There is growing concern about the potential health impact of engineered nanomaterials (Schulte et al., 2008; Brouwer, 2010). Inhalation of airborne nanoparticles is a potentially significant exposure route. There is relatively little information on the nature and magnitude of aerosol releases during manufacture and handling of nanoparticles; however, recent work (Birch et al., 2011; Dahm et al., 2012) suggests that localized exposures to nanoparticles can and do occur in the workplace. Industrial production presents the potential for worker exposure, especially for those involved in the manual handling, transfer, or conveyance of these materials (Peters et al., 2009; Evans et al., 2010). Inhalation of aerosolized material is expected to be the primary route of exposure and hence of most concern although dermal, ocular, and ingestion routes may also contribute. Controlled laboratory measurements may provide interim guidance as to the relative severity of airborne release of nanoparticles in occupational settings. Such measurements, together with information on the potential hazard (toxicity) of the material, form the basis of exposure control banding (Paik et al., 2008).The testing of nanoparticle powders is especially challenging, due to the small quantities of test materials available for testing (cost considerations) and the potential toxicity of these materials. This challenge is similar to that posed by potent (and frequently costly) pharmaceutical powders.

A variety of airborne dust generating mechanisms may be present in the workplace. These may include accidental spills (Sutter et al., 1982); falling powders, via pouring or transferring processes (Heitbrink et al., 1992); and other agitation, as in powder conveyance (Cheng, 1973). Grinding or milling of bulk materials both generates and suspends small particles as airborne dusts (Pensis et al., 2010). Mechanical airborne reentrainment of previously spilled or deposited powder (perhaps through dry cleanup procedures) is also a potential source. An undesirable end result is that a portion of the bulk powder becomes and remains airborne in the workplace environment.

Dustiness has been defined as ‘the propensity of a material to generate airborne dust during its handling’ (Plinke et al., 1992; Lidén, 2006). Dustiness can be tested by ‘standardized techniques’ that involve the application of a specified type and amount of mechanical energy to a specified amount of test material for a specified time (in order to overcome the adhesive binding forces within the test powder) and thus disperse preexisting particles from the test powder into the air; the amount of dust released is then quantified (Plinke et al., 1992; Lidén, 2006). A dustiness test is not intended to generate new particles.

Lidén has argued (Lidén, 2006) that dustiness is not an intrinsic physico-chemical property of a material: the particle size distribution, humidity, and nature of the adhesive forces (binding the particles within the powder) all influence the dustiness. While the particle size distribution is not an intrinsic physico-chemical property of the material, it is a statistical property of a given powder. Humidity, for example, is an external factor that can be controlled during testing. The nature of the adhesive forces is influenced by physical form and chemistry; with appropriate control of environmental conditions (e.g. humidity), it should be possible to obtain reproducible results. Thus, while not a material property, dustiness is a property of a given powder that should be quantifiable (and reproducible) under a given controlled testing protocol.

Powders typically consist of primary particles, which are aggregated (sintered or tightly bonded) and then further agglomerated (bound by forces of varying strengths). Typically, it is fairly easy to break up the loose agglomerates (held together by van der Waals forces, for example), while breaking up the aggregates into the individual primary particles often requires considerable mechanical action (via operations like ball milling); the fracture or further diminution of primary particle size can only be achieved with extreme difficulty.

The intent behind dustiness testing is that the energy supplied should not be enough to divide the primary particles (e.g. by grinding, cutting, or crushing) within aggregates, but liberate some fraction of the loosely bound preexisting primary particles and agglomerates from the bulk powder. The fraction of airborne dust liberated from the bulk powder will be related to the chosen test conditions. The more energetic the testing protocol, the greater the fraction of airborne dust liberated from the test powder. The object of the test should be to mimic the energy supplied in a typical environmental or occupational setting, so that an assessment can be made of exposure. However, the test should be controllable and should not be subject to undue influence of uncontrolled external parameters.

Dustiness testing should be able to provide a comparison of the relative dust exposure potential of different materials. In practice, it should alert the potential for worker exposure, predict the level of that exposure, and therefore indicate a required level of control to reduce or eliminate that exposure (Heitbrink et al., 1989; Heitbrink et al., 1990; Mark, 2005; Brouwer et al., 2006; Pensis et al., 2010). Heitbrink et al. (1989) correlated dustiness measurements and worker exposure through bag packing; the results from dustiness tests were used to predict order of magnitude of worker exposures. A simulated workplace study of scooping/weighing/adding and cleaning/sweeping of powders by Brouwer et al. (2006) found that dustiness was a major determinant of worker exposure and accounted for approximately 70% of exposure variability. Dustiness has also been suggested to be relevant for assessing the potential risk of dust explosions (Cashdollar, 2000; Turkevich et al., in preparation).

A variety of methods have historically been utilized to measure dustiness. The BOHS Working Group has described many of these methods (BOHS, 1985). Several attempts have been made at standardization (ASTM, 1980; DIN, 1999 and 2006), but these have not been widely accepted. For a review of methods, see Hamelmann and Schmidt (2003, 2004, 2005).

Many of these techniques require the use of relatively large quantities of powder, typically, 10²–10³ grams per test (CEN, 2006). These quantities generally preclude the testing of nanoscale powders due to their prohibitive costs.

Further complicating matters is that no clear relationship has yet been established linking inhalation exposure to dustiness as determined by any of these methods (Cowherd et al., 1989; Heitbrink et al., 1989, 1990; Class et al., 2001; Breum et al., 2003; Brouwer et al., 2006; Petavratzi et al., 2006; Tsai et al., 2011). While dustiness is one factor related to worker exposure potential, the quantity of material, type of operation (energy), differences in worker practice, use of exposure controls, and environmental conditions (e.g. humidity, ventilation) may also play a significant role for example, and it may be difficult to account or control for all of these factors.

Historically, dustiness testing has utilized configurations that have imparted fairly gentle mechanical agitation to the powder. These tests have been devised to simulate various industrial procedures. These configurations also tend to require larger quantities of material (>10g), and the aerodynamics tend to involve large-scale eddies.

These methods can be qualitatively divided into two classes:

These methods have been compared by various groups using a variety of powders (Heitbrink 1990; Carlson et al., 1992; Plinke et al., 1992; Bach and Schmidt, 2008; Schneider and Jensen, 2008; Jensen et al., 2009). Modeling of the aerosolization and dust generation of powders has also been attempted (Plinke et al., 1994a,b; Lanning et al., 1995; Ibaseta et al., 2008).

European standard EN 15051 (CEN, 2006) specifies two reference test methods for dustiness testing (the rotating drum and continuous falling powder tests). The rotating drum method is frequently employed due to its ability to simulate a wide range of workplace material handling processes (Mark 2005; Petavratzi et al., 2006). In both methods, the energy provided for aerosolization is gravity driven and possibly enhanced by agglomerate impact. The maximum velocities of the particles achieved are v ~1 m s⁻¹.

A qualitatively different method was introduced (Boundy et al., 2006) in order to test pharmaceutical powders. The aim was to utilize small quantities (~5mg) of powder under confined conditions (cost considerations and reproducibility, and also so as not to expose the test operator to potentially toxic or pharmacologically active material). A powder is placed in a holding tube (d = 0.44cm) exterior to but piercing the dispersion chamber. Air is introduced into the dispersion chamber (5.7 l) via the holding tube at a volumetric flow rate Q = 60 l min⁻¹, resulting in nozzle air flow v ~70 m s⁻¹. Aerosolization presumably occurs via aerodynamic lift and pneumatic drag mechanisms acting on the powder; particulate velocities are one to two orders of magnitude larger than the gentler falling powder and rotating drum methods discussed above. In addition, the aerosolization proceeds under turbulent conditions (Reynolds number, Re = Qd/vA ~8000, where Q is the volumetric flow rate, d and A the tube diameter and cross-sectional area, respectively, and v the kinematic viscosity), whereas in the gentle tests, the airflows can be considered at larger scale and in the laminar regime. For a small number of the more cohesive powders, impaction residue has been detected on the far wall of the chamber (see Fig. 2h for example), but it is not believed that this mechanism significantly contributes to the aerosolized dust.

The method provides reproducible results and is relatively quick and easy to use (Boundy et al., 2006). The method does involve more aggressive air flows than those typically encountered in large-scale workplace activities, but is expected to resemble energetic dust dispersion activities (e.g. the use of compressed air to clean contaminated worker clothing or work surfaces).

There were two objectives to this study. The first was to better characterize the Venturi device, through video photography and direct reading respirable mass measurements (described in the methods section). The second objective was to rank a variety of fine and nanoscale powders according to their airborne dust generating abilities. The relationship between dustiness, and the properties of humidity, BET specific surface area and particle size was additionally investigated.

The Venturi dustiness device, utilized in these dustiness measurements, has been described in detail elsewhere (Boundy et al., 2006) but is briefly discussed below. A photograph of the chamber (without the humidity preconditioning antechamber) is presented in Fig. 1.

Air is introduced into the chamber via a side port penetrated by a metallic ‘tee’ shaped Venturi nozzle (Fig. 1 right); powder is placed in the vertical tube of the ‘tee’ (removable cap at top) and falls down the tube to reside somewhere on the curved base of the vertical tube. When air is introduced into the chamber, the primary flow is along the horizontal tube of the ‘tee’, although a small flow is also drawn down the vertical tube of the ‘tee’ (the cap is perforated by a small pinhole). This combined air flow (60 l min⁻¹ at injection) aerosolizes the powder and sweeps it into the chamber.

Air is removed from the chamber (V = 5.7 l) via three ports in the top: out flows (4.2 l min⁻¹) through the cyclone (top left port) and (2.0 l min⁻¹) through the cassette (top right port) are continuous throughout the test; a third outflow (53.8 l min⁻¹) through the central top port (observed to the left of the handle) occurs only during injection. Injection thus occurs at 60 l min⁻¹ for 1.5 s; dust is collected for 4min following injection, with the flow divided between the cassette and cyclone filters.

Experiments were conducted in Cincinnati OH, USA, which presented large seasonal variations in ambient laboratory humidity (dry winter, humid summer). It is well recognized that humidity or moisture content of powders can affect their dustiness (Plinke et al., 1995), so controlling for such effects is prudent. This was achieved in two steps. Firstly, small batches of each powder were added to clean glass vials and allowed to sit within a temperature controlled oven set to 50°C overnight, prior to weighing, to remove residual moisture. Secondly, a humidity-controlled antechamber, adjacent to the dustiness testing device, with glove and access ports and constructed from static dissipative acrylic (Plexiglas), was utilized to condition powder samples prior to dispersion (for at least 1h) at the desired humidity.

An alternative approach that is stipulated in EN 15051 (CEN, 2006) would be to test powders in the moisture state in which they arrive. The ‘as received’ approach is important for determining dustiness of powders from a particular stage of a process; under that protocol, sealed samples are sent directly to the laboratory for testing without further conditioning. Moisture content for each material is then determined and reported (CEN, 2006). While different humidity conditions may be relevant for different workplaces or operations, it is important that changes in humidity, which might alter the cohesive forces in the powder, be controlled for testing, so as not to introduce additional variation in the dustiness values. In our study, the goal was to investigate dustiness without moisture introducing additional variability.

In our experiments, humidities of 20, 50, and 80% (±5%) were achieved, by selective mixing of dry and (water) saturated air, fed into the antechamber by needle valves, followed by high efficiency particulate air filtration. The antechamber additionally served as a clean conditioned environment from which air was drawn into the dustiness chamber (through the metallic ‘tee’) during dispersion and sampling. The clean conditioned air is of lesser importance for mass measurements, but for number concentration measurements, this step is crucial to ensure a sufficiently low particle background. Flow into the antechamber was controlled so that a slight positive pressure was attained, ensuring negligible ambient air infiltration.

The dustiness tester was connected either to house vacuum or to two vacuum pumps connected through a vacuum regulator and gauge. Dispersion and sampling flows were controlled through needle valves built into the tester. Flow timing was automatically controlled within the tester by timed relays and solenoid valves. Sampling flows were calibrated each day using a Gilibrator and adjusted to 2.0 l min⁻¹ for the closed-face metallic Slaton filter cassette, for total aerosol, and 4.2 l min⁻¹ for the metallic BGI GK 2.69 cyclone with filter cassette, conforming closely to the ISO respirable sampling convention (Kenny and Gussman, 1997).

An individual test consisted of two consecutive 5mg dispersions (totaling 10mg of powder dispersed for each test) sampled onto the same filter media. The two 5mg aliquots of powder were added to two aluminum weighing boats, using a microbalance (Mettler AT20, Columbus, OH), and stored and transported via glass Petridishes prior to nozzle loading. The balance was situated within a constant humidity-controlled weighing enclosure. Filters (37mm polyvinylchloride) were pre- and postweighed using the same microbalance and transported in clean static dissipative filter holders. Each test batch typically consisted of three consecutive individual tests, with two aliquots of powder (and two filters) required per test; two additional blank filters served as a control, totaling six aliquots of powder and eight filters per batch. Preweighed powders were conditioned within the antechamber to the tester for at least 1h prior to loading the nozzle for the first test. Six tests were typically performed for most powdered materials.

CleaningThe chamber, nozzle, and samplers were cleaned between each test with an ultra-low penetration air filtered vacuum and wiping. A more thorough cleaning procedure, involving water, mild detergent, and isopropyl alcohol was used whenever a new test powder was introduced. A thorough description of the standard operating procedure is provided in Boundy et al. (2006).

Initial dispersed masses, M _disp, were measured with weigh boats prior to loading in the holding tube; masses, M _cycl and M _cass, determined from pre- and postweighed filter media, were blank corrected. Total dustiness, D _tot, and respirable dustiness, D _resp, were determined as follows:

where f _cycl and f _cass are the flow rates through the cyclone and cassette, respectively. The above ratios (mass of aerosolized dust to mass of loaded powder) are converted to percentages with multiplication by a factor of 100. Dustiness results are typically expressed as mg kg⁻¹ elsewhere (i.e. normalized in some way to the amount of starting material).

Statistical analyses of the dustiness results were performed within Microsoft Excel.

The dispersion of several powders was video-photographed (SONY HandyCam DCR-VX 2000 with 3CCD progressive scan) at 30 frames s⁻¹ to visualize the aerosolization and mixing processes within the transparent glass chamber. Figure 2 presents the progression of one such dispersion of Aeroxide P25 fumed TiO₂: the injection ‘tee’ is to the right of each figure, the back wall of the glass chamber is to the left, the cyclone protrudes at top upper left and the closed-face cassette can be observed upper right. Figure 2a (t = 0.03 s) documents the initiation of injection: a light is illuminated from the right whose reflection (white with blue halo) is visible on the back wall of the chamber at left; at this moment, all three flows (2.0 l min⁻¹, 4.2 l min⁻¹, 53.8 l min⁻¹) are open. Figure 2b (t = 0.07 s) documents the powder jet emerging from the injection ‘tee’. Figure 2c (t = 0.10 s) shows the spreading of the powder jet. Figure 2d (t = 0.13 s) shows continued spreading of the powder jet; streamers of heavier particles (large agglomerates) can be seen at the bottom of the cloud. Figs 2e,2f,2g (t = 0.17, 0.20, 0.23s) show continued spreading of the powder jet. In Fig. 2h (t = 0.27 s), impacted powder residue can be seen on the back wall (just below the illuminating light); such impacted material has been observed for only a small number of powders, and it is believed that this mechanism does not contribute significant material to the dust cloud. In Fig. 2i (t = 0.33 s), the powder cloud has dispersed throughout the chamber, and clean air, introduced by the injection tube, clarifies the central portion of the chamber. The dust cloud in the chamber mixes, and, in Fig. 2j (t = 0.77 s), a momentary second injection (of very fine material) seems to have occurred. The dust cloud continues to mix. In the final frame, Fig. 2k (t = 1.50 s), the injection air (53.8 l min⁻¹) has ceased; however, in Fig. 2l (t = 1.53 s), a final particle jet occurs (probably material dislodged by the momentary backflow when the 53.8 l min⁻¹ injection air flow is closed off, but which is then swept into the chamber by the remaining 6.2 l min⁻¹). While the injection process is complicated, even within this relatively simple geometry and protocol, it is clear that in this test, aerosolization of the test powder occurs under considerably more energetic flow conditions than with dustiness tests previously described (rotating drum and continuous falling powder).

The time evolution of the dust cloud was monitored in the following experiment with carbon nanofibers (Pyrograph III). This experiment formed part of an instrument calibration assessment for a parallel workplace study (Evans et al., 2010). The closed-face cassette was removed, and a photometer (DustTrak Model 8520, TSI Inc., Shoreview, MN) was connected (externally) to this port via a 3-way valve, with the 10-mm Dorr-Oliver cyclone providing a respirable size-selective inlet at 1.7 l min⁻¹ within the chamber. The respirable BGI cyclone was operated simultaneously and the material collected, determined gravimetrically as per the standard protocol. Following the usual dispersion, a 3-way valve was switched, such that air was drawn from the chamber into the photometer; the total volume exchange rate during this calibration test is Q = 1.7 l min⁻¹ + 4.2 l min⁻¹ = 5.9 l min⁻¹. Photometer mass concentration estimates were measured and logged every second.

Figure 3 presents a time series of respirable mass concentration. The photometer measurements provided a reasonable estimate of the respirable mass concentration for these carbon nanofibers and therefore no further calibration or correction factor was required (Evans et al., 2010). The exponential decay of concentration (τ = 55.6 s = 0.927min, R ² = 0.995) is consistent with the dominant particle loss mechanism being dilution (removal of contaminated air through sampling and replacement with clean air: Q = V/τ = 5.7 l/0.927min = 6.2 l min⁻¹). Gravitational settling and losses to the chamber walls did not appear to significantly contribute (~5%) to the concentration decay within the chamber.

Size distribution measurements by aerodynamic diameter were conducted on a subset of the gravimetrically tested powders, using an electrical low-pressure impactor (ELPI; Dekati, Tampere, Finland) and an aerodynamic particle sizer (APS, model 3021 TSI, Shoreview, MN, USA). These measurements provided overlapping particle size distributions (by number) from 27nm to 20 µm (ELPI, 27nm to 10 µm and APS from 500nm to 20 µm).

For these measurements, the two filters (cassette and cyclone) were removed and a single sampling probe was connected to both the APS and ELPI instruments; these instruments required 5 l min⁻¹ and 10 l min⁻¹ flows, respectively (replacing the 2 l min⁻¹ through the cassette and the 4.2 l min⁻¹ through the cyclone); a static dissipative Plexiglas (instead of the standard glass) chamber was used for these measurements. In the standard configuration, the sampling (through both the cassette and the cyclone) commenced simultaneously with the powder injection; with the APS and ELPI instruments, the sampling occurred only after injection was completed. Otherwise, the test was conducted as before, but using the 3-way valve arrangement.

It should also be noted that the initial loading was reduced to 1mg for the Pyrograf III carbon nanofibers, HiPCO single-walled carbon nanotubes (SWCNTs) and Aerosil 50 OX fumed SiO₂ samples; all other samples were loaded with 5mg of powder for a single dispersion. Particle size distributions (number of particles as a function of aerodynamic diameter) were measured over the entire duration of the test; reported are the distributions averaged over the first minute for each instrument.

Twenty-seven powders were evaluated. Candidate materials included single- and multiwalled carbon nanotubes (MWCNTs), carbon nanofibers (CNFs), carbon blacks; oxides of titanium, aluminum, silicon, and cerium; metallic nanoparticles (nickel, cobalt, manganese, and silver), silicon carbide, Arizona road dust; nanoclays; and a mixed metal oxide, lithium titanate. These provided a broad array of available powders. (Details of the provenance of these materials and additional physical parameters are provided as supplementary data, available at Annals of Occupational Hygiene online). Specific surface areas were determined by BET analyses (Brunauer et al., 1938) and are listed in Table 1.

Twenty-seven fine and nanoscale materials were evaluated at 50% relative humidity (Table 1 and Fig. 4). Mean average values, and standard deviations, are provided. The lactose standard, widely utilized within the pharmaceutical industry, was included in order to make contact with the earlier work conducted with this device (Boundy et al., 2006). The values of D _tot = 5.2 (±0.4)% and D _resp = 0.9 (±0.1)% for the lactose in this study are fully consistent with those reported by Boundy et al. (2006). Since the measurements presented here were made over the course of approximately 4 years, and by several operators, the Arizona Road Dust was adopted as an internal standard to verify proper operation of the equipment.

For eight materials, total and respirable dustiness was studied as a function of relative humidity at 20, 50, and 80%. Results are presented in Table 2. With the exception of the Microgrit cerium oxide, where a marked decrease in both total and respirable dustiness was observed at the elevated relative humidity, there was no detectable variation of dustiness with humidity. While it is evident that not all powdered materials may be influenced by humidity, it is a factor that needs to be controlled during testing.

Returning to the 50% relative humidity measurements (Table 1), total and respirable dustiness varied over two orders of magnitude. This is conveniently displayed as respirable versus total dustiness using logarithmic scales (Fig. 4). Powders are color coded by material class (carbonaceous, fumed oxides, nanoscale metals, fine oxides, miscellaneous). Each material class is represented over the entire range of dustiness values (see discussion below). Error bars indicate 1 SD in the dustiness measurements. These statistical errors are typically large for low values of dustiness (lower left on Fig. 4) and small for high values of dustiness (upper right in Fig. 4). The relative mass resolution by gravimetric analysis is ~10^–5 g, which is comparable to the statistical variability (~0.1%) in powders with low dustiness values.

The relatively small intertest variability (i.e. high consistency) for the majority of the test materials is indicative of the intrinsic reproducibility of the Venturi device. Boundy et al. (2006) reported similar findings. The few cases of high intertest variability are probably due to inhomogeneity of the test material (between test aliquots); test materials that contain granular particles or clumps tend to disperse irreproducibly. Presieving (i.e. with 200 mesh) of the powders prior to testing may improve reproducibility (e.g. as described in ISO 14488, 2007).By far, the most dusty material examined was the SWCNT manufactured by the HiPCO process, with D _tot = 37.9 (±3.4)%, D _resp = 31.8 (±3.3)%. This is of particular concern, because of the high respirable fraction. Furthermore, animal studies have demonstrated significant adverse health endpoints, including inflammation, rapid pulmonary fibrosis, granulomas, oxidative stress, and mutagenesis (Shvedova et al., 2005, 2008; NIOSH, 2010) with this material. This motivates a higher level of exposure control for workers handling this material. Conventional exposure controls appear to be effective in reducing exposures to carbon nanotubes and nanofibers, where effectively implemented (Dahm et al., 2012).

The next most dusty material, the Carbon Black, Printex 90, had comparable D _tot = 30.9 (±3.1)%, but significantly lower D _resp = 12.9 (±1.4)%. It is probably noteworthy that both HiPCO SWCNTs and Printex 90 materials belong to the carbonaceous category. As we proceed to less dusty materials, all categories of materials X (D _tot, D _resp) are represented: Microgrit SiC (27.4, 9.6), AluC Al₂O₃ (26.0, 12.4), Aerosil R812 SiO₂ (22.1, 10.7), Ni (16.7, 7.5), Aerosil R202 SiO₂ (16.6, 6.2), Aeroxide P25 TiO₂ (15.7, 7.2), and Aerosil 380 SiO₂ (15.0, 5.5). No one material class stands out as being excessively dusty (i.e. significantly more dusty than the others); conversely, no one material class stands out as being relatively ‘dust free’. It is instructive to examine these results by material class.

Within the carbonaceous class, total dustiness ranges from 0.3% (Sterling V) to 37.9% (HiPCO SWCNT); this represents a range of two orders of magnitude. Similarly, respirable dustiness ranges from 0.2% (Sterling V) to 31.8% (SWCNT HiPCO); again, this represents a range of two orders of magnitude. It is interesting that respirable dustiness appears to be related to total dustiness.

Within the fumed oxide class, total dustiness ranges from 3.0% (Aerosil 50 OX SiO₂) to 22.1% (Aerosil R812 SiO₂); this represents a range of one order of magnitude. Similarly, respirable dustiness ranges from 1.5% (Aerosil 50 OX SiO₂) to 10.7% (Aerosil R812 SiO₂), an order of magnitude variation.

Only four nanoscale metals were studied. The total dustiness ranges from 1.7% (Ag) to 16.7% (Ni), again, an order of magnitude range. The respirable dustiness ranges from 0.4% (Ag) to 7.5% (Ni), an order of magnitude variation.

Within the fine oxide powders, the total dustiness ranges from 0.3% (Kemira TiO₂) to 11.1% (Altair lithium titanate spinel), almost two orders of magnitude. The respirable dustiness ranges from 0.1% (Kemira TiO₂) to 3.9% (Arizona Road Dust), again more than an order of magnitude.

In the remaining miscellaneous category, the total dustiness ranges from 3.6% (nanoclay PGV) to 27.4% (Microgrit SiC), an order of magnitude. The respirable dustiness ranges from 0.4% (nanoclay PGV) to 9.6% (Microgrit SiC), more than an order of magnitude. Again, there appears to be a relationship between total and respirable dustiness.

As anticipated, for all materials and experiments, D _resp < D _tot, i.e. the respirable dustiness is always less than the total dustiness (this is a nice internal check on the method, since D _resp and D _tot are obtained from independent measurements). Furthermore, total and respirable dustiness appear to be correlated. The correlation between total and respirable dustiness is made evident in Fig. 4, where respirable dustiness, D _resp, versus total dustiness, D _tot, is plotted (log-log scale). This may be fit to a power law,

with n = 1.04 (1 SD in the exponent, ω_n = 0.07; R ² = 0.90). Since the exponent n ~1, the total and respirable dustiness are linearly related across the wide range of materials tested. The D _resp/D _tot column of Table 1 gives the ratio of respirable to total dustiness. For all the materials studied, D _resp/D _tot = 0.39±0.15 (1 SD). Some materials studied lie outside this range. Materials with an anomalously high D _resp/D _tot ratio are HiPCO SWCNT (0.84) and Sterling V carbon black (0.61); materials with an anomalously low D _resp/D _tot ratio are Mitsui VII MWCNTs (0.17), Holland lactose (0.17), nanoclays PGN (0.21), and PGV (0.10). These anomalous materials are further discussed in the size distribution measurements below.

The linear relationship observed between respirable and total dustiness was unexpected and actually quite remarkable. This relationship is fortuitous and most likely results from the selection of materials in this study (mostly fine and nanoscale powders).

The results between the Venturi method presented here may be compared with results reported elsewhere where the same powders were tested. Tsai et al. (2009) reported gravimetrically determined dustiness from the rotating drum (CEN, 2006) for the Aeroxide P25 fumed TiO₂ powder: 6713 (±546), 576 (±37), and 15 (±37) mg kg⁻¹ for the inhalable, thoracic, and respirable dustiness, respectively. For comparison, results from this study (Table 1) are 1.57 (±0.51) × 10⁵ mg kg⁻¹ for total dustiness and 7.2 (±2.1) × 10⁴ mg kg⁻¹ for respirable dustiness are obtained. If total dustiness (provided by the closed face cassette) crudely approximates that of the inhalable fraction, then the Venturi method yields a 20-fold larger total (inhalable) dustiness; and lastly, the Venturi respirable dustiness is three orders of magnitude greater. NIOSH has published a current intelligence bulletin on occupational exposures to TiO2 (NIOSH, 2011).

O’Shaughnessy et al., (2012) constructed a small-scale falling powder test and used APS counts to estimate respirable mass for several powders, two of which were also investigated here. They report respirable dustiness values of 45.3 (±8.5) mg kg⁻¹ for Aeroxide P25 fumed TiO₂ and 42.9 (±6.0) for Printex 90 carbon black. For comparison, results from this study (Table 1) are 7.2 (±2.1) × 10⁴ mg kg⁻¹ for Aeroxide P25 fumed TiO₂ and 1.29 (±0.14) × 10⁵ mg kg⁻¹ for Printex 90 carbon black. Again, the Venturi dispersion yields respirable dustiness values three orders of magnitude larger than the small-scale falling powder values. A combination of energetic dispersion and small chamber volume (5.7 l, when compared with sampled air volume, 4min × 6.2 l min⁻¹ = 24.8 l) for the Venturi device results in efficient airborne collection. It is the energetic dispersion coupled with efficient collection within the Venturi device, which makes gravimetric dustiness determination for mg quantities of fine and nanoscale materials possible.

Relationship with BET Surface Area and Primary Particle Size. The dustiness values do not appear to be correlated with primary particle size. Primary particle size is directly related to the specific surface area provided by BET analysis (Hinds, 1999). Values for BET-derived specific surface area are provided in Table 1 for each test material. Fig. 5 displays (i) the total dustiness and (ii) the respirable dustiness, as a function of BET specific surface area. Attempts to fit to power laws resulted in the following:

with n = 0.21 (R ² = 0.068), and

with n = 0.32 (R ² = 0.13). There was significant scatter in this data (very poor R ²). It would thus appear that neither total dustiness, D _tot, nor respirable dustiness, D _resp, is predictable from primary particle size.

Particle Size Distribution Measurements. The D _resp/D _tot relationship (discussed earlier in the gravimetric dustiness measurements) was further studied by making aerodynamic particle size distribution measurements, by number, for a subset of powders. The size distributions (Fig. 6) are peak-normalized and are ordered by D _resp/D _tot (PGV with D _resp/D _tot = 0.10, at top, to SWCNT-HiPCO with D _resp/D _tot = 0.84, at bottom). All D _resp/D _tot values are provided in Table 1.

The APS measurements cut off below 0.5 µm. Except for the Mitsui VII MWCNTs, Pyrograf III CNFs, and HiPCO SWCNTs, the APS and ELPI results track each other very well. It is noteworthy that these three anomalous materials are all carbonaceous and display relatively complex airborne morphologies (Shvevoda et al., 2008; Porter et al., 2010; Ku et al., 2006). It is speculated that the discrepancy between the particle size distributions, as measured by these two instruments, may arise from poor counting efficiency of the APS in the submicrometer region for optically black (absorbing) materials with complex airborne morphologies. Further experimental study is clearly warranted.

There is a general qualitative trend that materials with low (high) D _resp/D _tot have particle number distributions with modes at larger (smaller) aerodynamic diameter (Fig. 6). Unfortunately, a more quantitative relationship between D _resp/D _tot and the particle number distribution cannot be deduced. Aerodynamic particle size distribution modes range from ~300nm to several micrometers as the D _resp/D _tot ratio increases. It is also noteworthy that no aerodynamic particle size mode below 100nm was observed for any of the materials measured, despite the relatively energetic dispersal mechanism employed with the Venturi testing device. Thus a substantial sub-100nm fraction is unlikely to be encountered in the workplace from these materials.