Evaluation of the new capture vaporizer for aerosol mass spectrometers (AMS) through field studies of inorganic species

ABSTRACT

The aerosol mass spectrometer (AMS) and aerosol chemical speciation monitor (ACSM) are widely used for quantifying aerosol composition. The quantification uncertainty of these instruments is dominated by the collection efficiency (CE) due to particle bounce. A new “capture vaporizer” (CV) has been recently developed to achieve unit CE. In this study, we examine the performance of the CV while sampling ambient aerosols. AMS/ACSMs using the original standard vaporizer (SV) and CV were operated in parallel during three field studies. Concentrations measured with the CV (assuming CE = 1) and SV (using the composition-dependent CE of Middlebrook et al.), as well as SMPS and PILS-IC are compared. Agreement is good in all cases, verifying that CE ∼ 1 in the CV when sampling ambient particles. Specific findings include: (a) The fragmentation pattern of ambient nitrate and sulfate species observed with the CV was shifted to smaller m/z, suggesting additional thermal decomposition. (b) The differences in fragmentation patterns of organic vs. inorganic nitrate and sulfur species are still distinguishable in the CV, however, with much lower signal-to-noise compared to the SV. (c) Size distribution broadening is significant, but its impact is limited in field studies since ambient distributions are typically quite broad. Consistent size distributions were measured with the SV and CV. (d) In biogenic areas, UMR nitrate is overestimated based on the default fragmentation table (∼factor of 2–3 in SOAS) for both vaporizers, due to underestimation of the organic interferences. We also report a new type of small interference: artifact chloride signal can be observed in the AMS when high nitrate mass concentration is sampled with both the SV (∼0.5% chloride/nitrate) or CV (∼0.2% chloride/nitrate). Our results support the improved quantification with the CV AMS and characterize its chemical detection properties.

Copyright © 2017 American Association for Aerosol Research

EDITOR:

• Paul Ziemann

1. Introduction

Aerosol mass spectrometers (AMS) and aerosol chemical speciation monitors (ACSM) commercialized by Aerodyne Research Inc. (ARI) are used widely to measure the mass concentrations and size distributions of non-refractory species in submicron-particles (Jayne et al. 2000; Allan et al. 2003a,b; Canagaratna et al. 2007; Jimenez et al. 2009; Ng et al. 2011; Crenn et al. 2015, and references therein). With the vaporizer that is installed in almost all commercial instruments to date (referred to as “standard,” or SV in this work), the quantification of ambient aerosol mass concentration requires the use of a collection efficiency (CE) to correct for the fraction of particles that bounce off the vaporizer and are therefore not efficiently detected (Huffman et al. 2005; Quinn et al. 2006; Matthew et al. 2008; Middlebrook et al. 2012).

CE in the SV depends on aerosol phase, aerosol acidity/neutralization of sulfate content, ammonium nitrate (NH₄NO₃) fraction, and typically contributes most to the uncertainty in the quantification of AMS concentration measurements (Huffman et al. 2005; Bahreini et al. 2009; Middlebrook et al. 2012). Bahreini et al. (2009) estimated an uncertainty of AMS OA concentrations around ±38% and ±34% for inorganic aerosols. Middlebrook et al. (2012) developed a formulation to estimate CE for ambient particles from their chemical composition by comparing the mass concentration measured by AMS with independent measurements. Note that this CE cannot be used for the aerosol quantification from chamber studies or pure compounds or mixtures in the laboratory. A separate determination of CE is needed for such studies (Docherty et al. 2012; Robinson et al. 2016). Multiple studies using the Middlebrook CE formulation have shown good agreement with other instruments (e.g., Jimenez et al. 2016 and references therein). However, the remaining uncertainty impedes progress on other quantification issues and obscures detection of other instrument or analysis problems, as anomalous results that are often incorrectly attributed to an anomalous CE.

To improve AMS and ACSM quantification, a new “capture” vaporizer (CV) has been designed to reduce or eliminate the CE correction (Jayne and Worsnop 2016). Instead of an inverted cone as with the SV, the CV was designed to have a cage with a narrow entrance in order to increase the number of collisions and hence the residence time of particles in contact with the hot vaporizer surfaces, thus reducing the fraction of particles that bounce without evaporation (Figure 1a). Compared to almost all the SV in commercialized instruments are made of porous tungsten, the CV is made of solid molybdenum instead. When using the new PM_2.5 lens developed for AMS systems (Peck et al. 2016), use of the CV to minimize loss due to bounce for supermicron particles will be more important (Xu et al. 2017). Previously, the CV performance was examined in laboratory studies with several standard compounds (Hu et al. 2016b; Xu et al. 2017). Hu et al. (2016b) and Xu et al. (2017) independently found identical CE (= 1) for purely flash-evaporated species NH₄NO₃ and higher CE (>0.7) for less volatile ammonium sulfate [(NH₄)₂SO₄] and sodium nitrate (NaNO₃) in the CV vs. SV, demonstrating the improved particle collection of the CV. The fragmentation patterns for pure nitrate and sulfate salt aerosols are shifted to smaller ions in the CV due to the additional thermal decomposition, arising from longer residence time and increased collisions on the vaporizer surfaces.

Figure 1. (a) Schematic of the standard vaporizer (SV) and capture vaporizer (CV). (b) Setup used for the ambient comparisons of two AMSs equipped with SV and CV, respectively, in the field studies. All the sampled air was dried to be below 30% RH before reaching the instruments with Nafion dryers.

For the SV, ambient CE is often higher for ambient particles than for pure laboratory particles (Matthew et al. 2008; Middlebrook et al. 2012). Thus, a characterization of CV quantification for ambient particles is needed. It is also important to assess how the change in mass spectral patterns affects quantification, species identification, and potential interferences in the field. Finally, the performance of the AMS particle sizing measurement with the CV is evaluated. To test the performance of the CV, we operated AMS/ACSMs with SV and CV in parallel during three field studies under different ambient conditions, including one where aerosol was sampled from the output of an oxidation flow reactor (OFR) that processed ambient aerosols (Hu et al. 2016a). The performance of the CV in those studies is the focus of this manuscript.

2. Experiment and instrumentation

2.1. Field measurement setup

The three field studies analyzed in this work are introduced here:

The Southern Oxidant and Aerosol Study (SOAS) field study was carried out in the SE US in the summer (1 June to 15 July) of 2013 (Carlton et al. 2017). Results shown in this study were from the SEARCH Centreville supersite (CTR) in a semi-polluted mixed forest in Alabama (32.95° N, 87.13°W) (Hu et al. 2015, 2016a). A schematic plot of the instrumental setup is illustrated in Figure 1b. Two AMSs equipped with an SV and CV and a scanning mobility particle sizer (SMPS, model: 3080, TSI, Shoreview, MN, US) were operated in parallel and sampled ambient air and OFR-processed air alternating every 4 min.

Both ambient and OFR sampling lines were dried by Nafion driers (e.g., MD-110-24S-4, Perma Pure LCC) to keep RH below 30% before gas and aerosol entered instruments. The purpose of the drying is to ensure dry particles were measured with the AMS or SMPS, which is especially critical for estimating total dry volume from SMPS data, and to ensure that the Middlebrook et al. (2012) parameterization of CE in the SV, which was derived for dry ambient particles, can be applied to this dataset. The OFR data used in this study are from ambient air after exposure to high levels of nitrate (NO₃) radical. NO₃ radicals were generated through rapid thermal dissociation of injected N₂O₅ (Brown and Stutz 2012; Palm et al. 2017). The NO₃-OFR was developed to investigate secondary organic aerosol (SOA) formation from biogenic VOCs with NO₃ radicals. However, the feature of the OFR data that make them useful for this work is that the processed air exhibited sporadic rapid changes in aerosol nitrate fraction, due to formation of high concentrations of NH₄NO₃ particles formed from nitric acid (HNO₃) that was either injected along with the N₂O₅ or formed by hydrolysis of N₂O₅ on wetted OFR walls in some cases (Palm et al. 2017). This provided an excellent test case for the application of composition-dependent CE in the SV by comparing its time series with those from the SMPS and CV.

Two additional studies were conducted by sampling outdoor air through inlets at the University of Colorado (Boulder, CO, USA) and Aerodyne Research (Billerica, MA, USA). Instrumentation setup in those two studies was similar to SOAS (Figure 1b), except that SMPS and OFR data were not collected. In the Boulder study, substantial amount of OA from primary sources was observed. This is in contrast to SOAS, where biogenic SOA was dominant (Hu et al. 2015; Xu et al. 2015). In the Billerica study, ACSMs instead of AMSs were used. A summary of field sampling locations, times, and instrumentation details is given in Table 1.

Table 1. Brief overview of three field experiments shown in this study.

			AMS type and Tv
Field study name	Location	Sampling period	SV	Tv (°C)	CV	Tv (°C)	Time resolution	PToF in AMS with CV
SOAS	Centerville site in SE forest, AL, USA	5 June–10 June 2013	HR-ToF	600	HR-ToF	550	2 min	NO
SOAS	Centerville site in SE forest, AL, USA	19 June–16 July 2013	HR-ToF	600	C-ToF	550	2 min	NO
Boulder ambient study	CU laboratory on the campus, Boulder, CO, USA	31 March–1 April 2013	HR-ToF	525	HR-ToF	520	2 min	Yes
Billerica ambient study	laboratory at Aerodyne Research Inc. (ARI), Billerica, MA, USA	31 August–4 September 2012	ACSM	600	ACSM	600	30 min	NO

2.2. AMS/ACSM measurements

In the SOAS study, a high-resolution time-of-flight AMS (HR-ToF-AMS; DeCarlo et al. 2006) equipped with an SV was used for the entire campaign. The CV data for the SOAS study was obtained with a compact time-of-flight AMS (C-ToF-AMS; Drewnick et al. 2005) for most of the campaign, and with a second HR-ToF-AMS for a few days (7–10 June). The Boulder field study was conducted with two HR-ToF-AMS instruments. In the Billerica study, two Aerosol Chemical Speciation Monitors (ACSM; Ng et al. 2011) were used, which provides a good opportunity to evaluate the performance of CV using unit mass resolution (UMR) analysis with different aerosol/background method and sequence than in the AMS. Concentrations and size distributions of aerosol species collected with the AMS were analyzed with standard ToF-AMS analysis software packages (Squirrel version >1.52M and PIKA version >1.12). The ACSM datasets were analyzed with ACSM analysis software (version = 1.5.5.0).

Before each study, the alignment of the particle beam (“Aerodynamic lens alignment”) was verified on each AMS/ACSM to ensure the focused particle beam impacted in the center of the vaporizers. For AMS, aerosol mass concentrations were calculated using the standard method for ambient studies, i.e., as the difference signal between ambient particles (particle beam “open”: 6 or 5 s) and background signal (particle beam “closed”: 4 or 5 s) in mass spectrum (MS) mode (Jimenez et al. 2003). ACSM does not have an internal mechanism to open/close the particle beam, so the background signal is instead measured by sampling ambient filtered air (Ng et al. 2011). This design feature, together with the slower time response of the ACSM mass spectrometer, results in a longer time step for ambient/background switching (∼30 s open/30 s filtered in this study). The ionization efficiency (IE) in the AMS was calibrated with monodisperse 400 nm dried NH₄NO₃ particles every few days during the field studies. Both brute-force single particle (BFSP, DeCarlo et al. 2006) and condensation particle counter (CPC) calibrations (DeCarlo et al. 2006; Ng et al. 2011) were performed for the SV-AMS while only the latter method was applied to the CV-AMS and ACSM. The BFSP method, while preferable, is not possible for a CV-AMS since single particle events are not always detectable due to the temporal broadening of single particle signals in the CV (Hu et al. 2016b). The relative IE (RIE) of ammonium and sulfate in the AMS was measured with pure NH₄NO₃ and (NH₄)₂SO₄ particles (Canagaratna et al. 2007). Compared to the default SO₄ RIE of 1.2, the CV-AMS always exhibited higher sulfate RIEs (1.7–2.4); however, ∼1 was found in the CV-ACSM. This difference between CV-AMS and -ACSM may be due to the different background method, or possibly to differences in ionizer and extractor size and geometry.

The HR and UMR fragmentation table treatment of m/z 18 (H₂O⁺) and m/z 32 (S⁺) from sulfate in AMS with CV was adjusted according to the ratio reported in Hu et al. (2016b). Other fragmentation table entries were unchanged from the default table applied for analysis of SV-AMS datasets. The default RIE of ambient OA (1.4) and chloride (1.3) were used in all the datasets for both vaporizers. The RIE of chloride in both vaporizers from HR-ToF-AMS (∼1.3–1.4) was calibrated in the laboratory after the field campaigns, which are similar to the default value. The particle size measured by the AMS was calibrated with dried monodisperse pure NH₄NO₃ or (NH₄)₂SO₄ particles from 70 to 550 nm, size-selected with a differential mobility analyzer (DMA).

2.3. SMPS measurement

The sizing measurement of the SMPS was verified with monodisperse polystyrene latex spheres (PSLs; Duke Scientific, Palo Alto, CA, US) with nominal standard diameters of 250–400 nm. The flow rates of the sheath and sample flows of the SMPS were also verified. The submicron volume concentration was estimated from SMPS measurement by integration of the measured number distributions. The mass concentration was calculated by multiplying the volume concentration by a bulk aerosol density estimated from the AMS chemical composition (assuming internally mixed particles). The aerosol density can be calculated by Equation (1[1] ) (DeCarlo et al. 2008):[1] where [OA] is the mass concentration of organic aerosol, and is the density of each component (g cm⁻³). was estimated from its oxidation level as measured by the SV-AMS (Kuwata et al. 2012). The elemental ratios used for the density estimation were based on the updated calibrations of Canagaratna et al. (2015). The calculated density from the updated calibration method actually shows slightly better agreement with measured density of chamber SOA (Slope = 0.98) than when using the original Aiken method (Slope = 0.95; Aiken et al. 2008), as shown in Figure S1. A time series of OA density (changing with the ambient OA elemental composition) was used in Equation (1[1] ). In SOAS, the average estimated ambient OA density was 1.34 ± 0.005 g cm⁻³ (variability in time, ±1σ std. err.). The inorganic densities used in Equation (1[1] ) were estimated from Haynes (2015) and held constant. The 1σ uncertainty of the total aerosol density (12%) was estimated by Montecarlo error propagation, assuming 15% (5%) error in OA (inorganics) density. The total uncertainty of the SMPS mass concentration was estimated to be 28% by propagating the uncertainties from aerosol density (12%) and SMPS volume estimation (25%).

2.4. Organic nitrate detection

To interpret the fragmentation patterns of ambient inorganic nitrate, it is necessary to first characterize the detection of organic nitrates in the CV, since those species were significantly present during, e.g., SOAS (Ayres et al. 2015; Lee et al. 2016).

The main fragment ions of nitrate for NH₄NO₃ particles are NO⁺ and NO₂⁺, which typically contribute >95% of the total nitrate signal in both vaporizers (Hogrefe et al. 2004; Hu et al. 2016b). However, nitrate groups (−ONO₂) from organic nitrate (RONO₂) can generate the same ions (Farmer et al. 2010). For the same instrument and operating conditions, the NO₂⁺/NO⁺ ratio for organic nitrate is typically a factor of ∼2–3 lower than for pure NH₄NO₃ (Rollins et al. 2009; Farmer et al. 2010; Fry et al. 2011, 2013). RONO₂ is thought to decompose thermally at the AMS vaporizer into gas-phase NO₂ and RO. NO₂ can then thermally decompose to gas-phase NO. Thus, the gas-phase products of thermal decomposition of RONO₂ are a combination of NO₂ and NO. For NH₄NO₃, the products are thought to be gas-phase HNO₃, NO₂, and NO in the SV (Drewnick et al. 2015; Hu et al. 2016b; Pieber et al. 2016). The low NO₂⁺/NO⁺ for organic nitrate is likely due to the dominance of gas-phase NO₂ and NO, which produce much lower NO₂⁺/NO⁺ (∼0.3–0.5 and zero, respectively) than HNO₃ (∼1.17–2) after 70eV EI ionization (Friedel et al. 1959; Hu et al. 2016b; Linstrom and Mallard 2016). Thus, using measured NO₂⁺/NO⁺ ratios from pure NH₄NO₃ and ambient NO₂⁺/NO⁺ ratios measured in the AMS, NH₄NO₃, and organic nitrate fractions can be estimated (Fry et al. 2013). Note that to use this method, other inorganic nitrate salts (e.g., NaNO₃ or Ca(NO₃)₂) need to be negligible within the AMS submicron size ranges, since they show comparably low NO₂⁺/NO⁺ ratios (0.01–0.1) as organic nitrate (Allan et al. 2004a; Farmer et al. 2010; Hu et al. 2016b).

Furthermore, we investigated whether those differences of NO₂⁺/NO⁺ ratios between pure NH₄NO₃ and organic nitrate can still be discerned with the CV. In chamber studies where large amount of organic nitrates (>1–8 µg m⁻³ –ONO₂ functional group mass) were formed through α-pinene+NO₃ radical chemistry, the NO₂⁺/NO⁺ ratio in the CV (∼0.0045) were 10 times lower than from pure NH₄NO₃ (∼0.06, Figure 2a). This suggests that organic nitrate and pure NH₄NO₃ in CV can be still separated in the mixed nitrate aerosol based on their differences of NO₂⁺/NO⁺ ratios. However, the application of this method is more likely to be limited by the signal-to-noise levels at low concentrations, compared to using the method for SV data. Based on the results in Figure 2a, we estimated that an order-of-magnitude higher nitrate concentration (∼100 ng m⁻³) is needed for the CV than for the SV (10 ng m⁻³) to distinguish NO₂⁺/NO⁺ from NH₄NO₃ vs. organic nitrate, as shown in Figure 2b.

Figure 2. (a) Time series of NO₂⁺/NO⁺ ratios for CV and SV, total nitrate for secondary organic aerosol produced from α-pinene oxidation by NO₃ radical in environmental chamber experiments. The nitrate present was entirely organic. The difference of total nitrate signal between the SV and CV at high nitrate concentration loadings was likely caused by the different AMS aerodynamic lens inlet size cuts as the particles grew very large at the end of the experiment. (b) 1-min measurement noise in NO₂⁺/NO⁺ ratios, quantified as the standard deviation of 1-min data points during periods of constant concentration in (a). Regression lines for each trace show that RONO₂ can be theoretically identified with one vaporizer until the noise level becomes comparable with the actually NH₄NO₃-RONO₂ difference, which corresponds to ∼10 ng m⁻³ for SV and 100 ng m⁻³ for CV. Note that the identification can be more difficult in field data for a variety of reasons.

Figure 3. Ratios of intensity for major sulfur-containing fragment ions (SO₂⁺, SO₃⁺, HSO₃⁺, and H₂SO₄⁺) to SO⁺, for ammonium sulfate and three sulfone molecules in (a) the SV and (b) the CV. Sulfones tested were benzyl sulfone (C₁₄H₁₄O₂S), diphenyl sulfone (C₁₀H₁₂O₂S), and bis (4-hydroxy-phenyl) sulfone (C₁₀H₁₂O₄S). The uncertainty of each ratio was obtained from the regression slope uncertainty when analyzing multiple 20 s laboratory measurements.

2.5. Organic sulfur detection

Ambient aerosols are known to sometimes contain organic sulfur species, and their presence might be detectable using fragment ratios, analogously to organic nitrates. It is of interest to evaluate the ability of the SV or CV to allow such investigations. As an example, we examined organic sulfones (R-S(=O)₂, e.g., dimethyl sulfone or bis-hydroxymethyl sulfone). Sulfones have been reported to be oxidation products of dimethyl sulfide (DMS) or SO₂ from power plant and urban plumes (Eatough and Hansen 1984; Berresheim et al. 1993). We examined the H_ySO_x⁺/SO⁺ ratios in three different organic sulfone standards in the laboratory, i.e., benzyl sulfone (C₁₄H₁₄O₂S), diphenyl sulfone (C₁₂H₁₀O₂S), and bis (4-hydroxy-phenyl) sulfone (C₁₂H₁₀O₄S), together with inorganic (NH₄)₂SO₄ in both the SV and CV (Figure 3). The SO₃⁺/SO⁺ ratios from different types of organic sulfone (10⁻³) were found to be more than 2–3 orders of magnitude lower than those in (NH₄)₂SO₄ (0.4) in the SV, while other H_ySO_x⁺/SO⁺ ratios showed diverse trends. This substantially low abundance of SO₃⁺ has also been observed in SV-AMS study for hydroxymethanesulfonate (HMS), which is a tracer of aqueous-phase chemistry (Gilardoni et al. 2016). The differences of H_ySO_x⁺/SO⁺ fragmentation patterns have also been observed between organosulfate and ammonium sulfate in the laboratory (Huang et al. 2015). Although lower H_ySO_x⁺/SO⁺ ratios were observed for both sulfones and (NH₄)₂SO₄ in the CV due to additional thermal decomposition, the differences of SO₃⁺/SO⁺ ratios between organic sulfone (∼10⁻⁶ to 10⁻⁵) and (NH₄)₂SO₄ (10⁻²) are still discernable in the CV. However, for typical mass concentrations of ambient aerosols, those differences in the CV may be more limited by signal-to-noise levels, similar to organic/inorganic nitrate separation in the CV.

3. Results and discussion

3.1. Fragmentation patterns

3.1.1. Sulfate

The ambient fragmentation patterns of sulfate for both vaporizers in the Boulder and SOAS studies are shown in Figure 4. The ambient H_ySO_x⁺/SO⁺ ratios for both vaporizers were consistent with those from pure (NH₄)₂SO₄ calibrations (Hu et al. 2016b; Xu et al. 2017), supporting the dominance of (NH₄)₂SO₄ over organic sulfur species in both campaigns. The CV exhibited slightly higher SO₂⁺/SO⁺ ratios (1.47–1.72) than the SV (1.26–1.34), while lower SO₃⁺/SO⁺ and HSO₃⁺/SO⁺ ratios (<0.03) were observed in the CV than SV (0.2–0.4). The quantification of UMR SO₃⁺ and HSO₃⁺ are more easily influenced by organic aerosols than UMR SO₂⁺ and SO⁺ due to the much larger contributions of organic ions at those m/z, as shown in Figure S2 (C₅H₄O⁺ and C₆H₈⁺ at m/z 80; C₅H₅O⁺ and C₆H₉⁺ at m/z 81). Although those contributions are approximately subtracted with the fragmentation table approach (Allan et al. 2004b) that method can have errors when the organic fractions of the UMR signals are large. Thus, in ambient plumes with high OA concentrations and relative high OA/SO₄ ratios such as biomass burning plumes, UMR sulfate in the SV may have higher errors (Milic et al. 2016). UMR quantification of sulfate with the CV could more easily be based entirely on SO⁺ and SO₂⁺ (by modifying the fragmentation table), eliminating this error.

Figure 4. Scatterplots of major ion ratios in the SV vs. the CV (a vs. b) for sulfate, (c vs. d) for nitrate, and (e vs. f) for ammonium during the SOAS and Boulder ambient studies. The ratios of fragment ions from pure (NH₄)₂SO₄ and NH₄NO₃ particles during SOAS calibration experiments are also shown. For clearer comparison, all signals were normalized to the maximum signal of each species on the x-axis during each study. Note that the ambient HR nitrate, sulfate, and ammonium shown from the CV in SOAS are from the 3 day long HR dataset (see methods section). The UMR data for CV are from C-ToF. All SV data are HR. All lines were determined using orthogonal distance regression with fits forced through zero. Regression slopes are given in the figure legends.

No significant variations in time of the sulfate fragmentation pattern were observed in both vaporizers during the SOAS study. Organosulfate groups (−OSO₃) from species derived from isoprene epoxydiols oxidation consistently accounted for a very small fraction (∼5%) of measured total sulfate signal in AMS (Hettiyadura et al. 2015; Hu et al. 2015; Liao et al. 2015; Riva et al. 2016). This is consistent with the small variation of the sulfate fragmentation pattern in SOAS. Additional investigation of organosulfate detection with the CV is recommended.

3.1.2. Nitrate

The ratios of fragment ions of NO₂⁺/NO⁺ from the SOAS and Boulder studies are shown in Figure 4. Generally, ambient NO₂⁺/NO⁺ ratios for the SV (0.2 and 0.34, respectively) were higher than for the CV (0.03–0.04), which is consistent with the results for pure NH₄NO₃ on both vaporizers (Hu et al. 2016b). Much lower NO₂⁺/NO⁺ ratios were seen in the ambient SOAS study for the SV dataset than the ratios from pure NH₄NO₃ particles during calibrations (Figure 4), indicating a substantial RONO₂ contribution to the ambient nitrate of SOAS study. By applying the RONO₂ method of Fry et al. (2013) to SOAS, ∼80% of ambient submicron nitrate was estimated to be RONO₂, which agreed reasonably well with RONO₂ measured with some independent methods (Ayres et al. 2015; Lee et al. 2016). During SOAS, the NO₂⁺/NO⁺ (HR) ratios for the CV (0.03) were lower than for pure NH₄NO₃ (0.04, Figure 4), which is qualitatively consistent with the substantial presence of RONO₂ during this study. However, NH₄NO₃ and RONO₂ cannot reliably be discriminated using the Fry et al. method with the CV data due to the very low ambient nitrate mass concentrations (40 ± 20 ng m⁻³), which is lower than the estimated limit of detection for this method (100 ng m⁻³) discussed in Section 2.4. The NO₂⁺/NO⁺ ratio for ambient nitrate in the CV in SOAS (0.03, Figure 4d, when ∼80% of the nitrate is organic) is higher than that obtained from the pure organic nitrate (∼0.0045) in the chamber studies (Figure 2). Possible reasons for this difference include, (1[1] ) different instrument conditions, (2) different type of organic nitrates between chamber and ambient studies, (3) organic interference at m/z 30 and m/z 46 in the UMR ambient dataset causing a larger quantification uncertainty of those ratios, or (4) other small unknown effects that become important at low signal levels.

3.1.3. Ammonium

Fragment ions from ammonium are thought to arise mainly from the direct ionization of gas-phase NH₃ formed from thermal decomposition of the salts (Hu et al. 2016b). In both the SV and CV, the fragmentation pattern of ammonium in the ambient SOAS datasets show ratios of NH₃⁺/NH₂⁺ (∼1.2–1.3) and NH⁺/NH₂⁺ (∼0.06) that are consistent with ratios from pure NH₄NO₃ calibration particles measured during the campaign (Figure 4). For SOAS with varied nitrate/sulfate mass ratio in the NO₃-OFR (Figure S3), constant NH₃⁺/NH₂⁺ ratios were observed, consistent with previous results (Hu et al. 2016b; Jimenez et al. 2016; Xu et al. 2017).

3.2. Quantification of CE

The comparisons of ambient total and inorganic species mass concentrations for both vaporizers are shown in Figures 5–9. Tables 2 and 3 show the regression slopes and correlation coefficients (R) of those comparisons from the different studies. A composition-dependent CE was applied to the SV-AMS data (Middlebrook et al. 2012). The time dependence of the estimated CE for both the ambient and NO₃-OFR datasets of the SV-AMS during SOAS are shown in Figure S4. For the ambient data, most of the CE values were between 0.5 and 0.6 with periodic increases up to 0.7–0.8. All datasets using the CV were analyzed using a CE of 1. In the following discussion, we summarize and discuss the comparison results by species.

Figure 5. Regression slopes and correlation coefficients (R) of total submicron aerosol mass concentration between AMS and SMPS in ambient and NO₃-OFR datasets during SOAS study. The same parameters are shown for comparisons of the main chemical species between SV-AMS and CV-AMS, and PILS-IC vs. CV-AMS in different studies. This plot is the summary plot of Tables 2 and 3. An uncertainty band of 44% is shown, which is calculated based on the quantification uncertainties of AMS and SMPS. The total uncertainty for the AMS vs. PILS-IC comparison is similar.

Figure 6. Ambient time series and scatterplots of ambient (a) total, (b) sulfate, (c) nitrate, (d) ammonium, and (e) chloride between the SV and CV during the SOAS study. Two SMPS datasets and one PILS-IC dataset from the SOAS field study are also shown here to compare to total and speciated inorganic AMS mass. All black solid lines were determined using orthogonal distance regression with fits forced through zero intercept and in all figures shown below. Regression slopes are given in the figure legends. The gray solid lines and in all figures shown below are the orthogonal distance regression lines without intercept constraining to be zero. The values from both regression methods are listed in Table S2.

Figure 7. Time series and scatterplots of (a) total, (b) sulfate, (c) nitrate, (d) ammonium, and (e) chloride for the SV and CV after NO₃-OFR during the SOAS study. SMPS data coupled with the NO₃-OFR in the SOAS field study are also shown.

Figure 8. Ambient time series and scatterplots of (a) total, (b) sulfate, (c) nitrate, (d) ammonium, and (e) chloride for the SV and CV in the Boulder study.

Figure 9. Ambient time series and scatterplots of (a) total, (b) sulfate, (c) nitrate, and (d) ammonium for the SV and CV in the Billerica study.

Table 2. Summary of orthogonal distance regression slopes (with fits forced through zero intercept^*) and correlation coefficients (R) of total and inorganic species measured by AMS or ACSM between the SV and CV and PILS-IC vs. CV in different field studies. Those results are also illustrated in Figure 5.

	SOAS ambient		SOAS NO3-OFR		Boulder study		Billerica study		SOAS ambient
Species	HR-ToFSV vs. C-ToFCV		HR-ToFSV vs. C-ToFCV		HR-ToFSV vs. HR-ToFCV		ACSMSV vs. ACSMCV		PILS-IC vs. C-ToFCV
	Slope	R	Slope	R	Slope	R	Slope	R	Slope	R
Total	0.90	0.98	0.98	0.98	1.1^a	0.98	1	0.94	N/A^c	N/A^c
Sulfate	1.07	0.98	1.1	0.96	0.93	0.96	0.73	0.97	1.09	0.92
Nitrate	0.91	0.97	1.02	0.99	0.89	0.99	0.81	0.76	N/A^d	N/A^d
Ammonium	1.03	0.96	1.03	0.99	0.96	0.99	0.94	0.65	1.16	0.90
Chloride	0.65	0.33	2.4	0.82	0.52^b	0.60	Below DL	Below DL	N/A^e	N/A^e

* During the field study, the zeros of both AMS for all species were checked frequently based on sampling particle-free air with filters. Thus, we do believe that forcing the intercepts through zero is superior for this dataset. We do not think the regression intercept is meaningful, but rather it just reflects random effects. For completeness, the results of regression with non-forced intercepts are documented in Table S2 and generally show small differences in slopes.

^a Intercept = −0.82, max concentration in SV = 6 µg m⁻³.

^b Intercept = 0.003, max chloride concentration in SV = 0.024. µg m⁻³; DL = detection limit.

^c Not available.

^d Not available. Nitrate measured by PILS-IC is exclusively inorganic, While AMS measured total nitrate. For the SOAS study, ∼80% of total submicron nitrate signal is organic nitrate in nature (Ayres et al. 2015). A separation of organic and inorganic nitrate using CV-AMS is not possible for SOAS data due to the low signal-to-noise ratio of NO₂⁺/NO⁺. Thus, a direct comparison of PILS-IC and CV-AMS inorganic nitrate is not possible.

^e Not applicable. The chloride signal is too low to draw any firm conclusions on comparison.

Table 3. Summary of orthogonal distance regression slopes (with fits forced through zero intercept) and correlation coefficients (R) for SMPS vs. AMS for total aerosol mass. Those results are also illustrated in the Figure 5.

	SOAS ambient air		SOAS NO3-OFR
	Slope	R	Slope	R
SMPS1 vs. AMSSV	0.98	0.91	0.99	0.96
SMPS2 vs. AMSSV	1.03	0.91	NA	NA
SMPS1 vs. AMSCV	0.89	0.89	1.01	0.95
SMPS2 vs. AMSCV	0.90	0.89	NA	NA

3.2.1. Total NR-PM₁ aerosol

Good agreement of total aerosol mass concentrations between two SMPS and CV-AMS was also observed in SOAS, with regression slopes between 0.89 and 1.0 and (Figure 5 and Table 3). The uncertainty (accuracy) of the AMS to SMPS ratio was estimated to be 44% by propagating the uncertainties of AMS quantification (36%) and SMPS total mass estimation (28% in Section 2.3). The consistency of total aerosol mass concentrations measured between SMPS and CV-AMS suggests the CV is capable of quantifying non-refractory sub-micron ambient aerosols with negligible particle bounce losses. The CE of ambient aerosol in the CV is roughly equivalent to 1. More ambient comparisons between SV and CV, in conjunction with SMPS, in areas with externally mixed aerosol are recommended.

Good correlation of the total NR-PM₁ concentrations between SV and CV was also found for the SOAS studies (R > 0.94, Figures 5–7 and Table 2). The total mass concentrations of aerosols detected by the CV (CE = 1) were within 12% of those measured by the SV when the composition-dependent CE was used. During the three ambient studies, the main chemical component of non-refractory submicron aerosol was OA (>50%, as shown in Figure 10). Therefore, the good correlation of total aerosols between the SV and CV is also indicative of good agreement of OA in those studies. The small intercept of total aerosols in the comparison for the Boulder study (Figure 8a) was mainly caused by the enhanced signals of OA ions below m/z 44 in the CV. The Boulder study is the only study where a significant intercept was observed. The campaign-average chemical compositions of ambient aerosols measured by the SV and CV were very similar (Figure 10), indicating that aerosol composition measurements with the CV can be accurate when ammonium, sulfate, and chloride RIE are measured. A detailed analysis of the comparisons for OA composition and quantification will be presented in a future publication.

Figure 10. Average chemical compositions of non-refractory submicron aerosols during SOAS, Boulder, and Billerica studies for the SV and CV. Note that the average chemical composition in SOAS shown here is for the period 19 June–16 July 2013.

3.2.2. Sulfate

In all studies, the time series of sulfate from the SV and CV all tracked well, with all Pearson's R values above 0.96. The regression slopes (0.93–1.1) between the SV and CV for the SOAS and Boulder studies were within 10% of the 1:1 line (Figures 5–9). Good agreement of sulfate with another independent instrument, particle-into-liquid sampler coupled to ion chromatography (PILS-IC, Guo et al. 2015) vs. CV during SOAS study is also observed (slope = 1.09 and R = 0.92), as shown in Figure 6b2. In the Billerica study (Figure 9), the lower regression slope (0.73) between two ACSMs may be due to the uncertainty of default sulfate RIE (1.2) applied in the SV that was not calibrated prior to the study. A sulfate RIE of 1 for the SV-ACSM (which is within the range of calibrated RIEs observed in previous studies) would result in a comparable agreement to the AMS datasets (Crenn et al. 2015). Laboratory-generated pure (NH₄)₂SO₄ showed a CE below 1 in the CV, ∼0.8 (Hu et al. 2016b) and 0.9 (Xu et al. 2017) at normal vaporizer temperatures (500–600°C). However, the good agreement of ambient sulfate for the CV and SV in different studies suggested that the CE of ambient sulfate from the CV was ∼1. Similarly, the CE of ambient sulfate when using the SV (>0.5) is also usually higher than that of pure (NH₄)₂SO₄ (0.2–0.4) due to less bounce (Matthew et al. 2008). The better detection of sulfate in the ambient aerosols than pure sulfate species is likely related with its internal mixing with nitrate and OA, which reduces the bounce fraction (Canagaratna et al. 2007). Xu et al. (2017) also showed an increased sulfate CE with higher fractions of NH₄NO₃ in the NH₄NO₃ and (NH₄)₂SO₄ mixtures for the CV in the laboratory studies.

Besides the comparison of sulfate difference signal (open minus close signal) discussed above, it is also interested to compare the sulfate background detection. It can help understanding if there is slow desorption for sulfate in the CV due to its longer residence time than SV. The time series of particle beam open (ambient + background) and closed (background) sulfate signals in SOAS are shown in Figure S5. Consistent with laboratory experiments (Hu et al. 2016b), much lower ambient sulfate background was found for the CV than for the SV. Experiments have been conducted in the laboratory to investigate the lifetime of sulfate signal decay by maintaining the particle beam close/open time intervals at several minutes (Hu et al. 2016b). The lower background for ambient sulfate in the CV is mainly caused by the faster sulfate signal decay in the CV (τ < 2 s) than that in SV (τ ∼ 13 s) at vaporizer temperatures around 500–600°C. The interval of MS mode switching during ambient measurement in both AMS was around 4–6 s. Systematically higher closed/open ratios were also observed in the ambient SV datasets (∼0.17 vs. 0.05–0.12 in MS mode in laboratory studies) than those for the CV (∼0.04 in the ambient dataset and ∼0.01 in the laboratory studies, Hu et al. 2016b).

3.2.3. Nitrate

The time series of HR nitrate mass concentration between the two vaporizers agreed well with each other with a narrow range in slopes of 0.89–1.02 (correlation coefficient R > 0.97, Table 2) for the SOAS (including the NO₃-OFR measurement) and Boulder studies. A very good linear correlation for nitrate between the two vaporizers was observed over a large range of concentrations (0–60 µg m⁻³, Figure 7). The good correlation indicates the CE of ambient nitrate is also approximately unity in the CV. Note that nitrate signals for all studies shown here are the total nitrate signal without separating organic nitrate and inorganic nitrate.

Larger deviations in the comparisons of NO₃ mass concentrations (slope = 0.81) were observed for the UMR Billerica dataset (both UMR ACSM) studies when the default fragmentation table was used (Figure 9c). A similar regression slope (0.82) was also found in the mass comparison of UMR nitrate (default fragmentation table applied) between both vaporizers in SOAS, as shown in Figure S9. The high-resolution fitting showed that at m/z 30 and 46 where most nitrate signal is detected, there are other organic ions (e.g., m/z 30: CH₂O⁺ and CH₄N⁺; m/z 46: CH₂O₂⁺), as shown in Figure S6. This larger slope difference could be due to changing interferences of organic ions to mass spectra obtained from SV and CV slightly differently that are not captured with default fragmentation table.

CH₂O⁺ was found to be a prominent ion at m/z 30 in the SOAS study, which has also been observed in other biogenically influenced areas with low nitrate concentrations, such as in the Amazon region (de Sá et al. 2016). In the presence of these organic interferences, the total nitrate mass concentrations calculated for the SV in SOAS, if only UMR analysis with the default fragmentation table were applied, would result in an overestimation by a factor of ∼2.3 compared to using HR analysis (Figure S10).

These overestimations suggest that the default UMR fragmentation table, at least for studies dominated by biogenic emissions (e.g., southeast US or Amazon), will require updated corrections regardless of vaporizer configuration. Based on results from the SOAS study, modified organic interference subtractions in the fragmentation table are shown in Table S1. By using the revised fragmentation tables, much better agreement of UMR and HR nitrate mass concentrations was found for both vaporizers (slope = 0.8–1) than when using the default UMR fragmentation table (∼2.3–2.7). This modified table is calculated only based on SOAS study. More detailed examination on AMS quantification for UMR inorganic nitrate from other areas with strong biogenic emissions and other types of emissions are recommended.

3.2.4. Ammonium

Ammonium also showed good agreement between the SV and CV with a regression slope ranging between 0.94 and 1.03 with R > 0.96 in the HR and C-ToF datasets and R = 0.65 in the UMR datasets of the ACSM comparison (Figure 5 and Table 2). Good agreement of ammonium between CV and PILS-IC during SOAS study are also found (slope = 1.16; R = 0.90, Figure 6d2). The lower correlation coefficient of ammonium in the Billerica ambient dataset is probably due to low signal-to-noise ratio, due in part to the noise left after subtracting average interferences of air fragments in the ACSM (e.g., m/z 16 O⁺ to NH₂⁺ or m/z 17 OH⁺ to NH₃⁺), the lower sensitivity of the ACSM compared to the AMS, and the relatively low ammonium concentrations (Figure 9d).

3.2.5. Chloride and chloride artifacts

Due to the extremely low concentrations of chloride in ambient air of these field studies (<0.1 µg m⁻³), it is difficult to make any firm conclusions on the quantification of ambient chloride in the SV vs. CV in this study. However, a reasonable linear correlation for chloride was still found between the two vaporizers (e.g., R = 0.6 in Boulder study). The deployment of the CV sampling ambient chloride should be further explored in areas with higher submicron chloride mass concentrations.

Extra chloride signals (maximum 0.28 µg m⁻³ and 0.1 µg m⁻³ in SV and CV, respectively) were observed when sampling NO₃-OFR output for both vaporizers compared to ambient chloride. The regression slope between the SV and CV in the NO₃-OFR (2.4) was systematically higher than from the corresponding ambient dataset (0.7, Figures 6e vs. 7e). No known source of chloride aerosol was present in the NO₃-OFR. We hypothesize that this enhanced chloride signal was caused by surface chemistry on the vaporizer. As Drewnick et al. (2015) and Hu et al. (2016b) have shown, chloride salts (e.g., NaCl or NH₄Cl, or HCl or other species formed from their decomposition) appear to be very sticky on the vaporizer surface and/or ionization chamber walls, and thus some chloride residue may always be present on the vaporizer. Thermal decomposition products from nitrate on the SV include gas-phase HNO₃(g), NO₂(g), or NO(g). HNO₃(g) may react with chloride on the vaporizer analogously to the atmospheric heterogenous reaction, e.g., NaCl(s) + HNO₃(g) → HCl(g) + NaNO₃(s) (Harkel 1997). The gas-phase HCl(g) would then lead to extra chloride ion signal. Since this possible artifact signal is quite small relative to nitrate, this effect would be most apparent when sampling high NH₄NO₃ mass concentrations (e.g., 10–60 µg m⁻³ in NO₃-OFR). Good linear correlations between nitrate and chloride signals in the SV and CV both support this possible mechanism (Figure 11).

Figure 11. Scatterplots between nitrate equivalent mass concentration (RIE = 1) of chloride and nitrate detected by AMS with the SV and CV measuring outflow of the NO₃-OFR during the SOAS study.

The linear regression slope in Figure 9 between chloride and nitrate (nitrate equivalent mass concentrations, both assuming RIE = 1) suggests the enhancement of chloride in the CV (0.2% chloride/nitrate) is substantially lower than in the SV (∼0.5% chloride/nitrate). The lower enhancement of chloride in the CV is probably because (1[1] ) the main thermal decomposition product of nitrate in the CV is gas-phase NO(g) with less gas-phase HNO₃(g) formed (Hu et al. 2016b); (2) potentially lower chloride residual in the CV because of shorter sampling history of the CV than the SV used in this study.

In addition to the SOAS OFR results, we have also observed increased chloride signals when sampling pure organic acid particles with both vaporizers (0.002–0.02% per nitrate mass, as shown in Figure S12). This indicates that artifact chloride signal might occur for acids other than HNO₃(g). However, we cannot rule out the possibility of chloride contamination in the organic-containing solution given its very small magnitude. The chloride enhancement was found to be negligible when sampling pure (NH₄)₂SO₄ particles in both vaporizers. AMS users should consider performing an RIE calibration for chloride, using ammonium chloride, before field and laboratory studies, although it may also result in an increased intensity of the interference discussed in this section.

3.3. Evaluation of composition-dependent CE

The SOAS NO₃-OFR dataset showed a large variation of NH₄NO₃ fraction in its output (1%–80%), which provided a good opportunity to examine the validity of the composition-dependent CE parametrization of Middlebrook et al. (2012). AMS/SMPS mass ratios as a function of NH₄NO₃ fraction in total aerosol are shown in Figure 12. A clear NH₄NO₃-fraction-dependent CE was observed in the SV, which is similar to the SV results of Middlebrook et al. (2012). However, AMS/SMPS mass ratios (∼1) in the CV did not show a dependence with varying NH₄NO₃ fraction, suggesting complete collection of ambient aerosol ().

Figure 12. Ratio of AMS to SMPS mass concentrations as a function of the measured ammonium nitrate mass fraction, for both the SV and CV. The CE parameterization of Middlebrook et al. (2012) is also shown. The error bars are the standard deviation of the averaged decile values.

To further explore this effect, a constant CE of 0.55 was applied to the SV-data (Figure S13). This value was determined as the average of the dynamic composition-dependent CE used in Figure S4. Systematically higher nitrate and ammonium concentrations (a factor of 1.5) are observed in the SV with constant CE compared to the CV (Figure S13). The high regression slopes were mainly biased by the overestimated concentration when NH₄NO₃ fractions were high. Ambient organic nitrates typically only dominate total nitrate at lower ambient nitrate concentration (<0.5 µg m⁻³; Day et al. in preparation). Thus, even if organic nitrates at high concentrations led to a similar increase of CE as NH₄NO₃ in Middlebrook et al. (2012) (which has not been shown to our knowledge), their low ambient fractions will generally not result in an increased CE. For sulfate species, most of the high sulfate mass concentrations occurred during time periods with low NH₄NO₃ fractions. Thus, the regression slope of sulfate with constant CE applied was similar to that with composition-dependent CE used. This comparison provides strong support for the application of the composition-dependent CE for accurate quantification with SV in areas with variable (and relatively high: >50%) NH₄NO₃ fractions.

3.4. AMS size distribution measurements

For monodisperse inorganic standard species, the CV showed additional broadening of the size distribution measurements, compared to the SV (Hu et al. 2016b; Xu et al. 2017). This is consistent with the delayed vaporization due to multiple bounces and longer thermal decomposition timescales within CV. In ambient air, aerosol size distributions are usually relatively broad (Seinfeld and Pandis 1998). The capability of the CV to characterize typical ambient accumulation-mode size distributions is examined in this section.

Particle time-of-flight (PToF) and size distributions of ambient total and inorganic species from both vaporizers during the Boulder study are shown in Figure 13. An apparent PToF shift (∼100 µs) in the CV was observed compared to that in the SV. This lag due to longer residence time in the CV is consistent with the PToF comparison from laboratory-generated standard species. The measured peak in the CV is up to 30% wider than in the SV (Table 4), consistent with the broadening effect observed for monodisperse particles of inorganic species (a factor of ∼1.8–5.5; Hu et al. 2016b).

Figure 13. (a) Measured particle time-of-flight distributions for ambient aerosols using the SV and CV in the Boulder study. (b) Comparison of the calibrated size distributions. The sizing with the CV was calibrated with pure ammonium sulfate particles. The relationships between particle velocity and size are shown in Figure 14.

Table 4. Summary of peak widths of ambient aerosol PToF and size distributions in the Boulder study. FWHM method was used to calculate peaks widths here. The corresponding PToF and size distributions are shown in Figure 13.

Species name	SV (ms)	CV (ms)	SV with NH4NO3 velocity cal. (nm)	CV with (NH4)2SO4 velocity cal. (nm)
Total	1.54	1.94	553	459
SO4	1.89	1.95	544	496
NO3	1.24	1.67	445	367
NH4	2.07	1.97	513	339

Calibrations of PToF size distributions based on monodisperse NH₄NO₃ and (NH₄)₂SO₄ particles were carried out for both the SV and CV. The PToF size calibration curves from both vaporizers are shown in Figure 14. In the SV, the PToF velocity calibration curve from NH₄NO₃ was consistent with that from (NH₄)₂SO₄, while the PToF size calibration curves from NH₄NO₃ and (NH₄)₂SO₄ in the CV were quite different because of the longer average residence time of (NH₄)₂SO₄ than NH₄NO₃. Most (NH₄)₂SO₄ signal in the SV is detected with limited broadening compared to NH₄NO₃ because the signal is dominated by the particles that do not bounce (Robinson et al. 2016). In contrast, (NH₄)₂SO₄ particles in the CV may have several opportunities to vaporize as it rattles around in the vaporizer cavity. This explanation is consistent with the large broadening of pure, monodisperse (NH₄)₂SO₄ size distributions (∼ a factor of 5.5 for 250 nm particles) shown in Hu et al. (2016b).

Figure 14. Particle velocity vs. vacuum aerodynamic diameter of ammonium nitrate and ammonium sulfate in the (a) SV and (b) CV. The gas velocity within the aerodynamic lens (Vl), calculated from the lens flow rate and pressure, and the gas velocity upon exiting the nozzle (Vg) estimated by the measured arrival time of airbeam signal, were constrained in the fitting. Other parameters were obtained by curve fitting (according to Jayne et al. 2000; Allan et al. 2003a).

By using the NH₄NO₃ size calibrations to calculate the CV size distributions of ambient aerosols, a large shift of about 250 nm was observed compared to the SV (Figure S14). When the calibration curve from (NH₄)₂SO₄ is applied, the updated size distributions from the CV is more consistent with the SV distributions (Figure 13b). This indicates that for typical ambient particles, mainly composed of OA and sulfate, most of the particles have vaporization/detection timescales closer to pure (NH₄)₂SO₄ in the CV. We note that the CV size calibrations were performed here using the peaks of the AMS PToF signals from monodisperse particles. For SV the recommended calibration method is to use the arrival time plus ½ of the chopper opening time, rather than the peak. However, the latter method is not desirable for the CV due to the larger and more variable broadening observed.

The good agreement of the updated size distribution between the SV and CV supports that CV-AMS particle sizing is still useful for ambient measurements. The peak width of the size distributions using a calibration from the (NH₄)₂SO₄ calibration curve (415 ± 74 nm) was close to that in the SV (513 ± 49 nm, Table 4). The slightly narrower peak width of total aerosol for the CV here is not necessarily indicative of better sizing resolution detection but rather due to uncertainties in the pure (NH₄)₂SO₄ sizing calibration and its application to ambient data, as (NH₄)₂SO₄ data show both a large delay as well as broadening of the peaks, which may be greater than for mixed-composition ambient particles. We recommend using (NH₄)₂SO₄ calibrations when analyzing data from sulfate/organic dominated ambient particles. Further systematic characterization of CV size distribution detection is recommended.

4. Conclusions

In this study, we have examined the performance of the AMS equipped with a CV in regards to chemical quantification, characterization, and sizing of ambient inorganic and total aerosols with four field datasets. AMSs with SV and CV installed were calibrated and sampled in parallel. A summary of the advantages and disadvantages of the SV and CV for AMS/ACSM measurements, based on the results in this study and the previous laboratory study (Hu et al. 2016b), is shown in Figure 15.

Figure 15. Summary of the advantages and disadvantages of SV and CV usage in AMS/ACSM measurement based on this field study and Hu et al. (2016b) laboratory study. Note that some effects (such as Pieber and chloride artifacts) need characterization in other instruments and studies.

The overall results showed good agreement of mass concentration of total and speciated inorganic composition detected between the SV and CV, and also SMPS or PILS-IC vs. CV-AMS (<16% deviation, R = 0.89–0.99 of HR comparisons for major inorganic species). For UMR datasets, good agreement and correlations of total and inorganic aerosols were also found (slopes are within 0.73–1.1, R > 0.65). The field studies indicate that CE ∼ 1 for ambient aerosol in the CV. The comparison results also support that the composition-dependent CE due to high nitrate fractions for ambient aerosol measurements when using an SV works well, and its robustness was demonstrated for a study with rapidly varying NH₄NO₃ fractions.

Compared to the SV, fragmentation patterns of nitrate and sulfate in the CV were altered due to additional thermal decomposition processes, with a shift in the ion distributions to smaller ions. The fragmentation pattern of ammonium showed little difference because the gas-phase NH₃(g) from NH₄NO₃ and (NH₄)₂SO₄ thermal decomposition was not influenced by additional thermal decomposition since it appears to be evaporated and ionized directly. We investigated the fragmentation patterns of organic vs. inorganic nitrate for both vaporizers. The differences in NO₂⁺/NO⁺ ratios between organic and inorganic species observed in the SV can still be observed in the CV. This indicates that the distinguishing chemical information from those ions is still present with the CV. However, with background ambient aerosol concentrations, the quantification of the variations of those ratios may sometimes be limited by the signal-to-noise level in the CV due to the much lower concentrations of the larger ions. H_ySO_x⁺/SO⁺ ratios between organic sulfone vs. inorganic (NH₄)₂SO₄ showed similar phenomenology as nitrate in the SV and CV.

In biogenic areas, UMR nitrate mass is overestimated when using default fragmentation table (∼factor of 2–3 in SOAS) for both vaporizers, due to underestimation of the organic interferences. Updates to the fragmentation tables for both vaporizers are proposed in this study. Since the CV might be installed extensively in the commercial ToF-ACSM and Q-ACSM with PM_2.5 lens, attention needs to be paid to the organic/air interferences in UMR inorganic quantification when inorganic concentration is lower compared to OA. A small artifact chloride signal can be observed when high NO₃ (or organic acid) mass concentrations are sampled into the AMS with the SV or CV, presumably due to the reaction of gas-phase HNO₃ released from thermal decomposition with residual chloride on the vaporizer. This effect is similar to the recently reported enhancement of CO₂⁺ signal when sampling nitrate into AMS due to particle–vaporizer interactions (Pieber et al. 2016). For the levels observed here, this effect will always be negligible and much smaller than the measurement uncertainties for AMS total mass. However, it could be a non-negligible and even dominant interference for the measured AMS chloride when the sampled air contains high nitrate and low chloride levels, as demonstrated with an NO₃-OFR dataset in this study. However, less chloride enhancement (0.2% per nitrate mass) was observed in the CV than in the SV (0.5% per nitrate mass), so in this case the CV simplifies quantification. Further study of this artifact is recommended.

Size distributions can be measured with the CV-AMS, with similar results to the SV-AMS. However, it is important to calibrate the size measurement with monodisperse particles. Using (NH₄)₂SO₄ for size calibrations when analyzing data from sulfate/organic dominated ambient particles is suggested. Further study of this mode and calibration with mixed particles is recommended.

Acknowledgments

We sincerely thank Hyungu Kang, Jason Schroder, Brett Palm, and Amber Ortega of CU-Boulder and Suzane S. de Sá of Harvard University for experimental support in specific datasets and Hongyu Guo and Rodney Weber from Georgia Tech for SOAS PILS-IC data support.

Funding

This study was partially supported by NSF AGS-1243354 and AGS-1360834, EPRI 10004734, NASA NNX15AT96G, DOE (BER/ASR/SBIR) DE-SC0011105, DE-SC0016559, and DE-SC0001673, and NOAA NA13OAR4310063.