Investigation of the Influence of Powder Moisture on the Spreadability Using the Spreading Tester

An inert gas atomized IN718 powder grade of voestalpine Böhler Edelstahl GmbH & Co KG, available under the designation L718 AMPO, was used for this study. Fig. 1 shows the mass-related particle size distribution measured by laser diffraction using a Sympatec HELOS/BF system. The distribution is minimally skewed left but overall, very uniform and narrow. The particle size ranges from 15 to 55 µm and the mass-median-diameter (d_3, 50) of the powder is 30.4 ± 0.1 µm.

In Fig. 2 an SEM image of the L718 AMPO powder used, taken with a JEOL JCM-6000, can be seen. The shape of the particles is mostly well spherical, but, nevertheless, there are also a few non-spherical particles as well as fused agglomerates consisting of smaller particles (<10 µm). Furthermore, some of the particles are partially covered with satellites, and yet, the overall shape is considered to be more or less spherical and uniform.

The flow time according to ISO 4490:2018 was also measured. The powder was found not to flow freely through the calibrated Hall funnel. Nevertheless, it could be spread using the Spreading Tester without any problems.

To investigate the influence of adsorbed water on the spreadability of the used powder and therefore on the quality of a spread powder layer, the powder was conditioned once in a dry and once in a humid state. For the dry state, the powder was stored in a desiccator above silica gel for four weeks, which is intended to remove all moisture from the environment, making the powder “dry”. For the humid state, the powder was also stored in a desiccator for four weeks, but above a saturated sodium chloride solution, through which a relative humidity of 75% could be achieved [15], making the powder “humid”.

To test the spreadability of powders, a special tester was developed—the so-called Spreading Tester shown in Fig. 3. It is controlled by a LabVIEW program, and thus parameters, such as the number of spread layers, layer thickness, spreading velocity, the mimicked machine, and much more, can be varied. In addition to these software parameters, other parameters, such as the distance between the recoater blade and the building platform—also known as the gap size—the angle of the recoater blade, or even the recoater blade itself, can be changed. Yet, in this study an EOS M280 machine was mimicked using an HSS recoater blade with a rhombic shape as shown in Fig. 4.

In Fig. 5 the simplified spreading strategy of an EOS M280 machine can be seen. First (1), the building platform moves down and thereby adjusts precisely the layer thickness. Next (2), the recoater blade moves forward, spreading the powder on the building platform. Now, the powder would be selectively molten, but since the Spreading Tester is only used to spread layers of powder, this step is not necessary. At the end (3) the building platform moves downwards to protect the layer when the recoater blade moves back to its initial position.

In this study, two different evaluation methods were used. The first one investigates the formation of empty spots in a single spread layer, whereas the second one investigates the surface roughness and flatness of several spread layers. Both methods are now explained in more detail.

In other words, examining the formation of empty spots means examining the surface coverage of a single spread powder layer. It is of course essential to spread a dense powder layer in order to selectively melt enough material in the actual layer and thus prevent defects during printing. To do so, two aspects are decisive, namely the gap size (δ) and the layer thickness (d_l). Thus, the layer thickness in the first layer is determined by δ, and it depends on the layer thickness if, in each further layer, it is equal to the gap size because of two facts: the first one is that there are voids between the particles in the layer even in case of maximum packing density (74% for spheres of constant particle size) and therefore, after selectively melting the powder, the remaining now solidified area is lowered compared to the surrounding powder as shown in Fig. 6. There the initial gap size (δ_i), the height of the selectively molten material in the nth layer (h_M, n) and the height difference by which the surface of the molten material is lowered due to the voids also for the nth layer (h_V, n) are marked. Then to grant a uniform process in every further layer, the layer thickness has to be adapted to the gap size as shown in Fig. 7. Of course, this can also be done the other way round so that the gap size is defined through the layer thickness. Thus, only if the layer thickness is equal to the height of the molten material in the first layer (d_l = h_M, 0), the upper end of molten material would be at the same level as the initial position of the building platform in the next layer. Therefore, the relationship given in Eq. 1 must be fulfilled, in which v_f is the amount of the void fraction between the powder particles.

Now, if the gap size was smaller than all of the particles, no powder particles could pass (Fig. 8a). The resulting layer would then probably have a high amount of empty space (approaching 100%) when spread on solid ground like the building platform or, later, the molten material from the layer below.

If the gap size during spreading on these grounds is within the range of the size of the particles in the powder used, at least the smaller particles can pass. Yet, this will result, on the one hand, in a segregation of the powder (Fig. 8b), whereby smaller particles remain in the spread layer and bigger ones remain in the powder heap. On the other hand, a bigger particle can get stuck in front of the recoater blade and thus block it for smaller particles to pass (Fig. 8c), which would result in the formation of empty spots as simulated in [16].

All these aspects do not apply, however, as soon as the powder is spread in a loose powder bed, as the powder can be compacted or easily moved in this bed as it is not fixed locally (this does not apply to electron beam melting, as there the powder is slightly sintered before each layer). As shown in [17], not only the gap size influences the formation of these empty spots but also the spreading velocity. Thereby, it was observed that the faster a powder is spread, the less dense the resulting layer.

Since the powder is not melted during the examination with the Spreading Tester, the correlation from Eq. 1 does not have to be taken into account, but the following applies to grant a uniform process: the gap size is equal to the layer thickness. In order to investigate the formation of empty spots, just one single layer of powder is spread for which the gap size has previously been adjusted precisely. After that the builder—on which the layer has been spread—is removed from the tester and placed under a digital microscope (Keyence VHX-5000). There, three-dimensional images are taken at nine spots (shown in Fig. 9) in the spread powder layer using 800 × magnification. These images each consist of 30 single images that are stitched together and therefore have a size of about 1.1 × 1.1 mm. This image acquisition process is always done carefully in order not to change the morphology of the spread layer.

In Fig. 10 an example of these three-dimensional images is shown. Taking a closer look at the magnified area in the upper left corner, it can be seen that the software of the used digital microscope somehow smoothens the surface, since the area right next to the particles is not dark blue—as it should be if the height at this position were zero. Therefore, the building platform’s height is not always zero, but somehow distributed within a few microns (which is in fact an—inevitable—artefact generated by the software). Now to evaluate the amount of empty space between the particles, the height data of each pixel in the image are used. These data can be exported by the microscope software and then be plotted as shown in Fig. 11. Obviously, the height distribution consists of two single peaks where the lower peak can be assigned to the empty spots and the higher one to the particles. The amount of empty space (ψ) can now be quantified by first fitting peaks with a Gaussian shape into the height distribution and then by calculating the percentage of the area below the lower peak relative to the total area below all fitted peaks, as shown in Eq. 2.

For the spreading test, a gap size of 40 µm and a spreading velocity of 150 mm/s were used. Since the Spreading Tester is only placed in a digestor in a lab, the atmosphere around it cannot be adjusted manually and thus the adsorbed water probably desorbs during spreading. In order to minimize this error, the powder was removed from the conditioning environment at the very last second and filled very quickly into the feeder. Then the test was started. Thus, for testing the formation of empty spots, in which only one single layer is spread, the conditioned state of the powder can be assumed to remain practically unchanged until after the test had been performed.

For describing and assessing the quality of the top spread layer using a powder bed system, a new method has been developed, which is discussed in brief: In contrast to the investigation on the formation of empty spots, several layers are spread. After that the top spread layer is examined using the digital microscope as well, whereby three-dimensional images (Fig. 13) in the same way as previously described and the same nine spots as shown in Fig. 9 are taken. A Python program written by the author then scans the three-dimensional images in two directions (horizontal and vertical) and thereby examines one line profile after the other searching for maxima occurring at the same location in both directions. By doing so, each of these maxima that are present in both scans is equivalent to one particle in the surface, as this is a unique feature in this kind of surface, also seen in Fig. 12. Therein also a perfect surface for this scenario can be seen, which consists of only spherical particles with the same size of the mass-median-diameter (d_3, 50) and is packed closest (hexagonal or cubic, both achieving the same packing density [18]). This method is described in more detail in [19‐21]. The surface evaluated with the method shown in Fig. 13 is shown in Fig. 14, where each particle found is marked with a black plus.

Using the particles detected in the surface as well as the height data of the recorded surface, several parameters can be calculated that allow us to assess the quality of a spread powder layer. For this, four different parameters have been determined in this study, whereby the size of the examined area of course plays a decisive role, which is why it was always kept the same. Thereby, the height (∆H_Par) as well as the distribution of the height (∆Q_Par) of the found particles, the surface area (A), and the enclosed volume of the surface (V_equ) are used as follows:

∆H_Par is defined as the difference in height within the surface between the lowest and highest particles visible from above and is called deviation of height. The parameter is used to describe the flatness of a surface and states that all the larger the deviation, the less flat the surface. Here, however, it should be considered that only images of the same size of the measured area should be compared, since the probability of a very large deviation of height increases with the investigated area.

∆Q_Par is classified as the difference between the third quartile and the first quartile relating to the height distribution of the found particles and is called difference in quartiles. This is also referred to as the interquartile range, which is often used in statistics to describe the width of the distribution. The parameter states that the higher it is, the higher the interquartile range and therefore the wider the distribution. Put simply, the higher ∆Q_Par, the wider the range in which the middle 50% of the particles appear and the rougher the surface. This parameter is virtually independent of the investigated area.

A is the surface area itself, which is calculated by triangulation of the surface. It states that the higher the surface area, the rougher the surface.

V_equ is the so-called equalizing volume. It is determined by placing a virtual perfectly flat plane in the surface. The enclosed volume between the plane and the surface, above and below the plane, that is equal is specified as V_equ. Thus, the volume enclosed above the plane corresponds to the volume of the powder and the volume enclosed below the plane corresponds to the volume of the cavities. The parameter states that the larger the volume, the rougher the surface.

For the spreading tests, a layer thickness as well as a gap size of 40 µm were chosen and the spreading velocity was set to 150 mm/s. As mentioned before the atmosphere around the Spreading Tester cannot be adjusted manually. Thus, again the powder was removed at the very last second from the conditioning environment, filled very quickly into the feeder and the test was started. Yet, since for this investigation a total of 40 layers were spread, it could not be guaranteed that the conditioned state of the powder remained practically unchanged until after the test had been performed, as this takes some time. As a result, this test was even more similar to the real L‑PBF process in which a dry inert gas atmosphere is used during printing, resulting in a significantly lower humidity in the atmosphere than during the spreading test, which was also investigated in [22].

In Fig. 15 the evaluated amount of empty space (ψ) of the dry (0% relative humidity, white bars) and humid (75% relative humidity, black bars) conditioned powder is shown. On the left side of the figure, the evaluated data for each spot and, on the right, the averaged value from all spots is depicted. The evaluated data of each spot shows that the amount of empty space for the humid conditioned powder is always higher in comparison to the dry one. This means that, in case of the humid conditioned powder, the spread layers are significantly less covered with powder. There is also a correlation between the position of the spot and the amount of empty space, which is caused by the tilt of the building platform, which is inevitable as it is not perfectly parallel to the recoater blade and cannot be adjusted with the required precision. Yet, since the building platform was never moved during these experiments, the tilt was always the same and the measured values can be compared to each other without any hesitations. In the area of spot number 5, there is a cover in the centre of the building platform to protect the spindle underneath, as can be seen in Fig. 9. This cover is not perfectly tight and is therefore slightly raised, resulting in a significantly decreased gap size, which is why the evaluated values there are significantly increased.

To sum this up, moisture adsorbed on the powder surface has a clearly negative impact on the coverage of a single spread powder layer as long as the moisture is retained during spreading. Higher relative humidity levels thereby lead to less covered spread surfaces, which would be detrimental for the powder bed fusion process. Thus, if the relative humidity was at the same level during the entire build job inside the machine, it would not be successful.

Since more than one layer was spread here, it should be mentioned that, even despite the non-measurable flow time, the powder could be spread without any problems.

In Fig. 16 the evaluated deviation of height (∆H_Par) can be seen. In contrast to the investigation on the formation of empty spots described previously, there is no clear trend that the deviation of height is always higher for one of the differently conditioned powders. Thus, the average value is considered, which indicates that the spread surface using the humid conditioned powder is slightly flatter. Nevertheless, the values are in the range of scatter of the other.

The evaluated interquartile range (∆Q_Par) is shown in Fig. 17. As with the deviation of height, no clear trend can be observed regarding improved values of any of the conditioned states. Thus, looking at the averaged values on the right, this indicates that the spread surface using the humid conditioned powder is marginally smoother. But again, the values are still in the range of scatter.

Fig. 18 shows the evaluated surface area (A). Again, there is no clear trend regarding improved values of any of the conditioned states. Therefore, looking at the averaged values on the right, it is also indicated that the spread surface using the humid conditioned powder is slightly smoother, whereby again the values are more or less within the scatter of the other.

In Fig. 19 the evaluated equalizing volume (V_equ) is shown. Here, the same trends referred to the single positions seen for the surface area before can be observed. This further means that there is no clear trend regarding improved values of any of the conditioned states. Looking at the averaged values on the right, it is again indicated that the spread surface using the humid conditioned powder is slightly smoother, whereby the values are more or less within the scatter of the other.

To conclude this test, all parameters indicate the same trend that, if a larger number of layers has been spread, the top spread surface for the humid conditioned powder will be slightly flatter and smoother, whereby the parameters describing the surface are always very similar and within the scatter range of each other. Therefore, it can be said that the relative humidity at which the powder is stored does not influence the quality of the top spread layer after a total of 40 layers as long as the relative humidity around the spread powder bed is much lower. It is therefore assumed that the desorption of the adsorbed water on the powder surface happens relatively fast. Yet, when combining both investigation methods—spreading just one single layer and spreading 40 layers—it can be concluded that there must be a critical humidity level at which the quantity of adsorbed water is so high that the spreading of layers cannot be sufficiently successful. Further, if this high level of humidity were to be present in the chamber during the entire printing process, then humidity would certainly have a clearly negative influence on the whole process, as shown in the investigation of the formation of empty spots. Yet, although this statement clearly applies at least to the examined L718 AMPO powder, it cannot be claimed that it applies to other powders as well.

Furthermore, there are two additional aspects that have to be considered. The first one is that it takes about 30 min to fill the powder into the machine before a real build job is started. During this time, the balance between adsorbed water and relative humidity in air is already established and the former is thus changed, provided that the powder is stored at a different humidity level than that prevailing in the room where the 3D printer is located. The second one is that the process chamber is flooded and purged with dry inert gas to establish the required low oxygen level inside, which takes an additional 30 to 45 min. Due to this constant dry gas flow inside the machine, the water will desorb on its own, maybe even before manufacturing starts. Thus, such a conditioned powder would only be processed about an hour later and would have enough time and opportunities to desorb the adsorbed water, thus more or less approaching the same state as powders stored in dry conditions.