Regular Article
Fatigue Crack Propagation in Pearlitic Steel under Pressurized Gaseous Hydrogen: Influences of Microstructure Size and Strength Level
2023 Volume 63 Issue 7 Pages 1251-1262
Details
2023 Volume 63 Issue 7 Pages 1251-1262
For the wall-thickness reduction of the components destined for pressurized gaseous hydrogen, widespread use of high-strength martensitic steels has long been desired. However, their strong susceptibility to hydrogen-assisted fatigue crack growth (HA-FCG) is still limiting their proactive applications. Here, we instead focused on pearlite as another potential reinforcing agent for the development of new hydrogen-compatible steels with acceptable cost performance. Fatigue crack growth (FCG) behavior of three eutectoid steels with different microstructure sizes (i.e., ferrite/cementite interlamellar spacing, colony and block sizes) and strength levels was investigated in a 90 MPa hydrogen gas, an essential evaluation when attempting to perform a defect tolerant design of the components used for high-pressure gases.
The pearlitic steels clearly exhibited the acceleration of their FCG rate in hydrogen gas up to a hundred times that in air, wherein its magnitude was greater in the material with finer microstructure and concomitant higher strength. The delamination of ferrite/cementite lamellae, which were inclined largely from the loading axis, was determined to be the primary cause of such HA-FCG in pearlitic steels. Nevertheless, the extent of FCG acceleration was minor with respect to martensitic steels. The fact was ascribed to the barrier role of the cementite platelets oriented nearly perpendicularly to the crack as well as to the geometrical retardation effects arising from the crack deflection and blanching. Ultimately, pearlite was superior to martensite from the perspective of HA-FCG resistance; besides, the superiority was more substantial as the loading rate became slower.
Toward the forthcoming development of a carbon-neutral society, hydrogen (H) is attracting great attention as a sustainable energy carrier.1,2,3) Hydrogen is stored and supplied as a gaseous phase pressurized up to an order of 100 MPa in the in-service hydrogen-related infrastructures represented by hydrogen refueling stations for fuel cell vehicles. In the pressure vessels or pipelines for the storage and transportation of hydrogen gas, the strength design considering the propagation of fatigue crack from the internal to external surfaces is required because the components are subjected to dynamic loading through the repetition of the pressurization and de-pressurization cycles.4,5,6) However, hydrogen dissolution into structural metals is known to accelerate the propagation rate of fatigue crack, i.e., the phenomenon recognized as hydrogen-assisted fatigue crack growth (HA-FCG),7,8,9,10,11,12,13,14) concerning the emergence of an unpredicted premature failure in hydrogen energy-related infrastructures.
The well-known Paris law generally describes the growth behavior of relatively large fatigue cracks longer than 1 mm. A formula da/dN = CΔKm yields upon the crack propagation distance per cycle and stress intensity factor range are respectively defined as da/dN and ΔK, as well as the material parameters, C and m, are given.15,16) In the majority of structural steels, the fatigue crack growth (FCG) is immune to the influences from internal microstructures and relevant yield/tensile strength of the materials under the absence of environmental effects.17,18,19) Meanwhile, when the FCG operates in a pressurized gaseous hydrogen environment, such a structure-insensitive property conversely turns into structure-sensitive to a significant extent.5,11,19,20,21,22) This strong microstructure-susceptible nature of the HA-FCG urges manufacturers to securely select compatible materials for hydrogen-relevant infrastructures. Nevertheless, at the same time, the fact implies that we can more or less suppress the magnitude of the FCG acceleration by pertinently controlling the steel microstructures via adequate thermo-mechanical processing.
The HA-FCG characteristics under a gaseous environment have been elaborated in steels primarily consisting of ferrite and martensite.7,8,9,10,11,13,14,21) In these basic phases with body-centered cubic (BCC) crystal structure, the FCG acceleration rates reach 10–1000 times those in ambient air, altering the fracture modes from ductile striations to intergranular (IG) or quasi-cleavage (QC).7,9,10,13,14,21) Furthermore, the magnitude of acceleration has been revealed to be a positive function of the material’s strength level, wherein a sharp augmentation arises when the tensile strength exceeds 900 MPa.19,20,21) The origin and underlying mechanisms responsible for the H-induced IG and QC fractures are also continuing the critical interests of researchers, proposing some key mechanisms in terms of grain boundary (GB) decohesion by segregated hydrogen atoms,23,24,25) nano-scale void nucleation along GBs,26,27,28) and local enhancement of dislocations activity ahead of the crack tip.12,29,30) On the other hand, little attention has been placed on the HA-FCG property and accompanying fracture morphologies of pearlite,5,11,19,22) a commonly contained microstructure in the various commercially available BCC steels.
There is an increasing demand from industries to facilitate the use of stronger steels for reducing the manufacturing cost of hydrogen-bearing components, although the 900 MPa border mentioned above is limiting the wider application of martensitic steels.20,21) Notwithstanding, despite its appreciable strength level equivalent to martensite, pearlite exerts a much superior resistance to hydrogen-assisted fracture under static tensile and constant stress loading conditions.31,32) Such a lower susceptibility to hydrogen has been attributed to the absence of prior austenite grain boundaries (PAGBs), which act as a preferential propagation pathway of hydrogen-induced cracks in martensite, as well as to the barrier effect of hard cementite (Fe3C) lamellae oriented perpendicularly to the crack propagation direction.31,33) Also, in the FCG tests performed under a 90 MPa hydrogen gas environment, the authors recently reported that a lower FCG acceleration rate could be obtained in pearlitic steel with 1080 MPa tensile strength compared with martensite having an identical strength level and even with ferrite.19) Similar investigations were also performed using several hypoeutectoid carbon steels, identifying the mitigated FCG acceleration with an increased pearlite volume fraction.22) On the basis of these experimental findings, pearlite can somehow be proactively utilized in the microstructure design for the development of new hydrogen-resistant steels. In particular, it can be an easily-available material-reinforcing agent in pipeline steels primarily composed of ferrite-pearlite dual structures.4,8,11)
This paper is a continuation of our investigations on fully pearlitic steels,19) wherein the sizes of hierarchical microstructures and the associated change in the strength level were newly involved as the parameters characterizing the materials. FCG properties of three eutectoid steels with 900–1080 MPa tensile strength were studied in a pressurized gaseous hydrogen environment, followed by clarification of the factors controlling the microstructure- or strength-dependences of HA-FCG from the perspective of fracture surface morphology, crack path, deformation substructure around crack wake, and some geometrical effects arising from the crack shape.
The material used in this study was a hot-rolled tool steel plate (JIS-SKS5) with a thickness of 7 mm, which had spherodized microstructure in its as-received state. The chemical composition of the steel was Fe-0.79C-0.24Si-0.43Mn-0.012P-0.004S-1.18Ni-0.43Cr-0.01Cu (mass%) as indicated in the inspection sheet supplied by the manufacturer. The plate was austenized at 850°C for 1 hour, followed by isothermal transformation treatments at three different temperature-time conditions: 690°C for 5 hours, 660°C for 30 min, and 550°C for 20 min (Fig. 1(a)), in order to make the materials with various microstructure sizes and strength levels. For the materials transformed at 660 and 550°C, the austenizing and isothermal heat treatments were both performed by salt bath soaking. Meanwhile, the heat treatment for the one transformed at 690°C was thoroughly done in a vacuum furnace.
Temperature-time chart of the heat treatments used for producing the pearlitic steel with three different microstructure sizes (a) and the configuration of compact-tension (CT) specimen used for the fatigue crack growth tests (b). The loading axis of the CT specimen coincides with the rolling direction of the as-received plate material.
The microstructure of pearlite, an aggregate of fine ferrite/cementite lamellae, is further hierarchically divided into two classes with different length scales: colony, where the lamellae are aligned into a unified direction; block, which consists of several colonies yet the ferrite phase exhibits almost an identical crystal orientation. Figure 2 shows the microstructures of the present three materials observed by a scanning electron microscope (SEM) after etching with 3% nital solution as well as by electron backscattered diffraction (EBSD) technique. As the isothermal transformation temperature was decreased, ferrite/cementite interlamellar spacing, colony size, and block size were all refined, the quantitative results of which are summarized in Table 1. The interlamellar spacing was measured according to the method described by Caballero et al.,34) while the colony and block sizes were defined as the average line intercepts. Table 1 also includes the mechanical properties measured by the tensile tests in ambient air with the initial strain rate of 5.6×10−4/s. The yield stress, σy, and tensile strength, σB, became higher with the refinement of microstructure sizes. Afterward, these three are designated as Fine, Medium, and Coarse materials after their fineness of the microstructure sizes (see Table 1).
Microstructures of fine (a)(d), medium (b)(e), and coarse (c)(f) pearlitic steels produced by the heat treatments shown in Fig. 1(a). (a)–(c) were acquired by EBSD, while (d)–(f) were captured by SEM after etching with nital solution. The black and white lines in (a)–(c) delineate the pearlite block (> 15°) and low angle (5–15°) crystal boundaries, respectively, whereas white lines in (d)–(f) describe pearlite colony boundaries. The pictures’ vertical and horizontal axes correspond to the plate material’s rolling and thickness directions, respectively. (Online version in color.)
| Material | Transformation temperature (°C) | Interlamellar spacing (nm) | Colony size (μm) | Block size (μm) | Yield stress (MPa) | Tensile strength (MPa) |
|---|---|---|---|---|---|---|
| Fine | 550 | 120 | 3.7 | 11 | 650 | 1080 |
| Medium | 660 | 190 | 5.8 | 23 | 440 | 950 |
| Coarse | 690 | 230 | 7.8 | 32 | 420 | 900 |
FCG tests were carried out in laboratory air and 90 MPa gaseous hydrogen environment at room temperature in accordance with the ASTM-E647 standard.35) Small-sized compact-tension (CT) specimen with the configuration shown in Fig. 1(b) was extracted so that its loading axis coincides with the rolling direction, then used for the FCG tests. The tests were conducted under a sinusoidal waveform and a load ratio, R of 0.1, using a servo-hydraulic testing machine with a 50 kN load capacity attached to a 100 MPa pressure vessel. For acquiring the relationship between da/dN and ΔK, the tests with constant load range, ΔP, (ΔP-constant tests) were first performed with the loading frequency, f, of 5 and 1 Hz for air and hydrogen gas, respectively. In addition, the tests in which ΔK was controlled constant (ΔK-constant tests), i.e., the load, P, was continually decreased with the crack advance, were carried out with f = 0.01–1 Hz for the purpose of evaluating the loading rate effect on the FCG rate in hydrogen gas under specific mechanical conditions. The crack length during the test was monitored in terms of the unloading elastic compliance method via a crack-mouth opening displacement (CMOD) measurement by a clip-on gauge. Crack closure behavior was also elucidated for some specimens based on the measured relationship between load, P, and CMOD. The detailed method of this analysis will be provided in the later part.
2.3. Observations of Fracture Surfaces and Cracking PathwaysThe fracture surfaces on the specimens after ΔP-constant tests were observed by a field-emission SEM, JEOL JSM-7001F, operated at an acceleration voltage of 15 kV. On the other hand, the specimens subjected to ΔK-constant tests were cut along their mid-thickness sections and then polished with colloidal SiO2 solution in order to ensure damage-free surfaces. The cross-sectional shapes of the cracks, microstructural cracking pathways, and deformation substructures in the crack wakes were characterized by EBSD and electron channeling contrast imaging (ECCI) in the same SEM with the fracture surface observation. The acceleration voltage for ECCI was 30 kV, while it was operated at 15 kV and a beam step size of 300 nm for EBSD.
Figure 3 depicts the da/dN-ΔK relationships of the three materials measured by ΔP-constant tests in the air and 90 MPa hydrogen gas. At the ΔK up to 30 MPa·m1/2 where the conventional Paris law typically yields, no detectable microstructural influence manifested on the da/dN-ΔK curves in the air. While at the other extreme, the da/dN in hydrogen gas was clearly augmented compared with that in the air in all Fine, Medium, and Coarse materials. The microstructure and strength level exerted a significant impact on the magnitude of such FCG acceleration. The gap of da/dN between the two test environments was amplified with the increase of ΔK.
Relationship between fatigue crack growth rate per loading cycle, da/dN and stress intensity factor range, ΔK of fine, medium, and coarse pearlitic steels in a laboratory air and 90 MPa hydrogen gas environment. The previous data of pure ferritic iron36) and martensitic steels20,21) having equivalent tensile strength levels with the present fine pearlitic steel obtained in 90–95 MPa hydrogen gas is plotted together to highlight the influence of microstructure. (Online version in color.)
In Fig. 3, the results of pure ferritic iron36) and quenched-tempered medium-carbon martensitic steels20,21) having similar tensile strength to the present pearlitic steels in 90–95 MPa are presented together to highlight the impact of the difference in microstructures. Comparing particularly at the ΔK of around 20 MPa·m1/2, the FCG rates of Fine, Medium, and Coarse materials in hydrogen gas were far below those of ferritic iron and martensitic steels, demonstrating the superior HA-FCG resistance of pearlite to the other two basic microstructures. Nonetheless, although the acceleration rates in Medium and Coarse materials remained around five times with respect to the in-air condition, that in the Fine material, having the finest microstructure and highest strength, was larger by up to 30 times at relatively high ΔK. This strength-dependent character of the HA-FCG magnitude in pearlite is in common with the general tendency observed in martensitic steels.10,20,21,37)
Even in the air, acceleration of FCG deviating from the linear Paris law behavior can be observed in Medium and Coarse materials when ΔK exceeded 30 MPa·m1/2. Such a deviation has its root in the mediation of static-type fracture accompanying brittle cleavage owing to the low toughness of pearlite with a coarser block size (or larger prior austenite grain size).38,39,40) In other words, the da/dN measured at ΔK > 30 MPa·m1/2 does not purely reflect the cyclic component of crack growth but also contains the contribution from the crack advance under monotonic loading irrespective of the presence and absence of hydrogen. For this reason, the focus of the present investigation was only placed on the domain of ΔK ≦ 30 MPa·m1/2.
In Fig. 4, the relative FCG rate in hydrogen gas with respect to those in the air, (da/dN)H/(da/dN)Air, measured via ΔK-constant tests are depicted at three different ΔK values: 20, 25, and 30 MPa·m1/2. For the ΔK of 25 and 30 MPa·m1/2, the results of medium-carbon martensitic steels (JIS-SCM435 and SCM44020,21)) with various strength levels, which were acquired in similar environmental conditions, are also plotted for comparison. In general, when the K value exceeds its threshold for the onset of static hydrogen-assisted cracking, KIH, high-strength martensitic steels with > 900 MPa tensile strength exhibit a so-called time-dependent HA-FCG: (da/dN)H/(da/dN)Air monotonically augments with the slowing of loading frequency.10,21,37) The origin of time-dependent cracking lies in the repetition of hydrogen accumulation at the maximum hydrostatic stress field ahead of the crack tip as a rate-controlling step, followed by the nucleation of precursory micro-cracks at PAGBs or block boundaries of martensite and their subsequent coalescence with the main crack.21,37) Figures 4(b) and 4(c) indeed demonstrate the occurrence of such time-dependent crack growth, particularly in the martensitic steels with σB > 900 MPa. Meanwhile, at least under the ΔK levels of 20 and 25 MPa·m1/2 for all three pearlitic steels as well as under ΔK = 30 MPa·m1/2 for Medium and Coarse materials, no time-dependent HA-FCG was recognized despite their tensile strength exceeding 900 MPa: crack propagation was totally cycle-dependent wherein the da/dN at a lower frequency was coincident with that at f = 1 Hz. However, the situation changed in Fine material when ΔK was 30 MPa·m1/2, manifesting a time-dependent character of FCG due to its high tensile strength beyond 1000 MPa. Based on these findings, it can be concluded that, while pearlite is basically superior to martensite in its balance between strength and HA-FCG resistance, a concern about a catastrophic failure arises under the satisfaction of the critical levels of strength and crack tip stress intensity. Even so, those critical σB and K for the onset of time-dependent fracture seemed higher in pearlite than those in martensite.
Acceleration rate of FCG in Fine, Medium, and Coarse pearlitic steels in 90 MPa hydrogen gas environment at ΔK = (a) 20, (b) 25, and (c) 30 MPa·m1/2 with respect to the FCG rates in the air. Previous results of medium-carbon martensitic steels with various strengths20,21) are included together in (b) and (c) for comparison. (Online version in color.)
Figure 5 shows some examples of the fracture surfaces of the three pearlitic steels at ΔK = 20 MPa·m1/2 after the ΔP-constant tests in the air. Randomly oriented stripe patterns covered almost all parts of the fracture surface, which might arise from the tearing fracture of the ferrite/cementite lamellar structure. Even though the length scale of these stripes shrunk as the size of the microstructure was refined, the distinction was not altered regardless of the material type and ΔK level.
Fracture surfaces of fine (a), medium (b), and coarse (c) pearlitic steels in laboratory air at ΔK = 20 MPa·m1/2 and f = 5 Hz, demonstrating the randomly oriented striped patterns which originate from the rupture of ferrite/cementite lamellar structure. The crack growth direction is from bottom to top.
The fracture surfaces formed in hydrogen gas at the same ΔK as Fig. 5 are presented in Fig. 6. Since the distinctive fractographic characters were almost identical in all three materials, only some examples of the Medium material are provided in Figs. 6(d)–6(f) as the high magnification images. Whereas the striped patterns due to the tearing of pearlite lamellar were occasionally observed even in hydrogen gas (Fig. 6(d)), considerable parts of the fracture surface were covered by planar faceted morphology, which is surrounded by dashed lines in Figs. 6(a)–6(c). Figure 6(e) magnifies one of such facets emphasizing their planar nature even in a sub-micron scale. The scaling of these individual facets tended to shrink with the reduction of microstructural sizes. Furthermore, the facets possessed broad planar dimensions when they were nearly parallel to the global fracture surface plane. Meanwhile, when they were relatively inclined from the global fracture surface, stair-like features were evident, as shown in Fig. 6(f).
Fracture surfaces of fine (a), medium (b)(d)(e)(f), and coarse (c) pearlitic steels in 90 MPa hydrogen gas environment at ΔK = 20 MPa·m1/2 and f = 1 Hz. Owing to the influence of hydrogen, planar facet-like distinction, which was not detected in laboratory air, was identified as surrounded by dashed lines in (a)–(c). (d)–(f) magnify some representative fractographic features marked by arrows in (b). The crack growth direction is from bottom to top.
The areal fractions of these planar facets in hydrogen gas to the overall fracture surfaces were measured, and the results are plotted in Fig. 7 for three materials as a function of ΔK. For the construction of Fig. 7, the mid-thickness part, as well as the portions ±1 mm apart from the mid-thickness, were captured with a fixed magnification (500 times for Coarse and Medium materials; 1500 times for Fine material), and the plot points and error bars were derived as the averages in the three fields of views and the ranges from the minimum to maximum, respectively. We were compelled to use a higher magnification only for Fine material because its small scaling of facets made it infeasible to distinguish them with a lower magnification, although low magnification was better from a statistical viewpoint. The areal fractions of facets were first augmented from ≈ 20% with the increase of ΔK, then saturated around 30–35% (i.e., fluctuated only within the error bands) at the ΔK larger than 20 MPa·m1/2 in all three materials.
Area fraction of planar facets, which represent the delamination of ferrite/cementite lamellar, on the fracture surfaces in 90 MPa hydrogen gas as a function of stress intensity factor range. (Online version in color.)
Figures 8(a)–8(c) shows the appearance of crack propagation paths in the air during the ΔK-constant tests at ΔK = 20 MPa·m1/2 and f = 1 Hz, which were analyzed on the mid-thickness portions of the CT specimens by using EBSD. Regardless of the local alignment of the ferrite/cementite lamellar structure, the cracks in air tended to propagate straightly inside the pearlite blocks, besides some small portions of block boundary fracture were also included where the block boundary was incidentally located in the proximity of the crack. The relatively smooth crack path was the character in the air from the macroscale viewpoint.
Cross-sectional EBSD micrographs around the cracks in (a)(d) Fine, (b)(e) Medium, and (c)(f) Coarse materials propagated in (a)–(c) the air and (d)–(f) 90 MPa hydrogen gas environment under ΔK = 20 MPa·m1/2 and f = 1 Hz. The inverse pole figure (IPF) color is defined as the orientation along the loading direction. The arrows in (d)–(f) denote the locations where the cracks exhibited blanching. (Online version in color.)
As an example, the kernel average misorientation (KAM) map of the same area with Fig. 8(b) and a magnified ECCI micrograph of a representative region (marked in Fig. 9(a)) are depicted in Figs. 9(a) and 9(b). The color changes in the KAM map denote the average misorientation between the targeted EBSD scan grid and its surrounding neighbors, qualitatively reflecting the accumulated density of geometrically necessary dislocations (GNDs).41) As can be identified in Fig. 9(b), the crack in the air propagated via traversing the ferrite/cementite lamellar, wherein high KAM values and bamboo-like sub-grain structures in between the cementite platelets were recognizable in its vicinity. This lamellar-traversing (LT) type of crack propagation, which was a primary fracture mode in the air, might be the origin of the pearlite tearing striped patterns recognized on the fracture surface (Fig. 5(b)). Even though any extra images are not shown here, it was confirmed that the crack propagation morphology in Fine and Coarse materials was almost the same with Medium material.
(a) kernel average misorientation (KAM) map around the crack of Medium material propagated in the air under ΔK = 20 MPa·m1/2: the field-of-view is the same as Fig. 8(b). (b) magnifies the region surrounded by a rectangle in (a), demonstrating the presence of bamboo-like sub-grain structure inside the ferrite layer that accompanied considerable development of KAM values with 2–3°. (Online version in color.)
The cross-sectional images of the cracks grown in hydrogen gas at ΔK = 20 MPa·m1/2 and f = 1 Hz are shown in Figs. 8(d)–8(f) so as the impact of hydrogen can be visible by comparing them with Figs. 8(a)–8(c). As for the crack propagation mode in the air, major parts of the cracks in three materials went through the insides of pearlite blocks and occasionally along block boundaries. However, one notable distinction in hydrogen gas was that the crack exhibited frequent deflection and blanching, giving rise to more undulated fracture paths than those in the air. More quantitatively, Fig. 10 presents the length ratio of the actual crack path, aac, to its projection to the loading axis, apr, of the three materials in the air and hydrogen gas. The aac/apr was increased by almost 30% in hydrogen gas, the magnitude of which was equivalent irrespective of the material type.
Ratio between the actual crack length, aac, and projected crack length, apr, in the air and 90 MPa hydrogen gas at ΔK = 20 MPa·m1/2, indicating the H-effect on the magnitude of crack deflection and blanching. (Online version in color.)
In Figs. 11(a)–11(d), the KAM map of the fracture path in Medium material, i.e., the same field-of-view as Fig. 8(e), is presented, in addition to the ECCI images of some distinctive regions which are surrounded with white rectangles in Fig. 11(a). Although the major part of the crack wake in the air (Fig. 9(a)) exhibited more or less high KAM values (i.e., 2–3°), such misorientation development was relatively uneven along the fracture path in the case of hydrogen gas environment: the crack wake was decorated by the regions with low and high KAM values as marked in Fig. 11(a) by red and yellow arrows, respectively. As the character inside the high KAM value area (Fig. 11(b)), the pearlite lamellar was aligned nearly perpendicular to the crack growing direction, developing the deformation microstructure with sub-grains similar to the one observed in the air (Fig. 9(b)). Meanwhile, the lamellar was largely inclined from the loading axis or sometimes almost parallel to the global crack growth direction in the area with low KAM values (Figs. 11(c) and 11(d)), wherein the crack locally propagated along the lamellar orientation with weaker development of dislocation substructures. This lamellar-delaminating (LD) type of cracking was a specific failure mode in hydrogen gas, an origin of the planar facets observed on the relevant fracture surfaces (Fig. 6). These characteristics crack propagation morphologies were also observed in Fine and Coarse materials, as well as in the pearlite grains in several hypoeutectoid carbon steels.22)
Details of the crack propagation path of Medium material in 90 MPa hydrogen gas under ΔK = 20 MPa·m1/2 and f = 1 Hz. (a) is the KAM map of the same field-of-view as Fig. 8(e), while (b)–(g) are the magnification of the regions surrounded by white rectangles in (a) and (e). The arrows and dashed lines with yellow, red, and white colors indicate the parts exhibiting lamellar traversing (LT), lamellar delamination (LD), and block boundary fracture, respectively. The LT-fractured parts are decorated by high KAM values with 2–3°, while LD- and block boundary-fractured parts showed relatively weak KAM development. (Online version in color.)
An elaboration on the crack tip zone, the result of which is shown in Figs. 11(e)–11(g), unveiled some key features that can deepen our understanding of the crack propagation processes under the presence of hydrogen. In front of the main crack at the left-hand side in Fig. 11(e), some isolated sub-cracks were nucleated along the pearlite lamellar structure with occasionally zigzag manners or along the pearlite block boundary. Supposedly, such a zigzag cross-sectional appearance corresponds to the stair-like faceted distinction of the LD parts observed on the fracture surface (Fig. 6(f)). Focusing on the main crack front and the sub-crack formed along the lamellar, it can be seen that they were arrested when they entered the region where the cracks and local lamellar alignments were mutually perpendicular (i.e., the portions marked with yellow arrows and dashed lines in Figs. 11(e)–11(g)). A comparison of these two ECCI micrographs to the KAM map (Fig. 11(a)) also enlightened us about the considerable development of misorientation in association with such crack-arresting events.
Based on the experimental results provided above, it has been uncovered that the steels with fully pearlitic microstructure basically exhibit superior resistance to HA-FCG than ferritic iron and martensitic steels with equivalent strength levels. Nonetheless, the superiority became less pronounced with the increase of tensile strength of pearlite and with the augmentation of stress intensity, manifesting a time-dependent character of H-assisted cracking as the case of martensite when the tensile strength was greater than 1000 MPa and ΔK > 30 MPa·m1/2. In the following, the underlying rationales for these dual aspects are discussed from the perspective of geometrical effects as well as from microstructural anisotropy.
4.1. Impact of Crack GeometryAs demonstrated in Figs. 8(d)–8(f), the fatigue cracks in hydrogen gas propagated accompanying frequent and large extent of deflection and blanching while those in the air were relatively straight. These intermittent crack deflection/blanching potentially caused the temporal decrease in the effective Mode I stress intensity factor, reducing the net driving force to propel the crack tip perpendicularly to the loading axis.42,43) Moreover, when the complete engagement of the mating crack faces under unloading becomes difficult due to the crack shape asperity, earlier contact of the crack faces ensues before the load reaches its minimum value, Pmin. Such a situation may instigate zero effective Mode I loading at the local crack tip zone, even though a considerable load is still being applied externally to the specimen: a phenomenon called roughness-induced crack closure (RICC).44,45,46)
Under the use of a clip-on gauge during the FCG test, the signature of crack closure generally manifests in the form of knee-point on the P-CMOD curve as schematically drawn in Fig. 12(b).46) The load corresponding to this knee-point, Pop, can be assumed to be the external force required to open the contacted crack faces, a moment at which the net driving force for Mode I crack propagation commences to arise. By subtracting the forces below Pop, the effective load range, ΔPeff = Pmax−Pop (Pmax is the maximum load), which is used to determine the effective stress intensity factor range, ΔKeff, can be obtained instead of ΔP = Pmax−Pmin by eliminating the ineffective part of loading that stems from RICC. Figures 12(c)–12(d) indeed demonstrates the presence of knee-point on the P-CMOD response of all the three materials in hydrogen gas at ΔK = 20 MPa·m1/2, in which the procedure for constructing these diagrams is as follows: the experimentally acquired P-CMOD curve during unloading was fitted with a regression line by using the data within the range of 50–95% of ΔP (see Fig. 12(b)); the COMD deviation rate of the experimental data from the regression line is plotted as its relationship with P. As can be seen in Figs. 12(c)–12(e), the experimental P-CMOD data include the fluctuations from the regression line almost within ±2%. Thus, in this study, the Pop was defined at which the CMOD deviation rate exceeded 2%. By repeating this procedure for multiple ΔK levels, the da/dN-ΔKeff curves shown in Fig. 12(a) were finally constructed.
(a) fatigue crack growth curves of three pearlitic steels plotted as a function of effective stress intensity factor range, ΔKeff, for screening the influence of crack closure. For comparison, the da/dN-ΔK data in Fig. 3 are also represented. The method for determining the crack opening load, Pop, at each ΔK level is schematically drawn in (b), while some examples of the practical P-CMOD responses at ΔK = 20 MPa·m1/2 derived by the method are shown in (c)–(e). (Online version in color.)
In Fig. 12(a), the original da/dN-ΔK data of the pearlitic steels, as well as of ferritic and martensitic materials, which were already shown in Fig. 3, are overlaid. Comparing the da/dN-ΔK and da/dN-ΔKeff curves in the pearlitic steels each other, no RICC appeared in the air as deduced from the relatively straight crack configuration (Figs. 8(a)–8(c)). On the other hand, the da/dN in hydrogen gas exhibited an upward shift upon using ΔKeff, concluding that RICC more or less contributed to the low magnitude of HA-FCG in pearlite, irrespective of the microstructural size and strength level. Nevertheless, even in the evaluation by ΔKeff, the da/dN in the pearlitic steels was still somewhat lower than those in ferrite and martensite, particularly in Medium and Coarse materials. Thus, the superior HA-FCG resistance of pearlite cannot thoroughly be rationalized from the viewpoint of the geometrical effects.
4.2. Effects of Microstructural AnisotropyA key microstructural perspective for interpreting the extent of HA-FCG was the emergence of dual types of fracture morphologies under the presence of hydrogen: LT type with striated features on the fracture surface; LD type accompanying planar facets. Whereas the former was common in both testing environments, the latter was encompassed only when the materials were tested in hydrogen gas. The authors preliminary performed the same experiments on some carbon steels with mixed microstructures of ferrite and pearlite, elucidating that the appearance of facets was a consequence of H-induced fracture along ferrite/cementite phase boundaries in the pearlite grains.22) Based on the similarity between the facets recognized in the previous22) and present (Fig. 6) studies, the LD in the three pearlitic steels can be deemed to also originate from the interphase delamination event. Note that the cracking along pearlite block boundaries was partially detected in the air and hydrogen, although its evidence was hard to be captured on the fracture surfaces.
Around the crack wake in the air, where LT-type fracture was prevalent, large KAM values (i.e., substantial evolution of local plasticity) were involved with the underlying deformation microstructure characterized as sub-grains (Fig. 9). In general, the significance of plasticity and dislocation structure developments in the crack wake is determined by the amount of accumulated strain inside the plastic zone ahead of the propagating crack tip.22,47,48,49,50) More specifically, the plastic zone size, rp, in most steels and alloys is the order of 10–100 μm, while crack propagation distance per loading cycle is less than a few μm. The crack thus has to pass through the pre-existing plastic zone over a number of cycles accompanying the formation of a fresh plastic zone in front of the moving crack tip. In the course of this process, the plastic zone is inevitably subjected to cyclic plasticity, as with the case in the low cycle fatigue of smooth specimens. Considering a fixed inspection volume located at the forefront of the plastic zone, discrete dislocations are first introduced, eventually transforming into more organized structures, i.e., cells and sub-grains, with the approaching of the crack tip and resultant increase of stress/strain amplitude as well as the number of strain cycles.22,47) In this context, the deformation structure at a particular part of the crack wake may reflect how many loading cycles were required for the crack to propagate through that specific portion of the material. If the crack, for example, propagated faster, the structural evolution should be weakened due to the fewer applied strain cycles to the plastic zone.
According to Irwin’s first approximation (i.e., rp = (1/3π)(Kmax/σy)2, where Kmax is the maximum stress intensity factor16)), the plastic zone sizes under plane-strain stress state resembling the mid-thickness part of the CT specimen are approximately 120, 270, and 300 μm at ΔK = 20 MPa·m1/2 for Fine, Medium, and Coarse materials, respectively. By contrast, the relevant FCG rate in the air was less than 0.1 μm (Fig. 3). This gap means that more than 1000 cycles of plastic strain were applied to the primary plastic zone when the crack passed through it, giving rise to the development of high KAM values and sub-grain structures presented in Fig. 9. The sub-grains in the present steels exhibited bamboo-like form in the ferrite layer interbedded by cementite, while those were equiaxed in ferritic iron and steels.7,22,48) Such a specific morphology of sub-grains or dislocation cells is a characteristic of pearlite when it is strained with a considerable number of cycles and high stress/strain amplitude.51) As demonstrated in Fig. 11(a), the KAM development around the LT-fractured parts was still significant even in hydrogen gas, whereas it remained relatively low in the case of LD (Figs. 11(c) and 11(d)). On the basis of the premise above, this difference implies that the crack propagated with slow and fast growth rates through the LT and LD regions, respectively. In fact, the reduction in the crystal misorientation inside the crack wake has been an evident consequence pertaining to the H-induced FCG acceleration in some ferrite-based materials.7,9,22)
In order to figure out the impact of microstructural anisotropy of pearlite on the resistance to H-assisted crack propagation, Tomatsu et al. carried out micro-scale bending tests of a cathodically-charged notched cantilever under various combinations of crack growth direction and lamellar orientations.33) They identified superior fracture resistance when the stacking direction of the lamellar and the crack growth direction was mutually parallel (i.e., LT in the present expression), besides the resistance became the lowest accompanying ferrite/cementite interface failure when the crack plane laid parallel to the lamellar plane (i.e., LD). These microscopic insights can reasonably explain the present characteristics around the LT: the fatigue cracks slowly propagated by standing against the high resistance, cumulating a number of plastic strain cycles and resultant microstructural (i.e., KAM) evolution (Fig. 11(a)). The high fracture resistance of the LT parts can be attributed to the low capability of cementite to absorb hydrogen at ambient temperature,52,53,54) which preserves it as the hard phase yet immune to H-induced embrittlement. That is, even if the interbedded ferrite layer could still be H-sensitive, the crack tip must encounter solid barriers with a regular interval, a rationale behind the superior resistance to HA-FCG of pearlite than ferrite and martensite from a microstructural perspective. The highly fracture-resistant nature of axially aligned pearlite with H-absorption was also reported recently by Ueji et al., even in the macroscale specimen of a caliber rolled C–Si–Mn–Cr steel.55) Even though our two-dimensional observation was not capable of assessing the fracture behavior of the other case where the lamellar is stacked parallel to the CT specimen thickness, the previous study notably clarified a comparable fracture resistance of such parts to LT33) because the tearing of cementite into its longitudinal axis is similarly required for the crack to propagate forward.
In addition to the weak development of KAM values, another notable signature of the LD-type fracture was that they predominantly manifested when the pearlite lamellar was nearly parallel to the global crack plane or largely inclined from the loading axis (Fig. 11(a)). Notably, the latter situation was one of the primary reasons for the crack deflection shown in Figs. 8(d)–8(f). Such an angular dependence of the emergence of LD implicates a brittle nature of ferrite/cementite interface debonding, in which the normal stress acting perpendicularly to the lamellar plane played a primary driving force for the onset of fracture. Note, however, that the critical debonding stress is not a simple function of the lamellar inclination angle and externally applied load because of the possible variations in the microstructures surrounding the crack tip zone, local crack geometry, and the resultant heterogeneity of the stress field. An interface separation along the ferrite/cementite phase boundary can be attributed to the well-known hydrogen-enhanced decohesion (HEDE) hypothesis.23,24,25) It has been experimentally as well as analytically confirmed that the ferrite/cementite interface functions as a preferential trapping site for hydrogen atoms,52,53,54,56,57) reducing the interfacial cohesive force52) and encompassing the specific failure mode infeasible under the absence of hydrogen. Moreover, Takai et al. and Yu et al. found an increasing hydrogen absorption capacity after the deformation of pearlite due to an accumulation of lattice defects (e.g., dislocations and vacancies) at the ferrite/cementite interfaces and their trapping of hydrogen atoms.58,59,60) Considering the lamellar involved within the plastic zone ahead of the crack where cyclic strain is cumulated, a similar situation could preliminarily be realized prior to the oncoming of the crack tip. The combination of the reduction in interfacial cohesive force and the reduced integrity of the interface due to H-trapping defects might make the separation of the interface much easier, finally making the LD the potential trigger of H-assisted FCG acceleration in pearlite. In the present context, the LD-type fracture should be enhanced with the increase of stress intensity, material’s strength level, and hydrogen gas pressure, which augment the normal stress and the amount of segregating hydrogen on the interface. The area fraction of LD facets and the magnitude of FCG acceleration were indeed escalated with the increases of ΔK and tensile strength (Figs. 3 and 7), although the facet fraction was saturated around 30–35%, supposedly due to the maximum probability of the favorably oriented lamellar at the crack tip fracture process zone. The influence of hydrogen gas pressure has also been verified in the authors’ previous study regarding a steel with ≈ 70% volume fraction of pearlite.22)
If the favorably oriented lamellar colony is located incidentally at the crack tip, the crack may grow by separating the phase interfaces inside the colony. Otherwise, it is feasible that the lamellar located slightly ahead of the main crack tip is delaminated and generates precursory cracks, as observed in Figs. 11(e)–11(g), coalescing subsequently with the main crack. When a whole lamellar colony was cleanly fractured, the resultant fracture surface forms a broad planer facet as shown in Fig. 6(e). Meanwhile, when the lamellar was inclined yet the LD crack tended to propagate with maintaining its perpendicular orientation to the externally applied load, an intermittent delamination with the stair-like crack morphology (Fig. 11(f)) might be operated. Supposedly, the easiness and frequency of the precursory delamination were higher in the materials having a smaller microstructural length scale, one of the possible reasons for the greatest FCG acceleration rate in Fine material. Once the favorable lamellae for LD are fractured and a certain fraction of the crack front encounters the fracture-resistant regions (i.e., LT parts) again, the crack propagation could be slowing down, accompanying a substantial extent of plasticity (Fig. 11(a)). At a relatively low ΔK level, the LT regions might be capable of prompting a complete arrest of the brittle LD-crack under a static load, thereby no further FCG acceleration occured even if the frequency was lowered and thus the load-holding time was prolonged, i.e., the HA-FCG was cycle dependent (Fig. 4). However, the escalation of ΔK and the material’s strength level render the hydrogen accumulation and resultant fracture along ferrite/cementite interfaces autocatalytic during the load-holding process, giving rise to time-dependent and catastrophic crack propagation as demonstrated in Fig. 4(c).
The block boundary fracture of pearlite was also a feature detected in the cross-sectional analyses irrespective of the presence and absence of hydrogen (Figs. 9(a) and 11(a)). Such a failure mode seems to be a conceivable outcome since the stress/strain concentration could be raised due to the mismatch of plasticity between the differently oriented adjacent lamellar colonies.61) Nevertheless, despite the fact that the region with block boundary failure in the air was accompanied considerable amount of plasticity (Fig. 9(a)), those in hydrogen gas exhibited relatively weak KAM evolution (Fig. 11(a)), implying that hydrogen altered the fracture event into the more brittle way. Hagi et al. successfully identified a greater hydrogen trapping capability in spherodized cementite having curved phase boundaries than in pearlite with planar ferrite/cementite interfaces.54) Extending their assertion, hydrogen trapping could similarly be intense at the block boundary where planar cementite is terminated with its curved interface, a presumption consistent with Chan and Charles, who pointed out the strong trapping at ferrite/pearlite or pearlite/pearlite grain boundaries.62) If the block boundary is located feasibly nearby the propagating crack tip, the crack in hydrogen is likely to select it as the weakest link for fracture. Nonetheless, it should have been a minor case considering the much higher probability for the crack to encounter the ferrite/cementite interfaces that are selectable as propagation pathways. Based on the fracture surface morphology (Fig. 6), the failure mode most responsible for the FCG acceleration was apparently LD, and the block boundary fracture might have merely a small contribution. Rather, these multimodally available fracture pathways could be another cause of crack deflection and blanching. An example of the Y-shaped crack deflection where the crack blanched into LD and block boundary is seen at the region just below the area (c) in Fig. 11(a).
4.3. Strategies for the Practical ApplicationsSynthesizing the above findings, one can define some strategies to utilize pearlite in hydrogen-bearing components in which the failure stemming from FCG is concerned. First, the interlamellar spacing should be moderate to retain the material’s strength below 1000 MPa, even though finer lamellar is known to improve the resistance to H-assisted fracture in a conventional tensile test.60) Concomitance with finer prior austenite grain size is also desirable for increasing the nucleation sites of pearlite nodules,64,65) resulting in a smaller block size that improve the ductility and toughness.63) Furthermore, the application of cold- or warm-rolling to the as-transformed pearlite may also be an effective way to reduce the fraction of lamellae that are largely inclined from the loading axis.66) Verifying the potencies of these candidate methods is a future direction of our research.
Fatigue crack growth (FCG) properties of three pearlitic steels with different microstructure length scales and concomitant various strength levels were studied in a 90 MPa hydrogen gas environment. The experimental results were rationalized in terms of the geometrical as well as microstructural aspects of the cracking pathways, leading to some primary conclusions below:
(1) The FCG rate was more or less accelerated in hydrogen gas compared with the in-air condition to a few to a hundred times, regardless of the microstructural sizes and strength levels. The acceleration was more pronounced with increasing stress intensity factor and material’s tensile strength, although its magnitude was smaller than other conventional steel microstructures, i.e., ferrite and martensite.
(2) In the materials with a tensile strength of less than 1000 MPa, the FCG in hydrogen gas exhibited a totally cycle-dependent character under the examined ΔK range, even when the loading frequency was lowered. Nevertheless, such a stable nature of the crack propagation broke down in the highest-strength material beyond 1000 MPa, resulting in catastrophic time-dependent fracture, particularly at the high ΔK domain. Even so, the critical tensile strength and K value for the onset of time-dependent fracture were higher in pearlite than those in martensite.
(3) The fatigue cracks in hydrogen gas were accompanied by frequent deflection and blanching due to the existence of multiple microstructurally favorable fracture pathways. The resultant geometrical asperities on the crack wake caused a considerable roughness-induced crack closure, partially playing a role in retaining the FCG rate in pearlite lower with respect to other microstructures.
(4) From a microscale perspective, the H-assisted failure modes in pearlite during FCG were classified into two types, i.e., lamellar traversing (LT) and lamellar delamination (LD), although some limited parts exhibited the fracture along block boundaries. The brittle LD was a root cause of H-assisted FCG acceleration in the pearlitic steels, while LT was associated with substantial plasticity and might act as a decelerator of crack propagation.
This work was financially supported by the 30th ISIJ Research Promotion Grant. The authors are grateful to Dr. Masami Nakamura from Kobe Materials Testing Laboratory (KMTL) group, Japan, for his helpful assistance in our FCG experiments in high-pressure hydrogen gas. Y.O. also acknowledges Dr. Kaneaki Tsuzaki from National Institute for Materials Science (NIMS), Japan, for the fruitful discussion in interpreting our experimental results from the microstructural viewpoint.
