Characterization of ultrafine particles emitted during laser-based additive manufacturing of metal parts

Atmospheric cabinet concentrations of elements emitted as ultrafine PM during additive laser processing are given in Table 3.

Primary particle formation and size-distribution

For all three instruments studied, the collected PM consisted of complex aggregates/agglomerates with fractal-like geometry (Fig. 1). No more than ten coarser particles with geometric projected diameters between 0.7 and 2 µm were observed on each filter. The elemental compositions were similar to the bulk material and no crystalline phases were identified. The presence of these may be due to sputtering from the melted alloy. No larger particles were seen. An equivalent projected area diameter of primary particles measured by TEM are shown in Fig. 2 and primary particle size-distribution summary statistics is presented in Table 4. The overwhelming number of particles formed in the three processes had equivalent projected area diameters within the 4–16 nm size range, with median sizes of 8.0, 9.4 and 11.2 nm for EOS M 270 dual mode, InssTek MX-Mini and LC-10 IPG-Photonics, respectively. The largest primary particles identified in the size-measurements had diameters of 50.4, 82.0 and 77.5 nm, correspondingly. Compared to previous research of laser ablation of metals where a maximum of the particle size distribution at 6–11 nm, dependent on laser intensity, were observed, the sizes of the primary particles in the laser additive processes studied in this work are similar³². It has previously been shown that the PM generated during manual metal arc, metal inert gas (MIG) and tungsten inert gas welding operations consists of agglomerates with primary particle diameters in the range of 5–40 nm with very few above 50 nm³³.

Numerical modelling of gas-phase processes

To understand particle growth and oxidation it is essential to locate their trajectories in the zones of heated laser spots. Although it is difficult to visualize tracks of nano-sized particles directly because of their size, as well as gas flow dynamics in vicinity of processed zone, it is, however, possible to perform a close-to-real-life numerical simulation of these gas-phase processes during laser surface treatment of the substrate. In Fig. 3 the presence of toroidal eddies surrounding a hot vertical jet of metal vapour is demonstrated. These vortices remain unchanged in vicinity of a laser spot during all the process of sintering and form a recirculation zone around the heat-affected region. A close look with streamlines plotted from a base of this hot up-stream (Fig. 3b) shows the recirculation zone more clearly. Nano-sized particles due to their extremely low mass (about 5 × 10^–16 ng) will exactly follow gas streamlines, finally trapping into that toroidal eddies. However, the particles do not stay in the recirculation zone permanently: they grow and drift to peripheral regions of the vortex, and finally leave it. According to estimations based on our numerical simulation, particle mean residence time in a vortex is about 0.5 ms for a particle of initial diameter of 10 nm and density of 7850 kg/m³.

Similar behavior of particles is expected for the DED machine. To verify that, gas flow and temperature dynamics for the InssTek MX Mini machine have also been simulated. Laser power of 200 W is focused in a Gaussian beam having a spot diameter of 1 mm, whereas surface absorbance and penetration depth are the same as in the modeling of the PBF-LB/M machine. Although complete information on inner design of a nozzle mounted in that machine is not available, we primarily oriented on its general view and typical conventional flow rates used in three-stream coaxial nozzles. Computational domain is initially filled with air, and all in-nozzle inlets consist of 99.9% pure argon. We consider even this “approximate” case is still usable to estimate flow pattern near the laser spot. Although the cladding head moves horizontally with speed of 1 cm/s, the flow field is relatively steady. Vaporization-induced puff above the treated surface of steel has a height of about 1 mm. Again, streamlines sampled close to the heat-affected zone represent the gas recirculation regions which are shown in black color on top of the vector velocity field, demonstrating the occurrence of a toroidal vortex caused by the hot vertical gas stream (Fig. 4). In contrary to the PBF-LB/M process, this gas jet cross-collides with a flow moving in opposite direction produced by the cooling gas. Nano-particles (NP) of partially condensed metal vapor should thus be trapped and turned back to the laser-affected zone again, but some of them will follow peripheral streamlines and slide along the treated surface. Estimated particle in-eddy residence time approximately equals 1.5 ms which is about 3 times more than in the PBF-LB/M process.

The boundary zone of the vortex is located close to the mixing zone of the surrounding air (see Ar mass fraction in its mixture with air in the right part of Fig. 4, where “Ar-air” interface is marked in green color) with a likely forced oxidation of airborne particles because of their possible interaction with oxygen (O) from air.

Various fabrication methods of NP based on vapor deposition have been developed. Laser ablation is a method where very high energy is focused to a solid material for evaporation of light-absorbing materials where the vapor phase is thermodynamically unstable. Under chemical supersaturation vapor-phase atoms/molecules will rapidly and uncontrolled be condensed with a coagulation rate proportional to the square of their number concentration. At high temperatures particles coalesce faster than they coagulate; at lower temperatures loose agglomerates with rather open structures are formed³⁵. In high temperature aerosol reactors NP (< 5 nm) coalesce almost takes place instantaneously even at temperatures considerably lower than the melting temperature of the bulk material due to e.g. the reduced melting temperature of NP³⁶. The primary particle growth is also dependent on the metal vapour concentration. It has been shown in a simulation of fume formation mechanism in arc welding that particles are mainly generated by FeO nucleation with small sizes formed at low Fe concentrations/low temperatures and larger primary particles at higher Fe concentrations which are not fully oxidized because of their lower surface to volume ratio³⁷.When aggregates/agglomerates are allowed to be exposed to high temperatures the whole or part of the particle may be restructured during sintering even until a fully coalesced sphere is formed³⁶. Similar particle formation mechanisms in PBF-LB/M processes are expected (Fig. 1b,c) where both restructured and fully coalesced spheres are present in particles consisting of both loosely and sintered primary particles. In our simulation of the gas-phase processes it is shown that when metal powder is rapidly heated by the laser, a mushroom-like cloud of hot metal vapour is formed just above the laser-processing zone of the metal surface. The evaporation rate in PBF-LB/M is more intensive than in DED because of smaller laser spot diameters (providing higher dissipated power density) used in PBF-LB/M. When released from the metal surface, the vapor rapidly expands into the surrounding atmosphere where fast-moving gas jets with surrounding toroidal eddies are formed with a typical speed of 930 m/s.

These vortices remain unchanged in the vicinity of the laser spot forming a recirculation zone around the heat-affected zone. Thus, the vapor will rapidly be transported to relatively cold regions with following condensation. In the case of DED, condensed low-mass particles follow the streamlines of the toroidal vortices located on the boundary of vapor cloud. This boundary surface is located remarkably close to the argon-air interface allowing further low-temperature oxidation of particles and their subsequent growth. Once particle diameter reaches its critical value (due to continuous cooling and oxidation), the particle may release from these vortices because of inertia forces and gas flow instabilities with abortion of further growth.

Evolution of processes in gas metal arc welding fume has been investigated by Vishnyakov et al.^38,39 by numerical modeling. According to their model it was shown that the primary particle chemical composition and the particle size distribution strongly depends on the vapor–gas mixture cooling rate. Such a dependency was explained in their model by the decrease of the vapor–gas mixture cooling rate when the shielding gas temperature was increased. Therefore, duration of particles growth via vapor condensation and coalescence is increased with subsequent increase of particle size. The number based primary particle size distributions presented by³⁸ compare well with our results of particle size-distributions during additive processing (Table 4 and Fig. 2). This may indicate that the mechanism of formation and growth of primary particles during AM is similar to the processes occurring during arc welding.

Analysis of gas flow in a cross-section of the domain shows that there are two primary gas jets affecting the flow pattern: carrier gas-powder stream and shielding gas protecting the optics. It looks like increased flow rates of shielding gas (simultaneously maintaining the carrier gas flow unchanged) will lead to efficient removal of airborne particles out the zone of processing preventing their remelting in a circulated toroidal vortex of the carrier gas. Higher flow rates will also possibly lead to decrease of particle oxidation due to their reduced residence time in the air-argon mixture on the boundary of nozzle-produced gas streams and ambient atmosphere. This is a typical scenario of gas-flow-particle interaction. However, to establish exact relation between gas flow rates and powder removal rate a series of additional numerical simulations under various operating conditions will be done in further.

Elemental composition of primary particles

At high magnification (Figs. 5, 6, 7) it is noticeably that the primary particles are sintered to each other leading to aggregates which are held together by chemical and/or sinter forces^40,41. The primary particles in the present size range are usually spheres with a core shell structure (Figs. 5, 6, 7). In these samples the 20 nm sized particles predominantly consist of the main alloying elements Fe, Cr and Ni in addition to Mn, Si and O which are all more or less homogeneously distributed in the particles as shown in the elemental maps in Figs. 5 and 6 for EOS M 270 and InssTekMx-Mini, respectively. For the 30–50 nm sized particles generated during operation of the LC-101PG-Photonics machine, there is indications of less O in the core, but an O enriched shell around the particles as illustrated in Fig. 7.

Since it was not possible to quantitatively measure the elemental composition of the individual primary particles due to their small size (< 20 nm) the major element content of the aggregates was determined by scanning the electron beam over an area of approximately 100 nm × 100 nm with roughly the same number of primary particles (Fig. 8, Table 5). In addition, the larger particles (30–50 nm) from LC-10 IPG-Photonics were analyzed. Aluminum, carbon, copper and tin which were detected in all samples were excluded, as they are artifacts from the TEM grids and the substrate, respectively. Since the EDS X-ray analysis system has a default spectra for the elements, the measurement differences between these spectra used for estimating the elemental composition and our TEM grid samples, limits the possibility to obtain quantitative data because appropriate ZAF correction factors could not be obtained. Especially for O, this is critical⁴². The amount of Fe and Cr in PM, (shown as Fe/Cr mass ratios in Table 5) quantified from SEM spectra of agglomerates and analysis of air filters is comparable to the content in stainless steel alloys used in the experiments. The ratios for Ni and Mn indicate fractionation of both elements with increased content in PM for the InssTek MX-Mini process. Especially for Mn, the fractionation is significant. For the other methods, valid elemental ratios were not obtained due to absence of usable quantitative information in the measurements.

By modeling Fe fume formation in arc welding³⁷ it has been shown that Fe atoms react in the vapour phase with O to form vaporous Fe monoxide (FeO) at 2500 K followed by nucleation of the FeO to liquid particles at temperatures around 2000–2500 K; remaining Fe atoms are condensing on the particle surface. Further on, the liquid particles are oxidized to Fe₃O₄ around 2000 K with solidification around 1800 K. Fe is further oxidized to Fe₂O₃ around 1500 K, but this is only for the smallest primary particles due to diffusion limitations. Thus, at the end of the formation process, the composition of the primary particles depends also on the initial Fe and O amounts. Larger primary particles are not fully oxidized since their surface to volume ratio is lower which prevent O diffusion into the core of the particles. The low O content relative to Fe present in large particles (approx. 50 nm) from LC-10 IPG shown in Fig. 7 may be explained by Sandibondi’s model³⁷.

Selected area electron diffraction (Fig. 1a,b,c) reveals that the PM consist of crystalline phases. Furthermore, the particles were investigated with EBSD in SEM which earlier has been used to characterize phase compositions of particles⁴¹. The EBSD patterns, shown in Fig. S2 and S3, are a Fe₃O₄ spinel phase for the EOS M 270 dual mode and α-Fe for the LC-10 IPG-Photonics. EBSD is a single particle analysis principle and it might very well be that a mixture of the two phases are present for all three techniques. Because of the small size of the primary particles, it is expected that the EDSD signals originate from the particle core. The shell surrounding the particles are clearly an oxide layer and could be Fe₃O₄. These results support also the formation process suggested by Sanibondi’s model³⁷.

Health aspects

Inhalation is the most relevant exposure route for occupational exposure to ultrafine PM. The lung deposition characteristics, the potential toxic effect induced and the kinetic fate and possible translocation to other organs are predominantly mostly determined by the agglomeration/aggregation status of the inhaled ultrafine PM⁴². Scheckman and McMurry⁴³ have shown experimentally using a silicone rubber lung cast model that silica agglomerates with primary particle diameter of 10 nm deposited more efficiently than sodium chloride (NaCl) particles and oleic acid (OA) spheres with equal mobility and aerodynamic sizes in the size ranges 30–300 nm. At larger primary particle sizes the deposition pattern for the agglomerates was closer to match that of NaCl and OA. This may indicate that the ultrafine PM characterized in the present work, if inhaled, would deposit more efficient than predicted using the International Commission on Radiological Protection and/or multiple-path particle dosimetry models^44,45.

If it is considered that deposited agglomerates/aggregates consist of about 10 nm sized primary particles as characterized in the present study, that are more or less loosely bound to each other, the crucial question for assessing possible biological effects of these particles upon inhalation is, if the agglomerates remain as agglomerates or if the agglomerates break down to smaller aggregates or even primary particles in contact with the lung surface—which may significantly influence their toxicological properties⁴⁶.

Buckley et al.⁴⁷ exposed rats nose-only to aerosols of radioactive iridium-192 particles with sizes ranging from 10 to 75 nm demonstrating a slow lung clearance and increasing concentrations of particles with decreasing particle size in secondary target organs as liver and kidney with a translocation efficiency of max 0.5% of the lung burden. If this low dose build-up of particles in other organs than lung could lead to any systemic effects is still unclear in their opinion.

Animal experimental studies, in vitro and in vivo, have demonstrated the tendency of nano-sized particles to form larger size agglomerates following deposition and an increase of particle number due to disintegration of agglomerates seems not to be of high relevance⁴².

Therefore, particle sizes measured airborne in the respiratory zone of individuals seem to be a reasonable estimate of the size related properties of particles in the lungs.

To our best knowledge the primary spherical particles emitted in this work with a core composition of mainly Fe, Cr, Ni and Mn and an oxidic coated surface have not been toxicology tested. However, comparable primary particles in the same size range and chemical composition are generated during solid stainless steel wire welding⁴⁸. When such particles have been investigated in lung cells and reporter cell lines, they showed no toxic effects to the reporter cells, no cytotoxicity, genotoxicity and no generation of reactive oxygen species⁴⁹.

Source: Ecology - nature.com