Surface chemistry in Ti-6Al-4V feedstock as influenced by powder reuse in electron beam additive manufacturing

Abstract

X-ray photoelectron spectroscopy (XPS) as well as scanning and transmission electron microscopy (SEM/TEM) analysis was carried out on four Ti-6Al-4V powders used in electron beam powder-bed fusion (PBF-EB) production environments: virgin low oxygen (0.080 wt% O), reused medium oxygen (0.140 wt% O), reused high oxygen (0.186 wt% O), and virgin high oxygen (0.180 wt% O). The two objectives of this comparative analyses were to (1) investigate high oxygen containing Grade 23 Ti-6Al-4V powders which were further oxidized as a function of reuse and (2) comparing the two virgin Grade 23 and Grade 5 powders of similar oxygen content. The microstructure of the low oxygen virgin Grade 23 powder was consistent with martensitic α′ microstructure, whereas the reused powder displayed tempered α/β Widmänstatten microstructure. XPS revealed a decrease in TiO₂ at the surface of the reused powders with an increase in Al₂O₃. This trend is energetically favorable at the temperatures and pressures in PBF-EB machines, and it is consistent with the thermodynamics of Al₂O₃ vs. TiO₂ reactions. An unexpected amount of nitrogen was measured on the titanium powder, with a general increase in nitride on the surface of the particles as a function of reuse in the Grade 23 powder.¹

1. Introduction

The ongoing development of metal additive manufacturing (AM), particularly electron beam and laser powder bed fusion (PBF-EB and PBF-L, respectively) of Ti-6Al-4V alloy, has inspired new technologies and advancements in AM, with applications growing by the day. That being said, AM Ti-6Al-4V alloy has yet to be fully adopted for fatigue- and fracture-critical applications due to an incomplete understanding of the processing–structure–property–performance relationships (PSPP) that exist across the many different technologies that additively produce Ti-6Al-4V parts [1]. Successful AM of Ti-6Al-4V parts is already challenging with respect to optimization of post-processing treatments, with only recent developments to hot isostatic pressing (HIP) treatments leading to isotropic microstructures while retaining as-built strengths [2]. In conjunction with microstructural effects to material performance, chemical homogeneity in Ti-6Al-4V feedstock also plays a large role in AM material performance. A recent review by the present authors highlights the current state of the art with respect to Ti-6Al-4V powder oxidation and reuse in both PBF-EB and PBF-L processes [3]. The two PBF processes, (PBF-EB and PBF-L) differ from one another such that the PBF-EB is carried out in a partial vacuum at elevated temperatures (≈ 600 °C–800 °C), whereas PBF-L is conducted at atmospheric pressure at room temperature with an inert cover gas, usually Ar. The high background processing temperature inherent to PBF-EB allows for the Ti-6Al-4V powder bed to accumulate more oxygen (either in the form of oxides or interstitial oxygen) from various impurity sources relative to the PBF-L process [3], and has been shown to have a gettering effect on the atmosphere of the chamber, leading to higher levels of oxygen content in the powder closest to the build plate (beginning stages of the build) [4]. Typically in full scale AM production environments, oxygen content of reused powder batches is only assessed in bulk form. A thorough understanding of the surface chemistry is needed to fully understand the PSPP relationships in PBF Ti-6Al-4V given that surface oxides typically do not melt and tend to float around melt pools (as slag), and could accumulate at the surface of melt pools leading to unwanted, periodic inclusions in the bulk given the extremely high number of melting events in a layered additive process. The focus of the present investigation will be the surface chemistry of Ti-6Al-4V powders used in the PBF-EB process, sampled from a PBF-EB Ti-6Al-4V production environment.

Historically, increased interstitial elements in titanium alloys have led to higher strength and reduced ductility [5], with strength contributions originating from increased oxygen, specifically in wrought Ti-6Al-4V [6]. Studies by Tang et al. [7] and Quintana et al. [8] show increased strength and minimal losses in elongation in AM Ti-6Al-4V as a function of increased powder reuse (and therefore oxygen content) for both PBF-EB and PBF-L, respectively. In most cases for PBF of Ti-6Al-4V, increases in oxygen content occur in the powder feedstock, likely due to the powder handling and reuse of the Ti-6Al-4V feedstock for both processes, with additional avenues for oxidation occurring in PBF-EB due to layer preheating [3]. However, a study of PBF-EB Ti-6Al-4V performance as a function of powder reuse by Popov et al. [9] further elucidates the strength–ductility trade-off as a function of increased oxygen content and recycle time, where a reduction in elongation is observed with a high number of powder reuses. Popov et al. [10] also show via crystallographic investigation that these materials still may be viable for static-strength use-cases, even with increased oxidation from powder reuse. Increased oxygen content in AM Ti-6Al-4V has also been shown to reduce impact toughness when manufactured via PBF-EB [11], to reduce fatigue resistance in PBF-L [12], and lead to anisotropic tensile behavior when produced via direct energy deposition [13]. Oxygen is an α stabilizer, and can be desirable to increase the strength of Ti-6Al-4V alloy. However in this stage of technological development, oxygen is an unwanted and uncontrollable by-product of the powder reuse and handling process. This unwanted oxidation can lead to detrimental variations in material performance. This was recently observed by the height-dependent oxidation that occurs in PBF-EB Ti-6Al-4V virgin build powder beds [4], which contributed to a significant variation in tensile strengths of resultant, virgin as-built parts based on build height [14]. The stochastic variation in tensile properties of PBF-EB Ti-6Al-4V was also shown to increase when virgin powder was blended with reused powder that still conformed to standard limits [15]. Recent work has even shown that viable scaffolds for bone tissue regeneration can still be fabricated in PBF-EB with reused Ti-6Al-4V powder [16], therefore it is critical to understand all of the oxidation mechanisms present in these materials to enable fracture-critical use cases.

From a manufacturing perspective, the increase in oxygen content as influenced by powder reuse can lead to costly discarding of Ti-6Al-4V powder when the chemical content of the powder no longer meets the criteria set forth in the standards for PBF of Grade 23 (ASTM F3001 [17]) and Grade 5 (ASTM F2924 [18]) titanium. This results in individual manufacturers and research bodies optimizing their powder handling process in-house. In the case of PBF-EB, which will be the focus of the current work, the users of PBF-EB technology are required to use the Arcam EBM Powder Recovery System (PRS)² in order to blast away the sintered powder that adheres to the as-built part, resulting from the layer preheating/sintering step that is employed in PBF-EB. A recent study by Sun et al. investigated Ti-6Al-4V chemistry, morphology, internal microstructure, and size distribution as a function of reuse cycles [19]. Sun reveals that the powders are generally more irregular in morphology after recycling, as well as having a narrower size distribution. More interestingly, Sun et al. revealed the oxide thickness to be relatively similar across all of the reused and virgin particles, concluding that oxygen preferentially diffused into the β phase due to the faster diffusion of oxygen through the body-centered cubic (BCC) phase than the hexagonal closed packed (HCP) α phase. The authors also observed differences in the microstructures of reused particles—some displaying tempered α/β Widmänstatten microstructure, while others showing morphology resembling martensitic α′ [19]. However, closer inspection with Auger electron spectroscopy of the surface characteristics in reused Ti-6Al-4V powder revealed anisotropy in oxide layer thickness as well as localized aluminum oxide (Al₂O₃) enrichment across particles [20].

Previous work does not elucidate the mechanisms by which oxygen is incorporated into PBF-EB Ti-6Al-4V powder over many reuses. This work investigates these incorporation mechanisms by characterizing surface chemistry via XPS and microstructure via SEM/TEM for powders of differing oxygen content and levels of reuse. The results of this work can be used to further optimize the PBF-EB Ti-6Al-4V process, including powder reuse methodology.

2. Materials and methods

2.1. Ti-6Al-4V powders

Standard PBF-EB atomized powders over a range of oxygen content and reuse extent were obtained for this study from an ISO 13485 [21] certified medical device production environment. The reused powder was collected after various builds. Given that the reused powder batches used in this study were collected at different times throughout the industrial workflow powder life cycle, only bulk oxygen content was tracked as an indicator of lifetime. The sample designations are as follows:

• Grade 23, virgin, low oxygen = VL (gas atomized)

• Grade 23, reused, medium oxygen, = RM (gas atomized)

• Grade 23, reused, high oxygen = RH (gas atomized)

• Grade 5, virgin, high oxygen = VH (plasma atomized)

The sample designations of V and R refer to virgin or reused powder, wheras the -L, -M, and -H refer to low, medium, and high oxygen content. RM consists of the VL lot mixed with virgin powder of a different lot to bring down the overall oxygen concentration, however RH powder is the end-of-life powder with high oxygen stemming from VL powder that has been heavily reused. The usage of the powder lots in the AM production environment were monitored for increased oxygen content from reuse. The powder reuse process consisted of recovering the sintered powder using the Arcam PRS, with additional ‘refreshing’ of the powder with virgin powder on an as-needed basis (mixed via drum mixer for 45 min) to remain within the ASTM F2924 oxygen limit (0.20 wt% O) for the production environment whence they came. The builds using the reused powder were completed on an Arcam A1 EBM machine. Powder chemistry information was provided by the manufacturer in the form of inert gas fusion measurements using a LECO machine. The particle sizes for all of the powders ranged from 45 μm–106 μm ≥ 96.0% by mass.

2.2. Electron microscopy

Powder morphology was imaged using a scanning electron microscope operated at 20 kV, whereas backscattered electron (BSE) imaging was carried out with a Robinson backscattered electron detector with an accelerating voltage of 30 kV. Flat polished specimens were prepared by hot mounting the Ti-6Al-4V powder in conductive phenolic resin powder, followed by metallographic abrading and polishing with 1 μm diamond suspension. The mounted samples were then vibratory polished for approximately 24 h in 0.05 μm colloidal silica.

Focused-ion beam (FIB) machining was used to extract TEM foils from specific locations on the mounted samples previously characterized by SEM for transmission electron microscopy (TEM) analysis. A dual-beam SEM/FIB was used to prepare the specimens, which were then examined in a TEM operated at 200 kV. TEM foils were specifically extracted from the edge of the mounted particles in order to examine the surface of the particles.

2.3. X-ray photoelectron spectroscopy (XPS)

Samples were prepared by adhering double sided copper tape to Au coated Si wafers. The powder was added in excess onto the adhesive surface and gently pressed into the tape using a spatula. The excess powder was removed and loaded onto the sample bar. From each vial, 4 separate samples (5 for the VL specimen) were prepared in this fashion for XPS analysis, mounted onto a sample bar, and pumped down to ultra-high vacuum. Detailed information on the XPS data acquisition and analysis is provided in tabular form in the Supplemental Information (SI). Please refer to Tables S1–S2 for details.

XPS was performed with an Axis Ultra DLD X-ray photoelectron spectrometer from Kratos Analytical (Chestnut Ridge, NY). Samples were pumped down to ultra-high vacuum conditions via a load lock and into the analysis chamber (base pressure approximately 5 × 10⁻¹⁰ torr). Photoelectrons were emitted using monochromatic, Al kα operating at a power of 150 W (10 mA emission; 15 kV anode bias) and collected along the surface normal from an area of analysis defined by the hybrid lens and slot aperture. Photoelectrons were analyzed at a pass energy of 40 eV (survey scans collected at 160 eV) with spectra acquired at 0.1 eV increments (1 eV for surveys). To eliminate differential charging, the samples were neutralized with low energy electrons while the stage was disconnected from ground. Spectra were acquired for the O 1s, V 2p, Ti 2p, N 1s, C 1s, and Al 2p regions.

After the XP spectra were acquired, the data was processed using CasaXPS (Teignmouth UK). Each region was fit with a Shirley background and charge referenced based on the fitted, Ti 2p3/2 metal component at 453.9 eV [22,23], unless otherwise noted. While efforts were made to generate control spectra from standards such as Ti-6Al-4V and TiO₂, the generated spectra did not match up well with the AM powders. Therefore, peaks were modeled based off of literature references from comparable methods [22]. Further details on the peak fitting, which was employed to determine shifts in oxidation, can be found in the SI (Figures SI1 and SI2). From the fitted data, areas were calculated and corrected with the instrument manufacturers provided elemental relative sensitivity factors for each component and converted into atomic percentages. Values and uncertainties reported in the manuscript are representative of the average atomic % and one standard deviation of 4–5 samples (1 measurement/sample).

3. Results

3.1. Bulk powder analysis

The chemical composition as measured via inert gas fusion are reported from the manufacturer certifications of the powders, and are presented in Table 1. Together, the VL, RM, and RH powders represent real-world production environment powder lots that are beginning, middle, and end-of-life, respectively. The oxygen concentration measured via LECO for each powder lot is listed in Table 2. The particle sizes for all of the powders ranged from 45 μm–106 μm ≥ 96.0% by mass. The virgin Grade 23 powder, VL-1 is what is referred to as VL in this work. An additional virgin Grade 23 powder (VL-2) lot was mixed into the VL lot by the part manufacturer prior to this work to bring the overall oxygen concentration down as the powder was reused and crept up in oxygen towards levels outside of the ASTM F2924–14 standard [18]. This mixture represents RM, with ‘M’ indicating a ‘medium’ oxygen content for the production cycle. The RH powder is the highly reused powder from the VL lot.

Table 1.

Chemical composition in weight % of the virgin Grade 5 and Grade 23 Ti powders as measured via inert-gas fusion. Tr. = trace elements. VL-1 refers to the virgin powder lot used in experimentation and analysis for VL and RH. VL-2 refers to the virgin powder lot that was mixed in with the aged powder from VL-1 to form RM. ASTM standard specification maximum limits (unless otherwise noted) for PBF Ti-6Al-4V Grade 23 (ASTM F3001 [17]) and Grade 5 (ASTM F2924 [18]) are listed with the measured values.

Grade	Al	V	Fe	O	C	N	H	Y	Tr.	Tr. Max.	Ti
Std. 23	5.5<x<6.50	3.5<x<4.5	0.25	0.13	0.08	0.05	0.012	0.005	0.10	0.40	bal.
VL-1	6.41	3.90	0.17	0.08	0.01	0.01	0.001	<0.001	<0.10	<0.40	bal.
VL-2	6.41	4.03	0.20	0.10	<0.01	<0.01	0.001	<0.001	<0.10	<0.40	bal.
Std. 5	5.5<x<6.75	3.5<x<4.5	0.30	0.20	0.08	0.05	0.015	0.005	0.10	0.40	bal.
VH	6.52	3.91	0.18	0.18	0.02	0.01	0.002	<0.001	<0.10	<0.40	bal.

Table 2.

Oxygen and nitrogen concentrations of the powders in this study measured via inert-gas fusion.

Sample	wt.% O	wt.% N
VL	0.080	0.010
RM	0.140	–
RH	0.186	0.042
VH	0.180	0.010

The high oxygen content of the heavily reused RH powder is comparatively similar to the virgin VH powder. The microstructure, phase fraction, and surface chemistry however, are not as easily compared via bulk chemical analysis methods such as inert-gas fusion. The Grade 5 and 23 powder lots are very chemically similar, with the main difference being the oxygen content (presented in Table 2), as mandated by the ASTM F2924 and F3001 standards for a powder-bed fusion process. The RH powder was sent for additional testing for N content and was measured to be 0.418 wt% N.

3.2. Electron microscopy

The morphology of the powders used in this study presented as secondary electron (SE) images in Fig. 1. The gas atomized, virgin Grade 23 powder, VL is pictured in Fig. 1a, where there are many satellite particles present attached to the larger particles. The appearance of some of these particles can be described as, elongated, agglomerated and/or having satellites, using the nomenclature as described in Ref. [9]. The recycled powder presented in Fig. 1b, RM, displayed similar characteristics to the VL powder, albeit containing fewer attached satellite particles and agglomerates. The shape of the powders is also more deformed, displaying rougher surfaces than the virgin lot, and appearing slightly deformed from the spherical initial shape. Fig. 1c displays the RH powder, where the particle morphology is similar to that of Fig. 1b. The plasma atomized virgin Grade 5 powder, VH is pictured in Fig. 1d, where dendritic features from rapid solidification were observed on the surface of the particles. This powder morphology is noticeably different from the gas-atomized VL powder, appearing significantly smoother than the VL lot, which is resultant from the plasma atomization process. These differences between VL and VH stem from the differences in the gas versus plasma atomization process; plasma atomization tends to produce powders with smoother surfaces and lower occurrences of non-spherical particles. The VH powder also contains bonded satellites, as well as fused particles.

Fig. 1.

Secondary electron images of the powder lot morphology. (a) VL, (b) RM, (c) RH, and (d) VH.

The microstructure of the flat polished powder cross-sections are presented in Fig. 2. The microstructure of the VL particle, shown in Fig. 2a, consists of fine acicular α morphology. The needle-like structures closely resemble martensitic α′ microstructure, typically resulting from rapid solidification processes. Given that the VL powder is directly from the gas-atomization process and has not been exposed to the electron beam or high background processing temperatures like RM and RH, it is likely that this microstructure is mostly α′, however without definitive crystallographic analysis, it can only be stated as likely based on the microstructural features. Random thickness sampling of the α (or α′) laths in the VL microstructure gives approximate thicknesses ranging from ≈ 100 nm–700 nm.

Fig. 2.

Backscattered electron images of the cross-sectioned and polished powder lots. Images display a representative particle from each lot. (a) VL, (b) RM, (c) RH, and (d) VH.

The RM powder imaged in Fig. 2b displays a more tempered microstructure, with α/β Widmänstatten morphology distinguished by the α laths with brighter regions between them, caused by the Z-contrast of the V-rich β phase. These microstructural transformations were also observed from virgin to recycled powder by Sun et al. in their powder recycling study for PBF-EB Ti-6Al-4V [19]. The α lath thickness ranges from ≈ 400 nm–1200 nm for the RM powder. The microstructure of the RH powder in Fig. 2c is very similar to the RM powder, however the outer morphology and shape of the particle is noticeably distorted from the heavy reuse. The microstructure of the virgin VH powder is shown in Fig. 2d, displaying a finer Widmänstatten microstructure than the RM or RH powders.

In an effort to visualize the effects of powder recycling on the surface oxidation, TEM images of the four powder conditions are presented in Fig. 3. The images represent a flat cross-section of the powders similarly to Fig. 2, but magnified at the surface interface between the metal and conductive Bakelite mounting media. In Fig. 3a, a thin layer of mounting resin remains on the surface of the VL particle from the FIB thinning process, but no evidence of a discrete surface layer can be observed at these magnifications. In Fig. 3b, the resin has more cleanly broken away from the RM particle surface, leaving a clean surface with little evidence of a discrete surface layer. In Fig. 3c, the bright spot was determined to be an artifact from the FIB milling process. In Fig. 3d, the resin is evenly thinned and still attached to the VH metal particle’s surface. Additional measurements using energy dispersive X-ray spectroscopy revealed that there was no apparent oxide concentration at the surface of the particles when measured via this technique.

Fig. 3.

TEM bright-field (a) and dark-field (b–d) micrographs of the cross-sectioned and polished powder surfaces. Micrographs display a representative particle from each lot. (a) VL, (b) RM, (c) RH, and (d) VH.

3.3. XPS

Representative XP spectra for the different Ti-6Al-4V powders are presented in Fig. 4, including the O 1s, Ti 2p, N 1s and Al 2p regions. V 2p (not shown) was also detected with a 2p_3∕2 component peak maximum between 516 eV–517.3 eV consistent with oxidized vanadium. Indeed, all metal components of the Ti-6Al-4V powders exhibited both oxidized and zero-valent components within their spectra. The Ti 2p spectra and the Al 2p spectra below provide examples of this. In the case of Ti 2p, the spectra was fit for 4 different functionalities, Ti⁰, Ti²⁺, Ti³⁺, and Ti⁴⁺, at locations proximate to previous literature results [22,23]. Please see the Supplemental Information for representative examples of the fitting. The Al 2p region was also separated into zero-valent (Al⁰) and Al³⁺ species, which was also consistent with previous approaches [24]. The O 1s region consisted of two clear components, labeled O–Ti and O–Al, at 530 eV ± 0.1 eV and 532 eV ± 0.2 eV, respectively, which are attributed to titanium oxides and aluminum oxides, respectively. There was also C 1s component that consisted of carbonyls attributable from surface contamination. The last spectra in Fig. 4 is the N 1s regions, which has two components attributed to nitride and amine species. For representative examples of fitted spectra (Fig. SI 1) and unfitted C 1s stackplots (Fig. SI 2), please refer to the Supplemental Information.

Fig. 4.

Representative XP spectra of the chemical composition of the Ti-6Al-4V powders. Top to bottom: VH, RH, RM, and VL.

Fig. 5 presents the elemental distribution of species of the different Ti-6Al-4V powders evaluated in this study. All samples had a significant carbon contribution in all powders, presumably due to surface contamination. However, in order to retain the chemical information from the Ti-6Al-4V powders, surface cleaning using an ion mill was not conducted. Oxygen was also present in high surface concentrations for all samples which is consistent with the spectra. With respect to the low percent contributions, the elemental distribution of Ti 2p decreased slightly with increased use of the reused powders while the Al 2p and N 1s percentages increased, both of which are consistent with the spectra presented in Fig. 4. Interestingly, the surface elemental distribution of the low percentage components between native virgin VL and VH powders were strikingly different, which will be discussed later.

Fig. 5.

High and low percentage elemental contributions to the XPS measurements.

As demonstrated in Fig. 4 spectra, there were numerous interesting results in the distribution of chemical species from a magnitude perspective and a proportionality of oxidized species, as demonstrated in Figs. 6 and 7, respectively. For example, the Ti 2p components for Ti-6Al-4V powders decreased in overall oxidation with increased use for the reused samples as shown in the bar chart (Fig. 5) seen as an increase in Ti⁰: Ti (Total) ratio in Fig. 7. The Al 2p region also showed a trend towards proportionally greater amounts of unoxidized aluminum with increased use, while the magnitude of Al_xO₃ surface concentrations actually increased significantly with the highest powder use. Comparing the two virgin powder samples, the magnitude of surface oxidized aluminum was much greater for the VH specimen, but the proportion/distribution of zero-valent:oxidized Al 2p functionality was found to be comparable. Lastly, the N 1s spectra for the reused powders had > 1% nitrogen, split between an amine-like component and a nitride-like component. With increased use, the nitride component increased while the amine component decreased. While the two virgin specimens appeared to have proportionally the same distribution of nitride:amine, the VH powder specimen had less than half of the total nitrogen content.

Fig. 6.

Aluminum, titanium, and nitrogen valence component contributions to the elemental composition percentages. Al (O), Al (III), Ti (0), and Ti (IV) refer to Al⁰, Al³⁺, Ti⁰, and Ti⁴⁺ respectively.

Fig. 7.

Ratio of elemental valences and species collected via XPS for Ti, Al, V, and N by powder type. Multiple markers indicate repeat measurements. Al (O), Al (III), Ti (0), and Ti (IV) refer to Al⁰, Al³⁺, Ti⁰, and Ti⁴⁺ respectively.

To summarize the XPS results, two different Ti-6Al-4V powders were examined, one after different reuse durations in PBF-EB. The reused powder increased in Al 2p and N 1s, while the Ti 2p diminished minimally with increased use. While the Ti 2p component remained only minimally changed in total surface concentration, there was a redistribution of oxidation states with a shift away from signal intensity in Ti⁴⁺ and towards Ti³⁺, Ti²⁺, and Ti⁰ with increasing use. Al 2p also demonstrated the same trend from a proportional standpoint, although as previously stated the magnitude of all species (Al³⁺ and Al⁰) increased. Continued use of the Ti-6Al-4V powders seemed to increase nitride content while decreasing amines. Lastly, comparison of the two virgin powders, VL and VH, demonstrated a stark difference in surface Ti:Al ratios (Fig. 7). However, distribution of oxidation states within a given element were nearly identical (Al, N) or comparable (Ti) with respect to the high use state of reused powders. Lastly there was an inverse relationship observed between the Ti⁴⁺ and the Al³⁺ surface concentrations (Fig. 7, lower right), which will be further discussed later.

The inelastic mean free paths (IMFPs) for the elements were estimated from Ref. [25] and are presented in Table 3. Additionally, the IMFPs were converted into estimates for information depth incorporating 95% of the inelastically scattered photoelectrons (ID₉₅) using Eq. 22 provided in a recent manuscript by Powell et al. [26]. While this is an estimate that makes assumption, we feel that this provides a useful point of comparison which we will further discuss later.

Table 3.

Inelastic mean free paths estimated from Ref. [25] and ID95₉₅ calculated from Ref. [26].

Element	Orbital	KE (eV)	IMFP (nm)	ID95 (nm)
O	1s	957	2.482	7.435
V	2p3	971	2.509	7.516
Ti	2p	1028	2.618	7.483
N	1s	1090	2.735	8.193
C	1s	1201	2.942	8.813
Al	2p	1413	3.329	9.973

4. Discussion

The electron microscopy of the powder lots in Fig. 1 revealed a general decrease in sphericity for the reused powders as a function of use. The gas atomization process employed in the manufacturing of these powders led to the initial powder morphology, however slight changes in exterior coarseness of the particles become visible with reuse. There was also a decrease in the size of the satellites likely from the sieving process, which has been observed as the powder size distribution narrows from reuse [7]. The microstructure of VL powder was consistent with α′ microstructure resulting from high cooling rates inherent to the powder atomization process. The study by Sun et al. characterized six virgin Ti-6Al-4V powder lots, which also revealed the α′ microstructure, as caused by the rapid solidification involved in the atomization process [28], which is suggestive that the micrograph displayed in Fig. 2a is consistent with α′ microstructure. There are also lighter contrast web-like regions of the particle in Fig. 2a, which could possibly be grain boundaries rich in α [29], however more crystallographic investigation is required and is beyond the scope of the current study. The contrast mechanism in BSE imaging depends on atomic number and electron channeling, the latter of which is orientation dependent [30].

As the RM, and RH powders were subjected to high background processing temperatures (approximately 650 °C) utilized in the PBF-EB process, the microstructure relaxes to α/β Widmänstatten microstructure in the RM powder, that slightly coarsens in the RH powder lot with increased exposure to high temperatures. There were several instances of particle bonding in the RH powder due to solid state sintering during the layer preheating employed in PBF-EB [31]. Further microstructural analysis using TEM of the powder particle surface (beam normal to the flat polished specimen) elucidated that an oxide layer could not be resolved with SEM/TEM at the magnifications employed in this study.

A discontinuous oxide layer across the particle surfaces is likely present in the powders examined in the current study, however this was not apparent during TEM imaging. Along with bulk oxygen chemistry measurements of reused Ti-6Al-4V powders [7], surface oxidation has been previously observed using XPS and Auger electron spectroscopy (AES) [20], where the increase also occurred as a function of reuse times and therefore, as a result of elevated temperatures and atmospheric exposure during handling. The AES measurements by Cao et al. revealed a discontinuous oxide layer across highly reused particles ranging from 4.1 nm to 17.8 nm [20]. A previous study on wrought Ti-6Al-4V has shown that through thermal oxidation, the rutile phase of Al₂O₃ becomes the predominant surface phase which leads to increased surface roughness and microhardness when exposed to temperatures of 600 °C to 800 °C in a water vapor environment. Table 3 lists the IMFPs and ID₉₅ values of elements measured via XPS, and therefore the depth, ID₉₅, at which these concentrations were measured, ranging from approximately 7.4 nm to 10.0 nm.

The bar graphs in Fig. 5 appear to show a general increase in oxygen from VL to RH. For the RM specimen, oxygen measured at the surface appears to be lower than the virgin powder. In order to verify that these trends are real (based on the assumption that C is a contaminant, and that varying concentrations of C could potentially alter the observed trends), plots excluding carbon and normalized to 100 at. % for the remaining elements (O, Al, N, V, Ti) are presented in the Appendix A as Fig. 9. The normalized plots show relatively little change in surface oxygen by itself for VL, RM, and VH, with a decrease in oxygen for RH. This could be due to the majority of the oxygen at the surface of the heavily reused powder being in the form of oxide as opposed to interstitial oxygen. The remaining trends for Al, N, V, and Ti for the normalized data (Fig. 9) are consistent with the trends in Figs. 5 and 6.

For the VL/M/H powders, there was a general decrease in Ti at the surface, with slight decreases in Al and V for the RM powder whereas there were increases in Al, V, and N in the RH powder. The highly reused RH powder ultimately resulted in slightly less Ti at the surface than the virgin powder, but nearly doubled in aluminum content. The VH powder on the other hand was measured to have less Ti, V, and N and significantly more Al than any of the VL, RM, and RH powders. The oxygen content for the RH and VH powders were comparable, which is similar to the bulk oxygen measurements from inert gas fusion (Table 2).

The bar graphs shown in Fig. 6 represent the oxidation state components pertaining to the overall measurements of the elements on the surface. The measured Al consisted of mostly Al³⁺, relating to Al₂O₃, whereas Al⁰ represents metallic Al, and has been previously observed in surface studies of Ti-6Al-4V powders when stored in air [32]. There was a general increase in Al³⁺ as the powder was reused, while the Ti⁴⁺ (TiO₂) decreased. An Ellingham diagram, adapted from Ref. [27], is presented in Fig. 8 to highlight the partial oxygen pressures for equilibrium oxide production for Al₂O₃ and TiO₂. From the diagram, it can be seen that Al thermodynamically reduces Ti, which means that the free energy for Al₂O₃ production is more negative than that for TiO₂ production. The decrease in TiO₂ at the surface is consistent with the thermodynamics of TiO₂ reduction in the presence of Al metal at the high background processing temperatures inherent to PBF-EB. Therefore, as the powder was cycled through the machine, exposed to the beam, and subsequently handled in atmosphere, the oxide at the surface began shifting from TiO₂ to Al₂O₃, and is displayed in terms of atomic percentages for the Ti⁴⁺ and Al³⁺ oxidation states in Fig. 7.

Fig. 8.

Ellingham diagram for Ti, Al, and V. Adapted from Ref. [27]. Pressures indicated on the dotted free energy curves are partial pressures of oxygen. Marked points are the melting points of metallic aluminum and titanium.

An interesting, albeit unexpected, observation from the chemical analysis were the nitrogen concentrations at the surface that were greater than vanadium for all of the powders. The N content increased from the VL to the RH powders, and was measured to be slightly less than either of the VL or RH powders in the mixed RM lot. Organic nitrogen and nitrides were previously observed in a XPS study of porous Ti-6Al-4V dental implants [33], however the literature is sparse with regard to the surface nitridation of Ti-6Al-4V powders used in AM. Fig. 6 displays a decrease in the amine component (NH₃) with an increase in the nitride component.

While the intensity was attributed to oxides for the Ti²⁺ and Ti³⁺ component oxidation states in the fitting process, it is important to acknowledge that there could be contributions present from nitrogen functionalities, such as TiN or TiN_x species. The VL and VH powders showed similar chemistry pertaining to the ratio of nitride to amine. Fig. 7 displays the ratios of the nitrogen components between each other as a function of powder reuse. As the powder is reused, the ratio of nitride to amine increases. In bulk Ti-6Al-4V, the diffusivity of N is relatively small compared to the diffusivity of oxygen. However, with increased nitridation at the surface of the powder lots, nitrogen is physically being introduced into the bulk material during powder bed fusion. This may or may not lead to deleterious effects in the performance of AM Ti-6Al-4V parts. However, it presents an interesting problem as to what degree are these nitride inclusions affecting material performance, given that TiN has drastically different material properties than Ti-6Al-4V and is commonly used in wear applications as a hard coating. A 2016 white paper by Renishaw PLC [34], showed nitrogen content increased in reused PBF-L Ti-6Al-4V powder, eventually reaching material specification limits. Renishaw noted that it took more reuses for nitrogen to reach allowable limits compared to oxygen, which suggests they feel oxygen is still the limiting element when considering usable lifetime of a powder lot. However, they only consider the Grade 23 limit for oxygen (0.13 wt% O [17]). Since the Grade 5 limit for oxygen is much higher than Grade 23 (0.20 wt% O [18]) and the nitrogen limit does not change for Grades 5 and 23 (0.05 wt% O [17,18]), it seems further investigation is necessary before concluding that nitrogen is not the limiting element in determining the usable lifetime of PBF-L and PBF-EB Ti-6Al-4V lots of powder. The heavily reused powder in this study, RH, contained 0.418 wt% N measured via inert-gas fusion. This value is close to the ASTM F2924 and F3001 limits for allowable nitrogen content (0.05 wt% N), and could result in non-conformance to the standard if a powder batch is allowed to age at similar rates to oxygen.

Processing of reactive metals (such as Ti alloys) typically involves in-situ gas purification using Ti as a getter to scrub the cover gas of impurities [35–38]. The high-background processing temperatures used in PBF-EB are generally in the range of 650 °C to 800 °C, to which the powder in the bed is subjected for the entire time. The Ti-6Al-4V powder sits at elevated temperatures, and therefore reacts with the chamber atmosphere impurities during the build process. This however may not be the only source of oxidation/nitridation in this work, as water vapor can be trapped in powder lots during atmospheric powder handling and reuse methodologies, which can also lead to increased oxidation in the powder lots as a function of reuse [39]. Oxidation can also occur at different rates depending on the build height of the powder bed [4]. Overall, increased O/N from Ti-6Al-4V powder reuse can lead to interstitial diffusion into titanium alloys, which can cause changes in strength and ductility [5], and can be deleterious to material performance if not carefully controlled.

5. Conclusion

Microstructural and surface analysis was conducted on virgin Grade 5 (high O) and Grade 23 (low O) Ti-6Al-4V powders, with additional focus on the Grade 23 powders that have increased oxidation through normal reuse in a qualified AM production environment. The most significant results are as follows:

• Scanning electron microscopy of the powders revealed a change in the microstructure from virgin to reused powders from martensitic acicular α′ to Widmänstatten α/β microstructure. The magnifications employed for this work using TEM could not able to resolve the presence of the oxide layer.

• X-ray photoelectron spectroscopy showed significant chemical trends across the surface of the reused powders. Through reuse, the amount of TiO₂ decreased, while Al₂O₃ increased on the surface due to thermodynamic favorability.

• The Ti-6Al-4V powders showed increased surface nitridation with increasing reuse.

These results present questions as to what effect these nitride inclusions have on material performance, and whether nitrogen content should be tracked with more scrutiny in bulk AM Ti-6Al-4V materials.

Appendix A

See Fig. 9.

Fig. 9.

Elemental and component concentrations excluding C measurements, normalized to 100 at. %.

Appendix B. Supplementary information on fitted and unfitted spectra, data collection parameters, and data analysis parameters

Supplementary material related to this article can be found online at https://doi.org/10.1016/j.apsusc.2022.154280.