Abstract
The purpose of this research work is to characterize and inform the design of (mechanical) property-graded bulk structures made from a single metallic alloy via a laser powder bed fusion (LPBF) process, with an end goal of creating repeatable/reproducible functionally-graded additively manufactured (FGAM) parts. This paper specifically investigates the manufacture of stainless steel (SS) 316L structures via a pulsed selective laser melting (SLM) process, and the underlying causes of property variations (within a functionally-acceptable range) through various material characterization techniques. For this, a design of experiments spanning the volumetric energy density (VED) based process parameter design space was utilized to investigate the range of functionally-acceptable physical/mechanical properties achievable in SS 316L. Five sample conditions (made via different process parameter combinations) were down-selected for in-depth microstructure analysis and mechanical/physical property characterization; these were suitably selected to impart a wide and controllable property range (209–318 HV hardness, 90–99.9% relative density, and 154–211 GPa modulus). It was observed that property variations resulted from combinations of porosity types/amounts, martensitic phase fractions, and grain sizes. Based on these findings, property-graded standard test specimens were designed and manufactured for further investigation—tensile specimens having a monotonic hardness change along its gauge length, four-point bending specimens with varying elastic moduli as a function of the distance from the neutral axis, and Moore’s rotating beam fatigue specimens with moduli variations based on the distance from the center. Altogether, this work lays the foundation for understanding and designing the local and global mechanical performance of FGAM bulk structures.
1 Introduction
With controllable layer-wise fabrication, metal additive manufacturing (AM) presents unrivaled design freedom to manipulate site-specific properties when manufacturing 3D parts directly from computer-aided design files. Selective laser melting (SLM) is one such metal AM technique that can impart intended property variations within the printed part via layer-wise process tailoring [1]. SLM-based property-gradation efforts have been implemented via approaches such as by assimilating various alloys and altering process parameters (for instance, laser beam profiling, scan strategies, etc.) [2–6]. To this end, functionally-graded additive manufacturing (FGAM) [7], is a technique to build a part layer-by-layer by intentionally varying the geometry/material organization within the part to meet specific local/global functional requirements. Currently, only a handful of efforts exist on FGAM fabrication with a single metallic alloy via SLM [8,9].
Over the years, researchers have expended significant effort in evaluating the printability of SS 316L via SLM for obtaining near full density parts [10,11] and to better understand its processing-structure-property-performance (PSPP) framework [12,13]. When considering the microstructure of SS 316L, it exhibits predominantly austenite grains, whose morphology is highly dependent on cooling rates [14] and thermal cycling [15,16]. The general theme of the majority of the SLM investigations on SS 316L has been to alter material/process parameters to maximize/minimize a certain physical and/or mechanical property—this has often led to looking at the process parameter space with a view to maximize density, minimize certain pore types/amounts, etc. [17] in order to obtain near full density and/or high-hardness parts. In contrast, our efforts build on the prior work of such researchers so as to understand how to intentionally vary properties within a certain functional range (including “achieving” certain porosity, lower hardness, etc.) to be able to eventually engineer FGAM part designs—this necessitates a dedicated process mapping, one that has a different end goal when compared to other seemingly-related efforts.
Some of the disparate efforts where a material/process condition could be used to intentionally alter a resulting property are outlined next—one could alter average cooling rates experienced within melt pools by tweaking energy-related process parameters, and in turn affect melt pool characteristics, solidification microstructures, and residual stresses [18,19]. Considering beam profiling, changing to an elliptical laser beam profile reduced the average grain area by almost half, reducing texture and increasing the Taylor factor that, in turn, enhances the mechanical properties of the parts. Considering hardness, variations in the Vickers microhardness values for SLM processed SS 316L have been widely reported in the literature [14,20].
Altogether, there are numerous approaches by which certain mechanical properties could be intentionally graded within a single alloy SLM build (same composition, in contrast to compositional grading which is doable via directed energy deposition). However, the potential for property grading of a single alloy FGAM structure, and specifically for 316L via SLM has yet to be systematically explored—this is the theme of this study.
1.1 Functional Ranges of Properties Achievable Via Selective Laser Melting.
To utilize FGAM structures for industrial purposes, a functionally-useful property range must be defined first for various physical and mechanical properties such as density, hardness, modulus, etc. Engineering applications that demand unique property combinations are suitable avenues where these functional ranges can have inherent advantages. For instance, a relative density between 90% and 100% could still be a structurally valid range for non-critical, multipurpose parts that require both strength and compliance. Being more mature and controllable, SLM is a suitable processing technique to further examine the usable property ranges of alloys to impart certain unique property combinations.
Considering related work, the early work of Mumtaz and Hopkinson [21] used SLM to fabricate bi-material specimens with Zirconia and Waspaloy using a high-power Nd:YAG laser. The experimental work by Niendorf et al. [8] demonstrated the use of SLM to produce a step-change SS 316L structure by using two different laser powers (400 W and 1000 W) with differing powder layer thicknesses (50 µm and 150 µm) while keeping the volumetric energy density (VED) constant. They found columnar coarse grains for regions processed with 1000 W and fine grains for 400 W—this showcased a local difference in properties (hardness, strength, and ductility). Popovich et al. [9] examined a step property change of Inconel 718 by using two different laser intensity profiles (Gaussian and flat-top) against laser powers (250 W and 950 W) and found a sharp transition from fine to coarse grains with corresponding changes in hardness (∼20% reduced hardness for coarse-grained regions). Hengsbach et al. [22] fabricated a material transition between tool steel H13 and SS 316L by utilizing a reconfigured recoater design for SLM to accommodate two different materials. More recently, Attard et al. [23] exhibited microstructural control for IN718 via altering the thermal history, by tweaking process parameters (scan strategies, hatch spacing, laser power). They obtained two different microstructures (highly columnar and quasi-equiaxed) and concluded that the preheat level from the previous layer was the critical factor in forming the ultimate microstructure.
1.2 Motivation and Challenges.
This work is an attempt to enable bioinspired material system designs based on natural materials (such as mammalian teeth) which have continuous gradients in spatial mechanical properties tailored for specific functionality. Based on the above discussion, SLM is a process candidate for limited property grading by intentional alteration of porosity, grain morphology, etc. Single alloy gradation has applicability where it would be a catastrophe to have a mismatch between different materials due to differing thermal expansions and/or strains. Moreover, single alloy grading can limit restrictions and vulnerabilities of extraneous welds/joints and minimize performance-related uncertainties as in multi-material FGM.
This research work presents a systematic study to demonstrate the SLM fabrication of property-graded 316L bulk structures. Here, we provide a better understanding of the control of processing and structure to impart mechanical properties in a repeatable and reproducible manner by varying energy density-related process parameters. In-depth material characterization was conducted, and comparisons were made concerning the grain size, phase fractions, and porosity types/amounts, in addition to physical/mechanical property testing from the perspective of developing gradient bulk structures. The present study contributes to the knowledge base by linking processing and structure to properties in the context of functionally-graded samples, besides tailoring local/global performance.
2 Materials and Methods
2.1 Stainless Steel 316L Powder.
Commercially available N2 gas-atomized AISI 316L stainless steel powder sourced from Renishaw PLC. was used as the raw material in this study, with the as-supplied powder size distributions of D(10) = 19.17 µm, D(50) = 26.25 µm, and D(90) = 37.96 µm. Microscopy of the virgin powder showed a high fraction of nearly spherical particles that is appropriate for better packing density of the powder bed and for good powder flowability. Further, average particle count distributions for the as-received powder showed an average particle size of 18–20 µm at D10 and 38–40 µm at D90; this confirms that the majority of the power particles are within the recommended size range needed for powder-bed SLM.
2.2 Selective Laser Melting Machine and Process Specifications.
A Renishaw AM 400 platform was employed to fabricate the samples used in this study. The laser had a wavelength of 1070 nm, and a spot diameter (based on D4σ) of approximately 70 µm with a TEM00 Gaussian beam profile. The substrate material was also SS 316L that was commercially cold-rolled and annealed. The build plate was not preheated since the samples and substrate were of the same material (SS 316L), whereby sample adhesion and part warping were not major concerns (each sample had a small XY footprint of 5 mm by 5 mm); further, the bottom layers of each sample (which might have experienced a significant thermal gradient along the Z-direction) were not relevant/used for this study (was sawed-off). A powder layer thickness of 50 µm and a hatch spacing of 110 µm were set along with other constant parameters. Laser beam scanning was done with a meander scan pattern (Fig. 1) which is preferred for small and homogeneous cross sections. A clockwise scanning angle of 67 deg was implemented between adjacent layers (Nth and N + 1st) to reduce the anisotropy in the build direction (XY plane). The laser power was varied in increments of 25 W between 150 W and 250 W. The scan speed was varied from 500 mm/s to 800 mm/s during the build by changing the exposure time from 60 µs to 100 µs (constant point distance of 50 µm) resulting in a VED range of 34–91 J/mm3 (VED (Ev) is a composite process parameter that utilizes laser power, scan speed, layer thickness, and hatch spacing to calculate the total energy input per unit volume. Being a combinatorial quantifier, it helps to understand melt pool behavior which in turn affects the solidification microstructure and hence properties). Post building, air blasting was used to remove the loose powder particles, and samples were carefully picked from the base plate. No post-processing was performed on the samples.
2.3 Testing and Characterization.
Bulk and optical densities were calculated for each sample that was 7 mm high, and had a square base of 5 mm by 5 mm. For bulk density (immersion-based Archimedes method), the weight of each sample was measured in air and distilled water. The distilled water was held at a constant temperature of 23.1 °C, with a density of 0.9975 g/cm3 considering temperature/fluid-density dependencies. The density was calculated by dividing the mass of the sample in water by the volume of water displaced. The relative density is expressed as a percentage of the nominal density of SS 316L (7.99 g/cm3). Two repetitions were made for each sample, with an apparatus measuring accuracy of ± 0.1 mg. For measuring optical densities, a LEICA DM 2500 M (Wetzlar, Germany) light optical microscope was used to capture micrographs of both XY plane (transverse) and XZ plane (longitudinal) views of samples (polished internal sections). Subsequently, ImageJ software was used to determine the (2D plane) porosity based on the average pore area fractions on the micrographs and repeated three times for each sample.
Vickers microhardness (HV) measurements were taken on polished sample surfaces using a Beuhler Wilson VH1102 microhardness tester. Following ASTM E92, a square-based pyramidal diamond indenter was used to probe samples while maintaining the distance between successive indents at ∼1 mm, and indents were kept at least 1.5 mm away from the sample edges. The applied load was 0.5 kgf (denoted as HV0.5) with a dwell time of 10 s. The reported microhardness was the average of three indents from different parts of the sample surface. In each case, care was taken to conduct the indent away from any major pores/defects, as viewable via the magnified live view of the microhardness tester. Nanoindentation tests were carried out at ambient air/temperature using the Hysitron TI 950 Triboindenter. A Berkovich type diamond indenter with a tip radius of 150 nm was used to probe a total of eight indents on each sample with loading/unloading rates of 0.05 mm/s and dwell time of 12 s. The maximum penetration depth was approximately 1.5 µm, and the applied load was around 250 mN. The collected data were adjusted for elastic recovery, considering the unloading residual depth with less than 10% of the maximum load applied. Finally, as-printed samples were sectioned and then hot mounted in Bakelite. These were then ground, polished, and electrochemically etched using 10 wt% oxalic acid and 90 wt% de-ionized water under 7 V DC for 20 s. Post etching, the samples were rinsed with water and acetone, and then dried using compressed air to prepare them for characterization using optical and electron microscopy.
3 Results and Discussion
3.1 Light Microscopy.
Morphological analyses were conducted on 28 as-built SS 316L samples to investigate the processing window to obtain repeatable and reproducible physical/mechanical properties, as detailed in Refs. [1,24]. Of primary interest was to check whether “sufficient” melting occurred for each sample. For this, both XY and XZ planes of each as-built sample were examined. Figure 2 shows three representative sample surfaces in the transverse (XY) plane with an area of ∼30 mm2 on which it is easier to compare burn marks, unfilled hatch spaces, scan tracks, etc. Further comparison of XY and XZ planes for select samples is provided in Fig. 4. A color-coded scale bar depicts the VED range. This experimental design was centered on recommendations from the OEM (Renishaw), which suggested a VED of Ev = 48.48 J/mm3 (P = 200 W, V = 750 mm/s). The samples processed within the green VED scale were expected to result in an overall well-melted surface with fewer defects. Based on the further analyses of each of these, five specific parameter combinations were chosen, such that they yielded a wide but usable range of properties (hardness, density/porosity, Young’s modulus), which were then employed to design and additively manufacture functionally-graded structures for tailoring mechanical performance.
3.2 Physical and Mechanical Properties
3.2.1 Bulk Density and Porosity Measurements.
Figure 3 represents the (Archimedes) densities of all 28 samples, with curve fits for two representative power conditions against VED; note that the overall range of sample densities observed was within 90–99.9%. The results agreed with the literature (low VED typically resulting in undermelted material and lack of fusion (LOF) defects, while too high of a VED (though reducing LOF pores and balling) causing material evaporation and (partial) re-melting, resulting in keyhole defects); this emphasizes the role of processing parameters in the formation of defects and resulting physical/mechanical properties. It is worth noting that at a constant power of 150 W, a monotonic change in density (90–99%) was achievable by changing the VED, whereas, at a higher power of 225 W, the observed densities were mostly 99%+. For the former case (150 W), the density trend could hence be reasonably captured via a second order fit for the 90–99% range using an equation (see inset in Fig. 3), where relative density is in percentage, and VED is in J/mm3. It should also be noted that though a single numerical value is obtained for each measurement, some density variation is expected within (attributed to differing phase fractions and pore extents/forms)—in other words, this is to be taken as a composite value which is still a reasonably good measure of the overall bulk density and can be linked to the processing condition.
For further probing, optical images of polished surfaces of five select samples were obtained in both XY (transverse) and XZ (longitudinal) planes, as shown in Fig. 4 (each captured at the same magnification for comparison).
Among these, sample 4 and sample 2 had the lowest and the highest energy densities, respectively. For the lower energy input sample 4 (Ev = 34.09 J/mm3, P = 150 W, V = 800 mm/s), there was insufficient melting of powder particles as well as inadequate melt pool overlaps, leading to significant visible holes, voids, and LOF defects. This sample showed the highest porosity of 9.22%. This is due to a low power value and high scan speed that may have produced discontinuities in the melt pool tracks, resulting in low density and high porosity. In contrast, for the higher energy input processed sample 2 (Ev = 63.02 J/mm3, P = 200 W, V = 577 mm/s), porosity decreased to 2.77%, which is due to sufficient melting and fusion, affecting the melt pool shape, grain size/texture, and residual thermal stresses. Note that the lowest porosity was 1.79% for sample 3, at the “recommended” VED.
Figure 5 illustrates the results of an image analysis approach (ImageJ) [25] to evaluate the densities, where polished internal surfaces were analyzed. As before, we observed a density range of 90–99.9% when measuring optically across both XY and XZ planes. Similar trends were observed for the two methods (optically on internal planes versus bulk immersion in liquid), and the minor differences were attributed to the selected smaller measurement areas in the case of 2D image-based density compared to the 3D bulk volume/density in the case of the immersion-based method. It was also observed that optically measured densities had higher values in most cases than the gravimetric measurements. Further, Fig. 6 illustrates the results of all investigated samples’ average mean pore radius and mean circularity against the chosen VED range. Circularity was computed as pore Perimeter2/4π * Area using ImageJ’s built-in tool by analyzing each of the 2D images in grayscale. Circularity ranges from 0 to 1, where lower values represent high aspect ratio polygons and values closer to one represent a perfect circle. In general, the results showed more angular/irregular pores for samples processed at low VED and more circular/rounded pores for high VED samples.
It is known that going above or below a certain VED threshold can impact final build quality [26]. In concurrence, at low VED values, samples exhibited lower densities (high porosities), indicating the presence of LOF type defects due to insufficient VED. The LOF pores (non-spherical) were evident up to the VED saturation point of 50 J/mm3, below which there was a significant drop in circularity (below ∼0.8). When considering mean pore radius, it was significantly increased, up to ∼14 µm. Furthermore, samples exhibited relatively small-circular pores for a VED range between 50 J/mm3 and 70 J/mm3. However, at too high of a VED (70 J/mm3 and above), which could likely cause material evaporation, samples showed somewhat larger pores (non-spherical) keyhole pores. From these observations, it can be confirmed that VED could be considered as a composite factor that can be used to control the defect characteristics in the printed part. However, for a more robust understanding, each individual parameter and their interactions with others should be considered as recommended in a similar study [27].
3.2.2 Hardness Measurements.
Data for the 28 samples’ hardness measurements have been detailed in the authors’ previous study [24]. Figure 7 plots the average hardness against VED. Typically, hardness is affected by porosity and microstructural aspects of the material.
Samples processed with lower powers (150–200 W) showed relatively higher hardness variations than those processed at higher powers (225–250 W). The highest average microhardness measured across the samples was 318 HV0.5 (P = 200 W, V = 698 mm/s, Ev = 52.10 J/mm3), which was closest to the one reported by Saeidi et al. [28] of 320 HV. They reasoned the high hardness on two main factors, the first being the existence of cellular subgrain microstructure (with a size range of 0.5–1 µm) that is believed to provide extra strength, and the second being oxide inclusions that block the dislocation movement which in turn results in higher strength. The lowest microhardness measured was 209 HV0.5 (P = 175 W, V = 800 mm/s, Ev = 39.77 J/mm3), which was also in line with the values reported by Tucho et al. [29] and Sun et al. [30]. The obtained hardness range for samples (209–318 HV0.5) was significantly higher when compared to conventionally cast SS 316L (150–160 HV), which is attributed to more refined (smaller) grains in SLM due to higher thermal gradients [31].
Generally, higher VED results in re-melting the already solidified layer and promotes grain refinement. Finer (smaller) grains enhance hardness, whereas the coarser (larger) grains yield lower strength to withstand deformation during indentation (Hall–Petch relationship) [32]. The phases (other than austenite) can also impact the observed hardness differences. As with density, a second-order equation could identify trendlines for the hardness. However, it is important to note that hardness is dependent on a combination of factors such as pore fractions, phase fractions, grain sizes, etc. The findings will form a valuable data catalog for informed selection of processing parameters to yield repeatable and reproducible SS 316L samples within a functionally-acceptable range within the processing bounds. In this light, the Vickers microhardness trend for the 175 W condition could be reasonably captured via a second-order equation by correlating it to VED (see inset in Fig. 7). Figure 8 illustrates the significant hardness and density variations achievable across all samples by imparting different volumetric energy densities; also shown are “iso-VED” lines along with dominant defect regions observed. The saturation point for VED was found to be ∼44 J/mm3, below which there was a significant drop in density (below ∼97%). In general, decreasing the laser exposure time to increase VED leads to more complete melting of powder particles and obtaining near full density samples. As highlighted by the two arrows in Fig. 8, by changing a single parameter (scan speed) at a given power level, one could monotonically change the hardness (at 175 W) or density (150 W)—these could serve as preliminary control parameters for grading properties in FGAM parts.
3.2.3 Nanoindentation Hardness and Modulus.
Table 1 outlines the results of the nanoindentation tests conducted on the chosen five SS 316L samples. The lowest and the highest nano hardness values were 3.83 ± 0.17 GPa and 4.81 ± 0.52 GPa for samples 1 and 2, respectively. In the case of modulus, the lowest and highest values were 187.4 ± 5.79 GPa and 211.1 ± 9.62 GPa for sample 4 and sample 3, respectively. Notice that sample 4 also has the lowest density, 90.78 ± 0.09%. The observed mismatch between the modulus and hardness trends at the nanoscale could be due to the tested grains having different crystallographic orientations [33], besides each property being a result of a number of differing factors. When considering the hardness trends at the micro versus nano scale, it is to be noted that a Vickers indentation typically spans multiple grains thereby providing an averaged/composite hardness quantification, while a nano indentation often probes a grain or grain boundary thereby providing a more localized material response. Moreover, the location-dependent anisotropy, inhomogeneous melting of powder particles, and indentation near defects also contribute to the variations in the obtained results.
Also indicated are the processing conditions versus the resulting density and microhardness for the five selected samples. These five were specifically chosen from among the 28 samples primarily based on the microhardness numbers—besides maximizing the total hardness range (209–318 HV) by choosing the extremities, the middle three were chosen so as to yield a more or less equally-spaced hardness interval of ∼20 HV, i.e., 256, 271, 294 HV. As highlighted by the maroon envelope, a monotonic change in hardness could thus be engineered by utilizing these specific process parameters sets; note that except for sample 1, the rest have a low standard deviation (<5%). In summary, the density, hardness, and nanoindentation measurements yielded a range of properties that could be used to tune property-graded SS 316L samples. For this, one would need to select a process parameter combination that intentionally yields a different (but repeatable) property value within a functionally-acceptable range.
3.3 Microstructural Characterization.
The microstructure of the selected five samples of SS 316L was further investigated at different scales. Representative SEM images for sample 4 (Ev = 34.09 J/mm3, ρ = 90.78%) and sample 3 are presented in Fig. 9. Several features such as laser scan tracks, hatch lines/spacing, layer thickness, melt pools/boundaries, and grain boundaries are visible. The difference in build quality is evident between samples 3 and 4 due to a significant change in the VED. Sample 3 shows more complete fusion of powder particles, whereas sample 4 shows several incompletely fused and unmelted regions. Nonetheless, half-cylindrical contours of melt pool boundaries are visible in both samples.
Several features such as laser scan tracks, hatch lines/spacing, layer thickness, melt pools/boundaries, and grain boundaries are visible. The difference in build quality is evident between samples 3 and 4 due to a significant change in the VED. Sample 3 shows more complete fusion of powder particles, whereas sample 4 shows several incompletely fused and unmelted regions. Nonetheless, half-cylindrical contours of melt pool boundaries are visible in both samples. Common grain structures observed were equiaxed, columnar, and mixed. Microstructures exhibited long elongated grains parallel to the BD and consistent with earlier studies on SLM fabrication of 316L [30,34–37]. Interestingly, more pore-type defects were observed for sample 1 and sample 4 (processed with low energy), including LOF defects and keyhole porosity. The keyhole pores were relatively large and are predominantly observed in samples processed at higher energy inputs, such as in sample 2 and sample 5 [38]. A cell growth mechanism, known as side-branching, was seen inside the melt pool [39]. This occurs especially at melt pool boundaries due to a significant local thermal gradient and is responsible for an alteration in the direction of growth for coarse elongated grains, which typically favors growing perpendicular to the half-cylindrical contours of the melt pool boundaries (fusion lines). Equiaxed grains are visible at the center of the melt pool and are believed to have been formed at the end of the solidification event. Moreover, epitaxial grain growth was also seen across the melt pool boundaries. The average planar grain sizes (diameter) were measured following the intercept method per ASTM standards [40]. Typically, equiaxed grains have grain size (diameter) in the range of 2–10 µm, whereas columnar grain size (diameter) can be greater than 10 µm [41]. We also noticed grains smaller than 1 µm which was expected due to the high cooling rates. The grain sizes presented should be considered with caution.
3.3.1 Identification of Dominant Phases.
SLM-processed SS 316L forms different phases, including face-centered cubic austenite, body-centered cubic ferrite, and body-centered tetragonal martensite [42]. As an example, Fig. 10 illustrates the phases identified for the longitudinal (XZ) plane of sample 5. Phase identification involved examining a significant number of etched optical/electron microscopy images at different magnifications in conjunction with XRD (diffraction) and EDS (spectroscopy) measurements on each specimen for both transverse/longitudinal planes (not included, for paper brevity). Typically, (non-magnetic) austenite exhibits angular/straight grain boundaries with twins and often shows sharp polygonal vertices. In contrast, (magnetic) ferrite is fairly equiaxed with rounded grain boundaries and with no twinning present. Martensite exhibits a needle-like (acicular) structure and is indicative of rapid cooling. There are also other artifacts often observed in SS 316L made via SLM such as fine cellular structures and carbide precipitates along boundaries. Ferrite phase formation in austenitic stainless steel is typically due to diffusion or segregation [43]. However, the volume fraction of ferrite depends on the cooling rate; a higher cooling rate obstructs diffusion. Both austenite and ferrite phase fractions affect the ductility and hardness of SS. On the other hand, martensite is formed due to diffusion-less transformation. As the formation of the martensitic phase from austenite involves interatomic movements, different martensite crystal structures are possible [44] and in two morphological forms (laths and plate), the formation of which depends on the carbon content. Lath martensite is typically seen in low C (<0.6 wt%) stainless steel, such as 316L, whereas plate martensite is seen in high carbon steels [45].
The processing- and microstructure-based dominant dependencies of property variations were mapped across all 28 samples to identify cause-effect pairs for property changes and their implications as shown in Fig. 11.
The underlying reasoning for differences observed in phase dominance can be explained via cooling rates. For instance, scan speed bounds of ∼550 mm/s or less yields ferrite dominance, whereas VED between ∼44 J/mm3 and 64 J/mm3 leads to the formation of highly martensitic grains. Lower VED results in shallower melt pools, yielding more randomly oriented, equiaxed/fine ferrite grains. These lower VED bounds can be achieved at higher scan speeds by keeping low laser power or vice versa. Similarly, melt pools are relatively deep at a higher VED and form coarser/elongated austenitic grains. The dominance of martensite within VED bounds has resulted in a higher hardness zone within the processing bounds. Ferrite (equiaxed grains) seemed to be dominating when the scan speed bounds were too low (∼500 mm/s) or too high (∼700 mm/s). It could be due to more spherical grains formed due to shallower melt pool features. For austenite (angular grains), the presence was outside the VED threshold of 44–64 J/mm3, correlating well with relatively coarser grains observed across the multiple samples’ optical images. They could be formed due to the lower cooling rates, allowing grains to grow. Finally, martensite (needle-like structures) was observed within the VED range of 40–60 J/mm3; martensite formation is affected by rapid cooling rates and is seen to increase the hardness of such samples.
Table 2 details the five selected samples which are grouped based on the constituent phases identified. The differences in the phases directly affect mechanical properties, especially hardness. This is evident for samples 3–5 which show traces of martensite—as expected these samples have higher hardness than samples 1–2 which do not show any indications of martensite presence; it is to be also noted that a large number of SEM images were examined for each case.
# | XY (transverse) plane observations | XZ (longitudinal) plane observations | Major phases identified | Grain size (µm) (Std. Dev.) | Remarks |
---|---|---|---|---|---|
1a | Visible carbon grains at grain boundaries, ferrite (lighter), and austenite | Long columnar grains across melt-pool, dense microstructure | Ferrite, austenite | ∼2–4 (±0.5) | Ferrite dominates |
2a | Unmelted regions, ferrite and austenite, gas pores | Long columnar austenite grains, some ferrite | Ferrite, austenite | ∼3–5 (±0.5) | Austenite dominates |
3b | Ferrite, austenite, carbon at the grain boundaries | Long columnar austenite grains, martensite visible | Ferrite, austenite, some martensite | ∼2–4 (±0.5) | Austenite dominates with some martensite |
4b | Lots of ferrite grains (smaller ones), a few austenite grains, dendrites, some martensite | Columnar grains crossing melt-pool boundaries, high-angle, low-angle grain boundaries visible | Lots of ferrite, austenite, some martensite | ∼2–4 (±0.5) | Ferrite dominates with some austenite |
5b | High ferrite, some martensite, gas pores, melt-pool overlapped | Long columnar grains, denser microstructure, melt-pool overlap, keyhole defect, gas pore | Ferrite, austenite, some martensite | ∼3–5 (±0.5) | Austenite dominates with some martensite |
# | XY (transverse) plane observations | XZ (longitudinal) plane observations | Major phases identified | Grain size (µm) (Std. Dev.) | Remarks |
---|---|---|---|---|---|
1a | Visible carbon grains at grain boundaries, ferrite (lighter), and austenite | Long columnar grains across melt-pool, dense microstructure | Ferrite, austenite | ∼2–4 (±0.5) | Ferrite dominates |
2a | Unmelted regions, ferrite and austenite, gas pores | Long columnar austenite grains, some ferrite | Ferrite, austenite | ∼3–5 (±0.5) | Austenite dominates |
3b | Ferrite, austenite, carbon at the grain boundaries | Long columnar austenite grains, martensite visible | Ferrite, austenite, some martensite | ∼2–4 (±0.5) | Austenite dominates with some martensite |
4b | Lots of ferrite grains (smaller ones), a few austenite grains, dendrites, some martensite | Columnar grains crossing melt-pool boundaries, high-angle, low-angle grain boundaries visible | Lots of ferrite, austenite, some martensite | ∼2–4 (±0.5) | Ferrite dominates with some austenite |
5b | High ferrite, some martensite, gas pores, melt-pool overlapped | Long columnar grains, denser microstructure, melt-pool overlap, keyhole defect, gas pore | Ferrite, austenite, some martensite | ∼3–5 (±0.5) | Austenite dominates with some martensite |
Samples with mostly ferrite and austenite phases in the microstructure.
Samples with martensite, ferrite, and austenite phases in the microstructure.
4 Next Steps—Functionally-Graded Additively Manufactured Parts
Altogether, the above effort served to probe the broad process parameter design space, down-select combinations from within to impart a wide but controllable property range (especially hardness), and hence facilitate property-graded structures for tailorable performance. It should be mentioned that the original motivations for this work were inspired from the hardness/modulus distributions within mammalian teeth. Figures 12 and 13 illustrate three instances where this capability was leveraged: property-graded tensile, bending, and fatigue SS 316L samples fabricated using SLM.
As illustrated in the top row of the collage (Fig. 12), ASTM standard dog-bone samples (of three different sizes) were fabricated via SLM where the gage volume had five sub-zones, each made via a different process parameter combination, and hence exhibiting a different property value (viz., hardness, density); these were designed such that there was a monotonic hardness change in the build direction. When subjected to standard tensile testing until fracture, these yielded various strength/stiffness and deformation profiles, along with unique behavior such as location shifts of maximum strain (attributed to strain hardening of a zone and hence another zone becoming a new “weakest link”), zonal interfaces acting as a graded band, etc. Thus, select zones within the gage volume of tensile parts could be engineered to solely/combinatorially support increasing strains—this could be tailored to cause failure within a select region or to strain harden a specific zone (yielding certain controlled deformation), and so on.
In regard to the bending samples shown in the bottom row of the collage (Fig. 12), ASTM standard bending test samples were fabricated via SLM that had three sub-zones, with each made via a different process parameter set; these were designed such that the outer zones (top/bottom regions) had different properties when compared to the zone housing the neutral axis (middle region) indicated by E/HVout and E/HVin, which refers to the different modulus/hardness. Also shown is a snippet from a four-point bend test on one of these samples in which digital image correlation (DIC) mapping shows longitudinal strain distributions which are dictated by different effective flexural moduli. When considering fatigue sample design (Fig. 13), a similar philosophy was adopted to create ASTM standard parts (for RR Moore rotating beam fatigue testing) whereby a different property was imparted to the outer versus inner zones via intentional process parameter selection; an outer zone with a different effective stiffness or hardness is expected to have a significant effect on fatigue life, since crack formation and propagation is typically initiated at the outer surface/sub-surface. Altogether, engineering the inner versus outer stiffness ratios could help adaptively stiffen the structure or allow for controlled deformation and/or help with light-weighting, minimize cross-sectional areas, and so on.
The local/global performance of such rudimentary FGAM structures will be compared with their “homogeneous-property” counterparts via various characterization techniques, including 2D-DIC and fractography, among others. These will be reported in detail in a subsequent publication. Further investigation is warranted to explore how a truly gradient property change could be spatially achieved rather than the step-changes (at the mm-scale) exhibited in this work, and how to engineer zonal interfaces for the intended benefit, among others. Therefore, the efforts presented in this paper essentially set the stage on how to obtain property-graded FGAM designs via informed process parameter selection while utilizing an SLM process with a single alloy.
5 Summary and Conclusions
The present study was focused on understanding the causalities within the PSPP framework for the SLM of SS 316L to realize FGAM bulk structures. The focus was to elucidate the relationships between AM process conditions and variations in macro- and microstructures and how such physicochemical traits affect the resulting physical/mechanical property distributions. Specifically, this study investigated the role of VED-based process parameters and their processing bounds for SLM-processed SS 316L in functionally usable property bracket and their influence on the resulting solidification macro/microstructures. The following conclusions are derived from this work:
Density can be reliably controlled (within 90–99.9%, defined as the “functional” range) by altering just one process variable, namely, laser exposure time (which directly affects laser scan speed). At a constant laser power of 150 W, by increasing the scan speed from 500 mm/s to 800 mm/s, the density can be monotonically decreased from 99.9% to 90%. This would be a preferred and more reliable approach rather than altering multiple parameters simultaneously.
– This aligns with the fact that as the exposure time (and scan speed) increases, it proportionally reduces the VED imparted. This was confirmed by the evidence of a more significant percent fraction of LOF pores that are primarily attributed to not having the needed energy to sufficiently melt the power.
Across the (28) samples probed, it was possible to reliably obtain/impart a hardness variation spanning 209–318 HV0.5, with the lowest and highest hardness obtained at P = 175 W, V = 800 mm/s and P = 200 W, V = 698 mm/s, respectively. Unlike density, imparting intended hardness differences (209–318 HV0.5) needed more than just a single parameter to be controlled. Nonetheless, a monotonic hardness decreases from 294 HV0.5 to 209 HV0.5 can be reliably obtained at a laser power at 175 W and increasing the scan speed from 500 mm/s to 800 mm/s.
– Unlike relative density (primarily due to pore size/fraction), which a single process parameter can monotonically control (viz., scan speed), hardness is dependent on a combination of factors—pore fraction, phase fraction, grain size, etc., justifying the need to use multiple parameters.
– This is also evidenced by the higher hardness observed in samples that contained a “detectable” volume fraction of martensite versus those with just austenite and ferrite.
Altering VED (Ev) enables intended variations in as-built part density/porosity. Higher porosities and defects were found in samples processed with Ev lower than ∼44 J/mm3, which resulted in a significant drop in densities (below 97%). Solidification defects such as LOF were substantial in these low-density samples. The higher density samples showed smaller micro-pores, suggesting that the variation in density is primarily rooted in pore morphologies.
– This was evidenced by the LOF pores (non-spherical) observed at low VED versus gas pores (spherical) observed at too high of a VED.
Microscopy revealed differences in microstructural morphology to identify varying fractions of austenite, ferrite, and martensite phases. The presence of martensite (with a needle-like (acicular) structure) was affected by the VED magnitudes. Samples processed with Ev within 44–54 J/mm3 have shown martensite phase fractions. Nevertheless, at higher VED (Ev ∼ 60 J/mm3), there were gas pores (entrapped gases) observed due to either pre-existing gas inside the powder (gas atomization processing or chamber oxygen) and/or vaporized material during fusion which could lead to changes in the local thermal field (cooling times) and therefore, alter the formation of solidification microstructure. Similarly, the absence of martensitic phase fractions at lower VED (Ev ∼ 40 J/mm3) could result from disruptions and thermal insulations due to the considerable lack of fusion (irregular pores) defects and voids.
It is deduced that the overall property variations result from a combination of porosity types/amounts, martensitic phase fractions, and grain sizes. Altogether, this work lays the foundation for understanding and designing the local and global mechanical performance of FGAM bulk structures.
Acknowledgment
The work described in this paper was partially supported by an NSF grant, as well as in part by appointment to the Oak Ridge National Laboratory Advanced Short-Term Research Opportunity (ASTRO) Program (MDF), sponsored by the U.S. Department of Energy and administered by the Oak Ridge Institute for Science and Education.
Conflict of Interest
There are no conflicts of interest.
Data Availability Statement
The datasets generated and supporting the findings of this article are obtainable from the corresponding author upon reasonable request.