Hydrogen Permeation in X65 Steel under Cyclic Loading

This experimental work analyzes the hydrogen embrittlement mechanism in quenched and tempered low-alloyed steels. Experimental tests were performed to study hydrogen diffusion under applied cyclic loading. The permeation curves were fitted by considering literature models in order to evaluate the role of trapping—both reversible and irreversible—on the diffusion mechanism. Under loading conditions, a marked shift to the right of the permeation curves was noticed mainly at values exceeding the tensile yield stress. In the presence of a relevant plastic strain, the curve changes due to the presence of irreversible traps, which efficiently subtract diffusible atomic hydrogen. A significant reduction in the apparent diffusion coefficient and a considerable increase in the number of traps were noticed as the maximum load exceeded the yield strength. Cyclic loading at a tensile stress slightly higher than the yield strength of the material increases the hydrogen entrapment phenomena. The tensile stress causes a marked and instant reduction in the concentration of mobile hydrogen within the metal lattice from 55% of the yield strength, and it increases significantly in the plastic field.


Introduction
On offshore structures, cathodic polarization is usually applied to prevent their generalized corrosion, to reduce the overall loss of metal due to anodic dissolution, and to enhance their corrosion-fatigue limits. However, a correct cathodic protection design is necessary to avoid hydrogen embrittlement (HE) risk [1,2]. Problems can arise on the welded legs of a platform, risers, and on the free span of sea-lines, which are exposed to lateral marine currents; these induce oscillation by vortex shedding [3,4]. Corrosion-fatigue in seawater becomes evident at a frequency below 10 Hz [5,6]. Very high crack growth rates are observed below 1 Hz, at intermediate values of the stress intensity factor range (∆K) and under high stress ratio-R (i.e., high levels of stress) [7]. Cathodic protection plays a positive role in crack initiation, in the absence of propagating defects, by preventing localized attacks [8], but it can increase the fatigue crack growth by promoting atomic hydrogen formation on the surface of the metal [9,10]. Cathodic overprotection (E < −1100 mV vs. SCE) is particularly deleterious for the insurgence of HE phenomena, as reported in the most widespread design guidelines [11][12][13]. It is expected that there would be more evident research on high-strength steels, because they are generally considered more susceptible to hydrogen embrittlement [14][15][16].
Hydrogen assisted cracking has been extensively studied during the last 50 years, but the mechanism is not yet completely known. Several studies formulated a certain number of theories Other studies evaluated hydrogen diffusion under static load conditions and slow strain rate conditions [66][67][68][69].
In this paper, hydrogen diffusion under applied cyclic loading on sorbitic X65 grade steel is studied. Permeation transients are analyzed by considering literature models in order to evaluate the role of trapping-both reversible and irreversible-on the diffusion mechanism during the application of load in the elastic and plastic field. In addition, the effect of the application of different cyclic loading conditions on steady state hydrogen permeation flux is analyzed.
Despite the topic having already attracted the attention of several researchers, the paper would increase the knowledge of the effect of cyclic loading on hydrogen diffusion to give more insights into the topics related to hydrogen embrittlement of high-strength low-alloyed steel.

Experimental Section
Commercial pipeline steel (X65 grade according to API 5L standard) having a sorbitic microstructure ( Figure 1) [70] was considered during this experimentation. The microstructure was revealed after polishing with emery papers up to 1200 grit and degreased with acetone in an ultrasonic bath. Other studies evaluated hydrogen diffusion under static load conditions and slow strain rate conditions [66][67][68][69].
In this paper, hydrogen diffusion under applied cyclic loading on sorbitic X65 grade steel is studied. Permeation transients are analyzed by considering literature models in order to evaluate the role of trapping-both reversible and irreversible-on the diffusion mechanism during the application of load in the elastic and plastic field. In addition, the effect of the application of different cyclic loading conditions on steady state hydrogen permeation flux is analyzed.
Despite the topic having already attracted the attention of several researchers, the paper would increase the knowledge of the effect of cyclic loading on hydrogen diffusion to give more insights into the topics related to hydrogen embrittlement of high-strength low-alloyed steel.

Experimental Section
Commercial pipeline steel (X65 grade according to API 5L standard) having a sorbitic microstructure ( Figure 1) [70] was considered during this experimentation. The microstructure was revealed after polishing with emery papers up to 1200 grit and degreased with acetone in an ultrasonic bath. The specimens were taken from a commercial thick wall seamless pipe (outer diameter equal to 360 mm, thickness equal to 57 mm). They were obtained from the center of the pipe thickness ( Figure  2) to ensure that hydrogen permeation occurrence is always at the same depth. Hydrogen permeation tests were carried out on unloaded specimens (U, rectangular sheet, Figure 2; Specimen 1) and dogbone specimens for cyclic loading tests (C, Figure 2; Specimen 2 according to EN 10002-1 standard).
Specimens were cut by using a cutting machine for metallographic sample preparation to avoid the occurrence of residual stresses inside the material. The specimens were then milled to a uniform thickness of 1 ± 0.01 mm at the exposed area. The thickness was measured by means of Mitutoyo's High-Accuracy Digimatic Micrometer (Kawasaki, Japan). Before the tests, the specimens were grounded with emery paper and polished with diamond paste of up to 1 µm finishing. The specimens were taken from a commercial thick wall seamless pipe (outer diameter equal to 360 mm, thickness equal to 57 mm). They were obtained from the center of the pipe thickness ( Figure 2) to ensure that hydrogen permeation occurrence is always at the same depth. Hydrogen permeation tests were carried out on unloaded specimens (U, rectangular sheet, Figure 2; Specimen 1) and dog-bone specimens for cyclic loading tests (C, Figure 2; Specimen 2 according to EN 10002-1 standard). Tensile strength was measured on specimens with the same dimensions as those used in all the permeation tests under cyclic loading conditions. The tensile tests, performed on an oleo-dynamic universal testing device, showed values of the tensile yield strength (TYS) and the ultimate tensile strength (UTS) equal to 450 and 570 MPa, respectively. The different loading levels in the cyclic loading tests were then determined according to these two values. Specimens were cut by using a cutting machine for metallographic sample preparation to avoid the occurrence of residual stresses inside the material. The specimens were then milled to a uniform thickness of 1 ± 0.01 mm at the exposed area. The thickness was measured by means of Mitutoyo's High-Accuracy Digimatic Micrometer (Kawasaki, Japan). Before the tests, the specimens were grounded with emery paper and polished with diamond paste of up to 1 µm finishing.
Tensile strength was measured on specimens with the same dimensions as those used in all the permeation tests under cyclic loading conditions. The tensile tests, performed on an oleo-dynamic universal testing device, showed values of the tensile yield strength (TYS) and the ultimate tensile strength (UTS) equal to 450 and 570 MPa, respectively. The different loading levels in the cyclic loading tests were then determined according to these two values.
A Devanathan-Stachurski cell was used. The layout of the cell was modified for the tests under cyclic loading. Poly-methyl-methacrylate (PMMA) compartments and flanges opportunely designed to hold the specimen by means of flat neoprene gaskets were applied as sealers. The exposed area in both the charging side (cathodic compartment) and detection side (anodic compartment) was 1.76 cm 2 , considering the neoprene deformation after the cell assembly. The ratio between the radius and thickness of the membrane was greater than 5:1, as recommended by [61]. Both the charging and detection side solutions were thermally regulated at a constant temperature of 23 ± 0.5 • C, as the results can be significantly affected by temperature variations.
A sodium hydroxide 0.1 M aerated solution was considered in the detection side, in stagnant conditions. A potentiostatic anodic polarization of +340 mV vs. Ag/AgCl (3 M KCl) was applied by using an AMEL General Purpose Potentiostat/Galvanostat (Model 2049, Milan, Italy). The anodic surface was not Pd-coated because rupture of this electrodeposited thin Pd layer is expected under loading deformation [71], even in the elastic field. An MMO-activated titanium net was used as a counter electrode in both the compartments. Once anodic current densities lower than 0.05 µA/cm 2 were recorded-well below the limit of 0.1 µA/cm 2 suggested by [61]-hydrogen charging was activated in the charging side of the cell. The cathodic surface of the specimen was then polarized at constant current density.
An aerated borate solution (0.3 M H 3 BO 3 + 0.075 M Na 2 B 4 O 7 ) buffered at pH 8.4 [72] was inserted in the charging compartment. As stated by Nagumo [73], the borate buffer solution pH is stable during long-term testing. An immersion pump guaranteed solution recirculation in the cathodic compartment. The flow rate was fixed at values around 1 m/s to further minimize alkalization effects.
The anodic current density in the detection compartment was measured with a frequency of 1 Hz to synchronize the applied load with the resulting current signal.
The loading sinusoid was characterized by a maximum value in the range of 55-120% TYS, an amplitude equal to ±10-20% TYS, and a frequency from 1 to 10 −2 Hz. The tests were summarized as follows. • A first series of tests was performed by applying the cyclic load before the activation of the charging cathodic current, and then it was maintained simultaneously to hydrogen permeation transient in order to evaluate the effect of cyclic loading on the apparent hydrogen permeation coefficient (Table 1). These tests were also performed in the absence of a hydrogen charge to evaluate the effect of the cyclic load on the passivity current density (Table 2). • A third series of tests was performed in order to evaluate the transient of the hydrogen steady state flux and of the passive current, increasing on the specimen the mean value of the applied load from 0% to 110% TYS, maintaining constant the amplitude and the frequency (Table 3). • Some tests were carried out on a specimen subject to the cyclic loading for the first time from 0% to 55% TYS and from 55% to 90% TYS, then unloaded and loaded again from 0% to 90% TYS.
The experiments were almost twice conducted, and the results showed a good reproducibility. The data were processed with a method based on the mathematical model proposed by Grabke and Riecke [74], which takes into consideration the method implemented by Oriani [75] and integrates it by considering the presence of both reversible (shallow) traps and irreversible (deep or strong) traps. Figure 3 shows the effect of the cyclic loading on permeation flux. The model proposed by Grabke and Riecke [74,76] was applied for fitting the transient data. More detailed information on data processing is reported in [64]. Based on this model, the effect of reversible traps can be described as a shift of the time needed to reach the steady state value. The effect of irreversible traps can be mainly noticed by analyzing the slope of the curve.

Results and Discussion
The hydrogen permeation transient was expressed as a function of cyclic loading parameters. The relation between apparent diffusivity (D app ) and cyclic loading conditions is shown in Figure 4. Data are compared to the unloaded condition as a function of the ratio between the applied nominal stress (σ) and the TYS value. elastic field (55% TYS), although to a lesser extent. This becomes much more evident when exceeding the stress at yield. The sudden decrease in Dapp values can then be ascribed to the creation of a large quantity of new reversible traps, i.e., new dislocations [79].
For the considered loading conditions, plastic deformation determines a significant increase in the trap density, which reduces the diffusible hydrogen in the lattice and causes a reduction in the hydrogen flux during the transient [80].  the stress at yield. The sudden decrease in Dapp values can then be ascribed to the creation of a large quantity of new reversible traps, i.e., new dislocations [79]. For the considered loading conditions, plastic deformation determines a significant increase in the trap density, which reduces the diffusible hydrogen in the lattice and causes a reduction in the hydrogen flux during the transient [80].   D app decreases with the maximum tensile stress, with values in the range between 1.5 × 10 −11 and 6.5 × 10 −12 m 2 /s. The decrease in D app as the applied maximum stress increases can be mainly related to the growth in the dislocation density [76][77][78]. The decrease in D app can already be noticed in the elastic field (55% TYS), although to a lesser extent. This becomes much more evident when exceeding the stress at yield. The sudden decrease in D app values can then be ascribed to the creation of a large quantity of new reversible traps, i.e., new dislocations [79].
For the considered loading conditions, plastic deformation determines a significant increase in the trap density, which reduces the diffusible hydrogen in the lattice and causes a reduction in the hydrogen flux during the transient [80].
The effect of an alternate component of loading at ±120% TYS (frequency = 1 Hz) and the steady state current is shown in Figure 5. The current signal is modified similarly to the applied load. The same behavior can be detected for passive current density in the absence of hydrogen. Fast Fourier transform (FFT) and inverse fast Fourier transform (IFFT) were then applied to the sinusoidal signals to obtain the average value and amplitude of i a,∞ and i P,∞ for each value of the amplitude and frequency of the load. observed in the absence of hydrogen.
The effect of the instant variations of the maximum load under cyclic loading conditions on ia,∞ in the presence of hydrogen was evaluated with some tests performed at the lowest frequency (10 −2 Hz) and with an amplitude of the cyclic load equal to ±10% TYS (Figure 7). All the permeation transients are obtained from the same specimen, starting from the unloaded condition and then increasing the average value of loading.
As a title of the example, the curve indicated as "C_100-110" refers to a variation in the applied maximum stress from 100% to 110% TYS. The loading variation was applied consequently once ia,∞ was achieved under cyclic loading conditions at an applied maximum stress of 100% TYS.   Figure 6 reports the increase in the current response (∆i (FFT)) both for i a,∞ and i P,∞ , in function of the load amplitude (∆σ) and frequency (f). The current variation induced by loading (amplitudes: ±10% and ±20% TYS; significant frequencies: 10 −2 , 10 −1 , 1 Hz for each load amplitude) on i a,∞ in the presence of hydrogen is slightly lower than the variation induced on i P,∞ in the absence of hydrogen.  The variation in the passive current density with the cyclic loading increases with the frequency. This effect could be ascribed to the repassivation of new exposed metal surfaces due to the loading and the variation of the internal energy of the oxide film in the elastic field. The cyclic loading could slightly deplete the metal lattice of hydrogen atoms, which are trapped in shallow traps, with the result of a decrease in the permeation flux during the cyclic loading [44]. This grows with the increase in the passive current, and the result is a growth of the anodic current variation lower than that observed in the absence of hydrogen.
The effect of the instant variations of the maximum load under cyclic loading conditions on i a,∞ in the presence of hydrogen was evaluated with some tests performed at the lowest frequency (10 −2 Hz) and with an amplitude of the cyclic load equal to ±10% TYS (Figure 7). All the permeation transients are obtained from the same specimen, starting from the unloaded condition and then increasing the average value of loading.  For the loading variation from 0% to 55% TYS, no appreciable variations in the hydrogen flux can be noticed. As loading variations approached the yield limit (from 55% to 90% TYS and from 90% to 100% TYS) and exceeding that limit (from 100% to 110% TYS), an instantaneous peak in the current at the loading variation is detected, and a marked drop to values well below the initial steady state flux is seen. Then, a transient can be evidenced, and the current slowly returns to values approaching the steady state reached prior to the loading variation. As a title of the example, the curve indicated as "C_100-110" refers to a variation in the applied maximum stress from 100% to 110% TYS. The loading variation was applied consequently once i a,∞ was achieved under cyclic loading conditions at an applied maximum stress of 100% TYS.
For the loading variation from 0% to 55% TYS, no appreciable variations in the hydrogen flux can be noticed. As loading variations approached the yield limit (from 55% to 90% TYS and from 90% to 100% TYS) and exceeding that limit (from 100% to 110% TYS), an instantaneous peak in the current at the loading variation is detected, and a marked drop to values well below the initial steady state flux is seen. Then, a transient can be evidenced, and the current slowly returns to values approaching the steady state reached prior to the loading variation.
However, when the same loading variations were applied on specimens without hydrogen charging-i.e., on the i P,∞ (Figure 8)-again on the same specimen and continuously, the instantaneous peak is always present at the loading variation. This suggests that this phenomenon is due to the instant local rupture of the oxide film and not to a variation in the hydrogen flux. However, when the same loading variations were applied on specimens without hydrogen charging-i.e., on the iP,∞ (Figure 8)-again on the same specimen and continuously, the instantaneous peak is always present at the loading variation. This suggests that this phenomenon is due to the instant local rupture of the oxide film and not to a variation in the hydrogen flux. Therefore, after the rupture of the oxide film, repassivation takes place along with a temporary decrease in the permeation flux, due to the creation of new trapping sites that temporarily reduce the concentration of diffusible hydrogen in the metal lattice. This effect seems to be quite negligible for loading variations in the elastic field. It becomes even higher just approaching the yield limit, and it further increases well exceeding this value.
In order to discriminate hydrogen trapping effects, the value of ia,∞ ( Figure 9) has to be reduced to take into account the increasing value of iP,∞ as loading increases due to this activationrepassivation mechanism occurring at steel surface.  Therefore, after the rupture of the oxide film, repassivation takes place along with a temporary decrease in the permeation flux, due to the creation of new trapping sites that temporarily reduce the concentration of diffusible hydrogen in the metal lattice. This effect seems to be quite negligible for loading variations in the elastic field. It becomes even higher just approaching the yield limit, and it further increases well exceeding this value.
In order to discriminate hydrogen trapping effects, the value of i a,∞ (Figure 9) has to be reduced to take into account the increasing value of i P,∞ as loading increases due to this activation-repassivation mechanism occurring at steel surface.
However, when the same loading variations were applied on specimens without hydrogen charging-i.e., on the iP,∞ (Figure 8)-again on the same specimen and continuously, the instantaneous peak is always present at the loading variation. This suggests that this phenomenon is due to the instant local rupture of the oxide film and not to a variation in the hydrogen flux. Therefore, after the rupture of the oxide film, repassivation takes place along with a temporary decrease in the permeation flux, due to the creation of new trapping sites that temporarily reduce the concentration of diffusible hydrogen in the metal lattice. This effect seems to be quite negligible for loading variations in the elastic field. It becomes even higher just approaching the yield limit, and it further increases well exceeding this value.
In order to discriminate hydrogen trapping effects, the value of ia,∞ ( Figure 9) has to be reduced to take into account the increasing value of iP,∞ as loading increases due to this activationrepassivation mechanism occurring at steel surface.  The effect of the decrease in diffusible hydrogen concentration in the lattice is due to the generation of new traps by the plastic deformation; this phenomenon is also present in the elastic field, and it increases as the applied loading increases, as expected. Once the new generated traps are filled, the steady state current returns to the initial value.
Some tests were carried out on specimens subject to the cyclic loading for the first time, which were then unloaded and loaded again with the same cyclic load applied before (Figure 10) to support this hypothesis. As a result, no transient decrease in the permeation flux was detected in this case, probably due to the local strain-hardening of the material and no creation of new trapping sites.
Materials 2020, 13, x FOR PEER REVIEW 10 of 15 The effect of the decrease in diffusible hydrogen concentration in the lattice is due to the generation of new traps by the plastic deformation; this phenomenon is also present in the elastic field, and it increases as the applied loading increases, as expected. Once the new generated traps are filled, the steady state current returns to the initial value.
Some tests were carried out on specimens subject to the cyclic loading for the first time, which were then unloaded and loaded again with the same cyclic load applied before (Figure 10) to support this hypothesis. As a result, no transient decrease in the permeation flux was detected in this case, probably due to the local strain-hardening of the material and no creation of new trapping sites. A decrease in the hydrogen stationary flux by only applying the external load was observed in previous works of the same authors under constant load [7], according to other results also [69]. In constant strain rates conditions [7,[81][82][83], the hydrogen permeation flux decreases with the strain rate. At very high strain rates, decrease in the hydrogen permeation flux can be noticed as plastic deformation increases due to the filling of new traps, thus causing a decrease in the hydrogen concentration inside the metal lattice. At strain rates below 10 −6 s −1 , the rate of trap generation is lower A decrease in the hydrogen stationary flux by only applying the external load was observed in previous works of the same authors under constant load [7], according to other results also [69]. In constant strain rates conditions [7,[81][82][83], the hydrogen permeation flux decreases with the strain rate. At very high strain rates, decrease in the hydrogen permeation flux can be noticed as plastic deformation increases due to the filling of new traps, thus causing a decrease in the hydrogen concentration inside the metal lattice. At strain rates below 10 −6 s −1 , the rate of trap generation is lower than the saturation rate, and hence depletion cannot occur; the trapping effect becomes negligible with respect to diffusion. The hydrogen concentration in plastically strained zones was well evidenced by means of photoelectrochemical techniques [84].
The embrittlement effect evaluated by means of slow strain rate tests and crack growth rate tests under corrosion-fatigue conditions can be related to the hydrogen apparent diffusion coefficient only for steel having similar microstructures [24,25,84,85]. The hydrogen embrittlement effects were evaluated by considering the presence of brittle areas on the fracture surfaces. No evidence of HELP was observed as the stress-strain curves of the specimens with ductile fracture were always similar than those at air. The decreases in the total elongation at break and in the reduction of area were always observed once brittle initiation of the fracture occurred. In many cases, it was always noticed that the role of plastic strain was relevant as HE was always initiated at the necking. Crack propagation was never observed to occur in absence of dynamic loading, i.e., under constant loading conditions. In this work, the continuous straining in the plastic field occurs mainly during the increasing of loading in the first sinusoidal waveform. The effects of local straining are evident in the hydrogen permeation transient once the specimen is charged with hydrogen after loading. On the contrary, the load acts mainly on the stationary value of the hydrogen permeation flux, which modifies as a function of the frequency. At very low loading frequencies, it can be assumed that the loading conditions are like those under a monoaxial slow strain rate. The trapping effect becomes prevalent at high frequencies and decreases at low frequencies.

Conclusions
The results underline the role of cyclic loading on hydrogen transport mechanisms in commercial API 5L X65 steel. The following conclusions can be drawn: • The apparent diffusion coefficient (D app ) ranges between 6.2 × 10 −11 and 8.9 × 10 −11 m 2 /s, with an average value of 7.5 × 10 −11 m 2 /s. Under cyclic loading conditions, the application of a tensile stress up to yield determines a slight decrease in D app . When exceeding such a value-i.e., at 110% TYS-D app decreases to values around 1.4 × 10 −11 m 2 /s due to the creation of a large quantity of new reversible and irreversible trapping sites.

•
An applied cyclic load causes a very small decrease in the hydrogen permeation current at the steady state, mainly hindered by the increase in the passive current. This is slight as the cycle frequency decreases, and it significantly increases as the load amplitude increases.

•
Increase in the maximum load causes a temporary decrease in the permeation flux, which becomes larger and larger as the applied load increases. Once the new traps generated are saturated with diffusible hydrogen, the steady state current returns to the value prior to load increase.

•
No transient modifications in the permeation flux are detected after subsequent load variations on the specimens already subject to the same load variation (already strain-hardened), probably because of the local strain-hardening of the material and no creation of new trapping sites.