Influence of Plastic Deformation on the Hydrogen Embrittlement Susceptibility of Dual Phase Steels

The susceptibility of advanced high-strength steels (AHSS) to hydrogen embrittlement (HE) limits the broad utilization of these materials for body-in-white (BIW) components. The considerable decrease of both ductility and toughness due to local hydrogen accumulation inside of formed components may cause unpredictable time-delayed failure. In particular deep-drawn and punched AHSS components are prone to hydrogen absorption. This work investigates the influence of plastic deformation on hydrogen absorption of dual phase (DP) steels. For that purpose, tensile samples were machined out of three commercial 1.2 mm-thick DP sheets with ultimate tensile strengths of 626 MPa, 826 MPa and 1096 MPa. Samples were uniaxially pre-strained to 2 %, 5 %, 10 %, 15 % and 20 %. After pre-straining the samples were electrochemically charged with hydrogen, and the actual hydrogen contents were determined using a thermal desorption analyser (TDA). Before and after charging, the hardness of the samples was measured and the uniaxial quasi-static tensile properties were determined. In order to quantify the influence of plastic deformation on HE, slow strain rate tests (SSRT) were performed. The results of the tests were correlated with the fraction of martensite determined for each of the three steels.

and proposed a new type of hydrogen embrittlement mechanism, namely the hydrogen-enhanced strain-induced vacancy (HESIV) mechanism. It assumes that hydrogen fosters and stabilizes the formation of excess vacancies in the crystal structure and that the accumulation to vacancy clusters causes brittle macroscopic appearance, although significant plastic straining proceeds fracture on the macroscopic scale. DP steels are widely used in the automotive industry and fundamentals of the HE susceptibility of AHSS are well studied [12][13][14][15]. However, since researchers have mostly been focusing on straining pure ferritic iron, lack of knowledge exists with respect to the influence of plastic deformation on HE of industrial relevant DP steels with high strength and reduced ductility. In comparison to ferritic iron, the composite-like microstructure of DP steels consists of both soft ferrite and hard martensite, which can easily be characterized using scanning electron microscopy (SEM) or light optical microscopy (LOM). Therefore, the present work aims for investigating the role of deformation on the hydrogen solubility of three industrial DP steels. For that purpose, the mechanical properties were characterized before and after hydrogen charging. The charging time until saturation was determined based on pre-recorded absorption curves. The HE susceptibility was measured by slow strain rate testing (SSRT) of hydrogen-charged samples. Comprehensive thermal desorption analysis (TDA) was performed to determine the hydrogen absorption capacity depending on the microstructure, deformation and hydrogen content. Thermal desorption spectroscopy (TDS) and Kissinger's theory supported the categorization of the trapping sites in the microstructures with respect to binding energy and trap density.

Materials
Microalloyed DP steels DP600, DP800 and DP1000 with carbon contents < 0.15 wt.%, < 0.18 wt.% and < 0.20 wt.%, respectively, were investigated. Details about the chemical compositions are given elsewhere [16]. The as-delivered sheets with thickness of = 1.2 mm were galvanized. Microstructure characterization was performed by etching and LOM as published in a previous work [15]. The martensite area fractions M are summarized in Table 1.

Electrochemical charging
Electrochemical hydrogen charging was performed by immersing samples in H2SO4 electrolyte containing 1 g/l Thiourea (TU, CH4N2S). Before hydrogen charging, the samples were pickled for 1-2 min in 20 vol% HCl containing 5 g/l Hexamethylenetetramine (HMTA, C6H12N4) to remove zinc coatings and oxide layers from the surfaces. HMTA was added to the solution to prevent corrosion of the samples and to avoid hydrogen uptake during pickling [16]. A cylindrical platinum counter electrode was used for electrochemical charging. The mesh electrode had a diameter of 40 mm and height of 50 mm. The current density = 0.1 A/dm² (1 mA/cm²) was constant. Hydrogen charging was performed for 2 hours in order to reach homogeneous hydrogen distribution within the samples. After charging, the samples were rinsed with ethanol and dried with compressed air. The laboratory temperature of 22 ± 3 °C and the electrolyte temperature were monitored during each charging sequence.

Quasi-static tensile and hardness testing
Quasi-static uniaxial tensile tests according to standard DIN EN ISO 6892-1 were conducted using a Zwick/Roell Z100 uniaxial testing machine for characterizing the basic mechanical properties of the as-delivered DP steels. Smooth samples were machined according to DIN 50125 -H (gauge length 0 = 80 mm, gauge width = 20 mm and sample thickness = 1.2 mm). The tensile directions of the samples were 0° (RD), 45° (DD) and 90° (TD) to the rolling direction. Before and after hydrogen charging, Vickers hardness HV5 was determined at the surface of the samples using an Emco-Test DuraScan G5 hardness tester. The Vickers hardness of each sample was determined by averaging the hardness values of 120 indentations. Before charging, samples used for hardness measurements were ground and polished, which reduced the thickness of the samples to = 1.1 mm. Charging the as-

2078
Achievements and Trends in Material Forming polished samples was performed as described in section "Electrochemical charging". The handling time of the samples between end of charging and start of the hardness measurements was less than 5 min.

Slow strain rate testing
Slow strain rate testing (SSRT) was performed to characterize the ductility and the strength after hydrogen charging. The tests were performed using a Zwick/Roell Z100 uniaxial tensile machine with the crosshead speed of 0.5 mm/min (~10 -4 s -1 ). Smooth tensile samples were machined according to DIN 50125 -H with the gauge length of 0 = 50 mm, the gauge width of = 12.5 mm and sample thickness = 1.2 mm. The sample axis was parallel to RD. Hydrogen charging was performed individually for each sample, as described in section "Electrochemical charging". The handling time of the samples between end of charging and start of SSRT was less than 2 min. SSRT was performed under atmospheric conditions and without additional zinc coating of the samples. For evaluating the influence of the strength and the microstructure on the HE susceptibility of the industrial DP steels, the HE index, %, was calculated as where 50 and 50,H are the fracture strains measured without charging and after hydrogen charging, respectively. With increasing HE susceptibility % increases towards unity.

Thermal desorption analysis and spectroscopy
For measuring the hydrogen content H , thermal desorption analysis (TDA) was performed using a Brucker G8 Galileo analyser. The machine is equipped with a thermal conductivity sensor (TCS) and an infrared furnace (IF). The helium gas-calibrated TCS has an accuracy of around 0.1 wppm/g sample. Before charging, the weight of each sample was measured using a high-precision scale. The charging procedure is described in the previous section "Electrochemical charging". The handling time of the samples between end of charging and start of TDA was less than 2 min. The hydrogen content was measured by heating the charged samples in the IF from room temperature (RT) to 600 °C linearly with a constant heating rate of 1 K/s. The hydrogen content H,BM of the as-delivered base material (BM) was determined using rectangular samples with length of 100 mm in RD, width of 15 mm in TD and thickness of 1.2 mm. The rectangular samples were wire-cut. For studying the influence of plastic deformation on the hydrogen content H,PS , tensile samples as used for SSRT were pre-strained as illustrated Figure 1. The increase of the absorption capacity due to pre-straining was evaluated by calculating the difference in the hydrogen content ∆ H as In addition, microstructural trapping sites were characterized by thermal desorption spectroscopy (TDS). TDS spectra were recorded from RT to 900 °C using different heating rates , namely 0.3 K/s (18 K/min), 0.6 K/s (36 K/min) and 1.0 K/s (60 K/min). For measuring the sample temperature accurately thermocouple of type K was placed next to the sheet samples in the IF. The gas flow between the sample position in the IF to the TCS lasted 16 s and was considered in the evaluation. As demonstrated by Drexler et al. [17], the evaluation procedure according to Kissinger's theory can be applied for single or well separated TDS peaks. The activation energy D of the chemical diffusion coefficient [17] was determined by where is the universal gas constant and max is the temperature at the peak maxima, which was determined in a straight forward manner as function of the applied heating rates . The activation energy according to Kissinger' theory was often misinterpreted as the thermal energy, which is necessary for hydrogen desorption from trapping sites. Due to the more generalized thermodynamic concept of Kissinger's theory [18], D is the activation energy for bulk hydrogen diffusion, if hydrogen diffuses from the interior of the sample to the surface [19,20]. Figure 2: a) Engineering stress-strain curves measured in different directions. b) Yield strength and ultimate tensile strength as function of martensite area fraction [2,12,[21][22][23]. c) Ultimate tensile strain and fracture strain as function of martensite area fraction [2,13,21,22,24,25]. d) Measured Vickers hardness before and after charging. Figure 2a shows the engineering stress-strain curve measured in each of the three directions RD, DD and TD. The ultimate tensile strength was comparatively low in DD, while the ultimate tensile strengths in RD and TD were higher and almost equal. The martensite area fractions M and the corresponding measured mechanical properties in RD are summarized in Table 1. The martensite area fractions were 25 ± 5 % for DP600, 44 ± 5 % for DP800 and 59 ± 5 % for DP1000. Figure 2b shows the linear increase of both the yield strength and the ultimate tensile strength with increasing M . Linear fitting of the measured yield strength p0.2 and of the ultimate tensile strength m gives

Achievements and Trends in Material Forming
The 2 -values, which quantify the quality of the fits, are 0.962 for Eq. (4) and 0.999 for Eq. (5). Literature data added to Figure 2b are in good agreement with the measured trend of the tensile properties. Differences in strength of around 200 MPa between literature sources arise from different alloying concepts and processing routes of the steels. The m / p0.2 ratios decrease just slightly from 1.8 for DP600 to 1.6 for DP800 to 1.5 for DP1000 and are independent of the investigated directions. The ratio corresponds with the strain hardening coefficient . With increasing strain hardening the resistance against plastic deformation increases and causes a more homogeneous distribution of plastic deformation within components. The macroscopic effect of localized plastic deformation and its role on the inhomogeneous hydrogen distribution in notched samples was recently demonstrated for AHSS in the work of Drexler et al. [26]. Figure 2c presents the measured ultimate tensile strains g and fracture strains l , which decrease with increasing martensite area fraction M . Linear fitting of g and l gives The 2 -values of the fitted linear curves are 0.978 and 0.999 for Eq. (6) and Eq. (7), respectively. g is always smaller than l and the differences are 8.7 % for DP600, 4.7 % for DP800 and 3.3 % for DP1000. The difference in strain is a measure for necking in the tensile test before fracture. With only small or without any necking brittle fracture would occur. Literature data added to Figure 2c are less consistent with the measured trend of the strains. Although, the correlation of the mechanical properties of DP steels with the martensite area/volume fraction is well known [15], such steels contain also significant amounts of microalloying elements. The steels used in the present study were alloyed with titanium and niobium to nucleate carbides in the matrix during tempering [27][28][29]. Those carbides contribute to the total strength due to precipitation strengthening [30,31]. They are also known as beneficial for the HE resistance of AHSS [19,20,[32][33][34]. For a comprehensive investigation of the influence of hydrogen on the mechanical properties of DP steels, Vickers hardness (HV5) measurements were conducted on as-polished surfaces before and after charging. As shown in Figure 2d, the average Vickers hardness before hydrogen charging was 192 HV5 for DP600, 244 HV5 for DP800 and 325 HV5 for DP1000. Hydrogen charging caused an increase of the Vickers hardness by about 6 HV5. Although this effect is small, the hardness increase seems to be independent of the microstructure. A similar observation was done by Depover et al. [35] applying nanoindentation and hydrogen plasma charging. To ensure that the increase in hardness attributes to hydrogen in the microstructure and not to the formation of oxide layers, further hardness measurements were performed after aging for 24 hours at room temperature (RT). During this time hydrogen desorbed from the surfaces of the samples, as measured by Drexler et al. [15]. 24 hours after charging the hardness decreased again to the initial values of the as-polished samples. Hence, the increase in hardness was reversible and it can be attributed to the hydrogen content of the microstructure. The role of hydrogen on the strength of ferritic and martensitic steels is a long-lasting discussion and for a comprehensive overview the authors reference to the book of Nagumo [11]. Macroscopic hardening, as observed in the present work, could be explained by a strong interaction of interstitial hydrogen atoms with dislocations. Most likely the strain fields around hydrogen atoms and around dislocations attract each other [36,37], which cause the formation of Cottrell clouds being dragged by moving dislocations and increase the critical resolved shear stress. For evaluating the role of the material strength on the HE susceptibility, smooth tensile samples were charged with hydrogen. In contrast to notched sample geometries, smooth samples have homogenous stress distribution, since they do not face the problem of hydrogen accumulation by maxima in the hydrostatic stress fields and by localized plastic deformation at the notch [2,3,26]. For guaranteeing also homogeneous hydrogen distribution after charging, the minimum time to reach saturation was determined as shown in Figure 3a. For better comparison of the charging time dependency, the absorption curves were normalized with the stationary hydrogen contents of 2.07 wppm for DP600, 2.20 wppm for DP800 and 2.67 wppm for DP1000, as given in Table 1. Due to the high bulk diffusivity and the low sheet thickness of 1.2 mm, the hydrogen content increased rapidly during charging and the stationary value was reached after the charging time of around 75 min. Thus, a conservative hydrogen charging time of 120 min was applied for all subsequent samples. As shown in Figure 3a, DP600 reached its hydrogen saturation faster than DP800 and DP1000. The kinetics of hydrogen uptake depend on the hydrogen absorption rate at the surface and on the hydrogen trapping capacity of the microstructure. In general, hydrogen trapping causes an increase of the hydrogen solubility and a decrease of the bulk diffusivity [5,6]. The bulk diffusivity has to be distinguished from the tracer diffusivity of hydrogen in ferritic iron [5,38]. For DP steels it is in the range between 10 -5 mm²/s and 10 -2 mm²/s depending on the hydrogen content [12,22,39,40]. Figure 3b shows the recorded stress-strain curves of the SSRT. While the bold lines were determined without hydrogen charging, the thin lines represent the SSRT after hydrogen charging. The SSRT were repeated threetimes for each material and the reproducibility was good. Due to the high hydrogen contents of more than 2 wppm (Table 1), the fracture strain was significantly reduced for each of the three DP steels,

Key Engineering Materials Vol. 926
while the ultimate tensile strength was less affected. As indicated by the measured increase in hardness, also an increase in strength can be seen in Figure 3b. Nevertheless, further measurements are necessary for improving the statistical certainty and for quantifying the change in yield strength and strain hardening by hydrogen.
In the present work the HE susceptibility of DP steels was characterized based on the fracture strain, since the fracture strain is reported in literature as the most HE-sensitive quantity [21,22,25]. As summarized in Figure 3c, due to HE the fracture strain decreased from 30.3 % to 12.8 % for DP600, from 21.3 % to 8.4 % for DP800 and from 15.3 % to 1.1 % for DP1000. However, the fracture strain seems to be less microstructural sensitive (as shown in Figure 2c) and depends on the sample geometry and strain rate. This makes a direct comparison with literature data difficult. The applied crosshead displacement speed of 0.5 mm/min, which corresponds to an average strain rate of 10 -4 s -1 , was relatively high compared to measurements on DP steels reported in literature [21,23]. Only Koyama et al. [24] used even higher strain rates of 10 -3 s -1 for evaluating the HE susceptibility. The strain rate dependency of the HE susceptibility is well known. The characterization of a true threshold as function of the hydrogen content would require either the use of smaller strain rate (e.g., about 10 -7 s -1 ) or the application of constant load tests (CLT) [2] instead of SSRT. Drexler et al. [26] proved that the strain rate limit for bulk diffusion-controlled SSRT is theoretically about 10 -7 s -1 . Lower strain rates cause a macroscopically stationary hydrogen distribution throughout the SSRT. However, desorption during SSRT with small strain rates affects the hydrogen distribution in the samples. Moreover, these tests are generally time-consuming and thus expensive. Zinc coatings, which are applied after charging, can prevent hydrogen losses during testing, but also cause a decrease of the hydrogen content by desorption during sample handling between charging and galvanization. Furthermore, reliable theories to model the effect of zinc coatings on the hydrogen distribution within the samples and on the measured hydrogen susceptibility are still missing in literature to the authors knowledge [15]. Anyway, the effect of HE was very pronounced in the present investigations using higher strain rates for SSRT, which is partially related to minimum hydrogen losses during sample handling and testing. Therefore, the evaluation of the HE index, as demonstrated in Figure 4a, can be regarded as a qualitative measure for the susceptibility, but cannot be interpreted as a true threshold. According to Eq. (1) the HE index increases with increasing M and thus with increasing m . Although the correlation between l and M was linear as shown in Figure 2c, the increase of the HE index was non-linear. In other words, the susceptibility was more pronounced at higher strength, as shown in Figure 4b. Figure 4: Hydrogen embrittlement indices of dual phase steels as function of a) martensite area fraction and b) ultimate tensile strength [2,13,23,24,41].

a) b)
Most scholars evaluated the HE susceptibility of DP steels by using Eq. (1). Only Liu et al. [41] referenced the HE index to the reduction in fracture area A instead of the fracture strain l . Anyway, a comparison is valid, because Depover et al. [21] demonstrated that both fracture strain and fracture area give similar experimental trends. Literature data were added to Figure 4a and 4b. Obviously, HE of DP steels can occur independently of the ultimate tensile strength, but materials with m > 1000

Achievements and Trends in Material Forming
MPa seem to be much more susceptible, as the HE indices are larger than 50% and as they increase more with further increase of m . The HE indices were high for the investigated hydrogen contents of around 2 wppm, but a correlation with the martensite area fraction in the microstructure could not be clearly found. For better interpretation, the applied methods for evaluating the HE susceptibility need to be better understood and improved to be more sensitive to microstructural changes rather than sensitive to changing testing parameters. The interaction of plasticity and hydrogen content during SSRT is one of the main issues, which is further investigated in the following sections. The large scatter of the measured HE indices could be related to differences in charging and testing parameters as well as in the microstructural resistance to HE. Furthermore, different charging procedures were applied. For example, Wang et al. [23] and Liu et al. [41] equipped their tensile testing machines with electrochemical cells to perform in-situ charging rather than charging before SSRT. Since SSRT combined with in-situ charging requires additional time for hydrogen uptake and bulk diffusion, they applied lower strain rates of 10 -6 s -1 and of 10 -8 s -1 , respectively.

Measurement of Hydrogen Absorption Capacity
a) b) Figure 5: a) Hydrogen content of the non-strained base material (BM), and after straining without hydrogen (BM+Strain) and after straining with hydrogen: (BM+H+Strain). DP600 was strained to 10%, DP800 was strained to 5% and DP1000 was strained to 1%. b) Increase in hydrogen absorption capacity as function of plastic strain.. Figure 5a, the hydrogen content H,BM of the base material (BM) increases almost linearly with m . Drexler et al. [15] assumed that hard martensite grains absorb almost three-times more hydrogen than soft ferrite grains and that the rule of mixture applies. Straining of the BM before charging (BM + Strain) increased the hydrogen content. DP600 with largest l was strained to 10 %, while DP800 and DP1000 with lower l were strained to 5 % and to 1 %, respectively. The hydrogen content H,PS was correlated with the plastic strain. H,PS increased from 2.07 to 3.48 wppm for DP600, from 2.20 to 3.05 wppm for DP800, and from 2.67 to 2.75 wppm for DP1000. Straining of charged samples (BM + H + Strain) had an even more pronounced effect: H,PS increased from 2.07 to 9.09 wppm for DP600, from 2.20 to 4.41 wppm for DP800, and from 2.67 to 3.02 wppm for DP1000. Between straining and hydrogen charging, the samples were aged at RT for 24 hours. Therefore, hydrogen was able to desorb from the surface before an additional charging sequence was applied. TDA measurements performed after aging at RT for 24 hours revealed a remaining hydrogen content of < 0.1 wppm. Thus, the additional charging sequence was not the reason for the excessive hydrogen absorption. As shown in Figure 5b, the hydrogen absorption capacity ∆ H depends linearly on the plastic strain pl . The slopes of the fitted lines increase from 0.14 to 0.63 wppm/% when hydrogen is present during straining. This increase in the hydrogen absorption capacity associated with plastic straining cannot be explained simply by dislocation multiplication. Strain induces most likely additional point defects, when hydrogen is present. The results confirm the interaction of hydrogen content and plastic straining for industrial DP steels. Straining of AHSS with m > 1000 MPa makes investigations difficult. It is shown that studying DP steels with different m allows extrapolation to higher plastic strains. The interaction of hydrogen content and plastic straining has several consequences with respect to the HE susceptibility:

As shown in
• The increase of the hydrogen content indicates an increase of the defect density in the microstructure that traps hydrogen. • The difference between in-situ and ex-situ charging to evaluate the HE susceptibility is more complex and cannot be described only by the difference in bulk diffusion. Furthermore, steady hydrogen absorption at the surface while samples are plastically strained most likely cause inhomogeneous distribution of point defects and thus hydrogen distribution inside the samples. The samples were uniaxially strained in RD, which caused a linear increase of the hydrogen content with plastic strain (see Figure 5b). Plastic straining introduces additional dislocations most likely in soft ferrite grains, which can trap large amounts of hydrogen. Hard martensite grains contain normally high dislocation densities close to the theoretical limit of 10 15 m/m³. Figure 6a shows the change of recorded TDS spectra of DP1000 when samples were strained to 10 % before charging. The TDS peak maxima shift with straining to higher temperatures and also the areas under the curves increase. The latter correlates with the hydrogen content and the applied heating rates ranging from 0.3 to 1.0 K/s. An evaluation of the TDS peak maxima according to Kissinger's theory is shown Figure 6b, which is also known as Choo-Lee plot. The evaluated activation energy of D ≈ 28 kJ/mol is virtually independent of the plastic strain. This value is very similar to the activation energies frequently quoted by Choo and Lee [42], who were first measuring the effect of dislocations on the activation energy for hydrogen bulk diffusion in pure iron. Drexler et al. [5,6] confirmed their results by evaluating permeation curves of cold formed iron samples. The increase in trap density due to plastic straining is assumed to shift the linear curves in the Choo-Lee plot to lower −1 -values. In other words, the energy gain of hydrogen by trapping at defects depends on the type and not on the density of the defects.

2086
Achievements and Trends in Material Forming Figure 7: TDS spectra of DP600 as function of strain.
The interaction of hydrogen with dislocations is crucial to fully understand the HE mechanisms also on macroscopic scale. The effect of straining of hydrogen-charged samples was even more pronounced for hydrogen absorption, which cannot be explained only by dislocation multiplication. In addition, point defect formation during dislocation motion must be considered. Figure 7 shows preliminary TDS spectra of strained and hydrogen-charged (BM + H) samples compared to TDS spectra recorded for unstrained (BM) samples. Two clearly separated peaks occur in the (BM + H) TDS spectrum, which were confirmed by three independent measurements. While straining of uncharged samples to 10 % increased the first TDS peak as outlined by the multiplication of dislocations, straining of hydrogen-charged samples lead to a second TDS peak at 450 °C. The first peak moderately increased and was shifted to higher temperatures. The second peak indicates that additional point defects such as vacancies [43,44] may nucleate in the microstructure during straining if hydrogen is present. Further measurements will be conducted to investigate the nature of the second peak and to correlate the results of SSRT with recorded TDS spectra.

Summary
The present work investigates the role of hydrogen and plastic deformation on the hydrogen embrittlement (HE) susceptibility. For that purpose, three industrial DP steels with thickness of 1.2 mm and ultimate tensile strengths of 626 MPa, 826 MPa and 1096 MPa were investigated by tensile testing, Vickers hardness measurements and slow strain rate tests (SSRT). Comprehensive TDA and TDS measurements were performed and the results were correlated with the HE susceptibility. The following conclusions can be drawn according to the results of these experiments: • Classification of the HE susceptibility of DP steels revealed a strong correlation with the ultimate tensile strength. However, the correlation with the martensite area fraction is less pronounced. • Hydrogen causes a reversible increase in hardness, which is explained by a strong interaction of hydrogen with dislocations increasing the critical resolved shear strength. • Plastic strain increases the hydrogen absorption capacity of DP steels. The increase is influenced by the DP microstructure, the plastic strain level and the hydrogen content being present during straining. • The activation energy according to Kissinger's theory was 28 kJ/mol and stays unaffected by the level of plastic strain. • An additional high temperature peak was measured in the recorded TDS spectra. Most likely this peak is caused by the strong interaction between hydrogen and moving dislocations intensifying the dislocation multiplication during plastic straining and enhances the straininduces nucleation of further point defects. In future investigations will be conducted to reveal the nature of the second TDS peak and to derive new damage models for evaluating the HE risks of cold formed DP steel components.