Hydrogen embrittlement of low alloys steels at three different strength levels (745 Mega Pascals [MPa], 904 MPa, and 1,166 MPa) were evaluated under cathodic polarization. Crack growth rate measurements were performed under constant stress intensity (K) conditions, as a function of applied K values as well as applied potential to characterize the behavior of the three different steels. At −1,050 mVSCE saturated calomel electrode (SCE), the threshold stress intensity (Kth) value increased from 44 MPa√m to 60 MPa√m as the yield strength decreased from 1,166 MPa to 745 MPa. The crack growth rate at 66 MPa√m and −1,050 mVSCE decreased from 3 × 10−5 mm/s to 4 × 10−8 mm/s as the yield strength decreased from 1,166 MPa to 745 MPa. For the 1,166 MPa steel at low values of K, the crack growth rate decreased by two orders of magnitude as the potential decreased from −1,000 mVSCE to −950 mVSCE. At higher values of K, the effect of potential on the crack growth rate was not as significant. The 745 MPa steel in general exhibited slow crack growth rate values (2 to 4 × 10−8 mm/s) over the range of K values and applied potentials in which it was evaluated. Water adsorption on fresh metal surfaces in the estimated crack tip chemistry was modeled using density functional theory. The variation in crack growth rate with applied potential at low and intermediate values of K correlated with the fractional coverage of water adsorption on the fresh metal surface. It is proposed that the water reduction reaction and the subsequent generation of hydrogen are the rate limiting steps in the slow subcritical crack growth rate processes for low alloy steels under the conditions evaluated. For the higher values of K, where the crack growth rate showed a weak dependence on applied potential, water reduction, and generation of hydrogen are likely not the rate limiting steps.
Low alloy steels are commonly used in subsea fastener applications. These materials are subject to cathodic protection from subsea anodes, and hydrogen embrittlement is a significant concern for low alloy steels in these applications.1 Hydrogen embrittlement susceptibility at −1,200 mVSCE was found to depend on the yield strength of the material, which also correlated with hardness in the low alloy steels.2 Early work by Townsend using a rising K method showed that there was a significant decrease in susceptibility at about Hardness Rockwell C-scale (HRC) 34, based on tests performed on specimens on which Zn coatings were applied.3 More recent work based on rising step loaded tests has evaluated the effect of hardness as well as applied potential on the hydrogen embrittlement resistance. A 60% reduction in fracture resistance compared to the in-air values was taken as a benchmark to suggest that HRC39 would help delineate susceptibility to H embrittlement.4 At potentials of −900 mVSCE and above there is no evidence of susceptibility to hydrogen embrittlement even at HRC44.4 The measured Kth for 1,000 MPa steels, in seawater increased as the potential increased from −1,050 mVSCE to −800 mVSCE.5 At −800 mVSCE there was no evidence of any cracking. The measured values of Kth for AISI 4340 (UNS G43400(1)) steel in artificial seawater at −1,000 mVSCE was about 50 MPa√m.6 The role of applied potential was also evaluated on welded steel and a similar effect was observed.7 However, in environments containing sulfate reducing bacteria, the measured Kth values were lower than in seawater environments and also showed a weak dependence on the applied potential.5 The susceptibility of AISI 4340 steel measured via slow strain rate tests also decreased sharply when the potential was increased above −800 mVSCE.6 Similar effects were observed in samples precharged in acidic environments and tested in air.8 Decreasing strain rates were also observed to produce increasing fractions of intergranular cracking along prior austenitic grain boundaries.8 While there has been a significant amount of work performed on unederstanding the susceptibility of low alloy steels in seawater under cathodic polarization, most of the work has focused on performing slow strain rate tests and in some cases Kth values have been measured. There is very little information on the crack growth rate behavior of low alloy steels under cathodic polarization conditions.
Hydrogen embrittlement of low alloy steels in hydrogen charging environments in high-pressure hydrogen has been evaluated over a range of yield strengths particularly with a focus on the very high yield strengths (>1,200 MPa). The measured Kth were found to be very sensitive to the dissolved hydrogen concentration and the yield strength. Cracking was found to be primarily intergranular, and was associated with prior austenite grain boundaries.9 It was speculated that segregation of minor impurities like P, S, and Mn along prior austenite grain boundaries made them more susceptible to hydrogen embrittlement.9-11 Most of this work was, however, focused on steels with yield strengths in the range of 1,400 MPa and higher.10-11 There is limited work that has been conducted to evaluate the performance of low alloy steels in the strength range of 700 MPa to 1,200 MPa.12 Recent work performed in high-pressure H2 on low alloy steels with strengths in the range of 700 MPa to 1,200 MPa indicates that the Kth decreases with increasing yield strength at 102 MPa pressure of H2.12 Crack growth rate of F22 (YS of 508 MPa) precharged with a fixed concentration of internal hydrogen was a function of the applied displacement rate (even though the values of Kth were independent of the displacement rate under the conditions evaluated).13 It was found that while lowering the applied displacement rate did not change the measured Kth in lower-strength materials, the measured crack growth rates decreased with decreasing K-rate.14 It was also found that Kth values measured from fixed crack mouth opening displacement tests typically gave higher values of Kth than those obtained from rising displacement tests.14 While noting that in internally charged hydrogen there is a fixed concentration of hydrogen, as opposed to environmentally testing, where crack tip charging does play a significant role. It is clear that in lower-strength steels subject to hydrogen embrittlement, the crack growth rate is a strong function of the loading mode. This is rarely taken into consideration either in the mechanistic interpretation of the data or in applying the data in design practices.
There is a paucity of research on understanding the crack growth rate behavior of low alloy steels in the strength range of 700 MPa to 1,200 MPa under cathodic charging conditions, and applied loading rate. The current paper presents crack growth rate data under static K conditions as a function of applied potential at different values of K for three steels in the range of 700 MPa to 1,200 MPa. The results are interpreted in terms of the roles of various rate controlling steps. Density functional theory (DFT) calculations are used to infer surface chemical processes that may be responsible for the dependence of crack growth rates on the applied electrochemical potential.
Crack growth measurements were performed on three different low alloy steels of the ASTM A320 (UNS G41400) L7 and (UNS G43400) L43 grades.15 The chemical compositions and mechanical properties of the three different steels evaluated are shown in Table 1. All material was supplied in bolt form with a nominal diameter of about 76.2 mm (3 in). The ID number for all specimens is also indicated. The log IDs will be used to refer to all the materials going forward in this paper.
All specimens were extracted in the L-R orientation from the bolt material, with the notch running from the outer diameter of the bolt towards the center. All tests were performed on compact tension (CT) specimens. The 1/2T-CT specimens with a nominal B = 12.5 mm and W = 25.4 mm were machined from the two higher-strength materials (2859 and 2683). The 3/4T-CT specimens with a nominal B = 19.05 mm and W = 37.5 mm were machined from the lowest-strength material (2666). All specimens were side grooved to a depth of 5% of thickness from each side (total of 10%) to help keep the crack propagation planar. The samples were fatigue precracked in-air, with the final precracking step performed at a ΔK of 16.5 MPa√m (Kminimum/Kmaximum, R = 0.1).
All tests were performed in either a plexiglass cell or an alloy C-276 (UNS N10276) cell. The capacity of the cell was about 12 L, and was filled with about 10 L of solution. The test cell was in line with a load frame. The samples were exposed to the solution on all sides. The samples were loaded into the clevis and isolated from the load train using ceramic coated loading pins. Adhesive Teflon† tape was used on the clevises to prevent any possibility of galvanic coupling between the sample and the clevis. The test temperature was maintained constant during the course of the test using a chiller that circulated a glycol water mixture through the cooling coils located in the cell. The coils carrying the cooling/heating fluid were uniformly spaced through the height of the cell to maintain a temperature of 4°C (±1°C), over the course of the test.
All tests were performed in 3.5 wt% NaCl, with the pH adjusted to 8.2 with NaOH. The test solution and cell were deaerated using high-purity nitrogen (N2), prior to transferring the solution under a nitrogen blanket into the test cell. The solution was deaerated with high-purity N2 over the course of the test. A platinized Nb mesh was used as a counter electrode, and a saturated calomel electrode (SCE) was used as a reference electrode. All tests were started at an initial potential of −1,050 mVSCE. The choice of test conditions was based on typical anticipated environments in subsea oil and gas applications.
Crack Length Measurements
Crack length measurements were performed using reversing irect urrent otential rop (DCPD) method. A pair of heat shrunk 1-mm Pt wires were used to apply the current and a pair of 0.5-mm heat shrunk Pt wires were used to measure the potential drop. The Pt wires were spot welded to the sample, and an epoxy coating was used at the spot weld. The function of the epoxy coating was to prevent galvanic coupling between the Pt wire and the sample as well as providing strain relief. A current of 4 A was applied on the sample and the current was reversed about every 1 s to minimize thermal junction effects. The absolute values of the measured potentials were averaged to obtain a single data point. Typically, anywhere from 100 to 1,000 data points were averaged to create a single datum. A reference sample that was nominally identical to the test sample was exposed to the same conditions as the test sample. The signals from the reference was used to correct for temperature fluctuations in the cell and drift in the voltage signals with time. The measured signals from the sample was normalized with the signals from the reference sample and converted to crack length based on the Johnson equation.16
All tests were performed under K control. The measured crack length was used to estimate the load required to maintain the desired K level. Load adjustments were made when the K deviated by more than 0.1%. No changes in load were made for any apparent decrease in crack length.
A single test was performed on each of the materials under K control to estimate the crack growth rate under constant K conditions. The intent of the test method used was to measure crack growth rate independent of K-rate. Crack growth rate and Kth response of low alloy steels of intermediate strength has been shown to be dependent on the loading mode.12,17 In all cases, tests were initiated under ΔK control (R = 0.5) using an asymmetrical saw tooth wave form with a 9:1 rise time to fall time. The samples were typically started by cycling at 10 s periods, and once sufficient crack growth was established, they were transitioned to 100 s periods and finally to 1,000 s periods. At 1,000 s periods, hold times of 9,000 s were introduced. After sufficient crack growth (∼50 μm) was established, hold times of 86,400 s were introduced before finally transitioning to constant K conditions. A schematic illustration of the wave form and the associated loading format associated with the cyclic and cycle + hold periods is shown in Figure 1. The purpose of the transitioning described was to generate uniform crack fronts during constant K portions of the test.
The crack growth rate measured prior to transitioning to constant K is referred to in this paper as cyclic crack growth rate (CCGR). Once sufficient crack growth (typically about 50 μm) was established under constant K conditions, changes were made to the applied potential. Typically, the value of 50 μm was about 10 times the noise of the DCPD system. Significant work in the nuclear industry has been performed to demonstrate that these type of crack extensions are stable and can be measured accurately.18-21 Typically, the measured crack length and the crack length estimated from DCPD were in good agreement (within about 10%). Typically, the potential was increased from −1,050 mVSCE to more noble values in steps of 50 mV, until the crack growth rate reached low values (∼10−8 mm/s). This period of constant K is referred to here as static crack growth rate (SCGR). At the completion of the crack growth rate measurements as a function of applied potential for a given value of K, the potential was adjusted back to −1,050 mVSCE. The change in potential to −1,050 mVSCE was also accompanied by an increase in the K value to the next level, and the commencement of cycling at 10 s periods. The cycling sequence described previously was repeated to transition to constant K at the new value of K and the sequence of varying potential to more noble values was repeated. The set of K values and potentials evaluated for each material in this test program is shown in Table 2.
A single fatigue crack growth rate measurement was performed on the intermediate strength steel 2683 where the effect of ΔK (at an R-ratio of 0.5), frequency, wave form, and applied potential were explored.
The highest value of K in the above tests for each material (66 MPa√m for 2859, 77 MPa√m for 2683, and 80 MPa√m for 2666) met the remaining ligament and thickness criteria in ASTM 1681,22 shown in Equation (1): where
W-a: remaining ligament
K: Stress Intensity Factor
σYS: Yield strength (The average of YS and UTS cannot be used in the current work since the UTS/YS < 1.3).
DFT provides the ability to compare the fundamental electronic energies between different atomic rearrangements of materials and chemical systems. By creating a slab model (i.e., two dimensionally periodic) for a metallic surface, one can then compute the potential energy difference between surface states with different configurations of surface atoms adsorbed from the environment. In prior work by Taylor,23 atomic configurations associated with the adsorption of water, hydroxide, oxide, or chloride on the bare metal Fe(110) surface were computed using the planar augmented wave, Perdew-Becke-Ernzerhof (PAW-PBE) method. Further details of the DFT calculations are provided a previous paper.23 A first-principles thermodynamic approach that included the effects of temperature, pH, electrochemical potential, and solution activities of ions (i.e., chloride concentration) was then used to connect the fundamental adsorption energies obtained from the DFT calculation to Gibbs free energies of adsorption.23
In a subsequent paper, Taylor, et al., demonstrated that the Gibbs free energies so obtained could then be applied to generate adsorption isotherms that allow the prediction of the surface coverage of environmental species on the metallic surface.24 Whereas the Langmuir model assumed in that work is potentially probably too simplistic, since it avoids interactions between adsorbates, it should serve as a reasonable first-approximation in the limits where fresh metal surface is exposed (i.e., under crack tip conditions). The adsorption model developed24 was applied in this work using the data from the Fe(110) surface to consider the surface coverage of chloride, hydroxide, oxide, and water as a function of the applied electrochemical potential, setting the pH to 11, the chloride concentration to 1.0 M, and the temperature to 300 K to mimic the expected crack tip conditions.25-27
Low-Strength Steel – 745 MPa (2666)
The SCGR tests were performed on the low-strength steel over a range of K values. The test was started at a Kmax of 66 MPa√m, and the measured crack length as a function of time is shown in Figure 2. The results indicate that the crack growth rate for the low-strength material at 66 MPa√m is about 3 × 10−8 mm/s. There is limited data on the crack growth rate of low-strength steels under cathodic charging conditions. An example time interval of the crack growth rate data is reported in Figure 2. It shows that in the range of 10−8 mm/s crack growth rate is significantly less than those typically reported using slow-rising displacement or fixed crack mouth opening displacement (CMOD) test methods.13-14 While most tests in literature were performed at room temperature, the hydrogen diffusivity at 25°C is 1.27× higher than at 4°C28 and this difference cannot explain and difference of several orders of magnitude. Assuming a √DH dependence for crack growth rate, would lead to a crack growth rate that is 1.12× lower at 4°C.
The K value was increased to 77 MPa√m, and the crack was transitioned as described in the experimental section to constant K conditions. The measured crack growth at 77 MPa√m is 2.7 × 10−8 mm/s, which is comparable to the values measured at 66 MPa√m.
The crack was then transitioned to 80 MPa√m, at −1,100 mVSCE to co]nstant K conditions, and the crack growth rate was measured as shown in Figure 3. A stable crack growth rate of 3.7 × 10−8 mm/s was measured, similar to the values at the lower K-levels evaluated. The test was then transitioned to −1,050 mVSCE, and the resulting crack growth rate was similar at about 4 × 10−8 mm/s.
A summary of the crack growth rate as a function of applied potential for 2666 is shown in Figure 4. It is clear from the limited data that is seen, over the range of conditions evaluated, that there is no significant effect of either the applied potential or the K-value. The measured crack growth rates are also low relative to what is typically quoted for hydrogen-assisted crack propagation13-14 and is in the range of 10−8 mm/s.
Intermediate-Strength Steel – 904 MPa (2683)
The fatigue crack growth rate (FCGR) behavior of 2683 was evaluated over a range of frequencies and applied potentials. The test started at a Kmax of 44 MPa√m, R = 0.5 (ΔK = 22 MPa√m) and the frequency was reduced from 0.1 Hz to ∼10 μHz with a triangular/trapezoidal wave form. The rise time to fall time was 9:1 for the triangular wave form. Hold periods of 9,000 s and subsequently 86,400 s were introduced at 1 mHz (900 s rise time/100 s fall time) to study the effect of hold time. The frequency of the wave form was defined as 1/(trise+tfall+thold).
The FCGR per unit cycle (CGRC) decreased with frequency and there was a clear evidence of a plateau in CGRC at about 1 mHz (Figure 5). The plateau CGRC value was about 20 times higher than the in-air value from BS7910.29 The K value was subsequently increased to 55 MPa√m, R = 0.6 (ΔK = 22 MPa√m) with a sine wave form. The results indicate that at the higher value of Kmax, there is no evidence of a plateau in the CGRC even at 0.1 mHz (Figure 5). The effect of wave form on the CGRC was evaluated at 55 MPa√m (ΔK = 22 MPa√m) by performing a frequency scan using a triangular wave form. The results indicate that there is no significant effect of the wave form (triangular vs. sine) on the fatigue crack growth rate (Figure 5). Recent work on effects of waveform suggests that a positive sawtooth profile resulted in CGRC’s that twice as high as those compared to a negative saw tooth.30
A repeat frequency scan was performed using the trapezoidal wave form on the same sample at a few select frequencies, and the data appear to be reproducible (Figure 5). The effect of ΔK (at Kmax = 55 MPa√m) was evaluated by performing a frequency scan over a limited range of frequencies at 15.4 MPa√m. The measured CGRC appears to have contributions from both ΔK and Kmax. This is evident at the higher values of Kmax, at −1,050 mVSCE, where there is no evidence of a plateau in the CGRC. Under these conditions, it is likely that the CGRC is largely controlled by Kmax, as evidenced by the weak dependence of ΔK. However, at a lower value of Kmax (44 MPa√m), the CGRC is primarily controlled by ΔK, with no evidence of contribution from static crack growth.
The effect of applied potential on the frequency scan behavior was evaluated at −850 mVSCE at a Kmax of 55 MPa√m, ΔK = 22 MPa√m over a range of frequencies, and the results are shown in Figure 6. The results indicate that CGRC increases with decreasing frequency similar to the observations at −1,050 mVSCE. However, unlike in the case at −1,050 mVSCE, there is clear evidence of a plateau in the CGRC at about 10 mHz. The plateau value of CGRC is about 10 times higher than the corresponding in-air value from BS7910.29
The effect of applied potential on the CGR per unit time (CGR) in FCGR tests was investigated at 0.1 mHz and a ΔK of 22 MPa√m as seen in Figure 7. The results indicate that as the potential is decreased from −1,050 mVSCE to −850 mVSCE, CGR decreases by about 10 times. The trend in CGR with applied potential is consistent with data in literature,31 for a number of steels, where CGR decreases with applied potential.31 In the current work, even at −850 mVSCE, the plateau in CGR is about 10 times higher than the in-air values (∼1.5 × 10−3 mm/cycle). A possible reason for this observation is that the tests were performed at higher Kmax of 55 MPa√m and down to low frequencies in the range of about 1 mHz to 0.1 mHz. Prior work in literature has typically been performed at frequencies down to 0.01 Hz.29
The absence of a plateau in the frequency scan suggests that there be an effect of static crack growth rate at the Kmax value of 55 MPa√m. This may be contributing in addition to the cyclic component of the crack growth. To explore the effect of possible static crack growth rate at a Kmax of 55 MPa√m, the test was transitioned to a constant K condition at −1,050 mVSCE. In the SCGR part of the test, stable static crack growth rate of 1.3 × 10−5 mm/s was obtained under these conditions as seen in Figure 8. While the potential was transitioned to −950 mVSCE, the crack growth rate appeared to drop sharply. However, it should be noted that the measurement was performed over a relatively short period of time. A more detailed exploration of the potential effect was performed on a different sample and is described subsequently.
Scanning electron microscope (SEM) micrographs of the sample tested in environment is shown below in Figure 9. The micrograph shows clear evidence of both intergranular and transgranular cracking.
The SCGR on 2683 was investigated in greater detail, by performing tests at constant K over a range of applied potentials. The results of the SCGR tests at 60 MPa√m is shown below in Figure 10. The data clearly indicates crack growth rate that is on the order of about 2 × 10−6 mm/s at −1,050 mVSCE and −1,000 mVSCE. However, at −950 mVSCE, there is a sharp drop in the crack growth rate by about 100 fold to 2 × 10−8 mm/s.
The crack was transitioned to a higher value of K (66 MPa√m) at constant K as described in the experimental section. The results indicate that the crack growth rates over all the potential ranges evaluated is higher at 66 MPa√m compared to the values at 60 MPa√m. Similar to the results at 60 MPa√m, there is a sharp decrease in the crack growth rate at −950 mVSCE to 7.6 × 10−8 mm/s. The crack was then transitioned to constant K conditions at 77 MPa√m and the results at about 1,175 h. The results indicate that the crack growth rates in the range of −1,050 mVSCE through −950 mVSCE are high and in the range of 10–5 mm/s to 7 × 10−6 mm/s. At potentials below −950 mVSCE, there is a sharp decrease in crack growth rate to 1 × 10−7 mm/s at −900 mVSCE and to 1.2 × 10−8 mm/s at −850 mVSCE.
A summary of the SCGR data as a function of K and applied potential is shown in Figure 10. Kth was identified as the value at which there was a clear evidence of stable crack growth under constant K conditions. In this case the value of Kth is in the range of about 50 MPa√m to 55 MPa√m at −1,050 mVSCE. It is clear, that with decreasing cathodic potential, the Kth (K1EAC) increases and is about 77 MPa√m in the potential range of −900 mVSCE to −850 mVSCE. The measured crack growth rates at −1,050 mVSCE are in the range of 10−6 mm/s to 10−5 mm/s. However, at lower potentials, the crack growth rates are substantially lower and are in the range of 10−7 mm/s to 10−8 mm/s. The lower crack growth rates on low alloy steels have not been reported in H charged environments.
A summary of the crack growth rate as a function of K at the various potentials is shown in Figure 11. The results indicate that there is no significant difference between the measured crack growth rates at −1,050 mVSCE and −1,000 mVSCE over a range of K-values. The crack growth rate has a K5 dependence at both these potentials. At a lower potential of −950 mVSCE, the K dependence, in the range evaluated is extremely steep. At the high value of K the crack growth rate is not very sensitive to the applied potential (in the range of −1,050 mVSCE to −950 mVSCE). There was insufficient data at −900 mVSCE and −850 mVSCE to generate K-dependent behavior.
The crack morphology observed on the fracture surface at the 66 MPa√m and at elevated potentials in the range of −1,050 mVSCE and −1,000 mVSCE is shown in Figure 12. The fracture surface exhibits evidence of intergranular cracking as indicated by the yellow arrows.
High-Strength Steel – 1,166 MPa (2859)
The SCGR test on the specimen was initiated at a K value of 44 MPa√m Figure 13. There is clear evidence of stable crack propagation in the range 5 × 10−6 mm/s at −1,050 mVSCE and −1,000 mVSCE, but a sharp decrease in the crack growth rate to 5 × 10−9 mm/s was observed at −950 mVSCE.
The K value was transitioned from 44 MPa√m to 55 MPa√m after about 400 h, as described in the Experimental Procedures section. Once stable crack propagation was established at −1,050 mVSCE, the potential was decreased from −1,050 mVSCE to −750 mVSCE in steps of 50 mVSCE. The crack growth at −1,050 mVSCE is about 1.5 × 10−5 mm/s and decreasing potential from −1,050 mVSCE to −800 mVSCE reduced the crack growth rate to about 2.5 × 10−6 mm/s, changing the potential to −750 mVSCE led to a slight increase in the crack growth rate. The measured current at −800 mVSCE fluctuated between anodic and cathodic values, consistent with the fact that this is very close to the expected corrosion potential under these conditions. At −750 mVSCE, the measured current was positive consistent with the fact that the potential was anodic.
The K value was transitioned from 55 MPa√m to 66 MPa√m at −1,050 mVSCE at about 510 h. Once stable crack propagation was established at −1,050 mVSCE, the potential was decreased from −1,050 mVSCE to −800 mVSCE in steps of 50 mVSCE. The crack growth rate at −1,050 mVSCE was 3.8 × 10−5 mm/s and decreased by about a factor of 3 to 10−5 mm/s at −850 mVSCE. Changing the potential to −800 mVSCE, resulted in a slight increase in crack growth rate to 1.2 × 10−5 mm/s. The results at 66 MPa√m, suggest that at the high value of K the effect of applied potential is not significant consistent with the observations on 2683 at 77 MPa√m.
A summary of the SCGR as a function of applied potential at various values of K is shown in Figure 14. The role of applied potential is pronounced at the lowest value of K (44 MPa√m), where the crack growth rate falls sharply at −950 mVSCE, similar behavior was observed in the intermediate strength steel though at a slightly higher value of K (55 MPa√m). While at the higher values of K, the crack growth rate is a not a strong function of applied potential. However, at the lower values of K (55 and 66 MPa√m) the crack growth rate does exhibit a sharp change with applied potential.
The effect of applied potential on the K vs. CGR is shown in Figure 15. The crack growth rate over a range of applied potential appears to have a K4 dependence.
SEM micrograph of the fracture surface is shown in Figure 16. The SEM micrographs indicate evidence of intergranular crack morphology over all ranges of applied K. This behavior is similar to what was observed for the intermediate strength (904 MPa) steel (2683).
The primary focus of the current work has been to understand the role of SCGR as a function of applied potential. No specific effort was made to understand or quantify the Kth behavior. Prior work has suggested in high-pressure H2 on low alloy steels has indicated that the measured values of Kth are a function of the loading format.12-14 In lower-strength material the rising displacement tests resulted in significantly lower values of Kth compared to constant displacement Karrest values measured.12,14 However, recent work has indicated that the measured values of Karrest are a function of the initial applied K32 suggesting that there may be significant variables that need to be understood in the constant displacement tests to make an appropriate comparison between the different test methods. The data generated in the current program was performed at constant K. In addition to avoiding the influence of applied K-rate on the measured response, they also provide representative service conditions.
While, the primary intent was to understand SCGR as a function of applied potential, the SCGR was also a strong function of YS as shown in Figure 17. There was only one common value of K (66 MPa√m) where the crack growth rate was measured for all materials. The results clearly indicate a dramatic effect of YS on the CGR behavior. While there are microstructural differences between the steels, the YS appears to play a dominant role on the SCGR response. It is most evident at −1,050 mVSCE, where the crack growth rate varies by several orders of magnitude (3 × 10−5 mm/s to 3 × 10−8 mm/s). The effect of YS on the crack growth rate is consistent with the increasing susceptibility with YS of low alloy steels in both high-pressure H29-11 and under cathodic polarization.2,4,33-34 However, what is of interest in this particular case is the low crack growth rates that are obtained in 2666 (745 MPa).
The effect of K on the crack growth rate at −1,050 mVSCE is shown in Figure 18. It is clear that the crack growth rate of 2859 (1,166 MPa) is significantly higher than the other two steels and exhibits a shallow K dependence. The intermediate strength steel 2683 (904 MPa) exhibited a stronger dependence on K in the range evaluated particularly in the range of 55 MPa√m to 66 MPa√m. There was no significant effect of K as the value was increased from 66 MPa√m to 77 MPa√m. The crack growth rate of 2666 (745 MPa) did not exhibit any significant K dependence in the range evaluated and was associated with very low crack growth rates, that are typically not measured in hydrogen embrittlement studies.14,17,33,35
Based on the above results, it is reasonable to conclude that the crack growth rates of 2859 (1,166 MPa) steel over the K ranges measured, are representative of what would be considered typical stage II crack growth rate. However, in case of 2683 (904 MPa) the crack growth rate measured appears to be spanning the range from stage I to stage II, with the values at the highest K, likely representing stage II crack growth rates. In the case of the lowest-strength material, the crack growth rates are likely associated with stage I-based crack growth rates. The lack of dependence of crack growth rate on K is likely due to a narrow range of K values evaluated for the lower-strength material. The values of crack growth rates at the higher K values for 2859 (1,166 MPa) and 2683 (904 MPa) are about 2 to 3 × 10−5 mm/s, which is lower than values for similar strength steels in high-pressure H2.14 The differences could be associated with different hydrogen charging conditions. However, the values of crack growth rate measured on 2683 (904 MPa) in the range of 55 MPa√m to 66 MPa√m, and across all K-ranges for 2666 (745 MPa) are significantly lower than what has been reported in literature for low alloys steels subject to hydrogen embrittlement. It should be noted that a bulk of the literature has focused on very high-strength steels. Recent work on low alloys steels (albiet in high-pressure H2) has reported crack growth rate values that are significantly higher (10−4 mm/s to 10−3 mm/s).10,14
A significant amount work has been performed by Wei & co-workers on high-strength steels on the effect of various parameters like temperature,36 H2 partial pressure,36-37 and water vapor content.38-39 The focus of their work was on stage II crack growth rate to identify the rate limiting steps for crack propagation. They proposed three possible rate limiting steps: transport of species to the crack tip, reaction at the crack tip surface to generate hydrogen, and diffusion of hydrogen through the fracture process zone.40-43 More recent work on a number of materials tested in a range of environments has focused on developing rising displacement tests to generate stage II crack growth rate and has suggested that the fastest possible crack growth rate is rate limited by the diffusion of hydrogen through the fracture process zone.35,44 The crack growth rate was modelled by diffusion of hydrogen through a critical distance (typically assumed to be on the order of about 1 μm).44-47 The concentration of hydrogen is influenced by the hydrostatic stresses and the binding energy of the traps in the microstructure.44-48 Recent modeling based on plastic strain gradient has been used to estimate hydrostatic stress on the order of 12 to 25 times the yield strength (σYS) over the critical distance.49 This framework has been applied to martensitic steels, nickel-based alloys to evaluate stage II crack growth rate.44-47 However, the framework does not explain a key observation namely the influence of the applied loading rate on the crack growth rate. The effect of applied K-rate on crack growth rate has been observed in a range of materials and environments. This suggests that “K” is not a unique descriptor for crack growth rate. One possible explanation of the K-rate effect may be the effect of strain rate on the dislocation transport of hydrogen.50 However, dislocation transport of hydrogen in steels is not firmly established and, as discussed below, would lead to the prediction of high CGR values.
Diffusion coefficients of hydrogen in low alloy steels can range between 10−7 cm2/s and 10−8 cm2/s depending on the concentration of hydrogen and the trap density.51-52 The value of critical distance ranges from 1 μm to 10 μm depending on the system, a lower value has been used for high-strength materials such as ultra-high-strength steels,47 while higher values has been used to rationalize the Kth of ferritic steels.13 An estimate of crack growth based on diffusion model can be described using Equation (2):35,46
where Deff,H is the diffusion co-efficient of hydrogen, Xc is the critical distance, Ccrit is the critical concentration of hydrogen for crack advance, and CHσ is the concentration of hydrogen in the fracture process zone at the crack tip amplified due to the hydrostatic stress. Using even the lowest values of diffusion coefficient (10−8 cm2/s) and the highest value of critical distance (10 μm), the estimated crack growth rates would be on the order of 10−3 mm/s to 10−4 mm/s, depending on the ratio of Ccrit/CHσ. In the limiting case in general the CHσ≫Ccrit, which would lead to crack growth rates on the order of 10−3 mm/s. Using values of CHσ ∼1.25 Ccrit predicts a crack growth rate in the range of 1.25 × 10−5 mm/s which cannot explain the values of crack growth rate in the range of 10−6 mm/s to 10−8 mm/s. It should be noted that most diffusion analysis CHσ is in the range of 7 to 100 Ccrit for high-strength steels.46 Any consideration of dislocation transport of hydrogen would further increase the predicted CGR. The measured crack growth rates, even in the plateau region, for the highest-strength material 2859 (1,166 MPa) at −1,050 mVSCE is 10−5 mm/s, which are significantly lower than the predicted value. The values of crack growth rate measured for the other steels are substantially lower in the range of 10−6 mm/s to 10−8 mm/s.
In high-strength steels precharged with high concentrations of hydrogen, it may be possible to model the fastest possible crack growth rate as being controlled by diffusion of hydrogen through the fracture process zone.33,35 However, this model does not provide a realistic and representative crack growth rate for lower-strength steels under cathodic charging conditions. It is important to point out in the development of the diffusion based models, it is pointed out that in the environments where surface reactions are the rate limiting steps, lower crack growth rates may be obtained.33,35
Furthermore, utilizing such high predicted crack growth rates one would expect considerably lower life time than seen in actual service. An important aspect that has not received attention that is highlighted in the current work is the stage I crack growth rate under constant K conditions. Typically, K vs. da/dt relationships that have been developed for both stage I and II in either high-pressure H214,17 or under cathodic polarization13,53 have been obtained under either rising K or decreasing K conditions. The measured crack growth rates under these conditions are influenced by the applied K-rate.13-14,17 There has, however, been a significant amount of work performed on understanding slow subcritical crack growth rate independent of loading rate in stainless steels in high-purity water21,54 and in steels in near-neutral and high pH environments.55-58
The observation of “slow” crack growth rate in these materials suggests that the rate limiting step associated with these types of crack growth rates is not diffusion of hydrogen through the fracture process zone. The other possible rate limiting steps associated with crack propagation are:
generation, adsorption, and subsequent absorption of hydrogen into the steel.
development of local strains/strain rate that is sufficiently high to interact with the hydrogen present to cause crack propagation.
interaction of hydrogen with the local strain to cause/sustain crack propagation.
While there was no direct evaluation of the applied strain rate on the crack growth rate in this work, the FCGR tests serve to represent the effect of a cyclic strain rate on the crack growth rate process. The effect of low cyclic loading serves to generate fresh metal surfaces similar to rising displacement tests. Apart from the differences in the loading rates in the rising displacement tests v. low cycle fatigue (0.5 MPa√m/h13,44-45 vs. 6 MPa√m/h to -10 MPa√m/h), a critical difference is resharpening the crack tip on unloading during low cycle fatigue. Not withstanding these differences, it is clear that the crack growth rate at low frequencies, and at −850 mVSCE is on the order of 10−7 mm/s. The role of hydrogen generation on fresh metal surfaces plays an important role in the fatigue crack growth rate process. This has been evaluated by characterizing the crack tip chemistry and potential under cathodic polarization.50-52 The pH at the bottom of cracks under cathodic polarization was in the range of 10 to 12.50,52-53 There is no significant potential drop at the crack tip in the potential range of about −700 mVOCP to −1,050 mVSCE.50,52-53 At potentials lower than −1,050 mVSCE there is evidence of an IR drop primarily attributed to formation of hydrogen bubbles.52 The charge passed during a cyclic loading has been related to the fatigue crack growth rate suggesting that the generation of hydrogen on fresh metal surfaces at the crack tip may be the rate limiting step for crack propagation.50 Prior work has suggested that steels in the range of 1,200 MPa exhibit a strong strain rate and potential dependence, on the susceptibility to hydrogen embrittlement.6 The results suggest that there was no significant effect of strain to failure at low applied potentials up to about −1,000 mVSCE.6 As the potential was decreased from −1,000 mVSCE to −850 mVSCE, the strain to failure increased similar to the decrease in the fatigue crack growth rate shown in Figure 7. It was suggested based on hydrogen permeation work performed under straining conditions, as well as on fresh metal surface that generation and absorption of hydrogen were critical steps with absorption of hydrogen suggested as the rate limiting step for crack propagation.6,42
At the alkaline pH’s that are expected at the crack tip, the primary cathodic reaction generating hydrogen is the reduction of water. The reduction of water on steel surfaces in alkaline conditions, has been studied and the primary path way has been attributed to either chemical recombination mechanism or rate limiting discharge followed by chemical desorption. Scully6,42 and Frankenthal54 have suggested that the rate limiting discharge followed by recombination and desorption is likely the path way for the reduction reaction
Scratch experiments performed suggest that the current density associated with hydrogen evolution on bare metal is about two orders of magnitude higher than that on obtained on oxide covered steels.55 It has been suggested that the increase in exchange current density by two orders of magnitude is responsible for the increased current density on the fresh metal surfaces.6,42 The Tafel slope measured on bare surface vs. an oxide covered surface is similar suggesting that the symmetry of the water reduction reaction on bare surfaces is similar to that on oxide covered surfaces.6,42
The exchange current density on the bare surface is likely influenced by reaction rate of water reduction as well as the concentration of adsorped water on the surface of the bare steel. Typically, the activity of water is not considered in the analysis, as the activity is assumed to be unity. However, when fresh metal surfaces are created, the adsorption of various species is a competitive process that is a function of the applied potential.24-59 Recent DFT modeing work, has suggested that at alkaline pH’s, the adsorption of water is a function of applied potential.24 The effect of pH on the surface adsorption in the cathodic potential ranges of interest is not significantly affected in the pH range of 7 to 10. The adsorbed water coverage is 100% up to potentials of about −1,050 mVSCE and starts to decrease rather sharply at −950 mVSCE.24 At potentials of about −800 mVSCE the surface coverage is associated with chloride and oxide.24 Water adsorption as a function of applied potential at pH 11 was calculated using DFT modeling, and the results are plotted with the CGR obtained at 0.1 mHz (ΔK = 22 MPa√m/R = 0.6) as a function of applied potential is shown along with the water coverage in Figure 19. The results clearly indicate that as the water coverage decreases with decreasing applied cathodic potential, the CGR also decreases. This suggests that the adsorption of water, which is essential for the subsequent reduction to generate adsorbed H followed by absorption in the steel may be a critical step in controlling the crack growth rate. It is recognized that using Fe(110) surface in the DFT modelling is an idealized representation of polycrystalline steels. In addition to crystallographic variations, impurities, dislocations, and even absorbed hydrogen could influence the exact surface adsorption values. However, the analysis provides a frame work in which to understand the surface adsorption phenomenon that is critical to hydrogen generation at the crack tip.
The above rationale that adsorption of water on fresh metal surfaces is a critical part of the crack growth rate processes was evaluated by comparing the trend in crack growth rate with adsorption of water for both 2859 (1,166 MPa) and 2683 (904 MPa) (Figure 20). It is clear from the data for the intermediate strength material (2683), at lower values of K, the crack growth rate drops sharply at −950 mVSCE. The sharp decrease in crack growth rate coincides with the decrease in adsorption of water from 100% to about 95% at the lower value of K. However, at higher values of K (77 MPa√m) the decrease in crack growth rate is less pronounced. At the higher value of K, the water coverage at which a significant decrease in crack growth rate occurs is much lower than that the lower K values. This is consistent with the fact that at higher levels of strain/strain rate applied, a lower generation/entry rate of hydrogen at crack tip is able to sustain similar crack growth rate.
The hydrogen embrittlement susceptibility of three low alloy steels strengths ranging from 700 MPa to 1,200 MPa were evaluated. The effect of applied potential and K values on the crack growth rate behavior were quantified. The primary conclusions from this work are as follows.
There is a strong effect of strength on the measured Kth values, which increase from 44 MPa√m to 66 MPa√m at −1,050 mVSCE as the yield strength decreases from 1,166 MPa to 745 MPa. Crack growth rates measured at high values of K (66 MPa√m and higher) and negative cathodic potentials were similar for the 1,166 MPa and 904 MPa steels and were in the range of 10−5 mm/s.
In the 1,166 MPa steel crack growth rate measurements at low and intermediate values of K exhibited a strong dependence on applied potential, with crack growth rates decreasing to the low 10-8 mm/s (1,000 times decrease) at −950 mVSCE. However, at high values of applied K, the effect of applied potential was not as significant. The crack growth rate of the 904 MPa exhibited a potential dependence over the K-range evaluated. The crack growth rate of the 745 MPa steel exhibited low values (10-8 mm/s) over the range of K values evaluated.
At different values of K, and applied potentials, all the steels exhibited slow crack growth rate ranging from 10−6 mm/s to 10−8 mm/s. This suggests that under these conditions and at less cathodic potentials, hydrogen generation may be the rate limiting step in the crack growth process. The measured crack growth rates as a function of applied potential appear to be correlated to water adsorption, estimated from DFT calculations on fresh metal surfaces. The water reduction reaction at the crack tip and the generation of hydrogen are proposed to be the rate limiting steps in the crack propagation processes.
UNS numbers are listed in Metals and Alloys in the Unified Numbering System, published by the Society of Automotive Engineers (SAE International) and cosponsored by ASTM International.
The authors would like to acknowledge the support of Schlumberger for funding the work. The authors would also like to thank Brandon Gerst for performing the experiments, and Glen Clark for the SEM analysis.