Airborne engineered nanomaterials such as single-walled carbon nanotubes (SWCNTs), multi-walled carbon nanotubes (MWCNTs), functionalized MWCNT, graphene, fullerene, silver and gold nanorods were characterized using a tandem system of a differential mobility analyzer and an aerosol particle mass analyzer to obtain their airborne transport properties and understand their relationship to morphological characteristics. These nanomaterials were aerosolized using different generation methods such as electrospray, pneumatic atomization, and dry aerosolization techniques, and their airborne transport properties such as mobility and aerodynamic diameters, mass scaling exponent, dynamic shape factor, and effective density were obtained. Laboratory experiments were conducted to directly measure mobility diameter and mass of the airborne nanomaterials using tandem mobility-mass measurements. Mass scaling exponents, aerodynamic diameters, dynamic shape factors and effective densities of mobility-classified particles were obtained from particle mass and the mobility diameter. Microscopy analysis using Transmission Electron Microscopy (TEM) was performed to obtain morphological descriptors such as envelop diameter, open area, aspect ratio, and projected area diameter. The morphological information from the TEM was compared with measured aerodynamic and mobility diameters of the particles. The results showed that aerodynamic diameter is smaller than mobility diameter below 500 nm by a factor of 2 to 4 for all nanomaterials except silver and gold nanorods. Morphologies of MWCNTs generated by liquid-based method, such as pneumatic atomization, are more compact than those of dry dispersed MWCNTs, indicating that the morphology depends on particle generation method. TEM analysis showed that projected area diameter of MWCNTs appears to be in reasonable agreement with mobility diameter in the size range from 100 – 400 nm. Principal component analysis of the obtained airborne particle properties also showed that the mobility diameter-based effective density and aerodynamic diameter are eigenvectors and can be used to represent key transport properties of interest.

There is a growing concern over the potential health risks from exposure to airborne nanomaterials in industrial environments. Respiratory deposition of airborne nanomaterials following exposure to them during manufacturing and handling in the workplace is also of high concern (

The objective of this study was to measure transport properties of various airborne engineered nanomaterials such as SWCNTs, multi-walled carbon nanotubes (MWCNTs), functionalized MWCNTs, graphene, fullerene, and silver or gold nanorods, and also, to characterize their morphological descriptors to find a relationship among their transport properties and morphology-based descriptors. Laboratory experiments were designed and conducted to measure diffusion diameter and mass of airborne nanomaterial aerosols generated using different techniques by tandem mobility-mass approach.

Fundamental transport properties such as mobility and mass of aerosolized nanomaterials were directly measured using tandem mobility-mass approach (

Effective density – The effective density of a particle used in our study is defined as the particle mass divided by the particle volume based on mobility diameter (_{eff} is the effective density, m_{p} is the particle mass, and d_{mob} is electrical mobility diameter.

Aerodynamic diameter – This diameter is obtained from the relationship among mobility diameter, mass, and effective density (_{ae} is the aerodynamic diameter, ρ_{0} is the standard reference density of a particle (1.0 g cm^{−3}), and C(d_{ae}) and C (d_{mob}) are slip correction factors using d_{ae} and d_{mob}, respectively. The aerodynamic diameter was obtained by an iterative solution.

Mass scaling exponent – We use a term “mass scaling exponent” instead of a fractal dimension, because it is not known if the particles used in this study are pure fractals. However, the mass scaling exponent of a particle is analogous to the fractal dimension of pure fractals (_{f} is the mass scaling exponent.

Dynamic shape factor (DSF) – The dynamic shape factor is defined as the ratio of the actual resistance force of the nonspherical particle to the resistance force of a sphere having the same volume and velocity as the nonspherical particle (_{ve}_{ve})_{ve}_{ve}_{p}), which is known to be ~ 2.0 g cm^{−3} for MWCNTs used in this study (

Envelop diameter was defined as a diameter of a sphere with the same projected area as that of a smallest ellipse which inscribes the particle of interest.

Aspect ratio was defined as a ratio of major axis to minor axis of the smallest ellipse which inscribes the particle of interest. It is worth noting that the aspect ratio defined in this way is different from the definition for aspect ratio of a single fiber which is defined as the ratio of fiber length to fiber diameter. For agglomerated fibrous particles like MWCNTs, the aspect ratio using major and minor axes of the enveloping or encompassing ellipse is a measure of elongation of the overall shape of the agglomerate.

Projected area equivalent diameter was defined as the diameter of a sphere with the same project area as the particle of interest. The ImageJ software (ImageJ, U. S. National Institutes of Health, Bethesda, Maryland) was used to perform image analysis to determine project area of the particle.

Open area was defined as one minus a ratio of the projected area of the particle to the area of the ellipse which inscribes the particle of interest. This definition is only meaningful for agglomerates. This serves as a measure of porosity or openness of particle structure.

TEM-based projected area-scaling exponent was calculated from TEM images of mobility-classified particles, using the program ImageJ, and the box-counting function. The box-counting method of determining the TEM-based projected area-scaling exponent used a shifting grid algorithm rather than a fixed grid and therefore was more sensitive to the complexity of the surface area (

Single-walled carbon nanotubes (SWCNTs), multi-walled carbon nanotubes (MWCNTs), functionalized MWCNT, graphene, fullerene, and silver nanorods were studied. Silver nanorods (L ~ 6.1 µm, purity 99.9+%; stock #: 0475NW2), MWCNT (95+%, OD 10–20 nm, length 10–30 µm; stock #: 1205YJ), MWCNT (95+%, OD 60–100 nm, length 5–15 µm; stock #: 1234NMG), short MWCNT-OH (95+%, OD 50–80 nm, length 0.5–2.0 µm; stock #: 1253YJF) were purchased from Nanostructured & Amorphous Materials, Inc (Houston, TX), and fullerene (C60 , 99.5 wt%) and Graphene nanoplatelets (GNPs, Grade 3) from Cheaptubes Inc (Brattleboro, VT). The GNP has surface areas 600 – 750 m^{2}/g, 4–5 layers, an average thickness of 8 nm, and typical particle diameters of less than 2 microns (Cheaptubes). SWCNTs were purchased from the Carbon Nanotechnologies, Inc. (CNI, Houston, TX), which were produced by the HiPCO™ technique, employing CO in a continuous-flow gas phase as the carbon feedstock and Fe(CO)_{5} as the iron containing catalyst precursor. Mitsui MWCNTs was provided from Mitsui & Co. (Ibaraki, Japan). Gold nanorods (OD 25 nm, length 256 nm; part#: A12N-25-1400) were purchased from Nanopartz Inc (Loveland, CO).

Different generation methods were used to aerosolize particles from liquid suspensions and dry powders. The generation methods fall into two categories: dry powder dispersion and liquid-based generation methods. Each nanomaterial was aerosolized using one or all of the following techniques, depending on the amount of material available and the suitability of each technique to generate aerosol with desired characteristics:

Vortex shaking (VS)

Dry dispersion of MWCNTs powders such as MWCNT (OD 10–20 nm), denoted as MWCNT1 in this study, MWCNT (OD 60–100 nm), which is called MWCNT2, and Mitsui MWCNT (OD 50–60 nm), named as MWCNT3, were performed using a vortex shaking method (

Electrospray (ES):

Suspensions of short MWCNT-OH, graphene, fullerene, and silver nanorods were electrosprayed using an electrospray generator (

Pneumatic atomization (PA):

Suspension of the same MWCNT (OD 10–20 nm) used for dry dispersion, i.e., vortex shaking method, was prepared and pneumatically atomized using a constant output atomizer (model 3073, TSI Inc). These generation techniques together capture a wide range of particle morphology; nebulization resulting in more compact morphologies (due to particle restructuring during droplet evaporation) and the direct dry aerosolization leading to more open structures. The morphology of airborne nanomaterial particles aerosolized during most workplace activities or processes are expected to be somewhere in between the two limits represented by these two techniques (i.e., VS and PA).

An experimental approach for determining diffusion and aerodynamic diameter, effective density, and dynamic shape factor of airborne nanomaterial particles is described in this section.

The experimental setup used in this study is shown in

Size distributions of aerosols generated by different methods were measured by a scanning mobility particle sizer (SMPS, Model 3936, TSI Inc, St Paul, MN). Typical aerosol flow rate to the SMPS was 1.0 lpm and sheath flow rates were in the range of 5.0 to 10 lpm depending on desired particle size range and resolution. The DMA-classified aerosol was also collected on a TEM grid using an impactor-based electrostatic precipitator (

DMA-classified agglomerates were introduced into an aerosol particle mass analyzer (

Effective densities, dynamic shape factors, diameter-mass scaling exponents, and aerodynamic diameters of mobility-classified particles were calculated from particle mass and mobility diameter measured using tandem mobility-mass approach, as shown in

Typical size distributions of airborne nanomaterials generated by each of three generation methods (VS, PA, and ES) are shown in

A vortex shaker (Vortex-Genie 2, Scientific Industries, Inc., Bohemia, NY) was operated at variable speeds of 600–3200 rpm, executing a large (r = 6 mm) orbit for aggressive vortex shaking. We typically operated the vortex shaker at 70–80 % maximum rotation speed. With these operating parameters, the vortex shaking method has provided well-dispersed aerosol particles for different materials, including carbon nanofibers (

Curves for idealized fractal and fiber particles are also shown for comparison. Fiber was assumed to be straight and randomly oriented during its transport. Also, it was assumed that the fractal particles have uniform primary particle diameter of 20 nm which was the same as the tube diameter of the MWCNT1 particles studied in this work. The aerodynamic diameter _{ae}_{ae}_{ve}_{ve,}_{ae}_{ae}_{ve}^{−3} and 19.3 g cm^{−3} for silver and gold, respectively). _{g}) is a measure of distribution of mass around the center of mass within aggregate particle structure. For pure fractals, the radius of gyration, defined as the root mean square radius that quantifies the overall size of the aggregate, is related to mobility diameter and mass scaling factor (_{g}) is higher than mobility diameter for all three nanomaterials, which is consistent with similar tendency of large fractal-like agglomerates (_{f} =1.75 and 2.15, respectively) are also shown in _{f}), showing different values compared to fractal dimension of pure, self-similar fractal particles. MWCNT1 generated by vortex shaking has a mass scaling exponent of 2.17 while the same material generated by pneumatic atomization has the exponent of 2.49, indicating that the particles generated by liquid atomization have relatively compact structure. Graphene particles, which have thin planar structure, show the lowest mass scaling exponent of 2.09.

^{−3}, and from 0.80 to 0.19 g cm^{−3}, with increasing diameter from 100 to 500 nm, respectively. The effective density of MWCNT1 particles generated by atomization decreases from 0.71 to 0.27 g cm^{−3} as particle diameter increases from 100 nm to 500 nm, indicating relatively compact structure compared to those generated by vortex shaking. Silver nanorods and gold nanorods have an effective density in the range of 0.84 to 1.22 g cm^{−3} and 0.78 to 0.93 g cm^{−3}, respectively. Considering the bulk material density of silver and gold are 10.5 and 19.2 g cm^{−3} respectively, the effective density of these particles is an order of magnitude smaller than their bulk material densities. Also, short MWCNT-OH particles have higher effective density for larger particles, e.g., 300 nm, unlike other particles, because the particles tend to be highly agglomerated as particle size increases, indicating that those particles are compact and this result is consistent with TEM image as shown in ^{−3} and it is smaller than those of MWCNT1 and MWCNT2 at the same size by a factor of 2.4 and 4.3, respectively, due to two-dimensional disc-like structures of graphene particles.

_{proj}_{proj}_{ve}_{proj,}_{ae}_{ae}_{ve}_{proj} and d_{mob} in _{proj}-d_{mob} relationship expected of pure fractal particles.

_{f} (see _{f, 3} / D_{f, 2}=1.1 is shown for comparison. The graphene data fall on this curve.

Estimation of lengths of carbon nanotubes were obtained. _{mass} on y-axis) and on projected area of a particle from TEM images (L_{proj} on x-axis) for different nanomaterials. L_{mass} on y-axis, the total tube length, was obtained with the following equation, _{c}L_{mass} ρ_{p}_{p}_{c}_{t}_{proj} on x-axis, the total tube length, was calculated using the relation, _{proj}=L_{proj} d_{t}_{proj}

The relevance of morphology of aerosol studied in this work to actual workplace nanomaterial aerosol depends on various factors, including the difference in mechanism of aerosolization (with respect to the energy available for aerosolization and the agglomerate breakup), and alternation of particle morphology post-aerosolization via coagulation with background aerosols (that are not specific to the source of nanomaterials). As noted earlier, the VS method used in this study is likely incapable of breaking the agglomerates apart and perhaps represents most dry dispersion methods involving mechanical agitation. The degree of alteration of particle morphology via coagulation with background aerosol can be significant depending on the particle size distribution of background aerosol. Order-of-magnitude calculations using size distributions of typical ambient/outdoor aerosol size distribution and typical workplace aerosol in nanomaterial manufacturing (available from literature), show that alteration of morphology via coagulation will likely be insignificant in many cases. (See for example the SI for order-of-magnitude analysis). For most workplaces, where airborne nanomaterial is the dominant source of aerosol, the range of morphologies studied in this work are likely to be relevant.

It is expected that high aspect-ratio particle orientation in electric field plays an important role in determining the mobility diameter. To investigate the particle orientation effect, we compared measured mobility diameter and aspect ratio with theoretically calculated values based on Li et al. (2013) and

We performed PCA on the measured properties to identify minimum set of orthogonal particle properties that could be used to clarify or distinguish different nanomaterials. We applied PCA to the entire data set that included the following measured or deduced particles properties: mobility diameter, aerodynamic diameter, volume equivalent diameter, mass, effective density, dynamic shape factor, and friction coefficient.

_{ae}, d_{ve}, d_{mob}, ρ_{eff}, and DSF) in the eigenvector space of PC1 and PC2; d_{ae}, d_{ve}, and d_{mob} have heavy loadings along PC 1. On the other hand, ρ_{eff} and DSF have heavy loadings along PC2 and capture information largely independent of d_{ae}, d_{ve}, d_{mob}. We used aerodynamic diameter (d_{ae}) and effective density (based on d_{mob}; ρ_{eff} ) as surrogates for PC1 and PC2 and plotted the data of ρ_{eff} as a function of d_{ae} as shown in

The particles belonging to the third group of particles (b is close to zero) have low aspect ratio structure. These particles are characterized by low inertia and their effective density does not substantially change with increasing agglomerate size. An alternative representation using DSF and particle mass as orthogonal components shows similar grouping of nanomaterials (see

The complete dataset for particle properties measured in this study is included in the supplemental information (

Airborne engineered nanomaterials such as single-walled carbon nanotubes (SWCNTs), multi-walled carbon nanotubes (MWCNTs), functionalized MWCNT, graphene, fullerene, and silver nanorods were characterized using a tandem system of a differential mobility analyzer and an aerosol particle mass analyzer to obtain their airborne transport properties and compare those to morphological descriptors based on transmission electron microscopy (TEM) images. From the measurement of mobility diameter, particle mass, and TEM analysis, equivalent diameters such as aerodynamic diameter, envelope diameter and projected-area diameter, effective density, dynamic shape factor, mass scaling exponent and open area of the particles were obtained in the submicrometer size range. In addition, principal component analysis was used to show that aerodynamic diameter and effective density (based on mobility diameter) can be used as two orthogonal particle properties that capture key transport properties of interest for most particle deposition systems. The exponent b, which defines the correlation between ρ_{eff} and d_{ae} (ρ_{eff} ~ d_{ae}
^{b}) could be used to classify materials in three distinct groups with different characters. Aerodynamic diameter was found to be smaller by a factor of 2 to 4 than mobility diameter for all nanomaterials studied in this work except silver and gold nanorods below 500 nm, emphasizing the need to use mechanism-specific equivalent diameters when modeling particle deposition in respiratory system or in other engineering systems such as particulate filters. Comparison with fractal theory showed that the particles agglomerates of aerosolized nanomaterials are not pure fractals.

The authors would like to thank Mr. Greg Deye and Dr. Chen Wang at NIOSH for helpful discussions on this work, and Joe Fernback for TEM images. This work was funded by the National Institute for Occupational Safety and Health through the Nanotechnology Research Center (NTRC) program (Project CAN 927ZJLS).

The mention of any company or product does not constitute an endorsement by the Centers for Disease Control and Prevention. The findings and conclusions in this paper are those of the authors and do not necessarily represent the views of the National Institute for Occupational Safety and Health.

Experimental setup for particle generation and measurement of particle mobility diameter and mass.

(a) Aerodynamic diameter vs. mobility diameter for different nanomaterial aerosols (b) Two times radius of gyration vs. mobility diameter. In the legend of the figure (a) VS stands for vortex shaker, ES for electrospray, and PA for pneumatic atomizer. In _{f}) are included for comparison. The fitting line for measured data was obtained using a model: y=ax^{b}; a=2.2702 ± 1.9028, b=1.00519 ± 0.1334, R^{2}=0.8604.

(a) Effective density as a function of mobility diameter (b) dynamic shape factor as a function of mobility diameter for different nanomaterial aerosols. In the legend of the figure VS stands for vortex shaker, ES for electrospray, and PA for pneumatic atomizer.

Characteristic diameter vs. mobility diameter for aerosol particles of the same MWCNT material. (a) Projected area diameter vs mobility diameter for MWCNT-VS and MWCNT-PA. and (b) Aspect ratios for the aerosols shown in (a). The number in legend box is nominal tube diameter from the manufacturer. Also shown are curves for idealized fractal and fiber particles. Fractal was assumed to be transparent. Fiber was assumed to be straight and randomly oriented. _{ae}_{proj}_{ve}_{proj}_{ae}_{ae}, d_{ve}_{ve}, _{proj}_{ae} of the fiber particles were calculated from cylindrical geometry and Cox’s theory (_{pr}^{0.043}, and _{m}

(a) Loading plot showing relationship between variables in the space of the first two principal components, and (b) effective density (ρ_{eff}) vs. d_{ae}. In plot (a), we clearly see that d_{ae}, d_{ve}, and d_{mob} have heavy loadings for principal component 1, and that ρ_{eff} and DSF have heavy loadings for principal component 2. Plot (b) shows distinct grouping and mapping of data for each material. Data for CNF were obtained from

Nanomaterials and aerosolization methods used in this study

Material | Name | Aerosolization | Physical size |
---|---|---|---|

Single-walled carbon nanotubes (OD 1.4 nm) | SWCNT | Dry dispersion | |

Multi-walled carbon nanotubes (OD 10–20 nm) | MWCNT1 | Vortex shaking (VS), pneumatic atomization (PA) | Length: 10–30 µm |

Multi-walled carbon nanotubes (OD 60–100 nm) | MWCNT2 | Vortex shaking | Length: 5–15 µm |

Mitsui Multi-walled carbon nanotubes (OD 50–60) | MWCNT3 | Vortex shaking | Length: 2–3 µm |

Functionalized MWCNT (OD 50- 80 nm) | MWCNT-OH | Electrospraying (ES) | Length: 0.5–2.0 µm |

Silver nanorods | SN | Electrospraying | Length: < 6 µm |

Fullerene | C60 | Electrospraying | |

Graphene | Graphene nanoplatelets (GNPs) | Electrospraying | Average thickness: ~ 8 nm, particle diameter: < 2 µm |

Gold nanorods (OD 25 nm) | GN | Electrospraying | Length: 256 nm |

Mass scaling exponent, dynamic shape factor and open area for nanomaterials tested in this study

Material | Mass | Projected | Dynamic shape | Open area |
---|---|---|---|---|

SWCNT, VS | 2.57 | 2.34 – 3.26 | - | |

MWCNT1, VS | 2.17 | 1.43 – 1.79 | 2.16 – 3.22 | 0.82 – 0.91 |

MWCNT1, PA | 2.49 | 1.60 – 1.91 | 1.83 – 2.39 | 0.68 – 0.78 |

MWCNT2, VS | 2.46 | 1.71 – 2.68 | - | |

MWCNT-OH, ES | 3.09 | 1.64 – 1.74 | 1.57 – 2.03 | 0.52 – 0.68 |

Silver nanorods, ES | 2.84 | 1.33 – 1.68 | - | |

C60, ES | 2.12 | 2.03 – 2.73 | - | |

Graphene, ES | 2.09 | 1.86 – 1.92 | 1.48 – 4.02 | 0.33 – 0.43 |

Gold nanorods, ES | 2.96 | 1.60 – 1.80 | - |

Fitting curves for effective density vs. aerodynamic diameter for different nanomaterials

Model equation | ρ_{eff} = a*d_{ae}^{b} | ||||
---|---|---|---|---|---|

Nanomaterial | Parameter, a | Parameter, b | R-Squared | Observations | p-value |

MWCNT1, VS | 246176.7 | −1.48802 | 0.99274 | 4 | 0.004 |

MWCNT1 & MWCNT-OH, PA | 10537.1 | −0.6452 | 0.69701 | 6 | 0.048 |

CNF, VS | 47468.4 | −0.78436 | 0.82248 | 13 | 3.01E-06 |

SWCNT, VS | 58.7 | 0.08343 | 0.03284 | 4 | 0.867 |

Silver nanorods, ES | 326.9 | 0.2298 | 0.7786 | 4 | 0.190 |

Graphene, ES | 2.6754 | 1.5444 | 0.3449 | 4 | 0.162 |