Tailoring of Mg and MgLi thin-film corrosion rates with dielectric barrier discharge plasma treatment

Magnesium and magnesium alloys are widely studied biodegradable materials as candidates for medical applications. Applications of interest reach from biodegradable orthopedic implants and stents up to the possible field of therapeutical treatment.^1–5 To ensure the duration of the integrity of the implant required by the application or the therapeutic activity, the degradation rate needs to be adjustable. Since the corrosion rate of Mg and Mg alloys is often too high for the applications,^6–9 several strategies to reduce the rate have been developed: These include alloying of the bulk material,^10–12 influencing the microstructure,^12–14 surface treatments, and coatings.^15–17 The surface treatments range from deposition of inorganic^18–20 and organic^21–23 deposition coatings over chemical conversion by immersion in solution^24,25 to ion implantation^26,27 or plasma electrolytic oxidation.^28–30 

In this study, the focus of the application is on miniaturized implants fabricated by MEMS (micro-electro-mechanical systems) technology which can serve as, e.g., biodegradable materials for brain implants, either as structural implants, substrates for additional materials or reservoirs for therapeutically active ions. As an example of degradable films that release ions which can act as treatments, MgLi thin films are studied since Li is used in treatments for mood disorders such as bipolar disorder.^31,32 The possibility to include additional ions in a coating to reduce the corrosion rate specifically for the application as reservoirs is limited since they could influence the therapeutic effect if they are released during the degradation. While small Mg structures with a thickness of 10–100 μm can be prepared by thin-film deposition techniques such as sputtering,^33–35 the addition of a thick coating would additionally counteract the effort to reduce the implant size. Thus, a treatment that leads to a thin layer only consisting of the alloy elements itself, possibly also in the form of, e.g., oxides and carbonates, is beneficial. Possible techniques to form those layers include, e.g., chemical treatment in solution to form MgO or Mg(OH)_2,^36,37 ion implantation of nitrogen,³⁸ or plasma treatments.^39–42 Kocijan et al. showed that plasma treatment in O₂ and H₂ leads to pin–hole free oxide layers and, thus, lowering the corrosion rate.³⁹ The formation of an oxide film is also studied for plasma treatment with Ar/O₂ by Tiyyagura⁴⁰ while Nakazawa et al. examined the implementation of nitrogen and oxide after the treatment of a Mg surface with an atmospheric plasma jet.⁴¹ Additionally, carbon contaminations on the samples are removed by plasma processes.³⁹ Li et al. show that a dielectric barrier discharge (DBD) plasma treatment can reduce or even prevent the corrosion of MgLi alloys with a Li content of 14.2% (m/m) significantly by the formation of Li₂CO₃ and oxide-containing layers.⁴² Atmospheric pressure plasmas are highly interesting to influence surface properties in an economic way. In comparison to other techniques, DBDs can easily be scaled up to surface treatments in industrial scales.^43,44 By DBD treatment oxygen groups or compounds can form on the surface by the interaction with elevated oxygen species from the plasma.⁴⁵ In the literature, DBDs are also discussed to clean the surface of Mg from remaining contaminations⁴⁶ or to lead to the reaction of carbon-containing components to carbonates additional to the formation of an oxide layer.⁴⁷ Since for thin films no prior grinding and polishing can be easily performed, the process can, thus, provide advantages additionally to the reduction of the corrosion rate due to the formation of protective layers including oxides and carbonates.

This study aims to evaluate the effect of a DBD plasma treatment to tailor the corrosion resistance of thin films of Mg and MgLi. Since the MgLi samples in this study have a Li content of only 1.6% (m/m), it does not lead to additional Li rich phases which could lead to high Li concentrated protective surface films as seen in previous studies,⁴² but might influence the activity and microstructure. Additionally, the nature of thin films with μm thickness does not allow the implementation of thick protective layers. Thus, the effect of the treatment forming layers below 1 μm thickness on the corrosion rate is studied in Hanks’ balanced salt solution to simulate the environment in medical applications. Additionally, the layer formed during the treatment is analyzed regarding the structure and chemical composition. By including Mg and MgLi [1.6%(m/m)], the influence of alloying and varying microstructure can provide a detailed understating of the overall process.

Magnesium and magnesium-lithium films were prepared by magnetron sputtering (Von Ardenne CS730S), using targets of pure Mg and MgLi [2.5%(m/m) Li] from FHR. As a substrate, 4 in. silicon wafers were cut into 15 × 15 mm² samples and coated with aluminium (Al) and aluminium nitride (AlN) for comparability to freestanding thin films. The sputtering was carried out with a base pressure of < 5 × 10⁻⁷ mbar and an Ar pressure of 2.3−2.6 × 10⁻³ mbar with a gas flow of 25 SCCM Ar. A final sample thickness of 10–20 μm of Mg and MgLi thin films with a Li mass fraction of 1.6% (m/m) was reached. For one measurement set, all samples were prepared in the same sputtering process, thus, with the same thickness for untreated and treated samples. Additionally, freestanding thin films were prepared with the same sputtering parameters, following the process described by Haffner et al.³⁴ After structuring on the wafer by UV-lithography and etching, a sacrificial AlN layer was added before the final layer of Mg or MgLi was deposited. For the etching of the sacrificial layer, the samples were afterwards immersed in 20 wt. % KOH solution.

The thin-film samples were treated in the gap of a self-built symmetric volume dielectric barrier discharge setup. All treatments were performed in stagnant ambient air. The freestanding thin films were placed on a Si chip (15 × 15 mm²) and fixed on the edges to ensure a flatter surface during the treatment. A laboratory power supply (SM7020-D, Delta Elektronika) and a function generator (DDS function generator 4025, Peak Tech) were connected to the high frequency high voltage power supply (Minipuls 4, GBS Elektronik). The resulting sinusoidal voltage signal of the high frequency high voltage power supply with an output of 1:2000 was monitored via an oscilloscope (UTD2025CL, UNIT). The treatment time of the plasma was controlled by an inhibiting signal by a microcontroller board (Arduino nano every) connected to the high frequency high voltage power supply. The DBD setup connected to the power supply can be seen in Fig. 1.

The parameters of the plasma power supply were chosen to allow the formation of filaments all over the sample surface during the complete treatment time, giving a state of saturation (Table I). An overlay of the plasma filaments at all times of the treatment can be seen in Fig. 1(b). The parameters were adjusted if the setup had to be adapted in between measurements of different sample sets due to functioning reasons to ensure a homogeneous surface treatment. For each sample and measurement type, treatments with different treatment times of 5–25 s were carried out with the same parameters to ensure comparability.

The surface of the thin films was imaged and analysed using a Zeiss Ultra 55 Plus scanning electron microscope (SEM) and an Oxford Instruments ULTIM MAX 65 energy-dispersive x-ray spectroscope (EDX). An accelerating voltage of 3 kV for imaging and 10 kV for EDX measurements was used. Additionally, the structure of the thin films was analysed by x-ray diffraction (Smart Lab 9 kW, Rigaku) with a parallel beam and monochromatic Cu Kα radiation on a θ/2 θ-scan with a range of 20°−90° with a speed of 5−10°/min and a step size of 0.03°. The chemical composition of the sample surface was characterized by x-ray photoelectron spectroscopy (XPS). For this purpose, an XPS UHV system from Omicron Electron Spectroscopy Ltd. with a 240 W Al anode was used. Survey scans to screen for the elements present at the surface were conducted at a pass energy of 100 eV, a step-size of 0.5 eV, and averaged over three sweeps. High-resolution scans of the characteristic core hole-level spectra used for the chemical analysis were conducted at a pass energy of 30 eV, 15 sweeps, and a step-size of 0.05 eV. For data analysis, the software casa xps (Version 2.3.23PR1.0) was utilized and charge correction was done by shifting the C 1 s main peak to 284.8 eV and adjusting all the corresponding spectra accordingly. The corrosion rate was determined by potentiodynamic polarisation measurements in a 155 mmol Hanks’ balanced salt solution (H1387, Sigma-Aldrich with added sodium bicarbonate) at a pH of 7.4 ± 0.2 (CO₂ regulation) and a temperature of 37 ± 1 °C. A VersaSTAT 3–300 potentiostat (AMETEKSI) and a three-electrode setup with an Ag/AgCl reference electrode, a Pt mesh counter electrode, and the sample included into a sample holder with an exposed area of 0.916 cm² were used. After 5 min of measuring the open circuit potential (E_OCV), a linear voltage sweep from −0.3 V vs E_OCV to +0.3 V vs E_OCV was performed. Further information can be found in Ref. 48. Additionally, the same measurement was carried out after 1h of immersion time.

Representative cross sections and surface images of the sputtered thin films of Mg and MgLi are shown in Figs. 2(a)–2(d). While for both materials a columnar growth is visible, the surface exhibits different structures for both sample types with a more structured surface for Mg and larger grainlike areas for MgLi. EDX analysis of the surface of the thin film identify an oxide and carbonate signal for both sample types with no significant difference between Mg and MgLi [2%–3%(m/m) C and 0.5%–1%(m/m) O]. No further quantification of oxide or carbonate components is carried out due to the possibility of the influence of contaminations on the exact intensity. In Fig. 2(e), XRD diffractograms show signals of the thin film itself and the substrate. Mg and MgLi have a hexagonal closed packed (hcp) structure with a strong preferred orientation for pure Mg.

Samples of Mg and MgLi on the substrate and freestanding films of MgLi were treated with an atmospheric pressure dielectric barrier discharge plasma in air with a saturation of filamentary discharges, as seen in Fig. 1(b).

Exemplary potentiodynamic polarization curves for an untreated and 15 s treated sample after 5 min of immersion are shown for Mg and MgLi in Fig. 3(a). The corrosion current densities and corrosion rates can be determined via Tafel extrapolation.^48,49 The cathodic branch was used for the estimation of the corrosion rate for the following studies due to the larger linear area. The current density of the anodic branch is influenced by additional hydrogen evolution, film formation, and passivation regions,^50–52 thus, a corrosion rate determined on the anodic branch may differ from the rate determined from the cathodic branch. MgLi samples were treated for 5, 10, 15, and 20 s. To exclude the effect of Li on the influence of the plasma treatment, Mg samples were treated for 5 and 15 s for comparison. The corrosion rates for the measurements are shown in Fig. 3(b). A significant decrease in the corrosion rate for both material type is apparent, lowering the corrosion rate (CR) during a treatment of 15 s from CR_untreated= 2.19 mm/yr to CR_treated= 0.45 mm/yr for MgLi and from CR_untreated= 1.69 mm/yr to CR_treated= 0.38 mm/yr for Mg. Since the effect for both sample types is similar and Li or the different surface structure does not seem to influence the effect of the treatment significantly, further studies were only carried out with MgLi samples.

MgLi thin films with a lower corrosion rate were measured with the corresponding corrosion rates given in Fig. 3(c). The main decrease in corrosion rate already occurs after a treatment time of 5 s with a reduction of the corrosion rate of about 46% (0.23 ± 0.03 mm/yr to 0.13 ± 0.01 mm/yr). Since the corrosion resistance differs only slightly for longer treatments, the times were set to 5 and 15 s. Since the potentiodynamic polarization measurements show the degradation of samples only for the short term, additional measurements are necessary to confirm the improvement of corrosion resistance over longer time periods. Weight-loss measurements over longer terms are difficult due to the low sample weight of thin films and the cleaning resulting in the removal of corrosion products and possible products formed during the treatment, thus, not allowing the final identification of the corroded mass. Therefore, additional potentiodynamic polarization measurements were carried out after 1h of immersion to allow the prior formation of corrosion products on the surface of the film which can protect the film from further corrosion.^53,54

As shown in Fig. 3(c), the corrosion rate still decreases after the DBD treatment significantly to approximately half of the corrosion rate. Thus, the treatment not only leads to a passivation decreasing the first corrosion before a protective corrosion layer is formed but also leads to a stronger protective layer throughout the degradation.

No significant change in the corrosion potential E_corr can be identified for MgLi thin films after the treatment for 15 s in comparison to the untreated samples. While an E_corr of −1.85 ± 0.08 V is measured for the untreated samples directly after immersion, E_corr for samples with a 15 s treatment measured is −1.85 ± 0.01 V, thus, the potential varies for untreated samples, while it is more stable for the treated samples. After 1h of immersion, the potential is slightly increased to −1.78 ± 0.02 V for untreated samples and −1.79 ± 0.03 V for samples after 15 s of treatment. The increase after longer immersion time can be assigned to the lower activity of the material, possibly due to the depletion of Li on the surface⁵⁵ and the formation of other products during corrosion.^53,54 Thus, the DBD treatment itself does not influence the potential of the material.

For possible applications of Mg or Mg-based alloys as implants for small-size applications and not coatings on other materials, the thin films need to have a thickness in the μm range without an additional substrate. The process to produce the thin films used in this study is described by Haffner et al.³⁴ Since the sacrificial layer of Al and AlN is dissolved in KOH, the films are also exposed to KOH for the duration of the lift-off, resulting in a changed surface. A treatment with KOH is reported to lower the corrosion rate even without additional treatment if the sample is anodized due to the formation of more stable MgO and Mg(OH)_2;³⁶ however, this effect is not observed here, possibly due to the insufficient thickness and density of the layer formed during the simple immersion. The effect of the plasma treatment is lower on the freestanding samples; however, a decrease in the corrosion rate is still visible [Fig. 3(c)]. Additional to the change of surface structure, the samples on substrate also had a flatter surface than freestanding thin films which were only attached to Si chips, thus, ensuring a more homogeneous treatment. Thus, an optimization of the treatment and sample fixation for freestanding thin films could improve the degradation rate decrease further.

The corrosion studies prove that only very short DBD plasma treatments of Mg and MgLi are required to significantly influence the corrosion of the films. Since the treatment mainly influences the surface of the films, a passivation and formation of a protective film is expected to be the reason for the change in the corrosion rate and is, therefore, analyzed further.

The degradation of Mg-based alloys such as MgLi is highly influenced by surface films formed before or during the corrosion. Those films hinder further ion release or reaction on the surface of the samples. For Mg and MgLi, layers formed in air and during corrosion including many components such as MgO, Mg(OH)₂, MgCO₃, LiO₂, LiOH, or Li₂CO₃ are discussed to influence the corrosion rate.^54–58 

Surface images and cross sections of 15 s plasma treated samples show the formation of a thin surface film during the treatment as shown in Fig. 4. A more homogeneous surface with reduced roughness in comparison to untreated samples is visible after the DBD treatment. The layer formed during the treatment which can be identified by a structural difference from the main sample in Fig. 4(a) has a thickness in the range of 50−100 nm. Since the formed surface film leads to a flatter surface, it reduces the surface area for corrosion and prevents the start of corrosion at preformed pits such as, e.g., the edges of the surface structures shown in Figs. 2(b) and 2(d). A homogeneous corrosion with less pitting and a smaller active surface can significantly influence the corrosion rate.^59–61 The film can only form by reactions of the elements of the film (Mg and Li) with elements available in air since no additional gases are added. EDX studies show an increase in oxygen and carbon components measured on the surface in comparison to untreated samples, especially on the surface [approximately 3%−4%(m/m) C and 5−5.5%(m/m) O], thus, the film could consist of components such carbonates, oxides, or hydroxides. However, since the film has only a thickness in the nm range, EDX analysis cannot give further insights. Therefore, XPS analysis was carried out on a treated and untreated MgLi sample to identify the components formed during the treatment. The wide spectra and Mg 2p, Li 1 s and C 1 s, O 1 s spectra are shown in Fig. 5. See the supplementary material Table SI for the atomic compositions.

While for the untreated sample, a Li signal but only a small Mg is shown in Fig. 5(b), for the treated sample, a significant increase of the Mg 2p signal can be found. Additionally, the carbon spectrum in Fig. 5(c) shows an overall decrease in carbon-containing components. The lower overall peak width and proportionally higher decrease in carboxylate signal in comparison to carbonates could possibly be the result of the removal or a reaction to carbonate in the layer. The formation of carbonates is also found in the O 1 s spectrum in Fig. 5(d). The lower carbon signal and formation of carbonates could be a result of a cleaning effect, reaction of other compounds to carbonates and carbonate formation during the treatment.^42,46,47

Different components were discussed in the literature to lead to a protective layer during corrosion of Mg or MgLi.^54–58 For MgLi, Li₂CO₃ formed during DBD treatment or during the corrosion was discussed to influence the corrosion rate for Li alloys with high Li fraction [14.2% (m/m)].⁴² Those layers could also be formed for material with low Li fraction.⁶³ Li₂CO₃ can be formed in air, leading to a higher Li fraction on the sample surface and a Mg deficiency in the surface layer in comparison to the bulk material on samples without treatment.⁵⁵ This could lead to the low Mg signal measured by XPS on the untreated sample. Apparently, after the treatment, the surface layer formed does not consist of Li₂CO₃ since the Mg signal is stronger and a decrease in Li is found by XPS. Thus, Li₂CO₃ is not the main important component for the increased corrosion resistance for the MgLi thin films studied. This is also confirmed by the corrosion rate decrease for samples of Mg without Li addition.

Since the surface layer is formed in air, possible elements included could be oxygen, carbon, or nitrogen. For the corrosion of Mg, the formation of MgO and further reaction to Mg(OH)₂ occurs. While the layer of MgO is dense but due to a Poisson–Bedworth ratio <1 unstable, the further reaction to Mg(OH)₂ when in contact with water leads to a porous layer⁶⁴ which can only slightly hinder corrosion. The stability of MgO could be increased by Li doping;⁵⁶ however, since there is no or only 1.6%(m/m) Li included, no stability can be assumed. The wide spectrum does not show a significant nitrogen signal and, thus, no components including N can be directly assigned as part of the protective film. The increase of the carbonate signal and the presence of more Mg on the surface of treated samples in comparison to untreated samples could hint to a formation of MgCO₃. MgCO₃ is stable and can reduce the corrosion rate of Mg^57,65–67 if the layer is dense.

The lower corrosion rate is, thus, a result of a change in surface structure and a formation of a protective layer, possibly including MgCO₃. However, as known for the layer formation on Mg and MgLi in air and in corrosion, the protective layers formed are often a system of different layers including different components^55,57,58 and cannot be directly identified as one component only. The combination of a decrease in roughness and a dense layer of stable carbonates and oxides leads to a slower degradation not only for the beginning of the immersion but also leads to an increased corrosion resistance after longer immersion time.

This study presents a fast and simple method to reduce the corrosion rate of Mg-based thin films using dielectric barrier discharge plasma treatment. hcp Mg and MgLi thin films were prepared by magnetron sputtering, leading to different surface roughness and microstructures. After 15 s of DBD plasma treatment, the corrosion rate of all as-sputtered samples was reduced by around 45%–50%, even when the sample was immersed for an hour before the potentiodynamic polarization measurement. Thus, a stable protective layer is formed during the treatment which is also denser and more stable than the passivating films formed during the corrosion. Analysis of the surface by SEM/EDX and XPS shows the removal of loose contaminations and a layer formed on the surface which can influence the corrosion rate by leading to a smoother surface and components including magnesium carbonate preventing fast corrosion. The treatment was also tested on freestanding thin films to confirm the effect of the treatment on samples for the possible application of ion release as therapeutical treatment.

The atmospheric DBD plasma treatment presents a cheap and easily industrially scalable way to prevent the burst release of Mg thin films. The effect reaching a plateau after 5 s also makes the method reproducible, as the treatment effect saturates quickly.

This work was supported by the DFG in the framework of the research training group 2154—Materials for Brain (Project No. 270394294).

The authors have no conflicts to disclose.

Lisa Hanke: Conceptualization (equal); Formal analysis (equal); Investigation (equal); Writing – original draft (equal); Writing – review & editing (equal). Torge Hartig: Conceptualization (equal); Investigation (supporting); Methodology (lead); Writing – original draft (supporting); Writing – review & editing (supporting). Felix Weisheit: Formal analysis (supporting); Investigation (equal); Writing – review & editing (supporting). Tim Tjardts: Formal analysis (supporting); Investigation (supporting). Tim Pogoda: Investigation (supporting); Methodology (supporting). Franz Faupel: Funding acquisition (equal); Supervision (equal). Eckhard Quandt: Funding acquisition (equal); Supervision (equal); Writing – review & editing (supporting).

The data that support the findings of this study are available from the corresponding author upon reasonable request.