Aqueous-phase chemistry of glyoxal with multifunctional reduced nitrogen compounds: a potential missing route for secondary brown carbon

The aqueous-phase chemistry of glyoxal (GL) with reduced nitrogen compounds (RNCs) is a significant source for secondary brown carbon (SBrC), which is one of the largest uncertainties in climate predictions. However, a few studies have revealed that SBrC formation is affected by multifunctional RNCs, which have a non-negligible atmospheric abundance. Hence, we assessed theoretical and experimental approaches to investigate the reaction mechanisms and kinetics of the mixtures for ammonium sulfate (AS), multifunctional amine monoethanolamine (MEA), and GL. Our experiments indicate that light absorption and growth rate are enhanced more efficiently in the MEA–GL mixture relative to AS–GL and MEA–AS–GL mixtures and MEA reactions of the chromophores than in the analogous AS reactions. Quantum chemical calculations show that the formation and propagation of oligomers proceed via four-step nucleophilic addition reactions in three reaction systems. The presence of MEA provides the two extra branched chains that affect the natural charges and steric hindrance of intermediates, facilitating the formation of chromophores. Molecule dynamics simulations reveal that the interfacial and interior attraction on the aqueous aerosols with MEA is more pronounced for small α-dicarbonyls to facilitate further engagement in the aqueous-phase reactions. Our results show a possible missing source for SBrC formation on urban, regional, and global scales.

1 Introduction

Brown carbon (BrC) represents the most important source of carbonaceous aerosols, with profound implications for the global climate, air quality, and human health (Laskin et al., 2015; Marrero-Ortiz et al., 2018; Li et al., 2022; Yan et al., 2018; Yuan et al., 2023). Chemical transport models reveal that a non-negligible radiative forcing by BrC ranges from 0.05 to 0.27 W m⁻² averaged globally (Tuccella et al., 2020; Wang et al., 2018; De Haan et al., 2020; Zhang et al., 2020; Laskin et al., 2015; Moise et al., 2015). Large differences in these estimated data result from the uncertainties of BrC on its formation mechanisms, chemical composition, and optical properties (An et al., 2019; Shi et al., 2020; Kasthuriarachchi et al., 2020; Corbin et al., 2019). This affects the understanding of the radiative effect in current climate models (Liu et al., 2020; Zhang et al., 2020, 2023). Compared with primary BrC, the sources and formation of secondary brown carbon (SBrC) are more complex and lack understanding in detail (Lin et al., 2015; Yuan et al., 2020; Srivastava et al., 2022). Hence, in recent years, great efforts have been made to better understand the chemical composition and formation mechanisms of SBrC chromophores.

There is compelling evidence that the heterogeneous reactions of reduced nitrogen compounds (RNCs) and small α-dicarbonyls have been recognized as significant sources of SBrC (Hawkins et al., 2018; De Haan et al., 2018; George et al., 2015). These SBrC chromophores are normally conjugated and are possible heteroaromatic species, such as imidazole (IML) and its derivatives (De Haan et al., 2009b, a; Yang et al., 2022). Numerous previous studies paid much attention to BrC from the secondary processes of small α-dicarbonyls with ammonium sulfate (AS) and methylamine (MA) (De Haan et al., 2020, 2019, 2009a; Lin et al., 2015). For example, nearly 30 chromophores were detected in the AS-methylglyoxal (MG) mixture by HPLC–PDA–HRMS, and nitrogen-containing compounds account for more than 70 % of the overall light absorption within the 300–500 nm range (Lin et al., 2015). Some studies have also revealed that the absorption of BrC generated in the AS–MG or MA–MG mixture increases with pH value (Hawkins et al., 2018; Sedehi et al., 2013). Also, the iminium pathway is predominant, while pH < 4 to form IML and its derivatives is suppressed at pH 4 (Nozière et al., 2009; Sedehi et al., 2013; Yu et al., 2011). Hence, pH value has a large effect on the formation of SBrC chromophores, but the chemical mechanisms of BrC formation under the different pH values remain unclear, hindering a systematical understanding of its integrated atmospheric chemistry and non-negligible environmental impacts.

On one hand, multifunctional RNCs (such as ethanolamines and amino acids) display strong atmospheric activity in terms of the formation of SBrC with a non-neglected atmospheric concentration (Huang et al., 2016; Ge et al., 2011; Powelson et al., 2014; Trainic et al., 2012; Laskin et al., 2015; Ning et al., 2022). For example, rapid BrC formation was detected in glycine reactions with small α-dicarbonyls, and sub-micrometer amino acid particles exhibited high growth upon exposure to small α-dicarbonyls (Powelson et al., 2014; Sedehi et al., 2013; De Haan et al., 2009b; Trainic et al., 2012). On the other hand, monoethanolamine (MEA) is an amine-based solvent for post-combustion CO₂ capture (PCCC) technology with a relatively high vapor pressure, emitting 80 t per year into the atmosphere for each 1×10⁶ t of CO₂ removed per year (Karl et al., 2011; Puxty et al., 2009; Shen et al., 2019). Recent field measurement has shown that MEA is the second-most abundant organic amine in PM_2.5 in Shanghai besides MA (Huang et al., 2016). However, to the best of our knowledge, few previous results are available on the participation of MEA in the SBrC formation with small alpha-dicarbonyls, and studies of its potential role in the atmosphere and human health have not been attempted.

Hence, we elucidated the chemical mechanisms of BrC chromophores from the mixtures of typical reaction of RCNs (i.e., MEA and AS) with small α-dicarbonyls using combined theoretical and experimental methods. Herein, glyoxal (GL) is selected as the representative of small α-dicarbonyls due to its high global emissions and significant contribution to BrC (Fu et al., 2008; Myriokefalitakis et al., 2008; Shi et al., 2020; Nie et al., 2022; Gomez et al., 2015). The chemical composition of the BrC chromophores was characterized by mass spectrometry at different initial pH values and the optical properties were measured using UV–Vis spectrophotometry. Possible pathways were calculated using density functional theory, and the mechanism of BrC chromophore formation was also simulated. The effects of multifunctional amine in the formation of SBrC chromophores were elaborated further. Additionally, the potential implications of multifunctional amine for climate radiative forcing were stated and discussed briefly.

2 Experimental methods and theoretical calculations

2.1 Experimental section

The procedures of each experiment are summarized in Fig. S1 in the Supplement. All reagents were used as described in the Supplement. Three mixtures were prepared under atmospheric-relevant aqueous conditions to generate SBrC: AS–GL, MEA–GL, and MEA–AS–GL. Briefly, the AS–GL (1 M) mixture was prepared by adding AS to aqueous GL (in ultrapure water) for a final concentration of 1 M of each reactant in the volumetric flasks. For the two MEA-containing mixtures, MEA was acidified with diluted sulfuric acid (20 %) to prevent GL from reacting with MEA in alkaline conditions. The acidified MEA was then combined with aqueous GL similar to that described for the AS–GL (1 M) mixture. All three solutions mentioned above were then diluted to reach a final concentration of 1 M in three 50 mL volumetric flasks. To explore the effects of pH values, the three mixtures were prepared with an initial pH value of 3 or 4 via the addition of sulfuric acid (20 %) or sodium hydroxide solution (2 M) prior to the mixing of RNCs and GL (Kampf et al., 2016; Yu et al., 2011). Each mixture was transported into brown vials, which has been proven to avoid the photolysis and light-induced reactions of light-absorbing products (Kampf et al., 2012), to guarantee efficiently produced chromophores in droplet evaporation collecting on the timescales of seconds (Zhao et al., 2015; Lee et al., 2014).

The absorption spectra of all mixtures were recorded by using a UV–Vis spectrophotometer (Agilent Cary 300, USA). All experimental solutions were diluted by a factor of 200 or 400 before each measurement to avoid saturation of the absorption peaks. The diluted samples were added into a quartz cuvette with 1 cm optical path length right away to prevent the diluted samples from photolysis. The spectra recorded between 200–500 nm are shown in Fig. 1, and the blank experiments of the GL and RNC solution were performed and are presented in Fig. S2. The absorption spectra of all samples were measured three times. The wavelength-dependent mass absorption coefficients (MACs) of experimental solutions were calculated from the initial base-10 absorbance (A₁₀) as follows:

$$\begin{array}{}\text{(1)}& \text{MAC}\left(\mathit{\lambda }\right)={\displaystyle \frac{A_{\mathrm{10}}^{\text{solution}}\left(\mathit{\lambda }\right)\times \ln \left(\mathrm{10}\right)}{b\times C_{\text{mass}}}},\end{array}$$

where C_mass is the mass concentration of reactants and b is the path length (Aiona et al., 2017; Chen and Bond, 2010). The different dilution factors were normalized by using the MAC formula.

Figure 1 The MAC values for AS–GL, MEA–GL, and MEA–AS–GL mixtures at the initial pH of 3 and 4 at 1 d (a) and 15 d (b).

The samples used for mass spectrometry analysis were diluted by a factor of 800 or 1000 followed by syringe filtration. The filters were stored in brown chromatography injection vials to block the light. Ultra-performance liquid chromatography, coupled with hybrid quadrupole-exactive Orbitrap mass spectrometry (UPLC-Q-Orbitrap HRMS, Thermo Scientific™, USA) (Wang et al., 2017), was employed to obtain structural data of chromophores in this study. MS² analysis was used for all chromophores with a weight error of less than 10 ppm compared with the theoretical mass to obtain fragment information for the identification of structure analysis. A detailed description of the mass spectrometry and chromatographic conditions is in the Supplement.

2.2 Quantum calculations and molecular dynamics simulations

Quantum chemical calculations were performed using the Gaussian 09 package (Frisch et al., 2013). Structures for all stationary points (SPs), including reactants, intermediates, transition states (TSs), and products, were optimized using the hybrid density functional of the M06-2X method (Zhao and Truhlar, 2007) with a 6-311G(d,p) basis set, i.e., at the M06-2X/6-311G(d,p) level (Ji et al., 2017). The solvent effect was considered using the solvation model based on density (SMD) to simulate the aqueous environment (Gao et al., 2016; Marenich et al., 2009). Harmonic frequency calculation was carried out at the same level as structural optimization to verify whether an SP was a TS (with one and only imaginary frequency) or a minimum (without imaginary frequencies) (Ji et al., 2022). The intrinsic reaction coordinate calculation was performed to confirm that the TSs connected with the corresponding reactants and products. A single-point energy (SPE) calculation was executed using the M06-2X method with a more flexible 6-311+G(3df,3pd) basis set to obtain more accurate potential energy surfaces (PESs). For the pathways with TSs, the rate constants (k) were calculated via the conventional transition state theory (TST) (Evans and Polanyi, 1935; Eyring, 1935; Galano and Alvarez-Idaboy, 2009; Gao et al., 2014). To simulate real atmospheric conditions in the solution, the calculated k values were refined by solvent cage effects (Okuno, 1997) and diffusion-limited effects (Collins and Kimball, 1949), of which the calculation details of diffusion-limited rate constant k_d can be seen in the Supplement. For the pathways without TSs, the corresponding k values are predominated by the diffusion-limit effect, which is equal to the diffusion-limited rate constants.

The classical molecular dynamics (MD) simulation was performed using the NAMD package (Phillips et al., 2005) to simulate the heterogeneous processes of GL from gas to the AS and MEA particles. The AS particle is composed of 39 SO${}_{\mathrm{4}}^{\mathrm{2}-}$, 78 NH${}_{\mathrm{4}}^{+}$, and 2046 H₂O in a box size of 40 × 40 × 40 ų, while the MEA particle consists of 39 MEA and 2036 H₂O. The 5 ns equilibration at the time step of 1 fs was executed in the isothermal–isochoric (NVT) ensemble (T = 298 K) to ensure the thermodynamic equilibrium of particles (Shi et al., 2020; Zhang et al., 2019). The MD simulation of 2 ns is run via the NVT ensemble. MEA and GL were described using the CHARMM force field (Jorgensen et al., 1996) and H₂O was described using the TIP3P model (Martins-Costa et al., 2012). The fixed charges on NH${}_{\mathrm{4}}^{+}$ and SO${}_{\mathrm{4}}^{\mathrm{2}-}$ are scaled by 0.75 to account for the electronic polarizability (Leontyev and Stuchebrukhov, 2011; Mosallanejad et al., 2020). The periodic boundary conditions were selected for three dimensions. In order to calculate the kinetic trajectories of GL from gas to two target particles, the free-energy profile along the distance of the center of mass between each particle and GL was calculated via umbrella sampling (Torrie and Valleau, 1977) and the weighted-histogram analysis method (Kumar et al., 1992) based on the above equilibrated molecular dynamics trajectories. The bias potential force constant was equal to 10 kcal mol⁻¹ Å⁻².

3 Results and discussion

3.1 Mass absorption coefficients of BrC chromophores

The mass absorption coefficients (MACs) identified in AS–GL, MEA–GL, and MEA–AS–GL mixtures at the initial pH of 3 and 4 (denoted as pH = 3 and pH = 4) are shown in Fig. S3. The maximum adsorption peaks are located at 207, 212, and 209 nm for AS–GL, MEA–GL, and MEA–AS–GL mixtures at pH = 3, respectively, and the corresponding location is not changed at pH = 4. The MAC values of the maximum adsorption peaks are in the range of 1080–17909 cm² g⁻¹ for the three mixtures. In addition, each mixture has an absorption peak between 285–324 nm (Fig. S4) with a range of 42–228 cm² g⁻¹, which is consistent with the MAC values measured by Powelson et al. (2014)​​​​​​​ at the reaction time of 4 d but is smaller than the values measured by Zhao et al. with a long reaction time of 2–3 months (Zhao et al., 2015). The MAC values at 207–212 and 285–324 nm exhibit a similar trend (Fig. S4). Therefore, to easily compare the absorbance in the three mixtures, we focus on the adsorption peaks in the range of 207–212 nm, which exhibits an obvious variation, and the effect of the initial pH on reaction systems is also discussed in this range. The MAC values at pH = 4 are higher than those at pH = 3 for three mixtures. For example, the MAC value in the AS–GL mixture is 2037 cm² g⁻¹ at pH = 4, which is almost 2 times higher than that at pH = 3. Hence, the initial pH values of the solution mainly affect the MAC values rather than the locations of absorption peaks.

In order to explore the influence of the initial pH values on the MAC values, a comparison of MAC values at the initial pH 3 and 4 is performed for all three mixtures (Fig. 1a). Figure 1a shows a comparison of the MAC values for all three mixtures at the initial pH of 3 and 4. The MAC values of maximum adsorption peaks increase from the AS–GL to MEA–GL to MEA–AS–GL mixture, ranging from 1080 to 6345 cm² g⁻¹ at pH = 3 and from 2037 to 7617 cm² g⁻¹ at pH = 4. The highest MAC value of MEA–AS–GL is explained by the different initial total concentration of reactants (see the ”Experimental methods and theoretical calculations” section) since the initial concentration of AS and MEA in the MEA–AS–GL mixture is 2 times higher than that in the MEA–GL or AS–GL mixture. In addition, the MAC value of maximum adsorption peak in the MEA–AS–GL mixture is higher than the sum of the values in the MEA–GL and AS–GL mixtures, and the location of the maximum absorption peak in the MEA–AS–GL mixture is between those in the MEA–GL and AS–GL mixtures. This implies that extra chromophores are yielded in the MEA–AS–GL mixture in addition to producing the same chromophores as AS–GL and MEA–GL mixtures.

To compare the formation rate of chromophores between the different mixtures, the growth rates (GRs) of the maximum absorption peaks as a function of reaction time are shown in Fig. 2. The trend of the GR variation with reaction time at pH = 3 is similar to that at pH = 4, while the GRs of the three mixtures at pH = 4 are larger than those at pH = 3 at the beginning of the reactions. The GRs are nearly invariant after 6–9 d, implying that the chromophore formation for the three mixtures is irreversible. The MEA–AS–GL mixture exhibits larger GRs than other mixtures at the beginning of the reaction because of its higher initial concentration of reactants. As the reaction proceeds, the GRs of the MEA–GL mixture are increased and finally larger than those of other mixtures. Hence, MEA reactions form the chromophores more efficiently than the analogous AS reactions.

Figure 2 Dependence of the growth rates (blue line) and pH values (green line) on reaction time for AS–GL, MEA–GL, and MEA–AS–GL mixtures.

The GR dependence on the pH values of the three mixtures is also plotted as a function of reaction time, as shown in Fig. 2. The pH values rapidly degrade within the first 2 d in the three mixtures, which is the same trend as GRs that decrease by a factor of more than 1–3 at pH = 3 and 4. This trend is explained by ambient pH values since a known byproduct (i.e., formic acid) is formed (De Haan et al., 2009b, 2020; Galloway et al., 2009; Hamilton et al., 2013; Kampf et al., 2012; Yu et al., 2011). Note that the trend of GRs shows a decrease from the MEA–AS–GL and MEA–GL to AS–GL mixtures at the beginning of the reaction time, while the MAC values of the MEA–GL mixture are larger than those of two mixtures accompanied by the more rapid decrease in pH values in the solution after the reaction is equilibrium (Figs. 1b and 2), suggesting that chromophore formation of the three mixtures depends on the ambient pH value.

3.2 Chemical composition characterization of BrC chromophores

The chemical composition characterization of formed BrC chromophores was conducted by UPLC-Q-Orbitrap HRMS. The formulas, $m/z$ values, characteristic fragments, and structures of chromophores and intermediates are identified based on obtained mass spectrum data in the AS–GL, MEA–GL, and MEA–AS–GL mixtures (Table S1). The corresponding MS and MS² spectra of chromophores and intermediates are exhibited in Figs. 3 and S8–S12. For all mixtures, imidazole (IML) compounds are identified with a characteristic peak at $m/z$ 69.045 in the MS² spectra. Therefore, various IML compounds are observed based on several representative peaks at $m/z$ 69.045, including imidazole (IML_AS and IML_MEA), imidazole-2-carboxaldehyde (IC_AS and IC_MEA), and their hydrated forms (HIC_AS and HIC_MEA) for AS–GL and MEA–GL mixtures (Table S1, Figs. 3a–b and S8–S9). For the MEA–GL mixture, extra catenulate intermediates without IML-structure characteristics are obtained at $m/z$ values of 102.055 and 120.065 (Table S1, Figs. 3a and S10), corresponding to C₄H₇O₂N (IA_MEA) and C₄H₉O₃N (AHA_MEA and ID_MEA) compounds, respectively. However, no catenulate intermediates in the AS–GL mixture are observed in this study because of their low concentrations and short lifetimes, although they are observed by previous studies using MS/AMS and ¹H nuclear magnetic resonance spectroscopy (Galloway et al., 2009; Lee et al., 2013; Yu et al., 2011). In addition, as shown in Figs. 3b and S11, some IML-based products at $m/z$ values of 145.061, 135.066, and 193.072 were obtained in the AS–GL mixture, corresponding to hydrated N-glyoxal-substituted imidazole (HGI_AS), 2,2'-biimidazole (BIM_AS), and its glyoxal substituted analog (GBI_AS), respectively. As discussed above, an important distinction between the AS–GL and MEA–GL mixtures is the formation of bicyclic IML products (Figs. 3a–b), indicating that the optical properties of chromophores are mainly determined by mono-imidazole compounds rather than bicyclic IML compounds.

Figure 3 Mass spectra monitoring of chromophores for (a) MEA–GL, (b) AS–GL, and (c) MEA–AS–GL mixtures.

Figure 4 Possible pathways leading to chromophores for (a) MEA–GL and (b) AS–GL mixtures (oriented by gray arrows). Detailed PESs of the four NA reactions are presented in dotted boxes. The number denotes the values of ΔG_r and ΔG^‡ (in brackets) for each reaction step (in kcal mol⁻¹), and all energies are relative to the corresponding reactants.

To further explore the difference of identified products in the MEA–GL and AS–GL mixtures, possible pathways leading to the identified intermediates and chromophores are illustrated in Fig. 4 along with the reaction energies (ΔG_r) of all pathways calculated at the M06-2X/6-311+G(3df,3pd)//M06-2X/6-311G(d,p) level. As shown in Fig. 4, the formation and propagation of oligomers were proposed to proceed via four-step nucleophilic addition (NA) reactions. For the MEA–GL mixture, three catenulate intermediates (AHA_MEA, IA_MEA, and ID_MEA) are successively yielded by the nucleophilic attack of MEA at the reactive carbonyl site via dehydration and hydration, with the total ΔG_r value of −7.8 kcal mol⁻¹ (Fig. 4a). Subsequently, two-step NA reactions between ID_MEA and MEA and between DI_MEA and GL-diol (DL), followed by protonation and dehydration, yield two intermediates (HA_MEA and PIC_MEA) in sequence. Although the third NA reaction between DI_MEA and DL is endothermic (ΔG_r = 12.7 kcal mol⁻¹), the total ΔG_r value of DI_MEA formation in the MEA–GL mixture is −18.7 kcal mol⁻¹ for proceeding the NA reaction to yield PIC_MEA. Similarly, the formation of PIC_AS in the AS–GL mixture is also thermodynamically feasible, with the total ΔG_r value of −10.9 kcal mol⁻¹. However, PIC_MEA or PIC_AS is thermodynamically unstable since there is a large exothermicity of the subsequent reaction pathway (ΔG_r = −78.6 or −50.0 kcal mol⁻¹) for proceeding cyclization leading to the formation of IC_MEA or IC_AS. It should be noted that for the AS–GL mixture, the fate of IC_AS is dependent on the competition between the pathways of hydration to yield HIC_AS and NA reaction with DL to form BI_AS, while for the MEA–GL mixture, there are no nucleophilic sites of IC_MEA for further oligomerization to form bicyclic IML compounds because IC_MEA is imidazolium cation. Similarly, IC_MEA also undergoes a hydration reaction to form HIC_MEA with a similar structure to HIC_AS. Subsequently, HIC_AS and HIC_MEA are decomposed to yield IML_AS and IML_MEA, respectively, accompanied by the formation of formic acid (ΔG_r = −10.2 and −15.6 kcal mol⁻¹), which is the reason for the decrease in pH in Sect. 3.1. However, as a reaction byproduct, formic acid hardly participates in the formation of light-absorbing products, so it has little influence on the reaction mechanisms. Current results further explain our experimental results mentioned above that higher MAC and larger GR values are observed in the MEA–GL mixture than in the AS–GL mixture.

Figure 5 Possible pathways leading to chromophores for MEA–AS–GL mixture (oriented by gray arrows). Detailed PESs of the four NA reactions are presented in dotted box. The shaded area is the overlapping part with the pathways of MEA–GL and AS–GL mixtures. The number denotes the value of ΔG_r and ΔG^‡ (in brackets) for each reaction step (in kcal mol⁻¹), and all energies are relative to the corresponding reactants.

For the MEA–AS–GL mixture, the products in AS–GL and MEA–GL mixtures are also observed (Fig. 3c). Beyond that, four extra IML compounds are also observed at $m/z$ values of 113.071, 141.066, 159.076, and 171.076, corresponding to IML (IML_MAG), imidazole-2-carboxaldehyde (IC_MAG) and its hydrated form (HIC_MAG), and N-glyoxal-substituted imidazole (GI_MAG) (Figs. 3c and S12). An extra -C₂H₄O group exists in the geometries of the above four IML compounds relative to the products of the AL–GL mixture, indicating that there are cross-reactions between MEA and AS in the MEA–AS–GL mixture. As shown in Fig. 5, the cross-NA reaction between ID_AS and MEA or ID_MEA and AS possesses a negative ΔG_r value of −4.8 or −5.4 kcal mol⁻¹, followed by dehydration to form the same intermediate diimine (DI_MAG). This implies that the cross-reactions in the MEA–AS–GL mixture are thermodynamically favorable. Therefore, the formation and propagation of chromophores in the MEA–AS–GL mixture also proceed via NA reactions, which is the key route for the formation of BrC chromophores.

As shown in Fig. 3c, no bicyclic IML compounds are produced in the MEA–AS–GL mixture because the precursors of bicyclic IML compound (i.e., imidazole-2-carboxaldehyde) are fully hydrated under more acidic conditions than the AS–GL mixture (see pH values in Table S2). This leads to the formation of N-glyoxal-substituted imidazole (i.e., GI_MAG) instead of bicyclic IML compounds. The similar phenomenon is also found in previous studies (Ackendorf et al., 2017; Kampf et al., 2012; Yu et al., 2011) that bicyclic IML compounds are hardly yielded from imidazole-2-carboxaldehyde in acidic conditions. As discussed above, imidazole-based structural characteristics in chromophores are maintained in the presence of MEA, but the nucleophilicity of chromophores is reduced because the nucleophilic sites are occupied. Additionally, the positively charged quaternary amine salts (such as IC_MEA and GI_MAG) are also yielded in MEA–GL and MEA–AS–GL mixtures, and thereby the chemical composition and optical properties of chromophores are affected.

3.3 Chemical reaction mechanism leading to BrC chromophores

As discussed above, the four-step NA reactions are the key pathways to forming and propagating oligomers, including intermediates and chromophores for the three mixtures. Therefore, all possible pathways involved in the four key NA reactions of the three mixtures are calculated using the density functional theory. The corresponding PESs established by the M06-2X/6-311+G(3df,3pd)//M06-2X/6-311G(d,p) level are also presented in the dotted boxes of Figs. 4–5. The optimized geometries of key stationary points, including TSs, intermediates, and products, are depicted in Figs. S13–S15 at the M06-2X/6-311G(d,p) level. We first performed quantum chemistry calculation to evaluate the direct nucleophilic attack of GL by MEA or AS, which proceeds a large activation energy (ΔG^‡) value of 6.3 or 8.6 kcal mol⁻¹ followed by H-shift reaction to yield AHA_MEA or AHA_AS, with also a large ΔG^‡ value of 15.2 or 18.2 kcal mol⁻¹ (see NA1a' and NA2a' in Fig. 4). The high ΔG^‡ values and large endothermicity of the direct NA reactions leading to AHA_MEA and AHA_AS imply that their occurrences are kinetically and thermodynamically hindered.

Hence, we explored the cationic oligomerization of chromophore formation under acidic conditions, which involves three essential steps: (1) protonation and dehydration to form cationic intermediates (CIs) or carbenium ions (CBs), (2) the nucleophilic attack of CIs or CBs by MEA and AS, and (3) the formation of intermediates and chromophores by deprotonation or dehydration. As shown in Figs. 4–5, each pathway involved in the cationic-mediated reaction mechanism proceeds without a TS, except for the deprotonation of CIs in line with the results of previous studies (Ji et al., 2020, 2022). However, the deprotonation of CIs by sulfate ion (SO${}_{\mathrm{4}}^{\mathrm{2}-}$) possesses a negative ΔG^‡ value in this study, implying an approximate barrierless process of this kind of deprotonation.

For the first-step NA reaction (NA1a in Fig. 4) in the MEA–GL mixture, the electrophilic cationic site of CB_DL is attacked by the nucleophilic -NH₂ group of MEA with the ΔG_r value of −40.3 kcal mol⁻¹. CB_DL is broadly produced from GL and reflected from the large particle growth and formation of IML products (Ji et al., 2020; Li et al., 2021). The deprotonation of CI_MEA1 possesses a negative ΔG^‡ value of −4.5 kcal mol⁻¹, and a pre-reactive complex is identified prior to the corresponding TS (detailed in the Supplement). Similarly, the other two NA1b and NA1c reactions (Fig. 4) also include protonation, dehydration, nucleophilic attack, and deprotonation to yield HA_MEA and PIC _MEA. Kinetic data listed in Table S3 show that the rate constants of most pathways involved in the NA1a-1b and NA2a-2c reactions fall in the range of ∼ 10⁹ M⁻¹ s⁻¹. Similar results can be drawn for the AS–GL mixture, suggesting that the electrostatic attraction is a significant factor to affect the NA reactions.

To further evaluate the cationic reaction mechanism, the natural bond orbital (NBO) analysis reveals that the N atom of NH₃ exhibits a more negative charge (−1.1 e) relative to MEA (−0.9 e), suggesting the stronger electrostatic attraction between CB_DL and NH₃ to yield CI_AS1 in the first-step NA reaction (see NA1a and NA2a in Fig. 4). However, the second-step NA reaction between CB_MEA and MEA is promoted by MEA because the presence of MEA enhances the positive charge in CB_MEA (0.6 e), facilitating the electrostatic attraction (see NA1b and NA2b in Fig. 4). For the third-step NA reaction (see NA1c and NA2c in Fig. 4), due to the steric hindrance, the deprotonation of CI_MEA8 possesses a larger ΔG^‡ value relative to that of CI_AS8. Hence, the NA reactions are regulated by both the electrostatic attraction and the steric hindrance effect.

The fourth-step NA reactions in the MEA–GL and AS–GL mixtures exhibit two distinct chemical reaction mechanisms in cyclization to yield N-heterocycles (see NA1d and NA2d in Fig. 4). The protonation of PIC_MEA and PIC_AS occurs at the hydroxyl group to form CI_MEA9 and CI_AS9. For the MEA–GL mixture, the barrierless dehydration and cyclization of CI_MEA9 occur in one step to yield N-heterocycle (i.e., IC_MEA), with the total ΔG_r value of −78.6 kcal mol⁻¹ (NA1d in Fig. 4a). However, for the AS–GL mixture, the cyclization of PIC_AS to IC_AS includes protonation, dehydration, cyclization, and deprotonation. Note that cyclization and deprotonation proceed via two TSs in sequence, with the corresponding ΔG^‡ values of 3.9 and −0.6 kcal mol⁻¹ (NA2d in Fig. 4b), respectively, forming IC_AS. As discussed above, cyclization in the MEA–GL and AS–GL mixtures is the rate-limiting step to chromophore formation.

For the MEA–AS–GL mixture, AHA_MEA/AS and ID_MEA/AS are yielded via the same first NA reactions (NA1a/2a) as MEA–GL and AS–GL mixtures. Also, the formation of ID_MEA/AS proceeds via protonation and dehydration to form CB_MEA/AS. However, the second NA reaction includes the cross-NA reaction of CB_MEA with AS (NA3b-1) or CB_AS with MEA (NA3b-2) to produce extra oligomers (i.e., HA_MAG1 and HA_MAG2) in contrast to MEA–GL and AS–GL mixtures. Hence, the fate of CB_MEA/AS is dependent on the competition reaction between the pathways of self-NA reaction to form HA_MEA/AS (NA1b/2b) and cross-NA reaction to yield HA_MAG1/2 (NA3b-1/2). The ΔG_r values of the cross-NA reactions to yield HA_MAG1 and HA_MAG2 are −30.3 and −30.4 kcal mol⁻¹, respectively, comparable with those of self-NA reactions. This suggests that both NA reactions to form HAs are equally accessible. Subsequently, HA_MAG1/2 undergoes dehydration to form DI_MAG and further proceeds to the third NA reaction to yield PIC_MAG, in line with the mechanisms of the third NA reactions for the MEA–GL and AS–GL mixtures. The cyclization of CI_MAG10 (the fourth NA reaction) possesses two successive TSs, similar to that of the AS–GL mixture but different to that of the MEA–GL mixture. The corresponding ΔG^‡ values are obtained as 5.0 and 1.6 kcal mol⁻¹, respectively, which are larger than those of the AS–GL mixture. In summary, compared with the AS-containing mixtures, the presence of MEA provides the two extra branched chains in N atoms, which affects the natural charges and molecular steric hindrance of intermediates, to thereby facilitate the intramolecular interaction between N and C atoms to form SBrC chromophores.

4 Conclusions and atmospheric implications

BrC chromophores play an important role in the Earth's radiative balance, air quality, and human health. However, the formation mechanisms of BrC chromophores are not fully understood, hindering a comprehensive assessment of BrC chromophores on atmospheric chemistry and environmental impacts. Hence, using combined theoretical and experimental methods, we investigated the aqueous chemistry of typical RNCs with GL and evaluated the impact of typical multifunctional RNCs on the formation of BrC chromophores. Experimental studies show that the MAC values of chromophores are affected by the initial pH value for AS–GL, MEA–GL, and MEA–AS–GL mixtures and the growth rates of chromophores are enhanced in the presence of MEA. The optical properties of chromophores are regulated by monocyclic and bicyclic IML compounds in the AS–GL mixture but by monocyclic IML compounds in MEA-containing mixtures (i.e., MEA–GL and MEA–AS–GL). Combined with the results of quantum chemical calculations, chromophore formation is characterized by nucleophilic addition with large exothermicity and strong electrostatic attraction among the MEA-derived intermediates, which are also enhanced by MEA.

In addition, to simply evaluate the impacts of MEA and AS on chromosphere formation in the aqueous aerosols and fog/cloud droplets, a study of the dynamic process of GL from the gas to aqueous phase was carried out (Fig. S16). The free-energy difference reflects whether the liquid particles with MEA and AS (denoted as MEA and AS particles) prefer to adsorb and accommodate GL. As shown in Fig. S16, a larger decrease in the free energy (−3.7 kcal mol⁻¹) occurs when GL approaches the interface of the MEA particle relative to the AS particle, indicating a thermodynamically favorable process. Subsequently, the stabilized GL enters the interior region of the MEA and AS particles, with slightly endothermic reaction (1.6 and 2.4 kcal mol⁻¹). A smaller free-energy difference from the interface into the interior region of the MEA particle implies that the interfacial GL is more readily promoted to enter the interior region of the particle when the particles contain MEA compared with AS. Hence, the interfacial and interior attraction on the MEA particle is more pronounced for small α-dicarbonyls to facilitate further engagement in the aqueous-phase reactions with RNCs in the particle.

The formation of SBrC from multifunctional RNCs and small α-dicarbonyls occurs widely on aqueous aerosols and fog/cloud droplets under typical atmospheric conditions. Compared with the ubiquitous coexistence between AS and small α-dicarbonyls from global aerosol measurement, the SBrC aerosol formation from multifunctional RNC mixtures should be paid attention to during serious haze formation in China because of its atmospheric reactivities and non-negligible concentrations. Our results also imply that SBrC aerosols, if formed from the aqueous reactions between MEA and GL, likely contribute to atmospheric warming because the presence of MEA enhances the MACs of the mixture. Hence, the recognition of this aerosol formation mechanism in the radiative transfer atmospheric model is needed, representing a possible missing source for BrC formation on urban, regional, and global scales.