The motivation for the work reported here was derived from a recent test evaluation program for trace detection of explosives, both as vapor and as particles. A problem arose in terms of generating a test aerosol with a target size distribution while maintaining an upper limit on overall mass concentration of an analyte material. The work reported here demonstrates a method for achieving independent experimental control over the total mass concentration and the particle size distribution.
Computational fluid dynamic modeling of the chamber carrier internal gas flow and experimental measurements of particles in the chamber showed that at typical average gas velocities, any droplets or particles of aerodynamic diameter between 0.1 µm and approximately 3 µm will essentially follow the internal airflow and be carried out of the chamber [7
]. For gravitational settling and inertial impaction to play any significant role in bringing analyte particles into contact with detector or other chamber substrates, target particle diameters will need to exceed ≈3 µm. One option to improve deposition efficiencies is to deliberately engineering formation of analyte particles with sizes greater than 3 µm. However, since mass scales with the particle volume (approximately size cubed), the requirement to maintain low analyte concentrations is sharply at odds with that approach since with homogeneous composition analyte particles, maintaining a constant higher mass fraction would require particle number concentrations to decrease below levels feasible to either generate or verify.
A method to resolve the conflict between particle size and analyte mass concentration involves increasing the particle size by adding an inert material with the analyte. This approach decouples the quantity (mass) of analyte delivered per particle from the total aerodynamic size of the particle, and permits a continuously adjustable range of analyte mass delivered by each particle, while the particle size would be independently controlled by the quantity of the inert core material to result in the desired particle size range. This paper describes a specific implementation of this approach in which we chose to generate droplets of analyte solution that contain an inert core composed of solid spheres of known size. Once the solvent evaporates, the residual particle is a solid sphere, or cluster of such spheres, with a layer of analyte on its surface. Appreciable mass of analyte materials is used for this proof-of-principle study in order to create composite particles with quantifiable parameters. This technique can be applied to create an approximately constant composite particle size (total mass) for varying concentrations of the analytes as in explosives applications where the mass of the analytes are very low (nm thick layers). This composite particle generation method can be used to create particles of two indices of refraction for elastic scatter studies or to simulate bacterial spores or other heterogeneous aerosol particles. This technique can also be applied to particles where the composite particle size can be padded with inert material of varying densities to create composite particles of similar size and analyte concentration but varying densities that will exhibit dissimilar flow characteristics. The following sections provide details of a proof-of-principle demonstration of this concept.
2. Experimental approach
To demonstrate the capability of generating micron-sized aerosol particles with a mixed composition of chosen analyte and inert core, we used two different fluorescent materials with emission in distinguishable spectral wavelength regions, although this technique is not limited to fluorescent materials. The inert material component was formed of spherical PMMA beads doped with 1,4-bis(5-phenyl-2-oxazolyl)benzene (POPOP) dye, referred to as PMMA-P [7
], while the material representing the analyte was soluble, powdered ovalbumin (grade III) [8
]. The separation of the fluorescent emission between POPOP (dominant fluorescence band centered at 410 nm) and ovalbumin (dominant fluorescence band centered at 330 nm) provides easy discrimination between the two materials. For this study an alternate simpler setup was used instead of the chamber described above. The experimental setup was comprised of an aerosol generator, a drying column with forced air and a detector stage. Two types of aerosol generators were used to create droplets/sprays of solutions and suspensions of the materials of interest. The first aerosol generation device was a micro-droplet dispensing generator from MicroFab [9
], which is capable of producing single droplets on demand, at rates up to 3500 drops/sec. Sizes of the droplets emerging from the micro-droplet generator are monitored using a video camera and image processing software [9
] provided as an integral part of the equipment. The second method was a Sono-TekTM
] ultrasonic spray generator, which creates droplets from a high-frequency vibrating wetted tip. This type of generator is capable of generating aerosols at high concentrations since many droplets are created simultaneously during one vibration cycle. For both droplet generation methods the initial droplet diameters were typically between 60 to 80 μm. A particle is formed from the dissolved and/or suspended material as water evaporates from each droplet and the residue dries. By controlling the concentration of the dissolved analyte material in solution, the choice of the inert PMMA-P bead size and their concentration, the size of the resulting aerosol particles could be varied over a desired range from 1 to 7 μm. The output of the generators was coupled to a glass drying tube (90 cm long; ID of 5 cm) that was heated to 85°C. A continuous flow of 1 L/min of dry nitrogen was used as the carrier gas. The output of the drying tube is split into two flow branches and sampled by two detection instruments; the Aerosol Interrogation Module (AIM) [11
11. V. Sivaprakasam, T. Pletcher, J. E. Tucker, A. L. Huston, J. McGinn, D. Keller, and J. D. Eversole, “Classification and selective collection of individual aerosol particles using laser-induced fluorescence,” Appl. Opt. 48(4), B126–B136 (2009). [CrossRef] [PubMed]
] and the Aerodynamic Particles Sizer (APS) [12
], with each instrument receiving 0.3 L/min and the APS receiving make-up air to total the required flow rate of 1 L/min. The drying tube is open to the atmosphere through a HEPA-filter to ensure minimal pressure differential along the sampling column.
The particle materials were chosen to have fluorescent properties that could be optically analyzed and quantified based on their emission spectra and intensity using the previously developed AIM device. The AIM device interrogates and characterizes individual aerosol particles in the 0.5 to 10 μm size range using elastic scatter from a 785 nm CW laser, and laser-induced fluorescence measurements from two pulsed UV lasers at 266 nm and at 355 nm. The observation of 785 nm elastic scatter from individual particles acts to trigger sequential single pulses from each of the two UV excitation lasers, and also provides a relative indication of the particle size (optical cross-section). Fluorescence emission induced, by the laser pulses from each particle was collected and detected in four broad spectral bands from 300 to 600 nm using a photomultiplier tube for each band. The emission intensity measured in two UV bands centered at 324 nm (303 to 344 nm) and 390 nm (380 to 400 nm) due to the 266 nm excitation source were used for the analysis described here. Although recorded, emission in the longer wavelength bands from 266 nm and 355 nm laser excitations are not discussed here for simplicity. Additionally, as mentioned earlier an aerodynamic particle sizer (APS 3321) instrument [12
] concurrently measured the aerodynamic diameter of the particles delivered to the AIM device. Throughout this work, the APS measured aerodynamic diameter is used for particle sizing.
3. Size and fluorescence measurements of analyte and inert core materials
Ovalbumin, like most proteins is an intrinsically fluorescent molecule in the UV, and was chosen as a soluble compound to represent the analyte in this study. As a first set of experiments, the micro-droplet dispensing generator and drying tube arrangement was used with solutions of only ovalbumin at concentrations of 6, 36 and 170 μg/ml. The residual aerosol particles formed from these droplets were analyzed one at a time by both the AIM device and the APS. Fluorescence intensities measured in the 324 nm and 390 nm emission bands from the particles for the three concentrations are shown as scatter plot data in Fig. 1
Fig. 1 266 nm laser excited fluorescence in spectral bands centered at 324 and 390 nm from ovalbumin particles generated by the MicroFab micro-droplet generator. Each point in the plot corresponds to the fluorescence emission from a single particle. Starting solution concentrations of ovalbumin are presented: 6, 36 and 170 µg/ml. The generated droplets from these solutions dried into aerosol particles with particle size modes of 1, 2.1 and 3.5 µm respectively, and the size distributions measured by the APS instrument are shown in inset.
, where eachdata point represents a single particle. Additionally, aerodynamic particle size distributions (concentration vs. size) simultaneously measured by the APS for the three concentrations are shown in the inset. The fluorescence intensities are sufficiently compact and clustered that the three particle populations are clearly separated for the chosen concentrations. The fluorescence intensity in the 324 nm band is about a factor of 5 stronger compared to the 390 nm band. Likewise, the particle size distributions are distinct with little overlap as shown by the plots in the inset in Fig. 1
that display a single mode at 1, 2.1 and 3.5 μm for the 6, 36 and 170 mg/l ovalbumin concentrations respectively, each having a FWHM of less than 25% of their mode value. These particle sizes scale approximately as the cube root of the ovalbumin concentration, consistent with formation of a particle from a single droplet. Each particle population is labeled with its corresponding size distribution mode in the emission measurement plot.
For the inert core simulant, 2.1 μm diameter PMMA spherical beads doped with 0.3% POPOP dye (PMMA-P) were used since the fluorescence of this material is spectrally distinct from the analyte simulant, ovalbumin. Fluorescence and size measurements were also acquired for the PMMA-P particles. A series of four solutions with increasing concentrations were prepared, which yielded a mean value of: 2, 5, 10 or 20 PMMA-P beads in a spherical volume nominally 60 μm in diameter. Droplets of the samples were disseminated using the micro-droplet generator, as with ovalbumin previously. Single particle fluorescence measurements of the PMMA-P clusters are shown in Fig. 2
Fig. 2 266 nm laser excited fluorescence in spectral bands centered at 324 and 390 nm from the PMMA-P particle clusters. Each point in the plot corresponds to the fluorescence emission from a single particle. Starting solution concentrations correspond to averages of 2, 5, 10 and 20; 2.1 µm PMMA beads doped with 0.3% POPOP dye (PMMA-P) in a nominally 60 µm diameter droplet. Solid contour lines containing 50% of the particles for each sample is shown. The aerosol particle size distributions as measured by the APS instrument are shown in inset.
. As in Fig. 1
, each point in the scatter plot corresponds to the measurements made on a single aerosol particle, and the data points are color coded as shown. The PMMA-P beads exhibits strong fluorescence in the 390nm band, which is two orders of magnitude larger than the fluorescence signal measured in the 324 nm band. Unlike ovalbumin however, in this case the four populations are heavily overlapped, and solid contour lines containing 50% of the population have been added to demark boundaries for each concentration. Size distributions of the resultant aerosols measured using the APS for the four solution concentrations are also shown as an inset in Fig. 2
. For the two lowest concentrations (with an average of either 2 or 5 beads per droplet), the measured size distributions are bimodal, with the dominant mode falling between 2 and 3 μm, corresponding to droplets containing one or more PMMA-P beads. The smaller diameter maxima with a mode size under 2 μm were formed primarily of residues of surfactant that was present in the PMMA-P suspension. From this data, the fraction of the particle distributions equal to and smaller than 1.8 μm in diameter are 28%, 4%, 0.3% and 0.3% for the average concentrations of 2, 5, 10 and 20 beads per drop respectively. This shows that for the two highest concentrations a negligible percentage of droplets were formed with no PMMA-P beads. In Fig. 2
(inset), the primary mode of the size distribution increases with increasing bead concentration, and their corresponding widths are broad and overlapping. As will be discussed in the following section, the widths of these distributions can be predicted for a process in which the number of PMMA-P beads per droplet is governed by Poisson statistics. Due to the wide and overlapping particle size distributions of these PMMA-P particle clusters generated from droplets, the fluorescence distributions of the aerosols generated from the four concentrations are highly overlapped as well.
An analysis was performed comparing the APS measured particle diameter with the computed diameter based on the concentration of the analyte and the drop sizes formed by the droplet generator. As mentioned earlier, the MicroFab micro-droplet generator includes a video camera and image processing software to monitor the size of the generated droplet, which is measured for each starting solution and used to compute the expected diameter of the resultant dry particle. Although the densities of the various solution/suspension samples were within 10% of each other, different sample compositions resulted in subtle effects in terms of fluid viscosity and surface tension that make fine-tuning of the droplet generator drive voltage and pulse width necessary to achieve stable operation for each sample composition. Consequently, slight differences in mean droplet diameter occurred between the different sample compositions. As an example, the suspension of PMMA-P clusters resulted in 63 μm mean diameter droplets, while pure ovalbumin solutions yielded a mean diameter of 71 μm with a standard deviation of about 6%. The analysis is performed for particles generated from: ovalbumin solutions and 2.1 μm PMMA-P bead suspensions and the results are listed in Table 1
Table 1. Comparison between the computed and measured diameter of ovalbumin and PMMA-P particles generated with micro-droplet generator.
| | |
. The concentration of the analyte, or the number of 2.1 μm sized PMMA-P beads present in a single generated drop, is listed in column 2. The expected diameter of the resultant dry particle, computed from the concentration, the density of the material and the measured droplet size is tabulated in column 3. Column 4 contains the geometric median diameter computed from the APS measured aerodynamic data that has been Stokes-corrected [12
] for the instruments flow velocity with the assumption that the particles are spherical. The percent difference between expected and measured diameters in each case is listed in column 5 with a mean of around 22% for ovalbumin and 30% for PMMA-P. Considering that the computed diameter is based on the simple expectation that total volume is equal to the sum of the volume of the individual PMMA-P beads and not accounting for the voids in between the beads, the results are in good agreement with APS measured diameters. Our observations are also in agreement with prior studies that report a smaller than expected diameter for agglomerates of spheres and other non-spherical geometries in ultra-Stokesian flow regime, and deformed liquid drops in the accelerating flow of APS [13
13. P. A. Baron, “Calibration and use of the aerodynamic particle sizer (APS 3300),” Aerosol Sci. Technol. 5(1), 55–67 (1986). [CrossRef]
16. Y. S. Cheng, B. T. Chen, H. C. Yeha, I. A. Marshall, J. P. Mitchell, and W. D. Griffiths, “Behavior of compact nonspherical particles in the TSI aerodynamic particle sizer model APS33B: Ultra-Stokesian drag forces,” Aerosol Sci. Technol. 19(3), 255–267 (1993).
]. The magnitudes of these differences (column 5) are not unreasonable given the uncertainties in droplet size determination from imaging, and absolute concentration. However, the fact that all of the measured values are consistently lower than the computed estimates with a range of 12 to 39% suggests a systematic bias leading to larger particle estimates in the computed values.
In Fig. 3
Power law fit of the 266 nm laser excited fluorescence emission from the ovalbumin particles (displayed in Fig. 1
) and PMMA-P particles (displayed in Fig. 2
) as a function of the median particle size obtained from the APS data shown in the inset of Figs. 1
. For both samples, the 324 nm band emission is colored violet and the 390 nm band emission is colored light blue. In all cases fluorescence response dependence on particle size is approximately volumetric (I ∝ d3
). The approximate per particle fluorescence cross-section is also labelled on the right-axis.
the median fluorescence response and standard deviations measured for both 324 and 390 nm emission bands for the three concentrations of ovalbumin from Fig. 1
and for the four PMMA-P concentrations from Fig. 2
are shown plotted as a function of median particle size (median value used since some of the size distributions are bimodal). By performing a linear least squares fit to each series, a simple power law exponent can be obtained, and its value is labeled on each graph in Fig. 3
. Since ovalbumin particles are composed of homogenous material, both the fluorescence and aerodynamic size data showed narrow standard deviations in Fig. 1
reflected by the relatively small error bars in Fig. 3
. Power lawdependence values for the 324 and 390 nm ovalbumin emission bands were 3.3 and 3.0 respectively with respect to the particle diameter. This value is in good agreement with a volumetric dependence of fluorescence intensity, (I ∝ D3
) expected for homogeneous aerosol measurements and those observed in prior studies [17
17. V. Sivaprakasam, H. B. Lin, A. L. Huston, and J. D. Eversole, “Spectral characterization of biological aerosol particles using two-wavelength excited laser-induced fluorescence and elastic scattering measurements,” Opt. Express 19(7), 6191–6208 (2011). [CrossRef] [PubMed]
]. The standard deviations for the PMMA-P clusters’ fluorescence response were larger than that for ovalbumin samples as shown in Fig. 2
, yielding larger error bars in Fig. 3
. The resulting power law exponents were 2.9 and 2.6 for the 324 nm and 390 nm fluorescence bands, respectively, which indicates a volumetric dependence of the fluorescence intensity. The slightly lower values are likely due to the fact that a cluster of discrete particle components will provide an effectively lower density compared to a homogeneous particle of the same material with the same aerodynamic size, consistent with previous observations of inhomogeneous aerosols [17
17. V. Sivaprakasam, H. B. Lin, A. L. Huston, and J. D. Eversole, “Spectral characterization of biological aerosol particles using two-wavelength excited laser-induced fluorescence and elastic scattering measurements,” Opt. Express 19(7), 6191–6208 (2011). [CrossRef] [PubMed]
]. The approximate corresponding per particle fluorescence cross-sections in the respective fluorescence bands are also labelled on the right-axis of Fig. 3
4. Analysis of size and fluorescence data for inert core material
Further consideration of the broad size distributions obtained for the PMMA-P bead clusters as a function of suspension concentration, led to the development of a model for expected cluster size. Given the mean number of PMMA-P beads expected to be present in a suspension volume represented by each droplet, the number distribution of PMMA-P beads in each drop is governed by Poisson statistics. To evaluate the distribution of dried aerosol particles measured by APS, Poisson probability distributions were computed with expected mean number of beads present in each drop as 1.15, 1.8, 4 and 9, corresponding to the median aerodynamic diameter of the particles measured for each of the four concentrations of PMMA-P solutions, rather than the theoretically expected mean numbers of particles listed in Table 1
. These measured values were chosen as input for this computation since we are interested in modeling the spread of the measured size distributions, and using the theoretical median values will make the comparisons less accurate. The computed distributions are plotted in Fig. 4
Fig. 4 Size distributions of the four concentrations of PMMA-P beads (nominally 2, 5, 10 and 20 2.1 μm PMMA-P beads in a 60 µm diameter droplet) as measured by the APS instrument are plotted in solid lines. The corresponding computed Poisson probability distribution curves are plotted in dashed lines. All the traces are normalized to unity for visual comparison.
as dashed lines, omitting the probability of having zero PMMA-P particles since in principle aerosol particles resulting from this value would not be measured by APS(particle diameter = 0). For comparison, the aerodynamic particle diameter data for each of the four concentrations of PMMA-P beads, previously shown in Fig. 2
inset, is also plotted in Fig. 4
as solid lines. All the traces are normalized to 1 for visual comparison, and although there are some differences, the widths of the distributions are in good agreement between the data and computational results. Additionally, the calculated probability of having zero PMMA-P beads using the theoretically expected mean number of beads in each drop is 22%, 0.2%, 0%, 0% for the four concentrations, which agrees with the observed percentages ofparticles smaller than a single PMMA-P bead (composed of only surfactant) of: 28%, 4%, 0.3% and 0.3%. These results imply that the dominant process responsible for the wide range of particle sizes produced is due to Poisson fluctuations in the physical number of PMMA-P beads in each droplet even though the droplet size is held reasonably constant as discussed earlier. Variation in generated droplet size was also added to the computational model as a Gaussian distribution combined with the Poisson distribution of PMMA-P beads, and the results were found to have negligible change in the spread of the size distribution of the resultant dried particles compared to simply using a constant, mean value for the droplet size.
The fluorescence measured from the four concentrations of PMMA-P cluster aerosols, as discussed in Fig. 2
, shows significant overlap between their emission intensity distributions. One-dimensional histograms of the fluorescence intensity data for the four PMMA-P bead concentrations are plotted in Fig. 5
One-dimensional histograms of the fluorescence emission in spectral bands centered at 324 nm (A) and 390 nm (B) of PMMA-P clusters from Fig. 2
are plotted in solid lines. The computationally modeled fluorescence responses; based on particle size and fluorescence yield are plotted as dashed lines in (A) and (B). Modelling the AIM’ device’s responses using a lognormal distribution function with adjustable standard deviations are plotted as dotted lines in (A) and (B). All the traces are normalized to 1 for visual comparison.
for: (A) the 324 nm band, and (B) the 390 nm band. These values can also be estimated by using the independently measured size distributions for the four samples shown in Fig. 4
caused by Poisson fluctuations in the generation method, combined with the measured fluorescence dependence on particle size illustrated in Fig. 3
, and taking into account the smaller particle diameter measured by TSI compared to expected values discussed in Table 1
(average value: 30%). The predicted fluorescence distribution based on this computational model is also plotted in Figs. 5(A)
using dashed lines with the same color-code for the four samples. Visual comparison between the measured florescence response (solid lines) and the model predictions (dashed lines) shows good agreement, especially for the larger clusters. This demonstrates that the broad, overlapped fluorescence distributions of the four samples are primarily due to wide particle size distributions caused by statistical fluctuations in the number of beads per cluster. However, for the smaller particles the modeled curve is narrower than the measured fluorescence response indicating that the measurement may have an additional source of uncertainty for weaker signals. This uncertainty in the fluorescence measurement can also be modeled as a response of the AIM device using a lognormal distribution function with a varying standard deviation that is a function of intensity for each specific emission channel. The standard deviation offers a free parameter, then that can be used to optimize the fit of the model results to the experimental data. Taking this instrument response into account, the fluorescence model result for the four concentrations and the two fluorescence bands are plotted as dotted lines in Figs. 5(A)
, and one can see that the fit of the fluorescence response for low PMMA-P concentrations is improved.
5. Results and analysis for composite aerosols
Once the basic emission properties of ovalbumin and PMMA-P cluster particles were determined, it was possible to continue feasibility evaluation of an inert core material with a soluble analyte material to produce a composite aerosol. Composite aerosol particles were produced from mixtures of different ovalbumin solutions with fixed suspensions of PMMA-P beads. The concentration of PMMA-P beads used in this study corresponded to the nominal average of 10 beads in a 60 µm diameter droplet, and this base suspension is referred to here as C-10. The ovalbumin concentrations were selected as: 0, 6, 36 and 170 μg/ml; the same concentrations previously used to generate well separable pure ovalbumin aerosol particles (Fig. 1
). The selected PMMA-P concentration, C-10, is sufficiently high that droplets produced with no PMMA-P beads are a negligible fraction of the total population as previously discussed (see Fig. 2
). The C-10 suspension was sonicated to insure uniform dispersion in the liquid, and during dissemination the suspension reservoir was continuously agitated to prevent concentration gradients due to gravitational settling. The mean diameter of the generated droplets was measured to be 65 μm. After droplet drying, the resulting aerosol particles were measured using the AIM device and the APS instrument simultaneously as previously reported. The diameters measured by the APS are compared in Table 2
Table 2. Comparison between the computed and measured diameter of composite particles generated with micro-droplet generator.
| | |
, with the expected diameters computed based on their concentration and measured initial droplet size following a similar process to that in Table 1
. It is not possible to perform Stokes correction to the aerodynamic diameter or compute the geometric diameter from the measured diameter for particles composed of two materials due to the unknown composite density. However, the computed geometric diameter can be converted into Stokes corrected aerodynamic diameter based on the measured value of each component separately, to provide an approximate aerodynamic diameter for the composite. The resultant volumes of the components were added, and the resultant diameters of the composite particles are listed as aerodynamic diameters in column 3 and compared to measured aerodynamic values from APS in column 4. The percent difference between expected and measured diameters in each case is listed in column 5 with a mean of 33%, consistent with the deviations observed and discussed in Table 1
Figure 6 Fig. 6
(A). 266 nm laser excited fluorescence in spectral bands centered at 324 and 390 nm from 2.1 µm PMMA-P beads suspended in solutions of ovalbumin using the MicroFab micro-droplet generator are plotted in primary colors. The PMMA-P solution corresponds to an average concentration of 10 PMMA-P beads per 60 µm drop (C-10) and ovalbumin solution concentrations used were 0, 6, 36 and 170 µg/ml and are labeled a, b, c and d respectively. Particle size distributions are shown in inset. For comparison, the fluorescence of same concentrations of ovalbumin only are plotted in lighter shades of the corresponding colors and labeled e, f and g. SEM images of typical cluster particles are shown on the top. Figure 6(B)
shows the one-dimensional histogram of the particle fluorescence in the 324 nm emission band, with the composite particles’ response plotted in solid lines, ovalbumin in dashed lines and the convolution of the fluorescence response of PMMA-P (a) and ovalbumin (e, f and g) in dotted lines while Fig. 6(C)
shows the histogram of the particle fluorescence in the 390 nm emission band, with the composite particles’ response plotted in solid lines, ovalbumin in dashed lines and the convolution of the fluorescence response of PMMA-P (a) and ovalbumin (e, f and g) in dotted lines.
is a mosaic image that provides multiple perspectives of the experimental composite particle aerosol data. The central exposition of data is labeled (A), which is similar to Figs. 1
in that it presents single particle fluorescence intensities as scatter plots for the two emission spectral bands, and also shows the simultaneously measured aerodynamic diameter data as an inset. Data from Fig. 1
for pure ovalbumin aerosol particles from solution concentrations of 6, 36 and 170 μg/ml, has been reproduced in this plot, labeled as samples e, f and g respectively, to provide a comparison for composite particle data. Additionally, the C-10 concentration data set of PMMA-P beads from Fig. 2
has also been reproduced on this graph, labeled as sample a. Finally three data sets of composite particles, labeled: b, c and d, are shown. These combined results in Fig. 6(A)
show that our composite particles have emission in the longer wavelength band, centered at 390 nm, due primarily to dye-doped PMMA-P bead clusters, and therefore remains approximately constant for the four samples a, b, c and d since the PMMA-P concentrations are substantially the same among them. At the same time, emission intensity in the shorter wavelength band, centered at 324 nm, is determined primarily by the amount of ovalbumin that has been added to the PMMA-P cluster cores, and therefore increases as the concentration of ovalbumin increases among the samples. To a first-order approximation, fluorescence from composite aerosol particles can be regarded as the sum of the fluorescence intensity contribution in each band from each of the two constituents, ovalbumin and dye-doped PMMA-P, but for which, the contribution of each constituent is primarily associated with only one spectral band.
This simple and intuitive expectation of the aerosol particle formation process is further supported by SEM analysis of collected samples of the generated aerosol clusters, of which representative images are shown on the top right of the Fig. 6
for the samples c and d. Images of these samples were taken from substrate surfaces exposed to aerosol flow during the sample generation process and are typical of the image results. Both images portray particles that appear to be composed of an agglomeration of smaller, roughly 2 µm diameter spherical particles of nearly uniform size covered by a homogenous film that fills in the voids and spaces between these primary particles. Moreover, there appears to be more of the film material for the sample image labeled d compared to c, resulting in a more rounded and smoother overall texture. This visual assessment is consistent with our interpretation that the smaller, spherical constituents are, in fact, PMMA-P 2.1 µm diameter beads that have formed clusters from a single droplet, with residue from dissolved ovalbumin covering the bare cluster and filling-in the voids and spaces among the beads. The increased concentration of ovalbumin (by about a factor of six) from sample c to sample d is also consistent with the smoother and more rounded appearance of sample d.
This simple conceptual model that fluorescence from composite aerosol particles can be regarded as the sum of the fluorescence intensity contribution in each band of the two constituents, ovalbumin and dye-doped PMMA-P (i.e., sum of samples a and e yields b; a and f yields c; and a and g yields d in Fig. 6(A)
) can be examined more closely. Separate one-dimensional histograms of the two fluorescence channels are plotted in Fig. 6(B)
and Fig. 6(C)
for 324 nm and 390 nm spectral bands respectively. The histograms of the composite particles are plotted as solid lines, while the fluorescence of the corresponding ovalbumin concentrations are plotted as dashed lines in same colors as the scatter-plot data. The convolution of the fluorescence distribution of the components, PMMA-P and varying concentrations of ovalbumin is performed and plotted as dotted lines in Fig. 6(B)
and Fig. 6(C)
in respective colors. As the measured mean droplet sizes vary for the three materials, the convolution was performed on appropriately weighted fluorescence distributions, i.e., the volumes of the composite droplets and thus the fluorescence intensities are 10% larger on average compared to that of bare PMMA-P beads but 23% smaller than that of pure ovalbumin particles. The 324 nm band fluorescence from the composite particles correlates well with the convolved signals establishing that the fluorescence of the composite particle is roughly the sum of the fluorescence of the components for the two higher concentrations and to a lesser extent for the lowest concentration. However, looking at the histogram of the fluorescence at 390 nm in Fig. 6(C)
, the fluorescence distributions of the composite particles (b, c and d) shows an increase in mean values of about 30% compared to the convolved spectra. From this, one could infer that there is approximately 1.3 times more PMMA-P beads present in the composite particles compared to the clusters composed of PMMA-P beads only. This observation is also corroborated by similar trend in the size distribution measured from the APS unit. The size distributions of the composite particles shown in inset of Fig. 6(A)
are roughly 30% higher than the convolution of the size distribution of PMMA-P (inset Fig. 2
) and the three concentrations of ovalbumin (inset Fig. 2
) would predict. A possible explanation is that the presence of ovalbumin in the solution creates an increased probability for PMMA-P bead aggregation prior to droplet formation, which, in turn, could skew the number of PMMA-P beads present in the dried composite particle even though the mass concentration in the solution remains unchanged.
6. Results and analysis for alternate droplet generation method
Similar experiments were conducted using particles generated from the Sono-tek® ultrasonic spray generator to evaluate the potential of generating composite particles with a larger production volume. As before, clusters of PMMA-P beads were generated as aerosol particles from liquid droplets with concentrations that resulted in an average 2, 5, 10 and 20 beads per 60 μm droplets and the measured size distribution has similar characteristics to the particle size distribution from micro-droplet generator shown in Fig. 2
. Following a similar analysis as before; particles equal to and less than 1.8 μm aerodynamic diameters are indicative of droplets that contained no PMMA-P beads and consequently the dried aerosol particles were primarily composed of surfactant, and account for 31%, 19%, 15% and 9% respectively for the four PMMA-P bead concentrations used. Even at the highest PMMA-P bead concentrations there are substantial numbers of particles that do not contain any PMMA-P beads. The higher percentage of particles that do not have any PMMA-P beads generated by the ultrasonic spray generator droplets compared to the particles generated by the micro-droplet generator is primarily due to the difference in the dissemination technique. The ultrasonic spray generator is an atomizer; the liquid stream is broken up by mechanical vibration, generating smaller satellite droplets as well as primary droplets, whereas the micro-droplet generator employs a piezo-electric crystal to apply pressure on the glass nozzle and results in the formation of droplets of more uniform size, and a much lower probability of satellite droplets.
The C-10 PMMA-P solution was disseminated with the ultrasonic generator, along with solutions of 0, 20, 126 and 360 μg/ml ovalbumin concentrations. These ovalbumin concentrations were chosen empirically to produce the same size distributions of ovalbuminparticles as were obtained with the micro-droplet generator in Fig. 1
. The measured fluorescence intensities for composite aerosols with the different concentrations of ovalbumin are shown in Fig. 7
Fig. 7 266 nm laser excited fluorescence in spectral bands centered at 324 and 390 nm from 2.1 µm PMMA-P beads suspended in solutions of ovalbumin using the Sono-tek ultrasonic generator. The PMMA-P solution corresponds to an average concentration of 10 PMMA-P beads per 60 µm drop (C-10) and ovalbumin solution concentrations used were 0, 20, 126 and 360 µg/ml.
with the APS measured size distribution shown as an inset. Unlike the fluorescence measured for similar particles that were generated with micro-droplet generator discussed in Fig. 6(A)
, the fluorescence for these composite particles show two distinct distributions. One mode is from particles composed of the combination of PMMA-P and ovalbumin and the second is from particles only containing ovalbumin. The fluorescence of the particles composed of both materials show an approximate constant intensity distribution in the 390 nm spectral band primarily due to the presence of PMMA-P beads that provide the dominant contribution in that band. Additionally, in the shorter wavelength 324 nm band the emission is dominated by fluorescence proportional to the ovalbumin mass of the sample. The fraction of particles that are composed of only ovalbumin is 15% for all of the three composite particles which is consistent with the 15% of particles containing only surfactant for the C-10 sample. Thus, the Sono-tek generator can produce composite particles, but onlywith a noticeable percentage of pure analyte particles also. This generation of two populations was not observed with the particles generated with the micro-droplet generator as the probability of generating particles with only surfactant was much lower due to the narrower size distribution of particles generated at the concentrations of interest for this study.
A quantitative analysis was performed comparing the measured diameter with the computed diameter based on the concentration of the analyte and the drop sizes formed. For the ultrasonic generator nozzle, the actual droplet sizes could not be directly measured, and an estimated droplet size of about 80 μm, is used based on the manufacturer’s specifications. The analysis was performed for particles generated from: pure ovalbumin solutions, 2.1 μm PMMA-P bead suspensions and combinations of both materials, and the results are listed in Table 3
Table 3. Comparison between the computed and measured diameter of various types and sizes of particles generated with Sono-tek generator.
| | |
. The expected diameter of the resultant dry particle computed from the concentration and the density of the material used or the aerodynamic diameter in case of the composite particle is tabulated in column 3. Column 4 contains the APS measured mean aerodynamic diameter for composite particles and Stokes-corrected [12
] and density-corrected geometric mean diameter calculated based on the assumption that they are spherical particles for ovalbumin and PMMA-P aerosol clusters. The percent difference between expected and measured diameters in each case is listed in column 5. The measured diameter is smaller than the expected diameter in all cases by about 50%. This variation obtained for the ultrasonic generated particles are larger compared to that of the micro-droplet generator, and could be simply due to the assumed droplet size.
We have demonstrated an ability to generate aerosol particles composed of two types of materials: a cluster of inert beads surrounded by, or covered with, an analyte material to form a composite particle. As an example, we chose commercially available dye-doped PMMA-P beads as a core material, and selected ovalbumin to represent a molecularly homogenous analyte material. Liquid drops containing either or both materials were disseminated using two types of droplet generators. The composite aerosol particles were demonstrated to be a combination of both materials based on single particle fluorescence measurements and particle sizing. This technique offers a solution to situations where aerosol particles are needed in which particle size and quantity of analyte mass can be independently controlled. For low concentrations of the analyte, as in the explosives application the analyte concentration could be varied essentially independently of the overall composite particle size. Although not attempted in this study, generating composite aerosols with only a single-bead as an inert core with a thin analyte coating would also be possible using this approach of allowing the solvent in droplets to evaporate to produce aerosol particles. To achieve this condition (of single bead cores), there would also necessarily be a relatively high percentage population of aerosol particles composed only of pure analyte. However, if the percentage mass of the analyte was small compared to the inert core, one could use a virtual impaction filter to separate out the smaller, analyte-only aerosol population. For the work reported here, the probability of generating analyte-only particles was reduced to negligible levels by forming inert cores as a cluster of multiple beads. Moreover, a droplet-on-demand, or constant frequency, type of generator was shown to produce very uniform (monodisperse) droplet sizes compared to an ultrasonic generator design that produce multiple droplets simultaneously with log-normal droplet size distributions. However, even with monodisperse droplets, the resulting dried aerosol particle size distributions will inevitably be broadened due to small-number statistical probabilities that govern the number of beads that reside in each cluster.