Fatigue Resistance of an Anodized and Hardanodized 6082 Aluminum Alloy Depending on the Coating Thickness in the High Cycle Regime

In this study, the high cycle fatigue behavior of an anodized 6082 aluminum alloy is investigated. Main focus is on the most relevant influencing factors for crack initiation and propagation under cyclic loading and damage mechanisms considering coating type, thickness, and residual stresses. The bare substrate is compared to anodized and hardanodized specimens with three coating thicknesses, for each coating type, in the range from 20 to 70 μm. Coating hardness and microstructure as well as residual stresses are analyzed. Fatigue and fracture behavior under alternating tension–compression loading is determined. Dependent on the coating thickness, the fatigue strength is reduced by 8%–50% after anodizing and by 50%–62% after hardanodizing. As the coating thickness is equal to the initial crack length from a fracture mechanical point of view, stress intensities at the crack tips are higher for thicker coatings respectively longer initial crack lengths. Therefore, propagation of fatigue‐induced cracks from the coating into the substrate is promoted for a higher coating thickness resulting in premature failure. A significant correlation between the coating thickness and tensile residual stresses induced by both coatings in the subjacent substrate is not found and residual stress influence on the overall fatigue strength is only minor.


Introduction
High-strength aluminum alloys of the age-hardenable 6xxx series are widely used as structural components in the automotive and aircraft industries.Their excellent strength to weight ratio combined with a good fatigue resistance enables a broad range of application, which is often only limited by their corrosion and wear resistance. [1]To enhance the tribological and corrosive properties for meeting the requirements of the respective application anodizing is a classical approach. [2,3]By this electrochemical surface treatment, the aluminum substrate is converted to alumina producing strong adherent coatings with cellular-arranged pores without influencing the microstructure of the substrate. [4,5]n addition to the electrolyte composition and the process parameters, e. g., the electrical regime, the anodizing temperature is a major influencing factor for the coating properties.Anodic coatings formed at room temperature are usually between 10 and 30 μm thick, depending on the application. [6]In contrast, decreasing the process temperature to À10 to maximum 5 °C and therefore inhibiting the simultaneous dissolution of the coating during the anodization process enables the formation of coating thicknesses between 25 and 100 μm. [5,6][13] The decrease in fatigue resistance is a result of premature crack initiation in the porous and brittle ceramic coating and the promoted crack propagation into the substrate due to the good adhesion of the anodic conversion coating leading to fracture. [14,15]Fatigue-induced failure is further supported by the tensile residual stresses in the anodic coating, which result from the coating formation and the growth of the cellular-arranged pores. [16,17]These tensile residual stresses are most pronounced at the pore base and at the interface to the aluminum substrate. [18,19]he detrimental effect on the fatigue strength is enhanced by an increase in coating thickness [20,21] as the number and size of irregularities in the anodic coating increase as well. [21]urther, own previous studies [22,23] show that the coating thickness can be considered as initial crack length from a fracture mechanical point of view as an initiating crack in the coating immediately extends toward the substrate stopping at the interface.For a given coating thickness a, the crack can propagate into the substrate, if the cyclic loading Δσ exceeds the critical value Δσ c at which the threshold against crack propagation ΔK th is exceeded.Therefore, the fatigue strength of an anodized component is directly influenced by the coating thickness for an equal substrate material determining the threshold against crack propagation.
Despite this knowledge, there is a lack in a systematic and comprehensive approach focusing on these factors and their interactions as well as their correlation to the fatigue strength.To the best of our knowledge only in the studies by Cirik et al., [21] Wragg et al., [24] and Lonyuk et al., [20] different thicknesses of anodic respective hardanodic coatings were investigated.However, a systematical comparison between anodic and hardanodic coatings is missing.Further, as Cirik et al. [21] and Wragg et al. [24] used sealed specimens, the comparability to non-sealed specimens is limited due to the significant influence of the sealing treatment on the high cycle fatigue behavior. [25,26]Furthermore, the influence of the coating type and the coating thickness on the failure behavior under fatigue loading needs to be addressed with regard to the preliminary damage of the coating.It is necessary to understand the relationship between these influencing factors and the fatigue resistance, as there are application fields where the corrosion and wear resistance enabled by an anodic coating are needed but the components are also undergoing cyclic loading.
Therefore, the main focus of this study is the systematic investigation of the influence of the coating thickness on the high cycle fatigue behavior with regard to damage mechanisms and residual stresses, when comparing anodic and hardanodic coatings.These two questions are addressed in particular: 1) which factors are most relevant for crack initiation and growth from the coating into the substrate under fatigue loading and 2) which influence has the coating thickness, when considered as initial crack length for crack growth.

Material
The widely used 6082 aluminum alloy in commercial purity was used for this study as substrate material (see chemical composition in Table 1 provided by the vendor Bikar Metalle GmbH, Germany).As this medium-strength, age-hardenable alloy was well suited for anodic processing, influencing factors resulting from insufficient coating formation were minimized.For the investigation, the substrate alloy was peak-aged at 170 °C for 65 h after solution annealing at 530 °C for 60 min and water-quenched to room temperature.The parameters for the peak-aging treatment were chosen based on own determined aging time-hardness-curves for the used charge of the substrate alloy.The tensile properties of the peak-aged substrate alloy listed in Table 2 were determined using three cylindrical specimens with a geometry shown in Figure 1a in accordance to DIN 50 125. [27]Quasi-static tensile testing was performed at a strain rate of 10 À3 s À1 and at room temperature using a servoelectric universal testing machine (ZwickRoell, Germany).From the peak-aged bare substrate alloy, axial fatigue specimens with a geometry shown in Figure 1b were precision-turned achieving a surface roughness of 0.38 AE 0.16 μm R a prior to the following electrochemical treatment.

Anodizing
As pretreatment prior to the electrochemical processing, the machined axial fatigue specimens were cleaned and degreased using Oxidite-C14 solution (MacDermid Enthone Industrial Solutions, United States) for 5 min at 50 °C.Etching was done in 3 vol% aqueous sodium hydroxide solution (NaOH) at 50 °C for 1 min followed by pickling in 1.1 nitric acid (HNO 3 ) for 15 s at room temperature.Specimens were rinsed in deionized water between each step.
For the subsequent electrochemical treatment, the used electrolytes, bath temperatures, and current densities are listed in Table 3.These parameters were chosen according to previous   own studies [7,26,28] showing their suitability to form functional coatings to enable a comparability.Anodizing and hardanodizing was performed in aqueous sulfuric acid due to the industrial relevance of this electrolyte.As the achievable coating thickness depended on the process temperatures, three different coating thicknesses were chosen for each investigated coating type.To enable a comparability, two thickness values should be similar as the third was not achievable using the respective other process temperature.Table 4 lists the coating thicknesses aimed for, the resulting chosen processing time, which was estimated based on own studies and the designation of the different testing conditions, to which is referred to herein after.10 specimens for each coating type and thickness were electrochemically treated.
After the electrochemical treatment, the specimens were rinsed in deionized water and dried in air.The anodized specimens were not sealed to prevent an additional influence of the sealing treatment on the mechanical properties, as it is shown in an own previous study. [26]

Residual Stress Measurement
To determine residual stresses induced by the electrochemical treatment, X-ray diffraction (XRD) technique was used.The measurements were conducted using rolled sheet material of the investigated 6082 aluminum alloy in peak-aged condition.Due to the small diameter of the fatigue specimens and the resulting pronounced surface curvature, residual stress measurement solely in the coating was not possible for this specimen geometry.The aluminum sheets (25 Â 100 Â 1.5 mm) were anodized in the same manner as described for the fatigue specimens, but the processing time was adjusted to form coating thicknesses similar to them on the fatigue specimens.Using an X-ray diffractometer D8 Advance series 2 (Bruker-AXS, Germany) and Co Kα-radiation (40 kV, 40 mA, point focus) with a 0.5 mm pinhole collimator, polycap optics, and a Lynxeye XE detector with an aperture angle of 2.1°, the sin 2 ψ method was applied.As anodic and hardanodic coatings are amorphous and therefore a measurement using XRD technique is not possible, residual stresses induced by these coatings were measured in the subjacent aluminum substrate.The measured material and the corresponding elastic components used for the residual stress evaluation by the software Leptos (Bruker AXS, Germany) are listed in Table 5.

High Cycle Fatigue Testing
The bare and all anodized conditions were tested to determine the high cycle fatigue strength in dependence of the coating type and the coating thickness.In accordance to DIN 50 100, [29] the pearl chain method and 10 specimens for each condition were used to estimate the fatigue strength.High cycle fatigue testing was performed using alternating tension-compression loading (load ratio R = À1).A resonant testing machine (RUMUL Testronic, Russenberger Prüfmaschinen AG, Switzerland) operating at a testing frequency of approximately 100 Hz for the bare and electrochemically treated aluminum specimens was used.The high cycle fatigue tests were stopped if either an endurance limit of N E = 10 7 was reached and therefore the specimen was a run-out or a drop in resonant frequency of 1 Hz indicating a crack was detected and the specimen had failed.The endurance limit of N E = 10 7 was chosen in accordance to DIN 50 100 [29] as the substrate material was an aluminum alloy.At least two tested specimens obtaining the same cycles to failure N f confirmed the stress amplitude σ a in MPa resulting in the endurance limit.For the investigation of the fracture surfaces, the fatigued specimens stopped at 1 Hz resonant frequency drop were loaded further at the same stress level the crack was initiated to force crack propagation and the last small connected part was broken by hand to get two fractured halves.

Coating Characterization and Microstructural Analysis
Coating thickness and hardness were determined for the electrochemically treated specimens using 10 specimens for each condition.For this purpose, cross and longitudinal sections were cut from the tapered length of the fatigue specimens and metallographically polished.Using an optical microscope Olympus GX51 (Olympus Deutschland GmbH, Germany), the thickness of the applied coating was determined by 40 measurements for each specimen.For the determination of the coating hardness, an instrumented hardness tester Fischerscope HM2000 XYm (Helmut Fischer GmbH þ Co. KG, Germany) and 12 measurements for each specimen were used.
To analyze the microstructure of the coating and the fatigueinduced damage a scanning electron microscope (SEM) LEO 1455VP (LEO Elektronenmikroskopie GmbH, Germany) and a quadrant backscattering detector (QBSD) were used.Prior to the investigation by SEM untested and fatigued, metallographically polished specimens of each investigated condition were vapor deposited with a thin gold layer for electric conductivity using an Emitech K550 coater (Quorum Technologies Ltd., UK).In addition, energy-dispersive spectroscopy was performed  on the metallographically polished samples to estimate the chemical composition of the oxide coatings using spot measurements.
In addition to the microstructural characterization of the electrochemically treated, metallographically polished fatigued specimen, one specimen of each investigated condition was fatigue loaded until complete fracture at approximately 10 5 cycles.The fracture surfaces after fatigue failure were analyzed using a digital microscope VHX-500 (Keyence Deutschland GmbH, Germany), which allowed for images with an overall depth of focus on the fracture surfaces by automatically stacking of single images.

Residual Stresses
The determined residual stresses for the bare and electrochemically treated aluminum sheets using XRD technique are listed in Table 6.For the amorphous anodic and hardanodic coatings, the measuring site was the subjacent aluminum substrate.The bare aluminum substrate was almost residual stress free, which was also the case for the substrate subjacent to the thinnest anodic as well hardanodic coating, although the latter one was twice as thick.For the anodic coating there is a tendency for higher tensile residual stresses with increasing coating thickness from 20 to 40 μm.However, this tendency was not noticeable for the hardanodic coating.As described, the 20 μm thick hardanodic coating did not influence the residual stress-free state of the aluminum substrate, but for the thicker hardanodic coatings with 40 and 70 μm, the induced tensile stresses in the substrate were almost identical and the highest for all coated conditions.It has to be noted that the thicker oxide coatings may have a more pronounced effect on the measured residual values, as the penetration of the y-rays is hindered, when compared to thinner coatings.

Thickness and Hardness
The average, maximum, and minimum values of the anodic and hardanodic coating thickness measured on the metallographically polished specimens are listed in Table 7.As intended, a higher processing time resulted in a higher coating thickness.
For both coating types the thickness, which was aimed for, respectively, was achieved sufficiently.In general, an increase in coating thickness is accompanied by an increased deviation of the thickness, which was noticeable for both coating types.However, when comparing the anodic and the hardanodic coating with the same thicknesses each, the thickness deviation of the hardanodic coating is more pronounced.
The hardness of the investigated coating types dependent on the thickness are listed in Table 8 showing the related average, maximum, and minimum values.For both coating types, an increase in coating thickness resulted in a distinct decrease in hardness, which is more pronounced for the hardanodic coating.However, the overall deviation in hardness was almost similar for each coating type independent of the coating thickness.

Initial Coating Appearance
The cross sections of the initial anodic and hardanodic coatings prior to fatigue loading are shown in the micrographs in Figure 2. In general, independent of the coating type, an increase in coating thickness resulted in an increased number of defects and irregularities like micropores or microvoids across the coating thickness.Further, some precipitates from the aluminum substrate are incorporated in both coating types, which is noticeable in the micrographs.From the incorporated precipitates as well as the microvoids fine short transverse cracks are originated frequently.When comparing the anodized to the hardanodized conditions with the respective coating thickness, the hardanodic  when referring to the average hardness value for the thinnest coating of the respective type.
coating exhibits a more pronounced preliminary damage with regard to the number of irregularities.Furthermore, the examination of the coating cross sections revealed a more frequent occurrence of radial cracks in the hardanodic coating, even for a thickness of 20 μm, and their number increased with an increasing coating thickness.For the thickest anodic coating with 40 μm, a few radial cracks were also present in the initial coating.
The chemical composition of both oxide coatings determined by energy-dispersive X-ray spectroscopy was 46-47 wt% O, 47-48 wt% Al, and 5-6 wt% S. No significant difference between both coating types and thicknesses was found.

High Cycle Fatigue Behavior
The results from the fatigue tests in the high cycle regime are displayed in the diagrams in Figure 3 for the uncoated and the coated aluminum substrate.The experimental determined fatigue data points were curve-fit using a logarithmic model provided by Rateick et al. [9] This empirical model allows for a high compliance of the experimentally determined data points with the calculated progression of the fatigue life as it was developed by Rateick et al. [9] for a hardanodized aluminum alloy and was successfully used in previous own studies representing the fatigue curve of coated specimens. [22,26,28] The fatigue strength, expressed by the cycles to failure N f , is estimated using the applied maximum stress amplitude σ max and the three curve-fit parameters A, B, and C, which are listed for all tested conditions in Table 9.The fatigue limit at N E = 10 7 cycles under symmetrical tension-compression loading of the uncoated and anodized conditions is listed in Table 10.
When compared to the bare aluminum substrate, anodizing and in particular hardanodizing lead to a deterioration in fatigue strength in general.For the anodized aluminum alloy, the fatigue strength is decreased by 8% for 10 μm to 50% for 40 μm coating thickness.Therefore, the thickness of the anodic coating is a major influencing factor for the resulting fatigue limit and with an increase in coating thickness the fatigue strength is decreased.This clear tendency is not that distinct for the hardanodized conditions, when comparing to the anodized ones.Although the fatigue strength is significantly decreased after hardanodizing, the actual range of deterioration is between a decrease of 50%-62%, when compared to the bare substrate.Further, interestingly, the thinnest as well as the thickest hardanodic coating with 20 and 70 μm, respectively, results both in the lowest fatigue strength of all tested conditions and exhibits a quite similarly fatigue behavior.For these both conditions, the stress amplitude σ a at the fatigue limit of N E = 10 7 cycles is 50 MPa, which is 15 MPa lower, when compared to the 40 μm hardanodic coating.However, for the thinnest hardanodic coating, the course of the fatigue curve is flatter, when compared to both thicker coatings with 40 and 70 μm.
It is further noticeable that the fatigue curves of the anodized and hardanodized conditions tend to be flatter in general when comparing to the bare substrate, which is a result of a higher deviation of the fatigue strength after the coating process in particular at lower stress amplitudes.

Coating Characterization after Fatigue Loading and Failure Behavior
The micrographs in Figure 4 exemplarily show the fatigueinduced damage in the anodic and hardanodic coatings.After fatigue loading, all investigated coating conditions exhibit periodically occurring radial cracks independent of the coating thickness (see Figure 4a).Most of these radial cracks stop at the interface to the aluminum substrate, but for the thickest anodic coating with 40 μm and the hardanodic coatings, these radial cracks have frequently propagated short ranging into the substrate (see Figure 4c,d,f ).However, for the anodic as well as the hardanodic coating, fatigue failure was caused by the longranging propagation of one or more radial cracks from the Table 10.Stress amplitude σ a in MPa at the fatigue limit of N E = 10 7 cycles under symmetrical tension-compression loading (load ratio R = À1) of the uncoated and anodized peak-aged 6082 aluminum alloy.coating toward the substrate leading to fracture (see Figure 4b,e).
In contrast, the microvoids and the fine short transverse cracks originating from these, which were already present in the initial coating condition, were not influenced by the fatigue loading and did not change in size.Radial cracks induced by fatigue loading originating from the microvoids were not observed and the present microvoids did not influence the split-like crack path of the radial cracks.
In Figure 5 and 6, the fracture surfaces after fatigue loading are shown for all tested conditions and the crack initiation sites are marked with white circles and white darts, respectively.For the bare substrate, the fatigue crack was initiated at one single point (see Figure 5a).This behavior was also observed for the anodized alloy independent of the coating thickness (see Figure 5b-d).This indicates that only one of the periodically occurring cracks in the anodic coatings was capable to propagate long-ranging toward the substrate leading to fatigue failure.In contrast, for the hardanodized conditions, multisite crack initiation is noticeable, which is in particular pronounced for the coating thicknesses of 20 and 70 μm (see Figure 5e-f ).This fracture behavior is in accordance to the frequently observed short-and long-ranging growth of radial cracks in the aluminum substrate, shown by the SEM micrographs in Figure 6.Further, these two hardanodized conditions with 20 and 70 μm coating thickness exhibited minimal fatigue strength, when comparing to all investigated conditions.and d-f ) hardanodic coating with different thickness on the peak-aged 6082 aluminum alloy after fatigue loading.Periodically fatigue-induced radial cracks are visible mostly stopping at the interface, which was more prominent for the anodized specimens, or short ranging in the aluminum substrate, which was more pronounced for the hardanodic coatings.The long-ranging propagation of these radial cracks from the coating toward the substrate caused fatigue failure independent of the coating type and respective thickness.

Discussion
The microvoids present in the initial coatings independent of the coating type occur during the early anodization process phase, due to the preferential dissolution of the aluminum-rich matrix, which is concentrated around the precipitates in the aluminum substrate leading to their decohesion. [20,30,31]With a longer anodization time and therefore resulting higher coating thickness, the number of irregularities increases. [32]This increasing number of irregularities in both coatings explains the decrease in hardness with an increase in coating thickness for the anodic as well as the hardanodic coatings.Cirik et al. [21] further observed larger irregularities with increasing coating thickness, but this was not the case in our study as the size of the flaws, which is determined by the size of the precipitates in the aluminum substrate, was the same for all coating thicknesses and types.However, they also observed a higher number of radial cracks in the thicker coatings, which is in line with our investigations.
Further, hardanodic coatings are more prone to radial cracks as the presence of them is not only a result of a thicker, more brittle coating but results from the hard anodizing process itself.The low anodizing temperature leads to thermal stresses in the coating, when the specimen is reheated to room temperature after hardanodizing at 5 °C.Due to the lower thermal expansion coefficient of the alumina coating, when compared to the aluminum substrate, the brittle coating cracks as a result. [20,33]The generally higher hardness of the hard anodic coatings is a result of their denser pore structure, when compared to the anodized specimens. [34]Kwolek et al. [35] also observed a decrease in hardness with an increase in the thickness of a hardanodic coating, but they give no further explanation.
The measured tensile residual stresses in the aluminum substrate subjacent to the coating are in good accordance to the literature. [16,17]Ide et al. [18] further showed that the tensile stresses are higher for thicker coatings.As they have investigated coating thicknesses in the nanometer range, the transferability to our study is limited, but for the anodic coatings, this tendency was also observed by us.In contrast, for the hardanodic coatings, it seems that the tensile residual stresses in the substrate reach a saturation value.It is sufficient to say that tensile residual stresses in the aluminum substrate promote the propagation of cracks from the coating into the substrate and therefore premature fatigue-induced failure.However, the influence of the tensile residual stresses seems to be only minor, when compared to the overall influence of the coating thickness on the crack formation and propagation and therefore on the fatigue strength.
In general, the fatigue strength is reduced by anodizing and hardanodizing as the brittle and porous coatings diminish the resistance against crack initiation and, further, the resistance against crack propagation is reduced due to the good adhesion of theses coatings to the aluminum substrate promoting the propagation of cracks from the coating toward the substrate. [14,15,21]As a further result, the fatigue strength curves of the anodized and hardanodized conditions are tending to be flatter when compared to the bare substrate, which was also observed in other studies. [36,37]This can be explained by the high sensitivity of these brittle ceramic coatings to stress intensities resulting in a premature formation of cracks and therefore a more pronounced deviation of the fatigue strength at one load level.This effect was also noticed by Baragetti et al. [38] for thin hard diamond-like carbon films.
Our investigation clearly shows that an increase in the thickness of anodic coatings leads to a significantly reduced fatigue strength, which can be attributed to the more pronounced preliminary damage with increasing coating thickness resulting from the anodizing process on the one hand decreasing the resistance against crack initiation.This is also in line with the findings by Wragg et al. [24] On the other hand, as the coating thickness is equivalent to the initial crack length from a fracture mechanical point of view, [22,23] the local stress intensity at the tip of a radial crack in the coating stopping at the interface to the substrate is much higher for thicker coatings respectively longer initial crack lengths for identical applied fatigue loads.For thicker coatings, this leads to the facilitation of crack propagation into the substrate resulting in premature fatigue failure and decreased fatigue strength.Additionally, as Cree et al. showed in their studies, [39,40] the crack propagation rate is generally higher for anodized specimens, when compared to the bare substrate, which contributes further to the diminished fatigue life after the electrochemical treatment.43] However, as hardanodic coatings are more brittle, when compared to anodic coatings, and exhibit a more pronounced preliminary damage in form of thermal residual stress-induced radial cracks, the number and density of these defects are very important for the resulting fatigue strength in addition to the coatings thickness.
The preliminary damage in the hardanodic coatings lead to their generally low fatigue strength.The lower dependency of the fatigue resistance from the coating thickness, when compared to anodic coatings, might be a result of the high number of process-induced cracks.The applied macroscopic stress is distributed to these, so the stress concentrations at the single crack tips are smaller due to their high number.The additional stress concentration induced by the coating thickness respective to the initial crack length is minor in comparison leading to a smaller sensitivity of the hardanodic coating to the thickness.The pronounced influence of the coating structure and in particular anodizing process-induced cracks was also detected in previous own studies when comparing different coating types. [23,28]In addition to the thermal residual stress-induced radial cracks, other defects like microvoids or micropores effect the fatigue strength of the hardanodic coatings.For thinner coatings, the defects, of which size is mostly determined by the size of the precipitates in the aluminum substrate, are larger in relation to the overall coating thickness and therefore these defects and irregularities have a more pronounced influence.This effect can explain the low fatigue strength of the hardanodic coating with 20 μm thickness, when compared to the 40 μm thick hardanodic coating.In contrast, for the 70 μm thick hardanodic coating, the long initial crack length is the primarily fatigue life limiting factor.
Regarding the fracture behavior, more pronounced preliminary damages in combination with a high coating thickness are factors, which promote multisite crack growth into the substrate as it was observed for the hardanodic coatings and in particular for the conditions exhibiting the lowest fatigue strength.The observation that in hardanodized specimens failure occurs by the growth of multiple cracks is in line with other studies. [33,36,44]In contrast, the anodized conditions show also multisite crack initiation in the coating but these cracks stop at the interface.Only at one crack tip, the stress concentration exceeded the threshold for fatigue propagation, so this dominant crack was able to growth further into the substrate leading to fatigue failure.On the fracture surface, this is expressed by one single crack front for the anodized specimens, which was also observed in other own studies. [21,22]Zhu et al. [45] demonstrated in in situ tensile and fatigue tests that a critical value for the cracking of the oxide coatings exists.They determined a critical stress of 157 MPa for a 20 μm thick anodic coating.If this critical stress is exceeded, numerous radial cracks appear in the oxide coating and a single crack propagates toward the substrate leading to fatigue failure.This damage mechanism is in line with our observations for the anodized samples.However, in our case, the critical stress value determined by Zhu et al. [45] is not reached for any tested coating thicknesses.We assume this might be an effect of the used specimen geometry and size as well as the used load ratio.In contrast to Zhu et al., [45] which tested at 0.1, we determined the fatigue strength at alternating tension-compression loading, which may lead to a more pronounced coating shattering due to the added compression loading.Further, the tested oxide coatings in their study were hydrothermally sealed, which can also have a major influence on the fatigue resistance. [26]The observed failure behavior of the anodized specimens was also shown by Vie et al. [46] They combined acoustic emission and microstructural analysis confirmed that only after few cyclic loadings numerous cracks are already initiated in the brittle anodic coating and that after 15% of the fatigue life the crack number reaches a saturation value.Due to further cyclic loading, although there are multicrack initiation sites in the coating, only one main crack propagates toward the substrate leading to fatigue failure and fracture.
Apparently this might be in contrast to studies by Shahzad et al., [14,47] which show multisite crack growth for anodized specimens.But, as they compared bare, pickled and anodized specimens, they detected a similarly fracture behavior of pickled and anodized specimens showing multisite crack initiation.Picklinginduced pits on the specimen surface acted as stress raisers and cracks initiated from them leading to a significantly diminished fatigue strength.This fracture behavior was also observed for the compared anodized specimens exhibiting a fatigue life which was only reduced slightly further.Therefore, we assume that the pickling treatment used in our study was not as critical and invasive so the single-site crack initiation in the substrate was detected for the bare and the anodized specimens.

Conclusions
In this study, the influence of the thickness of an anodic and a hardanodic coating on the fatigue resistance in the high cycle regime was investigated.Three different coating thicknesses were compared for each coating type and the hardness and the microstructure of the formed coatings were analyzed.The fatigue strength and the fracture behavior were compared to the bare aluminum substrate.Conclusions can be drawn as follows: 1) in general, the fatigue strength is significantly decreased by anodizing and in particular hardanodizing resulting from the porous and brittle nature of these conversion layers supporting crack initiation in the coating.For a coating thickness of 20 μm, the fatigue strength was reduced by on third by anodizing and was halved by hardanodizing.2) The coating thickness is the most relevant factor for the crack propagation into the substrate and therefore fatigue failure.As it is equal to the initial crack length from a fracture mechanical point of view, a higher coating thickness leads to higher local stress intensities at the crack tip.As a result, the propagation of fatigue-induced cracks in the coating toward the substrate is promoted.Consequently, for anodic coatings, there is a clear tendency that an increasing coating thickness decreases the fatigue strength further.For a 10 μm thick anodic coating, the fatigue strength was decreased by 8% and by 50% for 40 μm.Hardanodizing led to a deterioration in fatigue strength by 50%-62% for 20%-70 μm thick coatings, when compared to the bare substrate.3) In addition to an increasing coating thickness, also the resulting increasing number of irregularities and anodizing process-induced preliminary damage, which is more pronounced for hardanodic coatings, leads to an enhanced deterioration of fatigue strength.Further, higher local stress intensities at the crack tips resulting from an increased coating thickness support multisite crack growth, which was in particular observed for the hardanodic coatings.4) The tensile residual stresses in the subjacent substrate induced by the anodic and hardanodic coatings are hardly influenced by the coating thickness and their influence on the overall fatigue life is only minor in addition to supporting the propagation of crack in the coating toward the substrate in general.

Figure 1 .
Figure 1.Geometry of the specimen used for a) tensile and b) high cycle fatigue testing.Measurements are in mm.

Figure 2 .
Figure 2. Scanning electron microscope (SEM) micrographs (quadrant backscattering detector [QBSD]) of the initial a-c) anodicand d-f) hardanodic coating with different thickness on the peak-aged 6082 aluminum alloy prior to fatigue loading.An increased coating thickness resulted in a higher number of irregularities, which is more pronounced for the hardanodic coating.

Figure 3 .
Figure 3. Fatigue behavior described by the achieved cycles to failure N f at an applied stress amplitude σ a in MPa of the a) anodized and b) hardanodized peak-aged 6082 aluminum alloy in dependence of the coating thickness.Anodizing and in particular hardanodizing lead to a deterioration in fatigue strength, which is further enhanced by an increasing thickness of the anodic coating.

Figure 4 .
Figure 4. SEM micrographs (QBSD) of the a-c) anodic and d-f ) hardanodic coating with different thickness on the peak-aged 6082 aluminum alloy after fatigue loading.Periodically fatigue-induced radial cracks are visible mostly stopping at the interface, which was more prominent for the anodized specimens, or short ranging in the aluminum substrate, which was more pronounced for the hardanodic coatings.The long-ranging propagation of these radial cracks from the coating toward the substrate caused fatigue failure independent of the coating type and respective thickness.

Figure 5 .
Figure 5. Digital micrographs of the fracture surfaces of the peak-aged 6082 aluminum alloy after fatigue failure: a) bare substrate, b-d) anodized, and e-g) hardanodized.Crack initiation sites are marked with white dotted circle.The bare and the anodized substrate show independent of the coating thickness single-point crack initiation.In contrast, multisite crack initiation occurred for the hardanodized conditions.

Figure 6 .
Figure 6.Digital micrographs of the crack initiation sites on the fracture surfaces of the peak-aged 6082 aluminum alloy after fatigue failure: a) bare substrate, b-d) anodized, and e-g) hardanodized.Crack initiation sites are marked with white dart.The bare and the anodized substrate show independent of the coating thickness single-point crack initiation.In contrast, multisite crack initiation occurred for the hardanodized conditions.

Table 1 .
Chemical composition of the 6082 aluminum alloy in commercial purity used as substrate material, provided by the vendor.

Table 2 .
Tensile properties of the peak-aged 6082 aluminum alloy used as substrate material.The deviation is given in absolute values.

Table 3 .
Investigated coating types, electrolyte composition, and processing parameters.

Table 4 .
Investigated coating types and thicknesses aimed for, used processing times and the designation, which refers to the tested condition.

Table 5 .
Material and corresponding elastic components used for the determination of residual stresses by X-ray diffraction using the software Leptos (Bruker AXS, Germany).

Table 6 .
Residual stresses in the bare and anodized aluminum sheets measured by XRD.

Table 7 .
Coating thickness of the anodic and hardanodic coating on the 6082 aluminum alloy for the investigated conditions.

Table 8 .
Martens hardness of the anodic and hardanodic coating on the 6082 aluminum alloy for the investigated conditions.

Table 9 .
Curve-fit parameters for the uncoated and anodized 6082 aluminum alloy.R 2 is the coefficient of determination.