Rapid Size-Resolved Aerosol Hygroscopic Growth Measurements: Differential Aerosol Sizing and
3.5 Instrument Characterization .1 Size Limits of OPC Detection .1 Size Limits of OPC Detection
The range of particle sizes in which characterization experiments are carried out depends on the OPC size detection limits. The optical properties of salts vary, so the size detection limit of the OPCs is a function of the composition of an aerosol sample. By sampling a set of 16 different species (10 inorganic, 6 organic), the minimum particle diameters detected with 50% efficiency in the OPCs were found to be 135 + 8 nm and 234 + 16 nm for high- and low-gain settings, respectively. The variation in OPC response as a function of refractive index is contained within the standard deviations, which reflect the variability among the ensemble of salts tested. The maximum size that can be detected accurately by the OPC is approximately 450 + 30 nm with the high-gain setting, and beyond 1 μm with the low-gain setting. However, uncertainties associated with multiply charged particles are enhanced when detecting particles of increasing size, especially near 1 μm. Also, electrical saturation of the OPCs results in enhanced uncertainties starting at Dp > 450 nm (high gain) and Dp > 700 nm (low gain).
3.5.2 Time Resolution
The time resolution of the instrument is determined by three factors: 1) the transport delays between the point where particles enter the DASH-SP or, for a steady aerosol, where they enter the DMA, and the OPC detectors; 2) delays associated with programmed changes in the sampled particle size; and 3) the time required to obtain
statistically significant counts in each OPC channel. A maximum transition time of 17 s is needed when switching between different DMA-selected sizes for normal operating conditions in order to ensure particles of the previous size have exited the system; this addresses the first two factors governing time resolution. The tubing length of each stream from the 5-way splitter (see Figure 3.1) to each humidification chamber and OPC is equivalent, which results in the same particle transit time in each channel. Tests were performed to estimate the minimum number of particles at a given size required to obtain a response with varying degrees of statistical significance. These tests indicate the length of time the instrument should sample at a given set of operating conditions to obtain representative growth factors. Sample sets of over 10,000 pulses were obtained for a selected particle size for different salts (Dp = 150 – 404 nm). Pulses were randomly removed from each data file in a way to determine the variation of the modal pulse height and standard deviation as a function of decreasing counts at a fixed particle size. The ratio of standard deviation to modal pulse height is usually within 10% for pulse numbers ranging from 50 to 10,000. The variation of the modal pulse height is + 1% when comparing pulse counts of over 10,000 to any number of counts down to 50, below which the difference increases relatively quickly. In order for the standard deviation of observed pulse heights to be within 10%, 5%, and 1% of the original standard deviation calculated for a population of over 10,000 particles, an average of 60, 220, and 530 pulses are required, respectively. Figure 3.2 displays the time required to satisfy these different limits of the original modal pulse height and standard deviation for a wide range of particle number concentrations at a selected dry diameter. For example, for 10 particles cm-3 at a DMA-selected size, a minimum of 5 s is needed to be within 1% of the
modal pulse height had over 10,000 particles of the same composition been sampled. An additional order of magnitude of time is required to be within 1% of the original standard deviation as compared to 1% of the original mode. As will be shown subsequently, the mode is the critical parameter in quantifying growth factors via the iterative data processing technique. Assuming a pulse height distribution with a distinct and dominant mode (there are often less dominant modes due to multiply-charged particles), the data processing algorithm will determine the correct pulse height distribution mode regardless of the variance in the distribution. Thus, sampling time at a given Dp/RH combination should be reduced to improve time resolution at the expense of variance, which ultimately has no effect on the accuracy of quantified growth factors.
3.5.3 DASH-SP Accuracy, Precision, and Uncertainties
The uncertainty of the DASH-SP measurement of growth factor depends on the uncertainties in the RH and sizing (DMA and OPC) measurements. Critical to the accuracy of the RH measurements is the calibration procedure. One “reference” RH sensor (Model HMP50YCC1A2X, Vaisala) was initially calibrated over six saturated salt solutions within the 8 – 97% RH range. The remaining 12 RH sensors were then calibrated as an ensemble against the “reference” sensor in a temperature-controlled chamber through which a controlled mixture of dry and humid air was circulated. The accuracy of the RH measurements was evaluated by comparing theoretical deliquescence RH (DRH) for multiple salts with measurements. The DRHs of ammonium sulfate, sodium sulfate, and potassium chloride aerosols agreed with literature values (Seinfeld and Pandis 2006) to within + 1.0%, while sodium chloride aerosols agreed to within +
1.6%. The overall RH measurement uncertainty was calculated by taking the summed root mean square errors of the measured OPC sample and sheath flow RHs, and combining this with the uncertainty associated with the RH sensors. The overall uncertainty was usually + 1.5%, while the RH sensor precision was + 0.7%.
Temperature fluctuations and gradients within each of the four humidification chamber-OPC units can lead to uncertainties due to the RH changing. Temperature sensors monitoring the sheath, sample, and exit flows of each OPC exhibited no significant change (∆T < 1° C) during laboratory scans that lasted as long as ~ 10 min, but temperatures did show a systematic increase (∆T > 1° C) over longer periods of time.
During aircraft flights lasting 5 h in which the DASH-SP was deployed, the twelve temperature sensors recorded the same level of variation. For example, the twelve sensors during one representative flight exhibited averages between 24.0 – 25.0° C and standard deviations between 3.0 – 3.5° C. However, the RH sensors exhibited significantly less variation due to the feedback control of the humidification by the instrument software:
the RH standard deviation ranged between 0.1 – 1.8% for OPCs with upstream humidification chambers equilibrated at RHs of 74%, 85%, and 92%. Therefore, the RH variation resulting from temperature fluctuations in the humidification chamber-OPC units is normally within the uncertainty of the RH sensor measurements.
The instrument software calculates the DMA voltage required to classify a selected dry size for the specific geometry and operating parameters (T, P, flow rates) of the DMA. Since the DMA sheath flow rate is usually 10 times larger than the aerosol flow rate and relatively large particles are studied, the width of the sampled electrical mobility window is + 10%. DMA performance was frequently evaluated with calibrated
PSL particles ranging in diameter between 152 - 1101 nm. Accurate DMA operation is exhibited when the maximum particle concentration is observed at the DMA-selected size that matches the actual size of the PSL particles, rather than a smaller or larger size.
The maximum particle concentration was observed to occur at DMA-selected sizes within + 3% of the actual size of the PSL particles tested. It should be noted that the PSL particles studied have a reported size uncertainty ranging from + 1 - 3%. The average uncertainty associated with growth factor measurements is calculated to be + 4.3%, with an uncertainty as high as 8% at a dry size of 150 nm at lower RHs (< 80%).
Tests were done to check for contamination both in the form of leaks and in the source of Milli-Q water used in atomizing solutions. When feeding filtered air into the DASH-SP frequently over a period longer than 30 days, no particles were detected by the CPC and fewer than 1 cm-3 were detected by the OPC (high gain). Particles detected by the OPC corresponded to sizes below the 50% detection limit of the OPC (Dp < 135 nm).
Pure Milli-Q water was atomized into the DASH-SP system; no particles were detected by the CPC beyond 135 nm, and < 1 cm-3 were detected by the OPC (high gain). Below 135 nm, background particle concentrations measured by the CPC reached a value as high as 30 cm-3 at 5 nm; the OPC response at these lower sizes remained at < 1 cm-3. No background particle signals were observed using the low-gain OPC configuration for these tests. Since the experiments conducted with salts usually involved OPC number concentrations between 100 - 1000 cm-3 at a specific size, any background particles from the water source and inside the system itself are deemed negligible. Moreover, since no particles larger than 135 nm were observed by the OPC for these tests, contamination is
negligible for ambient sampling, as the system is designed to sample at minimum dry Dp
exceeding 135 nm (150 nm for this study).
3.5.4 Stability
The stability in dry and wet particle data was studied. For dry data, stability was quantified over a time span of four months by calculating pulse height standard deviation relative to pulse height average (σ/x) for both gain settings and different particle sizes (Figure 3.3). Above 200 nm, at the high-gain OPC setting, σ/x is < 10%, with the species of greatest atmospheric relevance (ammonium sulfate, ammonium nitrate, sodium chloride) being < 3%. For the low-gain setting, σ/x is < 10% at 350 nm and < 5%
beyond 350 nm for all species studied. For wet studies, the stability in growth factor at different initial dry sizes and RHs between 30 – 92% was quantified for several species, where σ/x was usually < 3% for all species studied. In summary, the stability of the DASH-SP improves for both dry and wet particles with increasing size above the 50%
detection limit of the OPC.