Molecular characterization of organic aerosol in the Himalayas: insight from ultra-high-resolution mass spectrometry

An increased trend in aerosol concentration has been observed in the Himalayas in recent years, but the understanding of the chemical composition and sources of aerosol remains poorly understood. In this study, molecular chemical composition of water-soluble organic matter (WSOM) from two filter samples collected during two high aerosol loading periods (denoted as P1 and P2) at a high-altitude station (Qomolangma Station, QOMS; 4276 m a.s.l.) in the northern Himalayas was identified using electrospray ionization Fourier transform ion cyclotron resonance mass spectrometry (ESI-FTICR MS). More than 4000 molecular formulas were identified in each filter sample which were classified into two compound groups (CHO and CHON) based on their elemental composition, with both accounting for nearly equal contributions in number (45 %–55 %). The relative abundance weighted mole ratio of O∕C_w for P1 and P2 was 0.43 and 0.39, respectively, and the weighted double bond equivalents (DBE_w), an index for the saturation of organic molecules, were 7.12 and 7.87, respectively. Although the O∕C_w mole ratio was comparable for CHO and CHON compounds, the DBE_w was significantly higher in CHON compounds than CHO compounds. More than 50 % molecular formulas in the Van Krevelen (VK) diagram (H∕C vs. O∕C) were located in 1–1.5 (H∕C) and 0.2–0.6 (O∕C) regions, suggesting potential lignin-like compounds. The distributions of CHO and CHON compounds in the VK diagram, DBE vs. number of C atoms, and other diagnostic diagrams showed high similarities among each other, suggesting their similar source and/or atmospheric processes. Many formulas formed from biogenic volatile organic compounds (e.g., ozonolysis of α-pinene products) and biomass-burning-emitted compounds (e.g., phenolic compounds) were found in the WSOM, suggesting the important contribution of these two sources in the Himalayas. The high DBE and high fraction of nitrogen-containing aerosol can potentially impact aerosol light absorption in this remote region. Further comprehensive study is needed due to the complexity of organic aerosol and limited molecular number identified in this study.

1 Introduction

Relatively high aerosol concentration events have been frequently observed over the Himalayas during the pre-monsoon period (March to June) (Bonasoni et al., 2010). The aerosol plumes are originated from the southern regions of the Himalayas such as northwestern India and/or the Indo-Gangetic Plain based on air mass back trajectory analysis and satellite observation (Liu et al., 2008; Lu et al., 2012; Lüthi et al., 2015). Due to increased consumption on fuels (including biofuels and fossil fuels) by industry and residents in recent decades, air pollution has been a serious issue in south Asia (Gustafsson et al., 2009). Under favorable atmospheric circulation, air pollutants emitted or formed in these regions can be transported fast to the Himalayas and Tibetan Plateau (HTP; Xia et al., 2011).

Enhanced aerosol concentration for the remote region of the HTP is thought to have many negative climate and environment effects. For example, the transported aerosol could heat the air in the higher layer of the troposphere over the HTP and impact the monsoon system of south Asia and accelerate the melting of glaciers in the Himalayas (Lau et al., 2006; Ramanathan et al., 2007). This heating effect is predominantly from light-absorbing particular aerosol (LAPA) such as black carbon (BC) and brown carbon which are types of organic aerosol (OA; Ram et al., 2010; Zhang et al., 2015, 2017). BC is from incomplete combustion and dominates the absorption of LAPA; brown carbon can originate from primary emission and/or secondary processes, and has made an increasing contribution (up to ∼20 %) to light absorption in recent years (Laskin et al., 2015, and reference therein). Due to the light absorption of brown carbon being strongly dependent on their molecular structure, light-absorbing compounds at a molecular level have been explored during recent years, and it was found that nitrogen-containing compositions are important brown carbon compounds (Lin et al., 2016, 2017). Many studies show that biomass burning is an important source of brown carbon (e.g., Saleh et al., 2014; Washenfelder et al., 2015), which is very common in developing regions in the southern Himalayas. In comparison with other regions, high elevation and mixed biomass fuels in the southern Himalayas could make the evolution and chemical composition of OA from biomass burning emission more complicated (Stockwell et al., 2016; Fleming et al., 2018; Jayarathne et al., 2018).

The details on the molecular composition of OA are important for understanding the sources and chemical evolution of OA (Laskin et al., 2018). Previous studies conducted in the HTP have focused on a limited number of molecular markers such as organic acids, which are closely related to biomass burning emission (Cong et al., 2015), and some toxicology species such as polycyclic aromatic hydrocarbons (PAHs) and persistent organic pollutants (POPs), which are related to anthropogenic activities (Wang et al., 2015, 2016). In addition, online measurement using the Aerodyne high-resolution time-of-flight aerosol mass spectrometer (HR-ToF-AMS) has provided more details on the chemistry and sources of OA with a high time resolution (Xu et al., 2018). However, different instruments have limitations on OA detection, and an ultra-high mass resolution of mass spectrometry which can identify many molecular formulas is lacking.

Fourier transform-ion cyclotron resonance mass spectrometry (FTICR MS) coupled with a soft ionization source, such as electrospray ionization (ESI), can be used to identify the individual molecular formula of extremely complex mixtures because of its ultra-high resolution and mass accuracy (Mazzoleni et al., 2010). Similar methods have been used for the identification of components in aqueous secondary OA (SOA) and in ambient samples, and allow the identification and separation of thousands of compounds in a sample (e.g., Mazzoleni et al., 2010; Altieri et al., 2012; Mead et al., 2013). Types of methods such as double bond equivalents (DBEs), elemental ratios, and Kendrick mass defects (KMDs) can be applied to deduce the chemical characterization of obtained molecular formulas. In this study, we focus on the molecular composition of water-soluble organic compound in fine particle aerosol in the Himalayas using ESI-FTICR MS and evaluate the sources, chemical processing, and potential impact of aerosol in this region.

2 Methodology

2.1 Aerosol sampling

A field study was conducted at the Qomolangma Station (QOMS; 28.36^∘ N, 86.95^∘ E; 4276 m a.s.l.) located at the toe of Mt. Qomolangma from 12 April to 12 May 2016 using a suite of online instruments (X. Zhang et al., 2018), including a HR-ToF-AMS (Aerodyne Research Inc., Billerica, MA, USA) for 5 min size-resolved chemical compositions (organics, sulfate, nitrate, ammonium, and chloride) of non-refractory submicron particulate matter (NR-PM₁) and a photoacoustic extinctiometer (PAX; DMT Inc., Boulder, CO, USA) for BC mass concentration. In addition, a low-volume (16.7 L min⁻¹) particular matter (PM) sampler (BGI, USA, model PQ 200) with an Aerodyne diameter cutoff of 2.5 µm at the inlet was used to collect PM_2.5 filter samples on pre-baked quartz fiber filters (47 mm, Pall Life Science, NY, USA). Due to the low aerosol loading at this remote region, a 2-day sampling strategy was adapted for each filter collection from 08:00 to 07:45 the following day (local time). A total of 18 filter samples were collected during the field study with three procedure blanks which were used to assess potential contamination during sampling and transportation. The sampling air volume ranged from 35.1 to 48.1 m³ at ambient conditions. Two samples collected during 25–27 April (P1) and 29 April–1 May (P2), respectively, were used in this study due to the relatively higher aerosol loading based on HR-ToF-AMS results (Sect. 3.1) and distinct particulate matter on the filter. One procedure blank was also adopted in this study like that of aerosol samples to subtract the potential background.

2.2 Chemical analysis

For FTICR MS analysis, filter samples were extracted in 20 mL Milli-Q water in an ultrasonic bath for 30 min and filtered using 0.45 µm pore-size Acrodisc syringe filters to remove water-insoluble matter (Pall Science, USA). The sample tubes were immersed in the mixture of ice–water during ultrasonic extraction to prevent potential chemical reaction. Prior to FTICR MS analysis, the extraction was concentrated and purified using PPL (Agilent Bond Elut-PPL cartridges, 500 mg, 6 mL) solid-phase extraction (SPE) cartridges for water-soluble organic matter (WSOM) to avoid possible ESI artifacts. PPL cartridges generally have the best properties for WSOM enrichment for subsequent FTICR MS analysis (Raeke et al., 2016). In addition, we control the concentration of SPE effluent to be ∼0.2 mg mL⁻¹, which was not too concentrated for artifact adducts. Note that through SPE cartridges, the most hydrophilic compounds, such as inorganic ions, and low-molecular-weight organic molecules, such as organic acids and sugars, were removed, whereas the relatively hydrophobic fraction was retained. Details on the SPE method using PPL cartridges and analysis using FT-ICR MS can be found in our previous paper (Feng et al., 2016). Briefly, the mass spectrometry analyses of these samples were performed using a SolariX XR FTICR MS (Bruker Daltonik GmbH, Bremen, Germany) equipped with a 9.4 T refrigerated actively shielded superconducting magnet (Bruker Biospin, Wissembourg, France) and a Paracell analyzer cell. The samples were ionized in positive ion modes using the ESI ion source (Bruker Daltonik GmbH, Bremen, Germany). A typical mass-resolving power of >400 000 was achieved at m∕z 400, with an absolute mass error of <0.5 ppm. The ions detected in filter blank were subtracted and molecular formulas in the samples were assigned to all ions with signal-to-noise ratios of greater than 10, with a mass tolerance of ±1.5 ppm, using custom software. Molecular formulas with their maximum numbers of atoms were defined as 30 ¹²C, 60 ¹H, 20 ¹⁶O, 3 ¹⁴N, 1 ³²S, 1 ¹³C, 1 ¹⁸O, and 1 ³⁴S. Identified formulas with H∕C, O∕C, N∕C, S∕C, and DBE∕C, ranging 0.3–3.0, 0–3, 0–0.5, 0–0.2, and 0–1, were selected, and formulas containing isotopomers (i.e., ¹³C, ¹⁸O, or ³⁴S) were not considered. Compounds were detected as either sodium adducts, $[\mathrm{M}+\mathrm{Na}{]}^{+}$, or protonated species, $[\mathrm{M}+\mathrm{H}{]}^{+}$. Although ammonium could also readily be adduct to assist in the formation of positive ions, this possibility in our sample was because of the low concentration of ammonium (Fig. 1). We report all detected compounds as neutral species, unless stated otherwise. Note that the ESI+ mode more easily detects basic functional group compounds, and the reported organic molecules here are only part of organic aerosol which are biased ionized in ESI+ mode. In addition, highly functionalized compounds could be detected in both positive and negative modes (Lin et al., 2012).

Figure 1 The average mass concentration (mass avg.) and chemical composition of PM₁ during P1 (a) and P2 (b) periods, respectively, measured by the HR-ToF-AMS and PAX. Note that the compounds of PM₁ include sulfate, nitrate, ammonium, chloride, BC (black carbon), organics, BBOA (biomass-burning-emitted OA), NOA (nitrogen-containing OA), and OOA (oxidized oxygenated OA).

2.3 Data processing

The assigned molecular formulas were examined using the Van Krevelen diagram (Wu et al., 2004), DBEs, KMD series, and aromatic indices (AI_mod). The O∕C and H∕C ratios were calculated by dividing the number of O and H atoms, respectively, by the number of C atoms in a formula. DBE analysis was used to determine the number of rings and double bonds in a molecule. The DBE was calculated using Eq. (1):

$$\begin{array}{}\text{(1)}& \text{DBE}=\mathrm{1}+c-h/\mathrm{2}+n/\mathrm{2},\end{array}$$

where c, h, and n are the numbers of C, H, and N atoms, respectively, in the formula.

The weighted DBE (DBE_w), O∕C (O∕C_w), and H∕C (H∕C_w) were calculated using Eq. (2):

$$\begin{array}{}\text{(2)}& X_{\mathrm{w}}=\sum (w_i\cdot X_i)/\sum w_i,\end{array}$$

where X_i and w_i are the parameters above and the relative intensity (RI) is for each individual formula, i.

The Kendrick mass (KM) and KMD for CH₂ series, used to search for potential oligomeric units (Hughey et al., 2001), were calculated using Eqs. (3) and (4):

$$\begin{array}{}\text{(3)}& {\displaystyle }& {\displaystyle }\text{KM}=\text{observed mass}\times \mathrm{14}/\mathrm{14.01565},\text{(4)}& {\displaystyle }& {\displaystyle }\text{KMD}=\text{NM}-\text{KM},\end{array}$$

where 14 is the nominal mass (NM) of CH₂, 14.01565 is the exact mass of CH₂, and NM is KM rounded to the nearest integer.

Furthermore, a second-order mass defect analysis using the base units of CH₂ and H₂ was applied following the method described in Roach et al. (2011), which greatly simplified the visualization of complex mass spectra.

AI_mod is a measure of the probable aromaticity of a molecule assuming that half the O atoms are double bonded and half have only σ bonds (Koch and Dittmar, 2006). AI_mod was calculated using Eq. (5):

$$\begin{array}{}\text{(5)}& {\mathrm{AI}}_{\mathrm{mod}}=(\mathrm{1}+c-\mathrm{0.5}o-\mathrm{0.5}h-\mathrm{0.5}n)/(c-\mathrm{0.5}o-n),\end{array}$$

where c, o, and h are the numbers of C, O, H, and N atoms in the formula. AI_mod ranges from 0 for a purely aliphatic compound to higher values for compounds with more double bonds and that are more aromatic.

Figure 2 The air mass back trajectory analysis using the HYSPLIT model (Draxler and Hess, 1998) during (a) P1 and (b) P2. The air mass trajectories were recovered back to 72 at a 1 h interval from the sampling site (QOMS) at 1000 m above ground level using a 1^∘ resolution Global Data Assimilation System (GDAS) dataset (https://ready.arl.noaa.gov/gdas1.php; last access: 10 September 2017). The cluster analysis for these trajectories was completed based on the directions of the trajectories (angle distance) and colored according to air pressure (vertical profile). The firespot observed from MODIS (https://firms.modaps.eosdis.nasa.gov; last access: 10 September 2017) and the average wind rose plot colored by wind speed (WS) for during each filter sampling period are also shown. The firespot is sized by fire radiative power.

3 Results and discussions

3.1 Chemical characterization of PM₁ during P1 and P2 measured by the HR-ToF-AMS

The average mass concentration and chemical composition measured by the HR-ToF-AMS during P1 and P2 periods are shown in Fig. 1. The mass concentrations of PM₁ were 9.2 and 10.6 µg m⁻³, respectively, which were at the high range of all filters (1.3–10.6 µg m⁻³) because of a continuous long-range transport event at the QOMS (X. Zhang et al., 2018). Due to our sample processing error, the mass concentration of filter measured gravimetrically could not be used, and thus the fractions of PM₁ to PM_2.5 are not available. However, most of WSOM in PM_2.5 is in the accumulation size mode (less than 1 µm) which could be detected by the HR-ToF-AMS (Zhang et al., 2005). The chemical composition of PM₁ during P1 and P2 was all dominated by OA (55 % and 57 %), followed by BC (26 % and 22 %), sulfate (7 % and 8 %), nitrate (5 % and 6 %), and ammonium (5 % and 6 %). The OA was comprised of biomass-burning-emitted OA (BBOA), nitrogen-containing OA (NOA), and more oxidized oxygenated OA (MO-OOA) decomposed using positive matrix factorization (PMF) analysis (Fig. 1). Details on PMF analysis can be found in X. Zhang et al. (2018). The mass contribution of BBOA was higher during P2 than P1 (32 % vs. 22 %), whereas the contribution of MO-OOA was higher during P1 (24 % vs. 16 %). The mass spectra of OA for these two filter periods were closely similar to a Poisson correlation efficiency (r) of 0.9. The elemental ratios of oxygen (O) to carbon (C) of OA were 1.04 and 0.97 for P1 and P2 periods (Improved Ambient method; Canagaratna et al., 2015), respectively, and the ratios of hydrogen (H) to C were 1.26 and 1.32 accordingly. These suggest that the OA during P2 was relatively less oxidized than that during P1 (t test, p<0.05). The six category ions (${\mathrm{C}}_x{\mathrm{H}}_y^{+}$, ${\mathrm{C}}_x{\mathrm{H}}_y{\mathrm{O}}_{\mathrm{2}}^{+}$, ${\mathrm{C}}_x{\mathrm{H}}_y{\mathrm{O}}_{\mathrm{1}}^{+}$, C_xH_yN⁺, C_xH_yO_zN⁺, and HO⁺) detected by the HR-ToF-AMS for these two filter periods were all dominated by ${\mathrm{C}}_x{\mathrm{H}}_y{\mathrm{O}}_{\mathrm{2}}^{+}$, following by ${\mathrm{C}}_x{\mathrm{H}}_y^{+}$, ${\mathrm{C}}_x{\mathrm{H}}_y{\mathrm{O}}_{\mathrm{1}}^{+}$, C_xH_yN⁺, C_xH_yO_zN⁺, and HO⁺. The air mass trajectory analyses using the hybrid single particle Lagrangian integrated trajectory (HYSPLIT) model for P1 and P2 periods show air mass mainly originated from west and southwest of the QOMS across north and northwest India where there were many firespots during these two periods (Fig. 2). The air mass during P2 was partly (13 %) transported with a low wind speed and across a short distance (less than 100 km), and could contain some fresh OA as illustrated by the higher fraction of BBOA.

Table 1 Chemical characterization of all molecular assignments detected in WSOM for P1, P2, and common ions. Relative intensity weighted (“w”) for each data subset (O∕C, H∕C, OM∕OC, DBE, and DBE∕C) is given.

Table 2 Chemical characterization of the molecular assignments detected in aerosol samples from selected studies (adapted and modified from Table 3 in Dzepina et al., 2015). Note that all values are presented as arithmetic means, which are given for convenience of comparison. The data for our study are the arithmetic means for P1 and P2.

NA: not available

3.2 The chemical characteristics of WSOM from ESI-FTICR MS

A total of 4295 and 4770 molecular formulas were identified by ESI-FTICT MS over the mass range of 100–700 Da for P1 and P2, respectively. The identified molecular formulas were grouped into two subgroups based on their elemental composition, i.e., CHO and CHON, all of which had equally important contributions (45 %–55 %) in number (Fig. 3). Note that individual species in the ESI-FTICR MS mass spectra could have many different isomeric structures, in which case the percentages only reflect the number of unique molecular formulas in each category. The mass spectra of these two samples were highly similar in the distributions of molecular formulas (Fig. 3). The average weighted element ratios of P1 and P2 were 0.43 vs. 0.39 for O∕C_w, 1.36 vs. 1.31 for H∕C_w, and 1.72 vs. 1.68 for OM∕OC_w (Table 1), suggesting a relatively higher oxidation and saturation degree for P1 than P2. These trends are consistent with the results of the HR-ToF-AMS, although the elemental ratios are different between them, which is due to the difference in the detection range of m∕z and the ionization efficiency of different mass spectrometry (ESI vs. electron impact) (Yu et al., 2016). The elemental ratios of WSOM from ESI-FTICR MS in our study are similar to those results observed in aerosol samples in remote sites using ESI-FTICR MS (e.g., 0.35–0.53 for O∕C) (Table 2). The O∕C and H∕C in Van Krevelen diagrams (Wu et al., 2004) for these two filters and the subgroup molecular formulas show similar distributions, and are all concentrated in ranges of 1.2–1.8 for H∕C and 0.3–0.7 for O∕C (Fig. 3), suggesting their similar aerosol sources and atmospheric processes. The similar distributions for these two filters are also observed in plots of KMD vs. KM and DBE vs. C (Fig. 3).

Figure 3 The combination plot for all molecular formulas of P1, P2, and common ions including the high-resolution mass spectrum, the Van Krevelen diagram, Kendrick mass defects (KMDs) vs. Kendrick mass (KM), and double bond equivalents (DBEs) vs. number of carbon atoms.

Structural information for the assigned molecular formulas is inferred from the DBE_w value which was higher for P2 than for P1 (7.87 vs. 7.12) (Table 1). Compared with other studies, the DBE_w values in our filter are relatively close to the results from biomass burning aerosol and aerosol samples from remote sites (Table 2) (Dzepina et al., 2015). The DBE_w values for each molecular subgroup were higher for CHON than for CHO (Table 1), especially for P2 (8.32 vs. 7.38), suggesting more rings and double bonds in CHON molecular formulas. The AI_mod, reflecting the minimum number of carbon–carbon double bonds and rings (Koch and Dittmar, 2006), was correspondingly higher in P2 as illustrated by its higher contribution of olefinic (75.0 % vs. 73.9 % for P2 and P1) and aromatic compounds (10.3 % vs. 7.7 % for P2 and P1) (Table 1). For aromatic compounds (AI_mod≥0.5) in P2, 52 % of them were CHON formulas (45 % for P1). Higher DBE and AI_mod values in CHON compounds suggest more unsaturated compounds in them which could contain a certain number of chromophores. The distribution of DBE vs. the carbon number of two filters showed a systematic increase in a concentrated region and a high similarity with each other. This similarity further suggests the consistent source and chemical processes for the aerosol of these two filters.

There were 3955 common molecular formulas between these two filters with the number contribution of CHO of 50.7 % and CHON of 49.3 %. These common molecular formulas accounted for 92.1 % (P1) and 82.9 % (P2) of two filters, respectively. There were 340 unique molecular formulas in P1, with 73 % being CHO compounds, whereas there were 815 unique molecular formulas in P2, with 65.3 % being CHON compounds. For more confidence in molecule assignment, we focus on the common molecular formulas detected in these two samples in the section below. Note that the mass spectrum of common ions was calculated from the average RI from two mass spectra and normalized to the highest peak.

3.3 The potential sources and formation processes

3.3.1 CHO compounds

CHO compounds have been frequently detected in ambient aerosol samples (Altieri et al., 2009b; Mazzoleni et al., 2010; Lin et al., 2012; Fleming et al., 2018), which could comprise high molecular weight humic-like substances (HULIS) or oligomers, and in primary emission or secondary formation of different aerosol sources (Mazzoleni et al., 2012; Wozniak et al., 2014; Lin et al., 2016; Cook et al., 2017). In our samples, the weighted molecular weight of CHO compounds was 361.9, with an average C atom of 19.3±5.3 per molecule; the most abundant O atoms were located in the range 5–10, with an average value of 7.8±2.9 per molecule (Fig. 4a and b). The oxygen distribution is also evidenced by the longest homologous series in the second-order mass defect analysis (CH₂-H₂), which were all O₅-O₁₀ compounds (Fig. 4d). The DBE of CHO increased with carbon number, with a DBE_w value of 6.75 (Fig. 3); the carbon-normalized DBE (DBE∕C) was 0.39±0.14. These two values were close to the results from biomass burning aerosol samples in other studies and are in a high range of published data (Table 2) (Lin et al., 2012; Mazzoleni et al., 2012), suggesting relatively high aromaticity in our samples. The carbon oxidation state (OS_C) values (Kroll et al., 2011), a useful metric for the degree of oxidation of organic species in the atmosphere, exhibited between −1 and 0 for high number density formulas, with 25 or fewer carbon atoms, suggesting that they are semi- and low-volatile organic compounds, corresponding to “fresh” (BBOA) and “aged” (LV-OOA) SOA by multistep oxidation reactions (Fig. 4c).

Figure 4 The molecular information for common CHO compounds. (a) The number frequency distribution of carbon (n_c) and oxygen (n_o). (b) The 3-D plot for n_o, n_c, and double bond equivalents (DBEs) colored by their relative intensity. (c) Scatter plot of carbon-based oxidation state (OSc) vs. n_c colored by the distribution of number of molecules. (d) The second-order mass defect analysis (MD² (CH₂, H₂) vs. MD¹ (CH₂)) using the base of CH₂ and H₂. The longest homologous series are marked with the group number and formula type.

Table 3 List of identified BVOC products and biomass-burning-emission-related compounds in this study.

There are several possible sources and chemical formation pathways for high-oxygen-containing CHO compounds. Ozonolysis of α-pinene has been found to form highly oxygenated molecules with some important products such as C₁₇H₂₆O₈ (m∕z 358) and C₁₉H₂₈O₇ (m∕z 368), which are possible esterification products of cis-pinic (C₉H₁₄O₄; m∕z 186) and diaterpenylic acid (C₈H₁₄O₅; m∕z 190) (Kristensen et al., 2013). The first three formulas were all found in our common CHO molecules with high relative abundance (Table 3). The appearance of these formulas together with high relative abundance gave us more confidence in the products of ozonolysis of α-pinene. Ozone concentration in the Himalayas during the pre-monsoon period was the highest based on online measurement at the Nepal Climate Observatory at Pyramid (NCO-P) during 2006–2008 (61±9 ppbv) (Cristofanelli et al., 2010). Biogenic volatile organic compounds could be transported from the low elevation regions in subtropical India and biogenic secondary organic aerosol has been found to be an important source in the Himalayas (Stone et al., 2012). A number of other previously reported monoterpene oxidation product formulas were also observed in our study (Table 3) (Claeys et al., 2007; Kleindienst et al., 2007; H. Zhang et al., 2018). In addition, some biomass burning aerosol markers were also found in CHO compounds. Sun et al. (2010) and Yu et al. (2014, 2016) observed that aqueous-phase oxidation of lignin produces phenol (C₆H₆O), guaiacol (C₇H₈O₂), and syringol (C₈H₁₀O₃), yielding a substantial fraction of dimers and higher oligomers, with key dimer markers identified as C₁₆H₁₈O₆ and C₁₄H₁₄O₄. The dimer markers C₁₆H₁₈O₆ and C₁₄H₁₄O₄ were also present in our sample with high RI (5.5 % and 27.5 %). The high relative intensity of these compounds indicates that fog and cloud processing of phenolic species (biomass burning aerosol) could be an important mechanism for the production of low-volatility SOA in the Himalayas. Compounds observed in biomass burning emission (cow dung and brush wood) during residential cooking in Nepal were also found in our samples (Fleming et al., 2018).

Figure 5 The molecular information for common CHON compounds. (a) The number frequency distribution of carbon (n_c) and oxygen (n_o). (b) The 3-D plot for n_o, n_c, and double bond equivalents (DBEs) colored by their relative intensity. (c) Scatter plot of carbon-based oxidation state (OSc) vs. n_c colored by the distribution of number of molecules. (d) The Van Krevelen diagram by H∕C vs. N∕C colored by number of oxygen (n_o). The different sizes of dot marker in (c) and (d) represent the 1N and 2N compounds.

Figure 6 (a) The number frequency distribution of n_o for 1N and 2N compounds and (b) the longest 10 CH₂ homologous series compounds in 1N compounds.

3.3.2 CHON compounds

The frequency distribution for n_o and n_c in CHON formulas was shown in Fig. 5a, which show peaks between 6 and 10 and 15 and 20, respectively. The DBE of CHON formulas ranged from 4 to 10, with DBE_w being 7.79 (Fig. 5b and Table 1). In the CHON class, compounds contained one or two nitrogen (1N or 2N) atoms, with 1N compounds accounting for 70.5 % and 2N for 29.5 %, respectively. Most (93.6 %) 1N compounds contained more than 3 oxygen atoms and up to 13 oxygen atoms, whereas about 62.5 % of 2N compounds contained more than 6 oxygen atoms (Fig. 6a). The average O atoms contained in each molecular formula were therefore higher for 1N compounds than 2N compounds (8.1±2.9 vs. 6.3±2.3). The ratios of O∕C_w and OS${}_{{\mathrm{C}}_{\mathrm{w}}}$ for 1N compounds were accordingly higher than those for 2N compounds (0.42 vs. 0.37 for O∕C_w and −0.48 vs. −0.54 for OS${}_{{\mathrm{C}}_{\mathrm{w}}}$, respectively), suggesting a higher oxidation state for 1N compounds (Fig. 5c). In contrast, the DBE_w and AI_mod values for 2N compounds were higher than for 1N compounds (Table 1). The higher H∕C_w for 2N compounds (Fig. 5d) suggests that 2N compounds could contain many aromatic N-heterocyclic compounds. For 1N compounds, longer and higher relative intensity CH₂ homologous series compounds were found based on the Kendrick mass defect plot (Fig. 6b); 1073 of the 1373 detected 1N compounds can be grouped into 145 homologous groups. The abundant long CH₂ homologous series in 1N compounds contained 7–10 O atoms, while that in 2N compounds contained 5–8 O atoms (Fig. 6).

Many CHON compounds could be identified in ESI+ mode, such as reduced N functional compounds and nitroaromatic compounds (Altieri et al., 2009a; Laskin et al., 2009; Lin et al., 2012; O'Brien et al., 2013; Wang et al., 2017). Laskin et al. (2009) identified the amount of N-heterocyclic alkaloid compounds from types of fresh biomass burning aerosol. Lin et al. (2012) and Wang et al. (2017) also identified many CHON compounds in fresh and aged biomass burning aerosol. Oxygenated organic nitrogen compounds in ambient aerosol (Dzepina et al., 2015), rainwater (Altieri et al., 2009a), and fog water (Mazzoleni et al., 2010) from biomass-burning-emission-influenced regions were also observed. Lin et al. (2017) found aged biomass burning aerosol in the present urban oxidants (such as NO_x) could result in a higher fraction of CHON compounds compared to the fresh biomass burning aerosol. Considering the high influence of biomass burning emission in the Himalayas (X. Zhang et al., 2018), the CHON compounds in our study were probably related to biomass burning emissions. Recent studies have proven that burning of mixed biomass fuels in Nepal could emit amounts of nitrogen species such as NH₃, NO_x, HCN, benzene, and organics, and the emission factors for these species are higher than for wood (Stockwell et al., 2016; Jayarathne et al., 2018). In addition, it is likely that smoldering burning of bio-fuels in high elevation areas is also responsible for the presence of many nitrogen-containing compounds in BBOA (Chen et al., 2010). Nitroaromatic compounds such as methylnitrocatechols (C₇H₇NO₄) are introduced to be tracers for biomass burning secondary organic aerosols (Iinuma et al., 2010). Although the C₇H₇NO₄ formula is not found in our measurement, C₁₄H₁₄N₂O₈ was found in our measurement, which was probably its dimer formula. In addition, the homologous series compounds in which C₇H₇NO₄ serves as the core molecule were also found in our samples. Some high relative abundance CHON molecular formulas identified in a recent paper from biomass burning aerosol were also found in our measurement (Table 3) (Song et al., 2018).

Besides primary emission and/or secondary formation from biomass burning emission, nitrogen-containing OA could also be formed through other chemical processes. For example, biogenic volatile organic compounds (BVOCs) can react with NO₃ radicals or RO₂+NO into organic nitrate (Ng et al., 2017). Although organic nitrate is not favored to be ionized in positive ESI-MS (Wan and Yu, 2006), organic nitrate formed from BVOC could be highly functionalized (Lee et al., 2016) and ionized in positive MS through other alkaline functional groups. Recent studies have shown that BVOCs, including isoprene (C₅H₈) and monoterpenes (C₁₀H₁₆), dominate the organic nitrate formation in the southeastern United States under the condition of the mixed anthropogenic NO_x and BVOCs (Xu et al., 2014; Lee et al., 2016; H. Zhang et al., 2018). Several molecular formulas formed from monoterpene and NO₃ radicals were found in our study (Table 3) (Lee et al., 2016; H. Zhang et al., 2018).

4 Implications and limitations

This study analyzed the WSOM using ESI(+) FTICR MS in fine particulate aerosol from the Himalayas and found that the molecular compositions of WSOM were mainly comprised of two group compounds (CHO and CHON) with equally important contributions. The two compound groups could be originated from biomass burning emission and BVOC oxidation products because many markers for these two sources were found in these molecular compounds. All our compounds had relatively high DBE values, suggesting a potential high light absorption feature. Due to the relative higher mass concentration and higher contribution of nitrogen-containing compounds (9 % of PM₁) during these two periods based on HR-ToF-AMS results (X. Zhang et al., 2018), it is believed that aerosol transported to the Himalayas plays an important role in atmospheric radiative forcing.

Ramanathan and Carmichael (2008) found a distinct warming effect of light-absorbing aerosol over the Himalayas through estimating aerosol radiative forcing by BC. However, brown carbon had not been considered before, which could also be an important light-absorbing aerosol due to its high mass loading (Laskin et al., 2015). Zhang et al. (2017) estimated the light absorption contribution of brown carbon from the inland Tibetan Plateau, which was up to ∼13 % of that of BC. The high DBE and nitrogen-containing OA in our study suggested aerosol in the Himalayas could also contain amounts of light-absorbing organic matter because the light absorption properties of organic molecules are closely related to the number of double bonds and rings in the molecule and nitrogen atoms. Many studies have found that the dominating chemical molecules in brown carbon were related to nitrogen-containing aerosol (e.g., Lin et al., 2016). This kind of aerosol combined with BC could have a higher radiative forcing than before in this area.

More comprehensive methods are needed in the future for identifying BrC in the Himalayas due to the chemical complexity of BrC. For example, BrC extraction is highly dependent on the solvent used, and water-insoluble OA can contribute higher light absorption than water-soluble OA (Chen and Bond, 2010). In addition, Budisulistiorini et al. (2017) found that a number of compounds can dominate the light absorption of BrC, although they make a minor contribution to the aerosol mass. Therefore, it is important to know the exact chromophores of BrC which can be obtained by combining light absorption measurement with a photodiode (PDA) detector and chemical composition with a high-resolution mass spectrometer with high performance liquid chromatography (HPLC-PDA-HRMS system) (Lin et al., 2016). For mass spectrometry analysis, different ionization sources which are also favorable for different compounds, such as ESI, only detect a part of polar compounds; non-polar compounds which could dominate the contribution of BrC can be measured using atmospheric pressure photo ionization (Lin et al., 2018). A recent study indicated that over 40 % of the solvent-extractable BrC light absorption is attributed to water-insoluble, non-polar to semi-polar compounds such as PAHs and their derivatives (Lin et al., 2018). In contrast, the polar, water-soluble BrC compounds, which are detected in ESI, account for less than 30 % of light absorption by BrC (Lin et al., 2018).