3.1. Materials and Microstructure
The microstructure of C45 steel after spheroidizing annealing (S state) is shown in Figure 2
a and Figure 3
Etched surface microstructures prior to the fatigue tests (a) initial state after soft annealing containing spheroidal carbides (b) after 6 HPT rotations (c) after 10 HPT rotations (d) initial state with tempered microstructure (e) after 6 rotations (f) after 10 rotations.
The soft annealed state contains spheroidal carbides distributed in a uniform coarse grained microstructure with well-defined grain boundaries (Figure 2
a and Figure 3
b). The grains are micrometer sized (~15 µm) and initial hardness is 169 HV. This state will be referred to as s-ini. Typical images of the S state after warm HPT for six and ten rotations (these states will be referred to as s-HPT-6 and s-HPT-10) are shown in Figure 2
b,c and Figure 3
c–f. After six rotations, the microstructure is refined and new grain boundaries appear (Figure 2
b and Figure 3
c). The grain size shrinks to approximately 0.8 µm and the grains show a slight elongation along the HPT shear direction (Figure 3
c) with a hardness of 289 HV. After ten rotations, the ferrite grain size decreases further to approximately 0.2 µm and the elongation is more pronounced (Figure 3
e,f). Hardness rises to 511 HV. The continuous grain refinement is best visible in Figure 3
The spheroidal carbides are still clearly visible in s-HPT-10, but the larger ones in Figure 3
e are partially fragmented and appear blurred. In addition, an increasing number of small carbides are distributed between the ferrite grains.
Orientation (a,c,e) and phase (b,d,f) maps (ACOM-TEM) overlaid with the reliability of the spheroidized microstructures (a,b), after 6 (c,d) and 10 HPT rotations (e,f) all prior to the fatigue tests. The color code of the orientation maps is given in (a). The projection direction is normal to the paper plane. The Fe phase is given in red and green represents the carbide phase. The reliability is given in black for both types of maps.
Needle-shaped microstructures are visible in Figure 2
d and Figure 4
a,b for C45 steel after tempering (T state). Carbides are finely dispersed between the needles (Figure 4
b). The samples have an initial hardness of 388 HV before HPT. This state will be referred to as t-ini. After six HPT rotations (t-HPT-6), the needles become slightly refined (Figure 2
e and Figure 4
c), their long axis is oriented along the shear direction and the formation of new low and high angle grain boundaries is visible. The hardness increases to 457 HV in t-HPT-6. After ten rotations (t-HPT-10), the microstructure shows strong refinement by SEM (Figure 2
f) with no visible needles anymore. ACOM-TEM reveals an ultrafine grained structure with elongated grains and the cementite is finely dispersed in the ferrite matrix. The final hardness at the fatigue samples extraction point is 758 HV. However, the microstructure and hardness show variations along the radius of the HPT disk due to the shear strain variation (compare Equation (1)). The cementite distribution is most homogeneous after 10 rotations.
After HPT, shear bands are visible by optical microscopy on the surface, fracture surface, and cross section of all specimens. Such shear bands have been reported earlier for HPT processed nanocrystalline metals such as palladium [25
]. Interesting to notice is that, in the present case of the C45 steel, non-metallic inclusion were sometimes sheared by the shear bands as seen in the FIB cross-section in Figure 5
Orientation (a,c,e) and phase (b,d,f) maps (ACOM-TEM) overlaid with the reliability of the tempered microstructures (a,b), after 6 (c,d) and 10 HPT rotations (e,f) all prior to the fatigue tests. The color code of the orientation maps is given in (e). The projection direction is normal to the paper plane. The Fe phase is given in red and green represents the carbide phase. The reliability is given in black for both types of maps.
SEM images of a sheared non-metallic inclusion on the (non-etched) surface of t-HPT-10: (a) top view; (b) FIB cut through the inclusion.
Four-point-bending fatigue tests were carried out with both carbide morphologies in the initial coarse grained (CG) and the HPT deformed states (Figure 6
). The edge stress amplitude was plotted against the number of cycles to fracture for both initial carbide morphologies. The arrows in the squares indicate runouts after 107
cycles. The small numbers beside the arrows show the number of overlapping runout-points. The crack initiation site type is indicated on top of the diagrams. A few specimens could not be investigated (n.I.) after fatigue, e.g., because they did not break completely.
S-N curves of (a) CG and UFG spheroidized SAE 1045 revealing endurance limits of 316 MPa, 493 MPa and 837 MPa and of (b) UFG and CG tempered SAE 1045 showing endurance limits of 640 MPa, 769 MPa and 850 MPa.
a shows the values for the spheroidizing annealed state in the CG (s-ini) and the UFG (s-HPT-6 and s-HPT-10) condition. For the s-ini state, a fatigue limit of 316 MPa was determined by the staircase method. Fractured specimens were observed after more than 10,000,000 cycles at stress amplitude of 325 MPa, while at stress amplitude of 300 MPa only one specimen fractured. In contrast, s-HPT-6 with a hardness of 289 HV showed fatigue fracture at lifetimes below about 400,000 cycles. The endurance limit (494 MPa) was significantly higher compared to s-ini. Crack initiation was mainly at the surface as indicated in the diagram. After 10 HPT rotations, resulting in a hardness of 511 HV, the fatigue limit further increased to 837 MPa. One specimen failed at 800,000 cycles, all other either failed within the first 80,000 cycles or ran out. Below 800 MPa, no fatigue failure was observed for s-HPT-10. The crack initiation site changed with an increasing number of HPT-rotations from the surface without non-metallic inclusions for s-HPT-6 to crack initiation at non-metallic inclusions at the surface in s-HPT-10 (see also Figure 7
b shows the S-N-curves for t-ini, t-HPT-6 and t-HPT-10. A fatigue limit of 640 MPa was determined for t-ini, corresponding to a hardness of 388 HV. In this case, crack initiation mostly occurred at non-metallic inclusions at the surface. With an increased hardness of 457 HV, the fatigue limit of t-HPT-6 rose to 769 MPa. For higher fatigue loads, a specific crack initiation location could not be identified. The exceptions are cyclic edge stresses between 780 MPa and 900 MPa, where all cracks initiated at non-metallic inclusions. For the t-HPT-10 specimens, offering the highest shear strain and hardness, the S-N curves and the crack behavior seem to be completely different. In contrast to the initial state and t-HPT-6, where specimens either failed before 200,000 cycles or ran out, many t-HPT-10 samples failed at a higher number of cycles. In addition, this state also revealed a number of very early cracks, which reduced the fatigue limit to a value of 850 MPa, lower than expected for a hardness of 758 HV, when considering a linear correlation between hardness and fatigue limit. Cracks were mostly initiated at the surface of the t-HPT-10 specimens.
3.3. Fractographic Investigations
All fracture surfaces of the broken specimens were investigated by SEM after the bending fatigue tests. In bending specimens, a stress gradient over the thickness leads to tension on one side, a neutral plane, and compression on the other side. The tension side, where fatigue cracks always start in this kind of material, is always shown at the bottom for the SEM images in this paper.
In the following fractographic investigations, only the HPT-10 states are included because of their large scatter in lifetime compared to all other states.
a shows the fracture surface of one of the s-HPT-10 specimens after a fatigue load of 850 MPa. A large fatigue fracture surface (Figure 7
b) is visible, which is larger than typically found in conventional high strength materials. The fatigue fracture surface shown in Figure 7
b is very homogeneous and appears smooth. The crack initiation occurred at a non-metallic inclusion. In the transition area between fatigue and residual fracture surface, some short cave-like shear bands are visible. Such shear bands are observed in fracture surfaces of HPT processed materials, for example in nanocrystalline palladium after HPT [25
]. In this specimen, shear bands seem not to have influenced the crack initiation or early fatigue crack growth because they are observed only in the late fatigue crack growth areas well as in the residual fracture area.
Fracture surface of s-HPT-10 (a) whole fracture surface (b) fatigue fracture surface; some shear bands resulting from HPT are shown with arrows.
illustrates the fracture surface of a t-HPT-10 sample loaded at a stress of 900 MPa. SEM investigations reveal shear bands in the fracture surface. The cave-like characteristics appear over the whole residual fracture surface. With a length of up to 200 µm, they seem to be larger than those in the spheroidizing annealed state, which are only up to 40 µm long. The fatigue fracture surface in Figure 8
b is smaller than in the s-HPT-10 state although the sample was loaded at nearly the same stress level. We also observe crack initiation at a non-metallic inclusion. The fracture appearance is rather homogenous with a flat fatigue fracture surface. In addition, some shear bands are visible in the fatigue fracture surface but similar to the spheroidizing annealed state, they do not influence the crack initiation or propagation.
Fracture surface of t-HPT-10 (a) whole fracture surface (b) crack initiation area at the surface without nonmetallic inclusion and fatigue fracture surface.
shows a comparison of the residual fracture surfaces of the two states with tempered and spheroidizing annealed microstructure after HPT. There are significant differences in the morphology. In Figure 9
a, s-HPT-10 shows well-defined dimples, which indicate a homogeneous ductile fracture behavior.
Residual fracture surface of (a) s-HPT-10 and (b) t-HPT-10.
In contrast, Figure 9
b reveals a rather rough residual fracture surface with sharp edges, which are common for this material state and indicate a brittle material behavior. Most likely, these sharp edges are shear steps from shear banding. Figure 10
a,b show an FIB cross-section of the t-HPT-residual fracture surface at a characteristic cave, which was formed by a shear band. This indicates a correlation between shear bands observed at the fracture surface and the microstructure underneath. The shear band at the fracture surface is connected to a line-shaped inhomogeneity inside the microstructure. A similar line pattern is also visible in a polished and Nital-etched cross section of the HPT specimen in Figure 10
c prior to fatigue loading. These lines are extended along the shear direction during HPT and become visible after etching in SEM in this case in the t-HPT-10 state. However, similar shear bands can also be seen in the micrographs of the s-HPT-10 states.
SEM images of shear bands in t-HPT-10 samples (a,b) inside the residual fracture surface and (c) on the polished and etched cross section.
Fracture surface of t-HPT-10 state which shows shear bands (a,b) loaded at 900 MPa and (c,d) loaded at 850 MPa (red open arrows at the top indicate the direction of shear bands in correlation to the specimens surface indicated by the yellow open arrows. The crack initiation site is always at the bottom of the pictures).
The fracture surface of a tempered specimen after HPT, which failed very early at 900 MPa, is shown in Figure 11
a,b. The picture shows a cliff-like fracture surface with a high topology in the residual and also in the fatigue fracture surface. The fatigue crack seems to change direction and jump between different levels. Usually, fatigue cracks grow mostly perpendicular to the tensile stress in one defined direction, as visible in Figure 7
and Figure 8
. The fact that this is not the case here is an indication of some inhomogeneities inside the material prior to fatigue loading which affected the crack initiation site and the crack path. The whole fracture surface offers different characteristics of the shear bands compared to the fatigue samples with higher endurance limits. The shear bands seem to influence the crack growth and the lifetime of the fatigue sample because there is a link between the shear bands inside the residual fracture surface and the different levels and layers in the fatigue fracture surface.
A similar behavior is presented in Figure 11
c,d, showing the fracture surface of a t-HPT-10 state sample, which cracked after 6,000,000 cycles. It is obvious that shear bands are present in the area of fatigue crack growth, especially at the crack initiation site, where an extension of a shear bands are points of multiple crack initiation. These shear bands have been produced during HPT and seem to have promoted the crack initiation during fatigue load. Further crack growth in the fatigue fracture surface is similar to that shown in Figure 11
a,b, exhibiting a high topology. The fatigue crack growth is dominated by the shear bands (marked with red arrows in all pictures). The two arrows in each part of Figure 11
indicate that there is an angle of misorientation between the expected shear plane at HPT (yellow arrows) and the shear band plane in the respective specimen (red arrows). This is the case for both specimens, which both exhibited early failure during fatigue testing, and it establishes a significant difference in contrast to samples in the spheroidization annealed or tempered state, which offer a homogeneous fracture behavior in the S-N curve without early cracks (t-HPT-6). For both specimens shown in Figure 11
, a complex structure of different fracture planes is visible in the fatigue fracture surface. The surface of each single plane itself seems to be very flat, as expected for fatigue failure in ultrafine grained materials, resulting in a strong topology with clustering in different spatial directions.
shows the crack initiation site marked in Figure 11
c,d in higher magnification after rotating the specimen by 20° out of plane. Some penetration lines are visible at the specimen surface where shear bands from HPT deformation seem to have been re-activated during fatigue loading. The assumed path of the shear bands inside the specimen is indicated by parallel lines, which fits the direction estimated from the overall fracture surface in Figure 11
Shear bands visible at the sample’s surface in the area of the crack initiation site.
shows cross sections of typical crack initiation sites for the HPT states after ten HPT rotations. BF-TEM and ACOM-TEM images were obtained to correlate the crack path, which is in both cases horizontal in the upper part of the images, with the microstructure underneath. In Figure 13
a, a non-metallic inclusion, located on the left side of the shown area, was responsible for crack initiation in the s-HPT-10 state. Grain size and morphology close to the crack do not differ significantly from the pre-fatigue s-HPT-10 state.
Cross section of the crack initiation (a) non-metallic inclusion for s-HPT-10 and (b) crack path near crack initiation from the surface for t-HPT-10. The overlays are orientation maps in projection direction normal to the paper plane with the reliability in black. The color code of the orientation maps is given in (Figure 3a).
Similarly, no significant grain coarsening or grain refinement has been observed for the t-HPT-10 state in Figure 12
b. Nevertheless, in both cases, the crack path is influenced by the microstructure as it (partly) deviates from a straight crack front when crossing grain boundaries.
The present investigation of CG and UFG medium carbon steels shows a correlation between hardness and endurance limit in bending fatigue tests. Four material states with ultrafine grained microstructure (s-HPT-6, s-HPT-10, t-HPT-6, and t-HPT-10), produced from two different initial carbide morphologies (s-ini + t-ini), were investigated. The refinement of the microstructure during HPT was in accordance with hardness measurements. Both s-ini and t-ini samples revealed a finer grain size and higher hardness with increasing shear strain due to torsional deformation. However, for the samples prepared by six rotations of HPT, the increase in hardness or grain refinement is not as high as expected from the applied shear strain. Nevertheless, it cannot be excluded that the anvils might have been partially sliding over the specimen during HPT deformation, which would explain the limited hardness increase between the initial states and the HPT-6 states.
indicates the relationship between hardness and endurance limit in our investigations. It was created on the basis of the endurance limits presented in Figure 6
. A linear correlation can be seen for the hardness range up to about 500 HV. The fatigue limit primarily correlates with the hardness which itself is strongly affected by the microstructure such as the carbide and grain morphologies. The initial and HPT-6 states fit this correlation very well. Following the investigation by Murakami [26
], there must be a decrease of the fatigue limit with increasing hardness values over about 500 HV. McGreevy analyzed this behavior in [27
] considering the competing roles of microstructure and flaw size.
Relationship between hardness and endurance limit revealing a linear correspondence.
Responsible for a deviation from the linear relationship are inherent material flaws such as non-metallic inclusions, which become more and more dominant at higher hardness. The subsequent decrease of the endurance limit, as suggested in Figure 14
, depends on the material state and processing route. In high strength steels, the so-called process flaws resulting from the respective treatment are responsible for such behavior. For conventional steels, it is mostly the heat treatment which leads to process flaws. In our case, the HPT treatment replaces heat treatment for reaching a high strength. Thus, instead of the traditional process, flaws known from the literature, such as micro-cracks or weakened former austenite grain boundaries in carburized steels [28
], here we are confronted with shear bands or cracks created during HPT, which act as process flaws. Other possible process flaws could not be clearly identified in the metallographic sections after the fatigue testing. Nevertheless, one result of the present investigation is that, in contrast to a material that only exhibits inherent or intrinsic flaws, here the process flaws in the material state with the highest hardness (t-HPT-10) lower the fatigue limit. The S-N curves already provide evidence for this, as contrary to our expectation, no non-metallic inclusions were responsible for crack initiation in the t-HPT-10 state, but the cracks mostly initiated from the surface. This is untypical for high strength steels and indicates that there must be something inside the microstructure that is more inhomogeneous or detrimental than the non-metallic inclusions.
A more detailed investigation of the fracture surfaces confirms that the s-HPT-state exhibits a very homogenous and typical fracture appearance for high strength steels with flat fatigue fracture surfaces. The dimples that are present on the residual fracture surface indicate a ductile fracture behavior. This is known from ultrafine-grained materials and has been linked to good fatigue properties [29
Comparing Figure 8
and Figure 11
, the fracture surfaces of t-HPT-10 do not appear as uniform as those of s-HPT-10. The main difference is that while the fatigue fracture surface in Figure 8
is not affected by inhomogeneities, Figure 11
reveals a strong topology with fracture planes of different directions covering the whole surface, including the fatigue fracture surface. In both cases shown in Figure 11
, the residual fracture surface has shear steps and no dimples indicating a brittle fracture behavior. In this context, Figure 10
provides an opportunity to understand the morphology of the shear bands prior to fatigue testing. Figure 10
a shows an FIB cross-section of one of these cavities, which exhibits the typical fracture appearance of the shear bands [25
]. There is a narrow plane-shaped inhomogeneity beneath the surface inside the microstructure. This could be a link to shear bands generated during HPT as similar stripes are present throughout the microstructure before fatigue loading, as visible in the optical micrograph in Figure 10
c. This observation can be made not only in the brittle tempered state but similarly also in the ductile spheroidizing annealed microstructure with excellent fatigue properties. It follows that s-HPT-10 and t-HPT-10 do contain shear bands prior to fatigue testing but only in the tempered state they affect the fatigue properties. Therefore, the brittle behavior is the main reason for fatigue limits which are lower than expected in consideration of the hardness, because the brittle material has a more limited ability to reduce stress around process flaws owing to plastic flow. In this context, the s-HPT-10 state is an ideal material in respect of fatigue because it exhibits a high hardness and also a high ductility resulting in excellent fatigue performance with higher tolerance against crack initiation at inhomogeneities.
However, an open question is: why do early cracks or cracks at low stress amplitudes occur in some specimens of t-HPT-10? This can be explained comparing Figure 8
and Figure 11
, which both show shear bands in the fracture surface. Shear bands represent inhomogeneities, which, in combination with a brittle material, are detrimental to the fatigue performance, as shown in Figure 14
. However, when comparing fracture surfaces of late and early failure, it is visible that the specimens with the worst fatigue properties exhibit a rotation of the shear bands, which is not parallel to the nominal HPT-shear planes. Unusually turbulent flows, manifesting themselves in the appearance of whirls observed at the macroscopic [30
] and microscopic [31
] scale in severely deformed materials, are most likely the reason for this rotation. This effect seems to increase the influence of the shear bands on fatigue failure.
The rotation of the shear bands increases the possibility for cracks to grow along the shear bands. Investigations from Miller [32
] support this argumentation. He proposed that cracks are always present with every kind of inhomogeneity and the only determining factor for the fatigue resistance is the question whether they reach a critical length that gives them the possibility of further growing. In the case of ultrafine grained high strength steels, grain boundaries, which are the traditional microstructural barrier against crack growth, are not efficient anymore because of the size of the inherent flaws such as non-metallic inclusions. In contrast to the grain size the inclusions do not decrease in size with increasing hardness. The inclusions can be regarded as cracks that are larger than the grain size and cannot be stopped easily by a single grain boundary. Process flaws, such as the identified shear bands, are, depending on their appearance, more detrimental than the non-metallic inclusions. However, important for reaching a critical length is not only the absolute length of a shear band. Crucial is also their direction in relation to the maximum applied stress direction [26
]. With this argument it becomes clear that the rotated shear bands are more critical than those in the HPT planes. The flat fatigue crack path along the shear planes also indicates a higher fatigue crack growth rate, which lowers the lifetime of the samples, as shown in the S-N diagrams in Figure 6
. This observation is known from the literature as well as the fact that the threshold for long crack propagation is lower for favorably oriented shear bands [34
]. Evidence for this explanation and for the crack initiation being affected by the shear bands and their orientation is presented in Figure 12
, where the shear bands are visible on the fracture surface and also at the lower side of the specimen with the highest fatigue stresses. This characteristic appearance has already been reported in literature [35
] and seems to be beneficial for crack initiation.