Investigating the Effect of pH on Corroding Magnesium by Dynamic pH Sweep Coupled with Electrochemical Measurements

Magnesium is the most reactive engineering material due to the low equilibrium potential for magnesium oxidation (−2.37 V_SHE) and the poor protective properties of the oxide/hydroxide film that naturally forms on the surface.^1-3  The corrosion potential of magnesium and magnesium alloys in aqueous environments is generally well below the equilibrium potential for hydrogen evolution; consequently the main cathodic reaction is hydrogen evolution and the access of oxygen to the corroding surface, a limiting factor for many corrosion scenario, plays a significant role only in relatively dry conditions, such as atmospheric corrosion.⁴  As a result, the corrosion rate of a fully immersed magnesium electrode is generally relatively high and, when corrosion occurs in a set volume of electrolyte, the pH of the electrolyte tends to increase relatively rapidly with time, to attain a steady value between 9 and 10 in 3.5% NaCl.

The Pourbaix diagram of magnesium indicates that magnesium tends to become passive at high pH (between the nominal pH 8.5 and 11, depending on the concentration of Mg ions) and this can potentially represent an issue from the perspective of corrosion testing, as the corrosion process modifies the environment (by increasing the pH) which in turn modifies the corrosion process (by promoting passivity). As a result, the data obtained, in principle, might be affected significantly by the ratio between the corroding surface and the volume of the environment. The situation described is even more complex considering that the local pH at the corroding surface might be significantly higher than the bulk solution pH, and that loosely adherent corrosion product of Mg ions containing gels might play a role during the corrosion process.

Based on the above considerations, it is evident that decoupling the intrinsic corrosion behavior of a corroding magnesium surface from the modifications that it unavoidably induces in the environment would be beneficial both from the fundamental perspective of understanding the corrosion behavior and for the purpose of a developing reliable and reproducible corrosion testing method. This can be pursued by various approaches, including using pH buffers,^5-9  rotating electrodes,^5,10  or flowing electrolytes.^11-13  For example, it is well documented that the use of pH buffers or use of flowing electrolyte have a very significant effect on the corrosion behavior of magnesium. In the presence of pH buffers that prevent alkalinization, the corrosion rate is significantly enhanced and some of the common features associated with magnesium, for example, the superfluous hydrogen evolution during anodic polarization,^11,14-22  can be suppressed.²⁰  In the presence of alkaline conditions, the passivity of magnesium is established rapidly and corrosion is minimal.²³  When performing electrochemical measurement under vigorous flow, such as by using a rotating electrode, the results are somewhat similar to when a buffer that prevents alkalinization is used.^5,10 

Both of these methods, however, have some disadvantages. The use of buffers introduces foreign species in the environment (the buffer) which might have a complex effect on the corrosion process.²⁴  Thus, the behavior observed is affected to a lesser extent by the pH variation with time, but to a significant effect by the buffer chemistry. The use of vigorous agitation can avoid the issue of the foreign species, but the perturbation of the corrosion process due to the hydrodynamic conditions is significant, and the response is affected both by the corrosion properties of the surface and of the hydrodynamic regime.^5,10  Further, measurements performed by rotating electrodes are generally short-term measures and not long-term corrosion tests.

Another approach to avoid the issue of macroscopic environmental pH changes during the corrosion testing of magnesium could be to continuously pump new test solution into a test cell and extract a corresponding volume of used test solution, such as after an initial transient, the pH in the test cell becomes stationary. This approach is in principle effective, as it avoids both the use of foreign buffer species and the solution in the test cell is almost quiescent. However, because pumps and solution reservoir are required, it is experimentally more complex compared with testing in a normal container, and it is rarely done in practice. Further, although in principle this setup could be used to investigate the effect of a dynamic pH variation, this would come at the expenses of a significant complexity in the experimental setup, because the pH scale is logarithmic and it is therefore difficult to dose acid or alkaline solution such that the pH could be varied over a wide range.

In this work, an experimental setup is introduced for dynamically controlling the pH of the environment where a magnesium specimen is exposed. The method exploits the pH changes associated with the evolution of oxygen and hydrogen from an inert electrode and it can be used to control the pH in a closed container. Its use is demonstrated by performing cyclic pH sweep experiments coupled with a series of corrosion potential and linear polarization measurements, such that the dependence of corrosion potential and polarization resistance on pH and experimental time can be disclosed.

The experimental setup developed in-house to perform electrochemical testing under variable pH conditions is schematically presented in Figure 1. An electrochemical cell, divided in two compartments, was obtained by 3D printing. The division between the two compartments was realized such that a polymeric sponge could be placed within an adequate 3D printed support. The purpose of dividing the cell in two compartments was to allow ionic exchange between the solutions in the two compartments but to prevent macroscopic mixing. In each of the two compartments, one graphite electrode was placed (oxygen/hydrogen evolving electrodes in Figure 1). The graphite electrodes’ couple was used to evolve oxygen in one compartment and hydrogen in the other by electrochemical water splitting. As a result of electrochemical polarization of the graphite electrodes couple, the pH in the two compartments varied, increasing in the compartment where hydrogen was evolved (graphite negatively polarized) and decreasing in the compartment where oxygen was evolved (graphite positively polarized). In one of the two compartments (identified as test cell herein), a pH probe was placed to acquire the pH value. A pH control analog interface was developed in-house in order to acquire the pH value from the probe and to control the direction and the magnitude of the polarization between the two graphite electrodes. A mechanical stirrer was also placed nearby the pH probe, such that the pH value in the bulk was relatively homogeneous. The pH control analog interface was connected to an analog-to-digital and digital-to-analog interface (NI USB-6001^†), which was controlled by an in-house developed LabVIEW^† software. The cell geometry, the electrodes’ size and position, and the control software were optimized to minimize the overpotential applied to the graphite electrodes without increasing excessively the time required to complete a full pH cycle. The minimization of the applied overpotential is beneficial because, in chloride-containing electrolytes, the chlorine evolution reaction may compete with oxygen evolution at large values of overpotential, particularly at low pH. Even if the issue of chlorine evolution was considered and minimized, it cannot be excluded that some chlorine was generated from the graphite electrodes when anodically polarized, particularly when the pH in the test cell was low. Thus, the presence of chlorine gas dissolved in the test solution might have resulted in increased availability of cathodic reactant and it cannot be excluded that this might have impacted to some extent on the corrosion behavior observed. In the test cell, an additional graphite electrode, a reference electrode, and a magnesium specimens were placed (Figure 1). The three electrodes were connected to an in-house built potentiostat analog interface which was connected to the analog-to-digital and digital-to-analog interface. All connections between cell and instrumentation were controlled by individual relays. As a result, an integrated potentiostat and pH controller device was available to perform the electrochemical measurements under variable pH conditions. Because the cell was entirely 3D printed in non-transparent material, it was not possible to image the magnesium surface during testing.

The integrated potentiostat and pH controller device was operated by the in-house software by applying in sequence a pH control step, an open-circuit measurement step, and a linear polarization step. The pH control step comprised two substeps that were iterated until the required target pH was attained. During the first pH control substep, all electrodes, except the pH probe, were isolated by opening the corresponding relays, and the pH signal was acquired for 5 s. During the second pH control substep, all electrodes were isolated except the two oxygen/hydrogen evolving electrodes, which were polarized for 15 s. During this 15 s period, and only during this period, the stirrer motor was also operated. Following iteration and once the target pH was attained, the open-circuit measurement was initiated. During the open-circuit measurement, all electrodes except the magnesium specimen and the reference electrode were isolated by opening the corresponding relays. Once the open-circuit step terminated, the linear polarization step started. In this case, all electrodes were isolated by opening the corresponding relays, except the magnesium specimen, the reference electrode, and the graphite counter electrode. The pH control, open-circuit potential measurement, and linear polarization measurement steps were iterated in this sequence for the entire duration of the experiment. The software adjusted the target pH as a function of time, such that an approximately linear pH sweep was applied to the test cell, and hence to the corroding magnesium electrode (Figure 2).

All experiments were performed in naturally aerated 3.5 wt% NaCl solution at room temperature. The solution volume used in each test was approximately 900 mL, distributed equally in the test cell and in the balancing cell.

Working electrodes were prepared from 2.54 cm² 99.95% Mg rods sections (main impurities in ppm as determined by inductively coupled plasma atomic emission spectroscopy [ICP-AES]: 56 Al, 12 Mn, 29 Fe, 53 Zn, 2.6 Cu). One of the circular surfaces of the magnesium rod section was exposed to the test solution, whereas the other surfaces were masked with beeswax, after connecting an electrical cable. The specimens were polished by 1200 grit polishing paper, using water as lubricant and stored in laboratory air. Immediately prior to testing, the specimens were briefly repolished under dry conditions to generate a fresh metal surface. A commercial calomel electrode was used as reference. The pH of the solution in the test cell was swept between the values of 4 and 10, starting at the natural pH and going toward higher values first. Five complete pH sweep cycles were performed for each experiment. The nominal period of the triangular pH waveform was set to 25,000 s, however, this slightly increased at longer test times. This is likely to be due to pH buffering effects from the increase in the amount of corrosion products in the test solution.

The duration of the open-circuit measurement was 120 s and individual corrosion potential transients were recorded. At the end of each open-circuit measurement, the last potential value was recorded and used to produce the aggregated datasets presented in some of the following figures.

Linear polarization measurements were performed at the sweep rate of 1 mV/s, initiating and terminating at the open-circuit potential. The amplitude of the potential sweep was 20 mV, i.e., 10 mV below and 10 mV above the open-circuit potential. Three full cycles were acquired consecutively and individual current responses were recorded. During the first cycle, the working electrode potential was swept in the cathodic direction first. The value of the polarization resistance was estimated by taking the slope of the linear fit (least squares method) of the complete 3 cycle dataset plotted in a V/I plane. At the end of each linear polarization experiment, the value of the polarization resistance was recorded and used to produce the aggregated datasets presented in some of the following figures. Individual linear polarizations measurements were recorded.

Figure 2(a) presents the time evolution of the pH value during the cyclic pH sweep experiment (the figure was obtained by recording only the last value at the end of each pH control step). As it is evident from the figure, the initial pH was near neutral and subsequently was cycled between the value of 10 and the value of 4. The pH was varied approximately linearly as a function of time, although perfect linearity was difficult to achieve, due to the variation in duration of the individual pH transients at different pH values and as a result of the increased amount of corrosion products in the test solution. The numbers in Figure 2(a) refer to the pH and times when the electrochemical measurements presented in the following figures were acquired.

Figure 2(b) presents a representative sequence of pH transients during the first cycle of the pH sweep experiment. Here, each segment represents the time evolution of the pH during an individual pH control step. Except for the first few transients at the beginning of the experiment, where some small pH overshoots were observed, generally the individual transients resulted in a relatively precise pH control with a tolerance which was generally below 0.2 points of pH. The initial overshoots in the pH are likely to be due to the fact that the diffusion regime across the two compartment of the cell might take some time before attaining a relatively steady state. However, such initial overshoots are small compared to the amplitude of the applied sweep and are unlikely to affect significantly the overall behavior.

In Figure 2(c), the individual transients to reach the nominal values of 4, 7, and 10 (points 4₁ to 4₅, 7₁ to 7₆, and 10₁ to 10₅ highlighted in Figure 2[a]), preceding the electrochemical measurements reported in the following figures, are presented (time in logarithmic scale). It is evident that the time required to achieve the target pH value of 10 was significantly higher than the time required to achieve pH 4 and 7. Such different transient as a function of pH is likely to be at the origin of the difficulty in applying a perfectly linear pH ramp. The figure also highlights that the deviation of the final pH from the target values of 4, 7, and 10 was generally below 0.25.

Figures 3(a), (b), and (c) present the corrosion potential records acquired at pH 7, 10, and 4, respectively. In the figure, the value of the corrosion potentials has been offset for the value of the corrosion potential at the beginning of the record to facilitate visual comparison. Thus, the corrosion potential variation with respect to the corrosion potential at the beginning of the record is reported. At pH 7 (Figure 7[a]), the first potential record (7₁) displayed a relatively large drift compared to the following (7_2-6). Notably at longer experimental times, well-defined individual potential transients become apparent, typically associated with localized corrosion events. The overall behavior was similar for the records acquired at pH 10 and 4 (Figures 3[b] and [c]). Notably, the overall shape and amplitude of the drift during the entire record did not appear to be substantially affected by the pH value.

The current responses acquired during linear polarization at pH 7, 10, and 4 are presented in Figures 4(a), (b), and (c) respectively. In Figure 4(a), the linear polarization labeled 7₁, which was acquired at the early stages of the experiment at pH 7 (see Figure 2[a]) displayed a significant deviation from linearity and a large hysteresis, indicating a complex electrochemical behavior and a large pseudo-capacitance. This is unsurprising considering the electrochemical response of magnesium during the early stages of immersion; for example electrochemical impedance measurements in similar environment reveal a large capacitive contribution and a significant inductive response.^16,25  These are expected to result in nonlinear response during the relatively fast (1 mV/s) linear polarization applied. Subsequent polarization resistance measurement at pH 7 following a pH sweep to pH 4 and back to 7 (curve 7₃) did not result in an increased non linearity and hysteresis; on the contrary, the behavior appeared almost perfectly linear. A similar behavior was observed for the linear polarizations performed at the pH 10 (Figure 4[b]), namely, a significant deviation from linearity during the early stages (curve 10₁), followed by a more linear behavior during the later measurements. At pH 4, only a slight deviation from linearity was observed for the first linear polarization (curve 4₁), whereas the following measurements displayed an almost perfect linear behavior (curves 4₂ through 4₅). Overall, the shape of the linear polarization response appeared to be more affected by the time passed from the beginning of the test rather than by the value of the solution pH.

Figure 5 displays pH, polarization resistance, and corrosion potential as a function of experimental time for two tests conducted under nominally identical conditions. The pH/time curves were broadly similar, indicating that the setup used to perform the experiment was capable of inducing relatively reproducible pH conditions. For both of the tests, the polarization resistance increased over time from a value of approximately 100 Ω·cm² measured immediately after immersion, to values between 600 and 1,600 at later stages. Fluctuations in the value of the polarization resistance were observed after the third pH cycle, with higher polarization resistance measured when the pH was high and lower polarization resistance measured when the pH was low. It should be noted that the initial values of polarization resistance (up to approximately 20,000 s) should be considered qualitative, due to the significant deviation from linearity. However, considering the individual response, it is evident that the value of polarization resistance measured during the early stages of the test was certainly lower than the more precise value estimated at later stages. The corrosion potential displayed a similar behavior, i.e., it was relatively steady during the first part of the test, and fluctuated during the central and final parts of the test.

Figure 6 shows the dependence of the polarization resistance on the environmental pH. From the graph, two regions can be identified: an early stage region, where the polarization resistance did not correlate with the environment pH but essentially increased with time regardless of the instantaneous pH; and a second region, where the polarization resistance increased with increasing pH. Some hysteresis is evident in the curves, but a clear trend associated with the number of cycle was not evident. Broadly, it appears reasonable to conclude that the response after the third cycle was not substantially affected by the previous pH history.

Figure 7 shows the dependence of corrosion potential on the solution pH. Generally, the potential is high and steady for the initial part of the test and subsequently appears to be correlated with the pH, with higher values of potential measured at higher pH. Overall, the corrosion potential drifted toward more negative values with increasing number of pH cycles.

In this work, an experimental setup exploiting the pH changes due to hydrogen and oxygen evolution from electrochemically polarized inert electrodes for controlling dynamically the pH of a corrosion testing solution was developed. The experimental results show that the pH can be controlled with a tolerance of the order of 0.25 pH units (Figure 1), and that the applied pH signal is relatively reproducible (Figure 5). It is evident, however, than the linearity of the pH signal deteriorates significantly with test time, and the time required to perform a full cycle increases. This behavior is likely to be due to the increased presence of magnesium species in the test solution, which provide some buffering effects during the pH sweep. From the practical viewpoint, however, such issue does not appear to be substantial, and does not prevent the acquisition of meaningful electrochemical measurements under pH sweep conditions.

From the electrochemical measurements on the magnesium electrodes, it is evident that experimental time and pH value both affect the corrosion behavior of the magnesium surface. However, the observations indicate that the behavior can be broadly divided in two stages (Figure 5): a first stage when the polarization resistance monotonically increases, regardless of the instantaneous pH value, and a second stage where the polarization resistance fluctuates in response to the pH fluctuations. The duration of the first stage was between 25,000 s and 50,000 s, whereas the second stage lasted for the remaining duration of the test.

The behavior observed can be rationalized considering that initially the magnesium surface has a silvery appearance and it is covered by a relatively thin air-formed film, with relatively poor protective properties.¹⁸  Upon immersion, the film is locally attacked and a corrosion front develops. The corrosion front is characterized by the occurrence of the anodic reaction of magnesium oxidation and vigorous hydrogen evolution in the form of hydrogen streams.^{14-15,18,23,25}  The hydrogen evolution at the corrosion front is due to the attack of the pre-existing air-formed film, which results in close contact between the magnesium and the electrolyte.^19,25-27  The corrosion front progressively propagates horizontally on the silvery electrode surface, leaving behind a darker surface,^21,28-29  which is covered by corrosion products and it is cathodically active compared to the silvery surface ahead of the corrosion front.^{15,18,25,30-31}  Over time, the initially silvery surface is completely converted in a darker surface, covered by corrosion products. Once the surface is entirely dark, the corrosion sites tend to localize, and the horizontal propagation of a corrosion front is no longer evident.

During the early stages of immersion, the corrosion behavior is similar to active corrosion, as evident from the potential transients where the potential drift is large and individual potential transients associated with large individual corrosion events are not evident. Accordingly, the initial linear polarization curves show large values of current for relatively low applied potential and a significant hysteresis due to a large pseudo-capacitance. This indicates that during the initial stages of testing the surface has a high pseudo capacitance, typical of active corrosion. This suggests that during the initial stage of horizontal propagation of the corrosion front, the electrochemical response is largely dominated by the low-impedance corrosion front, and the two filmed regions (air-formed film and corrosion product film, ahead and behind the corrosion front) provide little contribution to the overall response. The response of the horizontally propagating corrosion front is in essence similar to active corrosion, as the film is transiently disrupted at the active corrosion front, where simultaneous magnesium oxidation and formation of hydrogen streams occur. During this initial stage, the instantaneous value of pH has little effect on the value of polarization resistance (Figures 5 and 6) because the pH is likely to affect mostly the behavior of the filmed regions, but the electrochemical response is dominated by the nonfilmed corrosion front. It is, however, likely that above a critical pH, the air-formed film becomes protective and horizontal propagation of the corrosion front is hindered or stopped.^9,23  However, this has not been observed in the pH range inspected in this work.

At longer experimental times (i.e., after 25,000 s to 50,000 s), the behavior of the corroding surface was much closer to the behavior of a filmed surface undergoing localized corrosion; individual transients associated with corrosion events become apparent in the corrosion potential curves (Figure 3), and an almost perfectly linear behavior was observed from the linear polarization curves (Figure 4). Almost no hysteresis was evident in the current responses, which is consistent with a substantial reduction in the surface capacitance due to the formation of a relatively thick film. This suggests that, after the initial stage of horizontal propagation of the corrosion front, a corrosion product film is formed on the entire surface. The active corrosion sites are now localized and stable. Corrosion might proceed below the surface, but large areas of actively propagating corrosion front are no longer present a the outer surface. Hence, the electrochemical response is now dominated by the behavior of the corrosion products film that covers the entire electrode surface. Such film is in dynamic equilibrium with the environment and a variation in the environment pH induces a variation in the electrochemical response.

Thus, in summary, after an initial stage of active corrosion, associated to the horizontal propagation of the corrosion front and to the consequent silvery to dark transition, the electrode surface was covered by a corrosion product film in dynamic equilibrium with the environment, which determined the electrochemical response behavior. During the first stage, the pH of the environment did not appear to have a substantial effect on the electrochemical response because of the active nature of the corrosion processes at the horizontally propagating corrosion front, not mediated by a relatively stable film.

After the entire surface was attacked and the corrosion product film was established, the effect of the pH on the corrosion behavior became substantial. In particular, the polarization resistance was affected by the instantaneous pH value (Figure 5), with variation of approximately a factor of two between the values measured at pH 4 and those measured at pH 10. Interestingly, the polarization resistance response did not appear to have a significant memory effect, i.e., it did not appear to increase or decrease substantially as based on the previous number of pH cycles. This suggests that the film that controls the corrosion rate on the magnesium surface is in dynamic equilibrium with the environment and, at least the layer that limits the corrosion rate, does not build up at each cycle.

From the practical perspective of corrosion testing of magnesium, the results indicate that initially the pH does not have a substantial effect on the corrosion behavior of the magnesium surface. Hence, controlling accurately the pH when performing short-term measurement (i.e., on a completely or partially silvery electrode) is probably not a strong requirement, and the ratio between the volume of the test solution and the exposed area does not play a substantial role. Vice versa, the pH has a significant effect at later stages (i.e., on a completely darkened electrode), where the polarization resistance positively correlates with the pH value. From the perspective of long-term corrosion testing, it is then important to obtain reproducible pH conditions, as the value of pH has a direct impact on the film behavior and on the electrochemical measurement.

• A setup to control the pH of a solution used for corrosion testing of magnesium has been developed. The setup exploits the pH changes associated with oxygen and hydrogen evolution from electrochemically polarized inert electrodes. The possibility of applying a cyclic pH sweep to a corroding environment has been demonstrated, and its impact on the corrosion of magnesium has been discussed. It was found that initially the electrochemical response of magnesium is not substantially affected by the instantaneous value of the environment pH, because the interaction between air-formed film and environment has little effect on the initial propagation of the corrosion front within the pH range inspected. However, at longer test time, the electrochemical response is dominated by the response of the corrosion product films, which is in dynamic equilibrium with the environment. Thus, the environment pH had a significant effect on the electrochemical response of the magnesium surface and the polarization resistance correlated with the pH. From the practical perspective of magnesium corrosion testing, the results suggest that precise pH control is not needed for relatively short-term measurements (i.e., on completely or partially silvery electrodes), but it is important for long-term corrosion testing (i.e., on completely darkened electrodes).

The authors acknowledge the support of The Engineering and Physical Sciences Research Council (LightForm - EP/R001715/1 Programme Grant).