Journal of Nanoparticle Research

, 11:1625

Investigation of airborne nanopowder agglomerate stability in an orifice under various differential pressure conditions


    • Institute of Energy and Environmental Technology (IUTA)Air Quality & Sustainable Nanotechnology Unit
  • Sandra Wagener
    • Institute of Energy and Environmental Technology (IUTA)Air Quality & Sustainable Nanotechnology Unit
    • Department of GeographyHumboldt University
  • Christof Asbach
    • Institute of Energy and Environmental Technology (IUTA)Air Quality & Sustainable Nanotechnology Unit
  • Heinz Kaminski
    • Institute of Energy and Environmental Technology (IUTA)Air Quality & Sustainable Nanotechnology Unit
  • Heinz Fissan
    • Institute of Energy and Environmental Technology (IUTA)Air Quality & Sustainable Nanotechnology Unit
    • Center for Nanointegration Duisburg-Essen, CeNIDE
  • Thomas A. J. Kuhlbusch
    • Institute of Energy and Environmental Technology (IUTA)Air Quality & Sustainable Nanotechnology Unit
    • Center for Nanointegration Duisburg-Essen, CeNIDE
Special issue: Environmental and human exposure of nanomaterials

DOI: 10.1007/s11051-009-9731-x

Cite this article as:
Stahlmecke, B., Wagener, S., Asbach, C. et al. J Nanopart Res (2009) 11: 1625. doi:10.1007/s11051-009-9731-x


The stability of agglomerates is not only an important material parameter of powders but also of interest for estimating the particle size upon accidental release into the atmosphere. This is especially important when the size of primary particles is well below the agglomerate size, which is usually the case when the size of primary particles is below 100 nm. During production or airborne transportation in pipes, high particle concentrations lead to particle coagulation and the formation of agglomerates in a size range of up to some micrometers. Binding between the primary particles in the agglomerates is usually due to van der Waals forces. In the case of a leak in a pressurized vessel (e.g. reactor, transport pipe, etc.), these agglomerates can be emitted and shear forces within the leak can cause agglomerates to breakup. In order to simulate such shear forces and study their effect on agglomerate stability within the airborne state, a method was developed where agglomerate powders can be aerosolized and passed through an orifice under various differential pressure conditions. First results show that a higher differential pressure across the orifice causes a stronger fragmentation of the agglomerates, which furthermore seems to be material dependent.


AgglomeratesAggregatesShear forcesDispersionBinding energyOrifice flowEnvironmentEHS


Nanoagglomerate powders are used in a variety of applications such as colour pigments in paints (e.g. Liz-Marzán 2004; Quinten 2001), as carriers for instance during drug delivery (e.g. Panyam and Labhasetwar 2003; Soppimath et al. 2001) or sun screens (Schulz et al. 2002). Recently, the use of powders containing nanoscale particles is becoming more and more common, as they can provide new functionalities and applications (Maynard 2007), such as in scratch-resistant paint or self-cleaning surfaces. Nanoscale titanium dioxide (TiO2) particles are, e.g., used in cosmetics (Kaur and Agrawal 2007), since they become transparent as opposed to microscale particles, but maintain their other functionality. Other materials promise to yield the same effects as powders consisting of primary particles in the micrometer size range by using much less material, i.e. during use in an automotive catalytic reactor. Furthermore, due to their unique properties, they show the perspective of new functionalities and, therefore, new applications (Anselmann 2001; Siddiquey et al. 2008). The production of nanoagglomerate powders has been significantly increased over the past few years. Hence, the likelihood of accidental release of these particles is also increased and possible implications need to be investigated.

Primary particles are produced, e.g., by pyrolysis inside a closed vessel. Particle production in the gas phase usually takes place at atmospheric or sub-atmospheric pressure. Recently, new methods of particle production applying erosive laser plasmas or applying a gas dynamic reactor working in an overpressure environment are established (Gnedovets et al. 1996; Dannehl 2007; Grzona et al. 2009). Therefore, the possibility of an accidental particle release from a leak in a pressurized vessel has to be taken into account.

In the dry state and high concentrations usually dominant during production, the primary particles tend to form agglomerates (consisting of primary particles and/or tightly bound aggregates) of a size of up to some micrometers (Hinds 1999). Chemical bonds may also be formed by chemical or physical reactions (e.g. sintering during the passage of a high temperature zone inside the vessel), which lead to the formation of tightly bound aggregates consisting of several primary particles. These aggregates are usually stable in a sense that they cannot be broken down into primary particles once they are formed. Agglomerates, found in nanopowders, are commonly agglomerations of aggregates and/or primary particles. The most prominent mechanism for uncharged particles to form agglomerates is coagulation induced by Brownian motion of the particles. Binding between the particles takes place due to van der Waals forces. The effect of van der Waals force on bonding of spherical particles was described in detail by Hamaker (1937). Furthermore, electrostatic or, in the case of magnetic particles, magnetostatic interactions may lead to bigger agglomerates. The binding energy in an agglomerate is greatly enhanced in the presence of water within the powder. The effect of a meniscus building up between two primary particles may lead—depending on the distance between the particles and the relative humidity—to an immense increase of the binding energy (e.g. Lian et al. 1993; Megias-Alguacil and Gauckler 2009; Rabinovich et al. 2005).

A variety of methods exists to study the binding energy between primary particles or aggregates. The most basic method is to investigate the energy necessary to split up a couple of two primary particles. This can be achieved by using atomic force microscopy (AFM; Blum 2006), when one of the particles is attached to the tip and the other is attached to a substrate. By increasing the distance between the tip and substrate, the necessary energy to split up the agglomerates can directly be measured. The drawback of this method is the limited number of experiments which can be conducted in an adequate time and the simplification of studying the bonding between only two particles or agglomerates/aggregates.

Another method to study the binding energy of airborne agglomerates is the production of primary particles of the desired size and the subsequent controlled agglomeration (Froeschke et al. 2003; Seipenbusch et al. 2007). The agglomerates are formed in the airborne state and are subsequently deposited by impaction under various low pressure conditions on a grid for transmission electron microscope (TEM) investigations. The differential pressure leads to an acceleration of the particles and thus to a high impact velocity. This corresponds to a high impact energy which can lead to a deagglomeration of the agglomerates. Subsequently, this effect can be studied by high-resolution electron microscopy, in which a gap between primary particles indicates the breakup of the bond between these particles. By comparing the number of broken bonds of the impacted agglomerates with the number of bonds of ‘softly’ deposited agglomerates, the binding energy can be calculated for a known deposition velocity and an estimated agglomerate mass. In contrast to the AFM method, statistical information of the binding energy can be achieved, but the number of experiments is still limited. The effect of impact fragmentation on compact agglomerates is also studied numerically (e.g. Thornton et al. 1996; Sator et al. 2008), which helps to understand the microscopic factors influencing the fragmentation behaviour.

In both aforementioned cases, only a limited number of agglomerates can be investigated due to the tedious analysis of the experimental results. Furthermore, these experiments were designed to reflect ‘perfect agglomerates’, in the sense, that they are intentionally produced for these studies to determine material properties, whereas the intention of the study presented here was to investigate the stability of industrially produced nanopowders.

In this article, we present a novel method applicable to aerosolized nanopowders, which is based on the deagglomeration in an orifice under various differential pressure conditions. The fact that agglomerates do get destructed during the passage of an orifice is common knowledge and has been applied before for the deagglomeration of micron-sized agglomerates (Fonda et al. 1999). Similar experiments were also conducted by Kurkela et al. (2008) on particles in the same size range. They studied the deagglomeration process within a turbulent stream and found out that this method can effectively be used to separate compact particles from each other. The study presented in this paper allows the comparison of the stability of different nanoparticle agglomerates with each other. The deagglomeration of some typical nanopowders mainly based on metal oxides was studied and the results are presented.

Experimental setup

In order to mimic the flow and shear forces and the resulting deagglomeration in a leak in a pressurized vessel, nanoagglomerate powders were aerosolized and passed through an orifice under various differential pressure conditions. The number size distribution of the aerosol after passing the orifice was measured and compared to the size distribution of an ‘unstressed’ aerosol.

The experimental setup is shown in Fig. 1. Experiments consist of three steps: in the first step, the powder, as delivered from the manufacturer, was aerosolized. This is done in a pressurized beaker by magnetically stirring the powder under a constant carrier gas flow. In this study, particle free, dry nitrogen with a relative humidity well below 10% was used as carrier gas to eliminate possible unknown enhancement of the binding energy caused by water vapour.
Fig. 1

Schematic measurement setup. After generation within a pressurized beaker, the aerosol containing nanoparticle agglomerates passes an (critical) orifice under different overpressure conditions. The resulting number size distribution of the aerosol is measured after a compensation tank using an SMPS at ambient pressure

After generation of the aerosol, airborne agglomerates with an aerodynamic diameter above approximately 1.3 μm are removed by two consecutive cyclones (in Fig. 1 shown as one cyclone). After this pre-separation of bigger agglomerates, the aerosol passes the orifice with an aperture of 508 μm diameter and sharp edges (O’Keefe controls, type: K4-20) into a compensation tank with a volume of 10 L which is used to damp possible influences of concentration spikes. The whole setup (besides the beaker made of pressure resistant glass) is electrically grounded to minimize particle losses. Due to the shear forces within the orifice, a deagglomeration of the agglomerates into smaller fragments or even primary particles occurs. The magnitude of the shear forces is defined by the pressure gradient within the (critical) orifice, i.e. with increasing overpressure the flow Reynolds number and thus the turbulence inside the orifice is increasing.

The effect of shear forces was studied by applying overpressures starting from 10 kPa and then 20 kPa up to 140 kPa in 20 kPa steps. Above an overpressure of about 100 kPa, the orifice becomes critical. The pressure downstream of the orifice was maintained at ambient level. In order to maintain a constant flow rate of 2 L min−1 (at operating pressure) through the cyclones, a second orifice is used to release excess air (the size of this orifice is depending on the applied overpressure).

In order to investigate the effect of the different shear forces, a measurement of the ‘as prepared’ aerosol is also conducted for reference. In this case, the aerosol enters the compensation tank with a flow rate of 2.0 L min−1 without passing the orifice. Only the pre-separation process in the cyclones takes place. The resulting size distribution of these measurements will be referred to as reference number size distribution or reference measurement within this paper. Altogether measurements with nine different pressure levels including the reference measurement were conducted for each nanopowder under investigation.

The number size distribution of the resulting aerosol in the compensation tank was measured using a conventional Scanning Mobility Particle Sizer (SMPS) from TSI (Model 3080, long-DMA model 3081 in connection with a water based UCPC, model 3786) with a covered size range from 14.1 up to 736.5 nm electrical mobility diameter with a maximum resolution of 64 size channels per decade. The aerosol flow rate was set to 0.3 L min−1 with a sheath flow rate of 3.0 L min−1. The scan time to obtain one number size distribution was set to 240 s to minimize possible scan time effects (Russel et al. 1995). The next scan started after 20 s retrace and additional 40 s wait time.

In addition to the pre-separation of the cyclones, an impactor with an orifice size of 710 μm was used throughout the measurements as a pre-separator at the SMPS inlet. This impactor was carefully chosen, considering the largest electrical mobility diameter and the particle density. This was crucial, since the particle generation system produced fairly large amounts of particles larger than the upper size limit of the SMPS. If sampled by the SMPS, these would bias the results upon multiple charge correction.

The compensation tank was flushed with particle free nitrogen prior to particle generation, assuring a background particle concentration level below 1,000 particles cm−3. Test aerosol concentrations were at least between one and three orders of magnitude higher during actual measurements, which were conducted after the number size distribution within the compensation tank had reached stable conditions. The resulting number size distributions were then repeated five times for each pressure step to obtain a reliable data base and to calculate an average number size distribution.

The nanopowder agglomerates under investigation are summarized in Table 1. Five metal oxides and one carbonate with different properties were investigated. The size range of the primary particles is on the order of 10–70 nm. The shape of the primary particles as investigated by high-resolution scanning electron microscopy (SEM) and confirmed by TEM ranges from nearly perfect spheres (the different TiO2 modifications) over sphere-like particles (the different cerium dioxide (CeO2) modifications show edges on the TEM micrographs) to elongated particles with an estimated aspect ratio of about 2–5 in the case of strontium carbonate (SrCO3). Except for SrCO3 which was produced by a wet chemical process and dried afterwards, all powders were produced by pyrolysis.
Table 1

Overview of nanoparticle materials under investigation in this study



Primary particle size (nm)

Primary particle shape

BET surface area (m2/g)

Cerium dioxide



Spherical (partly irregular)


Cerium dioxide



Spherical (partly irregular)


Titanium dioxide





Titanium dioxide





Titanium zirconium aluminium oxide





Strontium carbonate (hydrophobized)





Overall five metal oxides and one carbonate were investigated during the measurements. The primary particle size and BET surface area were provided by the different powder manufacturers, the primary particle shape was analysed using TEM-images provided by the manufacturers

Data evaluation

For data evaluation, a resolution of 16 channels per decade (this corresponds to a size range from 16.5 nm up to 697.8 nm) was used throughout our study. In Fig. 2, an example of the average number size distributions as a function of the electrical mobility diameter of CeO2-1 is shown for the reference measurement and for four different overpressure values (for the sake of clarity, only the reference measurement and measurements at 20, 60, 100 and 140 kPa are displayed within this graph). Figure 2 shows that the reference size distribution has a mode diameter of about 500 nm, which shifts towards smaller diameters with increasing overpressure. Furthermore, the number concentration within each channel and thus the total concentration increase with increasing overpressure.
Fig. 2

Number size distribution of CeO2-1 for the reference measurement and four different overpressure levels

Due to differences in the total concentration for each substance and also for each pressure step (which results from the aerosol generation process and the passage through the orifice), the obtained number size distributions were further normalized. This allowed for direct comparison of the different pressure steps for one substance and comparable results for the different nanopowders.

The relative number size distribution was therefore calculated in a first evaluation step by dividing the concentration in each size class by the total number concentration of all size classes. The resulting relative size distributions are shown in Fig. 3. In this diagram, the shift of the modal diameter with increasing overpressure becomes more visible. Furthermore, the graph shows that the relative number concentration for the higher size classes decreases with increasing overpressure, whereas the relative number concentration in the lower size classes increases with increasing overpressure. This gives rise to the assumption that larger agglomerates break up in the orifice, thus producing smaller agglomerates, aggregates or primary particles. Of course, there are other processes like diffusion or impaction taking place within the orifice influencing the number size distribution of the particles, but a change towards smaller particles can only be observed either by a breakup of the agglomerates or by a removal of bigger particles due to impaction. Since the total particle concentration is increased (see Fig. 2), impaction of particles cannot have a major influence on the observed size changes.
Fig. 3

Relative number size distribution of CeO2-1 calculated from values in Fig. 2. In this diagram, the shift of the modal diameter towards smaller particles with increasing overpressure is better visible

In order to better display this process, the relative change of particle concentration within one size class was calculated in a second step for each given overpressure. The relative number concentration in each size channel of one pressure step was therefore divided by the corresponding relative number concentration of the reference size distribution. The results for CeO2-1 are shown in Fig. 4.
Fig. 4

Ratio of the relative number size distribution of CeO2-1 at four different overpressure levels with regard to the relative number size distribution of the reference measurement. An increase in the relative particle concentration for mobility diameters below about 300 nm takes place

Figure 4 shows the ratio of the four different overpressure measurements with regard to the reference measurement. A ratio above one indicates that the relative number concentration within the according size class is increased and a value below one indicates that it is decreased. It becomes apparent that the fraction of particles with an electrical mobility diameter below approximately 300 nm is increases with increasing overpressure, whereas the fraction of particles above approximately 300 nm is decreases with increasing overpressure. This shows that agglomerates are fragmented in the orifice and that this effect is pressure-dependent. For exposure assessment following an accidental release in a pressurized nanopowder production or transport vessel, it may therefore not be sufficient to evaluate only particle sizes as known from the size distribution inside the vessel as these may change when the particles pass a leak.

One complication during the experiments outlined here are possible particle losses due to impaction or diffusion in the sampling lines which were ignored during our study. These particle losses might occur within the orifice (Chen et al. 2007), within the compensation tank or the measurement unit. Also, resuspension most likely in the vicinity of the critical orifice by the turbulent flow might occur (Reeks et al. 1988). Since the whole setup was electrically grounded, it can be assumed that electrophoretic losses are negligible. Possible particle losses after the processes taking place within the orifice, however, only affect number size distributions shown in Fig. 2. Since it can be assumed that the occurring particle losses are due to the aerodynamic behaviour of the particles, the percentage of particles lost is only a function of the mobility diameter of the particles and not of the total number of particles within one size range. Therefore, due to the calculation of a relative particle concentration for the measured size range (which is the same for all pressure steps) and the subsequent calculation of the ratio of relative number concentrations for the different pressure steps with regard to the reference measurement, the size-dependent particle losses after the introduction of the aerosol into the compensation tank cancel each other out during the following calculations. That is, the information given in Figs. 4, 5, 6 and 8 is also independent of particles losses within the tubing and measurement equipment, but reflects directly the changes taking place due to the passage of the orifice.
Fig. 5

Comparison of the ratio for the 140 kPa overpressure measurements with respect to the reference measurements for six different nanopowders
Fig. 6

Mean values of the ratio of relative particle concentrations for the different overpressure measurements with regard to the reference measurements for a particle size of 100 nm (top) and 600 nm (bottom), respectively

Results and discussion

By applying the method described above, the change in the number size distribution of a variety of nanopowders was investigated. The results for an overpressure of 140 kPa are summarized for six different powders in Fig. 5. In the case of the lower size channels, the error bars get larger and there are some data points missing (see for instance the data of TiO2-1 or SrCO3). This is due to the low number concentration within these size channels for the reference measurement, which leads to a higher uncertainty for these size classes. If no data for the reference measurement are available at all, the data points for the ratio of the overpressure measurement are subsequently also unavailable.

As one can see in Fig. 5, the different substances show a generally similar behaviour, except for SrCO3. All other substances show a clear decrease in the relative number concentration for a size above roughly 300 nm electrical mobility diameter. This is caused by deagglomeration of the agglomerates within this size range. Also, all other substances show a pronounced increase of the relative number concentration below about 300 nm due to the fact that smaller fragments are introduced by the deagglomeration process. For SrCO3, the number concentration is increased only in a size range of approximately 100–300 nm, but it steeply decreases below 100 nm. In the size range above approximately 300 nm, the relative concentration is close to unity. It is assumed that the bonding energy between primary particles is higher (as compared to spherical primary particles) for these agglomerates, due to their elongated morphology, allowing for larger contact area. Therefore, agglomerates do not break up as easily, causing the concentration of larger particles to remain almost constant. Still it is surprising that the concentration of small particles decreases so drastically. This may be caused by different losses during the measurement of the reference size distribution and those during overpressure (particle losses within the orifice). Furthermore, it is noteworthy that concentrations of SrCO3 in the sub-100 nm size range were generally quite low leading to an increased uncertainty of the measured number size distributions.

In any case, Fig. 5 indicates that the deagglomeration behaviour in the orifice seems to be material-dependent. The two different CeO2 modifications show the highest degree of deagglomeration with a factor of about 7–8 for the relative increase of particle number in a size range of about 50–80 nm. The number concentrations for the modifications of TiO2 showed in the same size range only a 5- to 6-fold increase. Titanium zirconium aluminium oxide (TiZrAlO) shows only an increase by a factor of about three. For the case of SrCO3, the maximal increase by a factor of two in the relative number concentration is reached for a size of 130 nm electrical mobility.

In order to illustrate the relative particle concentration change with pressure for the six substances, the mean ratios of the concentration change for two different sizes are shown as a function of overpressure in Fig. 6. The mean value at 100 nm is calculated as the geometric mean value of the four size classes around 100 nm (80.6, 93.1, 107.5 and 124.1 nm), and the mean value at 600 nm is calculated as the geometric mean value of the three size classes around 600 nm (523.3, 604.3 and 697.8 nm). As can be expected from Fig. 5, the same trends regarding the behaviour of the different substances are also visible in Fig. 6. This is indicated by the fitting of the data points. It should be noted that linear fitting was used only as a first approximation to describe the deagglomeration behaviour as a function of the pressure difference. With more data points available in the future, the relationship may become non-linear. The horizontal black lines correspond to a ratio of one, thus indicating no change in the relative particle concentration. The magnitude of the increase of the relative number concentration for the smaller fragments (top) and the deagglomeration of bigger agglomerates (bottom) are illustrated in this figure. Values below one for the case of the smaller fragments, as well as values above one for the case of the bigger agglomerates are due to measurement uncertainties.

As described above, SrCO3 shows only a little increase of the concentration in the size range of 100 nm at higher overpressure values and almost no change in the 600 nm size range, indicating that there may not be any deagglomeration. For the case of the different TiO2 modifications, a more or less continuous increase (with some deviations for single overpressure steps) in relative particle number concentration occurs in the 100 nm size range. The highest increase in this size range was observed for the two different CeO2 modifications, indicating that these substances are more easily fragmented than the other substances examined here. This can also be seen at the data for the 600 nm size range, where CeO2 shows the highest degree of deagglomeration.

The differences in the deagglomeration behaviour of the substances evaluated within this study might at least be partly due to the shape of the primary particles. Therefore, as a starting point, we only discuss the effect of particle and agglomerate morphology on the deagglomeration behaviour and disregard possible influences of particle production and the resulting surface chemistry of the particles. Figure 7 shows transmission electron micrographs of three different substances (CeO2-1, TiO2-1 and SrCO3). In the case of CeO2-1, it is revealed that the primary particles have a sphere-like morphology with some irregularities. In principle, the same morphology is found for CeO2-2. TiO2-1 consists of agglomerates with primary particles having a spherical morphology, which is also true for the second modification of TiO2 and TiZrAlO. The primary particles of SrCO3 have, as revealed by the TEM-images, an elongated shape with an estimated aspect ratio of about 2–5. Also, the primary particles tend to form bigger aggregates as it is revealed by the TEM-images.
Fig. 7

Transmission electron micrographs for three different substances. The morphology of the primary particles consists of partly irregular particles (CeO2-1, left), more perfect spheres (TiO2-1, middle) to elongated particles (SrCO3, right). Please note the different scale bars

In the case of SrCO3, we consequently find only a low tendency of deagglomeration. An elongated shape of the primary particles and aggregates might lead to tightly bound agglomerates due to an enhanced contact area and, therefore, an enhanced binding energy. Once dispersed by the magnetic stirrer, these agglomerates remain intact during the higher mechanical stress occurring within the orifice. In the case of the different TiO2 modifications, the primary particles have a spherical shape and presumably only one contact point between each other. In this case, the binding energy is supposed to be lower which is also reflected by the higher deagglomeration as measured within this study. The highest degree of fragmentation into smaller particles was measured for the two CeO2 modifications. This might also be explained by the shape of the primary particles which is nearly spherical. But in contrast to TiO2 rough edges are also visible in the transmission electron micrographs. Since the van der Waals force depends strongly on the distance between the particles, an increase of the distance by some edges with a lower contact area might lead—even in the case of nanoparticles—to a higher fragmentation when shear forces are applied. This effect overcomes the generally increased binding energy of smaller particles by van der Waals forces. For the case of bigger particles in the micrometer size range, this effect is used by the introduction of nanoparticles to enhance the flow ability of such powders (Zimmermann et al. 2004). Furthermore, in the case of TiO2 particles, the TEM-images reveal a higher degree of sintering between the primary particles as compared to CeO2 particles thus leading to more stable agglomerates.

Figure 8 shows the fraction of particles below 100 nm (with respect to the total number concentration of the SMPS measurement) as a function of overpressure. The characteristics of the different substances are nearly the same as it is obvious by a comparison of Figs. 6 and 8. In the case of the reference measurement, Fig. 8 shows that all substances have a fraction around 1% of particles with an electrical mobility diameter below 100 nm. In the case of SrCO3, this fraction is mainly below this value for all overpressure conditions, indicating no significant fragmentation. For the different TiO2 modifications, the fraction of particles below 100 nm slightly increases to about 3% and in one case to about 5% of the overall number concentration at an overpressure of 140 kPa. CeO2 shows a higher fraction of particles below 100 nm even at moderate overpressure conditions and a value of up to 12% at 140 kPa overpressure (Table 2).
Fig. 8

Fraction of particles with a size below 100 nm (with regard to the total concentration)

Table 2

Comparison of the linear fit data based on the data within Fig. 8 of the increase of particles below 100 nm for the different nanopowder materials


Slope of linear fit (%/kPa)

Standard deviation of linear fit (%/kPa)

Correlation coefficient R





















SrCO3 (hydrophobized)




In order to establish a more reliable ranking of the different materials under investigation a linear fit of the data in Fig. 8 was made as a first approximation of the increase of particles below 100 nm as a function of increasing pressure difference. A higher slope indicates a stronger deagglomeration of the material. As can be expected from the results discussed above, the two CeO2 materials have the highest slope with modification 2 (slope: 0.0783%/kPa) showing about a doubled deagglomeration behaviour when compared with modification 1 (slope: 0.0326%/kPa). These values might be linked to the BET surface area where the modification 2 has a value which is about 50% of the value of modification 1, i.e. the tendency of deagglomeration seems to be inversely proportional to the BET surface area for these materials. Since the BET surface area of a given powder is a complex function of primary particle morphology and the morphology of the resulting aggregates and agglomerates (and cannot be calculated straightforward by particle diameter and density) it cannot be used to predict the dispersability of different powders. This is obvious for the case of TiO2-2 (slope: 0.0315%/kPa) with a BET surface in the same range as CeO2-2 but a slope in the same range as CeO2-1.

In the case of SrCO3 with a slope of 0.001%/kPa, there seems to be only a vague dependency of the deagglomeration on increasing pressure difference. This is also confirmed by the correlation coefficient R: in the case of SrCO3, there is nearly no linear relationship between the data points (R = 0.1531). For TiO2-1, the relationship is also rather bad with a value of R = 0.5648. The other substances show a good correlation coefficient with values between R = 0.8446 up to R = 0.9296, indicating that a linear fit might be used as a first approximation for the increase of particles below 100 nm with increasing pressure difference.

These results provide evidence for a connection between the shear forces occurring within a (critical) orifice and the number size distributions of the aerosolized nanopowders. Due to the influence of shear forces, agglomerates can get separated into smaller fragments, depending on their material. Therefore, the relative number of particles in the higher size classes is decreased, and the relative number of particles in the lower size classes increases by a factor of up to 10. Furthermore, the influence of the shear forces increases with increasing overpressure, as it can be expected due to the higher energy within the orifice.

From a microscopic point of view, an agglomerate of a size of some 100 nm consists of thousands of primary particles. In the simplest case of spherical (equally sized) particles, one particle can have up to six bonds with the surrounding particles. The binding energy resulting from van der Waals forces between the particles depends strongly on the number of bonds and thus the total binding energy strongly depends on the unknown local morphology of the agglomerate. A detachment of (smaller) fragments, i.e. primary particles, aggregates or smaller agglomerates, most likely takes place at the weakest link between two fragments. In the case of agglomerates with a diameter of some 100 nm, this might be between two more or less equally sized fragments. Small fragments are more likely detached from the outer parts of an agglomerate. Due to the vast number of primary particles and aggregates forming an agglomerate, the process of deagglomeration is stochastic. Therefore, a downsizing of the agglomerates for a given shear force is obvious, but no preferential size class for the fragmented parts can be expected. This behaviour is observed within these experiments: with increasing overpressure, the size distribution is continuously shifted towards smaller particles.

The differences in the deagglomeration between the substances presented here can presently not be traced back to intrinsic material parameters like the Hamaker constant of the materials, as this is presently not well-determined for most nanomaterials. The binding energy between primary particles of the agglomerates remains also unknown and cannot be calculated due to the complex shape of the agglomerates and the unknown number of primary particles. Also, other factors such as surface chemistry and local primary particle morphology (both depending on the formation process of the particles) influence the binding energy of a given nanopowder.


We have presented a method to compare the stability of airborne nanoparticle agglomerates. After the powder containing agglomerates and aggregates of primary particles within the nanometre range is dispersed under different overpressure conditions it passes an (critical) orifice. This method can be considered to simulate a leak in a pressurized nanoparticle production or transport line. During the passage through the orifice, the agglomerates are fragmented due to the shear forces present within the orifice. The magnitude of the shear forces and thus the degree of fragmentation depends on the overpressure at the inlet of the orifice. A clear trend to a higher degree of fragmentation is found with increasing overpressure. Furthermore, the degree of deagglomeration seems to be material-dependent. This method is sensitive to the measurement of small particles and is used for a differentiation between nanopowder materials. For exposure assessment, especially important following an accidental release from a pressurized vessel, these results show that it is not sufficient to know the particle size distribution inside the vessel but also to evaluate the resulting size distribution after passing the leak.

Further studies on the deagglomeration taking place during agglomerate flow through an orifice are currently under way. Within these studies, other materials consisting of primary particles within the nanometre size range are being tested to provide a more detailed ‘ranking’ of these materials with regard to their morphological features (i.e. primary particle size and shape, morphology of agglomerates). Also, these measurements will provide a more quantitative analysis of the degree of fragmentation as a function of more general quantities like the flow Reynolds number.


The financial support of this study by the German Ministry for Education and Research (BMBF) within the NanoCare project is gratefully acknowledged.

Copyright information

© Springer Science+Business Media B.V. 2009