Selective laser melting of CoCrMo: an evaluation of part density distribution across the build plate

. Selective laser melting (SLM) employs the use of powder as the feedstock to manufacture the desired geometries in a layer-by-layer fashion. The part density of manufactured components is one factor used to evaluate part quality since the density has been shown to influence the mechanical properties of SLM components. This study investigated the effect of build location on the part density of selective laser melted CoCrMo. The results showed that the part density varied across the build platform due to the powder packing efficiency. The main role in the density variation is due to the inconsistent powder distribution and packing efficiency across the build plate, negatively affecting the part density.


Introduction
Components manufactured by selective laser melting (SLM) have been shown to be reputable. They are being employed in several important industries, such as biomedical and aerospace, where part quality forms an essential requirement [1], [2]. In order to improve part quality, multiple studies have been focused on the effect of processing parameters on part quality in SLM [3]- [7]. Most of these studies focussed around employing different process parameters such as laser power, scanning speed and strategies along with various layer thicknesses. However, limited research has been carried out on the effect of a component's build location on part quality in terms of part density.
SLM employs the use of powder as the feedstock to manufacture the desired geometries layer by layer. Each layer requires a laser to selectively melt the consecutive layers of powder according to a predefined pattern to form the desired geometry. The laser beam acts as the heat source and directly interacts with the powder in a localised region, generating a melt pool and resulting in a solid track [8]. Once the predefined laser pattern is completed, the solid tracks form a solid layer, and a new layer of powder is deposited onto the build plate. The powder is collected from a supply chamber by a recoater arm and is spread across the build platform through a raking mechanism, with the excess powder being conveyed into the overflow chamber. After depositing the new layer of powder, the process is repeated until the desired geometry is formed. During the powder layer deposition, the weight force of the powder decreases above the current powder layer as the recoater arm moves along the build platform. As this weight force decreases, the powder layer being deposited may be unevenly distributed across the build platform resulting in different packing efficiencies [9]. Moreover, when deciding on a layer thickness to be used, the upper threshold of the powder's particle size distribution (PSD) should be considered. When particles are deposited onto the build plate at a fixed layer thickness, only particles that fall within the layer thickness height will be deposited, whilst larger particles will be scraped across the build area. Therefore, the layer thickness is a limiting factor when it comes to the upper threshold of PSD ranges [10]. Therefore, it is important to understand the mechanisms of powder spreading, which causes variations of the powder distribution and ultimately variations in the manufactured component and its quality.
The part density of manufactured components is one factor used to evaluate part quality since the density has been shown to influence the mechanical properties of SLM components [11]- [13]. The powder bed density (PBD) is an indication of the packing density of the powder during the SLM manufacturing process; since the densification of components involves the melting of particles, the PBD has an influence on the part density. The PBD is highly dependent on the powder deposited, which includes the raking mechanism, the powder PSD, particle morphology and layer thickness. Several studies have shown that an increase in PBD is achieved through the addition of finer particles; hence an improved particle packing efficiency and part density were achieved [12], [14]. In order to evaluate the PBD, Jacob et al. [15] developed an in situ method to measure the PBD of multiple layers deposited on a powder bed fusion (PBF) machine build plate during the manufacturing process. The main idea behind the approach is to capture the metal powder spread across the build plate within an enclosure during the PBF process. The PBD distribution was measured across the build platform and showed that the PBD increased as the distance from the powder supply chamber increased. In terms of part density distribution across the build platform, Pal et al. [9] analysed the density, porosity and tensile properties of specimens produced at different build plate locations. Part densities of specimens manufactured nearest to the powder supply chamber were higher than those of the specimens manufactured nearest to the overflow chamber. Similar results were seen for the mechanical properties. The cause for the inconsistency of results was due to differences in the packing densities of the powder particles along the powder spreading direction.
In the current study, the effect of build location on part density was investigated on CoCrMo parts that were manufactured using SLM. The variation in part density that occurs during manufacturing across the build plate was investigated by focusing on the influence that extrinsic powder characteristics have on the part density at different locations.

Materials and methods
Specimens were selectively laser melted from a CoCrMo alloy (CO-538-1) supplied by Praxair Surface Technologies and manufactured to fall within the ASTM F75 standards. The chemical composition of the powder as reported by the supplier is shown in Table 1: The SLM machine used to manufacture the specimens was the ORLAS Creator equipped with a 250 W 1070 nm Yb fibre laser. The SLM machine has a build volume of ø100 mm x 100 mm and utilises a rotational recoater arm. Therefore, to evaluate the part density and powder distribution across the ø100 mm build platform, twelve cuboid specimens and five PBD capsule specimens were distributed across the build plate, as shown in Figure 1. The dimensions of the cuboidal specimens were 10 x 10 x 10 mm 3 and the PBD specimens have an internal cavity of 6 cm 3 to capture powder for analysis. Specimens were manufactured with the same laser energy process parameters across two layer thicknesses of 25µm and 30µm, respectively. The SLM manufacturing parameters are shown in Table 2:  Archimedes' principle was applied in measuring the density of the cuboidal specimens. The weight in air and under distilled water was measured three times with an electronic scale with an accuracy of ±0.1 mg, and the mean values were expressed as the respective mass. The temperature of the distilled water was measured to the nearest 0.1 °C to calculate the actual density. From the calculated water density and cuboid specimens' mass in air and immersed in water, Archimedes' principle was used to calculate the part density.
The PBD specimens were designed and tested according to the findings of Jacob et al. [15]. The PBD specimen was designed with a 0.2mm lid that encapsulates powder during the manufacturing process within the 6 cm 3 internal cavity to analyse the PBD. Encapsulated powder within the PBD specimen was removed by removing the 0.2mm lid with a punch and hammer. The manufactured PBD specimens encapsulated powder mass was measured by determining the difference in mass between the closed specimen with encapsulated powder and the empty specimens in air. Whilst the true volume of the cavity within the specimen was determined by filling the cavity with distilled water at a known temperature and density. From the powder mass and volume, the PBD was determined.
The encapsulated powder from each PBD specimen was analysed through static automated imaging. A Malvern Morphologi G3 automated imaging machine was used to gather morphological data on the powders. This morphological data included the PSD, circularity, aspect ratio and convexity of the powders. Gathered information from the PBD and powder within was used to support the findings in the part density variations across the build platform. The density, PBD and powder analyses were performed across the build plate as discussed above. The results indicated that the density distribution across the build plate was inconsistent, with the powder deposited across the build plate having a significant influence. For each set of cubes manufactured, the density variation was labelled as minor if less than 0.1%, medium if between 0.1% and 0.5% and major if greater than 0.5%.

Results
The results of each density distribution were plotted in MATLAB according to a colour scale to represent the results. Blue corresponds to minor density variation, whilst red corresponds to major density variation. The SLM machine's actual build area is 100 mm in diameter; however, due to the positioning of the cube specimens, an octagonal shape was evaluated, covering the majority of the build platform. A representation of the part density distribution across the build platform for both layer thicknesses is shown in Figure 2.

Figure 2: Part density variation across the build platform at A) 25µm and B) 30µm
From Figure 2 above, it is evident that the part density is relatively consistent across the build plate at 25µm with only minor to medium density variations experienced. However, at a layer thickness of 30µm, the density variations increased with major density variations of up to 0.8% seen. A summary of the minimum, average and maximum densities across the build plate for both layer thicknesses is shown in Table 3:  Table 3 shows that the part density recorded is higher at a layer thickness of 25µm than that experienced at 30µm. The relative bulk part densities and PBD were calculated based on the true density of the powder. The true density of the powder was determined through helium pycnometry to a value of 8.3644 g/cm 3 . A maximum relative density of marginally over 100% was measured, this is possible since the reference material was measured by helium pycnometry which does not account for trapped internal porosity of the measured material. The manufacturing of powders still produces small internal porosity within and the helium pycnometry does not account for this. Therefore, the "true reference density" is marginally Similar to how the density variation was analysed, the PBD was analysed in MATLAB. As can be seen in Figure 3, the five PBD specimens allowed for a s quare area to be analysed across the build platform.

Figure 3: PBD variation across the build platform at a layer thickness of A) 25µm and B) 30µm
Similar to the density distribution in Figure 2, the PBD distribution across the build platform for a 25µm layer thickness is more consistent than the PBD distribution at a 30µm layer thickness. When looking at the results across the 30µm layer thickness, the largest density variation and PBD variation occurs at a similar position in the upper region of the build platform. From both layer thickness, the highest density and PBD was seen to the left of the build plate, where the powder is deposited first onto the build platform. A summary of the minimum, average and maximum PBD results across the build platform for both layer thicknesses is shown in Table 4: At a layer thickness of 25µm, both the average part density and the average PBD is higher when compared to the results across the 30µm layer thickness. As mentioned earlier, the powder encapsulated within the PBD specimens were analysed to gather a further understanding of the variations across the build plate and the findings are shown in Figure 4. The powders were analysed for PSD and shape factors such as the circularity and aspect ratio. Similar to how the results of the part density and PBD were displayed through MATLAB figures, the same was done with the powder shape factors. For the results shown in Figure 4, the red colour displays results of particles that are highly spherical, typically close to a value of 1, whereas the blue colour represents particles that are the least spherical when compared to the highest spherical particle apparent.  When comparing the results of the part density, PBD and powder analysis at a layer thickness of 25µm, one region of the build platform stands out the most, that being the left to the central region. In this region, the lowest part density and poorest particle shapes are found. For the layer thickness of 30µm, similar results are seen. A summary of the minimum, average and maximum shape factors across the build plate is shown in Table 5. The results presented above show how the packing efficiency is influenced by the powder particle shape, and this ultimately affects the part density. The results are discussed below further.

Discussion
The primary cause of the variation in part density is due to the powder packing density variations. The regions of the build platform that experienced the highest part density represented higher powder bed densities due to more spherical particles present. When powder particles are small to medium in size and highly spherical, they present a uniform packing efficiency which allows for steady penetration of the laser during melting [16]. With a steady penetration of the laser, a stable melt pool is generated, and less spattering of powder particles occurs. Larger irregularly shaped particles present within the powder layer reduce the packing efficiency and creates larger open spaces within the powder layer. These open spaces increase the laser penetration depth and cause a rapid increase of thermal energy within the heat-affected zone. A rapid increase of thermal energy causes the gas trapped within the interparticle spaces to suddenly expand and results in a large amount of material spattering [9], [17]. Spattering causes material loss within the melt pool and results in a smaller melt pool. From the results, it is evident that the PBD of the 30µm layer thickness was poorer, thus having a decreased packing efficiency. Therefore, the spatter is expected to be higher due to larger open spaces between the powder particles; this is true, as can be seen in Figure 5. Moreover, with the layer thickness being larger, the volumetric energy density will be lower, possibly causing a more unstable melt pool with increasing spatter. From Figure 5 C and D, it is clear that the 25µm layer thickness had less build-up of spatter material as compared to 30µm. With the lower layer thickness having less spatter, it shows that the melt pool is more stable with sufficient thermal energy being transferred into the powder bed. Therefore, if the laser process parameters were optimised for the 30µm layer thickness, a more stable melt pool would be expected with reduced spattering and improved part density. From the results, it is conclusive that the powder particle packing efficiency influences part density. When the laser processing parameters are optimised for the desired layer thickness, less part density variation is expected across the build platform.

Conclusion
This study investigated and reported on the part density variation across the build platform of selective laser melted CoCrMo. The part densities of the specimens depended on the powder deposited at that region of the build plate. The variations were due to poorer powder particle packing efficiencies that reduced the PBD and ultimately affected the part density. Results obtained from this study are on a specific SLM machine, but the gained knowledge could be applied to other powder bed fusion systems that utilise a recoater spreading system for powder deposition.
A further consideration for the study of powder variations across the build plate could be due to the interaction between the powder and recoater arm. The ORLAS Creator SLM machine has a circular recoater arm pivoting at the build chamber's centre. Reaction forces on the powder will differ depending on the distance from the point of rotation, and the recoater arms velocity will also have an influence. Therefore, with different reaction forces, the packing efficiency of the powder bed could be influenced. From the results presented in this study, an influence on packing efficiency affects part density.
Part density was measured by Archimedes' principle, which measures the bulk part density. Further analyses could be investigated in the future through 3D tomography, which will determine the part density along with the shape and distribution of the porosity. By visualising the shape and distribution of the porosity, process parameters could be changed to reduce porosity and ultimately achieve higher part densities.