Environ. Sci. Technol. 2006, 40, 5029-5035
Single-Particle Detection Efficiencies of Aerosol Time-of-Flight Mass Spectrometry during the North Atlantic Marine Boundary Layer Experiment MANUEL DALL’OSTO, ROY M. HARRISON,* AND DAVID C. S. BEDDOWS Division of Environmental Health and Risk Management, School of Geography, Earth and Environmental Sciences, The University of Birmingham, Edgbaston, Birmingham B15 2TT, United Kingdom EVELYN J. FRENEY, MATHEW R. HEAL, AND ROBERT J. DONOVAN School of Chemistry, University of Edinburgh, West Mains Road, Edinburgh, EH9 3JJ, United Kingdom
During the North Atlantic marine boundary layer experiment (NAMBLEX) sampling campaign at Mace Head, Ireland, both continental and maritime air masses were sampled. Aerosol was characterized both with a TSI 3800 time-of-flight mass spectrometer (ATOFMS) and a MOUDI microorifice impactor, and particle number counts were measured independently with an aerodynamic particle sizer. The data have been analyzed in order to elucidate factors determining the particle detection efficiencies of the ATOFMS. These are broken down according to the efficiency of the inlet system, the hit efficiency on particles which enter the sensing zone of the instrument and the sensitivity of the measured ion signal to the chemical species. A substantial matrix effect depending on the chemical composition of the aerosol sampled at the time was found, which is reflected in variations in the hit efficiency of particles entering the sensing zone of the instrument with the main desorption-ionization laser. This is in addition to the strong inverse power-law dependence of inlet transmission efficiency on particle diameter. The variation in hit efficiency with particle type is likely attributable to differences in the energetics of laser energy absorption, ablation, and ion formation. However, once variations in both inlet transmission and hit efficiencies are taken into account, no additional matrix dependence of ATOFMS response is required to obtain a linear relationship between the ion signal and the concentration of a particular chemical species. The observations show that a constant mass of material is ionized from each particle, irrespective of size. Consequently the integrated ion signal for a given chemical component and particle size class needs to be increased by a factor related to the cube of particle diameter in order to correlate with the airborne mass of that component.
* Corresponding author e-mail: R.
[email protected]. 10.1021/es050951i CCC: $33.50 Published on Web 07/07/2006
2006 American Chemical Society
Introduction Atmospheric aerosols are a very topical subject of research because of their recognized impact upon human health and global climate (1-3). Most aerosol analysis techniques require the collection of integrated samples which are then returned to the laboratory for analysis. While the use of cascade impactors can generate size-specific information, these methods do not provide any insights into the composition of individual particles nor of whether the aerosol is internally or externally mixed. Many of these off-line techniques also have a very poor time resolution (4). The advent of techniques of single particle analysis offers great insights into the source apportionment and atmospheric chemistry of aerosols. Instruments such as the TSI 3800 aerosol time-of-flight mass spectrometer (ATOFMS) offer the possibility of characterizing large numbers of individual particles in terms of both their aerodynamic diameter and major component chemical composition. The particle diameter is derived by measuring the time-of-flight between two low-powered lasers, while positive and negative ion mass spectra are generated by using a powerful laser for particle desorption and ionization and subsequent ion time-of-flight detection (5). There are, however, a number of factors affecting the potential to extract fully quantitative information on particle size distributions and chemical composition from single particle instruments such as the ATOFMS. These include the following: (1) the transmission efficiency of the nozzle and skimmers used to create the particle beam in the instrument’s inlet; (2) the efficiency of the desorption/ionization laser in creating mass spectra; (3) the sensitivity of the mass spectral ion signals to individual chemical species. The first factor is expected to be heavily dependent on particle size (6, 7). Allen et al. (6) showed that the particle detection efficiency of an ATOFMS obeyed an inverse power law on diameter in the range of 0.32-1.8 µm, decreasing by 2 orders of magnitude over this range. The second factor depends crucially on the ability of the laser desorption/ ionization (LDI) energy to couple to the particle material and cause ablation and ionization in the vaporization plume. Variation in this process is a commonly cited limitation to chemical quantification by LDI single particle mass spectrometry. Although at high laser fluences (>4 × 1010 W cm-2) quantitative data on elemental composition can be produced (8), at lower laser fluences (which yield a greater amount of molecular information) the LDI process is likely heavily influenced by particle morphology and matrix composition (9-13). A further problem with LDI methods is that the ion signals of identical particles vary considerably from shot to shot because of inhomogeneities within the laser beam (14, 15). To date only a few measurements on ATOFMS instrument sensitivity to particle chemical composition have been reported. Relative efficiencies of ion yield between species (factor 3 above) are often based on laboratory-generated particles that are monodisperse, spherical, and of identical chemical composition. In general, alkali metal ions, ammonium nitrate, and aromatic compounds appear to be ionized efficiently, while ammonium sulfate gives a much weaker response as demonstrated in both laboratory and field studies (15-18). Prather and co-workers have demonstrated the potential of the ATOFMS to track a range of aerosol-phase chemical concentrations, including nitrate and ammonium, as measured by co-located impactors or automated monitors (19-21). VOL. 40, NO. 16, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5029
However, none of the previous studies report systematic investigation of factor 2 above. This is the primary focus of the work reported here. We demonstrate a systematic variation in the efficiency of the LDI laser to yield mass spectra with particle composition, regardless of the sensitivity of the resulting mass spectral ion signal to a particular chemical species. In addition, we report here the first independent verification of the conclusion of Bhave et al. (21) that the ATOFMS instrument ablates/ionizes a constant mass of material from each particle in the size range investigated. Our analyses are based upon ATOFMS, APS, and MOUDI data collected as part of NAMBLEX (North Atlantic marine boundary layer experiment) carried out at Mace Head on the west coast of Ireland during August 2002 (22). Air masses arriving at Mace Head can yield information on marine sources, emissions from Europe, and long-range transport across the ocean. A description of the measured data appears elsewhere (23).
Experimental Section ATOFMS. The ATOFMS provides continuous, real-time detection and characterization of single particles from polydisperse samples, providing information on particle size and composition. Since aerosol time-of-flight mass spectrometry has been described in detail elsewhere (5, 24), only a short description of the commercial version (TSI, Model 3800) will be presented here. Briefly, air is introduced into a vacuum system region through a converging nozzle. At this point, two differentially pumped regions separated by skimmers form a narrow collimated particle beam, which travels through a sizing region where the aerodynamic diameter of individual particles is determined by detecting scattered light from two timing lasers positioned at a known distance apart. After being sized, the particles enter the mass spectrometer source region. Here, a pulse from a Nd:YAG laser (frequency quadrupled, λ ) 266 nm) is triggered at the appropriate times based on the transit time of the particle measured in the sizing regionsto desorb and ionize material from the sized particle. The mass-to-charge ratios of both positive and negative ions of single particles are then determined simultaneously in two time-of-flight reflectron mass spectrometers. Particles for which both size and mass spectral (positive and/or negative) data are collected are classified as “hit”. Particles which are sized but do not produce a mass spectrum are classified as “missed”. The total number of particles that are detected through the aerodynamic inlet is thus the sum of hit plus missed. The ATOFMS inlet was dismantled and the nozzle cleaned every 5 days. No compensation for potential ATOFMS busy time (21) was undertaken for two reasons. First, it is not realistic to attempt to derive a universal absolute scaling factor because such a factor will necessarily be instrumentdependent. Second, any allowance for busy time in deriving a scaling factor will cancel out in the rescaling of ATOFMS data to compare with the MOUDI data. The ATOFMS sampled air from a manifold attached to the 22 m sampling tower. The manifold inlet was switched alternately between 7 and 22 m, but there was no significant difference in particle population between the two heights (23). ATOFMS data were collected continuously between Aug. 1-21, 2002 (Julian days 213-233). Within the size range of 0.52-2.94 µm (corresponding to APS size cut points), 1 234 830 particles were recorded of which 177 960 (14.4%) were classified as hits. APS. During this study ATOFMS counts were scaled with an aerodynamic particle sizer (APS, Model 3320, TSI Inc. St. Paul, MN), which sizes particles on the same principle (timeof-flight of individual particles in an accelerating flow field) as the ATOFMS and at a similar nozzle air flow (1.0 L/min 5030
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 16, 2006
for APS, 0.9 L/min for ATOFMS). This avoids potential issues of bias between aerodynamic and optical diameters when using particle sizers that utilize optical scattering (25). The APS reports particle size distributions from 0.5 to 20 µm in 52 channels. Counting efficiency problems in the Model 3320 have been reported (26, 27), now resolved with the newer TSI 3321 APS (28). During this study corrected data from the model 3320 were used. Briefly, APS number distribution data were converted into a mass distribution and compared with a microorifice uniform deposit impactor (MOUDI) deployed over the same period (August 2002) in the same location, and a calibration curve was derived (by University College, Galway). The correction was 0.8 µm. The APS sampled air from the same manifold as the ATOFMS. No APS data were available for 11th or 21st August (Julian days 223 and 233). MOUDI. A type 110 microorifice uniform deposit impactor (MOUDI) (MSP Corp., Minneapolis, MN) was used to collect particles in 10 size fractions (29). Sampling flow was supplied by a Gast twin piston pump. The cutoff parameters of the MOUDI stages were adjusted according to the actual flow rate of 25.5 L/min employed, yielding the following 50% cut-points for particle aerodynamic diameter (µm): 19.5, 10.7, 6.7, 3.4, 2.0, 1.1, 0.61, 0.33, and 0.20. The fractionated particles at each MOUDI stage were collected on Teflon filters previously treated with silicone grease to avoid particle bounce. The MOUDI sampled from the top of the tower (23), and its inlet was positioned under a simple plastic shelter to protect it from rain while allowing free ventilation. The sizesegregated particle ensembles of each impaction substrate were analyzed by the same procedure described by Huang et al. (30) and Allen et al. (31). Ion chromatographic analysis was performed for sodium (Na+), potassium (K+), calcium (Ca2+), magnesium (Mg+), chloride (Cl-), methanesulfonate (CH3SO3-), sulfate (SO42-), and nitrate (NO3-). Ammonium (NH4+) was analyzed by using a flow injection fluorescence technique. Sampling time was 24 h from 10 a.m. each day between August 12 and 21, 2002 (JDs 224-233). Thus where ATOFMS compositional data are compared with MOUDI data in the rest of this paper, the word “day” signifies from 10 a.m on that day and not from midnight.
Results and Discussion The efficiency of the ATOFMS skimmer system in transmitting particles from the initial inlet to the sizing region where they pass through the two low-intensity laser beams was established by comparing data from the ATOFMS and the APS. The inverse particle transmission efficiency, E, of the ATOFMS was calculated as a function of particle size in the range of 0.523-2.943 µm via
E)
NAPS NATOFMS
(1)
where NAPS is the APS number concentration and NATOFMS is the ATOFMS number concentration (hits plus misses) in the given time interval and particle size bin. In deriving values of E, the criterion was applied that the ATOFMS should have counted at least 10 particles per hour per particle size bin. This required merging the APS data in the size range of 0.523-2.943 µm into 8 size ranges from an initial 24 bins, i.e., 0.523-0.649, 0.649-0.806, 0.806-1.00, 1.00-1.241, 1.241-1.54, 1.54-1.911, 1.911-2.207, and 2.2072.943 µm, and integrating the 5 min output into 1 h time intervals. Figure 1a shows the inverse transmission function, E, across the full size range analyzed. There is a minimum in
FIGURE 2. Hit efficiency of particles detected during NAMBLEX in the size range of 1.2-1.9 µm. Unscaled ATOFMS data represent hourly average temporal trends of pure and aged sea salt particles.
TABLE 1. Variation in ATOFMS Particle Hit Efficiency, H, for Different Days of the Campaign Derived Using the Complete Aerodynamic Diameter Range 0.52-2.94 µm and for the Subset 1.2-1.9 µma hit efficiency (%)
FIGURE 1. Inverse particle transmission efficiency (E) versus aerodynamic diameter (Da) for the TSI 3800 ATOFMS instrument sampling ambient aerosol at Mace Head in August 2002 during NAMBLEX. The size ranges are (a) 0.52-2.94 and (b) 0.52-1.91 µm. Uncertainty bars are (2 std dev of daily averages. the function around 1.6 µm, corresponding to a maximum in nozzle transmission and particle-sizing efficiency. Moffet et al. (7) recently reported a similar turning point. In Figure 1b the inverse transmission efficiency data within the more limited size range of 0.523-1.911 µm have been fitted by the exponential relationship to particle diameter Da:
E ) RDaβ
(2)
where R and β are fitted parameters. Our factor, E, although qualitatively similar to the factor φ determined by Allen et al. (6), differs in that Allen et al. (6) based the function upon the hit particles only, whereas our function describes the sum of hit and missed particles (i.e., all of those detected by the sizing lasers). The scaling parameter R is dependent on instrument features and is not universal. However, the exponent value of -4.9 ( 0.9 ((2 standard deviations (std dev)), although instrument-dependent, confirms that ATOFMS sensitivity to particle transmission and detection is a strong inverse function of particle diameter within this size range. Subsequently, the particle hit efficiency, H, was quantified, as follows:
H)
hit × 100 (hit + missed)
(3)
The particle hit efficiency was examined as a function of the particle chemical composition as previously determined by Dall’Osto et al. (23). Figure 2 shows a time series of hit efficiency over 5 days of the campaign (Aug. 13-17, 2002) together with the time series for aged and pure sea salt particles (23). There appears to be a clear similarity but not a close correlation between the time trends for particle hit efficiency for the size range of 1.2-1.9 µm and the counts of aged but not pure sea salt particles. By selecting periods
main chemical species present dust sea salt carbon a
Julian days
0.52 < Da (µm) < 2.94
1.2 < Da (µm) < 1.9
224, 225, 228 226, 227, 229 213-216
17 ( 3 5(1 31 ( 1
17 ( 3 5(1 36 ( 2
Values are means (95% CIs of hourly data.
from throughout the campaign, during which one type of particle predominated (23), hit efficiencies for different kinds of particles were defined and appear in Table 1. The table shows that hit efficiency did not differ significantly between particles in the size range of 1.2-1.9 µm or across the whole particle size range analyzed (0.5-2.9 µm). Clearly there is a very major compositional effect upon hit efficiency which does not appear to have been recognized in other studies which focused more on organic and ammonium nitrate aerosols (6, 18, 21), perhaps because such large differences in aerosol composition were not encountered in those studies. The low hit efficiencies on the pure sea salt days toward the end of the campaign were not due to a time-dependent deterioration in, for example, nozzle cleanliness, because the hit efficiency increased again after JD 229 when mixed rather than pure sea salt aerosol was again encountered. In Figure 3, ATOFMS scaling factors, φ, defined as in eq 4, and directly analogous to those of Allen et al. (6) are shown for three different days, JD 216, JD 225, and JD 229.
100 H
φ)E
(4)
As seen, the scaling factor is at its minimum for JD 216, which was a period dominated by carbon-containing particles. On the other hand, JD 225 was characterized by dust particles, whereas on JD 229 the particles were comprised mainly of pure sea salt which shows a much lower hit efficiency and therefore a larger scaling factor. The explanation appears to be that the alkali halide salts require a high laser fluence at 266 nm because of their relative transparency at this wavelength and high lattice energy (12, 32). Particles containing carbonaceous materials are likely to have both higher absorptivity and lower ablation energies. An additional VOL. 40, NO. 16, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5031
FIGURE 3. Scaling factors derived for Julian days 216, 225, and 229. explanation may relate to differences in particle morphology with particle type. Nonspherical particles may be less wellfocused through the ATOFMS inlet than spherical particles. Our data do not show a dependence of inlet transmission efficiency (1/E in our notation) on particle type beyond the range of transmission efficiencies measured experimentally and illustrated (with uncertainty bars) in Figure 1a, so any differences in morphology of ambient particles appear not to significantly affect their detection by the ATOFMS sizing lasers. However, not all sized particles necessarily pass through the ablation laser beam, so it is possible that if the extent to which a particle is defocused is related to its morphology, and if different types of particles consistently have different morphologies, then the observed compositional dependence of hit efficiency is also consistent with a morphology-dependent causation. Even if this is a contributing explanation, the overall effect remains the same: hit efficiency has a compositional dependence. Relative humidity might be expected to have some influence on the ionization process through the hygroscopicity of the particles. Inorganic aqueous aerosols typically have higher ionization thresholds and produce lower ion currents (33). Moffet et al. (7) recently found lower ATOFMS efficiency during episodes of high relative humidity (RH), causing condensation of water on particle surfaces. The relative humidity of the inlet was not controlled during our study. The RH of ambient air sampled during NAMBLEX was high and varied little, averaging 87% and never falling below 75%. Ambient temperature also varied little (15 ( 3 °C, 2 std dev) and was consistently well below the high temperature (30 ( 3 °C) of the container van housing the final part of the sampling line, the ATOFMS, and several other instruments. Air with 87% relative humidity at 15 °C will have an RH of 35% at 30 °C, while, in the extreme case of air saturated at 20 °C, the RH at 30 °C will be only 55%. Therefore, although the RH in the final inlet line was not measured, we expect it always to have been lower than ambient and consider it unlikely that particles entered the ATOFMS as solution droplets. Murphy et al. (34), likewise sampling a temperate marine boundary layer into a warm container, state the same conclusion. In a subsequent ship-borne campaign, we have measured an RH of ∼40% in the sampling line compared with an ambient RH of ∼70-80%. We therefore consider that chemical matrix effects influencing absorption of the LDI laser are the most probably explanation of the lower hit efficiencies encountered in clean air. We recognize, however, that both explanations remain viable and that this matter remains controversial. Comparison between MOUDI and ATOFMS. MOUDI data showed similar results from previous studies conducted in 1996 and 1997 at the same location (30, 31, 35). A detailed 5032
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 16, 2006
FIGURE 4. Size distribution of sodium, potassium, magnesium, and chloride for (a) Julian day 227 (08/15/2002) and (b) Julian day 225 (08/13/2002). description of the MOUDI data is not required here other than to support the interpretation of different dominant particle types on different days of the NAMBLEX campaign. The size distributions of Na+, K+, Mg+, and Cl- for JD 227 (08/15/2002) are shown in Figure 4a as an example. It can be seen that all the ions present a large monomodal distribution peaking at 2.5-4.5 µm, reflective of sea salt aerosols, the dominant class of particles for this day. Similar concentrations and size distributions for these ions were also apparent on Julian days 226 and 229. In contrast, the size distributions of Na+, K+, Mg+, and Cl- for Julian days 224 and 225 showed a bimodal distribution with modes at 1.5 and 8.5 µm, as illustrated in Figure 4b for JD 225 (08/13/2002). During these 2 days a distinct type of dust particle was detected with the ATOFMS, characterized by a strong aluminum/silicate signature. Five-day back trajectories indicated an origin from the Azores high-pressure region which can draw air from North Africa. We have previously speculated that these particles originated from the Sahara region (23). Julian day 228 (08/16/2002) was characterized by Ca-rich dust particles. The other days of the MOUDI dataset presented more complex size distributions due to the mixtures of different types of particle such as dust and the pure and reacted sea salt particles present. The MOUDI data provide a calibration comparison for the ion intensity signals measured by the ATOFMS, once the influence of inlet transmission and hit efficiencies are taken into account. Investigation of ATOFMS ion signal linearity has, to date, been largely based upon laboratorygenerated particles (15), although in one paper (21) ATOFMS chemical sensitivities were developed from co-located ATOFMS and impactor measurements taken during field experiments. A field-based approach requires the collection of a much larger number of single particle mass spectra in order to have a statistically robust number of particles from the complex mixture of different particles in the atmosphere.
FIGURE 5. Average ATOFMS Na+ ion signal per hit particle in each size range for days dominated by sea salt (pure and reacted). NAMBLEX provided a good opportunity in that a very large number of particles were detected and there were sustained periods of three main categories of chemical composition during the period that the impactor operated. Although some nine ions were analyzed in the MOUDI samples, due to the predominance of marine air masses during the period of impactor sampling, only the concentrations of sodium, chloride, and potassium were persistently high enough (>0.005 µg m-3 per MOUDI stage) to provide a good comparison with the ATOFMS data (although the latter’s concentrations were only sufficiently high in the coarser particle size ranges). A problem can arise from the high sensitivity of the ATOFMS to these chemical species resulting in an ion signal at the relevant diagnostic m/z value (23 ( 0.5, -35 ( 0.5, and 39 ( 0.5 Da for Na+, Cl-, and K+, respectively), exceeding the dynamic range of the data acquisition board; a phenomenon known as “peak clipping”. The absolute area of the ion signals was used to quantify the ATOFMS response to the corresponding species. Absolute area was chosen in preference to relative area as it is likely to be less affected by the total ion signal of the spectrum. Only peak areas with peak height of less than 250 units were considered in order to avoid saturated signals and the peak clipping phenomenon. About 18% of the sodium-containing particles used in the quantitative comparison to the MOUDI data exceeded the dynamic range of the data acquisition board, a greater proportion than in the work of Bhave et al. (21). However, the correlation between numbers of particles per day with saturated and nonsaturated peaks was very high (R2 > 0.9), and therefore we consider it likely that the integrated signal from the peaks of less than 250 units would be proportional to the total signal. Our results cannot, however, be regarded as absolute. The m/z values used to define chloride- and potassiumcontaining aerosol can also be attributed to other chemical components such as some organic clusters, e.g., C3H3+ (36). However, organic compound signals were relatively rare in the spectra observed in this period at Mace Head, and we think it unlikely that there is a very significant interference. About 39 and 21% of the chloride- and potassium-containing particles, respectively, used in the quantitative comparison with the MOUDI exceeded the dynamic range of the data acquisition board. Stages 4 (0.61-1.1 µm), 5 (1.1-2 µm), and 6 (2.0-3.4 µm) were considered in the analysis, but it must be noted that the finest particle size range contained much less alkali halide aerosol than the two larger particle size ranges (23) so there are less data for the comparison. (Stage 4 data were excluded altogether for the potassium comparison). Figure 5 shows the average ion signal per hit particle for Na+, binned by particle size range, for days dominated by sea salt. The key observation from the coincident, and horizontal, distribution of symbols is that the same amount
FIGURE 6. Total ATOFMS ion signal, scaled for inverse particle inlet efficiency and hit efficiency, for all hit particles per day in each size range, plotted against the corresponding MOUDI concentration (where, for example, 2.4E+09 represents 2.4 × 109): (a) Na+, (b) Cl-, and (c) K+. Error bars carried through from the inverse transmission efficiency function are plotted on the Cl- plot for illustration and are similar for the other species. Uncertainty in MOUDI data are not shown. of material is ablated/ionized from particles of this type regardless of particle size. (Note that the intensity of the LDI laser was maintained constant throughout data collection). This means that no additional variation in ATOFMS ion signal is introduced by the LDI process itself for these particles. The invariance in average ion signal with particle size implies that there is a particle-size dependence in the quantitative relationship between ATOFMS signal and the chemical content of the particle; i.e., the ATOFMS ablates a larger fraction of a smaller particle than of a larger particle. The incomplete vaporization of particles by LDI has been noted previously (37, 38). Quantitatively, a constant mass of ablated material equates to an inverse cubic dependence of fraction ablated material on particle diameter. This is demonstrated more clearly in Figure 6, which shows the sum, according to eq 5, of the measured ATOFMS signals for Na+, Cl-, and K+ for all particles in each of the size ranges for each VOL. 40, NO. 16, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5033
day, plotted against the MOUDI-derived concentration. Prior to the summation, each particle’s ion signal, Respj, was scaled according to the particle detection efficiency described above.
∑φ Resp j
j
j⊂i
(5)
V
(V is the volumetric flow rate of the ATOFMS instrument which remained constant throughout the campaign; the label j refers to an individual particle within each ensemble i comprising one day and one particle size range.) Thus, our procedure differs from that developed by Bhave et al. (21) in that our particle number scaling factor takes into account the variation in hit efficiency for particles of different composition. The data from each size range in Figure 6 lie on separate but approximately straight lines. The analysis shows that to obtain a linear relationship between ion signal and concentration of chemical species requires that variation in both inlet transmission and particle hit efficiencies be taken into account, but that an additional matrix-dependent LDI sensitivity is not required; i.e., variation in hit efficiency accounts satisfactorily for the sensitivity to a particular chemical species in a range of particle types. By continuing analogy with the work of Bhave et al. (21), we derive a function, ψj, accounting for the dependence of ion signal on particle size by fitting eq 6 simultaneously to all data per chemical species in Figure 6,
∑φ Resp ψ j
mi )
j
j⊂i
V
+ i
(6)
(7)
The regression for each chemical species yields the speciesdependent parameters δ and γ. The γ parameter represents the inverse of the sensitivity of the ATOFMS to the mass of a particular chemical species. Note that the interpretation of an LDI insensitivity with particle type is incorporated in the fact that γ is made constant for all particle types, i.e., a particle-type dependence in γ does not need to be instigated to yield the linearity observed in Figure 6. This γ invariance was not previously explicitly recognized by Bhave et al. (21), probably because they analyzed particles predominantly of one type only. The γ values obtained here cannot be meaningfully compared with each other or to those of Bhave et al. (21) because different subsets of particles were used for each species according to the extent of detector saturation (“peak clipping”) and because of different detector sensitivities. The δ parameter is, however, a physically meaningful quantity which shows the relative dependence of ATOFMS response to a particular species in different sized particles. Best-fit values of δ for Na+, Cl-, and K+ are given in Table 2, together with values for NH4+ and NO3- reported by Bhave et al. (21). The dependence of ψ on diameter approximately to the power 3 is consistent with a uniform signal per particle (Figure 5), as in a given aerosol mass particle number decreases with diameter to the power one-third. Our values of δ for Na+, 5034
9
chemical species
δ
no of particle ensembles in the regression
sodium, Na+ chloride, Clpotassium, K+ ammonium, NH4+ nitrate, NO+
2.7 ( 0.3 2.6 ( 0.3 2.7 ( 0.3 2.4 ( 0.4 2.4 ( 0.4
30 (this study) 30 (this study) 20 (this study) 30 (ref 21) 33 (ref 21)
Cl-, and K+ thus support the interpretation of Bhave et al. (21) of uniform ablation of NH4+ and NO3-. In conclusion, by analyzing a dataset of single particles consisting of distinct episodes of different particle composition type, we have shown there to be a strong particle compositional effect for the particle hit efficiency but no significant additional composition effect for the ablation and ionization process once the desorption/ionization energy couples to the individual particle.
Acknowledgments The authors are grateful to the Natural Environment Research Council for funding this work as part of the NAMBLEX Consortium project coordinated by Professor Dwayne Heard (University of Leeds). They also thank Dr. Darius Ceburnis (University College, Galway, Ireland) for providing APS data.
Literature Cited
j
where mi is the impactor concentration for each ensemble of single particles, i is the residual to be minimized in the regression, and the ATOFMS inverse sensitivity function to each chemical species, ψj, is expressed in terms of aerodynamic diameter by eq 7 in which γ and δ are speciesdependent constants.
ψ ) γDaδ
TABLE 2. Best-Fit Values of δ in the ATOFMS Inverse Instrument Sensitivity Function ψ Determined As Described in the Text
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 16, 2006
(1) Harrison, R. M.; Yin, J. X. Particulate matter in the atmosphere: which particle properties are important for its effects on health? Sci. Total Environ. 2000, 249, 85-101. (2) Penner, J. E.; Hegg, D.; Leaitch, R. Unraveling the role of aerosols in climate change. Environ. Sci. Technol. 2001, 35, 332A340A. (3) Pope, C. A. Review: Epidemiological basis for particulate air pollution health standards. Aerosol Sci. Technol. 2000, 32, 4-14. (4) McMurry, P. H. A review of atmospheric aerosol measurements. Atmos. Environ. 2000, 34, 1959-1999. (5) Gard, E.; Mayer, J. E.; Morrical, B. D.; Dienes, T.; Fergenson, D. P.; Prather, K. A. Real-time analysis of individual atmospheric aerosol particles: Design and performance of a portable ATOFMS. Anal. Chem. 1997, 69, 4083-4091. (6) Allen, J. O.; Fergenson, D. P.; Gard, E. E.; Hughes, L. S.; Morrical, B. D.; Kleeman, M. J.; Gross, D. S.; Galli, M. E.; Prather, K. A.; Cass, G. R. Particle detection efficiencies of aerosol time of flight mass spectrometers under ambient sampling conditions. Environ. Sci. Technol. 2000, 34, 211-217. (7) Moffet, R. C.; Shields, L. G.; Berntsen, J.; Devlin, R. B.; Prather, K. A. Characterization of an ambient coarse particle concentrator used for human exposure studies: aerosol size distributions, chemical composition, and concentration enrichment. Aerosol Sci. Technol. 2004, 38, 1123-1137. (8) Reents, W. D.; Schabel, M. J. Measurement of individual particle atomic composition by aerosol mass spectrometry. Anal. Chem. 2001, 73, 5403-5414. (9) Mansoori, B. A.; Johnston, M. V.; Wexler, A. S. Quantitation of ionic species in single microdroplets by online laser desorption/ ionization. Anal. Chem. 1994, 66, 3681-3687. (10) Carson, P. G.; Johnston, M. V.; Wexler, A. S. Laser desorption/ ionization of ultrafine aerosol particles. Rapid Commun. Mass Spectrom. 1997, 11, 993-996. (11) Reilly, P. T. A.; Lazar, A. C.; Gieray, R. A.; Whitten, W. B.; Ramsey, J. M. The elucidation of charge-transfer-induced matrix effects in environmental aerosols via real-time aerosol mass spectral analysis of individual airborne particles. Aerosol Sci. Technol. 2000, 33, 135-152. (12) Thomson, D. S.; Middlebrook, A. M.; Murphy, D. M. Thresholds for laser-induced ion formation from aerosols in a vacuum using ultraviolet and vacuum-ultraviolet laser wavelengths. Aerosol Sci. Technol. 1997, 26, 544-559. (13) Schoolcraft, T. A.; Constable, G. S.; Jackson, B.; Zhigilei, L. V.; Garrison, B. J. Molecular dynamics simulations of laser disintegration of amorphous aerosol particles with spatially non-
(14)
(15)
(16) (17) (18)
(19) (20)
(21)
(22)
(23)
(24)
uniform absorption. Nucl. Instrum. Methods Phys. Res., Sect. B 2001, 180, 245-250. Wenzel, R. J.; Prather, K. A. Improvements in ion signal reproducibility obtained using a homogeneous laser beam for on-line laser desorption/ionization of single particles. Rapid Commun. Mass Spectrom. 2004, 18, 1525-1533. Gross, D. S.; Galli, M. E.; Silva, P. J.; Prather, K. A. Relative sensitivity factors for alkali metal and ammonium cations in single particle aerosol time-of-flight mass spectra. Anal. Chem. 2000, 72, 416-422. Liu, D. Y.; Wenzel, R. J.; Prather, K. A. Aerosol time-of-flight mass spectrometry during the Atlanta Supersite Experiment: 1. Measurements. J. Geophys. Res. 2003, 108, art-8426. Lake, D. A.; Tolocka, M. P.; Johnston, M. V.; Wexler, A. S. The character of single particle sulfate in Baltimore. Atmos. Environ. 2004, 38, 5311-5320. Wenzel, R. J.; Liu, D. Y.; Edgerton, E. S.; Prather, K. A. Aerosol time-of-flight mass spectrometry during the Atlanta Supersite Experiment: 2. Scaling procedures. J. Geophys. Res. 2003, 108, art-8427. Liu, D. Y.; Prather, K. A.; Hering, S. V. Variations in the size and chemical composition of nitrate-containing particles in Riverside, CA. Aerosol Sci. Technol. 2000, 33, 71-86. Fergenson, D. P.; Song, X. H.; Ramadan, Z.; Allen, J. O.; Hughes, L. S.; Cass, G. R.; Hopke, P. K.; Prather, K. A. Quantification of ATOFMS data by multivariate methods. Anal. Chem. 2001, 73, 3535-3541. Bhave, P. V.; Allen, J. O.; Morrical, B. D.; Fergenson, D. P.; Cass, G. R.; Prather, K. A. A field-based approach for determining ATOFMS instrument sensitivities to ammonium and nitrate. Environ. Sci. Technol. 2002, 36, 4868-4879. Heard, D. E.; Read, K. A.; Al-Haider, S.; Bloss, W. J.; Johnson, G. P.; Pilling, M. J.; Seakins, P. W.; Smith, S. C.; Sommariva, R.; Stanton, J. C.; Still, T.; Ingham, T.; Brooks, B.; De Leeuw, G.; Jackson, A. V.; McQuaid, J. B.; Morgan, R.; Smith, M. H.; Carpenter, L. J.; Carslaw, N.; Hamilton, J.; Hopkins, J. R.; Lee, J. D.; Lewis, A. C.; Purvis, R. M.; Wevill, D. J.; Brough, N.; Green, T.; Mills, G.; Penkett, S. A.; Plane, J. M. C.; Saiz-Lopez, A.; Worton, D.; Monks, P. S.; Fleming, Z.; Rickard, A. R.; Alfarra, M.; Allan, J. D.; Bower, K.; Coe, H.; Cubison, M.; Flynn, M.; McFiggans, G.; Gallagher, M.; Norton, E. G.; O’Dowd, C. D.; Shillito, J.; Topping, D.; Vaughan, G.; Williams, P.; Bitter, M.; Ball, S. M.; Jones, R. L.; Povey, I. M.; O’Doherty, S.; Simmonds, P. G.; Allen, A.; Kinnersley, R. P.; Beddows, D. C. S.; Dall’Osto, M.; Harrison, R. M.; Donovan, R. J.; Heal, M. R.; Jennings, S. G.; Noone, C.; Spain, G. The North Atlantic Marine boundary layer experiment (NAMBLEX). Overview of the campaign held at Mace Head, Ireland, in summer 2002. Atmos. Chem. Phys. 2006, 6, 22412272. Dall’Osto, M.; Beddows, D. C. S.; Kinnersley, R. P.; Harrison, R. M.; Donovan, R. J.; Heal, M. R. Characterization of individual airborne particles by using aerosol time-of-flight mass spectrometry (ATOFMS) at Mace Head, Ireland. J. Geophys. Res. 2004, 109, D21302. DOI: 10.1029/2004JD004747. Prather, K. A.; Nordmeyer, T.; Salt, K. Real-time characterization of individual aerosol particles using time-of-flight mass spectrometry. Anal. Chem. 1994, 66, 1403-1407.
(25) Murphy, D. M.; Cziczo, D. J.; Hudson, P. K.; Schein, M. E.; Thomson, D. S. Particle density inferred from simultaneous optical and aerodynamic diameters sorted by composition. J. Aerosol Sci. 2004, 35, 135-139. (26) Armendariz, A. J.; Leith, D. Concentration measurement and counting efficiency for the aerodynamic particle sizer 3320. J. Aerosol Sci. 2002, 33, 133-148. (27) Stein, S. W.; Gabrio, B. J.; Oberreit, D.; Hairston, P.; Myrdal, P. B.; Beck, T. J. An evaluation of mass-weighted size distribution measurements with the Model 3320 aerodynamic particle sizer. Aerosol Sci. Technol. 2002, 36, 845-854. (28) Peters, T. M.; Leith, D. Concentration measurement and counting efficiency of the aerodynamic particle sizer 3321. J. Aerosol Sci. 2003, 34, 627-634. (29) Marple, V. A.; Rubow, K. L.; Behm, S. M. A microorifice uniform deposit impactor (MOUDI)-description, calibration, and use. Aerosol Sci. Technol. 1991, 14, 434-446. (30) Huang, Z.; Harrison, R. M.; Allen, A. G.; James, J. D.; Tilling, R. M.; Yin, J. X. Field intercomparison of filter pack and impactor sampling for aerosol nitrate, ammonium, and sulphate at coastal and inland sites. Atmos. Res. 2004, 71, 215-232. (31) Allen, A. G.; Harrison, R. M.; James, J. D.; Davison, B. M.; Robertson, L.; Hewitt, C. N. Impactor aerosol size distributions at two north Atlantic coastal sites. J. Aerosol Sci. 2000, 31 (Suppl. 1), 372-373. (32) Kane, D. B.; Johnston, M. V. Size and composition biases on the detection of individual ultrafine particles by aerosol mass spectrometry. Environ. Sci. Technol. 2000, 34, 4887-4893. (33) Neubauer, K. R.; Johnston, M. V.; Wexler, A. S. On-line analysis of aqueous aerosols by laser desorption ionization. Int. J. Mass Spectrom. Ion Processes 1997, 163, 29-37. (34) Murphy, D. M.; Thomson, D. S.; Middlebrook, A. M.; Schein, M. E. In situ single-particle characterization at Cape Grim. J. Geophys. Res. 1998, 103, 16485-16491. (35) O’Dowd, C. D.; Hameri, K.; Makela, J. M.; Pirjola, L.; Kulmala, M.; Jennings, S. G.; Berresheim, H.; Hansson, H. C.; de Leeuw, G.; Kunz, G. J.; Allen, A. G.; Hewitt, C. N.; Jackson, A.; Viisanen, Y.; Hoffmann, T. A dedicated study of new particle formation and fate in the coastal environment (PARFORCE): Overview of objectives and achievements. J. Geophys. Res. 2002, 107, art8108. (36) Silva, P. J.; Prather, K. A. Interpretation of mass spectra from organic compounds in aerosol time-of-flight mass spectrometry. Anal. Chem. 2000, 72, 3553-3562. (37) Carson, P. G.; Johnston, M. V.; Wexler, A. S. Real-time monitoring of the surface and total composition of aerosol particles. Aerosol Sci. Technol. 1997, 26, 291-300. (38) Carranza, J. E.; Hahn, D. W. Assessment of the upper particle size limit for quantitative analysis of aerosols using laser-induced breakdown spectroscopy. Anal. Chem. 2002, 74, 5450-5454.
Received for review May 18, 2005. Revised manuscript received March 3, 2006. Accepted May 24, 2006. ES050951I
VOL. 40, NO. 16, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5035
