Can We Trust Real Time Measurements of Lung Deposited Surface Area Concentrations in Dust from Powder Nanomaterials?

A comparison between various methods for real-time measurements of lung deposited surface area (LDSA) using spherical particles and powder dust with specific surface area ranging from 0.03 to 112 m 2 g –1 was conducted. LDSA concentrations measured directly using Nanoparticle Surface Area Monitor (NSAM) and Aerotrak and were compared to LDSA concentrations recalculated from size distribution measurements using Electrical Low Pressure Impactor (ELPI) and Fast Mobility Particle Sizer (FMPS). FMPS and ELPI measurements were also compared to dust surface area concentrations estimated from gravimetrical filter measurements and specific surface areas. Measurement of LDSA showed very good correlation in measurements of spherical particles (R 2 > 0.97, Ratio 1.0 to 1.04). High surface area nanomaterial powders showed a fairly reliable correlation between NSAM and Aerotrak (R 2 0.73– 0.93) and a material-dependent offset in the ratios (1.04–2.8). However, the correlation and ratio were inconsistent for lower LDSA concentrations. Similar levels of correlation were observed for the NSAM and the FMPS for high surface area materials, but with the FMPS overestimating the LDSA concentration. The ELPI showed good correlation with NSAM data for high LDSA materials (R 2 0.87–0.93), but not for lower LDSA concentrations (R 2 0.50–0.72). Comparisons of respirable dust surface area from ELPI data correlated well (R 2 > 0.98) with that calculated from filter samples, but materials-specific exceptions were present. We conclude that there is currently insufficient reliability and comparability between methods in the measurement of LDSA concentrations. Further development is required to enable use of LDSA for reliable dose metric and regulatory enforcement of exposure.


INTRODUCTION
There is increasing evidence that the pulmonary toxicological response of ambient air-pollution and manufactured nanomaterials may, at least partially, be driven by the specific surface area dose of the test materials (Oberdorster, 2000;Maynard and Kuempel, 2005;Duffin et al., 2007;Jacobsen et al., 2009;Giechaskiel et al., 2009;Donaldson et al., 2013;Saber et al., 2014). Therefore, it is of high interest to include airborne particle surface area measurements to offer a potentially more biologically relevant metric in exposure and risk assessment. To meet these new developments, it is also of interest to include surface area measurements in dustiness testing; the latter tests are performed to rank the ability of powders to generate dust and used in e.g., new modeling approaches and regulatory exposure assessment (BS EN:15051, 2006;Aitken et al., 2011;Jensen et al., 2015).
The surface area of airborne dust particles may be determined by either direct or indirect methods. There are currently no commercially available real-time methods to determine the geometric surface area concentration of airborne particles. In principle the most correct method would be direct determination of the specific surface area on filters using the BET inert gas-adsorption technique. Even-though this method has been demonstrated (LeBouf et al., 2011a;Lebouf et al., 2011b), the method is in reality not yet straight-forward and changes could potentially occur in the aerosol during filter-sampling and storage that might change the surface area. Therefore, this method may not be generally applicable.
Another approach is to estimate the surface area concentrations based on airborne particle number size distributions assuming particle shape. Particle numbersize-distributions can be measured using a range of different optical, aerodynamic, electrical mobility and charging techniques. Therefore, the comparability in these measurements and the reliability in conversions from number and volume concentrations metrics is still not well documented for non-spherical particles such as aggregates and agglomerates commonly observed in dustiness tests, covering a wide range of sizes from nano-to µm-size (Ibaseta and Biscans, 2007;Schneider and Jensen, 2008;Jensen et al., 2009;Tsai et al., 2009;Levin et al., 2014;Koivisto et al., 2015;Koponen et al., 2015). Even in case of direct release from an airborne synthesis processes, primary particles can already start to agglomerate or aggregate in the reactor and may further coagulate at or very near the point of release with similar result (Makela et al., 2009;Hämeri et al., 2009;Schneider et al., 2011;Koivisto et al., 2012;Koivisto et al., 2015). Measurements are therefore challenging when the aerosol structure is not known, because this in principle prevents the possibility to convert number-size-distribution measurements into reliable surface area values (Fissan et al., 2013).
To overcome the uncertainties in size-distribution conversion, the lung deposited surface area (LDSA), i.e., the fraction of the total airborne particle geometric surface area concentration that would deposit in the human lung, of aerosols can be measured in real-time. Based on particle unipolar diffusion charging and subsequent current measurement with electrometers (Fissan et al., 2007;Shin et al., 2007), this principle is used by several instruments such as the Aerotrak (Aerotrak 9000, TSI inc, MN, USA) and the Nanoparticle Surface Area Monitor (NSAM 3550, TSI Inc, MN, USA). It should be noted that this surface area is not the same as the surface area determined by nitrogen or krypton gas-adsorption using the BET technique for powders (Bau et al., 2010).
Recently, a few studies reported measurements comparing the results of different number size-distribution and surface area monitors challenged to equidimensional or spherical and aggregated particles. For aerosol monitors, Asbach et al. (2012) investigated the correlation between Aerotrak (Aerotrak 9000, TSI inc, MN, USA) and Fast Mobility Particle Sizer (FMPS 3091, TSI inc, MN, USA) with polydisperse aerosols of different sizes and morphologies (35 nm cubic NaCl, 190 nm spherical Di-Ethyl-Hexyl-Sebacat (DEHS) oil particles and 50 nm agglomerated soot particles; all modal diameters), and found good agreement (mostly within ± 15% difference) and correlation for NaCl (R 2 ≥ 0.985) and soot (R 2 ≥ 0.997) but poorer agreement for DEHS with deviations up to 50% (R 2 ≥ 0.989). The comparability in the tests with DEHS decreased when the spherical particle sizes exceeded 400 nm, which has been reported to be the upper limit for accurate measurement with this type of instruments. Levin et al. (2015) compared the particle sizing and counting of the FMPS, Electrical Low Pressure Impactor (ELPI, Dekati, Finland) and a Scanning Mobility Particle Sizer (SMPS) for spherical particles in the size range of 50-900 nm. While ELPI and SMPS data agreed well within the tested size-range, the FMPS was unable to size particle larger than 200 nm correctly. This led to it severely underestimating the particle size and overestimating the concentration. Leavey et al. (2013) compared the Aerotrak with LDSA concentrations calculated from size distributions measured with a SMPS for a range of different particle types and morphologies with mean particle diameters ranging from 30 to 140 nm. They found a strong correlation (R 2 0.78-0.99) between the two methods but a moderate dependency on particle type. In most cases, the SMPS data overestimated the LDSA as compared to the Aerotrak. Leskinen et al. (2012) compared LDSA concentrations measured with NSAM and SMPS and found that for TiO 2 agglomerates and dust agglomerates and found that values from NSAM were higher than those of the SMPS. Todea et al. (2015) compared the measured LDSA from Aerotrak and NSAM for monodisperse spheres and agglomerates with the calculated value based on number concentration from a Condensational Particle Counter (CPC) and knowledge of particle size. They found that for spherical particles the NSAM and Aerotrak data both generally agreed within ± 30% with the calculated value from the CPC in the size range of 20 to 346 nm. For spheres larger than 400 nm, LDSA concentration was underestimated by both instruments. For agglomerated particles, the instruments based on unipolar diffusion charging generally overestimated the LDSA concentration compared to the CPC value, but mostly still within ± 30%.
These previous results indicate that both method-and size-specific differences exist in the results given by different measurement devices used for sub-µm sized particles. Additional differences may arise when such devices are used to measure agglomerates with even wider size-distributions and different highly complex agglomerate structures, which may be found in dust generated from nanopowders. Since most applicable instrument(s), and potentially also the reliability of measurement results, will depend on the characteristics of the airborne particles and their sizedistribution and all aerosol instruments in principle are calibrated and report their data assuming spherical equivalent diameters, it is difficult to measure and report true particle exposure values. This is because the assumptions required for generating the correct values are usually not known or are very complex in true exposure situations, and it is still poorly understood how instrument responses vary with the many different material characteristics and properties.
The aim of this study was to investigate the comparability of the results in a performance test of four commonly used real-time monitoring particle measuring instruments when challenged to pigment and nanomaterial powder dusts which resemble dust from occupational powder handling. For comparison with a morphologically simple aerosol, we also demonstrate the performance of the measurement devices challenged to a spray product forming nm-size condensate particles (Norgaard et al., 2009). Because all currently commercially available surface area monitors report the biologically relevant alveolar lung deposited surface area, this work also involves an analysis of potential procedures of conversion between measured number size distribution data and LDSA and their comparability. Finally, the realtime measurement data are compared with interpretations from gravimetric measurements of collected respirable dust samples.
The test powder materials were specifically selected to challenge the monitoring instruments based on their variations in primary particle size and specific surface areas as determined by the BET technique. Dust aerosols released from such powders may pose some measurement problems due to their level of agglomeration, size-distribution and multimodality, which may not be revealed in tests using standard testing particles such as PSL, NaCl or DEHS. However, documentation of these problems and their levels is important to understand the ability to use such instruments for monitoring and regulatory exposure assessment.

Instrumentation
Four different real-time monitoring instruments were used in this study as listed below. Two of them are particle sizing instruments while the other two are LDSA concentration monitors. All instruments were calibrated by the manufactures prior to the study.
• Fast Mobility Particle Sizer (FMPS; TSI model 3091, TSI Inc., Shoreview, MN, USA). In the FMPS, particles are charged through a dual unipolar diffusion charger prior to size classification according to particles electrical mobility based on their trajectory in an electrical field. The electrical charge carried by the size classified particles is measured with 22 electrometers which currents are corrected for multiple charges and image charges and further inverted to size distribution having 32 bins between 5.6 and 560 nm. The FMPS was used to measure particle number size distributions. The alveolar deposited surface area concentration and respirable surface area concentration were further calculated assuming spherical particles with unit density. Sampling for the FMPs is done using a PM 1 -cyclone to remove micron-sized particles. • Electrical Low Pressure Impactor (ELPI, Dekati, Finland). In the ELPI (Keskinen et al., 1992;Marjamaki et al., 2005), particles are charged using a unipolar diffusion charger after which the particles are collected in a low pressure cascade impactor which fractionates the particle sizes based on their aerodynamic properties into 12 channels spanning from 30 nm to 10 µm. The electrometer detected current in each stage is corrected for multiple charges, image charges and inverted to a size distribution. In this study the ELPI was used to measure particle number size distributions and further calculate alveolar deposited surface area concentration and respirable surface area concentration. • Nanoparticle Surface Area Monitor (NSAM; TSI model 3550, TSI Inc., Shoreview, MN, USA). The NSAM uses unipolar diffusion charging and a Faraday cup electrometer to measure a current which was found to be proportional to the lung deposited surface area concentration. By changing the voltage of the instrument ion trap, different proportionalities can be applied to mimic that of alveolar or tracheobronchial deposition as described by ICRP (1994). The inlet of the NSAM is mounted with a PM 1cyclone. In this study, only the alveolar deposited surface area measured by the NSAM was investigated. • Aerotrak (Aerotrak 9000, TSI Inc., Shoreview, MN, USA). The Aerotrak is a field-portable version of the NSAM. In all other functions it is identical to the NSAM. The inlet of the Aerotrak is mounted with a PM 1 -cyclone. The Aerotrak has previously been compared with the NSAM and good agreement has been confirmed (Bau et al., 2012).

Case of Spherical Particles Challenging the Instruments
The first experiment was carried out in a 20 m 3 stainless steel chamber which was ventilated using a HEPA-filtered outdoor air which was well mixed within the chamber using two fans. A well-studied surface cleaning and maintenance product Multicover (Nanocover, Denmark) was applied for 10 seconds from a distance of 20 cm onto a metal sheet (Norgaard et al., 2009;Norgaard et al., 2010). Approximately 5 minutes after spraying a nucleation event took place, which formed nm-size droplet particles anticipated to be due to nucleation through limonene oxidation that slowly grew in size through condensation.
Airborne particles were sampled from the room at the same point using separate conductive tubes for each instrument. The tube lengths were adjusted so that residence times were equal for each instrument to avoid differences in particle losses. As spray condensation particles are considered to be highly spherical, this experiment on one hand served as a calibration check of the instruments as well as a challenge to a simple aerosol within the optimum application ranges of all instruments.

Case of Dust Particles Challenging the Instruments
Seven different powders (Table 1) were selected so that primary particle sizes varied from 16 nm to over 100 nm, specific surface areas from 0.03 to 113 m 2 g -1 , (Volume-Specific Surface Area (VSSA) from 0.3 to 437 m 2 cm -3 calculated using the nominal material densities from 2.7 to 10.4 g cm -3 ). The variation in specific surface areas was used to indicate the potential level of aggregation in the powder particles. The extraordinary low specific surface area was observed for the PVP-stabilized AgNP powder, which appears as large aggregates with a water-soluble PVP matrix. The highest specific surface area material was UF TiO 2 , which also had the smallest primary particle size (16 nm).
The powders were aerosolized using the miniaturized EN15051 dustiness drum (Schneider and Jensen 2008). The system has been made to correspond to the EN15051 standard drum so relative dimensions have been maintained and the 33 drops per minute is ensured by three lifter vanes in the drum and an 11 rpm rotation. The inlet air to the drum was controlled at 50% RH and HEPA-filtered to ensure a particle-free background. The measurements were split into two separate sampling trains in order to maintain 11 L min -1 flow through the drum (Fig. 1). Before starting the test, 2 g of powder was loaded into the drum, which was then rotated for 60 seconds to ensure that all surfaces were saturated with particles. After the drum had been emptied, 6 g of test material were loaded into it and 60 seconds of background measurements were done without the drum rotating to ensure a particle free test atmosphere. The experiment was then initiated by rotating the drum for 60 seconds during which particles were emitted and led through the airflow to the sampling train. After the drum was stopped, measurement of remaining airborne particles was conducted for another 120 seconds. This completed the rotational test. The measurement was then repeated an additional two times and the results here are presented as averages over the three experiments.
In the first experimental set-up ( Fig. 1(a)), respirable dust (as defined by EN 481 (1993), D 50 = 4 µm) was collected with a GK2.69 cyclone at 4.2 L min -1 (BGI Inc., Waltham, MA, USA), and particle concentrations were measured with the FMPS, NSAM and Aerotrak. In addition, an Electrostatic Precipitator (ESP, Asbach, 2015) was used to deposit particles onto 25 mm silica wafers. These were then used in an un-coated state for analysis by Scanning Electron Microscopy (SEM) using a FEI ESEM Quanta 200 FEG (FEI, Eindhoven, the Netherlands) and Energy-Dispersive X-ray Spectroscopy (EDX) to ensure that the particles of interest were studied. The respirable dust mass was determined using a Sartorius microbalance (Type R162 P; Sartorius GmbH, Göttingen, Germany), after conditioning the filters and control filters in a weighing room (22C; 50%RH) before and after sampling. The measured mass was used to calculate the gravimetrical powder dustiness index and categorize the powder dustiness level according to EN15051.
In the second experimental set-up ( Fig. 1(b)), particle size distributions were measured by using the ELPI and FMPS. The FMPS was used to validate that the measurements were comparable with the measurements performed in the first experimental set-up.

Particle Concentrations Analysis 1) Comparison of LDSA Concentrations
For both FMPS and ELPI, alveolar particle number concentration fractions were calculated by multiplying measured particle number size distributions with the alveolar deposition fraction, DF alv (ICRP, 1994). The alveolar deposition fraction is based on aerodynamic diameter; therefore we assumed spherical particle shape and unit density while calculating the FMPS alveolar particle number concentrations. Particle number size distributions measured  by the ELPI were also multiplied with the NSAM PM 1cyclone penetration efficiency (Table S1) to simulate the cyclone placed upstream of the NSAM, Aerotrak and FMPS. For direct comparison of the measured LDSA concentrations between the 4 instruments, time series data were converted into running 5 second averages. Due to NSAM and Aerotrak measuring alveolar deposited surface area, LDSA alv , the ELPI and FMPS (without PM 1 fractions) had to be recalculated according to: with n(D p ) being the number of particles with diameter D p , DF alv and PM 1 being the alveolar deposition fraction and cyclone penetration fraction for the specific particle size, respectively.

2) Comparison of Respirable Surface Area Concentrations
A second analysis was completed to investigate the comparability between the theoretical geometric surface area of the respirable dust collected on the filter for dustiness categorization and the theoretical accumulated respirable surface area concentrations of the airborne dusts measured with the FMPS and ELPI.
On the one hand, respirable LDSA values were compared with the respirable particle surface area calculated from the respirable gravimetric measurements [g] multiplied by their respective specific surface areas [m 2 g -1 ] ( Table 1).
On the other hand, FMPS and ELPI particle number size distributions were converted to geometric surface area size distributions, assuming spherical particles (SA sph ). The latter is then integrated during the sampling period (180 s), and, transformed to the corresponding respirable fractions, RF. It should be noted that the FMPS was included in comparison, despite not covering the entire range of the respirable fraction. It is therefore expected that the FMPS data will underestimate the surface area in cases where a considerable number of particles are larger than 560 nm.
where n(D p , t) is the number size distribution (FMPS, ELPI) at time t, Q drum the total air flowrate of the drum, Q instr the instrument flow and τ the sampling time.
Another approach is based on particle volume size distribution, combined with the respirable fraction, material density, ρ, and specific surface area, SSA: Unlike the first approach this uses the assumption of spherical shape to calculate a total volume which then is recalculated into a surface area using known properties of the material. Fig. 2 shows an overview of the measured gravimetrical respirable dustiness indices for the 7 materials used in the study along with the regions of classification of powder dustiness according to EN15051. The materials covers a wide range of dustiness from Very Low (TiO 2 Pigment, TiO 2 AFDC, CaCO 3 #2) to Low (AgNP) and High (ZnO, CaCO 3 #1, UF TiO 2 ).

Real-Time Measurements
Particle size distribution measurements by the FMPS and ELPI showed a good agreement on the droplets formed in the chamber measurements ( Fig. 3(a)). The small rotating drum shows in general good repeatability even though differences may occur in the modal diameter as well as the shape of the (number size) distribution. However, large discrepancies/deviations between FMPS and ELPI number size distributions were observed for all non-spherical powder particles generated using the small rotating drum (Figs.  3(b)-3(h)). More precisely, the FMPS shows consistently higher particle number concentrations within its measurement size range (< 560 nm) as compared to the ELPI. In the FMPS, particle concentrations often abruptly decreased in particle sizes below 100 nm (Figs. 3(a), 3(b), 3(d), 3(g), and 3(h)) while particles concentrations were specifically predominant in the size region of 150-200 nm. On contrary, the ELPI did not show any increased particle concentrations in this size range (Figs. 3(b)-3(h)). For powders, the ELPI peak particle number concentrations were generally located in the size range of 0.5 to 2 µm. It should be remembered that the FMPS classifies particles according to their electrical mobility equivalent diameter while the ELPI measures aerodynamic equivalent diameter, which are equal if particles are spherical and have unit density (1 g cm -3 ). In the chamber experiment, particles are presumably spherical with a density close to 1 g cm -3 . Thus, the FMPS and ELPI particle number size distributions were found to be similar ( Fig.  3(a)). However, in the rotating drum experiments, airborne particles are highly agglomerated, with particle effective densities that are likely to be significantly different from unity. For powder particles, an order of magnitude difference in measured modal diameter was observed. This difference cannot be explained alone by particles shape and density; it would require two orders of magnitude difference in effective density as estimated by using the relation between the mobility diameter and aerodynamic diameter as described by Kelly and McMurray (1992). The deviating FMPS measurement is explained by Levin et al. (2015) where it is shown that FMPS is unable to measure distributions of spherical particles with a Geometric Mean Diameter (GMD) higher than 200 nm. This is partly because the electrical mobility size dependency flattens out within the FMPS size range and particles become harder to size-separate based on this technique. Furthermore, secondary effects in the FMPS current inversion seem to affect the final size distribution. For distributions having a GMD above 200 nm, it still shows them as having a GMD of 200-300 nm, and overestimates the number concentration. Furthermore, Shin et al. (2010) showed that agglomerated particles have higher unipolar charging efficiency than spherical particles having the same mechanical mobility. Thus, after unipolar charging, the electrical mobility of an agglomerate is higher than the electrical mobility of a spherical particle having equivalent mechanical mobility as the agglomerate. Increased agglomeration level is seen in the FMPS measurements as a decrease in electrical mobility size and an increase in concentration. It also moves the minima in electrical mobility into the FMPS size range, making size-classification based on this ambiguous . In the ELPI measurements, agglomeration increases the measured concentration due to increased charging capabilities.

Electron Microscopy
Typical dust particles from each material used in the dustiness drum and collected using the ESP is shown in Fig. 4. The AgNP appears in irregular aggregates with no clear primary particles visible by SEM (Fig. 4(g)) while other powder particles appear as aggregates of fused primary particles. The special appearance of the AgNP is explained by the abundant matrix encapsulating the AgNP (see Nymark et al. (2013) for TEM images of the primary AgNP particles). It was impossible to generate statistical data on particle size-distributions and morphologies due to insufficient number of particle counts on the images obtained. However, qualitatively, the fractal dimensions of the particles studied are all rather close to 3, which correspond to compact particles (Virtanen et al., 2004). This would mean that the unipolar charging done in the instruments should not be affected too much by the particle morphology.

Surface Area Comparison
The first step in the comparison was a benchmark comparison between the NSAM and Aerotrak, which are specifically designed for LDSA concentration measurements. Since both instruments are based on the same design, they were expected to give very similar results. Fig. 5(a) shows the LDSA concentration measured by the NSAM plotted against that of Aerotrak for all test particles. The LDSA concentration measured from spherical test particles was well within the manufacturer's given accuracy (both instruments are specified to have a ± 20% measurement accuracy in the region of 20 to 200 nm) and had an average ratio of 1.06 (Table 2). In general the surface area monitors showed good comparability with a few exceptions. For low surface area materials, CaCO 3 #2 and TiO 2 Pigment, the Aerotrak gave notably higher concentrations of surface area and the correlation between the instruments was poor. TiO 2 AFDC, while having a poor correlation, gave values within ± 20% of the NSAM. For the powders with higher specific surface area, the Aerotrak appears to give slightly higher values than NSAM, but with better correlation factors. This shows that, even though the instruments are based on the same principle, the response to various particle shapes may differ. This may be due to differences in charger ion concentration and following calibration factors.
The calculated LDSA concentration measured using FMPS data are shown in Fig. 5(b), with correlation factors and ratios in Table 2. It shows that the LDSA concentration measured using the FMPS appears to correlate well with that of the NSAM for certain powder materials such as CaCO 3 #1, ZnO, AgNP and UF-TiO 2 throughout their whole spectra of LDSA concentration levels. For most powders the FMPS appears to consequently give higher values than the NSAM, with the exceptions of UF-TiO 2 where the ratio is 0.77. The disagreement between the methods increases slightly at lower surface areas, i.e., for TiO 2 Pigment. No powder material has an average ratio within the ± 20% boundaries. The underestimation of the NSAM as compared to the FMPS could be explained by the fact that NSAM underestimates LDSA concentrations for particles larger than 400 nm (Asbach et al., 2009). The assumption of spherical particle shape and unit density could also affect the results.
The LDSA concentration of particles below the 1 µm cut-off measured with the ELPI is shown in Fig. 5(c). The correlation of LDSA concentrations between the ELPI calculated values and the NSAM was fair (R 2 0.87-0.93) for the high surface area powders but poorer (R 2 0.5-0.72) for the low surface area ones ( Table 2). The ELPI calculated LDSA concentrations were underestimated as compared to the ones measured by the NSAM. Only UF-TiO 2 has an average ratio within the ± 20% boundary. It should be noted that the ELPI was not measuring on the same aerosol as the surface area monitors but on separate repetitions of the same experiment, due to limitations in the total available volume-flow through the drum. However, measurements done with FMPS during both setups showed a good comparability between the two setups (R 2 > 0.95).   6 shows that the accumulated respirable dust surface area calculated from gravimetric samples and calculated using Eqs. (2) and (3) from the ELPI and FMPS measurements were only similar for a few powders. With the exception of AgNP, the respirable dust surface areas calculated directly by assumption of spherical particles (Eq. (2)) is greatly underestimated as compared to the respirable dust surface area assigned from gravimetric measurements. This exception is likely due to its very low SSA as a bulk material. When the specific density and SSA is considered (Eq. (3)) the values for airborne dust as measured by the ELPI generally correlates well with the values for filter-based calculations  (2) and (3), and calculated value from filter sampling.
(R 2 0.98, Ratio 1.17). The corresponding values for the FMPS do not have a general trend towards approaching the gravimetrical value (Ratio 0.23). From this it is evident that using the relative density in the assessment of airborne dust surface area measured with the ELPI improves the comparability with the surface area derived from the respirable dust mass. However, large differences between the two estimates of released dust surface area may still arise for certain materials.

CONCLUSIONS
This study analyzes the comparability of lung deposited surface area (LDSA) concentrations as measured by or derived from four commonly used devices and approaches (FMPS, ELPI, NSAM, and Aerotrak) for both spherical particles and seven different powders aerosolized using the small rotary drum to simulate realistic workplace-like nanostructured dust particles. Test aerosol LDSA concentrations ranged from 0.1 to 10 4 µm 2 cm -3 as measured by NSAM, where spherical particles were mainly below 100 nm, while powder-borne particles covered the range from 0.1 to 10 µm, based on ELPI measurements. Powders dustiness indices varied from Very Low up to High according to the EN:15051 classification.
Test of spherical droplets after application of a surface coating product showed a good correlation (R 2 > 0.97, Ratio between 1.0 and 1.04) between NSAM, FMPS and Aerotrak for alveolar LDSA.
Dustiness testing of the seven highly different powders showed only fairly reliable correlation (R 2 between 0.73 and 0.93) and concentration ratio (1.04-2.8) between the two surface area monitors tested for the high surface area range of materials. However, the comparison was inconsistent for the lower surface area concentrations and varied considerably. Similar levels of correlation were observed comparing the data from the NSAM and the FMPS for high surface area materials, but with the FMPS overestimating the concentration. Finally, the re-calculated ELPI LDSA concentrations showed good correlation at high concentrations (R 2 between 0.87 and 0.93), with somewhat lower correlation at the lower end (R 2 between 0.50 and 0.72). The LDSA concentration ratios were between 1.12 to 5.19 and 1.35 to 4.43, respectively.
Using a different approach, we evaluated the comparability between accumulated surface areas derived from on-line particle size-distribution measurements using FMPS and ELPI and a theoretical respirable dust surface area calculated based on the mass of respirable dust collected on filters. The results show that dust surface area converted from ELPI data may generally correlate well (R 2 > 0.98, Ratio 1.17) with the theoretical mass-based surface area of the generated dust particles. However, some large deviations were observed for the same calculations using FMPS data which may be due to particle morphology (agglomeration/aggregates), effective densities, and other previously mentioned issues regarding online measurement of particle size-distributions and surface areas. Further studies are required to validate the results using this approach for assessment of airborne dust particle LDSA with ELPI as it shows promise.
As an over-arching conclusion, the measured LDSA concentrations variation between the instruments exceeded any accepted limit for comparability between the instruments and was strongly material dependent. The observed lack of data agreement between the different state-of-the-art measurement equipment is important and the apparent discrepancies encourage the need for further metrological research. If LDSA is to become a metric of choice in toxicology and occupational exposure assessments as well as regulatory enforcement, then a method of choice has to be decided or developed as current techniques give material dependent results and do not have the appropriate comparability. Results from different techniques can therefore not be trusted to represent comparable levels of LDSA concentrations. The results should have implications for development of new measurement devices developed.