Effect of atomization on surface oxide composition in 316L stainless steel powders for additive manufacturing

The initial oxide state of powder is essential to the robust additive manufacturing of metal components using powder bed fusion processes. However, the variation of the powder surface oxide composition as a function of the atomizing medium is not clear. This work summarizes a detailed surface characterization of three 316L powders, produced using water atomization (WA), vacuum melting inert gas atomization (VIGA), and nitrogen atomization (GA). X‐ray photoelectron spectroscopy (XPS) and scanning electron microscopy analyses were combined to characterize the surface state of the powders. The results showed that the surface oxides consisted of a thin (~4 nm) iron oxide (Fe2O3) layer with particulate oxide phases rich in Cr, Mn, and Si, with a varying composition. XPS analysis combined with depth‐profiling showed that the VIGA powder had the lowest surface coverage of particulate compounds, followed by the GA powder, whereas the WA powder had the largest fraction of particulate surface oxides. The composition of the oxides was evaluated based on the XPS analysis of the oxide standards. Effects of Ar sputtering on the peak positions of the oxide standards were evaluated with the aim of providing an accurate analysis of the oxide characteristics at different etch depths.

The initial oxide state of powder is essential to the robust additive manufacturing of metal components using powder bed fusion processes. However, the variation of the powder surface oxide composition as a function of the atomizing medium is not clear.
This work summarizes a detailed surface characterization of three 316L powders, produced using water atomization (WA), vacuum melting inert gas atomization (VIGA), and nitrogen atomization (GA). X-ray photoelectron spectroscopy (XPS) and scanning electron microscopy analyses were combined to characterize the surface state of the powders. The results showed that the surface oxides consisted of a thin The field of powder bed additive manufacturing (AM) has recently seen enormous growth, which has been associated with an increase in the number of powder producers, which utilize various melting practices and atomizing media. As of today, most of the produced powder for the powder-based metal AM are vacuum melted and atomized using high purity Ar gas. However, lower-cost powder produced using open air melting and nitrogen gas atomization, or even water atomized grades, can be expected to start making their way into the AM machines as the process becomes more robust and efficient, opening itself up to more powder grades. This is a viable option for many iron-based alloys, such as stainless steels, engineering steels, and some tool steels. This development, however, is not suitable for highly reactive alloys, such as Ti-alloys.
There are, however, potential drawbacks to using reactive atmospheres during atomization. From the powder metallurgical field, it is known that water-atomized pre-alloyed iron powders with strong oxide forming elements like, Cr and Mn form, besides the dominant Fe-rich thin oxide film, minor amounts of oxide particles. [1][2][3] If not reduced, these more stable oxsides remain within the consolidated sample. [4][5][6] Similar correlations have been drawn within the powder bed fusion AM processes, but these have mostly concerned powders that have been recycled. [7][8][9] There is hence a lack of published papers focusing on the surface oxide of original powder and the effect of atomizing media on surface chemical composition of the powder in the field of additive manufacturing. There is also lack of knowledge concerning the interaction of the high-power beam (e.g., laser or electron) with powder surfaces covered by oxide species in the actual AM fabrication process. Some research on direct energy deposition has indicated that oxides tend to migrate to open surfaces, 10 whereas other studies have indicated the presence of oxide films on fracture surfaces, causing a premature failure of similarly built alloy IN718. 11,12 These examples emphasize the importance of the surface oxide characteristics of the powder. Thus, before implementing such powders into powder bed systems, it is beneficial to know the surface chemical characteristics of classical gas-and water-atomized grades of iron-base powder in relation to the industry standard of vacuum melting inert gas atomization (VIGA) powder.
Nyborg et al 13 showed that surface characterization techniques, such as X-ray photoelectron spectroscopy (XPS) and Auger electron spectroscopy, are able to provide valuable information concerning the surface state of powder particles. Ion etching, together with mathematical models taking effects such as the angular dependence of the photoelectron flux and the ion sputter yields into account, provided means of performing accurate depth profiles, facilitating the characterization of oxides on various kinds of powder. The early work by Nyborg et al 14,15 showed that the presence of silicon oxide on the surface was minimal for inert gas-atomized ferrous powders, regardless of its content in the alloy. The same author 16,17 later also showed that oxidation of Si could be forced by altering the processing conditions. The relative presence of tetravalent Si would increase as a function of the decreasing partial pressure of oxygen in the atomizing atmosphere and with increasing cooling rate of the metal particles in case of water-atomizing. A recent study 18 that compared gas-atomized to water-atomized powders provided similar observations; no Si oxide was found on a gas-atomized powder, whereas Mn together with Fe and Cr formed the surface oxide instead. Additionally, the content of Mn in the surface oxide layer on the gas-atomized powder was found to increase as the powder particle size decreased, in line with the observations by Norell and Nyborg for a gas-atomized 12% Cr-steel powder. 14 For water atomized powders, the oxidation is governed by the cooling rate, whereby higher degree of selective oxidation is coupled with smaller oxide thickness as cooling rate increases. For gas-atomized powders, O 2 activity in the atomizing gas was shown to be the dominant factor. 2,14,15,17,19 This study aims to investigate how the atomizing media affect the surface oxide condition of 316L powder variants in comparison to a VIGA powder. As highlighted in a recent review, the powder cost is the second or third largest cost when producing AM parts. 20,21 Therefore, as a means of improving the utilization of additive manufacturing, it is important to highlight the similarities and differences between the powder qualities in ways other than flowability, bulk chemistry, and packing density. The approach involves a dedicated application of surface chemical analysis by means of X-ray photoelectron spectroscopy (XPS) and high-resolution scanning electron microscopy (HR-SEM) combined with X-ray microanalysis. That offers a means of coupling the surface morphology (oxide features, etc.) of powders to the chemical composition of the surface oxide products (given by XPS), thus providing a detailed description of the surface state for powders. It is envisaged that the study provides a basis for the expanded application of normal inert gas-atomized and water-atomized grades of powder in additive manufacturing by powder bed fusion.

| MATERIALS AND METHODOLOGY
The scope of this study was to investigate the effects of the atomizing medium on the surface oxide of the powder as well as its bulk chemical composition. Therefore, three grades of stainless steel 316L powder were selected. These were produced using VIGA, air melting and nitrogen gas atomization (GA), and water atomization (WA). The powder materials were provided by Höganäs AB. The VIGA powder was supplied in the 15-45 μm sieve cut, whereas the two other types of powder were supplied in the 20-53 μm sieve cut. The chemical composition of the powder materials was measured by three techniques: inductively coupled plasma-optical emission spectrometry (ICP-OES) using a SPECTRO ARCOS for the majority of alloying elements (e.g., Cr, Ni, and Si), combustion gas analysis for the interstitial elements C, and S using LECO CS844 and hot fusion analysis for the interstitial elements O, and N using LECO ON836.
A general overview of the morphology of oxides found on the particle surfaces was provided by HR-SEM, using a Leo Gemini 1550 instrument, coupled with energy dispersive x-ray spectroscopy (EDS) with an Inca X-Max detector. Further characterization was performed by XPS using a PHI 5500 instrument, followed by transmission electron microscopy (TEM), FEI Tecnai T20, involving prior sample preparation using focused ion beam (FIB), FEI Versa3D, to create milled liftout from an individual metal particle.
The powder samples were mounted on carbon tape and investigated using an HR-SEM equipped with an in-lens detector and imaged using an accelerating voltage of 15 kV. Special attention was dedicated to the oxide particulate features on the surfaces of the powder. The particulate features were further analyzed using EDS microanalysis, whereby the accelerating voltage was reduced to 5 kV to promote a smaller interaction volume to probe as little of the powder bulk as possible. L α lines were used for the quantification of Fe, Cr, and Mn.
The PHI 5500 instrument XPS was equipped with monochromatic Al K α source as well as standard nonchromatized dual Mg K α / Al K α and charge compensation was applied when needed. The spectra were then shifted according to set the position for adventitious C1s peak at 284.8 eV. However, the method of peak alignment using the C1s peak has an inherent error of ± 0.1 eV, 22 with some reports of a variance of up to ±0.3 eV. 23 This means that positioning of characteristic peaks cannot be done with higher precision than this, and peaks closer than the quoted variance cannot either be ascertained.
The beam size was 0.8 mm, thus analyzing a large number of particles simultaneously. This brings the benefit of a statistically reliable result of the surface oxide composition. The generated photoelectrons were collected in a hemispherical analyzer at a pass energy of 23.5 eV using a step size of 0.1 eV. The peaks were fitted using the Multipak software provided by Physical Electronics Inc. Peak shapes were fitted using an asymmetric peak geometry. The background estimation was performed using the iterated Shirley approach within the Multipak software. The binding energy scale of the system was checked prior to the analysis using the peak positions of Au 4f 7/2 , Ag For the lift-out, the water atomized powder was mounted onto a carbon tape and introduced into a FIB-SEM. A layer of Pt was deposited onto the surface of a metal particle and milled normal to the bulk of the particle until a lamella was nearly free. An Omniprobe was then attached to the lamella and the last attachment point to the metal particle milled away. The lamella was then attached to a copper grid and subsequently thinned to electron transparency.
The TEM was equipped with a LaB 6 cathode, and it was operated at 200 kV. It was used for capturing bright field images of the metal particle lift-out. The same instrument also has scanning transmission electron capabilities and is capable of EDX analysis. This was used to analyze the composition of the oxides from the lift-out. Table 1 presents the results from the chemical analysis of the analyzed grades of powder. All the powder materials were within the 316L specification. The major differences between them were the contents of silicon and manganese. The WA powder contained the least amount of manganese, while also containing more silicon. The VIGA powder had only traces of silicon. Differences inherent to the different atomizing media were detected in the analysis of the lighter elements, such as the higher nitrogen content in the GA powder and the higher oxygen content in the WA powder, the latter being predominantly a result of the larger specific surface owing to its more rough character for a given size range.

| SEM and EDS analysis of the surface oxides
The powders were further analyzed using SEM; see Figure 1. Submicron oxide particulate features were found on the powder surfaces of all studied variants. Representative particulate features are highlighted with arrows in Figure 1. The contrast and morphology of the oxide particulate features varied among the different kinds of powder. Even for the same powder, different oxide morphologies were found. An example of this can be seen on the GA powder, in which the surface was decorated with both small bright features and larger dark spherical features. The oxide particulates found on the VIGA powder were seemingly flat, like a splat, with some tendency to form interdendritically on the surfaces. However, the occurrence of the oxide particulates was low, and the surfaces seemed relatively clean. The GA sample had similar features to those found on the VIGA powder. In addition to these features, there were oxide particulates with a darker contrast. These had a hemispherical morphology and appeared to be larger compared to the other oxide particulate features. Such features were not as common as the previously mentioned oxide features.
Finally, the water atomized powder surface was also decorated with particulates. The particulates in this case were of a flake morphology, with clearly defined edges. Sometimes, the splats and the darker Results from the chemical analysis of the examined powders, in weight percent hemispherical objects, similar to the ones observed in the gas atomized powders, were also found on the surfaces of the water atomized grade.
EDS line scans were performed on the particulate features to provide semi-qualitative information concerning their composition.
Due to the size of the particulates, the accelerating voltage of the microscope was reduced to 3 (VIGA) and 5 kV to accommodate for this. Figure 2 Table 3.
To assess the influence of Ar ion sputtering on the XPS peak characteristics (peak position, width, and shape), the oxide standards  Table 4 containing a summary of the peaks. show a good fit to the multiplet structure, which can be seen in Figure 6, and is therefore likely a mixed Fe oxide. The Fe metal peak was also weakly discerned in the Fe 2p 3/2 peak in all of the samples, showing that the surface oxide was thinner than 3λ ox in parts of the surface, which for the 2p 3/2 -signal from the Fe-oxide would be about 3 × 1.5 nm = 4.5 nm. 13 After ion etching to 10 nm, only metallic Fe was detected.
Chromium in its oxide state was detected in all samples, with some variation in the peak positions, as can be seen in Table 4. At 10 nm, metallic Cr was found to be dominating as compared to the oxide state for the GA and VIGA powder samples, whereas the spectra of the WA powder still had significant contribution from the oxide state. Intense Mn peaks representing the oxide state were detected in both the VIGA and the GA powder, but absent in the WA powder.
The peaks appearing in the Mn-spectra range in the WA powder after ion etching to 3 and 10 nm were due to the Ni Auger emission, which also explained the shift (−0.6 eV) of the peaks in the Mn-spectra range recorded from the GA and VIGA powder samples after ion etching to 10 nm. Nickel appeared in the metallic state at 3 nm for all the powder grades and the Ni Auger peak was calculated to be half the intensity of the Ni 2p 3/2 peak. Silicon was present in its oxide state on the GA, and the WA powders, with traces also detected in the topmost surface of the VIGA powder. The peak at 99 eV that appeared at 3 nm and increased in intensity at 10 nm is elemental Si.
To be able to delineate composition gradients on the powder surfaces from potential Ar + induced damage, the sputter cleaned states were compared to the oxide standards, some of which were similarly sputter cleaned; see Table 2 and Figure 3. Several elements in the 316L powders were observed to undergo a chemical shift (see Table 4) relative to the as-received state; specifically, these were Fe, Mn, and Cr. The Fe peak shifted nearly −1 eV, whereas the Cr and Mn peaks had minor shifts, which were consistent with the sputtered standards. Thus, the as-received states of the powder were used to assess the proper state of the oxide species, as the observed shift due to Ar ion sputter damage cannot be disregarded.   Although the enrichment factor highlights the element enrichment at the surfaces, it underestimates the presence of elements that are also abundant in the alloy matrix (e.g., Fe and Cr). In addition, enrichment factors do not provide information concerning the chemical state of a specific element. To properly assess the state and quantity of these elements, each of the spectra from the narrow scans were deconvoluted and related to the total intensity of the peak.

| XPS depth profiling of the powder samples
These data were normalized and can be found in Figure 9. Here, the relative cation content is displayed as a function of the etch depth. To read the figure correctly, it is important to note that the data in Figure 9 does not take into account that the oxide is gradually removed as the surface is ion etched, but it can be viewed as a means of telling which oxide species extend to greater depth than others.  Note. LOD indicates that the peak was under the limit of detection. Peak positions after 1 nm etching are provided to show the peak shift relative to the as-received state. for the VIGA and GA powder, respectively. The discrepancy between these values and the values calculated using the relative intensity of oxygen is due to the influence of the particulate compounds, as described above. Thus, larger difference between the values determined for Figures 5 and 9D indicates that the coverage of particulate compounds is larger relative to the iron oxide layer, and vice versa.

| TEM of FIB lift-out
An FIB lamella for TEM studies was prepared from the WA powder to verify and delineate the results acquired from the XPS measurements.
To alleviate the Ga + induced amorphization and damage, and to increase the chances of extracting an actual surface oxide, a thicker oxide particulate was chosen for the lift-out, see Figure 10. Figure 11 presents a TEM bright field composite image of the lift-out; the arrows indicate (1) the Pt layer, which was deposited on top of the powder particle (dark grey contrast), and (2) the oxide particulate, which was the target for the lift-out. EDS spectra of the particulate compound were recorded, in which it was found that the oxide contained roughly F I G U R E 6 Spectral fit of the Fe 2p 3/2 peaks of the VIGA, GA and WA powder using the multiplet structure described in Table 3, the fit envelope is drawn in red F I G U R E 7 Depth profiles and quantification of the present elements for the studied powders: (A) VIGA, (B) GA, and (C) WA equimolar amounts of Cr and Si; see Figure 11C. The particulate was found to be heterogenous, with smaller nanometric spherical particulates within the thicker layer; see Figure 11B. The smaller particulates were found to be enriched in Si (not shown here). It appears that the oxide had grown inwards into the powder particle, as it extends beyond the powder particle surface, which is outlined with a dashed white line in Figure 11.
High resolution TEM was used to investigate the interface between the metal and the oxide particulate. Figure 12 shows the interface between the particulate and the matrix, and the oxide particulate was found to be amorphous; see fast Fourier transformations (FFT) insets.

| DISCUSSION
The results from the SEM imaging and the XPS measurements accord with the common consensus of the surface oxide characteristics in case of both inert gas atomized powder and water atomized powder of stainless steel. Both kinds of inert gas-atomized powder (VIGA and GA) had particulate compounds rich in Mn and Cr cations; the GA did, however, has a significant amount of oxidized Si present (>6 at.%), which has not been previously reported. This suggests that the atomizing gas was very low on residual O 2 , forcing elements with the highest oxygen affinity, in this case Si, to oxidize. 16 The peak fitting and deconvolution to define specific chemical states proved to be more difficult, as some shifting of the peak positions was observed, as well as the shift of C1s, which was used as a reference. The two inert gas atomized samples displayed many simi-  Putting everything into the context of AM, the crucial aspect is to determine whether differences in the surface oxide composition are significant between the different atomization methods and how these differences could potentially affect the processability of such powders in the case of powder bed fusion processes. The difference between the VIGA powder and the GA powder was considered minimal, particulate coverage, and the amount of oxidized Si were the sole differences. Previous research 39 shows that Mn can readily combine with Si to form low-temperature melting silicate phases. The presence of a Mn-silicate on the top surface would provide an explanation for the Si and Mn surface enrichments found on the gas atomized powder. However, the presence of such phase on a 316L powder surface in similar amounts was found not to be detrimental in the case of L-PBF, as such particulates were found as fine inclusions with a similar or identical composition inside the dimples on the fracture surface, with no registered impact on mechanical properties due to their low content in the material. 40 Similarly, no nitrogenrich inclusions were found within the dimples of the same study, hence the nitrogen remains in solid solution throughout L-PBF processing. 40 Furthermore, in SiO 2 -Cr 2 O 3 systems, such low melting phases do not form, and thus, the oxidation products found on the WA sample would not pose a risk.

| CONCLUSIONS
This work has examined the effect of the atomizing medium on the surface oxide state of three different kinds of 316L powder. These grades were produced using vacuum induction melting inert gas atomization, conventional nitrogen gas atomization, and water atomization.
A detailed surface characterization was performed using high resolution SEM imaging coupled with XPS analysis. The results were further verified using TEM investigations of a lift-out from a WA powder particle. The results indicated a significant effect of the atomizing media on the surface oxide chemistry and can be summarized as follows: • The VIGA powder was covered by a homogeneous Fe 2 O 3 oxide layer with a thickness of about 4 nm. Submicron oxide particulates rich in Cr and Mn cations with traces of a Si cation were detected, covering a minor part of the surface (below 5%).
• The GA powder has similar surface oxide composition to the VIGA powder, with the iron oxide layer having a thickness of about 4 nm. Particulate oxide features were found to be rich in Cr, Mn, and Si cations as well. The fraction and size of the oxide particulates covering the powder surface is higher than in the case of VIGA powder, evaluated to be up to 5%.
• The WA powder surface is characterized by a higher fraction of particulates rich in Cr-and Si-rich oxides, with a thin Fe-oxide layer of about 3 nm covering a minor part of the surface. The surface coverage of particulate oxides was hence significantly higher than for the two gas atomized powder grades.
• The XPS C1s peak measured on powder surfaces was found to be shifting significantly relative to the commonly used value of 284.8 eV for adventitious carbon. This makes energy referencing using C1s signal for metal powder ambiguous when needed, and its use must be closely scrutinized.
F I G U R E 1 2 Interface between the oxide particulate and the metal of the powder particle, with two FFT insets showing the crystalline metallic matrix and the amorphous oxide layer • The TEM analysis of the lift-out confirmed the findings of the XPS and showed that oxide particulates are rich in Cr and Si. In addition, the nucleation of the fine oxides inside oxide particulates was shown.