Uranium Aerosols at a Nuclear Fuel Fabrication Plant: Characterization Using Scanning Electron Microscopy and Energy Dispersive X-Ray Spectroscopy

Spectrochimica Acta Part B 131 (2017) 130–137

Contents lists available at ScienceDirect

Spectrochimica Acta Part B

journal homepage: www.elsevier.com/locate/sab

Uranium aerosols at a nuclear fuel fabrication plant: Characterization using scanning electron microscopy and energy dispersive X-ray spectroscopy

E. Hansson a,b,⁎, H.B.L. Pettersson a, C. Fortin c, M. Eriksson a,d a Department of Medical and Health Sciences, Linköping University, 58185 Linköping, Sweden b Westinghouse Electric Sweden AB, Bränslegatan 1, 72136 Västerås, Sweden c Carl Zeiss SAS, 100 route de Versailles, 78160 Marly-le-Roi, France d Swedish Radiation Safety Authority, 17116 Stockholm, Sweden article info abstract

Article history: Detailed aerosol knowledge is essential in numerous applications, including risk assessment in nuclear industry. Received 31 March 2016 Cascade impactor sampling of uranium aerosols in the breathing zone of nuclear operators was carried out at a Received in revised form 28 February 2017 nuclear fuel fabrication plant. Collected aerosols were evaluated using scanning electron microscopy and energy Accepted 3 March 2017 dispersive X-ray spectroscopy. Imaging revealed remarkable variations in aerosol morphology at the different Available online 6 March 2017 workshops, and a presence of very large particles (up to≅100 × 50 μm2) in the operator breathing zone. Charac- teristic X-ray analysis showed varying uranium weight percentages of aerosols and, frequently, traces of nitrogen, Keywords: ﬂ Uranium uorine and iron. The analysis method, in combination with cascade impactor sampling, can be a powerful tool Aerosol for characterization of aerosols. The uranium aerosol source term for risk assessment in nuclear fuel fabrication Impactor appears to be highly complex. Microscopy © 2017 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license X-ray (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction particle size distributions vary with production parameters and, natu- rally, between stages in the nuclear fuel cycle [2–7].Hence,theproper- In fabrication of nuclear fuel, the presence of uranium aerosols can ties of uranium aerosols will differ between sites using different prove hazardous with respect to radiation and chemical toxicity follow- production methods, but detailed descriptions of uranium aerosols in ing inhalation exposure. The main risk scenarios are chronic exposure of nuclear fuel fabrication are scarce in the literature. workers, acute exposure of workers, exposure of the public from normal The UO2 powder for production purposes can be characterized by operations and exposure of the public due to accidental release of urani- the mass median diameter (MMD), and airborne radioactive matter um compounds. The aerosol source term description is fundamental in by the activity median aerodynamic diameter (AMAD). The latter can such risk assessments as it predicts the behavior of aerosols with respect be evaluated by sampling aerosols with cascade impactors [8]. The aero- to dispersion as well as deposition in the airways and subsequent bio- dynamic particle diameter, dae, is described as logical excretion [1]. Scanning electron microscopy (SEM) combined ¼ ½ρ=ðÞρ χ 1=2 ð Þ with energy dispersive X-ray spectroscopy (EDX) is suitable for dae de 0 1 distinguishing uranium aerosols from other aerosols present in an industrial environment. where de is the diameter of the spherical particle with the same volume Characterization studies of uranium particles have been reported in as the particle considered, ρ (g/cm3) is the density of the irregular par- 3 numerous articles and reports over the last 50 years. In the field of nu- ticle, ρ0 the reference density (1 g/cm )andχ the dynamic shape factor clear fuel fabrication, much research has focused on the production pa- (dimensionless) [8–11]. The dynamic shape factor depends on particle rameters of the produced uranium dioxide (UO2) powder, e.g. morphology, and is defined as the ratio of the drag force on the particle flowability, density and sinterability. Such studies have shown that of interest to the drag force on a spherical particle with the same volume. In an industrial environment, few particles are spherical, and a value of 1.5 is typically assumed, i.e. the drag force on the average par- ⁎ Corresponding author at: Department for Radiation Protection and Environment, Westinghouse Electric Sweden AB, 72163 Västerås, Sweden. ticle is assumed to be 50% higher than for a spherical particle with the E-mail address: [email protected] (E. Hansson). same volume [8,10].

http://dx.doi.org/10.1016/j.sab.2017.03.002 0584-8547/© 2017 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). E. Hansson et al. / Spectrochimica Acta Part B 131 (2017) 130–137 131

Mass and activity distributions of aerosols often, but not always, fol- the powder goes through roller compacting and granulation, and lubri- low log-normal distributions [8,12]. Such tendencies for UO2 aerosols cant is added. BA pellet waste is oxidized and recycled at the workshop. have previously been reported [13–15]. Several authors have reported The waste has an MMD of 8–10 μm [25]. AMADs from nuclear fuel workplaces, discussing particle size distribu- At the conversion and powder preparation workshops most of the tions of the sampled material in various detail [16–20]. The elemental uranium is sealed, but various compounds might be exposed to the composition of aerosols might affect aerodynamic properties, and also work environment due to small leakages, maintenance and sampling. serve as an indicator of chemical compound, which is important in Open handling occurs in the pelletizing and BA pelletizing workshops. many applications, including risk assessment [9]. The 235U enrichment of the uranium handled at the site varies between The present work is a case study of uranium aerosols sampled in the depleted uranium (b0.71%) up to 4.95% (mass percentages). The aver- operator breathing zone at a nuclear fuel fabrication plant using cascade age 235U enrichment for the year of 2014 was approximately 3.7%, as impactors. Using electron microscopy and energy dispersive X-ray spec- measured by inductively coupled plasma mass spectrometry [26].Pres- troscopy, the uranium aerosol source term was characterized with re- ence of 236Uand232U is negligible (b0.1% of alpha activity). spect to morphology, size distribution, elemental composition and dynamic shape factor. To the best of our knowledge, no such studies 2.2. Aerosol sampling have been carried out on uranium aerosols sampled in the operator breathing zone at a nuclear fuel fabrication plant. The information is im- Cascade impactors accomplish a separation of particles based on portant in order to correctly carry out risk assessments with respect to aerodynamic diameter by pumping air through the impactor, which is inhalation exposure of workers and the public. divided into several stages. Air flow velocity is increased at each stage, which is prepared with an impaction substrate. Each impactor stage 2. Materials and methods has a specific cut-point, defined as the aerodynamic diameter of particles with 50% probability of impaction [8]. The inertia of large particles 2.1. Uranium source description will cause them to impact onto substrates at the early stages of the impactor, whereas small particles require higher velocities for impaction Several production methods are available for production of UO pel- 2 to occur. All air is filtered through a final collection filter (typically lets for light-water nuclear reactors. The fabrication plant in the present glass fiber material) before leaving the impactor, collecting the remain- study is run by Westinghouse Electric Sweden AB and processes hun- ing particles that were not deposited by impaction. dreds of tons of uranium annually in the different workshops: conver- Marple 298 impactors (Thermo Scientific, Prod. No. SE298) which sion, powder preparation, pelletizing and burnable absorber (BA) operate at 2.0 L/min, were used for sampling in the present study (ex- pelletizing. cept Sampling 2 which was carried out for SEM imaging only). The The conversion is carried out using a wet chemical process where choice of impactor was based on the following merits: 1) its portability UO is formed from uranium hexafluoride (UF ) via ammonium uranyl 2 6 enabled sampling in the operator breathing zone, 2) eight impaction carbonate (AUC). UF is added to a vessel, where AUC is formed and pre- 6 stages (A–H) with relevant cut-points (21.3, 14.8, 9.8, 6.0, 3.5, 1.6, 0.9 cipitated after an exothermic reaction with ammonium carbonate. After and 0.5 μm) and 3) the impactor has been verified in the literature to drying, the AUC powder is fed into a fluidizing bed furnace, where UO is 2 have sharp cut-points [14,27]. formed by reduction. The conversion workshop is complex, with several A Gilian 5000 pump was used and calibrated with an Alicat MB-50 side processes which enable reuse of waste uranium and chemicals. As a SLPM-D orifice flow controller in accordance with the instruction man- result, several additional uranium complexes can be present in the ual [28].Theflow rate was checked before and after each sampling cam- workshop: uranyl fluoride (UO F , ammonium diuranate (ADU) (very 2 2) paign using the same orifice flow controller and a rotameter. Flow rates small amounts from purification of waste uranium), uranium trioxide showed negligible variation (typically b1%). (UO ), uranium octoxide (U O ), uranyl nitrate hexahydrate (UNH) 3 3 8 Eight sampling campaigns were conducted to collect aerosols to be and uranyl peroxide (UO ·2NH ·2HF·2H O) [21,22]. The wet chemical 4 3 2 analyzed with SEM/EDX with the following objectives: AUC route of conversion generates a UO2 powder with a larger average particle size than from the alternative processes (dry route conversion 1. Investigate aerosol morphologies at all four workshops (breathing of UF to UO or wet chemical ADU conversion) [4,5]. The site in the 6 2 zone and complementary sampling at sites of particular interest). present study produces UO powder with an MMD of typically 20 μm, 2 2. Determine the size distribution, elemental composition and dynamic as measured by laser diffraction [23]. It has been shown that UO aero- 2 shape factor, χ, of uranium aerosols in the operator breathing zone at sols from the AUC route of conversion are larger than aerosols from the the pelletizing workshop. This workshop was prioritized from a ra- ADU route of conversion [24]. Interestingly, we have not found any re- diological risk assessment perspective. ports on the characterization of airborne AUC. 3. Evaluate SEM/EDX as a method for determining the elemental com- The powder preparation workshop prepares the UO powder for 2 position of aerosols at the conversion workshop with respect to fluo- pelletizing. This is done by verification of low levels of humidity in the rine and nitrogen, giving an indication of material chemical form. powder (for criticality safety reasons), blending to obtain the desired enrichment and blending with appropriate amounts of U3O8 (for The following impaction substrates were used: sticky carbon tape sintering properties). In addition, powder for the BA pelletizing work- (Ted Pella, Inc., Prod. No. 16085-1) (Sampling 1, 3, 4 and 8), mixed cellu- shop is milled. Waste materials such as grinding waste and defect pel- lose ester membrane (MCE) (Thermo Scientific, SEC-290-MCE) (Sam- lets from the pelletizing workshop are oxidized to U3O8 to be used for pling 2 and 5) and glass fiber (Thermo Scientific, SEC-290-MCE) powder blending. Milled UO2 powder and oxidized waste have MMDs (Sampling 6 and 7). Carbon tape is, due to its conductivity, ideal for of 3–4 μmand5–7 μm, respectively [23,25]. SEM/EDX analysis, and was chosen for all breathing zone sampling. The main pelletizing workshop produces the majority of the fuel pel- MCE and glass fiber were used for complementary sampling. Final collec- lets by pressing UO2 into pellets that undergo sintering at ≅1700 °C in a tion filter were glass fiber filters (Whatman GF/A, 1.6 μmporesize).Each hydrogen atmosphere to obtain the required density. The ceramic sampling campaign was designed to sample enough particles for a repre- pellets are then ground to the proper dimensions and finally undergo sentative SEM/EDX analysis, but avoiding particle overlap on the impac- a visual inspection before encapsulation into fuel rods. tion substrates and was thus based on knowledge of airborne uranium The BA pelletizing workshop produces pellets in a similar way to levels at the site. The sampling campaigns are summarized in Table 1, that of the main pelletizing workshop. The already milled powder is and a full description is given in Table S1 (Appendix). Impaction patterns blended with gadolinium oxide (Gd2O3)andU3O8.Beforepressing, were generally homogeneous, as illustrated in Fig. S1 (Appendix). 132 E. Hansson et al. / Spectrochimica Acta Part B 131 (2017) 130–137

Table 1 by the software agreed with manual counting with point EDX measure- Brief description of the conducted sampling campaigns. ments during scan set-up. Sampling no. Description The acquisition time for EDX measurements was set to 6 s per particle. The particle discrimination level was set by adjusting the magnifica- 1 Conversion workshop, operator breathing zone μ 2 Conversion workshop, ventilation air tion so that objects larger than 0.1 m could be detected. Visual 3 Powder preparation workshop, operator breathing zone inspections showed that the number of particles smaller than 0.1 μm 4 Pelletizing workshop, operator breathing zone was very small. Each impaction substrate held thousands of particles, 5 Pelletizing workshop, pellet inspection work station so in order to perform scans within a reasonable time frame, a represen- 6 Pelletizing workshop, pellet grinding 7 Pelletizing workshop, ventilation air tative section of each impaction substrate was selected for scanning. 8 BA pelletizing workshop, operator breathing zone Each section was defined in INCAFeature as an area, consisting of a number of fields of fixed size. Each scan was run until all non-discriminated particles in all fields had been analyzed, in total 0.6–1.2 h per impactor 2.3. SEM/EDX analysis substrate. A total of 377, 239, 167, 478 and 392 (impaction Stages C–G, respectively) uranium particles were automatically scanned and ana- SEM/EDX analyses of sampled aerosols were carried out using a Carl lyzed using INCAFeature. Zeiss EVO LS15 Scanning Electron Microscope at the International The equivalent circle diameter (ECD) was used as an indicator of Atomic Energy Agency (IAEA) Environment Laboratories. The X-ray de- particle size. Particle thickness was not evaluated. Elements considered tector used for the EDX analyses was an Oxford Instruments X-Max 50 in the elemental composition analysis from the automatic counting detector (electrically cooled silicon drift detector with an active surface were: oxygen, fluorine, sodium, aluminum, silicon, calcium, chromium, of 50 mm2 and equipped with a polymer window) with a resolution of manganese, iron, nickel, copper, zink and uranium. Carbon was exclud- 125 eV on the manganese Kα line. The software used for SEM imaging ed due to its abundance in the impaction substrate. Results were was SmartSEM (version 5.06) and INCAFeature (version 5.03) for the expressed as normalized weight percentages, as per INCAFeature de- EDX analyses. The INCAFeature software used a cobalt standard for cal- fault. The cobalt standard was used to regularly measure beam intensity ibration of the EDX system and quantification was carried out by peak and check detector calibration. deconvolution, digital filtering and a matrix correction using the XPP method (exponential model of Pouchou and Pichoir Matrix Correction, 2.3.3. Conversion workshop – aerosol elemental composition correcting for effects of atomic number, absorption and fluorescence Particles from impaction Stages B, E and F from Sampling 1 were ran- on X-ray emission from a sample) [29]. The XPP method is designed domly chosen for evaluation of elemental composition, especially nitro- to improve quantification of low Z elements in high Z matrices. A gen and fluorine. These impaction stages correspond to inhalable (Stage 20 kV accelerating voltage was used to enable excitation of both urani- B) and respirable (Stages E–F) size fractions. Automatic quantification of um L and M electrons. these elements could not be carried out using INCAFeature due to low Additional SEM/EDX measurements of reference particles (Section nitrogen and fluorine concentrations in combination with scan settings. 2.3.5) were carried out at the Carl Zeiss France S.A.S Laboratories using Instead a manual inspection of EDX spectra was carried out. EDX spectra the same microscope model, combined with an Oxford Instruments X- frequently had to be re-acquired due to inhomogeneous distributions of Max 80 detector (80 mm2 active surface). The same software versions elements within each particle and alteration of particle structure due to of SmartSEM and INCAFeature were used. interaction with the electron beam. The latter phenomenon occurred by For sets using carbon tape, images were acquired by the burning of low-Z components of the particle. In those cases, peaks were backscattered electron detector (BSD) in high vacuum mode. For sets only visible in the beginning of the spectrum acquisition, before they using MCE and glass fiber, images were acquired in variable pressure faded into the continuum of the spectrum. Electron beam interaction mode in order to minimize charging effects on the sample. with AUC has previously been observed to cause reduction to UO3 and/or U3O8 [21]. Spectra were regularly acquired by focusing the electron beam on the impaction substrate, with no particles nearby, in order 2.3.1. All workshops – aerosol morphology to rule out distortions from the sample support. Nitrogen and fluorine Uranium aerosol morphology was evaluated by scanning impaction was never observed in the background support spectra. surfaces and final collection filters from all sampling campaigns. Parti- cles of unknown or deviating morphologies were analyzed using 2.3.4. Pelletizing workshop – aerosol dynamic shape factor INCAFeature to verify uranium contents. Uranium particles were im- The dynamic shape factor was calculated for the pelletizing work- aged and non-uranium particles ignored. Point EDX measurements shop (Sampling 4), impaction Stages C–G. Particle volume was assumed were carried out to determine the elemental composition. This informa- to equal the median projected particle area (generated by the tion was used to determine the origin of each particle studied. INCAFeature software) multiplied by particle height which was assumed to equal half median particle width (also generated by

INCAFeature). Equivalent spherical diameters, de,werederivedand 2.3.2. Pelletizing workshop - aerosol size distribution and elemental shape factor estimates were carried out for different densities using composition Eq. (1). Aerosol densities of 1.6 g/cm3, 2.4 g/cm3,5.7g/cm3 and 3 The INCAFeature software was used to determine particle size distri- 10.5 g/cm were assumed as they correspond to milled UO2 powder bution and elemental composition of aerosols at the pelletizing work- fill density, regular UO2 powder fill density, un-sintered pellet density shop (Sampling 4) by automatically scanning five impaction and sintered pellet density, respectively [23,25]. substrates (Stages C–G). Stages A, B and H could not be automatically scanned due to setup difficulties (Stages A and B contained too few par- 2.3.5. Reference particles ticles over too large areas, and Stage H had too much particle overlap). The potential bias (further discussed in Section 3.2) from electron The final collection filter was not scanned due to the uneven glass beam interaction with the support material was investigated using cer-

ﬁber surface. tiﬁed high purity U3O8 reference particles (Standard Reference Material In order to detect uranium particles and discriminate non-uranium U-010, National Bureau of Standards). Particle sizes ranged between particles, the SEM was operated with low brightness and high contrast ≅0.5–10 μm and were attached to sticky carbon tape (Ted Pella, Inc., settings which were optimized for each sample. The optimization was Prod. No. 16085-1). Agglomerates of particles were present and particle carried out by ensuring that the number of uranium particles detected shapes were somewhat irregular. E. Hansson et al. / Spectrochimica Acta Part B 131 (2017) 130–137 133

Scans were carried out using the same support material and a 20 kV but no fluorine. The amorphous structure suggests an early process accelerating voltage. The INCAFeature protocol for automated scans step, perhaps related to precipitation of particles. We are not aware of described in Section 2.3.2 was used with a 2 s acquisition time due to similar uranium aerosols described in the literature. the larger detector area. Particles below 2 μm ECD were difficult to ana- Rod-like uranium aerosols were observed at several impaction lyze using the automated protocol due to contrast settings and particle stages (including the final collection filter) at the conversion workshop overlap. Small particles were thus analyzed in manual mode. A total of (Sampling 1–2) either as agglomerates (Fig. 1b) or individual particles 76 and 27 particles were analyzed in automatic and manual mode, (Fig. 1c). EDX analyses showed a presence of fluorine and nitrogen. respectively. The uranium weight percentages estimated by The origin of rod-like aerosols is unknown and we are not aware of sim- INCAFeature were compared to the expected theoretical level of 85% ilar uranium aerosols described in the literature. Possibly it originates for pure U3O8. from the handling of uranyl peroxide from the hydrogen peroxide process system, where uranium is extracted for recycling from process so- 3. Results and discussion lutions containing fluorine and nitrogen. The formation of particles is a highly complex process, depending on numerous parameters. A forma- 3.1. All workshops – aerosol morphology tion of somewhat similar cylindrical particles has been demonstrated

for UO4 particles from a hydrogen peroxide solution, as well as UO3 Particle morphology was observed to vary with impaction stage cut- and U3O8 under specific conditions [30–31]. The EDX analyses of the re- point and sampling location. The pre-pelletizing workshops, i.e. conver- maining particles in Fig. 1c, including the square-shaped particle, sion and powder preparation, showed a looser structure, more agglom- showed nitrogen traces, but no fluorine, indicating AUC. AUC crystals eration and a more irregular morphology compared to the pelletizing have been shown to occur in shapes similar to the particle in this and BA pelletizing workshops. Late impaction stages (E–H) showed image [2,32]. less agglomeration than the early stages. On the individual aerosol Another finding at the conversion workshop, Sampling 1–2, was the level, the different workshops showed remarkable variation, especially presence of spherical uranium particles (Fig. 1d). EDX analyses showed, at impactor Stages A–B. Structures of the uranium aerosols included, in addition to uranium, traces of fluorine and nitrogen. The spherical apart from discrete particles, many different geometrical structures, ag- shape and size strongly indicates UO2F2. Several studies have shown glomerates of particles and particles attached to lower-Z materials. that formation of UO2F2 from UF6 in a humid environment results in Some structures, especially from the conversion workshop, were dif- spherical shapes of about ≅0.2–2.0 μm diameter [33–35].Thepresence ficult to interpret. Fig. 1a, from Sampling 2 shows a uranium aerosol of nitrogen is slightly surprising, but could be explained by attachment were the EDX analysis indicated a presence of iron, traces of nitrogen of nitrogen-rich materials present in the atmosphere. The low-contrast

Fig. 1. Aerosol morphology at different sampling locations. a–e: conversion workshop, Sampling 1–2; f–h: powder preparation workshop, Sampling 3; i–k: pelletizing workshop, Sampling 4–7; and l: BA pelletizing workshop, Sampling 8. 134 E. Hansson et al. / Spectrochimica Acta Part B 131 (2017) 130–137 particles in the lower part of the image contained iron but no uranium. Most particles at the pelletizing and BA pelletizing workshop could Remaining particles contained uranium and nitrogen, indicating AUC. be correlated to sintered material (e.g. Fig. 1i). The aerosol morphology Fig. 1e shows an example of a particle consisting of uranium particles varied less compared to the other workshops but showed a very large attached to a low-Z particle. This particular particle contained nitrogen, size range. A large cluster of aerosols of varying sizes (Fig. 1j) sampled iron, chromium, aluminum and ﬂuorine. The particle origin is unknown, near one of the pellet grinders contained some very small (≅0.1 μm) but the presence of ﬂuorine and nitrogen indicate early conversion particles. The largest sampled aerosol was a pellet shard (Fig. 1k) steps or uranyl peroxide. collected in the operator breathing zone at the pelletizing workshop, Particles sampled at the powder preparation workshop were more measuring almost 100 × 50 μm2 (thickness unknown). Both pelletizing straightforward to identify as pellet shards and waste from pellet grind- workshops showed agglomeration of particles, as illustrated in Fig. 1l. ing from the pelletizing workshop, and un-milled UO2 powder from the Agglomeration tended to be more frequent at the BA pelletizing work- conversion workshop. Fig. 1f shows typical pellet grains and pores from shop, which could be explained by the addition of lubricant. The rate sintering in the pelletizing workshop. Some particles showed a more of agglomeration was not further investigated. rounded shape, as illustrated by Fig. 1g, which is typical for UO2 powder from the conversion workshop. The small particles in Fig. 1garelikelyto 3.2. Pelletizing workshop – aerosol size distribution and elemental originate from waste material from pellet grinding, and have possibly composition undergone oxidation, suggesting U3O8. Some particles showed more obvious indications of oxidation, such as Fig. 1h, where the rugged sur- Fig. 2 shows aerosol ECD distributions and uranium weight percent- face strongly suggests partial oxidation of a pellet shard, with obvious ages as counted by INCAFeature for Sampling 4. Overlap between the pores from the sintering process showing. impaction stages is evident and also expected by cascade impactor

Fig. 2. Distribution of uranium aerosol equivalent diameter (de) and uranium weight percentage for Sampling 4. Stages C–G correspond to 377, 237, 167, 478 and 392 particles, respectively. Note that de (ECD) does not equal dae (aerodynamic diameter) (Eq. (1)). A bin width of 0.25 μm was used in each histogram. For clarity, y-axes in Panels a) and i) were adjusted. The values of these bins were 0.32, 0.55 and 0.26, respectively. Three particles of 10.2, 11.2 and 13.8 μm ECD (Stage C) are outside the scale of the x-axes. Data have not been corrected for the oxygen bias discussed in the text. E. Hansson et al. / Spectrochimica Acta Part B 131 (2017) 130–137 135 cut-point definition. Variations in particle shape and density will affect the aerodynamic properties of particles, and may further add to the observed overlap (Eq. (1)). Similar patterns were observed in a pre-study at the same plant [36]. The ECD distributions must be interpreted with care, since particle counting is dependent on several parameters, e.g. brightness/contrast settings that must be optimized for each sample. Small differences in brightness/contrast can affect detection limits, however the electron beam was stable in the presented measurements and we believe the effects of such a bias is small in the present work. It was noted that some large particles near the scanned field edges were not counted by INCAFeature, possibly biasing particle size distributions. Agglomerates of uranium particles and small uranium particles attached to large non-uranium particles were sometimes counted as multiple individual uranium particles. This is the explanation to the many particles b1 μminFig. 2a–b. Observed uranium weight percentages were lower than expected Fig. 4. Uranium weight percentage of uranium aerosols with 2–4 μm ECD collected at (theoretically 88% for UO and 85% for U O ). The uranium weight per- 2 3 8 impaction Stages C–G. Error bars correspond to ±1 standard deviation. The solid line centages in Fig. 2 appeared to be lower for small particles, and were ob- represents the expected, uncorrected, measured uranium weight percentage for U3O8 served to drop with each impaction stage. This is probably due to a bias aerosols, based on our reference particle measurements. The dotted and dashed lines from oxygen in the impaction substrate (carbon was not included in the represent the theoretical uranium weight percentage for UO2 and U3O8, respectively. quantification) and a presence of other elements. For micron-sized par- Data have not been corrected for the oxygen bias discussed in the text. ticles a majority of the electron beam interaction may come from interactions in the impaction substrate, which will affect measured X-ray operating principles and could indicate a presence of multiple uranium intensities [37]. An observation similar to Fig. 2 has been reported else- compounds. Quantification of elemental composition of uranium parti- where for uranium particles when the same accelerating voltage was cles using SEM/EDX is well-known to be challenging, as emphasized by used [38]. e.g. Ciurapinsky et al. [39]. Our measurements are semi-quantitative

As illustrated in Fig. 3, automated measurements of U3O8 refer- and thus the results must be interpreted with care. ence particles (ECD N 2 μm) showed uranium weight percentages Uranium aerosols at the pelletizing workshop frequently contained of70.3±2.8%(1standarddeviation),i.e.≅ 15 percentage points iron and, to some extent, chromium and nickel, as shown in Table 2. lower than the theoretical level (85%). Manual scans of particles These elements probably attach to the uranium pellets as they slide smaller than 2 μm indicated lower uranium weight percentages, i.e. along transportation bands. Nickel is also found in the grinding equip- a more pronounced bias from oxygen in the substrate. ment. In certain batches, chromium is added to the powder to improve Uranium weight percentages were further investigated by studying sintering properties. Weight percentages of these elements in uranium particles of the same ECD (2–4 μm, a size range present at all impaction aerosols were in many cases substantial (N50%) (Table S2 (Appendix)). stages) present on different impaction stages (Fig. 4). Particles on the later impactor stages (lower cut-points) unexpectedly appeared to have lower uranium weight percentages. We cannot explain this with 3.3. Conversion workshop – aerosol elemental composition the aforementioned oxygen bias since reference particle measurement indicated an approximately uniform bias of 15 percentage points in EDX analysis showed a much more complex elemental composition the 2–4 μm ECD size interval. We hypothesize that 2–4 μm ECD particles at the conversion workshop compared to the other workshops. Elemen- collected at late impaction stages have a looser structure (and thereby a tal composition alone is insufﬁcient to distinguish between the many lower effective density and aerodynamic diameter of particles collected potentially present uranium compounds, but a presence of nitrogen at late impaction stages (Eq. (1)) than equally sized particles at early im- and/or ﬂuorine does, however, give an indication of chemical form. paction stages. This would be in agreement with cascade impactor Fig. 5a shows the EDX spectrum of a particle with an obvious presence

of nitrogen and fluorine, indicating UO2F2, AUC and/or uranium peroxide. Fig. 5b shows the EDX spectrum of a uranium particle where nitrogen and fluorine peaks are absent, indicating uranium oxide. It was found that for Sampling 1, impaction Stages E and F, about 50% of the uranium aerosols contained nitrogen and/or fluorine. The re-

maining 50% indicated uranium compounds such as UO2 and/or U3O8. At impaction Stage B, N90% of the uranium particles contained nitrogen and/or ﬂuorine. For Sampling 2, about 70% of uranium aerosols indicated nitrogen and/or ﬂuorine, but were not correlated to impactor cut- point.

Table 2 Percentage of uranium aerosols (Sampling 4) that also contained either iron, chromium or nickel at impactor Stages C–G.

Impactor stage Uranium aerosols containing either Fe, Cr or Ni

Fe Cr Ni

C 86% 31% 29% Fig. 3. Uranium weight percentages of U O reference particles. “+” and “o” markings 3 8 D 79% 2.5% 2.1% represent automated and manual scans, respectively. The dashed line represents the E 34% 1.2% 0.6% theoretical uranium weight percentage for U O . Two particles with ECD 37.2 μm and 3 8 F 49% 0.6% 0% 61.3 μm are outside the scale of the x-axis. Their corresponding uranium weight G 18% 0% 0% percentages are 74.7% and 75.6%, respectively. 136 E. Hansson et al. / Spectrochimica Acta Part B 131 (2017) 130–137

uranium particles with high chromium contents. The hypothesis is supported by the high chromium contents at Stage C (Table S2). A density of 10.5 g/cm3 corresponds to sintered pellet material, and was believed to give the best shape factor estimate. However, the shape factor estimates turned out to be much higher than the expected value of ≅1.5, suggesting a lower density. There have been reports of dynamic shape factors in the range of 2.2–3.6 for plutonium/uranium oxide aerosols of 0.5–1.5 μm aerodynamic diameters [40]. However, these were long chains of particles and the particles in the present work are expected to be associated with a lower shape factor. Fig. 6 suggests that uranium aerosols in the pelletizing workshop have a density lower than 10.5 g/cm3and a shape factor N1.5. This is supported by the electron microscope images frequently showing high-

ly irregular particles. The correlation between dae, de, density and shape factor appears to be very complex.

4. Conclusions and perspectives

The present work constitutes, to the best of our knowledge, the first SEM/EDX characterization of uranium aerosols sampled with cascade impactors in the operator breathing zone at a nuclear fuel fabrication plant. These characterization studies will add to the understanding of uranium aerosols in nuclear fuel fabrication, affecting applications such as risk assessments. Remarkable variations of uranium aerosol morphologies were found at the different workshops, and a presence of very large particles (up to ≅100 μm) in the breathing zone of operators could be verified. The exact fl Fig. 5. SEM/EDX spectra of a: a uranium aerosol containing nitrogen and uorine; and b: a formation process of all aerosols is not known, and needs further uranium oxide particle. The uranium M-lines are seen in both particles, however, the intensity from the carbon support makes the uranium intensity appear lower in b) investigation. compared to a). This is auto scaled by INCAFeature. At the pelletizing workshop, a clear size fractionation of aerosols was visible, as was an overlap over particle sizes. It was shown that 18–86% of the collected uranium aerosols also contained iron, chromium and/or 3.4. Pelletizing workshop - aerosol dynamic shape factor nickel. Uranium weight percentages appeared to be lower for late impaction stages. This is partially explained by a bias from electron beam Calculating the aerosol dynamic shape factor using Eq. (1) for the interactions with oxygen in the impaction substrate. Measurements of pelletizing workshop proved to be difficult; in particular the estimates U3O8 reference particles indicated that the bias causes the software to of de and particle density assumptions. generate uranium weight percentages 15 percentage points lower Fig. 6 illustrates shape factor estimates for densities corresponding than the theoretical level of 85%. This is valid for particles larger than to milled UO2 powder fill density, regular UO2 powder fill density, un- 2 μm ECD. The bias is greater for smaller particles. However, aerosols sintered pellet density and sintered pellet density as described in in the same size interval (2–4 μm ECD, where the bias was approximate- Section 2.1. A dynamic shape factor below 1 is not possible by definition, ly uniform), but collected at different impaction stages, unexpectedly which suggests faulty de estimates for Stage C. This can be explained by showed different uranium weight percentages. This could indicate a that a few large chromium particles carrying numerous small uranium variable density of uranium aerosols, but the finding requires further in- particles were misinterpreted by INCAFeature as multiple small vestigation. The aerosol dynamic shape factor proved difficult to evaluate, but there are indications that the generally assumed value of 1.5 is an underestimation. Size distribution and shape factor need further investigation for other workshops. Many particles at the conversion workshop showed traces of nitrogen and fluorine. Such particles occurred much more frequently (90%) at the early impaction stages compared to the later impaction stages (50%). SEM/EDX is a powerful tool for evaluation of aerosol parameters such as size distribution, elemental composition and shape factor. Future perspectives include studies of chemical form and crystalline states using micro X-ray fluorescence (µXRF) and micro X-ray diffraction (µXRD) analysis.

Acknowledgements

The authors are grateful to the International Atomic Energy Agency for allowing use of their SEM/EDX equipment at the Environment Labo- ratories in Monaco. Westinghouse Electric Sweden AB is acknowledged for participation in the study, providing samples and funding. Olav Axelsson Memorial Foundation is acknowledged for funding (grant number LIO-520711). The Swedish Radiation Safety Authority is thanked for funding of the pre-study to the present work (grant number Fig. 6. Shape factor estimates for the pelletizing workshop (Sampling 4) for impaction SSM2014-4843). Special thanks go to Isabelle Levy and Francois

Stages C–G. Error bars correspond to estimated uncertainty (±15%) in de. Oberhansli at the IAEA for assistance with the SEM, Magnus Hedberg E. Hansson et al. / Spectrochimica Acta Part B 131 (2017) 130–137 137 and Noëlle Albert at the European Commission, DG Joint Research Cen- [18] M.D. Dorrian, M.R. Bailey, Particle size distributions of radioactive aerosols measured in workplaces, Radiat. Prot. Dosim. 60 (1995) 119–133. tre, Directorate G - Nuclear Safety and Security for their commenting on [19] E. Ansoborlo, Particle size distributions of uranium aerosols measured in the French the manuscript, and to Jörgen Gustafsson and numerous colleagues at Nuclear Fuel Cycle, Radioprotection 32 (1997) 319–330. Westinghouse Electric Sweden AB for invaluable discussions and [20] E. Ansoborlo, V. Chazel, M.H. Hengé-Napoli, P. Pihet, A. Rannou, M.R. Bailey, N. Stradling, Determination of the physical and chemical properties, biokinetics, and comments. dose coefficients of uranium compounds handled during nuclear fuel fabrication in France, Health Phys. 82 (2002) 279–289. Appendix A. Supplementary data [21] L. Hälldahl, In situ studies of the decomposition of ammonium uranyl carbonate in an electron microscope, J. Nucl. Mater. 126 (1984) 170–176. [22] Westinghouse Electric Sweden AB, Beskrivning av bränslefabrikens funktion och Supplementary data to this article can be found online at http://dx. uppbyggnad, Säkerhetsredovisning för bränslefabriken(SRB Avsnitt 1.4, Internal doi.org/10.1016/j.sab.2017.03.002. documentation, report ID: SRB 1.4) 2006. [23] D. Bjernedal, Senior Engineer, Pellet Fabrication, Westinghouse Electric Sweden AB, Personal Communication, 2016. References [24] B. Raghunath, R. Ramachandran, S. Majumdar, Particle size distribution of UO2 aerosols, J. Nucl. Mater. 185 (1991) 299–301. – [1] W. Hofmann, Modelling inhaled particle deposition in the human lung areview,J. [25] T. Fries, Ceramics specialist, Westinghouse Electric Sweden AB(Personal communi- – Aerosol Sci. 42 (2011) 693 724. cation) 2015. [2] M. Jensen, Investigations on Methods of Producing Crystalline Ammonium-uranyl [26] Westinghouse Electric Sweden AB, Årsrapportering 2014 av utsläpp och Carbonate and Its Applicability for Production of Uranium-dioxide Powder and omgivningskontroll enligt SSMFS 2008:23(Internal documentation, report ID: BSA Pellets(Denmark. Forskningscenter Risoe. Risoe-R; No. 153) 1967. 15-016) 2015. [3] L. Hälldahl, Studies of Reactions Occurring in the AUC process: From UF6 to Sintered [27] K.L. Rubow, V.A. Marple, J. Olin, M.A. McCawley, A personal cascade impactor: de- UO2 Pellets, University of Stockholm, Chemical Communications, 1985 (ISBN: 91- sign, evaluation and calibration, Am. Ind. Hyg. Assoc. J. 48 (1987) 532–538. 7146-647-9, No. 2). [28] Thermo Fisher Scientific Inc., Marple Personal Cascade Impactors Series 290, Instruc- [4] H. Bergqvist, W. Sahle, Electron Microscopy Characterization and X-ray Measure- tion Manual, Part Number 100065-00, 2009. ments of UO2 Powder and UO2 Based Fuel Pellet Materials, Royal Institute of Tech- [29] Oxford Instruments, INCA Energy, Operator Manual(Issue 2.1) 2006. nology, FNM/ICT, 2010 (KTH Report No. 100331). [30] K.W. Kim, J.T. Hyun, K.Y. Lee, E.H. Lee, K.W. Lee, K.C. Song, J.K. Moon, Effects of the [5] International Atomic Energy Agency, The Front End of the Uranium Fuel Cycle, 50th different conditions of uranyl and hydrogen peroxide solutions on the behavior of IAEA General Conference, 2006 (Documents, available 2016-09-22 at: https://www. the uranium peroxide precipitation, J. Hazard. Mater. 193 (2011) 52–58. iaea.org/About/Policy/GC/GC50/GC50InfDocuments/English/gc50inf-3-att6_en.pdf). [31] S. Manna, B.R. Saswati, J.B. Joshi, Study of crystallization and morphology of ammo- [6] M. Kraiem, S. Richter, H. Kühn, E.A. Stefaniak, G. Kerckhove, J. Truyens, Y. Aregbe, In- nium diuranate and uranium oxide, J. Nucl. Mater. 424 (2012) 94–100. vestigation of uranium isotopic signatures in real-life particles from a nuclear facility [32] A. Mellah, S. Chegrouche, M. Barkat, The precipitation of ammonium uranyl carbon- – by thermal ionization mass spectrometry, Anal. Chem. 83 (3011 3016) (2011). ate (AUC): thermodynamic and kinetic investigations, Hydrometallurgy 85 (2007) [7] R. Hou, T. Mahmud, N. Prodromidis, K.J. Roberts, R.A. Williams, Synthesis of UO2F2 163–171. nanoparticles in a tubular aerosol reactor: reactor design and experimental investi- [33] R. Kips, A. Leenaers, G. Tamborini, M. Betti, S. Van den Berghe, R. Wellum, P. Taylor, gations, Ind. Eng. Chem. Res. 46 (2007) 2020–2033. Characterization of uranium particles produced by hydrolysis of UF6 using SEM and [8] W.C. Hinds, Aerosol Technology: Properties, Behavior, and Measurement of Air- SIMS, Microsc. Microanal. 13 (2007) 156–164. borne Particles, second ed. John Wiley & Sons, Inc., 1999 (Chapters 3.5 (Nonspher- [34] G.L. Wagner, S.A. Kinkead, M.T. Paffett, K.D. Rector, B.L. Scott, A.L. Tamasi, M.P. – – ical particles, pp. 51 53), 5.5 (Inertial impaction, pp. 121 128), 6.4 (Particle bounce, Wilkerson, Morphologic and chemical characterization of products from hydrolysis pp. 146–147)). of UF6, J. Fluor. Chem. 178 (2015) 107–114. [9] ICRP, Individual Monitoring for Internal Exposure of Workers(ICRP Publication 78, [35] V. Stebelkov, Informativeness of Morphology Analysis of Uranium Microparticles Ann. ICRP 27) 1997. from Industrial Dust at Nuclear Plants, GLOBAL, Tsukuba, Japan, 2005. [10] ICRP, Human Respiratory Tract Model for Radiological Protection(ICRP Publication [36] E. Hansson, H.B.L. Pettersson, M. Eriksson, Uranium Aerosol Characteristics at a Nu- 66, Ann. ICRP 24) 1994. clear Fuel Manufacturing Site - Particle Size, Morphology and Chemical Composi- [11] ICRP, Occupational Intakes of Radionuclides: Part 1(ICRP Publication 130, Ann. ICRP tion, Swedish Radiation Safety Authority, 2015 (Research Report No. 2015:18). 44(2)) 2015. [37] C.U. Ro, J. Osán, R. Van Grieken, Determination of low-Z elements in individual en- [12] R. Akselsson, M. Bohgard, A. Gudmundsson, H.C. Hansson, B. Martinsson, B. vironmental particles using windowless EPMA, Anal. Chem. 71 (1999) 1521–1528. Svenningsson, Aerosoler, Lunds Tekniska Högskola and NOSA, 1994. [38] V.G. Dyukov, E.N. Evstaf'eva, V.A. Stebelkov, A.A. Tatarintsec, V.V. Khoroshilov, X-ray [13] K.V. Prasad, A.Y. Balbudhe, G.K. Srivastava, R.M. Tripathi, V.D. Puranik, Aerosol size microanalysis of uranium dioxide particles by using low electron beam energy, Bull. distribution in a uranium processing and fuel fabrication facility, Radiat. Prot. Russ. Acad. Sci. Phys. 78 (2014) 851–853. – Dosim. 140 (2010) 357 361. [39] A. Ciurapinski, J. Parus, D. Donohue, Particle analysis for a strengthened safeguards [14] Y.S. Cheng, J.L. Kenoyer, R.A. Guilmette, M.A. Parkhurst, Physicochemical character- system: use of a scanning electron microscope equipped with EDXRF and WDXRF ization of capstone depleted uranium aerosols II: particle size distributions as a spectrometers, J. Radioanal. Nucl. Chem. 251 (2002) 345–352. – function of time, Health Phys. 96 (2009) 266 275. [40] M.D. Allen, O.R. Moss, J.K. Briant, Measurements of Dynamic Shape Factors of LMFBR [15] D.S. Surya Narayana, A.R. Sundararajan, J. Harvey, Characterization of uranium oxide Aggregate Aerosols, Battelle Pacific Northwest Labs, Richland, WA (USA), 1980 – aerosols, J. Aerosol Sci. 25 (1994) 909 922. 147–149 (PNL—3300(PT.1)). (Available 2015-11-13 at https://inis.iaea.org/search/ [16] T. Kravchik, S. Oved, O. Paztal-Levy, O. Pelled, R. Gonen, U. German, A. Tshuva, Deter- search.aspx?orig_q=RN:11558295). mination of the solubility and size distribution of radioactive aerosols in the uranium processing plant at NRCN, Radiat. Prot. Dosim. 131 (2008) 418–424. [17] H. Connelly, R.G. Jackson, Review of Respirable Particle Size Range, AMEC, Refer- ence: RP0605-86C, 2013.