Effect of Gradient Energy Density on the Microstructure and Mechanical Properties of Ti6Al4V Fabricated by Selective Electron Beam Additive Manufacture.

Selective Electron Beam Additive Manufacturing (SEBAM) is a promising powder bed fusion additive manufacturing technique for titanium alloys that select particular area melting in different energy density for producing complexly shaped biomedical devices. For most commercial Ti6Al4V porous medical devices, the gradient energy density is usually applied to manufacture in one component during the SEBAM process which selects different energy density built on particular zones. This paper presents gradient energy density base characterization study on an SEBAM built rectangular specimen with a size of 3 mm × 20 mm × 60 mm. The specimen was divided into three zones were built in gradient energy density from 16 to 26.5 J/mm3. The microstructure and mechanical properties were investigated by means of scanning electron microscopy, X-ray diffraction, transmission electron microscopy and mechanical test. The α′ martensitic and lack of fusion were observed in the low energy density (LED) built zone. However, no α′ phase and no irregular pores were observed both in overlap energy density (OED) and high energy density (HED) built zones located at the middle and bottom of the specimen respectively. This implies the top location and lower energy density have positive effects on the cooling rate but negative effects on densification. The subsequence mechanical properties result also supports this point. Moreover, the intermetallic Ti3Al found in the bottom may be due to the heat transfer from the following melting layer. Furthermore, the microstructure evolution in gradient energy built zones is discussed based on the findings of the microstructure and thermal history correlation analysis.


Introduction
Selective electron beam additive manufacture (SEBAM) is one of powder bed fusion technology that builds metallic parts with complex geometries by using the electron beam as the heat source to selectively preheat and melt over the metallic powder bed via layer by layer based on 3D CAD model input. SEBAM has attracted much attention over the past few years for its advantages, such as high material utilization, low porosity and fast production rate without post heat treatment in particular energy density was selectively built in different zones (top, middle and top) of one rectangular sample by SEBAM method using Ti6Al4V (ELI) alloy powders on the effect of the microstructure and mechanical properties. It is supposed to be of significance to evaluate the performance of SEBAM-built parts. More importantly, a thorough understanding of gradient energy on microstructure and mechanical properties obtained from this study will aid to further propel the practice application of SEBAM-built metallic parts.
Materials 2020, 12, x FOR PEER R EV IEW 3 of 15 top) of one rectangular sample by SEBAM method using Ti6Al4V (ELI) alloy powders on the effect of the microstructure and mechanical properties. It is supposed to be of significance to evaluate the performance of SEBAM-built parts. More importantly, a thorough understanding of gradient energy on microstructure and mechanical properties obtained from this study will aid to further propel the practice application of SEBAM-built metallic parts. Figure 1. Schematic diagram of commercial porous fusion cage structure was fabricated by selective electron beam additive manufacture (SEBAM) method using gradient energy density on three different zones. Solid and porous parts were made independently by higher and lower energy density. The overlap site with 0.2 mm in width between solid and porous parts was made sequentially by higher and lower energy density.

Sample Preparation
In order to investigate the microstructure and micro-harness evolution of one SEBAMed (Kaohsiung, Taiwan) metallic part was built by SEBAM method in gradient energy density. A rectangular sample which 3 mm (thickness) × 20 mm (width) × 60 mm (height) in size was used and selectively in setting three different energy density namely high energy density (HED), low energy density (LED) and overlap energy density (OED) built on bottom (0 ≦ z ≦ 20), middle (20 ≦ z ≦ 40) and top 40 ≦ z ≦ 60 of the specimen respectively along the z axis building direction as shown in Figure 2. The energy density parameters in detail were described in the next section.

Selective Electron Beam Additive Manufacturing
The Extra Low Interstitial Ti-6Al-4V (Grade 23) powder supplied by Arcam AB company was used; the powder size distribution was quoted as 45 to 100 μm. The chemical composition of Ti-6Al-4V powder was supplied as showed in Table 1 followed as ASTM F3001. Recycling of non-melted powder and/or sintered powder (Figure A1 in Apprendix A) was achieved via powder recovery system and a mechanical vibrating sieve which mesh size ≦ 150 μm. All SEBAM parts were fabricated on the Arcam Q10 machine (Producer: Arcam EBM GE Additive company, Gothenburg, Sweden). The SEBAM process was implemented in the vacuum chamber with pressure below 5×10 -3 Figure 1. Schematic diagram of commercial porous fusion cage structure was fabricated by selective electron beam additive manufacture (SEBAM) method using gradient energy density on three different zones. Solid and porous parts were made independently by higher and lower energy density. The overlap site with 0.2 mm in width between solid and porous parts was made sequentially by higher and lower energy density.

Sample Preparation
In order to investigate the microstructure and micro-harness evolution of one SEBAMed (Kaohsiung, Taiwan) metallic part was built by SEBAM method in gradient energy density. A rectangular sample which 3 mm (thickness) × 20 mm (width) × 60 mm (height) in size was used and selectively in setting three different energy density namely high energy density (HED), low energy density (LED) and overlap energy density (OED) built on bottom (0 ≤ z ≤ 20), middle (20 ≤ z ≤ 40) and top 40 ≤ z ≤ 60 of the specimen respectively along the z axis building direction as shown in Figure 2. The energy density parameters in detail were described in the next section.  Figure 3 shows representative specimens of the gradient energy from high energy density to low energy density built zones along the z axis building direction. The irregular pores which size exceeds 100 μm was only observed in the LED built specimen (Figure 3a) which is caused by the lack of fusion. Above line energy density more than 100 J/m is necessary for full densification of Ti6Al4V and below 100 J/m SEBAMed specimens contain more than 1% porosity as earlier reported [13]. This means that irregular pores did not obviously exist both in the HED and OED built zones, implying that highly dense sample can be fabricated by the SEBAM process if the energy density were more than 100 J/m. The irregular pores would greatly reduce yield stress, tensile stress and ductility [13].

Effect of Gradient Energy Density on the Defects
In contrast the typical spherical pores several μm in size were found in all HED, OED and LED built samples in Figure 3b-d. It means no relation between the spherical pores and energy density and is mainly caused by the entrapped argon gas inside the Ti6Al4V powder during the production of plasma wire gas atomization [22]. The presence of limited small pores cannot be eliminated by hot isostatic press treatment [23] and did not significantly affect tensile properties. It might be decreased fatigue life.

Selective Electron Beam Additive Manufacturing
The Extra Low Interstitial Ti-6Al-4V (Grade 23) powder supplied by Arcam AB company was used; the powder size distribution was quoted as 45 to 100 µm. The chemical composition of Ti-6Al-4V powder was supplied as showed in Table 1 followed as ASTM F3001. Recycling of non-melted powder and/or sintered powder ( Figure A1 in Appendix A) was achieved via powder recovery system and a mechanical vibrating sieve which mesh size ≤ 150 µm. All SEBAM parts were fabricated on the Arcam Q10 machine (Producer: Arcam EBM GE Additive company, Gothenburg, Sweden). The SEBAM process was implemented in the vacuum chamber with pressure below 5 × 10 −3 mbar in the beginning and finished with a pressure of 2 × 10 −5 mbar. Each layer was setting preheated 730 • C by fast scanning with the defocused electron beam and layer thickness of 50 µm. Gradient energy density in building parameters was set as follows: the standard parameter of the beam current from 3 mA to 15 mA and scan speed from 1500 mm/s to 4530 mm/s with fixed layer thickness 50 µm, beam diameter of 100 µm and hatch spacing of 150 µm were performed. The energy density can be approximated as: where energy density E, operation voltage V, beam current C, layer thickness t, hatch distance h and scan speed S in Table 2. The high energy density (HED) resulting from the reduced scan speed and increased beam current was therefore 63% higher compared to low energy density (LED) setups. For the transition overlapping 2 mm in width was applied overlap energy density (OED), i.e., HED in the first scan and LED in the after.

Microstructure Observation
Optical microscopy and scanning electron microscopy (SEM) of SEBAMed samples prepared by grinding and polishing-diamond abrasive disc with water used as coolant in the grinding process and polished with silicon carbide and 0.3 µm Al 2 O 3 suspension. Optical and SEM samples were etched in 2.5% HNO 3 + 5% HCl + 92.5% ethanol reagent. Secondary electron image SEM was carried on JEOL JSM-6380 (Tokyo, Japan) at 15 kV. Phase analysis was conducted using X-Ray Diffraction (XRD) (Panalytical B.V., Almelo, The Netherlands) and transmission electron microscopy (TEM) (PHILPLIES CM200 and JEOL 2000 FX, Tokyo, Japan). The characteristic CuKα radiation (λ = 1.5412 Å) in the 2θ range from 20 to 100 degrees having voltage 40 keV and current 40 mA was used. TEM samples were prepared using the following procedures: thin samples sections were manually ground to 0.06 ± 0.02 mm in thickness with silicon carbide paper and then the thin foils for TEM were electro polish using Automatic Twin-Jet Electropolisher Model 110 (Yokohama, Japan) at 25 V in a solution bath consisting of 12% perchloric acid, 15% glacial acetic acid and 75% ethanol reagent.

Mechanical Properties
Vicker microhardness (500 g, 15 s hold) of 15 individual measurements in each HED, OED and LED were performed on the metallographic samples using an Akashi MVK-H100 (Osaka, Japan) machine. The tensile tests on an MTS-10t (Eden Prairie, MN, USA) Tester using cylinder specimen with gauge length of 30 mm of specimen 3 ( Figure A2 in Appendix B) according to ASTM-E8M specifications. The specimens were subjected to impact testing with the use of a Tinius Olsen model IT 504 polymeric impact testing machine (Tinius Olsen, Horsham, PA, USA). All tests were conducted at a pendulum capacity of 15 J, a drop height of 609.6 mm and velocity of 3.46 m/s. Figure 3 shows representative specimens of the gradient energy from high energy density to low energy density built zones along the z axis building direction. The irregular pores which size exceeds 100 µm was only observed in the LED built specimen (Figure 3a) which is caused by the lack of fusion. Above line energy density more than 100 J/m is necessary for full densification of Ti6Al4V and below 100 J/m SEBAMed specimens contain more than 1% porosity as earlier reported [13]. This means that irregular pores did not obviously exist both in the HED and OED built zones, implying that highly dense sample can be fabricated by the SEBAM process if the energy density were more than 100 J/m. The irregular pores would greatly reduce yield stress, tensile stress and ductility [13].

Effect of Gradient Energy Density on the Defects
In contrast the typical spherical pores several µm in size were found in all HED, OED and LED built samples in Figure 3b-d. It means no relation between the spherical pores and energy density and is mainly caused by the entrapped argon gas inside the Ti6Al4V powder during the production of plasma wire gas atomization [22]. The presence of limited small pores cannot be eliminated by hot isostatic press treatment [23] and did not significantly affect tensile properties. It might be decreased fatigue life.

Effect of Gradient Energy Density on the Microstructure Evolution
In order to understand the influence of the gradient energy density on three different sites, the microstructure was studied along the build direction. Figure 3b shows partially acicular α′ martensitic and α + β were developed by LED built zone where is at the top of the specimen. At the middle of the specimen built in the OED parameter, a lamella mixture of α+β structure was observed (Figure 3c). A coarser α+β lamella structure in the prior β grain was also observed at the bottom where is built in HED parameter (Figure 3d). No obvious difference in SEM analysis was found both in OED and HED built zones. However, the acicular α′ martensitic only was observed in LED built zone, which indicates cooling rate in LED zone was higher than OED and HED built zones. Figure 4 shows the XRD analysis of the gradient energy density built by SEBAM method using Ti6Al4V powder. The characteristic peaks of dominant α phase with (101), (002) and small fraction of β phase with (110) were observed in all energy density built zones. Minor fraction of β phase in both OED and LED built zones compared to HED built zones. It is difficult to distinguish α and α′ phase by XRD analysis. Based on the mentioned above, the acicular α′ martensitic phase was only found in the LED built zone where the present characteristic peaks of α phase overlapped with α′. Interestingly, the intermetallic phase Ti3Al was observed in HED built zone. Further TEM analysis in Figure 5 revealed that the nano-size intermetallic Ti3Al dispersed near the α grain boundary, and a fine lamella α+β structure. Moreover the crystallographic relationship between α, Ti3Al and β phase as  [24]. The analysis of the formation of nanosize Ti3Al without the post heat treatment or repetitive melting (intrinsic heat treatment) during the SEBAM process will be discussed in the following section.

Effect of Gradient Energy Density on the Microstructure Evolution
In order to understand the influence of the gradient energy density on three different sites, the microstructure was studied along the build direction. Figure 3b shows partially acicular α martensitic and α + β were developed by LED built zone where is at the top of the specimen. At the middle of the specimen built in the OED parameter, a lamella mixture of α+β structure was observed (Figure 3c). A coarser α+β lamella structure in the prior β grain was also observed at the bottom where is built in HED parameter (Figure 3d). No obvious difference in SEM analysis was found both in OED and HED built zones. However, the acicular α martensitic only was observed in LED built zone, which indicates cooling rate in LED zone was higher than OED and HED built zones. Figure 4 shows the XRD analysis of the gradient energy density built by SEBAM method using Ti6Al4V powder. The characteristic peaks of dominant α phase with (101), (002) and small fraction of β phase with (110) were observed in all energy density built zones. Minor fraction of β phase in both OED and LED built zones compared to HED built zones. It is difficult to distinguish α and α phase by XRD analysis. Based on the mentioned above, the acicular α martensitic phase was only found in the LED built zone where the present characteristic peaks of α phase overlapped with α . Interestingly, the intermetallic phase Ti 3 Al was observed in HED built zone. Further TEM analysis in Figure 5 revealed that the nano-size intermetallic Ti 3 Al dispersed near the α grain boundary, and a fine lamella α+β structure. Moreover the crystallographic relationship between α, Ti 3 Al and β phase as  [24]. The analysis of the formation of nanosize Ti 3 Al without the post heat treatment or repetitive melting (intrinsic heat treatment) during the SEBAM process will be discussed in the following section.

Effects of Gradient Energy Density on the Mechanical Properties
A continuous microhardness on the HED, OED and LED built zones were measured in Figure  6. Apparently, it was divided into two groups via gradient energy density built. The microhardness values in average 320 to 350 HV also revealed no obvious difference between HED and OED built zones. Nevertheless, a relatively high microhardness (from 375 to 420 HV) was observed in the LED built zone. The higher microhardness in the LED built zone was caused by the acicular α′ martensitic microstructure (Figure 3b) and it is consistent with those reported for selective laser melting built Ti6Al4V alloy [25,26]. Moreover, the measure average α lath width in the HED and OED built zones were determined to be 0.69 ± 0.04 μm and 0.59 ± 0.14 μm respectively. The result is in good agreement with no obvious difference in mircohardness in the two zones. However, the microhardness in the OED built (double melting) zone showed the finer α+β lamella and grain bound of α structure than in the HED built zone. The result was in contrast to the paper showed the coarsening microstructure was caused by double melting in the same build height [14]. This implies that double melting is not only seen to be relevant to grain size. The effect of thermal history from the different building height might have a powerful impact on grain size [18].

Effects of Gradient Energy Density on the Mechanical Properties
A continuous microhardness on the HED, OED and LED built zones were measured in Figure 6. Apparently, it was divided into two groups via gradient energy density built. The microhardness values in average 320 to 350 HV also revealed no obvious difference between HED and OED built zones. Nevertheless, a relatively high microhardness (from 375 to 420 HV) was observed in the LED built zone. The higher microhardness in the LED built zone was caused by the acicular α martensitic microstructure (Figure 3b) and it is consistent with those reported for selective laser melting built Ti6Al4V alloy [25,26]. Moreover, the measure average α lath width in the HED and OED built zones were determined to be 0.69 ± 0.04 µm and 0.59 ± 0.14 µm respectively. The result is in good agreement with no obvious difference in mircohardness in the two zones. However, the microhardness in the OED built (double melting) zone showed the finer α+β lamella and grain bound of α structure than in the HED built zone. The result was in contrast to the paper showed the coarsening microstructure was caused by double melting in the same build height [14]. This implies that double melting is not only seen to be relevant to grain size. The effect of thermal history from the different building height might have a powerful impact on grain size [18].   Table 2. The presence of the obvious plastic deformation both in OED and HED curves were distinct with LED curve which  Table 2. The presence of the obvious plastic deformation both in OED and HED curves were distinct with LED curve which almost no plastic deformation was observed. Both OED and HED curves had no obvious difference from the UTS and elongation, however in the finally the HED curve was slightly reduced the strain. Higher elastic modulus of OED and HED built samples compared to LED built sample. Interestingly, the highest hardness was in the LED built zone due to rapid cooling leading to acicular α martensitic but it had the smaller UTS and elongation compared to OED built and HED built sample. To further understand the variation of the mechanical properties, the fracture morphology was observed in Figure 7b-d. The transgranular ductile dimple tearing resulting from the coalescence of microvoids fracture surface was observed in both OED and HED built samples. Crack propagation of fine dimples at the tensile fracture indicated the extent of plastic deformation. Besides, according to the Hall-Petch strength mechanism, smaller grain size provided more grain boundaries, which can impede the movement of the dislocation, demonstrated as OED built sample has slightly higher tensile stress and elongation compared to HED built sample. In contrast, the fracture surface included cleavage, unmelted and partially melted powder were shown in LED built sample. This implied insufficient energy density that was unable to melt the deeper layer had a negative effect on tensile stress and elongation properties. This result is consistent with others reported [13,27].

Formation α′ Martensitic in LED Built Zone
The thermal gradient, G (K/m), and solidification velocity, R (m/s) can be used to predict solidification microstructure where G/R controls the mode of solidification (morphology) and G × R (cooling rate) in SEBAM process [31]. For an increasing energy density, the cooling rate became slower as was shown in solidification map for SEBAM Ti6Al4V melted [31]. Moreover, the cooling rate of the top was higher than the bottom also reported in SEBAMed Ti6Al4V parts [18]. In the present study, the SEM investigations in Figure 3 revealed that formation of acicular α′ martensitic in LED-built zone where at the top of the specimen along the build direction. Both the low energy built and the top of the building height caused a high cooling rate. Moreover, during the SEBAM process, as the electron beam scanned over the Ti-6Al-4V powder, a melt pool formed and then was rapidly solidified into β grains. The prior β grains underwent a rapid cooling (a critical cooling rate of >410 °C/s) transforming into martensitic α' phase. Moreover, it its known that the build temperature of 600~650 °C in the chamber was below martensitic start temperature (Ms) 800 °C. Thus, α' martensitic phase should be formed after melting. The average fracture toughness properties in different energy density were shown in Table 3. Both OED and HED built samples showed the higher impact energy compared to the LED built sample which had the lower fracture toughness properties (2 to 3.2 J). These are similar results to the tensile stress in the fracture surface. Both fracture surfaces as transgranular ductile dimple fractures from the coalescence of microvoids in OED and HED built samples were observed in Figure 7b-f. However, the lower value of impact energy of fracture surface in LED built sample exhibited pores, unmelted powder and partial melt region (See Figure 7d,g). This means lower energy density may contributed to large area of pores and unmelted powder [13,27].

Formation Gradient Microstructure on the Gradient Energy Density Built Zones
The gradient microstructure was built by gradient energy density from the bottom to top of the specimen. In this work, the microstructure of the particular zones in different energy density built will get the gradient microstructure (α + α + β in LED built zone, slightly finer α + β lamella structure in the OED built zone and α+β lamella with minor fraction of nanosize Ti 3 Al intermetallic in HED built zone were observed in Figures 3-5). Moreover, all the specimens showed the microhardness values between 304 to 420 HV. The highest microhardness was contributed to the presence of α martensitic phase (Figure 3b) in LED built zone. However, no α phase was observed and only fine α+β lamella were observed both in the OED and HED built zones where the average microhardness of 340 HV was shown in Figure 6. The presence of limited small amount Ti 3 Al did not significantly affect the microhardness in the HED built zone compared to the microhardness in the OED built zone. In the present study, the thermal events were more complex due to a combination of the building height and gradient energy density.It is showed that the difference of previously report as varying microstructure only in changing building direction [15] process parameters [28,29], building high [18,30], overlap distance [14], heat treatment [19], intrinsic heat treatment [24]. Therefore, the analysis of formation of gradient microstructure/phase in gradient built zones will be discussed in the following section.

Formation α Martensitic in LED Built Zone
The thermal gradient, G (K/m), and solidification velocity, R (m/s) can be used to predict solidification microstructure where G/R controls the mode of solidification (morphology) and G × R (cooling rate) in SEBAM process [31]. For an increasing energy density, the cooling rate became slower as was shown in solidification map for SEBAM Ti6Al4V melted [31]. Moreover, the cooling rate of the top was higher than the bottom also reported in SEBAMed Ti6Al4V parts [18]. In the present study, the SEM investigations in Figure 3 revealed that formation of acicular α martensitic in LED-built zone where at the top of the specimen along the build direction. Both the low energy built and the top of the building height caused a high cooling rate. Moreover, during the SEBAM process, as the electron beam scanned over the Ti-6Al-4V powder, a melt pool formed and then was rapidly solidified into β grains. The prior β grains underwent a rapid cooling (a critical cooling rate of >410 • C/s) transforming into martensitic α phase. Moreover, it its known that the build temperature of 600~650 • C in the chamber was below martensitic start temperature (Ms) 800 • C. Thus, α martensitic phase should be formed after melting.

Formation α+β in OED Built and HED Built Zones
The observation of the α+β lamella structure inside OED built (Figure 3c) and HED built (Figure 3d) samples also proves the aforementioned process. Following the above mentioned formation of martensitic α , β phase was transformed into an α (β → α ) reaction on cooling. In addition, α martensitic will decompose into α and β phases, i.e., α → α + β, when it is subjected to isothermal annealing in the α + β two phase field. It was reported that α in the thicker sample would be able to decompose into α/β phases due to higher thermal mass causing the slower cooling. During the EBM process, an electron beam was constantly scanning over the Ti6Al4V powder bed and the heat was mainly transferred from top to bottom through the built parts, the temperature of the newly deposited layers must be higher than 650 • C in the middle and bottom of the specimen that was sufficient to enable α decomposition.

Formation Ti 3 Al in HED Built Zones
The formation of the intermetallic Ti3Al phase along the building direction of the bottom of the specimen was revealed using X-ray diffraction by the presence of (001) characteristic peak as shown in Figure 4 as well as by select area diffraction carried out during TEM investigations ( Figure 5). According to previous investigations of SLMed Ti6Al4V and SLMed Ti6Al4V with intrinsic heat treatment, the driving force for Ti 3 Al formation due to the fast solidification during the SLM process, concentration of aluminum and oxygen [24]. Moreover, Al depletion at the interface during EBM process was reported [29]. In this work, diffraction pattern evidence of Ti 3 Al was only observed in HED built zone (the bottom of the specimen) mostly affected by heat transfer from the following layer. Since precipitation of this phase can occur during aging at 500-600 • C for several hours [31], it is reasonable to conclude that this range of temperatures was reached by the scanning of following layers, lead to the precipitation of an intermetallic Ti 3 Al.

Cooling Path in Gradient Energy Density and Subsequence Heat Transfer
From the microstructure point of view as mentioned above in three different zones. The α + α + β in the LED built zone where it is the top of the specimen, α + β lamella structure in OED built zone where it is the middle of the specimen and α + β lamella with small amount of Ti3Al in HED built zone where it is the bottom of the specimen, were observed due to different thermal history. It implied the cooling rate of the top (LED built zone) was faster than the middle (OED built zone) and the bottom (HED built zone) of the specimen. Further the evidence of slight smaller width of α lath in the middle in Figure 3 implied slight higher cooling rate in the middle compared to the bottom. Based on the literature reported the complex thermal events in the EBM process can be simplified in three main stages [19]. The first step is a rapid cooling from the molten state to the layer temperature, followed by a quasi-isothermal stage at the local cooling until completion of the build, and finishing with a slow cooling to room temperature. Gradient energy density and different locations causing different thermal events were predicted in Figure 8 base on the microstructure analysis as mentioned above. Moreover, the heat coming from the following layer melting will transfer to the previous layer causing the similar effect as aging heat treatment was observed only at the bottom. Those observed results proved the effect combined gradient energy density and different locations on microstructure and phase transformation due to different thermal history. and LED parameters built in top, middle and bottom of the specimen respectively. The difference in cooling rate result from different energy density and spatial location thus contribute to different microstructure. i.e., α′ + α + βin LED built zone, α + β lamella structure in the OED built zone and α + β + Ti3Al intermetallic phase in the HED built zone were observed.

Conclusions
In this work, gradient energy density was carried out on a specimen from the bottom to the top by SEBAM technology. The effect of energy density on the microstructure and mechanical properties of SEBAMed Ti6Al4V was investigated by SEM, X-ray, TEM, microhardness, tensile and impact tests. Based on the experimental results, the following conclusions can be drawn:

•
Gradient energy density (16 to 26.5 J/mm 3 ) built sample with dimension of 20 mm × 60 mm × 3 mm was fabricated by SEBAM. Small spherical pores coming from powder were found to be independent of the energy density and locations in the built part. Large irregular pores were observed in LED built zone due to lack of fusion but no internal crack and large pores were observed in OED or HED built zones.

•
The α′ + α + β, α + β and α + β + Ti3Al graded microstructure was observed from the top to the bottom. The graded microstructure was caused by the complex heat transfer due to gradient energy density and locations during building. Correspondingly, gradually increased microhardness was achieved from the bottom to the top and there was no obvious difference from the bottom to the middle.

•
Lack of fusion defects resulted in a decrease in plasticity in the LED built zone. The unmelted powder and voids in following fracture morphology were observed by SEM.

•
A small amount intermetallic Ti3Al was observed only in the bottom (HED built zone). The formation of Ti3Al is attributed to the heat transfer from the following layer melting.    [19] of LED, OED and LED parameters built in top, middle and bottom of the specimen respectively. The difference in cooling rate result from different energy density and spatial location thus contribute to different microstructure. i.e., α + α + β in LED built zone, α + β lamella structure in the OED built zone and α + β + Ti 3 Al intermetallic phase in the HED built zone were observed.

Conclusions
In this work, gradient energy density was carried out on a specimen from the bottom to the top by SEBAM technology. The effect of energy density on the microstructure and mechanical properties of SEBAMed Ti6Al4V was investigated by SEM, X-ray, TEM, microhardness, tensile and impact tests. Based on the experimental results, the following conclusions can be drawn:

•
Gradient energy density (16 to 26.5 J/mm 3 ) built sample with dimension of 20 mm × 60 mm × 3 mm was fabricated by SEBAM. Small spherical pores coming from powder were found to be independent of the energy density and locations in the built part. Large irregular pores were observed in LED built zone due to lack of fusion but no internal crack and large pores were observed in OED or HED built zones.

•
The α + α + β, α + β and α + β + Ti 3 Al graded microstructure was observed from the top to the bottom. The graded microstructure was caused by the complex heat transfer due to gradient energy density and locations during building. Correspondingly, gradually increased microhardness was achieved from the bottom to the top and there was no obvious difference from the bottom to the middle.

•
Lack of fusion defects resulted in a decrease in plasticity in the LED built zone. The unmelted powder and voids in following fracture morphology were observed by SEM.