Densification, microstructure evolution and fatigue behavior of Ti-13Nb-13Zr alloy processed by selective laser melting

Powder Technology 342 (2019) 11–23

Contents lists available at ScienceDirect

Powder Technology journal homepage: www.elsevier.com/locate/powtec

Densification, microstructure evolution and fatigue behavior of Ti-13Nb-13Zr alloy processed by selective laser melting Libo Zhou a, Tiechui Yuan a,⁎, Ruidi Li a,⁎, Jianzhong Tang b, Guohua Wang b, Kaixuan Guo b, Jiwei Yuan c a b c

State Key Laboratory of Powder Metallurgy, Central South University, Changsha 410083, PR China Zhuzhou Printing Additive Manufacturing Co. LTD, Zhuzhou 412000, PR China Guizhou R&D Center of Titanium Materials Co. LTD, Zunyi 563004, PR China

a r t i c l e

i n f o

Article history: Received 23 January 2018 Received in revised form 6 September 2018 Accepted 24 September 2018 Available online 26 September 2018 Keywords: Selective laser melting Ti-13Nb-13Zr Densification Microstructural Fatigue behavior

a b s t r a c t Ti-13Nb-13Zr alloy was fabricated by selective laser melting (SLM), and the densification, microstructure evolution, nanohardness, tensile strength and fatigue behavior of the alloy were systematically investigated. A narrow, feasible process window (laser power 325 W, scanning speed 1000 mm/s, scanning distance 0.13 mm, and layer thickness 0.03 mm) was accordingly determined. With the increase of scanning speed, the coarse acicularshaped α′ grains changed to refined acicular-shaped α′ grains and then defect formed in the microstructure. The optimally prepared Ti-13Nb-13Zr sample had a very high hardness of 519.448 HV and tensile strength of 1106.07 MPa, which are superior higher than that prepared by traditional thermomechanical technique (280 ± 15 HV and 732 MPa). Owing to the change of BCC to HCP structure ([111] β→ [1-21-3] α′), the accumulation of dislocations and the finer grains, the SLM-processed samples exhibited higher fatigue strength than that the alloy processed by cast and was commensurate with the cast Ti-6Al-4 V. Moreover, the influences of phase transition on fatigue behavior have been discussed carefully. © 2018 Elsevier B.V. All rights reserved.

1. Introduction Titanium and its alloys are widely used in medical applications because of their unique properties, such as corrosion resistance [1], mechanical strength [2,3] and superior biocompatibility [4–6]. Ti-6Al4 V alloy was the first Ti alloy used as biomaterial which was originally developed for aerospace applications [7]. However, in recent years, there has been a lot of negative comments about vanadium ion and aluminum ion on causing health problems like cytotoxic effects, adverse tissue reaction and neurological disorders [5,7]. Moreover, the mismatch of elastic modulus between Ti-6Al-4 V (120 GPa) and human bone (30 GPa) could lead to a resorption of the adjacent bone tissue [4,5]. The elastic modulus of Ti-13Nb-13Zr (65 GPa), compared with the elastic modulus of Ti-6Al-4 V (120 GPa), is closer to that of human bone [8]. Consequently, the stress-shielding effects and mismatch problem are reduced. What's more, all the three constituents in the Ti13Nb-13Zr alloy meet the criterial for biomaterials in terms of biocompatibility, mechanical consideration, resistance to corrosion, and ionic cytotoxicity [9,10]. Therefore, in recent years, Ti-13Nb-13Zr alloy has attracted a growing number of researchers. Most of works reported on the process of Ti-13Nb-13Zr alloy are focused on the severe plastic deformation (SPD) or arc melting. Baptista ⁎ Corresponding authors. E-mail addresses: [email protected] (T. Yuan), [email protected] (R. Li).

https://doi.org/10.1016/j.powtec.2018.09.073 0032-5910/© 2018 Elsevier B.V. All rights reserved.

et al. [11] studied the fatigue behavior of arc melted Ti-13Nb-13Zr alloy and the result turns out that Ti-13Nb-13Zr shows high fatigue resistance when compared with the CP Ti and annealed Ti-6Al-4 V. Lee et al. [12] researched the microstructure tailoring of Ti-13Nb-13Zr to enhance the strength and ductility for biomedical applications. Urbańczyk et al. [13] demonstrated that the corrosion resistance of Ti-13Nb-13Zr alloy could be improved through the plasma electrolytic oxidation (PEO) process. Geetha et al. [14] revealed that the even distribution of alloying elements in Ti-13Nb-13Zr alloy bring superior corrosion resistance. Bobbili et al. [15] investigated the dynamic recrystallization behavior of Ti-13Nb-13Zr and found that increasing the deformation temperature and decreasing the strain rate can promote the process of dynamic recrystallization. Ivana et al. [16] investigated the metallic ion release of ultrafine-grained Ti-13Nb-13Zr alloy processed by high pressure torsion (HPT) and they found that the number of released ions of the Ti-13Nb13Zr alloy processed by HPT was higher than that produced by traditional casting. Suresh et al. [17,18] researched the equal channel angular extrusion (ECAE) of the Ti-13Nb-13Zr alloy and confirmed that corrosion resistance of the ultrafine-grained alloys obtained by ECAE was higher than the non-treated material. All of the forming methods discussed above can only process standard implants. However, as a biocompatible metal, Ti-13Nb-13Zr can be processed into the complex shapes of functional implants is very important. Among different possible methods to manufacture complex construction, selective laser melting is a specifically versatile and advantageous

12

L. Zhou et al. / Powder Technology 342 (2019) 11–23

way [19]. Using a laser to melt and solidify the required geometry layer-by-layer, SLM is able to produce intricate construction from metal powder materials [20]. SLM possesses a number of advantages over conventional manufacturing techniques, such as high material utilization rate, direct production based on the no geometric constraints mode [21,22]. The studies about SLM processed titanium and its alloys have attracted much attention [23,24]. Nevertheless, to the best knowledge of the authors, there are still no comprehensive previous studies focused on the fatigue behaviors of SLM processed sample, which are very important for biomedical materials. In this study, in order to fully illuminate the fatigue behaviors of SLM processed sample, the densification, microstructure and mechanical properties in terms of hardness, tensile strength and fatigue behavior, various parameters of SLM processing Ti-13Nb-13Zr were investigated. 2. Experimental procedures The powder was gas atomized from a Ti-13Nb-13Zr ingot and had a chemical composition as listed in Table 1. The differences between the analyzed composition of alloying elements and the nominal composition are very small, b0.4% for Nb, and 0.2% for Zr, respectively. The oxygen and nitrogen concentrations were too low to influence the alloy performance. The powder is spherical (Fig. 1(a)) and its particle size distribution is also shown in the figure as an inset. All the samples were fabricated using a SLM apparatus (FS271M Farsoon, Inc., China) under a high-purity Ar atmosphere. The machine was equipped with a 480 W Yb:YAG fiber laser, which had a spot size of 80 μm. Cubes (10 mm × 10 mm × 10 mm) and standard tensile and fatigue samples were fabricated using a laser power (P) from 175 to 375 W and a scanning speed (v) from 800 to 1200 mm/s. The layer thickness (d) and scanning distance (h) were kept constant at 30 μm and 130 μm, respectively. The Eq. (1) [25] E ¼ P=vhd

ð1Þ

was used to evaluate the laser energy that input to the powder layer during process. Each layer was alternated by 60° and scanned using a continuous laser mode shown in Fig. 1(b).

Table 1 Analyzed chemical compositions of Ti-13Nb-13Zr alloy (mass fraction, %). Element

Ti

Nb

Zr

O

N

Content

Balance

12.66

13.12

0.038

0.013

The Archimedes' method was applied to measure the density of specimens. Phase analysis was performed by X-ray diffraction (XRD, Rigaku D/MAX-2250) with a Cu Kα radiation. Metallographic samples were etched by the Kroll's reagent, which were prepared according to the standard procedures. The optical microscope (Germany Leica), a FEI Quanta FEG 250 filed emission gun scanning electron microscope (FEG-SEM) and a JEOL 2100F transmission electron microscope (TEM) were used to observe the microstructures of the samples. The samples for TEM observation were prepared under the following conditions: electrolytic polishing with a mixture of 60% methanol, 34% normal butanol and 6%perchloric acid at 243 K in temperature and 23 V in voltage. Four samples were prepared for tensile tests under each condition using an Instron 3369 machine with a crosshead speed of 1 mm/min. A loading–unloading test mode was used with a force of 30 mN and holding time of 10 s. The recorded load and indentation depth data were used to build the loading–unloading plots. The indentations were verified by atomic force microscope (America Veeco AFM). The fatigue behavior was tested under the condition of 20 Hz with a stress ratio R = 0.1 according to DIN 50113. The samples were cycled with constant stress amplitude until failure and the S\\N curves were plotted. The initial load was set at a maximum level estimated at about 80% yield strength (YS). To investigate the details and the mode of fracture, the fatigue fractography observations were performed using SEM. The tensile direction of the specimens in the tensile and fatigue tests in this work is in the scanning direction. 3. Results 3.1. Optimisation of processing parameters The SLM processing parameters highly affect the quality of products. In this work, the laser power and the scanning speed were systematically varied. The parameters were set as follows: laser power 175–375 W; scanning speed 800–1200 mm/s; fixed layer distance 0.03 mm and scanning distance 0.13 mm. Fig. 2(a) indicates this parameter window mainly covers the low energy density (E b 70 J/mm3), medium energy density (70 J/mm3 b E b 90 J/mm3) and high energy density (E N 90 J/mm3). Fig. 2(b) shows the typical surface morphologies of the SLM-processed Ti-13Nb-13Zr samples including three regimes at a relatively low magnification. In the SLM process, laser scanning was performed line by line and the laser energy melt a row of powder particles and formed a continuous liquid track in cylindrical shape [25]. The surfaces of Ti-13Nb-13Zr samples processed by SLM show no apparent pores and cracks for all tests. At relatively low energy density (Area I in Fig. 2(a)), several clusters of micro-sized balls were formed on the surface which increased the roughness of the product significantly. The liquid solidification front became considerably disordered which results in the formation of interrupted scan tracks and the formation

Fig. 1. (a) The morphology and particle size distribution of the starting Ti-13Nb-13Zr powder, (b) Schematic diagram of scanning mode.

L. Zhou et al. / Powder Technology 342 (2019) 11–23

13

Fig. 2. Schematic of process parameters (a) and corresponding morphologies of surfaces (b). The labeled zones I: energy density lower than 50 J/mm3, II: energy density between 50 and 100 J/mm3, III: energy density higher than 100 J/mm3.

of large-sized balls on the tracks (Figures a and b in Fig. 2(b). Interestingly, increasing the energy density to point c, the surface showed a regular and clear liquid solidification and no any balls. Stable and continuous scan tracks with sound inter track bonding were formed (Figure c in Fig. 2(b)). At an even high energy density (Points d and e in Fig. 2(a)), the balls were observed at the liquid solidification front again (Figures d and e in Fig. 2(b)), which indicates the microscopic scale balling effect. A close examination of Fig. 2(b) revealed that significant balling and turbulent liquid solidification front occurred at too high scanning speed and resultant low energy density, while too low scanning speed and resultant high energy density also resulted in the microscopic balling phenomenon. During the SLM process, when localized powder is melted, a significant temperature gradient will be formed in the molten pool, which will result in surface tension gradient and resultant Marangoni convection [26]. When the convection form within the pools, the thermocapillary force will increase, which results in instability of the flow. The faster applied scanning speed induces the more significant Marangoni convection and the more unstable liquid stream. The radially inward flow was formed as the liquid stream flows from the low surface tension area to the high surface tension area [27]. The spheroidize phenomenon appeared at the center of the laser beam due to the radially inward flow (Fig. 2(b) a). Because of the presence of the balls, the liquid stream is significantly disturbed which results in the interruption after solidification (Fig. 2(b) a and b). The lower scanning speed induces the longer dwelling time of laser beam stay on the surface of molten pool and subsequently a longer liquid lifetime, which leads to the more obvious overheating of the liquid flow and increases the instability of the melt pool. Because of the reduction of surface energy, small-sized liquid droplets tend to splash from the liquid front being solidified (Fig. 2(b) d and e) [25].

3.2. Densification As we all know, the densification of SLM-processed samples has a significant effect on its performance [28]. Fig. 3 shows the influence of energy density on the relative density of SLM samples produced with varying laser powers (175–375 W) and scanning speeds (800–1200 mm/s) under fixed hatch distance 0.13 mm and scanning distance 0.03 mm. Fig. 3(a) shows the relationship between the relative density and the energy density of the SLM-fabricated samples and the blue circles A, B, C and D represent the samples processed with a laser power of 325 W and a scanning speed of 900 mm/s, 1000 mm/s, 1100 mm/s and 1200 mm/s, respectively. The corresponding microstructures (the blue circles in Fig. 3(a)) of cross-sections are shown in Fig. 3(b). At an energy density input of 69.44 J/mm3 (P = 325 W, v = 1200 mm/s), a small number of micropores with a size of ~5 μm formed in the interlayer (Fig. 3(b)(A)), thereby a rather low relative density of 97.19% was achieved. When the scanning speed decreased from 1200 to 1000 mm/s (P = 325 W), the energy density input increased to 83.33 J/mm3, which resulted in the relative density increasing from 97.19% to 97.89%. The corresponding microstructure showed that there are no more pores or cracks in the metallurgically bonded layers (Fig. 3(b)(C)). At an even lower scanning speed of 900 mm/s and high energy density of 92.59 J/mm3, irregular-shaped defects were present in the cross-sectional microstructures (Fig. 3(b)(D)), producing 2.47% porosity in the SLM sample. The laser power has similar influence as scanning speed on the density of the SLM-processed samples. For instance, an increase of laser power from 175 W to 325 W (v = 1000 mm/s) led to the relative density increasing from 96.35% to 97.89%. However, at an even higher laser power of 375 W, the relative density decreased to 97.16% (Fig. 3(a)). From Fig. 3(a), it can be seen that when the energy density input was between 30 and 70 J/mm3, the relative density of

Fig. 3. (a) Influence of energy density on relative density of SLM samples produced with different laser powers and scanning speeds, (b) OM images of SLM-processed samples with laser power P = 325 W under different scanning speeds: (A) v = 1200 mm/s; (B) v = 1100 mm/s; (C) v = 1000 mm/s; (D) v = 900 mm/s.

14

L. Zhou et al. / Powder Technology 342 (2019) 11–23

Fig. 4. XRD patterns of raw powder and SLM-processed Ti-13Nb-13Zr samples with various conditions obtained in wide range of 2θ (a) and details of angle (36°–42°) (b), (52°–60°) (c).

the samples was in the range of 94.4% to 97.4%. When the energy density was in the range of 90–130 J/mm3, the relative density varied between 97.1%–97.4%. The densest samples (relative density 97.4%– 97.8%) can be obtained when the energy density input was 70–90 J/mm3. The micropores are formed when the energy density is low because of the discontinuous scan tracks caused by the balling effect [28]. Furthermore, since SLM is a layer-by-layer process, the balling effect makes the fresh powder are hardly to deposited uniformly on the previously melted layer [29]. The higher the scanning speed applied, the more obvious the phenomenon observed. It is difficult to fill the interball pores on the surface of the melted layer completely, thus interlayer micropores are formed and a limited densification response occurred (Fig. 3 point (A)). On the other hand, the formation of microcracks caused by thermal cracks at a high energy density is ascribed to the residual thermal stresses [28]. The conversion of loose powder to dense liquid leads to the densification in SLM process. Once the powder has been fully melted, shrinkage occurs rapidly. The relationships between shrinkage and residual stress during SLM have been introduced by Dai and Shaw [30]. They revealed that the primary mechanism of residual stresses during SLM is the thermal contraction. Furthermore, Zhu et al. [31] illustrated that when increasing the energy density, the thermal shrinkage increased. Therefore, huge thermal stress tends to generate and accumulate in the SLM-processed samples during solidification. The residual stresses include tensile stresses that distributed at the top or bottom of the sample, and the compressive stresses at the central [25,28]. In this situation, microcracks (shown in Fig. 3(b) (D)) are typically caused by the tensile stresses. In general, those parameters are considered to be suitable for production of bulk samples since the continuous single-tracks can be formed under these conditions. The effect of processing parameters on block material properties worth further investigation. Therefore, the samples were processed at a constant laser power of 325 W and varied scanning speeds between 900 mm/s and 1200 mm/s with fixed hatch distance of 0.13 mm and scanning distance of 0.03 mm, which cover the three areas in Fig. 2. Thus, to discuss the details of the performance of SLM-processed Ti-13Nb-13Zr, careful phase identification, microstructure observation and mechanical properties tests will be performed in the following study. 3.3. Phase analysis and microstructure evolution The preparation process of Ti-13Nb-13Zr alloy by SLM involved complete melting and solidification. The solidification behaviors of the molten pool including thermal history, solidification rate and liquid flow determine the phase transformation in the final solidified materials [25,32]. In order to understand the phase transformation in the SLM-fabricated Ti-13Nb-13Zr, in the present study, the XRD patterns

of the raw powder and the SLM-fabricated Ti-13Nb-13Zr samples with various processing parameters in the 2θ range of 10°–80° are depicted in Fig. 4(a). After SLM process, the intensities of α′(101) and α′(002) decreased while the intensity of β(110) increased, and a new phase β(200) peak appeared, which indicated that the majority of α′ martensite phases have in situ decomposed into α′ and βα, owing to the repetitive melting and solidification in the SLM process [33,34]: L→β→β þ α 0 →β þ α 0 þ βα Note: L presents the liquid, β presents the β phase crystallized from the liquid directly and βα presents the β phase formed by the decomposition of α′ during the repetitive melting. The detailed XRD patterns in the 2θ angle of 36°–42°, as depicted in Fig. 4(b), revealed that the diffraction peaks for BCC β (110) and HCP α′ (101) apparently deviated from the standard diffraction peaks for BCC Ti (β(110) located at 2θ = 38.481°) and HCP Ti (α′ (101) located at 2θ = 40.170°) which were marked by broken lines. The 2 θ values of α′ (002) in SLM-fabricated samples are lower than that in powder because of the solid solutioning process. Fig. 4(b) shows that a broadened peak appeared as a shoulder on the right side of β (110) peak. It is believed that the broadened peak is linked to the nano-domain structure which was formed at high cooling rate [35]. When the energy density was 83.33 J/mm3, the diffraction peaks for BCC β (110) became considerably broadened, which implied the refined microstructures formed under this condition. The detailed XRD patterns in the 2θ range from 52° to 60° are depicted in Fig. 4(c), which showed that the location of β (200) peak shifted to lower 2θ value with increasing the energy density (Table 2), indicating that the lattice constant of the β phase increased with the energy density. According to Bragg's law,

2d sin θ ¼ nλ ðn ¼ 1; 2; 3; …Þ the decrease of 2θ value at high energy density (Fig. 4(c)) implies an increase of the lattice plane distance d, which is considered to be caused by the solubilized Nb and Zr. Increasing the energy density leads to a longer liquid lifetime that will enhance the solubilized of Nb and Zr. The microscopic volume expansion which induces the stress that Table 2 XRD data on the displacement of identified peak β (200) with various energy densities.

2θ location (°)

Standard (PDF no. 44–1288)

E = 69.44 J/cm3

E = 75.76 J/cm3

E = 83.33 J/cm3

E = 92.59 J/cm3

55.541

55.513

55.392

55.331

55.216

L. Zhou et al. / Powder Technology 342 (2019) 11–23

15

Fig. 5. (a) XRD PFs showing the crystallographic orientation of the SLM-processed Ti-13Nb-13Zr with E = 83.33 J/mm3, (b) ODF sections with phi2 from 0° to 90° of the texture components of β phase of sample processed by E = 83.33 J/mm3.

influences the lattice parameters on the grain boundaries usually occurs in the solid solution [35]. The β textures of material processed by E = 83.33 J/mm3 are presented by {110}, {200} and {211} pole figures in Fig. 5(a). All texture components of sample are concentrated around building direction. The maximum intensities are 2.98, 1.98 and 1.48 times of the random intensities in the {110}, {200} and {211} pole figures, respectively. Fig. 5(b) shows the detailed β phase texture analysis on sample processed by E = 83.33 J/mm3 with the ODF sections of φ2 from 0° to 90°. Obviously, select distribution of texture orientations are presented with the maximum texture index of sheet texture in 3.92 × random at φ2 b 10° or φ2 N 80° and fiber texture in 3.26 × random at φ2 = 45°. What's more, the texture component shown in Fig. 5(b) is a cycle and with a center of symmetry (φ2 = 45°), which can be attributed to the select distribution of texture [34]. Fig. 6 shows the microstructures of SLM-processed Ti-13Nb-13Zr samples under various processing conditions. The large β grains oriented along the building direction, are marked by dotted lines (Figs. 6 (a), (b) and (c)), which are consistence with the results obtained in Fig. 5(a). The elongated columnar grains formed with length up to 100 μm (Fig. 6(a) and (b)). This observation displayed that the grain size in the building direction was greater than the layer thickness (30 μm), which indicated the remelting- resolidification of the previous layer and a diffusion phenomenon. The melt pool boundaries were clearly visible in the SEM images. Most grains grew across the boundaries of melt pool towards the top center of each melting track. Fischer et al. [34] obtained the similar growth pattern, and studied the in situ elaboration of Ti\\26Nb alloy by SLM. At a relatively low scanning speed of 900 mm/s and resultant high energy density of 92.59 J/mm3, relatively coarse acicular-shaped grains which were identified as α′ martensite by XRD (Fig. 4(a)) were presented (Fig. 6(a)). As the scanning speed

increased to 1000 mm/s, the crystalline structures of martensitic α′ phase were refined (Figs. 6(b)). At an even higher scanning speed (≥1100 mm/s), the grain size of martensitic had no significant decrease, while the shaped entrapped gases were observed (Fig. 6 (c) and (d)). The gas in the gaps of the powder particles is dissolved in the melting pool and may still exist after solidification due to the rapid cooling. [25,28]. What's more, because of the intense laser beam, the melting pool temperature is generally high, which results in the increased solubility of the gas in the liquid, and thus the gas enrichment is promoted. During the SLM process, the rapid movement of laser beam leads to rapid melting and fast solidification of Ti-13Nb-13Zr alloy. The larger cooling rate induces the greater undercooling effect, and thus the grain size will be finer [34]. The cooling rate of the material is extremely high during the SLM process because of the large scanning speed (900–1200 mm/s). The high scanning speed leads to the high heating/ cooling rate (103–108 K/s) [28,36], resulting in very fine microstructure of Ti-13Nb-13Zr and increased possibility of developing nonequilibrium phases with fine grained microstructures [37]. In this work, when the scanning speed was 900 mm/s, the thermal accumulation caused by the energy thermalization occurred within the molten pool, therefore, the quenching had no significant effect on the solidification rate, leading to the formation of relatively coarse martensite (Fig. 7 (a)). When the scanning speed is increased to 1000 mm/s, the temperature gradient rises and the grains will be refined obviously (Fig. 7(b)). Further increasing the scanning speed to 1100 mm/s or 1200 mm/s, the size of martensite with no significant change. Taking the results discussed above into consideration, it can be deduced that the solidification rate and undercooling degree increase with the laser scanning speed, and the microstructure of SLM-processed Ti-13Nb-13Zr is changed as follows: relatively coarsened acicular-shaped α′ grains →

16

L. Zhou et al. / Powder Technology 342 (2019) 11–23

Fig. 6. SEM images showing characteristic microstructures of SLM-processed Ti-13Nb-13Zr samples under different processing parameters: (a) E = 92.59 J/mm3, v = 900 mm/s; (b) E = 83.33 J/mm3, v = 1000 mm/s; (c) E = 75.76 J/mm3, v = 1100 mm/s; (d) E = 69.44 J/mm3, v = 1200 mm/s.

refined acicular-shaped α′ grains → acicular-shaped α′ grains with spherical pores caused by the entrapped gas. 3.4. Mechanical properties To understand the influence of process parameters on the mechanical properties of SLM-produced Ti-13Nb-13Zr, nanoindentation tests, tensile tests and fatigue behaviors were performed on the samples. Fig. 8 depicts the nano-indentation load–depth curves measured on the polished sections of samples from the top view. The indentation depths of the samples processed with reasonable energy density of 83.33 J/mm3 (532.33 nm) was lower than those of the samples prepared with inappropriate energy densities (550.15 nm at 75.76 J/mm3, 568.03 nm at 92.59 J/mm3 and 579.28 nm at 69.44 J/mm3). Figs. 8 (c) and (d) show the AFM images of indentation of the samples processed at E = 83.33 J/mm3 and E = 69.44 J/mm3, respectively. It was found that indentation depth of the sample processed at E = 83.33 J/mm3 was lower than that of the sample processed at E = 69.44 J/mm3, which was coincided with the load–depth curves (Fig. 8 (a)). The Vickers hardness values of 519.448 HV for the sample processed at E = 83.33 J/mm3 was the highest, as revealed in Fig. 8(b). More importantly, all of the SLM-processed Ti-13Nb-13Zr samples including those fabricated by inappropriate energy density showed much higher Vickers hardness values than the sample processed by equal channel angular extrusion (ECAE) technique (145 ± 5 HV) [17] and traditional thermomechanical processing technique (280 ± 15 HV) [6]. The large residual stress in the SLM-processed sample and the grain refinement resulting from the large cooling rate are responsible for this enhancement of microhardness [38,39]. Fig. 9 presents the variation of tensile strength for the SLMprocessed Ti-13Nb-13Zr samples. It shows that the strains of the samples processed at a low energy density of 69.44 J/mm3 (2.3%) and a high energy density of 92.59 J/mm3 (2.8%) were lower than those of the samples prepared at the energy densities (3.5% for 75.76 J/mm3

and 3.1% for 83.33 J/mm3), indicating the improvement of densification (Fig. 3(a)). The corresponding tensile strength values (1053 MPa at 69.44 J/mm3 and 1064 MPa at 92.59 J/mm3) were lower than those obtained at energy density 83.33 J/mm3 (1106 MPa) and 75.76 J/mm3 (1085 MPa), as also revealed in Fig. 9. As previously discussed, the impertinent of energy density is detrimental due to the presence of some detrimental phenomena, such as balling phenomenon (Figs. 2 (b) points a and b), spherical pores caused by the entrapped gas (Fig. 6 (c) and (d)) and grain coarsening (Fig. 7). The relationship between grain size and strength (Hall-Petch relationship) has been described elsewhere [40]. Nevertheless, all of the SLM-processed Ti13Nb-13Zr samples, whatever the process parameters of the samples are, had a higher tensile strength than the Ti-13Nb-13Zr alloy processed by the conventional powder metallurgy (PM) which with a typical tensile strength of 732 MPa [12] and had a commensurate value with those processed by eight-pass caliber rolling (1020 ± 20 MPa) [41] and dynamic globularization (1119 MPa) [42] (Table 3). In general, the significant grain refinement resulting from the laser high cooling rate favors an increase in the tensile strength. Fig. 10 shows the typical tensile fractographies of the SLM-processed Ti-13Nb-13Zr samples. It can be inferred from Fig. 10(a) that some incompletely melted Ti-13Nb-13Zr powders and micro-porosities were the main causes of early fracture of the sample processed at E = 69.44 J/mm3. Generally, the number of defects turn out to be the largest along the building direction in the additive manufactured samples [43]. The energy density of 69.44 J/mm3 was not sufficient enough to melt the powder completely, resulting in the weak connection between the layers. Consequently, the sample is fractured at an early stage during the plastic deformation. Isolated and randomly distributed micropores can be observed in the fracture surface (Fig. 10(a)). There were also some smooth areas related to dimple fracture and quasi-cleavage fracture (Fig. 10(a)). As the energy densities reached values of 83.33 J/mm3 and 75.76 J/mm3, bonding between the grains and layers was improved remarkably and the fracture surface displayed much

L. Zhou et al. / Powder Technology 342 (2019) 11–23

17

Fig. 7. TEM images of SLM-processed Ti-13Nb-13Zr samples at different processing parameters: (a) E = 92.59 J/mm3, v = 900 mm/s; (b) E = 83.33 J/mm3, v = 1000 mm/s; (c) E = 75.76 J/mm3, v = 1100 mm/s; (d) E = 69.44 J/mm3, v = 1200 mm/s.

Fig. 8. (a) Loading–unloading curves of samples processed with various SLM parameters, (b) Vickers hardness (HV) values of samples processed by various SLM parameters, AFM images of indentation of samples processed with various SLM parameters: (c) E = 83.33 J/mm3; (d) E = 69.44 J/mm3.

18

L. Zhou et al. / Powder Technology 342 (2019) 11–23

Fig. 9. Tensile true strain-true stress curves for SLM-processed Ti-13Nb-13Zr samples.

more dimples and less quasi-cleavage facets (Figs. 10(b) and (c)). The fracture surface showed the mixture of dimples in varying sizes, randomly quasi-cleavage facets and isolated microvoids (Fig. 10(b) and (c)). The nucleation, growth and coalescence of the micropores occurred in the tensile process, which would result in the initiation of microscopic cracks. After the tension further rises, the microscopic cracks tend to propagate in the direction of the stress axis and eventually induce the detrimental effect of strain to failure [39]. On the premise of sufficiently high densification and no cracks or pores were formed (Fig. 3), the higher scanning speed led to the higher residual stress in the specimens, which could enhance the tensile properties [44,45]. The grain refinement which is achieved by the transformation of BCC structure into HCP structure induces the strengthening via the formation of Ti martensite, [44]. The dislocation density increase in the walls of boundaries is and other reason for the strengthening [44]. Moreover, the higher scanning speed resulted in a higher cooling rate, which caused a reduction in the grain size, and thus affected the mechanical properties [44]. This can be explained why the tensile strength of samples obtained by the E = 83.33 J/mm3 shows little higher than E = 75.76 J/mm3. At an even high energy density of 92.59 J/mm3, because of the overheating and thermal accumulation, the voids and superburning were observed on the fractography (Fig. 10(d)). Owing to the presence of defects, the sample is fractured at the initial stage of plastic deformation, during which the plastic strain can't be accumulated (Fig. 9). The fatigue tests were performed with the stresses of 650, 550, 450, 350 and 250 MPa at room temperature and frequency of 20 Hz. The

Table 3 Tensile properties of Ti-13Nb-13Zr processed by different technologies. Condition

Micro-structure

Yield strength (MPa)

Tensile strength (MPa)

Elongation (%)

Reference

SLM Swaging WRQa WRAb CAc MPCRd DGe

α′ + β α′ α + α′ + α″ + β α + α″ + β – α+β α+β

887 ± 10 510 820 1000 850 935 ± 10 1010

1106 ± 10 732 1000 1100 980 1020 ± 12 1119

3.1 ± 0.3 30 13 7 6 7.8 ± 0.1 8.4

This work [12] [13] [13] [13] [41] [42]

a b c d e

Warm caliber-rolled material after air-cooling. Warm caliber-rolled material after water-quenching. Ti-13Nb-13Zr solution treated and subsequently aged following the ASTM F1713-08. Eight-pass caliber-rolling. Dynamic globularization.

obtained S-N curves are shown in Fig. 11(a). It is observed that, similar to the tensile test results, the sample obtained at the energy density of 83.33 J/mm3 had the best fatigue properties among the four conditions. The fatigue strength of the four SLM-processed Ti-13Nb-13Zr samples decreases continuously as the number of cycles to failure increases. From the Fig. 11 [46], it is observed that the SLM-processed Ti-13Nb13Zr samples exhibited higher fatigue strength than the Ti-13Nb-13Zr alloy processed by cast and had the commensurate value with the Ti-6Al-4 V alloy. Fatigue test data must be interpreted by the results of these properties in terms of microstructural evolution, hardness and tensile strength. Niinomi et al. [47] suggested that the increase of the tensile strength of an alloy can improve its resistance to fatigue crack initiation and expansion, which in turn increase the fatigue strength. The finer α′ martensite forms during SLM processing, leading to higher strength and hardness, and finally reformed fatigue properties [44]. The reduced sizes of grains in bimodal structures act as barriers to the fatigue crack propagation, which can delay the final fatigue fracture [44,48]. Generally, the smaller the grain size, the better the fatigue resistance will be. Thus, the enhanced fatigue strength of the sample obtained at the energy density of 83.33 J/mm3 is due to the finer grains in microstructure (Fig. 7(b)), higher hardness (Fig. 8) and strength (Fig. 9). Considering the mechanical properties shown in Figs. 8–11, the fatigue strength has direct relationship with the hardness and tensile property, which are determined by the microstructures. It can be concluded that the fatigue strength of bimodal microstructures can be optimized by controlling the microstructural features such as grain size and martensite phase volume fraction. Most fatigue fractures consist of three steps: nucleation of the crack at the appropriate location, propagation of crack resulting from the coalescence of micro-pores, and final fracture that takes place when the cracked cross section cannot withstand the cyclic loading anymore [46,48]. This vulnerability is caused by many defects upon the surface of materials. The defects are originated from fatigue cracks that grow at the same time until they are joined together and induce the final failure. Although mechanical polishing can remove some obvious external defects, surface defects are remained because of the occasional voids appear in the material removal process. Fig. 12 shows the SEM images of the fatigue fracture surfaces. Like other fatigue fractures [46], the fracture of SLM processed Ti-13Nb-13Zr sample started from the surface (Figs. 12(a) and (c)) and micro-voids (Fig. 12(b)). Lin et al. [46] studied the fatigue behavior of Ti-13Nb-13Zr prepared by casting, and revealed that on the cleavage type fracture surface in the crack propagation zone, a “river pattern” was often observed, which was a common characteristics of fatigue failure in a microscopic scale. While in the present studies, the fatigue striations (as shown in the circle of the Fig. 12(d)) are not as obvious as the fracture surface of Ti-13Nb-13Zr processed by casting, indicating that the striations on the fracture surface of SLM-processed Ti-13Nb-13Zr are apparently finer than those processed by casting [49]. It was reported that the presence of surface/surface pores induced by SLM-process, the location of pores and the nature properties of materials are three factors that significantly affect the fatigue performance [46,50]. The presence of microvoids can lead to stress concentration at the wall of pores (Fig. 12(b)). Microvoids were created and joined together to make crack propagation, when the cycle stress was applied. During the fatigue damage process, cycle stress is acted on both the surface and internal defects, and it is believed that the selection of the main crack initiation point is always competitive, which depends on the size and the stress level of the potential crack point [44]. In Fig. 12(c), some flat facets have the same size and shape as lamellas. It was reported that crack nucleation mostly happened in (0001) α, and crack propagation paralleled to the texture of (0001) α, then, the crack coalescence occurred and eventually failure happened [51,52]. Shear processes cross the lamella are difficult but are relatively easy to parallel to the lamellar interfaces [48]. Therefore, the crack propagation is paralleled to the layered interfaces, leading to the formation

L. Zhou et al. / Powder Technology 342 (2019) 11–23

19

Fig. 10. Fracture surface morphologies of SLM-processed Ti-13Nb-13Zr samples: (a) E = 69.44 J/mm3; (b) E = 75.76 J/mm3; (c) E = 83.33 J/mm3; (d) E = 92.59 J/mm3.

of facets in the fracture surface, which deteriorates the fatigue resistance of the material. 4. Discussion Nowadays, there is no clear picture of how the phase transition during the SLM process influences the fatigue behavior. As discussed above, the high cooling rate and the transition of BCC into HCP structure resulted in the grain refinement, and the dislocation density increasing in the walls of boundaries could increase the tensile strength and hardness which directly affect the fatigue strength. The detailed transformation mechanism of α′ and β phases in the SLMprocessed Ti-13Nb-13Zr at E = 83.33 J/mm3 was investigated by TEM observation and was illustrated in Fig. 13. The bright-field image (Fig. 13(a)) indicated that the β phase is the matrix and α′ phase is randomly dispersed in the β matrix. The result was consistent with the XRD measurement (Fig. 4). Fig. 13(b) presents the selected area electron diffraction (SAED) pattern. Different planes of β and α′ crystals can be

detected from the diffraction rings, which implies its polycrystalline structure [53]. From the SAED patterns d-spacings of 0.232 nm, 0.136 nm and 0.076 nm were measured, based on the standard XRD diffraction pattern (JCPDS Cards No. 44–1288), which correspond to (110), (211) and (411) β. In addition, a d-spacing of 0.123 nm implying (201) α′ and a d-spacing of 0.065 nm implying (114) α′ were obtained, based on the standard XRD diffraction patterns (JCPDS Cards No. 44–1294 and 51–0631). Figs. 13(c) and (d) show the SAED patterns of β phase along the [012] zone axis and α′ along the [2–1–10] zone axis in Fig. 13(a), respectively. High resolution transmission electron microscopy (HRTEM) was also performed on zones A, B and C in Fig. 14(a) in order to understand the transformation mechanism of α′ and β phases. From Fig. 14(a), the α′ and β phases can be easily found. The fast Fourier transform patterns (inset in Fig. 14(a)) determined the zone axis of [111] and [1–21–3], which allowed the direct imaging of BCC and HCP stacking, respectively. Moreover, one of the orientation relationship of α′ and β can be expressed as [111] β// [1-21-3] α′. According to the Ti-Nb binary

Fig. 11. (a) S-N fatigue curves of SLM-processed Ti-13Nb-13Zr samples and obtained from Ref [46]. Inset of (b) shown the photo of fatigue samples of Ti-13Nb-13Zr processed by SLM.

20

L. Zhou et al. / Powder Technology 342 (2019) 11–23

Fig. 12. Fatigue fracture surfaces of SLM processed Ti-13Nb-13Zr sample at E = 83.33 J/mm3.

Fig. 13. (a) Bright-field TEM image of SLM-processed Ti-13Nb-13Zr samples at E = 83.33 J/mm3; (b) SAED pattern obtained from (a); (c) SAED pattern of matrix β phase along [012] zone axis; (d) SAED pattern of α′ phase along [2-1-10] zone axis.

L. Zhou et al. / Powder Technology 342 (2019) 11–23

21

Fig. 14. HRTEM images showing β and α phases (a), two β grains interface (b), and stacking-fault (c), magnified inversed fast Fourier transform image (d) showing stacking sequence of the SFs, surrounded by red box in Fig. 14(c). The inset FFTs are those of the boxed regions. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

phase diagram (not shown), the solidification pathway is most likely generated as L → L+[111] β and [111] β→ [1-21-3] α′ and then the α′ phases sized in dozens of nanometers are randomly separated out in the β matrix. The HRTEM image of the interface between the two grains is illustrated in Fig. 14(b). Meanwhile, the fast Fourier transform patterns (inset in Fig. 14(b)) illustrate the [1-21-3] α′ and [111] β. The plane spacings were calculated to be 0.107 nm for [1-21-3] α′ and 0.11 nm for [111] β. It shows that when the BCC structure transformed to the HCP structure, the lattice shrink amounted to +1.03%, which verified the result that changing the BCC into HCP structure could fine grains. Furthermore, the α′ precipitate phases are nucleated preferentially at the grain interface, which is in accordance with the results in Ref. [54]. The phase transformation can probably be induced by the L → L+[111] β → [111] β+ [1-21-3] α′. The β phase transforms into the α′ phase orderly and the α′ phase precipitates at the β boundaries. However, due to the difference between the interfaces spacings, the mismatch caused by the dislocations occurs between the phase interfaces [55]. During the SLM process, because of the high cooling rate, a huge thermal gradient can be generated, which in turn causes the formation of residual stress inside the Ti-13Nb-13Zr alloy, therefore resulting in a high density of dislocations [56,57]. Multiple partial dislocations nucleated and propagated, leaving behind the stacking faults as shown in Fig. 14(c). The magnified inverse fast Fourier transform image (Fig. 14(d)) shows the transformation of BCC to local HCP stacking sequence with the faults. Other obstacles generated in the partial dislocation interaction hinder the dislocation motion, resulting in the further hardening process. The stable stacking-fault defects formed by the partial dislocation interaction can result in the significant strain hardening

effect. Besides, the movement of the dislocations will be hindered by the stacking-fault defects [58]. Due to the fine grains and the changing of BCC into HCP structure, the SLM-produced Ti-13Nb-13Zr alloys show much higher fatigue strength than the cast Ti-13Nb-13Zr alloy.

5. Conclusions The densification, microstructural evolution and mechanical properties including nanohardness, tensile strength and fatigue behavior of the SLM-produced Ti-13Nb-13Zr alloy have been studied. The main conclusions are drawn as follows. (1) At a lower laser scanning rate (v) of 900 mm/s and a higher energy input (E) of 92.59 J/mm3, the microstructure of the obtained alloy showed microcracks and balling phenomenon because of the thermal stress and long liquid duration. While at a higher v of 1200 mm/s and a lower E of 69.44 J/mm3, the micropores formed due to the Marangoni convection and balling effects. Both the defects lowered the densification of the alloy. The relative density of (97.8 ± 0.5) % for Ti-13Nb-13Zr alloy was obtained at a suitable process window (1000 mm/s, 83.33 J/mm3). (2) The microstructures of the SLM-processed Ti-13Nb-13Zr alloy evolved as follows: relatively coarsened acicular-shaped α′ grains → refined acicular-shaped α′ grains → acicular-shaped α′ grains with spherical pores caused by the entrapped gases, with the decrease of E, owing to the high cooling rate and thermal and kinetic undercooling.

22

L. Zhou et al. / Powder Technology 342 (2019) 11–23

(3) The highest hardness (519.448 HV) and tensile strength (1106 MPa) were obtained at the reasonable process parameters, which were higher than those of the alloy processed by traditional thermomechanical technique (280 ± 15 HV) and (732 MPa), because of the formation of finer grains. (4) The SLM-processed Ti-13Nb-13Zr sample exhibited higher fatigue strengths than that processed by cast and had commensurate values with the cast Ti-6Al-4 V alloy, owing to the change of BCC into HCP phase (L → L+[111]β → [111]β + [1-21-3]α′) and finer grains as well as the accumulation of dislocations. Acknowledgements The authors are grateful for the financial support from the National Key R&D Program of China (grant Nos. 2017YFB0305401, 2016YFB1100101), Guizhou Science and Technology Plan Project (grant No. GHK support [2017]2013), Zhuzhou Printing Additive Manufacturing Co. LTD and the National Natural Science Foundation of China (grant Nos. 51571214, 51474245, 51871249) for this work. References [1] M. Long, H.J. Rack, Titanium alloys in total joint replacement-a materials science perspective, Biomaterials 19 (1998) 1621–1639. [2] Y. Song, D. Xu, R. Yang, et al., Theoretical study of the effects of alloying elements on the strength and modulus of β-type bio-titanium alloys, Mater. Sci. Eng. A 260 (1999) 269–274. [3] Z. Cai, T. Shafer, I. Watanabe, et al., Electrochemical characterization of cast titanium alloys, Biomaterials 24 (2003) 213–218. [4] J.A. Davidson, A.K. Mishra, P. Kovacs, et al., Biocompatible low modulus titanium alloy for medical implants, Bio-Med. Mater. Eng. 4 (1994) 231–243. [5] M. Geetha, A.K. Singh, K. Muraleeharan, et al., Effect of thermomechanical processing on microstructure of a Ti–13Nb–13Zr alloy, J. Alloys Compd. 329 (2001) 264–271. [6] L. Ren, K. Memarzadeh, S.Y. Zhang, et al., A novel coping metal material CoCrCu alloy fabricated by selective laser melting with antimicrobial and antibiofilm properties, Mater. Sci. Eng. C 67 (2016) 461–467. [7] M.A. Khan, R.L. Williams, D.F. Williams, The corrosion behaviour of Ti–6Al–4V, Ti– 6Al–7Nb and Ti–13Nb–13Zr in protein solutions, Biomaterials 20 (1999) 631–637. [8] M. Geetha, A.K. Singh, R. Asokamani, et al., Ti based biomaterials, the ultimate choice for orthopaedic implants – a review, Prog. Mater. Sci. 54 (2009) 397–425. [9] M.A. Hussein, C. Suryanarayana, N. Al-Aqeeli, Fabrication of nano-grained Ti–Nb–Zr biomaterials using spark plasma sintering, Mater. Des. 87 (2015) 693–700. [10] J.A. Davidson, P. Kovacs, Biocompatible Low Modulus Titanium Alloy for Medical Implants, US, 1992. [11] C.A.R.P. Baptiasta, S.G. Schneider, E.B. Taddei, et al., Fatigue behavior of arc melted Ti–13Nb–13Zr alloy, Int. J. Fatigue 26 (2004) 967–973. [12] T.Y. Lee, Y.U. Heo, C.S. Lee, Microstructure tailoring to enhance strength and ductility in Ti–13Nb–13Zr for biomedical applications, Scripta Mater. 69 (2013) 785–788. [13] E. Urbańczyk, A. Krzakala, W. Simka, et al., Electrochemical modification of Ti– 13Nb–13Zr alloy surface in phosphate based solutions, Surf. Coat. Tech. 291 (2016) 79–88. [14] M. Geetha, U.K. Mudali, A.K. Gogia, et al., Influence of microstructure and alloying elements on corrosion behavior of Ti–13Nb–13Zr alloy, Corros. Sci. 46 (2004) 877–892. [15] R. Bobbili, V. Madhu, Dynamic recrystallization behavior of a biomedical Ti–13Nb– 13Zr alloy, J. Mech. Behav. Biomed. Mater. 59 (2016) 146–155. [16] I. Dimić, I.C. Alagić, B. Völker, et al., Microstructure and metallic ion release of pure titanium and Ti–13Nb–13Zr alloy processed by high pressure torsion, Mater. Des. 91 (2016) 340–347. [17] K.S. Suresh, M. Geetha, C. Richard, et al., Effect of equal channel angular extrusion on wear and corrosion behavior of the orthopedic Ti–13Nb–13Zr alloy in simulated body fluid, Mater. Sci. Eng. C 32 (2012) 763–771. [18] K.S. Suresh, N.P. Gurao, S. Singh, et al., Effect of equal channel angular pressing on grain refinement and texture evolution in a biomedical alloy Ti-13Nb-13Zr, Mater. Character 82 (2013) 73–85. [19] M. Ni, C. Chao, X.J. Wang, et al., Anisotropic tensile behavior of in situ precipitation strengthened Inconel 718 fabricated by additive manufacturing, Mater. Sci. Eng. A 701 (2017) 344–351. [20] R.D. Li, M.B. Wang, T.C. Yuan, et al., Selective laser melting of a novel Sc and Zr modified Al-6.2 Mg alloy: processing, microstructure, and properties, Powder Tech 319 (2017) 117–128. [21] R.D. Li, P.D. Niu, T.C. Yuan, et al., Selective laser melting of an equiatomic CoCrFeMnNi high-entropy alloy: processability, non-equilibrium microstructure and mechanical property, J. Alloys Compd. 746 (2018) 125–134. [22] G. Yablokova, M. Speirs, J.V. Humbeeck, et al., Rheological behavior of β-Ti and NiTi powders produced by atomization for SLM production of open porous orthopedic implants, Powder Tech 283 (2015) 199–209.

[23] H. Attar, M. Bonisch, L.C. Zhang, et al., Selective laser melting of in situ titanium– titanium boride composites: processing, microstructure and mechanical properties, Acta Mater. 76 (2014) 13–22. [24] H. Attar, K.G. Prashanth, L.C. Zhang, et al., Comparison of wear properties of commercially pure titanium prepared by selective laser melting and casting processes, Mater. Lett. 142 (2015) 38–41. [25] D.D. Gu, Y.C. Hagedorn, W. Meiners, et al., Densification behavior, microstructure evolution, and wear performance of selective laser melting processed commercially pure titanium, Acta Mater. 60 (2012) 3849–3860. [26] A. Simchi, H. Pohl, Effects of laser sintering processing parameters on the microstructure and densification of iron powder, Mater. Sci. Eng. A 359 (2003) 119–128. [27] H.J. Niu, I.T.H. Chang, Instability of scan tracks of selective laser sintering of high speed steel powder, Scripta Mater. 41 (1999) 1229–1234. [28] H. Attar, M. Calin, L.C. Zhang, et al., Manufacture by selective laser melting and mechanical behavior of commercially pure titanium, Mater. Sci. Eng. A 593 (2014) 170–177. [29] T. Gustmann, A. Neves, U. Kuhn, et al., Influence of processing parameters on the fabrication of a Cu-Al-Ni-Mn shape-memory alloy by selective laser melting, Addit. Manufact. 11 (2016) 23–31. [30] K. Dai, L. Shaw, Thermal and stress modeling of multi-material laser processing, Acta Mater. 49 (2001) 4171–4181. [31] H.H. Zhu, L. Lu, J.Y.H. Fuh, Study on shrinkage behaviour of direct laser sintering metallic powder, Mech. Eng. 220 (2006) 183–190. [32] M. Fischer, D. Joguet, G. Robin, et al., In situ elaboration of a binary Ti–26Nb alloy by selective laser melting of elemental titanium and niobium mixed powders, Mater. Sci. Eng. C 62 (2016) 852–859. [33] W. Xu, M. Branda, M. Qian, et al., Additive manufacturing of strong and ductile Ti– 6Al–4V by selective laser melting via in situ martensite decomposition, Acta Mater. 85 (2015) 74–84. [34] W. Li, Y. Yang, J. Liu, et al., Enhanced nanohardness and new insights into texture evolution and phase transformation of TiAl/TiB2 in-situ metal matrix composites prepared via selective laser melting, Acta Mater. 136 (2017) 90–104. [35] L.S. Wei, H.Y. Kim, S.C. Miyazaki, Effects of oxygen concentration and phase stability on nano-domain structure and thermal expansion behavior of Ti–Nb–Zr–Ta–O alloys, Acta Mater. 100 (2015) 313–322. [36] L. Thijs, K. Kempen, J.P. Kruth, et al., Fine-structured aluminium products with controllable texture by selective laser melting of pre-alloyed AlSi10Mg powder, Acta Mater. 61 (2013) 1809–1819. [37] M. Das, V.K. Balla, D. Basu, et al., Laser processing of SiC-particle-reinforced coating on titanium, Scripta Mater. 63 (2010) 438–441. [38] E.C. Santos, K. Osakada, M. Shiomi, et al., Microstructure and mechanical properties of pure titanium models fabricated by selective laser melting, Mech. Eng. 218 (2004) 711–719. [39] L.B. Zhou, T.C. Yuan, R.D. Li, et al., Selective laser melting of pure tantalum: densification, microstructure and mechanical behaviors, Mater. Sci. Eng. A 707 (2017) 443–451. [40] A.V. Sergueeva, V.V. Stolyarov, R.Z. Valiev, et al., Advanced mechanical properties of pure titanium with ultrafine grained structure, Scripta Mater. 45 (2001) 747–752. [41] T.Y. Lee, K.T. Park, C.S. Lee, et al., Microstructural evolution and strain-hardening behavior of multi-pass caliber-rolled Ti–13Nb–13Zr, Mater. Sci. Eng. A 648 (2015) 359–366. [42] C.H. Park, J.W. Park, C.S. Lee, et al., Enhanced mechanical compatibility of submicrocrystalline Ti–13Nb–13Zr alloy, Mater. Sci. Eng. A 527 (2010) 4914–4919. [43] L.C. Zhang, D. Klemm, J. Eckert, et al., Manufacture by selective laser melting and mechanical behavior of a biomedical Ti–24Nb–4Zr–8Sn alloy, Scripta Mater. 65 (2011) 21–24. [44] N. Davari, A. Rostami, S.M. Abbasi, Effects of annealing temperature and quenching medium on microstructure, mechanical properties as well as fatigue behavior of Ti6Al-4V alloy, Mater. Sci. Eng. A 683 (2017) 1–8. [45] M.J. Donachi, Titanium a technical guide, ASM Int. (2003) 238–267. [46] C.W. Lin, C.P. Ju, J.H.C. Lin, A comparison of the fatigue behavior of cast Ti–7.5Mo with c.p. titanium, Ti–6Al–4V and Ti–13Nb–13Zr alloys, Biomaterials 26 (2005) 2899–2907. [47] M. Niinomi, Fatigue performance and cyto-toxicity of low rigidity titanium alloy, Ti– 29Nb–13Ta–4.6Zr, Biomaterials 24 (2003) 2673–2683. [48] M.T. Jovanoic, S. Tadic, S. Zec, The effect of annealing temperatures and cooling rates on microstructure and mechanical properties of investment cast Ti–6Al–4V alloy, Meter. Des. 27 (2006) 192–199. [49] C.R. Books, A. Choudhury, Fracture mechanisms and micro-fractographic features, in: C.R. Brooks, A. Choudhury (Eds.), Metallurgical Failure Analysis, McGraw-Hill, New York, 1993. [50] J.C. Ting, F.V. Lawrence, Fatigue Fract. Eng. A crack closure model for predicting the threshold stresses of notches, Mater. Struct. 16 (1993) 93–114. [51] M.R. Bache, W.J. Evans, Impact of texture on mechanical properties in an advanced titanium alloy, Mater. Sci. Eng. A 319 (2001) 409–414. [52] K.L. Biavant, S. Prommier, C. Prioul, Local texture and fatigue crack initiation in a Ti6Al-4V titanium alloy, Fatigue Fract. Eng. M 25 (2002) 527–545. [53] W. Li, J. Liu, S.F. Wen, et al., Crystal orientation, crystallographic texture and phase evolution in the Ti–45Al–2Cr–5Nb alloy processed by selective laser melting, Mater. Charact. 113 (2016) 125–133. [54] Z.B. Zhao, Q.J. Wang, Q.M. Hu, et al., Effect of β (110) texture intensity on α-variant selection and microstructure morphology during β→α phase transformation in near α titanium alloy, Acta Mater. 126 (2017) 372–382.

L. Zhou et al. / Powder Technology 342 (2019) 11–23 [55] C.C. Chen, C. Zhu, E.R. White, et al., Three-dimensional imaging of dislocations in a nanoparticle at atomic resolution, Nature 496 (2013) 74–77. [56] D. Tomus, Y. Tian, P.A. Rometsch, et al., Influence of post heat treatments on anisotropy of mechanical behaviour and microstructure of Hastelloy-X parts produced by selective laser melting, Mater. Sci. Eng. A 667 (2016) 42–53.

23

[57] D.Y. Zhang, W. Niu, X.Y. Cao, et al., Effect of standard heat treatment on the microstructure and mechanical properties of selective laser melting manufactured Inconel 718 superalloy, Mater. Sci. Eng. A 644 (2015) 32–40. [58] Z.J. Zhang, M.M. Mao, J.W. Wang, et al., Nanoscale origins of the damage tolerance of the high-entropy alloy CrMnFeCoNi, Nat. Commun. 6 (2015) 10143.