High time-resolved measurement of stable carbon isotope composition in water-soluble organic aerosols: method optimization and a case study during winter haze in eastern China

Abstract

Water-soluble organic carbon (WSOC) is a significant fraction of organic carbon (OC) in atmospheric aerosols. WSOC is of great interest due to its significant effects on atmospheric chemistry, the Earth's climate and human health. The stable carbon isotope (${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}$) can be used to track the potential sources and investigate atmospheric processes of organic aerosols. However, the previous methods measuring the ${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}$ values of WSOC in ambient aerosols require a large amount of carbon content, are time-consuming and require labor-intensive preprocessing. In this study, a method of simultaneously measuring the mass concentration and the ${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}$ values of WSOC from aerosol samples is established by coupling the GasBench II preparation device with isotopic ratio mass spectrometry. The precision and accuracy of isotope determination is better than 0.17 ‰ and 0.5 ‰, respectively, for samples containing WSOC amounts larger than 5 $\mathrm{\mu }\mathrm{g}$. This method is then applied for the aerosol samples collected every 3 h during a severe wintertime haze period in Nanjing, eastern China. The WSOC values vary between 3 and 32 $\mathrm{\mu }\mathrm{g}$ m${}^{-\mathrm{3}}$, whereas ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ ranges from $-\mathrm{26.24}$ ‰ to $-\mathrm{23.35}$ ‰. Three different episodes (Episode 1, Episode 2 and Episode 3) are identified in the sampling period, showing a different tendency of ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ with the accumulation process of WSOC aerosols. The increases in both the WSOC mass concentrations and the ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ values in Episode 1 indicate that WSOC is subject to a substantial photochemical aging during the air mass transport. In Episode 2, the decline of the ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ is accompanied by the increase in the WSOC mass concentrations, which is associated with regional-transported biomass burning emissions. In Episode 3, heavier isotope (${}^{\mathrm{13}}\mathrm{C}$) is exclusively enriched in total carbon (TC) in comparison to WSOC aerosols. This suggests that the non-WSOC fraction in total carbon may contain ${}^{\mathrm{13}}\mathrm{C}$-enriched components such as dust carbonate, which is supported by the enhanced ${\mathrm{Ca}}^{\mathrm{2}+}$ concentrations and air mass trajectory analysis. The present study provides a novel method to determine the stable carbon isotope composition of WSOC, and it offers a great potential to better understand the source emission, the atmospheric aging and the secondary production of water-soluble organic aerosols.

1 Introduction

Water-soluble organic carbon (WSOC) contributes to a large fraction (9 %–75 %) of organic carbon (OC) (Sullivan et al., 2004; Decesari et al., 2007; Anderson, et al., 2008) and affects substantially the global climate change and human health (Ramanathan et al., 2001; Myhre, 2009). Due to its hydrophilic nature, WSOC has a great impact on the hygroscopic properties of aerosols and promotes the increase of cloud condensation nuclei (CCN) activity (Asa-Awuku et al., 2011). WSOC is a contributor to cardiovascular and respiratory problems because it is easy to be incorporated in biological systems such as human blood and lungs (Mills et al., 2009).

WSOC can be emitted as primary organic carbon (POC) and secondary organic carbon (SOC) produced from atmospheric oxidation of volatile organic compounds (VOCs) (Sannigrahi et al., 2006; Weber et al., 2007; Zhang et al., 2018). Due to the hygroscopic property of the WSOC, the origins of POC may be from biomass burning or marine emissions. However, the SOC may stem from various sources including coal combustion, vehicle emissions, biogenic emissions, marine emissions and biomass burning (Decesari et al., 2007; Jimenez et al., 2009; Kirillova et al., 2010, 2013; Bozzetti et al., 2017a, b).

Stable carbon isotopic composition (${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}$) can provide valuable information to track both potential sources and atmospheric processes of carbonaceous aerosols (Rudolph, 2007; Pavuluri and Kawamura, 2012; Kirillova et al., 2013, 2014). Carbonaceous aerosols from coal combustion have an isotope signature from $-\mathrm{24.9}$ ‰ to $-\mathrm{21}$ ‰ (Cao et al., 2011). Particulate matter emitted from motor vehicles exhibits with isotopes from $-\mathrm{26}$ ‰ to $-\mathrm{28}$ ‰ (Widory, 2006). Due to the different pathways of metabolism, C${}_{\mathrm{3}}$ and C${}_{\mathrm{4}}$ plants exhibit significant differences of ${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}$ (approximately $-\mathrm{27}$ ‰ for C${}_{\mathrm{3}}$ and $-\mathrm{13}$ ‰ for C${}_{\mathrm{4}}$; Martinelli et al., 2002; Sousa Moura et al., 2008). Laboratory studies demonstrate that there is no significant isotope fractionation ($\pm \mathrm{0.5}$ ‰) between the produced aerosols and the C${}_{\mathrm{3}}$ plant material (Turekian et al., 1998; Currie et al., 1999; Das et al., 2010). The burning of C${}_{\mathrm{4}}$ plants results in ${}^{\mathrm{13}}\mathrm{C}$ depletion (< 0.5 % to 7.2 %) of the produced aerosols (Turekian et al., 1998; Das et al., 2010). Marine organic aerosol sources have a carbon isotope signature of $-\mathrm{22}$ ‰ to $-\mathrm{18}$ ‰ (Miyazaki et al., 2011) and play an important role in the aerosols at coastal sites. In contrast, carbonate carbon exhibits with a pretty high isotopic ratio of $-\mathrm{0.3}$ ‰ (Kawamura et al., 2004) and generally shows a large proportion in dust aerosols. Thus, the isotope signatures of particulate matter emitted from these various sources may have different effects on the characteristics of ${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}$ in ambient WSOC.

In addition, atmospheric processes like secondary formation and photochemical aging may change the constitution and properties of WSOC, as well as the stable carbon isotope of WSOC (${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$). According to the kinetic isotope effect (KIE), the reaction rate of molecules containing heavier isotopes is usually lower than the molecules containing lighter isotopes (Atkinson, 1986; Fisseha et al., 2009; Kirillova et al., 2013). The change in reaction rate primarily results from the greater energetic need for molecules containing heavier isotopes to reach the transition state (Brüniche-Olsen and Ulstrup, 1979). Consequently, the oxidants preferentially react with molecules with lighter isotopes (inverse kinetic isotope effect), which would result in an enrichment of ${}^{\mathrm{13}}\mathrm{C}$ in the residual materials and a depletion in ${}^{\mathrm{13}}\mathrm{C}$ of the particulate oxidation products (Rudolph et al., 2002). Therefore, organic compounds formed via secondary formation are generally depleted in ${}^{\mathrm{13}}\mathrm{C}$ compared with their precursors (Sakugawa and Kaplan, 1995; Fisseha et al., 2009), and this isotope depletion has been proved by both field measurements and laboratory studies (Pavuluri and Kawamura, 2012). For example, the studies of KIE clearly indicate that the compounds formed via the oxidation are depleted in the ${}^{\mathrm{13}}\mathrm{C}$ compared with their precursors during the reaction of VOCs with OH and ozone (dominant atmospheric oxidants) (Rudolph et al., 2000; Iannone et al., 2003; Anderson et al., 2004; Fisseha et al., 2009). An enrichment of ${}^{\mathrm{13}}\mathrm{C}$ in the particulate organic aerosol may occur in the atmospheric aging processes, such as interactions with photochemical oxidants (e.g., hydroxyl radical and ozone), during the long-range transport. For instance, studies have demonstrated that the substantial enrichment of ${}^{\mathrm{13}}\mathrm{C}$ in the residual, aged aerosols (e.g., isoprene, a precursor of oxalic acid; Rudolph et al., 2003) occurred after a long-range transport. In that case, the stable carbon isotope can be used to study the sources and the atmospheric processes that contribute to the carbonaceous aerosols.

Several studies report the temporal and spatial variation, complex chemical species, light absorption and thermal characteristics of WSOC, as well as its relationship with other compounds in fine particles (Wang et al., 2006; Wozniak et al., 2008; Martinez et al., 2016; Zhang et al., 2018). However, only few studies focus on the analysis of ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ (Fisseha et al., 2006; Kirillova et al., 2010; Lang et al., 2012; Zhou et al., 2015; Suto and Kawashima, 2018). This is partially due to the limited techniques to analyze the ${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}$ signatures of WSOC in ambient aerosols, as their concentrations are usually very small. In the recent years, some efforts have been made to measure the ${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}$ values of WSOC. Bauer et al. (1991) used potassium persulfate to convert organic carbon in natural waters into ${\mathrm{CO}}_{\mathrm{2}}$ for ${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}$ measurements. This wet oxidation method requires more than 0.5 mM C and 1 h for the pretreatment (from sample injection to the isolation of purified ${\mathrm{CO}}_{\mathrm{2}}$). Fisseha et al. (2006) boiled the oxidizing solution for 45 min to remove the organic matter, and the total time required for the pretreatment (for 15 samples) is 1.5 h. Kirillova et al. (2010) developed a combustion method that applies the aerosol extract without filtration for isotope measurement and involves complicated processes such as freeze-drying of the aerosol extract under vacuum for 16 h. This combustion method is the most widely used for the ${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}$ measurements in WSOC aerosols (Kirillova et al., 2010, 2013, 2014; Miyazaki et al., 2012; Pavuluri Kawamura, 2017). Although these methods are able to provide the ${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}$ values of WSOC in natural waters and/or ambient aerosols, the analytical methods require either a large amount of WSOC (from 100 $\mathrm{\mu }\mathrm{g}$ C to 0.5 mM C) or time-consuming preprocessing. And some of the methods oxidize the WSOC extract without filtration and/or decarbonation in the pretreatment, which would result in higher uncertainty of the ${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}$ results. The high detection limit of the previous methods is difficult to determine the ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ in aerosol samples with low carbon concentrations. In that case, an easily operated method detecting the ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ values in aerosol samples with a low detection limit and high precision is urgently needed. The objectives of this study are as follows: (1) to provide an accurate, precise and easily operated method to measure the WSOC and ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ in ambient aerosol samples and (2) to apply this method for analyzing the high time-resolution aerosol samples during a severe haze and discuss the potential sources and the atmospheric processes of WSOC. In addition, the concentrations of inorganic ions and air mass back trajectories coupled with MODIS fire maps are also analyzed to substantiate the results obtained from the ${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}$ analysis.

2 Methods

2.1 Standards

Four working standards were used in this study: potassium hydrogen phthalate (KHP), benzoic acid (BA), sucrose (${\mathrm{CH}}_{\mathrm{6}}$) and sodium oxalate (${\mathrm{C}}_{\mathrm{2}}$). KHP and BA are widely used as the standards of WSOC measurements (Kirillova et al., 2010), and they were used here as the WSOC test substances. Also, their isotope signatures are close to the ${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}$ values of aerosol samples (Fisseha et al., 2009; Miyazaki et al., 2012; Suto and Kawashima, 2018). Sucrose and oxalic are taken as the standards to represent the characteristics of the components in atmospheric WSOC (Pathak et al., 2011; Pavuluri and Kawamura, 2012; Liang et al., 2015; Fowler et al., 2018). The carbon isotope composition of these four standards is as follows: $-\mathrm{12.20}$ ‰ (${\mathrm{CH}}_{\mathrm{6}}$), $-\mathrm{13.84}$ ‰ (${\mathrm{C}}_{\mathrm{2}}$), $-\mathrm{27.17}$ ‰ (BA) and $-\mathrm{30.40}$ ‰ (KHP), respectively. The wide range of the delta values of the working standards is able to cover the majority of the ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ values in ambient aerosol samples. Standards were resolved in Milli-Q water (18.2 M$\mathrm{\Omega }$ quality) to make standard solutions of 0.25, 0.75, 1.5, 3, 6, 12 and 24 $\mathrm{\mu }\mathrm{g}$ mL${}^{-\mathrm{1}}$, which means a carbon content of 1, 3, 6, 12, 24, 48 and 96 $\mathrm{\mu }\mathrm{g}$ in 4 mL standard solution to test the procedures during the pretreatment.

2.2 Aerosol samples

The aerosol samples were collected during a severe haze in January (from 14 to 28 January) of 2015 in a suburb of Nanjing, a megacity in eastern China. The sampling site is located at the agrometeorological station in the campus of the Nanjing University of Information Science and Technology. It is close to a busy traffic road and surrounded by a large number of industrial factories. The PM${}_{\mathrm{2.5}}$ samples were collected on pre-combusted quartz-fiber filters ($\mathrm{180}\times \mathrm{230}$ mm) every 3 h with a high-volume aerosol sampler (KC100, Qingdao, China) at a flow rate of 1 m${}^{\mathrm{3}}$ min${}^{-\mathrm{1}}$. After sampling, all the filters were wrapped in the aluminum foil, sealed in air-tight polyethylene bags and stored at $-\mathrm{26}$ ${}^{\circ }$C for later analysis. A field blank was obtained by placing the blank filter in the filter holder for 10 min without sampling.

2.3 Chemical analysis

The PM${}_{\mathrm{2.5}}$ concentrations were observed at Pukou environmental supervising station. Concentrations of total carbon (TC) and ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{TC}}$ values were analyzed with and elemental analyzer combined with an isotope ratio mass spectrometer (EA-IRMS,Thermo Fisher Scientific, Bremen, Germany). The WSOC mass concentrations were measured with the total organic carbon (TOC) analyzer (Shimadzu). Ion concentrations were obtained from the ion chromatography system (IC, Thermo Fisher Scientific, Bremen, Germany). The meteorological data were observed close to the sampling site (Envis automatic meteorological station).

2.4 Sample pretreatment

The wet oxidation method was used to covert the WSOC to ${\mathrm{CO}}_{\mathrm{2}}$ (Sharp, 1973), and the resulting ${\mathrm{CO}}_{\mathrm{2}}$ levels were measured by an isotope ratio mass spectrometer (IRMS). The overview of the optimized method for measuring WSOC and ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ in the aerosols is shown in Fig. 1. The process of the pretreatment consisted of six steps: WSOC on a 20 mm diameter disk was extracted with 6 mL Milli-Q water in an ultrasonic water bath for 30 min (step 1–2). The WSOC extract was filtered with a 0.22 $\mathrm{\mu }\mathrm{m}$ syringe filter to remove the particles in step 3. A total of 2.0 g potassium persulfate (${\mathrm{K}}_{\mathrm{2}}{\mathrm{S}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{8}}$, Aladdin Industrial Corporation, Shanghai) and 100 $\mathrm{\mu }\mathrm{L}$ phosphoric acid (85 % ${\mathrm{H}}_{\mathrm{3}}{\mathrm{PO}}_{\mathrm{4}}$, AR, ANPEL Laboratory Technologies Inc., Shanghai) were dissolved in 50 mL Milli-Q water to make the oxidizing solution. The oxidizing solution made within 24 h was added into the filtered WSOC extract as shown in step 4 of Fig. 1. The phosphoric acid was added to remove the inorganic carbon resolved in the solution, and the persulfate was added for the preparation to convert the organic compounds to ${\mathrm{CO}}_{\mathrm{2}}$. The vials were sealed tightly with the caps as soon as the oxidizing solution was added to the WSOC extract.

2.5 Determination of the carbon content and stable carbon isotopic ratios

${\mathrm{CO}}_{\mathrm{2}}$ gas produced in the headspace of the prepared sample was extracted and purified by the GasBench II system (Thermo Fisher Scientific, Bremen, Germany) and was introduced into an isotope ratio mass spectrometer (IRMS) (MAT 253, Thermo Fisher Scientific, Bremen, Germany) for ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{CO}{}_{\mathrm{2}}}$ analysis. The extracted gas was purified with a Nafion water trap to remove the water vapor, and then the gas was loaded into a 100 $\mathrm{\mu }\mathrm{L}$ sample loop through an eight-port Valco valve. After 120 s loading time (the duration time from the beginning of the analysis to the first rotation of the eight port in the GasBench II), the eight-port Valco valve rotated every 70 s to inject the sample gas from the loop into a GC column (PoraPLOT Q fused-silica cap, 25 m, 0.32 mm; Agilent Technologies). The GC column was set at 40 ${}^{\circ }$C for the ${\mathrm{CO}}_{\mathrm{2}}$ separation from the matrix gases. The separated ${\mathrm{CO}}_{\mathrm{2}}$ was introduced into another Nafion water trap and subsequently entered into the IRMS through an open split. The ${\mathrm{CO}}_{\mathrm{2}}$ gas in each vial was detected 10 times in 15 min, showing 10 sample peaks after five reference peaks. The peak areas and the isotope compositions of the 10 sample peaks were given correspondingly, and the results of the first two sample peaks were abandoned considering the possible effect on the memory of the system. The average peak area and the isotope composition of the last eight peaks were taken as the result of a certain sample determined by the GasBench II–IRMS.

3 Method optimization

The wet oxidation method was adapted from the stable isotope analysis of organic matter in ground water (Lang et al., 2012; Zhou et al., 2015). Several tests were performed to adjust the optimal conditions for measuring WSOC aerosols with relative low carbon amounts.

3.1 The carbon content in the procedural blank

In order to quantify the low concentration of WSOC in aerosols, it is critical to reduce the carbon content in the procedural blank for minimizing the detection limit of the method. To achieve this goal, the procedural blanks were analyzed to test for contamination that the reagents could have introduced to the results (shown in Table 1). The average carbon content in the procedural blank was about 0.5 $\mathrm{\mu }\mathrm{g}$ C (corresponding with a peak area of 0.23 Vs) with a ${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}$ value of $-\mathrm{27.04}\pm \mathrm{1.28}$ ‰ ($n=\mathrm{15}$). The carbon content and the isotope compositions of Milli-Q water and the agents dissolved in the oxidizing solution were also determined in order to identify the source of contamination in the procedural blank. The peak area of Milli-Q water was not detected (Table 1) after going through all the processes in the pretreatment without adding any other materials, suggesting that no contamination had been introduced from the Milli-Q water. Furthermore, the contamination from 85 % ${\mathrm{H}}_{\mathrm{3}}{\mathrm{PO}}_{\mathrm{4}}$ with different purities (acid-1: analytical reagent, AR; acid-2: high performance liquid chromatography, HPLC) was compared. The carbon content in the 85 % ${\mathrm{H}}_{\mathrm{3}}{\mathrm{PO}}_{\mathrm{4}}$ dissolved in Milli-Q water was 0.03–0.04 $\mathrm{\mu }\mathrm{g}$ C and showed no significant discrepancy between different purities.

Table 1

Various blank preparations and results.

[Figure omitted. See PDF]

With respect to isotope composition, KHP standards heated for 15, 30 and 60 min all show stable results with similar standard deviations (from 0.51 to 0.57; see Table S1). The heating times of 15 and 30 min are not long enough for the complete oxidation, which is shown by the presence of a lower carbon content (Fig. 2). Therefore, heating for 60 min at 100 ${}^{\circ }$C is found to be the most suitable duration to produce constant results without gas leak and isotope fractionation.

3.4 Waiting time and instrument settings

The waiting time of the mixture (the aerosol extract and the oxidizing solution) between steps 4 and 5 in Fig. 1 was tested to prevent the ${\mathrm{CO}}_{\mathrm{2}}$ loss during the flushing. Some of the compounds in aerosol samples could be oxidized at room temperature. The ${\mathrm{CO}}_{\mathrm{2}}$ generated from the mixture before heating could be lost during the flushing step (Sharp, 1973). The ambient sample was tested to detect the room-temperature-oxidized ${\mathrm{CO}}_{\mathrm{2}}$ (Fig. S1 in the Supplement). Replicates of the ambient aerosol extract (from one filter) were mixed with the oxidizing solution, and the mixtures of the aerosol extract and the oxidizing solution were flushed with He to exclude the effect of ${\mathrm{CO}}_{\mathrm{2}}$ (both in the headspace and in the mixture) as soon as possible. After flushing, the mixtures were stored at room temperature from 1 to 31 h before analysis without heating. The carbon content produced in the mixtures that were stored less than 12 h before analysis was smaller than 0.02 $\mathrm{\mu }\mathrm{g}$, which contributes to $\sim \mathrm{7}$ % to the carbon content in the procedural blank (0.5 $\mathrm{\mu }\mathrm{g}$ C). But when the waiting time was extended to 31 h, up to 2.3 $\mathrm{\mu }\mathrm{g}$ C (about 5 times more than the procedural blank) was oxidized into ${\mathrm{CO}}_{\mathrm{2}}$. The room-temperature-oxidized ${\mathrm{CO}}_{\mathrm{2}}$ produced during the waiting time could be flushed out by the He in the later procedure and could result in significant isotope fractionation in the delta results. Therefore, the mixture should be flushed with He within 12 h to avoid the ${\mathrm{CO}}_{\mathrm{2}}$ loss and isotope fractionation.

In addition, various combinations of shorter loading times (30–90 s) and/or fewer sample peaks (i.e., five sample peaks) were tested with reference gas (${\mathrm{CO}}_{\mathrm{2}}$ mixed with He) to shorten the analysis in the system. However, the amount of ${\mathrm{CO}}_{\mathrm{2}}$ in the reference gas detected by the mass spectrometry was about 2 $\mathrm{\mu }\mathrm{g}$ C lower compared the results obtained with longer loading times and more sample peaks. And there was a decrease in isotope value ($\sim \mathrm{0.4}$ ‰) when the loading time was shorter or the sample peaks were less than 10. Thus, a 120 s loading time and 10 sample peaks are necessary for the precise results, and the standard deviation is < 0.03 ‰ for the 10 sample peaks within a run.

3.5 Calibration of the results

3.5.1 Quantification of the carbon content

The sample peak area is proportional to the carbon content in the vial and is used to quantify the amount of ${\mathrm{CO}}_{\mathrm{2}}$ in the inflow of IRMS. The average value of the peak areas for the last eight sample peaks is taken as the peak area of a certain sample. The first two sample peaks are excluded to avoid the effect of the residual ${\mathrm{CO}}_{\mathrm{2}}$ of the former vial. We established a carbon content standard curve (linear equation) by measuring the peak areas of ${\mathrm{CO}}_{\mathrm{2}}$ gas samples containing 1–24 $\mathrm{\mu }\mathrm{g}$ C (Fig. S2). It has to be noted that the gas samples with a larger carbon content are not tested for the difficulty of injecting a too large volume of mixed ${\mathrm{CO}}_{\mathrm{2}}$ and $\mathrm{He}$ gas. Then the amount of ${\mathrm{CO}}_{\mathrm{2}}$ oxidized from the unknown samples can be quantified with the following linear equation: carbon content ($\mathrm{\mu }\mathrm{g}$) $=$ peak area (Vs) $\times (\mathrm{2.50}\pm \mathrm{0.08})-(\mathrm{0.62}\pm \mathrm{0.39}$), $R^{\mathrm{2}}=\mathrm{0.98}$. The standard curve (linear equation) of the peak areas against the carbon content in the WSOC solution (KHP solution containing 1–100 $\mathrm{\mu }\mathrm{g}$ C) is also established (Fig. S2). And a linear equation similar with the peak areas against ${\mathrm{CO}}_{\mathrm{2}}$ gas is obtained: carbon content ($\mathrm{\mu }\mathrm{g}$) $=$ peak area (Vs) $\times (\mathrm{2.34}\pm \mathrm{0.01})-(\mathrm{0.86}\pm \mathrm{0.14}$), $R^{\mathrm{2}}=\mathrm{1.00}$.

Then the conversion efficiency of the WSOC extract containing 1–100 $\mathrm{\mu }\mathrm{g}$ C can be roughly calculated as $\mathrm{104}\pm \mathrm{3}$ %. A high conversion efficiency demonstrates a complete conversion and the negligible isotope fractionation during the oxidation. In that case, the carbon content in the WSOC extract of unknown samples can be calculated based on the standard curve of peak areas against the carbon content in the WSOC extract. And the standard curve quantifying the carbon content has to be established with every batch of unknown samples to assure the complete conversion.

3.5.2 Blank correction

The blank contribution to the WSOC mass concentrations and the ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ values are evaluated with the peak area and the isotope value of the procedural blank. The peak area (average value of the last eight peaks) from the measurement is proportional to the carbon content in the vial and then is taken to represent the ${\mathrm{CO}}_{\mathrm{2}}$ amounts in the inflow of IRMS. The procedural blank can be corrected according to the mass balance as follows:

1$$\begin{array}{rl}{\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{\mathrm{meas}}\times A_{\mathrm{meas}}=& {\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{\mathrm{corr}}\times \left(A_{\mathrm{meas}}-A_{\mathrm{blk}}\right)\\ & +{\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{\mathrm{blk}}\times A_{\mathrm{blk}},\end{array}$$ where ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{\mathrm{corr}}$, ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{\mathrm{meas}}$ and ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{\mathrm{blk}}$ are the blank-corrected ${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}$, the measured ${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}$ of the samples and the ${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}$ of the procedural blank, respectively. $A_{\mathrm{meas}}$ and $A_{\mathrm{blk}}$ denote the peak areas of the samples and the blank, correspondingly.

In order to calibrate the contribution of the procedural blank to the isotope results, $A_{\mathrm{blk}}$ and ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{\mathrm{blk}}$ were calculated with an indirect method (Polissar et al., 2009). KHP (${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}$ $=-\mathrm{30.40}$ ‰) and ${\mathrm{CH}}_{\mathrm{6}}$ (${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}$ $=-\mathrm{12.20}$ ‰) with various concentrations were measured to calculate $A_{\mathrm{blk}}$ and ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{\mathrm{blk}}$. The wide range of their isotopes can basically cover the ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ values in most ambient aerosol samples. According to Eq. (1), ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{\mathrm{meas}}$ can be written as the following: 2$${\displaystyle }{\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{\mathrm{meas}}={\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{\mathrm{corr}}+A_{\mathrm{blk}}({\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{\mathrm{blk}}-{\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{\mathrm{corr}})/A_{\mathrm{meas}}.$$ According to Eq. (2), there is a linear relationship of the ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{\mathrm{meas}}$ values and the reciprocal of peak areas (1/$A_{\mathrm{meas}}$). Based on the Keeling plot theory, linear equations of the ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{\mathrm{meas}}$ values and 1/$A_{\mathrm{meas}}$ for the two standards can be set up separately (e.g., ${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}$ and 1/$A_{\mathrm{meas}}$ values obtained from the measurement of ${\mathrm{CH}}_{\mathrm{6}}$ and their linear relationship are shown in Fig. S3). The slopes ($k_{\mathrm{1}}$ and $k_{\mathrm{2}}$) and the intercepts ($b_{\mathrm{1}}$ and $b_{\mathrm{2}}$) of this linear relationship can be expressed with ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{\mathrm{blk}}$, $A_{\mathrm{blk}}$ and ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{\mathrm{corr}}$ as follows. $$\begin{array}{}\text{3}& {\displaystyle }\begin{array}{rl}& k_{\mathrm{1}}=A_{\mathrm{blk}}\times \left({\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{\mathrm{blk}}-{\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{\mathrm{corr}-\mathrm{std}\mathrm{1}}\right)\\ & k_{\mathrm{2}}=A_{\mathrm{blk}}\times \left({\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{\mathrm{blk}}-{\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{\mathrm{corr}-\mathrm{std}\mathrm{2}}\right)\end{array}\text{4}& {\displaystyle }\begin{array}{rl}& b_{\mathrm{1}}={\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{\mathrm{corr}-\mathrm{std}\mathrm{1}}\\ & b_{\mathrm{2}}={\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{\mathrm{corr}-\mathrm{std}\mathrm{2}}\end{array}\end{array}$$ ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{\mathrm{corr}-\mathrm{std}\mathrm{1}}$ and ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{\mathrm{corr}-\mathrm{std}\mathrm{2}}$ are the blank-corrected ${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}$ values of two standard materials, for example, ${\mathrm{CH}}_{\mathrm{6}}$ and $\mathrm{KHP}$. Thus, $A_{\mathrm{b}}$ and ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{\mathrm{b}}$ can be calculated as follows. $$\begin{array}{}\text{5}& {\displaystyle }& {\displaystyle }{\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{\mathrm{blk}}=(k_{\mathrm{2}}\times b_{\mathrm{1}}-k_{\mathrm{1}}\times b_{\mathrm{2}})/\left(k_{\mathrm{2}}-k_{\mathrm{1}}\right)\text{6}& {\displaystyle }& {\displaystyle }{A}_{\mathrm{blk}}=(k_{\mathrm{2}}-k_{\mathrm{1}})/(b_{\mathrm{1}}-b_{\mathrm{2}})\end{array}$$ Thus, the blank contribution is able to be calibrated with the equation below: 7$$\begin{array}{rl}{\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{\mathrm{corr}}=& ({\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{\mathrm{meas}}\times A_{\mathrm{meas}}-{\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{\mathrm{blk}}\times A_{\mathrm{blk}})/\\ & (A_{\mathrm{meas}}-A_{\mathrm{blk}}).\end{array}$$ For example, ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{\mathrm{blk}}$ and $A_{\mathrm{blk}}$ are calculated to be $-\mathrm{27.43}$ ‰ and 0.3 Vs ($\sim \mathrm{0.5}$ $\mathrm{\mu }\mathrm{g}$ C) based on the results of KHP and ${\mathrm{CH}}_{\mathrm{6}}$ (shown in Fig. 3). The carbon content in the procedural blank contributed to 1 %–10 % of the carbon content of an ambient aerosol sample. Although the ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{\mathrm{blk}}$ and $A_{\mathrm{blk}}$ are not strongly varied values, they need to be measured before every batch of the ambient samples to assure the stable status of the system (IRMS) and the proper processes during the pretreatment.

Figure 3

Isotope results before and after the two-step correction of the four standards (a KHP, b BA, c ${\mathrm{CH}}_{\mathrm{6}}$, d ${\mathrm{C}}_{\mathrm{2}}$). Red circles with a spot represent the two-step corrected isotopic ratios; the other symbols represent the raw data from the GasBench II; the dotted line represents the blank-corrected ${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}$ values tested by EA.

[Figure omitted. See PDF]

3.5.3 Calibration of isotope results

In order to calibrate the isotope results, four working standards (KHP, BA, ${\mathrm{CH}}_{\mathrm{6}}$ and ${\mathrm{C}}_{\mathrm{2}}$) containing different carbon contents were measured with EA-IRMS and GasBench II–IRMS. The standards measured with the elemental analyzer (EA) were combusted at 1000 ${}^{\circ }$C to convert the organic materials into ${\mathrm{CO}}_{\mathrm{2}}$ for the measurement in IRMS without pretreatment. More than 10 repetitions of each standard were measured in this way, and the average delta values (after blank correction) of each standard are defined as correct values here. On the other hand, the average isotope compositions (after blank correction) of 10 repetitions obtained from the wet oxidation method (determined with GasBench II) are defined as measured values. Thus a calibration curve can be established on the basis of the measured values and the correct values (Fig. S4). For instance, the isotope results can be calibrated as follows:

8$${\displaystyle }{\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{\mathrm{cali}}=k\times {\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{\mathrm{blk}-\mathrm{corr}}+b,$$ where ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{\mathrm{cali}}$ is the isotope composition after the isotope calibration, ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{\mathrm{blk}-\mathrm{corr}}$ is the blank corrected isotope composition determined with GasBench II, $k$ and $b$ are the slope and the intercept obtained from the calibration curve. Similar with the blank correction, the isotope calibration curve needs to be established with each batch of the ambient samples to assure the stable status of the IRMS and the proper processes during the pretreatment.

Figure 4

Correlation of WSOC mass concentrations measured with the GasBench II–IRMS and TOC analyzer.

[Figure omitted. See PDF]

In this way, the isotope results can be calibrated. The raw data, the isotope composition after the blank correction and the isotope calibration determined with GasBench II are compared in Fig. 3. The correct values of standard carbon isotopes are plotted in Fig. 3. The isotope results after two steps of correction (the blank correction and the calibration of isotope results) are closer to the correct values (isotopes measured with EA), and the blank contribution is drastically eliminated. But for the standards containing carbon content smaller than 5 $\mathrm{\mu }\mathrm{g}$ C, the contribution of the procedural blank (with an isotope ratio about $-\mathrm{27.43}$ ‰) is still significant. According to the isotope variation of the ambient aerosols, the analysis of isotope compositions is not reliable if the repetitions of the standards show a difference larger than 1 ‰ (SD > 0.5 ‰). After correction, the standard deviations of isotope results of each standard are better than 0.17 ‰ (regardless of the carbon content of a certain standard) when the carbon content is larger than 5 $\mathrm{\mu }\mathrm{g}$ C. In that case, the detection limit of this method is 5 $\mathrm{\mu }\mathrm{g}$ C and the results (both carbon content and the isotopic ratios) of WSOC lower than 5 $\mathrm{\mu }\mathrm{g}$ C are not reliable.

3.6 Quality control and quality assurance procedures

A batch of working standards with different carbon content were measured to evaluate the optimized method in this study (data shown in Fig. 3). The qualities of the unknown samples were assured with a standard curve established with the peak areas and the corresponding input carbon content of WSOC extract (e.g., in Fig. S2). The conversion efficiency of the WSOC oxidation was found to be $\mathrm{104}\pm \mathrm{3}$ %. The average recovery of the working standards and the ambient samples were tested to be $\mathrm{97}\pm \mathrm{6}$ % and $\mathrm{99}\pm \mathrm{10}$ %, respectively. The conversion efficiency and the recoveries suggest complete oxidation of WSOC extract without significant isotope fractionation in the pretreatment. The blank contribution was evaluated with the peak area and the isotopic ratio, and these values are calculated with the indirect method introduced in Sect. 3.5.2. According to the carbon content (0.3–0.5 $\mathrm{\mu }\mathrm{g}$ C) and the isotope composition ($\sim -\mathrm{27.43}$ ‰) of the procedural blank, the WSOC detection limit of this method is 5 $\mathrm{\mu }\mathrm{g}$ C, 10 times more than the carbon content in the procedural blank. The blank-corrected isotope compositions should be calibrated again with the calibration curve as described in Sect. 3.5.3 to obtain the isotopic ratios of the unknown samples.

In order to obtain the carbon content and the corrected isotope compositions of the unknown samples, at least two kinds of standards need to be measured before every batch of the unknown samples. The range of the carbon content and the isotope compositions of the standards are required to cover the range of WSOC and ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ in the ambient samples, e.g., KHP, BA and ${\mathrm{CH}}_{\mathrm{6}}$. Hence, the concentration standard curve, the linear equations for the blank correction and the isotope calibration curve are able to be established according to the results of the standards. Besides, one standard should be measured after every 10 unknown samples to assure the stable status of the equipment.

As for the isotope measurement, the precision of the last eight sample peaks is < 0.15 ‰ within a run for standards containing more than 1 $\mathrm{\mu }\mathrm{g}$ C; between runs, the deviation of the standards with different carbon content (> 5 $\mathrm{\mu }\mathrm{g}$ C, $n\ge \mathrm{10}$) was < 0.17 ‰. The accuracy is estimated to be better than 0.5 ‰ by comparing the calibrated ${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}$ results from the GasBench II and the blank-corrected isotopic ratios from EA. Isotope results tested by GasBench II are slightly lower compared to the results of EA. The ambient aerosol filters were tested repeatedly to evaluate the reproducibility of the ambient samples. The standard deviation of the WSOC concentrations and the isotope results of the repeated ambient samples are $\mathrm{0.25}\pm \mathrm{0.04}$ $\mathrm{\mu }\mathrm{g}$ C ($n\ge \mathrm{3}$) and $\mathrm{0.14}\pm \mathrm{0.07}$ ‰ ($n\ge \mathrm{3}$), respectively. To conclude, the method presented is considered to be precise and accurate in the detection of low abundance WSOC and isotopes in aerosol samples.

To test the applicability of this method in measuring the atmospheric WSOC, the ambient aerosol samples collected in Nanjing were analyzed. And the WSOC concentrations were measured with a TOC analyzer (Shimadzu) for comparison. Figure 4 shows the scatter plot of WSOC concentrations measured with the two peripherals (TOC analyzer and GasBench II–IRMS). The strong correlation ($R^{\mathrm{2}}=\mathrm{0.95}$, $p\mathit{<}\mathrm{0.01}$) and the slope (0.97) demonstrate the reliability of measuring WSOC with the presented method. It suggests complete oxidation of WSOC in aerosol samples, which means no significant carbon isotope fractionation happens during the preparation. Moreover, the ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ values (between $-\mathrm{26.24}$ ‰ to $-\mathrm{23.35}$ ‰) of ambient aerosols are close to the published data (from $-\mathrm{26.5}$ ‰ to $-\mathrm{17.5}$ ‰) (Kirillova et al., 2013, 2014). In that case, the ${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}$ values determined from this method are considered to be effective for ambient WSOC.

4 Sources and atmospheric processes of WSOC

4.1 Temporal variation

Time series of PM${}_{\mathrm{2.5}}$, ${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}$ values, chemical tracers and meteorological data observed at the sampling site during the studied period are illustrated in Fig. 5. WSOC ranges from 3.0 to 32.0 $\mathrm{\mu }\mathrm{g}$ m${}^{-\mathrm{3}}$, occupying $\mathrm{49}\pm \mathrm{10}$ % of total carbon in PM${}_{\mathrm{2.5}}$. The stable carbon isotopes of WSOC and TC vary between $-\mathrm{26.24}\mathrm{‰}$ to $-\mathrm{23.35}$ ‰ and $-\mathrm{26.83}$ ‰ to $-\mathrm{22.25}$ ‰, respectively. ${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}$ values shift over 2 ‰ in 24 h and over 1 ‰ in 3 h, which was not able to be captured in lower time-resolution samples (e.g., 12 or 24 h). In that case, this data set can be interpreted with more detailed information about the WSOC sources and the atmospheric processes. Biomass burning tracer (nss-${\mathrm{K}}^{+}$), dust tracer (${\mathrm{Ca}}^{\mathrm{2}+}$), MODIS fire spots and air mass trajectories were analyzed to investigate the potential sources of WSOC. Nss-${\mathrm{K}}^{+}$ is used as a proxy of biomass burning (Zhang et al., 2013). Nss-${\mathrm{K}}^{+}$ concentrations are evaluated from ${\mathrm{Na}}^{+}$ concentrations in the samples according to their respective ratios (${\mathrm{K}}^{+}/{\mathrm{Na}}^{+}$ $=\mathrm{0.037}$ $w/w$) in seawater (Osada et al., 2007):

9$${\displaystyle }\mathrm{nss}-{\mathrm{K}}^{+}=\left[{\mathrm{K}}^{+}\right]-\mathrm{0.037}\cdot \left[{\mathrm{Na}}^{+}\right],$$ where [${\mathrm{K}}^{+}$] and [${\mathrm{Na}}^{+}$] are the total mass concentrations of ${\mathrm{K}}^{+}$ and ${\mathrm{Na}}^{+}$ of the aerosol samples. The concentration of nss-${\mathrm{K}}^{+}$ ranges from 0.16 to 6.70 $\mathrm{\mu }\mathrm{g}$ m${}^{-\mathrm{3}}$ with an average of 1.31 $\mathrm{\mu }\mathrm{g}$ m${}^{-\mathrm{3}}$. The high concentrations and the intense increase on 24 January indicate a significant biomass burning event and will be discussed later.

As shown in Fig. 5, ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{TC}}$ and ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ show similar patterns during the sampling period. In general, ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{TC}}$ is slightly lower than ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$, and the trend is also observed elsewhere (Fisseha et al., 2009). The difference is related to the sources and the atmospheric processes during the formation and transformation of carbonaceous particles in the atmosphere. The C${}_{\mathrm{4}}$ plant biomass burning and the marine organic materials are the sources with relatively enriched ${}^{\mathrm{13}}\mathrm{C}$. Smith and Epstein (1971) suggest that C${}_{\mathrm{4}}$ plants have a mean ${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}$ isotope signature of $-\mathrm{13}$ ‰. And the isotope composition of carbon emitted from phytoplankton, an example of primary marine aerosol, is about $-\mathrm{22}$ ‰ to $-\mathrm{18}$ ‰ (Miyazaki et al., 2011). However, January is not a specific time period for the growing or combustion of C${}_{\mathrm{4}}$ plants in eastern China, indicating a small possibility of C${}_{\mathrm{4}}$ plant biomass burning as a major source of WSOC aerosols. In addition, both WSOC and non-WSOC components can be emitted from biomass burning; thus the C${}_{\mathrm{4}}$ plant combustion would generally result in the enrichment of ${}^{\mathrm{13}}\mathrm{C}$ in both TC and WSOC. The air parcel transported from marine areas normally has little effect on the aerosols during winter in Nanjing (Qin et al., 2016), suggesting the negligible contribution of marine emissions to WSOC during the sampling period. Therefore, the WSOC sources with higher isotope signatures (compared with non-WSOC sources) are not able to explain the higher values of ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ over ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{TC}}$.

Figure 5

Time series of PM${}_{\mathrm{2.5}}$, WSOC, precipitation, ${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}$ values, nss-${\mathrm{K}}^{+}$, ${\mathrm{Ca}}^{\mathrm{2}+}$, TC, wind speed and wind direction at the sampling site during the studied period. The time period framed with the rectangles is defined as Episode 1 (green), Episode 2 (red) and Episode 3 (orange). The dotted line in 5e is the average value of nss-${\mathrm{K}}^{+}$ during the studied period. The high concentration and intense increase in nss-${\mathrm{K}}^{+}$ in Episode 2 indicate a significant biomass burning (BB) event and is marked with “BB event” in (e). The similar trends of ${\mathrm{Ca}}^{\mathrm{2}+}$ and TC suggest a dust event in Episode 3.

[Figure omitted. See PDF]

Apart from the sources, the secondary formation (Saarikoski et al., 2008; Jimenez et al., 2009; Hecobian et al., 2010) of WSOC is reported to affect the isotope compositions. Precursors like VOCs can be oxidized with the hydroxyl radicals and ozone to produce WSOC in the atmosphere (Pathak et al., 2011). Laboratory and field studies demonstrate that the lighter isotopes have the priority to be oxidized and produce particulates with lower isotopic ratios. For example, the oxidation of VOCs in the atmosphere would result in the ${}^{\mathrm{13}}\mathrm{C}$ depletion in the products and the ${}^{\mathrm{13}}\mathrm{C}$ enrichment in the residual VOCs (Rudolph et al., 2002). In other words, the secondary formation tends to lower the ${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}$ value of ambient WSOC; thus the secondary formation could not explain the ${}^{\mathrm{13}}\mathrm{C}$ enrichment in WSOC compared to TC.

Studies demonstrate that the photochemical aging process during the long-range transport causes significant enrichment in ${}^{\mathrm{13}}\mathrm{C}$ (Aggarwal and Kawamura, 2008; Wang et al., 2010). The isotope fractionation is up to 3 ‰–7 ‰ of the residual during the photolysis of oxalic acid, a dominant species in WSOC aerosols (Pathak et al., 2011). Due to the hydrophilic property, WSOC is associated with the aerosol aging processes. The WSOC $/$ OC ratio is normally considered to represent the aging status of aerosol samples (Agarwal et al., 2010; Pathak et al., 2011), and it increases with the photochemical aging process. The ratio of WSOC $/$ OC is $\mathrm{0.67}\pm \mathrm{0.12}$ (Fig. S5) in this study, which is higher than the aged aerosols with WSOC $/$ OC $=\mathrm{0.41}$ reported elsewhere (Huang et al., 2012). The high ratio of WSOC $/$ OC indicates aged aerosols during the sampling period. Thus the photochemical aging process could partially explain the reason of higher values of ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ (compared with ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{TC}}$).

According to the principle of mass balance, ${}^{\mathrm{13}}\mathrm{C}$-depleted sources of non-WSOC can also result in the depletion of ${}^{\mathrm{13}}\mathrm{C}$ in TC. TC consists of OC, EC and carbonate carbon (CC) (Huang et al., 2006), and OC can be divided into WSOC and water-insoluble OC (WIOC) according to the hydrophilic character (Eq. 10). In most circumstances, CC is negligible to the amount of TC in PM${}_{\mathrm{2.5}}$ (Ten Brink et al., 2004; Huang et al., 2006). Thus non-WSOC component can be presented as Eq. (11). $$\begin{array}{}\text{10}& {\displaystyle }& {\displaystyle }\mathrm{TC}=\mathrm{OC}+\mathrm{EC}+\mathrm{CC}=\mathrm{WSOC}+\mathrm{WIOC}+\mathrm{EC}+\mathrm{CC}\text{11}& {\displaystyle }& {\displaystyle }\mathrm{TC}-\mathrm{WSOC}=\mathrm{WIOC}+\mathrm{EC}\end{array}$$ WIOC and EC generally originate from primary emissions (Park et al., 2013; Y. L. Zhang et al., 2014), and the ${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}$ values are better in representing their sources. In that case, the ${}^{\mathrm{13}}\mathrm{C}$-depleted source, which only contributes to non-WSOC components, such as WIOC emitted from the vegetation, is likely to be another reason of ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{TC}}$ depletion during the sampling period.

4.2 Three episodes

During the sampling period, three significant haze events (Episode 1, Episode 2 and Episode 3) are observed in Nanjing. These three episodes show different tendencies of ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ variation during the accumulation of WSOC aerosols (see Fig. 5). Episode 1 and Episode 2 are compared here due to the distinct ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ trends with WSOC accumulation. In Episode 3, ${}^{\mathrm{13}}\mathrm{C}$ is found to be enriched in TC compared to WSOC (${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ < ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{TC}}$, $p\mathit{<}\mathrm{0.01}$), in contrast to the trend of isotope compositions during other periods (${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ > ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{TC}}$, $p\mathit{<}\mathrm{0.01}$).

4.2.1 Episode 1

As for Episode 1, the ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ values increase with the mass concentrations of WSOC ($r=\mathrm{0.84}$, $p\mathit{<}\mathrm{0.001}$; see Fig. 6d), indicating that the sampling site is impacted by ${}^{\mathrm{13}}\mathrm{C}$-enriched WSOC sources and/or photochemical-aged aerosols. As shown in Fig. 6a, air mass trajectories of WSOC with higher ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ values (> 24 ‰) originate mainly from northern China, and the northerly wind prevails at this site (Fig. 5g). During the long-range transport, the studied WSOC mass concentration increases with the ${}^{\mathrm{13}}\mathrm{C}$ enrichment of WSOC due to the isotope fractionation in the photochemical aging process. This is supported by the increasing ratio of WSOC $/$ OC (from 0.73 to 0.91) in Episode 1 (Fig. S5).

Figure 6

The 48 h air mass back trajectories at 500 m and MODIS fire maps for the three episodes and the corresponding relationship between WSOC and ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$. Panels (a–c) represent the back trajectories and the fire maps of Episode 1, Episode 2 and Episode 3, separately. The colors of the back trajectories are marked according to the time of the specific trajectory. Red points represent the fire spots in each episode obtained from the Fire Information for Resource Management System (FIRMS) derived from the Moderate Resolution Imaging Spectroradiometer. The ranges of the ${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}$ values of the back trajectories are labeled: the marked isotopic ratios are the ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ values (for a and b) and the ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{TC}}$ values (for c). Panels (d–f) are the correlation between WSOC and ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ in each episode.

[Figure omitted. See PDF]

According to the higher isotopes (${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ > $-\mathrm{24}$ ‰ ) and the corresponding trajectories (Fig. 6a), C${}_{\mathrm{4}}$ plant biomass burning (${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}$ $\sim -\mathrm{12}$ ‰; Martinelli et al., 2002; Sousa Moura et al., 2008) and coal combustion (${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}$ $\sim -\mathrm{24.9}$ ‰ to $-\mathrm{21}$ ‰; Cao et al., 2011) are considered to be possible sources of WSOC. Nss-${\mathrm{K}}^{+}$ mostly originates from plant combustion (Zhang et al., 2013) and is analyzed as a proxy of biomass burning. However, during this period the nss-${\mathrm{K}}^{+}$ level ($\mathrm{0.56}\pm \mathrm{0.41}$ $\mathrm{\mu }\mathrm{g}$ m${}^{-\mathrm{3}}$) was not significantly increased and was generally lower than the average value (1.31 $\mathrm{\mu }\mathrm{g}$ m${}^{-\mathrm{3}}$, Fig. 5e), indicating that the C${}_{\mathrm{4}}$ plant biomass burning is not a major source of WSOC. The main crops growing in northern China are mainly C${}_{\mathrm{3}}$ plants such as wheat and rice instead of C${}_{\mathrm{4}}$ plants during the sampling period (Chen et al., 2004). And the biomass burning contribution of C${}_{\mathrm{3}}$ plants would even lower the ${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}$ values of WSOC. Furthermore, there are only few MODIS fire spots along with the trajectories from northern China (Fig. 6a). In that case, open-field biomass burning is not considered as a major source of WSOC at the sampling site during Episode 1.

Furthermore, the WSOC mass concentrations and the ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ values decrease synchronously with the change of the wind direction (from north to southeast) after Episode 1. The southeast wind breaks the continuous transport of WSOC from northern China. And the relatively lower ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ values are then observed, suggesting a regional isotope signal of WSOC without the substantial aging. Besides, the WSOC $/$ OC ratio declines obviously with the isotope after Episode 1 (Fig. S5), indicating less contribution of aged aerosols to the sampling site. Therefore, the elevated ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ values with the increased WSOC mass concentrations in Episode 1 are mainly affected by the aged aerosols transported from northern China.

4.2.2 Episode 2

The ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ values show an opposite trend with WSOC mass concentrations ($r=-\mathrm{0.54}$, $p\mathit{<}\mathrm{0.01}$; see Fig. 6e) in Episode 2. At the beginning of Episode 2 (22 January), the sampling site was mainly affected by the air mass from the north of Nanjing, and the WSOC displayed relatively higher ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ values at the same time (Fig. 6b). After 22 January, the shift of the wind direction and the air mass trajectories correspond well to the decline of the ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ values (Fig. 6b). The large amount of fire spots in the potential source regions suggests the significant impact of open-field biomass burning. It should be noted that the stable carbon isotope composition of C${}_{\mathrm{3}}$ plant combustion is relatively low (i.e., ${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}$ $\sim -\mathrm{27}$ ‰; Martinelli et al., 2002; Sousa Moura et al., 2008). The ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ values decrease and the WSOC mass concentrations peak to the maximum when the air mass travels throughout the regions with a great many hot spots. The concentration of nss-${\mathrm{K}}^{+}$ has a positive correlation with WSOC concentration ($r=\mathrm{0.82}$, $p\mathit{<}\mathrm{0.001}$) and a negative correlation with ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ ($r=-\mathrm{0.45}$, $p\mathit{<}\mathrm{0.05}$) during Episode 2. And the concentration of nss-${\mathrm{K}}^{+}$ increases up to 6.70 $\mathrm{\mu }\mathrm{g}$ m${}^{-\mathrm{3}}$, about 5 times the average value, indicating a significant biomass burning contribution (Fig. 5e). The decrease in the ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ values and the increase in the biomass burning tracers (i.e., nss-${\mathrm{K}}^{+}$) suggest that the biomass burning emission is a major contributing factor to WSOC aerosols. Also, the WSOC $/$ OC ratio declines from 0.88 to 0.53 (Fig. S5), indicating that the increased WSOC is rather from fresh biomass-burning aerosols without a substantial aging process.

4.2.3 Episode 3

The ${}^{\mathrm{13}}\mathrm{C}$ is clearly enriched ($p\mathit{<}\mathrm{0.01}$) in TC ($-\mathrm{23.5}\pm \mathrm{0.43}$ ‰) compared to WSOC ($-\mathrm{25.17}\pm \mathrm{1.08}$ ‰) during Episode 3 (see Fig. 6f). This might be related to a ${}^{\mathrm{13}}\mathrm{C}$-enriched source and/or the aging process of the non-WSOC fraction in TC. The non-WSOC fraction consists mainly of WIOC, EC and carbonate carbon (CC). Among these carbonaceous species, carbonate carbon (CC) exhibits much higher ${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}$ values than EC and OC (Kawamura et al., 2004). CC could be a significant fraction of dust aerosols, even though it is a very small part of TC in PM${}_{\mathrm{2.5}}$ in most cases.

To study the dust contribution in Episode 3, ${\mathrm{Ca}}^{\mathrm{2}+}$ is determined as an indicator of dust (Jankowski et al., 2008; Huang et al., 2010). ${\mathrm{Ca}}^{\mathrm{2}+}$ and TC show similar patterns ($R^{\mathrm{2}}=\mathrm{0.84}$, $p\mathit{<}\mathrm{0.01}$), indicating dust origins in this period. The argument is also supported by the 48 h backward trajectory analysis (Fig. 6c). It shows that the air mass mainly originates from a semi-arid region, Mongolia. The photochemical aging of dust aerosols during the long-range transport from Mongolia to Nanjing could possibly promote the ${}^{\mathrm{13}}\mathrm{C}$ enrichment. Briefly speaking, the enrichment of ${}^{\mathrm{13}}\mathrm{C}$ in TC over WSOC is due to a dust event transported to the studied site.

According to the mass balance, the isotopic ratio of TC affected by CC in the dust aerosols can be expressed as follows:

12$${\displaystyle }{\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{TC}}=f_{\mathrm{cc}}\times {\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{CC}}+\left(\mathrm{1}-f_{\mathrm{cc}}\right)\times {\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{NC}},$$ where the ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{TC}}$, ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{CC}}$ and ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{NC}}$ are the measured stable carbon isotope of TC, the isotopic ratio of CC in dust aerosols and the isotope composition of non-CC fractions. The $f_{\mathrm{cc}}$ represents the CC contribution to TC. The CC contribution during Episode 3 is roughly estimated based on a few assumptions: (1) the increase in TC and ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{TC}}$ is only affected by the dust origin, (2) the average value of ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{TC}}$ ($-\mathrm{25}$ ‰) during the studied period (except Episode 3) is taken as the value of ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{NC}}$, and (3) ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{CC}}$ $=\mathrm{0.3}$ ‰ in dust sources (Kawamura et al., 2004). With these considerations, CC is estimated to contribute up to 10 % of TC according to the Eq. (12).

5 Conclusions

An optimized method for the determination of WSOC mass concentrations and ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ values in aerosol samples with GasBench II–IRMS is presented. A two-step correction is applied to correct the blank contribution and to calibrate the isotope results. The procedural blank is estimated to be 0.5 $\mathrm{\mu }\mathrm{g}$ C with isotope composition of $-\mathrm{27.43}$ ‰. The detection limit is demonstrated to be 5 $\mathrm{\mu }\mathrm{g}$ C according to the measurement of working standards with various carbon contents. The method yields a high recovery of the standards ($\mathrm{97}\pm \mathrm{6}$ %) and ambient samples ($\mathrm{99}\pm \mathrm{10}$ %). According to the high recoveries, the isotope fractionation during the pretreatment tended to be negligible. The precision and the accuracy is better than 0.17 ‰ and 0.5 ‰, respectively. WSOC concentrations determined with this optimized method are consistent ($R^{\mathrm{2}}=\mathrm{0.95}$) with the results of the TOC analyzer. Compared with the previous methods, the optimized method presented in this study is more precise and accurate and requires less time-consuming pretreatment.

The method presented was applied to analyze the ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ of the high time-resolution aerosol samples collected during a severe winter haze in eastern China. WSOC ranged from 3.0 to 32.0 $\mathrm{\mu }\mathrm{g}$ m${}^{-\mathrm{3}}$, and ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ varied between $-\mathrm{26.24}$ ‰ and $-\mathrm{23.35}$ ‰. ${}^{\mathrm{13}}\mathrm{C}$ was more enriched in WSOC than TC in the majority of the sampling period, indicating aged aerosols and/or ${}^{\mathrm{13}}\mathrm{C}$-depleted primary sources of non-WSOC component. Three haze events (Episode 1, Episode 2 and Episode 3) are identified with different tendencies of ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ during the accumulation of WSOC aerosols. Similar patterns of the WSOC concentrations and the ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ values in Episode 1 are demonstrated to be affected by the air mass transported from northern China. The increase in ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ indicates that the WSOC aerosols from the studied site are subject to a substantial photochemical aging process during the long-range transport. The contrasting trend of the WSOC and ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ values in Episode 2 is interpreted as the contribution of regional C${}_{\mathrm{3}}$ plant biomass burning sources. In Episode 3, the heavier isotope (${}^{\mathrm{13}}\mathrm{C}$) is clearly enriched in total carbon (TC) compared to WSOC fraction due to the dust contribution.

The optimized method is demonstrated to be accurate and precise in the detection of the WSOC mass concentration and its isotope compositions (${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$) in aerosols. Our results indicate that the high time-resolved measurement of ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{-\mathrm{WSOC}}$ can be used to distinguish different atmospheric processes such as photochemical aging and aerosol sources (e.g., biomass burning and dust). However, a quantitative understanding of sources and formation processes of WSOC aerosols is still a great challenge. To reduce the knowledge gaps, a combination of multiple methodologies is needed in future studies, such as high time-resolved measurement of radiocarbon (${}^{\mathrm{14}}\mathrm{C}$) and stable carbon isotope compositions (${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}$), and the real-time measurement of chemical compositions (e.g., aerosol mass spectrometers, AMS or thermal desorption aerosol gas chromatograph-AMS).

Data availability

Data are available from the corresponding author on request ([email protected]).

The supplement related to this article is available online at: https://doi.org/10.5194/acp-19-11071-2019-supplement.

Author contributions

YLZ conceived and designed the study; YLZ, FC and WZ designed the experimental strategy; WZ, YZ and YX performed the sampling and isotope measurements; YLZ and WZ analyzed the experimental data; YLZ and WZ proposed the hypotheses; WZ wrote the paper with YCL, and all other co-authors contributed to writing.

Competing interests

The authors declare that they have no conflict of interest.

Special issue statement

This article is part of the special issue “Multiphase chemistry of secondary aerosol formation under severe haze”. It is not associated with a conference.

Financial support

This study is supported by the National Natural Science Foundation of China (grant nos. 91644103, 41977305, 41761144056, and 41603104), the National Key Research and Development Program of China (grant no. 2017YFC0212704) and the Provincial Natural Science Foundation of Jiangsu (grant no. BK20180040).

Review statement

This paper was edited by Jingkun Jiang and reviewed by two anonymous referees.