Optimum Process Parameters for Direct Metal Laser Sintering of Ti6Al Powder Blend

Titanium aluminides have become the preferred titanium-based alloys for high temperature applications due to their resistance to oxidation at elevated temperatures. However, the inherent limitations of the conventional methods of manufacturing have adverse effects on the mechanical properties of the alloy and limit its applications. The current study focused on determining the optimum process parameters that could be used to produce a Ti6Al alloy with required microstructural properties and complex geometrical configurations using the direct metal laser sintering method. Single tracks were produced at laser powers of 150 W and 350 W over a wide range of scanning speeds. Continuous tracks were achieved only at a laser power of 150 W at corresponding scanning speeds of 1.0 m/s to 1.4 m/s. A cross sectional analysis was conducted on the single tracks and 1.2 m/s emerged as the optimum scanning speed. 3D objects were manufactured at optimum process parameters of 150 W, 1.2 m/s and a hatch distance of 80 μm. The microstructure of the 3D objects was homogenous which attests that the direct metal laser sintering method could be used to produce Ti6Al parts with the desired mechanical properties and geometrical complexity.


Introduction
Titaniu m has been the preferred metal of choice for many engineering applications due to its outstanding specific strength, corrosion resistance and strength at high temperatures [1]. Ho wever, the structural integrity of titanium is limited by the increased oxidation of the alloy at temperatures above 600º C [2]. Titaniu m alu min ide (TiAl) alloys have been proven to overcome the oxidation disadvantage which has provoked intense academic and industrial research into developing TiAl-based alloys for high temperature engineering applications. TiAl alloys can lead to substantial payoffs in aircraft engines, industrial gas turbines and automotive parts due to their unique thermo mechanical properties. TiA l based alloys have a strong potential to increase the thrust-to-weight ratio in an aircraft engine [3]. These alloys could reduce the structural weight of high-performance gas turbine engines by 20-30%, wh ich wou ld enhance engine performance and fuel efficiency and potentially be a replacement of Ni-based superalloys that are nearly twice as dense (heavy) as TiAl-based alloys [4].
However, the conventional methods of manufacturing the alloy have limitations in the form of metallurg ical defects, such as porosity, shrinkage and inhomogeneous microstructure which adversely affect the mechanical properties of the alloy [5]. Because of these limitations, manufacturing of TiAl co mponents has focused on producing simp le shapes and moderate dimensions only, to reduce the risks of metallurg ical defects. However, the dynamics in the market have demonstrated that there is a high demand for large manufactured TiA l co mponents for engineering applications at high temperatures [6]. To satisfy the industrial demand and the requirements of producing large TiA l parts with geomet rical, technical and functional properties of comp lex shapes and homogenous microstructure, an alternative processing route(s) that satisfies the industrial requirements has to be exploited.
Direct metal laser sintering (DM LS), a subset of additive manufacturing (AM) which has received considerable attention due to its unique capabilities of manufacturing large co mp lex shapes, could be used to produced TiAl parts with structural integrity for high temperature applicat ions. The DMLS manufacturing technology is recognised as a key technology in the Fourth industrial revolution, wh ich can complement the inherent limitations of the conventional methods of manufacturing [7]. It is a monolithic manufacturing technology that manufactures 3D objects additively layer by layer wh ich is diametrically opposed to the conventional subtractive manufacturing methods of manufacturing [8]. The technology can be used to manufacture 3D objects with co mplex geo metry and thin intricate walls, wh ich is impossible to achieve using the conventional methods of manufacturing [9]. The manufacturing of titaniu m alu minide based components with specific geo metrical and functional near net shape characteristics for high-temperature operations in gas turbines, aero-engines and automobiles would great ly reduce the time spent on manufacturing, reduce assembly and maintenance cost, avoid waste of manufacturing materials, imp rove performance reliability and weight reduction, hence the cost of manufacturing operations will be greatly reduced. The use of such intricate geometries would lead to fuel consumption and emission reduction and sustainable development at a lower cost [8].
It is already envisaged that the successful production of TiAl near net shapes would enable the production of aircraft engines with greater propulsion efficiency of a 20% reduction in fuel consumption, a 50% reduction in noise, and 80% reduction in NOx emissions compared to similar conventional engines [10]. Fro m the above-mentioned, it is clear that there is a need for the development of TiAl components with near net shape for high value engineering applications at high temperatures. The study was aimed at determining optimu m process parameters that could be used to manufacture Ti6Al alloy parts with comp lex geometry and homogenous microstructure for engineering applications at high temperatures.

Materials and Methods
The experiment was conducted using pure Ti (CP Ti, grade 2) and 99.8% pure Al spherical gas atomized powders procured from TLS Technik Gmb H. The powder was prepared for the in-situ alloying process by mixing 94 wt% Ti and 6 wt% Al. The Ti part icle size was <45 µm and the Al powder particle size ranged from 20 -45 µm.
An EOSINT M280 machine supplied by EOS Gmb H was used for the experiment. The laser spot diameter of the mach ine is ~80 μm. A Ti6Al4V substrate was used with a uniform powder deposition thickness of 60 µm. Single tracks were produced at laser powers of 150 W and 350 W over a wide range of scanning speeds (V = 0.4 -3.4 m/s). All tracks were of length 0.02 m. For each scanning speed, three single tracks were produced.
The samples were mounted and metallurg ically prepared for optical microscopic analysis according to well-known procedures described in the literature [11,12]. Cubes of dimension 10 mm x 10 mm x 10 mm (length x width x thickness) were produced for microstructural analysis. The samples were metallurg ically prepared and etched with Kroll's reagent for examination. The d istribution of Al in Cp Ti was determined in a scanning electron microscope (SEM) with an X-ray energy dispersive spectrometer (EDS). The surface roughness of the samples was measured with a Surftest SJ-210 portable surface roughness tester.

Single Tracks
Single tracks are the basic building units for additive manufacturing using the DM LS process. The geometrical characteristics of the single tracks have a decisive effect on the mechanical propert ies and surface morphology of DM LS built parts [9]. It is, therefore, ob ligatory to determine the optimu m process parameters that could be used to produce continuous single tracks that would ultimately be used to produce the 3D object. The morphology of the solidified mo lten liquid (single tracks) produced in the current experiment that was conducted at laser powers of 150 W and 350 W with a wide range of scanning speeds is shown in Fig. 1.
Fro m the top surface analysis, the single tracks were found to be continuous only at scanning speeds of 1.0 m/s, 1.2 m/s and 1.4 m/s at a laser power of 150 W. The continuity or discontinuity (irregular or balling) of a track is determined by the hydrodynamic move ment of the mo lten liquid wh ich solidifies to form the tracks [13]. The melting process begins with a rapid rise of the surface temperature of the powder particles resulting in the surface melting of the powder particles. The heat then flo ws to the core of the powder and melts the powder to form a mo lten pool. The characteristics of the molten pool are governed by the principal process parameters (laser power, spot size and scanning speed) and the powder layer thickness. These principal p rocess parameters determine the laser energy input which influences the geometrical characteristics of the solidified single tracks [14]. A continuous track is formed when the is an optimal co mb ination of the powder layer thickness, laser power and scanning speed. Discontinuous tracks are produced when the combination of the principal process parameters is not optimal [8]. The literature reveals that several authors attempted to explain the phenomenon behind the behaviour of the molten pool and the solidified single t racks. Several theoretical and experimental models have been proposed to predict the behaviour of the molten pool and its solidification mechanism [13, 15,16]. There are satellites at the edge of the single tracks ( Figure 2) as reported by previous authors [9,17]. Satellites are formed due to the incomplete melt ing of the powder particles in the peripheral zone of the laser spot. Since they are formed towards the end of the solidification process, there is insufficient transfer of the laser energy to melt them co mpletely, as a result, they stick on to the surface of the tracks [17]. Satellites are generally considered as surface defects, which can have adverse effects on the mechanical properties of the final 3D objects [9].
The behaviour of the single tracks as a function of scanning speed was investigated and the results are shown in Figure 3. It was observed that at both laser powers of 150 W and 350 W the widths of the tracks reduced with increasing scanning speeds. Increasing the scanning speed reduces the laser energy transfer, which leads to a reduction in the temperature of the mo lten pool. The lower melting temperature would produce a limited mo lten flu id of high viscosity which would obstruct the free movement of the mo lten liquid hence reduction in the width of the single tracks. Conversely, a lo wer scanning speed leads to higher laser energy density and the resultant increase in temperature of the mo lten pool. The higher temperature within the mo lten pool would lead to a larger amount of liquid phase with lo w viscosity which would automatically enhance the easy flowing of the molten pool and the subsequent increase in the widths of the single tracks. Yadroitsava et al. [9] experimented with different powder layers of Ti6A l4V and reported similar observations. Dzogbewu et al. [7] also focused on determining the optimu m p rocess parameter for Ti15Mo and noted a reduction of the track widths with increasing scanning speeds and increasing of the track widths with decreasing scanning speeds.  Since the DM LS process is a layer-by-layer process, the cross sections of the single tracks were examined to determine the extent to which the laser energy density was able to melt the powder and penetrate into the substrate (previous layer). To ensure there is strong metallurg ical 'welding' between the layers, the laser energy density must be enough to melt the powder and the substrate (previous layer). As depicted in Figure 4, the laser energy density decreases with increasing scanning speeds. Yadroitsev (2009) [18] noted that it is required that the laser energy is enough to re-melt the previous layer to ensure the stability of the molten pool.
During the DM LS process, three kinds of melt ing phenomenon can occur. The first phenomenon occurs when the laser energy density is not enough to melt the powder and penetrate the substrate, which is normally termed as poor penetration (Fig.5A). The second is when the laser energy is enough to melt the powder and penetrate the substrate to the desired depth, which is known as conduction mode (Fig.5B). The third melting condition is known as keyhole mode conduction (Fig.5C). A keyhole mode normally occurs when the co mbination of the selected process parameters leads to high laser energy absorption into the substrate and subsequent 'drilling' of the substrate. The drilling due to the absorption of the high laser energy contributes significantly to pore format ion in the final DM LS products. Eagar and Tsai [18] and Yang et al. [17] noted that for a conduction mode, the mo rphology of the cross-section of the solidified molten pool should form a semi-circular 'U' shape ( Fig.5B), wh ile the morphology of the keyhole mode should resemble a 'V' shape (Fig.5C). King et al. [16] also pointed out that the keyhole mode would occur if the penetration depth is greater than half the width of the track. Several authors have proposed theoretical and experimental models to predict the threshold that leads to the keyhole mode [19,20].
For the current experiment, only the single track was produced at laser power 150 W and scanning speed of 1.2 m/s morphology resemb led a U shape which corresponds to the conduction mode (optimu m process parameter - Fig.5B). A ll the other process parameters did not meet the characteristics that defined optimu m process parameters. At 150 W laser power with corresponding scanning speeds fro m 1.4 m/s -2.2 m/s and at 350 W laser power with corresponding scanning speeds fro m 1.6 m/s -3.4 m/s, poor penetration into the substrate was observed (Fig.5A). This is the effect of high scanning speeds, resulting in insufficient time for the laser beam to melt the powder completely and penetrate the substrate. The keyhole mode (Fig.5C) was observed at 150 W laser power and corresponding scanning speeds of 0.4 m/s -1 m/s and 350 W at scanning speeds fro m 1 m/s -1.2 m/s. Th is could be attributed to the slow scanning speeds. At slow scanning speed, the laser beam radiation dwells at a particu lar spot for a relatively long time, wh ich results in high laser energy absorption into the substrate. This resulted in the laser beam melting the powder and drilling very deep into the substrate forming the V shape keyhole mode profile (Fig.5C). As was observed in Fig. 3, the penetration depths of the single tracks decreased with increasing scanning speeds, since the time for the laser radiat ion at a part icular spot reduced with increasing scanning speed. This observation concurs with the reports of prev ious authors [7,15]  The height of the tracks was also investigated since it determined the surface morphology of the final 3D parts. Due to the single track side-by-side and subsequent layer-by-layer manufacturing method of the DMLS process, the nature of the track height affects the even deposition of the powder on the previous layer. A rough surface would lead to inho mogeneous powder deposition, which would t rigger inconsistency in melt flow in the subsequent layers, resulting in a relatively rougher surface of the subsequent layer. Therefore, it is obligatory to optimize the melt flow to ensure the production of single tracks with suitable geometrical characteristics in order to produce a non-porous dense DMLS object.
However, the temperature distribution of the Gaussian laser beam that melts the metallic powder is not uniform [21], wh ich leads to a non-uniform temperature gradient of the molten pool. The un-even temperature in the molten pool provokes a surface tension gradient between the center and the edges of the molten pool, wh ich induces ripple (un-even surface) formation as the molten metal solidifies [22,23]. According to Körner et al. [24] and Körner et al. [25], the Gaussian laser beam melting process is very co mplex, and it is governed by laser beam absorption, Marangoni flow, viscosity, surface tension, capillary effects, gravity, etc., wh ich leads to stochastic melt tracks with irregular, corrugated track heights. From Fig.3, it could be observed that the track heights were irregular at both laser powers and scanning speeds which concur with the observation of most previous authors [11,17]. However, single t racks produced at laser power of 150 W (Fig.3A) demonstrated a relatively less irregular behavior than single tracks produced at 350 W (Fig.3B). It could be inferred fro m this observation that within an optimu m process parameter window the heights of the single tracks are less irregular, hence tracks produced at optimu m process parameters would demonstrate less irregular t rack heights, which implies that samples that are produced at optimu m process parameters would demonstrate relatively s mooth surfaces. Yang et al. [17] produced Ti6Al4V single t racks at 100-400 W laser powers and reported a nearly uniform track height. The authors explained that the uniform track heights were due to the low (20 -40 µm) powder layer thickness.

Single Layers
Based on the hierarchical design princip les proposed by Yadroitsev et al. [9] for determining optimu m p rocess parameters for the DM LS process, single layers were produced (Fig.6) at laser power of 150 W with the corresponding scanning speed of 1.2 m/s to determine the optimu m hatch distance and the homogeneity of the Al in the Ti alloy mat rix. Two scanning strategies (single scan and re-scan) were emp loyed to produce the single layers at three different hatch distances (80 µm, 90 µm and 100 µm). The optical and scanning electron microscope investigations revealed that the powders melted completely to form ho mogeneous layers of the alloy (Fig.6). There was complete overlapping of the single tracks to form dense single layers which were metallurg ically fused onto the substrate. Satellites were observed on the surfaces of the single layers after the single scan. These satellites were removed after the rescanning process. The surface roughness of the samp les after the single scan was recorded as 10 ± 1.2 for the Rz, and after the rescanning it reduced to 6 ± 0.9. The d ifferent hatch distances did not have any significant effects on the distribution of the Al in the Ti alloy matrix and the surface roughness of the single layers. All the samples demonstrated similar surface morphology at all three different hatch distances.

3D Object Analysis
To ensure a high degree of overlapping, cubes were produced (Fig.7A) at hatch distance of 80 µm to investigate the properties obtained by using the optimu m p rocess parameters (150 W, 1.2 m/s and 80 µm) to build a 3D object. Using the optical microscope, it was noted that the powder melted co mpletely and there were no visible satellites at the top, side and front planes (Fig.8). The front (y-z p lane) and the side (x-z plane) view microstructures showed overlapping of subsequent scanned tracks due to the optimu m hatch distance. The overlapping edges of the single tracks were curved and were noted on the front and the side views. These curves signify the merging of each layer into the nearby layers ( Fig.8B & C).
Fro m Fig.8A, α+β microstructure is observed, with the primary α grains in the transformed β matrix. In Fig.8B and 8C, needles of fine acicu lar martensite was found inside the prior β grains of the Ti alloy, which signif ies the high cooling rate that characterizes the DM LS process [26]. It was reported that the heating and cooling proces s that occurs simultaneously in the build chamber during the DM LS process occurs at (~ 10 4 − 10 6 ℃/ ) [27,28]. It is this rapid heating and cooling that lead to the formation of the martensitic microstructure of DMLS built parts. The martensitic microstructure which limits the ductility of a DM LS built part can be overco me by post-thermal treatment process [29]. It is also worth mentioning that a section of the materials science research commun ity are of the view that the high rate o f heating and cooling reported during the DMLS p rocess is quite ambiguous and still a matter of research [30].  Backscattered electron (BEC) and secondary electron (SEI) images of the three views (top, side and front) reveal dark spots on the micrographs (Fig.9). These spots are possible pores in the 3D part. These pores are found in the (x-y plane) in-between the solid ified layers. Porosity in-between solidified layers are known as interlayer porosity and can be caused by non-uniform powder delivery. During the layer-by-layer building process of DM LS, if the surface of the previously solidified layer is uneven, then the next powder distribution would be uneven and interlayer pores can form. The non-uniform track height (Fig.3) can lead to such occurrences during the DMLS melting process.
Further analysis was conducted to confirm the p resence of the micropores in the DM LS bu ilt part. The samples were re-polished and optical micrographs (Fig.10) were taken since the optical microscope gives a clear contrast view of such defects. From the optical micrographs (Fig.10) it was confirmed that micropores were p resent in the DM LS 3D built part. The pores were less than 20 µm when measured on micrographs of all three cross -sectional views, hence the samples were considered as well-built dense 3D objects as reported elsewhere [31]. It was empirically proven and generally accepted that DMLS samples with pores size less than 20 µm wou ld demonstrate stable mechanical properties fo r engineering applications and are therefore considered as dense well-built 3D objects [8].
EDS elemental mapping analyses were done for the three views of the 3D part (Fig.11). The Al distribution in the Ti matrix was even and homogeneous. The elemental mapping (Fig.11) and the microstructure in Fig.8 clearly demonstrate that the DMLS process was able to produce a homogeneous alloy of Ti6Al. The EDS investigation revealed an average of 4% Al in the Ti matrix of the 3D part, which implies that 2% of the Al was lost during the DM LS process. This phenomenon of Al loss could be attributed to the thermophysical differences between the two materials as given in Table 1. As indicated in Table 1, Al has a higher specific heat capacity and a higher thermal conductivity than Ti, which results in Al absorbing more laser radiat ion than Ti. The high absorptive capacity of Al means it tends to melt faster than Ti. Secondly, the melt ing point of Al (660 º C) is lower than Ti (1668 º C), hence it will absorb the laser radiation faster and melt at a lower temperature than Ti. As result, Al could evaporate in the process which explains the 2% reduction of Al in the bulk material. The lower density of Al (2710kg/ m 3 ) when co mpared to Ti (4500 kg/ m 3 ) could also have contributed to the loss of Al. Since the density of Al is only 60% that of Ti, Al would be in relat ively h igher concentration towards the top of the melt pool than at the bottom, which could have increased the Al evaporation rate.
The Ti-Al phase diagram in Fig.12 is used to determined which phase is expected at equilibriu m for 6 wt% Al content of Ti at temperature in °C. Two stable equilibriu m states are possible at 6 wt% A l content, namely hexagonal close-packed (hcp) α-Ti which is stable fro m room temperature up to 882 0 C and body-centered cubic (bcc) β-Ti wh ich is stable fro m 882 0 C up to the melting temperature [34]. The Al will form a comp lete substitutional solid solution in the Ti matrix [35].

Conclusions
The successful production of homogeneous, non-porous dense 3D objects (cubes) with the optimu m p rocess parameters demonstrated that the DMLS method of manufacturing is capable of producing the Ti6Al alloy with the desired microstructural properties for high temperature applications. Using the DMLS process would permit the manufacturing Ti6Al co mponents with comp lex geometrical configurations as required in the aircraft and the automotive industries. The net shapes would enhance the cost-effectiveness of the manufacturing process. The reduced weight of the near net shape Ti6Al DM LS built parts could enhance the energy efficiency of turb ines constructed with such parts for high temperature applications.