Polyamide 12 powder ageing in laser beam-based powder bed fusion and its combined effects on powder and component characteristics

In polymer additive manufacturing, the term “Selective Laser Sintering” (SLS) is widely used for the laser-beam-based powder bed fusion technique. It is characterised by a layer-by-layer deposition of powder, which is selectively melted by a laser to form the three-dimensional component. Its counterpart for metallic materials is usually termed “Selective Laser Melting”. Both processes are based on comparable principles (powder bed with layer-by-layer deposition of powder and selective melting using a laser) but differ significantly in technical details. The SLS technique is nowadays already in operation on the production level to manufacture parts and components in small and medium lot sizes. In this context, the term “Rapid Manufacturing” in contrast to the more established “Rapid Prototyping” has been created to indicate this progress (Gibson et al., 2015; Gornet et al., 2002; Schmid, 2018).

A key feature of the SLS process is that during the process, the powder is kept at a rather high-temperature level (within the so-called “Super Cooling Window” [Gibson et al., 2015; Grießbach, 2012; Mielicki, 2014; Drummer et al., 2010; Schmid et al., 2025]), which is quite challenging for polymer stability. To reduce the negative influence of these high temperatures to a minimum, conventional SLS machines used for industrial fabrication are usually flooded with nitrogen during the process. However, there is still a big impact of the process temperature on the remaining unmelted powder: the powder “ages” during each conducted build cycle. This ageing of the powder affects powder properties (Josupeit and Schmid, 2017; Gornet et al., 2002; Chen et al., 2018a; Wudy and Drummer, 2019; Wegner et al., 2014; Wudy et al., 2014; Mielicki, 2014; Martínez et al., 2012; Yang et al., 2023), processing parameters and stability (Gornet et al., 2002; Wegner et al., 2014; Mielicki, 2014) as well as the mechanical behaviour and quality of the parts and components produced (Gornet et al., 2002; Chen et al., 2018a; Wegner et al., 2014; Wudy et al., 2014; Yao et al., 2020; Gazzerro et al., 2023; Gomes et al., 2022). Although some studies dealing with the powder ageing topic in SLS, especially for Polyamide 12 (PA12), exist (Alo et al., 2023; Benz and Bonten, 2019; Dotchev and Yusoff, 2009; Chen et al., 2018a; Gomes et al., 2022), many questions are still not answered. This is mainly attributed to the fact that it is a complicated interaction of several competing and partly counteracting effects and processes, such as increased viscosity and porosity. Depending on the dominating effect, brittle or ductile trends are visible in the mechanical properties. Hence, to consider all relevant aspects, a lot of effort is required. In addition, there is a significant influence on the boundary conditions chosen for such an ageing study. Such a boundary condition is for example if the ageing was conducted in a heating oven in the lab (Josupeit and Schmid, 2017; Wegner et al., 2014; Mielicki et al., 2014b; Wudy et al., 2014; Mielicki, 2014; Martínez et al., 2012; Sanders et al., 2022; Vendittoli et al., 2024) or on the SLS machine (Josupeit and Schmid, 2017; Gornet et al., 2002; Chen et al., 2018a; Wudy and Drummer, 2019; Wegner et al., 2014; Wudy et al., 2014; Stiller et al., 2022; Seigler et al., 2024). Further examples are the atmosphere in which the ageing was conducted (nitrogen or air), the detailed processing parameters and conditions in case the ageing was conducted on an SLS machine and if so-called powder refreshing (the addition of unused powder to aged powder in a certain extent) was included in the study (Sanders et al., 2024; Dotchev and Yusoff, 2009; Zarringhalam et al., 2006) or not (Yao et al., 2020; Alo et al., 2023; Sanders et al., 2022). Many of the publications available so far focus only on specific aspects of powder ageing in the SLS process. However, due to the strong influence of the above-mentioned boundary conditions, their comparability is sometimes difficult.

Nevertheless, at least for PA12 aged in a nitrogen atmosphere under SLS processing conditions, it is a fact that ageing leads to a drop in the melt volume-flow rate (MVR)/melt mass-flow rate (MFR) values due to an increase in viscosity. It is widely accepted nowadays that the viscosity increase is an effect of polymer post-condensation in the process, which results in a significant increase in the molar mass (Zarringhalam et al., 2006; Wudy and Drummer, 2019; Chen et al., 2018a; Mielicki et al., 2014b; Yao et al., 2020; Stiller et al., 2022; Yang et al., 2023).

Why is powder ageing relevant for the SLS process? Aside from the detailed scientific discussion of powder ageing, it has a significant practical impact on the SLS technique. Firstly, the powder needs “refreshing” after each building job: after sieving the unmelted (old) powder of the previous building job, it is mixed with unused (new) powder with a ratio of old to new between 70:30 and 50:50 (either in a fixed ratio or optimised for a constant value of the melt flow ratio (Wegner et al., 2014; Gibson et al., 2015; Wudy and Drummer, 2019; Chen et al., 2018a; Josupeit and Schmid, 2017; Dotchev and Yusoff, 2009; Yang et al., 2023). As the average filling of the powder cake with components is in the order of 10%–15% (Josupeit and Schmid, 2017; Chen et al., 2018a) [other authors say below 30% (Wudy and Drummer, 2019; Wegner et al., 2014]), this leads to significant powder loss. Secondly, at a distinct level of powder ageing the remaining powder has to be discarded completely. Both, powder refreshing and the full discard are required to maintain the geometrical accuracy of the parts produced and to avoid an “orange peel” on their surface. However, the powder loss due to both procedures leads to significant additional process costs. With the cheapest powders for Polyamide 12 being in a price range of about 50 €to 100 €per kg, they are about 10 times more expensive than resins used for conventional processing techniques (such as extrusion and injection moulding).

Due to the rather high price, one important goal in the area of SLS has always been to find a cost-effective way to modify the discarded powder, so that it can be reused for printing. This, to the best knowledge of the authors, has not been fully successful so far. As mentioned above, the various processes on chemical and physical levels as well as their partly contradicting effects make a clear strategy difficult. The described lack of comparability between parts of studies on this topic and the fact that many authors focus on specific ageing effects make it even more difficult. In Table 1, an overview of the literature about ageing for PA powders for SLS is listed, showing, firstly, contradicting effects of ageing on different properties and, secondly, the lack of completeness of testing all relevant methods within one study. Contradicting effects can be traced back to different ageing (in the oven or on the machine) or testing conditions, which in turn limits the comparability between studies. Oven-aged samples are annealed and show increased stiffness and strength, whereas ageing studies on the machine with or without refreshing show no clear trend.

From this overview, three groups can be named: oven-aged, aged on the machine with refreshing and aged on the machine without refreshing. As an example of the oven-aged study, Sanders et al. (2022) tried to simulate multi-jet fusion conditions with this experiment. They concluded that due to oxygen, a thermo-oxidative degradation process took place visibly in discolouration and FTIR measurements. This does not represent the ageing of PBF-powder inside the machine, as this degradation mechanism is prevented by nitrogen atmosphere. In the second group, the machine-aged powder with powder refreshing (Dadbakhsh et al., 2017) conducted a study with different powder states (virgin, mixed and aged) showing the typical post-condensation effects due to the absence of oxygen, which can also be found for ageing studies skipping the refreshing. Despite similar trends, the strain at break seems to vary depending on the study, which is related to the dominating effects in the respective studies. In the work of Chen et al. (2018b), it is not clear if the powder was refreshed, nonetheless, their focus was on the crystalline structures and the processability and recyclability of PA6 in comparison to the commonly used PA12. Even though their study is remarkable, it lacks other characterisation methods for the thermal properties, which are relevant to the stability of the process. By refreshing the powder, a mixed state of the powder is present and affects the properties, thus, the underlying mechanisms can be covered. The last corresponds to our boundary conditions. In the studies from Wudy and Drummer (2019) and Wudy and Drummer (2016), five consecutive build cycles were conducted and both times they found a decrease in MFR and an increase in porosity, furthermore declining bulk density, stiffness and strength were measured. The general trends from Wudy and Drummer (2016) correlate well with the one presented in our study. Gazzerro et al. (2023), who performed at least five build cycles, found in contrast an enhanced stiffness but at the same time reduced crystallinity. Gomes et al. (2022) measured a shift to lower melting peak temperatures, whereas the powder size and the mechanical properties showed no significant trend. After eight build cycles, Alo et al. (2023) reported no change in the particle-size distribution, but the well-known increase in orange peel. Yao et al. (2020) reused their powder eight times and found first increasing mechanical properties which then dropped again. Even though similar but also contradicting findings are present in the literature, no study so far investigated thermal, rheological and mechanical properties with the same boundary conditions (PA12 powder aged during the process on the machine, without refreshing the powder) in one single study for nine build cycles and correlated the effects. Studies with comparable boundary conditions did not investigate the total sum of relevant parameters for sintering, such as packing density, thermal properties and melt flow changes.

Hence, the goals of this research were to qualitatively and quantitatively determine the most relevant effects of PA12 powder ageing in a well-defined SLS ageing study, reduce the external influences to a minimum (no powder refreshing) and keep the ageing study as authentic as possible (powder ageing directly determined on the SLS machine). Throughout the whole study for each build cycle, the powder and the parts were investigated with all relevant methods, of which in the literature only selected investigation methods were compared or correlated. For the characterisation of the different effects on qualitative and quantitative levels, both the powder (bulk density, scanning electron microscopy (SEM), particle size distribution, IR-spectroscopy, plate–plate rheometry, MVR and differential scanning calorimetry (DSC)) as well as the printed parts (DSC, surface quality and tensile tests) were studied.

The SLS processes were conducted on a laser sintering machine of the type “FS402P” modified by LSS (LSS Laser Sinter Service GmbH, Holzwickede, Germany). For the printing, a powder bed temperature of around 175°C (around 10°C below the melting peak temperature of new powder) and a pre-heating temperature of around 150°C were selected. To avoid thermo-oxidative ageing, the sintering process was conducted under a nitrogen atmosphere. The energy input from the laser was 0.017 J/mm² according to the Andrew Number (Schmid, 2018) and the layer thickness was set to 0.1 mm. To keep the powder ageing in the SLS process as uniform and reproducible as possible, the process parameters were kept constant for all SLS build cycles.

The alignment of the specimens in the build chamber is shown in Figure 1. Two categories of specimens were printed: tensile bars according to ISO 3167 Type A (total length 170 mm, testing area 80 × 10 × 4 mm³) (International Standard, 2002) and plates with a dimension of 50 × 50 × 3 mm³. In Figure 1, the tensile bars for the investigation of the mechanical properties are shown in green. In addition, plates (marked in orange) were added to study the effect of powder ageing on the surface quality and the printed part morphology. This plate was chosen due to its position in the middle of the build chamber. All the remaining plates were used for other investigations, which are not included.

The material for this study was “ALM PA 650”, a PA12 powder resin from ALM (Advanced Laser Materials, Temple, TX, USA). A set of new powder was selected at the beginning of the study and used for the first SLS build cycle. Subsequently, after each SLS process the remaining powder from the powder bed was collected and mixed with the remaining powder from the feed and overflow pistons. Afterwards, the powder was passed by a coarse meshed sieve to remove the sintered parts from the part cake. By this step, bigger agglomerates were sorted out. Subsequently, a small portion of the powder (around 200–250 g) was taken as a sample for later analysis. The remaining powder was used for the next sintering process without any powder refreshing. In this way, in total nine build cycles were printed before the amount of remaining powder was too small for a further cycle. For the discussion of the results, the abbreviation “BC” together with the respective number will denominate the corresponding build cycle (“BC 1”–“BC 9”). In addition, “BC 0” will refer to the new, unused material.

The bulk density of the powders was determined according to ISO 60 with and without pre-drying (80°C for 168 h in a drying oven) of the samples. To ensure complete drying, the weight loss was documented over the drying time. A longer time was chosen due to powder piles (containing the powder for all methods with dried powder), which could not be spread due to limited space in the oven. In retrospect, a drying time of 100 h was sufficient. At first, the measurement cylinder was measured for its volume and weighted in the empty state with a digital precision scale. During the filling of the cylinder, it was especially taken care not to compress the powder. After weighing the completely filled cylinder, the bulk density was calculated according to equation (1). For each build cycle, five samples were tested:

Furthermore, the particle size distribution for all build cycles was evaluated with a Bettersizer 2600 device (Bettersize Instruments Ltd., Dandong, China). The particle size was analysed once for each build cycle in the range of 0.02 µm up to 2,600 µm in wet dispersion.

To evaluate potential changes in the polymer chemistry, all build cycles were investigated by ATR-IR-spectroscopy (attenuated total reflection infrared spectroscopy) by using a Bruker Vertex 70 device (Bruker Corporation, Billerica, MA). For each build cycle, the powder was measured three times with 16 scans in each case. The wave number scan range was from 4,500 cm⁻¹ to 450 cm⁻¹ with steps of 2 cm⁻¹ measured at room temperature.

The development of the melt viscosity with an increasing number of build cycles was studied by two different testing techniques: rheometry and MVR measurement.

All rheometry experiments were carried out on a Physica MCR 501 (Anton Paar GmbH, Graz, Austria) rheometer with plate–plate configuration. The tests were performed in oscillatory mode, with a plate diameter of 25 mm, a gap between the plates of 1 mm and with constant overflow of nitrogen. For the tests, the magnitude of the frequency-dependent complex viscosity (|η*|) was measured within an angular frequency range of 0.1 s⁻¹ and 50 s⁻¹ and at a constant deformation amplitude of 0.1%. All measurements were conducted at a temperature of 210°C. Before the rheological characterisation, all powder samples were pre-dried in a vacuum drying oven at 80°C for 4 h and kept under vacuum until the measurement. To ensure a homogenous and constant melting, the powder applied on the rheometer was compacted in a two-step process, starting with preheating the powder for 1 min at a gap position of 2 mm after filling. Afterwards, the final measurement gap of 1 mm was set. It was taken care that the procedure was carried out in a reproducible way each time, that the timespan between the filling and the start of the experiment did not exceed 4 min, and that the nitrogen flow was applied for as long as possible during this period. If one criterion was violated, the test was discarded. This step was done until three valid runs were measured. The determined viscosity data points were fitted with the “Carreau-Yasuda” regression according to (Carreau, 1972; Yasuda et al., 1981) assuming the validity of the Cox–Merz rule, therefore ω = $γ˙$⁠, to obtain the zero shear viscosity (η₀) values for each sample. The equation for this regression is given in equation (2):

where $η(γ˙)$ is the shear rate dependent viscosity [Pa.s], η_∞ is the infinite shear viscosity [Pa.s], η₀ is the zero shear viscosity [Pa.s], λ is the relaxation time [s], $γ˙$ is the shear rate [s⁻¹], a is the width of transition range between zero shear viscosity and the power law regime [−] and n is the power law exponent [−].

From the zero shear viscosity (η₀) of each sample, the weight-average molar mass (M_w) was calculated according to equation (3) applying values for the coefficients k_e (3.5·10⁻¹³) and A (3.55) obtained from (Anton Paar, 2011):

where η₀ is the zero shear viscosity [Pa.s], k_e is the material and temperature dependent constant [−], M_w is the weight-average molar mass [g/mol] and A is the material and temperature dependent relaxation time exponent [−].

MVR or MFR measurements are commonly used for the quality control of SLS powders. All tests were conducted on a device called “MeltFlow” from KARG (Emmeram Karg Industrietechnik, Krailling, Germany) according to ISO 1133. Each sample weighed 4.3 ± 0.2 g and was tested at a temperature of 210°C with a load of 2.16 kg. The tests were conducted in a laboratory atmosphere. After a preheating time of 300 s, the weight pressed the melt through an 8 mm long nozzle with a diameter of 2.095 mm and a measuring length of 30 mm. For each printing iteration, three samples were examined and five measurement points per sample were obtained (resulting in 15 data points per iteration). The powders were specifically not dried before testing.

Potential changes in the peak temperatures (melting, crystallisation) and/or in the degree of crystallinity of both, the powder and the printed parts were examined by DSC. The powder was examined in undried as well as in dried state. For the latter one, the powder was conditioned in a drying oven for 168 h at 80°C. Afterwards, the dried powders were stored in a desiccator to avoid any uptake of humidity. For each build cycle, three powder samples with about 10 mg were weighed in and put in a 40 µl aluminium crucible with a perforated lid. The printed parts were examined in an undried state using the orange marked plate from Figure 1 for each build cycle. Out of each plate, a minimum of three samples weighing 5–7 mg were cut from one corner and put in a 40 µL aluminium crucible with a perforated lid. Contrary to the powder samples, the plates were not specifically dried, because we observed during this study that drying had no detectable influence on the DSC results for the PA12 examined.

The procedure was conducted according to ISO 11357–1 on a “Mettler Toledo DSC 1 Star” system (Mettler – Toledo Inc., Columbus, OH, USA). All temperature cycles (first heating, cooling, second heating) were done under a flooded nitrogen atmosphere with a rate of 10 K/min. The set temperature limits (25°C and 230°C) were held for 3 min before cooling or heating started. With equation (4), the degree of crystallinity (X_c) was calculated. ΔH₀ represents the melt enthalpy of the 100% crystalline PA12 with a value of 209.3 J/g according to Gogolewski et al. (1980):

SEM was included to study changes in the powder appearance and to qualitatively examine cryogenically broken fracture surfaces of the tensile specimens for their porosity. A Tescan Vega II scanning electron microscope (Tescan GmbH, Brno, Czech Republic) was used. The powder was dried and stored in a desiccator before analysing it. For the porosity analysis, broken halves of the tensile specimens (only those from the bottom of the build chamber), which remained after the later described tensile tests (Section 2.2.4), were used. Their shoulder region was cooled using liquid nitrogen, rapidly fixed in a bench vise and broken apart using a hammer. The so-created shoulder fracture surfaces were analysed in the SEM. They were preferred over the fracture surfaces produced in the tensile tests to avoid falsification of the analysis through excessive deformation during (not cryogenic) fracture.

With the “Alicona” InfiniteFocus microscope IFM G4 (Alicona Imaging GmbH, Graz, Austria) the surface topography was determined for the plates. The orange-marked plate in Figure 1 was selected for these tests. For the measurements, the following settings were used: a vertical and lateral resolution of 349 nm and 3.9 µm, respectively, a lens with a magnification of 10, an exposure time of 373 µs and a contrast of 0.98. With these settings an area of 50 mm in length and 6 mm in width was analysed in the middle of one sample for selected build cycles. As the samples were sandblasted after the printing (which made a comparison of the roughness useless), only the surface waviness was considered for the comparison of the different build cycles.

Tensile tests were conducted according to ISO 527–2 (ISO527-2, 2012) on a Zwick Z250 (Zwick Roell, Ulm, Germany) universal testing machine with mechanical clamping to determine Young’s modulus, tensile strength and strain at break. A macro extensometer determined the deformation in strain. For the tests, the green-marked tensile specimens in Figure 1 were used. Consequently, for each build cycle, eight specimens were tested, which were dried before testing for two weeks at 80°C in a vacuum drying oven.

The interpretation of the results and the significance of their differences were statistically ensured according to Tukey’s range test with α = 0.05.

A crucial factor for the resulting quality of the sintered part is the bulk density of the powder. In Figure 2, the bulk density is shown for all build cycles together with selected SEM images of new and aged powders (Figure 3). The new powder (BC 0) has the same bulk density of 0.475 g/cm³ for the dried and not dried state. In the SEM image of the new powder in Figure 3(a), the loose single powder particles with the typical potato shape are visible. With the reuse of the powder, the density reduces continuously until BC 5, after which no significant changes are observed anymore. A comparable decrease was found by Wudy and Drummer (2016). The humidity in the powder shows only a small impact on the density, around 0.005 g/cm³ for the BC 4 powder, for example. Comparing this trend with the particle size, the overall size of the particles did not significantly change with the increasing build cycle. This is revealed by both, the SEM images [Figure 3(a)] and the particle size analysis (Figure 4). However, an increasing extent of so-called “satellite” particles (Alo et al., 2023) stick to the surface and glue the regular particles to agglomerates, as is shown in the SEM images of BC 3 and BC 9. With the resulting partial change in shape, the packing density and thus the powder bulk density decreases. This leads to a higher porosity in the sintered parts, as it is qualitatively visible in Figure 3b. Generally, an increasing porosity level reduces mechanical parameters, such as strain at break, strength and Young’s modulus. As porosity cannot be fully avoided in SLS-printed parts (Khudiakova et al., 2020; Flodberg et al., 2018), a volume percentage <5% is aimed to guarantee proper mechanical behaviour.

The particle size distribution for all build cycles is shown in Figure 4. Overall a slight shift towards smaller particle sizes was detected with respect to the new powder and an increase in particles with a size of 0.2–0.3 µm. For all following build cycles the distribution is congruent. The same observation was made by Berretta et al. (2014), Dadbakhsh et al. (2017), Mielicki et al. (2014a), Alo et al. (2023) and Chen et al. (2018b), arguing with two counteracting effects, namely, coalescence on the one-hand side and fragmenting by expansion and shrinkage during the heating on the other-hand side. In addition to the thermal expansion, the evaporation of absorbed moisture or remnants of the alcohol from the dissolution-precipitation technique can cause cracking. In total, the fragmentation process is more present than the agglomeration, which explains the constant size distribution. Why the agglomerates from the SEM images of Figure 3(a), which were repeatedly observed, were not detected here is not fully understood. It is speculated that the agglomerates were destroyed during the production of the wet suspension required for the measurement. An alternative, yet less probable explanation is that, despite their size, their absolute number is small compared to the other powder particles.

Ageing of polymers is very often accompanied by changes in their chain length and weight-average molar mass. This effect can be detected and even quantified by viscosity measurements. Viscosity and the molar mass are connected as longer polymer chains reduce the mobility of these chains by entanglements and therefore increase the zero viscosity. Corresponding results (viscosity curves) from the plate–plate rheometer measurements are presented in Figure 5. They show the shear-thinning effect typical for polymer melts with higher angular frequencies and an increase in the viscosity with increasing build cycle number. The drop of the viscosity curves at the very low angular frequencies (roughly indicated by the grey area in Figure 5) is the result of an experimental limitation: the plate–plate rheometer used in this study was unable to create a stable measurement condition in this regime. Hence, these data points are invalid and were not considered in the approximation of η₀. This effect was more pronounced for the smaller viscosities, thus, lower build cycle numbers.

As described in Section 2.2.2, the viscosity curves were fitted with the “Carreau-Yasuda” regression to obtain the zero shear viscosity. The corresponding results are shown in Figure 6. As already observable for the flow behaviour in Figure 5, the zero shear viscosity increases with a higher build cycle indicating a rising weight-average molar mass. The corresponding weight-average molar mass values are shown in Table 2. They were calculated from the zero shear viscosity values applying equation (3) and quantitatively confirm the significant increase in (average) molecular chain length with increasing build cycle number. Hence, as it has already been documented in the literature (Yang et al., 2023; Chen et al., 2018a; Stiller et al., 2022; Dadbakhsh et al., 2017; Yang et al., 2020; Sanders et al., 2024; Sanders et al., 2022; Paolucci, 2019; Gomes et al., 2022), the high temperature in combination with the nitrogen atmosphere present in the SLS machine clearly induced post-condensation in the PA12 powder during each build cycle.

Besides the rheometer results, also the MVR results are included in Figure 6. MVR tests are a facilitated rheological testing method, which fits well for the demands in polymer powder bed fusion and is used frequently in this sector. The higher the MVR value, the lower the viscosity of the material. In Figure 6, the MVR values decrease with increasing build cycle number, thus confirming the results from the plate–plate rheometer. Up to BC3 the differences in the MFR are statistically significant.

In Figure 7(a), the mapped plate surfaces of BC 1, BC 3, BC 5, BC 7 and BC 9 are shown with the same scale for the height. Due to the increase in viscosity with rising build cycle number, the surface waviness increases leading to the famous orange peel, especially visible for BC 9 [Figure 7(b)]. It is a consequence of the limited flowability of the melt (Kerschbaumer et al., 2021), which hinders the compensation of surface unevenness created during melting with the laser (Dotchev and Yusoff, 2009). The roughness itself shows no strong difference because all samples were sandblasted in the post-processing resulting in unified surface roughness.

Infrared spectroscopy can detect changes in polymers, such as chemical ageing, on a molecular level. For this, all powders from all build cycles were analysed and the results are shown in Figure 8. The changes in peak height are related to the contacting conditions between powder and ATR crystal surface or CO₂ contamination during the measurement (peaks around 665 cm⁻¹, 2250–2350 cm⁻¹). Therefore, the spectra are quasi-identical and the powder did not undergo any detectable chemical change. The post-condensation of the chains proven by the increasing shear viscosity and the increasing molar mass (Figure 6, Table 2) is not visible in these spectra. This is not surprising, as this reaction does not alter the basic chemistry of the powder. Nevertheless, the increased chain length simultaneously reduces the number of reactive end groups, which is potentially detectable upon a more sophisticated analysis of the spectra (Bredács et al., 2023).

In Figure 9(a), the melting peak temperatures of the first and the second DSC heating runs are compared for the powder and the printed parts (printed plates marked orange in Figure 1). In DSC, the first heating run represents the thermal and mechanical history of the polymer examined. In our case with both, the powder and the printed parts examined, we had one polymer with two different thermal histories. In contrast, the second heating run in DSC, which follows a defined cooling run with controlled crystallisation of the material, allows for the study of the intrinsic material properties and their changes, indicating for example ageing of the material. Clearly visible, the melting peak temperature of the first heating run is 9–12°C (for the powders) and 8–9°C (for the plates) higher than the one of the second heating run. This results from the different thermal history of the material in the first (powder/printed part) and second (powder/printed part which was melted and slowly crystallised in the cooling run) heating run. For the powder, the melting peak temperature in the first heating run is rising steadily with increasing build cycle number. This is a consequence of the long-term high-temperature exposure of the powder in the SLS process. During the SLS printing job, also the not sintered powder in the powder bed is exposed to pretty elevated temperatures of around 175°C. This temperature is within the so-called sinter window (Schmid, 2018; Gibson et al., 2015) and above the crystallisation temperature of the material. It is a key requirement for a successful SLS print. In the second heating run, all powders have the same “cooling history”. They show no significant differences in the melting peak temperatures indicating that their melting behaviour was not irreversibly modified during the build cycles and that the increasing melting peak temperature in the first heating run is an annealing effect only. For the printed parts, the melting peak temperature in the first heating run is rather constant with only a slight decrease with increasing build cycle number. This small change is correlated with the melting peak temperature change in the powder (annealing). Because of the constant processing parameters, this change affects the melting of the powder during the printing job and, thus, the consolidation and final (thermal) properties of the printed part. Interestingly, for the first build cycles, the melting peak temperatures of powder and part are very similar, indicating a comparable thermal history for both. The difference in the melting peak temperatures of powder and part in the second heating run is not fully understood, as both have the same thermal history (cooling with 10 K/min in the DSC) and neither the corresponding crystallinity values [Figure 9(b)] nor the onset temperatures [Figure 9(c)] confirm this difference. As was observed for the powder, and also for the printed parts, no significant change in the (thermal) material properties is visible in the melting peak temperatures of the second heating run.

The crystallinity results of the powder [Figure 9(b)] show intense scatter, especially in the first heating run. This is because the powder form of the sample unavoidably violates a basic requirement in DSC to obtain well-repeatable results: it is not a single piece of bulk with a smooth contact surface for maximum thermal conductivity in the crucible. Nevertheless, from the diagram, one could speculate that the crystallinity in the first heating run decreases over the build cycles. A decreasing crystallinity and a simultaneously increasing melting peak temperature are, however, a contradiction. Considering this and the intense scatter, a further interpretation of these results is not useful. For the second heating run, the crystallinity values of the powders also show distinct scatter, but considerably less than for the first heating run. They remain roughly constant for all build cycles and thus suggest no change in the (thermal) material properties with increasing build cycle numbers. For the printed parts, the crystallinity in the first heating run is smaller than that of the powder and it shows a very similar trend as observed for the melting peak temperature. In the second heating run, the printed parts show crystallinity values comparable to those of the powders with no statistically significant change over the conducted build cycles.

For the SLS process, also the onset temperatures in DSC (melting and crystallisation) are relevant because they define the sinter window and thus the temperature range in which the powder bed has to be kept during printing. These onset temperatures are illustrated in Figure 9(c). Interestingly, the onset melting temperatures in Figure 9(c) reveal curve trends very similar to those of the melting peak temperatures in Figure 9(a). In the first heating run, there is an increase for the powders and a slight decrease for the printed parts with an increasing build cycle number, which are only partially significant. As for the melting peak temperature, also for the onset melting temperature, powder and part show very similar values for the initial build cycles. In the second heating run, the onset melting temperatures of powders and parts are almost identical with no significant change over the build cycles.

The melting and crystallisation onset temperatures in Figure 9(c) were used to calculate the sinter window values shown in Figure 9(d). Only the powder results are included in the diagram as the sinter window has no relevance for the printed parts. In the first heating run, the sinter window increases over the build cycles by around 10%. This increase is advantageous as it even facilitates the processing of the powder at higher build cycle numbers. However, it has to be kept in mind that the opposing effect of the increasing melting peak temperature is more important for the processing. Although it has no deeper meaning for the process (powder experiences only “1^st heating” in the SLS process), also the sinter window for the second heating run is included in Figure 9(d) for illustration purposes. The goal is to show the sensitivity of the sinter window on the thermal history of the polymer (powder). The sinter window in the second heating run is much smaller than in the first heating. In detail, the different cooling (crystallisation) history before the second heating run reduced the sinter window by more than 25% for the same polymer. Because of its importance for the process, the sinter window is a key parameter in the powder development. Although it is known to be variable to some extent, its tuneability in such a wide range only by varying the cooling conditions, at least for the PA12 resin examined, is surprising.

The most relevant results from the tensile tests are shown in Figure 10. As clearly visible in Figure 10(a), the Young’s modulus is constant at a value of around 1925 MPa up to the fourth build cycle, after which a decrease can be noted. With the seventh build cycle, the Young’s modulus shows a significant drop to 1550 MPa with no markable change afterwards. A comparable, but less significant, trend is observed for the tensile strength [Figure 10(b)]. Contrary, the strain at break [Figure 10(b)] follows the increase of the molar mass (documented in Table 2) and increases steadily with each build cycle. Generally, the observed effects of increasing molar mass (Table 2), decreasing part crystallinity [Figure 9(b)] and increasing part porosity [Figure 3(b)] are counteracting and competing with respect to the final mechanical behaviour of the printed parts. Although molar mass and decreasing part crystallinity increase the ductility, the higher part porosity reduces ductility and promotes brittleness. Obviously, in our case, the gain in ductility due to the rising molar mass and decreasing part crystallinity outweighs the embrittlement due to the porosity as was found in ageing studies of PA6 powder (Stiller et al., 2022) and also PA12 powder (Gomes et al., 2022; Yao et al., 2020).

The dimension of the tested tensile samples is given in Figure 11. Both width and thickness show a statistically significant continuous decrease with increasing build cycle number. Considering the constant laser sinter parameters used for all build cycles, this is correlated with the increasing melt peak temperature of the powder, which is illustrated in Figure 9(a) and the increase in the zero viscosity in Figure 6 and Table 2. This increase makes the gap between powder bed temperature and powder melt temperature bigger. Therefore, more laser energy would be required to compensate for this gap and to melt the same amount of powder at the later build cycles. As this compensation was not conducted, less powder was molten and, thus, smaller part (specimen) volumes were obtained.

Within this study, the ageing effects of PA12 powder during the SLS process were investigated. Unlike the conventional procedure, the PA12 powder was reused and sintered in total nine times without any refreshing. For the sake of comparability of the subsequent build cycles, the sintering parameters were kept constant throughout the whole study. For each iteration step, both, the powder and the printed parts were investigated using the same methods (bulk density, SEM, IR spectroscopy, plate–plate rheometry, MVR, optical analysis, DSC and tensile tests).

Overall, with an increasing build cycle number, the bulk density dropped statistically significantly from 0.475 g/cm³ down to 0.43 g/cm³, which was traced back to the formation of agglomerates, the MVR decreased from 37 g/10 min to 7 g/10 min because of increasing viscosity (from around 1,340 Pas up to 41,520 Pas) and molar mass (from approx. 24,550 g/mol to 64,510 g/mol), the DSC melting peak and onset temperatures rose slightly (about 3°C respectively), the tensile behaviour became more ductile (strain at break rose from around 4% to almost 12%), whereas the printed parts (tensile specimens) shrunk nearly by 10 mm² from first to last build cycle, got more pores and worse surface quality. In contrast to these statistically significant effects, the IR spectroscopy showed no change in the chemical structure (in addition to post-condensation detected by the rheological analyses) during the build cycles. The particle size distribution was not strongly influenced either. Generally, most results showed significant differences between each or more build cycles, whereas the changes in crystallinity were not substantial.

To conclude, for the PA12 powder ageing, all the detected effects can be related to three main changes in the powder: decreasing bulk density (through agglomeration), increasing melt viscosity (through polyamide post-condensation), and changes in the thermal behaviour, i.e. melting peak and onset temperatures (through thermal annealing of the powder). These three effects limit the reusability of the powder in practice and have to be overcome or partly compensated in the future for significantly improved recyclability. As a continuation of the presented results, process simulations are currently in progress to predict the ageing behaviour and its extent for PA12 based on real material data and processing parameters.

In the future, two paths can be taken for the aged powder: a treatment to achieve a closed-loop reuse of SLS powder or the processing in PIM (powder injection moulding). The first method has already been implemented by companies to reduce the refreshing rate to less than 15% and avoid the disposal of aged powder. Especially, the three main ageing effects (agglomeration, post-condensation, thermal annealing) must be controlled to avoid severe ageing of the powder. The second method was tried on a lab scale by producing granulate from the powder and processing it with conventional injection moulding or PIM. The first results are promising regarding ductility and optical impression. Nonetheless, both methods are up to now economically disadvantageous though better than simple disposal.

The research work of this paper was performed at the Polymer Competence Center Leoben GmbH (PCCL, Austria) within the framework of the funding program “Production of the Future” of the Federal Ministry for Climate Action, Environment, Energy, Mobility, Innovation and Technology and the Federal Ministry for Digital and Economic Affairs in the projects “3D-Print&Coat”, “fun-3D-manu”, “ThermoMelt4PA”, and “sustainableSLS” with contributions by Joanneum Research Forschungsgesellschaft mbH, Materials Science and Testing of Plastics/Montanuniversitaet Leoben, RPD Rapid Product Development GmbH and LSS Laser Sinterservice GmbH. The PCCL is funded by the Austrian Government and the State Governments of Styria, Lower Austria and Upper Austria.