This article presents the development of a Portable Aerosol Collector and Spectrometer (PACS), an instrument designed to measure particle number, surface area, and mass concentrations continuously and timeweighted mass concentration by composition from 10 nm to 10 μm. The PACS consists of a sixstage particle size selector, a valve system, a water condensation particle counter to detect number concentrations, and a photometer to detect mass concentrations. The stages of the selector include three impactor and two diffusion stages, which resolve particles by size and collect particles for later chemical analysis. Particle penetration by size was measured through each stage to determine actual collection performance and account for particle losses. The data inversion algorithm uses an adaptive gridsearch process with a constrained linear leastsquare solver to fit a trimodal (ultrafine, fine, and coarse), lognormal distribution to the input data (number and mass concentration exiting each stage). The measured 50% cutoff diameter of each stage was similar to the design. The pressure drop of each stage was sufficiently low to permit its operation with portable air pumps. Sensitivity studies were conducted to explore the influence of unknown particle density (range from 500 to 3,000 kg/m^{3}) and shape factor (range from 1.0 to 3.0) on algorithm output. Assuming standard density spheres, the aerosol size distributions fit well with a
Adverse health effects from the inhalation of particles are a complicated function of particle size, shape, composition, and exposure metric (e.g., number, surface area, and mass concentration) (
Commercial instruments provide a way to continuously assess aerosol concentrations of a given metric by size. Handheld photometers and condensation particle counters (CPCs) are used to continuously measure mass concentrations and number concentrations, respectively (
Samplers that collect particles for subsequent chemical analyses enable assessment of particle mass concentration by composition. Size and timeintegrated samplers, such as the 37mm filter cassette and inhalable IOM sampler, are widely used to measure personal exposures in the workplace, but yield only gross information on the size of the collected particles (
Aerosols can be mathematically described by multimodal lognormal (MMLN) distributions. Whitby used a trimodal distribution consisting of a nuclei mode (0.005–0.1 μm), an accumulation mode (0.1–2 μm), and a coarse mode (>2 μm) to describe measured size distributions of ambient aerosols (
Mathematical algorithms have been developed to fit sizeresolved aerosol data. Twomey compared two algorithms that estimated the parameters of a bimodal aerosol number distribution from aerosol measurements using diffusion batteries (
Size distributions of one concentration metric can be converted to those of other metrics. For example,
Our goal was to develop a single instrument, the Portable Aerosol Collector and Spectrometer (PACS), to continuously measure aerosol size distributions by number, surface area, and mass over a wide size range (from 10 nm to 10 μm) and to collect particles with impactor and diffusion stages for postsampling chemical analyses. Moreover, we aimed to accomplish this goal using two commercial handheld instruments. First, we describe the design and testing of the PACS hardware. Then, we describe a MMLN fitting algorithm that leverages the multimetric, lowresolution data from one sequence of PACS measurements to estimate aerosol size distributions of number, surface area, and mass concentration from 10 nm to 10 μm in near realtime. We refined the algorithm to obtain accurate and precise size distributions for four aerosols (clean background, urban and freeway, coal power plant, and marine surface). We also conducted a sensitivity study to assess the influence of unknown particle density and shape factor on the algorithm output. In a companion manuscript, we conduct laboratory tests by comparing information on size and composition obtained with the PACS to reference instruments (SMPS, APS, and NanoMOUDI).
The PACS consists of four main parts (
The aluminum size selector consists of six stages in a series: a bypass stage, three impactor stages, and two diffusion stages (
The two diffusion stages consist of circular, nylon meshes (41
In theoretical calculations, we assumed standard temperature (20 °C) and pressure (101.3 kPa), standard particle density (1,000 kg/m^{3}), and a hydrodynamic factor of 0.0942. The theoretical penetration curves for impactor stages were calculated with the designed
The valve system consists of six independent, custom pinch valves and a controller (
Each pinch valve includes a motor (Pololu 50:1 micro metal geared motor HP, Pololu Corporation, Las Vegas, NV, USA) connected to the pinch assembly. The direction of the current flow to the motor determines whether the assembly pinches the flexible tube open or closed. The amount of current delivered to the motor controls the magnitude of force applied to pinch the tubing. A custom circuit board designed using Multisim Version 13 (National Instruments Corporation, Austin, TX, USA) uses a microcontroller (Nano, Arduino, Ivrea, Italy) to process serial communications and appropriately signal the six motors through a motor driver (Pololu Dual HBridge Motor Driver, DRV8833, Texas Instruments, Dallas, TX, USA). The board also supports the power regulation for all of these components.
Two handheld instruments were selected for use as detectors in the PACS: a photometer (SidePak AM510, TSI Inc., Shoreview, MN, USA) and WCPC (Box Magic, Aerosol Dynamics Inc., Berkeley, CA, USA) (
The laminarflow WCPC, developed by
We measured the penetration by size and pressure drop of each stage of the separator using the experimental setup shown in
We generated three aerosol types to span the size range of interest. Fresh metal fume was produced with a spark discharge system, providing an ultrafine mode aerosol (
We measured particle penetrations by size through each stage for each aerosol type six times (
The penetration for each size bin of the SMPS and APS was calculated as the number concentration exiting the outlet divided by that entering the inlet. We calculated the sharpness
We calculated the
We measured the pressure drop of each stage three times (
We measured the response time to achieve a stable number concentration after a valve switch (
A custom software program was developed using Visual Basic (VB.Net Version, Microsoft Corporation, Redmond, WA, USA) to control the timing of valves and to acquire data from the photometer and WCPC. The user defines the delay after a valve is opened and the duration over which concentrations are averaged (15 s in the current work). The program sequentially opens one valve at a time, collecting and storing six mean number concentrations and six mean mass concentrations for a single scan through all stages (3 min in the current work). It then calls a MMLN fitting algorithm to translate these 12 measurements into aerosol size distributions of number, surface area, and mass concentration. The program then displays these data graphically and numerically by particle size mode to the user.
A flowchart of the fitting algorithm developed to determine the continuous aerosol size distributions of number
In Step 1, we estimate
For simplicity, we rewrite
For subsequent stages,
We also set the partial derivative of the squared difference between observed
For subsequent stages,
We applied the CLLS solver,
We then calculate bias of number and mass concentration in each PACS stage
In Step 2, we refine the estimates of
Then, we applied the HatchChoate equation to convert
Lastly, for each metric
We conducted tests to determine the step size and range for the gridsearch of
For each aerosol, we used the nine parameters (one
We evaluated the influence of the gridsearch step size on the accuracy and precision of the fit for the four aerosols. For
According to the above testing results, we selected the step size with the most accurate and precise fit. However, the computation time would dramatically increase due to the increase of gridsearch times of
We then evaluated the refined algorithm for the four typical atmospheric aerosols by comparing the fitting results to the observed ones as follows: (1) we compared the aerosol size distributions in three metrics, (2) we compared the nine parameters given by
We performed a sensitivity analysis to test the robustness of the algorithm in the presence of uncertainties arising due to unknown particle density and shape factor. The sensitivity study was conducted by changing the particle density from 500 to 3,000 kg/m^{3} with a step of 100 kg/m^{3}, and the shape factor from 1 to 3 with a step of 0.1. Therefore, 546 combinations (26 densities ×21 shape factors) of density and shape factor were selected to cover a wide range of aerosol types found in different environments. For example, the density of diesel fume ranges from 500 to 1,200 kg/m^{3} (
Particle penetrations by aerodynamic particle size for bypass and impactor stages, and geometric particle size for diffusion stages are shown in
For particles progressively smaller than ~37 nm, the penetration of the 10μm impactor (Stage 1,
As expected, the measured characteristic
A knowledge of actual aerosol penetration by size in each stage is important to reduce uncertainties in estimating aerosol size distributions from PACS data.
The particle losses we observed for the impactor stages are similar to those observed for other impactors. According to theory, particle losses occur due to gravitational settling, impaction, interception and diffusion (
Penetration by geometric particle size for the diffusion stages are shown in
The effective collection efficiency by equivalent mobility particle size of particles to the two diffusion stages combined (stages 4 and 5) is shown in
The measured cumulative pressure drop for each PACS stage is listed in
The number and mass concentrations of the combined test aerosol measured by the detectors after passing through each stage are shown in
There are three factors for determining the response time: (1) opening and closing valves; (2) response of the pumps to recover from pressure drop released by the stage; and (3) the clearing of the volume of air between the exit of the stage and the detector. Valve opening and closing is fairly rapid (~3s), so unlikely to be the largest contributor to overall response time. The time for the pumps to regain airflow is dependent on the pressure drop added/released by stage. Based on the airflow and the air volume between the exit of the stage and the WCPC, the estimated time for clearing the volume of air between the exit of the stage and the detector is ~3 s for the Bypass Stage, ~3.1 s for the Stage 1, ~3.3 s for the Stage 2, ~3.4 s for the Stage 3, ~3.4 s for the Stage 4, and ~3.5 s for the Stage 5.
These time delays and the associated averaging time define the minimum time required to obtain a full set of measurements with the PACS. If the averaging time is 15 s, then the minimum time required to obtain a set of measurements over all stages was 144 s.
In general, decreasing the step size of
We selected
We refined the algorithm using an adaptive process to decrease the computation time while still using the smallest step size for
The results of fitting aerosol size distributions using the refined algorithm are shown in
The results of the sensitivity study are depicted in
Independent of particle density and shape factor, fitted number and mass concentrations were within ±10% of known concentrations, a typical acceptance criterion used by EPA and NIOSH (see number and mass concentration plots in
In a single portable instrument, the PACS provides a way to continuously measure aerosol size distributions of number, surface area, and mass concentration over a wide size range while simultaneously collecting particles with impactor and diffusion stages for chemical analysis. The ELPI, an instrument that retails for ~$120,000, is the only other single instrument with similar capabilities. However, the low pressure impactor stages used to achieve separation of sub300 nm particles of the ELPI are expensive to manufacture and require a large, heavy vacuum pump, which dramatically reduces the portability of the system. The reliance on diffusion stages to separate these sized particles in the PACS dramatically reduces the cost of size separation and eliminates the need for high vacuum pumps, thereby promoting portability. We envision that the size selector in future versions of the PACS can be made by injection molding of conductive plastic. ++Further reducing weight and cost. Whereas the ELPI relies on highly sensitive electrometers to measure the concentration of particles, the detectors employed in the PACS (a photometer and a WCPC) are substantially less inexpensive and have been shown robust in field use. Moreover, we envision that these detectors could be combined in a commercial PACS version, further reducing costs associated with redundant user interfaces and pumps.
Similar information can also be obtained with multiple researchgrade instruments, such as the combination of an SMPS, APS, and nanoMOUDI (~$150,000 in total). Researchers combine the SMPS and APS to measure aerosol size distributions over a wide range (
Nevertheless, the PACS also has some limitations that constrain its intended use to measure continuous aerosol size distributions. As shown in
In this article, we described the development of hardware and software of a Portable Aerosol Collector and Spectrometer, the PACS. The PACS continuously measures aerosol size distributions by number, surface area and mass concentrations over a wide size range (from 10 nm to 10 μm), and collect particles with impactor and diffusion stages for postsampling chemical analyses. The penetration by size in all six stages were measured experimentally to have characteristic
We then developed an MMLN fitting algorithm to rapidly (<3 min) estimate aerosol size distributions in three metrics (number, surface area and mass concentration) with high resolution over a wide size range (from 10 nm to 10 μm) from number and mass concentrations measured with relatively inexpensive handheld detectors in the PACS. Fitted size distributions were in close agreement with observed distributions for all three metrics (number, surface area and mass concentrations) for aerosols found in highly diverse environments. The sensitivity studies indicated that the particle density and shape factor were of great importance to the fitting accuracy of the algorithm. These parameters can be estimated from physical and chemical analysis of particles collected with the size separator of the PACS. With the data analysis methods introduced, the PACS can provide novel exposure assessments, including aerosol size distributions of number, surface area and mass concentrations in a wide size range (from 10 nm to 10 μm). It also provides a way to collect particles to determine time and massweighted size distributions.
The authors greatly appreciate the technical support of Drs. Beau Farmer of TSI who provided the photometer and Susanne V. Hering of Aerosol Dynamics who provided the WCPC.
This work was supported by the smallbusiness innovative research (SBIR) project (AF131024) from the U.S. Air Force and a pilot project from the Heartland Center for Occupational Health & Safety (NIOSH T42OH008491). Some of the equipment used to conduct the study was borrowed from the Exposure Assessment and Modelling Facility of the University of Iowa Environmental Health Sciences Research Center (NIEHS/NIH P30 ES005605).
Color versions of one or more of the figures in the article can be found online at
Schematic diagram of the PACS with major components identified.
Photographs of the PACS, showing the assembled instrument (center) and each component around the perimeter.
Experimental setup used to measure particle penetration by size and pressure drop.
Fractional penetration measured for the six PACS stages (error bars represent the standard deviation of six measurements; dashed line indicates the measured
Number concentrations from the WCPC (a) and mass concentrations from the photometer (b) for the combined aerosols of fresh metal fume, aged metal fume and ARD (error bars represent the standard deviation of three measurements).
Effect of the gridsearch step size on the fitting results expressed as: (a)
Particle size distributions estimated with the PACS fitting algorithm for four atmospheric aerosols: (a) clean continental background; (b) urban and freeway; (c) coal power plant; (d) marine surface. The dotted lines (black) represent the predefined aerosol, and the solid lines (red) represent the distribution fit with the algorithm.
Physical characteristics, flow parameters, and experimental results for the impactor stages.
Stage  Physical characteristics  Flow parameters  Design  Experimental results  



 σ  Δ  
 
Bypass Stage 0  n.a.  n.a.  n.a.  n.a.  n.a.  n.a.  n.a.  n.a.  n.a.  0  15 ± 4 
Impactor Stage 1  3.15  4.47  2.78  313  150  10^{a}  ~10^{a}  0.22  2.6  0  7 ± 1 
Impactor Stage 2  0.67  2.70  6.81  1470  3309  1.0^{a}  1.0^{a}  0.22  1.6  0.65  10 ± 1 
Impactor Stage 3  0.48  0.48  5.67  2051  6447  0.3^{a}  0.4^{a}  0.22  1.5  2.88  8 ± 2 
Diffusion Stage 4  n.a.  n.a.  n.a.  25  n.a.  0.016^{b}  0.016^{b}  n.a.  n.a.  2.88  8 ± 1 
Diffusion Stage 5  n.a.  n.a.  n.a.  25  n.a.  0.110^{b}  0.056^{b}  n.a.  n.a.  2.88  6 ± 0 
aerodynamic diameter
geometric diameter; n.a.: not applicable.
Summary of fitting results for aerosols found in diverse environments, assuming standard density spheres.
Aerosol type 


 

 SA 

 SA 

 SA 
 
 
1. Clean background  0.0  3.5  0.0  11.9  6.2  3.3  0.98  1.00  1.00 
2. Urban and freeway  0.0  −3.2  0.0  5.3  7.9  9.4  1.00  0.99  0.99 
3. Coal power plant  0.0  −4.9  0.0  27.6  24.0  23.2  0.90  0.94  0.90 
4. Marine surface  0.0  2.5  −2.7  17.5  6.2  5.9  0.97  0.99  1.00 