Consistency and Traceability of Black Carbon Measurements Made by Laser-Induced Incandescence, Thermal-Optical Transmittance, and Filter-Based Photo-Absorption Techniques

Abstract

In this study, we show that black carbon (BC) mass concentrations measured by different techniques are consistent and traceable. First, we present the volatilities of 13 organic compounds passed through a heated inlet. These data were used to quantify the interference of organic aerosols on the BC measurement techniques. The masses of the refractory particles that incandesce (m*_ref) were used to calibrate BC mass measured by a single-particle soot photometer (SP2), which uses laser-induced incandescence. This calibration was influenced little by refractory organics and agreed well with that of fullerene soot, which indicates the consistency of the standards. We estimated the interference of pyrolyzed refractory organics on the BC measured with a filter-based absorption photometer continuous soot monitoring system (COSMOS) with a heated inlet to be small in Asia. This was also confirmed by the stable mass absorption cross section (MAC) obtained by the high correlations between BC mass concentrations measured by COSMOS (M _COSMOS) and those measured by the thermal-optical transmittance method (M _TOT) (Kondo et al. 2009). M _COSMOS was also compared with total BC mass concentrations measured with an SP2 (M _SP2) in Tokyo in 2009. M _COSMOS and M _SP2 were highly correlated (r ²= 0.97) and agreed to within about 10% on average. These results demonstrate that M _SP2, M _COSMOS, and M _TOT were nearly identical. Use of the masses of incandescing refractory BC particles for calibration of BC mass concentrations determined by different techniques gave consistent results.

1. INTRODUCTION

Black carbon (BC) aerosols have been extensively studied because they efficiently absorb solar visible radiation and contribute to the radiative forcing of the atmosphere on regional and global scales (IPCC 2007; Ramanathan et al. 2007). Elevated BC concentrations in source regions have deleterious impacts on human health (Lighty et al. 2000; McCreanor et al. 2007). Reliable measurements of BC mass concentrations are a prerequisite for improved evaluation of these effects. Various techniques have been used to measure BC mass concentrations in the field (Schmid et al. 2001; Hitzenberger et al. 2006; Slowik et al. 2007; Kanaya et al. 2008). Assessment of possible systematic differences between the measurements obtained by the various techniques is necessary, particularly with regard to their uncertainties and traceability to appropriate standards. To achieve this goal, we need to establish methods for absolute and systematic calibration of BC mass concentrations.

FIG. 1 Experimental setup for the measurement of the size distributions of organic aerosols with or without the heated inlet. Total numbers of particles were measured by a CPC after size selection by a DMA.

In recent years, a technique has been developed to measure the masses of individual BC particles by means of laser-induced incandescence (LII) (Gao et al. 2007; Schwarz et al. 2006; Schwarz et al. 2008; Moteki and Kondo 2007, 2008, 2010; Moteki et al. 2009). The measurement of BC by means of a single-particle soot photometer (SP2), which uses LII, has been shown to be unperturbed by internal mixing of BC. The technique is not influenced by the coexistence of other types of aerosols, including organic aerosols (OA). Therefore, we used this technique as a basis for measuring BC mass concentrations. We used ambient BC mass to calibrate our SP2 mass measurement.

The main focus of this article is to show the consistency of the different measurement techniques of ambient BC mass concentrations. The techniques used are the LII technique, the refractory mass method (RMM; Kondo et al. 2006), the thermal-optical transmittance (TOT) method (Birch and Cary 1996; Lim and Turpin 2002; Kondo et al. 2006), and the filter-based photo-absorption technique, specifically the continuous soot monitoring system (COSMOS) (Miyazaki et al. 2008; Kondo et al. 2009). The BC measurement techniques are summarized in Table 1. The symbols and acronyms used in this article are summarized in Table 2.

TABLE 1 Summary of the measurement techniques for BC mass concentration used in this study

Acronym	Detection method	Calibration	Mass
SP2	Incandescence	APM + ambient BC or fullerene soot	M SP2
RMM	Mass	APM	M ref
TOT	Conversion to CO2	CH4/He standard gas	M TOT
COSMOS	Light absorption	TOT or SP2 + ambient BC	M COSMOS
RMM: Refractory Mass Method; APM is calibrated with standard PSL particles.

TABLE 2 Summary of the variable symbols and acronyms used in this study

Terminology	Definition
M ref	Mass of refractory components in ambient aerosol extracted by a heated inlet (single particle)
m*ref	Mass of refractory components in ambient aerosol that incandesce and are extracted by a heated inlet (single particle)
m full	Mass of fullerene soot (single particle)
M SP2	Mass of ambient BC detected by SP2 (single particle)
M ref	Total mass concentration of refractory particles in the fine mode
M SP2	Total mass concentration of BC measured by SP2 in the fine mode
M TOT	Total mass concentration of BC measured by the TOT method in the fine mode
M COSMOS	Total mass concentration of BC measured by the COSMOS in the fine mode
D m	Mobility diameter of refractory aerosol
D me	Mass equivalent diameter of refractory aerosol
D BC	Mass equivalent diameter of BC
b abs	Absorption coefficient measured by COSMOS
C abs(D BC, λ)	Absorption cross section of an airborne BC particle with diameter D BC at wavelength λ
C emit(D BC, λ)	Emission cross section of an airborne BC particle with diameter D BC at wavelength λ
N	Complex refractive index of BC
σ*abs	Mass absorption cross section of an airborne BC particle (λ= 565 nm for this study)
σabs	Mass absorption cross section of a BC particle collected on a filter (λ= 565 nm for this study)
f fil	Correction factor of the magnitude of BC photo-absorption by multiple scattering in the filter medium
CMD	Count median diameter
MMD	Mass median diameter
APM	Aerosol particle mass analyzer
DMA	Differential mobility analyzer
SMPS	Scanning mobility particle sizer
CPC	Condensation particle counter
SP2	Single-particle soot photometer
LII	Laser-induced incandescence
COSMOS	Continuous soot monitoring system
TOT method	Thermal-optical transmittance method

The physical and chemical properties of BC and coexisting aerosols influence the uncertainties of the BC measurements and the degree of the interferences strongly depends on measurement principle. First, the variabilities in the microphysical properties of BC, including size, morphology, and mixing state, can cause changes to the response to BC instruments (Kondo et al. 2009; Nakayama et al. 2010; Moteki et al. 2010b).

Secondly, the coexistence of aerosols other than BC can cause significant uncertainties in the BC measurements, depending on their chemical properties. OA in particular is known to cause significant uncertainties, namely, uncertainties in the separation of BC and organic carbon (OC) in the TOT method (Boparai et al. 2008), enhancement of absorption (lens effect) (absorption photometers except for COSMOS) (Subramanian et al. 2007; Cappa et al. 2008; Lack et al. 2008), and charring of OA by heating the sample air (COSMOS).

In this study, we conducted detailed laboratory studies of the characteristics of vaporization and pyrolyzation of OA. Based on these data, we have assessed the effect of OA on ambient BC measurements. We show that BC mass concentrations measured by these techniques are consistent and traceable, as indicated by new measurements and our previously published results. The published results include detailed comparisons of the BC measurements by the TOT and filter-based photo-absorption techniques at six locations in Asia (urban and suburban Tokyo in Japan, Jeju Island in Korea, rural Guangzhou and Beijing in China, and Bangkok in Thailand). Here we investigate the relationships between the results obtained from the different techniques from the perspective of our current understanding. The effects of microphysical properties on these measurements are also taken into account.

TABLE 3 Ratios of the volumes of organic particles downstream of the heated (400°C) and unheated inlets (heated/unheated) for different mobility diameters

			100 nm		150 nm		200 nm
Classification	Compound	MW	300°C	400°C	300°C	400°C	300°C	400°C
Saccharides	levoglucosan	162	0.002	0.001	0.002	0.009	0.003	0.003
Saccharides	sucrose	342	0.027	0.003	0.097	0.004	0.061	0.005
Saccharides	glucose	180	0.002	0.002	0.002	0.001	0.002	0.002
Dicarboxylic acid	malonic acid	104	0.016	0.011	0.015	0.011	0.018	0.011
Dicarboxylic acid	succinic acid	118	0.007	0.004	0.008	0.004	0.010	0.005
Dicarboxylic acid	adipic acid	146	0.048	0.042	0.053	0.048	0.060	0.055
Dicarboxylic acid	pimelic acid	160	0.035	0.029	0.032	0.026	0.035	0.027
Dicarboxylic acid	azelaic acid	188	0.007	0.004	0.010	0.006	0.014	0.009
Aromatic acids	phthalic acid	166	0.015	0.011	0.017	0.012	0.008	0.010
Humic-like^a	SRHA II	NA	0.87	0.72	0.93	0.74	0.87	0.75
Humic-like^a	SRFA I	NA	0.72	0.43	0.74	0.44	0.75	0.48
Humic-like^a	SRFA II	NA	0.82	0.62	0.85	0.66	0.91	0.76
Humic-like^a	PPFA	NA	0.87	0.76	0.90	0.78	0.91	0.80
^aProduced by the International Humic Substances Society and desalted (H^+-saturated) by cation exchange.

2. HEATED INLET

We start with the description of a heated inlet, because it was used for the RMM and COSMOS measurements and for the calibration of an SP2, as discussed in section 3. The heated inlet was made of a stainless steel tube with a 3/8-inch outer diameter and 0.049 inches thick. A 21-cm section of the inlet line was heated at 400°C, and the residence time of particles in this section was about 0.3 s for a sample flow rate of 0.7 STP liter min^–1 (Kondo et al. 2009). Changes in the mass concentration of aerosols with diameters less than about 1 μm (PM₁) were measured using an Aerodyne aerosol mass spectrometer (AMS) (Kondo et al. 2009). Almost all the mass of inorganic compounds and more than 90% of the mass of organic compounds detected by the AMS evaporated in the heated inlet. However, we need to characterize the volatilities of different organic species through the heated inlet more quantitatively for the interpretation of the BC data obtained by the different techniques.

For this purpose, we measured the volatilities of 13 different organic compounds using the experimental setup shown in Figure 1. The temperature of the heated inlet was set to be 300° and 400°C. We tested three saccharides: levoglucosan (C₆H₁₀O₅), glucose (C₆H₁₂O₆), and sucrose (C₁₂H₂₂O₁₁); five diacids: malonic acid (HOOC(CH₂)(COOH)), succinic acid ((HOOC(C₂H₂)₂COOH), adipic acid (HOOC(CH₂)₄COOH), pimelic acid (HOOC(CH₂)₅COOH), and azelaic acid (HOOC(CH₂)₇COOH); one aromatic acid: phthalic acid HOOC(C₆H₄)(COOH)); and four humic-like substances (HULIS): Suwannee River Humic Acid II (SRHA II), and Suwannee River Fulvic Acid I (SRFA I), Suwannee River Fulvic Acid II (SRFA II), and Pahokee Peat Fulvic Acid (PPFA), as summarized in Table 3, together with their molecular weights (MW). The HULIS were from the International Humic Substances Society.

FIG. 2 Changes in the number size distributions of organic aerosols with changes in the temperature of the inlet (300° and 400°C) for organic species.

The organic compounds were dissolved in water without significant changes in water temperature to avoid possible changes in chemical characteristics. Aerosol particles generated by a nebulizer setup were dried using a diffusion dryer and then passed through a conductive elastic tube to a DMA (Model 3081, TSI Inc., MN, USA) for mobility-size selection. The flow of the nearly mono-modal organic particles was split into three flows (Figure 1). One flow was used to measure the total aerosol particle concentration in the sample air with a condensation particle counter (CPC; Model 3022, TSI Inc., MN, USA). The size of organic particles selected by the DMA ranged from 100 to 200 nm, depending on the organic species.

The number size distributions of organic particles pre-selected by the first DMA were measured with a wide range particle spectrometer (WPS; Model 1000 XP, MSP Corporation, MN, USA) operated with a flow rate of about 1 liter min^–1 (Figure 1). The size range measured was between 10 and 500 nm. The number concentrations of the particles measured by the DMA + CPC and WPS with the unheated inlet agreed to within 5%.

Figure 2 shows size distributions as a function of mobility diameter with and without the inlet heated to 300° and 400°C for 12 organic compounds for particles with a mobility diameter of 150 nm, pre-selected by the DMA. Because the WPS system was equipped with its own charger for its internal DMA (second DMA), doubly charged particles by the first DMA were measured as particles larger than 150 nm by the WPS, as notably seen for some species, including sucrose and levoglucosan. The size distributions shown in Figure 2 are corrected for the combined effect of charging by the first and second DMA systems.

The remaining volume fractions downstream of the heated inlet are summarized in Table 3. At 400°C, the diameters of levoglucosan, adipic acid, pimelic acid, azelaic acid, and phthalic acid decreased by factors of about 4 to 10. More than 99% of their initial volumes were lost by the heating. In contrast, the changes in the sizes were much smaller for the HULIS, as shown in Figure 2. Correspondingly the changes in their volumes were as low as about 20–60%. At 300°C, the changes in the volumes were about 10–30%. The observed low volatilities of the HULIS and the high volatilities of the other low-MW organics are generally consistent with those observed by the thermograms of these compounds collected on filters (Miyazaki et al. 2007; Boparai et al. 2008). We also discuss this point in detail in section 3.2.

FIG. 3 Experimental setup for the calibration of the SP2 using ambient BC and commercially available BC particles. The particles were mass selected by means of an APM.

3. BC MEASUREMENT TECHNIQUES

3.1. SP2 Method

We used an SP2 (Schwarz et al. 2006, 2008; Gao et al. 2007; Moteki and Kondo 2007, 2008, 2010; Shiraiwa et al. 2008) to measure the BC mass size distribution. Each BC particle sampled by the SP2 was irradiated by a laser beam at a wavelength of 1064 nm and thus heated to incandescence. The SP2 monitors LII signals in two distinct visible bands (λ= 300–550 nm and 580–710 nm). In addition, laser light scattered by individual particles was detected by two avalanche photodiodes to measure the sizes and number concentrations of light scattering particles.

BC particles can be identified by the ratio of LII intensities of the two visible bands, because the ratio is a proxy for the vaporization temperature of the particles. The vaporization temperatures of ambient BC were measured to be about 4000 K (Schwarz et al. (2006). The BC mass for each particle (m _SP2) was estimated from the LII peak intensity. We measured the LII peak intensities of monodisperse BC aerosols that were mass selected by an aerosol particle mass analyzer (APM; Model-302, KANOMAX, Inc., Japan) with a 400°C heated inlet (Moteki and Kondo 2007). Figure 3 shows the experimental setup for the measurements of the ambient BC and other types of BC including fullerene soot particles. The BC particles, except for ambient BC, were aerosolized from a water suspension with a pneumatic nebulizer. The diameters of the coated BC particles (shell diameters, D _p) were derived using the scattering signals for D _p. A detailed description of the method to derive D _p is given in the Appendix.

The dependence of the LII intensity on particle microphysical properties was formulated by Moteki and Kondo (2010) and is summarized in the Appendix. The relationships between the peak LII intensity and the mass of BC particles were measured for fullerene soot (Alfa Aesar, USA), colloidal graphite (Alfa Aesar), Aquadag (Acheson, USA), glassy carbon (Tokai Carbon, Japan), glassy carbon (Alfa Aesar, USA), Aqua-Black 001 (Tokai Carbon, Japan), Aqua-Black 162 (Tokai Carbon), and ambient BC in Tokyo (Moteki and Kondo 2010). The relationships were very similar for fullerene soot and ambient BC.

The density of fullerene soot was measured using a helium pycnometer (Gas Pycnometer: Model Accupyc 1340, Micromeritics, GA, USA). The volume of fullerene soot is measured by replacing its void space with helium. The density is determined with independent and simultaneous measurements of its mass, which was about 0.2 g. The measurements were made three times. Each measurement consisted of 10 measurement cycles, providing 10 data points. The average density measured in the three data sets was 1.718 ± 0.004 g cm^–3. The estimate of the mass equivalent diameter of BC (D _BC) is based on this value.

BC in ambient air was extracted by vaporization of the condensed compounds coating the BC particles by means of the heated inlet mentioned in section 2. The SP2 calibration using ambient BC was made for several hours on November 10, 2009. The purity of the extracted BC estimated from the scattering signals (Appendix) was very high, as discussed by Moteki et al. (2010a). While the heated-inlet-APM output includes both BC and possibly HULIS, the Gaussian mean mass of the output should be the same for all particle types, if they are externally mixed. Since only incandescent, temperature-ratio-filtered BC particles are detected by the SP2 and used for the SP2

calibration, HULIS externally mixed with BC does not affect this calibration.

Now we evaluate errors due to HULIS internally mixed with BC. The median shell/core ratio (D _p/D _BC) of the extracted BC has been estimated to be 1.00–1.03 for D _BC= 160–240 nm, indicating that the volume fraction of the coating materials was less than 9%. About 20–60% of the volume of the 4 kinds of HULIS was lost through the heated inlet (section 2. The density of the HULIS is 1.5 g cm^–3 (section 3.4. Assuming these values as typical for refractory organics and assuming the density of BC to be 1.7 g cm^–3, the error in extracting refractory BC mass is estimated to be less than about 10% (9 × (1.5/1.7) × 0.8 = 6%).

We also make another estimate. The median value of the shell/core ratio (D _p/D _BC) of ambient BC in Tokyo was about 1.1 in August and September 2009, as shown in section 4. This means that the ratios of the volume of coating materials to that of BC were about 0.3. It is very unlikely that all the coating compounds were refractory organic compounds, considering that the coating thickness of BC was correlated with the mass concentrations of PM₁ inorganics and organics (Shiraiwa et al. 2007). PM₁ organic mass fractions in Tokyo were about 0.4 in summer and winter (Kondo et al. 2010). The mass concentrations of the sum of PM₁ aerosols (non-refractory aerosols + BC) agreed to within 10% with the independent gravimetric measurement by a Tapered Element Oscillating Microbalance (TEOM; Rupprecht and Patashnick Co., Albany, NY) instrument in summer (Kondo et al. 2007), indicating that less than 25% (= 10%/0.4) of the mass of organics was refractory. Therefore, the influence of refractory organic compounds on the extracted BC mass should have been less than about 10% (= 25 × 0.3 × (1.5/1.7) × 0.8 = 5%), assuming a density of HULIS of 1.5 g cm^–3.

This extraction procedure means that the definition of the mass of ambient BC is the mass of incandescent aerosols refractory at 400°C (m*_ref). The m*_ref data for ambient BC were not interfered by other incandescing aerosols like metal particles, because the ratios of the LII in the two visible bands were always similar for laboratory BC particles. The uncertainty of m*_ref is about 10%.

Figure 4 shows plots of the LII peak intensity versus m*_ref and the mass of fullerene soot (m _full) measured by Moteki and Kondo (2010). We selected only ambient BC and fullerene soot among all the other types of BC for this plot in order to demonstrate clearly their similarity in the LII peak intensity–m*_ref

relationships. The values of m _full and m*_ref ranged up to 350 femtograms (fg) and 230 fg, respectively. The LII peak intensity increased linearly with mass up to about 20 fg. At higher mass, the relationships deviated from linearity. These results agree with the theoretically predicted results, as discussed above.

We attributed the similarity of the LII response of fullerene soot and ambient BC to the similarity in their microphysical properties, namely their refractive indices and particle shapes (Moteki et al. 2010a; Moteki and Kondo 2010). We used the LII peak intensity–m*_ref relationship for ambient BC to derive the BC size distributions throughout the present study. It should be stressed here that M _SP2 calibrated by m _full is nearly identical to M _SP2 calibrated by m*_ref. In this way, M _SP2 is fully traceable to the calibration standards (m*_ref). The D _BC detection range of the SP2 used for this study was 70–920 nm.

3.2. Refractory Mass Method

The size distribution of the mass of the PM₁ refractory particles (m _ref) was measured with a scanning mobility particle sizer (SMPS; Model 3034, TSI Inc., MN, USA) combined with a 400°C heated inlet on the campus of the Research Center for Advanced Science and Technology (RCAST), the University of Tokyo, Tokyo, in the summer of 2004. Ambient air was introduced into a cyclone with a 50% effective cut-off aerodynamic diameter of 1 μm (PM₁ cyclone) to exclude coarse particles. The observation site was located about 2 km from highways with large traffic volumes. BC at this location is strongly influenced by emissions from heavy-duty diesel vehicles (Kondo et al. 2006).

The SMPS is composed of a differential mobility analyzer (DMA; Model 3081, TSI Inc., MN, USA) and a CPC. The mobility diameter (D _m) of the SMPS was converted to mass equivalent diameter (D _me) with an APM. The DMA provides a slice of aerosol with a reduced range of D _m from the ambient distribution. The APM then identifies the single mass per particle that best represents the mass distribution of the slice. The actual mass per particle varies widely with the change in D _m, and this constitutes the M _ref–D _m and D _me–D _m relationships. The details of the method used to calibrate the DMA–APM system are given in Kondo et al. (2006).

Mono-modal size distributions were fitted to the data of mass size distribution of m _ref(dm _ref/dlogD _me) for D _m< 500 nm (D _me< 300 nm). The mono-modal mass size distribution of m _ref was integrated to obtain the total BC mass concentration in this mode (m _ref). This procedure excluded dust, carbonate, and sea salt with D _me> 300 nm from M _ref. In fact, Na⁺ constituted a very minor (about 1%) fraction of PM₁ aerosol mass (Takegawa et al. 2005), showing that M _ref was influenced little by sea salt particles.

The absolute accuracy of the APM depends on the accuracy of the estimate of the masses of polystyrene latex (PSL) particles used for calibration. The uncertainty in the masses of submicron PSL particles is about 5%, according to the manufacturer of the PSL. The accuracies of M _ref and M _ref are therefore about 5%. It should be noted here that M _ref and M _ref were measured gravimetrically and do not depend on the density of BC assumed to derive the D _me–D _m relationship.

FIG. 4 Relationship between the LII peak intensity measured by the SP2 and the BC masses of ambient BC and fullerene soot.

FIG. 5 Thermograms of the HULIS obtained in the laboratory.

3.3. TOT Method

The total mass of PM₁ BC was also measured by a semi-continuous elemental carbon (EC)-organic carbon (OC) analyzer (Model RT3052, Sunset Lab., Oregon). These measurements were conducted simultaneously with the M _ref measurements at RCAST. The BC mass concentrations (M _TOT) were derived by the TOT method on the basis of the National Institute for Occupation Safety and Health protocol (Kondo et al. 2006; Han et al. 2009). The accuracy of total carbon measurements is about 5%. The overall accuracy of the M _TOT measurement was estimated to be 22% from the uncertainties associated with the sensitivity calibration, aerosol sampling, and temperature protocol.

The volatilities of organic compounds can be measured from thermograms of TOT. We relate the volatility data obtained from the TOT and those measured with the heated inlet in section 2 to assess the uncertainty of the COSMOS measurement. In the oven of the TOT, the OC collected on filters evolved in He gas during the four heating steps of 300°C (Step-1: 75 s duration), 450°C (Step-2: 60 s), 600°C (Step-3: 60 s), and 870°C or 750°C (Step-4: 150 s) were defined as OC1, OC2, OC3, and OC4. The OC released at 870°C in He/O₂ mixtures (Step-5: 240 s) was defined as OC5. OC4 and OC5 agreed to within 10% for the different temperatures during Step-4, and we use the results for 870°C for discussion. During Step-5, BC was oxidized in the oven. Namely, OC5 represents the fraction of OC that is resistant to volatilization in the He mode (including carbon that undergoes charring) and requires the same conditions as BC evolution in the thermogram (Huntzicker et al. 1982; Cadle et al. 1980).

For atmospheric sampling, OC5 cannot be estimated in the same way as described above, because BC coexists with OC. Instead, we define pyrolyzed carbon (PC) as the amount of carbon released between the start of Step-5 and the time when the laser transmittance through the filter recovered to the initial value. During this time, we measured the thermograms of SRHA II, SRFA I, SRFA II, and PPFA, as shown in Figure 5. The transmittance of TOT largely decreased during Steps-3 and 4 and recovered during Step-5, indicating that these species were pyrolyzed mainly during Steps-3 and 4. The fractions of OC1-OC5 are summarized in Table 4, together with the values of levoglucosan, pimelic acid, and azelaic acid, shown in Miyazaki et al. (2007). The thermograms of some other low-MW organics were also given by Miyazaki et al. (2007).

There are some differences in the OC5 fraction for SRFA I measured by Miyazaki et al. (2006) and the present work (36% versus 54%). The causes of this difference are not understood, although some difference in the samples used could be a possibility. We used the present value in estimating the performance of COSMOS because the same sample was used for this purpose. It should be noted here that the OC5 fractions (F _OC5) exceeded 40% only for the HULIS, while the F _OC5 of the other organic species were less than 6%. These high F _OC5 values of the HULIS correspond quite well to their distinct low volatilities, discussed in section 2.

3.4. COSMOS Method

3.4.1. Measurement Principle

Concentrations of ambient BC have also been measured by means of photo-absorption photometry. This technique requires conversion of the measured absorption coefficient (b _abs(λ)) at wavelength λ to BC mass concentration assuming a certain mass absorption cross section (σ_abs). However, coating of BC with relatively volatile compounds and the coexistence of light-scattering particles can contribute greatly to variation in σ_abs (Lack et al. 2008; Cappa et al. 2008; Kondo et al. 2009). This difficulty is overcome by removal of almost all the mass of volatile aerosol components before the BC particles are collected on filters. The volatiles are removed by means of a 400°C heated inlet operated with a residence time of about 0.3 s, as discussed in section 2.

TABLE 4 Summary of thermograms of organic species standards. The data for levoglucosan, pimelic acid, and azelaic acid were taken from Miyazaki et al. (2007)

	Percentage of integrated peak area
Compound	OC1	OC2	OC3	OC4	OC5
Levoglucosan	62.2	17.8	8.1	8.9	3.0
Pimelic acid	43.2	6.4	22.3	25.3	4.8
Azelaic acid	1.8	5.9	29.3	57.6	5.4
SRHA II	0.6	9.3	11.6	25.5	53.0
SRFA I	1.3	10.4	16.6	18.1	53.6
SRFA II	1.3	10.7	15.8	24.6	47.6
PPFA	0.3	9.9	10.1	17.1	62.6

For the present study, we measured mass concentrations of BC with a filter-based absorption photometer COSMOS. Details of the basic performance of the COSMOS instrument have been reported elsewhere (Miyazaki et al. 2008; Kondo et al. 2009). In brief, the instrument measures the absorption coefficient (b ₀) at a wavelength of 565 nm from the change in transmission through a filter loaded with BC particles after removal of volatile compounds. Operationally, b ₀ is determined with the COSMOS system by the following equation,

where A is the area of the sample spot, V _s is the air sample volume during a given time period Δt (between t–Δt and t), and I _t–Δt and I_t are the average transmittances. The flow rate of the COSMOS instrument used to measure b ₀ was set to 0.7 liters min^–1 at STP. Then b ₀ was corrected to compensate for the effect of multiple scattering (Bond et al. 1999); the corrected value, b _abs, is given by

where f _fil represents the correction of the magnification of absorption by multiple scattering in the filter medium.

By means of laboratory experiments using size-selected nigrosin (C₄₈N₉H₅₁) particles, we observed that the amplification of BC photo-absorption by multiple scattering increases with the decrease in D _BC (Kondo et al. 2009; Nakayama et al. 2010). Large BC particles do not penetrate as deeply into the filter fibers as do smaller particles, and large particles are situated in locations where the radiation field is influenced by less-frequent multiple scattering. The decrease in the frequency of multiple scattering with the increase in D _BC decreases the photo-absorption by BC, resulting in a need of smaller correction for multiple scattering. This size dependence was also confirmed quantitatively by theoretical calculations for spherical photo-absorbing particles on the basis of a radiative transfer model (Moteki et al. 2010b). Here we used the following equation for f _fil based on Nakayama et al. (2010)

where Tr (=I_t /I _t=0) is the filter transmission and B= 1.397 instead of 1.22 (Bond et al. 1999), as discussed in detail by Nakayama et al. (2010). The correction factor E(MMD) is a function of the mass median diameter (MMD), decreasing with increasing MMD, and represents the size-dependent effect of multiple scattering. Here we set E= 1.554, corresponding to the parameters for a BC size distribution of MMD = 180 nm and geometric standard deviation (σ_gm)= 1.6.

One-minute averaged b _abs was converted to BC mass concentration (M _COSMOS) using σ_abs:

FIG. 6 Experimental setup for the measurement of the absorption coefficient of organic aerosols by means of two COSMOS instruments. COSMOS I was mounted downstream of a heater, and COSMOS II was operated without a heater. Total numbers of particles were measured by a CPC after size selection by a DMA.

FIG. 7 Plots of equivalent mass concentrations of BC measured by COSMOS with an inlet heated to 300°C and 400°C and without a heated inlet versus the volume concentrations of organic species. The equivalent mass concentrations of BC were derived from the observed absorption coefficients by assuming a mass absorption cross section of 5.4 m² g^–1.

Values of b _abs were measured simultaneously with M _TOT at six different sites in Asia. The sampled BC particles were fresh and aged and emitted by combustion of fossil fuel and biomass, depending on the locations and seasons. At these locations, b _abs was highly correlated (r ²= 0.85–0.93) with M _TOT, with a 7% variability in the slope of the b _abs–M _TOT correlation, which is the mass absorption cross section, denoted as σ_abs(TOT) (Kondo et al. 2009). Even in air strongly impacted by burning of biomass (mainly rice straw) in Bangkok, σ_abs(TOT) did not deviate more than 10% from the average value of 5.9 m² g^–1. Kondo et al. (2009) derived σ_abs(TOT) = 10.4 m² g^–1 using B= 1.22 and E= 1.0. The set of parameters B= 1.397, E= 1.554, and σ_abs(TOT) = 5.9 m² g^–1 gives the same BC mass concentration for the same value of b ₀. This is because only the ratio f _fil/σ_abs affects M _COSMOS, and the σ_abs was determined according to the change in f _fil so that f _fil/σ_abs is unchanged.

TABLE 5 Slopes of correlation plots of equivalent BC mass concentrations versus volume concentration measured by COSMOS I (heated inlet) and COSMOS II (without heated inlet)

	Slope (μg m^–3)/(10^–6 cm^3 m^–3)				Net increase in slope
Compound	Unheated	r^2	300°C	r^2	400°C	r^2	300°C	400°C
PSL	0.029	0.12	NA	0.021	0.26
levoglucosan	0.0002	0.03	0.0000	0.06	0.0000	0.00
sucrose	0.033	0.86	0.0000	0.001	0.0000	0.00
glucose	0.0095	0.68	0.0000	0.12	0.0000	0.00
malonic acid	0.0025	0.29	0.0000	0.003	NA	NA
succinic acid	0.0000	0.00	0.0000	0.05	NA	NA
adipic acid	0.015	0.10	0.0000	0.08	0.0000	0.00
pimelic acid	0.0000	0.00	0.0000	0.00	0.0000	0.00
azelaic acid	0.0000	0.00	0.0000	0.07	0.0000	0.00
phthalic acid	0.026	0.90	0.0000	0.00	0.0000	0.00
SRHA II	0.021	0.16	0.024	0.06	0.069	0.007	0.003	0.048
SRFA I	0.012	0.60	0.014	0.74	0.039	0.71	0.002	0.027
SRFA II	0.0092	0.001	0.014	0.32	0.073	0.78	0.005	0.064
PPFA	0.025	0.28	0.030	0.83	0.089	0.91	0.005	0.064

3.4.2. Laboratory Experiments

Some fraction of the organic compounds collected on a quartz-fiber filter are known to be transformed into photo-absorbing carbon compounds at temperatures above 300°–400°C in oxygen-free and oxygen-rich environments, and this transformation causes an uncertainty in the value of M _TOT (Novakov and Corrigan 1995; Schauer et al. 2003; Kirchstetter et al. 2004), depending on the detailed procedures during the heating. The errors in separating charred refractory organics from BC are considered to be the major cause of the uncertainties of TOT methods (Yu et al. 2002; Yang and Yu 2002; Hitzenberger et al. 2006; Chow et al. 2007; Boparai et al. 2008; Hadley et al. 2008). If this process occurs in the present heating system for airborne particles, it may cause corresponding uncertainties in the BC values measured with the COSMOS instrument, although the mechanism of the interference is very different from that of TOT. We have evaluated this effect using the experimental setup shown in Figure 6.

The experiments were carried out for the same organic species as discussed in section 2. For reference, PSL particles were also used. Aerosol particles generated by a nebulizer setup were dried using a diffusion dryer and then passed through a conductive elastic tube to a DMA for mobility-size selection. The flow of the nearly mono-modal organic particles was routed to a CPC, COSMOS I (with a heated inlet), and COSMOS II (without a heated inlet). The size of organic particles selected by the DMA ranged mainly from 100 to 200 nm, depending on the organic species.

The b ₀(λ) measured by COSMOS I and COSMOS II were converted to equivalent BC mass concentrations in μg m^–3 to quantify interpretation of the results. We used σ_abs= 5.9 m² g^–1 for the conversion. Figure 7 shows the correlations between equivalent BC mass concentrations and the volume concentrations of the generated organic aerosols (sucrose, levoglucosan, SRHA II, SRFA I, SRFA II, and PPFA) that were directed into the two COSMOS instruments. The volume concentrations (× 10^–6 cm³ m^–3) were estimated from the diameters and number concentrations of the organic aerosols. Table 5 lists the slopes of correlation plots of equivalent M _COSMOS versus volume concentrations.

Although COSMOS II showed significant response to sucrose, COSMOS I showed no detectable response. The COSMOS I slopes of low-MW organics, including sucrose, levoglucosan, glucose, adipic acid, and phthalic acid, did not exceed 0.0001 μg m^–3/10^–6 cm³ m^–3 at either 300°C or 400°C, consistent with their high volatilities through the heated inlet. The HULIS showed significant absorption for both the unheated and heated inlet. The increased absorption with the inlet heated at 400°C suggests that some fraction of the HULIS pyrolyzed, leading to larger σ_abs. The reduction in particle size can also lead to enhanced absorption by COSMOS, as discussed in the previous section. However, this effect is considered to be small in this case, considering that the responses of SRHA II and SRFA to the COSMOS with the unheated inlet did not show significant differences for different mobility diameters between 100 and 200 nm.

For a quantitative analysis, we relate the response of the COSMOS to the volatility shown in the thermograms of the organic compounds. The OC4 fractions of pimelic acid and azelaic acid were as high as 25% and 58%, respectively (Table 4). However, the F _OC5 values were as low as 6%. These organic species caused no detectable absorption in the COSMOS with the heated inlet. During Step-4, carbon is released by the evaporation of pyrolyzed carbon or by thermally dissociation of organic compounds. Some fraction of BC also evaporates when the Step-4 temperature is 870°C (Subramanian et al. 2006).

These results indicate that only the carbon released during Step-5 is relevant to the absorption to COSMOS. This carbon was pyrolyzed mainly during Steps-3 and 4, as discussed in section 3.3. This relation is used to assess the interference of refractory organics on M _COSMOS. First, we derive the sensitivities of the equivalent M _COSMOS to the HULIS mass. The slopes of the equivalent M _COSMOS–HULIS volume correlation at room temperature, 300°C, and 400°C are summarized in Table 5, together with the net increase in the slopes due to heating. The densities of SRFA and PPHA were measured to be about 1.5 g cm^–3 (Dinar et al. 2006). We derived the ratios of the equivalent M _COSMOS to the total HULIS mass (R ₄₀₀ (total mass)) from the slopes of the equivalent M _COSMOS–volume concentration and the HULIS densities. The slopes ranged from 0.039 (SRFA I) to 0.089 (PPFA) μg m^–3/(10^–6 cm³ m^–3) at the heated inlet temperature of 400°C. The R ₄₀₀ (total mass) values are calculated to be 0.025 (SRFA I)–0.059 (PPFA) μg/μg.

The mass fractions of carbon in these compounds were measured to be about 0.50 (Dinar et al. 2006). Based on this value, the ratios of the equivalent M _COSMOS to the HULIS carbon mass (R ₄₀₀(total-C)) are 2 ×R ₄₀₀ (total mass) and are between 0.050 and 0.12 μg/μg. Namely, 5–12% of the carbon mass of these species was measured as equivalent M _COSMOS, assuming that the absorption was entirely due to the charred HULIS.

At 300°C, the ratios (R ₃₀₀(total-C)) were 2–4%, with a R ₃₀₀ (total-C)/R ₄₀₀ (total-C) ratio of about 1/3. The net increases in the absorption at 300°C were even lower (about 10 times) than those at 400°C. These results indicate that dominant fractions of the HULIS passed through the inlet without having been pyrolyzed at 400°C and that the fractions increased at 300°C.

It is useful to derive the [equivalent M _COSMOS]/[OC5] ratios using F _OC5. They are given as R ₄₀₀ (OC5) =R ₄₀₀ (total-C)/F _OC5 and R ₃₀₀ (OC5) =R ₃₀₀ (total-C)/F _OC5, and their values are summarized in Table 6. The average R ₄₀₀ (OC5) and R ₃₀₀ (OC5) values were 0.17 and 0.05.

TABLE 6 R ₃₀₀ (OC5) and R ₄₀₀ (OC5) of HULIS

Compound	R 300 (OC5)	R 400 (OC5)
SRHA II	0.060	0.174
SRFA I	0.035	0.097
SRFA II	0.039	0.204
PPFA	0.064	0.190
Average	0.050	0.166

TABLE 7 PC/BC ratios at different locations in Asia

			Relative interference (%)
Locations	Season/Airmass	Average PC/BC	300°C	400°C
Tokyo	Winter	0.063 ± 0.053	0.4 ± 0.3	1.0 ± 0.8
	Summer	0.047 ± 0.057	0.3 ± 0.3	0.7 ± 0.9
Jeju Island, Korea	Spring
	Average	0.04 ± 0.03	0.2 ± 0.2	0.6 ± 0.5
	Chinese	0.03 ± 0.02	0.2 ± 0.1	0.5 ± 0.3
	Marine	0.08 ± 0.05	0.4 ± 0.3	1.3 ± 0.08
Rural Guangzhou, China	Summer	0.28 ± 0.35	1.4 ± 2.1	4.6 ± 5.3
Rural Beijing, China	Summer	0.23 ± 0.22	1.1 ± 1.3	3.8 ± 3.6
Suburban Bangkok,	Summer (rainy season)	0.08 ± 0.27	0.4 ± 0.2	1.3 ± 4.4
Thailand	April (intense BB)	0.24 ± 0.37	1.2 ± 2.2	3.9 ± 6.1

The insensitivity of COSMOS to the low-MW species that produce high OC1-OC4 signals is due to their high volatilities in the heated inlet under high O₂ concentration (Table 3). The R ₄₀₀ (OC5) value as low as 0.17 is mainly due to the short residence time (<1 s) in the heated inlet. This is in contrast to the much longer time (>60 s) OC is exposed at 740°–870°C in He (Step-4) in the oven of the TOT instrument. Even at 300°C in the COSMOS inlet, more than 94% of the volumes of low-MW organic species evaporated (Table 3). However, only 10–30% of the volumes of the HULIS were lost through the heated inlet at this temperature, leading to a R ₃₀₀(OC5)/R ₄₀₀ (OC5) ratio of about 1/3.

3.4.3. Estimate of Interference of Refractory Organics on M _COSMOS

It is possible to estimate the effect of charring of refractory organics in the heated inlet using R ₃₀₀ (OC5) and R ₄₀₀ (OC5). The use of this R ₄₀₀ (OC5) includes an assumption that the σ_abs of charred refractory organics is the same as that of the HULIS used for the laboratory experiment. We obtained thermograms of the TOT instrument at the five locations in Asia, adopting the same temperature protocols as used for the laboratory experiments. The average PC/M _TOT ratios were derived from the thermograms, as summarized in Table 7. We consider PC to be equivalent to OC5, as discussed in section 3.3. The average PC/M _TOT ratios were as low as 0.04-0.08 in Tokyo, Jeju Island, and Bangkok during non-biomass burning periods, and they were as high as 0.23–0.28 in rural Guangzhou and Beijing and also in Bangkok during biomass burning periods. Simultaneous COSMOS/PSAP measurements were made with an inlet heated at 400°C. The estimated interference due to refractory organics is as low as 0.5–4.2%, as summarized in Table 7. If the temperature is lowered to 300°C, the interference will be less than 1.7%.

M _COSMOS and M _TOT were highly (r ²> 0.92) correlated, and the derived σ_abs(TOT) was stable to within 7% at these sites (Kondo et al. 2009). M _TOT is directly influenced by the uncertainties in the estimate of PC, while only about 17% of PC contributes to the errors in M _COSMOS. Therefore the high M _COSMOS−M _TOT correlation and little dependence of σ_abs(TOT) on the PC/M _TOT ratios indicate the reliability of the PC estimate. This in turn indicates that σ_abs(TOT) represents BC mass concentrations disturbed little by refractory organics at these sites.

It has been reported that large amounts of organics are emitted from biomass burning (e.g., Mayol-Bracero et al. 2002). As an example, we use data obtained in Amazonia to assess the possible interference of organics on M _COSMOS for this case (Mayol-Bracero et al. 2002), assuming that the properties of HULIS in the Amazonian rain forest are similar to those of SRFA and SRHA. The average concentrations of BC, OC, and water-soluble organic (WSOC) in the fine mode were 7, 35, and 23 μgC m^–3 over the Amazon Basin in October 1999. Mayol-Bracero et al. (2002) roughly estimated that up to 50–60% of WSOC may be composed of HULIS. It is unlikely that other identified organic species produce high OC5 signals. The interference of the HULIS on M _COSMOS with the inlet heated at 300°C and 400°C is crudely estimated to be about 0.3–0.4 μgC m^–3 and 0.8–1.2 μgC m^–3, respectively, assuming the same OC5/OC ratios as obtained by our laboratory experiments. This corresponds to errors in M _COSMOS of about 4–6% and 11–17%, respectively.

4. COMPARISON OF BC MEASUREMENTS

4.1. RMM versus TOT

The observed M _ref values were highly correlated (r ²= 0.88) with the M _TOT values in the M _TOT range between 0.5 and 7 μg m^–3 in Tokyo. The obtained relation is expressed as

This high correlation and a linear regression slope near unity demonstrate that M _TOT and M _ref were nearly identical. It should be noted that the M _ref and M _TOT values were obtained every 1 h for 22 days under different meteorological conditions. M _ref should be influenced by refractory organics more sensitively than M _TOT, considering that significant fractions of the HULIS remained even at 400°C. The remaining mass concentrations of refractory organics was estimated from the average PC/M _TOT= 0.047 (Table 7). The remaining mass relative to M _TOT was calculated to be about 0.12 by using the average F _OC5, mass fractions of carbon in HULIS (∼0.5), and the fractions of volumes lost by heating (Table 3). This suggests that M _ref may have been overestimated by the same amount. If M _ref in Equation (5) is interpreted to represent the real BC mass concentrations, the factor of 0.96 becomes 1.08.

4.2. COSMOS versus SP2

4.2.1. Correlation between COSMOS and SP2 Measurements

Simultaneous measurements of BC were obtained with the SP2 and the COSMOS instrument for 9 days at RCAST between August 26 and September 12, 2009. A PM_2.5-cyclone was used to exclude coarse particles. The weather was either mostly sunny or cloudy with little precipitation during the measurement period.

The sensitivity of the SP2 was found to be stable to within about 10% for particle mass measurement by means of calibration before and after the ambient measurements. The instrument was calibrated with fullerene soot. The lognormal size distribution was fitted to the mass size distributions of BC (dM _SP2/dlogD _BC). For comparison with the COSMOS measurements, we have integrated the mass size distribution dM _SP2/dlogD _BC to derive the total BC mass concentration over the entire BC size range (m _SP2).

FIG. 8 Daily averaged number and mass size distributions of black carbon measured by SP2 in Tokyo in August and September 2009.

FIG. 9 Frequency distribution of the CMD and MMD of the lognormal size distributions fitted to 1-min averaged BC mass size distributions measured by SP2.

Figure 8 shows examples of the 1-day averaged number size distribution (dN/dlogD _BC) and dM _SP2/dlogD _BC values measured between August 27 and September 11, 2009, along with lognormal fitted curves. In most cases, the count median diameter (CMD) was smaller than 70 nm. The BC mass for particles with diameters less than 70 nm was estimated assuming the lognormal mass size distributions (lower panel of Figure 8). On average, about 9–13% of the total BC mass concentration was in the size range below 70 nm for MMD = 130–185 nm.

Figure 9 shows the frequency distribution of the CMD and MMD for lognormal size distributions with a 1-minute resolution. CMD was derived stably only for a small fraction (about 10%) of the data. The mean CMD was 64 ± 6 nm, with a mean σ_gc of 1.66 ± 0.12. The majority of MMD were in the range of 130–170 nm, and the mean MMD value was 146 ± 12 nm, with a mean σ_gm of 1.82 ± 0.14.

Figure 10 shows the median shell/core ratios (D _p/D _BC) for different ranges of D _BC. The median D _p/D _BC ratios were about 1.1, indicating that the BC particles were thinly coated. The possible effect of charring in extracting m*_ref is considered to be small from these results, as discussed in section 2.2.

FIG. 10 Frequency distribution of the shell/core (D _p/D _BC) ratios of BC particles with different D _BC size ranges.

FIG. 11 Time series plots of BC mass concentrations measured by SP2 and COSMOS, their ratios, and MMD in Tokyo from September 10 to September 12, 2009. The mass concentration of BC was derived from the observed absorption coefficients by assuming a mass absorption cross section of 5.4 m² g^–1.

We estimated the error in M _SP2 due to the uncertainty of the unmeasured BC is smaller than 5% by changing the parameters of the lognormal fits. We estimated the overall uncertainty of M _SP2 due to the uncertainties of the calibration and the estimate of the unmeasured BC to be about 11%.

Figure 11 shows the time series of M _COSMOS, M _SP2, M _COSMOS/M _SP2 ratios, and MMD averaged for 1 min for the data obtained from September 10–12. The MMD ranged between 135 and 170 nm, and M _COSMOS and M _SP2 tracked each other well in this range.

Here we used σ_abs= 5.4 m² g^–1, which was obtained by the comparison of COSMOS and TOT measurements in Tokyo (Kondo et al. 2009) and updated using the new f _fil.

Figure 12 shows the correlation between M _COSMOS and M _SP2 (1-min and 1-h averages) obtained during the observation period. The data were highly correlated (r ²= 0.96 and 0.97) at BC mass concentrations lower than 3–4 μg m^–3. The high correlation indicates that interference from refractory organic aerosols to M _COSMOS was small, considering that M _SP2 is not influenced by the coexistence of organic aerosols.

FIG. 12 Correlation plots of BC mass concentrations measured by SP2 and those measured by COSMOS in 2009 using 1-min average data (left panel) and 1-h average data (right panel). Mass concentration of BC by COSMOS was derived from the absorption coefficients by assuming a mass absorption cross section of 5.4 m² g^–1. The least squares fitted line is denoted as the solid line.

The average relationships of M _COSMOS with M _SP2 are expressed as follows.

We used bivariate regression analysis incorporating the errors in both variables. The high r ² and small offset indicate that BC observed by SP2 (by incandescence) and COSMOS (by absorption) were similar. The value of σ_abs derived from the correlation between b _abs and M _SP2 was σ_abs(SP2) = 5.5 ± 0.6 m² g^–1. The σ_abs(TOT) = 5.4 ± 0.4 m² g^–1 agreed very well with σ_abs(SP2). This agreement means that M _TOT agreed with M _SP2 to the same degree on average. The small difference of the σ_abs(SP2) and σ_abs(TOT) is well within the uncertainties of the TOT, SP2, and COSMOS measurements.

It is expected that M _COSMOS, M _TOT, and M _SP2 agree well in Asian regions free from strong influence of biomass burning, considering that σ_abs(TOT) was stable to within 10% in the 6 Asian locations. We also expect SP2 calibration with m*_ref or fullerene soot to be valid in these regions.

4.2.2. Size Dependence of M _COSMOS

We expect M _COSMOS to have predicted and well-defined uncertainty despite its overall good agreement with M _SP2. M _COSMOS/M _SP2 may vary with BC size distribution owing to the size dependence of σ_abs. For monodisperse BC, σ_abs depends on D _BC for two reasons. First, the mass absorption cross section for airborne BC particles (σ*_abs) depends on D _BC. The value of σ*_abs is reported to peak at a D _BC of 150–200 nm (Bond and Bergstrom 2006) for typical refractive indices of BC particles. However, ambient BC particles are nonspherical (Park et al. 2004), and reliable estimates of σ*_abs for nonspherical BC particles are very limited. Second, the decrease in the frequency of multiple scattering with the increase in D _BC decreases the photo-absorption by BC, resulting in a need of smaller correction for multiple scattering. The need for a smaller correction means that f _fil given by Equation (3) should decrease with increasing D _BC. If the MMD of dm _SP2/dlogD _BC increased from a standard value, the f _fil used for the present study would be overestimated, and the overestimation would lead to an underestimation of M _COSMOS.

FIG. 13 M _COSMOS/M _SP2 ratio versus the MMD of BC. The solid black line is the least squares fitted line, and the red squares denote the mean M _COSMOS/M _SP2 ratios for 10-nm MMD bins.

To investigate this size dependence, we plotted M _COSMOS/M _SP2 versus MMD (Figure 13). The MMD was determined by the size distributions obtained by the SP2. The ratios decreased with MMD at an average slope of –0.0014/nm; on average, the ratios increased by about 7% as MMD increased from 130 to 180 nm. We estimate that f _fil decreased by about 20% as MMD increased from 130 to 150 nm, on the basis of laboratory experiments and theoretical calculations (Kondo et al. 2009; Nakayama et al. 2010; Moteki et al. 2010b). This decrease in MMD resulted in the underestimation of M _COSMOS by the same amount. The estimate of the decrease in the M _COSMOS/M _SP2 ratio by Nakayama et al. (2010) includes the size dependence of σ*_abs for typical refractive indices of BC based on the assumption of spherical particle shape. The fact that the MMD dependence predicted by the laboratory experiments was significantly larger than the observed dependence may be due to the difference in the size dependence of σ*_abs assumed for the prediction.

5. DISCUSSION AND RECOMMENDATIONS

The refractory particles extracted by means of a heated inlet at 400°C provided a common reference for BC mass concentrations for all the instruments used in the present study. However, the responses of the different types of optical BC instruments may depend on the microphysical properties of BC (refractive index and shape). The peak LII intensity–m*_ref relationship depends on these parameters (Moteki and Kondo 2010). We used the m*_ref of ambient BC particles that incandesce in urban air for this study. Considering the good agreement of the peak LII intensity–m*_ref and peak LII intensity–m _full relationships, either fullerene soot or thermally extracted BC should be used for calibration of an SP2. Thinly coated BC has to be chosen for thermal extraction to minimize the interference of refractory organics.

M _COSMOS obtained with an inlet heated to 400°C correlated highly with M _TOT and M _SP2, with very good agreement between σ_abs (SP2) and σ_abs (TOT). The interference by charring of refractory organics on M _COSMOS was estimated to be less than 5% based on laboratory experiments combined with thermograms of TOT. It was also shown that the degree of the interference is reduced by a factor of about 3 by lowering the inlet temperature down to 300°C without causing interference from low-MW organics. We recommend the use of COSMOS at 300°C if the same configuration and operational conditions of the heated inlet are adopted.

If the filter material (Pallflex E70-2075W, Pall, USA) used in a photo-absorption photometer with a heated inlet is the same as that used for a COSMOS and particle soot absorption photometer (PSAP; Radiance Research, WA, USA), the constants used for the present study can be used as is; that is, σ_abs= 5.5 m² g^–1 combined with Equations (1)–(4) can be used. If the filter materials are different, σ_abs can be determined by comparison of b _abs with COSMOS measurements or with other standards, including a well-calibrated SP2 instrument.

As discussed in section 3.4. R ₄₀₀ (total mass) was as low as 0.03–0.06. This ratio decreased by 1/3 at 300°C. By contrast, the HULIS mass concentration contributes to M _ref more sensitively, and the sensitivity is larger at 300°C. It is likely that the good correlation between the M _ref and M _TOT values holds only at locations where the mass concentrations of refractory organics relative to M _TOT are small, such as in Tokyo.

6. SUMMARY AND CONCLUSIONS

Coexistence of OA causes significant uncertainties in BC measurements, depending on the techniques used. In order to assess the effects of OA, we measured the volatilities of 13 organic species with different MW through a heated inlet. The low-MW organics were evaporated almost completely. Only about 10–30% and 20–60% of the volumes of the HULIS were lost at 300°C and 400°C, respectively.

For the most reliable BC mass concentration measurements, we used the SP2, which is free of interference from aerosols other than BC. For calibration of the SP2, we extracted particles that survived evaporation after passing through an inlet heated to 400°C. The refractory particles that incandesced (m*_ref) were used to calibrate the LII peak intensity of the SP2. The uncertainty of m*_ref is estimated to be about 10%. Fullerene soot showed microphysical properties (refractive index and shape) similar to those of ambient BC. The peak LII intensity was similar for the mass of fullerene soot (m _full) that was the same as m*_ref; this result is consistent with the similarity in the microphysical properties of fullerene soot and ambient BC. Practically, m _full can be used as a primary standard. However, confidence in the absolute calibration of SP2 increases if both m _full and m*_ref are used and compared under various conditions, considering the limited measurements made thus far.

The HULIS underwent partial charring, causing additional absorption in the COSMOS. Only about 5–12% of the carbon atoms contained in HULIS contributed to the equivalent M _COSMOS with an inlet heated at 400°C. This interference is reduced by about a factor of 3 at 300°C. The charred carbon mass was found to be directly related with the OC5 signal of the TOT. The [equivalent M _COSMOS]/[OC5] ratios were about 0.17 and 0.05, respectively, for the inlet temperatures of 400°C and 300°C. The average errors due to charring of refractory organics on M _COSMOS were estimated to be smaller than 5% in Asia. We recommend setting the inlet temperature at 300°C for COSMOS measurements.

The values of b _abs were measured simultaneously with M _SP2 for 9 days in Tokyo; the values were highly correlated (r ²= 0.97). The high correlation between M _COSMOS and M _SP2 is understood considering little influence of the coexistence of organic aerosols on these measurements. The derived σ_abs(SP2) agreed with σ_abs (TOT) to within 2%. This means that the masses of PM₁ refractory aerosols that incandesce were very close to M _TOT.

The value of σ_abs(SP2) varied by only 7% as MMD increased from 130 to 180 nm and changed little with aging. It should be noted that these results are restricted to Tokyo and similar aerosol mixtures that are not strongly impacted by refractory organic compounds such as those expected from biomass burning.

These results demonstrate that M _SP2, M _TOT, and M _COSMOS are nearly identical to a first approximation in Asia in the absence of pronounced effects of forest fires and dust particles. Throughout these studies, the COSMOS measurements were stable and traceable to the calibrated SP2 measurements to within about 10%.

APPENDIX

A1. Method to Calculate Coating Thickness

The shape of the BC particles in the atmosphere is known to be fractal. Even if these BC particles are coated with volatile compounds, their structure is not concentric shell-core in a strict sense. Despite the difference in the detailed structures, coating thickness or shell/core ratio is a useful conceptual parameter to quantify the mixing state of BC. For the present study, we obtained the shell/core ratio by deriving the time-dependent scattering cross sections of BC particles from the scattering waveforms from an SP2 (Moteki and Kondo 2008). We derived the scattering cross section of coated BC measured at the leading edge of the laser beam (before evaporation) and that measured at the onset of incandescence (after evaporation of coatings) using the light-scattering algorithm for concentric spheres (Aden and Kerker 1951; Bohren and Huffman 1983). The shell/core ratio derived by the new algorithm depends little on the assumed refractive indices. Further, the morphological dependence of light scattering and thermal emission has little affect on the shell/core ratios. This is because both the diameters of the core and shell are scaled by the light-scattering intensity measured by the same detection method. If the core diameter is estimated from the L-II peak signal, the uncertainty of the derived shell/core ratios will be larger.

A2. Dependence of the LII Intensity on the Microphysical Properties of Black Carbon

The time (t)-dependent intensity of the incandescence signal S_i (t) measured by the SP2 is given as

where C _emit is the emission cross section of the thermal emission of BC in m², η is a wavelength (λ)-dependent function describing the transmissivity of the optical filter and the responsivity of the photodetector, V is the particle volume, N is the complex refractive index, Ω is the direction of radiation propagation, and P_e is the Planck function in units of radiance (Moteki and Kondo 2010). C _emit depends also on the shape of BC.

C _emit of an arbitrary body for radiation is equivalent to the photo-absorption cross section (Moteki et al. 2009). The photo-absorption cross section of BC (C _abs(D _BC, λ)), where D _BC is the mass equivalent diameter of BC, depends on N, the particle shape, and the mixing state (Moteki and Kondo 2010; Shiraiwa et al. 2009). The size parameter x (= kD _BC/2) is defined to categorize C _abs(D _BC, λ) at large and small values of D _BC relative to λ. Here k= 2π/λ is the wavenumber. For x< ∼1, the necessary conditions for the Rayleigh–Gans approximation are generally satisfied, and C _abs (D _BC, λ) is given by

where Im is the imaginary part of the quantity given in parentheses. Because of the fractal structure of ambient soot and fullerene soot, the Rayleigh–Gans approximation holds up to a BC mass of about 20 fg.

This expression indicates that C _abs is proportional to volume V (or mass), Im[(N ² –1)/(N ²+ 1)], and 1/λ. It does not depend on shape. For x> ∼1, generally the Rayleigh–Gans approximation does not hold, and C _abs depends on particle shape. These size and shape dependencies are the cause of the observed nonlinearity in the correlation between the LII peak intensity and M _SP2 at large D _BC values (Moteki and Kondo 2010).

Acknowledgments

The authors thank T. Ishigai for his support of the laboratory experiments. D. W. Fahey, J. P. Schwarz, and M. Kuwata provided very useful comments on the manuscript. This work was supported by the Ministry of Education, Culture, Sports, Science, and Technology (MEXT), the strategic international cooperative program of the Japan Science and Technology Agency (JST), and the global environment research fund of the Japanese Ministry of the Environment (B-083). This study was conducted as a part of the Mega-Cities: Asia Task under the framework of the International Global Atmospheric Chemistry (IGAC) project.