Aspects of the Powder in Metal Additive Manufacturing: A Review

The most widely used metal additive manufacturing processes utilize powder that is spread or fed onto a building platform. Although there are reviews of the literature on some aspects of the powder, many aspects have been un-der-reviewed or unreviewed. The present work is a review of the literature on these aspects. Articles published in the open literature through the end of February 2022 were collected by consulting highly regarded relevant biblio-graphic databases, such as Google Scholar and Science Direct. The aspects reviewed were emerging methods of powder production, methods used to improve the quality of a powder after production by a well-established method, influence of variables of well-established powder production methods on powder properties, influence of powder production method on powder properties, and influence of powder reuse on properties of powders of a wide collection of alloys. One key finding was that with regard to powder reuse, the only consistent finding is that it leads to increase in the oxygen content of the powder. Another key finding was that the literature on the aspects of the literature reviewed herein contains many shortcomings and gaps, which suggest potential areas for future research, such as techniques for optimization of process variables for a given combination of metal powder and powder production method and development of methods for production of powders of new/emerging metallic materials.


Introduction
The list of metal additive manufacturing (MAM) processes/technologies is very In powder-based MAM methods, the importance of the powder feedstock is underscored by the fact that its cost (typically, unit cost is in the range of $40 -500/kg) is a large proportion of the total normalized manufacturing cost of a part (defined as the sum of raw material, printing, and post-processing costs per volume of built part); for example, for both LB-PBF and EB-PBF, powder cost is ~40% [9] [23]. An additional indication of the importance of the powder is that its properties exert a marked influence on both the build rate of the part (and, hence, the production cost of the part) and its properties, especially its densification mechanism (and, hence, its porosity and mechanical properties) [9] [26] [27] [28] [29] [30].
There is a sizeable body of literature on many aspects of the powder feedstock used in MAM, and there are a number of reviews of this body but, with the exception of two reviews, their focus has been on three aspects. These are descriptions of and comments on advantages and shortcomings of established powder production methods ( Figure 2) [11] [31]- [41]; powder characterization methods [27] [30] [42]; and the influence of powder properties on properties of built parts [26] [27] [29] [30] [41] [43] [44]. One of the exception reviews covered an overview of powder-bed MAM methods, powder flowability, raking behavior/powder layer formation, and brief discussion of the influence of powder properties and powder reuse on properties of both the powder and the built parts [29]. In the other exception review, in addition to covering established powder production methods, the issue of powder reusability/recycling vis-à-vis properties of powders of Ti, Ni, and Al alloys was addressed, and some of the coverage was on the influence of powder reuse on properties of built parts [41]. when balls with D b = 6 mm were used, the morphology of the particles was changed, to spherical or near-spherical particles. These findings provided the rationale for using a dual-stage ball milling method, which yielded particles that were spherical or near-spherical (particle diameter (D) = 38 -150 μm).
At each of these stages, the morphology of the product had distinctive features. After reduction, the morphology did not change significantly and ~95% of O gas was removed; the granulated powder was spherical; and the individual powder particles sintered very well, with only a very small amount of adhesion taking place. The final powder was 100% dense and its O , C, and N contents were each lower than the levels given in the relevant materials standard specification (ASTM B299 [50]).

Continuous Process, TiRO TM
There are two steps in the continuous process (TiRO TM ), as was used for the production of commercially-pure Ti powder [51] [52]. The first step was conducted in a fluidized bed, where TiCl 4 gas and Mg powder reacted to form small-sized Ti particles (D = ~1.5 μm). The reaction mechanism involved deposition of reaction products around the Mg particles. At the end of this step, some of the Ti particles contained small amounts of unreacted Mg as well as some internal voids. The second step involved vacuum distillation, in which MgCl 2 particles (produced in the first step) were removed from the Ti particles. The vacuum-distilled Ti formed a lightly sintered "biscuit", the majority of which easily broke loose from the belt at the discharge end. This "biscuit" was very lightly ground to yield spheroidal particles (D m = ~200 μm). With further grinding, an angular Ti powder was produced. The final particle properties were influenced by grinding rate. For example, when the vacuum-distilled product was ground very vigorously (using a ring grinder), the O content of the final powder was very high, suggesting that this method of grinding may not be appropriate.

Variant of Ono and Suzuki Process
With this process, the production of Ti metal powder from FeTiO 3 involved nitridation (via TiN), sulfurization (via TiS x ), and the Ono and Suzuku (ONSU) process [53]. The starting material was TiN. Ti 2.45 S 4 and TiS 2 powders were formed from TiN at 1200˚C in 3.6 and 10.8 ks, respectively. For the sulfurization process, S 2 or CS 2 gas may be used. When S 2 gas was used, the sulfur powder was evaporated in a dual-zone furnace, at a low temperature (typically, ~300˚C). S 2 gas was transferred into the hot zone of the reactor, from where a high-purity Ar gas (≥99.9998%) stream moved it to the reaction site. The sulfides obtained in the first step were converted to a-Ti metal powders by an electrochemical reduc-World Journal of Engineering and Technology tion (ONSU process) in a molten CaCl 2 -CaS salt. The final α-Ti powder had spherical morphology and foil-like Ti sheets and an O content of <0.15 wt./wt%.

High-Temperature Remelting Spheroidization (HTRS) [55]
In the HTRS process ( Figure 5(b)), which takes place in a reducing atmosphere (75 vol./vol% argon and 25% vol./vol. hydrogen), the feedstock was placed in a hopper and then fed into a furnace tube by a screw feeder, after which the powder was dispersed in the air evenly. When the powder passed through the high-temperature zone in the furnace, the in-flight metal particles were rapidly heated and melted into liquid drops, which, due to surface tension, produced spherical particles. When HRTS was used to produce Cu particles, the process variables that influenced the values/range of values of particle size were: temperature (1100˚C -1350˚C), particle feed rate (0.1013 -0.3920 g/min), and sheath gas flow rate (150 or 50 mL/min when Ar gas or hydrogen gas was used, respectively).
Six aspects of the process were presented in detail, these being particle behavior in flight, settling process, heat transfer process, spheroidization, influence of feed rate, and influence of ambient temperature. Very brief summaries of these aspects, when Cu powder was produced, are now given.
For particle behavior in flight, Stokes formula was used to calculate the falling time and the instant velocity of Cu powder. After that, the temperature of the particles was obtained by combining the conservation energy equation with the heat transfer equation. For the settling process, three simplifications were used. First, since D was in the range of 45 -125 μm, the Brownian force and the thermophoresis force were neglected. Second, since the density of the Cu powder >> than that of the gas, the buoyancy, additional mass force, and Basset force were considered negligible. Third, the particles were considered subject only to hydrodynamic resistance. The required settling time and D were found to be inversely related; for example, the settling times of 125 μm-and 45 mm-sized particles were 0.95 s and 5.55 s, respectively. The calculation of heat exchange was based on two assumptions. First, heat conduction inside the particle was considered fast because D is small; as such, the temperature distribution in the interior of the particle was assumed to be uniform and the particle shape was taken to be spherical. Second, the powder was taken to comprise fine particles that were scattered evenly at an appropriate feed rate. Particles were regarded as a sparse phase compared to the gas phase. A heat balance equation was developed, which presented the relationship between particle temperature (T p ) and environment temperature, from which T p was obtained and found to vary with displacement. It was found that the residence time of the particles in the hightemperature region of the furnace is an important variable.
At 1350˚C, Cu spherical powder, with smooth surfaces and no satellite particles, homogeneous in microstructure, and fully dense with no pores, was obtained. For this powder, there was marked decrease in O content, marked increase in bulk density, and marked increase of flowability of HTRS-treated variant compared to corresponding values for the initial powder.
At a given particle feedstock feed rate (FR), HTRS treatment led to a narrowing of the particle size distribution (PSD) of the final powder. However, when FR was very high, the particles agglomerated. Also, at very high FR, the powder adhered to the furnace tube wall and, when sufficient build-up occurred, it fell off in large chunks, resulting in a significant reduction of the spheroidization rate (defined as the ratio of volume of spherical particles to that of non-spherical particles).
With increase in ambient temperature (T amb ), the O content of the HTRStreated powder increased very slightly while the spheroidization rate of the powder increased markedly. Most of the 45 -125 μm-sized particles spheroidized at T amb = 1300˚C, and, at T amb = 1350˚C, all the particles were perfectly spherical, with bright smooth surfaces. As such, it was suggested that it may not be necessary to use T amb > 1300˚C.

Post-Production Modification Methods
One variable investigated was the method of preparation of Ti-6Al-4V alloy powder feedstock through a process known as "in situ alloying of elemental powder mixtures" [56]. The four study groups in this study were 1) commercially-available powder produced using gas atomization (GA) ("pre-alloyed powder" group); 2) simple mixing of the elemental powders purchased from vendors: Ti powders sieved to obtain particle sizes in the range of 45 -75 μm, Al powder sieved at 45 μm, and V powders with PSD of D 10 , D 50 , and D 90 = 6.39, 21.2, and 45.8 μm, respectively ("simply mixed powder 1" group); 3) simple mixing of the Ti powder used in group 2 with fine Al powder (D 10 , D 50 , and D 90 = 2.8, 6.2, and 12.2 μm, respectively) and V powder (D 10 , D 50 , and D 90 = 6.39, 21.2, and 45.8 μm), respectively) ("simply mixed powder 2" group); and 4) the Ti, Al, and V powders were mixed using a method called "satellite mixing", which has three steps. First, the Ti powder, the fine Al powder and the V powder were dry mixed to yield a homogeneous mixture. Second, this mixture was mixed with a polyvinyl alcohol (PVA) solution until the liquid was fully dispersed through the mixture, after which the mixture was mixed in an agitator. Third, the mixture was dried in an oven and then passed through a 75-μm sieve. The morphology and flowability of each of the four powders were determined.

G. Lewis World Journal of Engineering and Technology
With regard to morphology, the particles in the pre-alloyed powder were uniformly spherical; in simply-mixed 1 powder, there were regions where Ti or Al aggregated; in the simply-mixed powder 2, random Ti particles were decorated by Al clusters; in the satellite-mixed powder, there were zones with high Al and V contents (a consequence of the presence of some large-sized V particles) and varying amounts of satellites on the Ti particles because of the irregular particle size and morphology of the particles and insufficient wetting of the powder mixture by the PVA. The difference in the morphologies of the four powders is reflected in their flowabilities (Table 1), which show that preparation of a feedstock by mixing the constituent element powders may not improve its processability for use in a powder-based MAM process.
Two key components of the inductively-coupled plasma process (ICPP) (a plasma spheroidization process developed by a commercially entity, Tekna Plasma Systems, Inc.) are an inductively coupled plasma torch placed on top of a water-cooled stainless steel chamber [57] (Figure 6(a)). In the process, a carrier/diluent gas and a water-cooled injection probe were used to inject the powder into the center of the discharge. As the individual powder particles were heated   [57] and the radiofrequency inductively coupled plasma spheroidization process (b) [59].
and melted in the plasma, they formed spherical dense molten metal droplets, which cooled and solidified as they dropped to the bottom of the chamber. The processed powders were recovered at the bottom of the chamber, while the plasma gases and any vapors or ultrafine powders exited through a side port and were directed to a sintered metal filter, from where the gases were sent to the vacuum pumping and exhaust system. The ICPP system may be operated over a wide range of plasma gas mixtures at atmosphere pressure or soft vacuum. The plasma can be used as a chemical reactor as well as an enthalpy source, which means that the system may be used to spheroidize a starting powder regardless of its method of production. The ICPP has been used to treat commercially-pure Ti powder [57], Grade 23 Ti-6Al-4V alloy powder [57], and Zr-4 alloy powder [58].

G. Lewis World Journal of Engineering and Technology
Properties of Grade 23 Ti-6Al-4V powder produced using the hydride-dehydride (HDH) method and, then, treated using ICPP, have been compared [57]. On the ICPP-treated powder, in addition to powder spheroidization, there were "satellites", which was attributed to the partial vaporization of the particles during its plasma heating and melting, followed by the condensation of the formed vapors on the surface of the particles in the cooling phase of their trajectory, resulting in soft deposits on the surface of the larger particles. When these deposits were removed, a high-quality powder was obtained (  [58]. A variant of ICCP, namely, a radio frequency inductively-coupled plasma system [59] (Figure 6(b)) was used to transform irregularly-shaped Ta powders to spherical ones that were contaminated with oxygen and contained a small number of irregular-shaped particles. Compared to the starting powder, the PSD of the spheroidized powder was narrower (D 50 for the starting and spheroidized powders were ~60 and ~55 μm, respectively) and possessed ~27% higher ρ app , ~17% higher ρ tap , and ~36% higher flowability.
The influence of treating Ti powder produced using HDH, in a high-temperature fluidizing process, on the morphology, PSD, flowability, and O content of the powder was investigated [60]. The powder was treated in a fluidized bed reactor (FBR), under high-purity Ar gas, flowing at a steady rate and a temperature (T bed ) that varied between 450˚C and 600˚C, for 10 min ("FBR powder"). The as-received powder particles were irregularly shaped and had sharp edges whereas, in the FBR powder, the particles had smoother edges (attributed to the drag force that occurred during the treatment) and some particles with D < 5 μm were either partially embedded in or adhered to the surface of larger-sized particles (attributed to bonding or coalescence between small-and large-sized particles facilitated by the high process temperature). The PSD of the FBR powder was narrower than that of the as-received powder, with increased narrowness of the PSD of the FBR powder occurring with increase in T bed . D m of FBR powder was larger than that of the as-received powder, but, with processing in the FBR, D m did not increase monotonically with T b (  [61]. When all the results were considered, for each of the powders, the influence of powder drying condition on ρ app was marginal and flowability of Ti-6Al-4V powder decreased with drying, with the best flowability being obtained when the powder was dried in air. For each of the other three powders, the influence of drying conditions on flowability was inconsistent. The influence of cryomilling (which is attrition ball milling in liquid nitrogen) on various properties of GA-produced 17 -4 stainless steel powder has been studied [62]. Two groups of cryomilled powder were obtained. In one group, the powder was not degassed at the end of the process. In the other group, the powder was degassed (powder was heated, under flowing argon, from room temperature to 325˚C, at 10˚C/min, held there for 6 h, and, then, cooled to room temperature) and, then, sieved with a −325 mesh screen. In terms of morphology, most of the GA-produced powder particles were round (mean sphericity and symmetry = 0.74 and 0.91, respectively) ( Table 2) and contained pores and satellite particles and some of the particles were hollow. Compared to the GA-produced powder particles, the cryomilled + degassed + sieved powder particles were less round (shaped like thin plates) and displayed noticeable increase in the sizes of the small particles (~59%), the medium particles (~40%), and the large particles (~25%) ( Table 2). Both powders contained the same phases and displayed very similar melting range (1456˚C -1474˚C) and solidification point (1455˚C ± 9˚C). The cryomilled + degassed + sieved powder had greater flowability than the GA-produced powder (~48%) ( Table 2), a possible explanation for this trend being that during cryomilling, the finest particles in a powder were removed.  to ~2.0, with increase in NMC. The same trends were seen for Cu powder, with the difference being that flattening led to a change of spherical powder to flake-like particulates because Cu is markedly softer than 316L stainless steel.

G. Lewis World Journal of Engineering and Technology
Subjecting HDH-prepared Ti powder to ball milling led to marked changes in various powder properties [64] [65]. Compared to the as-received powder, the ball-mill-treated powder had higher O content (by ~40%) (attributed to use of Ar gas as the shielding atmosphere); improved morphology (grinding of the edges of the powder with minimal agglomeration led to a change from nonspherical to near-spherical morphology); narrower PSD; and smaller D, which was attributed to collisions and friction that occurred during the milling (D 10 , D 50 , and D 90 decreased by ~49, ~20, and ~14%, respectively). In terms of microstructure, both powders had α-Ti structure but whereas the grains in HDH-produced powder were large (>1 mm) and possessed negligible microstrain, ball-mill-treated powder had a "core and shell" structure with nanocrystallites in the shell and stress-induced twins, resulting in considerable grain refinement (grain size: 92.89 ± 0.25 nm). Additionally, the ball-mill-treated powder had a lower Hauser Ratio (by ~13%), resulting in increase in flowability index (by ~8%), and displayed between small and very large changes in each of the three flowability/rheological properties determined (angle of repose decreased by ~16%, degree of compression decreased by ~12%, and specific surface area increased by ~83%).

Influence of Process Variables on Powder Properties
The influence of the composition of the gas used (two variants of nitrogen and Ar) in the production of AISI 316L austenitic stainless steel powder using GA World Journal of Engineering and Technology has been investigated [66]. There were three study groups, comprising powders produced using two variants of N 2 gas ("nitrogen gas A" powder (procured from one vendor) and "nitrogen gas B" powder (procured from a different vendor)) and Ar gas ("argon gas powder"). Clear trends were seen in some of the powder properties determined ( Table 3). The powders possessed comparable values of convexity and circularity but differences were seen in aspect ratio, with 65% of the particles in the Ar gas powder falling in the range of 0.9 -1.0 but only 55% of the nitrogen gas A particles or nitrogen gas B particles falling in that range. The PSDs of the three powders, as determined using the volume of particles, were similar but when the analysis was based on numbers of particles, there was a slightly higher number of fine particles in the Ar gas powder compared to either of the nitrogen gas powders. The highest and lowest internal porosities were found in the Ar gas and nitrogen B powders, respectively. Compared to the Ar gas powder, both ρ app and ρ tap for the nitrogen gas powders were lower but each of the flowability properties determined (Hausner Ratio, Hall flow rate, and angle of repose) was similar for all three powders. The avalanche angle for Ar gas powder was lower than that of either of the nitrogen gas powders, reflecting the difference in morphologies of the powders; however, some of the other rheological properties (such as specific energy (SE)) were similar for the three powders.
The packing density of the Ar gas powder was higher than that of either of the Permeability, at applied normal stress = 8 kPa (10 9 cm 2 ) 9.0 6.5 5.8 G. Lewis World Journal of Engineering and Technology nitrogen gas powders, a trend also attributed to differences in the morphologies of the powders. The permeabilities of these powders (a reflection of their packing structures) were in the order nitrogen gas A powder > nitrogen gas B powder > Ar gas powder.
The influence of several variables on PSD and D m of plasma rotating electrode process (PREP)-produced Ti-6Al-4V powder has been investigated [67]. The variables were: rotation speed (7000 -12,000 rpm), electrode diameter (15 -25 mm), plasma current (50 -70 A), gas (Ar and He), and gas flow rate (0, 70, 110, and 160 L/min). At a given electrode diameter, D m decreased monotonically with increase in rotation speed, up to 10,000 rpm, beyond which the decrease flattened out. At a given rotation speed, D m decreased with increase in electrode diameter. For a fixed combination of electrode diameter of 15 mm and rotation speed of 7000 rpm, D m increased with increase in plasma current, attributed to increase in the size of droplets from the fluid strips and possible increased recombination of the droplets in the process chamber. Compared to when Ar gas was used, D was smaller when He gas used because of the formation of a significantly greater inclined end face of the electrode (defined as the difference between the highest and the lowest parts of the residual end face of the electrode). It was suggested that this difference may be a consequence of the difference in thermal properties of the gases: specific heat of He gas is ~10 times that of Ar gas and thermal conductivity of He gas is ~9 times that of Ar gas. In either Ar or He gas, D m decreased with increase in gas flow rate and then increased with further increase in flow rate. This complicated pattern of variation was attributed to the disturbance and cooling effects caused by the flowing gas. Specifically, the disturbance effect, which led to a decrease of D m , was attributed to two phenomena: first, fluid granulation caused by the fluid stream being disturbed by the mechanical crashing of the gas on it, and, second, the mechanical crushing leads to decreased number of adjacent droplets. The increase in D m was attributed to the increased cooling rate of the molten metal. The gas flow rate at which the change of dominant effect (from disturbance effect to cooling effect) occurred was ~57% higher when Ar gas was used compared when He gas was used because the thermal conductivity of Ar gas is ~90% lower than that of He gas. The results with respect to influence of rotation speed and electrode diameter on D m and PSD, computed using a numerical method (computational fluid dynamics), were very close to those obtained from the experimental tests.

Influence of Powder Production Method on Powder Properties
There is a limited volume of literature on this topic [37] [39] [68]- [73].
Two Ti-6Al-4V powders, one produced using GA (GA powder) and the other using PREP (PREP powder), were compared with respect to morphology, surface chemical composition, and PSD [68]. μm. Two Ti-6Al-4V powders, one produced using electrode induction melting gas atomization (EIGA powder) and the other using PREP (PREP powder), were compared with respect to PSD, morphology, and microstructure [69]. With regard to PSD, while EIGA powder particles distributed in five size fractions (0 -60, 60 -80, 80 -120, 120 -180, and 180 -250 μm; D 50 = 93 μm), PREP particles distributed mainly in only two size fractions (80 -120 and 120 -180 μm; D 50 = 135 μm). EIGA powder particles were nearly spherical with several satellites, and had sub-grains that resulted in the formation of a cellular microstructure (especially in the larger-sized particles), a consequence of rapid cooling experienced during the powder production process. PREP powder particles were entirely spherical, smooth, and free of satellites, with sub-grains seen only in a small proportion of the particles. In both powders, the particles were composed of martensitic α' phase, a consequence of the very high cooling rate experienced during each powder production process (on the order of 10 3 -10 5 K/s).
Various properties of Ti-6Al-4V powders produced using three different methods were compared [70]. The methods were PA (PA powder), GA (GA powder), and PREP (PREP powder). The properties determined were morphology (through the use of scanning electron microscopy (SEM) and synchrotron x-ray computed tomography (SXRCT)), PSD, O content, flowability, ρ app , pore morphology within the powders, and porosity (Table 4). SEM showed that although each powder was spherical, PREP powder showed the highest surface smoothness and sphericity. While the PSDs of PA and GA powders were similar (characterized by bi-modal peaks), that of PREP powder was distinctly different (showing only one peak). PA powder had the smallest O content, but the flowabilities of the three powders were about the same, with the same trend found for ρ app . SEM results showed that while pores ranging in D from 10 -50 μm were found in GA and PA powders, no pores were found in PREP powder. SXRCT results showed that there were more satellites in GA powder than in either PA or PREP powders, quasi-spherical pores in GA and PA powders, irregularly-shaped pores in PREP powder, and internal pores in all three powders but more in GA and PA powders, with the smallest and largest internal pore sizes being found in PA and GA powders, respectively. For powders with D < 150 μm, mean porosity was in the order GA powder >> PA powder >> PREP, with the same trend seen in the mean Ar contents. Porosity increased with increased D, with the patterns for PA and PREP powders being the same (that is, monotonic increase). For each powder, Ar content increased with increase in D. Clear differences between the powders were seen in the variation of pore porosity: broadly speaking, at each D, porosity was in the order GA powder > PA powder > PREP powder. Results for variation of pore sphericity with D showed sphericity in the order GA G. Lewis World Journal of Engineering and Technology  powder > PA powder > PREP powder. The difference in sphericity between the different powders was attributed to differences in pressures of the gas trapped within the particles during the atomization process (~five times lower in the case of PREP powder particles compared to the case in both GA and PA powder particles). For each of the powders, three-dimensional (3D) reconstructed images obtained using SXRCT showed that the pore population increased with increase in D.
A comparison between GA-and water atomization (WA)-produced Ni-based alloy 625 powders was carried on the basis of morphology, structure, PSD, composition, and particle porosity [37]. GA powder was spherical whereas WA powder particles were angular and irregular in shape. In both powders, the particles had a dendritic microstructure. The mean O content of GA powder was ~99% lower than that of WA powder, with a similar trend seen for the C content World Journal of Engineering and Technology (~55% lower). For each of the crystallographic parameters determined, the values for the two powders were essentially the same. In differential scanning calorimetry tests, melting began in GA powder at 1304˚C and ended at 1334˚C whereas for WA powder, the corresponding temperatures were 1252˚C and 1305˚C, respectively, which indicated that the solidus temperature (T sol ) of GA powder was about 52˚C higher than that of WA powder. The difference in T sol between the two powders was attributed to the GA powder having a lower C content. For GA powder, its PSD was wider, D 50 was smaller, circularity was greater, and % mean porosity was lower compared to the corresponding values for WA powder (Table 4). For Ti-6Al-4V alloy powder produced using PREP and 5 "variations of the gas atomization (GA) process", PSD (obtained using SEM and a commercially-available particle size analyzer (Camsizer®)), morphology (using SEM, back-scattered electron SEM (BSE-SEM), and transmission electron microscopy (TEM)), microstructure, chemical composition, PSD, flowability, ρ app , and ρ app were obtained [71]. The variations in GA-produced powders reflect differences in processing conditions used by the 5 different vendors from whom the powders were procured (AP & C, ATI, Hoeganaes, Praxair, and Puris). In terms of PSD, PREP powder had a uni-modal distribution, each of 2 of the variants of GA powders had bi-modal distribution, and each of the other 3 variants of GA powders had discontinuous distribution. Camsizer® analysis results gave the range of D for PREP powder as 52 -114 μm, while, among the 5 GA powders, there were differences seen in this range (for 1 variant: 38 -102 μm and for another variant: 57 -95 μm). Mean D 50 for GA powders was considerably lower (~30%) than that of PREP powder (Table 4). For some of the powders, the PSD results from SEM and Camsizer® analyses were the same; for example, for one GA variant powder, very few particles were detected at D > 120 μm. However, for other powders, the PSD results obtained using SEM and those obtained using Camsizer® were different; for example, for PREP powder, D 50 was 64 mm and 79 μm from SEM and Camsizer® analyses, respectively.
In terms of morphology, PREP powder particles were spherical and had smooth surfaces; particles of Puris powder were less smooth than PREP ones and had satellite particles attached to many of the larger particles and not all of them were spherical; and particles of Praxis powder were, mostly, spherical, with some that were elongated.
With respect to microstructure, particles of each of the powders exhibited only the peaks associated with HCP a (or α') phase, an acicular α' martensite microstructure with lath thickness of ~200 nm, and twins within the laths. In terms of chemical composition, PREP powder had the highest mean O content, while, among the variants of PA powder, the mean content in Praxair and AP&C powders was markedly lower than that in the 3 other variants. Among all the powders, Ti, Al, and V contents were, essentially, the same. The flowability of PREP powder was within the range of that of the 5 variants of GA powder, with the same trend seen for ρ app and ρ tap (Table 4). The morphology, PSD, porosity, and rheological properties of PA-and GA-produced Ti-6Al-4V alloy powders have been compared [72]. In terms of morphology, PA particles were spherical, whereas GA powder particles were irregularly shaped. Clear differences between the particles from these two powders were seen in sphericity, PSD, ρ app , ρ tap , flowability, Hausner Ratio, and Carr Index (Table 4). Additionally, these workers [72] introduced a figure of merit (AMS) as a measure of the suitability of a powder for use in an LB-PBF process.
AMS was calculated as the arithmetic mean of the 7 rheological properties of the powder that were determined. AMS of PA powder is ~28% lower than that for GA powder (Table 4). It was postulated that the smaller AMS is, the greater is the suitability of the powder for an LB-PBF process. As such, it was suggested that PA powder was more suitable than GA powder.
3D size, shape, and internal porosity of particles of recycled Ti-6Al-4V powders that had been produced using PA and GA were determined [39]. This was achieved by use of a methodology that involved use of x-ray computed tomography (XCT) and various mathematical algorithms. The methodology was calibrated by limited visual examination of 3D images of individual particles, which allowed classification of the particles as either nearly spherical (SnS) or non-spherical (NS). The latter type can be single non-spherical particles or multi-particles (comprising 2 or more conjoined particles). Clear differences were seen between the powders in terms of the particle shape classification ( Table 5), showing that the sphericity of the PA powder particles was greater than that of the GA powder particles. However, the various aspect ratios of the two powders were comparable (Table 5) as were the various PSDs (that is, those based on W and VESD), L/T, L/W, and W). The superior quality of PA powder was also seen in the porosity results (Table 5).
For PA-and HDH-produced Nb-47Ti powders, clear differences were seen in each of the properties determined, with one exception [73]. PA powder particles were spherical, had mean ρ app of 3.45 g/cm 3 , Hall flowability of 20 s/50g, and mean O content of 0.12 wt./wt%, whereas HDH powder particles were irregularly shaped, had mean ρ app of 1.85 g/cm 3 , no measurable flowability, and mean O content of 0.34 wt./wt%. The exception was that the PSDs of the powders were similar. There was no discussion of these results except to show that the effective [105] [106], have also been studied.
For Ti-6Al-4V alloy powders intended for SLM and EB-PBF, powder reuse has a very marginal influence on PSD and aspect ratio [75]. With respect to flowability of the powder intended for EB-PBF, powder reuse impaired flowability (higher SE, higher basic flow energy (BFE), lower conditioned bulk density, and lower flow energy after 50 taps). The opposite outcome was found for the powder intended for SLM.
For ELI Ti-6Al-4V powder, the influence of reuse of the powder (up to 21 cycles) on O content, particle morphology, PSD, flowability, ρ app , and ρ tap has been determined [76]. With increase in the number of reuse cycles, O content increased in a monotonic manner, the PSD became narrower although there was very little change in D 50 (Table 6); the powder particles became less spherical and rougher, distorted after the 16th and 21st reuse cycles ( Figure 7); satellite particles were visible after the 2nd and 6th reuse cycles but there were few of these after the 11th reuse cycle ( Figure 7); flowability improved by a modest amount in that the flow rate increased by ~12%, ρ app remained about the same, and ρ tap decreased by ~3% ( Table 6). The following explanations were given for these trends: increase in O content was due to pick-up of oxygen when the powder was exposed to air, including when the unmelted powder was blown off from the part being built. As for the narrowing of the PSD, during the time that the unmelted powder was blown from a part being built, and, then, collected, some satellite particles became detached from the parent particle and was carried away by the gas being used in the MAM process. Improved flowability was not consistent with a priori expectation and was attributed to the removal of satellites from the particles and a reduction of moisture in the powder after prolonged exposure to vacuum, at elevated temperature, in the build chamber machine.
For PA-produced Ti-6Al-4V powder, powder reuse (up to 7 cycles) yielded changes in the phases in the powder [77]; namely, α or martensitic α' in the virgin powder; β phase after the 1st and 2nd cycles; α phase boundary after the 3rd cycle; near-equiaxed prior β grains and phase boundary after the 4th cycle; and   For Ti-6Al-4V [78], with reuse (over a maximum of 69 cycles), the change in the Ti, Al, and V contents was each negligible, but the O content increased progressively; and the particle morphology changed from spherical (virgin powder) to one that contained a variety of defects, such as elongated particles, bonded and agglomerated satellites, and satellites with particles, a consequence of the sieving procedure and overheating and smelting.
For ELI Ti-6Al-4V alloy powder, reused powder had higher O content and essentially unchanged nitrogen content [79]. Additionally, with increase in number of builds (1 -31), the PSD became narrower, with no change in D 10 or D 50 but decrease in D 85 ; the particles remained essentially spherical but coarser and there were fewer satellites; ρ tap decreased; and flowability improved (as a result of decreased number of satellites).
For GA-produced Ti-6Al-4V alloy grade 23 powder [80], after 15 reuse cycles, the PSD narrowed, with D 10 being approximately unchanged, D 50 being slightly smaller, and D 90 being markedly smaller; the particles remained spherical, with fewer satellites; O content increased while nitrogen content was unchanged; flowability improved (markedly lower BFE and markedly lower permeability); and thermal conductivity increased.
For Ti-6Al-4V powder [81], the virgin powder particles were nearly spherical and contained a few slightly elongated particles and a few broken particles, and were fully dense, with a feather-like microstructure. In the powder after the 100th reuse cycle, there were satellite particles attached to the powder surface and a few aggregates. Powder reuse led to narrowing of the PSD, decrease in D 50 (~21%), and a large drop in flowability (~54%) ( Table 6).
In the case of PA-produced Grade 5 Ti-6Al-4V powder [83], the virgin powder particles were smooth, with a high degree of sphericity and comprised a large range of particle sizes, satellite particles that adhered to the larger-sized particles, and a number of non-spherical particles. With increase in the number of reuse cycles (up to 30), the particles became more irregularly shaped, there was increase in particle surface damage, the number of smaller-sized particles decreased, and the PSD narrowed (from a bimodal distribution and tended to a Gaussian distribution after the 30th reuse cycle), with the trends in D 10 , D 50 , and D 90 being small increase, no change, and marked decrease, respectively. As for chemical composition, the O content increased markedly with increase in powder reuse cycles, with the content after the 11th cycle exceeding the ASTM F2924 limit [84] (0.2 wt./wt%). Most likely, the surface damage to the particles occurred during powder handling and recovery. Decrease in the number of small-sized particles may be the consequence of them being consumed during part building, fused onto the surface of larger-sized particles, or lost during powder recovery. This decrease may be the reason for the trend in D 10 while the trend in D 90 may be accounted for by size of the sieve used to collect the used powder (120 sieve), which removed powder particles and agglomerates with size > 125 μm.

G. Lewis World Journal of Engineering and Technology
Inconel 718 and Ti-6Al-4V powders were the subject of one study [86]. For Inconel 718 powder, with increase in the number of powder reuse cycles, there was minimal change in particle shape, increase in particle agglomeration (attributed to slow sintering in the powder mound in the additive manufacturing machine), slight narrowing of the PSD but not in a systematic manner, increase in the number of large-sized particles, and increase in D 50. For Ti-6Al-4V powder, there was also an increase of D 50 , but there was slight broadening of the PSD, and fast sintering led to large particles being sieved out at the end of each build, resulting in fewer sintered agglomerated particles.
For GA-produced Inconel 718 powder, powder reuse (up to 50 cycles) led to various changes in powder properties [93]. O content markedly increased (by ~85%) to a level that was still below that specified for use of this powder for additive manufacturing of parts for aeronautical applications. There was noticeable change in PSD (for example, increased D 10 , D 50 , and D 90 (Table 6)) and decreased circularity. Increase in particle sizes was attributed to the presence of spatters and satellites in the reused powder. Both ρ app and ρ tap of the reused showed very small decreases (~3% in each) and, hence, flowability increased (Table 6). It was suggested that the increased flowability may be linked to the decreased amount of small particles in the powder or particle oxidation.
For GA-produced nickel superalloy ATI 718 powder [94], powder reuse (up to 9 cycles) led to changes in particle size and morphology of the powder. Specifically, the number of large-sized particles increased and many of these particles had non-spherical agglomerated or fused particle morphologies, the PSD flattened, and O content increased considerably (by a factor of ~2) (it was suggested that the presence of features, such as films and spots, which consisted of Al oxides, may contribute to this increase), D 50 increased (by ~28%), and flowability decreased (by ~15%) ( Table 6).
In the case of GA-produced IN718 powder [95], with increase in powder reuse (between 1 and 14 times), the bulk composition of the powder was, practically, unchanged; the PSD shifted to the right (meaning that D 10   For GA-produced AISI 304L stainless steel powder [99], compared to virgin powder, recycled powder was more spherical, had noticeably higher O content (~35%), and displayed very small increases in both ρ app and ρ tap (~2% for each parameter), and increased flowability (for example, ~14% in decrease in avalanche energy) ( Table 6). With increase in the number of reuse cycles, there was G. Lewis World Journal of Engineering and Technology a small amount of both coarsening of the powder particles and increase of the aspect ratio of the particles. In terms of morphology, clear differences were seen in the powder after the 7th reuse cycle compared to the virgin powder; namely, recycled powder contained large-sized particles, some of which were highly spherical and others that were irregular.
For GA-produced 316L stainless steel powder, influence of powder reuse on PSD, porosity, pore composition, NH, and effective modulus (E r ) of the particle was determined [100]. Compared to the virgin powder, the reused powder (over the course of 10 printing cycles) displayed about the same "average" particle size, a larger number of small-sized particles, a broader pore size distribution, a slightly reduced pore size (by ~10%), higher pore density, higher Fe, Cr, and Ni contents on the interior wall of the pore, higher Mn, S, and Si contents in the outer layer around the pore and on the surface of the particle, and lower NH and higher E r (at a given depth of indentation). The larger number of small-sized particles was explained by their being blown away during the LB-PBF process (SLM). Higher pore density was attributed to the redistribution of the composition of the alloy powder during SLM, while the reduced NH and increased E r was attributed to the increased porosity of the particles.
For GA-produced AISI 316L stainless steel powder [102], with reuse (up to 15 printing cycles), O content increased markedly (by ~37%), D increased slightly (for example, D 50 decreased by ~9%; with some irregular-shaped particles seen), and flowability increased slightly (flowability time decreased by ~10%, which was attributed to decreased number of small-sized particles (which results in a reduction of interparticle separation) and increased number of particles whose surface was oxidized).
For AlSi10Mg_200 alloy powder [103], powder reuse did not affect particle morphology (spherical or near spherical, a few particles with rounded oblong shape, and a few satellites); surface finish (smooth); D 50 (~9 μm); microstructure (the silicon phase was distributed in the dark Al matrix in the form of a bright network of plates/islands that have thin and long arms around them); and phase composition (no shifting of Al peaks).
For AlSi10Mg alloy powder [105], with increase in number of builds (up to 8), the PSD shifted slightly to the left, indicating reduction of the larger-sized particles, with no change in D 10 , no change in D 50 , and a slight decrease in D 90 ; the particles remained predominantly spherical; and there was a very small increase in ρ app and a modest increase in ρ tap .
The morphology, PSD, circularity, chemical composition, and phase (crystal structure) of virgin powder and reused powders (50 cycles) of 316L stainless steel and AlSi10Mg intended for use in an LB-PBF process were compared [101].
For each of the powders, with reuse, PSD shifted to slightly higher values (Table   6). For 316L stainless steel powder, both virgin and reused powder particles were largely spherical or near-spherical, but there were satellites in the virgin powder.
The morphologies of the virgin and reused AlSi10Mg powders were the same, World Journal of Engineering and Technology except for the presence of satellites in the reused powder. While mean circularity of 316L stainless steel powder was, practically, unchanged with reuse (mean reduction ~2%), there was a larger decrease in that parameter in the case of Al-siMg10 powder (mean decrease ~17%). Each of the recycled powders retained its chemical composition, but with increased O content (by ~230% and ~102% for 316L stainless steel and AlSi10Mg powders, respectively (Table 6)). For each of the powders, powder reuse did not affect its microstructure (in the case of 316L stainless steel, a single γ-austenite FCC structure) and there were no precipitates and oxide layers. Because of the increase of O content in the recycled powder, which is often associated with the presence of contaminants inside the build chamber, it was noted that even though no phase change was observed in the XRD spectra, change that was below the limit of detection of the XRD equip- Si10Mg powder (~25% increase).
Very limited attention has been given to characterizing the physical degradation suffered by a metal powder over the course of reuse [85] [92] [106]. In each of these referenced studies, the degradation was monitored by using x-ray photoelectron spectroscopy to determine the thickness of the oxide on the particle, and consistent results were obtained. With powder reuse, the thickness increased (mean of ~16% in the case of Ti6Al-4V over 10 reuse cycles [85] and mean of ~850% in the case of AlSi10Mg powder over a reuse period of ~30 months [106]). This trend was consistent with the trend that for a combination of either alloy powder and reuse duration, O content of the powder increased markedly [92] [106].
From the review of the studies on Ti-6Al-4V powder, one noticeable feature is that while in some studies, it was found that with reuse, D 50  Considering all the results summarized in this Section, there is little consistency in the trends of the properties that were most commonly determined (Table 8). "Consistency" is herein taken to be the same direction of a trend, as

Appraisal of the Literature
In this Section, a number of shortcomings of and gaps in the literature are highlighted.
First, although extensive and detailed characterizations of powder particles have been conducted (Table 9), there is scope for more widespread use of more advanced characterization techniques such as synchrotron x-ray computed tomography scanning (SXCTS) for 3D reconstruction of the particles [70]; SXCTS and mathematical analysis for determination of aspect ratio of particles [39]; and synchrotron-based hard x-ray photoelectron spectroscopy for determination of the chemical composition across particle pores [100].
Sixth, from the fourth-and fifth-mentioned issues above, it is clear that there is need for a consensus to be reached on definitions of "reuse" and "reuse cycle".
After that, a recommendation should be made as to the minimum number of reuse cycles that should be used in studies on the influence of powder reuse on powder properties.
Seventh, there is a limited number of replication studies on the influence of powder reuse on powder properties (even in the case of Ti-6Al-4V alloy powder, which has attracted the most attention). One consequence of this shortcoming, coupled with the fourth-and fifth-mentioned issues stated above, is that, for a given alloy, it is very challenging to perform inter-study comparisons and, hence, to draw, isolate, and discuss consistent trends that may serve as the basis for improvements in the additive manufacturing of parts from the alloy.
Eighth, there is inconsistency on how the terms, "flowability" and "flowability rate" are used. It is recommended that "flowability rate" be used but with it defined as the inverse of the result obtained in a Hall test, as was done by Kroeger et al. [30]. In other words, present the result for flowability rate in the unit of 50 g/s or, simply, g/s.
Ninth, even though, in many studies, a sizeable number of properties of the powder particles were determined, there is very limited attempt in using these results to develop a methodology to guide powder selection. In fact, when such studies were conducted, the same approach was used [72] [87]. Brika et al. [72] presented such a methodology, culminating in the presentation of an "additive manufacturing suitability figure of merit (AMS)" applicable to an LB-PBF process, based on the arithmetic mean of the values of 7 flowability properties of a fresh powder (the closer AMS for a reused powder is that for a fresh powder, the more suitable the reused powder is for use in a laser powder bed fusion AM process). Similarly, Shangbag et al. [87] computed the arithmetic mean of 13 properties of fresh and reused powders and designated it an "electron beam powder bed fusion (EB-PBF) suitability factor (ESF)", with the lower the factor, the more suitable is the powder for use in an EB-PBF process. Two shortcomings of this approach are highlighted. First, equal weighting factors were assigned to each of the properties, and, second, well-established material selection methodologies, such as the Desirability Method [108], the VlseKriterijumska Optimazicija I Kompromisno Resenjein in Serbian (VIKOR) method [109], and the weighted aggregate sum product assessment (WASPAS) method [110], were not used. One of the many attractive features of each of these methodologies is that a distinction is made between properties that need to be maximized (in the case of powders, an example is circularity of the particle) and those that need to be minimized (in the case of powders, examples are powder SE, flow rate index, and repose angle) [111]).
Tenth, a few studies have been reported involving so-called "composite G. Lewis powder feedstocks". Such a feedstock is defined herein as one in which a starting powder is modified by mechanical alloying (ball milling) with either an additive (for example, 316L stainless powder ball milled with 1 or 2 wt./wt% Y 2 O 3 [112] or 1 wt./wt% CrN [113]) or deposition or coating of an additive on its surface (for example, 0.3 wt./wt% nano-sized Y 2 O 3 deposited on the surface of a ferritic stainless steel powder [114]). In these reports, very few properties of the composite powder have been presented and there are no comparisons between those properties and those of the starting powder counterpart. Additional work in this area should be on developing other methods of obtaining composite powder feedstocks. Eleventh, well-designed proof-of-concept studies should be conducted. One such study should compare powder produced using an established method (such as Ar gas atomization) versus powder produced using an emerging method (see Section 2) versus a composite powder feedstock, for a specified alloy. The comparison should include determination of powder properties (in particular, pore properties) as well as detailed characterization of the performance of a powder during part building using a specified powder-based MAM technology (for example, determination of melt pool geometry and spatter pattern if an LB-PBF method is used) as well as determination of a large assortment of properties of the built parts. These properties should include fatigue life under multiaxial loading conditions and spectrum loading conditions (using specimens that contain flaws, such as notches [115]), fatigue crack propagation rate [116], and corrosion performance (using a test medium that is appropriate for the anticipated service use of the built parts, such as phosphate buffered saline solution, at 37˚C, for alloys to be used to fabricate total joint replacements [117] [118]).
Twelfth, either limited or no attention has been paid to a number of issues. As such, there are many areas for future research. Example issues are 1) development of novel environmentally-friendly and low energy-intensive powder production methods [23]; 2) development of powder production methods that will yield powder particles that do not contain satellite particles; 3) development of international materials testing standards for determining rheological properties of powder (similar to ASTM 3049 [119]) and, as an adjunct to this effort, establishment of acceptable values of the properties (as has been done with O content, with the maximum allowable content being and 0.3 wt./wt% (per ASTM B988 [120]) and 0.2 wt./wt% (per ASTM F-3001 [121]) for Grade 5 and ELI Ti-6Al-4V alloys, respectively); 4) utilization of machine learning/artificial intelligence/deep learning methods to accomplish two purposes. The first would be determination of the key process variables, for a given powder production method. The second would be establishment of the influence of this collection of process variables on a large and varied assortment of powder properties, which would culminate in the optimization of these variables [122] [123]; 5) determination of influence of conditions under which a powder is stored prior to its use to build parts (notably, ambient temperature and relative humidity in the storage facility) on powder properties [124]; 6) development of novel methods of separation/fractionation World Journal of Engineering and Technology of as-received powder into a PSD that is suitable for a given powder-based MAM process, such as 20 -63 μm-sized particles for LB-PBF of light metals [125]; 7) development of methods for improving flowability of a powder without adversely affecting any of its other properties, notably, PSD (and, hence, D 50 ) [126]; and 8) methods for producing high-quality powders of new/emerging metallic materials, such as multi-principal element alloys/complex concentrating alloys/high-entropy alloys (for example, FeCoNiCrAlCu [127]), Al crossover alloys (for example, AlMg4.7Zn3.6Cu0.6 [128]), ferritic oxide-dispersion-strengthened alloys that are strengthened by a yttrium-based nano-sized oxide (for example, Fe-10Al-4Cr-4Y 2 O 3 [129]), and low-modulus β-Ti-based alloys (for example, Ti-38Nb-5Zr-2Ta-2Sn [130]), for use in powder-based MAM technologies.

Summary
The following are the main points made in the present review: 1) Methods of production of metal powders for use in powder-based MAM technologies are being expanded with the introduction of many emerging ones, such as plasma and gas hybrid atomization (PGHA), wire induction gas atomization, hydrogen-assisted magnesiothermic reduction, and a granulation-sintering-deoxygenation process. Extensive characterizations of metal powders produced from both established methods (such as argon gas atomization and water atomization) as well as emerging methods have been carried out. Powder properties determined include particle size distribution (PSD), phase composition, flowability, repose angle, and Carr Index. With each of the emerging methods, spherical powder was obtained.
2) In cases where powder produced using an established method is not spherical or very close to spherical, post-treatment methods have been presented for spheroidizing these powders. Two examples are a radiofrequency inductively coupled plasma process and ball milling.
3) Only a few studies have been conducted on the influence of process variables, for a given established powder production method, on powder properties. In one of these studies, it was found that, for AISI 316L austenitic stainless steel, the PSD (based on analysis of volume of particles) and flowability of powder produced using argon gas atomization were comparable to those of powder produced using nitrogen gas atomization but apparent density was lower and avalanche angle was greater for the powder produced using nitrogen gas. In another study, this one involving PREP-produced Ti-6Al-4V powder, median particle size decreased with the increase of electrode diameter, at a given electrode rotation speed. 4) Many of the studies on the influence of powder production method on powder properties have been on Ti-6Al-4V alloy powder and compared powders produced using GA, PREP, EIGA, and PA. One consistent trend is that compared to GA-produced powder, PA-produced powder had greater sphericity but comparable aspect ratio and PSD.
G. Lewis World Journal of Engineering and Technology 5) There is a sizeable volume of literature on the influence of powder reuse on a large assortment of properties of Ti-6Al-4V, Ni-based alloys (Inconel 718 and AT1718), stainless steels (316L, 304, and 17-4), and Al-based alloys (AlSi10Mg and Scalmaloy). However, because of variability in terms of number of studies (for a given alloy), powders studied, and powder properties determined, the only consistent trend is that for Ti-6Al-4V, Ni-based alloys, and stainless steels, powder reuse leads to marked increase in O content of the powder.
6) The literature on the aspects of the powder covered in the present review has many shortcomings and gaps, suggesting areas for future work. Examples of these areas are for one metal powder produced using one method, determination of a large and varied collection of powder properties as a function of powder reuse cycle (up to at least 50 cycles); for one metal powder, determination of a large and varied collection of properties when the production method is established (such as argon gas atomization) versus when it is an emerging method (such as PGHA); application of machine learning/artificial intelligence/deep technique method to determine the optimized values of all the key process parameters for a given combination of metal powder and powder production method; and development of methods of production of powders of new/emerging metallic materials, such as crossover Al alloys and low-modulus β-Ti alloys, for possible use in MAM technologies.