*Article*

## **The Influence of Microstructure on the Passive Layer Chemistry and Corrosion Resistance for Some Titanium-Based Alloys**

**Nader El-Bagoury 1,2, Sameh I. Ahmed 2,3, Ola Ahmed Abu Ali 1, Shimaa El-Hadad <sup>4</sup> , Ahmed M. Fallatah 1, G. A. M. Mersal 1,5, Mohamed M. Ibrahim 1,6, Joanna Wysocka 7, Jacek Ryl 7,\*, Rabah Boukherroub <sup>8</sup> and Mohammed A. Amin 1,9,\***


Received: 5 March 2019; Accepted: 9 April 2019; Published: 15 April 2019

**Abstract:** The effect of microstructure and chemistry on the kinetics of passive layer growth and passivity breakdown of some Ti-based alloys, namely Ti-6Al-4V, Ti-6Al-7Nb and TC21 alloys, was studied. The rate of pitting corrosion was evaluated using cyclic polarization measurements. Chronoamperometry was applied to assess the passive layer growth kinetics and breakdown. Microstructure influence on the uniform corrosion rate of these alloys was also investigated employing dynamic electrochemical impedance spectroscopy (DEIS). Corrosion studies were performed in 0.9% NaCl solution at 37 ◦C, and the obtained results were compared with ultrapure Ti (99.99%). The different phases of the microstructure were characterized by X-ray diffraction (XRD) and scanning electron microscopy (SEM). Chemical composition and chemistry of the corroded surfaces were studied using X-ray photoelectron spectroscopy (XPS) analysis. For all studied alloys, the microstructure consisted of α matrix, which was strengthened by β phase. The highest and the lowest values of the β phase's volume fraction were recorded for TC21 and Ti-Al-Nb alloys, respectively. The susceptibility of the investigated alloys toward pitting corrosion was enhanced following the sequence: Ti-6Al-7Nb < Ti-6Al-4V << TC21. Ti-6Al-7Nb alloy recorded the lowest pitting corrosion resistance (*R*pit) among studied alloys, approaching that of pure Ti. The obvious changes in the microstructure of these alloys, together with XPS findings, were adopted to interpret the pronounced variation in the corrosion behavior of these materials.

**Keywords:** titanium-based alloys; microstructure; passivity breakdown; pitting corrosion

#### **1. Introduction**

Titanium and its alloys are widely used in many industrial applications, because of their highly desirable properties, including very good mechanical properties, excellent corrosion and erosion resistance, and favorable strength to weight ratio [1]. In fact, titanium and its alloys have experienced increased use in the past years as biomaterials, because of their superior biocompatibility, high resistance to localized and generalized corrosion, and their good mechanical properties (fatigue resistance) [2]. Among all titanium and its alloys, the commonly used materials in biomedical area are commercially pure titanium (cpTi) and its (α + β) Ti6-Al4-V alloy [3–5].

Next to biomedical applications, aerospace sector has dominated titanium use, instead of heavy steel components, in the fabrication of crucial and decisive systems such as airfoils and airframes [6–9]. About 50% of titanium used in the aerospace industry is the (α + β) Ti-6Al-4V alloy. This alloy possesses a perfect combination of operational and technological properties [10,11]. Titanium alloys have also found widespread applications in a variety of fields such as in chemical and petrochemical sectors due to their excellent corrosion resistance [12]. The outstanding characteristics (such as high specific strength, high fatigue strength, good corrosion resistance, etc.) of the titanium alloys (particularly Ti-6Al-4V) are attributed to a very stable native oxide film (1.5–10 nm) formed on the Ti and Ti-alloy surface upon exposure to atmosphere and/or aqueous environments [13,14]. However, this thin oxide layer can be damaged and thus strongly impacts the bioactivity and other characteristics of the material. To improve the performance of Ti and Ti-alloys for biomedical and aerospace applications, oxidation (anodization) has been applied as a successful approach to improve the material properties [15,16].

The microstructure, formed during various processing methods, is found to greatly influence the mechanical properties of titanium alloys [17]. The microstructure type (bimodal, lamellar and equiaxed) affects the mechanical properties of Ti based alloys [18]. Even though the corrosion of Ti-alloys in different environments has previously beenstudied [19–23], to the best of our knowledge, the literature contains no reports onthe passive layer growth kinetics and breakdown, and subsequent initiation and propagation of pitting corrosion over the surfaces of Ti-6Al-7Nb, Ti-6Al-4V, and TC21 alloys. In this context, the mainobjective of our studywas to assess the effect of microstructure changes of tested alloys on their surface morphology and chemistry using different techniques such as scanning electron microscopy with electron dispersive X-ray spectroscopy (SEM/EDX), X-ray diffraction (XRD), and X-ray Photoelectron Spectroscopy (XPS). The influence of microstructure changes on the anodic behavior and passive layer growth kinetics and breakdown was also investigated. The corrosion resistances were compared using potentiodynamic polarization and impedance spectroscopy tools. All measurements were conducted in 0.9% NaCl solution at 37 ◦C.

#### **2. Materials and Methods**

The working electrodes investigated in this study consisted of three Ti-based alloys, namely Ti-6Al-4V, Ti-6Al-7Nb and TC21; their chemical compositions are presented in Table 1. The as-received titanium alloy samples were prepared by melting in a 500 kg vacuum induction furnace to obtain billets. These billets were then forged and machined into 10 mm diameter bars. The microstructures of these alloys werestudied by Meiji optical microscope (Meiji Techno Co., Ltd., Chikumazawa, Japan) fitted with a digital camera (Meiji Techno Co., Ltd., Chikumazawa, Japan). JEOL JSM5410 and Hitachi S-3400N scanning electron microscopes (LxRay Co., Ltd., Hyogo, Japan) (SEM) were also used for microstructure studies. For this purpose, the specimens were prepared following ASTM E3-11 standard metallographic procedures, and then etched in a mixture of 5 mL HNO3, 10 mL HF and 85 mL H2O. The alloys were machined in the form of rods to perform electrochemical measurements. These rods were mounted in a polyester resin offering an active cross-sectional area of ~0.2 cm2. Prior to conducting any electrochemical analysis, the surface of the working electrode was cleaned and polished using a silicon carbide paper (600-grit) installed on a polishing machine (Minitech 233). The surface was then washed copiouslywith distilled water and rinsed with absolute ethanol (SIGMA-ALDRICH, Steinheim, Germany).


**Table 1.** Chemical composition of investigated Ti alloys.

Electrochemical measurements were conducted in a standard, double-walled electrochemical cell (Princeton Applied Research, USA) with an inner volume capacity of 200 mL. Temperature of the test solution was maintained constant at the desired value by means of a temperature-controlled water bath (FP40-MA Refrigerated/Heating Circulator) (JULABO GmbH, Seelbach, Germany). The water, after being adjusted to 37 ± 0.1 ◦C, wasallowed to circulate through external jacket of the cell. The cover of the electrochemical cell had five openings with different sizes. Such openings were designed to be fitted to the working electrode, counter electrode (a long, coiled platinum wire), reference electrode (KCl-saturated calomel electrode (SCE)), a thermometer and a gas inlet/outlet for gas release. The reference electrode was placed in a Luggin capillary, the tip of which was adjusted to be close to the working electrode to minimize iR drop. The cell was connected to a Potentiostat (Autolab PGSTAT30) (Metrohm, Herisau, Switzerland). The test solution was a normal saline (0.9% NaCl). A Millipore Milli-Q water system (Merck Millipore, MA, USA) (18.2 MΩ cm) was used to freshly prepare the saline solution. The salt was of analytical grade and purchased from Sigma-Aldrich (Steinheim, Germany).

Linear sweep voltammetry (LSV), Tafel plots, and Electrochemical Impedance Spectroscopy (EIS) techniques were applied to investigate the uniform corrosion characteristics of the studied alloys. The susceptibility of these alloys to passivity breakdown was evaluated via conducting potentiodynamic polarization measurements. Uniform corrosion measurements were started by stabilizing the working electrode at the rest potential for 2 h, followed by conducting EIS measurements at the respective corrosion potential (*Ecorr*) every day for a week of exposure in 0.9% NaCl solution at 37 ◦C, covering a wide frequency range (100 kHz–10 mHz), with 15 mV perturbation amplitude. Uniform corrosion study was assessed by constructing Tafel plots via sweeping the electrode potential around the Tafel potential (*E* = *E*corr <sup>±</sup> 250 mV), applying a sweep rate of 1.0 mV s−1. After that, the electrode was removed from the cell (which was cleaned properly and re-filled with a new fresh test solution), cleaned and polished up to the mirror finish, as described above, and then inserted in the cell for cyclic polarization measurements. Chronoamperometry (CA) technique was also applied using a new set of cleaned and polished electrodes submerged in a cleaned cell filled with a new fresh solution.

Prior to performing cyclic polarization measurements, the working electrode was allowed to stabilize at rest potential for 2 h, then swept linearly, at a sweep rate of 1.0 mV s−1, starting from a cathodic potential of −2.0 V to +8.0 V vs.SCE. The potential sweep was then reversed back with the same sweep rate to reach the start point again, thus forming one complete cycle. To conduct chronoamperometry (current vs. time) measurements, a two-step route was applied. The working electrode was first held at a starting cathodic potential of −2.0 V vs. SCE for 60 s, and then polarized towards the anodic direction at a sweep rate of 1.0 mV s−<sup>1</sup> untilthe required anodic potential (*E*a). Finally, the anodic current was measured versus time (5.0 min) by holding the working electrode at *E*a. To ensure reproducibility, each run was repeated at least three times, where mean values of the various electrochemical parameters and their standard deviations were calculated and reported.

The XRD diffraction patterns were collected for the bulk samples using a SmartLab SE (Rigaku Americas Corporation, Oxford, MS, USA) X-ray diffractometer with Cu Kα (λ = 1.54056 Å) operated at 40 kV and 40 mA. The scanning speed was 0.2◦/min and the scanning angle ranged from 20◦ to 100◦ in *2*θ. Energy dispersive X-ray spectroscopy (EDS) measurements were utilized to determine microstructural composition of investigated alloys as well asevaluate changes in chemical composition as a result of exposure to corrosive media. S-3400N SEM (Hitachi, Tokyo, Japan) was equipped with an UltraDry detector from ThermoFisher Scientific (Waltham, MA, USA). High-resolution X-ray photoelectron spectroscopy (XPS) studies were carried out on an Escalab 250 Xi from Thermofisher Scientific (Waltham, MA, USA), equipped with Al Kα source. Pass energy was 20 eV and the spot size diameter was 650 μm. Charge compensation was controlled through the low-energy electron and low energy Ar<sup>+</sup> ions emission by means of a flood gun (emission current: 150 μA, beam voltage: 2.1 V, filament current: 3.5 A). Avantage software (Thermofisher Scientific, Waltham, MA, USA) was used for deconvolution purposes.

#### **3. Results and Discussion**

#### *3.1. Microstructure Investigation*

Based on the morphology of α phase, the microstructure of titanium alloys can be classified into equiaxed, lamellar and bi-modal microstructures [24]. The microstructure of Ti-based alloys can be controlled based on their chemical composition, i.e., based on the balance between the α phase stabilizing elements, such as Al, Sn and O, and the forming β phase elements such asV, Mo and Nb [25]. As shown in Figure 1, the microstructure of all studied titanium alloys consisted of bimodal structure of α/β phases. The initial microstructure of Ti-Al-V and Ti-Al-Nb alloys in as-received (forged) state was represented by equiaxed grains of primary αphase (dark), as well as β-transformed structure (light), as shown in Figure 1. The β phase formed in the microstructure of both alloys was globular in shape, but seemed larger in size in Ti-Al-V alloy than in Ti-Al-V alloy. The particle size of β phase in Al-Ti-V alloy was about 0.5–1.5 μm, while its size in Ti-Al-Nb alloy was slightly lower (about 0.25–1 μm), as depicted in Figure 1a,b.

**Figure 1.** Microstructure of the three investigated Ti alloys: (**a**) Ti-Al-V; (**b**) Ti-Al-Nb; and (**c**) TC21.

Similar to Ti-Al-V and Ti-Al-Nb alloys, the microstructure of TC21 alloy (Figure 1c) contained α and β phases, but displayed different morphologies and volume fractions. The TC21 alloy's β phase consisted of two shapes:an acicular-like structure (Figure 2a) and a blocky shape (Figure 2b). The thickness of the acicular β phase in the TC21 alloy's microstructure ranged from around 0.2 to 0.6 μm, while the size extent of the blocky β phase was about 0.75–1.5 μm. Moreover, the volume fraction of β phase in the microstructure of TC21 alloy was higher than that in the Ti-Al-V and Ti-Al-Nb alloys' microstructure, as depicted in Figure 1.

**Figure 2.** Morphology of β phase in TC21 alloy: (**a**) acicular-like structure; and (**b**) blocky shaped structure.

Table 2 illustrates the volume fraction of α and β phases in the microstructure of the studied titanium alloys. The microstructure of pure Ti has the highest volume fraction of the α phase (~100%) and the lowest volume fraction of β phase (~0.0%). The presence of Al (α-phase stabilizer) and V (β-phase stabilizer) as alloying elements in the chemical composition of Ti-Al-V alloy influenced the volume fraction of α and β phases. The values of the volume fractions of α and β phases (Table 2) in the microstructure of Ti-Al-V alloy were 65% and 35%, respectively. Replacing V with Nb, yielding Ti-Al-Nb alloy, resulted in an obvious enhancement in the volume fraction of α phase (increased to 77%) at the expense of that of the β phase, which decreased to 23%, as shown in Table 2. The volume fraction of both phases in the microstructure of TC21 alloy wasalso altered, probably due to the mutual combination of the alloying elements of that alloy (cf. Table 1). The volume fractions of α and β phases in the microstructure of TC21 alloy recorded almost equal values: 48% for α phase and 52% for β phase (Table 2).


**Table 2.** Volume fraction of α and β phases in the investigated Ti based alloys.

To further assess the influence of chemical composition on the microstructure and volume fraction of α and β phases, [Al]eq and [Mo]eq were calculated, where [Al]eq and [Mo]eq represent the alloying elements from α and β phases, respectively [5,26]. Table 3 illustrates the calculated values of [Al]eq and [Mo]eq for the tested Ti-based alloys, following Equations (1) and (2) [5,26].

$$[Al]\_{eq.} = [Al] + 0.33[Sn] + 0.17[Zr] + 10[O + \text{C} + 2\text{N}]\tag{1}$$

[*Mo*]*eq*. = [*Mo*] + 0.2[*Ta*] + 0.28[*Nb*] + 0.4[*W*] + 0.67[*V*] + 1.25[*Cr*] + 1.25[*Ni*] + 1.7[*Mn*] + 1.7[*Co*] + 2.5[*Fe*] (2)

**Table 3.** and [Mo]eq for the investigated alloys [26,27].


As shown in Table 3, Ti-6Al-7Nb and TC21 alloys recorded the highest value (8.59) of [Al]eq, while the lowest value (8.05) was measured for the Ti-6Al-4V alloy. Additionally, Ti-6Al-7Nb alloy achieved the maximum value of [Mo]eq, 3.94, whileTC21 alloy recorded 1.71. Table 3 also depicts the ratio [Al]eq/[Mo]eq for the tested alloys. Ti-6Al-7Nb alloy displayed the maximum ratio, 3.94, while a minimum ratio of 1.71 was determined for the TC21 alloy. The results shown in Table 3 agree well with those in Table 2. The calculated ([Al]eq/[Mo]eq) and (α/β) ratios weremaximum in case of Ti-6Al-7Nb alloy, and minimum for the TC21 alloy.

The chemical composition of both phases in all microstructures of the investigated alloys was analyzed using the EDS unit attached to SEM. The averaged chemical composition (wt. %) for each investigated alloy is displayed in Tables 4 and 5. It was evident that the Ti, Al, Sn and Zr elements tended to more segregate to α phase than to β phase [27]. However, V, Nb, Cr and Mo were β forming elements [28], meaning that higher ratios of these elements werefound in β phase rather than in α phase.The detailed EDS linescan/map analyses are discussed in the Supplementary Materials (Figures S1–S3).


**Table 4.** Chemical composition (wt %) of different phases in Ti-6Al-4V and Ti-6Al-7Nb alloys.

**Table 5.** Chemical composition (at %) of different phases in TC21 alloy.


#### *3.2. X-ray Di*ff*raction Studies*

Phase identification was performed by X-ray diffraction (XRD) patterns to define the phases comprising each alloy sample. The diffraction patterns recorded for the studied alloys are compared in Figure 3. The phases were identified by matching the characteristic peaks with the Joint Committee on Powder Diffraction Standards (JCPDS) files [29]. The phases α-Ti (JCPDS#00-044-1294) andβ-Ti (JCPDS#00-044-1288) were common and dominated the composition of the three studied alloys. The Ti-Al-V and TC21 alloys were found to contain solely α-Ti and β-Ti phases, respectively. On the other hand, Ti-Al-Nb alloy contained some Ti and Nb oxides, TiO (JCPDS#00-008-0117) and Nb6O (JCPDS#00-015-0258).

**Figure 3.** XRD diffraction patterns recorded for the Ti-6Al-4V, Ti-6Al-7Nb and TC21 samples.

An effective procedure for the simultaneous refinement of structural and microstructural parameters based on the integration of Fourier analysis for broadened peaks in the Rietveld method was first proposed by Lutterutti et al. [30] and is implemented in the Maud program [31]. Consequently, weight percent (wt.%), lattice parameters, isotropic crystallite size (D) and r.m.smicrostrain (με) were regarded as fitting parameters in the Rietveld adjustments and refined simultaneously. The structural information for all the refined phases was obtained from the Inorganic Crystal Structure Database (ICSD) [32]. The results obtained for the structural and microstructural analysis are summarized in Table 6 for all alloys. It is worth mentioning that all studied alloys were characterized with considerable degree of preferred orientation, which strongly modified the relative intensities of the Bragg reflections, especially for α-Ti and β-Ti phases. The MAUD program also incorporates correction for preferred orientation [33,34] in the Rietveld adjustments to obtain the best fitting parameters.


**Table 6.** The structural and microstructural parameters of the three alloys obtained by the Rietveld adjustment of the XRD patterns. Wt. % is the weight percentage of each phase, a and c are the cell parameters of the crystal lattice, D is the average crystallite-size and ε is the lattice microstrain.

The calculated diffraction patterns from the Rietveld adjustment are plotted with the observed ones for the three alloys in Figure 4. The average R-values obtained for the refinements were about Rwp (%) = 24–27 and Rb (%) = 15–20. The simultaneous refinements of both structural and microstructural parameters produced good matching of the calculated to observed profiles of diffracted intensities. In addition, the incorporation of the preferred orientation models enabled accounting for the variations of the peak intensities of α and β-Ti phases.

In the Rietveld adjustment of the Ti-6Al-4V alloy, the hcp α-Ti (Space group *P*63/*mmc*) together with the bcc β-Ti (Space group *Im*3*m*) dominated the composition of the alloy. In the second alloy, Ti-6Al-7Nb, the formation of some TiO (Space group *Fm*3*m*) and Nb6O (Space group *P*42*cm*) was observed and they formed larger crystallites than those formed in the Ti phases. The total weight percent of those oxide phases was less than 10% (Table 6). For the third alloy, TC21, only α- and β-Ti phases were observed in the XRD patterns. No oxide phases were detected due to the slight oxygen content of this alloy. Nevertheless, there weresome mismatches between the wt% values obtained from the Rietveld adjustments and the corresponding wt% values determined with other techniques. This was attributed to the behavior of the preferred orientation of the α-Ti phase observed for the reflection (100), which was relatively stronger for the Ti-6Al-4V and Ti-6Al-7Nb alloys than in the TC21 alloy.

As shown in Table 6, the last two alloys, Ti-6Al-7Nb and TC21, contained relatively higher portions of β-Ti than α-Ti in contrast to the first alloy, Ti-6Al-4V, which hadα-Ti content higher than β-Ti. As known from the literature, Al is an α-stabilizing while V, Nb, Mo and Fe are β-stabilizing elements. Nevertheless, the results indicate that Nb, Mo and Fe hadstronger capabilities to stabilize β-Ti phase than V. These findings corroborate microstructural studies (cf. Section 3.1).

**Figure 4.** The calculated (red line) and recorded (black dots) diffraction patterns for the three alloys as obtained from the Rietveld adjustments using the MAUD program; the positions of the Bragg reflections of each phase and the difference between the calculated and observed patterns are also presented at the bottom.

#### *3.3. Electrochemical Measurements*

#### 3.3.1. Uniform Corrosion Studies

Figure 5 illustrates the cathodic and anodic polarization curves for the studied alloys in comparison with pure Ti, after seven days of immersion in 0.9% NaCl solution at 37 ◦C. As shown in Figure 5, among the studied alloys, TC21 alloy exhibited the lowest cathodic and anodic overpotentials, corresponding to enhanced corrosion rate. On the contrary, Ti-6Al-7Nb alloy displayed the highest overpotentials, close to that of Ti, for both the cathodic and anodic processes, thus referring to its highest corrosion resistance.

**Figure 5.** Cathodic and anodic polarization curves recorded for the three tested alloys in comparison with pure Ti, after seven days of exposure in 0.9% NaCl solution at a scan rate of 0.5 mV s−<sup>1</sup> at 37 ◦C.

Indeed, Tafel slopes and respective calculated uniform corrosion rates for a metal covered with a semi-conductive passive film raisedstrong doubts, despite the utilization of the Tafel extrapolation method to determine corrosion rates for Ti and some Ti-based alloys [35,36]. In addition, in Figure 5, the polarization curves do not display the expected log/linear Tafel behavior. This was clear in both the anodic and cathodic branches of alloys Ti-6Al-4V and Ti-6Al-7Nb and as well as in the anodic branches of Ti and TC21 alloy, which exhibited some sort of curvature over the complete applied potential range. This in turn made evaluation of Tafel slopes by Tafel extrapolation method, and hence corrosion rates, inaccurate [37–39]. There is, therefore, an uncertainty and source of error in the numerical values of Tafel slopes (β<sup>a</sup> and βc), and possibly in the values of jcorr.

The curvature of the anodic branch mightbe attributed to the deposition of the corrosion products and/or passive film formation, as evidenced from XPS studies. With respect to the cathodic branch, since the solution was stationary, diffusion of the electrochemically active species was slow, and concentration polarization couldact to shorten the cathodic linear Tafel region. In the extreme case, linearity mightdisappear altogether, with the cathodic reaction proceeding under combined activation and diffusion control at Ecorr [37–39]. This counteracted the validity of the Tafel extrapolation method for measuring uniform corrosion rates, which was successfully applied for the charge transfer controlled processes.

EIS measurements were also conducted at the respective Ecorr during sample's exposure in 0.9% NaCl solution at 37 ◦C to confirm the polarization data and to assess the kinetics of the uniform corrosion process on alloy surface. The measurements were carried day-by-day allowing for monitoring of uniform corrosion susceptibility [40–42]. Figure 6 displays the impedance plots in Bode projection, recorded for the studied alloys during the second and the last day of exposure. Pure Ti (99.99%) was also included for comparison (see Figure 6a). The impedance spectra recorded on Day 1 (after initial 120 min of conditioning) were highly scattered due to non-stationary conditions at the metal/electrolyte interface, which is a common problem in EIS measurements. This issue became negligible after a few hours of exposure. For this reason, results recorded on Days 2–7 were considered for further analysis. All the impedance plots exhibited a single time constant (capacitive loop). The overall corrosion resistance of each investigated alloy was very high; the impedance modulus |Z| linearly increasedwith frequency decrease, reaching over 105 at 0.1 Hz. For each studied alloy, inclination of the phase angle θ was shifted towards lower frequencies on Day 7 of exposure, testifying tothe decreased corrosion process kinetics.

An electric equivalent circuit (EEC) was proposed to analyze the impedance results. When defining the adequate EEC, one must consider whether to include the space charge layer resulting from the semi-conductive nature of titanium oxides. In his studies, Blackwood concluded that thickness of the space charge layer is considerably less than the oxide film itself in the open circuit conditions [43]. On the other hand, its dominant influence was observed under anodic polarization conditions [43–46]. The impedance measurements of titanium oxide films investigated within this manuscript were studied under open circuit conditions, thus the parallel resistance and CPE represented primarily the dielectric properties of the passive layer. Due to absence of additional time constants in the analyzed frequency rage, a simple Randles circuit was proposed with constant phase element (CPE) selected instead of capacitance to account for the dispersion of the time-constant. The parallel resistance represents the sum of charge-transfer limiting effects through the metal/electrolyte interface, dominated by the passive layer resistance RF [47]. The aforementioned time-constant dispersion originated from thepresence of the charge space layer in the semi-conductive film, thesurface distribution of the time-constant due to the geometric heterogeneity (pits, scratches, and porosity), and diversified surface electric properties due to adsorption processes of passive layer breakdown [48].

**Figure 6.** Bode impedance plots on Days 2 and 7 of exposure, recorded for: (**a**) pure Ti; and three tested alloys: (**b**) TC21; (**c**) Ti-6Al-4V; and (**d**) Ti-6Al-7Nb. Studies performed at Ecorr in 0.9%NaCl solution at 37 ◦C. Points represent experimental results while the solid line was calculated based on R(QR) EEC.

The CPE impedance ZCPE = (Q(jω) n) <sup>−</sup><sup>1</sup> represents a capacitor with capacitance 1/Q for a homogeneous surface n→ 1. Thus, it is often believed that CPE component n is the heterogeneity factor and its variation can be monitored. CPE describes quasi-capacitive behavior of the passive layer. The effective capacitance Ceff can be calculated based on CPE using Hirschorn's model for surface distribution of time constants [49]. The EEC can be schematically written as RS(QRF), where RS is electrolyte resistance. The aforementioned single time-constant EEC covers all the applied frequency range. The fitting quality is represented by solid lines in Figure 6a–d.

Figure 7 depicts the electric parameters obtained on the base of RS(QRF) EEC and their changes during the one-week exposure. The higher was the RF, the lower was the corrosion current density, offering an easy comparison in uniform corrosion resistance of investigated alloys (see Figure 7a). Each investigated alloy was characterized with very high resistance, order of MΩ, owing to a presence of a passive layer tightly covering the metal surface. Nevertheless, for TC21 alloy, RFvalue wasone order of magnitude lower and slowly but consistently decreased throughout the exposure, revealing its lower corrosion resistance and corroborating DC electrochemical studies.

The analysis of constant phase element (CPE) alloweddrawing conclusions onthe passive layer homogeneity. The effective capacitance Ceff, calculated using the surface distribution model, was similar for allinvestigated alloys, falling in a range between 2 and 5 μF. The differentiation may result from differences in passive layer thickness d and to some extent from relative permittivity of alloying additives and their oxides ε<sup>r</sup> according to: C = ε0εrd/A, where ε<sup>0</sup> is the absolute permittivity and A is the electrochemically active surface area. A steady decrease of Ceff was attributed to an increase of passive layer thickness, denouncing further passivation of metal in investigated electrolytic conditions. The presence of stable corrosion pits would be visible in theform of rapid increase in Ceff [49,50] (likely observed in TC21 alloy on Day 4).

**Figure 7.** Monitoring of (**a**) passive layer resistance RF, (**b**) effective capacitance Ceff and (**c**) CPE exponent n calculated on the base of RS (QRF) EEC for each investigated alloy. The one-week exposure was carried out in 0.9% NaCl solution at 37 ◦C.

The initial value of CPE exponent n depends on factors such as surface phase distribution and geometric defects remaining after polishing. Its decrease throughout the exposure in corrosive electrolyte reflected the appearance of heterogeneities on analyzed sample surface, which in this case was primarily associated with initial phases of corrosion pits formation (see Figure 8c). This effect was clearly seen in SEM micrographs (see below). Notably, the value of n factor of Ti-6Al-7Nb alloy was both the highest and the least affected by exposure in corrosive media. The aforementioned observation indicated high surface homogeneity, which may be the reason behind outstanding corrosion resistance of this alloy.

**Figure 8.** Cyclic polarization curves recorded for the studied alloys in 0.9% NaCl solutions at a scan rate of 1.0 mV s−<sup>1</sup> at 37 ◦C.

#### 3.3.2. Cyclic Polarization Measurements

Figure 8 shows typical cyclic polarization curves recorded for the studied alloys between −2.0 V and + 8.0 V (SCE). Measurements were conducted in 0.9% NaCl solution at a scan rate of 5.0 mV s−<sup>1</sup> at 37 ◦C.

The polarization curve of TC21 alloy exhibited active dissolution near Ecorr, followed by an obvious enhancement in the anodic current with the applied potential due to thinning and weakening of the passive layer as a result of the aggressive attack of Cl− anions. In addition, Ti-6Al-7Nb and Ti-6Al-4V alloys showed active dissolution near Ecorr, but to a much lower extent than TC21, and, in addition, tended to passivate with a very low current covering a wide range of potential. These findings reflect the weaker passivity of TC21 and its higher tendency to corrode in this solution than Ti-6Al-4V and Ti-6Al-7Nb alloys. On the contrary, as expected, the anodic polarization curve of pure Ti exhibited typical passivity near Ecorr, referring to its high corrosion resistance.

Passivity of the studied alloys persisted up to reaching pitting potential (Epit). Remarkable changes occurred within the passive region at potentials exceeding Epit. These involved a sudden increase in corrosion current density and formation of a hysteresis loop on the reverse potential scan. These events werea clear sign ofpassivity breakdown, and initiation and propagation of pitting corrosion. Repassivation was only achieved when the reverse scan intersected the forward one within the passive region, below which the working electrode was immune to pitting [51].

A current intermission could be seen on the reverse scan of the three tested alloys. This current discontinuity aws quite clear on the reverse scan of the TC21 alloy, and couldbe observed for Ti-6Al-7Nb and Ti-6Al-4V alloys. We previously reported similar findings during pitting corrosion studies of Zn in nitrite solutions [52]. Recently, Zakeri et al. [53] explored the transition potential and the repassivation potential of AISI type 316 stainless steel in chloride containing media devoid of and containing 0.01 M thiosulfate.

Beyond pit transition potential, the rate of anodic dissolution was diffusion-controlled [51–53]. Such a current transient relationship, when satisfied, referred to an anodic diffusion control process [53]. On reversing the potential scan, the thickness of the salt (pitting corrosion product) film diminished. This decrease in salt film thickness enhanced with back scanning untila certain potential was reached, at which the cation concentration decreased below the saturated concentration. At this stage, salt precipitation was stopped, and the remaining metal salt film dissolved, making the bottom of pits free from salt film. This in turn established an ohmic/activation control (a linear decrease of current density with potential) regime.

Ti-6Al-7Nb alloy's passivity seemed stronger and more stable that of the Ti-6Al-4V alloy. The latter was characterized by a higher jpass, which enhanced with potential untilits Epit, which attained ~50 mV vs. SCE before that of the former. In addition, the pits existing on the surface of Ti-6Al-4V alloy were much more difficult to repassivate than those on the surface of Ti-6Al-7Nb alloy, as the hysteresis loop of the former was much larger than that of the later.

Another important pitting corrosion controlling electrochemical parameter is the pitting corrosion resistance Rpit = |Ecorr − Epit|, which defines the resistance against the nucleation of new pits [38]. Referring to Figure 8, it is clear that Rpit increased following the order: TC21 << Ti-6Al-4V < Ti-6Al-7Nb. The resistance against growth of the pits also controlled the susceptibility toward pitting corrosion. A specific routine of the software (Autolab frequency response analyzer (FRA) coupled to an Autolab PGSTAT30 potentiostat/galvanostat with FRA2 module) (Metrohm, Herisau, Switzerland.) was used to calculate the areas of the hysteresis loops, related to the charge consumed during the growth of such already formed pits. Again, the hysteresis loop of the TC21 alloy recorded the highest area (charge consumed) among the studied alloys, while the lowest value of the hysteresis loop charge consumed during was measured for Ti-6Al-7Nb alloy. Thus, the resistance against the growth of pre-existing pits was ranked as: Ti-6Al-7Nb > Ti-6Al-4V >> TC21. These findings mean that replacing V by Nb in Ti-6Al-4V alloy promoted alloy's repassivation, thus enhancing its pitting corrosion resistance.

#### 3.3.3. Chronoamperometry Measurements

Chronoamperometry (*j*/*t*) measurements were also carried out to confirm the above results and gain more information about the influence of alloyed V and Nb on the passive layer growth kinetics and breakdown. Figures 9a and 10b depict the *j*/*t* curves measured for the tested alloys at two different *E*<sup>a</sup> values, far below and close to *E*b. Measurements were conducted in 0.9% NaCl solution at 37 ◦C. The profile of the obtained curves was found to vary according to the chemical composition of the studied alloy and the position of *E*<sup>a</sup> versus *E*pit. When *E*<sup>a</sup> was located far cathodic to *E*pit, a *j*/*t* profile with two stages was obtained, as shown in Figure 9a. During the first stage, the anodic current (*j*a) declined with a rate depending upon chemical composition of the tested alloy, denoting passive layer electroformation and growth [39]. This decay in current then reached a steady-state value (*j*ss), an almost constant passive current related to *j*pass (cf. Figure 8), constituting the secondstage of the current. The constancy of *j*ss originated from a balance between the rates of the passive layer growth (current builds up) and its dissolution (current decays) [54,55].

**Figure 9.** Chronoamperometry (current–time) curves recorded for the studied solder alloys in 0.9% NaCl solution at applied anodic potentials of 2.0 V (**a**) and 4.0 V (**b**) vs. SCE at 37 ◦C: (1) pure Ti; (2) Ti-6Al-7Nb; (3) Ti-6Al-4V; and (4) TC21.

**Figure 10.** SEM micrographs taken in secondary electron mode for each investigated sample: (**a**) pure Ti as a reference; (**b**) TC21 alloy; (**c**) Ti-6Al-4V; and (**d**) Ti-6Al-7Nb at the end of one-week exposure in 0.9% NaCl at 37 ◦C. Magnification: × 500. In the inset, back-scatter electron topography mode images of selected surface defects. Magnification: × 2000.

Further, Figure 9a demonstrates that the rate of *j*<sup>a</sup> decay, and consequently the rate of passive layer growth, diminished upon alloying Ti with V and Nb. These results further confirm the influence of the alloying elements V and Nb, with V being more active than Nb, which, when added to Ti, weakened its passivity viadepassivation (destabilizing the passive oxide film through oxide film thinning/dissolution [56]). This in turn madethe passive film more susceptible to pitting.

At an *E*<sup>a</sup> value very close to *E*pit (Figure 9b), the *j*/*t* curves with three different stages (I–III) were recorded. Similar results were previously obtained in our lab [55,57]. Stage I refersto the passive layer electroformation and growth, as its current fellwith time [54,55,57]. Stage I ended at a certain time (*t*i), the incubation time, where Stage I's current reached its minimum value; *t*<sup>i</sup> is defined as the time the adsorbed aggressive Cl− anions must acquire to locally attack and subsequently remove the passive oxide film [54]. The magnitude of *t*i, more specifically its reciprocal value (1/*t*i), denotes the rate of pit initiation and growth [54,55], and measures the susceptibility of the oxide film to breakdown and initiate pit formation and growth.

Stage II began at *t*<sup>i</sup> and terminated at time τ, and its current aws termed *j*pit (pit growth current density). *j*pit increased from the moment just after *t*<sup>i</sup> and continuedto grow until τ, suggesting that the pit formation and growth dominated over passivation during this stage. Ultimately, *j*pit attaineda steady-state just after the time τ, denoting the onset of Stage III, and remained almost constant until the end of the run. The constancy of Stage III's current was attributed to the hindrance of the current flow (*j*pit) through the pits sealed off by the pitting corrosion products formed during the events of Stage II, namely pit initiation and growth [55,57]. This hindrance in *j*pit was balanced by a current increase due to metal dissolution, thus yielding an overall steady-state current.

Close inspection of Figure 9 reveals that *j*pit increased and *t*<sup>i</sup> shortened, thus referring to accelerated pitting attack, in presence of alloyed V. These results again support the catalytic impact of alloyed V towards pitting corrosion

#### *3.4. Surface Morphology and Composition*

After one-week exposure, the investigated samples were reexamined using SEM to evaluate the susceptibility to pitting corrosion. This procedure was carried out after rinsing in ethanol using ultrasonic cleaner. The results of the analysis are exhibited in Figure 10. Defects started to appear on the surface of each analyzed sample throughout the exposure. The micrographs in the inset of Figure 10 were taken using back-scatter electrons (BSE) in topography mode. This allowed bringing out the geometry of aforementioned defects. As can be seen, each analyzed defect formed a bulge above the alloy's surface, testifying to either repassivation once shallow corrosion pits formed or at an early stage of passive layer degradation. Ti-6Al-4V sample was characterized with both the highest amount and the largest defects, reaching 30 μm in diameter. On the other hand, the surface of pure Ti and Ti-6Al-7Nb appeared the most intact. No real corrosion pits were observed on the surface of either investigated alloy at the end of exposure in 0.9% NaCl solution at 37 ◦C, testifying tothe overall high pitting corrosion resistance.

Nevertheless, the passive layer must have weakened, hence it was possible for corrosion products to adsorb on the metal surface. EDS analysis was carried out on the defects observed on each investigated alloy to qualify their chemical constitution. The exemplary results, obtained for Ti-6Al-7Nb alloy, are summarized in Figure S4 (Supplementary Materials). The chemistry of defects observed for each investigated alloy was similar. The defects were primarily composed of carbon and oxygen, most likely forming metal carbonates typical for early pitting corrosion stages [58]. Small amount of chlorine was also found within defects. Its low amount was distorted by EDS depth of analysis ranging few microns.

The chemistry of the passive layer in each examined case was composed primarily of titanium (IV) oxides, as verified by a strong recorded Ti2p peak doublet, with Ti2p3/<sup>2</sup> component located each time at 458.6 eV [48,59,60] (Figure 11). Furthermore, there was no sign of titanium oxides at lower oxidation states corroborating the aforementioned result (see Supplementary Materials, Figure S5). Besides the titanium, other alloying additives also tookpart in the passivation process. The strongest signal among the alloying additives was recorded for aluminum oxide Al2O3 (Al2p3/<sup>2</sup> peak at 74.5 eV), ranging between 3.5 and 3.8 at.% for each sample [61,62]. The contribution of VO2 (V2p3/<sup>2</sup> at 516.4 eV) in Ti-Al-V and Nb2O5 (Nb3d5/<sup>2</sup> at 207.1 eV) in Ti-Al-Nb alloy did not exceed 0.7 at.% [60,63,64]. The passive film

formed on the surface of TC21 alloy was naturally more complex. Besides TiO2, it wascomposed of Al2O3 (3.8 at.%), Nb2O5 (0.3 at.%), ZrO2 (0.4 at.%, Zr3d5/<sup>2</sup> at 182.4 eV), Cr2O3 (0.8 at.%, Cr2p3/<sup>2</sup> at 576.0 eV), SnO2 (0.1 at.%, Sn3d3/<sup>2</sup> at 486.5 eV), MoO3 and MoO2 (0.2+0.2 at.%, Mo3d5/<sup>2</sup> at 232.9 and 229.2 eV, respectively) [65–68].

**Figure 11.** High-resolution XPS spectra recorded in (**a**) Ti2p, (**b**) Cl2p and (**c**) O1s energy range for each investigated alloy after seven days of exposure to 0.9% NaCl solution at 37 ◦C.

The high-resolution spectra analysis carried out in the Cl2p energy range confirmed the electrochemical and microscopic studies regarding chloride adsorption on the metal surface as a result of seven-day metal exposure to chloride-containing electrolyte. Full chemical analysis is summarized in Table 7. Metal chlorides were found on the surface of each investigated sample, which confirmed metal-chlorine bond formation, shownby a peak doublet: Cl2p3/<sup>2</sup> at 198.9 eV [48,55,69]. Nevertheless, the amount of adsorbed chlorides Was nearly 2.5 times higher for the TC21 alloy than pure titanium. The chloride concentration obtained for highly resistant Ti-6Al-7Nb alloy was nearly on par withTi sample, and slightly smaller than in the case of Ti-6Al-4V. An interesting conclusion couldbe drawn based on O1s peak analysis for each investigated sample. The spectra were conventionally deconvoluted into three components. Two dominant components located at 530.2 and 531.6 eV were ascribed to Me-O and Me-OH species, respectively. The second component intensity mightbe further influenced by the presence of C-O bonds in carbonates. Its formation may result from prolonged electrolyte exposure as well as adventitious carbon deposition due to air exposure [61,64]. The finding regarding carbonate adsorption on the metal surface wasfurther confirmed by a third O1s component at 532.8 eV, typical for C=O bonds but also chemisorbed water molecules. For clarity purposes, the analysis excluded data recorded for carbon C1s, which was found in large amounts, up to 30 at.%, at binding energies corroborating adventitious carbon and carbonates. Importantly, the highest amount of the adsorbed carbonate species was found on Ti-Al-V sample surface, which is in very good agreement with SEM micrographs presented in Figure 10. The least amount of carbonate species was once more found on the surface of Ti sample.


**Table 7.** XPS deconvolution results carried out in Ti2p, Cl2p and O1s energy range after seven days of exposure to 0.9% NaCl solution at 37 ◦C (in at.%).

#### **4. Conclusions**

The effect of microstructure on the uniform and pitting corrosion characteristics of Ti-Al-V, and Ti-Al-Nb alloys were studied. Pure Ti and TC21 alloy were included for comparison. Measurements were conducted in 0.9% NaCl solution at 37 ◦C employing various electrochemical techniques, and complemented with XRD and SEM/EDS analysis. The obtained results reveal that:


**Supplementary Materials:** The following are available online at http://www.mdpi.com/1996-1944/12/8/1233/s1. Figure S1—SEM image (a) and EDS spectrum (b) of phase in Ti-Al-V alloy; Figure S2—Line analysis of β phase in TiAlNb alloy; Figure S3—Microstructure of the Ti-Al-Nb alloy (a), and mapping of Ti (b), Al (c), and Nb (d) alloying elements. Figure S4—(a) SEM micrograph with marked areas for EDS analysis, (b) EDS examination at defect and at the surrounding, not corroded area. Figure S5—High-resolution Ti2p XPS spectra with the quality fit (light blue line) for each studied alloy: (a) pure Ti reference, (b) TC21, (c) Ti-6Al-4V, (d) Ti-6Al-7Nb.

**Author Contributions:** Conceptualization, M.A.A. and R.B.; resources, N.E.-B. and S.E.-H.; investigation, S.I.A. and O.A.A.A, (XRD), A.M.F. and M.M.I. (uniform corrosion), M.A.A. and G.A.M.M. (passive layer breakdown), N.E.-B., J.W. and S.E.-H. (microstructure), J.R. (XPS), and J.R. and J.W. (EIS); writing—original draft preparation, all authors; writing—review and editing, J.R., M.A.A. and R.B.; and project administration, M.A.A.

**Funding:** This study was funded by the Deanship of Scientific Research, Taif University, Saudi Arabia (Project No. 1-439-6070).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Relationship between Phase Occurrence, Chemical Composition, and Corrosion Behavior of as-Solidified Al–Pd–Co Alloys**

**Marián Palcut, Libor Duriška, Ivona ˇ Cerniˇ ˇ cková, Sandra Brunovská, Žaneta Gerhátová, Martin Sahul, L'ubomír Caploviˇ ˇ c and Jozef Janovec \***

Faculty of Materials Science and Technology in Trnava, Slovak University of Technology in Bratislava, J. Bottu 25, 917 24 Trnava, Slovakia; marian.palcut@stuba.sk (M.P.); libor.duriska@stuba.sk (L.D.); ˇ ivona.cernickova@stuba.sk (I.C.); brunovskas@gmail.com (S.B.); zaneta.gerhatova@stuba.sk (Ž.G.); ˇ marian.sahul@stuba.sk (M.S.); lubomir.caplovic@stuba.sk (L'.C.) ˇ **\*** Correspondence: jozef.janovec@stuba.sk; Tel.: +421-918-646-072

Received: 18 April 2019; Accepted: 20 May 2019; Published: 22 May 2019

**Abstract:** The microstructure, phase constitution, and corrosion performance of as-solidified Al70Pd25Co5 and Al74Pd12Co14 alloys (element concentrations in at.%) have been investigated in the present work. The alloys were prepared by arc-melting of Al, Pd, and Co lumps in argon. The Al74Pd12Co14 alloy was composed of structurally complex ε<sup>n</sup> phase, while the Al70Pd25Co5 alloy was composed of ε<sup>n</sup> and δ phases. The corrosion performance was studied by open circuit potential measurements and potentiodynamic polarization in aqueous NaCl solution (3.5 wt.%). Marked open circuit potential oscillations of the Al70Pd25Co5 alloy have been observed, indicating individual breakdown and re-passivation events on the sample surface. A preferential corrosion attack of ε<sup>n</sup> was found, while the binary δ phase (Al3Pd2) remained free of corrosion. A de-alloying of Al from ε<sup>n</sup> and formation of intermittent interpenetrating channel networks occurred in both alloys. The corrosion behavior of ε<sup>n</sup> is discussed in terms of its chemical composition and crystal structure. The corrosion activity of ε<sup>n</sup> could be further exploited in preparation of porous Pd–Co networks with possible catalytic activity.

**Keywords:** aluminum alloys; phase characterization; electrochemical corrosion; de-alloying

#### **1. Introduction**

Alloys with nominal chemical composition of approximately Al-30 at.% TM (TM stands for one or more transition metals) constitute a specific group of materials called complex metallic alloys (CMAs). These metallic materials contain, besides classical crystalline phases with simple unit cells, structurally complex intermetallic phases (SCIPs) [1]. The SCIPs are composed of giant unit cells and lack translational symmetry. Because of their complex atomic structure, the SCIPs are appealing for thin film applications, coatings, and reinforcement phases in composites [2].

The phase equilibria in the Al–Pd–Co system have been studied by Yurechko et al. [3,4], Cerniˇ ˇ cková et al. [5,6], and Duriška et al. [ ˇ 7]. The authors observed six stable ternary phases (W, Y2, U, V, F, C2) and a structurally complex ε-family. Selected phases occurring in the Al–Pd–Co system are summarized in Table 1 [3,8]. Their homogeneity ranges at 790 ◦C are shown in the corresponding isothermal phase diagram section (Figure 1). The cluster-based orthorhombic decagonal quasicrystalline approximant of the ε-family consists of five structures: two binary (ε6, ε28) and three ternary (ε16, ε22, ε34). Since the ε-family is considered to be a single phase from the thermodynamic point of view, it has been briefly denoted as εn. Although two lattice parameters (*a* and *b*) are identical for each structure within the family, the third lattice parameter (*c*) differs for each of the structures since it is associated with the cluster arrangement [3,9]. Contrary to Al–Pd alloys [10], Co was observed to substitute Pd in ε<sup>n</sup> in ternary Al–Pd–Co alloys. The Co solubility in ε<sup>n</sup> is up to approximately 15 at.% at 790 ◦C [3]. The Al–Pd–TM ε<sup>n</sup> phase is predominantly diamagnetic and has a good electrical conductivity. This phase is brittle and can be easily powdered. Furthermore, it contains Pd, a catalytically active element, which, in combination with a unique crystal structure, provides a variety of different adsorption sites. As such, the Al–Pd–Co SCIPs are interesting for catalytic applications [11].


**Table 1.** Overview of selected binary and ternary phases occurring in the Al–Pd–Co system and related binaries [3,8].

The corrosion activity of Al-based SCIPs is relatively unknown. It has been found that the electrochemical properties of CMAs differ from those of aluminum metal [12]. The previous studies of Al–Cu–Fe [13,14], Al–Cr–Fe [15], and Al–Cu–Fe–Cr [16] CMAs indicated that the relative amount of alloy phases and their chemical compositions had a major influence on their electrochemical behavior. It was presented that Cr additions significantly improved the corrosion resistance of Al–Cr–Fe and Al–Cu–Fe–Cr alloys [16]. Recent studies of Al–Co CMAs [17–21] have shown that the relative amounts of the alloy's phases and electrical contact between them played an important role in their corrosion performance. The anodic dissolution of different alloy phases was found to take place by a galvanic mechanism. The electrochemical nobility of Al–Co SCIPs was found to increase with increasing Co concentration. The phase crystal structure had only a secondary influence. An exception, however, was found for the structurally complex Z-Al3Co phase. This phase was found to be more corrosion resistant compared to Al5Co2 in chloride-containing environments [19]. The reason for this behavior could stem from the complex crystal structure of Z-Al3Co, formed by a complex monoclinic unit cell containing large pentagons composed of six small pentagons of monoclinic Al13Co4. The complex structure of this phase is probably stabilized by vacancies. The vacancies may influence the Co diffusivity leading to a protective layer formation on the sample surface.

**Figure 1.** Isothermal section of the Al–Pd–Co phase diagram at 790 ◦C, redrawn from Reference [3].

The corrosion behavior of Al–Pd alloys in various solutions has been studied in References [22–25]. The results showed a preferential Al dissolution from ε<sup>n</sup> (~Al3Pd). The corrosion attack of the structurally complex ε<sup>n</sup> in the Al–Pd alloys led to the formation of a porous, channel-like network [22–25]. This phenomenon is known as electrochemical de-alloying [26], i.e., a corrosion-driven process during which an alloy is decomposed by selective dissolution of the most electrochemically active element (Al). This process results in the formation of nano-porous metal networks composed of noble elements. In the NaCl aqueous solution, the de-alloying of Al–Pd alloys was found to be more pronounced in as-solidified alloys compared to as-annealed samples [24]. The de-alloying of Al–TM alloys has attracted much attention in recent years as a versatile tool for creating nano-porous metal networks with high catalytic activity [25]. Nano-porous ribbons of Pd, Au, Pt, and other precious metals have been fabricated through chemical de-alloying of rapidly solidified Al-based alloys under free corrosion conditions [27].

In the present work, the corrosion performance of Al70Pd25Co5 and Al74Pd12Co14 alloys (element concentrations are given in at.%) have been studied by potentiodynamic polarization in 3.5 wt.% NaCl aqueous solution for the first time. The aim of this work is to investigate the effect of both phase occurrence and chemical composition on the alloy's corrosion behavior. Furthermore, the effect of Co concentration on the corrosion behavior of ε<sup>n</sup> is studied.

#### **2. Materials and Methods**

The alloys with nominal compositions Al70Pd25Co5 and Al74Pd12Co14 were prepared by repeated arc-melting of Al, Pd, and Co granules (purity of 99.95%) in argon. After melting, the alloys were rapidly solidified on a water-cooled copper mold, cast in epoxy resin, and metallographically prepared by wet grinding and polishing down to a surface roughness of 1 μm. The as-solidified alloy's phase constitution and microstructure were studied by room temperature X-ray diffraction (XRD) and scanning electron microscopy (SEM), respectively. During XRD experiments, a Panalytical Empyrean PIXCel 3D diffractometer (Malvern Panalytical Ltd., Malvern, UK) with Bragg–Brentano geometry and Co Kα1,2 radiation was used. The measurements were conducted in the 2θ range between 20◦ and 60◦, with the step size 0.0131◦ and the counting time 98 s per step. For the microstructure observation, a JEOL JSM-7600F scanning electron microscope (JEOL, Akishima, Tokyo, Japan), equipped with an Oxford Instruments X-max 50 spectrometer (Oxford Instruments, Abingdon, UK) and operated by the INCA software (version 5.04), was employed. The microscope was operated at the acceleration voltage of 20 kV. The scanning was performed in regimes of secondary (SEI) and backscattered (BEI) electrons. Furthermore, a scanning transmission electron microscope JEOL JEM ARM200F (JEOL, Akishima, Tokyo, Japan), operated at 200 kV and equipped with a high-angle annular dark-field detector (HAADF/STEM), was employed to obtain HAADF images. The two-dimensional (2D) projections of crystal structures were calculated in PowderCell software (version 2.4) using the data derived from References [8,28,29].

The corrosion experiments were conducted at room temperature (21 ± 2 ◦C) in a 500 ml glass vessel filled with an aqueous electrolyte. A three-electrode setup was used. The working electrode consisted of the polished surface of the Al–Pd–Co alloy with an exposed area of about 1 cm2. A silver–silver chloride electrode immersed in a saturated sodium chloride solution (saturated Ag/AgCl electrode) was used as a reference electrode. The counter electrode was a platinum mesh (2 <sup>×</sup> 2 cm2). The corrosion experiments were conducted in an aqueous NaCl solution (concentration 0.6 mol dm<sup>−</sup>3). The solution was prepared immediately before the experiment by dissolving 35 g of NaCl in 1 liter of de-ionized water (conductivity <20 μS). The electrolyte was not de-aerated before the experiment to simulate real environmental conditions. The progress of the reaction was controlled by a PGU 10 V-1A-IMP-S potentiostat/galvanostat from Jaissle Electronic Ltd. (Waiblingen, Germany).

The surface topography of the corroded samples was analyzed by a Zeiss LSM 700 confocal laser scanning microscope (CLSM, Zeiss, Oberkochen, Germany). The ZEN 2009 software was used for the three-dimensional topographical resolution.

#### **3. Results and Discussion**

#### *3.1. Microstructure and Phase Occurrence before Corrosion Testing*

The microstructures of the as-solidified Al70Pd25Co5 and Al74Pd12Co14 alloys are illustrated in Figure 2. The XRD patterns corresponding to the above alloys are given in Figure 3a,b, respectively. The metal concentrations of microstructure constituents determined by SEM/EDX and their phase assignments are presented in Table 2.

**Figure 2.** BEI/SEM images of microstructure constituents in as-solidified Al70Pd25Co5 (**a**) and Al74Pd12Co14 (**b**) alloys. Black areas in (**b**) correspond to pores. Phases assigned to particular constituents are also marked.

**Figure 3.** XRD diffraction patterns of as-solidified Al70Pd25Co5 (**a**) and Al74Pd12Co14 (**b**) alloys.

**Table 2.** Metal concentrations and phase assignments of microstructure constituents observed in as-solidified Al70Pd25Co5 and Al74Pd12Co14 alloys.


The microstructure of the Al70Pd25Co5 alloy consisted of two different constituents (Figure 2a). The images were acquired in a BEI regime and therefore the bright regions have a higher Pd concentration compared to the dark constituents. The metal concentrations and volume fractions of the bright-grey microstructure constituent (Table 2) indicate that it corresponds to the δ phase (Al3Pd2). This assumption was also confirmed by X-ray diffraction (Figure 3a). The visually and chemically homogeneous dark-grey constituent was identified to be a mixture of ε<sup>n</sup> structures (Figure 3a). To index diffraction peaks of particular ε<sup>n</sup> structures, the data derived from References [8,28,29] were used.

In the XRD pattern of the Al74Pd12Co14 alloy (Figure 3b), a combination of ε6, ε16, and ε<sup>28</sup> structures was identified. In the related microstructure image, however, a chemically heterogeneous constituent has been observed (Figure 2b). The dark-grey areas had an increased Co concentration, while the bright areas showed a higher Pd concentration compared to the dark-grey areas. The atomic structure of the as-solidified Al74Pd12Co14 alloy was observed using HAADF/STEM. Three different structural motives have been recognized in the atomic structure of this alloy (ε6, ε16, and ε28, Figure 4). For each ε<sup>n</sup> structure, specific combinations of phason tiles are characteristic. ε<sup>6</sup> is formed by hexagons only, ε<sup>16</sup> is represented by the combination of pentagons and nonagons, while ε<sup>28</sup> comprises all three types of tiles. It has been suggested that transitions between various structures of the ε-family could be associated with a small rearrangement of clusters, resulting in changes in the occurrence and/or configuration of phason tiles. The arrangement of tiles in particular ε<sup>n</sup> structures, observed experimentally in this work, was also calculated using the data derived from References [8,28,29]. The 2D projection of the ε6, ε16, and ε<sup>28</sup> structures, presented in Figure 5, is in a good agreement with the HAADF/STEM image.

**Figure 4.** A high-angle annular dark-field (HAADF)/STEM image of the atomic structure of the as-solidified Al74Pd12Co14 alloy. Phason tiles, i.e., hexagon, pentagon, and banana-shaped nonagons, are highlighted by solid lines. Yellow, green, and orange structural motifs correspond to ε6, ε16, and ε28, respectively. For the color interpretation of this figure, the reader is referred to the web version of this article.

**Figure 5.** Two-dimensional projection of the crystal structure of ε6, ε16, and ε28. The phason tiling is denoted by dark-blue lines. For the color interpretation of this figure, the reader is referred to the web version of this article.

Structures of ε<sup>6</sup> and ε<sup>28</sup> were reported to be binary structural variants of εn, while ε<sup>16</sup> has been described as a ternary ε<sup>n</sup> structure [3,10,29,30]. In the latter structure, Co atoms substitute Pd. Therefore, the dark-grey areas (Figure 2b, Table 2), enriched with Co from the Co–Pd balance point of view, could be assigned to the ternary ε<sup>16</sup> structure in the as-solidified Al74Pd12Co14 alloy. Similarly, the bright areas in Figure 2b could be ascribed to the mixture of ε<sup>6</sup> and ε<sup>28</sup> structures, which lie closer to the Al–Pd binary edge of the Al–Pd–Co ternary system. The bright areas were located preferentially around pores. The pores were formed on the grain boundaries during solidification due to shrinking. Co and Pd concentrations of ε<sup>n</sup> changed since de-mixing took place during solidification. The Pd concentration in ε<sup>n</sup> increased towards the grain boundary. Thus, the Pd-rich ε<sup>n</sup> (ε<sup>6</sup> + ε28) were located preferentially around pores. The Co-rich ε<sup>n</sup> (ε16) was located in the center of the grain as this phase structure solidified from the melt. The overall chemical composition of the ε<sup>n</sup> phase in the Al74Pd12Co14 alloy is presented in Table 2. Due to the presence of ε16, the ε<sup>n</sup> phase in the Al74Pd12Co14 alloy had a significantly higher Co concentration compared to the Al70Pd25Co5 alloy where the ternary ε<sup>16</sup> phase has not been identified.

The distributions of particular structures within the ε<sup>n</sup> phase were previously studied in the Al–Pd and Al–Pd–Co systems; however, the exact boundaries between structures have not been determined yet. In the Al–Pd system, Yurechko et al. [10] proposed a hypothetical double-phase area (ε<sup>6</sup> + ε28) in between two single-phase areas (ε<sup>6</sup> and ε28). In the partial phase diagram published by Grushko [31], ε<sup>6</sup> and ε<sup>28</sup> have been positioned in a common "single-phase (ε<sup>6</sup> + ε28)" area consisting of two presumably separated subareas adherent to the particular structures. Earlier, the same distribution of ε<sup>6</sup> and ε<sup>28</sup> was studied by Balanetskyy et al. [28] in the Al–Pd–Fe system at 750 ◦C. Moreover, the homogeneity ranges of ε<sup>16</sup> and ε<sup>22</sup> have been defined. However, the strict boundaries between particular structures have not been described. In the Al–Pd–Co system, Yurechko et al. [3] estimated the boundaries of all the structures within the ε-family. Considering the results obtained using HAADF/STEM in this work and in [29], it can be assumed that the transitions between particular structures are rather open as schematically highlighted in gradient colors (green, red, yellow, and blue) in Figure 6. As follows from this figure, several ε<sup>n</sup> structures in the transient area can coexist. This situation can also be seen in the microstructure of the Al74Pd12Co14 alloy. The dark-grey areas, corresponding to the ε<sup>16</sup> structure, fluently transformed to the bright areas represented by the mixture of ε<sup>6</sup> and ε<sup>28</sup> structures (Figure 2b). The chemical composition of ε<sup>6</sup> is very close to the composition of ε28. Consequently, this bright-grey microstructure constituent in the Al70Pd25Co5 alloy (Figure 2a) can be considered to be homogeneous. Individual ε<sup>6</sup> and ε<sup>28</sup> structures can be recognized in the HAADF/STEM image only (Figure 4).

**Figure 6.** Schematic positions of binary and ternary phases in the isothermal section of a partial Al–Pd–Co diagram, redrawn from Reference [3]. For the color interpretation of this figure, the reader is referred to the web version of this article.

#### *3.2. Corrosion Behavior*

Immediately after the sample's immersion in aqueous NaCl, an open circuit potential (OCP) was recorded. The OCPs of the alloys are presented in Figure 7. A distinct behavior has been observed. While the OCP of the Al74Pd12Co14 alloy was relatively stable over time, irregular oscillations for the Al70Pd25Co5 alloy have been found. Furthermore, the OCPs of the Al70Pd25Co5 alloy were less negative and a difference of more than 200 mV was found compared to the Al74Pd12Co14 alloy.

**Figure 7.** Open circuit potential of the as-solidified Al74Pd12Co14 and Al70Pd25Co5 alloys in 0.6 M NaCl. For the color interpretation of this figure, the reader is referred to the web version of this article.

The as-solidified Al74Pd12Co14 alloy is a single-phase alloy. The OCP of this alloy therefore corresponds to the electrochemical activity of εn. The Al70Pd25Co5 alloy, on the other hand, is a double-phase alloy composed of ε<sup>n</sup> and δ (Al3Pd2). The less negative OCP of this alloy indicates a higher electrochemical potential of δ. Because of the potential difference between ε<sup>n</sup> and δ, local galvanic cells may have been formed on the surface of the Al70Pd25Co5 alloy.

Every physical contact between δ and ε<sup>n</sup> corresponds to an elementary galvanic corrosion cell. During galvanic corrosion, there is a net current flow between the cathodic microstructure constituent (δ) and its adjacent matrix (εn). The metal ions dissolve into the solution on the anode and electrons released flow to the micro-cathodic area for the reduction process. This causes a redistribution of electrical charge between anodic (εn) and cathodic areas (δ), thereby leading to a variation of the OCP. As the OCP is measured at the tip of the Haber–Luggin capillary, it represents the overall contributions of all elementary galvanic cells on the sample surface [32]. These contributions are not correlated. A high number of elementary galvanic cells between ε<sup>n</sup> and δ co-exist with each other in the microstructure of the Al70Pd25Co5 alloy (Figure 2a). Their interactions are combined and contribute to the overall corrosion behavior of this alloy.

A further insight into the peculiar corrosion behavior of the Al–Pd–Co alloys was obtained by potentiodynamic polarization. After the OCP measurement, a polarization scanning from −1000 mV to 0 mV (Ag/AgCl) was performed using a sweeping rate of 1 mV s<sup>−</sup>1. After reaching 0 mV (Ag/AgCl), the polarization direction was reversed and returned back to the initial potential (the direction of the polarization is indicated by open arrows in Figure 8). The resulting cyclic polarization curves are depicted in Figure 8. The forward curves are characterized by the corrosion minimum followed by an increase of the current density at potentials less negative than the corrosion potential. The current density increase was further followed either by stabilization (the Al74Pd12Co14 alloy) or even a slight decrease of the current density (the Al70Pd25Co5 alloy). This behavior indicates a passivation of the alloys. The transient behavior was further followed by a sudden current density increase at potentials

less negative than −400 mV (Ag/AgCl), indicating a breakdown of the passive film. Upon reverse polarization, a re-passivation of the existing pits occurred. In order to compare equally polarized samples, we reversed the scanning at the fixed potential. The forward curves presented in Figure 8 have been analyzed by Tafel extrapolation [33]. The electrochemical parameters of the alloys (corrosion potential, corrosion current density, and breakdown potential) are listed in Table 3. A re-passivation potential obtained from the reverse curve is also presented. However, caution is required when comparing the individual re-passivation potentials of the alloys. The currents at the vertex were higher for the Al74Pd12Co14 alloy and this might have influenced the pit depth and local chemistry [33]. In order to obtain more comparable *E*rp values, reversing the polarization at a constant current density would be necessary.

**Figure 8.** Cyclic potentiodynamic polarization curves of the as-solidified Al74Pd12Co14 and Al70Pd25Co5 alloys in 0.6 M NaCl. The polarization direction and positions of breakdown (Ebd) and re-passivation potentials (Erp) are indicated by arrows. For the color interpretation of this figure, the reader is referred to the web version of this article.

**Table 3.** Electrochemical parameters of as-solidified Al70Pd25Co5 and Al74Pd12Co14 alloys. Corrosion potentials (Ecorr) and corrosion current densities (jcorr) were obtained by Tafel extrapolation of forward curves (Figure 8).


Based on the above-presented results, a corrosion mechanism of the Al–Pd–Co alloys has been postulated. The corrosion mechanism is depicted in Figure 9. Pitting is a highly localized form of corrosion that happens in the presence of halide anions, such as Cl− [34]. Initially, a protective alumina scale has been formed on the sample surface, which is indicated by a current density plateau observed upon sample polarization for both alloys. This plateau is observed at potentials of −600 to −300 mV versus Ag/AgCl for Al70Pd25Co5 alloy, i.e., at potentials less negative than is the corrosion potential (Figure 8). In the presence of Cl−, however, this passive layer has been weakened. Aluminum forms unstable [AlCl4] − complexes that dissolve in aqueous solutions. The dissolution of the protective alumina scale in NaCl leaves a naked alloy surface susceptible to further corrosion attack (Figure 9).

**Figure 9.** Pitting corrosion mechanism of the Al–Pd–Co alloys. For the color interpretation of this figure, the reader is referred to the web version of this article.

Interactions between co-existing phases in double-phase alloys may play an important role during corrosion [32,35,36]. Once the pitting potential is reached during sample polarization, the compact passivation layer becomes locally disrupted (Figure 9). As a result, Al3<sup>+</sup> cations are released from the alloy into the solution in the course of the following reaction

$$\text{Al} \rightarrow \text{Al}^{3+} + 3\text{e}^- \tag{1}$$

Reaction (1) leads to positive charge enrichment within the dissolution zone [37]. As a consequence, Cl− anions of the electrolyte rapidly migrate into the dissolution zone as presented in Figure 9. The released Al3<sup>+</sup> cations become solvated by water molecules. Consequently, the hydrolysis of [Al(H2O)4] <sup>3</sup><sup>+</sup> in aqueous environment takes place in line with the following reaction [34]

$$\left[\text{Al}(\text{H}\_2\text{O})\_4\right]^{3+} + \text{H}\_2\text{O} \iff \left[\text{Al}(\text{H}\_2\text{O})\_3(\text{OH})\right]^{2+} + \text{H}\_3\text{O}^+\tag{2}$$

Hydroxo complexes of Al may further react with chloride and water according to the following reactions

$$\rm{[Al(H\_2O)\_3(OH)]^{2+} + Cl^- \rightarrow [Al(H\_2O)\_2(OH)Cl]^+ + H\_2O} \tag{3}$$

$$\text{[Al(H}\_{2}\text{O)}\_{2}\text{(OH)Cl]}^{+} + \text{H}\_{2}\text{O} \rightarrow \text{[Al(H}\_{2}\text{O)(OH)}\_{2}\text{Cl]} + \text{H}\_{3}\text{O}^{+}\tag{4}$$

By these reactions, hydrogen cations are released into the pit. Their accumulation yields to a local pH decrease within the dissolution zone, which is known as a self-acidifying effect [34,37]. The presence of H<sup>+</sup> in pits further accelerates the Al dissolution. At the cathode (δ, Figure 9), a reduction of water may take place in accordance with the following reaction

$$\rm{^2O\_2 + 2H\_2O + 4e^- \to 4OH^-} \tag{5}$$

At the pit walls and possibly in their immediate vicinity, since pH is reduced due to hydrolysis reactions (2) and (4), the most likely prevailing cathodic reactions are

$$\rm O\_2 + 4H^+ + 4e^- \rightarrow 2H\_2O \tag{6}$$

or

$$2\text{H}^+ + 2\text{e}^- \rightarrow \text{H}\_2\tag{7}$$

The latter reaction takes place in the case of pH having fallen to very low values. As a result, emerging bubbles of H2 evolve on the alloy surface.

The results presented in Table 3 show that the OCP of the double-phase Al70Pd25Co5 alloy is located between the breakdown and re-passivation potentials. This is also manifested by the OCP oscillations observed between −330 mV (Ag/AgCl) and −400 mV (Ag/AgCl, Figure 7), indicating periodic breakdown and re-passivation events on the sample surface. These observations indicate that this alloy was in a localized corrosion stage already upon the sample's immersion in the electrolyte, contrary to the single-phase Al74Pd12Co14 alloy. The passivation stage was found to be more pronounced in the case of the double-phase Al70Pd25Co5 alloy (Figure 8). This could be related to the presence of the noble δ phase in this alloy. For the mono-phasic Al74Pd12Co14 alloy, on the other hand, a higher corrosion current density (jcorr) has been found, reflecting a higher dissolution rate. Moreover, a more negative corrosion potential for this alloy has been found, indicating a higher corrosion susceptibility. The above-reported differences in the corrosion behavior of the alloys could result from their different microstructures. More information about the specific corrosion attack of different SCIPs has been therefore obtained by investigating the alloys' microstructures after electrode polarization.

The microstructures of the as-polarized alloys are documented in Figure 10. Metal concentrations of the phases after corrosion are summarized in Table 4. For both alloys, a preferential attack of ε<sup>n</sup> was found. δ as a nobler phase in the Al70Pd25Co5 alloy has been retained. A de-alloying of Al from ε<sup>n</sup> as well as formation of intermittent inter-penetrating channel networks have been observed in both alloys (Figure 10). In the single-phase Al74Pd12Co14 alloy, a higher density of intermittent inter-penetrating channels and pits has been found (Figure 10). Moreover, the channels formed a cross-linked network. This behavior is similar to the Al–Pd alloys, where the pits were observed in the interconnection between two channels [24]. In the Al70Pd25Co5 alloy, the pits were observed to be randomly distributed in the channels (Figure 10).

**Figure 10.** BEI/SEM images (**a**,**b**) and CLSM images (**c**,**d**) of as-polarized Al70Pd25Co5 (**a**,**c**) and Al74Pd12Co14 (**b**,**d**) alloys. Phases assigned to particular constituents are also marked. For the color interpretation of this figure, the reader is referred to the web version of this article.


**Table 4.** Metal concentrations and phase assignments of the as-polarized Al–Pd–Co alloys. Differences in metal concentrations between as-polarized and as-solidified alloys are also presented (compare with data in Table 2).

A dissolution of Al in the Al70Pd25Co5 alloy has been found (Table 4). Simultaneously, the Al concentration in ε<sup>n</sup> decreased from 72.5 to 69.0 at.% (Table 4). In the Al74Pd12Co14 alloy, a decrease of Al concentration in ε<sup>n</sup> from 73.9 to 71.1 at.% has been found. Thus, the level of Al de-alloying was higher in the double phase Al70Pd25Co5 alloy. Moreover, the pits found in this alloy were deeper compared to the Al74Pd12Co14 alloy. The formation of cracks observed in the as-polarized alloys could be governed by a combination of de-alloying kinetics and the release rate of internal stresses. As the electrochemical potential is raised in a positive direction, the dissolution rate of the alloy increases (Figure 8). This electrochemical force drives the surface at the de-alloying front further away from the equilibrium [37]. The removal of Al from the alloy phases leads to microcrack initiation. The residual stress accumulated in the alloys during rapid solidification is released during de-alloying.

Corrosion potentials and corrosion current densities of the as-solidified Al–Co [17,18,20,21,38], Al–Pd [24], and Al–Pd–Co alloys are compared in Figure 11. The corrosion potentials of the Al–Co alloys show a significant dependence on the Al atomic fraction. They become more negative with increasing Al concentration. The corrosion potentials of the Al–Pd alloys, on the other hand, are relatively constant with respect to the alloy's overall chemical composition. They are, in fact, more negative than the corrosion potentials of the remaining two groups of alloys. The corrosion currents of the Al–Pd alloys, on the other hand, are higher compared to the Al–Co and Al–Pd–Co alloys (Figure 11b). These observations suggest that the Al–Pd alloys are less corrosion-resistant compared to both the Al–Pd–Co and Al–Co alloys. The corrosion behavior of the Al–Pd–Co alloys is closer to the behavior of the Al–Co alloys. This observation is unexpected, since both alloy groups have a different chemical composition and phase constitution. Moreover, εn, the preferentially corroding phase in the Al–Pd–Co alloys, is not present in the Al–Co alloys. Therefore, it can be suggested that Al3Co SCIPs are nobler compared to binary ε<sup>n</sup> (Al3Pd). This is manifested by the less negative corrosion potentials of the Al–Co alloys compared to the Al–Pd alloys with a similar Al atomic fraction (Figure 11). Furthermore, the Co substitution for Pd significantly improves the corrosion resistance of εn. As such, it is not the crystal structure of the phase, but its chemical composition, that plays a major role in the corrosion behavior.

To further probe this hypothesis, we have plotted the corrosion data of other ternary Al–TM systems, found in the literature, together with those of the Al–Pd–Co system. The data survey [13,39–42] is presented in Figure 12. Although the data are scattered due to large variations in alloy chemical compositions, some trends can be identified. The as-solidified Al–Cu–Pd and Al–Cu–Fe alloys have lower corrosion currents compared to the Al–Pd–Co alloys [13,39]. The addition of Pd was found to slightly decrease the corrosion current of the Al–Cu alloys in chloride solution [39]. For the Al4Cu9 samples, however, not much effect from Pd has been seen [39]. The corrosion potentials of these alloys are found over a broad range of values. The scatter in Ecorr values, however, could be caused by variations in their chemical composition. The Al–Cr–Fe alloy is also presented in Figure 12 [40]. This alloy has a more negative corrosion potential due to the absence of noble elements, such as Pd. Furthermore, it has a low corrosion current due to the presence of chromium, which forms a passive layer on the sample surface.

**Figure 11.** Corrosion potentials (**a**) and corrosion current densities (**b**) of as-solidified Al–Co, Al–Pd, and Al–Co–Pd alloys. Lines are a guide to the eyes only. For the color interpretation of this figure, the reader is referred to the web version of this article.

**Figure 12.** Corrosion current densities versus corrosion potentials of selected ternary Al-based complex metallic alloys (as-cast and as-annealed alloys only). For the color interpretation of this figure, the reader is referred to the web version of this article.

Interesting is the corrosion behavior of as-solidified Al–Co–Ti alloys [41]. These alloys have corrosion potentials comparable to those of the Al–Pd–Co alloys (Figure 12). The concentration of Ti in these alloys was fixed at 2 at.% and the atomic concentration of Co varied between 5 and 30 at.% (Al–xCo–2Ti alloys). As such, the materials design of these alloys had features typical of the Al–Co alloys [41,42]. In general, the corrosion currents of the Al–Co–Ti alloys are higher compared to those of the Al–Pd–Co alloys. An exception was found, however, for the Al–15Co–2Ti alloy since this alloy had a very low corrosion current. This difference is, however, attributable to the fact that the intermetallic particles present in this alloy (Al9Co2, Al13Co4, and Al3Ti) were of different morphologies and volume fractions compared to the remainder of the alloys [41]. These observations indicate that the specific Co concentrations may greatly improve the corrosion performance of the Al–TM alloys. The ε<sup>n</sup> phase in the Al–Pd–Co alloys contains a significant amount of Co. The Co additions thus contribute to the corrosion resistance of the Al–Pd–Co alloys and this is especially obvious in the case of the double phase Al70Pd25Co5 alloy.

Al3Ti and Al3Fe are noble intermetallic phases with respect to the aluminum matrix [36]. The results presented in this work show that Al3Co is also relatively noble. These phases are nobler compared to binary ε<sup>n</sup> (Al3Pd). Co substitution for Pd thus significantly improves the corrosion resistance of εn. As such, it is not the crystal structure of the phase, but its chemical composition, that plays a major role in the corrosion behavior. The electrochemical behavior of constituent phases may change over time. In a recent study, Zhu et al. studied the evolution of corrosion behavior of intermetallic phases in Al alloys over time [43]. At the early stages, the corrosion attack occurred in the form of de-alloying. However, as the time progressed, the particles became nobler as a result of Al dissolution [43]. This particle ennoblement may accelerate the galvanic dissolution of the surrounding matrix. The corrosion behavior of constituent phases may also change as a result of long-term annealing. The long-term annealing causes element redistribution and reduces stresses accumulated during rapid solidification [19]. A comparative study of as-annealed, near-equilibrium Al–Pd–Co alloys is planned and results will be reported in a future publication.

#### **4. Conclusions**

In this work, the corrosion performance of as-solidified Al70Pd25Co5 and Al74Pd12Co14 alloys was studied by open circuit potential measurements and potentiodynamic polarization in aqueous NaCl (3.5 wt.%), following an in-depth structural characterization of the alloys. The alloys were prepared by arc-melting of Pd, Al, and Co lumps in argon. Based on the results, the following conclusions can be presented:


7. Specific Co concentrations may greatly improve the corrosion performance of the Al–TM alloys.

**Author Contributions:** Conceptualization, M.P., L.D., I.C., and J.J.; Funding acquisition, J.J.; Investigation, M.P., L.D., I.C., S.B., Z.G., M.S., and L.C.; Methodology, M.P., L.D., and I.C.; Project administration, J.J.; Supervision, J.J.; Writing—original draft, M.P.; Writing—review & editing, M.P., L.D., and J.J.

**Funding:** This work was supported by project no. 1/0490/18 of the Grant Agency VEGA of the Slovakian Ministry of Education, Research, Science and Sport, project APVV-15-0049 of the Slovak Research and Development Agency, and project NFP313010T606 (PROGMAT) supported by European Structural Investment Funds.

**Acknowledgments:** Shinichi Watanabe (JEOL Ltd., Tokyo, Japan) is acknowledged for his assistance with the HAADF/STEM measurements. This paper is dedicated to the memory of our deceased fathers, Ján Palcut and Jozef Gerhát.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Influence of Alloyed Ga on the Microstructure and Corrosion Properties of As-Cast Mg–5Sn Alloys**

#### **Jing Ren, Enyu Guo, Xuejian Wang, Huijun Kang, Zongning Chen \* and Tongmin Wang \***

Key Laboratory of Solidification Control and Digital Preparation Technology (Liaoning Province), School of Materials Science and Engineering, Dalian University of Technology, Dalian 116024, China; 18840830427@163.com (J.R.); eyguo@dlut.edu.cn (E.G.); wangxuejian0618@163.com (X.W.); kanghuijun@dlut.edu.cn (H.K.)

**\*** Correspondence: znchen@dlut.edu.cn (Z.C.); tmwang@dlut.edu.cn (T.W.); Tel.: +86-411-84709500 (Z.C.); +86-411-84706790 (T.W.)

Received: 18 September 2019; Accepted: 5 November 2019; Published: 8 November 2019

**Abstract:** In this paper, the microstructures and corrosion behaviors of as-cast Mg–5Sn–*x*Ga alloys with varying Ga content (*x* = 0, 0.5, 1, 2, 3 wt %) were investigated. The results indicated that Ga could not only adequately refine the grain structure of the alloys, but could also improve the corrosion resistance. The microstructures of all alloys exhibited typical dendritic morphology. No Ga-rich secondary phases were detected when 0.5 wt % Ga was added, while only the morphology of Mg2Sn phase was changed. However, when the addition rate of Ga exceeded 0.5 wt %, an Mg5Ga2 intermetallic compound started to form from the interdendritic region. The volume fraction of Mg5Ga2 monotonically increased with the increasing Ga addition level. Although Mg5Ga2 phase was cathode phase, its pitting sensitivity was weaker than Mg2Sn. In addition, the standard potential of Ga (−0.55 V) was lower than that of Sn (−0.14 V), which relieved the driving force of the secondary phases for the micro-galvanic corrosion. An optimized composition of 3 wt % Ga was concluded based on the immersion tests and polarization measurements, which recorded the best corrosion resistance.

**Keywords:** magnesium; immersion test; polarization; microstructure; corrosion resistance

#### **1. Introduction**

Magnesium alloys, with their excellent properties of low density and high specific strength, have received extensive attention in automobile, aerospace, electronics, and other industries [1–4]. Among many of the binary alloy systems, the Mg–Sn phase diagram demonstrates that the Mg-rich side has a very shallow α-Mg solvus curve, indicating that it is conducive to maximizing the precipitation of the thermally stable Mg2Sn particles. This infers that adding Sn potentially helps to enhance the mechanical integrity of the alloy at elevated temperatures [5–9]. On this account, the Mg–Sn-based alloy system is one of the most extensively studied alloys in recent years [6–13].

Aside from the mechanical strength, other factors that limit the application of magnesium alloys are their high chemical activity and poor corrosion resistance. It was documented, however, that alloying with some Sn results in superior corrosion resistance. Ha et al. [14] reported that Sn decreases the cathodic current density and inhibits the cathodic H2 evolution. Similar effects were also observed in Mg–5Al–1Zn alloy [11] and Mg–7Al–0.2Mn alloy [12]. The reason for inhibition of H2 evolution by alloying Sn was studied. Ha et al. [13] observed that the main reason for the decrease of H2 evolution rate was that Sn elements were enriched on the metal surface. They further reported that Sn had a higher potential H2 evolution than the Mg matrix [15,16]. Moreover, under the condition of a high cathode, Sn could form SnH4 [14]. That is to say, when the matrix is corroded, the element Sn is spontaneously enriched on the matrix surface, acting as a barrier to further corrosion [13].

The addition of Ga in magnesium alloy has important research value in seawater battery and sacrificial anode [17]. In those applications, Ga enhances the mechanical performance of Mg-based alloys by solid solution and precipitation strengthening [17,18]. It was suggested that Mg–Ga, Mg–In, and Mg–Sn alloys have potential use as biomaterials [15]. Among the aforementioned three binary alloys, Mg–Ga alloys appeared to be the best candidates, considering both the mechanical properties and corrosion behaviors in 0.9 wt % NaCl solution. Marta et al. [19] pointed out that the superior corrosion resistance of Mg–Ga is mainly related to the microgalvanic corrosion between α-Mg and the second phases, and the morphology and distribution of Mg5Ga2 phase.

The purpose of this study is to explore the effect of Ga on the microstructure and corrosion resistance of Mg–5Sn alloy. Both potentiodynamic polarization measurements and immersion weightlessness experiments were conducted to evaluate the corrosion property of the experimental Mg–5Sn–*x*Ga alloys. By investigating the microstructures of Mg–5Sn–*x*Ga alloys with varying Ga contents, the immanent relationship between the microstructure and corresponding corrosion behavior is established, and the results are discussed.

#### **2. Materials and Methods**

#### *2.1. Specimen Preparation*

As-cast Mg–5Sn–*x*Ga (*x* = 0, 0.5, 1, 2, and 3, all in weight percentage unless otherwise specified) alloys were investigated in this work. The experimental alloys were prepared by melting pure metals in an electric resistance furnace under the protection of an atmosphere containing CO2 and SF6 mixture in a ratio of 99:1. Pure Mg (99.99%) ingot was cut into small pieces and put in a magnesia crucible. Once the temperature of the melt reached 973 K, pure Sn (99.99%) and Ga (99.999%) granules were added to the melt. The melt was fully stirred with a graphite impeller for 180 s before it was poured into a cylindrical steel mold preheated at 523 K. The chemical compositions of the alloys were measured by X-ray fluorescence analysis (XRF-1800, Shimadzu, Kyoto, Japan), and the results are summarized in Table 1.


**Table 1.** Chemical compositions of the experimental Mg–5Sn–*x*Ga (*x* = 0, 0.5, 1, 2, 3 wt %) alloys analyzed by XRF tests.

#### *2.2. Microstructural Characterization*

The samples for microstructural observation were cut 10 mm from the bottom of the ingot. The exposed surfaces were ground by SiC abrasive papers up to 4000 grit and then polished up to 1 μm in a suspension of diamond pastes. The polished surfaces were etched in a mixed solution of picric and acetic acid (10 mL acetic acid, 4.2 g picric acid, 10 mL distilled water, and 70 mL ethanol). The microstructures and corrosion morphologies of the specimens were observed using an optical microscope (OM, Olympus GX51, Olympus Corp., Tokyo, Japan) and a field emission scanning electron microscope (FESEM, Zeiss supra 55, Zeiss Corp., Oberkochen, Germany), equipped with energy-dispersive X-ray spectroscopy (EDS). Phases were identified by an X-ray diffraction instrument (XRD, PANalytical Empyrean, Almelo, The Netherlands) with Cu Kα radiation at a scanning speed of 0.142224◦/s. An electron microprobe analyzer (EPMA, JXA-8530F PLUS, Tokyo, Japan) equipped with a wavelength dispersive spectrometer (WDS) was used to examine the elemental distribution of

selected phases. Grain size was measured using the OM micrographs of the alloys by the intercept method regulated in GB/T 6394-2002:

$$
\tau = \frac{L}{M \times N} \tag{1}
$$

where τ is the average grain size, *L* is the measured mesh length, *M* is the magnification for observation (50× in this work), and *N* is the number of intercept points on the measured mesh as indicated in Equation (1). The grain sizes were averaged with values measured from six micrographs for each alloy.

#### *2.3. Corrosion Tests*

The corrosion behaviors of the as-cast Mg–5Sn–*x*Ga (*x* = 0, 0.5, 1, 2, and 3 wt %) alloys were investigated using a potentiodynamic polarization test and immersion test. All experiments were carried out at room temperature (298 ± 1 K).

For the electrochemical test, a conventional three-electrode cell was employed. The cell was composed of a working electrode (sample), a reference electrode (Ag/AgCl electrode, saturated KCl with electrode potential of 0.1981 V vs. Standard Hydrogen Electrode), and a Pt plate counter electrode. The test was carried out using a Gamry electrochemical workstation (Reference 600, Gamry Instruments Inc., Warminster, PA, USA). The samples were ground using SiC abrasive papers up to 2000 grit, and the surface of a round specimen with an area of 1.0 cm2 was exposed to 3.5 wt % NaCl solution. After reaching a steady open circuit potential (OCP), the polarization test was initiated from −0.3 V versus the OCP level of the sample to +0.3 V vs. OCP at a scanning rate of 1 mV/s. Three samples were tested for each alloy to ensure repeatability. The corrosion current density (*i*corr, mA/cm2) of the investigated alloys, which was obtained by fitting the cathodic branch of the potentiodynamic polarization curve, was converted to the average corrosion rate (*Ri*, mm/y) according to the following equation [20,21]:

$$R\_i = 22.85 \times i\_{\text{corr}} \tag{2}$$

The immersion test was performed on three parallel samples for each alloy. Samples with dimensions of 20 mm wide, 25 mm long, and 3 mm thick were mechanically ground using SiC abrasive papers up to 2000 grit. The ground samples were immersed in a 3.5 wt % NaCl solution at 298 K for 72 h. After immersion, the samples were taken out and cleaned with chromate acid (200 g/L CrO3 + 10 g/L AgNO3) to remove the corrosion products [22,23]. The obtained samples were rinsed with distilled water, cleaned in alcohol, and dried in air. The final specimens were weighed on an analytical balance (ME204E, Mettler Toledo Corp., Greifensee, Switzerland), and the corrosion rate (*C*R) for each sample was calculated by Equation (3) [24,25]:

$$\mathcal{C}\_R = \frac{K \times W}{A \times T \times D} \tag{3}$$

where *<sup>K</sup>* is a constant (8.76 <sup>×</sup> 104), *<sup>W</sup>* is the mass loss in g, *<sup>A</sup>* is the exposed area of the sample in cm2, *<sup>T</sup>* is the exposure time in hours, and *D* is the density in g/cm3.

Another group of samples for observation of corrosion surface morphology, corrosion depth, and products identification were Φ 12 mm in diameter and 3 mm in thickness. These samples were prepared following the same procedure as the samples for microstructural observation mentioned in Section 2.2. Measures were taken to ensure that corrosion occurred only on the transverse surface of the cylindrical samples. After immersion in 3.5 wt % NaCl solution for 24 h, XRD was used to identify the phases in the corrosion product. The surface and cross-section morphologies of the corroded samples were observed in SEM and EPMA, respectively.

#### **3. Results and Discussion**

#### *3.1. Microstructure Analysis*

The OM micrographs of the as-cast Mg–5Sn–*x*Ga alloys with varying Ga contents are shown in Figure 1. All the alloys present a similar typical dendritic grain structure. The solute elements could have significant effects on the growth behavior of α-Mg grains and the final morphological patterns [26–29]. It was documented that Sn could form a composition undercooling zone of liquid ahead of the solid–liquid interface, leading to dendrite formation [30]. The average grain sizes were measured to be 300.4 ± 18.4, 227.1 ± 19.3, 205.7 ± 16.7, 183.3 ± 12.7, and 155.9 ± 8.6 μm for the Mg–5Sn, Mg–5Sn–0.5Ga, Mg–5Sn–1Ga, Mg–5Sn–2Ga, and Mg–5Sn–3Ga alloys, respectively. The result suggests that the grain size decreased monotonically with increasing Ga content, probably due to the increasing grain growth restriction factor (*GRF*), which can be expressed by Equation (4) for a binary alloy [26]:

$$GRF = m\_L\mathbb{C}\_0(k\_0 - 1)\tag{4}$$

where *m*<sup>L</sup> is the slope of liquidus (assumed to be straight), *k*<sup>0</sup> the equilibrium distribution coefficient, and *C*<sup>0</sup> the concentration of a solute. According to the Mg–Ga phase diagram [31], in dilute binary magnesium alloys, the *GRF* value for Ga element is calculated to be 4.04, which is similar to Al and Zn element (*GRF*Al = 4.32, *GRF*Zn = 5.31 [26]). Composition undercooling was established by the Ga concentration gradient in the diffusion layer adjacent to the solid–liquid interface, restricting grain growth as a consequence of slow diffusion; and thus, the growth rate is limited and grain size is refined.

As for the case of the Mg–5Sn–0.5Ga alloy, the quantitative results from the WDS analysis of the areas marked by the white rectangles in Figure 1f are given in Table 2. It was found that Mg–5Sn alloy with 0.5 wt % Ga did not cause the formation of any new phases. Most of the Ga and a part of the Sn were either dissolved in the α-Mg matrix or enriched at the interdendritic region (indicated by arrow *a*).

**Table 2.** The chemical compositions of the areas marked by the white rectangles in Figure 1f using the wavelength dispersive spectrometer (WDS) (in wt %).


Dozens of studies have shown that the microstructural features, such as the grain size and the morphology of second phase, may have impacts on the corrosion resistance of magnesium alloys [32,33]. Grain refinement led to the decreased corrosion rates, possibly because of the enhanced passivity of the oxide film [34]. Because the experimental alloy with 3 wt % Ga had the finest equiaxed grains, it was most likely that the as-cast Mg–5Sn–3Ga alloy would exhibit the greatest corrosion resistance.

**Figure 1.** OM micrographs of the Mg–5Sn–*x*Ga alloys: (**a**) Mg–5Sn, (**b**) Mg–5Sn–0.5Ga, (**c**) Mg–5Sn–1Ga, (**d**) Mg–5Sn–2Ga, (**e**) Mg–5Sn–3Ga, and (**f**) backscattered electron (BSE) micrograph of the Mg–5Sn–0.5Ga alloy.

The backscattered electron (BSE) micrographs of the Mg–5Sn–*x*Ga alloys with varying Ga content are shown in Figure 2a–e. The OM and BSE micrographs of the alloys confirm the presence of second phases and element enrichment in the interdendritic regions. The elemental mapping results of the Ga, Mg, and Sn in the Mg–5Sn–3Ga sample using EPMA are shown in Figure 2f–i. Contrasts along the interdendritic contours demonstrate that there are two types of phases with different chemistries. According to the quantitative results measured by WDS in Table 3, the grey eutectic phase indicated by the letter *a* is Mg5Ga2 intermetallic compound, and the brighter phase indicated by the letter *b* is Mg2Sn. The inserts in Figure 2a–e are zoom-in views of the two phases embedded in the interdendritic region. When 0.5 wt % Ga was added, no Ga-rich second phases were detected, according to the composition

analysis in Table 2. When the addition of Ga exceeded 0.5 wt %, Ga formed from the interdendritic region in the form of Mg5Ga2 phase, while Mg2Sn did so in the form of divorced eutectic phase.

**Figure 2.** BSE micrographs of the Mg–5Sn–*x*Ga alloys: (**a**) Mg–5Sn, (**b**) Mg–5Sn–0.5Ga, (**c**) Mg–5Sn–1Ga, (**d**) Mg–5Sn–2Ga, (**e**) Mg–5Sn–3Ga. The inserts are zoom-in views showing the second phases. (**f**–**i**) Elemental mapping of the Mg, Sn, and Ga in the Mg–5Sn–3Ga sample using electron microprobe analyzer (EPMA).

**Table 3.** The chemical compositions of the phases measured by WDS as indicated by the arrows in Figure 2f (in at %).


The XRD spectra in Figure 3a further confirmed that the Mg–5Sn alloy consisted of α-Mg matrix and Mg2Sn. When the Ga content exceeded 1 wt %, a new phase containing Ga was formed. Survey on Joint Committee on Powder Diffraction Standards indicates that the new peaks correspond to the reflections of the Mg5Ga2 compound, which normally exists in Mg–Ga binary [15,35] and Mg–Hg–Ga ternary [18] alloys. This is consistent with the quantitative results measured by WDS in Table 3. The increasing density of the Mg5Ga2 peaks is expected with the increase in the Ga content of the alloys. Figure 3b shows the area fraction of the second phases in the investigated alloy calculated by ImageJ 1.47 (US NIH, Bethesda, MD, USA) [36]. It is also clear that the area fraction of Mg5Ga2 particles increased with increasing Ga content, while the area fraction of Mg2Sn particles decreased accordingly.

**Figure 3.** (**a**) X-ray diffraction (XRD) patterns of the Mg–5Sn–*x*Ga alloys; (**b**) the area fraction of the second phases in the investigated alloy.

#### *3.2. Polarization Tests*

Figure 4 shows the potentiodynamic polarization curves of the as-cast Mg–5Sn–*x*Ga alloys. The corrosion current density (*i*corr), corrosion potential (*E*corr), and the average corrosion rate (*R*i) obtained from the potentiodynamic polarization curves are listed in Table 4. The *i*corr of the as-cast Mg–5Sn–*x*Ga alloys in 3.5 wt % NaCl solution increase in the order of Mg–5Sn–3Ga, Mg–5Sn–2Ga, Mg–5Sn–0.5Ga, Mg–5Sn–1Ga, and Mg–5Sn. The lowest corrosion current density and corrosion potential are 2.79 <sup>×</sup> <sup>10</sup>−<sup>2</sup> mA/cm2 and <sup>−</sup>1.684 V for Mg–5Sn–3Ga alloy, respectively.

**Figure 4.** Polarization curves of the Mg–5Sn–*x*Ga alloys in 3.5 wt % NaCl solution.


**Table 4.** Fitting results of polarization curves of the Mg–5Sn–*x*Ga alloys with varying Ga content in 3.5 wt % NaCl solution.

It is accepted that the corrosion of Mg-based alloys is generally caused by the cathodic reaction of hydrogen evolution. The anodic reaction of Mg dissolution can be described as [37,38]:

$$\text{Anodic reaction: } \text{Mg} \rightarrow \text{Mg}^{+} + \text{e}^{-} \tag{5}$$

$$\text{Mg}^+ + \text{H}\_2\text{O} \rightarrow \text{Mg}^{2+} + \text{OH}^- + \frac{1}{2}\text{H}\_2\tag{6}$$

$$\text{Cathodic reaction:}\ 2\text{H}\_2\text{O} + 2\text{e}^- \rightarrow \text{H}\_2 + 2\text{OH}^-\tag{7}$$

Ga is reported to have relatively high hydrogen overpotentials [15]. Alloys with Ga are thus anticipated to inhibit the cathodic reaction of Equation (6). The cathodic branches of the polarization curves show that the cathodic current density decreases with increasing Ga content, demonstrating that Ga indeed delays cathodic reaction and thus reduces corrosion rate.

Figure 5a shows the representative anode branches of the polarization curves of the Mg–5Sn–*x*Ga alloys in 0.01 M NaCl solution. Within the whole range of anode polarization, the relation between the passive current density (*i*passive) of the alloys at any potential remains unchanged, so the relation between the current density of the alloys at an eigenvalue potential can be selected to describe the change of the trend of the passive current density of the alloys after adding Ga element. This method is favored by scholars in many studies. Ha et al. [39,40] used this method to study the change trend of anode passive current density at −1.7 V and cathode hydrogen evolution current density at −1.9 V of Mg–5Sn–(1–4 wt %)Zn alloy system. Similarly, Kim et al. [41] studied the polarization curve of Mg–8Sn–1Zn alloy with this method. In order to explore the role that Ga played in the anodic passive layer, Figure 5b compares the passive current density values for the alloys of different Ga contents with the alloy without Ga addition. The *i*passive values were measured at the anodic potentials of −1.65 V. Evidently, compared with the Mg–5Sn alloy, the *i*passive values decreased after adding Ga element, among which the Mg–5Sn–0.5Ga alloy exhibited the lowest *i*passive value. Decreased *i*passive indicates enhanced passive film stability. As mentioned above, Mg5Ga2 phase will be formed when Ga addition exceeds 0.5 wt %, which belongs to the cathode phase and destroys the stability of passive film to some extent. Moreover, the passive film of magnesium alloy is extremely unstable and cannot protect the matrix, so the corrosion rate is mainly controlled by hydrogen evolution reaction of the cathode.

Another factor that also influences the corrosion rate is the microstructural features of an alloy [42–44]. With the increase in Ga element, the area fraction of Mg2Sn particles monotonically decreases, while the opposite trend is the case for the Mg5Ga2, as shown in Figures 2 and 3. Liu et al. [16] reported that both the hydrogen evolution rate and corrosion potential decreased with decreasing volume fraction of the Mg2Sn particles, which is consistent with the experimental results in this study.

**Figure 5.** (**a**) The representative polarization curves of the Mg–5Sn–*x*Ga alloys in 0.01 M NaCl solution. (**b**) Passive current density (*i*passive) values measured at −1.65 V based on (**a**). The average and standard deviation were obtained from at least five measurements.

#### *3.3. The Immersion Tests*

Figure 6 shows the average corrosion rates of the alloys with continuously increasing Ga content calculated by Equation (2). In essence, the variation trend of the corrosion rate obtained by immersion test is in accord with the potentiodynamic polarization results. The lowest weight loss rate is 1.15 ± 0.1 mm/y for the Mg–5Sn–3Ga alloy. This confirms that the addition of Ga improves the corrosion resistance of Mg–5Sn alloy.

**Figure 6.** Average corrosion rate of the Mg–5Sn–*x*Ga alloys after immersion in 3.5 wt % NaCl solution for 72 h.

After immersion at 298 K for 6 h, the corrosion initiation sites were observed using SEM and EDS. Figure 7 presents the representative BSE micrographs of three alloys (Mg–5Sn, Mg–5Sn–0.5Ga, and Mg–5Sn–3Ga) and the corresponding EDS spectra of the marked areas. Figure 7a,b reveals that the Mg2Sn phase was not covered with corrosion films, and cracks were formed in the vicinity of Mg2Sn phase. Combining with the EDS analysis of points *A* and *B* (Figure 7d,e), it is inferred that corrosion began in the magnesium matrix around the Mg2Sn phase. Moreover, the adjacent distribution of the two types of phases provides the possibility for analyzing the initial corrosion location. In the Mg–5Sn–3Ga alloy (Figure 7c), the phase at point *C* is rich in Ga element compared with its neighboring

phase (marked by point *D*), suggesting that points *C* and *D* are Mg5Ga2 phase and Mg2Sn phase, respectively. Also shown in Figure 7c is that the surface film on Mg2Sn phase was broken, while this was not the case for the film on Mg5Ga2 phase. Mg2Sn phase was detached from the matrix after immersion for 6 h. All the aforementioned observations indicate that Mg2Sn phase, rather than the Mg5Ga2 phase, provided the sites to initiate corrosion. Moreover, compared with the larger Mg2Sn phase, eutectic Mg5Ga2 phase is deemed to stay more firmly in the alloy matrix. Therefore, with the increase in Ga content, the increasing area fraction of Mg5Ga2 further improved the corrosion resistance of the alloys.

**Figure 7.** High magnification BSE micrographs showing the alloy surfaces after immersion in 3.5 wt % NaCl solution at 298 K for 6 h: (**a**) Mg–5Sn, (**b**) Mg–5Sn–0.5Ga, (**c**) Mg–5Sn–3Ga; (**d**–**g**) energy-dispersive X-ray spectroscopy (EDS) spectra of the points as indicated by the arrows in the BSE micrographs.

Figure 8 shows the evolution of the surface morphologies of the investigated alloys after immersing for different durations. The surfaces of the Mg–5Sn–*x*Ga alloys are covered by a mixture of dark film and bright corrosion product. A number of deep corrosion pits are observed after immersion (Figure 8a). It is noticeable that the white corrosion products increase with the decrease in Ga content.

It is noted in Figure 8 that the Mg–5Sn–0.5Ga alloy was more seriously corroded than the Mg–5Sn–1Ga alloy, while the immersion tests produced an opposite result. The reason is that in the Mg–5Sn–0.5Ga alloy, there is no Ga-rich second phase, which has a marginal effect on initiating corrosion. In addition, Sn and Ga elements are more passive than Mg (the standard potentials of Mg, Sn, and Ga are −2.36 V, −0.14 V, and −0.55 V, respectively) [45]. The dissolved Ga element was expected to increase the potential of α-Mg, reducing the potential difference between Mg2Sn and the matrix. When the Ga content exceeded 0.5 wt %, Mg5Ga2 was formed and acted as the cathode phase to accelerate the corrosion rate.

**Figure 8.** The evolution of the surface morphology of the investigated alloys after soaking for different durations in 3.5 wt % NaCl solution: (**a**) Mg–5Sn, (**b**) Mg–5Sn–0.5Ga, (**c**) Mg–5Sn–1Ga, (**d**) Mg–5Sn–2Ga, and (**e**) Mg–5Sn–3Ga, respectively.

Based on the above analysis, at the initial stage of corrosion, the initiation site of Mg–5Sn–0.5Ga alloy is likely to cause corrosion. However, the corrosion initiation sites of Mg–5Sn–1Ga multiply with immersion time, promoting further corrosion. Therefore, in the case of immersion test, the Mg–5Sn–0.5Ga showed superior corrosion resistance than the Mg–5Sn–1Ga alloy.

Precipitation of Mg(OH)2, as the main corrosion product on the surface, occurs when Mg2<sup>+</sup> from anodic dissolution meets OH− from water reduction [37]. The XRD patterns (Figure 9) of the investigated alloys after immersion for 24 h evidence the presence of Mg(OH)2 on the surfaces, and the peak intensities of Mg(OH)2 reflections decrease with increasing Ga content. This indicates that the corrosion resistance of the Mg–5Sn–*x*Ga alloys was improved by the addition of Ga. As the addition amount of Ga was very small, the corrosion products containing Ga were not detected by XRD.

**Figure 9.** XRD patterns of the Mg–5Sn–*x*Ga alloys after immersion in 3.5 wt % NaCl solution for 24 h.

Figure 10 shows the BSE micrographs of the Mg–5Sn–*x*Ga alloys in cross-section after immersion in 3.5 wt % NaCl solution for 72 h. The samples have good uniformity. After many experiments, it was found that the corrosion depth of all three samples with the same composition showed little discrepancy. As given in Figure 10, corrosion started from the matrix near the second phases and penetrated the matrix along the area of the element-rich region. The thickness of the corrosion layer was the result of measuring the deepest corrosion in a sample. The inserts are an enlarged view of the deepest corrosion of the samples. Again, the thickness of the surface film decreases with increasing Ga content, indicating the positive effect of Ga on influencing the corrosion resistance of the alloys. Figure 10a reveals that the corrosion layer of the Mg–5Sn alloy was 99.1 μm in thickness. The mismatch between the matrix and corrosion product would induce local mechanical stress. Such stress increases with the gradual increasing thickness of the film after immersion [46], and once it reaches a critical value, the surface film will break. The aggressive solution will penetrate the substrate surface through the broken membrane. The relatively complete surface film can protect the substrate to some extent. As a result, the pits grow along a vertical direction.

According to the immersion test, the Mg–5Sn–0.5Ga alloy is more corrosion-resistant than the Mg–5Sn–1Ga alloy. However, this is not consistent with morphologic observations made in the cross-section. The surface film of the Mg–5Sn–0.5Ga alloy (54.7 μm) is thicker than that of the Mg–5Sn–1Ga (38.8 μm) alloy. The calculated mass loss estimates the average corrosion rate in a whole immersion period, while the corrosion section reflects the local corrosion level. It is significant to see that local corrosion occurs around the second phase along the interdendritic region to the interior of the matrix in Mg–5Sn alloy (Figure 10a). The only second phase in the Mg–5Sn–0.5Ga alloy is Mg2Sn; hence, the two alloys should behave following the same corrosion mode. However, the oxidation state of Ga is higher than Mg in the Mg(OH)2 lattice, which locally induces positive charge in the brucite lattice. Such an increase in positive charge is apt to be balanced by the interaction between Ga element and the harmful Cl− anions, slowing down the penetration of Cl− into the hydroxide layer [15].

To sum up, the second phases, Mg2Sn and Mg5Ga2, exhibit more positive corrosion potentials than the α-Mg matrix [42]. They act as the cathode, and α-Mg the anode, resulting in galvanic corrosion in 3.5 wt % NaCl solution. It is worth noting that Mg5Ga2 phase has a volta potential difference (VPD) vs. matrix around +75 mV, which will cause marginal local corrosion if the immersion time is not

long [19]. Therefore, the presence of Mg5Ga2 phase increases the area ratio between cathode and anode, which is not conducive to the corrosion resistance, but the local corrosion sensitivity does not increase. Nevertheless, a problem with Ga addition is the uneconomic cost of metallic gallium. Therefore, higher addition levels of Ga content were not performed in this study.

**Figure 10.** Cross-section morphologies of the Mg–5Sn–*x*Ga alloys after immersion in 3.5 wt % NaCl solution for 72 h: (**a**) Mg–5Sn, (**b**) Mg–5Sn–0.5Ga, (**c**) Mg–5Sn–1Ga, (**d**) Mg–5Sn–2Ga, (**e**) Mg–5Sn–3Ga; the inserts are high-magnification scanning electron microscope (SEM) micrographs of the alloys; (**f**) a typical SEM micrograph showing the cracks on the surface of the Mg–5Sn–3Ga alloy.

#### **4. Conclusions**

The main findings of this work are:


**Author Contributions:** Conceptualization, J.R., E.G., X.W., and Z.C.; Formal analysis, J.R.; Funding acquisition, E.G., H.K., Z.C., and T.W.; Investigation, J.R.; Methodology, X.W.; Project administration, X.W.; Resources, E.G., X.W., H.K., and T.W.; Supervision, H.K., Z.C., and T.W.; Writing—original draft, J.R.; Writing—review and editing, J.R., X.W., and Z.C.

**Funding:** This research was funded by the National Key Research and Development Program of China [grant No. 2016YFB0701203]; the National Natural Science Foundation of China [Grant Nos. 51525401, 51601028, 51774065, 51974058, 51728101, and 51690163]; and the fundamental research funds for the central universities [Nos. DUT18RC(3)042, DUT17RC(3)108].

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


#### *Materials* **2019**, *12*, 3686


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Experimental and Theoretical Studies on the Corrosion Inhibition of Carbon Steel by Two Indazole Derivatives in HCl Medium**

#### **Shenying Xu 1,2, Shengtao Zhang 1,\*, Lei Guo 3,\*, Li Feng <sup>1</sup> and Bochuan Tan <sup>1</sup>**


Received: 2 April 2019; Accepted: 23 April 2019; Published: 24 April 2019

**Abstract:** In this work, two indazole derivatives, namely 5-aminoindazole (AIA) and 5-nitroindazole (NIA), were investigated as corrosion inhibitors for carbon steel in 1 M HCl solution by experimental and density functional theory (DFT) methods. The electrochemical results indicate that the inhibition ability follows the order of AIA > NIA, which is due to the stronger electron-donating effect of –NH2 of the AIA group than the –NO2 group of NIA. Besides, the frontier orbital theory shows that the AIA exhibits higher reaction activity than NIA, and a more negative adsorption energy for AIA was also obtained, which is consistent with the analysis of the electrochemical measurements. We draw the conclusion that the electron-donating effect makes it easier for AIA to donate electrons to iron atoms to form a stronger protective layer than NIA.

**Keywords:** corrosion inhibitor; carbon steel; indazole derivatives; electrochemistry; DFT

#### **1. Introduction**

Carbon steels are extensively utilized as structural materials in plenty of industrial fields, especially in corrosive environments. In view of this, some protective strategies have been employed to protect steel from corrosion. The addition of corrosion inhibitors has been proved to be an easy and highly effective way to achieve this [1–3]. Organic compounds containing heteroatoms such as N, P, S or O have been used as excellent inhibitors due to their strong electron-donating ability [4–6]. Additionally, due to their high adsorption capacity, indazole derivatives have attracted the attention of several researchers [7,8]. Our previous work revealed that indazole can absorb onto copper surfaces and exhibits a favorable inhibition efficiency for copper corrosion in a 3.0 wt.% NaCl solution [7]. Furthermore, the influence of active adsorption, which centers on the inhibition effectiveness of IA-based inhibitors, was also researched in aggressive solutions [9]. Potentially, it is important to explore the effects of functional groups of indazole-based inhibitors on inhibition effectiveness. Herein, 5-aminoindazole (AIA) and 5-nitroindazole (NIA), as shown in Figure 1, were discovered for the first time to be effective corrosion inhibitors for carbon steel in hydrochloric acid (HCl) solution.

**Figure 1.** Molecular structures of indazole derivatives, (**a**) 5-aminoindazole; (**b**) 5-nitroindazole.

Specifically, this work focused on the corrosion inhibition of AIA and NIA for Q235 carbon steel in 1 M hydrochloric acid solution by using multiple techniques including electrochemical measurements, scanning electron microscopy (SEM) and DFT calculations. On account of the experimental and theoretical results, the difference of the inhibition mechanisms between AIA and NIA molecules was revealed, which would provide some reference functions to develop more efficient inhibitors for corrosion protection.

#### **2. Experimental Section**

#### *2.1. Material and Samples Preparation*

The tested inhibitors were AIA and NIA (Aladdin Company, Shanghai, China). The chemical composition of Q235 carbon steel (JIS G3101) was 0.17% C, 0.47% Mn, 0.26% Si, 0.017% S, 0.0048% P and Fe. The specimens for electrochemical measurements were sealed in ethoxyline resin with a 1 cm2 area exposed as working area. Prior to experiments, all specimens were polished by emery paper from 400 to 1200 grit. Then, samples were cleaned by ethanol and distilled water and dried at room temperature. Analytical reagent grade 37.5% water-diluted HCl was used as the corrosive medium (1 M HCl).

#### *2.2. Electrochemical Measurements*

A three-electrode cell was used to do the electrochemical experiments, and all tests were performed in a CHI660B CHI660B electrochemical workstation (Chinstruments, Shanghai, China). Q235 carbon steel, a platinum plate of 1.5 <sup>×</sup> 1.5 cm2, and a saturated calomel electrode (SCE) with a Luggin capillary were the working electrode, counter electrode, and reference electrode, respectively. To obtain a steady open circuit potential (OCP), the electrodes were immersed into the corrosive solution for one hour before tests. The potentiodynamic polarization curves were obtained with the potential range from <sup>−</sup>250 mV to 250 mV vs. OCP at the scan rate of 0.2 mV s<sup>−</sup>1. Electrochemical impedance spectroscopy (EIS) experiments were performed in a frequency range from 10−<sup>2</sup> Hz to 10<sup>5</sup> Hz with amplitude of 5 mV at OCP (−0.45 V to −0.49 V). Z-view software was employed to fit the experimental data by an appropriate equivalent circuit.

#### *2.3. Scanning Electron Microscopy*

The surface morphologies of Q235 carbon steel specimens were captured with scanning electron microscopy (SEM, Joel-6490LV, Tokyo, Japan) with an accelerating voltage of 20 kV. Before SEM characterization, the samples were immersed for six hours in 1 M HCl with and without AIA or NIA inhibitor.

#### *2.4. Theoretical Simulations*

A Gaussian 03W program was employed for quantum chemical calculations based on DFT. The geometry optimized structures of AIA and NIA were obtained through the use of the B3LYP function with a 6-311++G(d, p) basis set. During this process, no imaginary frequency was confirmed and the structures were in their lowest-energy state. Furthermore, DFT was used to calculate the Mulliken charge, dipole moment (μ) and frontier molecular orbitals including the energy of the highest occupied molecular orbital (*E*HOMO), the energy of the lowest unoccupied molecular orbital (*E*LUMO) and energy gap (Δ*E* = *E*LUMO − *E*HOMO).

The interactions between AIA or NIA and the Fe (110) surface were modeled in a simulation box (12.4 Å <sup>×</sup> 9.9 Å <sup>×</sup> 24.1 Å) with periodic boundary conditions by the Dmol3 program of Material Studio software (BIOVIA, USA). A 4-layer 5 × 5 supercell (the lower two layers were constrained) with 20 Å vacuum slab was used to simulate bulk metal. DFT calculations were treated within the generalized gradient approximation (GGA) function of Perdew–Burke–Ernzerhof (PBE) and the double numerical basis set with polarization functions on hydrogen atoms (DNP). DFT semi-core pseudopots (DSPPs) were used for Fe treatment. The displacement convergence, gradient, and tolerances of energy were 5 ×

10−<sup>3</sup> Å, 2 <sup>×</sup> 10−<sup>3</sup> Ha·Å<sup>−</sup>1, and 1 <sup>×</sup> 10−<sup>5</sup> Ha, respectively. The interaction energy (*E*Fe-inhibitor) between the Fe (110) surface and inhibitor obeyed [10–12]

$$E\_{\text{Fe-inhibrtor}} = E\_{\text{Total}} - E\_{\text{Fe}} - E\_{\text{inhibrtor}} \tag{1}$$

where *E*Total is the energy of Fe(110) surface and the adsorbed inhibitor molecule, *E*inhibitor is the energy of the isolated inhibitor molecule and *E*Fe is the energy of the steel surface, respectively.

#### **3. Results and Discussion**

#### *3.1. Electrochemical Impedance Spectroscopy (EIS) Measurements*

Figure 2 shows the electrochemical impedance spectra plots of AIA and NIA, respectively. In Figure 2a,c, it can be seen that the radius of the capacitive resistance arc increased with the growing concentration of inhibitors, which indicates that the protective layer was formed on the steel surface by the adsorption of AIA or NIA and that the corrosion inhibition efficiency of AIA is better than NIA. In addition, the impedance spectrum exhibits a squashed semicircle, which is caused by the formation of a protective layer on the steel surface. Figure 2b,d is Bode plots in the presence and absence of AIA and NIA, respectively. The impedance values and phase angle values increase with the growing concentration of inhibitors. In addition, a time constant can be found in the phase angle, usually due to the relaxation effect of the corrosion inhibitor molecule adsorption [2,13].

**Figure 2.** The plots of Nyquist and Bode for carbon steel in 1 M HCl with and without different concentrations of 5-aminoindazole (AIA) and 5-nitroindazole (NIA) at 298 K. AIA: (**a**,**b**); NIA: (**c**,**d**).

The equivalent circuit (Figure 3) is used to fit the impedance spectrum data, and the fitting data are shown in Table 1. In Figure 3, *R*s is the solution resistance, *R*ct is the charge transfer resistance, and CPE is a constant phase element; the impedance of the CPE is expressed as follows [14,15]:

$$Z\_{\rm CPE} = \frac{1}{\chi\_0 (j\omega)^n} \tag{2}$$

The double-layer capacitance (*C*dl) can be calculated from CPE parameter values *Y*<sup>0</sup> and *n* by the following expression [16]:

$$\mathbb{C}\_{\text{dl}} = \frac{\mathbb{Y}\_0 \omega^{n-1}}{\sin(\frac{u\pi}{2})} \tag{3}$$

where *Y*<sup>0</sup> is the CPE constant, *n* is the phase shift, which can be explained as a degree of surface inhomogeneity, *j* is the imaginary unit and ω is the angular frequency. The inhibition efficiency ηEIS can be expressed by the following equation [17–19]:

$$
\eta\_{\rm EIS} = (\frac{R\_{\rm ct} - R\_{\rm ct}^0}{R\_{\rm ct}}) \times 100 \tag{4}
$$

where *R*ct and *R*ct,0 are the charge transfer resistances with and without AIA and NIA, respectively. In Table 1, we can see that as the concentration of the corrosion inhibitor increases, the value of the *R*ct becomes larger, and the value of ηEIS also increases simultaneously. When the concentration of the inhibitor was 2 mM, the *R*ct values of AIA and NIA are 238 and 156.2 Ω cm2, respectively. According to the Helmholtz model formula, the value of double layer capacitance (*C*dl) can be expressed as [20,21]

$$\mathcal{C}\_{\text{dl}} = \frac{\varepsilon^0 \varepsilon}{d} \mathcal{S} \tag{5}$$

where ε<sup>0</sup> is the dielectric constant of air and ε is the local dielectric constant. *S* is the surface area of the working electrode, and *d* is the surface film thickness. Compared with water molecules, the molecular volume of AIA and NIA is significantly larger, and their dielectric constant is smaller than that of water molecules. Therefore, with increasing concentrations of AIA or NIA, the two investigated inhibitors replace the water molecules on the surface of carbon steel continuously, and the value of *C*dl decreases. Hence, the smaller the *C*dl, the denser the protective film formed on the surface of the carbon steel by AIA and NIA.

**Figure 3.** Electrical equivalent circuit used to fit the electrochemical impedance spectroscopy (EIS) experimental data. CPE: constant phase element.


**Table 1.** Impedance data for Q235 steel in 1 M HCl with various concentrations of NIA and AIA at 298 K.

#### *3.2. Potentiodynamic Polarization Measurements*

Figure 4 shows the electrodynamic polarization curves of carbon steel at 298 K in 1 M HCl solutions with various concentrations of NIA and AIA. The relevant data (in Table 2) are obtained by the extrapolation method. The inhibition efficiency η<sup>P</sup> is calculated by the following equation [22–24]:

$$
\eta\_{\rm P} = (\frac{i\_{\rm corr}^0 - i\_{\rm corr}}{i\_{\rm corr}^0}) \times 100 \tag{6}
$$

where *i* 0 corr and *i*corr are the corrosion current density in 1 M HCl solutions without and with the two investigated inhibitors, respectively.

**Figure 4.** Anodic and cathodic polarization curves for carbon steel in 1 M HCl with various concentrations of (**a**) AIA and (**b**) NIA at 298 K. SCE: saturated calomel electrode.

**Table 2.** Relevant parameters for Q235 steel in 1 M HCl solution in the absence and presence of different concentrations of AIA and NIA at 298 K from polarization curves.


It can be seen in Figure 4 that, compared with the values in blank solution, the corrosion potential (*E*corr) and corrosion current density (*I*corr) of the polarization curves change obviously with the addition of the two investigated inhibitors. With the addition of AIA and NIA, the *I*corr decreases and the *E*corr moves in a positive direction, illustrating that the corrosion reaction is effectively controlled. Clearly, the investigated inhibitors not only reduced the corrosion of the cathode but also reduced the corrosion of the anode. In addition, the shapes of polarization curves are parallel with the increase of concentration for the two investigated inhibitors, indicating that the action mechanism is same under different concentrations of inhibitors. Generally, it is considered as a cathodic inhibitor when the potential change exceeds 85 mV, while it is a mixed inhibitor when the potential change is less than 85 mV [25,26]. From Table 2, the changed values of *E*corr for the two inhibitors are less than 85 mV, suggesting that AIA and NIA are mixed corrosion inhibitors. The changed values of β<sup>a</sup> and β<sup>c</sup> also reflect the cathodic and anodic corrosion rates being retarded by the studied inhibitors. Furthermore, the inhibition efficiency (ηp) is also improved with increasing concentrations of the two inhibitors, and the inhibition ability follows the order AIA > NIA, which may be due to the stronger electron-donating effect of –NH2 than the –NO2 [27,28]. The electron-donating effect of the –NH2 makes the electron cloud density of the whole AIA molecule larger than the –NO2 of NIA, which leads to AIA finding it easier to give electrons to iron atoms and having a better protective effect than NIA.

#### *3.3. Morphology Analysis*

To obviously display the inhibition differences between AIA and NIA inhibitors, the morphologies of untreated and treated carbon steel were studied. The obtained images are shown in Figure 5. The surface of fresh carbon steel is smooth, while large holes and cracks with a size of 10 μm were observed after immersion in 1 M HCl (Figure 5a,b). The pitting morphology is formed due to the elimination rate of corrosion products slower than the reaction rate between Cl− and Fe [29,30]. Besides, some small holes and cracks still appeared on the steel surface even after 2 mM NIA was added into the HCl solution (Figure 5d). However, a flat surface was achieved with the addition of 2 mM AIA (Figure 5c); meanwhile the steel surface was covered with the absorbed AIA. Hence, the AIA shows prior inhibition performance than NIA. These results are in good agreement with electrochemical measurements.

**Figure 5.** SEM images of (**a**) fresh carbon steel and carbon steel immersed in 1 M HCl solution (**b**) without and with 2 mM (**c**) AIA or (**d**) NIA.

#### *3.4. Computational Study*

Computational simulation is an effective way to explain reaction mechanisms. In this work, DFT calculations were used to reveal the adsorption and inhibition performances of AIA and NIA molecules. By comparison, the acidity coefficient (p*K*a) of both AIA and NIA displayed higher values than the pH of 1M HCl medium, proving the existence of the protonated molecule. Specifically, there are two p*K*<sup>a</sup> (1.89 and 3.42) for AIA, indicating the existing form of AIA-2H<sup>+</sup> in HCl solution. Instead, there is only one p*K*<sup>a</sup> for NIA, indicating the existence of NIA-H+. The optimized geometry structure and frontier molecular orbitals (the highest occupied molecular orbital (HOMO) and the lowest unoccupied molecular orbital (LUMO)) of AIA-2H<sup>+</sup> and NIA-H<sup>+</sup> are shown in Figure 6, and Table 3 shows the energy of HOMO (*E*HOMO) and LUMO (*E*LUMO). It is generally known that HOMO is related to the ability of a molecule to donate electrons, and a higher *E*HOMO value shows a stronger electron-donating ability [31,32]. By contrast, LUMO is associated with the electron-accepting ability of a molecule, and a lower value of *E*LUMO represents a strong electron-accepting ability [33]. As can be seen from Figure 6c,d, the LUMO of both AIA-2H<sup>+</sup> and NIA-H<sup>+</sup> is distributed uniformly around the whole molecule. Conversely, in comparison with AIA-2H<sup>+</sup> (Figure 6f), the HOMO of NIA-H<sup>+</sup> was mainly delocalized around the nitro-substituent (Figure 6e), indicating that the electron-donating ability of NIA mainly comes from the nitro-substituent. According to the frontier orbitals, both AIA and NIA tend to interact with the steel surface through the π bond of rings in the way of parallel adsorption configuration.

**Figure 6.** (**a**,**b**) Optimized geometric structures, (**c**,**d**) LUMO orbitals and (**e**,**f**) HOMO orbitals of AIA and NIA inhibitors.

**Table 3.** Quantum chemical parameters for AIA and NIA by using the B3LYP/6-311 + + G(d,p) method.


In addition, the HOMO–LUMO gap (Δ*E*) is an important parameter to evaluate the stability of inhibitors, and the lower value of Δ*E* indicates that the inhibitor molecule could more easily adsorb on the metal surface [34,35]. As shown in Table 3, both AIA-2H<sup>+</sup> and NIA-H<sup>+</sup> have lower values of Δ*E* (3.4 eV and 4.6 eV, respectively), resulting in their strong ability to accept electrons from the d-orbital of steel as well as the high stability of the [Fe-inhibitor] complexes; namely, the AIA exhibited higher reaction activity than NIA. At the same time, the dipole–dipole (μ) interaction between the inhibitor and metal surface could improve the inhibition efficiency [36]. Herein, the fact that μNIA is about three times μAIA indicates that AIA exhibits more appropriate adsorption between the AIA and metal surface than NIA.

The ionization potential (*I* = −*E*HOMO) and electron affinity (*A* = −*E*LUMO) could be used to derive the electronegativity (χ) and global hardness (γ). The fraction of the electron transfer (Δ*N*) between the inhibitor molecules and Fe surface is given by following equation [37–39]:

$$
\Delta N = \frac{\chi\_{\text{Fe}} - \chi\_{\text{inh}}}{2(\chi\_{\text{Fe}} + \chi\_{\text{inh}})} \tag{7}
$$

where χFe and γFe are the absolute electronegativity and hardness of the Fe atom; and χ*i*nh and γinh are the absolute electronegativity and hardness of the inhibitor molecules. A theoretical χFe value of bulk Fe is 7 eV/mol, whereas γFe is almost zero. χ*i*nh and γinh are related to *I* and *A* [40,41].

$$\chi = \frac{I + A}{2} \tag{8}$$

$$\gamma = \frac{I - A}{2} \tag{9}$$

The direction of electron transfer is manifested by positive or negative Δ*N* values [42]. From Table 3, both AIA and NIA are electron acceptors. It is noteworthy that the magnitude of Δ*N s* absolute value is not connected with inhibition efficiency.

AIA-2H<sup>+</sup> and NIA-H<sup>+</sup> were placed in a simulation box parallel with or perpendicular to the Fe(110) surface. The simulation results showed that both AIA-2H<sup>+</sup> and NIA-H<sup>+</sup> tended to adsorb in parallel on the Fe(110) surface, as shown in Figure 7. Namely, the indazole and aromatic rings were the adsorption sites, which was in agreement with previous reports. Besides, all the hydrogen atoms upturning after adsorption may be due to the hybridization between Fe and heavy atoms. The AIA molecule is possibly a more efficient inhibitor because of its more negative adsorption energy (−4.65 eV) than NIA (−4.05 eV). This is consistent with the analysis of the electrochemical measurements.

**Figure 7.** Stable adsorption configurations (side and top view) of (**a**) AIA-2H<sup>+</sup> and (**b**) NIA-H<sup>+</sup> molecules on the Fe(110) surface.

Figure 8 shows the projected density states of AIA-2H<sup>+</sup> and NIA-H<sup>+</sup> before and after adsorbing on the Fe(110) surface. By comparing these with the isolate inhibitors, the p orbitals of the adsorbed inhibitors almost disapear, revealing the strong interaction between AIA or NIA and the Fe(110) surface [43]. This is consistant with the inhibition efficencies obtained by experiments.

**Figure 8.** Density states projected of (**a**,**c**) AIA-2H<sup>+</sup> and (**b**,**d**) NIA-H<sup>+</sup> molecules before and after adsorbing on the Fe(110) surface.

#### **4. Conclusions**

In this study, two indazole derivatives, AIA and NIA, were proved to be excellent corrosion inhibitors for carbon steel in 1 M HCl. The inhibition performance was tested by electrochemical methods. Theoretical calculations were also performed to reveal the inhibition mechanism of AIA and NIA. The detailed results are as follows:


**Author Contributions:** S.Z. and L.G. proposed the concept and were involved in the design of the experiments. S.X. and L.F. performed the experimental work and wrote the main manuscript text. S.X. and B.T. evaluated the inhibition performance using theoretical calculations. All authors were involved in the drafting, revision and approval of the manuscript.

**Funding:** This research was supported by National Natural Science Foundation of China (21706195, 21878029), Science and Technology Program of Guizhou Province (QKHJC2016-1149), Guizhou Provincial Department of Education Fundation (QJHKYZ2016-105).

**Conflicts of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Mannich Base as Corrosion Inhibitors for N80 Steel in a CO2 Saturated Solution Containing 3 wt % NaCl**

**Mingjin Tang 1,\*, Jianbo Li 1,\*, Zhida Li 2, Luoping Fu 1, Bo Zeng <sup>1</sup> and Jie Lv <sup>1</sup>**


Received: 14 December 2018; Accepted: 30 January 2019; Published: 1 February 2019

**Abstract:** In this paper, a corrosion inhibitor containing nitrogen atoms and a conjugated π bond was synthesised, and its final product synthesised by the optimal conditions of the orthogonal test results is named multi-mannich base (MBT). The corrosion inhibition effect on the N80 steel sheet of the corrosion inhibitor was evaluated in a CO2 saturated solution containing 3 wt % NaCl; the corrosion rate was 0.0446 mm/a and the corrosion inhibition rate was 90.4%. Through electrochemical and adsorption theory study, MBT is a mixed corrosion inhibitor that mainly shows cathode suppression capacity. The adsorption of MBT on the surface of the steel sheet follows the Langmuir adsorption isotherm; it can be spontaneously adsorbed on the surface of the N80 steel sheet, which has a good corrosion inhibition effect. The surface of the N80 steel sheet was microscopically characterised by atomic force microscope (AFM). It can be seen from the results that the N80 steel sheet with MBT added is significantly different from the blank control group; the surface of the steel sheet is relatively smooth, indicating that MBT forms an effective protective film on the surface of N80 steel, which inhibits the steel sheet.

**Keywords:** corrosion inhibitor; electrochemical; AFM; CO2 corrosion

#### **1. Introduction**

To improve oil and gas recovery, CO2 injection is used in the exploitation of oil, natural gas, coalbed methane and shale gas [1]. However, CO2 injection is accompanied by CO2 corrosion [2]. Corrosion will decrease the mining yield, resulting in oil wells failure, shutting down and even cause safety accidents, which will have a serious impact on the development of the oilfield and the economic benefits of the market [3–6]. The most common and effective way to control corrosion is to use chemical additives that reduce corrosion at very small dosages [7,8]. In recent years, various corrosion inhibitors have been successfully applied to the industry, and the corrosion has been controlled obviously and effectively [9,10].

So far, phosphorus and nitrogen-containing compounds have been widely used as corrosion inhibitors, including organic phosphorus, nitrogen-containing compounds and phosphorus-containing organic water-soluble polymers [11]. Phosphorus-containing corrosion inhibitors have excellent corrosion inhibition properties. However, they have some fatal defects; for example, phosphorus compounds are difficult to degrade in water and usually lead to eutrophication. The widespread use of this chemical may cause oxygen deficiency in the water, which leads to the death of aquatic organisms [12]. Therefore, in the last few years, the industry has not only required the effectiveness of chemical compounds but also their safety [13]. Today, chemical emissions are strictly controlled through legislation. Therefore, finding an alternative solution, namely the effective control of corrosion by green corrosion inhibitors, is the most important research direction at the moment. Chemical products identified as "green" are based on criteria: non-toxic and degradable [14,15].

In this paper, salicylaldehyde, diethylenetriamine, formaldehyde and acetone were selected to synthesise corrosion inhibitor MBT containing nitrogen atoms with lone pairs and a conjugated π bond. The structure was characterised by infrared radiation spectroscopy (IR) and nuclear magnetic resonance spectroscopy (1H NMR). The product was evaluated for corrosion inhibition by a weight loss method. Finally, the mechanism of corrosion inhibition was studied by atomic force microscopy (AFM) and electrochemical methods.

#### **2. Experiment**

#### *2.1. Main Experimental Materials and Instruments*

Salicylaldehyde C7H6O2, diethylenetriamine C4H13N3, ethanol C2H6O, acetone C3H6O, formaldehyde HCHO: AR, Chengdu Kelong Chemical Reagent Factory, Chengdu, China.

WQF-520 infrared spectrometer: Beijing Ruili Analytical Instrument (Group) Co. Ltd., Beijing, China; Bruker AVANCE III HD 400 nuclear magnetic resonance spectrometer: Bruker Corporation, Billerica, MA, USA; IVICMSTAT electrochemical workstation: Ivium Technologies, Eindhoven, The Netherlands; SPM-9600 atomic force microscope: Shimadzu Corporation, Kyoto, Japan.

#### *2.2. Synthesis of Corrosion Inhibitor*

First, in a three-necked flask equipped with a condenser and a thermometer, 20 mL of ethanol was added as a solvent. When the reactant reached a certain temperature, a certain amount of salicylaldehyde and diethylenetriamine were added in a certain ratio. After a period of reaction, the reaction was terminated to give the product intermediate mannich base (MB-1).

Then, a certain amount of acetone and formaldehyde were added in order, being kept at a constant reaction temperature for a certain period, and then the reaction was terminated to obtain the target product mannich base (MB-2). The synthesis of MB-1 and MB-2 is shown in Schemes 1 and 2.

**Scheme 2.** The synthesis of MB-2.

#### *2.3. Weight Loss Measurements*

A hanging N80 steel test piece (50.00 mm × 13.00 mm × 1.50 mm: length, width, height) was ground with sand paper to give a surface with a uniform metal luster [16]. To calculate the superficial area of the test piece, the carbon steel was deoiled with putty powder and then immersed in ethanol for 5 min, dried with an air blower, wrapped in filter paper, maintained in a dryer and weighed to 0.0001 g.

Then the test piece was immersed in a bottle containing 500 mL of CO2 saturated solution and 3 wt % NaCl saline water with and without inhibitors at 70 ◦C for 72 h. According to the weight loss of each test piece and Equations (1) and (2), the rate of corrosion of the test piece and efficiency of the corrosion inhibitor at different concentrations could be calculated [4].

*<sup>v</sup>* <sup>=</sup> 8.76 <sup>×</sup> <sup>10</sup><sup>4</sup> <sup>×</sup> <sup>Δ</sup>*<sup>m</sup> <sup>ρ</sup>At* , (1)

where *ν*, Δ*m*, *ρ*, *A* and *t* are the corrosion rate (mm/a), weight loss (g), test sample density (g/cm3), exposed sample area (cm2) and immersion time (h), respectively. The inhibition efficiency (*η*) of the inhibitor was calculated by Equation (2):

$$
\eta\left(\%\right) = \left(1 - \frac{v}{v^0}\right) \times 100,\tag{2}
$$

where *ν* and *ν*<sup>0</sup> are the corrosion rates calculated from the weight loss with and without the inhibitor.

#### *2.4. Electrochemical Measurements*

The potentiodynamic polarisation curve method was used to measure the Tafel polarisation curve. When the self-corrosion potential (Ec) of the system was stable, the cathode and anode scanning was from 250 mV to 300 mV. The polarisation curves of the N80 steel sheets with different concentrations of corrosion inhibitors were measured [4]. According to the relevant theory of electrochemistry, the corrosion inhibition rate (*η*) of the corrosion process can be calculated by Equation (3) [17]:

$$
\eta = \frac{i^0 - i}{i^0} \times 100,\tag{3}
$$

where *i* <sup>0</sup> and *i* are the corrosion current densities without and with the inhibitor, respectively.

The electrochemical impedance spectroscopy (EIS, Ivium Technologies, Eindhoven, The Netherlands) experiments were performed by applying a sinusoidal voltage signal of 10 mV; the frequency range was 0.01–10−<sup>5</sup> Hz; and the alternating current excitation signal amplitude was 10 mV. Impedance diagrams of the N80 steel sheets with different concentrations of corrosion inhibitors were measured. The inhibition efficiency (*η*) was determined from the EIS data using Equation (4):

$$
\eta = \frac{R\_{\text{t}} - R\_{\text{t}}^0}{R\_{\text{t}}} \times 100,
\tag{4}
$$

where *R*<sup>t</sup> and *R*<sup>t</sup> <sup>0</sup> are the charge transfer resistance in the absence and presence of inhibitors, respectively.

#### *2.5. AFM Analysis*

Atomic force microscopy (AFM, Shimadzu Corporation, Kyoto, Japan) was used for characterizing the surface morphology and measuring surface roughness in a CO2 saturated solution containing 3 wt % NaCl without and with corrosion inhibitors. The two- and three-dimensional AFM images of steel specimens were taken in the range from 0 to 5 μm at room temperature [18].

#### **3. Results and Discussion**

#### *3.1. IR Characterisation Results*

The IR spectrum of MB-2 is shown in Figure 1. It is demonstrated that the stretching vibration peak attributed to the carbonyl group in salicylaldehyde does not appear at 1740 cm−<sup>1</sup> to 1720 cm−1. The stretching vibration peak attributed to the C=N bond at 1644 cm−<sup>1</sup> indicates that the salicylaldehyde reacts with diethylenetriamine to form a compound containing C=N and the aldehyde group disappears. The absorption peak at 3457 cm−<sup>1</sup> and 1270 cm−<sup>1</sup> is attributed to the stretching vibration of the O–H bond and C–O bond of phenol; the absorption peak at 2989 cm−<sup>1</sup> and 2923 cm−<sup>1</sup> is attributed to the stretching vibration of methyl (CH3) and methylene (CH2); the bending vibration of the N–H bond belonging to the secondary amine did not occur at 1580 cm−<sup>1</sup> to 1490 cm<sup>−</sup>1, indicating that the hydrogen on the secondary amine reacted with acetone and formaldehyde, and the secondary amine disappeared into a tertiary amine; the absorption peak at 1718 cm−<sup>1</sup> is attributed to the stretching vibration of C=O. IR characterization results showed that the synthesized MB-2 was the target product.

**Figure 1.** IR spectrum of MB-2.

#### *3.2. Optimization of MB-2 Synthesis Conditions*

Through orthogonal experiments, the three-factor four-level L16(43) orthogonal table was selected to investigate the influencing factors of MB-2 synthesis, including the ratio of reactants, reaction time and reaction temperature. Considering that both methyl groups of acetone react, the amount of acetone is fixed to adjust the ratio of MB-2 to formaldehyde. According to the product synthesised by the orthogonal experiment, when the amount was 200 mg/L, the corrosion inhibition performance was evaluated in a CO2 saturated solution containing 3 wt % NaCl.

The experimental factors and levels are shown in Table 1. The results of orthogonal experiments are shown in Table 2. It can be seen from Table 2 that the optimal reaction conditions for the synthesis of MB-2 are: reaction temperature 85 ◦C; reactant ration(MB-2):n(formaldehyde):n(acetone) = 2:2:1; and reaction time 6 h. The final product synthesised by the optimal conditions of the orthogonal test results is named MBT, and the reaction equation for the synthesis is shown in Scheme 3.


**Table 1.** Experimental factors and levels.


**Table 2.** Orthogonal experimental results.

**Scheme 3.** The synthesis of MBT.

#### *3.3. 1H NMR Analysis*

The 1H NMR spectrum of MBT is shown in Figure 2. It can be seen from Figure 2 that the absorption peak at chemical shift *δ* = 12.00~11.00 is attributed to –OH; the absorption peak at *δ* = 8.56~8.45 is attributed to –CH=N; the absorption peak at *δ* = 7.36~6.79 is attributed to benzene (–CH); *δ* = 3.68~3.65, 3.02, the absorption peak at 2.63 is attributed to –CH2 at different positions; and the absorption peak at *δ* = 2.5 belongs to the solvent peak of Dimethyl sulfoxide (DMSO) and is

characterized by IR. Combined with IR characterisation, the results indicate that MBT is the target product of synthesis.

**Figure 2.** 1H NMR spectrum of MBT.

#### *3.4. Concentration Effect of MBT on Corrosion Inhibition*

The corrosion inhibition performance of the corrosion inhibitor MBT was evaluated by the static weight-loss method. Test conditions: temperature 70 ◦C, pH = 5.6, ρ(NaCl) = 30 g/L, and constant temperature water bath 72 h. The corrosion inhibition effect of MBT on a N80 steel sheet in a CO2 saturated solution containing 3 wt % NaCl was investigated under different dosages. The result is shown in Figure 3.

**Figure 3.** Effect of the dosage of MBT on the corrosion rate and corrosion inhibition rate of the N80 steel sheet.

According to Figure 3, in the case of different dosages of MBT (0, 50, 100, 200, 300, 400, 500 mg/L), when the amount of MBT was 200 mg/L, the N80 steel sheet corrosion rate was 0.0579 mm/a. The corrosion rate of the steel sheet decreases with the increase of the corrosion inhibitor concentration, indicating that MBT can form an effective protective film on the metal surface. When the amount of corrosion inhibitor is 400 mg/L, the corrosion rate is 0.0446 mm/a, and the corrosion inhibition rate can reach 90.4%.

#### *3.5. Temperature Effect of MBT on Corrosion Inhibition*

In order to study the corrosion inhibitor for a CO2 saturated solution containing 3 wt % NaCl, the corrosion inhibition effect at different temperatures (40, 50, 60, 70, 80, 90 ◦C), the dosage of corrosion inhibitor in this experiment is 400 mg/L, and the evaluation time 72 h.

As shown in Figure 4, the increasing temperature increases the corrosion rate in the absence (rcoor0) and appearance of MBT solution (rcoor1). This is because as the temperature increases, the adsorption capacity of the corrosion inhibitor decreases, the desorption capacity increases, and the corrosion rate of the steel sheet itself increases as the temperature is higher.

**Figure 4.** Evaluation of temperature on corrosion rate and corrosion inhibition rate.

#### *3.6. Discussion on Corrosion Inhibition Mechanism*

#### 3.6.1. Polarization Curve Data Analysis

The inhibition mechanism of MBT was studied by the Tafel polarization curve method. Different concentrations (0 mg/L, 50 mg/L, 100 mg/L, 200 mg/L, 300 mg/L) of MBT were added, and the corrosive medium was a CO2 saturated solution containing 3 wt % NaCl at a temperature of 25 ◦C. The Tafel curve is shown in Figure 5.

It is shown in Figure 5 that when the different concentrations of MBT are added, the polarization curve moves downward as a whole, that is, the corrosion current density decreases. As the concentration of the corrosion inhibitor increases, the corrosion current density decreases. Since the corrosion current density is proportional to the corrosion rate, the concentration of the MBT increase made the corrosion rate drop, which is consistent with the results of the weight loss experiment. It can be seen from Figure 5, that after the addition of MBT, the polarization curves of the cathode and anode move in the direction of low corrosion current density, so MBT reacts to both the cathode and anode.

**Figure 5.** Tafel curve of N80 steel with different concentrations of MBT.

The corrosion potential (Ecorr), corrosion current density (Icorr), Tafel slope (βa and βc) and corrosion inhibition rate (*η*) calculated by the equation are shown in Table 3. It can be seen in the table that the corrosion inhibitor MBT moves the self-corrosion potential of the N80 steel sheet negatively; it shows that the corrosion inhibitor blocks the cathode process, and the self-corrosion potential moves in the negative direction (fluctuating around 30 mV) [19,20], so MBT is a mixed type corrosion inhibitor mainly for suppressing the cathode.


**Table 3.** Electrochemical parameters of different MBT concentrations.

#### 3.6.2. Electrochemical Impedance Spectroscopy Data Analysis

In order to further study the corrosion inhibition mechanism of MBT on steel sheets, the impedance of the system was measured by an AC impedance method. Figure 6 is an impedance spectrum of MBT with different concentrations added.

It can be seen from Figure 6 that the Nyquist diagram of the blank sample and the added corrosion inhibitor is a set of capacitive reactance arcs with a semicircular shape; however, the impedance of the system changes significantly after adding different concentrations of MBT compared with the blank. In addition, the capacitive anti-arc increases with the increase of corrosion inhibitor concentration, and the Bode-modulus value and the phase angle Bode-phase angle increase obviously, indicating that the corrosion inhibitor can be adsorbed on the surface of the N80 steel sheet to increase the transmission resistance to corrosion of the steel sheet, thereby reducing the corrosion rate of the metal. The EIS diagram was fitted using ZSimpWin software (ZSimDemo3.30d), and the equivalent circuit was obtained as R(C(R(QR))), as shown in Figure 7.

In Figure 7, RS represents the solution resistance; Rct represents the transfer resistance of the N80 steel sheet corrosion reaction charge; Rt and CPE represent the metal interface film resistance and constant phase element (CPE) to replace a double layer capacitance with a more accurate fit [21], the impedance of CPE was calculated following Equation (5); and Cdl represents the capacitance. The fitting parameters are shown in Table 4.

$$\mathcal{Z}\_{\rm CPE} = \frac{1}{\Upsilon\_0 j w^{\rm n}},\tag{5}$$

where Y0 is CPE constant; *j* is the imaginary number; *w* is the angular frequency; and *n* is the phase shift (which represents the deviation from ideal behavior).

**Figure 6.** Nyquist (**a**), Bode-modulus (**b**) and Bode-phase angle (**c**) diagrams of N80 steel sheets with different concentrations of MBT.

**Figure 7.** Fitting equivalent circuit diagram.

From the data analysis in Table 4, as the intrusion time of N80 steel increases, the charge transfer resistance Rct and film resistance Rt in the different corrosion inhibitor concentration solutions increase with the concentration of the corrosion inhibitor; the compactness of Rt and MBT adsorption film formation is related to the thickness, which indirectly indicates that the MBT molecule is effectively adsorbed by the N80 steel sheet. The solution resistance Rs does not change much and is significantly smaller than the film resistance Rt and charge transfer resistance Rct. Combined with the analysis of the Nyquist diagram, it is fully demonstrated that the corrosion inhibition effect is mainly absorbed by the corrosion inhibitor molecules on the metal surface, which changes the interface properties of the N80 steel surface and the electric double layer structure, thus achieving a good corrosion inhibition

effect [22]. The calculated corrosion inhibition effect is basically consistent with the results obtained by the above weight loss experiment and the polarization curve method.


**Table 4.** Electrochemical impedance spectroscopy fitting parameters of different concentrations of MBT.

#### *3.7. MBT Adsorption Mode*

The MBT molecule is adsorbed on the surface of the metal to form a dense adsorption film, which acts as a corrosion inhibitor, and the stronger the adsorption capacity of the molecule on the metal surface, the better the corrosion inhibition effect [23]. The adsorption of MBT molecules on the metal surface is based on the structure, spatial distribution of the groups in the MBT molecule and the surface morphology of the metal; the adsorption modes are Temkin, Langmuir, Frumkin and Freundkick.

The above four adsorption isotherms were used to fit the data obtained by the polarisation curve method and the impedance method. It was found to be in accordance with the Langmuir adsorption equation, and the Langmuir adsorption curve is shown in Figure 8.

**Figure 8.** Langmuir isothermal adsorption curve of corrosion inhibitor MBT.

From Figure 8, it can be concluded that Kads = 1.701 × 104 L/mol and R<sup>2</sup> = 0.998, indicating that C/θ has a relatively good linear relationship with concentration C. The results show that at 70 ◦C, in a CO2 saturated solution containing 3 wt % NaCl, the adsorption of MBT on the surface of the N80 steel sheet follows the Langmuir isotherm adsorption. The formula for the adsorption free energy ΔGads of the corrosion inhibitor MBT is as follows:

$$
\Delta \mathbf{G}\_{\rm ads} = -\text{RTln}(55.5 \mathbf{K}\_{\rm ads}),
\tag{6}
$$

Calculate ΔGads = −39.25KJ/mol < 0 according to the adsorption equilibrium constant Kads and Equation (6), indicating that the corrosion inhibitor MBT is capable of spontaneous chemical adsorption on the N80 steel sheet, which has a good corrosion inhibition effect.

#### *3.8. Corrosion Inhibition Surface Morphology Analysis*

In order to study the corrosion inhibition ability of the corrosion inhibitor MBT, the surface of the N80 steel sheet was characterised by AFM. The N80 steel sheet was immersed with different concentrations of corrosion inhibitor MBT in a CO2 saturated solution containing 3 wt % NaCl, and the water bath was kept at 70 ◦C for 72 h. The three-dimensional and planar topography of the N80 steel sheet surface is shown in Figure 9.

**Figure 9.** AFM three-dimensional image of N80 steel sheet: (**a**) blank sample; (**b**) initial uncorroded sample; (**c**–**f**) added 50 mg/L, 100 mg /L, 200 mg/L, 300 mg/L MBT.

Figure 9a shows the N80 steel sheet without MBT, it can be seen from the figure that the surface corrosion of the N80 steel sheet is serious and the roughness is high because the metal is strongly damaged in the etching solution; Figure 9b is the polished N80 steel sheet that has not been immersed in the etching solution, it can be seen from the figure that the surface of the steel sheet is relatively smooth and the roughness is very low. Based on Figures 9c–f and 10c–f, the N80 steel sheet added to the corrosion inhibitor MBT is significantly different from that of Figures 9a and 10a. As the concentration of the corrosion inhibitor MBT increases, the corrosion rate of the steel sheet reduces, the surface of the steel sheet is smoother and the roughness is also closer to that of Figure 9b. MBT at a concentration of 300 mg/L is exhibited in Figures 9f and 10f with sandpaper polishing marks. It shows that the surface of MBT carbon steel forms an effective protective film, which prevents the corrosion of the steel sheet.

**Figure 10.** AFM plane topography of N80 steel sheet: (**a**) blank sample; (**b**) initial uncorroded sample; (**c**–**f**) is added 50 mg/L, 100 mg/L, 200 mg/L, and 300 mg/L MBT.

By analysing the data of the atomic force microscope carried by the analysis software (NanoScope Analysis v1.40), we can obtain the mean square roughness (Rq), the average roughness (Ra) and the maximum height difference (Rmax) of the surface of the N80 steel sheet, which is shown in Table 5. It can be seen from the table, that after the N80 blank is etched, the Rq is 155.0 nm, the Ra is 113.0 nm and the Rmax is 2050.0 nm, which is quite different from the Rmax of 165.0 nm which has not been etched. After adding 300 mg/L corrosion inhibitor MBT, the Rmax of the N80 steel sheet is 192.0 nm, which is very close to the maximum height difference of the unetched N80 steel sheet. Combined with Figures 9 and 10, the surface roughness of the steel sheet can be seen. The appearance of the corrosion inhibitor forms a passivation film on the surface of the N80 steel sheet, which reduces the surface roughness of the steel sheet and effectively prevents the N80 steel sheet from being eroded by corrosive media.


**Table 5.** AFM parameters after soaking in different concentrations of corrosion inhibitor MBT.

Note: CM (corrosive media) refers to a CO2 saturated solution containing 3 wt % NaCl.

#### **4. Conclusions**

The overall aim of this study was to synthesise and assess the corrosion inhibition effect of MBT. After dosage temperature tests, the corrosion rate was 0.0446 mm/a and the corrosion inhibition rate was 90.4% while the dosage was 400 mg/L in a CO2 saturated solution of 3 wt % NaCl, in a 70 ◦C constant temperature water bath for 72 h. As evidenced by polarisation curve analysis, MBT is a mixed type corrosion inhibitor mainly used for suppressing the cathode. The adsorption of MBT on the metal surface contributes to a dense molecular film, changing the interface properties and electric double layer structure, thus achieving a good corrosion inhibition effect. The adsorption follows the Langmuir isotherm adsorption and the calculated ΔGads < 0 indicates that the chemical adsorption of the corrosion inhibitor is spontaneous on the N80 steel sheet. AFM is applied in microscopic characterisation on the surface. Results demonstrate that the N80 steel sheet added the corrosion inhibitor MBT is different from blank the sample, showing a smooth surface, indicating that MBT contributes to the protective film with a good corrosion inhibition effect on the surface of the steel sheet.

**Author Contributions:** Conceptualization, J.L. (Jianbo Li) and M.T.; Data curation, M.T. and L.F.; Funding acquisition, J.L. (Jianbo Li); Investigation, J.L. (Jianbo Li), M.T., Z.L., L.F., B.Z. and J.L. (Jie Lv); Writing—original draft, M.T. and Z.L.; Writing—review & editing, M.T., J.L. (Jianbo Li) and Z.L.

**Funding:** This research was funded by the National Science and Technology Major Project of China, grant number 2016ZX05016-004-008 and The APC was funded by the National Science and Technology Major Project of China.

**Acknowledgments:** The authors thank the Southwest Petroleum University, Institute, Chengdu, China, for providing the research facilities needed for the above study. This work was supported by the National Science and Technology Major Project of China (No. 2016ZX05016-004-008).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Corrosion Inhibition Mechanism and E**ffi**ciency Di**ff**erentiation of Dihydroxybenzene Isomers Towards Aluminum Alloy 5754 in Alkaline Media**

**Jacek Ryl 1,\*, Mateusz Brodowski 1,2, Marcin Kowalski 1,2, Wiktoria Lipinska 1,3, Pawel Niedzialkowski <sup>4</sup> and Joanna Wysocka <sup>1</sup>**


Received: 10 August 2019; Accepted: 19 September 2019; Published: 20 September 2019

**Abstract:** The selection of efficient corrosion inhibitors requires detailed knowledge regarding the interaction mechanism, which depends on the type and amount of functional groups within the inhibitor molecule. The position of functional groups between different isomers is often overlooked, but is no less important, since factors like steric hinderance may significantly affect the adsorption mechanism. In this study, we have presented how different dihydroxybenzene isomers interact with aluminum alloy 5754 surface, reducing its corrosion rate in bicarbonate buffer (pH = 11). We show that the highest inhibition efficiency among tested compounds belongs to catechol at 10 mM concentration, although the differences were moderate. Utilization of novel impedance approach to adsorption isotherm determination made it possible to confirm that while resorcinol chemisorbs on aluminum surface, catechol and quinol follows the ligand exchange model of adsorption. Unlike catechol and quinol, the protection mechanism of resorcinol is bound to interaction with insoluble aluminum corrosion products layer and was only found efficient at concentration of 100 mM (98.7%). The aforementioned studies were confirmed with Scanning Electron Microscopy and X-ray Photoelectron Spectroscopy analyses. There is a significant increase in the corrosion resistance offered by catechol at 10 mM after 24 h exposure in electrolyte: from 63 to 98%, with only negligible changes in inhibitor efficiency observed for resorcinol at the same time. However, in the case of resorcinol a change in electrolyte color was observed. We have revealed that the differentiating factor is the keto-enol tautomerism. The Nuclear Magnetic Resonance (NMR) studies of resorcinol indicate the keto form in structure in presence of NaOH, while the chemical structure of catechol does not change significantly in alkaline environment.

**Keywords:** aluminum alloy; corrosion inhibitor; alkaline environment; impedance analysis; adsorption; dihydroxybenzene

#### **1. Introduction**

Aluminum is the most widely spread metallic element on Earth [1], owing to its unique physico-chemical properties, such as low weight, high thermal and electrical conductivity, high linear expansion coefficient, and good corrosion resistance. It is non-magnetic, non-toxic and may be subjected to repeated recycling [2]. Aluminum and its alloys have been used in almost all industries, in particular in mechanical engineering, the defense industry, aviation, shipbuilding, food and chemical industries, and many others. It is a strategic resource whose consumption is a measure of countries' development and industrialization level. Over the past 50 years, the world production of aluminum has been constantly increasing, with the highest leap occurring in this millennium. Limited access to bauxite ores and their gradual depletion are the main obstacles in the further development of the aluminum industry. Constantly growing demand and utilization of aluminum requires more effective methods for its recycling or protection.

Aluminum and its alloys belong to the group of passivating metals, alongside titanium, chromium and high-alloy steels. Spontaneously formed native oxide layer on the aluminum surface is built primarily of aluminum oxide Al2O3. The layer is thermodynamically stable in the pH range between 4 and 9, where aluminum possesses the highest corrosion resistance. On the other hand, the passive layer is not thermodynamically stable in alkaline and acidic media. In the presence of hydroxyl ions, aluminum undergoes oxidation to form Al(OH)4 −, according to the mechanism proposed by Macdonald [3]. One of the most commonly accepted aluminum corrosion mechanisms in alkaline media may be simplified to the form of equations (1–4) [4]. Other, more detailed mechanisms suggests involvement of intermediate products and/or passive layer [5–10].

$$\text{Al} \rightarrow \text{Al}^{3+} + \text{3e}^- \tag{1}$$

$$\rm Al^{3+} + 3OH^{-} \rightarrow Al(OH)\_{3} \tag{2}$$

$$\text{Al(OH)}\_{3} \downarrow + \text{OH}^{-} \to \text{Al(OH)}\_{4} \text{{}^{\cdot}\text{}\tag{3}$$

The anodic dissolution is accompanied by cathodic process of water electrolysis with hydrogen generation, according to Equation (4):

$$2\text{H}\_2\text{O} + 2\text{e}^- \rightarrow \text{H}\_2\uparrow + 2\text{OH}^- \tag{4}$$

The problem of aluminum corrosion in alkaline media occurs in numerous practical cases, starting from its possible application as an anode material in energy storage devices, through pre-treatment processes prior to anodization or for aesthetic purposes, up to the alkaline character of various cleaning agents used on working elements and constructions [11–14]. One of the most frequently utilized methods for lowering the corrosion rate of aluminum in these environments is the application of corrosion inhibitors.

Various organic inhibitors have been reported to be efficient corrosion inhibitors of aluminum and its alloys. Their inhibition effect depends on the molecule structure, the functional groups being electron donors or acceptors, and the number of such groups per molecule [15,16]. It is worth pointing out that the most effective inhibitors are based on molecules containing heteroatoms, such as oxygen, nitrogen, phosphorus, sulfur and aromatic rings [17,18]. Carboxylic acids in particular have been shown as highly efficient corrosion inhibitors of aluminum and its alloys in aqueous alkaline environments [19–22]. Studies carried out on maleic, malic, succinic, tartaric, citric and tricarballylic acids have revealed changes in corrosion efficiency with the increased amount of –COOH groups and decreased amount of –OH groups within inhibitor molecule [20]. Similar reports involved restriction of aluminum corrosion in alkaline media by polyacrylic acids, where studies shown increase of inhibition efficiency with the increase of molecular weight of inhibitor molecule [23]. Compounds containing nitrogen and/or sulphur have also been proved to be efficient corrosion inhibitors, an example of which may be studies on 3-methyl-4-amino-5-mercapto-1,2,4-triazole (MAMT). Inhibitor molecules adsorption on protected metal surface occurs through amine and thiol functional groups [24].

Lashgari and Malek proved that phenol is a highly efficient corrosion inhibitor of aluminum [25]. Phenols are deprotonated in alkaline environments and transformed into inhibitory active forms of phenoxide and phenolate. Similar conclusions, based on theoretical studies, were later on drawn

for p-phenol derivatives, where the author confirmed that the mechanism of inhibition relies on a complicated cycle of protonation/deprotonation of inhibitor molecules in the inner area of the electrical double layer [26]. The process mentioned above leads to local neutralization of corrosive factors and their electrostatic repulsion in the vicinity of an active metal surface. Corrosion inhibition efficiency of p-phenol derivatives depends on several factors, including electron density on oxygen and hydrogen atoms in hydroxyl group, charge transfer, the energy of interaction, molecular activity, electric dipole moment and Gibbs free energy of the dissolution process.

The attention of corrosion scientists is nowadays increasingly focused on application of corrosion inhibitors of natural origin, obtained in accordance with the principles of green chemistry. Green corrosion inhibitors in the form of plant extracts are eco-friendly, non-toxic and biodegradable in neutral environments [27–35]. For example, *Phyllanthus amarus* leaf extract offers nearly 75% efficiency in 2M NaOH solution [36]. The extract contains alkaloids, cyanogenic glycosides, flavonoids, carbohydrates, sugar, proteins, triterpenoids and steroids. Functional groups –OH, –NH2, -SH, present in mentioned above compounds and π-bonds are most likely responsible for inhibition efficiency of *Phyllanthus amarus.* On the other hand, *Piper longum* seed extract, with 94% efficiency at concentration of 0.4 g·L−<sup>1</sup> in 1 M NaOH, contains piperine, piperlongumine, piplartine, piperlonguminine, piperundecalidinine and pipernonaline [37]. The high inhibition efficiency was explained with presence of N-heteroatoms and π-electrons in aromatic rings. The 92% inhibition efficiency of *Gossypium hirsutum* extract in 2 M NaOH most likely originates from presence of O, N or S in amino acids such as: cysteine, lysine, methionine, phenylalanine, arginine, threonine, tyrosine, tryptophan, valine, but also polyphenolic aldehyde and tannins [32]. The authors also observed that higher concentration of active substances in present in leaves rather than seeds of *Gossypium hirsutum.*

Green corrosion inhibitors in the form of extracts from natural products are characterized by a large number of chemical compounds. In such a complex mixture of potential inhibitory compounds it is particularly important to perform phytochemical studies in order to determine the active compounds and their mechanism of interaction, which in many cases appears to be an incredibly difficult task. Therefore, in order to avoid blind-picking during selection of natural extracts containing potentially efficient inhibitor compounds one must get to know in detail the mechanism of interaction of various types of functional groups with protected metal surface as well as how it is modified by other aspects of the molecule structure. A valuable approach towards more effective determination of the active inhibitory compounds may be found in designing the extraction process. Differentiation of the type of solvents or extraction conditions leads to selective extraction of certain groups of compounds. Ryl et al. [38] showed that preparation of bee pollen extracts through extraction in different solvents results in different corrosion inhibition efficiency towards AA5754 in bicarbonate buffer at pH = 11. It has been proved that these differences are caused by varying content of carboxylic acids and phospholipids, which acted as inhibitory active substances in bee pollen extracts.

A certain group of phenol derivatives exhibits very high corrosion inhibition efficiency towards aluminum and its alloys. This group includes catechol, cresol, chlorophenol, resorcinol, nitrophenol and aminophenol [39,40]. Talati and Modi [39] suggested that –OCH3, –NH2 and –CH2CHCH3 functional groups lower the inhibition efficiency of phenol, while –OH, –Cl, –NO2 increase it. Furthermore, they suggested three different adsorption mechanisms, namely: electrostatic forces, the formation of chelating agents with aluminum ions or covalent bond formation. The authors also observed that the inhibition efficiency decreases with the increase of electrolyte alkalinity. The synergistic interaction of resorcinol with Zn2<sup>+</sup> ions was further studied by Lakshmi et al. [40], which revealed significant increase in corrosion inhibition efficiency of aluminum. However, all the aforementioned studies were carried out using the gravimetric method, introducing significant uncertainty to the measurements. The formation of the insoluble corrosion products layer on aluminum surface hinders specific determination of weight loss of the analyzed samples.

Not only the type and the number of active functional groups but also their position in the molecule structure may have a significant influence on corrosion inhibition efficiency. The chemical structure of isomer molecules affects modification of their physic-chemical properties such as solubility, while steric hinderance may influence both kinetics and mechanism of the adsorption process. This subject has not been given sufficient attention in the corrosion research; however, several available reports prove the importance of substituents position in the molecule [41–43]. Fouda and Elasmy [41] presented their studies on phenylthiosemicarbazide derivatives as aluminum corrosion inhibitors in 2M NaOH, with efficiency ranging between 75.0% and 98.5%, depending on the derivative. Hassan et al. [42] confirmed that the efficiency of aromatic carboxylic acids depends on the number and position of carboxylic groups and the presence of other substituents in the aromatic ring. The increasing corrosion efficiency was as follows: benzamide < benzaldehyde < acetophenone < benzoic acid < benzophenone (99.99% efficiency).

The electrochemical impedance studies on thiosemicarbazone interaction with aluminum alloys in 1 M HNO3 revealed an almost 20% higher inhibition efficiency offered by para-substituted compounds in comparison to ortho-substituted ones [44]. There is no general relationship, though. The search for corrosion inhibitors of mild steel in 1 M HCl revealed that ortho-nitroaniline and ortho-methyloaniline show higher corrosion inhibition efficiency in comparison with both meta- and para-substituents, but in the case of phenlylenediamine, meta-substituted functional groups offered the highest efficiency [45]. A similar observation was made on aminophenol-N-salicylidene isomers' action towards zinc in 0.5 M H2SO4 [46]. The goal of this work is to evaluate the influence of position of hydroxyl functional groups within dihydroxybenzene molecule on the corrosion inhibition provided by the isomer towards aluminum in alkaline electrolytes. Dihydroxybenzene isomers (catechol, quinol and resorcinol) were utilized, as presence of hydroxyl functional groups is expected to provide reasonable inhibition efficiency in studied electrolytes. In our studies we have implemented newly developed instantaneous impedance tool, i.e., Dynamic Electrochemical Impedance Spectroscopy in galvanostatic mode (g-DEIS), which is capable of accurate and time-efficient determination of the adsorption mechanism differences [19,20,38].

#### **2. Materials and Methods**

#### *2.1. Materials*

The studied material was aluminum alloy 5754, which had the following alloying additives (in wt.%): Mg 3.6, Fe 0.3, Si 0.3, Cr 0.1, Mn 0.5, Ti 0.1 and Cu 0.1. Cylindrical samples were cut from a rod and subjected to pretreatment procedure in the form of grinding and polishing, carried out on Digiprep 251 (Metkon, Bursa, Turkey) polishing machine. Samples were grinded on a waterproof abrasive papers SiC 600 and 1500, polished with diamond suspensions 6 and 1 μm and mirror-finished on 0.05 μm silica. Following polishing, samples were cleaned and degreased in acetone using ultrasonic cleaner (Polsonic, Warsaw, Poland). Samples were exposed to corrosion studies with 0.5 cm2 surface area.

The primary electrolyte was the bicarbonate buffer solution. The buffer was prepared using 227 cm3 0.1 M NaOH and 500 cm<sup>3</sup> 0.05 M NaHCO3, diluted with deionized water to 1 dm<sup>3</sup> volume. The obtained buffer had pH = 11 and conductivity of 3.8 mS·cm−1. All were analytical purity Sigma Aldrich reagents (Sigma Aldrich, St. Louis, MI, USA). Three dihydroxybenzene isomers were evaluated, namely: resorcinol, catechol and quinol. Their chemical structures are presented on Figure 1. The corrosion inhibition efficiency of the aforementioned compounds was investigated at various inhibitor concentrations cinh = 1, 10 and 100 mM as well as with linearly changing inhibitor concentration during g-DEIS studies related to inhibitor injection into the corrosion cell.

**Figure 1.** The chemical structure of dihydroxybenzene isomers: (**a**) resorcinol; (**b**) catechol; (**c**) quinol.

#### *2.2. Electrochemical Measurements*

The electrochemical impedance studies were performed in a three-electrode setup. Investigated aluminum alloy 5754 was the working electrode (WE), Ag|Ag2O was the reference electrode (RE) (E◦ = +0.215 V vs. SHE) while Pt mesh served as the counter electrode (CE). The electrolyte volume in the corrosion cell was 10 mL.

Corrosion inhibition efficiency studies were performed by means of classic Electrochemical Impedance Spectroscopy (EIS) and Dynamic Electrochemical Impedance Spectroscopy in galvanostatic mode (g-DEIS) after the initial conditioning for 15 min. EIS was carried out in potentiostatic mode, at open circuit potential (OCP) conditions. The perturbation signal for EIS measurements was applied in the frequency range between 50 kHz and 40 mHz, with 10 points per decade of frequency and amplitude of 15 mV. Multisinusoidal perturbation signal for g-DEIS studies composed of 29 superimposed elementary signal in the frequency range between 4.5 kHz and 1.0 Hz, with 8 points per decade of frequency. Sampling frequency was 128 kHz. The amplitude of perturbation signal was controlled to assure the amplitude of response signal does not exceed 15 mV. The analysis window for the Short-time Fourier Transformation was 10 s in length. A similar measurement procedure was previously applied in corrosion studies [19,20,38,47–49].

To build the adsorption isotherm, studied corrosion inhibitor was injected from the secondary cell to the corrosion cell through BQ80S microflow peristaltic pump (Lead Fluid, Baoding, China). The flow rate was set as 0.02 mL·min<sup>−</sup>1. The concentration of studied inhibitor in the secondary cell was set in a way to assure corrosion inhibitor concentration in the corrosion cell (cinh) equal to 10 mM at the end of 6000 s long experiment. The corrosion cell was thermostated at 25 ◦C. The electrochemical setup used during all corrosion studies is schematically presented on Figure 2.

**Figure 2.** Schematic representation of the setup used during electrochemical studies.

#### *2.3. Equipment*

The EIS measurements were carried out using frequency-response-analysis on Gamry Reference 600+ potentiostat (Gamry Instruments, Warminster, PA, USA). The g-DEIS measurement system consisted of Autolab PGSTAT 302N (Metrohm, Herisau, Switzerland) galvanostat connected to PXI-4464 measurement card for AC signal generation and PXI-6124 card for AC/DC signal acquisition, both operating in PXIe-1073 chassis (all from National Instruments, Austin, TX, USA). The microflow peristaltic pump used was BQ80S (Lead Fluid, Baoding, China). The thermostat was Corio CD (Julabo GmbH, Seelbach, Germany)

Microscopy analyses of AA5754 corrosion process were performed on a Scanning Electron Microscope VP-SEM S-3400 N (Hitachi, Chiyoda, Japan), equipped with a tungsten source and operating at 20 kV accelerating voltage. SEM micrographs were done in secondary electron mode.

High-resolution X-Ray Photoelectron Spectroscopy (XPS) analyses were performed on Escalab 250 Xi multispectroscope (ThermoFisher Scientific, Waltham, MA, USA). The spectroscope is equipped in Al Kα X-ray source with a spot diameter of 250 μm. The measurements were carried out at 20 eV pass energy and 0.1 eV energy step size. The charge compensation was provided by means of low-energy electrons and low-energy Ar+ ions emission from the flood gun.

Nuclear Magnetic Resonance (NMR) spectra were recorded on AVANCE III 500 MHz NanoBay spectrometer (Bruker, Billerica, MA, USA). Tetramethylsilane (TMS) was used as an internal standard in all the measurements. The 0.5 mL solution containing 80 mM dihydroxybenzene isomer in D2O was filled with NaOH. The titrated compounds ratio to NaOH was 1:1, 1:5 and 1:10. 1H-NMR and 13C-NMR spectra were recorded. All the measurements were performed after 24 h in 25 ◦C and in the same volume of solvent.

#### **3. Results and Discussion**

#### *3.1. Dihydroxybenzene Isomers as Corrosion Inhibitors*

Electrochemical Impedance Spectroscopy (EIS) was used during the preliminary studies to assess the corrosion resistance of the investigated aluminum alloy immersed in an alkaline environment with the addition of each dihydroxybenzene isomer. The studies were carried out for three different corrosion inhibitor concentrations cinh, namely: 1, 10 and 100 mM. The impedance spectra, presented in the form of Nyquist plots, are plotted in Figure 3. It can be seen by the shape of the impedance semicircles, that studied derivatives offer different level of corrosion protection; however, detailed analysis required fitting of the obtained data with electric equivalent circuit (EEC).

**Figure 3.** EIS impedance spectra recorded for AA5754 exposed to bicarbonate buffer (pH = 11) with the addition of the studied dihydroxybenzene isomers: (**a**) resorcinol; (**b**) catechol; (**c**) quinol at various concentrations in range 1–100 mM.

There is only one clear time constant present on the obtained impedance spectra, suggesting that the charge transfer through the electrode interface and through the adsorbed inhibitor layer are characterized with similar relaxation times. The authors decided to apply a simple form of Randles electric equivalent circuit (EEC) due to the prominence of one time constant on the impedance spectra in the applied frequency range. The one and only alteration was to replace the capacitance parameter with a constant phase element (CPE) in order to properly consider investigated electrode surface heterogeneity, originating from alloy microstructure and roughness, but even more important presence of local adsorption sites of corrosion inhibitor at its low concentrations. The CPE impedance is given with Equation (5).

$$Z\_{\rm CPE} = \left[ Q(j\omega)^a \right]^{-1} \tag{5}$$

It should be noted that in the boundary case, if α = 1, the CPE impedance responds to a capacitor of capacitance Q. Therefore, CPE exponent α is often considered as the homogeneity factor, its decrease represents the increase of surface heterogeneity, while Q reflects the quasi-capacitance of the heterogeneous electrode. Furthermore, the CPE can be used to estimate the effective capacitance of studied electrode Ceff, with the use of Hirschorn's approximation for the surface distribution of the capacitance dispersion [50]:

$$C\_{eff} = Q^{1/a} \left(\frac{R\_t R\_{ct}}{R\_t + R\_{ct}}\right)^{(1-a)/a} \tag{6}$$

where Re is the resistance of the electrolyte and Rct is charge transfer resistance. The χ*2*-distribution of the selected EEC was in the range of 10<sup>−</sup>4, which is a good result, when taking into consideration EEC simplicity and system non-stationarity.

The shift in the calculated value of charge transfer resistance Rct may be utilized to estimate the inhibition efficiency IE%, using the well-known relationship (7) [19]:

$$IE\_{\%} = \left(1 - \frac{R\_{ct}^0}{R\_{ct}}\right) \tag{7}$$

where Rct<sup>0</sup> denounces the measured value of charge transfer resistance in the absence of the inhibitor. The results of impedance data analysis with the R(QR) EEC are summarized in Table 1.


**Table 1.** Electric parameters of the studied systems obtained from EIS results fitted with R(QR) EEC.

Interestingly enough, the addition of resorcinol at the lowest concentration (1 mM) provided the lowest level of corrosion resistance. This feature most likely originates from the altered mechanism of molecules adsorption on the surface of aluminum alloy and was an object of further studies. On the other hand, presence of catechol and quinol offers approx. 60% efficiency already at cinh = 1 mM, which does not significantly improve until reaching substantially higher concentrations. This may be observed in particular through effective capacitance Ceff changes, which decreased by nearly 30% already at the lowest inhibitor concentrations, compared to reference buffer electrolyte. The most important factor affecting this parameter is the thickness of the adsorbed layer on aluminum surface [20].

Importantly, the highest inhibition efficiency was obtained after the addition of 100 mM of resorcinol. While the difference between IE% at this concentration does not exceed 4%, it should be noted that it corresponds to over 3× and nearly 5× higher Rct values of AA5754 alloy in presence of resorcinol versus catechol and quinol, respectively. The increased efficiency of resorcinol at very high inhibitor concentrations was previously revealed in gravimetric studies [51]. In our opinion, its positive interaction is directly connected to the competitive formation of the corrosion products layer, making an additional barrier to aggressive environment. In the case of catechol and quinol, the barrier properties of the corrosion products layer are less evident, as discussed later.

#### *3.2. The Adsorbed Layer Chemistry*

The authors decided to focus on the core of electrochemical and physic-chemical studies on resorcinol and catechol isomers, which is due to the nearly identical response and adsorption mechanism between catechol and quinol. At the same time, there is a significant solubility difference between these two compounds in aqueous electrolytes at 25 ◦C, hindering possible utilization of quinol as efficient corrosion inhibitor. Quinol is sparingly soluble in water and show tendency for sedimentation over time. Its solubility in water is 5.9 g/100 g in comparison to 43 g/100 g for catechol and 110 g/100 g for resorcinol [52].

The chemistry of the adsorbed dihydroxybenzene layer and corrosion products layer on the surface of studied aluminum alloy 5754 was evaluated with the use of high-resolution XPS analysis. First, samples were pre-exposed to alkaline electrolyte with the addition of studied inhibitor at 10 or 100 mM concentration for a period of 24 h. The XPS spectra in the binding energy (BE) range of C1s, O1s, Al2p and Mg1s peaks were collected and deconvoluted using the fitting procedure described below. The results of the aforementioned deconvolution are summarized in Table 2.


**Table 2.** Surface chemical composition (in at.%) for AA5754 after 24 h exposure in bicarbonate buffer (pH = 11) with the addition of resorcinol or catechol, based on high-resolution XPS analysis.

Figure 4a presents the XPS spectra obtained in Al2p BE range. There is a clear shift in the peak position recorded between resorcinol-exposed and catechol-exposed aluminum samples. First, when exposed to bicarbonate buffer, but also with the addition of resorcinol, the primary peak doublet is located at higher BE values, with Al2p3/<sup>2</sup> peak at approx. 76.7 eV. This component was labeled Alox2 and ascribed to the non-stoichiometry aluminum corrosion product layer, often observed in studies of this metal in pH range 10–12 [53]. The second component (Alox1), corresponding to native Al2O3 passive film was shifted at −1.9 eV. The presence of native oxide film results from air exposure of samples. The share of non-stoichiometric oxides (Alox2:Alox1) is nearly 13:1 when compared to native oxides for sample exposed in bicarbonate buffer, and drops to 5:1 after addition of resorcinol, regardless the concentration. The results suggest presence of corrosion products layer, which contributes the decrease of aluminum corrosion rate in these environments.

On the other hand, the addition of catechol at cinh = 10 mM resulted in Alox2:Alox1 share of 1:3. The significantly lower contribution from Alox2 may suggest that inhibitory action of catechol efficiently reduces formation of the corrosion products on aluminum alloy 5754 surface. This observation was confirmed for higher concentrations of catechol, where Al2p signal is composed solely of native oxide film.

**Figure 4.** High-resolution XPS spectra recorded in (**a**) Al2p and (**b**) C1s binding energy range for aluminum alloy 5754 after 24 h exposure in bicarbonate buffer (pH = 11) and bicarbonate buffer with the addition of resorcinol or catechol at concentrations 10 and 100 mM.

When compared to bicarbonate buffer exposed sample, the C1s spectra reveal the significant presence of carbon species for both analyzed dihydroxybenzenes and at both concentrations, suggesting that inhibitor molecules indeed take part in formation of the adsorbed layer on aluminum alloy 5754 surface (Figure 4b). The chemistry of this carbon species is strongly altered, however. In the case of catechol, the amount of C-C and C-OH species, based on peaks at 284.7 and 286.0 eV, respectively, is nearly 3× compared to resorcinol at the same concentrations [54–56]. Furthermore, the C1s spectra reveal significantly higher amount of carbonyl and/or carboxyl species for catechol-containing electrolyte (at BE exceeding 287.6 eV). These results corroborate previous findings regarding possible formation of the adsorbed inhibitor layer directly on aluminum surface and negligible participation of the corrosion products layer. On the other hand, in absence of corrosion inhibitor the C1s contribution originates primarily from C-OH and C=O species, testifying the interaction of bicarbonate species with aluminum sample.

The O1s spectra deconvolution is in good accordance with previously discussed results. In the absence of dihydroxybenzene isomers, over 47 at.% of the aluminum surface consists of oxygen atoms, with peak binding energies characteristic either for aluminum metal oxides (531.3 eV) or hydroxides (531.6 eV) [19,20]. Further on, this value slowly decreases for resorcinol-containing electrolyte, still within the range of 40 at.%. However, in the presence of catechol, the amount of O2<sup>−</sup> species is significantly lower, at 10 and 100 mM. Here, the signal corresponding to peaks located at approx. 532.6 eV is still strong, but with an organic C-O interaction origin, instead. The last deconvoluted O1s component is located at binding energies exceeding 532.7 eV. Its source is adsorbed carboxyl species, but also chemisorbed water molecules [57]. The non-stoichiometric corrosion products layer reveals high hydration level, confirmed with SEM micrographs, therefore higher share of chemisorbed water on surface of bicarbonate buffer-exposed and resorcinol-exposed aluminum alloy 5754 is natural.

Finally, the amount of magnesium, the primary alloying additive, was only slightly altered between various investigated electrolytes. In each case magnesium was found in the form of hydroxides with Mg1s peak BE at 1303.1 eV, its higher contribution for resorcinol testifies the presence of magnesium oxides in the corrosion products layer, having a possible effect on the increased corrosion resistance [58].

The SEM micrographs shown in Figure 5 reveal significant difference in occurrence of the corrosion process for both studied dihydroxybenzene derivatives. The addition of resorcinol does not affect significantly surface topography when compared to reference aluminum alloy sample in buffer electrolyte (Figure 5a). The oxidation is primarily oriented around the alloy microstructure, which can be confirmed through local dissolution of alloy matrix surrounding intermetallic particles [58–60]. These particles appear on the micrographs with bright colors. Based on own studies and the literature survey discussed, the particles are primarily composed of aluminum and alloying additives of Fe, Cr and Mn—each cathodic in nature compared to metal matrix [60,61]. Furthermore, dense network of cracks visible on Figure 5a–c testifies high hydration level of adsorbed layer and corrosion products layer on the electrode surface.

**Figure 5.** SEM micrographs of aluminum alloy 5754 surface exposed to bicarbonate buffer (pH = 11) for 24 h: (**a**) in absence of corrosion inhibitor; with addition of resorcinol at concentration (**b**) 10 mM and (**c**) 100 mM or with addition of catechol (**d**) 10 mM and (**e**) 100 mM.

The presence of catechol or quinol in the electrolyte consequences in localized corrosion of aluminum, which is most often restricted to anodic Mg-rich phases and areas surrounding cathodic intermetallic particles. The absence of thick corrosion product layer effects in lack of cracks, which otherwise cover metal surface. The localized corrosion is naturally more evident at lower inhibitor concentrations (see Figure 5d), where the spherical shape of caverns should be connected with local areas of hydrogen evolution in coupled cathodic reaction [53].

#### *3.3. Thermodynamics of the Adsorption Process*

The thermodynamics of dihydroxybenzene isomers adsorption was evaluated based on instantaneous impedance studies, carried out in galvanostatic mode (g-DEIS). The studied inhibitor is injected with linear injection rate (0.02 mL·min−1); thus, the instantaneous value of inhibitor concentration in the corrosion cell is known. Dynamic impedance measurements allow for a precise evaluation of the instantaneous values of electric parameters, where the inhibition efficiency may be estimated from Rct using previously introduced Equation (7). Studies carried out in galvanostatic mode at iDC = 0 ensures constant measurement conditions and lack of an additional polarization component, resulting from corrosion potential changes during inhibitor injection. The approach is characterized with higher accuracy, allowing to obtain larger data set and evaluate the exact inhibitor concentration at which the linear character of the adsorption isotherm is modified due to full electrode surface coverage with inhibitor monolayer. The details of the experimental procedure are presented elsewhere [19,20].

The dynamic impedance spectra presented in the form of the Nyquist plot are shown on Figure 6a,b for resorcinol and catechol, respectively. The shape of the impedance spectra develops during corrosion inhibitor injection, where the increased impedance loop diameter testifies the increase of aluminum corrosion resistance. Fitting impedance spectra with R(QR) EEC allows for determination of dynamic changes of the electric parameters: Rct and CPE within a timeframe of an analytical window length of a Short-Time Fourier Transform function, equal to 10 s in this case. The fitting procedure was carried out using dedicated software based on Nelder-Mead algorithm and build in LabView environment. The resultant <sup>χ</sup>2-distribution was typically in 10−<sup>4</sup> range and did not exceed 2 <sup>×</sup> <sup>10</sup><sup>−</sup>3. Determination of the instantaneous Rct values allowed for calculation of momentary inhibition efficiency IE%, which is also the measure of surface coverage with inhibitor molecules θ (IE% = θ × 100%). According to the principles of the most commonly used Langmuir adsorption isotherm, the adsorption equilibrium constant, Kads, depends on surface coverage θ, which is given with Equation (8):

$$K\_{\rm ads}c\_{\rm inlh} = \left(\frac{\theta}{1-\theta}\right) \tag{8}$$

In the linear range of Equation (8), the adsorption isotherm may serve for calculation of the adsorption Gibbs free energy ΔG, using Equation (9):

$$
\Delta G = -RT\ln(K\_{\text{ads}} \times 55.5) \tag{9}
$$

Importantly, Langmuir isotherm conditions are only fulfilled for concentrations below full coverage with the adsorbent monolayer. One should note, that classical approaches towards adsorption isotherm determination are based on merely few measurement points, where the non-linear behavior resulting from aforementioned situation is difficult to track. On the other hand, the quasi-capacitance parameter obtained during g-DEIS impedance measurements allows estimating the exact concentration required for monolayer formation by inhibitor molecules [20].

**Figure 6.** The g-DEIS impedance graphs of aluminum alloy 5754 in bicarbonate buffer (pH = 11), presented in Nyquist projection versus (**a**) catechol and (**b**) resorcinol concentration changes during its injection to corrosion cell. (**c**) Langmuir model of adsorption isotherms drawn based on instantaneous Rct changes, in the inset the instantaneous changes of effective capacitance Ceff.

Figure 6c presents the adsorption isotherms obtained with g-DEIS approach and plotted according to the Langmuir model of molecules adsorption. The isotherms were drawn for resorcinol and catechol at concentrations in range 0–10 mM. It can be pointed out that both of these functions are characterized by loss of linear character and both have inflection at concentrations between 2 and 4 mM. However, the effective capacitance Ceff measurements plotted in the inset of Figure 6c reveals significant differences in the adsorption mechanism.

There are three primary factors affecting the value of capacitance according to Equation (10), namely, electrochemically active surface area A, permittivity ε, and layer thickness d:

$$\mathcal{C} = \frac{\varepsilon\_0 \varepsilon A}{d} \tag{10}$$

where ε<sup>0</sup> is vacuum permittivity. Normalization of heterogeneity factor affecting quasi-capacitance Q and estimation of the effective capacitance Ceff allows ignoring the effect of electrode heterogeneity. Previous studies have shown that with relatively short measurement duration the key factor influencing instantaneous Ceff of the adsorbed layer is its thickness [20,38].

In the case of a catechol-exposed AA5754 electrode, the Ceff value increases until reaching its maximum at cinh = 3.5 mM and then decreases. Therefore, it should be assumed that the full coverage of aluminum surface with corrosion inhibitor molecules occurs at this concentration and the following Ceff decrease results from the increase in adsorbed layer thickness. The adsorption isotherm still remains linear afterward, but no longer following the aforementioned restriction regarding surface coverage.

The situation is essentially different in the case of resorcinol-exposed AA5754 electrodes, where according to classic EIS and XPS studies the inhibitory action is generally lower. In the whole studied concentration range, up to 10 mM the Ceff value effectively increases. The competitive interaction of the corrosion products layer and the resorcinol adsorption layer may have its effect on difficulties in assessing the full coverage of the adsorption layer but also influences layer permittivity. As a result, the adsorption isotherm for resorcinol was following Langmuir model of adsorption at significantly higher inhibitor concentrations than catechol. On the other hand, the estimated inhibitor efficiency at the lowest concentrations is negligible and thus hard to measure. A conclusion should be made that the applicability of the Langmuir adsorption isotherm model for catechol and resorcinol lies in different inhibitor concentration ranges.

Following Equations (8) and (9), the calculated values of Gibbs free energy ΔG of the adsorption process are presented in Table 3. Their negative values in both cases confirm spontaneous adsorption of both studied dihydroxybenzene molecules on aluminum alloy 5754 surface. Nevertheless, the significantly different ΔG values between resorcinol and catechol suggest an altered adsorption mechanism.


**Table 3.** The applicability range of the Langmuir adsorption model and the obtained thermodynamic parameters for resorcinol and catechol.

The more negative Gibbs free energy values are typical in case of chemisorption and formation of chemical bonds between filled π-orbitals in the oxygen atoms and partially unoccupied π-orbitals in the d-block metals. This is the postulated adsorption mechanism of resorcinol. Naturally, the value of this thermodynamic parameter may be further influenced by reported presence of the nonstoichiometric corrosion products layer. On the other hand, our previous studied on carboxylic acids revealed that the less negative free Gibbs energies correspond not only to the electrostatic interaction of the physisorption process but also the ligand exchange model of adsorption, resulting in formation of coordination compounds at the metal interface [19,20]. This is the case of catechol interaction. The lower efficiency of ligand formation by resorcinol and quinol originates from the molecule geometry.

#### *3.4. The Keto-Enol Tautomerism*

During the long-term exposure tests, the authors observed changes in the color of the studied electrolytes over time, and resorcinol in particular (see inset of Figure 7). These changes were followed by alteration of the electrochemical characteristics over time. On the other hand, for catechol and quinol the long-term inhibition efficiency was significantly higher. The g-DEIS studies were carried out once more to track the exact change in the electrochemical behavior of aluminum alloy 5754 during 24 h exposure. The results of the impedance monitoring are presented in Figure 7.

Δ

**Figure 7.** The g-DEIS impedance graphs of aluminum alloy 5754 in Nyquist projection during 24 h exposure in (**A**) bicarbonate buffer (pH = 11); and with the addition of (**B**) 10 mM catechol and (**C**) 10 mM resorcinol.

Analysis of the impedance data makes it possible to draw conclusions regarding the long-term behavior of AA5754 under the studied electrolytic conditions. The slight increase of the semicircle diameter over time, seen on the Nyquist projection for buffer-exposed sample, makes it possible to conclude that the corrosion product layer forming on metal surface provides partial barrier properties and decreases the corrosion rate approximately 2.5×.

When exposed to buffer with addition of 10 mM catechol, the instantaneous charge transfer resistance is slightly higher (typically around 0.4 kΩ) and then gradually increases over time to reach significantly improved inhibition efficiency of ~98% after 24 h exposure. On the other hand, AA5754 exposed to electrolyte containing the same amount of resorcinol shows very small increase of charge transfer resistance over duration of the long-term exposure experiment.

The long-term exposure study allows drawing two important conclusions. First, there must be an additional interaction between studied inhibitor molecules and the electrolyte or the analyzed sample, which further differentiates the electrochemical characteristics of these dihydroxybenzene isomers over time. Second, when performing inhibitor efficiency measurements one has to take into consideration possible changes of investigated system characteristics. This is possible by carrying out fast measurements with techniques that allow non-stationary process analyses (such as g-DEIS). Alternatively, one could employ a sufficiently long conditioning period, which might be different for each studied system. The latter approach, although more accessible, may cause problems in terms of meeting the conditions for many adsorption models.

The authors claim that the mechanism leading to further differentiation of adsorption by catechol and resorcinol on the aluminum alloy surface is the keto-enol tautomerism, which may occur in aqueous alkaline environments. Nuclear magnetic resonance (NMR) studies were performed in order to verify this hypothesis.

1H-NMR measurements were performed in order to determine the presence of possible keto-enol forms in resorcinol in the alkaline conditions or to determine the formation of a salt of those compounds. The studies of proton transfer by 1H-NMR titration present a useful technique to determine the keto-enol equilibria [62,63]. The 1H-NMR spectra of resorcinol dissolved in D2O with addition of NaOH in molar ration of 1:5 and 1:10 are presented on Figure 8. The 1H-NMR spectra of resorcinol have previously been performed in D2O [64,65], while the titration of this compound by NaOH has not been investigated.

**Figure 8.** 1H-NMR spectra of resorcinol in D2O (black line) and in the presence of NaOH in D2O in the molar ratio 1:5 (red line) and 1:10 (blue line), respectively.

Two triplets are observed on spectra of resorcinol in D2O, which correspond to H5 and H2, while two doublets correspond to H4 and H6. The spectra of resorcinol after the addition of NaOH in molar ratio 1:5 changed diametrically. The shape, the chemical shifts and the multiplicity are different in comparison to the first one. Two main signals shifted towards negative values are now observed. This phenomenon indicates that the protons present in the structure of resorcinol are changed, further influencing the chemical shifts and the multiplicity. The addition of molar excess of NaOH in (1:10), does not cause any additional changes in 1H-NMR spectrum shape, but the peaks are further shifted.

The determination of the new resorcinol derivative structure was possible after measuring the 13C-NMR spectra, shown on Figure 9. These spectra were performed in D2O and after addition of NaOH in molar ratio 1:5 and 1:10, similar to previous experiment. It should be noticed that the peaks C4, C5, and C6 present in spectra before and after addition of NaOH do not change their position significantly. On the other hand, peaks C1 and C3, overlapping at 156.8 ppm, change their position to 167.4 ppm with the addition of NaOH, which may indicate that keto form is present in the structure of resorcinol regardless the molar ratio. Additionally, after addition of NaOH the shift of peak C2 is observed from 102.6 ppm to 106.0 ppm, this shift confirms the formation of keto form.

**Figure 9.** 13C-NMR spectra of resorcinol in D2O (black line) and in the presence of NaOH in D2O in molar ratio 1:5 (red line) and 1:10 (blue line), respectively.

In the next step, the 1H-NMR and 13C-NMR spectra were performed for catechol under the same experimental conditions. The 1H-NMR spectra of catechol have previously been performed in aqueous solution, but under acidic pH = 2.4 [66] and in CDCl3 [67]. The shape and chemical shifts of 1H-NMR spectra illustrated on Figure 10 are very similar to those in the literature.

**Figure 10.** 1H-NMR spectra of catechol in D2O (black line) and in the presence of NaOH in D2O in molar ratio 1:5 (red line) and 1:1 (blue line), respectively.

Two multiplets are present regardless of the solution, while the addition of NaOH in ratio 1:5 and 1:10, respectively, do not cause any changes in 1H-NMR spectra shape; however, with increasing molar ratio of NaOH, the spectra are shifting towards more negative values. The above changes clearly indicate that the chemical structure of the catechol does not change significantly, while the shifts may indicate formation of sodium salts of catechol [68].

13C-NMR spectra were performed in order to confirm the formation of sodium salt of catechol in alkaline conditions, as shown on Figure 11. The obtained data confirm previously drawn assumptions. 13C-NMR spectra reveal 13C chemical shifts for catechol in D2O with the presence of NaOH in molar ratio 1:5 and 1:10. The C1 and C2 atoms connected to the hydroxyl groups in catechol in investigated solutions both give signal at 143.9 ppm in absence of NaOH, but after its addition, the signal is shifted towards 152.6 ppm and 154.9 ppm for ratio 1:5 and 1:10, respectively. The shift of carbon signals directly indicates that catechol in alkaline solutions forms a salt [69]. It is worth noticing that a small change of peak position was also observed for carbon C4/C5 and C3/C6 after additions of NaOH.

**Figure 11.** 13C-NMR spectra of catechol in D2O (black line) and in the presence of NaOH in D2O in the molar ratio 1:5 (red line) and 1:1 (blue line), respectively.

#### *3.5. Dihydroxybenzene Isomers Adsorption Mechanism*

The overall interaction of the studied dihydroxybenzene isomers with aluminum alloy 5754 surface may thus be explained using the scheme presented on Figure 12. Figure 12a illustrates the case of aluminum corrosion in an alkaline environment, according to the two-step mechanism discussed in the introduction section: 1) attack of OH<sup>−</sup> ions on Al2O3 leading to its dissolution and Al(OH)3 formation, followed shortly after by 2) chemical formation of Al(OH)4 − ions. The corrosion inhibition mechanism for catechol considers formation of ligands with aluminum ions (see Figure 12b), which is hindered in the case of resorcinol. On the other hand, resorcinol depends on the formation of insoluble corrosion products layer, which to a large extent provides a barrier mechanism towards corrosive electrolyte. The molecules chemisorb on the corrosion products layer, which becomes very efficient only at high inhibitor concentrations.

**Figure 12.** Schematic representation of aluminum corrosion mechanism in aqueous alkaline solutions: (**a**) in absence of corrosion inhibitor and with the addition of (**b**) catechol; (**c**) resorcinol. Phase (1) represents the OH<sup>−</sup> attack leading to formation Al(OH)3, phase (2) describes the interaction between OH<sup>−</sup> and Al(OH)3, leading to the formation of Al(OH)4 − and a non-stoichiometric insoluble corrosion product layer.

The lower inhibition efficiency of resorcinol at concentrations not exceeding 10 mM is connected with the keto-enol tautomerism mechanism, occurring in aqueous alkaline media and lowering the molecule influence on the corrosion protection (see Figure 12c). Due to local differences in pH in anodic and cathodic zones of electrode/electrolyte interface, the dynamics of the keto-enol tautomerism may be locally altered. The NMR spectra revealed that the process takes place in wide pH range.

#### **4. Conclusions**

Understanding the interaction mechanism of inhibitor molecules with the protected metal surface is of key importance in the selection of the most efficient corrosion inhibitors, the most important, in particular, in the case of green inhibitors based on natural extracts. While it is widely known how different functional groups affect the adsorption mechanism, the differences introduced by its location within the inhibitor molecule are often omitted.

In this study, we revealed how the position of hydroxyl groups affects the adsorption mechanism of dihydroxybenzene isomers and offered corrosion resistance toward alumnium alloy 5754 surface in alkaline environment. The utilization of Dynamic Electrochemical Impedance Spectroscopy in galvanostatic mode (g-DEIS) for adsorption isotherm determination made it possible to confirm different forms of dihydroxybenzene interaction. All of the studied inhibitors followed the Langmuir model of adsorption, although we have observed that its applicability lies in different inhibitor concentration ranges.

Resorcinol was found to be characterized by the chemical adsorption mechanism. Its adsorption on aluminum surface is competitive to insoluble corrosion product layer formation, as shown with SEM and XPS studies. This interaction leads to the best inhibitor efficiency at the highest investigated concentration of 100 mM, but is not as efficient at lower concentrations. On the other hand, catechol and quinol follow the ligand exchange model of adsorption. This leads to more efficient adsorption and increases corrosion protection even at lower corrosion concentrations: 1 and 10 mM. The adsorption process dominates insoluble corrosion product layer formation, the presence of which on the analyzed surface was negligible.

The next significant difference lies in the long-term behavior and corrosion protection offered by dihydroxybenzene isomers in alkaline electrolyte. We report that resorcinol molecules undergo keto-enol tautomerism in sodium hydroxide solution, while the aforementioned process was negligible in the case of quinol and catechol. The tautomerism leads to the rebuilding of the inhibitor molecule, electrolyte discoloration, but does not have significant influence on the chemical adsorption mechanism by resorcinol over longer periods of time. It is even possible that the presence of keto-enol tautomerism itself is the reason behind hindered adsorption of resorcinol and offered corrosion resistance. Keto forms were not observed in the structure of catechol and quinol molecules. At the same time, their ability to complex metal ions leads to formation of layers with higher barrier properties and increased corrosion inhibition.

**Author Contributions:** Conceptualization, J.R.; Methodology, J.R. and J.W.; Investigation, M.B., M.K., W.L. (Electrochemical Studies, SEM), J.R. (XPS) and P.N. (NMR); Resources, J.R.; Writing—Original Draft Preparation, J.R., P.N., J.W. and W.L.; Writing—Review and Editing, J.R. and J.W.; Funding Acquisition, J.R.

**Funding:** The authors acknowledge the financial support of the Polish Ministry of Science and Higher Education from the budget funds in the period 2016-2019 under Iuventus Plus project IP2015 067574.

**Acknowledgments:** The authors acknowledge Pawel Slepski and Artur Zielinski from Gdansk University of Technology for development of the software dedicated to effective DEIS data collection and analysis.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **PDC Glass**/**Ceramic Coatings Applied to Di**ff**erently Pretreated AISI441 Stainless Steel Substrates**

#### **Milan Parchovianský 1,\*, Ivana Parchovianská 1, Peter Švanˇcárek 2, Günter Motz <sup>3</sup> and Dušan Galusek 1,2**


Received: 16 December 2019; Accepted: 24 January 2020; Published: 31 January 2020

**Abstract:** In this work, the influence of different cleaning procedures on adhesion of composite coatings containing passive ceramic and commercial glasses was investigated. Two compositions (C2c, D2-PP) of double-layer polymer-derived ceramic (PDC) coating systems, composed from bond coat and a top coat, were developed. In order to obtain adherent coatings, stainless steel substrates were cleaned by four different cleaning procedures. The coatings were then deposited onto the steel substrate via spray coating. Pretreatment by subsequent ultrasonic cleaning in acetone, ethanol and deionised water (procedure U) was found to be the most effective, and the resultant C2c and D2-PP coatings, pyrolysed at 850 ◦C, indicated strong adhesion without delamination or cracks, propagating at the interface steel/bond coat. In the substrate treated by sandblasting and chemical etching, small cracks in the bond coat were observed under the same pyrolysis conditions. After oxidation tests, all coatings, except for those subjected to the U-treated substrates, showed significant cracking in the bond coat. The D2-PP coatings were denser than C2c, indicating better protection of the substrate.

**Keywords:** cleaning; bond coat; PDC coatings; fillers

#### **1. Introduction**

Because of the increasing costs for metals, there is an effort made to enhance the service life of steel components exposed to aggressive environment, which is commonly used in exhaust gas elements, waste incineration plants or in chemical industry. Refractory stainless steels are highly oxidation and corrosion resistant materials. As metal wear and oxidation/corrosion cause significant economic losses, the development of thermal (TB) and environmental barrier coatings (EBC) is the matter of significant importance.

Due to their extraordinary properties at high temperatures and in chemically aggressive environments, non-oxide and oxide PDC ceramic coatings are suitable for increasing the oxidation and corrosion resistance of metals [1]. Preceramic polymers offer a lot of processing advantages that are not possible with traditional ceramics [2]. For example, organosilicon polymer precursors such as polysiloxanes [3], polycarbosilanes [4], or polysilazanes [5–7], represent a class of hybrid materials which, by suitable heat treatment (pyrolysis in a controlled atmosphere), provide high purity ceramic materials with an adaptable chemical composition and a well-defined structure. These polymers are characterised by an inorganic polymer chain composed of silicon atoms and organic substituents attached to the backbone. Polysilazanes are currently used as precursors for the synthesis of Si3N4 and SiCN ceramics, mainly due to the high ceramic yield after pyrolysis (often >80 wt. %) [8]. Polysilazanes are suitable materials for the preparation of protective coatings due to their excellent oxidation and corrosion resistance, UV stability and high hardness. These polymers have excellent adhesion to a wide variety of different substrates, e.g., metal, composite, graphite and glass. The PDC route provide the application of liquid or diluted polymers by easily scalable methods, such as dip-coating [9,10], spin-coating [11], doctor-blade method [12] or spray-coating [13,14]. The choice of a particular method depends on the future use of the coating, the type and shape of the body to be coated and the deposited layer, the size of the covered area, the thickness of the coating and its desired properties.

The main disadvantage of the organosilicon polymer precursors, however, is shrinkage, often more than 50 vol. %, that occurs during the transformation from polymer to an amorphous ceramic [15]. The undesirable shrinkage of the polymer leads to crack formation and, in extreme cases, complete failure of the coating. To overcome these unwanted problems, the coatings that consist of only liquid polymer have to be loaded with beneficial components called fillers. The fillers are active [16,17] or passive, and include a large variety of materials, including YSZ [18], Si3N4 [19], Al2O3 [20] and NbC [21] or commercial glasses [22]. The fillers partially or completely compensate the shrinkage, close the pores and increase the coating thickness [23].

The main function of passive fillers is to decrease the bulk fraction of the polymer used, to reduce the amount of gases generated during pyrolysis and, consequently, to alleviate the overall weight loss and shrinkage, and to eliminate the presence of macro-defects by filling the void space in the material without changing its volume. The glass fillers account for densification and sealing of the system, increasing the efficiency of EBC [24]. The service temperature and softening point of the glass filler particles should be matched to increase efficiency of the coating and, in optimum case, heal any defects formed during the coating operation. In our previous work [25], composite PDC coatings with passive fillers and commercial glasses have been developed. Despite using a range of passive fillers, the bulk shrinkage of the polymer precursor has in many cases led to the preparation of porous coatings. Also, the coating often delaminated from the metal substrate, and lost its protective action as EBC.

Another factor ensuring good adhesion of PDC coatings is based on providing an appropriate surface of stainless steel substrate. A number of various pretreatment procedures, such as sandblasting or etching the substrate by different chemical agents, have been described.

In this work, the influence of pretreatment of the AISI441 steel substrate such as sandblasting, etching of the surface or combination of these methods, were investigated in order to choose the most effective type of cleaning and prevent the delamination of the bond coat from the steel substrate. The oxidation tests were performed in order to evaluate the adhesion of the bond coat at higher temperature and longer operating times.

#### **2. Materials and Methods**

The preparation of the PDC coating systems consisted of 3 steps: (1) synthesis of passive fillers with compositions in the Al2O3-Y2O3-ZrO2 (AYZ) system by sol–gel Pechini method [26], (2) pretreatment of stainless steel by different methods and (3) preparation of double layer coatings consisting of the bond coat and top coat using a combination of a commercial polymer with passive and glass fillers.

#### *2.1. Preparation of the Precursor Powder*

A powder in the AYZ system with the composition (in mol. %) 61.49 Al2O3, 18.51 Y2O3 and 20 ZrO2 was used as a passive filler was prepared by the modified sol–gel Pechini method [27]. Y2O3 (99.9%, Treibacher Industrie AG, Althofen, Austria) was converted into nitrate by dissolving powder oxide in concentrated HNO3 (65% Centralchem, Bratislava, Slovakia). Al(NO3)3·9H2O (p.a., Centralchem, Bratislava, Slovakia) and ZrOCl2·8H2O (99.9%, Sigma-Aldrich Co. LLC., Darmstadt, Germany) dissolved in deionised water were then added to yttrium nitrate solution. A 1:1 molar ratio solution of C6H8O7 (99.8%, Centralchem, Bratislava, Slovakia) and C2H4(OH)2 (99%, Centralchem, Bratislava, Slovakia) in deionised water was then added dropwise to the mixture, which was then

refluxed under a condenser and heated in an oil bath at a temperature of 85–95 ◦C for 2 h. Subsequently, the solvent was evaporated under continuous stirring. The product was dried, calcined at 850 ◦C to a white powder and sieved through a 40 μm sieve. For better usability in relatively thin coatings, the AYZ powder was homogenised and granulated by a freeze-drying process. A flowchart of the process of preparation of AYZ powder is presented in Figure 1.

**Figure 1.** The flowchart of the processing of AYZ powder.

#### *2.2. Pretreatment of the Stainless Steel*

Ferritic refractory stainless steel grade AISI441, which is commonly used in exhaust gas elements, was used as the metal substrate. Prior to cleaning, the steel sheets were cut into 1 <sup>×</sup> 1.5 cm<sup>2</sup> plates to make the samples suitable for further characterisation and testing and to prevent damage to the coated samples by further cutting. This was followed by grinding and chamfering the edges and corners of each sample with sandpaper. To produce adhesive coatings without failure, the surface of stainless steel was treated and cleaned to achieve adhesive coatings with sufficient protective capability at temperatures up to 1000 ◦C. Four different cleaning procedures were applied, i.e., subsequent ultrasonic cleaning in acetone, ethanol and deionised water; sandblasting with glass beads; chemical etching with Kroll's reagent; and a combination of the last two methods. The description of cleaning procedures of the steel is summarised in Table 1.


**Table 1.** The description of cleaning procedures.

#### *2.3. Preparation of the Coatings*

A two-layer PDC coating, composed of a bond coat and a ceramic top coat, was applied. The bond coat was prepared from the commercial polymer Durazane 2250 (Merck KGaA, Darmstadt, Germany) by the dip-coating method (dip-coater RDC 15, Relamatic, Glattburg, Switzerland). The pyrolysis of the bond-coat was carried out in air (Nabertherm® N41/H, Nabertherm, Lilienthal, Germany) at a temperature of 450 ◦C for 1 hour, with heating and cooling rates of 3 K/min. The top coats were prepared from the commercial polymer—Durazane 1800 (Merck KGaA, Darmstadt, Germany), passive fillers and commercial glass. ZrO2 stabilised with 8 mol. % Y2O3 (8YSZ, Inframat® Advanced MaterialsTM, Manchester, CT, USA), AYZ powder prepared by Pechini method and a commercial glass (G018-281, Schott AG, Mainz, Germany) were used as passive fillers. The basic properties of the filler materials are listed in the Table 2.


**Table 2.** Basic properties of filler materials.

Commercially available glass was selected to form a viscous melt at the application temperature of the layers, thereby ensuring the healing of any defects and strengthening of the ceramic top layer. The combination of a liquid commercial polymer Durazane 1800 with glass frits and passive filler particles offers the possibility of designing a large range of compositions. Therefore, the composition of the top layer was designed to match the coefficient of thermal expansion (CTE) of the steel substrate and to reduce the CTE mismatch and increase the compatibility of the metal with the ceramic coating. The CTE of stainless steel was provided by the manufacturer (11.5 <sup>×</sup> <sup>10</sup>−6/K). The CTE of the prepared coatings were estimated by the rule of mixtures using the CTE of Durazane 1800 (3.0 <sup>×</sup> <sup>10</sup>−6/K), 8YSZ (11.5 <sup>×</sup> 10−6/K), AYZ (8.6 <sup>×</sup> 10−6/K) and glass G018-281 (12.1 <sup>×</sup> 10−6/K). Two compositions of top coat were tested, in the following text denoted as C2c and D2-PP. The prepared compositions are listed in Table 3.

**Table 3.** Compositions of the composite top coats after pyrolysis (vol. %).


In the case of the composition C2c, ZrO2 stabilised with 8 mol. % Y2O3 and glass frits were homogenised in a solution of di-n-butyl ether (Acros Organics BVBA, Geel, Belgium) and dispersant (DISPERBYK 2070, BYK-Chemie GmbH, Wesel, Germany). To break up the agglomerates, the suspensions were dispersed in the ultrasound and homogenised for 48 h by stirring with a magnetic stirrer. Subsequently, Durazane 1800 polymer, with 3 wt. % of dicumyl peroxide (DCP) (Sigma-Aldrich Co. LLC., Darmstadt, Germany), was added to the slurry, which was homogenised for an additional 48 hours in a plastic jar with ZrO2 balls (Ø1 mm). After homogenisation, the suspension was applied to the stainless steel with a bond coat by a spray-coating technique from both sides. The suspension was deposited onto steel substrates by spray coating using a spray coater model 780S-SS (Nordson EFD, East Providence, RI, USA). The nozzle diameter of spray gun was 0.71 mm (0.028"). The final suspension was sprayed under the air pressure of 2.2 bar. The distance between the spray gun and the sample was 10 cm. In the composition D2-PP, the AYZ powder prepared by the modified Pechini sol–gel method was used as an additional passive filler. The coated samples were then pyrolysed in air at 850 ◦C for 1 hour, at a heating and cooling rate of 3 ◦C/min. A flowchart of the coating processing is presented in Figure 2.

**Figure 2.** The flowchart of the coatings preparation process.

#### *2.4. Characterisation Methods*

X-ray powder diffraction analysis was used to assign the phase composition of the prepared AYZ powder. Diffraction records were measured on an Empyrean DY1098 powder diffractometer (Panalytical, B.V., Almelo, The Netherlands) with a Cu anode and with X-ray wavelength of λ = 1.5405 Å over 2θ angles of 10–80◦. Diffraction records were then evaluated using HighScore Plus (v. 3.0.4) using COD2019 (Crystallographic Open Database). Mean Roughness Depth (Rz) was measured using atomic force microscopy (AFM, Brooker Innova, Billerica, MA, USA). Rz was calculated by measuring the vertical distance from the highest peak to the lowest valley within five sampling lengths, then averaging these distances. The surface morphology of pretreated samples was examined by scanning electron microscopy (SEM, JEOL JSM 7600 F, JEOL, Tokyo, Japan). For detailed examination of the coating/metal interface, the cross sections were prepared via mounting in resin followed by grinding and polishing. The inspection of the coatings was then performed using an SEM equipped with an energy-dispersive X-ray spectroscopy (EDXS) detector (Oxford instruments, Abingdon, UK) and was focused on the evaluation of adhesion, homogeneity and possible failures of the coatings.

#### *2.5. Oxidation Tests*

The oxidation tests were carried out in a high temperature horizontal electric tube furnace (Clasic 0213T, Clasic, Praha, Czech Republic) in flowing atmosphere of synthetic air (purity 99.5, SIAD Slovakia spol. s.r.o., Bratislava, Slovakia) at a temperature of 950 ◦C with a heating rate of 5 ◦C/min and an exposure time of 48 h. The composition of synthetic air is as follows; nitrogen (78%), oxygen (21%), argon (0.9%) and other gases (0.1%).

#### **3. Results and Discussion**

#### *3.1. Characterisation of the AYZ Filler*

To achieve a homogenous precursor powder to be used as a filler in the prepared coating, the AYZ powder was prepared via a modified Pechini method [27]. The main advantage of this method is that the metallic ions are immobilised in a rigid polymer network, which ensures their homogeneous dispersion on the atomic scale without precipitation or phase segregation. This process allows complete control over the product stoichiometry, even for more complex oxide powders [27]. To facilitate the use of the AYZ powder in the coatings, the powder was refined by milling and freeze drying to avoid agglomeration of precursor powder and achieve the particle size below 10 μm. SEM micrographs of the AYZ powder after freeze drying are shown in Figure 3. From the as-prepared AYZ powder consists of irregular and angular particles resulting from the crushing process, with the sizes ranging from a few to several tens of micrometres.

**Figure 3.** SEM of AYZ powder.

The XRD pattern of AYZ powder after calcination is shown in Figure 4, confirming the presence of t-ZrO2, as well as yttrium aluminium garnet (Y3Al5O12, YAG) as the main crystalline phases formed during calcination. Smaller amount of the mellilite Y2Al2O6 phase was also observed after calcination. High background in the diffraction pattern indicates the crystalline phases were embedded in an amorphous (glassy) matrix.

**Figure 4.** The X-ray powder diffraction (XRD) pattern of AYZ powder after calcination at 850 ◦C.

#### *3.2. Treated Steel Surfaces*

The metal surface quality is significant characteristic influencing the adhesion of the protective coating. In the case of double-layer PDC coatings, the weakest point is usually the interface between the bond coat and the metal substrate, due to the presence of impurities or defects on the steel surface, which often lead to delamination of the bond coat during pyrolysis or after corrosion tests. To ensure a high adhesion of the bond coat with the steel, the stainless steel substrate has to be cleaned properly to degrease its surface and remove possible contaminants. The influence of different cleaning procedures on the surface morphology was therefore examined. The Mean Roughness Depth (Rz) was determined using AFM and the surface morphology of pretreated stainless steel samples was examined by SEM. Figure 5 presents the Rz of variously pretreated stainless steel surfaces. The AFM analysis confirmed that different treatment created different sizes of roughness: there is a relation between surface topography and the type of the used cleaning procedure. The Rz was in the range of 0.26 to 1.69 μm. A roughness similar to that observed for untreated surfaces was measured after ultrasonic cleaning, while sandblasting increased the roughness substantially. An increase of Rz was expected to result in stronger bonding at the steel-bond coat interface.

**Figure 5.** The roughness after pretreatment of stainless steel (U – ultrasonic cleaning, K – chemical etching with Kroll's reagent, S – sandblasting, S+K – sandblasting + chemical etching with Kroll's reagent).

AFM images of the stainless steel cleaned by different methods are shown in Figure 6. SEM-micrographs featuring surfaces of the stainless steel cleaned by different methods are shown in Figure 7. Chemical etching resulted in a surface with irregular topography, with slight roughening of the surface compared to ultrasonically cleaned substrates. More uniform and regular surfaces were obtained by sandblasting with glass beads, or a combination of sandblasting and chemical etching with Kroll's reagent. These treatments resulted in rough surfaces with rounded edges, with the average surface roughness 1.69 μm for sandblasted samples. Chemical etching of sandblasted substrates slightly decreased the average surface roughness, but the sandblasting induced roughening was still significant.

**Figure 6.** Atomic force microscopy (AFM) images representing roughness of the stainless steel surfaces treated by different cleaning procedures. The applied cleaning procedure is indicated by the abbreviation in the upper left corner of each figure (U – ultrasonic cleaning, K – chemical etching with Kroll's reagent, S – sandblasting, S+K – sandblasting + chemical etching with Kroll's reagent).

**Figure 7.** SEM micrographs of the stainless steel surfaces treated by different cleaning procedures. The applied cleaning procedure is indicated by the abbreviation in the upper right corner of each figure (U – ultrasonic cleaning, K – chemical etching with Kroll's reagent, S – sandblasting, S+K – sandblasting + chemical etching with Kroll's reagent).

#### *3.3. Characterisation of the Coatings*

The SEM cross-sectional micrographs of compositions C2c and D2-PP were obtained through the metal–ceramic interface to investigate the bonding between the bond coat and variously treated steel substrates (Figure 8 (C2c composition) and Figure 9 (D2-PP composition)). The weakest location in a typical double layer coating is usually the interface between the stainless steel and the bond coat, where the cracks or spallation can occur. The spallation of the bond coat is usually caused by thermal and elastic mismatch between the steel and bond coat, due to the presence of impurities at the steel surface, by changing the chemistry of the steel by chemical etching, or growth of stresses followed by the formation of thermally-grown oxides, mostly due to the weak adhesion of bond coat to steel. Irrespective of the applied surface treatment the bond coat did not delaminate from the steel surface after pyrolysis of the coatings at 850 ◦C, indicating its good adhesion. An undamaged bond coat approximately 1 μm thick was observed in all cases, which acts as an effective diffusion barrier to oxidation during pyrolysis. Only for the D2-PP coating deposited at the substrate etched by Kroll's reagent the bond coat peeled off from the surface. If the stainless steel was treated by sandblasting or chemical etching or their combination, a few small cracks (marked with white circles) were generated in the bond coat, perpendicular to the substrate surface. The crack formation was attributed to strong adhesion of the bond coat to the metal substrate and, at the same time, by the high volume shrinkage of the polysilazane during heat treatment. No crack formation in the bond coat was observed on samples pretreated via ultrasonic cleaning in acetone, ethanol and deionised water. Cracking could also occur due to uneven and rough sandblasted surface, with sharp edges and peaks acting as stress concentrators in the coatings [24]. The pretreatment by ultrasonic cleaning in acetone, ethanol and deionised water was found to be the most effective process, since no spallation or cracking was observed in the cross section of the bond coat after pyrolysis. Moreover, the ultrasonically cleaned steel surface was uniformly covered by the approximately 1 μm thick bond coat.

**Figure 8.** SEM cross sections of pyrolysed C2c coatings. The applied cleaning procedure is indicated by the abbreviation in the upper right corner of each figure (U – ultrasonic cleaning, K – chemical etching with Kroll's reagent, S – sandblasting, S+K – sandblasting + chemical etching with Kroll's reagent).

**Figure 9.** SEM cross sections of pyrolysed D2-PP coatings. The applied cleaning procedure is indicated by the abbreviation in the upper right corner of each figure (U – ultrasonic cleaning, K – chemical etching with Kroll's reagent, S – sandblasting, S+K – sandblasting + chemical etching with Kroll's reagent).

In addition, EDXS mapping was carried out on the cross section of D2-PP-coated steel cleaned by ultrasonic treatment in acetone, ethanol and deionised water to demonstrate the existence of an interface bond metal/base coat and base/top coat. EDXS element maps are shown in Figure 10. The bond coat contains mainly Si and O, since during pyrolysis in air the Durazane 2250 was converted to SiO2, as confirmed by SEM/EDXS measurement. The presented element maps of Si confirm the formation of the protective bond coat at the steel surface in agreement with the literature [15]. The bond coat enhances the bonding between the steel and the top coat, and it preserves the steel from oxidation/corrosion [15,27]. On steel exposed to ambient environment, a natural oxide layer with adsorbed water is always present. Because of the reactivity of Durazane 2250 with surface-bound –OH groups, steel forms direct metal–O–Si chemical bonds with the base coat, leading to excellent adhesion [28]. The reaction of Durazane 2250 with hydroxyl groups of the substrate surface is described by the following simplified reaction equations [29]:

$$\text{Fe-OH} + \text{\#Si-NH-Si} \rightleftharpoons \text{\#Fe-O-Si} + \text{H}\_2\text{N-Si} \tag{1}$$

$$\text{Fe-OH} + \text{H}\_2\text{N-Si} \rightleftharpoons \text{\#Fe-O-Si} \text{\#} + \text{NH}\_3 \tag{2}$$

and a direct chemical metal–O–Si bonds between the steel and the precursor-derived coatings are formed. Therefore, the adhesive strength of the PDC coatings to the metal surface is very strong. In the case of sandblasting, we assume that it was the surface roughness of the steel that caused the coating delamination, as sandblasting should not affect the concentration of hydroxyl groups present at the steel surface. However, strong adhesion due to formation of the metal–O–Si bonds causes immobility of the coating during pyrolysis, which does not allow the coating to adjust to volume shrinkage of the steel substrate. Moreover, the sharp borders and peaks of the substrate initiate the formation of cracks perpendicular to the substrate surface. The corrosion/oxidation medium is thus able to penetrate through the cracks to the metal surface causing coating delamination. No crack formation in the bond coat was observed in samples pretreated via ultrasonic cleaning. In the case of chemical

etching, we assume that this treatment negatively influenced the adhesion of the bond coat because the concentration of hydroxyl groups on steel surface was significantly affected by etching.

**Figure 10.** SEM/energy-dispersive X-ray spectroscopy (EDX) of D2-PP coatings documenting distribution of elements in the coating and in substrate.

The main elements of stainless steel Fe and Cr were detected below the protective bond coat. The main component of the bond was Si. The top coat consisted of Zr, Al, Y (from AYZ powder) and Ba (from commercial glass).

The top coat layers of both studied compositions—C2c (Figure 8) and D2-PP (Figure 9)—were almost dense, containing only small closed pores present predominantly at the boundaries between the filler particles and the matrix. The filler particles were well coated with the Durazane 1800 precursor, which builds up the matrix, and acts also as an adhesive between the particles. In the case of the C2c coating, all pores were almost spherical. This indicates that the closed pores could result from bubble formation due to the release of dissolved gases, as well as the expansion of insoluble gases (e.g., oxygen or air) entrapped in the initial pores. The pores could result also from the release of gases such as NH3, CH4, and H2 generated during the polymer to ceramic transformation [15]. Note that pore formation cannot be completely avoided when passive fillers are used in the processing of PDC coatings, and some residual porosity usually remains in the final ceramic top coats. Existing pores provide a transportation path for gaseous products of decomposition escaping the coating. In the case of the D2-PP coating, the addition of the AYZ powder with irregular and angular particles have helped to create a rigid and articulate structure. This structure allowed outgassing of the preceramic polymer pyrolysis products from the system thereby effectively reducing the size and amount of pores. The elimination of larger pores, and thus an increased density, led to a significantly more compact coating in comparison to the C2c composition pyrolysed under the same conditions. Although the top layers were not completely dense, microstructure with residual porosity was beneficial for the thermal stability of coatings, and contributed to the mitigation of residual stresses during the heating and cooling cycles [30].

The mismatch of the CTE between the steel and the coating, together with the Young modulus, are critical factors for the resistance of coatings exposed to high temperatures [31]. By reducing the

non-conformance of the CTE and the elastic modulus, the tension in the coating is reduced, resulting in a higher stability of the coatings during thermal loading. Moreover, glasses are suitable material for sealing in PDC coatings. By decreasing both quantities (CTE mismatch and Young modulus), the stresses in the layers can be reduced, leading to better coating thermal stability during the oxidation tests. Coatings with both ceramic and glass fillers should then exhibit better thermal compatibility with the steel substrate than the coatings with active and passive fillers without glass tested in our previous work [26]. As a result, no cracks were observed in top coats, preventing penetration of oxidation agents through the coating and attack of the steel substrate.

Static oxidation tests were performed to assess the efficiency of the selected types of steel substrate cleaning. The aim of the oxidation tests was to determine whether the tested cleaning methods would ensure sufficient adhesion of the bond coat to the metal substrate not only during pyrolysis, but also at long-term operating at high temperature. The tests were performed in the atmosphere of synthetic air at 950 ◦C at a heating rate of 5 ◦C/min, with maximum duration of 48 hours. SEM cross-sections of C2c and D2-PP compositions after oxidation tests are shown in Figures 11 and 12, respectively. In all cases, except for ultrasonic cleaning, spallation of the bond coat from the surface of stainless steel was observed. Chemical etching of steel with the Kroll's solution probably caused changes in the microstructure and chemical composition of the steel, resulting in peeling of the coating. For sandblasted samples, or a combination of sandblasting and etching, detachment of the bond coat resulted from high surface roughness: as discussed above, strong adhesion of the coating on a rough surface with sharp edges resulted in formation of cracks perpendicular to the substrate. The cracks serve as a gateway for inward penetration of oxygen, which directly attacks the substrate, creating a layer of oxidation products, and eventually leading the detachment of the bond coat from the substrate. Simple cleaning by ultrasonication in acetone, ethanol and deionised water was found to be the most effective way for achieving a sufficient bonding of the bond coat to the steel substrate after oxidation tests.

**Figure 11.** Cross sections of C2c coatings after oxidation tests. The applied cleaning procedure is indicated by the abbreviation in the upper right corner of each figure (U – ultrasonic cleaning, K – chemical etching with Kroll's reagent, S – sandblasting, S+K – sandblasting + chemical etching with Kroll's reagent).

**Figure 12.** Cross sections of D2-PP coatings after oxidation tests. The applied cleaning procedure is indicated by the abbreviation in the upper right corner of each figure (U – ultrasonic cleaning, K – chemical etching with Kroll's reagent, S – sandblasting, S+K – sandblasting + chemical etching with Kroll's reagent).

Originally, numerous pores were observed across the whole cross-section of the as-prepared PDC glass ceramic coatings. After oxidation tests, a significant increase in the porosity of the layers accompanied by the growth of pores and a decrease of the coating thickness was observed in C2c coating. Decrease of the layer thickness was attributed to differential sintering of individual components (mainly glass fillers) by viscous flow, complemented by inadequate removal of gases entrapped in the ceramic matrix. Due to the softening of the used glass fillers, the cracks in the layers gradually heal, indicating at least partial protection of the metal substrate. After oxidation tests, the D2-PP composition exhibited a compact structure with no increase of the size and amount pores. No cracking or delamination from bond coat was observed.

#### **4. Conclusions**

Suitable pretreatment of steel substrate, as well as the using a Durazane 2250 bond coat, are prerequisites for preparation of adherent composite coatings. The most effective cleaning process is a 3-step ultrasonic cleaning in acetone, ethanol and deionised water. Small cracks in the bond coat perpendicular to substrate were observed after pyrolysis in bond coats deposited at substrates treated by sandblasting and chemical etching. After oxidation tests, all coatings, except for those applied to substrates cleaned in an ultrasonic bath, delaminated or showed significant cracking of the bond coat. Combination of PDC with tailored fillers and glass systems enable the processing of dense and crack free coating system on stainless steel. Irregular and angular filler particles favour outgassing of the coating during pyrolysis, reducing the total porosity in the layer, and conferring better protection of the substrate against oxidative environment.

**Author Contributions:** M.P. performed cleaning procedures, prepared coatings and wrote the manuscript; I.P. prepared the precursor powder, performed cleaning procedures, prepared coatings and performed oxidation tests; P.Š. performed the SEM and EDXS analysis; G.M. and D.G. conceived the study, supervised the project, gave

research guidelines, and finalised the manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** Financial support of this work by the APVV 0014-15 grant and the Deutscher Akademischer Austauschdienst (DAAD) grant scheme is gratefully acknowledged. This paper was also created in the frame of the project Centre for Functional and Surface Functionalised Glass (CEGLASS), ITMS code 313011R453, operational program Research and innovation, co-funded from European Regional Development Fund.

**Acknowledgments:** The authors would like to thank Alexander Horcher and Mateus Lenz-Leite for their help to create the coating solution. The authors would also like to thank Jacob Andrew Peterson for grammar corrections in the manuscript and Branislav Hruška for AFM measurement.

**Conflicts of Interest:** The authors declare no conflicts of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*

## **Ni**/**cerium Molybdenum Oxide Hydrate Microflakes Composite Coatings Electrodeposited from Choline Chloride: Ethylene Glycol Deep Eutectic Solvent**

#### **Juliusz Winiarski 1,\*, Anna Niciejewska 1, Jacek Ryl 2, Kazimierz Darowicki 2, Sylwia Ba´slady ´nska 1, Katarzyna Winiarska <sup>3</sup> and Bogdan Szczygieł <sup>1</sup>**


Received: 30 January 2020; Accepted: 17 February 2020; Published: 19 February 2020

**Abstract:** Cerium molybdenum oxide hydrate microflakes are codeposited with nickel from a deep eutectic solvent-based bath. During seven days of exposure in 0.05 M NaCl solution, the corrosion resistance of composite coating (Ni/CeMoOxide) is slightly reduced, due to the existence of some microcracks caused by large microflakes. Multielemental analysis of the solution, in which coatings are exposed and the qualitative changes in the surface chemistry (XPS) show selective etching molybdenum from microflakes. The amount of various molybdenum species within the surface of coating nearly completely disappear, due to the corrosion process. Significant amounts of Ce3<sup>+</sup> compounds are removed, however the corrosion process is less selective towards the cerium, and the overall cerium chemistry remains unchanged. Initially, blank Ni coatings are covered by NiO and Ni(OH)2 in an atomic ratio of 1:2. After exposure, the amount of Ni(OH)2 increases in relation to NiO (ratio 1:3). For the composite coating, the atomic ratios of both forms of nickel vary from 1:0.8 to 1:1.3. Despite achieving lower corrosion resistance of the composite coating, the applied concept of using micro-flakes, whose skeleton is a system of Ce(III) species and active form are molybdate ions, may be interesting for applications in materials with potential self-healing properties.

**Keywords:** metal coatings; nickel; composite coatings; electrodeposition; XPS; polarization; electrochemical impedance spectroscopy (EIS)

#### **1. Introduction**

In recent years, the electrodeposition of metals from ionic liquids (ILs) has become more popular because of the many advantages, such as high solubility metals salts, high conductivity, and wide electrochemical potential windows. Ionic liquids also allow obtaining metal coatings, such as Na and Mg, which is impossible in conventional baths. ILs are non-toxic, non-flammable and can be used in a wide range of temperatures [1–4]. One of the most interesting analogues of ILs are the deep eutectic solvents (DESs). DES is a eutectic mixture of quaternary ammonium salt with hydrogen bond donor species or metal salts. DESs are relatively cheap and easy to prepare [5,6]. DES-based galvanic baths have the advantage that the metals do not passivate during deposition. DESs can be used to electrodeposition a lot of metal coatings: Ni, Cu, Zn, Cr, Sn, Co, Al, Ti, W, Ag, Mg, Se, and Pd [6]. Nickel coatings obtained from DES feature nano-crystalline morphology and lower

surface roughness than Ni-coatings from conventional aqueous Watt's baths which produce coatings with a micro-crystalline structure. Furthermore, the microhardness of Ni coatings deposited from DES is 100 HV (Vickers hardness) higher than coatings from aqueous solutions [7]. A used eutectic solvent has a significant impact on the morphology of nickel coatings (size of Ni crystallites) [8]. Moreover, from baths based on eutectic solvents it is possible to obtain Ni coatings on an aluminum substrate. Obtaining bright Ni coatings requires the use of brighteners, which differ from those used in aqueous plating baths [9,10]. The important influence on the coating's deposition from DESs is the water content in a plating bath. By changing the water content, one can control the potential for Ni deposition. Potential Ni impacts the self-limiting growth crystallite and passivation effect [11]. In addition to the above advantages, DESs liquids enable the electrodeposition of nickel alloy coatings: Zn-Ni [12], Ni-Co-Sn [13], Fe-Ni [14], Ni-Co [15], Ni-Mo [16], and nickel composite coatings. Ni-carbon nanotubes composite coatings [17,18] have much better tribological properties than pure Ni [18]. Furthermore, Ni-PTFE (poly(tetrafluoroethylene)) coatings show better wear resistance than nickel coatings and have hydrophobic properties [19]. The addition of SiC to the nickel matrix causes the increase of microhardness [20] and corrosion resistance [21]. Ni-SiO2 coatings also have better microhardness and wear resistance than Ni coatings [22]. Nickel coatings with a TiO2 dispersive phase show good electrocatalytic activity [23]. However, co-deposition with micro- or nano-particles of compounds of chemically active elements, e.g. cerium or molybdenum, may affect the protective properties of the composite coatings. Dong et al. prepared cerium molybdate particles with different morphologies, such as the bundle-like, flower-like, and microspheric structures [24]. The nanostructure of Ce-Mo shows excellent adsorption and catalytic properties [24]. Amino acids, for example: glycine, lysine, serine, and glutaminate used at synthesis of cerium molybdate have influence on its morphology [25]. Sajad et al. proved that by using lysine in synthesis one can obtain a pure product (cerium molybdate) which is not contaminated with cerium molybdate oxides [26]. Nanocapsules are also made up of shells from cerium molybdate, while the core was a corrosion inhibitor [27,28] or hollow nanospheres have an antibacterial effect on *Escherichia coli* in the absence of light [29]. Hence their potential application in medicine or the anti-corrosion industry [29]. Patel et al. developed a method of fabrication of cerium zinc molybdate nanopigment with anticorrosive properties [30]. Cerium molybdate nanowires were also added to the coatings obtained by the sol-gel method [31]. Therefore, cerium molybdate is very popular as a corrosion inhibitor in protective coatings for aluminum alloys [32] and magnesium [33,34]. Bhanvase et al. have investigated that 5% content of cerium zinc molybdate nanocontainers, significantly increases the corrosion resistance [35].

The use of cerium molybdenum oxide, proposed in this work, is novel in electrodeposition of nickel composite coatings. Ce and Mo compounds embedded in the coating can modify its protective properties by giving it potential self-healing properties (through the selective release of cerium and molybdenum ions). Moreover, this work uses a deep eutectic solvent based on choline chloride and ethylene glycol as an excellent environment for the co-deposition of microparticles. As a result, deposition is carried out in a stable suspension bath. Microflakes of fairly large size are intentionally used to make it easier to capture the effect of their interaction with a metallic nickel matrix and the corrosive environment. Emphasis is placed on understanding qualitative and quantitative changes in surface chemical composition—X-ray photoelectron spectroscopy (XPS), caused by the presence of microflakes, on the one hand, and the exposure to a corrosive environment, on the other. The corrosion resistance of a newly obtained material is monitored in-situ by electrochemical impedance spectroscopy (EIS) and basic *dc* (direct current) polarization methods. The morphology, structure, and topography of the coatings are investigated by scanning electron microscopy (SEM), X-ray diffraction (XRD), and contact profilometry.

#### **2. Materials and Methods**

Cerium molybdenum oxide microflakes were synthesized at low temperature from Ce(NO3)3 and Na2MoO4 precursors through the precipitation method. First, the two precursors were prepared separately by the dissolution of Na2MoO4 2H2O and Ce(NO3)3 6H2O in demineralized water in the molar ratio of 2:1, respectively. The solutions were next cooled down to 8 ◦C overnight in the fridge. Na2MoO4 solution was instilled slowly (60 mL min−1) into Ce(NO3)3 solution under continuous mechanical stirring (200 rpm), which resulted in the formation of a light yellow suspension. After 1 h mixing at 15 ◦C, the precipitate was filtered under the reduced pressure on a Buchner funnel. The grey-yellowish precipitate was washed several times with demineralized water and then dried in a vacuum dryer.

The plating bath for nickel electrodeposition (100 mL plating volume) was prepared by mixing choline chloride and ethylene glycol in a 1:2 molar ratio and addition of 1 mol dm−<sup>3</sup> NiCl2 6H2O at 70 ◦C. All these reagents were mechanically stirred until a homogeneous green liquid was obtained. Cerium molybdenum oxide microflakes were then added (2 g dm<sup>−</sup>3) and the bath was homogenized for 30 min (UP50H – Hielscher Compact Lab Homogenizer).

For the electrodeposition, copper disks (1.5 cm diameter, 0.1 cm thick) were used as cathodes. They were polished on 600–1200 grade abrasive paper. Then, the samples were degreased in an ultrasonic cleaner in methanol, acid etched (10 vol.% H2SO4), rinsed in demineralized water and again in methanol. Electroplating was carried out in a thermostatic electrolyzer consisting of two anodes (pure nickel) and a cathode placed between them. The process was carried out at a current density 6 mA cm−2, at 70 ◦C, for 1 h. After this time, the coatings were twice rinsed in demineralized water and methanol. At the end, the coatings were dried and stored in a vacuum desiccator until required.

Surface morphology was analyzed by Quanta 250 (FEI) scanning electron microscope equipped with an Octane Elect Plus SDD microanalyzer (operating at 25 kV, and at 10−<sup>4</sup> Pa pressure).

The crystal structure and phase composition of cerium molybdenum oxide microflakes and blank and composite nickel coatings were analyzed using powder X-ray diffraction. The XRD diffractograms were recorded on a D 5000 (Siemens) diffractometer with CuKα radiation (λ = 0.15409 nm) at room temperature in the 2θ range: 10–60◦ for blank and composite Ni coatings, and 5–60◦ for cerium molybdenum oxide microflakes. Phase identification was carried out by comparing the experimental patterns with the reference patterns collected in the Powder Diffraction Files database (International Centre for Diffraction Data PDF-2 base).

The X-ray Photoelectron Spectroscopy (XPS) measurements were performed using Escalab 250Xi (Thermo Fisher Scientific, United Kingdom). Al Kα monochromatic X-ray source with spot diameter of 250 μm was used. The set up pass energy was 20 eV. Charge compensation was controlled through low-energy Ar<sup>+</sup> ions emission by means of a flood gun, with the final calibration made using Ni2p3 metallic peak component at 852.6 eV. The deconvolution procedure was performed using Avantage software (Thermo Fisher Scientific, Waltham, USA).

The chemical analysis of corrosive solutions was performed by ICP-OES (Inductively Coupled Plasma-Optical Emission Spectrometry) and ICP-MS (Inductively Coupled Plasma-Mass Spectrometry) technique. The Ni concentration was determined using ICP-OES method (Vista MPX, Varian, Australia). Cu, Ce, Mo, and also Ni content was determined using ICP-MS method (XSeries2, Thermo Fisher Scientific, USA). The analyses were performed in the Chemical Laboratory of Multielemental Analysis, Wrocław University of Science and Technology, Poland, accredited by the Polish Centre for Accreditation (AB 696). The laboratory has a measurement procedure of determination of metals in water in its scope of accreditation. The procedure is based on EN-ISO 11885:2009, PN-EN ISO 17294-1:2007, and PN-EN ISO 17294-2:2006.

Corrosion measurements were carried out at 25 ◦C in 0.05 mol dm−<sup>3</sup> solution of NaCl in Autolab 400 mL corrosion cell using a Reference 1010E (Gamry, Warminster, USA) potentiostat/galvanostat/ZRA (zero resistance ammeter). The geometric area of the working electrode exposed to the solution was 1 cm2. A stainless steel rod (geometric area 5 cm2) and a saturated Ag|AgCl / 3M KCl electrode (Metrohm) mounted in a Luggin capillary were used as the counter and reference electrodes, respectively. Potentiodynamic polarization curves were recorded after 168 hours exposure starting from −0.1 V to +0.5 V vs. open circuit potential (*E*OC) at a scan rate of 0.166 mV s−1. The polarization resistance (LPR technique) was measured by polarization of the sample starting from −10 mV to +10 mV vs. *E*OC at a scan rate of 1 mV s<sup>−</sup>1. Impedance spectra for electrochemical impedance spectroscopy (EIS)

were recorded at *E*OC (potentiostatic mode) with a resolution of 10 pts/dec., in a frequency range from 100 kHz to 0.001 Hz and at a 10 mV signal amplitude. Equivalent circuit modeling, graphing, and analysis of impedance data was performed using ZView® software (Scribner Associates, version 3.5g). The deposition process was studied by cyclic voltammetry (CV) technique using Autolab RDE-2 electrode with a Pt tip (3 mm diameter) in a electrochemical vessel (20–90 mL, Metrohm) with thermostat jacket, Ag and Pt wires as the reference and counter electrodes, in the range of potentials from *E*OC to <sup>−</sup>1.5, then 1.5 and ending at *E*OC with a scan rate of 20 mV s<sup>−</sup>1. Pt tip instead of Cu was chosen to avoid the influence of any copper ions on the current-potential characteristics.

#### **3. Results and Discussion**

#### *3.1. Morphology, Topography, and Phase Structure of Blank and Composite Ni Coatings*

After electrodeposition, both blank and composite nickel coatings with a mean thickness of 10 μm were subjected to SEM and XRD analysis. The surface of the Ni coating consisted of spheroidal particles that form randomly arranged larger agglomerates (Figure 1a). The surface of the Ni/CeMoOxide coating, like the Ni coating, consisted of spherical particles, which were noticeably smaller in this case. In addition, there were randomly embedded microflakes of CeMoOxide within the surface. Single cracks were visible, mostly spreading from the places where the microflakes are built-in (Figure 1b). Additional topographic measurements of the surface of both coatings confirmed the increase in roughness of the composite coating, expressed by the Ra parameter equaled to 122.3±17.7 nm (related to 92.6 ± 10.1 nm for blank Ni coating). For comparison, Figure 1c shows the photomicrograph (recorded in a secondary electron - SE mode of SEM) of synthesized cerium molybdenum oxide hydrate powder. The plates are overgrown and form agglomerates with a length exceeding 20 μm, a width below 10 μm, and a thickness of 100–200 nm (Figure 1c).

**Figure 1.** Surface morphology of (**a**) blank and (**b**) composite Ni coatings deposited on the copper substrate at a current density of 6 mA cm−<sup>2</sup> at 70 ◦C for 1 h, and (**c**) as-synthesized cerium molybdenum oxide hydrate microflakes for comparison. Pictures were recorded in a SE mode of SEM.

*Materials* **2020**, *13*, 924

Figure 2 shows the diffractograms for of cerium molybdenum oxide microflakes and blank and composite nickel coating. The thickness of the prepared Ni coatings was sufficient to perform the XRD analysis. The dominant peaks at the diffractograms are characteristic of Ni (111) and (200) crystal planes visible at 2θ 44.6◦ and 51.9◦. The indicated reflections were well indexed to these collected in ICDD JCPDS card No. 00-004-0850 for Ni. Peaks at 2θ 43.3◦ and 50.4◦ originated from the Cu substrate (JCPDS card No. 00-004-0836). The other small peaks match well with the cerium molybdenum oxide. The diffractogram recorded for as-prepared microflakes is also shown in Figure 2. All the diffraction peaks correspond to Ce2(MoO4)3 4.5H2O phase (JCPDS card No. 00-031-0333) without the presence of other peaks from any impurities. The strong and sharp diffraction peaks indicate that the prepared cerium molybdenum oxide microflakes were well-crystallized. High-resolution diffractograms were recorded for the blank and composite Ni coatings to estimate the average crystallite size. A copper peak from the substrate overlapped the nickel peak (111), therefore, for the proper determination of FWHM (full width at half maximum) for the Ni (111) peak, a deconvolution using a pseudo-Voigt function was performed. The average crystallite size was calculated using the Scherrer Equation (1):

$$D = \frac{K \cdot \lambda}{B \cdot \cos \theta} \tag{1}$$

where D is the mean crystallite size in the direction perpendicular to the (hkl) plane of reflexes in nm, K is a Scherrer constant (0.9) [36], λ = 0.154 nm is the X-ray wavelength used in the measurement, B was calculated from Equation (2):

$$B = \sqrt{(\beta\_{FWHM}^2 - \beta\_0^2)}\tag{2}$$

**Figure 2.** X-ray diffractograms for blank and composite Ni coatings deposited on the copper substrate at a current density of 6 mA cm−<sup>2</sup> at 70 ◦C for 1 h. A powder diffraction pattern for as-synthesized cerium molybdenum oxide hydrate is also shown at the bottom of the figure for comparison.

(βFWHM and β<sup>0</sup> is the FWHM of diffraction peak at angle θ and the corrected instrumental broadening - in radian, respectively). After deconvolution, the average crystallite size was estimated at about 10.4 nm and 6.3 nm for blank Ni and composite Ni coatings respectively. The resulting D value indicates the nano-crystalline structure of the electrodeposited nickel. It can be seen that the cerium molybdenum oxide microflakes affects the structure of the coating by co-depositing and reducing the size of the crystallites.

On the basis of the observed changes in the morphology and structure of both types of coating, it is obvious that the cerium and molybdenum compound must have modified the course of cathodic processes. Therefore, cyclic voltammetry curves were recorded under deposition conditions in both

(blank Ni and suspension) plating baths. The experiment was repeated twice to avoid accidental results. Representative CV curves are shown in Figure 3. Analyzing the course of the curves, it can be seen that the composite coating deposition can proceed probably with greater efficiency. This can be demonstrated by a much larger anodic peak within the potentials of Ni (deposit) oxidation (Figure 3). It is however not excluded that the larger area of that peak may be associated with, e.g., the oxidation of molybdenum species released at high temperature (70 ◦C) from the microflakes suspended in the plating bath. However, the cathode curves were not significantly different (Figure 3).

**Figure 3.** Cyclic voltammetry curves recorded in a blank Ni plating bath and suspension bath containing 2 g dm−<sup>3</sup> Ce2(MoO4)3 4.5H2O microflakes, at 70 ◦C on a rotating disk electrode with Pt tip (3 mm diameter), at 600 rpm.

#### *3.2. Corrosion Resistance*

#### 3.2.1. *dc* Polarization

Polarization resistance (*R*p) as a function of exposure time in 0.05 mol dm−<sup>3</sup> solution of NaCl for blank Ni and composite (Ni/CeMoOxide) coatings is presented in Figure 4. In the case of a blank Ni coating, a clear increase in *R*<sup>p</sup> was observed. During exposure in a NaCl solution, *R*<sup>p</sup> increased up to 536.6 kΩ cm2 after about 144 hours, then, its value decreased. The *R*p-time dependency for the Ni/CeMoOxide coating was different. An increase of *R*<sup>p</sup> to above 100 kΩ cm<sup>2</sup> was visible within the first hours of exposure in a chloride solution. After approximately 72 hours, the *R*p value decreased to approximately 77 kΩ cm2, and then it showed a slight upward trend (Figure 4).

The corrosion potential (*E*corr) of the Ni/CeMoOxide coating increased quite rapidly in the first hours (Figure 5). After about three days it reached a constant level. The potential-time dependency for the Ni coating was different. An increase in *E*corr towards more positive values was noticeable throughout the entire time studied. It is possible that such a direction of changes resulted from qualitative changes in the morphology of the layer of nickel oxidation products.

Finally, after 168 hours of exposure—after reaching a certain stability of the potential (at least for the composite coating), polarization curves for Ni and Ni/CeMoOxide coatings were registered. Their course is presented in Figure 6. The anodic branch of the Ni/CeMoOxide coating, compared to the anodic branch of the Ni coating, is characterized by higher corrosion current density values, which suggests more active anodic process on the surface of the former. A section with slowly a growing density of the corrosion current is visible on the anodic side in the course of Ni coating (Figure 6), which may suggest some protective/barrier action of a passive layer.

**Figure 4.** Polarization resistance (*R*p) recorded for blank and composite Ni coatings during 168-hour exposure in 0.05 M NaCl solution.

**Figure 5.** Corrosion potential (*E*corr) recorded for blank and composite Ni coatings during 168-hour exposure to 0.05 M NaCl solution.

**Figure 6.** Potentiodynamic polarization curves recorded for blank and composite Ni coatings after 168-hour exposure to 0.05 M NaCl solution.

#### 3.2.2. Electrochemical Impedance Spectroscopy

Impedance spectra of both types of coating were recorded every 24 hours during 7 days of exposure of the samples in 0.05 mol dm−<sup>3</sup> NaCl solution. They have been presented on a complex Bode plot (Figure 7) to gain better visibility of the changes of the investigated system. The first analysis of the shape of the spectra leads to the conclusion that the corrosion process of Ni coating is characteristic of "corroding coating". The magnitude of impedance |Z| increased during exposure (Figure 7a), which may indicate increasing corrosion resistance, due to the formation of, e.g., a stable passive layer or increasing diffusion constrains (especially after 144 and 168 hours). A similar shape of impedance spectra and similar tendency (increasing) for the impedance modulus have been observed in previous work by Urcezino et al. [37] and J. Winiarski et al. [38].

**Figure 7.** Bode representation of the impedance spectra recorded for (**a**) blank and (**b**) composite Ni coatings for 168 hours exposure in 0.05 M NaCl solution.

Figure 7b presents impedance spectra for composite coating. This material behaved slightly differently, rather like "damaged coating", due to the presence of coating incontinuities. Furthermore, not very clearly separated time constants—caused by capacitive dispersion and overlapping of time constants—were observed (Figure 7b). Therefore, for further discussion, two electric equivalent circuits (EECs) were used for the calculation of the theoretical spectra: a single time constant model for a blank Ni (Figure 8a) and a double time constant model for a composite Ni (Figure 8b).

**Figure 8.** Electric equivalent circuits used for fitting the experimental spectra of (**a**) blank and (**b**) composite Ni coatings.

Both models use a constant phase element (CPE). The impedance of the CPE is defined by Equation (3), where: *Y*<sup>0</sup> is a time constant parameter (Ω−<sup>1</sup> cm−<sup>2</sup> sα), ω is the angular frequency of the AC signal and α is the CPE exponent.

$$Z\_{\rm CPE} = \mathcal{Y}\_0^{-1} (j\omega)^{-a} \tag{3}$$

The other elements used in the EECs presented in Figure 8a,b are: *R*s—the resistance of NaCl solution, *R*ct—the charge transfer resistance associated with nickel oxidation, *Y*0,dl and α corresponds to a double layer capacitance (*C*dl). In the circuit for a composite coating (Figure 8b), the properties of a developed surface layer of corrosion product was modeled using *R*film (the resistance of electrolytic solution in a porous layer) and *C*film (the dielectric properties of this layer). Both EECs yielded a very good fit to the experimental data (parameter χ<sup>2</sup> in order of 10–4–10–5) and low residual errors (0.2%–8%). However, it should be noted that the fitting was made after manually reducing the frequency range to ca. 10 kHz–0.08 Hz. This interference was intentional to avoid the influence of instability of the measured system, especially in the low frequency range, on the interpretation of the physical meaning of EECs elements and, finally, their values (collected in Tables 1 and 2).


**Table 1.** Values of electric elements calculated for EEC of blank Ni coating.


**Table 2.** Values of electric elements calculated for EEC of composite Ni coating.

For a blank Ni coating, the values of *R*ct increased with increasing exposure time and finally exceeded 0.5 MΩ cm2 after 7 days (Table 1). A similar tendency was observed for the *R*<sup>p</sup> determined on the basis of the LPR technique, see Figure 4. This marked increase in *R*ct was accompanied by only minor changes in CPE parameters (*Y*<sup>0</sup> and α), whose values indicate the absence of significant changes in (e.g., homogeneity) of the surface at which the corrosion process occurs. In the case of a composite coating (Table 2), it was also noted that the direction of changes of a charge transfer resistance was close to the polarization resistance values obtained from the LPR method (Figure 4). Initially, for the first 48 hours, there was a drop of *R*ct from 75 to 59 kΩ cm2, but after 72 hours exposure, the values of *R*ct increased with increasing time and finally exceeded 100 kΩ cm<sup>2</sup> after 7 days (Table 2). For this coating the contribution of a resistance connected with the presence of a developed surface layer of corrosion products is not to be missed—*R*film values increased from 5.6 to 8.8 kΩ cm<sup>2</sup> over 7 days exposure (Table 2).

#### *3.3. XPS Surface Analysis of Coatings*

In order to better understand the role of microflakes in the corrosion process, both composite and blank Ni coatings were analyzed by X-ray photoelectron spectroscopy (XPS). The high-resolution XPS analyses were performed in the binding energy (BE) range of Ni2p, Cu2p, and O1s photopeaks, with the results presented in Figure 9. Supporting studies were also carried out in the C1s BE range.

**Figure 9.** High-resolution XPS spectra recorded for blank Ni coating in (**a**) Ni2, (**b**) Cu2p, and (**c**) O1s peak binding energy range, before and after the exposure to 0.05 M NaCl.

The analysis performed for the blank Ni coating prior to corrosion stability tests reveal a weak signal from nickel in the Ni2p spectra (see Figure 9a), further deconvoluted using two peak doublets Ni(B) and Ni(C). The Ni2p3/<sup>2</sup> photopeaks of the aforementioned components are located at 854.0 and 856.3 eV, being highly characteristic for nickel(II) oxide NiO and nickel(II) hydroxide Ni(OH)2, respectively [38–41]. This is further confirmed by the location of the Ni satellite peak at 861.3 eV. The small signal intensity originates from substantial surface coverage with adventitious carbon layer, including oxidized carbon species. At the same time, the ion gun etching was not performed, due to significant observable changes in surface chemistry caused by the Ar<sup>+</sup> ions used for sputtering purposes. These findings are further confirmed by the O1s spectra visible at Figure 9c, where the O(B) peak was ascribed to the nickel hydroxide species, while the dominant O(C) component represents the organic carbon-oxygen bonds and carbonates formed during air exposure. The detailed analysis is summarized in Table 3.


**Table 3.** Spectral deconvolution performed for blank Ni coating.

The one weeklong exposure of the analyzed sample in 0.05 M NaCl solution resulted in partial modification of the observed sample surface chemistry. While the chemical state of Ni and O peaks appears to be unchanged, the photoelectron intensity count is significantly improved, suggesting partial removal of the carbon-based corrosion products layer. Nickel appears primarily in the form of Ni(OH)2 (symbol Ni(C)), with the NiO to Ni(OH)2 ratio slightly decreasing from 1:2 to 1:3. Furthermore, the third component Ni(A) emerges as a result of the exposure, its location is characteristic for metallic nickel. Again, the shape of O1s spectra corroborates the results, revealing an increased intensity of hydroxide species O(B) but also nickel oxide species O(A) at approximately 529.5 eV. The amount of adventitious carbon dropped twice.

A characteristic feature of XPS measurements carried out after blank Ni coating exposure to corrosive media is the appearance of the Cu substrate beneath the layer (see Figure 9b). Two peak doublets could be deconvoluted on the basis of the spectra shape: Cu(A) with Cu2p3/<sup>2</sup> peak at 932.5 eV and Cu(B) at 933.5 eV. The position of the Cu(A) component is characteristic both for metallic copper as well as for Cu2O [42], on the other hand, the Cu(B) component was ascribed to CuO oxides according to previous literature findings [43–46]. The total copper contribution reached 3.5 at.%.

Similar to blank Ni coating, detailed high-resolution XPS analysis was also carried out for the composite Ni coating, where the analysis was expanded with the Ce3d and Mo3d spectral range. The results of the aforementioned analysis are presented in Figure 10 with the spectral deconvolution summarized in Table 4.

**Figure 10.** High-resolution XPS spectra recorded for composite Ni coating in (**a**) Ni2p, (**b**) Cu2p, (**c**) Ce3d, and (**d**) Mo3d peak binding energy range, before and after the exposure to 0.05 M NaCl. The chemistry of synthesized cerium molybdenum oxide hydrate powder used for electrodeposition is also shown in (**c**) Ce3d and (**d**) Mo3d spectra ("reference" spectra).

Unlike the blank Ni coating, the nickel chemistry on the surface of the composite coating prior to exposure in corrosive media is well developed, with the total Ni contribution ranging from 38 at.% compared to merely 5 at.% observed in the absence of microflake functionalization. Furthermore, the metallic Ni peak Ni(A) is also well developed, proving significantly higher corrosion resistance of the protective layer under atmospheric conditions. Another difference is the strong contribution of the Ni(B) component, revealing increased NiO to Ni(OH)2 ratio of 1.2:1 (compared to 1:2 for blank Ni prior the exposure).

As mentioned previously, the galvanic Ni coating was modified using a cerium molybdenum oxide hydrate compound. The chemistry resulting from Ce and Mo surface modification is shown in Figure 10c,d, respectively. Comparison with the precursor compound analysis denoted as the reference confirms high spectral similarity, proving functionalization, did not introduce significant chemical changes. The cerium Ce3d spectra reveals high complexity, the deconvolution was performed using two peak doublets, Ce(A) and Ce(B). Both of these were ascribed to the Ce3<sup>+</sup> component on the basis of the value of peak BE (881.8 and 885.7 eV) and similar intensity, nearly at a 1:1 ratio [47–49]. It is possible, however, that composite coating before exposure contains some Ce4<sup>+</sup>, which can be inferred from the lowered position of the component at the energy 886 eV.

A similar comparison was made in the Mo3d spectral range for cerium molybdenum oxide hydrate compound and for Ni composite coating functionalized with these microflakes. Both analyses reveal the dominant presence of the component ascribed in Figure 10d as Mo(B). The peak location at 232.7 eV is a very good match with the literature BE value of Mo6<sup>+</sup>, typically reported in MoO3 oxides and (MoO4)3 <sup>2</sup><sup>−</sup> molybdates [50–52]. The second component is common for both the analyzed samples, Mo(A), is negatively shifted at approximately <sup>−</sup>2.1 eV, being characteristic for the Mo5<sup>+</sup> component. The Mo6<sup>+</sup>:Mo5<sup>+</sup> ratio of 2.8:1 remains nearly intact. Finally, the surface analysis of Ni composite coating revealed the appearance of Mo(C) component at 229.5 eV bound with Mo4<sup>+</sup> oxides. Its absence in cerium molybdenum oxide hydrate microflakes powder may indicate partial reduction of molybdenum during the electrodeposition process [53–55]. According to the literature, the reduction of molybdenum could take place during electrodeposition, because the shape of the voltammetric curve and currents achieved differed from those recorded in a blank Ni electrolyte (Figure 3).

The Ni chemistry on the surface of the composite Ni coating changed as a result of the exposure to the 0.05 M NaCl solution. The second analysis carried out after a one-week period revealed a double diminished Ni contribution associated primarily with the observable removal of the Ni(B) component from the coating surface. It is possible that NiO got, in part, reduced to the metallic Ni in local cathodic areas and, in part, transformed to Ni(OH)2. At the same time, the amount of molybdenum oxides within the protective coating nearly completely disappeared as a result of the corrosion process. The removal of cerium molybdenum oxide hydrate microflakes was further confirmed by significant removal of cerium species, however, it appears that the corrosion process is less selective towards it, since the spectra was still recognizable after the exposure, with nearly intact chemistry. Finally, the secondary effect of the corrosion process is the appearance of the Cu oxides, Cu(A) and Cu(B). Not only the total contribution of Ni, but also Cu, is significantly increased in comparison to blank Ni coating, indicating smaller thickness of the protective layer and/or locally reduced coverage resulting from consumption of cerium/molybdenum species during the corrosion process.


**Table 4.** Spectral deconvolution performed for composite Ni coating.

#### *3.4. Chemical Analysis of Corrosive Solutions*

To verify the hypothesis regarding the selectivity of Ce and Mo leaching from microflakes embedded in the composite coating, an additional experiment was planned consisting of a comparative analysis of the composition of NaCl solutions both before and after the corrosion process. For this purpose, blank Ni and Ni composite coatings were immersed for 7 days in 0.05 M NaCl solutions (100 ml volume for each sample). One series of samples was just immersed and left for 7 days, a second one was immersed with an additional continuous stirring of the solution. After 7 days, all solutions were filtered and analyzed by ICP-OES and ICP-MS. The focus was on the comparative analysis of: Ni, Cu, Ce, and Mo content, with blank 0.05 M NaCl as the reference. The results are summarized in Table 5.


**Table 5.** Concentrations of Ni, Cu, Ce, and Mo in a corrosive solution.

<sup>1</sup> ICP-OES analysis. <sup>2</sup> ICP-MS analysis. \* Solution was mechanically stirred for 7 days exposure of the coatings.

The results presented in Table 5 clearly indicate that Ce did not go into the solution in any significant amounts (similar level before and after corrosion process). At the same time Mo content increased from ca. 5 to 370–390 μg/L. This analysis shows that molybdenum is not only leached from microflakes, but also significant amounts of Mo passed into the corrosive solution. This behavior is quite possible, because the SEM observation of composite coating after 168 hours exposure clearly indicates that some of the microflakes have probably been leached away and those that were visible have been partially dissolved, leaving something like a skeleton/frame (Figure 11a). Furthermore, XPS analysis seems to confirm that molybdenum, when released from micro-flakes, does not form a stable oxide layer on the surface of the coating. Such selective dissolution probably caused voids (well visible in Figure 11a) in the coating, which resulted in a lower corrosion resistance of this composite material.

**Figure 11.** Surface morphology of (**a**) composite and (**b**) blank Ni coatings (deposited on the copper substrate at a current density of 6 mA cm−<sup>2</sup> at 70 ◦C for 1 h) after one-week exposure to 0.05 M NaCl solution.

The results related to Ni concentration (Table 5) very likely to confirm that Ni almost completely forms the layer of corrosion products. Only point corrosion products were visible on the surface of blank Ni coating (Figure 11b). Its (nickel) presence in the solution after corrosion at a level similar to that determined in the blank 0.05 M NaCl solution only confirms this assumption. Different behavior was observed for copper (originating from the substrate). In the case of blank Ni coating, Cu was not observed to pass into the NaCl solution, which, in the context of the results of the XPS surface analysis, only confirmed that this element is part of the relatively stable corrosion products on the surface of this coating. Only in the case of composite coating, copper was determined at the level of 14–29 μg/L, which means that this element partially passed in the form of an ion into the solution. This is even more possible because during the dissolution of the microflakes, the copper substrate was gradually exposed to chloride ions.

#### **4. Conclusions**

In summary, it can be concluded that:


**Author Contributions:** All authors have read and agree to the published version of the manuscript. Research concept and supervision, J.W.; methodology, J.W. (corrosion, electrochemistry, SEM, powder synthesis), J.R. (XPS), A.N. (electrodeposition, laboratory work, profilometry), K.W. (XRD), S.B. (ICP-OES and ICP-MS); writing—original draft preparation, J.W., A.N., J.R., K.W.; review, K.D., B.S.

**Funding:** This research was funded by a statutory activity subsidy from the Polish Ministry of Science and Higher Education for the Department of Advanced Material Technologies at Wrocław University of Science and Technology in 2020 year—grant number 8201003902.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Microstructure Characterization and Corrosion Resistance of Zinc Coating Obtained on High-Strength Grade 10.9 Bolts Using a New Thermal Di**ff**usion Process**

#### **Henryk Kania 1,\* and Jacek Sipa 2,\***


Received: 15 March 2019; Accepted: 25 April 2019; Published: 29 April 2019

**Abstract:** The article presents the results of research on the application of innovative thermal diffusion zinc coating technology with the recirculation of the reactive atmosphere to high-strength grade 10.9 bolts. The innovation of this method consists in the introduction of reactive atmosphere recirculation and the application of coating powder mix which contains zinc powder and activator. Recirculation of the reactive atmosphere ensures its uniform composition, while the presence of an activator intensifies the process of saturating steel surface with zinc, which boosts the efficiency of active agents. Coatings were created at 440 ◦C and a heat soaking time of 30–240 min. Coating structure (SEM) was exposed, chemical composition in microsites (EDS) was defined, and coating phase structure (XRD) was identified. The kinetics of coating growth were defined. It was found that the increment of coating thickness was controlled by square root of soaking time. Coatings obtained using innovative thermal diffusion zinc coating technology had a two-layer structure. At the substrate, a compact layer of phase Γ<sup>1</sup> (Fe11Zn40) was created, which was covered with a layer of phase δ<sup>1</sup> (FeZn10). The new method of thermal diffusion zinc coating will alow for the creation of coatings of very good corrosion resistance while maintaining strength properties of bolts defined as strength class 10.9.

**Keywords:** thermal diffusion coatings; grade 10.9 bolts; corrosion resistance

#### **1. Introduction**

Due to their versatility and reliability, bolted joints are among most frequent methods of joining steel structures. They demonstrate a number of advantages, which include easy and quick assembly at minimum cost of maintenance and control but also good strength at variable loads [1]. However, due to specific geometry of such joints, they are the most critical area of construction, which in numerous instances necessitates the application of high-strength bolts made of special, additionally heat-treated steel grades.

The durability of steel structures is determined by the impact of the environment whose aggressiveness speeds up corrosive wear. Therefore, adequate anti-corrosion protection of steel structures is an important factor and is decisive for safe exploitation. Steel structures are most often protected against corrosion by hot dip metallization. Zinc coatings are also used to protect bolts designed for joining structural elements [2]. However, the requirement of a proper thread match between the bolt and nut brings about the limitation of coating thickness [3]. Despite being stronger than construction steel, bolts become the weakest element of a structure in a corrosive environment.

Currently, zinc coatings on bolts are obtained by various methods. Hot dip galvanizing (HDG) is most frequently used to protect bolt surfaces [2]. Bolts are galvanized by a batch hot dip method in baskets with centrifugation [4]. Coating consists of a transient layer of intermetallic phase Fe-Zn, which is covered with an outer layer of zinc extracted from the bath [5,6]. The depth of the external layer is difficult to control, while removing an excess of zinc is not always successful at a conventional hot dip galvanizing temperature of 450 ◦C. Coatings obtained usually have a thickness of 45 to 65 μm. This makes it difficult to obtain a proper fit between bolts and nuts, while decreased coating thickness lowers the strength of the joint compared to other elements of the structure [7]. Much better results are achieved by high-temperature galvanizing. Galvanizing at a higher temperature allows for the effective removal of excess zinc from the surface of bolts. Coating obtained at a high temperature consists exclusively of intermetallic phase Fe-Zn. The thickness of intermetallic phase layers is more effectively controlled with process parameters. The corrosion resistance of high temperature coatings is approximately two times greater than that of conventional zinc coatings [8]. Removal of excess zinc allows one to maintain the thread fit, while increased corrosion resistance yields bolt strength comparable to other structural elements. High temperature galvanizing of bolts is effected at a temperature of ca. 560 ◦C [9]. In the case of high-strength quenched and tempered bolts, galvanizing at such a temperature poses a risk of compromising the high-strength properties. Furthermore, the pickling procedure in hydrochloric acid, normally applied in the procedure of surface finishing, creates a risk of hydrogen embrittlement due to hydriding in the case of high-strength bolts.

Another method of protecting bolts against corrosion is electro-galvanizing, which ensures obtaining a uniformly thick coating on a threaded surface. Due to small coating thickness (usually up to 15 μm), the corrosion resistance of the bolts, which is significant at the beginning of exploitation, decreases fairly quickly, which considerably shortens the life of the entire structure. The application of galvanized coatings does not require heating of treated elements during the process, which minimizes the risk of high-strength bolts losing their strength after quenching and tempering. However, this method leads to extensive contamination of the natural environment as well as possible hydrogen embrittlement in high-strength steel [10].

Much better results are achieved when zinc coatings on bolts are obtained by sherardizing. The sherardizing process is run in closed rotary retorts in which bolts are placed together with zinc powder and zinc oxide [11]. The retort is soaked at a temperature of 380–450 ◦C, which allows for the creation of a layer of intermetallic phase Fe-Zn on the bolts [12–14]. The protective layer created by sherardizing, similar to the one obtained through high-temperature galvanizing, consists exclusively of intermetallic phase Fe-Zn, which permits the control of case depth and ensures high corrosion resistance. Meanwhile, due to a much lower processing temperature, the high-strength bolts retain their strength properties [13]. It is usually enough to have the surfaces cleaned mechanically before sherardizing in order to prevent hydrogen embrittlement. The major drawback of sheradizing is its relatively long processing time. In order to obtain coatings of the required depth of 15–30 μm, 6 to 12 h will be needed, to be further extended if the process temperature is lowered.

#### **2. Innovative Thermal Di**ff**usion Zinc Coating Technology with Reactive Atmosphere Recirculation**

The technology of thermal diffusion zinc coating with recirculation of reactive atmosphere is a new technology in zinc coating. The process is run in a closed retort, which is loaded with products to be coated together with powder mixture. The rotating motion of the retort ensures continuous contact of powder mix with the surfaces of processed products. The powder mix is composed of zinc powder, ZnO as a filler, and NH4Cl as an activator. In the course of heating the powder mix in the retort, gaseous reaction products are created that move towards the surfaces to be coated. On steel substrate adsorption of zinc atoms created from the gaseous mixture as a result of reduction, dissociation, or exchange occurs.

Forming of thermal diffusion coating results from the chemical reaction of vapor deposition [15–17]. Previous studies have shown that NH4Cl decomposes according to the following reaction [16]:

$$\text{NH}\_4\text{Cl} \rightarrow \text{NH}\_3 + \text{HCl}\_4 \tag{1}$$

Zn powder reacts with HCl vapors to form ZnCl2 by the reaction:

$$\text{Zn} + 2\text{HCl} \rightarrow \text{ZnCl}\_2 + \text{H}\_{2\text{-}} \tag{2}$$

ZnCl2 is reduced by hydrogen to Zn as a result of heating process [16]. During heating, active zinc atoms can also be formed according to Equation [18], as follows:

$$\text{ZnCl}\_2 \rightarrow \text{Zn} + 2\text{Cl},\tag{3}$$

The concentration of active Zn atoms in the powder mixture is higher than on the steel surface. The concentration gradient causes internal diffusion of zinc, which in a gaseous form deposits on the steel surface [18].

Active zinc atoms easily diffuse into the steel substrate along existing vacancies, leading to distortion of the crystallographic structure of the steel substrate. The iron atoms on the surface of the substrate can be activated by the following reactions [18]:

$$2HCl + Fe \to FeCl\_2 + H\_{2\_2} \tag{4}$$

$$FeCl\_2 \to Fe + 2Cl,\tag{5}$$

Crystal lattice distortion accelerates the diffusion of active Fe atoms outside [18]. During heating, there is a two-way diffusion between Zn and Fe, causing the formation of Fe-Zn intermetallic phases [17].

The process, besides rotary movement of the retort, is aided by an innovative solution that consists of forced circulation of reactive atmosphere along the retort cylinder [19]. This ensures a better and more uniform contact of product surface with powder mix, both on external and internal surfaces as well as on surfaces of intricate shape. Introduction of atmosphere recirculation along the retort cylinder permits uniformity of reactive atmosphere within the entire volume of the retort, which leads to uniform temperature in the working space, lower consumption of powder mix, and more effective use thereof in comparisation to typical sherardizing [19]. Due to recirculation and better use of active agents, it is possible to create the layers of intermetallic phase Fe-Zn [20] on low-carbon steel at a much shorter time compared to conventional sherardizing technology. At the same time, the coatings obtained through this method demonstrate higher corrosion resistance than the hot dip coatings [21]. The new technology constitutes an alternative to traditional process of sherardizing, in which an activator is not used and mixture motion in the atmosphere is determined solely by the rotating movement of the retort.

#### **3. Materials and Methods**

Coatings for research were created on a prototype plant equipped with a furnace with a rotary working chamber of capacity 100 kg and forced recirculation of reactive atmosphere. The workload consisted of M10 bolts and threaded bars of strength grade 10.9 made of quenched and tempered 1.7225 steel (42CrMo4). The powder mix was composed of zinc powder with an addition of 15% of zinc oxide ZnO as a filler and 3% of NH4Cl as an activator. All that was soaked at the temperature of 440 ◦C and times of 30, 60, 120, and 240 min. Before filling the working chamber of the furnace, the ingredients of powder mix were dried at the temperature of 120 ◦C for 12 h in order to reduce moisture content to approx. 1%. The activator was added to the mixture right before processing due to its high hygroscopicity. Before processing, bolt surfaces were cleaned mechanically using sandblasting process to prevent the risk of hydrogen embrittlement.

In order to establish the structure of the thus-produced coatings, metallographic tests were performed with a light microscope. The microstructure and chemical composition in micro-areas of coatings were analyzed with a Hitachi S-3400 N (Tokyo, Japan) scanning microscope equipped with a X-ray energy dispersion spectroscope.

Phase analysis by X-ray diffraction was performed with a JEOL JDX-7S X-ray (Tokyo, Japan) diffractometer using a copper anode lamp (λCuK<sup>α</sup> <sup>=</sup> 1.54178 ´ Å) powered with 20 mA current at 40 kV, and with a graphite monochromator. Recording was performed with a stepwise approach of 0.05◦ step and counting time of 3 s in the range of 10 to 90◦ 2θ. Phases were identified with the help of the ICDD (Newtown Square, PA, USA) PDF-4+ database. Diffractometric tests were performed on skew-ground sample surfaces so that phase structure could be displayed in the entire cross-section of the coating.

Testing of resistance to the impact of neutral salt spray was performed in accordance with the standard EN ISO 9227 in a salt chamber CORROTHERM Model 610 by Erichsen.(Hemer, Germany) Testing was conducted in a 5% spray of sodium chloride in distilled water at the temperature of 35 ◦C. In order to determine mass changes, gravimetric analyses were performed in the course of testing following 48, 96, 164, 240, 480, 720, and 1000 h of the exposure of samples in the chamber. Corrosion tests were carried out on grade 10.9 screws with thermal diffusion zinc coating obtained at 4408C and heating time of 12 min.

Tensile strength testing was performed on a tensile strength testing machine Inspekt Table 100 by Hegewald und Peschke MPT GmbH (Nossen, Germany) with maximal load 100 kN.

#### **4. Results and Discussion**

#### *4.1. Surface and Cross-Sections Appearance*

External appearance of zinc coatings obtained on a grade 10.9 bolt with the new technology of thermal diffusion zinc coating with reaction of reactive atmosphere is presented in Figure 1. The coatings obtained do not show any discontinuity, while their surface looks matt grey. No traces of powder mix were found on the thread surface after zinc coating (Figure 1a). Additionally, in a cross-section the coating shows no discontinuity and densely covers the bolt surface both on the head (Figure 1b) and on the tip of thread (Figure 1c) as well as in the groove (Figure 1d). The grey and matt appearance of the coating and its cross-sectional morphology are a proof of intermetallic Fe-Zn phases in its structure. The appearance of cracks located mainly in the upper part of the coating can be observed in the coating structure, which may indicate the brittleness of the coating. The occurrence of transverse cracks is characteristic of the intermetallic phases of the Fe-Zn phases.

(**a**) (**b**)

**Figure 1.** *Cont.*

**Figure 1.** Appearance of surface (**a**) and cross-section of coating on bolt head (**b**), thread tip (**c**), and thread groove (**d**) of grade 10.9 bolt.

#### *4.2. Growth Kinetics*

Growth kinetics of coatings obtained on high strength grade 10.9 bolts at the temperature of 440 ◦C and soaking time from 30 to 240 min is presented in Figure 2. Coating thickness increases with the increase of soaking time. However, the increment of coating thickness becomes slower and slower. After 120 min, a coating of average thickness of 50.61 μm was obtained. During further soaking, the increment was over 2 x slower with thickness of 72.62 μm obtained after the time of 240 min (Figure 2a).

**Figure 2.** Growth kinetics (**a**) and plot of coating thickness as a function of t1/<sup>2</sup> (**b**).

The impact of temperature soaking time on coating thickness is shown in more detail as the dependence of thickness in the square root function of soaking time t1/2. For the experimentally defined dependence of coating thickness from soaking time, there is a nearly linear correlation of coating thickness against square root of soaking time t1/2. The determined trend function and high value of correlation factor R2 = 0.98 confirm the linear nature of coating thickness increment. Thus, the growth kinetics for a coating may be described by the following equation:

$$\text{by (thickness)} = 5.22 \times \text{t}^{1/2} - 7.18,\tag{4}$$

The presence of a factor equaling t1/<sup>2</sup> in the equation describing growth kinetics means that the coating increase is a process controlled by diffusion. Since those coatings are created as a result of zinc diffusion into the substrate, it may be concluded that the kinetics of that phenomenon are controlled by square root of time, which appears to be in compliance with literature data [22]. The equation also shows that the growth of the coating begins after the time that is necessary to reach the temperature allowing the diffusion of ingredients.

#### *4.3. Microstructure (SEM) and Microanalysis (EDS)*

The microstructure of a coating obtained at a temperature of 440 ◦C and time of 120 min is presented in Figure 3. Percentage contents of analysed elements are shown in Table 1. The coating is composed of two layers: outer layer defined as site A and a layer adjoining the substrate marked as site B. In the outer layer (Figure 3b), the presence of 7.2 wt % Fe and 92.8 wt % Zn was found in outer zone (point 1, Table 1) and 11.4 wt % Fe and 88.6 wt % Zn in the zone bordering on the underlying layer (point 2, Table 1). Chemical composition of outer layer is comprised within intermetallic phase δ<sup>1</sup> (FeZn10) [23]. Change of concentration of components on the cross-section of that phase is typical of phase δ<sup>1</sup> and may be an evidence of its varied morphology: phases δ1p of palisade structure in the upper zone and phase δ1k of compact structure in the substrate zone [5]. Chemical composition of the underlying layer (Figure 3c) is more homogenous. In the outer layer, there is 19.7 wt % Fe and 80.3 wt % Zn (point 3, Table 1), whereas the substrate layer contains 21.2 wt % Fe and 78.8 wt % Zn (point 1, Table 1). Chemical composition in the underlying layer corresponds to stability range for phase Γ<sup>1</sup> (Fe11Zn40) [6].

**Figure 3.** Microstructure (SEM) of coating obtained on high-strength grade 10.9 bolts via thermal diffusion method with reactive atmosphere recirculation: cross-section of the coating with selected areas (**a**), microstructure of the coating in area A (**b**), microstructure of the coating in area B (**c**).


**Table 1.** Chemical composition at selected micro-areas of coating obtained on high-strength grade 10.9 bolts (measurement points acc. to Figure 3).

#### *4.4. X-ray Phase Analysis (XRD)*

XRD analysis performed on the surface of a skew-ground coating (Figure 4) demonstrates phase δ<sup>1</sup> (FeZn10) and phase Γ<sup>1</sup> (Fe11Zn40). Considering chemical composition in the coating microsites (Table 1), it may be claimed that the outer layer of the coating is phase δ1, whereas the layer adjacent to the substrate is phase Γ1. The research did not confirm any presence of other phase Fe-Zn, although these are stabile in the conditions the coating is created. According to phase equilibrium system, phases Γ (Fe3Zn10) and ζ (FeZn13) are also stabile [24]. Configuration of all stabile phases (Γ, Γ1, and δ<sup>1</sup> i ζ) occurs in the diffusion layer of coatings obtained by HDG method at the temperature of 450 ◦C [6]. However, the mechanism of producing a hot dip galvanizing coating is more complex than that, and the transition layer is created as a result of simultaneous processes of diffusive growth, solutioning in liquid zinc, and recrystallization [24].

**Figure 4.** Diffractogram of the surface of skew-ground coating obtained on high-strength grade 10.9 bolts through thermal diffusion method with reactive atmosphere recirculation.

For technologies using powder mixtures as zinc carrier, the available literature does not allow an unambiguous determination of coating phase structure. In conventional sherardizing, a coating is created that is composed of a compact layer of phase δ<sup>1</sup> (marked as FeZn7) and a non-homogenous outer layer of phase ζ (FeZn13) [14]. However, Konstantinov [25] claims that the coating is built of phases Γ and δ1. Liu [26] obtained a similar structure, making his coatings with a pack cementation method in a powder mix with an addition of activator on 1.7225 (42CrMo4) steel. On the other hand, Wortelen [27] claims that the coating obtained through sherardizing in powder mix with an activator is composed of phases Γ, Γ1, δ1, and ζ. It must be pointed out that characteristic spectra for phases Γ and ζ coincide largely with the spectra of phases Γ<sup>1</sup> and δ1; therefore, XRD examination does not

allow for unambiguous exclusion of the presence of phases Γ and ζ. However, no other structural components were determined in the microstructure of the tested coating, whose chemical composition might correspond to the range of homogeneity of phases Γ and ζ. This was also confirmed in the research conducted by Chaliampalias [15,28], who obtained a coating composed of phases Γ<sup>1</sup> and δ<sup>1</sup> in the pack cementation process.

#### *4.5. Corrosion Resistance*

The surface appearance of grade 10.9 bolts with thermal diffusion after corrosion testing in neutral salt spray is presented in Figure 5. During exposure in a salt chamber, the thermal diffusion coating becomes covered with products of white rust. After 1000 h of exposure in the salt chamber, local rusty discolouring becomes visible on the surface (Figure 5a). White rust products accumulate in thread grooves and expose the thread tips, where the intensity of corrosion is clearly higher. Upon completion of corrosion test, no distinctive permeation of coating into substrate was visible, which was confirmed by structural research on a cross-section of coating (Figure 5b). After testing, a non-uniform depletion of coating thickness was manifested on the cross-section; however, the coating did not yet lose its continuity. The presence of rusty discolourations on the surface of zinc coatings is characteristic of corrosion of intermetallic phase Fe-Zn. Such a phenomenon is observed during exposure of HDG coatings to an environment that contains chlorides [8].

**Figure 5.** The appearance (**a**) and structure (**b**) of the thermal diffusion zinc coating after corrosion tests in a salt chamber.

Dependence of mass changes in tested coatings from the time of exposure in the salt chamber is presented in Figure 6a. It may be concluded that during exposure to neutral salt spray, the tested coatings are characterized by increment of mass, which is an evidence of accumulation of corrosion products on their surfaces. Changes of bolt mass, and at the same time corrosive wear of thermal diffusion coating, follow a linear relationship. The trend function, determined in the course of experimental research, and the high correlation factor R<sup>2</sup> = 0.97 (Figure 6a) constitute proper linear match with data and confirm the linear character of the wear of thermal diffusion coating.

P The linear character of coating wear is also evidenced by coating loss in its cross-section. Average coating thickness changes after 480, 720, and 1000 h of exposure in a salt chamber are presented in Figure 6b. Assuming average initial thickness of 51.52 ± 8.48 μm, they decreased to 18.03 ± 5.53 μm after a 1000 h corrosion test. The trend function, determined through measurements of average thickness of coating and good match of correlation factor R2 = 0.96, confirm the linear character of corrosive wear of coating in its cross-section. Extrapolation of trend line allows an estimation that the total loss of thermal diffusion coating on a bolt's surface and at the same loss of protective properties will happen after approx. 1540 h of exposure in the salt chamber. In comparison to conventional hot dip galvanizing zinc coatings, the durability of the thermal diffusion coating tested in a corrosion test in a salt chamber [22] results in an approx. 2–3-fold increase in the durability time to penetration

of the substrate surface (steel). While carrying out the corrosion test in a salt-spray chamber (acc. EN ISO 9227) traditional hot dip galvanizing coating of comparable thickness (average thickness 50.8 ± 6.42 μm), distinctive permeation of coating into substrate was visible after 480 h of exposure in neutral salt-spray environment [21]. For conventional hot dip zinc coating, much more severe corrosion losses can be observed in the initial stage of corrosion process when the outer layer of zinc is subject to corrosion. When this layer is worn out, the corrosion process proceeds more slowly in the Fe-Zn intermetallic layers [8]. Thermal diffusion zinc coating does not show increased corrosion intensity in the initial stage of corrosion process due to the lack of an outer zinc layer. The corrosion process proceeds only in the Fe-Zn intermetallic layers, whose thickness is higher in comparison with conventional hot dip coatings. The average thickness loss of the coating was used to estimate coating durability. Locally, in areas of lower thickness, the coating may be broken through to the base material. Due to the very good protection properties of the coating and the significant share of areas with a higher coating thickness, the protection of the steel in the initial stages of coating penetration will still be preserved.

**Figure 6.** Average change of mass in tested bolts grade 10.9 (**a**) and average change of thermal diffusion coating thickness (**b**) during exposure in salt spray chamber.

#### *4.6. Tensile Test*

A diagram of a static tension test on threaded bar grade 10.9 without coating and with thermal diffusion coating is presented in Figure 7. The values of tensile strength limit Rm and conventional yield limit Rp0.2 for three tests on each bar are presented in Table 2. The tensile strength of grade 10.9 threaded bar decreased slightly after creation of thermal diffusion coating. The average tensile strength limit was Rm = 1024 MPa for the uncoated bar and Rm = 1007 MPa for the thermal diffusion coated one, respectively. The process of creating a coating by the new method of thermal diffusion zinc coating did not compromise the tensile strength below the required limit. However, the ratio of conventional yield limit to tensile strength had a similar value of 0.94–0.95 both for the uncoated bar and the thermal diffusion coated one. Tensile strength of the threaded bar coated with the new method of thermal diffusion zinc coating with recirculation of reactive atmosphere fulfilled the tensile strength requirement for grade 10.9.

**Figure 7.** Tensile stress-displacement relationship curve of coated and uncoated 10.9 grade threaded bar.


**Table 2.** Results of tensile strength test of coated and uncoated 10.9 grade threaded bar.

In these tests, threaded bars M10 in strength grade 10.9 made of 42CrMo4 steel (1.7225) after quenching and tempering process were used. The total length of the sample was 220 mm, gauge length 160 mm. Special high nuts were screwed on the ends of the bar, and then the sample was placed in the holders of the testing machine and stretched until it was broken.

#### **5. Conclusions**

The innovative technology of thermal diffusion zinc coating with recirculation of reactive atmosphere provides effective anti-corrosion protection for high-strength grade 10.9 bolts while retaining their strength properties. The application of reactive atmosphere recirculation ensured homogenization of its composition and a better contact of product surface with powder mix, which permitted one to obtain continuous coatings of a compact structure in a shorter time compared to the conventional sherardizing method.

The research conducted allows for the following conclusions to be made:


**Author Contributions:** Conceptualization, H.K., J.S., Experimental research, J.S., Methodology, H.K., Research of structure and corrosion resistance, H.K., Tensile test, J.S., Data curation, H.K., Visualization, H.K. Analysis of results H.K., Project administration, J.S.; Funding acquisition, J.S.

**Funding:** The research on thermal diffusion zinc coating was conducted within the R&D project "Innovative Technology of Thermal Diffusion Zinc Coating of Key Structural Elements with Application of Reactive Atmosphere Recirculation" (project no. POIG.01.04.00-08-383/13) co-financed by European Union within Operating Programme Innovative Economy from the means of European Regional Development Fund.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **High-Temperature Oxidation of Heavy Boron-Doped Diamond Electrodes: Microstructural and Electrochemical Performance Modification**

#### **Jacek Ryl 1,\*, Mateusz Cieslik 1, Artur Zielinski 1, Mateusz Ficek 2, Bartlomiej Dec 2, Kazimierz Darowicki <sup>1</sup> and Robert Bogdanowicz <sup>2</sup>**


Received: 28 January 2020; Accepted: 19 February 2020; Published: 21 February 2020

**Abstract:** In this work, we reveal in detail the effects of high-temperature treatment in air at 600 ◦C on the microstructure as well as the physico-chemical and electrochemical properties of boron-doped diamond (BDD) electrodes. The thermal treatment of freshly grown BDD electrodes was applied, resulting in permanent structural modifications of surface depending on the exposure time. High temperature affects material corrosion, inducing crystal defects. The oxidized BDD surfaces were studied by means of cyclic voltammetry (CV) and scanning electrochemical microscopy (SECM), revealing a significant decrease in the electrode activity and local heterogeneity of areas owing to various standard rate constants. This effect was correlated with a resultant increase of surface resistance heterogeneity by scanning spreading resistance microscopy (SSRM). The X-ray photoelectron spectroscopy (XPS) confirmed the rate and heterogeneity of the oxidation process, revealing hydroxyl species to be dominant on the electrode surface. Morphological tests using scanning electron microscopy (SEM) and atomic force microscopy (AFM) revealed that prolonged durations of high-temperature treatment lead not only to surface oxidation but also to irreversible structural defects in the form of etch pits. Our results show that even subsequent electrode rehydrogenation in plasma is not sufficient to reverse this surface oxidation in terms of electrochemical and physico-chemical properties, and the nature of high-temperature corrosion of BDD electrodes should be considered irreversible.

**Keywords:** boron-doped diamond; high-temperature treatment; surface oxidation; microstructure defects; electrochemical activity

#### **1. Introduction**

Boron-doped diamond (BDD) surfaces are widely studied due to their unique electrochemical and physico-chemical properties [1]. While the most often reported applications are for sensing [2], energy storage [3], or water treatment [4], BDD does play an important role in high-temperature environments [5]. The electrically conductive nanocrystalline boron-doped diamond (BDD) layers can be applied as a protective coating of Si photoelectrodes in sun-driven photoelectrochemical cells in aqueous electrolyte solutions [6]. Furthermore, wastewater treatment with advanced oxidation processes at BDD usually strongly depends on the potentials and current densities. It has been shown that a larger current promotes stability for urea removal at high potentials [7], which could also induce

the thermal issues and electrode surface defects [8]. Moreover, the relatively high thermal conductivity coefficient (~700 W/mK) of BDD allows for its use as a heat spreader, replacing the commonly used metal spreaders such as copper, copper/refractory, or copper laminate in high power RF/microwave devices resulting in higher isolation of the ground plane at below 1.5 GHz [9,10].

BDD thin films are found to be particularly attractive as electrodes for electrolysis and electroanalytical applications due to their outstanding properties, which are significantly different from those of other conventional electrodes (e.g., glassy carbon or platinum electrodes [11]). In addition to the innate properties of diamond, such as high thermal conductivity, high hardness, and chemical inertness, the attractive features of conductive BDD include a wide electrochemical potential window in aqueous and non-aqueous media, very low capacitance, and extreme electrochemical stability [12]. BDDs are said to possess higher oxidation resistance at elevated temperatures in comparison to pure diamond [13], which is due to the formation of a B2O3 surface layer. Wang and Swain [14] reported a Pt/diamond composite electrode that exhibited superb morphological and microstructural stability during vigorous electrolysis in acidic media at a temperature of around 170 ◦C and current density around 0.1 A/cm2. In addition, BDD layers can operate at high temperatures in electronics, an example being diamond-based Schottky barrier diodes for high-power devices [15,16].

High-temperature treatment is also one of the reported routes to modify and oxidize the surface termination of BDD electrodes [17–19]. Other approaches include oxidation under electrochemical polarization [20,21], chemical agents [22,23], ozone [24], plasma and/or UV treatment [25,26], but also natural aging in air [27–29]. The surface termination type significantly differentiates the electric and physico-chemical properties of boron-doped diamond electrodes. Hydrogen termination (HT-BDD) leads to hydrophobic and non-polar behavior, as well as high surface electrical conductivity and low electric transfer resistance [22,25]. The presence of hydroxyl or carbonyl surface bonds and oxidized BDD termination (OT-BDD) results in hydrophilicity and much lower surface electrical conductivity [30,31]. On the other hand, the electrochemical potential window for OT-BDD electrodes is reported to be wider than in the case of hydrogen-terminated ones [31,32]. Importantly, Zielinski et al. [29] reported that oxidation homogeneity of polycrystalline electrodes as well as oxidation rate in general highly depend on the above-mentioned modification route and confirmed that various treatment procedures produce different surface terminating species (C–OH, C–O-C, C=O, C–OOH, etc.), which may have a significant influence on tailoring desired functionalization procedures and BDD electrode characteristics.

Among the multitude of reported oxidation routes, high-temperature oxidation is characterized by a few unique features [29]. First of all, it appears to be the most heterogeneous at low durations, which was claimed to be related to the diverged propensity of differently oriented BDD crystal planes towards surface oxidation. Furthermore, according to studies where X-ray photoelectron spectroscopy (XPS) was performed, high-temperature oxidation tends to produce a significantly larger amount of surface hydroxyl species than any other reported approach. Their high polarity in comparison to other surface species results in very small reported contact angles during drop shape analysis tests. Next to electrochemical anodization, high-temperature treatment is one of the most often used pretreatment methods to clean the impurities on as-prepared BDD electrodes after the chemical vapor deposition (CVD) process [19,33].

High-temperature oxidation may also lead to changes in surface morphology as well as the altered sp2/sp3-carbon ratio [34,35]. The surface structure of BDD electrodes exposed to elevated temperatures in an oxygen-containing atmosphere was presented by Jiang et al. [17], who noticed that during such a treatment the amorphous carbon phase converts to the diamond phase and that the diamond grain increases with the decrease of grain boundaries. In addition, the short high-temperature treatments (30 min) led to the arrangement of grain boundaries on the electrode surface, which had a positive effect on their conductivity. An interesting consequence of the high-temperature oxidation is the enhancement of the peak ratio between the diamond peak and the graphitic peak, showing a decreasing amount of sp2-carbon after the process, resulting in an increased carrier concentration [36,37]. However, prolonged high-temperature oxidation led to disordered grain boundaries and significantly

deteriorated conductivity of the BDD electrode. High-temperature treatment may also lead to a diffusion of silicon substrate through the columnar structure of the diamond film and corrosion [33].

In light of the above presented discussion, the aim of this work was to present the effect of high-temperature treatment in air at 600 ◦C on the utility properties of boron-doped diamond electrodes, in particular for electrochemical applications. To the best of our knowledge, there are no dedicated studies on BDD electroactivity and stability of the properties when subjected to prolonged high-temperature exposure, the homogeneity and the rate of high-temperature corrosion of BDD, and its reversibility.

#### **2. Materials and Methods**

BDD films were synthesized in a microwave plasma-assisted chemical vapor deposition system (SEKI Technotron AX5400S, Tokyo, Japan). The substrates were seeded by sonication in nanodiamond suspension for 30 min following the standard procedure [38]. The chamber stage was maintained at 700 ◦C during the deposition process and the growth time was 6 h. The boron level expressed as [B]/[C] ratio in the gas phase was 10,000 ppm (boron dopant concentrations 2 <sup>×</sup> 1021 atoms cm−3) [39]. A more detailed description of the thin film synthesis can be found elsewhere [40,41]. Then, the electrode surface was cleaned and hydrogenated. First, metallic impurities were dissolved in hot aqua regia (HNO3:HCl/1:3), followed by the removal of organic impurities by hot "piranha" solution (H2O2:H2SO4/1:3) at 90 ◦C. Microwave hydrogen plasma treatment was performed using 1000 W microwave power and 300 sccm of hydrogen gas flow for 10 min.

The high-temperature treatment was carried out in an MRT-20 furnace (Czylok, Jastrzebie-Zdroj, Poland). The BDD electrodes were treated at 600 ◦C for 3, 10, 30, or 90 min in air. Afterward, the samples were removed from the furnace and cooled in air. A similar procedure was carried out in other studies [29,30,33]. After the electrochemical and physico-chemical examination, the hydrogenation procedure described earlier was carried out for high-temperature-treated BDD samples in order to determine surface oxidation reversibility.

Electrochemical measurements were carried out in a three-electrode cell. The working electrode was Si/BDD, with Ag/AgCl used as a reference electrode and platinum mesh as a counter electrode. All reagents were analytical purity (Sigma-Aldrich, Saint Louis, MO, USA). The exposed BDD sample area was 0.25 cm2. Cyclic voltammetry (CV) studies were carried out in the polarization range between −0.9 V and 1.1 V vs. Ag/AgCl, at different scan rates between 5 and 800 mV/s. The electrolyte used was 0.5M Na2SO4 with 2.5 mM K3[Fe(CN)6] and 2.5 mM K4[Fe(CN)6]. Scanning electrochemical microscopy (SECM) studies were performed using a three-step motor system (Sensolytics, Bochum, Germany) with 1 μm resolution in each direction, coupled to an Autolab 302 N potentiostat equipped with bipotentiostat module (Metrohm, Herisau, Switzerland). The commercially available ultramicroelectrodes (UMEs) made of platinum wire sealed in glass were used, with a platinum disc diameter of 5 μm. Each probe was polished and rinsed with acetone prior to the study. The electrolytic solution was composed of 2.5 mM K4[Fe(CN)6] and 0.5M Na2SO4, purged with argon prior to BDD sample examination. The potential applied to the UMEs and the BDD sample was +0.4 and 0.0 vs. Ag/AgCl, respectively. In such a setup, the oxidation of redox species occurs at the UME tip. The SECM maps were recorded with a scanning speed of 5 μm/s and 1 μm steps in the x-y plane. A similar experimental approach was previously applied by authors [40].

Topographic and electrical microscopic measurements were made using an NTegra Prima device by NT-MDT (Moscow, Russia) in contact mode. CTD-NCHR-10 probes from Nanosensors (Neuchatel, Switzerland) were used with the following catalog parameters: L × W × T lever dimensions: 125 × 29 × 4 μm. Lever spring constant was equal to 71 N/m. The contact force determined from the approach curve was 8 μN. The tip radius of the curvature was in the 100–200 nm range according to manufacturer data. Conductivity measurements were made in the scanning spreading resistance mode using a constant voltage of 20 mV. Scanning electron microscopy (SEM) micrographs were taken with an S-3400

N microscope (Hitachi, Tokyo, Japan), with 20 kV accelerating voltage and operating in secondary electron mode.

High-resolution X-ray photoelectron spectroscopy (XPS) analyses were carried out in C1s binding energy range using an Escalab 250Xi multispectroscope (ThermoFisher Scientific, Waltham, MA, USA). The spectroscope was equipped with a monochromatic Al Kα energy source. The applied X-ray spot diameter was 650 μm and the pass energy through the hemisphere analyzer was 10 eV. Prior to operation, the spectroscope was calibrated on Cu and Au single crystals. Charge compensation was controlled through low-energy electron and Ar+ ion flow by means of a flood gun. Spectral deconvolution was performed with Avantage software (v5.973, ThermoFisher Scientific, Waltham, MA, USA) provided by the spectroscope manufacturer.

#### **3. Results and Discussion**

The heavy boron-doped diamond electrodes were subjected to electrochemical and physico-chemical examination in order to evaluate the effect of oxidation under high temperature on the charge transfer kinetics. Afterward, the oxidation and corrosion reversibility was evaluated and determined.

#### *3.1. Electron Transfer through High-Temperature-Treated BDD Interface*

Figure 1 shows the results of the cyclic voltammetry studies, which were carried out on BDD electrodes before and after high-temperature oxidation in air at 600 ◦C. The redox couple used within this study, [Fe(CN6)]3−/4−, is characterized by an inner-sphere electron transfer (ISET) mechanism, which is said to be more dependent on the electrode homogeneity and applied modification procedures, thus offering a more sensitive approach to track subtle changes of charge transfer kinetics [42]. Figure 1a reveals [Fe(CN6)]3−/4<sup>−</sup> oxidation/reduction kinetics for each sample, observed at a 50 mV/s scan rate. High-temperature oxidation leads to hindered charge transfer kinetics, demonstrated by the decrease in peak current iA and iC when comparing to the as-prepared BDDs, a feature observed even at the shortest treatment duration. The peak current decreased nearly by a factor of 2 after merely 3 min of sample exposure to high temperature, and by a factor of 4 after 10 min. Detailed analysis is presented in Table 1.

**Figure 1.** Cyclic voltammetry (CV) studies for boron-doped diamond (BDD) electrodes after high-temperature oxidation at 600 ◦C. (**a**) CV scans recorded at 50 mV/s for different oxidation durations; (**b**) CV anodic peak current vs. scan rate square root function for each studied electrode. Electrolytic solution: 0.5 M Na2SO4 with 2.5 mM K3[Fe(CN)6] and 2.5 mM K4[Fe(CN)6].


**Table 1.** Electrochemical properties (anodic peak current iA, anodic-to-cathodic peak current ratio iA/iC, peak separation ΔE) based on CV studies (at 50 mV/s scan rate) of BDD samples after various high-temperature treatment durations.

It should also be noted that the modification of BDD surface termination type translates into a significant peak separation ΔE increase, the parameter which is inseparably connected with reversibility of the corrosion process. In the case of a single electron transfer reaction, the fully reversible processes should have ΔE equal to 59 mV [43]. The as-prepared BDD sample is characterized by ΔE value of approximately 350 mV. The value is quite high taking into consideration other reported heavy boron-doped electrodes, which is due to the lack of any other electrode pretreatment procedures [41,44]. Nevertheless, the ΔE increase resulting from the applied high-temperature treatment should be explained by the move further away from the diffusion-controlled mechanism, due to the slowing down of the charge transfer kinetics. Prolonged high-temperature treatment is also characterized by the increased anodic and cathodic peak asymmetry, indicating a change in process kinetics, a feature typical for irreversible processes.

The discussed irreversibility of the studied redox process, as well as the decrease in electrode kinetics, are a testimony for surface modification from HT- to OT-BDD as a result of high-temperature treatment. The above-mentioned behavior results in BDD electrode corrosion, causing variability in electrode behavior and worsening its efficiency in electroanalytical studies.

The CV anodic peak vs. the scan rate square root function in the wide scan rate range is illustrated in Figure 1b. These plots show a strong linear trend, with local deviations explained by the heterogeneous nature of the polycrystalline electrodes [45]. To determine the standard reaction rate constant k0, a numerically determined current function is used, depending on peak separation. The standard rate constant was calculated with the approach proposed by Velasco [46] to estimate completely irreversible processes. The following equation was used:

$$k^{0} = 2.415 \exp\left(-0.02 \frac{F}{RT}\right) D^{\frac{1}{2}} \left(E\_p - E\_{\frac{F}{2}}\right)^{-\frac{1}{2}} v^{\frac{1}{2}} \tag{1}$$

where *Ep* and *Ep*/<sup>2</sup> are the potentials of the CV peak and half-peak, respectively, and v refers to the rate of change of potential; the diffusion coefficient was assumed as D <sup>=</sup> 6.67 <sup>×</sup> <sup>10</sup>−<sup>6</sup> cm2/s [47].

Figure 2 reveals the effect of high-temperature treatment on the local distribution of charge transfer kinetics during [Fe(CN)6] <sup>4</sup><sup>−</sup> oxidation, assessed with scanning electrochemical microscopy (SECM). The relative current values strongly depend on the distance between the SECM tip and the electrode and vary from sample to sample. The current value is less important than its local changes due to electrode charge transfer heterogeneity. A normalization procedure was applied for the Z-axis of these graphs in order to show local discrepancies in the oxidation currents. The procedure based on a negative shift of the tip current displayed on each map to the position where the lowest recorded value equaled zero. A similar procedure was successfully applied in previous studies on heterogeneous charge transfer through the BDD electrode [40].

The obtained SECM micrographs clearly illustrate that the surface distribution of the oxidation currents is significantly increasing already after a short 10 min high-temperature treatment (Figure 2b) as compared to the as-prepared BDD electrode (Figure 2a). The heterogeneous electron transfer results from an altered propensity towards surface oxidation by BDD grain of various crystallographic

orientations. A similar observation was previously reported for various methods of BDD surface treatment, in particular through electrochemical anodic polarization [40]. Importantly, increasing the high-temperature treatment length does not significantly increase surface homogeneity. Large tip current discrepancies are still observed after 90 min of oxidation (Figure 2c). At the same time, the as-prepared BDD electrode reveals positive feedback on the approach curve, suggesting low charge transfer resistance at the electrode interface [48]. The longer the high-temperature treatment, the higher was the negative feedback observed, whereas the sample oxidized for 10 min was characterized with large discrepancies of the approach curve shapes, depending on tip landing location. These results corroborate the decrease of the electron transfer kinetics with high-temperature treatment duration and the heterogeneous nature of the oxidation process.

**Figure 2.** Typical scanning electrochemical microscopy (SECM) maps revealing charge transfer heterogeneity at heavy-doped BDD electrode interface after (**a**) 10 min, (**b**) 30 min, and (**c**) 90 min of high-temperature oxidation at 600 ◦C in air. Electrolyte: 0.5 M Na2SO4 + 2.5 mM K4[Fe(CN)6].

#### *3.2. High-Temperature Oxidation Influence on BDD Physico-Chemical Properties*

The effect of high-temperature oxidation and corrosion on the microstructure of boron-doped diamond electrodes is shown on the SEM micrographs in Figure 3. The first 10 min of exposure to 600 ◦C do not lead to significant changes in grain structure; however, after this period the material undergoes gradual degradation.

**Figure 3.** Typical scanning electron microscopy (SEM) micrographs of heavy-doped BDD surface after (**a**) 10 min, (**b**) 30 min, and (**c**) 90 min of high-temperature oxidation at 600 ◦C in air.

After 30 min of high-temperature oxidation, small and shallow etch pits start to appear throughout the diamond surface (Figure 3b), yet the grain structure is still recognizable. It is said that diamond can react with O2 and water vapor contained within the atmosphere, to create surface etch pits [37]. Treatment in the oxygen-containing environment leads to high BDD surface etching and corrosion in the relatively short periods and a large surface area with respect to the geometric electrode area [49].

Furthermore, it is evident that certain crystallographic planes on the diamond surface are more prone to surface etching. Ohashi et al. [50] reported steam-activated high-temperature treatment in 600–900 ◦C to be an efficient nanotexturing tool for diamond surfaces. Notably, it was observed that, in the case of heavy-doped BDD films, the (111) facets are more prone to etching [19]. This observation lies in compliance with the fact that (111) facets have the dominant presence in the texture of studied BDD electrodes [51,52]. Other studies revealed that the propensity towards the modification of BDD surface termination is also dependent on crystallographic orientation [40,51].

Extended exposure at elevated temperature leads to advanced electrode decay, where the grain structure of the polycrystalline electrode almost completely fades away after 90 min of high-temperature oxidation (Figure 3c). Overall, the combined effect of surface area enhancement as well as previously reported [36,37] increased charge carrier concentration within the diamond structure may possibly result in an improved electrochemical response by BDD electrodes.

The detailed topography studies were carried out using atomic force microscopy (AFM) in contact mode. Next, scanning spreading resistance microscopy (SSRM), an AFM technique derivative, allows obtaining maps of local distribution of surface resistance, originating from changes in BDD termination type due to high-temperature treatment. The topographic images (Figure 4) were acquired simultaneously with surface conductivity maps (Figure 5).

The sequence of three-dimensional images of the surface shown in Figure 4 allows observing the evolution of surface morphology consistent with that obtained by scanning electron microscopy (refer to Figure 3). The first clearly represented pyramidal crystallites change into fractured, irregularly shaped structures with high-temperature treatment duration, clearly visible in the red-marked areas. It should be noted that this form of surface development affects the modification of the conductivity distribution through the geometric factor of the contact surface between the surface and various parts of the surface of the AFM pyramid tip [29]. Furthermore, possible surface development increases might contribute to the increase of the electrochemically active surface area.

**Figure 4.** The atomic force microscopy (AFM) topography maps for exemplary BDD samples subjected to high-temperature treatment for different durations: (**a**) reference, untreated sample; (**b**) 10 min; (**c**) 30 min; (**d**) 60 min.

**Figure 5.** Typical spreading resistance (SSRM) maps for BDD samples subjected to different durations of high-temperature treatment in air at 600 ◦C: (**a**) reference untreated sample; (**b**) 10 min; (**c**) 30 min; (**d**) 90 min.

For the statistical description of the sample topography, the average roughness parameter was used, defined as:

$$S\_a = \frac{1}{MN} \sum\_{k=0}^{M-1} \sum\_{l=0}^{N-1} \left| z(\mathbf{x}\_{k'}, y\_l) - \mu \right| \tag{2}$$

where *M* = *N* = 256 are the length and width of the analyzed image in pixels and μ is the average height, defined as:

$$\mu = \frac{1}{\text{MN}} \sum\_{k=0}^{M-1} \sum\_{l=0}^{N-1} z(\mathbf{x}\_{k\prime} y\_l) \tag{3}$$

The above statistical parameters were determined for areas of 3 × 3 μm, the scan was repeated 4 times for each sample in different places, and the average roughness contained in Table 2 is the average value from each set of topographic scans.

**Table 2.** Statistical parameters: average roughness Sa and mean surface resistance R of studied BDD samples, based on AFM and scanning spreading resistance microscopy (SSRM) analyses.


The value defined in Equation (2) as *Sa* determines the local height deviations from the average value for a given area. Maxima can be associated with the presence of high crystallites; however, their degradation, which progresses over the course of the heat treatment process, causes their degradation and additionally the formation of fragments of relatively small sizes, contributing to the reduction of maxima.

Figure 5 reveals the results of surface conductance analysis in accordance with the previously presented assumptions. The SSRM technique makes it possible to create a map of the quantity defined as the spreading resistance [53], characterizing the local changes in nanocontact conductivity between the probe tip and the sample surface. A simplified, commonly used formula [54]:

$$R\_s = \frac{1}{4\sigma a} \tag{4}$$

where *a* is the radius of curvature of the probe tip and σ is the specific conductivity of the sample material, indicates the need for a trade-off between spatial resolution and probe durability, related to the thickness of the conductive layer and the radius of curvature a. The probes used in this report have a significantly increased radius of curvature for typical topographic ones (10 nm); however, they provide sufficient resolution to visualize changes in conductivity structure.

There is a visible decrease in the surface resistance average value, as well as a significant change in the mutual relationship between areas with relatively high and low electrical conductivity. For samples subjected to longer exposure, a bimodal character of conductivity can be seen with a gradual increase in the share of low conductivity areas. The observable changes present a good representation of the transition from HT- to OT-BDD under the oxidation agent [29]. Furthermore, it may be observed that high-temperature treatment leads to significant heterogeneity in surface oxidation, where the areas of altered spreading resistance are corresponding to different grain areas of the polycrystalline electrode. This observation may suggest that a varied crystallographic orientation is affecting the oxidation propensity, thus translating to locally variable electrochemical activity, as demonstrated by SECM studies.

It should be noted that the recorded heterogeneity in conductive regions can only be treated as a rough estimate of the variable electrochemical activity due to the difference in the conditions on the nanocontact in the atmosphere of air and in an electrolytic environment. Nevertheless, changes in the spreading resistance should to some extent correlate with changes in the charge transfer resistance on the surface of BDD and additionally reflect the trend of its changes together with the variability of heat treatment conditions. In this sense, the above discussed heterogeneity may correspond to the Compton model of partially blocked electrodes [55,56].

The high-resolution XPS spectra of BDD electrodes after different durations of high-temperature treatment are presented in Figure 6. These data were recorded in C1s peak binding energy range. The spectra were then deconvoluted according to the fitting model presented below and the results of the analysis summarized in Table 3.

**Figure 6.** High-resolution X-ray photoelectron spectroscopy (XPS) spectra recorded in C1s binding energy range with applied spectral deconvolution. Spectra recorded for BDD samples: (**a**) as-prepared and after high-temperature treatment in air at 600 ◦C: (**b**) 3 min; (**c**) 10 min; (**d**) 30 min.


**Table 3.** Chemical composition (in %) of various carbon chemical states on the surface of untreated BDD electrodes and after high-temperature treatment in air at 600 ◦C, based on high-resolution XPS analysis.

The C1s spectra recorded for the as-prepared BDD electrode, not subjected to high-temperature oxidation, reveals relatively simple surface chemistry. The registered spectra may be deconvoluted using three separate components. The primary peak, denoted as C-C(1) and located at approximately 284.2 eV, lies in the energy range characteristic for sp3-carbon CH species on the hydrogen-terminated diamond surface and sp3-carbon within the BDD bulk. The exact location of this peak depends on crystallographic texture rather than boron dopant concentration [57,58]. Thus, the peak position is in good agreement with previous studies on heavy boron-doped diamond substrates [20,29,59]. The second notable component, C-C(2), is usually attributed to non-hydrogenated carbon atoms on the BDD surface but also adsorbed polyhydride carbon (CHx) species [20,60]. The position of this component is typically shifted by +0.7 eV vs. C-C(1), which is also in this case.

The oxidation of BDD surface termination, occurring as a result of high-temperature treatment in air, results in the substitution of hydrogenated terminal bonds with oxygen-containing species: hydroxyl C–OH (peak at 285.6 eV), but also carbonyl >C=O (at 287.0 eV) and carboxyl COOH (at 288.7 eV) groups [41,57,61]. The XPS analyses confirmed that hydroxyl species are the dominant ones on the surface of high-temperature-treated BDD electrode, unlike other types of surface oxidation treatments (electrochemical, oxygen plasma, ozone treatment, etc.) [29]. The HT- to OT-BDD transition is bound to the increase in surface hydrophobicity, as illustrated by the drop shape analyses in the inset of Figure 6a,d.

The total share of oxidized BDD surface area, OT-BDD, counted as a sum of the above-mentioned components is naturally increasing with the duration of sample exposure to the high temperature. The share of OT-BDD surface for the untreated samples was 5.7% and gradually increased up to 43.1% after 10 min of exposure. Prolonged treatment leads to further surface oxidation, but the total OT-BDD share tends to the plateau with only 52.4% oxidized surface after 90 min of treatment. This effect is partially bound to the fact that the XPS analysis is partially obtained from ~5 nm volume beneath the electrode surface. However, similar studies on BDD electrodes, but oxidized by electrochemical or oxygen plasma treatment, have led to significant diminishing of C-C(1) peak, down to 12%–14% [29]. Retaining around 25% of HT-BDD surface even after 90 min of high temperature is well represented and imaged on the SSRM micrographs (see Figure 5), which reveals a significant heterogeneity of oxidized areas on the BDD electrode surface. Following the electric BDD spatial heterogeneity is the electrochemical behavior, determining the conditions of the charge transfer mechanism by a partially blocked electrode.

It was also observed that prolonged high-temperature treatment durations have led to the disappearance of the spectral component, located at approximately 283.2 eV, and originating from surface sp2-carbon, which is well explained by a previously mentioned theory that such an exposure results in disordered grain boundaries and significantly deteriorated BDD conductivity.

The above discussed effect of high-temperature treatment in air at 600 ◦C on BDD surface chemistry is schematically presented in Figure 7. The initially hydrogen-terminated surface of the polycrystalline BDD electrode undergoes surface oxidation over very short durations, even after a few minutes of exposure to the oxidation agent. This effect is heterogeneous in nature and hypothetically limited to

specific crystallographic planes, based on previous studies [29,51,62]. The authors did not investigate the exact planes of high-temperature interaction within this study. The prolonged treatment leads to the appearance of etch pits on the diamond surface. This effect is said to increase the electroactive surface area; however, it was not confirmed since the OT-BDD surface is characterized by lower values of the standard reaction rate constant. Most importantly, the appearance of the etch pits is heterogeneous and depends on the crystallographic orientation, which is similar to surface oxygen termination. According to the literature survey, the etch pits are primarily initiated at (111) facets [19], which, on the other hand, are the least prone to surface oxidation [63].

**Figure 7.** Schematic visualization of the oxidation process under high-temperature treatment in air at 600 ◦C.

#### *3.3. Reversibility of the High-Temperature BDD Surface Oxidation*

A few available routes for surface rehydrogenation have been reported in the literature, where hydrogen plasma treatment is claimed to be the most efficient [64]. Despite growing awareness and interest in this topic, some studies reveal issues with the reversibility of oxidized BDD electrodes [31]. Various surface physico-chemical properties (such as surface chemistry or contact angle) are often reported to be on par with those observed for BDD electrodes prior to their oxidation; however, electrochemical and electric parameters seem to be a subject of irreversible change. The above-mentioned characterization is most often discussed in the case of electrochemically oxidized BDDs. We decided to evaluate the reversibility of the oxidation process under study since both the mechanism of oxidation and the resultant surface chemistry are claimed to be different in the case of high-temperature treatment than any other oxidation route.

Figure 8a reveals the CV scans at 50 mV/s for high-temperature oxidized (at various durations) and then plasma rehydrogenated BDDs, compared to the as-prepared BDD electrode. Our studies show that neither of the investigated electrodes is characterized by improved peak current values. On the other hand, a significant improvement is visible when compared to oxidized BDD electrode kinetics. It appears that very short high-temperature treatments might lead to a certain improvement in electrochemical characteristics, that is, the peak separation ΔE is significantly improved and closer to the theoretical value of diffusion-controlled electrode processes, with high peak symmetry and peak currents equal to nearly 90% of its original value. This feature might be explained by the removal of sp2-carbon species adsorbed on the BDD electrode surface after the CVD process, demonstrating a successful electrode cleaning. The longer the high-temperature treatment duration, the lower the value of both anodic and cathodic peak currents and the more irreversible the character of the oxidation process. This effect is naturally connected with the degradation of the BDD grain structure. While the deterioration of the electrode kinetics is progressive, an interesting feature may be observed for the sample after 90 min oxidation and rehydrogenation, which is characterized by the most narrowly observed ΔE value. Following the previously defined partially blocked electrode mechanism, it may be concluded that deep etching of BDD grains (refer to Figure 3c) leads to homogenization of the electrode surface and unification of the diffusion fields at the electrode interface, a conclusion supported by SSRM studies. The detailed analysis is summarized in Table 4.

**Figure 8.** The effect of BDD surface rehydrogenation in plasma: (**a**) CV studies after various durations of oxidation. A scan rate of 50 mV/s. Electrolytic solution: 0.5 M Na2SO4 with 2.5 mM K3[Fe(CN)6] and 2.5 mM K4[Fe(CN)6]; (**b**–**d**) exemplary results for electrodes after 30 min oxidation followed by rehydrogenation: (**b**) SSRM map, (**c**) contact angle analysis, and (**d**) high-resolution XPS C1s spectrum.

**Table 4.** Electrochemical properties (anodic peak current iA, anodic-to-cathodic peak current ratio iA/iC, peak separation ΔE) based on CV studies (at 50 mV/s scan rate) of BDD samples after various high-temperature treatment durations, followed by sample rehydrogenation in plasma.


A similar observation is revealed by SSRM maps, shown in Figure 8b for BDD electrodes after 30 min of high-temperature treatment and rehydrogenation. While the hydrogen plasma is increasing the surface conductance, the Rs is far off from its original value (10.416 MΩ) and the local distribution of electric properties on the BDD surface remains heterogeneous. This is an important observation, confirming that not only the electrochemical but also the electric parameter characteristics have been modified. However, the rehydrogenation process undoubtedly does have a positive effect on BDD surface chemistry, demonstrated by the XPS analysis (Figure 8d). The hydrogenation process led to a significant reduction in the OT-BDD share (from 49.9% to 20.6%) and, in particular, in the number of surface hydroxyl groups (from 35.6% to 15.7%), which decreased more than twice for prolonged oxidized samples. The reappearance of the hydrogen-terminated surface resulted in the increase of surface hydrophobicity, observed with the drop shape analysis (Figure 8c). However, the process is not fully reversible, thus leading to a conclusion regarding the corrosive nature of the high-temperature treatment, in particular at prolonged oxidation durations.

#### **4. Conclusions**

In summary, we performed a detailed study of the influence of high temperature on the oxidation processes for heavy boron-doped diamond electrodes. Electrode surface was investigated by both electrochemical and physico-chemical techniques to reveal its charge transfer kinetics and oxidation behavior.

High-temperature oxidation leads to:


The change in the electrode's surface electric properties by local surface oxidation is reflected in a more heterogeneous distribution of the diffusion field, observed by SECM. The longer the high-temperature exposure duration, the lower the value of both anodic and cathodic peak currents and the more irreversible is the character of the oxidation process, which is mainly attributed to the corrosion of BDD grain structure and the deterioration of electrochemical activity.

**Author Contributions:** Conceptualization, J.R. and A.Z.; Methodology, J.R. and A.Z.; Investigation, J.R. (XPS, SEM, electrochemical studies), M.C. (electrochemical studies), M.F. and B.D. (BDD synthesis and structural studies), and A.Z. (AFM and SSRM); supervision, J.R.; writing—original draft preparation, all authors; writing—review and editing, J.R., R.B., and A.Z.; funding acquisition: J.R., K.D., and R.B. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the National Science Centre, grant SONATA number 2015/17/D/ST5/02571, and by the National Centre for Science and Development, grant Techmatstrateg number 347324. The DS funds from the Faculty of Chemistry and the Faculty of Electronics, Telecommunication, and Informatics are also acknowledged.

**Acknowledgments:** The authors acknowledge Lukasz Burczyk for performing the SECM measurements.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Materials* Editorial Office E-mail: materials@mdpi.com www.mdpi.com/journal/materials

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34 Fax: +41 61 302 89 18

www.mdpi.com ISBN 978-3-03936-177-9