Evaluation of a particle trap laser desorption mass spectrometer (PT-LDMS) for the quantification of sulfate aerosols

Abstract

A particle trap laser desorption mass spectrometer (PT-LDMS) has been developed for the online measurements of the chemical composition of submicron aerosol particles. The PT-LDMS was evaluated by both laboratory and ambient measurements, with the focus being the quantification of sulfate aerosols. Ammonium sulfate ((NH₄)₂SO₄) is generally the predominant form of sulfate aerosols in urban air; hence, it is used as a material for laboratory experiments and calibration. Major fragments of (NH₄)₂SO₄ were observed at mass-to-charge ratios (m/z) of 48 (SO⁺) and 64 (SO₂⁺). The dependence of sensitivity (expressed as the ratio of m/z 48 signal to sulfate mass) on laser power and cell temperature was investigated. An intercomparison of PT-LDMS with a commercial sulfate particle analyzer (SPA) and filter sampling was performed in Tokyo. Good agreement was observed between SPA and filter analysis (slope = 0.98, r² = 0.99). Although the mass concentration of sulfate measured by PT-LDMS exhibited a tight correlation with that measured by SPA, the mass concentration measured by PT-LDMS tended to underestimate that measured by SPA (slope = 0.70, r² = 0.96). While the discrepancy can be mainly attributed to the difference in size cut between PT-LDMS (approximately PM₁) and SPA (PM_2.5), differences in vaporization efficiency were also found to be important.

Copyright © 2016 American Association for Aerosol Research

EDITOR:

• Paul J. Ziemann

1. Introduction

Online measurements of aerosols are important for improving our understanding of the formation and loss processes of ambient aerosols. In particular, quantification of mass concentration of sulfate aerosols is of high priority because sulfate aerosols are typically the major components of fine particulate mass in the troposphere (Altshuller 1973; Drewnick et al. 2003; Hogrefe et al. 2004). Ammonium sulfate ((NH₄)₂SO₄) is a predominant form of sulfate aerosols found in urban areas and in the regions significantly affected by anthropogenic emissions (Saltzman et al. 1983; Pilinis and Seinfeld 1987; Schwab et al. 2004).

The instruments currently employed for online measurements of sulfate aerosols are roughly categorized into dry and wet methods. Dry methods are often based on thermal desorption techniques and include the Aerodyne aerosol mass spectrometer (AMS; Aerodyne Research Inc., USA) (Jayne et al. 2000), thermal desorption chemical ionization mass spectrometer (TDCIMS) (Voisin et al. 2003), and sulfate particulate analyzer (SPA; Thermo-Scientific, USA) (Schwab et al. 2006). Wet methods are based on online water extraction techniques followed by chemical analysis; these methods include a steam jet aerosol collector (SJAC) (Khlystov et al. 1995) and particle-into-liquid sampler (PILS) (Weber et al. 2001). These are often coupled with ion chromatography for the quantification of bulk compositions of aerosols.

Among these instruments, the Aerodyne AMS has been utilized by a number of researchers, and useful insights into the formation of sulfate aerosols have been reported (Zhang et al. 2005, 2007). Generally, the routine calibration of an AMS is performed using pure ammonium nitrate (NH₄NO₃) particles because of the high particle collection efficiency (CE) for NH₄NO₃. The quantification of sulfate aerosols in ambient air is performed by assuming representative values of collection efficiency and relative ionization efficiency (RIE) (Allan et al. 2003). As discussed below, the thermal decomposition products of (NH₄)₂SO₄ are known to significantly vary depending on the temperature of vaporizer. In fact, earlier AMS studies indicated that the detection efficiency of (NH₄)₂SO₄ exhibited significant dependence on vaporizer temperature (Alfarra 2004). For these reasons, there is a need to revisit the quantification of sulfate aerosols by thermal decomposition techniques and to reconsider the choice of appropriate materials for calibration.

We have developed a particle trap laser desorption mass spectrometer (PT-LDMS), and demonstrated the proof of concept (Takegawa et al. 2012). The PT-LDMS developed has been categorized as a dry method, as mentioned previously. This is the first report of quantification of mass concentrations of ambient sulfate aerosols by PT-LDMS. Laboratory experiments for determining the operating conditions of PT-LDMS and intercomparisons with other instruments, including an SPA and a filter sampling method, are presented. Possible uncertainties in quantification by PT-LDMS and SPA are discussed in detail.

2. Experimental

2.1. PT-LDMS

The hardware and data acquisition software have been used according to the previous version reported by Takegawa et al. (2012) with some modifications. Herein, we summarize the important specifications. Aerosol particles are introduced into a vacuum chamber through an aerodynamic lens (Liu et al. 1995). The majority of gas molecules contained in sample air can be removed by a skimmer cone, which results in a concentration of aerosol particles along the centerline of the lens (Allan et al. 2003). Aerosol particles are collected on a particle trap comprising layers of micro-structured mesh, and are subsequently vaporized by a carbon dioxide (CO₂) laser (ULR-10, Universal Laser System, Inc., USA). The vaporization mechanism included both direct heating of aerosol particles due to light absorption and thermal conduction from the heated surface of particle trap, as described in Takegawa et al. (2012). The particle trap is enclosed by a gold (Au)-coated aluminum cell (a quartz cell was used in the previous version), which is connected to a quadrupole mass spectrometer (QMS; gas-tight ionizer type, PrismaPlus, Pfeiffer Vacuum, GmbH, Germany). The vaporized gas from collected aerosol particles is ionized by electron impact ionization (EI) with an electron energy of 60 eV. Ion currents at some selected m/z values (I_m/z) are detected by a secondary electron multiplier (SEM) using the multiple ion detection mode of QMS.

Compared with the previous version presented by Takegawa et al. (2012), the coating material of the trap was changed from pure platinum (Pt) to a mixture of Pt and tungsten (W) to increase melting point. The separation distance between the mesh layers was changed from 200 to 250 μm to facilitate machining. A 20-μm thick Au film was inserted between the particle trap and holder to improve thermal conductivity. The trap holder was cooled by circulating water at 15°C. A small gap was introduced between the holder and the cell for improving thermal insulation. These modifications enable heating of cell and cooling of trap. The former is necessary to reduce the adsorption of the evolved gas, while the latter is needed to reduce the loss of particle compounds by evaporation during collection.

The laser power (P_laser) was continuously monitored using a beam splitter and thermopile sensor. A beam homogenizer (Sumitomo Electric, Japan) was installed in optical path for generating a nearly top-hat-shaped laser beam, although the shape of the actual beam at the position of the particle trap was not precisely measured. This is beneficial in terms of both homogeneous heating of the particle trap and avoiding penetration of trap by laser beam. A radiation thermometer (IGA140, LumaSense Technologies Inc., USA) was placed outside the chamber for monitoring the temperature of particle trap. The thermometer observed near-infrared radiation emitted from the trap heated by the laser passing through a quartz window of vacuum chamber. Note that the temperature data should be treated as a lower limit because both the emissivity of the trap and the transmission efficiency of the quartz window were assumed to be unity.

The total measurement cycle was of 15 min, including the time for particle collection, laser desorption analysis, cooling of the trap, and buffer time for stabilization. After aerosol particles in sample air were collected on the particle trap for 7 min, the inlet valve was closed (SS-43GF4 with MS-142ACX, Swagelok, Inc., USA). The laser was switched on for 1.5 min to detect the gas evolved from collected particles 1 min after the inlet valve had been closed. The inlet valve was kept closed for 5.5 min after the laser was switched off for cooling the trap.

The inlet flow rate of PT-LDMS was calibrated using a standard flow meter (DryCal DC-Lite Air Flow Calibrator, Brandt Instruments, Inc., USA). The mass concentration is expressed as μg m⁻³ at 25°C and 1 atm.

2.2. Other instruments

The SPA quantifies the mass concentration of sulfate aerosols by a thermal decomposition/reduction method coupled with a sulfur dioxide (SO₂) monitor (Schwab et al. 2006). Sulfate aerosol particles are reduced to SO₂ in a converter tube heated to 1000°C and subsequently detected by a pulsed UV fluorescence method. The detected signals include signals from sulfate aerosols and those from gas-phase SO₂. We did not use denuder to remove gas-phase interference. The mass concentration of sulfate was calculated by subtracting the data obtained for particle-free air (zero air) from the data obtained for particle-loaded air.

Filter sampling and analysis were performed by Murata Corp (http://www.murata-s.co.jp/). PM_2.5 samples were collected using a PM_2.5 aerosol sampler (MCAS-SJ, Murata Keisokuki Service Co. Ltd., Japan) and subsequently analyzed by ion chromatography in the laboratory. Miyakawa et al. (2015) have described the collection and analysis procedure in detail.

2.3. Laboratory experiment

The dependence of sensitivity on P_laser and cell temperature (T_cell) was investigated in the laboratory prior to ambient measurements. The particle generation and evaluation system was practically the same as that described by Takegawa et al. (2005). It comprised collision atomizer (Model 3076, TSI, Inc., USA), diffusion dryer, differential mobility analyzer (DMA; Model 3081, TSI, Inc., USA), and condensation particle counter (CPC, Model 3022, TSI, Inc., USA). Monodisperse aerosols with known chemical compositions were generated and introduced into PT-LDMS.

We used (NH₄)₂SO₄ particles for the laboratory experiment because they are typically the predominant form of sulfate aerosols in urban air. As described by Takegawa et al. (2012), major fragments of (NH₄)₂SO₄ were observed at m/z 48 (SO⁺) and 64 (SO₂⁺). A mobility diameter (d_m) of 250 nm was chosen for most of the experiments because of the high particle transmission efficiency through aerodynamic lens at this diameter (Miyakawa et al. 2014). The effect of multiple charged particles was estimated using the same method as that described by Takegawa and Sakurai (2011) to calculate the mass of sulfate introduced into PT-LDMS.

In addition to (NH₄)₂SO₄, potassium nitrate (KNO₃) was used as the reference because of its simple molecular structure and fragmentation pattern (Takegawa et al. 2012). Details of the experimental results for KNO₃ particles are given in the online supplementary information (SI). The quantification of nitrate aerosols is also an important issue, but it is beyond the scope of this article. A detailed investigation of nitrate aerosols, including KNO₃ and NH₄NO₃, is currently underway and will be presented elsewhere.

2.4. Ambient measurements

Ambient measurements were performed in November 2013 at a building on the campus of the Research Center for Advanced Science and Technology, the University of Tokyo. Although the intercomparison was also performed in March 2013, as qualitatively discussed by Miyakawa et al. (2014), their data were not utilized in this study. Figure 1 shows the experimental setup for ambient measurements. Ambient air aspirated from a window was introduced into a buffer box so that all instruments could take samples from the same air volume. The inner surface of the buffer box was coated with aluminum foils and grounded to avoid electrostatic charges. The filter samples were collected every 12 h. The SPA and PT-LDMS were equipped with a PM_2.5 cyclone (URG-2000-30EH, URG, Inc., USA), while the size cut of PT-LDMS was approximated as PM₁ (limited by the transmission efficiency of aerodynamic lens).

Figure 1. Schematic diagram of the experimental set up for ambient measurements.

The zero levels of the PT-LDMS signals were measured every hour by replacing the sample air with zero air generated using a high-efficiency particulate air (HEPA) filter. The PT-LDMS was routinely calibrated using monodisperse (NH₄)₂SO₄ particles during the measurement period by the particle generation system described in Section 2.3.

The zero levels of SPA signals were measured every hour using a three-way valve and a HEPA filter placed upstream of the converter. The valve was switched between particle-loaded and zero air. Although the SO₂ monitor of SPA was calibrated using SO₂ standard gas prior to the measurement period, the overall calibration of SPA was performed using monodisperse (NH₄)₂SO₄ particles after the period. The three-way valve was not used for the correction of gas-phase SO₂ during calibration because substantial amount of SO₂ outgas from DMA was present, presumably originating from HEPA filters for generating sheath air. Instead, the contribution of gas-phase SO₂ was estimated by turning off the high voltage of DMA.

2.5. Data reduction

2.5.1. PT-LDMS

As described in Section 2.1, ion currents at some selected m/z values were detected (m/z 14, 16, 28, 30, 44, 48, 55, and 64). The m/z 14 signal was dominated by N⁺ originating from the nitrogen present in ambient air with the inlet valve open, and was used to correct the temporal drift of ion detection efficiency. A similar correction process was also carried out by Takegawa et al. (2012), who used m/z 32 (O₂⁺) instead of m/z 14. Although this section focuses on the quantification of sulfate aerosols, the equations presented here are also applicable to other compounds such as nitrate.

Table 1 summarizes the key parameters used in this study and their abbreviations. The ion current above the background level was accumulated over the laser irradiation time to obtain an integrated ion signal at each m/z . Note that the superscript X (= “amb” or “cal”) represents ambient or calibration conditions, respectively. The mass concentration of sulfate was calculated using the difference in the integrated ion signal between particle-loaded and zero air :[1]

Table 1. Key factors affecting quantification

Instrument	Parameter	Definition	Remark
PT-LDMS
	Im/z	Ion current at m/z	a
	M^XMS	Mass concentration measured by the PT-LDMS	a
	ΔQ^Xm/z	Net integrated ion signals at m/z	a
	ΔQ^Xm/z,particle	Integrated signal at m/z for particle-loaded air	a
	ΔQ^Xm/z,zero	Integrated signal at m/z for particle-free air (zero air)	a
	Sm/z	Sensitivity determined by calibrations	a
	Tr^X	Transmission efficiency of the aerodynamic lens
	CE^X	Collection efficiency of particles
	Y^X	Vaporization efficiency of particle compounds
	α^X	Adsorption fraction of SO2 on the cell surface
	IE^X	Ionization efficiency for evolved gas
	f^Xm/z	Fragment ratio at m/z
SPA
	M^XSPA	Mass concentration measured by the SPA	a
	Δχ^X	Net mixing ratio of SO2 measured by the SO2 monitor	a
	χ^Xparticle	Mixing ratio of SO2 for particle-loaded air	a
	χ^Xzero	Mixing ratio of SO2 for zero air	a
	FSPA	Calibration factor of the SPA
	ϵ^X	Conversion efficiency of sulfate to SO2
	χ^Xgas	Mixing ratio of gas-phase SO2 in sampling air
Others
	M^calCPS	Mass concentration of calibration particles quantified using the CPC	a
	V	Sampled air volume	a
	dm	Mobility diameter classified by the DMA	a
	ρ	Density of the calibration material	a
	NCPC	Number concentration measured by the CPC	a
	fcorr	Correction coefficient for the multiple-charge effect	a
	M^Xtrue	“True” mass concentration of aerosol particles

The superscript X (= “amb” or “cal”) denotes ambient or calibration conditions, respectively.

The subscripts represent the species, instruments, or mode of measurements.

^aThese values were explicitly determined by calibrations or ambient measurements.

where V is the sampled air volume, and S_m/z is the sensitivity of sulfate aerosols at the corresponding m/z value. is expressed as follows:[2] where is the integrated ion signal at m/z for particle-loaded air and is the integrated ion signal at m/z for zero air. Using monodisperse aerosol particles generated and quantified by the DMA and CPC system, the average mass concentration of sulfate aerosols introduced into PT-LDMS during a calibration cycle can be calculated as follows:[3] where d_m is the mobility diameter of calibration particles classified by DMA, ρ is the particle density, N_CPC is the number concentration of particles measured by CPC, f_corr is the correction coefficient for the multiple-charge effect estimated by a method similar to that of Wang et al. (2010) (see the SI), and R_SO4 is the molecular fraction of SO₄²⁻ in (NH₄)₂SO₄. The S_m/z values are then given as follows:[4]

These equations are also used to estimate possible uncertainties in quantification (Section 4.3).

2.5.2. SPA

The mass concentration of sulfate aerosols measured by SPA was calculated using the difference of the mixing ratio of SO₂ between particle-loaded and zero air (net SO₂ mixing ratio, Δχ^X):[5]

Δχ^X is expressed as follows:[6] where is the mixing ratio of SO₂ for particle-loaded air, and is the mixing ratio of SO₂ for zero air. F_SPA is the correction factor determined by calibration:[7]

Notably, both and were scaled to the same standard, which is critical for the interpretation of intercomparison results.

3. Results

3.1. Laboratory experiment

The radiation temperature of the particle trap was primarily controlled by the P_laser density and thermal properties of the trap (i.e., absorbance and thermal conductance). Figure 2 shows an example of P_laser and radiation temperature profiles at a certain laser power setting. The P_laser was modulated (overshoot at the beginning) to achieve a rapid evolution of radiation temperature. The P_laser and trap temperature (T_trap) were estimated using data points between 30 s and 90 s after the laser was switched on (i.e., “plateau time period”).

Figure 2. Example of T_trap (solid line, left axis) and P_laser (dashed line, right axis) profiles at T_cell of (a) 280°C and (b) 200°C.

We measured sensitivity under various P_laser (T_trap) and T_cell conditions. Figure 3 shows the evolution of the m/z 48 ion currents for (NH₄)₂SO₄ particles obtained at a P_laser of ∼4 W (T_trap of ∼550°C) and T_cell values of 200°C and 280°C. The accumulated mass of sulfate was 18 and 8 ng for T_cell = 200°C and 280°C conditions, respectively. The ion current for T_cell = 200°C was scaled down by a factor of 0.44 (= 8/18) to clarify the comparison of two signals. The ion current signal was characterized by a sharp peak immediately after the laser was switched on and a broad peak of 10 s after the laser was switched on. This result suggests that the particle vaporization process comprised at least two thermal desorption mechanisms, as mentioned earlier. Notable changes were observed for the ion current at m/z 48 for (NH₄)₂SO₄ in the latter period of laser irradiation (30–90 s) with increasing T_cell.

Figure 3. Evolution of mass normalized m/z 48 ion current for (NH₄)₂SO₄ particles obtained at a T_trap of ∼550°C and T_cell of 200°C (blue and turquoise lines) and 280°C (red and orange lines). The accumulated mass of sulfate was 18 and 8 ng for T_cell = 200°C and 280°C conditions, respectively. The ion current for T_cell = 200°C was scaled down by a factor of 0.44 (= 8/18) to clarify the comparison of two signals.

Figure 4a shows the relationship between T_trap and P_laser under three T_cell conditions. The relationship did not show significant dependence on the T_cell conditions. Figure 4b summarizes the results of temperature experiments. Although substantial scatter was observed in the data, the dependence of sensitivity on T_cell appeared to be more significant than that on P_laser (T_trap) for (NH₄)₂SO₄. Higher T_cell conditions tended to exhibit higher sensitivity for (NH₄)₂SO₄ particles. As mentioned in Section 2.3, major fragments of (NH₄)₂SO₄ were observed at m/z 48 and 64, but detectable amounts of other sulfur-related signals, such as m/z 80 (SO₃⁺) and 98 (H₂SO₄⁺), were not observed. The to ratio can be used as a diagnostic for desorption because this ratio can vary depending on the chemical form of sulfur compounds (SO₂, H₂SO₄, or others).

Figure 4. (a) Scatter plot of T_trap versus P_laser from laboratory experiments. (b) Dependence of sulfate sensitivity on P_laser at various T_cell values. T_cell values for the experiment were selected as 280°C (filled circle), 250°C (triangle), and 200°C (open circle). Dashed lines denote the regression lines of each dataset in (b). Error bars depict the uncertainty originating from the random error of PT-LDMS (4%), CPC (2%), radiation thermometer (1%), and thermopile sensor (1%).

Figure 5 shows the scatter plot of versus obtained during laboratory experiments. Previous studies have reported that the thermal decomposition products of (NH₄)₂SO₄ mainly comprise SO₂ above ∼400°C, and that the decomposition products vary below ∼400°C (Kiyoura and Urano 1970). According to the impact ionization (70 eV) mass spectrum database provided by the National Institute of Standard Technology (NIST) (http://webbook.nist.gov/chemistry/), the to ratio can differ significantly depending on the chemical form of sulfate compounds (SO₂, H₂SO₄, etc.). The observed value from (NH₄)₂SO₄ was ∼1.4 with a high correlation coefficient (r² ∼ 0.99). Although we did not directly identify the chemical form of sulfur compounds in the evolved gas, the tight correlation suggests that the thermal decomposition products of (NH₄)₂SO₄ can be regarded as uniform.

Figure 5. Correlation plots of versus from laboratory experiments (filled circle: T_cell = 280°C, open circle: T_cell = 250°C, and shaded circle: T_cell = 200°C). The data points for T_cell = 250°C include routine calibrations during ambient measurements. The dashed line denotes the regression line for entire data (r² = 0.998).

3.2. Ambient measurements

Figure 6 shows the time series of the average ion current at m/z 14 (reference signal) with the inlet open and the sensitivities for sulfate and nitrate during the laboratory experiment (October 2013) and ambient measurement (November 2013). The sensitivity of nitrate was determined using KNO₃ as a calibration material (Section 2.3). The reference signal changed in accordance with the changes in SEM voltage. The sensitivity for sulfate exhibited a large decrease over the course of few days following the start of ambient measurements, while the sensitivity of nitrate exhibited a slight decreasing trend. The sensitivity recovered after the replacement of particle trap; however, it exhibited a similar decreasing trend again. Such a drastic change was not observed during the laboratory experiment or ambient measurement in March 2013 (Miyakawa et al. 2014). A scanning electron microscopy observation indicated that the Pt/W coating was partially detached from the particle trap surface (not shown), which could have caused decrease in sulfate sensitivity. The particle trap was replaced when the sensitivity of sulfate was decreased to ∼200 pC ng⁻¹, although optimal timing of the replacement of particle trap is still under consideration.

Figure 6. Temporal changes in (a) the average of m/z 14 ion current, and (b) sensitivity of sulfate (filled circle) and nitrate (open square) during laboratory experiments and ambient measurements. The particle trap was replaced on November 9. The sensitivity data include laboratory experiments and routine calibrations. There was an overlap of ambient measurement and laboratory experiment periods because some laboratory experiments were performed during the ambient measurement period.

Figure 7 shows the scatter plot of versus obtained during ambient measurements. For comparison, the regression line from the laboratory experiment is also shown. The values for ambient aerosol particles showed good agreement with the laboratory data with a variation of 15%, indicating that the thermal desorption products from ambient sulfate aerosol particles were very similar to those from (NH₄)₂SO₄ particles generated in the laboratory.

Figure 7. Correlation plots of versus . The dashed line denotes the regression line shown in Figure 5.

Figure 8a shows the time series of and during ambient measurement period. The zero air signals were found to exhibit a positive correlation with ambient signals, suggesting that the zero air signals originated from residual sulfate compounds, which may have been incompletely vaporized by each laser shot. Figure 8b shows the time series of , , and during the measurement period. Note that is the ambient mass concentration measured by filter analysis. The values were calculated using the data shown in Figure 8a and by interpolating the sensitivity data shown in Figure 6. The values were calculated using Equations (5)−(7), and the calibration was performed by the same method of PT-LDMS. During the period of high sulfate mass concentrations (November 17–19), the PT-LDMS data tended to be lower than the data obtained from SPA and filter analysis. On the other hand, the difference in the mass concentration of sulfate between PT-LDMS and SPA/filter analysis during the low mass concentration period (November 14 and 20) was not as large as the difference during the high sulfate period.

Figure 8. (a) Time series of and . The filled and open circles denote the particle-loaded air signals and zero air signals, respectively. The dashed line represents the zero level. (b) Time series of (solid line), (shade line), and (circle marker). Error bars of depict the average time of filter sampling (12 h). Missing data points of PT-LDMS (e.g., November 12) were due to calibrations.

Figure 9 shows the scatter plot of , , and and that of and . Good agreement was observed between and , with a regression slope of 0.98 and an r² value of 0.99. Although exhibited a close correlation with (r² = 0.96) and (r² = 0.95), the PT-LDMS tended to underestimate filter analysis (regression slope = 0.67) and SPA (regression slope = 0.70). It should be noted that the regression slope was largely affected by the data points in the higher mass concentration range. When focusing on the lower mass concentration range, the data points became closer to the 1:1 correspondence line.

Figure 9. Correlation plots of (a) and versus analysis, and (b) versus . The shaded and solid lines in (a) represent the regression lines of versus (r² = 0.95) and versus (r² = 0.99), respectively. The shaded points in (b) represent all data, while open circles represent averages in the bin of 1 μg m⁻³. Solid line in (b) represents the regression line for averaged values (r² = 0.96). Dashed lines in (a) and (b) denote 1:1 lines.

4. Discussion

4.1. Decrease in the sensitivity of sulfate

The observed decrease in sulfate sensitivity during ambient measurements is discussed. Decrease in sulfate sensitivity was not observed during the laboratory experiments but became evident after the start of ambient measurements (Figure 6). The damage of particle trap may have been caused by the presence of refractory compounds in ambient aerosols that efficiently absorbed CO₂ laser. This may have resulted in local and extreme heating of particle trap and partial detachment of coating. Further investigation and improvement are needed to increase the durability of trap.

4.2. Limit of detection

The limit of detection (LOD) for sulfate aerosols by PT-LDMS is now evaluated. The LOD can be estimated as the equivalent concentration at which the difference in the peak height of ion currents between particle-loaded and zero air is twice the standard deviation (2σ) of background level. The LOD value estimated by this method was 0.07 μg m⁻³, on average. Alternatively, the upper limit of LOD can be estimated by calculating variability in mass concentration in ambient air at sufficiently low mass concentrations (November 12–13 in Figure 8). The average and standard deviation of mass concentration during this period were 0.5 and 0.04 μg m⁻³, respectively, which yields an upper limit of 0.08 μg m^–3 for LOD.

4.3. Possible uncertainties

4.3.1. PT-LDMS

Possible systematic uncertainties in the determination of sulfate mass concentrations by PT-LDMS are discussed based on theoretical expressions. Simply, we assume that was proportional to . can be theoretically expressed as follows:[8] where A is a constant for conversion from mass to ion signals, Tr^X is the transmission efficiency of aerodynamic lens, CE^X is the collection efficiency of particles, Y^X is the vaporization efficiency of particle compounds, α^X is the adsorption efficiency of SO₂ on cell, IE^X is the ionization efficiency of the evolved gas, and is the fragment ratio at m/z. is the “true” mass concentration of sulfate aerosols, which is unknown in principle. Note that each parameter in Equation (8) is not explicitly determined by calibration.

Based on Equations (1)–(4) and (8), the relative error in the mass concentration of sulfate aerosols can be expressed as follows:[9]

The uncertainty in the mass concentration of sulfate determined by PT-LDMS depends on the difference in the following factors between ambient and calibration particles: transmission efficiency of aerodynamic lens (Tr^amb/Tr^cal), particle collection efficiency (CE^amb/CE^cal), vaporization efficiency of particles (Y^amb/Y^cal), adsorption efficiency of evolved gas ((1 − α^amb)/(1 − α^cal)), and ionization efficiency and fragmentation of evolved gas (IE^amb/IE^cal and fr^amb/fr^cal, respectively). The uncertainty also depends on the accuracy of DMA sizing, CPC counting, and multiple charge correction, which can be cancelled out when taking into account the comparison between PT-LDMS and SPA. It is noted that these factors are also important for quantification using an Aerodyne AMS.

Table 2 summarizes the estimates of systematic errors. The size cut of PT-LDMS was approximated as PM₁ and limited by the transmission efficiency of aerodynamic lens, which is essentially the same as those used in Aerodyne AMS (Liu et al. 1995; Takegawa et al. 2009; Miyakawa et al. 2014). The transmission efficiency of particles with d_m = 250 nm can be regarded as unity, but the transmission efficiency for multiple charged particles included in the DMA exit (d_m = 412, 569 nm) may be less than unity. The overall Tr^cal was estimated to be ∼0.97 using the transmission efficiency curve given by Miyakawa et al. (2014). The overall Tr^amb was estimated to be ∼0.85, assuming a typical size distribution of aerosol particles in Tokyo (see the SI). This yields an estimate of ∼0.88 for Tr^amb/Tr^cal in the case of PM_2.5 ambient particles. Note that Tr^amb and Tr^cal values are strongly dependent on the accuracy of transmission efficiency of aerodynamic lens and the assumption of size distribution in ambient air.

Table 2. Possible error of each instrument.

Instruments	Factor		Relative error
PT-LDMS	Size cut^a	Ambient	−0.15
		Calibration	+0.03
	Vaporization^b	Ambient	−0.13
		Calibration	+0.03
	Contribution of refractory compounds^c		−0.06
Calibration^d	DMA sizing		±0.01
	CPC detection efficiency		−0.05
	Correction for multiply charged particles		±0.04

^aThe size cut error was estimated assuming a typical size distribution of aerosols in Tokyo (unpublished data, see the SI).

^bThe vaporization error was estimated based on the temporal variation of (see text for details).

^cThe refractory compound error was estimated using Na⁺, Cl⁻, and K⁺ from filter analysis and previous studies (Takegawa et al. 2005; Schwab et al. 2006). This error represents refractory sulfate compounds that are detectable by SPA but not by PT-LDMS (Na₂SO₄, K₂SO₄).

^dThe calibration error can be cancelled out when taking into account the comparison between PT-LDMS and SPA.

The collection efficiency of ambient aerosol particles is also difficult to quantify because it depends on the size distribution and physical properties of aerosol particles. Compared with liquid particles, solid particles, such as dry (NH₄)₂SO₄ particles, generated in the laboratory tend to exhibit a broader distribution on particle trap (Takegawa et al. 2012). Polydispersed, irregularly shaped particles, such as soot aggregate particles, may also exhibit a broader distribution than monodisperse, spherical particles because of the characteristics of aerodynamic lens (Zhang et al. 2004; Huffman et al. 2005; Salcedo et al. 2007). The degree of overlap between laser and particle beams can affect the temporal evolution of ion signals (Takegawa et al. 2012). A possible approach for estimating uncertainty is to compare the temporal evolution of ion signals at m/z 48 between ambient and calibration conditions, as shown in Figure 10. The evolution profiles were found to be similar, suggesting that the distribution of collected particles did not significantly vary between ambient and calibration conditions, and that CE^amb/CE^cal can be assumed to be close to unity. This assumption does not mean that both CE^amb and CE^cal were unity but that these were comparable during the measurement period.

Figure 10. The evolution of mass normalized m/z 48 ion current for calibration (blue and turquoise lines) and ambient (red and orange lines) conditions. The accumulated mass of sulfate was 10.2 and 5.8 ng for calibration and ambient conditions, respectively. The ion current for ambient condition was scaled up by a factor of 1.76 (= 10.2/5.8) to clarify the comparison of two signals.

The Y^amb/Y^cal ratio is affected by two factors: incomplete vaporization of (NH₄)₂SO₄ particles and contribution of refractory sulfate compounds, which were unlikely vaporized with their vaporization temperature exceeding T_trap (Na₂SO₄ and K₂SO₄). The ratio of zero air to particle-loaded air signals was ∼0.13 for ambient particles and ∼0.03 for calibration particles. As a result, Y^amb/Y^cal originating from incomplete vaporization of (NH₄)₂SO₄ was estimated to be ∼0.90.

Good agreement was observed for between ambient and calibration particles, suggesting that the major component of evolved gas in the case of ambient particles was very similar to calibration particles (Figures 5 and 7). Hence, we can assume that IE^amb/IE^cal, , and (1 − α^amb)/(1 − α^cal) were close to unity.

The positive and negative signs of systematic errors in Table 2 show that the mass concentration of sulfate in ambient air can be overestimated and underestimated, respectively. Errors for ambient and calibration conditions were partially cancelled out because of the similarities in the chemical form and size of sulfate aerosol particles between ambient and calibration conditions. Conversely, it is suggested that the choice of material and size of calibration particles is important for reducing measurement uncertainties.

4.3.2. SPA

Possible systematic uncertainties in the determination of sulfate mass concentrations by SPA are discussed. As mentioned in Section 2.4, was measured by using a HEPA filter, and was measured by turning off the DMA voltage. is expressed as follows:[10] where B is a constant for conversion from the mass concentration of sulfate to the mixing ratio of SO₂, ϵ^X is the conversion efficiency of sulfate to SO₂, and is the mixing ratio of gas-phase SO₂ in sample air. As for calibration, should be equal to because there is no difference in the airflow line between particle-loaded and zero air (the zero level was measured by turning off DMA voltage). On the other hand, and may be different due to the adsorption of SO₂ in HEPA filter, considering large surface area inside the filter. It is rather difficult to estimate the value because the amount of SO₂ adsorption on the filter may not correlate with the SO₂ mixing ratio in the sample air (Eatough et al. 1995). Although may also have been affected by the adsorption of SO₂ on other surfaces, including conversion cell and tubing, we assumed that this effect was relatively minor.

Based on Equations (5)−(7) and (10), the mass concentration of sulfate aerosols determined by SPA can be expressed as follows:[11]

Uncertainty in the mass concentration of sulfate determined by SPA depends on two terms. The first term originates from the conversion of sulfate to SO₂ and depends on conversion efficiency (ϵ^amb/ϵ^cal). The second term originates from SO₂ in the ambient air and depends on the mixing ratio of ambient SO₂.

The ϵ^amb/ϵ^cal ratio can be affected by the contribution of refractory sulfate compounds, which were unlikely vaporized with their vaporization temperature exceeding converter tube temperature of SPA (e.g., CaSO₄).

As mentioned above, the uncertainty originating from gas-phase SO₂ in ambient air was difficult to estimate because was not determined from measurements. If we assume an error of to be 1 ppbv, for example, it corresponds to an error of 4 μg m⁻³ in the mass concentration of sulfate aerosols. This uncertainty could potentially become important with increasing mixing ratios of SO₂.

4.4. Comparison between PT-LDMS and SPA

For comparison between PT-LDMS and SPA, it is necessary to consider difference in sulfate compounds measured by these instruments. The temperature of particle trap induced by CO₂ laser was ∼500°C, while the temperature of the converter of SPA was set at 1000°C. Among various refractory sulfate compounds, Na₂SO₄ (dissociation temperature of 800°C) was likely to be the most relevant form of compound, which was detectable by SPA but not by PT-LDMS. Schwab et al. (2006) reported that responses of SPA to Na₂SO₄ and K₂SO₄ were ∼20% and ∼60%, respectively. Based on their study, the contributions of Na₂SO₄ and K₂SO₄ were estimated to be ∼2% and ∼5%, respectively. Note that the contribution of Na₂SO₄ was estimated using Na⁺ and Cl⁻ from the filter data assuming that Na⁺ originated from both sodium chloride (NaCl) and Na₂SO₄ (Takegawa et al. 2005). The contribution of K₂SO₄ was estimated, assuming that all K⁺ originated from K₂SO₄.

The combined uncertainty of the factors described in the previous section and the above factor was estimated to be ∼20%, which may potentially explain the underestimation by PT-LDMS to a large extent. Further studies are needed to improve the quantification of sulfate aerosol under various environments.

5. Summary

This study reports the quantification of sulfate aerosols by PT-LDMS. The instrument was modified compared with the previous version described by Takegawa et al. (2012). Laboratory experiments were performed to investigate optimal conditions for ambient measurements and temperature dependence of sulfate sensitivity. The major decomposition product of (NH₄)₂SO₄ was similar to that of ambient sulfate in PTLDMS. The results also showed that the sensitivity of (NH₄)₂SO₄ particles tended to increase with T_cell, which may be explained by the decreased adsorption of the decomposition product of (NH₄)₂SO₄ on cell surface at high temperatures.

Ambient measurements for the overall test of the instrument, as well as the intercomparison of PT-LDMS with other measurement methods, were performed in Tokyo. for ambient aerosol particles showed good agreement with the laboratory data, to be within 15%, indicating that the thermal desorption products from ambient sulfate aerosol particles were very similar to those from (NH₄)₂SO₄ particles generated in the laboratory. The PT-LDMS exhibited tight correlation with SPA (r² = 0.96) but tended to underestimate the mass concentration of sulfate (slope = 0.70).

Major uncertainties in PT-LDMS during the measurement period originated from the size cut of aerodynamic lens (∼12%) and the difference in the vaporization efficiency of (NH₄)₂SO₄ particles between calibrations and ambient conditions (∼10%). In addition, uncertainty in the contribution of refractory sulfur compounds between PT-LDMS and SPA was estimated to be ∼6% in this case. These uncertainties were estimated by the consideration of numerous sources of systematic error and used for the interpretation of intercomparison results. The overall uncertainty due to above factors was estimated to be ∼20%, which may potentially explain the underestimation of PT-LDMS to a large extent. Further studies using polydisperse aerosol particles and other materials for calibrations are needed to improve the accuracy of quantification.

Acknowledgments

The authors thank Professor M. Koike of the University of Tokyo for useful comments.

Funding

This study was funded by the SENTAN program of the Japan Science and Technology Agency (JST) and the Grant-in-Aid for Scientific Research of the Japan Society for the Promotion of Science (25281004).