Evaluation of Low Cycle Fatigue Performance of Selective Laser Melted Titanium Alloy Ti–6Al–4V

: The material of Ti–6Al–4V has been widely applied in various industries, such as automobile, aerospace, and medical due to its high speciﬁc strength, superior thermal stability and strong corrosion resistance. In the recent decades, selective laser melting (SLM) has become an attractive method to fabricate Ti–6Al–4V parts, thanks to its signiﬁcant advantages in low material consumption, the high degree of freedom in design, low carbon footprint, etc. Predictability of SLM material fatigue properties is especially important for the safety-critical structures under dynamic load cases. The present research is aimed at evaluating the low cycle fatigue (LCF) performance of SLM Ti–6Al–4V under high loading states. LCF tests were performed for as-built and annealed SLM Ti–6Al–4V. Comparison between LCF properties of SLM Ti–6Al–4V and the wrought Ti–6Al–4V was also made. It was found that as-built SLM Ti–6Al–4V demonstrated a comparable LCF performance with the wrought material. The LCF life of as-built SLM Ti–6Al–4V was longer than that of wrought Ti–6Al–4V at lower strain amplitudes. However, the wrought Ti–6Al–4V had better LCF performance at higher strain amplitudes. The results revealed that the porosity in the as-built SLM material exerted much more impact on the degradation of the material at high strain amplitudes. Annealing deteriorated the LCF performance of SLM Ti–6Al–4V material due to the formation of coarser grains. The cyclic Ramberg–Osgood and the Basquin–Co ﬃ n–Manson models were ﬁtted to depict the cyclic stress–strain and the strain–life curves for the SLM Ti–6Al–4V, based on which the LCF performance parameters were determined. In addition, the fatigue fracture surfaces were observed by using scanning electron microscopy (SEM), and the results indicated that fatigue cracks originated from the surface or subsurface defects. This investigated the LCF performance of SLM Ti–6Al–4V materials, in which a set of monotonic tensile tests and strain-controlled LCF tests were performed on both as-built and annealed SLM Ti–6Al–4V samples. The cyclic softening / hardening properties of the SLM Ti–6Al–4V were characterized by stress–strain hysteresis loops and the progression curves of the stress amplitude with cycles. The cyclic stress–strain and strain–life curves of SLM Ti–6Al–4V were ﬁtted according to the Ramberg–Osgood and Basquin–Co ﬃ n–Manson models, respectively. In addition, the LCF failure mechanism was analyzed using scanning electron microscopy (SEM). The results were compared with those of the wrought Ti–6Al–4V specimens.


Introduction
Selective laser melting (SLM), as an additive manufacturing technology method, focuses on direct transformation and fabrication from a three-dimensional computer-aided design (3D-CAD) model to fully functional ready-to-use metal components [1]. During the SLM process, the metal powder This paper investigated the LCF performance of SLM Ti-6Al-4V materials, in which a set of monotonic tensile tests and strain-controlled LCF tests were performed on both as-built and annealed SLM Ti-6Al-4V samples. The cyclic softening/hardening properties of the SLM Ti-6Al-4V were characterized by stress-strain hysteresis loops and the progression curves of the stress amplitude with cycles. The cyclic stress-strain and strain-life curves of SLM Ti-6Al-4V were fitted according to the Ramberg-Osgood and Basquin-Coffin-Manson models, respectively. In addition, the LCF failure mechanism was analyzed using scanning electron microscopy (SEM). The results were compared with those of the wrought Ti-6Al-4V specimens.

Selective Laser Melting of Test Samples
In the present study, the printed samples were manufactured by a SLM machine EOSINT M280 using the gas-atomized Ti-6Al-4V powder whose chemical composition content is accordance with the ISO 5832-3, ASTM F1472, and ASTM B348 titanium alloy standards. The chemical composition of the powder is listed in Table 1. To eliminate the Ti-6Al-4V's oxidation, the oxygen level of the chamber was controlled under 0.1%. In order to avoid curved warping caused by thermal stress, the powder bed was preheated to 35 • C. The samples were made based on the EOS recommended SLM processing parameters shown in Table 2. All specimens were printed by using a stripe pattern scanning strategy so as to reduce the thermally induced stresses [19,20]. The rotation angle of the stripe direction is 67 • between consecutive layers, as indicated in Figure 1a, which resulted in a more isotropic microstructure and a higher density [21]. SLM Ti-6Al-4V samples. The cyclic softening/hardening properties of the SLM Ti-6Al-4V were characterized by stress-strain hysteresis loops and the progression curves of the stress amplitude with cycles. The cyclic stress-strain and strain-life curves of SLM Ti-6Al-4V were fitted according to the Ramberg-Osgood and Basquin-Coffin-Manson models, respectively. In addition, the LCF failure mechanism was analyzed using scanning electron microscopy (SEM). The results were compared with those of the wrought Ti-6Al-4V specimens.

Selective Laser Melting of Test Samples
In the present study, the printed samples were manufactured by a SLM machine EOSINT M280 using the gas-atomized Ti-6Al-4V powder whose chemical composition content is accordance with the ISO 5832-3, ASTM F1472, and ASTM B348 titanium alloy standards. The chemical composition of the powder is listed in Table 1. To eliminate the Ti-6Al-4V's oxidation, the oxygen level of the chamber was controlled under 0.1%. In order to avoid curved warping caused by thermal stress, the powder bed was preheated to 35°C. The samples were made based on the EOS recommended SLM processing parameters shown in Table 2. All specimens were printed by using a stripe pattern scanning strategy so as to reduce the thermally induced stresses [19,20]. The rotation angle of the stripe direction is 67° between consecutive layers, as indicated in Figure 1(a), which resulted in a more isotropic microstructure and a higher density [21].

Heat-Treatment and Post-Processing
As shown in Figure 1(b), the printed samples are hexagon rods with a length of 120 mm and diameter of the inscribed circle of 13mm. They were divided into two groups randomly, i.e., the annealed and as-built. Samples in the annealed group were further processed by heat-treatment (800 °C for 2 h) and furnace cooled. All printed samples in the two groups were then machined to LCF test samples as shown in Figure 2. Precisely speaking, on each of test sample, the LCF test was applied

Heat-Treatment and Post-Processing
As shown in Figure 1b, the printed samples are hexagon rods with a length of 120 mm and diameter of the inscribed circle of 13 mm. They were divided into two groups randomly, i.e., the annealed and as-built. Samples in the annealed group were further processed by heat-treatment (800 • C for 2 h) and furnace cooled. All printed samples in the two groups were then machined to LCF test samples as shown in Figure 2. Precisely speaking, on each of test sample, the LCF test was applied in the middle section with a length of 12 mm and diameter of 6mm. To ensure the LCF test quality, the average surface roughness of the test section on each sample was polished to 0.32 µm. In addition, to establish a benchmark for test results comparison and analysis, a group of wrought samples with the identical dimension in Figure 2 were produced.

Microstructure Observation
Samples in each of three test groups (two SLM-based, one wrought) were randomly selected for microstructure observation by using a Neophot 30 optical microscope (Carl Zeiss, Jena, Germany). For observation, the selected samples were processed through three preparation stages sequentially: grinding, polishing, and etching (1 mL HF + 2 mL HNO 3 + 50 mL H 2 O).

Phase Identification
Phase analysis was conducted by a D/Max 2500PC X-ray diffraction (XRD) system (Rigaku, Japan) on the as-built and annealed SLM specimens using a Cu Kα-source with a scan speed of 4 • /min and a step size of 0.02 • .

Monotonic Tension Test
From each of the three groups (two SLM-based, one wrought), three samples were randomly selected, on which tensions tests were conducted in line with the ASTM E8-09 standard. The rate of monotonic tension test was 1 mm/min.

Low Cycle Fatigue (LCF) Test
Strain-controlled LCF tests were performed under different strain amplitudes (ε a ) as shown in Table 3. 10 samples were used, including 4 as-built, 4 annealed, and 2 wrought samples. The loading ratio was R = −1. During the test, the strain applied to the test samples was varied by a triangle wave-based control and the frequency was 0.1 Hz, as shown in Figure 3. The experiments were conducted on an MTS 100 kN servo-hydraulic closed-loop test machine (MTS, Eden Prairie, MN, USA) at room temperature. The uniaxial strain was measured by Epsilon 12 mm extensometer with a resolution of 0.1 µm. Table 3. Strain amplitudes ε a in (%) of low cycle fatigue (LCF) testing.

As-Built SLM
Annealed SLM Wrought

Low Cycle Fatigue (LCF) Fractography Observations
The LCF fractures of SLM samples were analyzed using a TESCAN VEGA SEM (Tescan, Brno, Czech Republic) to investigate the failure mechanism. Figure 4 presents the respective microstructure of the as-built SLM, annealed SLM, and wrought samples before LCF test. Figure 4(a) indicates that the fine acicular α' martensite is formed during SLM processing, which results from the alloy's rapid melt and solidification due to the short interaction between laser and powder and high energy input [5,6,22]. Based on a simulation of the temperature field of SLM processing, Li et al. [23] observed that the cooling rate of Ti-6Al-4V in SLM significantly exceeds the critical cooling rate of 410 K/s required for martensitic transformation (β to α') of Ti-6Al-4V [24]. The microstructure of annealed samples is shown in Figure 4(b), which reveals that the acicular microstructure becomes even coarser than that of the as-built. Compared with needle shape lamellas observed in the SLM samples, the wrought samples have a typical α-β two-phase microstructure, as shown in Figure 4(c), wherein the β phases are distributed around the coarse platelet α phases. Figure 5 shows XRD patterns for as-built and annealed SLM Ti-6Al-4V. All peaks of the two types of samples are characteristic of hexagonal close-packed (hcp) structure of Ti, which indicates that the microstructure is dominated by α'/α phase even after annealing. In this study, the annealing treatment was carried out below β transus temperature and did not result in obvious decomposition of the as-fabricated α' phase. Therefore, β phase peaks are absent in the annealed SLM Ti-6Al-4V. The peak intensities in the annealed condition are significantly higher compared to the pattern of the as-built conditions, indicating coarser grains in the annealed samples.

Low Cycle Fatigue (LCF) Fractography Observations
The LCF fractures of SLM samples were analyzed using a TESCAN VEGA SEM (Tescan, Brno, Czech Republic) to investigate the failure mechanism. Figure 4 presents the respective microstructure of the as-built SLM, annealed SLM, and wrought samples before LCF test. Figure 4a indicates that the fine acicular α martensite is formed during SLM processing, which results from the alloy's rapid melt and solidification due to the short interaction between laser and powder and high energy input [5,6,22]. Based on a simulation of the temperature field of SLM processing, Li et al. [23] observed that the cooling rate of Ti-6Al-4V in SLM significantly exceeds the critical cooling rate of 410 K/s required for martensitic transformation (β to α ) of Ti-6Al-4V [24]. The microstructure of annealed samples is shown in Figure 4b, which reveals that the acicular microstructure becomes even coarser than that of the as-built. Compared with needle shape lamellas observed in the SLM samples, the wrought samples have a typical α-β two-phase microstructure, as shown in Figure 4c, wherein the β phases are distributed around the coarse platelet α phases. Figure 5 shows XRD patterns for as-built and annealed SLM Ti-6Al-4V. All peaks of the two types of samples are characteristic of hexagonal close-packed (hcp) structure of Ti, which indicates that the microstructure is dominated by α /α phase even after annealing. In this study, the annealing treatment was carried out below β transus temperature and did not result in obvious decomposition of the as-fabricated α phase. Therefore, β phase peaks are absent in the annealed SLM Ti-6Al-4V. The peak intensities in the annealed condition are significantly higher compared to the pattern of the as-built conditions, indicating coarser grains in the annealed samples.   Figure 6 illustrates the monotonic tensile stress-strain curves of the as-built, annealed, and wrought Ti-6Al-4V samples. It is shown that the SLM samples (i.e., as-built and annealed) have a greater ultimate tensile strength and yield strength than the wrought one. The as-built SLM sample shows the highest strength, while the wrought sample has the greatest ductility. The difference in    Figure 6 illustrates the monotonic tensile stress-strain curves of the as-built, annealed, and wrought Ti-6Al-4V samples. It is shown that the SLM samples (i.e., as-built and annealed) have a greater ultimate tensile strength and yield strength than the wrought one. The as-built SLM sample shows the highest strength, while the wrought sample has the greatest ductility. The difference in  Figure 6 illustrates the monotonic tensile stress-strain curves of the as-built, annealed, and wrought Ti-6Al-4V samples. It is shown that the SLM samples (i.e., as-built and annealed) have a greater ultimate tensile strength and yield strength than the wrought one. The as-built SLM sample shows the highest strength, while the wrought sample has the greatest ductility. The difference in their mechanical performance can be explained by the difference in the microstructure. Amongst the three types of sample, the as-built sample has the smallest grain size, and hence, the largest volume of crystal boundary to prevent alloy dislocation slip resulting in plastic deformation. Therefore, it has the highest tensile strength. After annealing, the grains of the SLM sample are coarsened, and in the context, the tensile strength of the specimens decreased. Low ductility of as-built SLM samples can be improved by increasing the volume fraction of β Ti. However, in this study, the annealing treatment did not result in obvious decomposition of the as-fabricated α phase into the α + β phases, therefore, there was a slight change in elongation. As shown in Figure 6, the slopes of elastic deformation stage of stress-strain curves for all three types of sample are approximately equal. This is due to the fact that the elastic modulus mainly depends on the binding force amongst atoms rather than the microstructure. The tension properties of as-built, annealed SLM, and wrought samples are summarized in Table 4. As seen from the results, the ratios between the ultimate tensile strength and the yield strength for the as-built SLM, the annealed SLM, and the wrought samples are about 1.16, 1.05, and 1.03, respectively. All these ratios are less than 1.2, which can indicate that the capability for cyclic strain softening is high [25]. their mechanical performance can be explained by the difference in the microstructure. Amongst the three types of sample, the as-built sample has the smallest grain size, and hence, the largest volume of crystal boundary to prevent alloy dislocation slip resulting in plastic deformation. Therefore, it has the highest tensile strength. After annealing, the grains of the SLM sample are coarsened, and in the context, the tensile strength of the specimens decreased. Low ductility of as-built SLM samples can be improved by increasing the volume fraction of β Ti. However, in this study, the annealing treatment did not result in obvious decomposition of the as-fabricated α' phase into the α + β phases, therefore, there was a slight change in elongation. As shown in Figure 6, the slopes of elastic deformation stage of stress-strain curves for all three types of sample are approximately equal. This is due to the fact that the elastic modulus mainly depends on the binding force amongst atoms rather than the microstructure. The tension properties of as-built, annealed SLM, and wrought samples are summarized in Table 4. As seen from the results, the ratios between the ultimate tensile strength and the yield strength for the as-built SLM, the annealed SLM, and the wrought samples are about 1.16, 1.05, and 1.03, respectively. All these ratios are less than 1.2, which can indicate that the capability for cyclic strain softening is high [25].     Figure 7 shows the LCF life, i.e., N f , for the three types of samples at different strain amplitudes. As reflected in the figure, N f declines with the growth of strain amplitude for all types of samples. Under the same strain amplitude level, the as-built SLM sample always has a longer LCF life than the annealed one, because it has a higher strength due to smaller grain structure. Meanwhile, comparing N f results between the strain amplitude level at 1.2% and 2.0%, it is indicated that the N f decrement rate of the as-built sample is much higher than that of the wrought sample. This is caused by the pores raised in the SLM process [8,17]. With the increase of strain amplitude level, the porosity of as-built SLM materials exerts a more significant role in the overall degradation of the material [26]. The N f decrement rate of the annealed SLM samples is lower than that of the as-built, which implies that the coarser grains presented in the annealed SLM microstructure play a more important role in reducing the LCF life of the samples than the porosity.  Figure 7 shows the LCF life, i.e., Nf, for the three types of samples at different strain amplitudes. As reflected in the figure, Nf declines with the growth of strain amplitude for all types of samples. Under the same strain amplitude level, the as-built SLM sample always has a longer LCF life than the annealed one, because it has a higher strength due to smaller grain structure. Meanwhile, comparing Nf results between the strain amplitude level at 1.2% and 2.0%, it is indicated that the Nf decrement rate of the as-built sample is much higher than that of the wrought sample. This is caused by the pores raised in the SLM process [8,17]. With the increase of strain amplitude level, the porosity of as-built SLM materials exerts a more significant role in the overall degradation of the material [26]. The Nf decrement rate of the annealed SLM samples is lower than that of the as-built, which implies that the coarser grains presented in the annealed SLM microstructure play a more important role in reducing the LCF life of the samples than the porosity. Figure 8 illustrates the stress-strain responses of the LCF test. In detail, Figure 8(a), (c), and (e) on the left refer to the hysteresis loops of as-built SLM, annealed SLM, and the wrought samples, respectively, under strain amplitude level at 1.2%, while the figures on the right (see Figure 8(b), (d), and (f)) represent the respective loops for the samples at 2.0% strain amplitude level. It is implied that the as-built SLM has the smallest plastic deformation during cyclic loading amongst the three types of sample, which further results in the as-built SLM having the most compressed loops amongst the samples. Moreover, it is shown that the SLM samples have a cyclic softening behavior. This proves that there is a correlation between the degree of cyclic softening/hardening and the ratio of the ultimate tensile strength to yield strength for SLM parts. In general, the wrought sample has the largest softening degree, while the degree on the as-built sample is the smallest. At low strain amplitude level (see Figure 8(a), (c), and (e)), the peak stresses reached in the first cycle for all three types of samples are approximately identical. This is because when the strain level is low, the peak stresses of the samples would not exceed their yield strengths and elastic modulus of three types of samples are approximately equal. On the contrary, at high strain amplitude level, the as-built SLM achieves the greatest peak stress in the first cycle amongst the three types of samples. Figure 9 demonstrates the stress amplitude variations of the three types of sample alongside the increase of cycles under different strain amplitude levels. The results indicate that for all three types of sample, the cyclic softening is increased alongside the growth of the strain amplitude. Moreover, the as-built SLM Ti-6Al-4V shows the most significant change in cyclic softening amongst the three samples between low and high strain amplitude levels.  Figure 8 illustrates the stress-strain responses of the LCF test. In detail, Figure 8a,c,e on the left refer to the hysteresis loops of as-built SLM, annealed SLM, and the wrought samples, respectively, under strain amplitude level at 1.2%, while the figures on the right (see Figure 8b,d,f) represent the respective loops for the samples at 2.0% strain amplitude level. It is implied that the as-built SLM has the smallest plastic deformation during cyclic loading amongst the three types of sample, which further results in the as-built SLM having the most compressed loops amongst the samples. Moreover, it is shown that the SLM samples have a cyclic softening behavior. This proves that there is a correlation between the degree of cyclic softening/hardening and the ratio of the ultimate tensile strength to yield strength for SLM parts. In general, the wrought sample has the largest softening degree, while the degree on the as-built sample is the smallest. At low strain amplitude level (see Figure 8a,c,e), the peak stresses reached in the first cycle for all three types of samples are approximately identical. This is because when the strain level is low, the peak stresses of the samples would not exceed their yield strengths and elastic modulus of three types of samples are approximately equal. On the contrary, at high strain amplitude level, the as-built SLM achieves the greatest peak stress in the first cycle amongst the three types of samples. Figure 9 demonstrates the stress amplitude variations of the three types of sample alongside the increase of cycles under different strain amplitude levels. The results indicate that for all three types of sample, the cyclic softening is increased alongside the growth of the strain amplitude. Moreover, the as-built SLM Ti-6Al-4V shows the most significant change in cyclic softening amongst the three samples between low and high strain amplitude levels.     The reduction in macroscopic elasticity modulus between the first cycle to the half-life for the three types of sample is shown in Figure 10. The macroscopic elasticity modulus was measured as the slope of the linear part of the stress-strain hysteresis loops. In general, there is more reduction in macroscopic elasticity modulus as the strain amplitude increases, which promotes greater cyclic softening in Figure 9. Reductions of macroscopic elasticity modulus are likely related to the decrease of the effective cross-section due to the formation of internal defects [27]. The details of this are necessary to further investigate.

Cyclic Stress-Strain Curves
Based on the LCF tests under multi-level strain amplitudes, a series of stabilized cyclic stressstrain hysteresis loops at half of the specimen's life can be obtained. By placing these loops in one reference coordinate system, as shown in Figure 11, the best fit line can be regressed according to the

Variation of the Macroscopic Elasticity Modulus
The reduction in macroscopic elasticity modulus between the first cycle to the half-life for the three types of sample is shown in Figure 10. The macroscopic elasticity modulus was measured as the slope of the linear part of the stress-strain hysteresis loops. In general, there is more reduction in macroscopic elasticity modulus as the strain amplitude increases, which promotes greater cyclic softening in Figure 9. Reductions of macroscopic elasticity modulus are likely related to the decrease of the effective cross-section due to the formation of internal defects [27]. The details of this are necessary to further investigate.  The reduction in macroscopic elasticity modulus between the first cycle to the half-life for the three types of sample is shown in Figure 10. The macroscopic elasticity modulus was measured as the slope of the linear part of the stress-strain hysteresis loops. In general, there is more reduction in macroscopic elasticity modulus as the strain amplitude increases, which promotes greater cyclic softening in Figure 9. Reductions of macroscopic elasticity modulus are likely related to the decrease of the effective cross-section due to the formation of internal defects [27]. The details of this are necessary to further investigate.

Cyclic Stress-Strain Curves
Based on the LCF tests under multi-level strain amplitudes, a series of stabilized cyclic stress-

Cyclic Stress-Strain Curves
Based on the LCF tests under multi-level strain amplitudes, a series of stabilized cyclic stress-strain hysteresis loops at half of the specimen's life can be obtained. By placing these loops in one reference coordinate system, as shown in Figure 11, the best fit line can be regressed according to the maximum cyclic stresses, which is generally termed as cyclic stress-strain curve and can be modeled in a similar function of monotonic Ramberg-Osgood [28] curve, as shown in Equation (1): where ε ea and ε pa refer to elastic and plastic components of strain amplitude, respectively, K denotes the cyclic strength coefficient, n is the strain cyclic hardening exponent, and E is the elastic modulus. The ε pa part in the above equation can be expressed in a linear format in Equation (2) when making a logarithmic transformation on both sides.
The parameters of the cyclic Ramberg-Osgood model for the SLM Ti-6Al-4V are shown in Table 5.
Metals 2019, 9, x FOR PEER REVIEW 11 of 16 maximum cyclic stresses, which is generally termed as cyclic stress-strain curve and can be modeled in a similar function of monotonic Ramberg-Osgood [28] curve, as shown in Equation (1): where ε ea and ε pa refer to elastic and plastic components of strain amplitude, respectively, K' denotes the cyclic strength coefficient, n' is the strain cyclic hardening exponent, and E is the elastic modulus. The ε pa part in the above equation can be expressed in a linear format in Equation (2) when making a logarithmic transformation on both sides.
The parameters of the cyclic Ramberg-Osgood model for the SLM Ti-6Al-4V are shown in Table  5.

Strain-Life Curves
The local approach is one of the typical fatigue design methods, and it is adopted in low cycle conditions to estimate the crack formation duration, which is generally conducted on the basis of strain-life curves. The strain-life curves can be approximately described by the Basquin-Coffin-Manson relation using a power function [29][30][31][32] given in Equation (3): where σ f ' and ε f ' are the fatigue strength and ductility coefficients, respectively, and b and c are the fatigue strength and ductility exponents, respectively. The two parts on the right side of Equation (3) represent the elastic and plastic strain amplitudes, which are expressed in Equations (4) and (5), respectively: When making logarithmic operations on both sides, they can become linear as well. Based on the LCF test results, the strain-life curves for as-built and annealed SLM are plotted in Figure 12

Strain-Life Curves
The local approach is one of the typical fatigue design methods, and it is adopted in low cycle conditions to estimate the crack formation duration, which is generally conducted on the basis of strain-life curves. The strain-life curves can be approximately described by the Basquin-Coffin-Manson relation using a power function [29][30][31][32] given in Equation (3): where σ f and ε f are the fatigue strength and ductility coefficients, respectively, and b and c are the fatigue strength and ductility exponents, respectively. The two parts on the right side of Equation (3) represent the elastic and plastic strain amplitudes, which are expressed in Equations (4) and (5), respectively: When making logarithmic operations on both sides, they can become linear as well. Based on the LCF test results, the strain-life curves for as-built and annealed SLM are plotted in Figure 12a,b, respectively. In each figure, the curve illustrates the relationship between total strain amplitude ε a , and the fatigue life N f , while the relationships of ε ea -N f and ε pa -N f are represented by the two straight lines which are obtained by using the least square regression method. The fatigue strength and fatigue ductility parameters of the Basquin-Coffin-Manson equation for the SLM Ti-6Al-4V are then calculated based on the regression lines, which are given in Table 6. For the analysis reference, the relevant curves for the wrought Ti-6Al-4V are cited from the study in Reference [32], shown in Figure 12c. It is reflected from Figure 12a,b that the transition lives of the as-built and annealed SLM Ti-6Al-4V occur at approximately 100 cycles. This implies that the LCF life of the SLM Ti-6Al-4V above 100 cycles is governed by elastic strain rather than plastic strain. Therefore, the LCF behavior of SLM Ti-6Al-4V is dominated by fatigue strength rather than fatigue ductility properties. However, the wrought material has a larger number of transition life than the SLM Ti-6Al-4V, which is approximately 338 cycles, as exhibited in Figure 12c. Therefore, plastic deformation has more great fatigue damage for the SLM material than for the wrought material. Figure 13 depicts a comparison of the total strain amplitude life of the SLM against the wrought Ti-6Al-4V. It is observed that the fatigue performance of the as-built SLM Ti-6Al-4V is at a similar level to the wrought material, and the annealed SLM Ti-6Al-4V has the worst fatigue performance amongst the three types of materials. Compared to the wrought material, the as-built SLM Ti-6Al-4V has a better fatigue performance at low levels of strain amplitude, but a worse LCF performance when the strain amplitude is high. This is because, under the small strain amplitude levels, the peak cyclic stress is close to the yield stress, and hence strain hardening has a significant effect on material deformation. As the strain amplitude increase, the plastic deformation mechanisms play a more important role, and the porosity in the as-built SLM material exerted much more impact on the overall degradation of the material.  Table 6. For the analysis reference, the relevant curves for the wrought Ti-6Al-4V are cited from the study in Reference [32], shown in Figure 12(c). It is reflected from Figure 12(a) and (b) that the transition lives of the as-built and annealed SLM Ti-6Al-4V occur at approximately 100 cycles. This implies that the LCF life of the SLM Ti-6Al-4V above 100 cycles is governed by elastic strain rather than plastic strain. Therefore, the LCF behavior of SLM Ti-6Al-4V is dominated by fatigue strength rather than fatigue ductility properties. However, the wrought material has a larger number of transition life than the SLM Ti-6Al-4V, which is approximately 338 cycles, as exhibited in Figure 12(c). Therefore, plastic deformation has more great fatigue damage for the SLM material than for the wrought material. Figure 13 depicts a comparison of the total strain amplitude life of the SLM against the wrought Ti-6Al-4V. It is observed that the fatigue performance of the as-built SLM Ti-6Al-4V is at a similar level to the wrought material, and the annealed SLM Ti-6Al-4V has the worst fatigue performance amongst the three types of materials. Compared to the wrought material, the as-built SLM Ti-6Al-4V has a better fatigue performance at low levels of strain amplitude, but a worse LCF performance when the strain amplitude is high. This is because, under the small strain amplitude levels, the peak cyclic stress is close to the yield stress, and hence strain hardening has a significant effect on material deformation. As the strain amplitude increase, the plastic deformation mechanisms play a more important role, and the porosity in the as-built SLM material exerted much more impact on the overall degradation of the material.

Low Cycle Fatigue (LCF) Fractography
In general, fatigue damage to metal material has three stages: the crack initiation caused by local plastic stress concentration, the crack propagation in the way of microcracks interconnection, and the instantaneous rupture taking place when crack propagation reaches a critical threshold. Ti-6Al-4V fabricated by SLM exhibits the same LCF failure mechanism as other metal materials. In all the fatigue test samples, the cracks always initiated from the surface or subsurface defects. SEM micrographs of LCF fracture surface for as-built SLM Ti-6Al-4V tested at 1.2% strain amplitude are given in Figure  14. In Figure 14(a) and (b), fatigue cracks start from the surface and propagate to the inside. During crack propagation, fatigue striations are generated, which are generally adjacent to the area of the crack initiation. Figure 15 shows the LCF fracture surface of the as-built sample tested at 2.0% strain amplitude. In Figure 15(a) and (b), a pore located about 120 μm below the surface can be noticed, which contributes to the crack initiation. Cyclic stresses are applied to both surface and interior of the samples simultaneously. The stress applied to the surface tends to be higher than that applied to the interior area. Meanwhile, in the internal area, every grain is surrounded by others, while the grains on the surface area have fewer border constraints formed by the surrounding and, hence, slip easily. In addition, machine traces and scratches result in stress concentration on the surface area and then accelerate the crack initiation. Besides, it has been proven that the pores near to surface have higher stress concentration and act as potential crack initiation sites [33,34]. The final rupture regions are shown in Figure 14(c) and Figure 15(c). It can be seen that the instantaneous rupture surfaces contain dimples which are characteristic of the static load fracture.

Low Cycle Fatigue (LCF) Fractography
In general, fatigue damage to metal material has three stages: the crack initiation caused by local plastic stress concentration, the crack propagation in the way of microcracks interconnection, and the instantaneous rupture taking place when crack propagation reaches a critical threshold. Ti-6Al-4V fabricated by SLM exhibits the same LCF failure mechanism as other metal materials. In all the fatigue test samples, the cracks always initiated from the surface or subsurface defects. SEM micrographs of LCF fracture surface for as-built SLM Ti-6Al-4V tested at 1.2% strain amplitude are given in Figure 14. In Figure 14a,b, fatigue cracks start from the surface and propagate to the inside. During crack propagation, fatigue striations are generated, which are generally adjacent to the area of the crack initiation. Figure 15 shows the LCF fracture surface of the as-built sample tested at 2.0% strain amplitude. In Figure 15a,b, a pore located about 120 µm below the surface can be noticed, which contributes to the crack initiation. Cyclic stresses are applied to both surface and interior of the samples simultaneously. The stress applied to the surface tends to be higher than that applied to the interior area. Meanwhile, in the internal area, every grain is surrounded by others, while the grains on the surface area have fewer border constraints formed by the surrounding and, hence, slip easily. In addition, machine traces and scratches result in stress concentration on the surface area and then accelerate the crack initiation. Besides, it has been proven that the pores near to surface have higher stress concentration and act as potential crack initiation sites [33,34]. The final rupture regions are shown in Figures 14c and 15c. It can be seen that the instantaneous rupture surfaces contain dimples which are characteristic of the static load fracture.

Conclusions
This paper investigated the microstructures, monotonic tension, cyclic stress-strain behavior, and LCF properties of the as-built and annealed SLM Ti-6Al-4V. The conclusions of the study are drawn as follows: 1. The as-built SLM Ti-6Al-4V consisted of fine acicular α' martensite, therefore displayed the higher strength and lower ductility than the wrought material, whereas after being annealed, the tensile strength of Ti-6Al-4V decreased, which mainly resulted from the coarsened grains formed during the annealing process. 2. Cyclic softening behavior was observed for both as-built and annealed SLM Ti-6Al-4V. With the growth of strain amplitude level, more elasticity modulus reductions occurred during the cyclic loading, which promoted a greater cyclic softening. 3. The cyclic Ramberg-Osgood and the Basquin-Coffin-Manson models were fitted to depict the cyclic stress-strain and the strain-life curves for the SLM Ti-6Al-4V, respectively. The parameters of LCF performance for SLM Ti-6Al-4V were then determined based on the fitted curves, which could be used in fatigue simulation analysis of SLM Ti-6Al-4V parts in real cases. 4. This paper showed that the as-built SLM Ti-6Al-4V had a comparable LCF performance to the wrought material. Compared to the wrought material, the as-built SLM material had better fatigue performance at low strain amplitudes but a worse LCF performance at high strain amplitudes, which is because that porosity in the as-built SLM material exerted much more impact on the overall degradation of the material at high strain amplitude levels. However, the annealing treatment caused SLM Ti-6Al-4V to have coarser grains and hence deteriorated LCF performance. To avoid grain growth while eliminating the residual stresses, the appropriate heat treatment process for SLM Ti-6Al-4V parts is worthy of being be investigated in future research.

Conclusions
This paper investigated the microstructures, monotonic tension, cyclic stress-strain behavior, and LCF properties of the as-built and annealed SLM Ti-6Al-4V. The conclusions of the study are drawn as follows: 1. The as-built SLM Ti-6Al-4V consisted of fine acicular α' martensite, therefore displayed the higher strength and lower ductility than the wrought material, whereas after being annealed, the tensile strength of Ti-6Al-4V decreased, which mainly resulted from the coarsened grains formed during the annealing process. 2. Cyclic softening behavior was observed for both as-built and annealed SLM Ti-6Al-4V. With the growth of strain amplitude level, more elasticity modulus reductions occurred during the cyclic loading, which promoted a greater cyclic softening. 3. The cyclic Ramberg-Osgood and the Basquin-Coffin-Manson models were fitted to depict the cyclic stress-strain and the strain-life curves for the SLM Ti-6Al-4V, respectively. The parameters of LCF performance for SLM Ti-6Al-4V were then determined based on the fitted curves, which could be used in fatigue simulation analysis of SLM Ti-6Al-4V parts in real cases. 4. This paper showed that the as-built SLM Ti-6Al-4V had a comparable LCF performance to the wrought material. Compared to the wrought material, the as-built SLM material had better fatigue performance at low strain amplitudes but a worse LCF performance at high strain amplitudes, which is because that porosity in the as-built SLM material exerted much more impact on the overall degradation of the material at high strain amplitude levels. However, the annealing treatment caused SLM Ti-6Al-4V to have coarser grains and hence deteriorated LCF performance. To avoid grain growth while eliminating the residual stresses, the appropriate heat treatment process for SLM Ti-6Al-4V parts is worthy of being be investigated in future research.

Conclusions
This paper investigated the microstructures, monotonic tension, cyclic stress-strain behavior, and LCF properties of the as-built and annealed SLM Ti-6Al-4V. The conclusions of the study are drawn as follows: 1.
The as-built SLM Ti-6Al-4V consisted of fine acicular α' martensite, therefore displayed the higher strength and lower ductility than the wrought material, whereas after being annealed, the tensile strength of Ti-6Al-4V decreased, which mainly resulted from the coarsened grains formed during the annealing process.

2.
Cyclic softening behavior was observed for both as-built and annealed SLM Ti-6Al-4V. With the growth of strain amplitude level, more elasticity modulus reductions occurred during the cyclic loading, which promoted a greater cyclic softening. 3.
The cyclic Ramberg-Osgood and the Basquin-Coffin-Manson models were fitted to depict the cyclic stress-strain and the strain-life curves for the SLM Ti-6Al-4V, respectively. The parameters of LCF performance for SLM Ti-6Al-4V were then determined based on the fitted curves, which could be used in fatigue simulation analysis of SLM Ti-6Al-4V parts in real cases.

4.
This paper showed that the as-built SLM Ti-6Al-4V had a comparable LCF performance to the wrought material. Compared to the wrought material, the as-built SLM material had better fatigue performance at low strain amplitudes but a worse LCF performance at high strain amplitudes, which is because that porosity in the as-built SLM material exerted much more impact on the overall degradation of the material at high strain amplitude levels. However, the annealing treatment caused SLM Ti-6Al-4V to have coarser grains and hence deteriorated LCF performance.
To avoid grain growth while eliminating the residual stresses, the appropriate heat treatment process for SLM Ti-6Al-4V parts is worthy of being be investigated in future research.

5.
It was observed that fatigue cracks always initiated from the surface or subsurface defects. Instantaneous rupture region exhibited the characteristics of the static load fracture.