodic potentials underwent a reduction in net current between 100 to 1000 seconds implying either eventual inhibition of anodic dissolution or a large increase in cathodic reaction rate such that net anodic current was significantly reduced. Visual inspection of the Mg anode during testing showed that the minimum value in net current density measured during this period of time occurred when the Mg surface became completely covered with an optically black corrosion product. It is noteworthy that the current polarity switches from net anodic to net cathodic current density at a potential of −1.625 V_SCE after approximately 1000 seconds similar to that reported elsewhere.²⁰ In contrast, the current time behaviors were relatively stable at applied potentials where the current density remained net cathodic at all times (Figure 1b).

Zoom In Zoom Out Reset image size

The net charge associated with hydrogen evolution plus the net charge measured by the potentiostat during anodic polarization are compared to mass loss as a function of potential in Figure 2a in accordance with Equation 1. This data is consistent with Mg²⁺ based dissolution where cathodic hydrogen evolution persisted on the corroding Mg surface at high anodic potentials. The volume of hydrogen captured was a maximum value at −1.4 V_SCE and decreased at more negative applied potentials. Hydrogen evolution reached a minimum value of approximately 1.5 C/cm² at the net cathodic potential −1.65 V_SCE and subsequently increased at more cathodic potentials as shown in Figure 2b. Note that mass loss data was excluded in Figure 2b because negative values were measured and analysis of −1.625 V_SCE is complex due to polarity reversal. It is also interesting to note that the minimum value of hydrogen captured was not measured at the OCP but rather at a potential roughly 50 mV cathodic than the OCP. Moreover, the accumulated charge density from evolved hydrogen at net cathodic potentials was much less than that expected based on the current measured by the potentiostat. It is interesting to note that the charge density ratio from HER to the charge density measured by the potentiostat, Q_H2:Q_{i net anodic}, was near 2.5 at −1.575 V_SCE and decreased with increasing applied anodic overpotentials to a value of 0.85 at −1.4 V_SCE. A measure of the NDE at anodic potentials is observed in Figure 3 where the NDE is defined as and is in agreement with the principles of the NDE. Note that the NDE is traditionally measured as a function of constant applied current.^28,48

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

A simple indication of enhanced catalytic activity toward HER on Mg is given by Figure 4. Figure 4a shows cathodic potentiodynamic polarization scans which were performed immediately upon completion of each respective potentiostatic hold; potentiodynamic polarization after 10 min and 24 hr OCP are also shown. This plot indicates that prior polarization had a significant effect on the net cathodic current density realized. Prior anodic current density produced a considerably larger net cathodic current density over a wide range of potentials. Moreover, what is evident from Figure 4 is that allowing a specimen to undergo free corrosion under open circuit conditions also leads to an enhanced catalytic effect – such that longer exposure times at OCP can produce a similar qualitative effect. In addition, cathodic polarization scans which were performed on samples subjected to the potentiostatic hold, cleaned in CrO₃, and tested in a film free condition yielded similar results. An example of this is given in Figure 4b.

Zoom In Zoom Out Reset image size

The combination of optical micrographs (Figure 4b–4d) and ICP-OES (Table IV) support enhanced cathodic activity on anodically polarized samples since the samples exposed at cathodic potentials exhibit very little damage and 1200 grit grind marks were still present. Indeed, very limited Mg cations were detected from ICP-OES at net cathodic potentials compared to mass loss during anodic polarization. It should be noted that Mg was the only element detected beyond a reasonable doubt by ICP-OES (Table IV) during anodic polarization. The measured intensities for all other elements were at or below the detection limit of the apparatus resulting in high error values. Therefore it is plausible that samples which have undergone anodic dissolution have surfaces enriched with more noble impurity elements, such as Fe. It should be noted that not all of the corrosion product could be recovered for solution analysis due to complications with chromic acid cleaning as further discussed in the Role of Fe enrichment section below.

In order to compare the net cathodic current density at a fixed potential sufficiently displaced from E_corr, the value of −1.9 V_SCE was selected, and the results are compiled in Figure 5. The increased cathodic activity is readily apparent for all prior anodic potentials pre-treated where the net cathodic current density is approximately 40 times greater than that experienced at cathodic potentials. The statistics from replicate experiments indicate the phenomenon of cathodic activation is highly reproducible.

Zoom In Zoom Out Reset image size

RBS analysis performed in this study is shown in Figures 6–8. Figure 6 displays RBS data for a ground specimen without chromic acid cleaning where the number of counts measured at a given energy level is shown in Figure 6a whilst Figure 6b shows the corresponding composition profile calculated from a simulation fit in the SIMNRA program. Note that the x-axis for all concentration vs. depth profiles have been expanded to focus on the Mg surface while utilizing a linear axis. This figure indicates that the freshly prepared surface (not exposed to electrolyte) consists predominantly of a Mg oxide/hydroxide covered surface with a trace levels of Si which we have ascribed to mechanical grinding with the Si incorporated in the oxide/hydroxide layer and at the film/metal interface; hydrogen is not included in the analysis as it is a forward scattered particle and will not appear in the backscattered spectrum. The broad low energy shoulder on the left of Figure 6a indicates bulk Mg with the low energy near surface oxygen peak superimposed. The oxygen peak indicates little energy loss from its theoretical level and thus indicating that the oxygen resides at the surface. Fitting of the oxygen peak utilized input data for the elastic scattering cross section measured at a scattering angle of 165°. This data was used because the SIMNRA program does not possess an elastic scattering cross section for a scattering angle of 160° which results in an overestimate of the oxygen peak intensity. The presence of elements in the zone just below the oxide is near the detection limit for all elements with an atomic mass ranging from Si (28.09 amu) to Fe (55.85 amu).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The RBS data of a ground sample followed by chromic acid cleaning (Figure 7) indicates a sharp peak consistent with near surface Cr. It is noted from the raw RBS data in Figure 7a that the oxygen peak is negligible, indicating that chromic acid is indeed capable of corrosion product removal with minimal adverse effects such as additional oxidation and scale formation (as advocated in ASTM standard G-1⁴³). In addition, the composition of Fe (Figure 7b) remains constant on the surface.

Zoom In Zoom Out Reset image size

In order to detect possible enrichment of elements other than Mg upon the dissolving metal surface, RBS was performed on a specimen exposed for 24 hrs potentiostatic polarization at −1.625 V_SCE in 0.6 M NaCl and cleaned using CrO₃ solution (Figure 8). This specimen was chosen for analysis because accurate RBS requires a flat test surface with minimal roughness which was not available on the heavily corroded surfaces of anodically polarized samples. In addition, this sample showed enhanced hydrogen collection during anodic polarization (Figure 2b), enhanced HER at −1.9 V_SCE (Figure 4a), and experienced polarity reversal during potentiostatic polarization (Figure 1b) which made it ideal for RBS analysis testing for possible surface enrichment by noble elements (Figure 4). Figure 8 indicates the presence of Cr as well as Fe and a near surface oxide/hydroxide as expected from Figures 6 and 7. However, the measured intensity of the broad Cr peak in Figure 8a decreased gradually over a broad range of energy in contrast to the sharp Cr peak observed in the non-corroded sample (Figure 7). This suggests that Cr is not wholly responsible for the gradual decrease in energy. The interaction energies of Cr and Fe are only separated by 49 keV, thus measured intensities for both elements will be superimposed. The decrease in energy observed is indicative of an enrichment of Fe in the residual oxide/hydroxide and on the metal surface decreasing as a function of depth into the sample. The simulation of this peak shows the composition of Fe on the surface is ca. 18.5x greater than that in the bulk metal (Figure 8b). Additionally, Figure 8c shows that the Fe:Mg ratio (i.e., C_Fe/C_Mg) at the lowest value on the surface is ca. 30x greater than in the bulk metal (2.38 × 10⁻³ versus 8.00 × 10⁻⁵).

Several independent groups have published evidence supporting hydrogen evolution on a corroding Mg anode to be a cathodically activated process, as opposed to a chemical reaction in solution.^{15–22,32,49–51} However, this is the first study with concurrent hydrogen evolution, ICP-OES, electrochemical, and chemical detection (RBS) of element enrichment as a physically based description concerning the origins of the NDE. The data presented in this study are in agreement with these prior reports. During anodic potentiostatic polarization tests, a reduction of the net anodic current was measured after 100–1000 seconds resulting in essentially complete coverage of the corroding surface with a black corrosion product. The origin for the color of this black corrosion product has not been fully described in the literature although some attempts have been made. For instance, Davidge reported dark brown and black samples of single crystal MgO doped with Fe at several different heat treatments.⁵² Reports by Williams^16–19 using SVET have revealed activation of cathodic regions behind the advancing anodic front at both OCP and while under potentiostatic polarization. It was also shown that the cathodic current density increased over time proportionally to the surface coverage of corrosion products. This, therefore, would account for the decrease in net anodic current density observed in the 100–1000 second time window during anodic potentiostatic scans. Once the entire surface is covered with corrosion product, localized corrosion currents can no longer be resolved with SVET and a more general form of corrosion begins. As such, the measured anodic current density increased before decreasing again due to alkalinization of the bulk NaCl solution to pH values in the stable region for Mg(OH)₂. A thick corrosion film is allowed to form which covers cathodes with implications of its own regarding the cathodic reaction rate. However, it should be noted that an anodically activated and cleaned sample has an equal or greater cathodic reaction rate than ones covered in Mg(OH)₂ (Figure 4b). Since Mg corrosion is under cathodic control, the anodic dissolution rate and net anodic current would be expected to decrease when the corrosion film covers the surface. Mg(OH)₂ is not a suitable passive layer due to its porous nature which leads to crack formation and spalling of the hydroxide revealing fresh areas of Mg metal to be oxidized,^16,53 whilst the Mg(OH)₂ may even allow electron transfer itself. The noise at the end of the −1.4 V_SCE and −1.45 V_SCE scans observed in Figure 1a can be interpreted as a result of crack formation and spalling of the oxide/hydroxide corrosion film. Breakdowns in the Mg(OH)₂ film would be expected to cause an increase in the anodic current density after which the anodic current density will decrease by some combination of enhanced catalytic activity and reformation of the oxide/hydroxide, the latter aspect contributing to the development of hydrogen as a by-product of the re-filming process.

The most interesting case of enhanced catalytic activity is inherent in the sample previously held at −1.625 V_SCE, which indicates a polarity reversal from net anodic to net cathodic current. Despite having a more positive potential than the other net cathodic potentials studied, a greater net cathodic charge density as well as volume of hydrogen captured was measured as shown in Figure 2. This is not surprising given an assumed increase in the HER exchange current density after undergoing anodic dissolution supported by RBS detection of Fe. Unfortunately, the hydrogen captured was not measured as a function of time so it is difficult to ascertain how much of the hydrogen was captured during the period of anodic polarization versus cathodic polarization. Nevertheless, larger quantities of hydrogen would be expected based on the measured net cathodic current.

Early studies by Hanawalt¹² established the effects of impurity concentration on the corrosion rate of pure Mg. In doing so, they discovered a tolerance limit at which point the corrosion rate increased multiple orders of magnitude. For Fe, this tolerance level was found to be about 170 ppmw (μg/g) or an atomic Fe:Mg ratio of 7.4 × 10⁻⁵. The Fe:Mg ratio on the Mg surface after anodic polarization at −1.625 V_SCE was found herein to be between 1.04 × 10⁻³ and 2.81 × 10⁻³. As the exchange current density for hydrogen evolution on Fe, Ni and Cr occurs at a much higher rate than on Mg,^54,55 an enrichment of more than an order of magnitude should produce a large effect on cathodic E-log(i) kinetics dominated by hydrogen even though surfaces have predominantly Mg composition.

The degree of enhanced cathodic activity on a corroding Mg anode is clearly shown in Figures 4 and 5. It is apparent from the data in Figure 5 that there is a limit to the possible enhanced catalytic activity. Taub et al.⁵⁶ reported increases in the HER kinetics with increasing elemental additions of Fe to Mg but observed no further increases in cathodic reaction rate beyond about 7–8 at% Fe in 2 M NaCl solution. It is reasonable to assume that the concentration of Fe enrichment on the Mg surface reaches a maximum value after undergoing extensive corrosion damage. Reasons for this are speculative but could involve reduced catalytic capability when incorporated in poorly conductive MgO or Mg(OH)₂ films or Fe-rich particles detached from the surface, and simply, poor enrichment efficiency. The present work indicates that the efficiency of Fe enrichment would in fact be very low, which is perhaps why lower resolution forms of spectroscopy to measure Fe enrichment have not been successful. We indicate an enrichment herein of about an order of magnitude, which is appreciable in the context of Mg corrosion, however the total Fe content remains well below 1 at.%. The degree of enhanced cathodic activity toward the HER also reaches a maximum value. It was observed in Figure 2a that the relative contribution of HER to the right hand side of Equation 1 decreased with increasing anodic overpotential. This observation was also reported by Frankel et al.¹⁵ as well as Bender et al.⁵⁷ and confirms the notion of a decrease in the contribution of enhanced catalytic activity at increasing anodic overpotential.

The Fe enrichment observed in RBS only accounts for Fe on the surface of the alloy but does not account for possible Fe-rich particles embedded in the thick oxide/hydroxide since most of this layer was removed CrO₃ cleaning. Moreover, RBS cannot detect whether the enrichment of Fe is by accumulation of Fe-rich particles on the alloy surface or by elemental enrichment in the α-Mg matrix. The limited solubility of Fe in Mg would point toward the accumulation of Fe-rich particles.³ One would expect Fe-rich particles to be present in the oxide/hydroxide layer based on the TEM studies performed by Taheri et al.³² Any Fe embedded in the corrosion layer would therefore be expected to be detected by ICP-OES measurements only if dissolved. Unfortunately not all of the product remaining on the Mg surface after corrosion could be dissolved with chromic acid and hence made available for ICP-OES. The only corrosion product analyzed was that which was present in solution and could be mechanically removed from the corroded Mg surface after testing. The remaining corrosion product from the sample was collected after CrO₃ acid cleaning and tested with ICP-OES. However, the presence of high Cr concentrations with respect to Fe caused interferences with the measured Fe intensities resulting in non-usable data. It is likely that there are some Fe cations in solution as a result of dissolved particles in the corrosion layer. However, the levels were at or below the detection limit.

Fe dissolution from the Mg surface is not expected to occur during the potentiostatic tests as the potentials tested in this study are more than 1 V more negative than the reversible potential of the Fe/Fe²⁺ reaction.⁴⁵ Moreover, Swiatowska et al.²² did not observe quantities of Fe in solution using an in-situ atomic emission spectroelectrochemical (AESEC) method at various anodic pulses varying between −1.433 V_SCE and −0.333 V_SCE in a 0.01 M NaCl electrolyte which supports this claim. If Fe and other impurity elements are not dissolving into solution then they must be enriching on the Mg surface, and this has been demonstrated in this study.

The results presented do not explain recent results for Mg-Ca and Mg-Li alloys since these elements presumably can dissolve preferentially to Mg yet the NDE is still observed.⁵¹ However, Fe enrichment in these alloys is still possible and may participate in manifestation of the NDE. Further investigation would be required. The efficiency of enrichment and the role of other alloying elements such as Ca additions are areas of important future work, and whether the enrichment alone is sufficient to account for the whole of the NDE phenomenon by the enhanced catalytic effect mechanism, will also be an important future clarification. The work herein has also contributed to the notion that the dissolution of Mg is predictable on the basis of Mg dissolving as Mg²⁺.