Journal of Nanoparticle Research

, Volume 11, Issue 7, pp 1637–1650

Relevance of aerosol dynamics and dustiness for personal exposure to manufactured nanoparticles


    • National Research Centre for the Working Environment
  • Keld A. Jensen
    • National Research Centre for the Working Environment
Special Issue: Environmental and human exposure of nanomaterials

DOI: 10.1007/s11051-009-9706-y

Cite this article as:
Schneider, T. & Jensen, K.A. J Nanopart Res (2009) 11: 1637. doi:10.1007/s11051-009-9706-y


Production and handling of manufactured nanoparticles (MNP) may result in unwanted worker exposure. The size distribution and structure of MNP in the breathing zone of workers will differ from the primary MNP produced. Homogeneous coagulation, scavenging by background aerosols, and surface deposition losses are determinants of this change during transport from source to the breathing zone, and to a degree depending on the relative time scale of these processes. Modeling and experimental studies suggest that in MNP production scenarios, workers are most likely exposed to MNP agglomerates or MNP attached to other particles. Surfaces can become contaminated by MNP, which constitute potential secondary sources of airborne MNP-containing particles. Dustiness testing can provide insight into the state of agglomeration of particles released during handling of bulk MNP powder. Test results, supported by field data, suggest that the particles released from powder handling occur in distinct size modes and that the smallest mode can be expected to have a geometric mean diameter >100 nm. The dominating presence of MNP agglomerates or MNP attached to background particles in the air during production and use of MNP implies that size alone cannot, in general, be used to demonstrate presence or absence of MNP in the breathing zone of workers. The entire respirable size fraction should be assessed for risk from inhalation exposure to MNP.


AgglomerateCoagulationDustinessEHSExposureNanoparticlesOccupational healthSurface deposition


Production and handling of manufactured nanoparticles (MNP) may result in unwanted worker exposure to airborne MNP. Exposure risk is determined by the source strength, the physical properties of MNP at the time of release, and the changes in these properties during transport from the source to the breathing zone of the worker as well as by particle removal and dilution processes. Considerable insight has been gained regarding factors, or exposure modifiers, that determine the mass concentration of conventional particles in the breathing zone. Especially for insoluble MNP, it may not be sufficient to quantify MNP exposure by a mass measure. There is emerging evidence that additional quantification by size, surface area, and agglomeration is needed to better characterize health relevant exposure (Oberdörster et al. 2005). New knowledge gained by exposure measurements and modeling is needed for identifying and quantifying the exposure modifiers that determine these additional characteristics of particles in the breathing zone. This article addresses the role of some of these exposure modifiers, i.e., coagulation, transport losses, and dustiness in two types of scenarios.

The first scenario is the release of insoluble MNP after formation and before being applied or collected. An example is release of MNP from a flame spray pyrolysis process due to failure or inefficiency of control measures (Luther 2004). During transport from the source to the breathing zone of workers the concentration, size distribution, and agglomeration of the MNP change due to homogeneous coagulation, scavenging by background aerosols, losses by surface deposition, and dilution/ventilation. The degree to which these processes will affect concentration, size distribution, and agglomeration at the receptor will depend on the relative time scales of the processes (Seipenbusch et al. 2008). This will be further discussed in the sections on coagulation and surface deposition.

The second scenario is handling of bulk MNP powder after collection. Industrial MNP’s usually build aggregates/agglomerates with sizes in the micrometer range (Luther 2004). During handling of bulk MNP, dust may be released to the workroom air. An important property of the bulk MNP is the amount of particles released during handling (dustiness), and in particular, to which degree the agglomerates break up and whether primary MNP can be released. This property will be determined by a broad range of mechanisms including those that affect agglomeration and de-agglomeration in powders. Debrincat et al. (2008), testing nickel flash furnace concentrate and recycled dust, found that for dry and non-magnetic particles, the van der Waals forces are the most likely mechanism responsible for holding agglomerates together, whereas the electrostatic forces mainly are responsible for bringing the particles together. However, testing fundamentally different types of particles (PSL, SiC, TiO2, Al2O3, and CaCO3) revealed that the van der Waals bonding could not predict the dispersability (binding energy) of the ceramic particles in an air jet (Endo et al. 1997). Capillary forces, on the other hand, appear relatively strong for larger micron-size particles (Plinke et al. 1994). In addition, formation of strong material bridging leading to aggregation or even particle growth may also occur in powders. Bridging is a well-known problem in the powder industry and leads to aggregation, in addition to inter-granular fusion of powders (Szepvolgyi et al. 2001; Gbureck et al. 2005; Brockel et al. 2006; Tomas 2007; Wahl et al. 2008). Modeling the interface energies for MgO nanoparticles has suggested that the activation energy of aggregation may approach zero when very small particles are aligned along crystallographically parallel planes (Spagnoli et al. 2008). The potential implication of this mechanism could be spontaneous aggregation in bulk powders. Figure 1 summarizes some mechanisms based on model and experimental studies (e.g., Plinke et al. 1994; Endo et al. 1997; Debrincat et al. 2008) as well as the micromechanical model of powder adhesion by Tomas (2007). As illustrated in Fig. 1, flexible fibers have a high risk of becoming entangled and form quite large and stable agglomerates. For powders, Tomas (2007) also considers interlocking by, e.g., chain-branched macromolecular protein particles and locks created due to overlaps by rough particle surfaces. Several types of chain-branched molecules exist, and for MNP, both types occur. Dendrimers and hyperbranched sulphide or selenide particles (Bierman et al. 2007) are examples of chain-branched particles. The rough interface may occur in all non-facetted particles. Soft-bridging (Fig. 1) describes adhesion of particles due to adsorbed liquids (e.g., adsorbed water due to particle hygroscopicity) or sticky surfaces. The final mechanism proposed is the adhesion caused by binding over neutral or greasy organic surface functionalizations. The latter could be relevant for, e.g., nanoclays functionalized with the soapy benzalkonium chloride.
Fig. 1

Schematic overview of physical properties with potential significant impact on MNP coagulation rates and inter-particle forces

It is thus hardly surprising that size-resolved dustiness cannot, with the present state of the science, be calculated from characteristics that can be obtained from analyzing the material at rest. In consequence, experimental methods have to be used. Handling activities can be simulated in dustiness tests, i.e., bench-scale tests using an agitation mechanism supposed to simulate a range of real handling scenarios. This will be further discussed in the section on dustiness.


For the first scenario, we will consider a simple case where MNP are emitted from a source into a room that is completely mixed at all times. The time-dependent change in concentration n(Dp, t) of particles with diameter Dp in the room can then be described as
$$ {\tfrac{{\partial n(D_{\text{p}} ,t)}}{\partial t}} = {\tfrac{{S(D_{\text{p}} ,t)}}{V}} + \left( {{\tfrac{{\partial n(D_{\text{p}} ,t)}}{\partial t}}} \right)_{\text{coag}} + \left( {{\tfrac{{\partial n(D_{\text{p}} ,t)}}{\partial t}}} \right)_{\text{loss}} $$
where S is the source rate (particles s−1) of MNP and V is the room volume. The subscript “coag” refers to changes due to coagulation and “loss” to changes due to surface deposition and ventilation. For simplicity, it is assumed that the make-up air is particle free.
The change in number concentration due to Brownian coagulation can be described in continuous form as
$$ \begin{aligned} \left( {{\tfrac{{\partial n(D_{\text{p}} ,t)}}{\partial t}}} \right)_{\text{coag}} = \,&{\tfrac{1}{2}}\int\limits_{0}^{{D_{\text{p}} }} {K\left({\root {3} \of {{D_{\text{p}}^{3} - q^{3} }},q} \right)n} \left({\root {3} \of {{D_{\text{p}}^{3} - q^{3} }},t} \right)n(q,t){\text{d}}q \hfill \\& - n(D_{\text{p}} ,t)\int\limits_{0}^{\infty } {K(q,D_{\text{p}} )n(q,t){\text{d}}q} \end{aligned}$$
where K(x, y) is the Brownian coagulation rate constant between particles of diameter x and y. The first term on the right accounts for the coagulation between two particles forming a new particle with volume equivalent diameter Dp. The second term accounts for particles with diameter Dp coagulating with all other particles. The Fuchs interpolation formula is commonly used as a starting point for calculating the coagulation rate constant. It can be modified to take into account agglomerate structure as parametrized by the fractal dimension and van der Waals/viscous forces (Jacobson and Seinfeld 2004). Coulomb and magnetic forces will not be addressed.

Maynard and Zimmer (2003) modeled experimental data obtained from the decay of an aluminium aerosol with a broad size distribution (5 nm–20 μm) in a 1 m3 stirred box. The initial concentration of aluminium aerosol was generated by high speed grinding. The size distributions were measured by SMPS configured with a nano DMA (TSI 3085), and a long DMA (TSI 3934) and with an APS (TSI 3320). Complete mixing was assumed. The loss rate by surface deposition was estimated from experiments where coagulation could be neglected due to low particle concentration. Coagulation was modeled for an experiment where the initial concentration was about 106 cm−3. Using the Fuchs interpolation and assuming a particle fractal dimension Df of 1.7 in the size range 20–500 nm, the temporal evolution of the size distribution could be modeled for residence times up to 4500 s. After a residence time of 500 s and without the coagulation term, the reduction in particle number was increasingly underestimated with decreasing diameter for diameters below 70 nm, and was over a factor of 20 at 5 nm.

Jacobson and Seinfeld (2004) calculated the effect of fractal geometry and van der Waals/viscous forces on the coagulation rate constant for 10- and for 100-nm particles in a background of 10–1000-nm particles. For fractal dimension Df, a correction corresponding to Df = 1.7 was used. The van der Waals force was calculated assuming A/kBT = 200, where A is the Hamaker constant, kB the Boltzmann’s constant, and T is the absolute temperature. Including either or both, always increased the coagulation rate constant. By including correction for both fractal dimension and van der Waals/viscosity forces, the coagulation rate constant increased by a factor of 3.75 for two particles, both of size 10 nm, and by a factor of about 16 for a 10-nm particle in a 1000-nm background. For 100-nm particles in a 1000-nm background, the corrections were smaller.

Seipenbusch et al. (2008) studied coagulation of nano-sized particles (NP) released from a source with and without presence of a coarser background aerosol. The release was simulated by injecting 7–8-nm (GSD = 1.3) Pt particles at a flow rate of 83 cm3 s−1 into a stirred chamber of volume 2 m3. Background aerosol was simulated by injecting oil droplets (DEHS, high concentration, median size 200–250 nm) or monodisperse silica particles (low concentration, 1000 nm). Two scenarios were simulated. Scenario A represented a moderately strong source (4 × 106 cm−3), active for several hours in an initially clean room. For the given experimental conditions, 16 min after activating the source, a single peak at 15 nm emerged, and subsequently, a larger peak that shifted to larger particles with time emerged, with a decrease in magnitude. Maximum total concentration in the chamber was ≈2 × 105 cm−3. Scenario B represented the release into a typical workspace with varying background concentrations. The background aerosol concentration and size were chosen to be representative of indoor conditions of typically 103 to 104 particles cm−3 and sizes in the range 100–1000 nm (accumulation mode). In the experiments, a well-mixed background of oil or silica particles was present prior to activating the source. The source was active during the entire test period while the background aerosol was not replenished. At high background concentration 3 × 105 cm−3, virtually all Pt particles became attached to the background aerosol, while at lower concentration the 15-nm peak was visible. By suddenly increasing the background concentration (5 × 104 cm−3) through a pulse injection of oil aerosol, the nanoparticle concentration dropped to its new steady-state level in less than 4 min, the time resolution of the experiment.

For the experiments by Seipenbusch et al. (2008), the source particles, i, were approximately monodisperse and much smaller than the (approximately monodisperse) background particles, j. Equations 1 and 2 for the evolution of the source particle concentration, ni, in the box can then be written as (Seipenbusch et al. 2008):
$$ {\frac{{\delta n_{i} }}{\delta t}} = {\frac{{Q_{\text{S}} }}{V}}n_{{i,{\text{S}}}} - n_{i} \left( {n_{j} K_{i,j} + {\frac{{Q_{\text{V}} }}{V}}} \right) - {\tfrac{1}{2}}K_{ii} n_{i}^{2} \quad ( = 0\quad {\text{at}}{\kern 1pt} {\kern 1pt} {\kern 1pt} {\text{steady}}{\kern 1pt} {\kern 1pt} {\kern 1pt} {\text{state}}) $$
where ni,S is the NP source concentration, QS is the volumetric source flow rate into the box, QV is the volumetric ventilation rate of the box, ni the concentration in the box of the primarily released NP, V is the box volume, Ki,j is the coagulation rate constant for NP and background particles, Kii is the coagulation rate constant for homogeneous coagulation (Kii as used by Seipenbusch et al. (2008) includes the factor ½ that corrects for counting collisions between two identical particles twice), and nj is the background aerosol concentration. It is assumed that background concentration is constant and that generation of particles by homogeneous coagulation can be neglected, i.e., that all source particles are scavenged by the larger background particles (this will not change diameter as a result). Dimensional analysis of the parameters in this model was used to derive a non-dimensional scaling parameter, Z, being the ratio of the loss rate by coagulation and the source rate
$$ Z = {\frac{{K_{ij} n_{j} n_{{i,{\text{S}}}} }}{{{\frac{{Q_{\text{S}} n_{{i,{\text{S}}}} }}{V}}}}} = {\frac{{{\text{VK}}_{ij} n_{j} }}{{Q_{\text{S}} }}} $$
It was found that the experimental data could be approximated by the relation
$$ {\frac{{n_{i} }}{{n_{{i,{\text{S}}}} }}} = {\frac{\text{const}}{Z}} $$

It was anticipated that additional measurement results may provide more accurate correlations in the future.

Equation 5 can be derived directly from Eq. 3 at steady state under the assumption that the ventilation term (QV/V) and the loss by homogeneous coagulation (Kiini2) can be neglected.

In Eq. 3, surface deposition was neglected. The loss rate by ventilation λv = (QV/V) can be generalized to a total loss rate, λ, by including the loss rate by surface deposition, λs, i.e., λ = λs + λv.

Equation 3 at steady state is quadratic in ni and can thus be solved easily. Figure 2 shows how ni varies for varying total loss rate, λ, other parameters fixed. We have used a box volume (2 m3) and source flow rate (83 cm3 s−1), source particle size (10 nm), background particle size 1000 nm, source concentration (106 cm−3) and background particle concentration (104 and 103 cm−3) similar to the conditions in the experiments by Seipenbusch et al. (2008). The coagulation rate constant, as given by Jacobson and Seinfeld (2004), both without and with correction for fractal dimension and van der Waals/viscosity force, was used. Results are shown for no correction for fractal dimension and van der Waals/viscosity force (Df = 3, coagulation constants Kii = 2 × 10−9 cm3 s−1, Kij = 3.75 × 10−7 cm3 s−1) and with correction (Df = 1.7, Kii = 8 × 10−9 cm3 s−1, Kij = 60 × 10−7 cm3 s−1). Actually, Kii was almost only influenced by the van der Waals/viscosity correction, while Kij was almost only influenced by the fractal correction.
Fig. 2

Steady state concentration of particles, Eq. 3. Source concentration ni,S = 106 particles cm−3, size 10 nm. Source volume flow rate = 83 cm3 s−1. Chamber volume 2 m3. Background particle size 1000 nm and concentration nj 104 and 103 # cm−3. No correction for fractal dimension Df and van der Waals/viscosity force (Df = 3, coagulation constants Kii = 2 × 10−9 cm3 s−1, Kij = 3.75 × 10−7 cm3 s−1), and with correction (Df = 1.7, Kii = 8 × 10−9 cm3 s−1, Kij = 60 × 10−7 cm3 s−1)

Seipenbusch et al. (2008) concluded that NP would not reach the receptor in the form of the primarily released aerosol, and thus, that workers would most likely be exposed to NP as agglomerates or as attached to background particles. The authors stressed that NP scavenged by the coarser background particles may be invisible in the particle size spectrum. Taking into consideration the model and parameter uncertainty, it was proposed that Eq. 3 or 5 could be used to predict at least what would be the governing scenario for a given workplace environment. Figure 2 suggests that for calculating the coagulation rate coefficient in Eqs. 35 correction for fractal dimension and van der Waals/viscosity force should be included.

Field measurements support the role of coagulation for the size distribution and structure of MNP in the breathing zone. Tsai et al. (2008b) studied a scenario somewhat similar to the scenarios described above. They measured emission of nanoalumina (primary particle size 27–65 nm) from a twin screw extruder for production of nanocomposites. Airborne particles were measured with FMPS (TSI 3091) and corrected for particles generated from the polymer and general background. Particles were also sampled for TEM analysis. At the source, particles were found in the size range 10–30 and 50–400 nm and only particles in the large size range were found in the breathing zone. The size distribution was qualitatively confirmed by TEM analysis. The authors concluded that this change was due to coagulation.

Very localized sources can generate meandering plumes of contaminated air that are quite persistent (Ayer and Yeager 1982) thereby locally maintaining a high concentration and thus high homogeneous coagulation rate. Furthermore, complete mixing on a time scale shorter than the time scale for coagulation or surface deposition cannot be assumed in large rooms and uni-directional air flows may be present. For modeling such cases, models providing spatial resolution will be needed. Kuhlbusch et al. (2008) used CFD to follow the fate of an accidental release of nanoparticles from a point source into a particle free workspace. The work space chosen for modeling consisted of a reactor with tubings, benches, and cabins. The source concentration was 1010 TiO2 particles cm−3, particle diameter 50 nm with geometric standard deviation 1.3. It was found that the mode diameter had almost doubled few meters from the source.

Surface deposition

The surface deposition velocity v (cm s−1) is defined via the surface deposition flux J (particles cm−2 s−1) and particle concentration n (cm−3) outside the surface boundary layer as
$$ J = vn $$
For diffusion through the surface boundary layer at steady state in the presence of an external force field inducing a drift velocity vf normal to the surface, the flux can be described as
$$ J = - (\varepsilon + D){\frac{\partial n}{\partial y}} + iv_{\text{f}} n $$
where ε is the particle turbulent diffusivity, D is Brownian diffusivity, and y is distance from surface. For a drift towards the surface i = 1, and if away from the surface, i = −1. A review of the various approaches to solving Eq. 7 where the friction velocity u* is used as scaling parameter can be found in, e.g. Lai and Nazaroff (2000) and Lai (2002), to which the reader is referred for details. Let v0 be the deposition velocity in the absence of drift velocity (e.g. vertical wall without electrical or thermal fields). Then, the deposition velocity vtot in the presence of a drift velocity vf can be obtained as
$$ v_{\text{tot}} = {\frac{{i \times v_{\text{f}} }}{{{ \exp }\left( {{\frac{{i \times v_{\text{f}} }}{{v_{0} }}}} \right) - 1}}} $$
The loss rate by surface deposition in a stirred box can be determined as
$$ \lambda_{\text{s}} = (A_{\text{floor}} v_{{{\text{tot}},{\text{floor}}}} + A_{\text{wall}} v_{{{\text{tot}},{\text{wall}}}} \, + A_{\text{ceiling}} v_{{{\text{tot}},{\text{ceiling}}}} )/V $$
where A is the surface area and V box volume. Equations 8 and 9 can be generalized to include surfaces of any orientation (Crump and Seinfeld 1981).

Experimental data on particle loss rates due to surface deposition by gravity, Brownian and turbulent diffusion show a large scatter between different studies. A common feature is that the diameter dependent loss rate has a U-shape with minimum at around 300 nm which is also predicted by Eq. 9. Loss rates of the order 10 h−1 have been reported for room-sized chamber studies for 10-nm particles, this being of the same order as loss rates for 10-μm particles. For 300-nm particles, loss rates in the range 0.01 to over 1 h−1 have been reported. For a review see Lai (2002). Gong et al. (2009) measured the surface deposition loss rates of diesel exhaust particles in ventilated car cabins. This represents scenarios with large surface to volume ratio, large surface roughness, and relatively high air velocities. They found loss rates of ≈10 h−1 for 30-nm particles, and ≈4 for 100-nm particles. In conclusion, there are realistic scenarios where surface deposition competes with scavenging (Fig. 2) and thus has to be included in scavenging models such as Eq. 3.

Thermophoresis can either increase or decrease surface deposition velocity depending on the sign of the temperature gradient at the surface. Modeling suggests that thermophoresis can affect local airborne concentration of MNP in work spaces (Kuhlbusch et al. 2008), but in general, information on surface temperatures and boundary layer flow fields in industrial settings to a degree of detail allowing the effect of thermophoresis to be modeled is not available.

Charged nanosized particles can obtain high migration velocities in electrical fields. Even for particles having a Boltzmann charge distribution, the presence of an electrical field at a surface increases the deposition rate but the increase is negligible for particles below 10 nm, because the fraction of such particles carrying a charge is very small (McMurry and Rader 1985). For particles at saturation charge and assuming a field strength of 20 kV m−1 at the surface and the right polarity, the deposition velocity can be ≈1 cm s−1 for 10-nm particles (Chen and Lai 2004). If such a field was generated over a total collection surface area of ≈500 cm2 in a chamber of 2 m3, it would cause a loss rate λs = 1 h−1. This would likely result in only a minor reduction of the mean residence time of primary MNP but would greatly enhance the rate of contamination of this surface.

Lai and Chen (2006) used CFD to calculate surface deposition of uncharged particles in a ventilated enclosure. They explicitly calculated how large a fraction, f, of particles would deposit on the internal surface, 1 − f being the fraction removed by ventilation. The enclosure measured 0.4 × 0.8 × 0.4 m with supply of air contaminated by particles in one end and exhaust in the opposite end. Two levels of air exchange rates λv were used; λv = 5 h−1 and λv = 10 h−1. For 10-nm particles they found f ≈ 0.3 decreasing to ≈0.08 for 30 nm and to ≈0.02 for 100-nm particles. For 1-μm particles, f ≈ 0.1 increases to f ≈ 0.8 for 10-μm particles. For the high air exchange rate, the fraction was almost unchanged for the 10- and 100-nm particles. This is because an increase in air exchange also increases deposition velocity. Even though this scenario does not represent a ventilated enclosure containing a source, the results suggest that inside ventilated enclosures containing high concentrations of MNP, significant MNP containing surface contamination can accumulate even after short periods of time. In large rooms where the room surface to volume ratio is smaller and where airborne MNP concentration will be much lower, surface deposition will be a minor determinant of airborne concentration, but MNP surface contamination can still build up over longer time periods.

Experiments by Kousaka et al. (1980) have shown that wind velocities of 100 m s−1 parallel to the surface did not resuspend agglomerates of nano-sized particles from a glass plate for agglomerate size below 1 μm. Experiments on resuspension caused by human indoor activities such as walking and cleaning have focused on particles above 1 μm. The resuspension rate in general increases for particles increasing in size from 1 μm to over 10 μm (Qian and Ferro 2008). It thus can be expected that human activities will resuspend MNP attached to super-micron particles. Tsai et al. (2008a) found that water-based cleaning of floor and all laboratory equipment in a laboratory where MNP were handled greatly reduced the background concentration of airborne particles in the size range 30–100 nm. This strongly suggests that agitation of surfaces caused by normal laboratory work also can resuspend particles in the nano size range. In conclusion, surfaces should be cleaned at regular intervals to prevent them from becoming secondary sources of MNP containing particles.

Dustiness testing

Several bench-scale tests of material dustiness have been used by industry and research laboratories (Gill et al. 2006; Petavratzi et al. 2007). The commonly used bench-scale dustiness testing with relevance to occupational exposure can be grouped according to the following four methods of agitation; single drop, continuous drop, rotating drum, and fluidization. In the single drop test, a given amount of material is dropped as a single bolus from a well-defined height over the surface of impact. It dates back to Andreasen et al. (1939) who studied how dust generation was influenced by drop height and the fraction of fine particles in the test material. It has been found that the continuous drop method, whereby the sample is released as a stream rather than a single bolus had greater reproducibility, and it was suggested that the continuous drop method would simulate more commonly found industrial scenarios (Davies et al. 1988). The rotating drum probably dates back to the “Roche” Friabilator (Shafer et al. 1956). It was chosen in favor of shaking, rolling etc, for testing tablet disintegration during packaging and handling. This principle exerted both an abrading and impact action on the test material. A rotating drum dustiness tester that separates the generated dust into the biologically relevant size fraction inhalable, thoracic, and respirable mass was developed in the UK (HSE 1996). Fluidization has been used by, e.g., Sethi and Schneider (1996) and Maynard (2002).

The BOHS Working Party on Dustiness which was established almost 30 years ago had the vision that it would be possible to identify test methods that gave adequate reproducibility and to relate a number of different such methods to a common 10-point Dust Index Scale using a set of standard powders (Davies et al. 1988). Later research has shown that this can only be obtained if the test methods are closely related. The European standard for dustiness testing (CEN 2006) specifies two reference test methods; the rotating drum and the continuous drop method. The standard methods provide material dustiness as respirable, thoracic, and inhalable mass per mass of tested material. Data are presented showing that these two methods do not rank dustiness similarly and in the standard it is acknowledged that “no single test method is likely to represent and reproduce the various types of processing and handling in industry”. Users should therefore choose the one of the two methods that is most appropriate for the material and handling process they wish to simulate. A criterion for equivalence is specified to allow use of other versions of the two test principles if proven equivalent against a specified set of test materials.

The need for methods that are closely adapted to specific activities has been stressed by, e.g., Bach and Schmidt (2008). In order to meet this need they proposed that several reference methods be included in the EN 15051 or, alternatively, a standard be developed that would only standardize the measuring circumstances and evaluation of results. Such an approach would allow a broad variety of techniques to be used. The disadvantage would be that there could be incentives to deliberately choose that test method that would give the most favorable dustiness value for a given material. To counteract this negative effect one could, e.g., develop a well-defined metric of closeness of test to a specific workplace activity and give guidance on an acceptable range of degree of closeness. Efforts should however still be made to improve the understanding of which factors drive a dustiness index so that one could extrapolate from as few reference methods as possible to any given specific activity. Four patterns of the time dependent release rates of respirable particle volume during the rotation period were identified among MNP, ultrafine, and other powders (Schneider and Jensen 2008; Jensen et al. 2008); Short initial burst, and decreasing, constant, and increasing rate with time. Thus, a single drop (and likely a continuous drop) and a rotating drum test will give results that are not comparable across materials from all four classes of temporal behavior. Schneider and Jensen (2008) have suggested that a dual dustiness characterization by continuous or single drop and by rotating drum might provide the information needed for predicting exposure for a very broad range of activities.

Heitbrink et al. (1989) noted that dustiness testing had greater discriminating power than work place measurements, because the influence of within and between worker variability caused by differences in activity and work practice and other external exposure modifiers is eliminated. This discriminating power of dustiness tests is useful for comparing products or in product development aiming at minimizing dustiness. Using the continuous drop method it has been found that dustiness of 135 different materials, measured as respirable mass ranged four orders of magnitude (CEN 2006). Dustiness thus is an important exposure determinant. Only few attempts have been made in establishing a practical relationship between measured dustiness and actual dust exposure at the work site (Cowherd et al. 1989; Heitbrink et al. 1990; Brouwer et al. 2006). A close relationship cannot be expected because there are many other exposure determinants than dustiness. As an example, the conceptual model of Tielemans et al. (2008) includes nine modifying factors: Intrinsic emission potential (related to dustiness), activity emission potential, local controls, separation, segregation, surface contamination, dilution, personal behavior, and respiratory protective equipment. Future efforts to establish a relationship between dustiness and exposure should include such other exposure modifying factors.

Introduction of MNP’s has created new interest in dustiness testing extended to provide information on size, surface area, and agglomeration of the particles released during handling of these new powder types. Maynard (2002) used a two-component fluidized bed to study the agglomerate size distribution originating from ultrafine TiO2 (Table 1). Particles were quantified by SMPS (TSI) and APS (TSI 3320). Two size modes were found (Table 1). The dust was also collected onto TEM-grids using a point-to-plane electrostatic precipitator. TEM analysis confirmed the presence of the two modes, but the physical diameter of the coarse mode was slightly larger. An additional mode at about 50 nm and containing particles with diameter down to about 10 nm was also identified. The authors hypothesized that this mode was an artefact caused by the breaking away and deposition of small (partially sintered) aggregates from larger deposited agglomerates due to aerodynamic or electrical forces. Using the same fluidized bed for testing SWCNT could not generate detectable amounts of particles above background (Baron et al. 2003).
Table 1

Modes of diameter distribution by number obtained from dustiness tests


Specific surface area m2/g

Test method


Modes of diameter distribution by number



Mobility <500 nm

Aerodynamic >0.5 μm

Nano-TiO2 (Degussa P25)

Two-component fluidized bed





(Maynard 2002)

Laser ablation generated SWCNT

Two-component vortex shaker

15 min test duration



Initial but rapidly decreasing release of particles in 100–400 nm mode. Slowly decreasing 2–3-μm mode

(Maynard et al. 2004)

Fumed alumina

Two-component vortex shaker

15 min test duration




(Baron et al. 2003)

High Pressure Carbon Monoxide SWCNT

Single component vortex shaker

5 min test. Agitation from none to strong



At low levels of agitation few particle released giving large measurement uncertainty. At high level of agitation a clear bi-modal size distribution

(Maynard et al. 2004)


SiO2 aerosil 200

Single drop

Drop height 50 and 100 cm



Electrical low pressure impactor (ELPI). Uncertain presence of mode below about 200 nm due to large measurement uncertainty and potential artefacts due to agglomerate breakage in the ELPI

(Ibaseta 2007)

TiO2 G5



Al2O3 (Baikalox A125)



Silica fume


Small rotating drum

1 min rotation



Mode diameters similar for the single drop test

(Schneider and Jensen 2008)


Ultrafine TiO2



















Nanoclay Nanofil®5-C


Small rotating drum

1 min rotation



Mode diameters similar for the single drop test

(Jensen et al. 2008)


Nanoclay Nanofil®5-C, compacted.





Nano-TiO2 (Degussa AEROXITE P25)

EN15051 rotating drum

30 min rotation



Only respirable fraction analyzed. Mode diameter increased (↑) or decreased (↓) within the stated range during the test period. Virtually no particles below 70 nm

(Tsai et al. 2008c)

Fine ZnO (Sun Beam, Grade A)



For details, see text

–, Not specified

aTheoretical value

A new method that could impart a higher energy and requiring only milligram amounts was developed for testing carbon nanotubes (Baron et al. 2003; Maynard et al. 2004). Their method was based on a vortex shaker. In the two-component version 70 micron bronze beads were added. However, due to problems with background generated by the beads and to better simulate handling the one-component version (no bronze beads) was also used. Particles were quantified by a SMPS (TSI 3080) configured with a Nano DMA (TSI 3085) and a Long DMA (TSI 3934) and APS (TSI 3320). The results are shown in Table 1.

Ibaseta (2007) used a single drop apparatus with variable drop height and quantified dust generation using the Electrical Low Pressure Impactor (ELPI), see Table 1. It was cautioned that agglomerate breakage and bouncing in the ELPI may have affected the results obtained from the smallest diameter stages in the ELPI.

The EN 15051 dustiness test methods and many others require hundreds of grams of test material. This is a disadvantage when testing MNP that can be both expensive and be potentially highly toxic. Schneider and Jensen (2008) thus developed a rotating drum test based on a down-scaled version of the EN 15051 rotating drum. One test requires only 6 g of test material. The generated dust was quantified by a FMPS (TSI 3091) and an APS (TSI 3321) and by collecting on a filter for weighing. The drum has three lifter vanes. At start of the test, the drum was rotated exactly 180° so that the test material dropped exactly once (single drop test). After the dust concentration had decayed to background the drum was rotated for 1 min (rotation test). The identified size modes of the generated dust for the rotation test are presented in Table 1. Schneider and Jensen (2008) and Jensen et al. (2009) found that the geometric mean diameter of the size modes was similar for the single drop and the 1-min rotation test indicating that the diameter of each mode was independent of duration of agitation.

The data by Schneider and Jensen (2008) have been used to calculate the surface-weighted diameter distributions of the dust generated from dustiness tests (Fig. 3). Surface weighting was based on the mobility respectively aerodynamic diameter assuming spherical shape. Figure 3 shows that for the tested materials, the surface of particles with electrical mobility diameter in the nano size range only constitutes a very small fraction of the total particle surface area.
Fig. 3

Surface-weighted diameter distribution of dust generated in dustiness tests based on counts accumulated over 1 min

Tsai et al. (2008c) tested TiO2 and ZnO2 using the EN 15051 rotating drum with a modified sampling train whereby the respirable dust was characterized using SMPS (TSI 3963), APS (TSI 3321), and MOUDI (MSP 110), see Table 1. It is interesting to note that even after 30 min rotation the mode diameters changed by less than 10%. The MOUDI results were compared with the APS distributions (converted to mass). According to the authors this comparison indicated that large agglomerates could deform in the acceleration nozzle thereby appearing smaller and that the large geometric diameter of TiO2 particles as compared with the aerodynamic diameter could lead to larger losses in the inner nozzle of the APS. It was not discussed whether breakage of agglomerates in the APS could occur.

A size mode in the sub-micron range has also been found at work places. Tsai et al. (2008b) determined personal MNP exposure during full-scale simulation of manual handling of nanoalumina and nanosilver in fume hoods. Real-time measurements were made with the FMPS (TSI 3091) at 1-s time resolution, and all measurements were corrected for average background. Nanoalumina had a primary particle size ranging from 27 to 56 nm and the particles formed agglomerates in the bulk material with nominal size 200 nm. Nanosilver particles had an average particle size of 60 nm. When handling 100 g of nanoalumina, the authors consistently found a size mode with diameter ≈200 nm. There was an inconsistent presence of a size mode spanning the size range 10–40 nm. When handling 15 g nanosilver a marked size mode was also identified. The diameter of this mode was in the range 70–200 nm. Particles above the range of the FMPS (560 nm) were not measured. However, SEM images showed presence of super-micron agglomerates of both alumina and silver.

Discussion and conclusions

Modeling of coagulation and other aerodynamic processes, experiments, and field measurements have provided useful knowledge about factors that determine the presence in the breathing zone of workers of MNP as primary particles, as agglomerates, or as attached to larger background particles. These studies have shown that homogeneous coagulation and scavenging by larger background particles can be fast processes in real MNP exposure scenarios. This suggests that workers in such scenarios are exposed to MNP that are present in larger size classes than those in which they were emitted into room air. More generally, the apportionment of MNP to the three particle structures; primarily released MNP, agglomerated primarily released MNP, and scavenged MNP is determined by MNP and background aerosol concentration, size distribution, and residence time in the room air of the particles. Banding of MNP production scenarios according to this apportionment will likely be possible. Scenario banding can facilitate selection of measuring strategy and interpretation of the measurement results for choice of risk management measures.

In order to make further progress in predictive exposure modeling, for design of measurement strategies, and for interpretation of measurement results, experimental and field studies of MNP source emission rates and size distribution and, e.g., charging and magnetic properties of the emitted MNP will be needed.

Surface contamination by MNP should also be systematically studied for both MNP production and user scenarios as it is a potential secondary source of airborne MNP containing particles.

Dustiness testing combined with online size-distribution measurements provides insight into the state of agglomeration of particles released during handling of bulk MNP and the role of duration and intensity of agitation for breaking agglomerates. All non-fibrous MNP size distributions measured with SMPS or FMPS listed in Table 1 showed distinct size modes with the smallest mode having a GMD above 100 nm. The emitted particles are thus in the “classical” size region. In addition, the mode diameters did not change during a 1-min rotation period or less than 10% during a 30-min rotation period. It is conceivable that this stability (or self-preservation of size distribution) is caused by some kind of equilibrium between agglomerate breakage/re-formation and coagulation in the interface between agitated powder surface and surrounding air. There is a need for further experimental and theoretical research in this area to achieve a more general understanding of size distribution and state of agglomeration of particles released during handling of bulk MNP.

Breakage of agglomerates also has implications for sampling and lung deposition. Maynard (2002) hypothesized that the shear stresses in some aerosol sampling, dilution, or measurement instruments might break agglomerates and Ibaseta et al. (2007) found indications that aggregates broke off from larger agglomerates of aggregates during their passage through the smaller stages of the ELPI. Stahlmecke et al. (2008) showed that the fraction of TiO2 below, e.g., 70 nm increased after exposure to shear forces generated in a small orifice at pressure differences 0.4 Bar. In order to better understand the properties of powder inhalers, de-agglomeration in powder inhalers has been studied by, e.g., Li and Edwards (1997). They performed numerical simulations of de-agglomeration due to the aerodynamic shear stress in the mouth and throat (particularly in the larynx). It was found that de-agglomeration of particles of diameter less than 10 μm can occur. Part of the non-respirable agglomerates thus may break up and pass trough a respirable size selector whereby the respirable fraction or the total surface in the respirable fraction may be overestimated. Similarly, the laryngeal jet may produce respirable agglomerates that were not present in the air prior to inhalation. Thus, respirable dust as determined by a size fractionating sampler or by a sizing instrument may not correctly assess the fraction that actually penetrates into the alveolar region. Whether this is a problem of practical significance needs to be investigated.

The dominating presence of MNP agglomerates or MNP attached to background particles in the air at production and user sites implies that size characterization of airborne particles alone in general cannot be used to demonstrate presence or absence of MNP in the breathing zone of workers. It also implies that until more knowledge has been obtained regarding the role of agglomeration and attachment to background particles for the toxicological properties of MNP and the degree of de-agglomeration in the lung fluids, it would be prudent to include the entire respirable size fraction in risk assessment of inhalation exposure to MNP.


This study is part of the NANOTOOL project funded by the Occupational Safety and Health Advisory Boards for the Industry, Teaching and Research, and the Ministry of Science, Technology and Innovation in Denmark.

Copyright information

© Springer Science+Business Media B.V. 2009