Secondary Logo

Journal Logo


Particle Size-dependent Dissolution of Uranium Aerosols in Simulated Lung Fluid: A Case Study in a Nuclear Fuel Fabrication Plant

Hansson, Edvin1,2; Pettersson, Håkan B. L.3; Yusuf, Ibtisam3; Roos, Per4; Lindahl, Patric5; Eriksson, Mats2

Author Information
doi: 10.1097/HP.0000000000001564
  • Open



Upon inhalation exposure to materials containing radioactivity, the lung equivalent dose (Hlung) can constitute a large fraction of the committed effective dose (CED) due to the lung’s high tissue weighting factor of 0.12 (ICRP 2007). This is particularly important in the case of alpha emitters, e.g., most uranium isotopes, where a radiation weighting factor of 20 is recommended for calculations of equivalent doses (ICRP 2007). Occupational inhalation exposure to uranium is a concern in nuclear fuel fabrication, where committed effective doses need to be evaluated to ensure worker safety and regulatory compliance. Calculations are normally carried out based on the Human Respiratory Tract Model (HRTM) developed by the International Commission on Radiological Protection (ICRP) (ICRP 1994, 2015). The present work is a case study in a nuclear fuel fabrication plant, aimed at evaluating how the aerodynamic size of uranium particles affects key parameters and dose calculations in the HRTM. This was carried out by size-fractionated aerosol sampling and dissolution rate experiments in simulated lung fluid.

The HRTM is a compartment model that predicts the biological fate of inhaled particulate matter following deposition in the respiratory tract. Clearance of deposited particles is governed by mechanisms such as coughing, mucociliary transport, and phagocytosis by macrophages (upon which dissolution in the phagolysosome will occur). In addition, particles will start to dissolve in the extracellular lung fluid followed by uptake to blood. The dissolution process has long been modeled using a biexponential function (Mercer 1967). In the ICRP terminology, absorption represents the transfer of material into blood, i.e., the combination of dissolution/dissociation of particles and the uptake of material to blood (ICRP 2015). The time-dependent dissolution of material deposited in the respiratory tract is modeled in its simplest form by a set of absorption parameters: a fraction of the material (fr) dissolves at a rapid rate (sr) and the remainder (1-fr) dissolves at a slow rate (ss). Dissolved matter is assumed to absorb from the respiratory tract to blood instantaneously if a bound state can be neglected (which is the case for uranium) (ICRP 2017).

In nuclear fuel fabrication, occupational inhalation exposure to uranium aerosols must be considered. For manufacturing of fuel for light-water reactors, exposure to several uranium compounds is expected and varies depending on the production method used. The calculation of CED and Hlung requires knowledge about the absorption parameters (fr, sr, ss). The absorption parameters are known to vary with different chemical compounds of uranium. The ICRP gives default absorption parameters for a variety of chemical compounds; e.g., uranium dioxide (UO2) and triuranium octoxide (U3O8) are considered Type M/S (intermediate moderate/slow absorption type) in the latest ICRP recommendations (ICRP 2017). The default absorption parameters are fr = 0.03, sr = 1 d−1, ss = 5×10−4 d−1. However, data on these parameters, derived from in vitro experiments, animal experiments and human contamination cases, indicate large variability for a given uranium compound (Davesne and Blanchardon 2014). In the review article by Davesne and Blanchardon (2014), results for ss for UO2 ranged from about 5 × 10-5 d−1 to about 5 × 10−3 d−1 (mean and median of 9.1 × 10−4 d−1 and 5.6 × 10−4 d−1, respectively). Process history of material (e.g., calcination temperature), experimental conditions, and model used for fitting were mentioned as causes of variability. It is desirable to determine material-specific absorption parameters to improve dose calculations (ICRP 2002).

The absorption parameters (fr, sr, ss) can be estimated in vitro by studying the dissolution of sampled particles in simulated lung fluid. Typically, collected particles are sealed within a filter membrane “sandwich” and allowed to dissolve in simulated lung fluid for a certain amount of time, after which the lung fluid is collected and replaced (Ansoborlo et al. 1999). The absorption parameters (fr, sr, ss) for the specific material can be derived by fitting a two-component exponential function to the undissolved fraction over time (eqn 1):


where U(t) and U0 are the amounts of radioactivity at time t and 0, respectively. The equation only describes removal of material by dissolution (instantaneous uptake to blood is assumed). The equation contains no information about particle clearance through other mechanisms, uptake rate of dissolved material to blood, or the presence of a bound state. In this type of experiment, the absorption parameters thus represent dissolution only. In the present work, the term absorption parameter is used for consistency with the ICRP terminology.

Numerous experimental setups and compositions of simulated lung fluids have been proposed over the years (Ansoborlo et al. 1999; Marques et al. 2011). The earliest published composition of simulated lung fluid is the Gamble’s solution (Gamble 1967). Concentrations of phosphates, carbonates, and chelating agents are likely to be important in dissolution tests (Ansoborlo et al. 1999). Typically, a static experimental setup is used for simplicity, although flow-through systems have been evaluated as well. In vitro experiments are considered useful to screen compounds at a workplace and provide approximations of absorption parameters, although in vivo data are preferred, if available (Ansoborlo et al. 1999). Similar experimental set-ups have been used to study dissolution characteristics in other solvents as well; e.g., Eidson and Mewhinney (1984) studied the dissolution of mixed-oxide fuels in 0.1M HCl in addition to simulated lung fluid.

Phagocytosis by alveolar macrophages is important for particle clearance from the respiratory tract, and several studies have used alveolar macrophages in vitro (Guilmette 1998; Ansoborlo et al. 1999). Ansoborlo et al. (1998) demonstrated that a culture medium with macrophages better represented in vivo data of rats following intratracheal instillation of UO4. Another study reported that phagocytosis occurs rapidly; e.g., more than 90% of lavageable particles (PuO2 and iron oxides) were found in macrophages by 24 h (Guilmette 1998). A review by Oberdörster et al. (2005) reported that approximately 80% of 0.5–10 μm particles were retrieved by 24 h in alveolar macrophages from rats exposed to different particles. For particles <0.1 μm, approximately 20% were retrieved by 24 h. Phagocytosis by macrophages can also occur in the alveolar interstitium (Liegeois et al. 2018). After formation of phagolysosomes, engulfed particles are exposed to a lower pH (about 5), proteolytic enzymes and oxygen radicals (Kreyling 1992). The different chemical environment in the phagolysosome is associated with different particle dissolution characteristics. Stefaniak et al. (2005, 2006) developed a simulated phagolysosomal fluid to study the dissolution of beryllium aerosols. Regarding uranium particles specifically, it has been shown that uranyl phosphate needles can form in the phagolysosome (Hengé-Napoli et al. 1996; Berry et al. 1997). Macrophages are clearly important for particle clearance but it can, however, be challenging to maintain normal functionality of macrophages in vitro, and systems without macrophages are therefore typically used in health protections programs (Ansoborlo et al. 1999).

Following inhalation exposure to aerosols, the spatial deposition of particles in the respiratory tract depends on the particle size distribution. This is of great importance since it affects Hlung, CED, and the urinary excretion pattern. For aerosols with an activity median aerodynamic diameter (AMAD) of 15 μm, approximately 3% of the total inhaled activity is expected to deposit in the alveolar region, whereas the corresponding figure for an AMAD of 5 μm is 10% (ICRP 2015). This results in different dose coefficients, 3.6 and 5.5 μSv Bq−1 for AMAD 15 μm and 5 μm, respectively (for Type M/S material and 234U inhalation intake). The urinary excretion rate (modeled using Taurus software, Base version 1.0; UK Health Security Agency, Didcot, Oxon, UK) is approximately a factor of 2 to 4 times lower (time-dependent) for the AMAD 15 μm compared to the AMAD 5 μm scenario.

Since the surface-to-volume ratio increases for small particles, one can suspect that small particles reaching the alveolar region will dissolve at a more rapid rate than large particles (Mercer 1967). In the HRTM, absorption is regarded as a two-stage process (dissolution/dissociation and uptake) (ICRP 2015). Uptake of dissolved material to blood is assumed to occur instantaneously (if the bound fraction is assumed to be zero, which is the case for uranium) in all regions of the respiratory tract (except the anterior nasal passage, ET1, where no absorption is assumed). It is recognized, however, that the absorption likely is faster in the alveolar region due to a thinner air-blood barrier (not accounted for in the model) (ICRP 2015). Particles in the bronchiolar compartment (ciliated airways) are transported to the bronchial compartment at a rate of 0.2 d−1 (corresponding half-time 3.5 d) (ICRP 2015). This is considerably faster than the transport rate of 0.002 d−1 (corresponding half-time 350 d) from the alveolar to the bronchiolar compartment. Thus, one can speculate that dissolution of particles that deposit in the alveolar region should be given more weight to the overall absorption than particles that deposit in other regions of the respiratory tract. Thus, dissolution rate experiments using only the smallest particle size fraction might yield more representative absorption parameters.

The importance of particle size in dissolution can be illustrated by revisiting the work of Mercer (1967), where the mass m of a single particle at time t is given as (eqn 2):


where m0 is the initial mass or the particle, αs, and αv are the surface and volume shape factors, respectively; k is a proportionality constant (mass per unit area per unit time); ρ is the particle density; and D0 is the initial particle diameter. Fig. 1 illustrates the theoretical dissolution of monodisperse particles (assuming smooth surfaces and spherical shape). A value of k was taken to be 1 × 10−7 g cm−2 d−1 [assumed to be constant and derived from the paper by Mercer (1967) by assuming parameter values close to the Type M/S material, i.e., ρ = 3 g cm−3, and a population of particles with biological half-time of 1,400 d and median diameter of 5 μm] (Mercer 1967).

Fig. 1:
Theoretical calculation of the undissolved fraction over time for monodisperse particles of various sizes.

From Fig. 1 it can be seen that particle size can be of great importance if the total mass contained in small (sub-micron) particles is considerable. The question arises regarding which size-range has the highest impact on lung dissolution rates. While large particles dominate the total mass, small particles are expected to dissolve more rapidly (as illustrated in Fig. 1). In the HRTM, large particles are less likely to deposit in the alveoli compared to small particles, and particles with aerodynamic diameters around 0.01–0.02 μm and 1–2 μm are of particular interest since they have deposition maxima in the alveolar region (ICRP 1994). Previous work at the site has demonstrated a very wide range of uranium aerosol sizes (below 0.1 μm up to well above 10 μm geometric diameter) using electron microscopy (Hansson et al. 2017). The standard assumption of 5 μm AMAD and a geometric standard deviation (GSD) of 2.5 covers a very wide range of particle sizes: 95% of the total activity is contained in particles between 0.8 μm and 31 μm (Hinds 1999). Thus, size fractionation (e.g., by using cascade impactors) might be appropriate for in vitro experiments in simulated lung fluid compared to using bulk samples where large particles dominate the mass.

However, experiments on the effect of particle size on dissolution rates in simulated lung fluid (and thus lung absorption parameters) are scarce in the literature. Guilmette and Cheng (2009) studied in vitro dissolution rates of depleted uranium aerosols collected using 5-stage cyclone samples (cut-points 7.9, 3.3, 2.3, 1.2, and 0.7 μm). The authors found that the dissolved fraction generally increased with lower cut-points, although exceptions were noted. Another study investigated the effect of the specific surface area (SSA) (range 0.7–15.5 m2 g−1) of U3O8 samples and found that the fr as well as the ss increased for samples with higher SSA (Chazel et al. 1998). However, the effect was as not clear for UO2 materials (SSA range 1.1–4.5 m2 g−1) (Chazel et al. 2000).

In the present work, the effect of aerodynamic particle size on dissolution in simulated lung fluid was investigated. This is important in order to learn about the role of particle size in the HRTM. Uranium aerosols were collected during work at the major workshops at a nuclear fuel fabrication plant. To obtain samples representative for worker exposure, the operator breathing zone at the main workshops was sampled using portable cascade impactors, accomplishing aerodynamic size fractionation. Lung absorption parameters as expressed in the HRTM were evaluated for different aerodynamic sizes (i.e., different impactor stage cut-points). The work was carried out in the form of a main experiment and two supplementary experiments to study the potential impact of pH drift and adsorption of uranium material on filter holders. In addition, the effect on the curve-fitting procedure used to determine the absorption parameters was evaluated depending on the experimental duration.


Aerosol sampling

The present study was carried out at a nuclear fuel fabrication plant run by Westinghouse Electric Sweden AB and producing fuel for light-water reactors. Operations with potential risk of internal exposure are carried out at four workshops. At the conversion workshop, batches of low-enriched (up to 4.95% 235U enrichment) uranium hexafluoride (UF6) are converted via ammonium uranyl carbonate (AUC) through a wet-chemical process and reduced to UO2 in fluidizing bed furnaces. In addition to the main processes, various side-processes associated with other uranium compounds exist. At the powder preparation workshop, various processes including blending of uranium powders, powder milling, and oxidizing of waste materials to U3O8 are carried out. The UO2 powder is pressed, sintered, ground, and inspected at the pelletizing workshop. In addition, burnable absorber (BA) fuel (addition of neutron absorber, gadolinium oxide, Gd2O3) is produced at a separate pelletizing workshop. The processes at the plant have been described in more detail in previous work (Hansson et al. 2017).

Previous work at the plant has shown bimodal particles size distributions, with 75–88% of the radioactivity in uranium aerosols at the plant workshops associated with an AMAD of 15–19 μm (depending on work tasks), i.e., a coarse particle size distribution (Hansson et al. 2020). Remaining activity (referred to as a fine fraction) was associated with an AMAD of 2–7 μm. In cascade impactors, aerodynamic fractionation is accomplished through a series (cascade) of impaction stages where an airstream is directed toward an impaction plate through a nozzle, forcing it to a 90-degree bend. Particles with small aerodynamic diameters align with the airstream more easily than large particles, which will attach by impaction onto the plate (which can be prepared with an impaction substrate). Each impaction stage is associated with a cut-point that corresponds to the aerodynamic diameter that has a 50% probability of impaction (Hinds 1999).

In the present work, 8-stage Marple 298 impactors (Prod. No. SE298; Thermo Scientific, Waltham, MA) operated at 2.0 L min−1 using Gilian 5000 pumps (Sensidyne, St. Petersburg, FL) were used. The impactor/pump system is portable, enabling sampling in the operating breathing zone throughout the workday, allowing sampling more representative of actual exposure compared to, for example, static air sampling. Air flow was checked before and after each sampling by a rotameter calibrated with a MB-50 SLPM-D orifice flow controller (Alicat Scientific, Tucson, AZ). Flow rate uncertainties were estimated to ±0.05 L min−1. Impaction stage cut-points were 21.3, 14.8, 9.8, 6.0, 3.5, 1.6, 0.9, and 0.5 μm. The last impaction stage was followed by a final collection stage, collecting remaining particles not captured by impaction. Prior to sampling, a mixed cellulose ester (MCE) (SEC-290-MCE; Thermo Fisher Scientific, Waltham, MA) impaction substrate was attached to each impaction plate. Polyvinyl chloride (PVC) filters (Thermo Fisher Scientific, SEF-290-P5) were used for the final collection stage. Impactor stages were not coated with an adhesive to avoid particle bounce, as this could distort the dissolution rate experiments. The reasoning behind this was that particles could be partially or entirely embedded in adhesive coating, thus making the particle surface area less available for contact with the simulated lung fluid and consequently making interpretation of results difficult.

Sampling campaigns with cascade impactors were carried out in the operator breathing zone at the different workshops (Table 1). Each sampling campaign was designed to collect sufficient amounts of uranium for the dissolution rate studies but avoid particle overload on the impactions stages (which can lead to increased particle bounce-off). Time periods for sampling were determined based on previous knowledge of typical levels of airborne uranium (Hansson et al. 2020). Sampling was carried out during work in the main production lines during normal production conditions. Two to four different operators participated in each round of sampling to increase sampling representativeness (e.g., reduce the effect from different work methodology). Sampling campaigns were carried out in the conversion, powder preparation, pelletizing, and BA pelletizing workshops, respectively. Active sampling times (Table 1) vary since time on the workshop floor varies between the different workshops, and sampling was not carried out during administrative work in the control rooms. Following each sampling campaign, filters were stored in vacuum-sealed containers in order to minimize oxidation of particles.

Table 1 - Description of the breathing zone cascade impactor sampling.
Sample-ID Workshop Sampling time period Active sampling time (h) Description of main work tasks
74 Conversion 48 h 6.4 Changing UF6 cylinders, UO2 powder homogenization, various sampling and transportation of different uranium materials.
76 Powder preparation 4 d 3.1 Powder milling, oxidizing of discarded pellets, oxidizing of waste material from pellet grinding, humidity check of uranium powder, powder blending and powder transportation.
72 Pelletizing 48 h 22.7 Pellet inspection and work with pellet grinders.
73 BA pelletizing 5 d 21.7 Powder pressing, work with the pellet grinder and pellet inspection

Dissolution rate experiments

Several experimental setups for in vitro dissolution of particles in simulated lung fluid have been published over the years (Ansoborlo et al. 1999; Marques et al. 2011). In the present work, Gamble’s solution (Table 2) was chosen, as it has been widely used in studies to simulate extracellular lung fluid (Ansoborlo et al. 1999). Each time new simulated lung fluid was prepared, CO2 was bubbled through the solution until a pH of 7.3 was reached.

Table 2 - Ingredients of the simulated lung fluid based on Ansoborlo et al. (1999).
Compound Quantity (mol L −1 )
NaCl 0.116
NH4Cl 0.010
NaHCO3 0.027
Glycine 0.006
L-Cystéine 0.001
Na3 Citrate 0.0002
CaCl2 0.0002
NaH2PO4 0.0012

Each impaction substrate and final collection filter was cut in half and placed in a “filter sandwich.” The other half of the sample was intended to be used for experiments on dissolution rates in simulated gastric and intestinal fluids (under preparation in a separate manuscript). The filter sandwich was prepared by placing each sample (impaction substrate or final collection filter) between two membrane filters (Supor-200, 0.2-μm pore size, 47-mm-diameter, P/N 60402; PALL Life Sciences, Westborough, MA). The membrane filters were sealed using custom-made filter holders (polysulfone, Quadrant 1000 PSU material; Total Plastics International, Kalamazoo, MI). Each sample was placed in 100 g of simulated lung fluid in a plastic container, which was sealed tight to avoid evaporation and placed in a laboratory oven at 37 °C (Venticell, MMM Medcenter Einrichtungen GmbH, Munich, Germany). At pre-determined time points (1 h, 3 h, 6 h, 9 h, 1 d, 2 d, 4 d, 8 d, 12 d, 16 d, 20 d, 26 d, 35 d, 50 d, 70 d, and 100 d), the filter sandwich was transferred to a new container with 100 g of fresh simulated lung fluid. Experimental durations in the literature vary, with times up to 100 d (Kravchik et al. 2008). To the best of our knowledge, no recommendations on experimental duration exist, and it has not been described whether experimental duration might affect the absorption parameters (e.g., if the ss is low, too short an experimental duration could introduce increased uncertainty in determination of the slow component).

After each sample transfer, the collected liquid sample was conserved by adding 1 mL of concentrated HNO3 (analytical grade). Liquid samples were analyzed by inductively coupled plasma mass spectrometry (ICP-MS).

Following the final liquid sample extraction, the filter sandwich was removed from the filter holder for sample preparation and alpha spectrometry. The filter holder was rinsed with 2% HNO3 (analytical grade) to remove any uranium that could have attached to the surfaces. In addition, the filter holders were leached in 2% HNO3 to ensure that no uranium remained after rinsing. However, it was noted that some uranium remained, which justified supplementary experiments (described below).

Supplementary experiments: repeated dissolution rate experiments

For the samples described in Table 1, it was noted that the pH drifted from 7.3 to about 7.6–7.7 (up to 8 or higher for late time-points) while the samples were in the laboratory oven. To evaluate the effect of pH drift, a less extensive version of the experiment was repeated. Breathing zone sampling was repeated at the pelletizing and BA pelletizing workshops (Appendix A).

Impaction substrates from every two impaction stages (cut-points: 21.3, 9.8, 3.5, and 0.9 μm) and final collection filters were evaluated in parallel in two experimental setups. One third of each substrate was evaluated in the laboratory oven and one third in a laboratory incubator (INCO 2, Memmert GmbH, Schwabach, Germany) allowing for 5% atmospheric CO2 in the chamber (the laboratory incubator was not available during the main experiment, in which case only the laboratory oven could be used). The remaining third was saved for experiments on dissolution rates in simulated gastric and intestinal fluids (under preparation in a separate manuscript).

For the repeated experiment, the pH for samples in the incubator was stable over time at 7.6 ± 0.1, whereas the pH for samples in the laboratory oven drifted from 7.7 ± 0.1 to 8.1 ± 0.2. This was higher than in the previous round of experiments, probably since the initial pH was not manipulated by bubbling CO2 through the sample (this was intentional, as to allow as straightforward a comparison between laboratory oven and incubator as possible).

The plastic containers in the incubator were open to allow for gas exchange, which led to approximately 1% evaporation loss per 24 h in in the chamber. Samples were collected at 1 h, 3 h, 6 h, 9 h, 1 d, 2 d, 5 d, 8 d, 16 d, 26 d, and 40 d.

Supplementary experiments: uranium adsorption on filter holders

Following the main experiment, the filter holders were leached in 2% HNO3 (analytical grade) to verify a low adsorption of uranium onto the filter holders. The solution was prepared and analyzed using alpha spectrometry. A negligible adsorption was expected onto the filter holder (polysulfone material), but the levels were considerable compared to the total amounts detected in the liquid samples. The median ratio of uranium on the filter-holder to the total dissolved amount in liquid samples was 0.29 (Q1–Q3 range: 0.12–1.84). Similar observations have been made elsewhere; e.g., Sdraulig et al. (2008) noted a significant fraction of uranium activity in a precipitate on filter holders and in the calculations allocated the material to the dissolved fraction proportionally to concentration. Cera et al. (2000) discovered the formation of a secondary solid phase of uranium following dissolution of UO2 pellets in simulated saline water.

Our observation could be described by dissolved material attaching to the filter holder (e.g., as a precipitate) or by particle leakage from the membrane filter. We decided to further investigate this. By preparing a solution with 235U in solution only, particle leakage can be safely disregarded. Thus, if uranium can be detected on the filter holder or membrane filter, it can originate from formation of a precipitate (other than liquid soaking into the membrane filter and droplets on the filter holders, but this would be expected to be constant).

Samples (15 in total) of fresh simulated lung fluid were prepared, and 0.3 μg of 235U from a certified uranium solution (Eckert & Ziegler Isotope Products, Valencia, CA; reference number 1263-94) was added to each sample. The mass corresponded to approximately the maximum mass of uranium dissolved in any liquid sample for the experiments of samples described in Table 1. The samples were placed in the incubator using the same settings as for the repeated dissolution rate experiment (previous section). Samples were collected at about 1 h, 2 h, 3 h, 4 h, 6 h, 21 h, 2 d, 4 d, 6 d, 8 d, 11 d, 15 d, 22 d, 32 d, and 39 d. Procedure blank samples were collected at 2 h and 11 d. Upon collection, the filter sandwich was immediately removed from the solution, and the membrane filter was removed from the filter holder. The membrane filter was acid digested in concentrated HNO3 and aqua regia following addition of 232U yield determinant (Isotrak, AEA Technology, PLC, Didcot, UK). The filter holder was leached in 2% HNO3 following addition of 232U yield determinant. The samples were analyzed using alpha spectrometry.

Analysis: ICP-MS

All sample solutions in the main experiment, including the two procedure blank series (i.e., leaching of a blank MCE impaction substrate and a PVC final collection filter) were analyzed using ICP-MS (Agilent 8800 Triple Quad; Agilent Technologies, Santa Clara, CA). When needed, samples were diluted to ensure analysis in pulse count mode only, avoiding calibration uncertainties between count and analogue mode. A 233U tracer calibrated against a 232U standard solution (National Physical Laboratory, Teddington, UK; product code R20-02) was used. Samples from the repeated experiments (section: Supplementary experiments: repeated dissolution rate experiments) were analyzed using an Agilent 8900 Triple Quad ICP-MS in single quadrupole mode (Agilent Technologies). A cross-comparison between the two ICP instruments was carried out by analyzing 40 samples from the main experiments on both instruments. The total amount of 238U agreed well (median ratio 0.997 with Q1–Q3 range: 0.965–1.088).

Mass numbers 232, 233, 235, and 238 were analyzed. Mass number 234 was analyzed occasionally, but levels were generally too low and would have required a different sample preparation. The dwell time was typically set to 0.1 ms for each mass number, and the number of replicates per mass was 4–5. A rinse protocol (3% supra pure HNO3) was run between each sample to minimize the contribution from sample residue in the instrument (e.g., tubes). In addition, acid blank samples were run regularly (typically every fourth sample) to determine the machine background level. Uranium standards (certified reference material 112-A and U500; New Brunswick Laboratory, New Brunswick, NJ) were run regularly to monitor mass-bias and ensure instrument stability.

Mass number 232 was used to evaluate the presence of 232Th, since the hydride 232Th1H+ can affect the count rate at mass number 233. The median count rate ratio for mass numbers 233 and 232 was 1.4 (Q1–Q3 range: 0.7–2.2) for all samples. The maximum ratio was 19. Shi et al. (2013) have demonstrated that the 232Th1H+/232Th+ ratio was 6.6 × 10−5 for a 232Th standard solution. For samples in the present work, the contribution from 232Th1H+ was thus considered negligible.

The following data processing was carried out to obtain the total amounts of 238U in each liquid sample:

  1. Replicates were evaluated for potential outliers (that might be present due to sample heterogeneities, which can cause sampling and injection of nanometer-sized particles in the plasma). Outliers identified using Grubb’s test (0.05 significance level) were excluded (Grubbs 1969);
  2. Count rates for each sample and mass number were corrected by subtracting the average instrument background level before and after sample analysis. When more than one consecutive sample had been analyzed, linear interpolation was used; and
  3. The final 238U result was calculated using the 238U/233U ratio and the known amount of tracer in the sample. Corrections taking into account tracer interference for mass number 238, procedure blank levels, and aliquot fractions were carried out.
  4. Only the data for 238U were used in the present work.

Analysis: alpha spectrometry

Filter samples were acid digested in conc. HNO3 followed by aqua regia after addition of 232U yield determinant (Isotrak, AEA Technology, PLC, Didcot, UK). Samples were prepared by radiochemical separation (Holm 1984) and electrodeposition onto planchettes (Hallstadius 1984). Alpha spectrometry measurements (analysis using ICP-MS would have introduced uncertainties due to required dilution of approximately a factor 500) were carried out using passivated implanted planar silicon detectors (Canberra PIPS, Mirion Technologies Inc, San Ramon, CA), ORTEC (Oak Ridge, TN) Octête MCA system, and ORTEC Maestro software. Counting times varied between 2–33 d.

A comparison of the alpha spectrometry and ICP-MS analysis was carried out by analyzing 8 liquid samples with both methods. The results agreed within 10% (mean ratio 0.93 ± 0.11 for 238U, only taking into account counting statistics), which was considered satisfactory.

Data evaluation and statistical analysis

Data for 238U were used to evaluate the undissolved fraction as a function of time (eqn 1). U0 was defined as the sum of (1) 238U in all liquid samples, (2) 238U in the filter sandwich at the end of the experiment, and (3) 238U after additional leaching of the filter holder. In addition to the undissolved fraction, the dissolution rate [i.e., 238U mass (μg) in the sample divided by the sample time (d) in solution] was evaluated. For 20 liquid samples (576 in total) among 15 filter series (36 in total), unexpected increases in the calculated dissolution rate that did not follow the general trend of the data were observed. These samples were considered outliers, and the dissolved amount of 238U was approximated based of the dissolution rates of adjacent samples instead. This is further discussed in the Uncertainties section.

Lung absorption parameters (fr, sr and ss) were derived by fitting experimental data to eqn (1) using a nonlinear regression model fit in MATLAB (version R2018b). A statistical evaluation of uncertainties was carried out by random sampling from assumed normal distributions of the ICP-MS and alpha spectrometry analyses, as well as the tracers used (3% and 2% for the 233U and 232U tracers, respectively, with assumed rectangular distributions). Uncertainties due to weighing were considered low in comparison to the other uncertainties and thus not considered. Simulations and curve-fits were repeated 5,000 times for each analyzed impactor stage [10,000 simulations per sample gave the same results for parameters (within 0.1%) and standard deviations (within 1%) based on a comparison of 5 randomly selected samples]. Lung absorption parameters were expressed as the mean with its standard deviation.

Dosimetric evaluation

It has been demonstrated that activity size distributions at the plant can be described with a bimodal size distribution (fine and coarse fraction), with some variations across the different workshops (Hansson et al. 2020). For the dosimetric evaluation, it was thus assumed that impactor stages with cut-points greater than 5 μm (i.e., the first four stages) represent the coarse fraction, and remaining impactor stages + final collection stage represent the fine fraction.

Dose coefficients (committed effective dose, CED and lung equivalent dose, Hlung) for each workshop were evaluated using the Taurus internal dosimetry software. The ICRP 130 particle transport model was used to simulate inhalation intake of 1 Bq of 234U (which typically constitutes >80% of the uranium activity at the plant). The average lung absorption parameters for the coarse and fine fraction, respectively, were used. The alimentary tract transfer factor (fA) was calculated as 0.02 × fr (ICRP 2017). An AMAD of 15 and 5 μm was used for the coarse and fine fraction, respectively. Default geometric standard deviation (GSD = 2.5), aerosol density (3 g cm−3), and shape factor (1.5) were used, as they were not possible to modify in the software.


Activity measurements

Radiometric data of undissolved uranium on the filter sandwiches (membrane filter with either impactor substrate or final collection filter), as determined by alpha spectrometry, are presented in Appendix B. The 238U activity was converted to mass using a specific activity constant of 12,436 Bq g−1 (, Nuclid Datasheet ++ application). Levels ranged between 0.05-4 μg 238U. Procedure blank levels were 0.01–0.03 μg 238U.

Levels of 238U in liquid samples for ICP-MS analyses ranged between 2–4,000 pg mL−1 (procedure blank levels: 3 ± 1 pg mL−1). About 1% of the samples (575 in total) were below the average procedure blank level, and 17% of the samples were below two times the average procedure blank level. The majority of these samples originated from four different sample series where at least 8 samples per series (16 in total) were below double the average procedure blank level. The affected series were the 0.5 μm stage at the conversion, pelletizing, and BA pelletizing workshop, and the 0.9 μm stage at the BA pelletizing workshop. This was due to the low sampling flow rate (2 L min−1), which resulted in insufficient material during the time of sampling for the sub-micron stages, which were associated with a small fraction of the total collected mass. Thus, derived lung absorption parameters for these series should be viewed with caution.

Lung absorption parameters

Undissolved fractions (stages with cut-points 21.3, 3.5, and 0.9 μm) for 0–100 d are given in Fig. 2. The full figures (all impaction stages) are given in Appendix C, Fig. C1.

Fig. 2:
Undissolved fraction of 238U for three impaction stages for each workshop. The most relevant ICRP default absorption types are included for comparison. Error bars correspond to ±1 standard deviation. Samples for the 0.9-μm fraction for the BA pelletizing workshop were frequently close to the detection limit (i.e., 238U levels close to procedure blank samples), and thus the 1.6-μm fraction was selected instead.

From Fig. 2 it appears that the dissolution characteristics can differ between impaction stages, which is expected since the SSA of small particles can be expected to be higher. However, many exceptions from the relationship can be seen; i.e., a lower cut-point does not always indicate a lower undissolved fraction at the end of the experiment (Appendix C, Fig. C1 and Fig. 3). Several explanations are plausible, e.g., the results may be due to the presence of multiple size distributions of particles with different properties (e.g., chemical compound, porosity, AMAD, and density). This is discussed in further detail in the Uncertainties section. A previous study noted a similar pattern (Guilmette and Cheng 2009). Interestingly, the undissolved fraction at the end of the experiment was higher for the final collection filter compared to the final impaction stage (cut-point 0.5 μm) for some workshops (Appendix C, Fig. C1 and Fig. 3). We interpret this as a result of particle bounce in the impactor, i.e., large particles that have bounced off the impaction stages and reached the final collection filter. Signs of particle bounce have been observed for the 0.9 μm and 0.5 μm impaction stages and the final collection filter (Hansson et al. 2020). As a result, absorption parameters derived from these stages might be less credible (e.g., underestimating the rapid fraction) than parameters derived from earlier impaction stages.

Fig. 3:
Absorption parameters as a function of impaction stage cut-point for the different workshops. The final collection filter does not have a cut-point, but a value of 0.25 μm was assigned to allow for inclusion in the figure. The most relevant ICRP default recommendations are presented (dashed lines) for comparison. Error bars correspond to ±1 standard deviation. The y-axis for slow rate, conversion, was cut due to a highly uncertain datapoint for the 0.5-μm cut-point stage (0.02 ± 0.21 d 1). The 0.5-μm stage at the conversion, pelletizing, and BA pelletizing workshop, and the 0.9-μm stage at the BA pelletizing workshop should be viewed with caution due to several samples close to procedure blank levels.

Absorption parameters per impaction stages and workshops are presented in Fig. 3. It is evident that the conversion workshop has the highest fr, indicating a presence of uranium materials that are soluble in the respiratory tract. This is expected due to the variety of different uranium compounds handled at the workshop. The fr appears to range between 0.2–0.8, suggesting material intermediate to Type M and Type F/M, with a sr more resembling Type F. The ss, however, appears to be close to Type M/S material. It is likely that the observations correspond to a combination of uranium compounds that are soluble (e.g., AUC) and relatively insoluble (e.g., UO2 powder) in the respiratory tract.

The remaining workshops handle UO2 and/or U3O8 in various forms (e.g., powder and/or sintered pellets). The absorption parameters at the powder preparation workshop appear to agree well with the ICRP default absorption type (Type M/S), with the exception of sr, which is closer to Type M or Type F absorption types. At the pelletizing workshops, the absorption parameters are closest to Type S absorption type (with the exception of sr). This is probably due to a large fraction on particles originating from sintered material (e.g., work near pellet grinders and pellet inspection), which has a lower SSA compared to the powder material. The sr resembled Type M or Type F materials. It is worth noting that in the previous ICRP recommendations, the sr was 100 d−1 for all absorption types.

In general, the correlation between absorption parameters and impaction stages appears to be somewhat inconsistent. This could indicate that the effect of aerodynamic size is of secondary importance compared to other variations (e.g., multiple size distributions and chemical compounds, further discussed in the Uncertainties section). It could also be a matter of SSA. If the SSA is large (i.e., porous particles), the particle size might be of secondary importance. In Fig. 3, a correlation between absorption parameter and impactor cut-point appears to be present for the ss at the pelletizing workshop. The particles are expected to originate predominantly from sintered material with low SSA; hence, the particle size could be of greater importance.

Dosimetric evaluation

Dose coefficients for assumed coarse (impactor stage cut-point ≥5 μm) and fine (impactor stage cut-point <5 μm) fractions for each workshop are presented in Table 3.

Table 3 - Dose coefficients (CED and Hlung per unit of activity intake) based on derived absorption parameters for each workshop, following inhalation of 234U. To allow for easier comparison, the same AMADs (15 and 5 μm) were used for all workshops, although some variations exist (Hansson et al. 2020). Absorption from the alimentary tract (fA) was assumed to be 0.02 × fr. Default GSD, density and shape factor was assumed, as they could not be modified in the Taurus software. The ICRP 130 particle transport model was used. Recommended ICRP parameters for a selection of absorption types are included for comparison (ICRP 2017). Numbers in brackets indicate 1 standard deviation.
AMAD = 15 μm 2 AMAD = 5 μm
fr (-) sr (d 1) ss (d 1) fA (-) CED
(μSv Bq 1)
Hlung 3
(μSv Bq 1)
fr sr (d 1) ss (d 1) fA (-) CED
(μSv Bq 1)
Hlung 3
(μSv Bq 1)
Conversion 0.29 (0.09) 28 (12) 6.3 × 10 4 (7.2 × 10 5) 5.8 × 10 3 (1.7 × 10 3) 1.8 6.8 0.53 (0.18) 20 (13) 1.6 × 10 3 (1.2 × 10 3) 1.1 × 10 2 (3.6 × 10 3) 1.7 9.2
Powder preparation 0.025 (0.002) 5 (1) 5.3 × 10 4 (8.6 × 10 5) 5.0 × 10 4 (3.8 × 10 5) 2.6 10 0.033 (0.005) 9 (5) 5.2 × 10 4 (2.5 × 10 4) 6.7 × 10 4 (9.2 × 10 5) 5.4 32
Pelletizing 0.008 (0.002) 13 (6) 5.5 × 10 5 (4.2 × 10 5) 1.7 × 10 4 (4.1 × 10 5) 6.5 29 0.01 (0.003) 9 (3) 1.3 × 10 4 (8.7 × 10 5) 2.0 × 10 4 (6.4 × 10 5) 12 68
BA-pelletizing 0.018 (0.003) 4 (1) 2.1 × 10 4 (3.3 × 10 5) 3.6 × 10 4 (6.6 × 10 5) 3.9 16 0.024 (0.006) 6 (3) 1.4 × 10 4 (1.1 × 10 4) 4.7 × 10 4 (1.2 × 10 4) 11 65
ICRP Type F 1 10 0.02 0.16 0.1 1 10 0.02 0.25 0.23
ICRP Type F/M 0.8 1 1.0 × 10 2 1.6 × 10 2 0.18 0.7 0.8 1 1.0 × 10 2 1.6 × 10 2 0.41 2.1
ICRP Type M 0.2 3 5.0 × 10 3 4.0 × 10 3 0.65 2.9 0.2 3 5.0 × 10 3 4.0 × 10 3 1.4 8.9
ICRP Type M/S 0.03 1 5.0 × 10 4 6.0 × 10 4 2.6 10 0.03 1 5.0 × 10 4 6.0 × 10 4 5.5 33
ICRP Type S 0.01 3 1.0 × 10 4 2.0 × 10 4 5.4 24 0.01 3 1.0 × 10 4 2.0 × 10 4 13 78
1The 0.5 μm stage was excluded due to substantial uncertainty (Fig. 2)
2Dose coefficients were averaged for AMAD = 10 μm and 20 μm, as 15 μm was not an available option in the software version used.
3Averaged for male and female

When exposure to uranium aerosols is evaluated, bioassay data (e.g., urine samples) are typically used. Although in vitro experiments using simulated lung fluid represent a simplified version of the respiratory tract, it might be justified to adjust certain parameters if this gives a better fit to the bioassay data (Marsh et al. 2005). Indeed, previous work has recommended conducting in vitro experiments but as a complement to in vivo data (Chazel et al. 2001).

Supplementary experiments: repeated dissolution rate experiments

Absorption parameters following the repeated dissolution rate experiments were difficult to evaluate due to very low levels of 238U in the samples. About 20% of the samples (220 in total) were below the average procedure blank level, and about 50% of the samples were below two times the procedure blank level. This introduces significant uncertainties when subtracting the procedure blank levels. Furthermore, it was noted that samples (including procedure blanks) collected at times 2 and 5 d showed higher than expected dissolution rates (μg 238U d−1) (i.e., an increased dissolution rate compared to adjacent samples, which could indicate cross-contamination). The amount of 238U in these samples and two additional samples showing signs of particle penetration was approximated based on the dissolution rates of adjacent samples.

One impactor sample (cut-point 9.8 μm, sampling series 81, Table A1) generated simulated lung fluid samples with 238U levels above the procedure blank level throughout the entire series for both laboratory oven and incubator. The undissolved fraction over time is shown in Fig. 4 and derived absorption parameters in Table 4.

Table A1 - Details regarding sampling for the supplementary experiments (repeated dissolution rates).
Sample-ID Workshop Sampling time period Active sampling time (h) Description of main work tasks
82 Pelletizing 48 h 22.5 Pellet inspection and work with pellet grinders.
81 BA pelletizing 5 d 12.8 Powder pressing, work with the pellet grinder and pellet inspection

Fig. 4:
Comparison of the undissolved fraction for a 9.8-μm cut-point impactor stage following sampling at the BA pelletizing workshop. Error bars correspond to ±1 standard deviation.
Table 4 - Comparison of absorption parameters for incubator and laboratory oven for the 9.8-μm cut-point impactor substrates, collected at the BA pelletizing workshop, in simulated lung fluid. Numbers in brackets indicate 1 standard deviation.
Incubator Laboratory oven
Rapid fraction, fr (-) 0.022 (0.001) 0.027 (0.002)
Rapid rate, sr (d 1) 0.70 (0.06) 0.65 (0.07)
Slow rate, ss (d 1) 1.3 × 10 4 (0.1 × 10 4) 1.3 × 10 4 (0.2 × 10 4)

The results in Table 4 and Fig. 4 indicate similar results regardless of whether the incubator or laboratory oven was used. This indicates that the pH variations observed in the main experiment (samples in Table 1) are of secondary importance. The fr in Table 4 is similar to what is presented in Fig. 3 and Table 3 for the BA pelletizing workshop, whereas the sr and ss are lower. Variations in day-to-day work tasks, and thus sampled aerosols, is a plausible explanation for the observed differences. It is also noted that the curve-fit for early samples in Fig. 4 is poor for early samples, which affects the fr and sr parameters.

Supplementary experiments: uranium adsorption on filter holders

As previously mentioned (Materials and Methods), leaching of filter holders from the main experiment revealed elevated levels of uranium. The supplementary adsorption experiments using a 235U solution indicated that this effect was very small: 0.4% of the uranium in solution was attached to the filter holder, and 0.4% to the membrane filter. Thus, <1% of uranium in each liquid sample can be expected to be lost to the filter holder and membrane filter sandwich. These losses corresponded mainly to the uptake of liquid rather than precipitate. Thus, we consider the effect of precipitation small for the present work. The observed uptake on filter holders was interpreted as an effect of particle leakage and not assigned to the dissolved fraction (but to the total activity).


There are many challenges when conducting experiments to determine the ICRP absorption parameters in vitro in simulated lung fluid. This includes aerosol sampling, conducting the dissolution experiments, analysis of samples, and data evaluation (e.g., curve-fitting). The analytical analysis (i.e., ICP-MS and alpha spectrometry) of the samples has far fewer problems, and the uncertainties from this step are small compared to the other steps. We find it important to share our experiences regarding these challenges and the supplementary test we have carried out in order to help other researchers in the field. Such sources of uncertainty are discussed below.

Sampling representativeness and particle losses

Cascade impactor sampling in the operator breathing zone provides sampling representative to exposure. There are, however, sources of uncertainty to bear in mind while using cascade impactors. The impactor stage cut-point is defined as the aerodynamic size of particles with 50% probability of impaction. This introduces some overlap between impaction stages, and in addition, particles can occur in agglomerates, and small particles can attach to large particles. Sampling of agglomerates present in air is correct from a dosimetry point of view but might obscure the correlation between undissolved fraction (and thus absorption parameters) and impactor cut-points. Moreover, phenomena such as particle bounce can occur in the impactor, resulting in large particles depositing at low cut-point impactor stages. The issue of particle bounce is well known to occur in impactor sampling. The bounce effect can be reduced by using adhesive coating. In the present work, adhesive coating was not used due to the risk of interference with the dissolution of particles in simulated lung fluid (e.g., if particles were embedded in coating material). In the present work, it cannot be ruled out that particle bounce could have obscured the correlation between absorption parameters and impactor stage cut-point, at least for the 0.9 μm, 0.5 μm, and final collection stages, where indications of a bounce effect has been previously noted (Hansson et al. 2020). The question of coating is not trivial considering choice of coating material and how to apply it in a reproducible way. For example, Khalili et al. (2018) demonstrated that the choice of coating material affected the fine particle fraction among other parameters, in next generation impactors. The impact of particle bounce/coating on dissolution rates could be studied in future work, e.g., by applying adhesive coating to a selection of impaction stages.

Experimental setup

There are several methods in use for evaluation of particle dissolution in simulated lung fluid, which can affect the experimental outcome (Ansoborlo et al. 1999; Marques et al. 2011). To the best of our knowledge, no gold standard exists. Variable factors include different formulas for simulated lung fluid, whether macrophages should be included, static/dynamic systems, pH, temperature, filter materials, experiment duration, etc. In our work, the pH was slightly higher than 7.3–7.4 (used in most studies). The effect of pH variations on the absorption parameters appeared to be small (Fig. 4). Previous work demonstrated that a pH shift from 7 to 8 reduced the dissolved fraction by a factor ≈0.8 for UO2 material (Duport et al. 1991). Sutton and Burastero (2004) used a chemical thermodynamic speciation code to model dissolution characteristics of uranium in simulated lung fluid. Their results predicted stable solubility in the pH range relevant for the present work and that carbonate complexes [primarily (UO2(CO3)3)−4] are the dominating species. Since the amount of carbonates is stable in the pH range observed in the present work, we argue that the observed pH fluctuations are of secondary importance.

Experimental duration

In the present work, the experiment was terminated after 100 d. However, derived absorption parameters varied depending on the time-period considered in the curve-fitting procedure. For example, the ss for the pelletizing workshop increased by almost a factor of 2 if only samples collected up to day 40 (instead of day 100) were considered. If only samples collected during the first 10 days were considered, ss increased by a factor of 4. The effect on fr and sr was less prominent. This observation could be of importance when bioassay data are evaluated. Considering that some materials (e.g., Type S) correspond to very long half-times (e.g., a ss of 1 × 10−4 d−1 corresponds to almost 7,000 d half-time), a long experimental duration seems appropriate in order to achieve a good curve-fit if the experimental conditions are stable.

Particle penetration through membrane filter

If particles penetrate through the membrane filter sandwich, this can distort the calculated undissolved fraction and thus absorption parameters. Indications of this phenomenon were identified by studying the dissolution rate (μg d−1) for each sample, i.e., the ratio of 238U (μg) in the sample divided by the sample time (d) in the solution. The dissolution rate was expected to be highest in the beginning and gradually decrease. This was indeed observed, but in some cases, deviations were seen (sudden increases in dissolution rate), which can be interpreted as particle penetration through the membrane filter sandwich (e.g., due to diffusion and/or disturbances when the filter sandwich was moved to a container with fresh simulated lung fluid). This was corrected for as described in the Materials and Methods section. This was particularly prominent for late samples (d 40 or later) at the powder preparation and pelletizing workshop. Simply excluding samples with a deviating dissolution rate would have distorted the undissolved fraction, which is an accumulation of previous samples.

For some data series, a general increase in the dissolution rate was observed after about 10 d, which could be interpreted as more subtle particle penetration. This could overestimate the amount of dissolved 238U (and thus underestimate the undissolved fraction) for late time-points. As a consequence, the ss parameter could be overestimated. However, excluding data-points (e.g., only using from the first 10 or 40 d of the experiment) would also lead to overestimation of the ss parameter due to bias in the curve-fitting procedure (as mentioned in the previous section). Thus, we considered it more appropriate to use data for the entire time period. This was obvious when curve-fitting was carried out after exclusion of late samples: the modeled undissolved fractions at late time-points was lower than actually measured, even though late samples could be affected by some degree of particle leakage. Information on particle leakage during this type of experiment is scarce in the literature.


We have demonstrated that dissolution characteristics for uranium aerosols in simulated lung fluid, sampled and size fractionated using cascade impactors, can differ depending on impactor stage cut-point. A larger mass fraction of sampled uranium was dissolved for stages with small cut-points compared to stages with large cut-points. This was expected, since small particles are expected to have a higher surface-to-volume ratio. Many exceptions were noted, however. We speculate that this is primarily an effect of multiple particle sources of differing physicochemical composition, which is probable since sampling was carried out in the operator breathing zone during ordinary work, which consists of multiple tasks. Other reasons may include effects from sampling artifacts, such as particle bounce.

Derived absorption parameters were in general in fair agreement with ICRP default parameters. The conversion workshop resembled Type M or Type M/S material and was probably a combination of uranium compounds that are soluble and relatively insoluble in the respiratory tract. The powder preparation workshop resembled Type M/S material, whereas the pelletizing workshop resembled Type S material. The sr was larger than the ICRP default parameters throughout all measurements but lower than previously recommended by the ICRP 66 publication.

The correlation between derived absorption parameters and impactor stage cut-point was not always clear. If stages with cut-points smaller than 5 μm were considered as a fine fraction and cut-points larger than 5 μm were considered a coarse fraction, differences in absorption parameters between the fractions were not statistically significant.

In conclusion, it might be appropriate to conduct size fractionation prior to conducting in vitro experiments. Aerodynamic particle sizes near the alveolar deposition maxima (about 0.01–0.02 μm and 1–2 μm) are probably of particular interest. However, the effect on absorption parameters in this study appeared to be relatively small.

The experimental duration can be an important factor to consider for future work, as it affects the curve-fitting; e.g., if fewer samples were considered in the curve-fitting procedure, higher slow rates were estimated. Future work includes using the derived absorption parameters when evaluating bioassay data from workers at the site, as well as dissolution rate experiments in simulated gastrointestinal fluids. The effect of particle bounce and impactor stage coating should be further studied to determine if it obscures the relationship between absorption parameters and aerodynamic particle size.


The Swedish Radiation Safety Authority is acknowledged for funding the present work (grant number SSM2016-589-2). Westinghouse Electric Sweden AB is acknowledged for participation in the study and funding of the PhD program.

Special thanks to Grzegorz Olszewski at Linköping University and numerous colleagues at Westinghouse Electric Sweden AB for invaluable discussions. The operators at the plant are thanked for assisting with sampling of aerosols.


Ansoborlo E, Chazel V, Houpert P, Henge-Napoli MH, Paquet F, Hodgson A, Stradling N. Assessment of physico-chemical and biokinetic properties of uranium peroxide hydrate UO4. Health Phys 75:389–397; 1998. DOI:10.1097/00004032-199810000-00005.
Ansoborlo E, Henge-Napoli MH, Chazel V, Gibert R, Guilmette RA. Review and critical analysis of available in vitro dissolution tests. Health Phys 77:638–645; 1999. DOI:10.1097/00004032-199912000-00007.
Berry JP, Zhang L, Galle P, Ansoborlo E, Hengé-Napoli MH, Donnadieu-Claraz M. Role of alveolar macrophage lysosomes in metal detoxification. Microsc Res Techniq 36:313–323; 1997. DOI:10.1002/(sici)1097-0029(19970215)36:4<313::aid-jemt9>;2-m.
Cera E, Grive M, Bruno J, Ollila K. Modelling of the UO2 dissolution mechanisms in synthetic groundwater—experiments carried out under anaerobic and reducing conditions. POSIVA Report ID 2000–10; 2000.
Chazel V, Houpert P, Ansorbolo E. Effect of U3O8 specific surface area on in vitro dissolution, biokinetics, and dose coefficients. Radiat Protect Dosim 79:39–42; 1998. DOI:10.1093/oxfordjournals.rpd.a032434.
Chazel V, Houpert P, Ansorbolo E, Henge-Napoli MH, Paquet F. Variation of solubility, biokinetics and dose coefficient of industrial uranium oxides according to the specific surface area. Radiat Protect Dosim 88:223–231; 2000. DOI:10.1093/oxfordjournals.rpd.a033039.
Chazel V, Houpert P, Paquet F, Ansoborlo E. Effect of absorption parameters on calculation of the dose coefficient: example of classification of industrial uranium compounds. Radiat Protect Dosim 94:261–268; 2001. DOI:10.1093/oxfordjournals.rpd.a006498.
Davesne E, Blanchardon E. Physico-chemical characteristics of uranium compounds: a review. Int J Radiat Biol 90:975–988; 2014. DOI:10.3109/09553002.2014.886796.
Duport P, Robertson R, Ho K, Horvath F. Flow-through dissolution of uranium-thorium ore dust, uranium concentrate, uranium dioxide, and thorium alloy in simulated lung fluid. Radiat Protect Dosim 38:121–133; 1991. DOI:10.1093/rpd/38.1-3.121.
Eidson AF, Griffith WC Jr. Techniques for yellowcake dissolution studies in vitro and their use in bioassay interpretation. Health Phys 1 46(1):151–63; 1984. DOI:10.1097/00004032-198401000-00013.
Gamble JL. Chemical anatomy, physiology and pathology of extracellular fluid. Boston: Harvard University Press; 1967.
Grubbs FE. Procedures for detecting outlying observations in samples. Technomet 11:1–21; 1969. DOI:10.1080/00401706.1969.10490657.
Guilmette RA. Issues and research on the biochemistry of inhaled actinides. J Alloy Compd 271–273:66–71; 1998. DOI:10.1016/s0925-8388(98)00026-7.
Guilmette RA, Cheng YS. Physicochemical characterization of Capstone depleted uranium aerosols IV: in vitro solubility analysis. Health Phys 96:292–305; 2009. DOI:10.1097/01.hp.0000291946.32717.2b.
Hallstadius L. A method for the electrodeposition of actinides. Nucl Instrum Methods Phys Res 223:266–267; 1984. DOI:10.1016/0167-5087(84)90659-8.
Hansson E, Pettersson HBL, Fortin C, Eriksson M. Uranium aerosols at a nuclear fuel fabrication plant: characterization using scanning electron microscopy and energy dispersive x-ray spectroscopy. Spectrochimica Acta Part B: Atomic Spectroscopy 131:130–137; 2017. DOI:10.1016/j.sab.2017.03.002.
Hansson E, Pettersson HBL, Eriksson M. Uranium aerosol activity size distributions at a nuclear fuel fabrication plant. Health Phys 119:327–341; 2020. DOI:10.1097/hp.0000000000001254.
Hengé-Napoli MH, Ansoborlo E, Claraz M, Berry JP, Cheynet MC. Role of alveolar macrophages in the dissolution of two different industrial uranium oxides. Cell Mol Biol 42:413–420; 1996. PMID: 8793194.
Hinds WC. Aerosol technology: properties, behavior, and measurement of airborne particles. New York: Wiley & Sons; 1999.
Holm E. Review of alpha-particle spectrometric measurements of actinides. Int J Appl Radiat Isotop 35:285–290; 1984. DOI:10.1016/0020-708x(84)90070-x.
International Commission on Radiological Protection. Human respiratory tract model for radiological protection. Oxford: Pergamon Press; ICRP Publication 66; Ann ICRP 24(1–3); 1994.
International Commission on Radiological Protection. Guide for the practical application of the ICRP Human Respiratory Tract Model. ICRP Supporting Guidance 3. Oxford: Pergamon Press; Ann. ICRP 32 (1–2); 2002.
International Commission on Radiological Protection. The 2007 recommendations of the International Commission on Radiological Protection. Oxford: Pergamon Press; ICRP Publication 103; Ann ICRP 37(2–4); 2007.
International Commission on Radiological Protection. Occupational intakes of radionuclides: part 1. Oxford: Pergamon Press; ICRP Publication 130; Ann ICRP 44(2); 2015.
International Commission on Radiological Protection. Occupational intakes of radionuclides: part 3. Oxford: Pergamon Press; ICRP Publication 137; Ann ICRP 46(3–4); 2017.
Khalili SF, Ghanbarzadeh S, Nokhodchi A, Hamed Hamishehkar H. The effect of different coating materials on the prevention of powder bounce in the next generation impactor. Res Pharm Sci 13:283–287; 2018. DOI:10.4103/1735-5362.228958.
Kravchik T, Oved S, Paztal-Levy O, Pelled O, Gonen R, German U, Tshuva A. Determination of the solubility and size distribution of radioactive aerosols in the uranium processing plant at NRCN. Radiat Protect Dosim 131:418–242; 2008. DOI:10.1093/rpd/ncn206.
Kreyling WG. Intracellular particle dissolution in alveolar macrophages. Environ Health Persp 97:121–126; 1992. DOI:10.1289/ehp.9297121.
Liegeois M, Legrand C, Desmet CJ, Marichal T, Bureau F. The interstitial macrophage: A long-neglected piece in the puzzle of lung immunity. Cell Immunol 330:91–96; 2018. DOI:10.1016/j.cellimm.2018.02.001.
Marsh JW, Bailey MR, Birchall A. A step-by-step procedure to aid the assessment of intake and doses from measurement data. Radiat Protect Dosim 114:491–508; 2005. DOI:10.1093/rpd/nch495.
Marques MRC, Loebenberg R, Almukainzi M. Simulated biological fluids with possible application in dissolution testing. Dissolut Technol 18:15–28; 2011. DOI:10.14227/dt180311p15.
Mercer TT. On the role of particle size in the dissolution of lung burdens. Health Phys 13:1211–1221; 1967. DOI:10.1097/00004032-196711000-00005.
Oberdörster G, Oberdörster E, Oberdörster J. Nanotoxicology: an emerging discipline evolving from studies of ultrafine particles. Environ Health Perspect 113:823–839; 2005.
Sdraulig S, Franich R, Tinker RA, Solomon S, O'Brien R, Johnston PN. In vitro dissolution studies of uranium bearing material in simulated lung fluid. J Environ Radioact 99:527–38; 2008. DOI:10.1016/j.jenvrad.2007.08.009.
Sutton M, Burastero SR. Uranium(VI) solubility and speciation in simulated elemental human biological fluids. Chem Res Toxicol 17:1468–1480; 2004. DOI:10.1021/tx049878k.
Shi Y, Collins R, Broome C. Determination of uranium, thorium and plutonium isotopes by ICP-MS. J Radioanalyt Nucl Chem 296:509–515; 2013. DOI:10.1007/s10967-012-2128-9.
Stefaniak AB, Guilmette RA, Day GA, Hoover MD, Breysse PN, Scripsick RC. Characterization of phagolysosomal simulant fluid for study of beryllium aerosol particle dissolution. Toxicol In Vitro 19:123–134; 2005. DOI:10.1016/j.tiv.2004.08.001.
Stefaniak AB, Day GA, Hoover MD, Breysse PN, Scripsick RC. Differences in dissolution behavior in a phagolysosomal simulant fluid for single-constituent and multi-constituent materials associated with beryllium sensitization and chronic beryllium disease. Toxicol In Vitro 20:82–95; 2006. DOI:10.1016/j.tiv.2005.06.031.


Breathing zone sampling was repeated at the pelletizing and BA pelletizing workshops (Table A1).


Radiometric data of undissolved uranium on the filter sandwiches (membrane filter with either impactor substrate or final collection filter), as determined by alpha spectrometry, are presented in Appendix B (Table B1).

Table B1 - Radiometric data (alpha spectrometry) for the undissolved amount of 238U. Numbers in brackets indicate 1 standard deviation from counting statistics.
Workshop Impactor stage cut-point (μm) 238 U (Bq) 238 U (μg)
Conversion 21.3 7.10 × 10 3 (1.41 × 10 4) 5.71 × 10 1 (1.13 × 10 2)
14.8 8.72 × 10 3 (2.42 × 10 4) 7.01 × 10 1 (1.95 × 10 2)
9.8 6.74 × 10 3 (1.96 × 10 4) 5.42 × 10 1 (1.58 × 10 2)
6.0 5.14 × 10 3 (1.19 × 10 4) 4.13 × 10 1 (9.60 × 10 3)
3.5 1.86 × 10 3 (7.09 × 10 5) 1.50 × 10 1 (5.70 × 10 3)
1.6 1.27 × 10 3 (9.00 × 10 5) 1.02 × 10 1 (7.24 × 10 3)
0.9 6.79 × 10 4 (5.22 × 10 5) 5.46 × 10 2 (4.20 × 10 3)
0.5 7.21 × 10 4 (8.48 × 10 5) 5.80 × 10 2 (6.82 × 10 3)
<0.5 2.74 × 10 3 (1.46 × 10 4) 2.21 × 10 1 (1.18 × 10 2)
Powder preparation 21.3 2.69 × 10 2 (8.34 × 10 4) 1.93 (6.00 × 10 2)
14.8 3.79 × 10 2 (3.72 × 10 4) 2.82 (2.77 × 10 2)
9.8 5.16 × 10 2 (1.81 × 10 3) 3.77 (1.33 × 10 1)
6.0 4.46 × 10 2 (1.86 × 10 3) 3.27 (1.36 × 10 1)
3.5 3.18 × 10 2 (8.11 × 10 4) 2.29 (5.85 × 10 2)
1.6 3.44 × 10 2 (1.15 × 10 3) 2.44 (8.18 × 10 2)
0.9 1.92 × 10 2 (5.04 × 10 4) 1.44 (3.78 × 10 2)
0.5 7.13 × 10 3 (2.86 × 10 4) 5.34 × 10 1 (2.14 × 10 2)
<0.5 4.93 × 10 3 (1.50 × 10 4) 3.62 × 10 1 (1.10 × 10 2)
Pelletizing 21.3 4.55 × 10 2 (9.96 × 10 4) 3.66 (8.01 × 10 2)
14.8 2.59 × 10 2 (5.80 × 10 4) 2.08 (4.66 × 10 2)
9.8 3.90 × 10 2 (9.70 × 10 4) 3.14 (7.80 × 10 2)
6.0 2.62 × 10 2 (8.90 × 10 4) 2.10 (7.16 × 10 2)
3.5 1.45 × 10 2 (3.31 × 10 4) 1.17 (2.66 × 10 2)
1.6 8.86 × 10 3 (2.29 × 10 4) 7.12 × 10 1 (1.84 × 10 2)
0.9 4.82 × 10 3 (1.15 × 10 4) 3.87 × 10 1 (9.22 × 10 3)
0.5 2.52 × 10 3 (1.34 × 10 4) 2.03 × 10 1 (1.08 × 10 2)
<0.5 9.23 × 10 3 (2.80 × 10 4) 7.42 × 10 1 (2.25 × 10 2)
BA pelletizing 21.3 4.91 × 10 3 (1.78 × 10 4) 3.95 × 10 1 (1.43 × 10 2)
14.8 8.81 × 10 3 (2.23 × 10 4) 7.08 × 10 1 (1.80 × 10 2)
9.8 8.93 × 10 3 (2.66 × 10 4) 7.18 × 10 1 (2.14 × 10 2)
6.0 9.16 × 10 3 (3.06 × 10 4) 7.36 × 10 1 (2.46 × 10 2)
3.5 9.64 × 10 3 (3.41 × 10 4) 7.75 × 10 1 (2.74 × 10 2)
1.6 6.91 × 10 3 (1.71 × 10 4) 5.56 × 10 1 (1.37 × 10 2)
0.9 2.40 × 10 3 (8.00 × 10 5) 1.93 × 10 1 (6.43 × 10 3)
0.5 1.32 × 10 3 (1.09 × 10 4) 1.06 × 10 1 (8.78 × 10 3)
<0.5 1.87 × 10 3 (1.22 × 10 4) 1.50 × 10 1 (9.77 × 10 3)
Procedure blanks
Membrane filters + MCE impaction substrate 4.30 × 10 4 (4.00 × 10 5) 3.46 × 10 2 (3.21 × 10 3)
Membrane filters + final collection filter 1.84 × 10 4 (3.87 × 10 5) 1.48 × 10 2 (3.11 × 10 3)


The full figures (all impaction stages) are given in Fig. C1.

Fig. C1:
Undissolved fraction of 238U for all impaction stages for each workshop. The ICRP default absorption types are included for comparison. Error bars correspond to ±1 standard deviation. The filter holders for the 14.8-μm stage at the powder preparation workshop and 0.5-μm stage at the conversion workshop were accidentally destroyed, and thus 238U from additional leaching could not be added to U0. The 0.5-μm stage at the conversion, pelletizing, and BA pelletizing workshop, and the 0.9-μm stage at the BA pelletizing workshop should be viewed with caution due to several samples close to procedure blank levels.

dose, internal; lungs, human; nuclear fuel cycle; uranium

Copyright © 2022 The Author(s). Published by Wolters Kluwer Health, Inc. on behalf of the Health Physics Society.