A. Morphology and dimensions of the melted zone

During the laser treatment processing, the temperature distribution of material cross-section was regular in laser beam direction. According to Li et al.^{Reference Li, Wang, Zhang, Han and Han12} who previously reported the temperature distribution during laser processing by finite element (FE), the isotherms presented an inverted parabola shape and the temperature gradually decreased outward from the heat source, as shown in Fig. 4. By the irradiation of high-energy laser beam, material elements were heated and subsequently resulted in phase transformation. In light of this, the cross-sections of created units (by LM and LC) showed inverted parabola shape, where the top of each unit was wider than its bottom, implying the temperature field feature, as the so-called units shown in Fig. 5. More specifically, a zone was experienced rapidly melting and solidification causing phase transformation while the other zone was undergone phase transformation without melting. According to the temperature field, the region under temperature higher than the melting point was defined as remelting zone (RZ), and the region under temperature lower than the melting point but higher than 727 °C underwent phase transformation without melting, was defined as transition zone (TZ). Additionally, the tempering effect, (confirmed by micro-hardness measurement in previous research^{Reference Jiang, Xue and Wang13}), was overwhelmed in the area under temperature lower than 727 °C. However, the tempering zone was not taken into account in this study, as the scale was very narrow. Therefore, by the changed temperature in cross-section, the distinct regions in units were apparent in visual, including RZ and TZ.

FIG. 4.

The draft cross-section of unit according to Li et al.^{Reference Li, Wang, Zhang, Han and Han12}

FIG. 5.

The actual cross-section morphologies of units with different coatings thickness in laser parametres: pulse duration 5.0 ms, frequency 9 Hz and laser input energy density 6.84 J/mm². (a) without coating; (b) 0.1 mm; (c) 0.2 mm; (d) 0.3 mm.

Table II indicated the dimensions of the laser remelting unit (LMU) and laser cladding unit (LCU). For both LM and LC, because the material element in TZ was not melted and subsequently failed to make a distinct boundary between TZ and the substrate, the measured dimensions were the statistic results of RZ. In comparison, the LMU melt pool was wider but shallower than that of the LCU; however, the area of LMU was apparently smaller than that of the LCU. Among the LCUs, the width was similar in different coating thickness, while the depth and area decreased in inverse proportion to coating thickness. Because the sample's surface was covered with black carbon powder, the covered surface was correspondingly subjected to higher power density as increased laser absorptivity, which resulted in large volume fraction of remelted region on LC sample compared to that on LM sample. These results were in accordance with reports by Benyounis,^{Reference Benyounis, Fakron, Abboud, Olabi and Hashmi14} in which low power density resulted in wider melt zone and little penetration while high power density produced a deep and narrow melted zone. What's more, with respect to the samples treated by LC, the created surface absorptivity and power density were completely identical. However, as known, a portion of heat would be consumed by the carbon coating during the thermal transmission in vertical direction. In light of the regular of LCU dimensions, this can be considered as the reason that the unit depth and area of LCUs were sensitive with the increase of coating thickness. In spite of the heat consumption by coating, the areas of all LCUs were always larger than the area of LMU. In other words, the level of increased unit area caused by increased surface absorptivity was more dominant than that of decreased unit area due to coating heat consumption.

TABLE II.

Dimensions of laser remelted GCI processed in the laser parameters: pulse duration 5.0 ms, frequency 9 Hz and laser input energy density 6.84 J/mm².

B. XRDs and microstructure characteristics

Owing to the high solidification rate of laser processing, the phases formed in the units were detected by means of XRD, and the microstructure characteristics of units were detected by SEM and FESEM. The parameters of major peaks were listed in Table III, in which demonstrated that the common 2-Theta of major peaks was about 44.6°. Nevertheless, slight differences in 2-Theta were an objective probability, due to the existence of residual stress (compressive or tensile) caused by phase transformation.^{Reference Grum and Šturm15} Broadening full width at half maximum (FWHM) of the treated specimens suggested the refinement of grain size, and LCU grain size likely decreased with increasing coating thickness, reaching similar FWHM as the LMU at coating thickness of 0.3 mm. Additionally, almost identical new phases, such as martensite and retained austenite, were demonstrated in the diffractograms (Fig. 6), which performed the role of reinforced phase in LMUs and LCUs. At the duration of laser treatment, the phase transition process could be validated as follows. When the surface got remelted, most of the α-Fe (bcc) transformed to γ-Fe (fcc) by trapping carbon, and subsequently transformed to martensite (M) or retained austenite (Ar) while some α-Fe kept its original phase in the absence of carbon diffusion. Consequently, the LMU microstructure [Fig. 7(a)] was irregular and nonuniform as absent carbon diffusion, which had a negative effect on mechanical properties. Conversely, the microstructures of LCUs were sufficiently uniform and regular in comparison with LMU, as illustrated in Figs. 7(b)–7(d). Due to the fact that external carbon was introduced into melted pool to supplement adequate carbon by thermal stress or laser beam impact, the ferrite in poor carbon region could trap the supplied carbon resulting in the transformation of martensite or austenite. Although the grain size in LCU (No. 5) was similar to that of LMU, LCU particles were more regular, uniform, and compact in comparison with the latter. However, in both LMU and LCU, similar dendrite structures throughout the melt zone were characterized by hypoeutectic dendrites with long primary and secondary arms. There was, additionally, complete dissolution of flake graphite and re-solidification of the hypoeutectic structure in RZ. Nevertheless, the spaces between secondary arms in LCU samples were much smaller than those in LMU. Under high magnification, the interior of the dendrites lacked M or P in the LMU [Fig. 8(a)], in accordance with previous research results,^{Reference Benyounis, Fakron, Abboud, Olabi and Hashmi14} but in LCU showed a considerable amount of martensite and pearlite [Figs. 8(b) and 8(c)].

TABLE III.

Information of lattice plane (110) according to XRD.

FIG. 6.

XRD before and after laser depositing: (a) untreated sample; (b) treated sample by laser remelting; (c) treated sample by LC.

FIG. 7.

Microstructures of units with different coatings thickness by FESEM: (a) without coating; (b) 0.1 mm; (c) 0.2 mm; (d) 0.3 mm.

FIG. 8.

The microstructure of units by SEM: (a) LMU; (b) LCU; (c) interior detail of dendrites by LC.

Figure 9 depicted the details of interfaces between RZ and TZ of LMU and LCUs by FESEM. For the LMU and LCUs, TZ and RZ were clearly observed to be metallurgically bonded, and compositions of cementite (Fe₃C) and abnormal ledeburite (Ld′) were formed in RZ. In addition, as visible in the micrograph of Fig. 9(a), a distinct interface was observed as well as martensite in TZ. With respect to LCUs, martensite content in TZ reduced with increased coating thickness until completely disappearing at 0.3 mm thickness, which instead was full of sorbite in the nominal TZ. Evidenced by these results, the thickness of coating was also considered to affect cooling rate to some extent. When the surface was exposed in the air or with thin coating, wide temperature gradient existed in RZ resulting in high cooling rate that induced the phase transformation. Surfaces with thick coating, conversely, showed relatively low cooling rate and subsequent narrow TZ with less martensite. Once coating increased to a critical value (0.3 mm), cooling rate was reduced even farther, and the martensite completely disappeared in TZ, owing to the fact that the thick coating consumed more heat or played an excellent heat preservation layer. This suggested that TZ formed in thick coating was suffered quenching and tempering effects during heating and cooling; in other words, that it normalized, resulting in the occurrence of sorbite.

FIG. 9.

The boundary of RZ and TZ of units in different coating thickness by FESEM: (a) without coating; (b) 0.1 mm; (c) 0.2 mm; (d) 0.3 mm.

C. Micro-hardness of units

According to refined crystalline strengthening mechanisms, microscopic particle refinement reflected the increased micro-hardness on the macro. Micro-hardness tests were performed on the LCU and LMU cross-sections, showing micro-hardness profiles along the path from the melt surface to the bottom of melt pool (Fig. 10). Interestingly, all plots rose and then dropped down from top to bottom of the melt pool, indicating the oxidation reaction of carbon in the surface or subsurface inducing the formation of decarburization layer. Similar results were observed by Wang and Bergmann.^{Reference Wang and Bergmann16} Another notable aspect of the test results was that micro-hardness significantly increased under laser deposition, and it was more effective by LC, as high as 780–850 HV_0.2 in RZ. Additionally, micro-hardness of both surfaces and subsurfaces of samples with thin coating was consistently higher than those with relatively thick coatings, ranked No. 3 > No. 4 > No. 5. Depending on the refined grain size and the microstructure in units, this phenomenon seemed to be reasonable.

FIG. 10.

The micro-hardness against the depth of units.

D. Improvement of tensile and compressive resistance

Figure 11 indicated the results of tensile and compressive tests executed on samples. Figure 11(a) markedly depicted the stress–strain plots of samples ranged Nos. 1–5. It was clearly seen that all the samples performed inconspicuous yield point and low fracture elongation, implying GCI to be a fragile material. According to the empirical view that yield stress approximated fracture stress for fragile materials, the yield stress of treated samples was subsequently sorted as: No. 3 > No. 4 > No. 5 > No. 2 > No. 1. That can be interpreted that laser deposition (either LC or LM) could significantly improve the yield stress of sample as the refinement of particles, and the improved level was intensively sensitive to grain size (explainable according to Hall–Petch Law^{Reference Hugo, Kung, Weertman, Mitra, Knapp and Follstaedt17}). Additionally, although the unit grain sizes of No. 5 and No. 2 were the same, the former exhibited higher yield stress due to its more compact and uniform crystal compared to the letter, as mentioned above. As a conclusion, the microstructure was another factor affecting the yield stress of sample. What's more, an unexpected result was demonstrated that fracture elongation was reduced by increased micro-hardness among LC treated samples. Even so, the fracture elongations of LC treated samples were consistently more favorable than those of LM treated samples or untreated samples. Figure 11(b) showed the axial displacement of the samples response to axial force during elastic deformation. This plot implied that the abilities of samples' axial deformation resistance had the same sequence with their yield stress. It can be concluded that the ability of compressive deformation resistance was determined by yield stress. Consequently, by the arrangement of refined particles and increased volume fraction of reinforced phase, the both of tensile and compressive deformation resistances were correspondingly improved.

FIG. 11.

The results of mechanical properties tests by MTS: (a) tensile plots; (b) compressive plots.

E. Mechanism of FWR improvement

Figure 12 demonstrated the average ML of each group at the end of test (about 1.26 × 10⁶ cycles). The results clearly showed that laser surface treatment was beneficial for FWR, particularly under LC. In addition, the influence was varying with various coating thickness that the level of improvement reduced in inverse proportion to coating thickness. There were several reasons behind that. Owing to the correspondingly improved tensile and compressive resistance of samples, the FWR enhancement presented in this study was therefore reasonable in accordance with a study by Kapoor,^{Reference Kapoor18} in which the Coffin–Manson Law was used to describe RCF failure ascribed to accumulated plastic deformation (combination of shear strain and axial strain). In the view of fracture, characterized by crack initiation and propagation, FWR was improved to some extent by increased fracture strength, as well. Generally, micro-cracks preferred to occur in graphite, which was usually regarded as the natural hole, suggesting that eliminating graphite can effectively prolong crack initiation.^{Reference Shi, Hu and Zheng5} What is more, the graphite also formed a connection bridge between adjacent cracks, allowing micro-crack propagation in the least energy as possible.^{Reference Chen, Zhou, Zhao, Zhang, Lu, Yang and Zhou19} When cracks propagated in the region without graphite (treated area), the generally developed paths of crack were intergranular and transgranular. By the refinement of crystal, the propagation of crack was more frequent deflection and more difficult to cross fine crystal compared to coarse crystal. Thus, the both ways of crack propagation would increase the absorbed energy, and subsequently improved the fatigue life. Therefore, once the graphite disappeared, crack initiation and propagation was effectively hindered and fatigue life was accordingly enhanced.

FIG. 12.

Average ML of each group ranged Nos. 1–5.

At the end of fatigue wear testings, the worn surfaces of samples were examined by optical microscope, as shown in Fig. 13. According to the worn surface profiles, large scale defects formed on the surface of the untreated sample [Fig. 13(a)], while relatively slight damage occurred on the substrate of treated samples [Fig. 13(b)]. Here, in terms of the substrate of treated samples, it can be concluded that its fatigue life was significantly prolonged as the existence of unit. The explanation had been stated by FE analysis in elsewhere^{Reference Chen, Zhou, Zhao, Zhang, Lu, Yang and Zhou19} that dominant contact stress occurred on the unit region resulting in the corresponding reduction of burden on the substrate. What's more, owing to the distinct microstructures of units, the released levels were remarkably various, resulting in the different level of FWR improvement. On the other hand, even though the unit was subjected to the major load, no prominent damage [Fig. 13(c)] was observed on unit parts under high magnification (500x), attributed to the reinforcement of unit zones. Differences were further demonstrated by 3-D profile scanning [Fig. 13(d)], in which severe damages were observed on the substrate but almost no defects were presented on the unit region in contrast.

FIG. 13.

The worn surface: (a) PM; (b) matrix of treated sample; (c) unit region; (d) contrast of unit and matrix by 3-D profile scanning.

F. Mechanism of damage formation

Different types of defects were visible on the worn surface, caused by a variety of factors. In Fig. 14(a), delamination, cracks, and pitting were observed on the region with high roughness, as the dominant representation of rolling fatigue failure. Commonly, these types of damages were the typical result of crack propagation. Differently, with respect to the nonlinear propagation of cracks, closed and unclosed cracks were presented. Unclosed cracks indicated the interrupted propagation processing, and as a result, the material element in the semiclosed region was removed from the surface. The phenomenon was clearly observable in the micrograph under high magnification [Fig. 14(b)]. In addition, metallic flakes were also observable on the surface where metallic tongue occurred (see Ref. Reference Franklin, Weeda, Kapoor and Hiensch20), because the formed cracks at subsurface reached the surface at a specific angle. By the development of these metallic flakes, a gradient surface was subsequently appeared, named as “delamination.” As shown in Fig. 14(b), two forceful cracks encountered at Region I, making the region more brittle and apt to develop into a defect-prone region. As cycle progressed, the defect at the intersecting region fanned out and made a ragged bottom, resulting in the typical spalling [Fig. 14(c)]. In addition, an apparent altitude intercept was revealed at the both side of semiclosed crack, in which the inside area of the crack was visibly lower than the outside one. This implied that the large sections of metal had been removed from the sample's surface, as a performance of the formation of typical delamination. In Region II, a crack branched from the major crack took circular form and prevented further damage by circular pitting [Fig. 14(b)]. Differently, circular pitting resulted from the propagation of a single crack was characterized by a flat bottom under high magnification [Fig. 14(d)]. So far, the reasons of cracks propagating in various paths had not been reported, and the authors conjectured that it was significantly affected by the stress intensity factor of crack.

FIG. 14.

The surface damages on the region with high surface toughness.

As far as relevant formation mechanisms, the fanned-out defect was caused by the interaction of two forceful cracks, while the circular defect was a result of the nonlinear propagation of a single crack. However, by the reduction of surface toughness, a completely different mechanism was responsible for the surface damage formation. As illustrated, the propagated direction of primary cracks were perpendicular [Fig. 15(a)] and parallel [Fig. 15(b)] to the rolling direction, respectively. The common continuous metal removal occurred along the one side of cracks during the processing of linear propagation (marked by the white dash lines in figures), which made the cracks look broader. To find out the details of the defect, a high-resolution defect was indicated in Fig. 15(c). As shown, this defect was characterized by an extended bottom to the contact surface from the bottom of the major crack at a certain angle. The result was in accordance with the found under lubricated condition in previous research, in which the initial crack was first directed toward the depth at an inclination of about 20°; when I _branch > 1, a branch began to rise toward the contact surface inclined by about 90°, causing metal removal (see Ref. Reference Donzella, Faccoli, Ghidini, Mazzù and Roberti21). It should be noted that the type of defect was only observed during linear crack propagation. In contrast, during nonlinear propagation processing, it was caused in response to the damages as shown in Fig. 14. The exact mechanism at work during this process was not fully distinguishable, but will be examined further in our future research.

FIG. 15.

The surface damages on the region with low surface toughness.