Extending the Passive Region of CrFeNi‐Based High Entropy Alloys

This study provides principles for designing new corrosion resistant high entropy alloys. The theoretical framework is a percolation model developed by Newman and Sieradzki that predicts the ability of an alloy to passivate, i.e., to form a protective surface oxide, based on its composition. Here, their model is applied to more complex materials than previously, namely amorphous CrFeNiTa and CrFeNiW alloys. Furthermore, the model describes a more complex passivation process: reforming the oxide layer above the transpassive potential of Cr. The model is used to predict the lowest concentration of Ta or W required to extend the passive region, yielding 11–14 at% Ta and 14–17 at% W. For CrFeNiTa, experiments reveal a threshold value of 13–15 at% Ta, which agrees with the prediction. For CrFeNiW, the experimentally determined threshold value is 37–45 at% W, far above the predicted value. Further investigations explore why the percolation model fails to describe the CrFeNiW system; key factors are the higher nobility and the pH sensitivity of W. These results demonstrate some limitations of the percolation model and offer complementary passivation criteria, while providing a design route for combining the properties of the 3d transition metal and refractory metal groups.


Introduction
The discovery and development of high entropy alloys (HEAs) as both bulk materials and coatings have greatly impacted materials science. [1]The formation of entropy-stabilized phases that combine many elements creates opportunities for new alloy compositions that were not thought possible without severe segregation.Furthermore, in traditional alloy design, most elements are DOI: 10.1002/adfm.202307897added in low concentrations, while in HEAs with their equimolar compositions, the concentrations of all elements are relatively high.This means that each element is more likely to directly impact the corrosion process, including the formation of a passivating oxide layer. [2]Because of the vast number of new, unexplored alloy systems, there is a general need for guidelines regarding how the elements in HEAs should be chosen and combined to obtain the desired properties. [3]ince corrosion resistance typically requires the formation of passivating oxides, the elements in the HEA that contribute to the passive layer are most important for its performance.Not surprisingly, by mixing multiple elements with highly stable and insoluble oxides, very high aqueous corrosion resistances can be achieved.This is often the case for elements within the group of refractory metal HEAs (RHEAs), which includes many metals known for their high corrosion resistances. [4]iZrNbTaMo, [5] Hf 0.5 Nb 0.5 Ta 0.5 Ti 1.5 Zr, [6] and VNbMoTaWAl [7] are examples of alloys that have been shown to greatly outperform the stainless steels (SSs) that were used as reference samples.It is, however, unlikely that this type of HEA could replace SSs in most applications.Steels are typically cheap, light, conductive, and (for austenitic steels) ductile, whereas the RHEAs are harder, denser, more brittle, and form highly resistive oxides during corrosion. [8]he 3d transition metal HEAs resemble SSs more closely.These HEAs have more modest corrosion resistances, comparable to SSs, [9] but are nevertheless of great technological interest due to their combination of properties. [3]Most 3d transition metal HEAs contain Cr [1] and in acidic environments, they rely on its ability to form a Cr-rich passive layer, [10] since the other elements have less stable oxides.Above 0.8 V versus standard hydrogen electrode (SHE), Cr exhibits a transpassive region where Cr is oxidized to Cr(VI) and soluble species are formed.This limits the use of 3d-metal HEAs in low pH and high potential environments.
Water splitting for hydrogen production and energy production in hydrogen fuel cells [11][12][13] are applications in need of materials that can withstand low pH and/or high potential conditions.The proton exchange membrane (PEM) electrolyzers [11] and PEM fuel cells [14] have quickly become leading in the commercial markets.The materials in these devices, including the bipolar plates which often are made of metallic materials, must withstand corrosive conditions.Electrolyzers have cell potentials of 1.48 V versus SHE as a minimum, but commonly operate at above 2 V versus SHE. [12,15,16]The anode in the electrolyzer will be polarized to a potential above 0.8 V versus SHE, i.e., above the transpassive potential for Cr.In a PEM fuel cell, the highest potential at the bipolar plate is slightly lower than 0.8 V versus SHE (at the cathode), but the cells operate at a temperature of 80 °C and low pH, which can also aggravate the corrosion.While the corrosion resistance of the materials must be high, the contact resistance of each component must be low (the bipolar plates must have values lower than 10 mΩ cm 2 according to the US Department of Energy [17] ) to avoid a decrease in the cell voltage.It is a challenge to find bipolar plate materials that meet the abovementioned requirements, while also being cost-effective and easy to process industrially.For the electrolyzers, the use of expensive noble metal coatings is still dominating over that of more common metals and alloys. [13,18]Although metals from the refractory metal family have been evaluated due to their high corrosion resistances, their contact resistances were found to be high after polarization and, therefore, unsuitable for the application. [15]he primary aim of this study is to find principles for designing new HEAs that surpass the corrosion performance of stainless steels at potentials above 0.8 V versus SHE while preserving the beneficial properties of the 3d metal group, such as the lower contact resistance.This problem will be addressed by adding one of the refractory metals to a 3d-metal base alloy: equimolar Cr-FeNi.The idea is to extend the passive region by allowing the refractory metal to passivate the alloy when Cr is dissolved.Two candidate elements were selected, Ta and W, both of which form stable passive oxides in acidic environments (Ta 2 O 5 and WO 3 , respectively).
For the alloy to remain passive above 0.8 V versus SHE, the concentration of the refractory metal must be high enough to sustain a passive layer on its own.There exists a model, first developed by Newman and Sieradzki, which predicts the lowest required concentration of a passive element using percolation theory. [19,20]A recent publication by the group proposed that this model could be used to design more complex alloys, such as high entropy alloys. [2]The secondary aim of this publication will, therefore, be to test the model on the CrFeNiTa and CrFeNiW systems.The influences of Ta and W on the corrosion process will be assessed to examine, if the possibility to extend the passive region depends on which refractory metal is chosen.A key difference between Ta and W is their standard reduction potentials, which should affect the corrosion mechanisms.The percolation model and the expected differences between Ta and W are further explained in the theory section of this paper.
To summarize, coatings of an equimolar base alloy, CrFeNi, with the addition of different concentrations of Ta or W will be synthesized using magnetron sputter deposition and characterized with diffraction, electron microscopy, and photoelectron spectroscopy.Electrochemical analyses in 0.05 m H 2 SO 4 will be performed to study the corrosion resistances of the alloys in an acidic environment.The minimum value of Ta or W to extend the passive region will first be estimated using percolation theory and then determined experimentally.The contact resistances of selected coatings will also be evaluated before and after polarization to different potentials.Postcorrosion surface characterization with photoelectron spectroscopy (XPS) and electron microscopy (SEM) will be combined with inductively coupled plasma mass spectrometry (ICP-MS) investigations of the electrolyte to understand the mechanisms leading to successful passivation of the coatings.

Phase Formation Prediction
Although high entropy alloys have become well-known as singlephase materials, [1] it has been shown that far from all equimolar alloys form a thermodynamically stable single phase. [29]Thermodynamic calculations can be used to predict the equilibrium phases for HEAs using thermodynamic data for the binary and ternary subsystems. [30]The results of such calculations for the alloys studied in this work are presented as phase fraction diagrams in Figure S1 (Supporting Information).None of the alloys in the CrFeNiW or CrFeNiTa systems are expected to form a single phase at any temperature.Furthermore, the large negative formation enthalpies of the possible binary alloys [29] mean that there should be a strong driving force for segregation, particularly in the CrFeNiTa system where Ta forms highly stable intermetallic compounds with both Fe and Ni.
Another aspect that may be important for the phase formation is the ability to avoid crystallization.It is well-known that alloys with many components can easily become amorphous. [31]Zhang et al. [32] classified many HEA compositions using two parameters that describe the atomic-size difference and the mixing enthalpy, respectively.These authors identified parameter ranges that favored either single-phase or multiphase formation, or the formation of amorphous alloys.By calculating the parameters proposed by Zhang et al., [32] it was found that the CrFeNiW system falls mostly in the range for single-phase or multiphase materials, whereas CrFeNiTa falls within the amorphous alloy range.The results of these calculations are shown in Figure S2 (Supporting Information).
It can thus be concluded that the present alloy systems are unlikely to form single phases.If they were synthesized through traditional bulk methods such as melting and casting, they would most likely decompose into multiple phases, which should decrease their corrosion resistance.However, by using a low-temperature synthesis method, such as magnetron sputter deposition, a meta-stable material can be formed that suppresses the phase separation.This effect may be enhanced by the parameters mentioned above.The Ta-containing alloys, which have the highest risk of segregation, are also predicted to form amorphous alloys more easily.

Predicting the Concentrations Required for Passivation Using Percolation Theory
Percolation theory is a mathematic model that describes the connections in a network (e.g., the lattice of a crystal) when the number of nodes or links between sites is changed.Below a critical value, p c , called the percolation threshold, the sites can only form small and disconnected clusters.Above p c , the sites connect to form a cluster that spans the entire lattice.This approach can be used to describe many processes in materials science. [33]ieradzky and Newman [19,20] proposed that the minimum concentration of Cr needed to achieve passivation in stainless steels (around 10 at% in ferrite and 13 at% in austenite) stems from a percolation phenomenon.In this case, the fraction of open nodes was defined as the concentration of Cr atoms in the material.Above p c , Cr can interconnect throughout the whole material by bridging through Cr─O─Cr units that eventually form a continuous Cr oxide layer.Below the threshold, Cr can only form smaller oxide clusters that can be removed via undercutting when the Fe is dissolved.This theory has successfully been applied to several binary alloys. [34,35]ie et al. [2] proposed the use of percolation theory when designing new corrosion resistant alloys, including high entropy alloys.Here, the minimum concentrations of Ta and W will be estimated using this approach.The primary purpose is to understand if we should expect an inherent difference in the minimum concentrations of Ta and W required for passivation.The secondary purpose is to test the proposed method by Xie et al. on amorphous materials, which, to the best of our knowledge, has not been done before.Ta and W should be the only passive elements at potentials above 0.8 V versus SHE.According to Sieradzki and Newman's theory, this means that passivation is possible if Ta or W can connect through Me─O─Me bonds, i.e., above p c .This is achieved only if the distance between the Ta or W in the alloy is shorter than the oxygen bridging distances, which are 4.28 and 4.00 Å, respectively, for Ta─O─Ta and W─O─W bonds.
Finding the percolation threshold, p c , requires a description of the structure.As will be shown in the Results Section, most of the materials obtained in this study were amorphous.Only the base alloy (FeCrNi) and one of the alloys with added W were single-phase fcc.In a crystal lattice, the atomic distances are well-defined, so finding the percolation threshold is simple; it requires only knowledge of the unit cell parameters.For amorphous materials, however, the structure is difficult to both describe and characterize.Many different models for the structure of metallic glasses have been proposed and their validity is still under discussion. [36,37]For some of these models, there are analytical solutions describing the connectivity between nearest neighbors. [38,39]It is, however, a considerably more complex task to describe such properties for extended-range interactions, i.e., up to the oxygen bridging distances described above.No such analytical solution exists to our knowledge.
To estimate p c we need to, instead, rely on results from experiments or computer simulations of amorphous materials.We have chosen four routes to estimate the percolation threshold of the present materials systems using data from the following studies • A computer simulation of random packing of spheres by Powell [40] • An experiment by Aste et al.where the packing of 140 000 spheres was characterized with X-ray tomography [41] • A molecular dynamics (MD) simulation by Jiang et al. [42] where the structure of amorphous AlCoCrFeNi (which has a similar atomic size difference to CrFeNiTa) was explored • An experimental study by Chen et al. [43] where amorphous Nb 40 Ni 60 (which has a slightly higher atomic size difference than CrFeNiTa) was characterized by total X-ray scattering None of the four data sets is a perfect representation of the present materials but they can provide a reasonable estimate that will aid the interpretation of our experimental results.These calculations are also an important step in testing the design approach proposed by Xie et al. on amorphous alloys.By comparing p c obtained using different data inputs, from idealized models to real materials, and by evaluating the scatter in the results, we can better understand the limitations of the method.
In the study by Powell, [40] p c for different radial distances (r) was one of the extracted results from the simulation and can be found tabulated.For the remaining three models, p c was not evaluated.To obtain an estimate of p c , we have used the relationship between p c and the coordination number, z, given by Equation (1) where d is the dimension and  c is the continuum percolation threshold ( c = 0.34 189 for spherical particles [44] ).This relationship was observed by Powell [40] and better described by Xun et al. [45,46] who also provided a theoretical explanation.Powell tested the relationship for his amorphous model and found that it fitted the data for z ≥ 12. Data for z as a function of r was supplied by Aste et al. [41] For refs [42] and [43] , it was extracted from graphs of the pair distribution function, requiring only the global density as additional input.
The four curves of p c as a function of the radial distance are displayed in Figure 1.The radial distance was normalized by dividing r by the average sphere/atomic diameter, d, to allow for a comparison between the models.The percolation threshold of a face-centered cubic (fcc) crystal is also displayed, based on values given by Shante et al. [33] The most striking difference between the crystalline and amorphous models is that p c for fcc decreases stepwise, while the amorphous decreases continuously.This is expected from the distinct atomic positions of crystals and the more diffuse order in amorphous materials.It entails that, in an amorphous material, p c is more sensitive to changes in atomic diameter or bridging distance.The lowest values were extracted from the MD simulation by Jiang et al., while the remaining three models give more similar and higher values.2][43] To find p c for our materials, the oxygen bridging distances were normalized by dividing by the atomic diameter, d calc , for each sample composition (the values are found in Table 1).We thus obtained ranges for the maximum distance for Me─O─Me connectivity, which were marked out in Figure 1.We can see that in the fcc lattice, the Ta and W atoms have to be either the nearest or next-nearest neighbors (1st or 2nd coordination shell), as the spacing between the 2nd next-nearest neighbors (3rd shell) would exceed the oxygen bridging distances.This places the 3D percolation threshold at 13.6 at% for both metals.For the amorphous models, p c varied over the sample series by more than 1 at% (within each model).The highest p c was found for the compositions with the highest amounts of Ta or W, which had the largest average atomic distances.For Ta, the range of p c considering all models and compositions was between 10.6 and 14.2 at%.
Table 1.Compositions and thicknesses of all the samples, measured by energy-dispersive X-ray spectroscopy EDS and SEM, respectively.Average atomic diameter/smallest inter-atomic distance, d, calculated from tabulated metallic radii (calc) and extracted from the diffraction data (exp) using Braggs law and the Ehrenfest relation. [55,56]Estimated percolation thresholds, p c , for Ta and W connectivity, given as mid-range values.For W, p c was between 13.6 and 17.4 at%.The mid-range value for each sample composition is displayed in Table 1.
We have thus evaluated p c based on four data sets, two idealized models based on the packing of equally sized spheres, as well as one simulated and one real material with atomic size differences comparable to the present sample sets.The packing of spheres is highly affected by size differences. [47]It will, therefore, be one of the largest sources of error for the idealized models.This effect has been demonstrated using Powell's model. [48]When the size difference increased, the 1st shell coordination number around the larger spheres increased and rapidly lowered the threshold for percolation through touching contacts.The behavior was, however, never tested for larger radial distances.Understanding this effect would require a dedicated simulation using the relevant size differences.Considering this effect, it is interesting that the model by Jiang et al yields the lowest values for p c , since it is the model with the most similar atomic sizes as the present systems.The metallic radius of Al is 1.43 Å, which lies between the radii 1.41 and 1.47 Å of W and Ta, respectively.Nevertheless, four completely different models yielded a range in p c of only a few at%.For the purpose of describing the passivation process, for which percolation theory is only a rough description, an error of this magnitude is acceptable.
A potentially larger source of uncertainty for the percolation model applied to passivation was discussed by Xie et al. [2] The percolation threshold in three dimensions, p c , represents only the lowest concentration at which passivation is possible.At this concentration, the remaining elements (Fe, Cr, and Ni) would then have to dissolve to a depth of several thousand atomic layers before passivation would occur after, for instance, a rupture of the passive oxide.Xie et al. [2] proposed to instead define the threshold value by a maximum number of dissolved layers.This increases the passivation threshold, an effect which can be described as a crossover between the 3D percolation in the bulk and the 2D percolation across the surface of the material (demonstrated by Sotta and Long [49] ).Depending on how much metallic dissolution is tolerated (Xie et al. suggested around 10-20 atomic layers), the required value for stable passivation could be up to 7 at% higher than the original p c .The description of the crossover effect is dependent on the crystal structure and has not been simulated for an amorphous system, which is why even an estimation is difficult in the present case.It should also be noted that this mechanism would depend highly on the state of the surface before passivation occurs, e.g., on how the corrosion test is performed.It is not certain that this is relevant for the present experiments.This is briefly discussed in the summarizing discussion (Section 3.6).
To summarize, the percolation thresholds were calculated and the results indicate that the threshold is expected to be around 3 at% higher for the amorphous CrFeNiW system compared to the amorphous CrFeNiTa system.The materials characterization presented in Section 3.1 will be used to assess the validity of these estimations.

Predicting the Oxide Layer Composition
Based on the Gibbs free energies of the metals and their respective oxides, the relationship ΔG°= −n F E°c ell can be used to calculate the corresponding standard reduction potentials for each metal.These can then be used to make a third prediction about the corrosion mechanism of these alloys, namely which elements in the alloy should be preferentially oxidized.
The reduction of Ta 2 O 5 to Ta has a standard reduction potential of −0.94 V versus SHE.This is comparable to the values for the reduction of Cr 2 O 3 to Cr, or Fe 2 O 3 to Fe, which are −0.76 and −0.89 V versus SHE, respectively.The reduction of WO 3 to W has a standard reduction potential of −0.27 V versus SHE, whereas Ni is not expected to form a passive layer below pH 9 (assuming [Ni 2+ ] < 10 −6 m). [50]The standard reduction potential for the reduction of Ni 2+ to Ni is, however, −0.25 V versus SHE.
These differences between the standard reduction potentials (or between the equilibrium potentials under nonstandard conditions) can strongly affect the composition of the oxide layer formed on the materials since the less noble elements should be preferentially oxidized.For stainless steels, it has been observed that the less noble Cr and Fe are present in the oxide layer and that this layer then is enriched in Cr since Fe dissolves at a higher rate.The nobler Ni is, on the other hand, enriched in the alloy below the oxide. [51,52]A similar layered structure has likewise been observed for high entropy alloys. [53,54]It would, therefore, be expected that Ta and W would be incorporated into the oxide layer at different stages and to different extents.The less-noble Ta should be more active than W in the formation of the oxide under open circuit potential (OCP) conditions and at low potentials (i.e., at potentials more negative than the WO 3 /W standard reduction potential).This means that, when the transpassive potential is reached during a potential scan, there should be more Ta than W in the oxide.One question that is discussed below is then if this difference affects the ability to extend the passive region to higher potentials.

Materials Characterization
The atomic compositions of all the samples, which were determined with energy-dispersive X-ray spectroscopy (EDS) in SEM, are presented in Table 1.Using XPS, the oxygen content in the bulk of the coatings was determined to be below 1 at% and the oxygen concentration was therefore omitted in the concentration calculations.Despite the use of an equimolar CrFeNitarget, the relative concentrations of Cr, Fe, and Ni were not perfectly equimolar.The relative amounts of these elements also varied within the sample series, particularly when W was added to the coatings.XPS was also used to measure the core level spectra of the five metals, as is shown in Figure S3 (Supporting Information).The binding energies (BE) of the primary peaks, obtained through peak fitting using asymmetric Donjac-Sunjic peak shapes and a Shirley background, are shown in Figure S4 (Supporting Information).The Fe2p 3/2 BE was found to vary by less than 0.1 eV between the samples.Slight shifts were observed for all other elements.The most prominent shift was found for Ni, where the BE was found within a range of 0.4 eV.
X-ray diffraction was used to study the phase content and the crystallinity of the coatings.The results are shown in Figure 2.Only two samples exhibited sharp diffraction patterns: the base alloy CFN0 and CFNW13.In the −2 measurement, presented in Figure 2a, these samples displayed a sharp peak corresponding to the 111-planes of the face centered cubic (fcc) phase, as well as a weaker 222-peak.This indicates that the coating had a strong 111 texture.The 111 peak was broader for CFNW13 than for CFN0.All the remaining samples containing Ta or W were X-ray amorphous, as seen in Figure 2b.The diffractograms only showed a broad and weak halo, which shifted in position with the Ta/W concentration.
The average interatomic distances between nearest neighbors, d exp , were extracted from the diffraction data.For the crystalline samples, d was extracted from the position of the 111-peak.For the amorphous samples, the distance to the first coordination sphere can be determined from the halo position using the Ehrenfest formula, [55,56] show below in Equation ( 2) where  is the scattering angle and  is the X-ray wavelength.The values are given in Table 1 and Figure 2c, where they are also compared to the weighted average atomic diameters, d calc , which represent the interatomic distances in an idealized case where the nearest neighbors are in touching contact.The experimental values follow the expected trends, with a deviation within ±2 pm for most samples.The values of d calc and d exp increased with the addition of Ta or W, and the Ta-containing coatings had larger d values than the W-containing coatings for the same Ta/W concentration.The sample for which d exp deviated the most from d calc was the crystalline CFNW13.This can be understood considering that the large W atom would make the structure less close-packed than in an fcc lattice with more equal-sized spheres.In this case, the amorphous structure could allow for closer inter-atomic contacts and thus be better represented by the average atomic diameter.
The structures of two selected samples, i.e., CFNT25 and CFNW25, were further investigated by electron diffraction in top-view transmission electron microscopy (TEM).Their selected area electron diffraction (SAED) patterns, shown in Figure 3, only displayed diffuse rings, confirming the lack of long-range ordering and the absence of nanocrystallites.The patterns of diffuse rings were similar in the two amorphous materials, which sug-gests that the ordering in the amorphous alloys was also similar.The amorphous structure was also investigated with high resolution scanning TEM (STEM) imaging, presented in Figure S5 (Supporting Information).No long-range ordering was visible, only small regions of local order which, in some places, spanned up to 3-4 atomic layers.This is not unexpected for amorphous materials.A structure with such small regions of order is best described as amorphous and should be well represented by experimental data used in the calculations in Section 2.2.The TEM-EDS maps, presented in Figure S5 (Supporting Information), showed no clustering of elements, only the independent fluctuation of each element which is expected from a random distribution.
The observed formation of amorphous and homogenous coatings, despite the high driving force for segregation, was due to the unique properties of magnetron sputtering.During sputtering, the atoms are ejected from the targets and then quenched from the gas phase directly onto the substrate, which is at close to room temperature.This rapid quenching can limit the diffusion so that the thermodynamically stable phase cannot form.Structures and phases that require shorter diffusion paths will then be favored.This has previously been utilized to synthesize single-phase coatings of alloys that are known to form multiple phases when they are cast, for example, CrCoCuFeNi [57] and AlCoCrCuFeNi. [58]In the present case, the alloys were quenched into an amorphous state.Because of the risks associated with phase separation (e.g., depletion of passivating elements in some phases and galvanic coupling) single-phase HEA are generally found to be more corrosion resistant. [59]In the present work, the complexity of multiphase formation has been avoided, which allows us to focus on the effects of the alloy compositions.
From the characterization, it can be concluded that the average atomic distances varied as expected from average atomic diameters with a deviation of only a few pm, supporting the validity of the values obtained in Section 2.2.It can also be concluded that the amorphous structures in CrFeNiTa and CrFeNiW have similar ordering.Finally, no segregation was detected with EDS, meaning that we should not expect a large change in the percolation threshold due to clustering.According to percolation theory there is, therefore, support that the threshold value for passivation of CrFeNiW should be a few at% above that for Cr-FeNiTa.In the next section, electrochemical analyses will be used to test this hypothesis.

The Minimum Concentration of Ta or W needed for Repassivation
Figure 4a,b displays the results of potentiodynamic polarization experiments with samples containing different concentrations of Ta and W. For comparison, the polarization curves for the base alloy, i.e., CFN0, and the pure Ta and W reference coatings have also been included.Selected results from experiments performed with different scan rates are presented in Figure S6 (Supporting Information).These will be discussed as needed, to clarify whether the positions of the peaks are purely potential-dependent or if there is also a time-dependence.
Polarization curves display the electrochemical response of samples exposed to an increasing potential, i.e., an increasing driving force for oxidation.The base alloy CFN0 exhibited a behavior similar to a typical Fe─Cr alloy. [2]A net cathodic current was observed until the corrosion potential (E corr ) was reached, after which the net current was anodic.The anodic current was, however, quickly suppressed and a plateau was reached.This plateau indicates the passive region, where a Cr-rich oxide covers the surface, and extends to around 0.8 V.The polarization curves for the W series indicate a similar passive behavior.The formation of a passive layer at potentials below 1 V versus Ag/AgCl is expected, since all the samples had Cr concentrations of 21 at% or above, which, at least for crystalline Cr-containing alloys, is well above the passivation threshold value.The pure W reference also appeared to passivate, although its E corr was found at higher potentials.The latter is, however, expected based the higher nobility of W compared to Cr and Fe.The samples in the Ta series and the pure Ta reference, on the other hand, display a different behavior.At potentials above E corr , the current values were very low and increased slowly during the scan.This is the typical behavior of samples that were already passive prior to reaching E corr .
At around 1 V, an oxidation peak was observed for all the coatings (except pure W and Ta).This potential is close to the expected potential where the oxidation of Cr from +III to +VI should take place [50] (i.e., the onset of the Cr transpassive region).The peak at about 1 V versus Ag/AgCl can, therefore, be explained by the oxidation of the Cr present in the surface oxide yielding soluble Cr(+VI) species.When different scan rates were used, the peak was found at the same potential, which indicates that the process was potentialdependent.
Above 1 V, the current increased for CFN0 until 1.5 V, after which a steep current decrease was observed.SEM micrographs of the surface of this sample, recorded after polarization to 1.7 V, showed that the coating was gone from the substrate, leaving only scattered particles with a size of about 10 nm behind (see Figure 5; and Figure S7, Supporting Information).The steep current decrease was hence due to the exposure of the substrate, SiO 2 , which is stable at these potentials. [50]It can therefore be concluded that the coating was almost completely dissolved, and that this process began as soon as the Cr transpassive potential was surpassed.
For the samples containing Ta or W, the dissolution of the coating appeared to have been suppressed, at least partially.In these cases, the peak at 1 V versus Ag/AgCl was followed by a valley.However, for samples with less than 15 at% Ta or less than 45 at% W, the current increased shortly thereafter.The second current increase was shifted to higher potentials when higher scan rates were used (see Figure S6, Supporting Information).This indicates that the current increase was not caused by the activation of a specific electrochemical reaction, but rather a consequence of the length of time spent above the transpassive potential (which becomes shorter when the scan rate is increased).For the samples with the lowest amount of Ta or W (i.e., CFNT11 and CFNW13), a steep current decrease was seen shortly before the end potential of the scan was reached.When different scan rates were used, this feature was always found to be associated with the same oxidation charge.This would be expected if the coating finally was lost due to a complete oxidation, as was observed for CFN0.The mechanism behind this will be further explored in Section 3.3.1.When more Ta or W was added, the second current increase was delayed, as was the full oxidation of the coating.For the CFNW45 sample, the curve flattened out after the first peak at 1 V versus Ag/AgCl so that there was no steep increase in the current at higher potentials.This indicates the presence of a second passive region.
The behaviors of the coatings containing 15 and 25 at% Ta require more investigation.After the peak at 1 V, the current decreased steadily, but at 1.6 V, it increased again, more rapidly than for the other samples and at the same potential for both samples.This increase was also found at the same potential regardless of the scan rate (see Figure S6, Supporting Information).Furthermore, the coatings appeared to be intact at the end of the curve.This indicates that this effect was more likely to be due to the oxygen evolution reaction (OER) than an oxidative dissolution of the coating.The CFNT15 and CFNT25 coatings hence remained passive even at potentials above 1.5 V.
Some key parameters, extracted from the polarization curves, are displayed in Figure 4c-f as a function of the concentration of either Ta or W. Figure 4c,d presents the influence of the Ta or W concentration on E corr and j corr , respectively.The j corr values were extracted assuming Tafel behavior only for the cathodic process, which means that the linear part of the cathodic curve was extrapolated to E = E corr and that the calculated current density at this point was assumed to represent j corr .A clear difference can be seen between the influence of the metal concentration on the E corr and j corr values for the two sample series.As E corr and j corr varied only slightly with the W content, the initial passivation and corrosion rates were evidently not strongly dependent on the W concentration.In contrast, an increase in the Ta concentration resulted in a gradual change in both E corr and j corr .E corr was thus shifted to higher potentials, i.e., from −0.33 to −0.01 V, whereas j corr decreased by three orders of magnitude upon increasing the Ta concentration from 0 to 25 at%.This clearly shows that the Ta concentration influenced the initial passivation process.
Figure 4e shows the charge obtained by integration of the current in the polarization curves between E corr and 0.8 V.The charge increased when the W concentration was increased, from 7.48 for CFN0 to 17.3 mC cm −2 for CFNW45.This effect could be explained by the decrease in relative Cr concentration that accompanied the addition of W. This, in turn, indicates that the presence of W does not greatly influence the passive layer, as the passivation is more reliant on Cr than on W. The addition of Ta, on the other hand, greatly decreased the charge, with was 0.360 mC cm −2 for CFNT25, once again showing that the Ta plays a greater role in the passivation compared to W. However, the CFNT11 sample clearly broke this trend yielding a charge of 14.9 mC cm −2 , which is almost two times higher than that for the CFN0 sample.This is an important observation since it shows that Ta concentrations below a certain concentration value may in fact make it more difficult to achieve passivation.
Figure 4f shows the charge of the first peak in the transpassive region as a function of the Ta or W concentration.The charge was integrated from 0.8 V versus Ag/AgCl and to the lowest point in the valley after the peak.It can be seen that the charge decreased when the concentration of either Ta or W was increased, but also that lower charges were reached in general for the Ta series.For the CFNT25 and CFNW25 coatings, which had the same concentration of refractory metal, the obtained charges were thus 7.0 and 19 mC cm −2 , respectively.These results indicate that the transpassive dissolution of Cr was better suppressed by Ta than by W.
Based on the data from the polarization curve experiments, it can be concluded that the transpassive dissolution and failure of the coatings could be prevented using a sufficiently high concentration of Ta or W. The threshold value for the passivation at high potentials appeared to be between 13-15 at% Ta and between 37-45 at% W, at least on the time scale of the employed polarization curve experiments.It can also be concluded that the initial passivation was more dependent on the Ta content than the W content, and that the transpassive dissolution of Cr could be suppressed using lower Ta than W concentrations.These effects can be explained by assuming that W was present at a lower degree than Ta in the initially formed oxide, as was predicted in Section 2.3.This hypothesis will be further tested in conjunction with the surface characterization presented in Section 3.3.Here, the mechanisms responsible for the failure and the extension of the passive region will also be explored.

The Evolution of the Surfaces During the Recording of the Polarization Curves
To further understand the phenomena seen in the transpassive region, four samples were selected for in-depth postcorrosion analysis.The first two samples were CFNT11 and CFNW25, which had refractory metal concentrations below the experimentally determined threshold values (i.e., 15 at% Ta and 45 at% W, respectively, see Section 2.2).The CFNT25 and CFNW45 were then selected as samples with concentrations above the threshold values.To follow the evolution of the corrosion process, the samples were polarized using a scan rate of 1 mV s −1 , starting from −0.7 V.The scan was then stopped at five selected potentials.These potentials were different for each sample and were selected to include one point at the end of the primary passive region (A), one point close to the highest point of the first peak (B), and one point during the subsequent current decrease (C).For the CFNT11 and CFNW25 samples, the points D and E were placed in the region of the second current increase.For CFNW45, D was placed at the small second current increase, whereas E was placed at the end of the curve at 1.7 V.For CFNW25, no corresponding second current increase was present, so D was placed at the same potential as for CFNW25, while E was placed at the end of the curve.At these potentials, the samples were removed from the cell and their surfaces were characterized with microscopy, photoelectron spectroscopy, and atomic force microscopy.The electrolytes were also analyzed with ICP-MS to determine the concentrations of the dissolved metals.
The results of the abovementioned analyses are presented in Figures 6 and 7. Figures 6a,d  The different circular layers in Figures 6c,f and 7c,f represent the results of the XPS measurements performed with different analysis angles, i.e., 10°, 30°, or 80°.The outer ring represents the result of the most surface sensitive measurement (i.e., the results obtained with an angle of 10°), while the innermost ring depict the composition of the bulk alloy measured using EDS.The concentrations obtained from the measurement at 80°represent the composition of the full oxide layer since the measurements probe the full depth of the oxide and the relative sensitivity factors account for the differences in mean free path of the different core levels.The measurements at the lower angles, however, should only be viewed as a way of examining which elements were enriched or depleted closer to the surface as the information depth varies between the different core levels.The samples were exposed to air for at least a 1 h prior to the XPS experiments, which may have changed the oxide compositions.The samples that were removed from the electrolyte at a potential in a stable passive region (such as point A) should remain stable in air, but those that were removed during active dissolution (such as points B-E for CFNT11) are more likely to have changed during the subsequent short open circuit period in the electrolyte and later during the contact with air.The latter samples are therefore less representative of the surface evolution during the experiment and should be interpreted with special care Figure 8.

Mechanisms Behind the Failure of the Coatings
The results for the two samples with concentrations below the threshold values, i.e., CFNT11 and CFNW25, indicate that the amount of dissolved metal ions in the electrolyte increased during the scan, as can be seen in Figure 6b,e.For both samples, it was the Cr and Ni that were dissolved most rapidly.Fe could unfortunately not be detected in most of the electrolyte samples, but at point D for CFNT11, the high detection limit for Fe was surpassed.For CFNT11, the Ta concentration was below the detection limit until point D was reached.The W concentration was found to be above the detection limit at all points but did not increase between points B and E as was seen for Cr and Ni.The oxide compositions for the CFNT11 and CFNW25 samples at the different potentials are shown in Figure 6c,f.At point A, before entering the transpassive region, the CFNT11 oxide contained 21 at% Ta, which is about double the bulk Ta concentration.The W concentration in the CFNW25 oxide was, on the other hand, 29 at%, which is closer to the bulk concentration.Ta was also more enriched in the outer layer of the oxide compared to the inner layer, while the W was slightly depleted in the outer layers.The rest of the oxide was mainly Cr and Fe, but some Ni was also found in the CFNW25 oxide.
When entering the transpassive region, the concentrations of Ta and W in the oxides increased gradually.The Ta concentration increased from 21 to 80 at% from point A to E, and the increased was more dramatic in the surface layers than in the inner layers.The W concentration likewise increased from 29 to 76 at% but was more evenly distributed through the depth of the oxide.It should also be noted that for the CFNT11 sample, the oxide was completely depleted of Cr at point E.
More information on the breakdown mechanisms was obtained via microscopy studies of the samples.Beyond point D both the CFNT11 and CFNW25 samples exhibited an increased surface roughness indicating active dissolution of the coatings (shown in Figures S9 and S10, Supporting Information).The samples were also studied after reaching the polarization curve end points, i.e., 1.7 V versus Ag/AgCl for CFNT11 and 2 V versus Ag/AgCl for CFNW25.The SEM images can be seen in Figure 5b,e.CFNT11 had fully dissolved under the O-ring, while the center of the corroded area featured a thin layer of a flaky, porous material, most likely due to a precipitation of the previously dissolved corrosion products.This is similar to the final state reached for the CFN0 coating.Point measurements performed with EDS showed that the surface composition was almost entirely Si and O, with only small amounts of Fe, Cr, and Ni and C contamination This confirms that the coating was completely detached or dissolved.
The polarization of the CFNW25 sample was continued to 2 V, where the current decreased as it did for CFN0 or CFNT11 when they were dissolved (see Figure S6c, Supporting Information).However, the end result was different for the CFNW25 coating.The microscopy studies, seen in Figure 5e, revealed that the coating had been replaced by a thick layer of a new material, which was found everywhere except under the O-ring.Using EDS, the material that was shown to consist of a W rich oxide containing smaller amounts of Fe (see Figure 8).The W-rich oxide can be seen in cross-section in Figure 5e; and Figure S7e (Supporting Information).The oxide was thicker (about 1.5 μm) than the original coating (650 nm) and was filled with cracks, indicating that a volume expansion had taken place during the polarization.The oxide was also separated into layers, which could be explained by repeated cracking caused by the volume expansion.In some places, some or all of the layers had flaked off the surface when the SEM analysis was performed, for example in Figure 5e.The increase in thickness indicates that the formed oxide was underdense.Adding the fact that the Cr and Ni had been fully dissolved, it can be concluded that the oxide must be nanoporous.Although the presence of a continuous W-rich oxide indicated that the percolation threshold had been surpassed, this had clearly not prevented the remaining elements from undergoing dissolution.The oxide that formed was hence nonprotective, i.e., nonpassivating.

Mechanisms Behind the Successful Extension of the Passive Region
In this section, the results for the samples with Ta or W concentrations above the passivation threshold value will be discussed.The results from the stepwise polarization of the CFNT25 and CFNW45 coatings are shown in Figure 7.The ICP-MS results show that the dissolution rates for the CFNT25 sample were lower than for any other sample, the concentrations of the four metals were, in fact, below the detection limits at all measurements points.Note that this does not mean that Figure 9. Results of selected XPS measurements.a) Ta 4f core level spectra for CFNT11 and CFNT11 samples polarized to different potentials, as well as for a pure Ta reference after 2 h at OCP and after polarization to 2 V. b) W 4f core level spectra for CFNW25 and CFNW45 samples polarized to different potentials, as well as for a pure W reference after 2 h at OCP and after polarization to 2 V.The spectra, which were recorded with an 80°analysis angle, were energy-calibrated using the metal signal from the bulk measurements.The vertical lines and gray regions are the binding energy ranges based on literature values for Ta [60][61][62] and W [60,63] in their metallic and oxidized states, respectively.The labels A, D, and E correspond to the same potentials values given in Figures 6 and 7.
there was no dissolution occurring.For example, when reaching 1 V versus Ag/AgCl, the Cr that was in the oxide must have been dissolved, which is why the Cr cation fraction decreased in Figure 7c.The lack of Cr signal simply means that the accumulated concentration over the course of the whole experiment was not high enough to be detected.For CFNW45, there was detectable dissolution of Cr, Ni and W but the dissolution rates did, however, not increase dramatically when the potential was increased, as was seen for the CFNW25 coating.
The Ta cation fraction in the CFNT25 oxide increased from 38 at% at point A to 55 at% at point E. It can also be noted that, while the fraction of Ta at point A was significantly higher for CFNT25 than for CFNT11, the final fraction (i.e., at point E) of Ta was lower.This indicates that the enrichment of Ta in the outer layer seen at point A successfully prevented a complete dissolution of Cr and Fe in the Cr transpassive region.The W content at point A was likewise higher in the CFNW45 coating than in the CFNW25 coating and increased from 45 at% at point A to 76 at% at point E.
Of the four investigated samples, the highest Cr content at point A was found for the CFNT11 sample exhibiting a Cr cation fraction of 59 at%.The corresponding values for the CFNW25, CFNT25, and CFNW45 samples were 40, 36, and 24 at%, respectively.This demonstrates that the charge of the peak at 1 V, seen in Figure 4f did not depend solely on the cation fraction of the Cr in the oxide at point A. For CFNT25, it appears that the high Ta content in the outer layer of the oxide was key to suppressing the dissolution of the nonpassive metals.Note also that the thickness of the oxide may differ between the samples, meaning that a higher relative amount of an element in the oxide should not necessarily be reflected by a higher absolute amount of that element in the electrolyte.
The SEM studies of the CFNT25 coating showed that the there was no difference between the pristine and the final state of the coating (compare Figures 3b with 5f).At the end of the polarization curve, the CFNW45 coating was slightly roughened compared to the pristine state, but there was no detectable change in the thickness of the coating (see Figure 5f).

Formation of Mixed Oxides
Figure 9a presents the Ta4f core level spectra obtained for the CFNT11, CFNT25, and a pure Ta reference coating, after polarization to different potentials, whereas Figure 9b presents the corresponding W4f spectra for the CFNW25, CFNW45, and a pure W reference coating.The spectra show peaks due to two distinct environments, and each contributing environment is also split into 7/2 and 5/2 spin doublets.The low binding energy peaks are the signals from the bulk alloy, while the higher binding energy peaks are compatible with binding energies reported for Ta 2 O 5 and WO 3 , respectively.
The binding energies for the oxide peaks were lower for the alloys compared to for the pure Ta and W references.The difference was larger for Ta (up to −1.1 eV) than for W (up to −0.5 eV).These shifts indicate that the Ta and W environments in the oxide were changed when Ta and W were alloyed with Cr, Fe, and Ni.It is consequently likely that mixed oxides were formed as opposed to segregated layers of Ta 2 O 5 and WO 3 surrounded by Cr and Fe rich oxides.For both the Ta and W series, the samples with the highest concentration of the refractory metal, i.e., CFNT25 and CFNW45, had binding energies closer to those of pure Ta and W, respectively.
The Ta and W references were measured both after exposure to the electrolyte for 2 h and after polarization to 2 V.While the relative oxide contributions increased as a result of the polarization of the samples, due to the increased oxide thickness, the binding energies for the oxide peaks remained the same.For the alloys, however, the positions of the peaks shifted slightly.This is expected based on the gradual change in oxide compositions seen in Figures 6 and 7.The shift was toward lower binding energies for CFNT25 and toward higher binding energies for CFNW45.

Establishing the Ta and W Threshold Values
The results discussed above show that the samples with 13 at% Ta began to dissolve in the Cr transpassive region, whereas the samples with 15% Ta remained passive.For the W series, the samples with 37 at% W dissolved, while the samples with 45 at% W remained passive.In this section, these four types of samples were subjected to potentiostatic polarization at potentials in the Cr transpassive region in different ways.The purpose was to investigate if the threshold value could be changed by arriving at a high potential via different paths, and if the passivity in the passive region was stable over longer times.
In Figure 10, the results of two different types of potentiostatic polarization experiments are presented.The coatings were either polarized at 1.5 V versus Ag/AgCl for 5 min directly after 2 h at open circuit in the electrolyte, or, alternatively, first polarized at −1.2 V versus Ag/AgCl for 4 min and then at 1.5 V versus Ag/AgCl for 5 min.When polarizing at −1.2 V, the oxide should, at least partly, be reduced since this potential should be below the equilibrium reduction potential for all five metals.Longer versions of these experiments, where the polarization at 1.5 V versus Ag/AgCl was 1 h, were also performed.The curves from these experiments are presented in Figure S11 (Supporting Information).
The samples CFNT15 and CFNW45, which had remained passivated during the potentiodynamic polarizations, displayed a similar behavior in the present experiment.In CFNW45, the difference was smaller between the sample that was polarized at 1.5 V versus Ag/AgCl with and without initial reduction.For CFNT15, this difference was larger.The reduced CFNT15 and CFNW45 samples initially exhibited higher oxidation currents at 1.5 V.This is expected since a more extensive oxidation would be required to achieve a passive state after having partially reduced the oxide.The currents for both CFNT15 and CFNW45, with and without reduction, decreased during the 5 min measurements, indicating a stable passive behavior.This was also observed when polarized to 1.5 V versus Ag/AgCl for 1 h.During the longer measurement, the initially reduced CFNT15 continued to exhibit higher oxidation currents during the entire measurement, while the currents of the reduced and nonreduced CFNW45 samples converged after the first 15 min.
For the CFNT13 and CFNW37 samples, i.e., the samples with Ta and W concentrations below the threshold values, the currents were generally higher than for the passivated samples (CFNT15 and CFNW45).For CFNW37, the current increased during the 5 min oxidation experiment.This was a sign of a gradual breakdown of the coating, indicating an inability to form a stable passive layer.The highest oxidation current was found for the CFNT13 sample after polarization to 1.5 V versus Ag/AgCl directly from the OCP.In this case, the current also fluctuated during the measurements.When the oxide was reduced, the CFNT13 oxidation current at 1.5 V versus Ag/AgCl was lower and more stable.This relative stability of the reduced sample could be due to an increase in the Ta content in the oxide when the oxide was reformed at high potentials.The lower Cr content could then lead to a less aggressive dissolution of Cr.However, when the samples were polarized at 1.5 V versus Ag/AgCl for 1 h, both the CFNW37 and CFNT13 coatings were eventually fully oxidized, re-www.afm-journal.degardless of whether an initial reduction at −1.2 V versus Ag/AgCl was used or not.The CFNW37 coating remained on the surface in the form of a porous oxide, while the CFNT13 coating was fully dissolved.

The Effect of Ta Content on the Contact Resistance
For the Ta containing samples which formed stable passive layers at high potentials, i.e., CFNT15 and CFNT25, interfacial contact resistance (ICR) measurements were performed to study how the composition of the alloys affected this resistance.Figure 11 shows the interfacial contact resistance (ICR) measured on pristine samples, on samples exposed to 15 min at 0.6 V versus Ag/AgCl and 15 min at 1.5 V, respectively.The two values, 0.6 and 1.5 V, are similar to the potentials the materials have to withstand at the cathode of a PEM fuel cell and the anode of a PEM electrolyzer, respectively.Based on the analysis in Section 3.3.2,a potential of 1.5 V versus Ag/AgCl should be located in the transpassive region for Cr, which is why the surface should, be more enriched in Ta.
The results show that an increased Ta concentration in the coating increased the ICR.The pure Ta coating showed significantly higher ICR values compared to those for the alloyed samples.For CFN0, CFNT15, and CFN25, an increase in the ICR value of 33-48% was seen after 15 min at 0.6 V, whereas the ICR value for the pure Ta was almost doubled.However, an even more significant increase in the ICR value was seen after 15 min at 1.5 V.For the CFN0 coatings, the electrolyte was tinted yellow after the hold at 1.5 V, and XPS and SEM confirmed the full dissolution of the coating (Figures S12 and S13, Supporting Information).For the Ta containing samples, the coatings were intact after the corrosion due to the passivating effect of Ta.The contact resistance of the CFNT15 and CFNT25 were higher than that of the exposed SS316L substrate of the CFN0 sample, but lower than for the Ta reference.The increasing ICR with increasing po- tential is due to the increased concentration of Ta in the passive film, as confirmed by the XPS data and discussed in the previous sections.

Summarizing Discussion: Criteria for Extended Passivation
After extensive electrochemical analysis and characterization of the surface, it can be concluded that a minimum concentration of Ta or W is needed to extend the passive region to higher potentials for both the CrFeNiTa and CrFeNiW alloy systems.Below these threshold concentrations, the coatings eventually detach or dissolve, despite the fact that the surfaces are enriched in Ta and W.This is in line with the percolation model for passivation of alloys.If the passivating element is not interconnected, i.e., percolating, through the whole alloy, it will only form small oxide fragments that can be removed by dissolving the Fe, Cr and Ni surrounding it, a process known as undercutting.This appeared to have been the case for the Ta series, where the coatings either achieved a stable passive layer (CFNT15 and CFNT25), or dissolved completely (CFNT11 and CFNT13), leaving behind only precipitated corrosion products scattered on the surface.The apparent minimum value, 13-15 at%, agrees well with the theoretically calculated 3D percolation threshold of 11-14 at%.
An interesting observation is that the estimation of the 3D percolation threshold was enough to predict the successful passivation of the surface.This demonstrates that, in the current experiments, the increase in percolation threshold due to 2D-3D crossover effects that was described by Xie et al. [2] does not apply.This can be understood as the surface is already enriched in the passivating element.
Another interesting observation was that the presence of Ta also appeared to make it more difficult to reach a stable initial passivation, as seen in the polarization curve for CFNT11 and the extracted values in Figure 4e.This could be due to the formation of disconnected clusters of Ta oxide which may hinder the formation of a uniform Cr-rich passive layer.
A very different behavior was observed for the CrFeNiW system.To begin with, it was predicted that the percolation thresholds would be only slightly higher (14-17 at%) than for the Cr-FeNiTa system.Yet, the experimental passivation concentration was three times higher for CrFeNiW than for CrFeNiTa.It was, in fact, higher than for any of the alloys included in the study by Xie et al. [2] Based on the materials characterization of the bulk of the coatings, there were no indications that such a large increase of the threshold value should be expected for the CrFeNiW system.
CFNW25 can be used an example of a coating that failed despite having a high bulk concentration of W. The coating was studied with microscopy, and it was found that the W did not fully dissolve stayed behind in the form of an oxide instead of being removed by undercutting.The fact that an interconnected network of W oxide could be formed means that the percolation threshold was surpassed as was predicted by the percolation theory.However, this did not lead to passivation.All of the Cr and Ni, and a part of the Fe were dissolved, and the W appeared to be fully oxidized.The last piece of necessary experimental information is the state of the CFNW13 coating at the end of the polarization curve.In the microscopy image of this coating in Figure 5d it can be seen that, when the current decreased in the polarization curve, the coating was completely dissolved, as for CFN0 or CFNT11.This is consequently an example of a coating with a W concentration below the calculated percolation threshold for W. In this case, the W oxide could not connect through the oxide was thus removed by undercutting.
The three cases mentioned above are schematically depicted in Figure 12.In case A, which is valid only for the CFNW13 coating, the W concentration is below the percolation threshold and the coating fails.In case C, which is valid only for CFNW45, the W concentration is above the percolation threshold and the coating is passivated.In case B, the W concentration is above the percolation threshold, but for some reason, stable passivation can still not be obtained.This case applies to the coatings with 19-37 at% W.
The final question is then why the W oxide that enriched the surface cannot passivate the alloy below a certain concentration.Why is the resulting oxide porous and not protective?A possible explanation is that the lower W concentration in the initial oxide, which leads to an enhanced dissolution of the remaining metals at the beginning of the transpassive region (see Figure 4f), caused an evolution of roughness and porosity that made the subsequent passivation more difficult.The slow increase in roughness seen during the sweep supports this interpretation.The formation of pores and channels can lead to a self-accelerating process, caused by concentrations gradients, most likely involving changes in the local pH.This would be aggravated by the fact that the solubility of WO 3 is pH dependent. [64,65]The dissolution rate is the lowest in 0.025 m H 2 SO 4 and increases if the pH is higher or lower. [66]uch a pH dependent dissolution rate could explain the complete dissolution of the oxide seen under the O-ring.If the dissolution of Cr can be limited at the beginning of the transpassive region, the roughening and subsequent pH changes would not be as severe, allowing passivation to be reached.This appears to be possible for a sufficiently high bulk concentration of W, since the oxide then is more enriched in W and can more seamlessly take over the passivation when Cr undergoes transpassive dissolution.
The effect observed for the CrFeNiW system is related to 2D-3D crossover effect the described by Xie et al. [2] and Sotta and Long. [41]Due to the lack of enrichment of W in the oxide, the formation of a new W-rich oxide can only occur through the dissolution of many atomic layers.It is only a higher concentration of W in the oxide that can stabilize the passivation process by limiting the amount of dissolution and thus the development of roughness and porosity.
Based on the comparison of the Ta and W containing coatings, a more general principle can be discerned.It appears that the transition from the initial oxide to the enrichment of the second passivating element should be seamless and not allow extensive dissolution of Cr.Only then can irreversible damage be prevented.This means that, in order to guarantee that the coating survives at higher potentials, the second passivating element should have similar or lower nobility than Cr.

Conclusion
Amorphous, homogenous coatings in the CrFeNiTa and CrFeNiW high entropy alloy systems were synthesized through magnetron sputtering and their corrosion resistance was evaluated.The main findings are summarized below.
• The minimum concentration needed for Ta and W to passivate the alloy was estimated with percolation theory.The threshold values were expected to be similar for Ta and W. The minimum value to allow passivation was estimated to 11-14 at% Ta and 14-17 at% W. • The values needed to achieve passivation were experimentally found to be between 13-15 at% Ta and 37-45 at% W. At these concentrations, the passive region was extended to potentials above the Cr transpassive potential.• Postcorrosion characterization showed that concentration of the elements in the surface oxide changes after reaching the transpassive potential.Cr, Fe, and Ni were dissolved, whereas Ta and W were enriched in the oxide.Ta was already enriched at potentials below the transpassive potential, while the W concentration in the oxide was closer to the W bulk concentration.Furthermore, Ta was more enriched in the top layers of the oxides.This meant that more W than Ta was needed to prevent the oxidation and dissolution of Cr at the transpassive potential.• A higher degree of dissolution of Cr, Fe, and Ni is expected to lead to more roughening of the surface in the Cr transpassive potential region.It is proposed that this hinders the repassivation of the surface of the samples.For a W concentration above the predicted percolation threshold (≈14-17 at%), but below the experimentally minimum value (≈45 at%), all the Cr and Ni, and some of the Fe, was dissolved, leaving behind a nanoporous W-rich oxide.This shows that the most basic assumption based on percolation theory was fulfilled: A connected network of W existed in the alloy.However, the porous oxide was nonprotective, and the alloys were not passivated.• Based on the experimental results a new materials design approach can be identified.The corrosion resistance of 3d transition metal alloys in the Cr transpassive region can be enhanced by the inclusion of an element from the refractory metal family, if this element has similar or lower nobility than Cr.This would mean that the resulting alloy should be functional at higher potentials than stainless steel, while having a low contact resistance compared to the pure refractory metal coating.This principle can hence be used in the design of high entropy alloys for high potential environments.

Experimental Section
Thermodynamic Calculations: Thermodynamic calculations using the CALculation of PHAse Diagrams (CALPHAD) approach were performed to predict the equilibrium phases for the alloy compositions used in this study.The calculations were performed in the Thermo-calc 2020a software [21] using thermodynamic data from the TCHEA3 database, [22] which includes critical assessments of all the relevant binary subsystems and four out of seven of the ternary subsystems.
Synthesis: Coatings with varying compositions were deposited on oxidized Si substrates and 316L stainless steel foil using direct current magnetron sputtering.In this process, an ultrahigh vacuum chamber (designed by Mantis Ltd.) with a base pressure below 10 −8 torr was used.Three magnetrons in a sputter-down position were focused on the middle of the substrate holder, which was rotating at 10 rpm.No active heating was applied to the substrate.However, the substrates were heated to ≈80 °C by the near-by plasma.The working pressure or Ar was 3 mtorr and a radio frequency (RF) substrate bias of −100 V was applied during all the depositions.Prior to the deposition, the substrates were cleaned for 15 min with an RF substrate bias set to −250 V. To obtain coatings with different compositions, the spark-plasma sintered 3 in.CrFeNi target (from Plansee SE) was kept at a constant power (100 W), while the power of either a pure W or a pure Ta target (both 2 in.targets from Kurt J Lesker, 99.95% pure) was varied between 8 and 40 W. The deposition rate increased with increasing Ta or W target power, and the deposition time between 40 and 60 min.This resulted in coatings with thicknesses between 430 and 680 nm (see Table 1).In addition, coatings of pure Ta and W were deposited to be used as references.
Characterization: Grazing incidence (GI) X-ray diffraction patterns were recorded using a Bruker D8 Discover diffractometer with Cu K  radiation.The primary optics included a Göbel mirror, a 0.1 mm horizontal slit, and a 10 mm collimator to limit the spot size to the approximate sample area.The secondary side was a parallel plate collimator with 0.4°d ivergence and a LynxEye XE detector used in the 0D mode.Scans were performed with a  value of 1°and a 2 range from 20°to 100°, with a step size of 0.05°and 25 s per step.−2 scans were performed with a Bragg-Brentano geometry using a Bruker D8 Advance diffractometer with Cu K  radiation and a Lynxeye XE-T detector.The scans (2 = 20°-100°) were performed with a step size of 0.02°and 0.2 s per step.
SEM was performed to study the thicknesses and the morphologies of the coatings before and after the corrosion tests, using either a Zeiss Merlin or a Zeiss LEO 1550 instrument.Both instruments were high-resolution FEG SEMs, and the images were recorded using the in-lens detector and an acceleration voltage of 3 kV.The samples were prepared by cracking the Si substrate cleanly in the (100) crystal direction and mounting them so that the fractured surface (i.e., the cross-section of the coating) was tilted 10°relative to the beam.The compositions of the coatings were determined with energy-dispersive X-ray spectroscopy (EDS), using the Zeiss LEO 1550 instrument, by placing the samples in top-view geometry and using an acceleration voltage of 15 kV.The EDS signal was recorded using an 80 mm 2 Silicon Drift Detector and analyzed with the Aztec software.
XPS analyses were performed with a Physical Electronics Quantera II Scanning XPS Microprobe instrument.The bulk of the coating was studied by presputtering the samples for 30 min with 200 eV Ar + .After the different corrosion procedures, the surfaces of the coatings were studied with no presputtering, using analysis angles of 10°, 30°, and 80°to vary the surface sensitivity.The pass energy was 69 eV in all measurements, and the step size was 0.125 eV.During the surface measurements, neutralization with both Ar + ions and electrons was used to control the charging.Peak fitting of the acquired data was performed using WaveMetrics IGOR Pro version 6.34A.The atomic concentration of each element was calculated using the relative sensitivity factors provided by Physical Electronics.
Transmission electron microscopy (TEM) analysis was performed on a Titan Themis 200 (FEI, currently Thermofisher) equipped with a SuperX EDS detector (G1, Oxford Instruments).Top-view TEM lamellae were prepared using a Crossbeam 550 microscope from Zeiss, with a mild Ga-ion polishing of 2 kV as a final step to limit damages from the ion beam. [23]The analyses were conducted at 200 kV and the data processing was achieved using Hyperspy python package [24] and CrysTBox. [25]SAED patterns were recorded at the middle of the top-view lamella.The diameter of the selected area aperture was 0.5 μm.
Electrochemical Tests: Electrochemical tests were performed using a three-electrode cell with a Ag/AgCl (3 m NaCl) reference electrode and a Pt wire counter electrode, respectively.The samples were mounted in a Teflon cell with a 4 mL electrolyte reservoir, and connected, using Cu tape, as the working electrode.The exposed sample area was 0.2 cm 2 .All tests were performed after 2 h of initial exposure to the electrolyte, which was 0.05 m H 2 SO 4 .After the immersion, the open circuit potential (OCP) was recorded for at least 2 min.Potentiodynamic polarization curves were recorded using a start potential of −0.7 V.The scan rate was 0.2, 1, or 5 mV s −1 and the end potential was between 0.8 and 2 V. Potentiostatic (chronoamperometric) polarization was performed to study the behavior of the samples at a constant potential.After the initial immersion time, certain samples were first polarized to −1.2 V versus Ag/AgCl for 4 min to partially reduce the oxide and then polarized to 1.5 V versus Ag/AgCl for up to 1 h.Other samples were directly polarized at 1.5 V.After the corrosion tests, the samples were rinsed with deionized water and dried under N 2 gas.Selected samples were then subjected to postmortem characterization with SEM, EDS, and XPS.The electrolyte was collected, and quantification of the dissolved metal ions was performed by Eurofins Environment Testing Sweden AB, according to ISO standardized procedures for nitric acid digestion (ISO 15587-2:2002 [26] ) and inductively coupled plasma mass spectrometry (ICP-MS) of metals in water (ISO 17294-2:2016 [27] ).
Interfacial Contact Resistance: The electric interfacial contact resistance (ICR) was measured by placing a Freudenberg 23H carbon paper between two coated 316L stainless steel samples, with the coated sides facing the carbon paper.This stack was then compressed between two gold-coated copper probes.The force applied was monitored by a scale and the resistance was measured by an Agilent 34420A resistance meter.The measurements were performed according to the methodology for bipolar plates developed in a previous work. [28]To study the effect of corrosion on the ICR, the ICR measurement was repeated after chronoamperometric experiments for the first 15 min at 0.6 V versus Ag/AgCl and then 15 min at 1.5 V versus Ag/AgCl in 0.05 m H 2 SO 4 .
Statistical Analysis: Statistical analysis was performed for the EDS, ICP-MS, and ICR results.For EDS, measurements were performed on three points on the surface of each sample.The standard deviation (0.01-0.1 at%) was much lower than the expected errors from peak fitting (a few at%), which is why this information is not included in the manuscript.For ICP-MS, each measurement was performed twice on homogenized electrolyte samples from a single electrochemical test.The standard deviations (up to 1-13%) were lower than the estimated accuracy of the measurement at close to detection levels (25-45%), which is why this information is not presented.For the ICR, the analysis was performed on three pairs of samples for each composition.

Figure 2 .
Figure 2. X-ray diffraction results displaying a) the −2 diffractograms of CFN0 and CFNW13, and b) the GI-XRD-scans ( = 1°) of all the other samples.In the −2 diffractogram, peaks from the coating (fcc) and from the Si substrate are seen.The Si(400) peak intensity has been cropped to better discern the coating signal.Some crystal planes give rise to several peaks due to Cu K  and W L contributions to the primary beam.c) Shows the average inter-atomic distances between nearest neighbors, extracted from the diffraction data (exp) and calculated as the weighted average of tabulated values for each element (calc), displayed as a function of the concentration of Ta or W. The intensities of the weak broad feature (halo) were much lower than that of the crystalline peaks.In (c), the intensity has been increased and normalized to more clearly visualize the change in position.

Figure 3 .
Figure 3. Results from electron microscopy studies of selected coatings showing SEM micrographs of a) CFN0, b) CFNT25, and c) CFNW25 displaying the morphology of the fractured samples in cross-section and the SAED patterns of CFNT25 and CFNW25 recorded in TEM.All amorphous samples had a similar smooth and featureless morphology to CFNT25 and CFNW25.

Figure 4 .
Figure 4. Potentiodynamic polarization data obtained for the coatings after 2 h at the OCP.Polarization curves for a) the Ta series and b) the W series, as well as the following values extracted from the polarization curves: c) E corr , d) j corr , e) the charge from E corr to 0.8 V, and f) the charge of the peak at around 1 V, calculated from 0.8 V versus Ag/AgCl to the local minimum in the current, all as a function of the Ta or W concentration in the coatings.

Figure 5 .
Figure 5. SEM images of selected coatings recorded at the end of the polarization curve experiments.The CFNW25 coating was scanned up to 2 V versus Ag/AgCl to study the effects of the failure of the coating.
as well as 7a,d show the locations of the A, B, C, D, and E points for the different samples.The corresponding (b) and (e) figures show the amount of metal detected in the electrolyte after polarization to the different potentials.Finally, the (c) and (f) figures display the metal ion fractions in the oxide, calculated from the peak fitting results of the Cr2p, Fe2p, Ni2p, Ta4f, and W4f XPS core-levels.The peak fitting is described in more detail in the Supporting Information, with examples of the deconvolution shown in Figure S8 (Supporting Information).

Figure 6 .
Figure 6.Results from the postcorrosion analysis of the CFNT11 a-c) and CFNW25 d-f) samples and electrolyte obtained after stopping at the five indicated potentials shown in the polarization curves in (a) and (d).(b) and (e) show the ICP-MS data, while (c) and (f) show the cation fractions calculated based on the XPS measurements using three different analysis angles, and the bulk concentrations measured by EDS.

Figure 7 .
Figure 7. Results from the postcorrosion analysis of the CFNT25 (a-c) and CFNW45 (d-f) samples and electrolyte obtained after stopping at the five indicated potentials shown in the polarization curves in (a) and (d).(b) and (e) show the ICP-MS data, while (c) and (f) show the cation fractions calculated based on the XPS measurements using three different analysis angles, and the bulk concentrations measured by EDS.

Figure 8 .
Figure 8. Top-view SEM image and EDS maps of the edge of the corroded area seen on the CFNW25 sample after polarization to 2 V.

Figure 10 .
Figure 10.Results from potentiostatic experiments at 1.5 V.The samples were polarized either directly from the OCP (no reduction) or after an initial 4 min step to −1.2 V versus Ag/AgCl (reduction).(a) shows the results for the CFNT13 and CFN15 samples, whereas the (b) shows the results for the CFNW37 and CFNW45 samples.

Figure 11 .
Figure 11.Interfacial contact resistance data for CFN0, CFNT15, CFNT25, on 316L SS substrates, obtained with a compression of 150 N cm −2 before and after electrochemical treatments.The values are averages based on three measurements with the error bars representing one standard deviation.

Figure 12 .
Figure 12.Schematic illustration of the proposed corrosion mechanisms valid for the CrFeNiW coatings at different stages during the recording of the polarization curve.Case A can be applied to CFNW13, Case B to CFNW19 -CFNW37, whereas Case C can be applied to CFNW45.The images are not to scale.The thickness of the oxide and the scale of the roughness and porosity are exaggerated for clarity.