Welding fume nanoparticles from solid and flux-cored wires: Welding fume nanoparticles from solid and flux-cored wires: Solubility, toxicity, and role of fluorides Solubility, toxicity, and role of fluorides

Welding fume particles are hazardous. Their toxicity likely depends on their composition and reactivity. This study aimed at exploring the role of sodium or other fluorides (NaF), which are intentionally added to flux-cored wire electrodes for stainless steel welding, on the solubility (in phosphate buffered saline) and toxicity of the generated welding fume particles. A multi-analytical particle characterization approach along with in-vitro cell assays was undertaken. The release of Cr(VI) and Mn from the particles was tested as a function of fluoride solution concentration. The welding fume particles containing NaF released significantly higher amounts of Cr (VI) compared with solid wire reference fumes, which was associated with increased cytotoxicity and geno- toxicity in-vitro. No crystalline Na or potassium (K) containing chromates were observed. Cr(VI) was incorporated in an amorphous mixed oxide. Solution-added fluorides did not increase the solubility of Cr(VI), but contributed to a reduced Mn release from both solid and flux-cored wire fume particles and the reduction of Cr (VI) release from solid wire fume particles. Chemical speciation modeling suggested that metal fluoride com- plexes were not formed. The presence of NaF in the welding electrodes did not have any direct, but possibly an indirect, role in the Cr(VI) solubility of welding fumes.


Introduction
Welders without sufficient protection equipment are exposed to potentially hazardous welding fumes. Despite significant efforts to improve occupational health safety, exposure to welding fume remains a serious problem even in developed countries with a long history to improve work environment conditions. For example, a recent report commissioned by the Swedish Work Environment Authority (Andersson et al., 2019) estimates 71 death cases annually (based on 2016 data) in Sweden that could be directly related to welding fume (considering ischemic heart disease and lung cancer as death causes). This is more than the total annual death cases related to work accidents in all occupations in Sweden. In the UK, 152 deaths (lung cancer only) are estimated annually due to occupational exposure to welding fume (Rushton et al., 2012).
Exposure to welding fume is associated with an increased risk of developing lung cancer (IARC, 2018;Honaryar et al., 2019), cardiovascular diseases (Ellingsen et al., 2019;Ibfelt et al., 2010), and a number of lung and airway diseases (Antonini et al., 2003;Coggon and Palmer, 2016;Zeidler-Erdely et al., 2012), partially due to suppression of the immune system (Zeidler-Erdely et al., 2012;Knobloch et al., 2020). Welding fume, without any distinction of the material or welding method, has been classified as Group 1 carcinogenic to humans by the International Agency for Research on Cancer (IARC) (IARC, 2018). It has been shown that there is an increased risk independent of type of welding fume (IARC, 2018;Honaryar et al., 2019). Exposure to small particles, including welding fume, is also in general associated with acute myocardial infarction and other ischemic heart disease (Wiebert et al., 2012;Mocevic et al., 2015;Sjögren et al., 2002). Further, air pollution, including particular matter which can reach the brain, has been linked to central nervous system diseases such as Alzheimer's disease, Parkinson's disease, and stroke (Block and Calderón-Garcidueñas, 2009). Welding fume consists of both potentially hazardous gases and particles, which often are nano-sized (Floros, 2018).
The authors have shown considerable variations in in-vitro toxicity (cytotoxicity, generation of reactive oxygen species, and genotoxicity) that correlate with the released amount of hexavalent chromium [Cr (VI)] and manganese (Mn) from stainless steel welding fume particles at physiological pH (McCarrick et al., 2019). These findings are in line with epidemiological evidence showing an increased lung cancer risk if exposed to high (≥ 1.8 mg/m 3 ) amounts of welding fume and high (≥ 1.4 µg/m 3 ) amounts of Cr(VI) in the work environment (Kendzia et al., 2020). Mn in, or released from, welding fume can cause neurotoxic effects (Taube, 2013). From an exposure point of view, national occupational exposure limits are often exceeded for Mn (Hanley et al., 2017;Jeong et al., 2016;Hedmer et al., 2014). In-vitro and in-vivo studies have shown that welding fumes from stainless steels containing nickel (Ni) and chromium (Cr), which under certain circumstances contains Cr(VI), are significantly more toxic compared to fumes generated from welding mild steel (Antonini et al., 2005(Antonini et al., , 1999Leonard et al., 2010;Olgun et al., 2020;Shoeb et al., 2017bShoeb et al., , 2017a. However, long-term effects (lung tumor promotions) were in an in-vivo study more strongly associated with less soluble metal compounds in model metal oxide particles, in particular iron (Fe) oxides at physiological pH (Falcone et al., 2018).
Gas metal arc welding (GMAW) and flux-cored arc welding (FCAW) are, due to high-productivity, among the most common welding methods (Floros, 2018;Praveen et al., 2005). These processes result in higher amounts of airborne ultrafine particles as compared to other welding methods, such as tungsten-inert gas welding (Gomes et al., 2012b). In GMAW and FCAW, an electric arc is formed between a consumable wire (electrode or filler material) and the base metal such as stainless steel under the protection of a shielding gas. Solid wire and flux-cored wire (FCW) electrodes are the electrodes for GMAW and FCAW. FCW forms a protective slag on the weld. FCW electrodes consist of a sheath metal filled with a flux powder. The flux contains different salts (often fluorides), metals, and minerals (metal oxides) of widely varying compositions for different applications (Floros, 2018;Guo et al., 2019;Westin et al., 2016). The amount of fluoride salts in FCW depends on the type of the FCW electrode, and typical fluoride salts are sodium fluoride (NaF), calcium difluoride (CaF 2 ), or perovskite fluoride KCaF 3 (Floros, 2018). Although the amount and composition of the welding fume particles depend on several welding parameters such as melting rate, potential, current, arc mode (e.g. spray arc), filler material feed rate, and shielding gas, the main factor that influences the solubility and composition of the welding fume particles is the composition of the filler material (electrode) (Floros, 2018;McCarrick et al., 2019;Mei et al., 2018). The flux consists of at least a dozen different raw materials and works in symbiosis with the strip composition, thickness, and various coatings for feedability. All of these parameters have individual, antagonistic, or synergistic influences on arc stability, fluidity and viscosity, slag formation, element loss, and weldability.
Since the release of Cr(VI) and Mn from welding fume particles are important processes that govern acute toxic effects, inflammatory markers, and genotoxicity, and since their solubility varies largely as a function of welding electrode and welding parameters (McCarrick et al., 2019;Antonini et al., 2004), it is crucial to understand the underlying chemical and physical reasons for Cr(VI) and Mn solubility in welding fume under physiologically relevant conditions. An evident correlation has been observed between fluoride salts [often containing sodium (Na) or potassium (K)] and soluble Cr(VI) in welding fume of stainless steel welded with certain FCW electrodes (Mei et al., 2018;Tandon et al., 1985;Floros, 2016). It is however unclear, if Na or K contribute to the formation of easily soluble chromate compounds (Floros, 2018;Tandon et al., 1985), whether fluoride ions could play a role, and/or whether other factors need to be considered. Knowledge of the underlying physicochemical mechanism of the formation and stabilization of soluble Cr(VI) in these welding fume particles would enable reduction of these hazards by means of appropriate engineering and chemical measures.
The aim of this study was therefore to explore the roles of NaF and welding fume composition of welding fume generated by welding stainless steel with a flux-cored wire electrode on the release and soluble fraction of Cr, Cr(VI), Mn, Ni, and Fe, cytotoxicity, and genotoxicity.

Materials and welding method
The base material was 5 mm austenitic 1.4307/AISI 304 L (denoted 304 L in the following) and 6 mm lean duplex 1.4162/UNS S32101 (denoted 2101). The composition of the base metals is given in Table A1 (supplementary material). Welding was performed using either solid wire electrodes (for samples S1 and S2) or FCW electrodes (for samples F1 and F2). The FCW electrode for F1 was optimized for a fast-freezing slag and the FCW electrode for F2 for a slow-freezing slag (for welding in flat position). The chemical composition of the filler materials used and the classification in accordance with the American Welding Society (AWS) standards AWS A5.9 and A5.22 are given in Table 1. Please note that the wire manufacturers do not provide the formulation of the flux in the FCW products as the combination of oxides and fluoride salts affect the weldability and formation of slag rather than the chemical composition. All wires had a 1.2 mm diameter and were delivered on 15 kg wire baskets.
Welding was performed bead-on-plate by experienced welding operators using the spray arc mode (high current resulting in high melting rates). The welding parameters are presented in Table 2. For FCW, argon with 18% CO 2 , is the optimum shielding gas for arc stability, sidewall fusion, and mechanical properties. The slag cover protects the melt from oxidation and only a minor increase in carbon uptake takes place during the welding process. When welding the gas shielded solid wires, however, it is necessary to avoid significant uptake of carbon and oxygen in the weld metal. Argon with 2% CO 2 offers good arc stability, while the corrosion resistance and mechanical properties are retained.
Collected fume particles were brushed from the cellulose filters, which were used for collection, directly after welding and delivered for analysis in sealed glass vessels. The vessels were kept in a desiccator (< 10% relative humidity) at room temperature prior to and in between the different measurements.

Electron microscopy
Particle morphology and elemental composition were investigated by means of transmission electron microscopy (TEM) with a JEM-2100F instrument equipped with energy dispersive X-ray spectroscopy (EDS, EDAX Genesis XM 460 system). The operating voltage was 80-200 kV and the lateral resolution was ≥ 0.14 nm. Images were captured in both secondary electron (SE) and back scatter electron (BSE) mode, as well as in STEM mode using the Gatan Ultrascan 100 16 MP camera. The preparation of the fume samples for TEM analysis included the preparation of a suspension of particles in ethanol, ultrasonic treatment for 5 min, and centrifugation for 5 min at 5000 rpm. A drop of the supernatant was then pipetted on the TEM copper grid (covered with carbon) and dried at ambient conditions. Image processing and particle size distribution calculations were performed by means of the ImageJ software for 247-682 independent particles for each sample.
Morphological information was obtained by using a Hitachi SU8230 Regulus Ultra High-Resolution Field Emission SEM (FE-SEM) equipped with EDS. The images were captured in both SE mode and BSE mode with a setting of 15 kV. EDS was performed as area analysis on nanoparticle aggregates and additionally as spot analysis on micrometersized spherical particles and nanoparticle aggregates in their vicinity. Image processing and particle size distribution calculations were also performed for 162-270 independent particles for each sample based on the SEM images.

X-ray diffraction (XRD)
The phase composition and crystal structure of the fume samples were characterized by XRD using a Bruker D8 ADVANCE diffractometer with a rotating cobalt radiation. The reference data was used from the PDF2 database. Rietveld refinements were conducted with the software HighScore Plus. The pseudo-Voigt function was used for the peak profile refinement. The average crystal size (d, in nm) in the welding fume particles was calculated by the Scherrer formula d = Kλ/βcosθ, where the dimensionless particle shape factor (Scherrer's constant) K = 0.94 for cubic crystals, the wavelength of X-ray radiation (of the Co source) λ = 0.1788965 nm (cobalt line), β is the width at half the maximum peak intensity, and θ is the angle of diffraction (Bragg angle), here chosen for the highest intensity peak (Alexander and Klug, 1950).

X-ray photoelectron spectroscopy (XPS)
The outermost surface (5-10 nm) composition was determined by means of XPS using an UltraDLD spectrometer, Kratos Analytical Manchester, UK, with a monochromatic Al Kα X-ray source (150 W). The fume particles were fixed on adhesive copper tape. Wide spectra and detailed spectra (pass energy of 20 eV) for Fe 2p, Cr 2p, Ni 2p, Mn 2p, Si 2p, K 2p, Bi 4f, Na 1s, F 1s, O 1s, and C 1s (as energy reference at 285.0 eV) were run for two different locations for each sample. Quantitative information was obtained by using linear background corrections and peak fitting based on the Gaussian-Lorentzian function.

Fourier-transform infrared spectroscopy (FTIR)
An infrared (IR) Fourier spectrometer NEXUS E.S.P. (Thermo Scientific, USA) was used to acquire 32 spectra with 4 cm − 1 resolution in the transmission mode of the welding particles mixed with potassium bromide (KBr) pressed into a pellet (2 wt% welding fume particles, 98 wt % KBr). The background was measured prior to each measurement and subtracted.

Cyclic voltammetry (CV)
The welding fume samples were immobilized on a paraffinimpregnated graphite electrode (PIGE) as working electrode for CV measurements in 8 M NaOH (pH 13) with a platinum wire as counter electrode, and an Ag/AgCl (sat. KCl) as reference electrode. The potential was swept cathodically from the open-circuit potential to − 1.4 V vs. Ag/AgCl sat. KCl and anodically to 0.2 V, or vice versa, in order to investigate solid-solid transformation peaks corresponding to reduction or oxidation of electroactive species in the surface oxide of the welding fume particles. Detailed methodological information is given in previously published studies (Mei et al., 2018;Cha et al., 2016).

Digestion for total metal content analysis
5 ± 0.5 mg of the welding fume samples was weighed using a microbalance with 2 μg readability (XP26 DeltaRange®, METTLER TOLEDO, Sweden), documented for each sample. 20 mL of diluted aqua regia (2.5 vol% HCl and 6.5 vol% HNO 3 ) was added before sonicated in an ultrasonic bath for 2 h. This resulted in a final solution temperature of 60-70 • C. The digested sample was then kept in the diluted aqua regia for at least 24 h prior to solution analysis (Section 2.10). Triplicate samples and one background sample (blank), without welding fume, were prepared for each welding fume sample. Ultrapure water (18.2 MΩ cm, Millipore, Sweden) was used as solvent for all solutions.

Release testing in phosphate buffered saline (PBS) with and without addition of NaF
5 ± 0.5 mg of the welding fume samples was weighed into acidcleaned dry Nalgene® polypropylene vessels, with the exact weight recorded. 50 mL of PBS (8.77 g/L NaCl, 1.28 g/L Na 2 HPO 4 , 1.36 g/L KH 2 PO 4 , adjusted with 50% NaOH to pH 7.2-7.4) were then added and the closed vessels were exposed for 24 h at 37 • C in darkness in an incubator (Edmund Bühler GmbH TH30, Bodelshausen, Germany) with bi-linear agitation (22 cycles/min, 12 • inclination). After exposure, the pH of each sample (triplicate samples and one blank sample) was measured and the solutions filtered (GE Healthcare Whatman, 25 mm diameter, 0.02 µm inorganic syringe membrane filter anotop™) and frozen prior to further analysis (Section 2.10).
The same procedure was followed for different added amounts of NaF (analytical grade) during exposure in PBS. 0, 1, 5, 10, and 50 mg F/L (from NaF) in PBS was used to investigate the effect of fluorides on the Table 1 Chemical composition of the individual heat number of the iron-based filler metals in wt%. The solid wires S1 and S2 show the wire analysis and the flux-cored wires F1 and F2 the analysis of the all-weld metal with Ar + 18% CO 2 as shielding gas.  release of metals from the welding fume samples S2 and F2. The effect of fluoride adsorption on the release of Cr(VI) was also attempted to be investigated by a double-step exposure. The welding fume sample S1 was exposed to PBS containing fluorides (50 mg F/L) in a centrifuge tube followed by centrifugation at 3000 rcf. After 10 min the solution was decanted, and fluoride-free PBS was added.

Cr(VI) extraction in carbonate solution
Extraction at alkaline conditions according to the ISO standards (ISO, 2017a(ISO, , 2005 was also conducted in order to compare the Cr(VI) solubility tests in PBS with industrial methods. 10 mg of the welding fume sample was extracted for Cr(VI) by adding 2.5 mL of a leaching solution (20 g/L NaOH and 30 g/L Na 2 CO 3 ) and 5 mL of ultrapure water, resulting in a final pH of 12.6. This solution was then heated near the boiling point (until visible bubbles) for 30 min in an acid-cleaned glass vessel covered with a microscope cover glass. After that procedure, the solution was filtered using a paper filter (VWR, qualitative filter paper, 413, particle retention 5-13 µm, European Cat. No 516-0816), and filled up to 25 mL with ultrapure water prior to any further dilution with ultrapure water and analysis for Cr(VI) by means of spectrophotometry (Section 2.10).

Analysis of metals in solution and Cr(VI) in solution
Flame atomic absorption spectroscopy (AAS) was employed to analyze total concentrations of Cr, Ni, Fe, and Mn in the solution samples (AAnalyst 800, Perkin Elmer, Waltham, MA). Prior to analysis, the unfrozen solution samples were acidified with ultrapure HNO 3 to a pH < 2. Ultrapure water and 3-4 standard solution samples prepared in 1% HNO 3 were used for calibration. One of these standard solution samples was checked regularly (every 5th sample) to detect deviations from the calibration curve and, if necessary, recalibrate. The detection limits, as estimated from three times the average standard deviations of the blank samples, were 0.02 mg Cr/L, 0.05 mg Fe/L, 0.04 mg Mn/L, and 0.03 mg Ni/L. 960 µL of the unfrozen PBS solution samples were mixed with 20 µL of diphenylcarbazide (DPC) solution and 20 µL phosphoric acid (70%). The DPC solution, composed of 1.0 g 1,5-diphenylcarbazide in 100 mL acetone, was acidified with one drop of glacial ultrapure acetic acid. The concentration of Cr(VI) was determined at 540 nm using a UV-vis spectrophotometer (Jenway 6300) following the procedure described in ISO (2017b). The calibration was based on PBS blank samples with additions of 0, 125, 247.5, 495 and 990 µg/L Cr(VI). The calibration curves were linear (R 2 ≥ 0.9997). The limit of detection, as based on three times the highest standard deviation of the blank samples, was ≤ 60 µg/L.

Preparation of fume (nano)particles for cell exposure
The welding fume particles were weighed and suspended in ultrapure water in a glass tube at a concentration of 1 mg/mL, followed by sonication in water bath for 2 × 10 min with vortexing prior, in between, and after. Immediately prior to exposure, the particle suspensions were further diluted to desired concentrations in HBEC cell medium. Fresh suspensions were prepared for every experiment.

Cytotoxicity (Alamar blue assay)
HBEC cells were seeded at 1 × 10 4 cells/well in transparent 96 well plates following exposure to fume particles in HBEC cell medium for 24 h in triplicates. Alamar blue assay was performed as previously described by the authors (Di Bucchianico et al., 2017). Briefly, following exposure, the medium was removed and replaced with 10% Alamar blue reagent in fresh medium and the plate was further incubated for 2 h. The fluorescence was measured using a Tecan Infinite F200 plate reader (Tecan Trading 125 AG, Mannedorf, Switzerland) at 485 nm excitation and 590 nm emission. The average fluorescence intensity for the triplicate in the negative control was set to 100% viability and the mean fluorescence values of the exposed wells were further normalized to this value and calculated as % viability. The results were then expressed as % cytotoxicity calculated by 100 minus % viability. A control experiment with up to 20 µg F/mL (from NaF) without welding fume particles was conducted in parallel. 20 µg F/mL corresponds to the approximate F content (20 wt%) observed in 100 µg/mL welding particles (the highest particle dose tested).

Genotoxicity (comet assay)
Prior to exposure of the welding fume particles, HBEC cells were seeded at 5 × 10 4 cells in 24 well plates for 3 h. The alkaline version of the comet assay was performed as previously described (Gliga et al., 2014). Subsequent to electrophoresis, the slides were neutralized, dried and stained with 1:10 000 dilution of SYBR Green in TAE buffer for 15 min and scored using a fluorescence microscope (Leica DMLB, Germany) with Comet assay IV software (Perceptive Instruments, Suffolk, UK). 50 cells were scored in duplicate for each sample.

Joint expert speciation system
The joint expert speciation system (May, 2015) was used to model the chemical speciation of Cr, Mn, Ni, and Fe in PBS and of Cr and Mn in PBS in the presence of 1-50 mg F/L. 25 • C was set as the temperature since the database is most reliable at that temperature. Other input values were pH 7.3, atmospheric pressure, and the calculated ionic strength of the system (180 mM). Solids were allowed to precipitate (which was investigated). The input values for the metals were 19. The redox potential was investigated for two conditions: pe 5 and pe 10, which correspond to normal and oxidative conditions in water at pH 7.3. Effects of fluoride concentrations were investigated at 0, 1, 5, 10, and 50 mg F/L.

Statistical methods
For in-vitro toxicity assays, one-way analysis of variance followed by Dunnett's post hoc test was used to test for significance between exposures and control (P < 0.05) using GraphPad Prism 5.02 statistical software (GraphPad Inc., La Jolla, CA). Results are expressed as mean values ± standard deviation (SD) of at least three independent experiments (n = 3). For other differences, a student's t-test of unpaired data with unequal variance (KaleidaGraph version 4.0, Synergy software, Reading, PA) was used, also with P < 0.05 counted as a statistically significant difference. Further details are reported elsewhere (McCarrick et al., 2019).

Results and discussion
3.1. Welding fume particle characterization

Electron microscopy and XRD
The high-resolution SEM images (Fig. 1) and TEM images (Fig. A1, supplementary material) revealed a primary particle size distribution in the nanometer range for all samples, ranging from 3 to 180 nm, with mean diameters of 11-47 nm (Table A2 and Fig. A1). All particle size histograms were unimodal, with a tail towards larger sizes (Fig. 1).
Overview SEM images shown in Fig. A2 (supplementary material) revealed the majority of the welding fume particles to be aggregated nanoparticles. This is in agreement with literature findings (Floros, 2018;McCarrick et al., 2019;Gomes et al., 2012bGomes et al., , 2012aMei et al., 2018) and is a result of solidification (and oxidation) processes of the metal vapor (Floros, 2018). In agreement with previous studies (McCarrick et al., 2019;Mei et al., 2018), spherical micrometer-sized particles were also present showing visible dendrites originating from the solidification of the molten phase (Floros, 2018). Examples of these particles are shown in Fig. 2. These particles seemed to be mainly metallic, while the nanoparticles had a lower atomic number on average, probably due to oxides, Fig. A3 (supplementary material). The composition of the micrometer-sized spherical particles was similar to what was expected from the base metal (Fig. A4), and surrounding nanoparticles were enriched in oxygen (Fig. A4). The nanoparticle aggregates for samples F1-F2 were significantly enriched in Na, K, and F, and depleted in the main alloying elements. This was most pronounced for sample F2, as judged from the EDS results shown in Table A3. Na and K may originate from fluorides that were intentionally added to the FCW for controlled fluidity, viscosity, and arc stabilization during the welding process. Another origin may be water glass, such as sodium and potassium silicates, that together with water act as binder during flux agglomeration. This water is evaporated by controlled drying in a furnace. Detected titanium possibly originated from rutile TiO 2 used for slag formation in flux-cored wires.
The d-spacing values of 0.489, 0.255, 0.241, 0.232 and 0.148 nm shown in Fig. 3 for the S1, S2, F1, and F2 samples, respectively, correspond to the peaks at 21.10, 41.26, 43.5, 45.43 and 73.94 2-theta in the XRD pattern (Fig. 4). The main diffraction peaks were assigned to a magnetite type phase of X 3 O 4 with X = Fe, Cr, Mn, and/or Ni for the solid wire fume particles, and to the same phase and NaF in the case of the FCW fume particles (Fig. 4 and Table A4). This is in agreement with literature observations (Floros, 2018;Sowards et al., 2008Sowards et al., , 2010Moroni and Viti, 2009). The content of NaF was significantly larger in sample F2 as compared to F1 (Table A4).
Both amorphous and crystalline phases were observed by high resolution TEM.

Fig. 5.
Composition (outermost 5-10 nm surface) of the welding fume samples (S1, S2, F1, and F2) measured by means of XPS. Ti and Ca were additional elements of minor presence (information from the wide spectra) for the F1 and F2 samples.
showed a higher NaF content compared to sample F1. It should be emphasized that the information depth of 5-10 nm in some cases may reflect the complete nanoparticle rather than its surface oxide only.
The broad and common IR band at 3440 cm − 1 corresponds to the stretching vibrations of adsorbed H 2 O molecules, while that at 1630 cm − 1 is most probably attributed to O-H bending vibrations (Aouissi et al., 2010). Residual water in flux-cored arc welds may originate from water and water glass in the agglomeration of the flux, but also from the cleaning of the wire surface. The bands in the region between 800 and 350 cm − 1 can be assigned to different metal-O lattice vibrations (Cr (Zecchina et al., 1971), Mn (Kang et al., 2007;Shen and Clearfield, 1986;Feng et al., 1992), Fe (Ellid et al., 2003), and Bi (Dimitrov et al., 1994)). Due to several possibilities and a mixed oxide, it is difficult to assign them to specific metals. Some single-phase oxides can be excluded: O-Fe (from Fe 3 O 4 ) and O-Mn groups are assigned to bands at 582 and 626 cm − 1 , respectively (Zhou et al., 2018;Zhang et al., 2018), but they are almost absent in Fig. 6, which could reflect that they may exist as different ferrites of Mn and/or other elements. Likewise, no characteristic peaks of hematite (Fe 2 O 3 ) (Ellid et al., 2003) or α-Cr 2 O 3 (Zecchina et al., 1971) were observed. However, when taking into account possible shifts of peak positions due to the presence of different metal oxides, the high intensity peak at 610 cm − 1 may correspond to the most intense peak of Cr 2 O 3 (NIST, 2020) or hematite.
For sample F2, the band at 375 cm − 1 and shoulder at 450 cm − 1 could also indicate Bi-F bonds (Kavun et al., 2000). Bi probably originates from small amounts of bismuth oxide Bi 2 O 3 , added to the flux for improved slag detachability (Westin et al., 2016).
The bands at 940 and 890 cm − 1 , only visible for F1-F2, could be assigned to Cr-O-Cr and Fe-O-H bending vibrations (Yao et al., 2003;Legrand et al., 2001), respectively. Absorption bands in the region 1100-950 cm − 1 may be related to groups of CrO 4 x− (x = 2, 3, or 4) (Yao et al., 2003). These bands originate from Cr-O-Cr vibrations in which the valency of the Cr ion can vary from 4 to 6 (Ratnasamy and Leonard, 1972;Fahim et al., 1981). The absorption band at 1020 cm − 1 could indicate the organization of CrO 3 and/or Cr 2 O 3 in a chromate-like structure (Fahim et al., 1981). However, more probably, this broad band (950-1250 cm − 1 ) present for all samples is mostly related to silicon oxides and its compounds (Golden et al., 1986;Deshmukh and Aydil, 1995). NaF is characterized by peaks at 536 cm − 1 (Ritchie and Lew, 1964), 836, 1211, 1386, and 1420 cm − 1 (Gubareva, 2011). These bands were largely missing or hidden in Fig. 6. The bands at 1380 cm − 1 (all samples, but broader for sample F1) and the strong and broad band at 1440 cm − 1 (sample F2) could not unambiguously be assigned but is most probably related to bonds with carbon or fluorine. The strong and broad band observed at 1440 cm − 1 for sample F2 derives most probably from carbonate (Bruckman and Wriessnig, 2013).

CV
In agreement with previous studies by some of the authors (Mei et al., 2018), all electroactive constituents of the surface-available phases of the welding fume particles consisted of a mixed oxide phase (Fig. 7a). Peaks that could be assigned for single-phase MnO 2 (Cha et al., 2016;Chouaib et al., 1981;Linhardt, 1998) and mono-or dichromates (Mei et al., 2018) were missing (Fig. 7a). These findings agree with the XRD measurements that did not show any evidence of either crystalline chromate species or MnO 2 . This mixed oxide phase consisted of Fe (Cha et al., 2016;Baek et al., 2001;Rico et al., 2009) and Mn (Chouaib et al., 1981;Linhardt, 1998) oxides in all cases, and in addition Bi (Pourbaix, 1984) oxides in the cases of the two FCW samples (F1 and F2) (Fig. 7b). Note that silicon (Si), titanium (Ti), and NaF species did not show any solid-solid transition peaks under the experimental conditions of the cyclic voltammetry measurements and could hence not be detected.
Missing oxidation peaks during an initial oxidation, observed for all four welding fume samples, Fig. A6, indicate that the surface oxide(s) cannot be further oxidized and are hence in their most oxidized forms. These observations are also in agreement with previous findings (Mei et al., 2018).

Total metal content and PBS-soluble metals in welding fume particles
The relative content of the main alloying elements (Fe, Cr, Ni, and Mn) of stainless steel in the welding fume was highest for the solid wire welding fume samples with 35-46 wt%, and significantly lower in the FCW welding fume samples with 14-26 wt% ( Fig. 8 and Table A5). This is in agreement with previously reported studies on a larger number of welding fume samples (McCarrick et al., 2019). Despite a lower metal content in the FCW welding fume samples, the solubility of Cr(VI) expressed per mass of welding fume was significantly higher than that of the solid wire samples in PBS ( Fig. 8 and Tables A6-A7). 37-58 wt% of the total Cr in the welding fume was solubilized in PBS for the FCW welding fume samples, while only 2.3-3.0 wt% of total Cr was released and stable as Cr(VI) in PBS for the solid wire welding fume samples (Table A7). These observations are in agreement with previous reports (McCarrick et al., 2019;Mei et al., 2018). Fig. 6. FTIR spectra (off-set for clarity) for the different welding fume samples (S1, S2, F1, and F2) in the range of 350-4000 cm − 1 . Y.S. Hedberg et al. It has previously been discussed whether PBS is a relevant simulant to test Cr(VI) solubility of welding fume particles (Mei et al., 2018). A bicarbonate/sodium hydroxide mixture has commonly been used as leachate medium following the ISO standards 15011 and 16740 (Cena et al., 2014(Cena et al., , 2015Dennis et al., 1996Dennis et al., , 2002aDennis et al., , 2002bDennis et al., , 1997. However, that alkaline procedure has been criticized for its ability to oxidize Cr to Cr(VI) (Pedersen et al., 1987). PBS is a simulant without antioxidants that mimics the ionic strength of human blood and has a relevant physiological pH. PBS has been chosen for its unlikelihood to oxidize Cr to Cr(VI) or to reduce Cr(VI) to Cr(III) (Mei et al., 2018;Pastore et al., 2004). In order to compare different studies, the ISO 16740 procedure was followed for the four welding fume samples and compared with the PBS procedure. The trends among the different welding fumes were equal for the different procedures (Fig. 9). Consistently less Cr(VI), compared to the PBS procedure, was released during the ISO 16740 procedure, which aims at extracting all Cr(VI) (Fig. 9). This could be explained by a longer extraction time (24 h vs. 30 min) and a lower particle loading (0.1 g/L vs. 1.3 g/L) for the PBS procedure compared with the ISO 16740 procedure. However, in all, results from the two procedures were relatively comparable, which facilitates correlations and comparisons with literature findings.

Effect of fluorides on Cr(VI) and Mn release from welding fumes in PBS
Since the welding fume samples S2 and F2 showed the highest Cr(VI) release (of total Cr) in PBS within their respective groups of solid wire and FCW fumes, Table A7, these two welding fume samples were selected to investigate the effect of added fluoride ions (from NaF) on their release of Cr(VI) and Mn into PBS. These lean duplex wires also Fig. 7. Representative cyclic voltammograms (reduction from open circuit potential, OCP, followed by oxidation) at pH 13 of the welding fume particles (S1, S2, F1, and F2). Magnified voltammograms in (b) for a clearer view on the oxidation peaks. The arrows mark the start potential (OCP), which was about − 0.2 V in all cases, and the reduction (a) and oxidation (b) peaks. contained about 3.5% and 5% more Cr than their respective austenitic counterpart. The release of these two species/elements was investigated, since they (i) showed the highest release in PBS and (ii) were earlier reported to link to different acute in-vitro toxic endpoints (McCarrick et al., 2019). The release of Cr(VI) of the sample F2 was not affected by varying fluoride concentration in PBS (Fig. 10). However, the soluble and released fraction of Cr(VI) of the sample S2 decreased with increasing fluoride concentration (Fig. 10). The amount of released (aqueous) Mn decreased furthermore for both particle fumes with increasing fluoride solution concentration (Fig. 10).
The two-step procedure, during which first fluorides (50 mg F/L) were adsorbed on the sample S1 followed by re-dispersion in PBS, resulted in non-detectable Cr(VI) levels (data not shown). Table 3 presents predictions of the chemical speciation of Cr, Mn, Fe, and Ni in PBS at different redox potentials (simulating any co-released species of oxidative nature) as determined by chemical speciation modeling (JESS). Mn and Fe were not predicted to be stable in solution at any conditions, while Ni was predicted soluble and Cr soluble as Cr (VI) at higher redox potential. A higher redox potential can be achieved by dissolved oxygen or co-released oxidative species, such as permanganate.
The effect of the presence of fluorides on the chemical speciation of Cr and Mn was also investigated by chemical speciation modeling. Data was available for solid Cr and Mn-fluoride compounds in the database. No effect was observed. The predicted chemical speciation of Cr and Mn in the presence of 1-50 mg F/L in PBS was equal to corresponding predictions in PBS.
Fluoride-metal complexes could have been formed and resulted in precipitation at higher fluoride concentrations. It is in this context important to note that 50 mg F/L corresponds to similar or a higher loading of the sum of Mn, Cr, Ni, and Fe mass in the given experimental conditions (0.1 g/L welding fume with 15-35 wt% metal content)hence, an excess of fluorides. If metal-fluoride complexes would form, precipitation could be expected. Alternatively, a protective surface layer of metal-fluoride complexes could be expected to hinder further dissolution of Mn and Cr(VI) from the welding fume particles. Such a case would be supported by the experimental observation of a reduced Cr(VI) and Mn solubility in some cases with increasing fluoride content in PBS (Fig. 10), and speciation database and literature observations of the formation of stable metal-fluorides (Fackler and Chawla, 1964;Parkinson and Williams, 1950;Wanklyn, 1969;Azad and Sreedharan, 1987). However, these complexes are according to the speciation prediction not the predominant species formed in PBS and under the experimental conditions. An alternative interpretation would be that the fluoride ions change the solution redox potential. This would not be possible directly (fluorides are not electroactive under the experimental conditions), but indirectly, e.g. by influencing the access to trivalent iron. Fluorides, especially in the presence of chlorides, can damage metal oxides and increase the dissolution of iron oxides (Mirjalili et al., 2013;Li et al., 2007;Pahlavan et al., 2016), which can have a reductive effect (Shettlemore and Bundy, 2001) and hence reduce the release of chromates and permanganates.

In-vitro cytotoxicity and genotoxicity
In agreement with previous observations (McCarrick et al., 2019), the FCW samples F1 and F2, which released the highest amounts and percentages of Cr(VI), were significantly more cytotoxic and genotoxic as compared to controls and to the solid wire samples S1-S2 (Fig. 11). Cytotoxic effects were significantly increased compared with the control for both FCW samples (F1-F2) at 25 µg/mL and above in a dose-response manner. Close to 100% cytotoxicity was observed in the human bronchial epithelial cells at 100 µg/mL of the F1 and F2 particles after 24 h

Table 3
Results of chemical speciation modeling at two different redox potential representing a normal redox potential in water (pe 5) and oxidative conditions (pe 10) in PBS (pH 7.3) at 25 • C and atmospheric pressure. Aqueous species are marked in bold. ssolid; SHEstandard hydrogen electrode.  Y.S. Hedberg et al. incubation. In contrast, no cytotoxic effect was observed for the sample S2, and significant increased cytotoxic effects were observed only for the highest exposure of 100 µg/mL for the sample S1 (Fig. 11). A dose-response increase in DNA damage was observed for the FCW samples F1 and F2 at 3 h exposure, where significant genotoxic effects were observed starting at a dose of 25 µg/mL and 50 µg/mL for the F1 and F2 samples, respectively (Fig. 11). No genotoxic effects were observed following exposure to solid wire S1-S2 samples. No cytotoxic effect was observed for the control experiment with up to 20 µg F/mL (from NaF) (data not shown).

Further discussion
The low release of Ni from welding fume particles into PBS observed experimentally can primarily be explained by its absence in the surface oxide (c.f. Section 3.1). A previous study reported an increased release of Ni for the smallest size fraction of the welding fume particles (Mei et al., 2018) due to a generally higher solubility of smaller particles that have a strongly increased surface-to-bulk ratio (Misra et al., 2012;Kuech et al., 2016). In agreement with expectations, Fe was not soluble under the given experimental conditions (Table 3, Fig. 8). Mn was not expected to be soluble either but was to some extent released and detected, possibly due to slow kinetics of precipitation. Cr was only at high oxidative potentials [as Cr(VI)] expected to be in an aqueous (non-solid) form (Table 3). Cyclic voltammetry revealed fully oxidized metal oxides on all welding fume particles. XPS revealed the presence of highly (IV/VI/-VII-valent) oxidized Mn species that could potentially provide a strong oxidative effect and induce the chemical environment necessary to release Cr(VI). The presence of several Mn species, including highly valent species, is in agreement with previous characterization findings of welding fume (Hanley et al., 2017;Andrews et al., 2015). The ability of Mn-containing oxides to oxidize Cr to Cr(VI) and keep it oxidized is a known geochemical phenomenon of Cr-containing minerals, soils, and their release of Cr(VI) to e.g. groundwater (Hausladen and Fendorf, 2017;Oze et al., 2007;Weaver et al., 2002;Wu et al., 2007).
This study stresses that studies of welding fumes should not only focus on separate effects but rather on synergistic effects of environmental conditions that influence the solubility of highly valent Mn and Cr(VI) species. Highly valent Mn oxides, especially soluble oxides, could potentially support the dissolution of Cr species as Cr(VI) under physiologically relevant conditions.
A clear correlation has previously been observed between the presence of NaF and/or K-containing fluorides in the flux and the release of Cr(VI) (c.f. Section 1). The welding fume sample F2, which contained significantly more NaF as compared to the other FCW sample F1, did not release higher amounts of Cr(VI) when normalized per welding fume mass, but did release higher amounts of Cr(VI) in PBS when normalized to the total Cr content. Both FCW samples released significantly more Cr (VI) into PBS when compared to their solid wire reference samples. This study suggests that this is not directly caused by the NaF or fluoride content in FCW fumes. In this study, Na did not contribute to the formation of crystalline and easily soluble chromates, such as Na 2 Cr 2 O 7 . The absence of these phases was confirmed by XRD and CV and the results are in agreement with previous observations for similar welding fumes (Mei et al., 2018), but in contrast to studies on other or similar welding fumes (Floros, 2018;Minni et al., 1984;Sowards, 2006). Instead, Cr(VI) was incorporated in a mixed oxide phase (a spinel), most probably amorphous, since Cr(VI) cannot be incorporated in a crystalline magnetite-type structure (Floros, 2018). Solution-added fluorides did not increase the release and solution-detected amounts of Cr(VI), but resulted instead in a reduced amount of Cr(VI) in solution for one of the two tested fume particle samples. These fluorides furthermore resulted in a reduced Mn release for both investigated fume samples. Considering the chemical speciation modeling results, this is most probably related to an influence of the fluorides on the redox potential, for instance via increased surface access and solubility of iron oxides. NaF was not cytotoxic itself. This study suggests hence that NaF or other easily soluble fluorides do not directly increase the Cr(VI) solubility of welding fume particles in PBS.
There is clear evidence that the solubility, especially of Cr(VI) and Mn, of welding fume particles in PBS and similar media is linked to acute in-vitro cytotoxic and genotoxic outcomes (McCarrick et al., 2019;Olgun et al., 2020). This study confirmed those observations. However, it should be emphasized that long-term effects, such as lung cancer, have been primarily observed for non-soluble species such as iron oxides (Falcone et al., 2018;Moulin et al., 1993). In addition, the small size of the welding fume particles, alone, is causing increased cell uptake and possible short and long-term adverse health effects (Block and Calderón-Garcidueñas, 2009;VanWinkle et al., 2009;Shi et al., 2012;Carlander et al., 2019). Still, there is an advantage to reduce the acute hazard of welding fume already at the source. The welding process and electrode composition appear to be the most important factors influencing the solubility of Mn and Cr(VI). This is in agreement with previous studies that also highlight the effect of the weldability, fume formation rate, and arc stability, which in turn are influenced by the shielding gas and filler metal formulation (Floros, 2016;Vishnu et al., 2018;Keane et al., 2016;Pires et al., 2007). As the composition of the four electrodes used in this study was considerably different, it was not possible to draw any conclusions regarding the slag concepts. This aspect would be very interesting to evaluate in future studies.

Conclusions
This study aimed at exploring the role of NaF (or other fluorides) in flux-cored wire electrodes, as compared to solid wire electrodes, on the stainless steel welding fume composition, the solubility of Cr, Cr(VI), Mn, Ni, and Fe released from welding fume particles into PBS, cytotoxicity, and genotoxicity. The following main conclusions can be drawn: Fig. 11. Cytotoxicity at 24 h (top) and genotoxicity at 3 h (bottom) in human bronchial epithelial cells as estimated based on the Alamar blue assay and the comet assay, respectively, for different doses and the four different welding fume samples. Triton X (1%) and NiO (25 µg/mL) were used as positive controls for cyto-and genotoxicity, respectively, and resulted in significant increase (P < 0.001). The results are presented as mean ± SD of three or four independent experiments (n = 3-4). Ctrlnegative control; *P < 0.05; **P < 0.01; ***P < 0.001.
The welding fume particles generated from wire electrodes containing NaF (and other fluorides) released significantly higher amounts of Cr(VI) into both PBS and an alkaline medium (ISO 16740). This was associated with increased cytotoxicity and genotoxicity in-vitro. These responses were not necessarily a direct cause of Na/K or F ions. Crystalline Na or K-containing chromates were not observed in this study. Cr (VI) was most probably incorporated in an amorphous mixed oxide (spinel). Fluorides did not increase the solubility of Cr(VI) in welding fume particles but were able to reduce the release of Mn for both solid wire and FCW fume particles and to reduce the release of Cr(VI) from solid wire fume particles into PBS. Chemical speciation modeling suggested that metal fluoride complexes were not formed. Instead, an indirect effect, such as higher surface access and solubility of iron oxides caused by fluorides, could be a contributing factor to a reduced solubility of Mn and Cr(VI) in the presence of excess NaF. This study suggests that highly valent soluble Mn species, such as permanganates, contribute to the solubility of Cr(VI) in welding fume particles under physiological relevant conditions.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.