Smog Chamber Study on the Ozone Formation Potential of Acetaldehyde

Ozone (O₃) is an important pollutant in the troposphere, especially in the urban atmosphere, which affects the quality of atmospheric environment (Dong et al., 2020; Kinose et al., 2020) and human health (Liang et al., 2020; Zhao et al., 2020). In the past few decades, ozone has received considerable attention (Carter et al., 1993; Derwent et al., 2007; Wang et al., 2017; Cao et al., 2020).

In recent years, although a series of control measures for air pollution have been formulated and implemented in various places, ozone pollution incidents still occur frequently (Chi et al., 2018; Ma et al., 2019; Guérette et al., 2020; Janik et al., 2020; Kalabokas et al., 2020; Seltzer et al., 2020). In the Beijing−Tianjin−Hebei region, the Yangtze River Delta, and the Pearl River Delta of China, air pollution events with ozone concentration exceeding 100%−200% of the air standard value have been observed many times (Wang et al., 2017). Cao et al. (2020) observed that the ozone concentration in the west China rain zone and its adjacent regions has exceeded the relevant ecological critical loads. In Sydney, Australia, the ozone concentration can even exceed 100 ppb when ozone pollution occurs (Guérette et al., 2020). Seltzer et al. (2020) found that the harmful impacts of long-term exposure to ambient ozone on human health increased from 2000 to 2015 in USA.

O₃ precursors are volatile organic compounds (VOCs) and nitrogen oxide (NO_x). Among thousands of VOCs species, carbonyl compounds such as aldehydes and ketones are considered to be the important precursors of ozone formation (Dong et al., 2014; da Silva et al., 2016). It is known that acetaldehyde, as one of the main carbonyl compounds, has a significant effect on the formation of ozone (Saengsai and Jinsart, 2015). Guo et al. (2014) found that acetaldehyde contributed 42% of the total carbonyl compounds to the summer ozone concentration in the rural areas of Southern China, while da Silva et al. (2016) found that acetaldehyde was the most important precursor of ozone in Rio de Janeiro as compared with other carbonyls such as formaldehyde and aromatics such as benzene and toluene. In fact, its contribution to local ozone was about 2.6 times that of formaldehyde.

Many chemical mechanisms have been widely used to simulate ozone formation (Shi et al., 2012), such as Master Chemical Mechanism (MCM) (Jenkin et al., 1997, 2003, 2015) and Statewide Air Pollution Research Center mechanism (SAPRC) (Carter and Heo, 2013; Venecek et al., 2018). It is often found that there are some differences in the simulation results between different chemical mechanisms (Zong et al., 2018). To improve the accuracy of simulation results, it is necessary to evaluate and improve the chemical mechanisms according to experimental results (Derwent et al., 1998; Carter and Heo, 2013; Zong et al., 2018). Carter and Heo (2013) have supplemented the SAPRC mechanism many times, and the MCM mechanism has also been modified in some studies (Jia et al., 2009; Jia and Xu, 2016). To the best of our knowledge, the only known experimental study on the ozone formation of acetaldehyde was carried out by Carter et al. (1993), whose purpose was to verify the accuracy of the simulation result of SAPRC. Their experiments mainly focused on the influence of effective light intensity under low relative humidity (~15%). Smog chamber experiments regarding the influences of relative humidity (RH) and precursor concentrations on ozone formation from the photochemical reaction of acetaldehyde have not been reported in detail. Also, whether the MCM mechanism can accurately simulate the reaction process under different reaction conditions remains to be evaluated by experiments.

In this study, the ozone formation potential of acetaldehyde is studied in detail through the smog chamber system. The main research purposes are as follows: (1) to determine the influence of initial CH₃CHO/NO_x ratios (R_CN) on ozone concentrations under different RHs, (2) to evaluate whether the current known chemical mechanism (MCM) can accurately simulate the photochemical reaction of acetaldehyde, and (3) to estimate the influence of acetaldehyde on the ozone concentrations in the real atmospheric environment.

All experiments in this study were carried out in a smog chamber system. The detailed description of this smog chamber system has been given in our previous studies (Jia and Xu, 2016, 2018). The reactor has a volume of 1.3 m³ in this study. The light source was supplied by two types of UV lamps F40 BLB (GE, USA) and UVA-340 (Q-Lab Corporation, USA), with center wavelengths of 365 nm and 340 nm, respectively. These types of UV lamps have been used in other similar studies to simulate the sunlight in the real atmospheric environment (Wang et al., 2015; Luo et al., 2019, 2020; Chen et al., 2020). The NO₂ photolysis rate constant was determined to be J[NO₂] = 0.20 min⁻¹ to represent the effective light intensity in the reactor, which is in the range of 0.139−0.44 min⁻¹ widely used in other similar photochemical reaction systems such as 1,2,3-trimethylbenzene-NO_x (Luo et al., 2019), toluene-NO_x (Chen et al., 2020), and styrene-NO_x (Luo et al., 2020).

The background air used in all experiments was zero air produced by the Zero Air Supply (Model 111 and Model 1150, Thermo Scientific, USA), in which the concentrations of O₃, NO and NO_x are below 1 ppb, and the concentrations of SO₂ and hydrocarbon (HC) are below 50 ppt and 100 ppb, respectively. The concentration of the NO₂ standard gas was 508 ppm (diluted by N₂, Beijing Huayuan Gas Co., Ltd., China). The purity of acetaldehyde was 99.9% (Aladdin Reagent Co., Ltd., China). The RH of background air was controlled by bubbling zero air through the highly purified water. A hygrometer (Model 645, Testo AG, Germany) was used to measure temperature and RH in the experiments, with the accuracy being ±0.1°C and ±1%RH. The concentrations of NO_x (NO and NO₂) and O₃ were measured online by a NO_x analyzer (Model 42C, Thermo Scientific, USA) and an O₃ analyzer (Model 49C, Thermo Scientific, USA), respectively. The concentration of NO₂ measured by the NO_x analyzer included all other nitrogen-containing substances (NO₂, HNO₃, etc.) except NO, which is expressed as NO_y below. The concentration of acetaldehyde was measured by GC-MS (7890A/5975, Agilent, USA).

Before each experiment, the reactor was flushed with zero air with the light on until the concentrations of NO, NO_y and O₃ in the reactor were all lower than the detection limit of the instruments (1 ppb). A solenoid valve was used to control the flow of the background air into the reactor. When the background air filled half of the volume of the reactor, acetaldehyde was introduced into the reactor with zero air through the tee on the intake pipe, and the NO₂ standard gas was directly injected from the sample inlet of the reactor with a special syringe. After the reactor was full, the reactor was left for 0.5 h under the condition of no light and no activity, and then the initial condition in the reactor, including the temperature of the reactor and the initial concentrations of NO, NO₂, O₃ and acetaldehyde, was determined. After the measurement of the initial condition, the light was immediately turned on, marking the beginning of the experiment. During the experiment, the temperature around the reactor was kept stable (±1°C). The experimental initial conditions are shown in Table 1. Experiments 1−8 are used to study the effect of the ratio of initial CH₃CHO concentrations to NO_x concentrations (NO_x = NO + NO₂) on ozone concentrations under low-RH (10.7%−13.5%), while Experiments 9−14 are used for high−RH conditions (77.5%−79.8%). Hereafter, the ratio of initial CH₃CHO concentrations to NO_x concentrations is termed R_CN. The variation of R_CN was mainly realized by changing initial NO_x concentrations under the fixed initial CH₃CHO concentrations. Because the ozone concentration at 6 h in the photochemical reaction was generally used to represent the actual ozone formation ability of VOC (Carter et al., 1993), an experiment in this study lasted for 6 h.

In order to evaluate the accuracy of the MCM mechanism regarding the photochemical reaction of acetaldehyde, the results from the smog chamber experiments were compared with the simulated results by using MCM v3.3.1 (Jenkin et al., 2015). In the simulation of MCM v3.3.1, the wall loss rates of NO_x, O₃, acetaldehyde, formaldehyde, etc. were taken into account, and the source of HONO from NO₂ was also included. In addition to these chamber-dependent reaction processes and species, the mechanism of the photochemical reaction of acetaldehyde in MCM v3.3.1 included 136 chemical reactions and 46 species (http://mcm.leeds.ac.uk/mcmv3.3.1/). To solve these many ordinary differential equations and run the kinetic model under MATLAB, a chemical reaction compiler (CRC) developed by Jia (2007) was used to automatically generate MATLAB code for the simulation of the kinetic processes of the chamber experiments. A variable-order solver soft package called ode15s in MATLAB was used.

Figure 1 shows variations of O₃, NO_y, NO and CH₃CHO concentrations with time from the experiments under different RHs and initial reactant concentration ratios. The growth rate of O₃ concentrations is the largest in the first 1−2 minutes from the start of the experiment (Fig. 1a). After that, the O₃ concentration maintains a rapid increasing trend, which is due to formation of RO₂ radicals generated from the photochemical reaction of CH₃CHO. Because of the photolysis of NO₂, the concentration of NO increases rapidly and NO_y decreases rapidly within the first 1−2 minutes. Then, the NO concentration decreases gradually, and meanwhile, the NO_y concentration increases slowly, which is due to the gradual conversion of NO to nitrogen-containing substances such as NO₂ and peroxyacetyl nitrate (PAN). However, due to the effect of wall losses, there is a decreasing trend of NO_y in the later stage of reactions. The CH₃CHO concentration decreases rapidly at the beginning of reactions, but the decreasing rate gradually becomes smaller in the later stage of most experiments.

Figure 1.  Variations of measured concentrations of O₃, NO, NO_y and CH₃CHO in the reactions with time under different RHs (a: 12.7%; b: 77.5%; c: 11.4%; d: 79.8%) and R_CN (a: 1.1; b: 1.3; c: 13.7; d: 13.0) (No wall loss correction for all concentrations).

The photochemical reaction process is significantly affected by light intensity, temperature, RH, reactant concentrations, etc. In this study, the impacts of the initial reactant concentration ratio R_CN and RH on the photochemical reaction of acetaldehyde were the primary foci of interest, and only these results are reported herein.

R_CN has a significant effect on NO_y, for which the different R_CN effects can be seen under different RH conditions (Fig. 1). Under low-RH conditions, R_CN greatly affects the increase rate of NO_y in the early stage of the reaction. The time required for the NO_y concentration to reach its maximum value in Expt. 8 (R_CN = 13.7, RH = 11.4%) is about 80 minutes earlier than that in Expt. 1 (R_CN = 1.1, RH = 12.7%). However, under high-RH conditions, R_CN mainly affects the pattern of the variation of NO_y with time. In Expt. 9 (R_CN = 1.3, RH = 77.5%), the NO_y concentration decreases almost linearly with time after the start of reactions, while in Expt. 14 (R_CN = 13.0, RH = 79.8%) it increases during the first one hour of reactions, and then becomes essentially unchanging. At the same RH conditions, the conversion rate of NO_x is different under different R_CN, which leads to the difference of the concentrations of chemical components in NO_y. In addition, the rate of conversion of NO₂ to HONO under different RHs is significantly different (Hu et al., 2011), and the different wall loss rate of NO₂ under different RHs was also observed (Jia et al., 2011). These are probably the reasons for the influence of R_CN and RHs on the NO_y concentrations.

The R_CN also has a significant impact on the O₃ concentration during the experiment. Under the condition of R_CN = 1.1 in Expt. 1 (Fig. 1a), the average formation rate of O₃ is about 0.79 ppb min⁻¹ in the first 180 minutes, which is close to 0.83 ppb min⁻¹ during the second 180 minutes from time t = 180−360 minutes. However, under the condition of R_CN = 13.7 in Expt. 8 (Fig. 1c), the average formation rate of O₃ is about 1.00 ppb min⁻¹ in the first 180 minutes, but it is only 0.35 ppb min⁻¹ during the later stage of 180 minutes from t = 180−360 minutes. Similarly, the average formation rate of O₃ in Expt. 9 (R_CN = 1.3) does not change significantly during the first and second 180 minutes of the experiment, while in Expt. 14 (R_CN = 13.0) it is larger in the first 180 minutes than in the second 180 minutes. With the increase of R_CN, the concentration of NO₂ decreases. Therefore, under high R_CN, such as Expts. 8 and 14, the low concentration of NO₂ reduces the increase of the formation rate of O₃ at the later stage of reactions.

Variation of the O₃ concentration with R_CN at time t = 6 h is shown in Fig. 2. Under low-RH conditions (Expts. 1−8), with the increase of R_CN from 1.1 to 3.4, the O₃ concentration at t = 6 h increases rapidly, while the O₃ concentration at t = 6 h decreases rapidly with the increase of R_CN in the range of 4.6−13.7. According to the variation pattern of O₃ concentration with R_CN in Expts. 1−8 in Fig. 2, it is estimated that the R_CN at the inflection point of the curve that describes variation of O₃ concentrations with R_CN is 3.2. Similarly, under high-RH conditions, the fitting result of R_CN for Expts. 9−14 is 2.8 at the inflection point.

Figure 2.  Variations of the O₃ concentration at time t=6 h with R_CN under different RHs (11.6%±1.1% in Expts. 1−8, 78.8%±1.0% in Expts. 9−14; the symbols represent experimental values, and the dashed and solid lines represent fitting results of Expts. 1−8 and Expts. 9−14, respectively).

It is known from Fig. 2 that the effect of RH on the O₃ concentration at 6 h is different under different R_CN. On the left side of the O₃ inflection point (R_CN < 3), there is no significant difference in the O₃ concentration at 6 h between low and high RHs. As R_CN rises, the difference increases. When R_CN > 4.6, the O₃ concentration at 6 h under high RH is about 16% lower than that under low RH on average. Under low R_CN (R_CN on the left side of the O₃ inflection point), the wall loss rate of O₃ increases under high RH, but the increase in the loss of NO₂ on the reactor surface can promote the formation of O₃. Thus, the RH effect on the O₃ concentration is generally limited. However, under high R_CN (R_CN on the right side of the O₃ inflection point), the increase of NO₂ wall loss under high RH will further reduce the O₃ formation. As a result, the O₃ concentration under high RH is lower than that under low RH.

RH can also affect the amount of acetaldehyde consumed in the reaction. According to the change of CH₃CHO concentrations, the proportion (γ) of CH₃CHO that participated in the reaction at the end of each experiment (the ratio of CH₃CHO consumed at the end of an experiment to its initial concentration) was calculated and is plotted against R_CN in Fig. 3. In the low-RH experiments (Expts. 1−8), the values for γ range from 25% to 33% under the different R_CN, for which γ is 31% and 29% on average under R_CN ≤ 3.4 and R_CN ≥ 4.6, respectively. However, in the high-RH experiments (Expts. 9−14), under R_CN ≤ 3.0, the average value of γ is 36%, which is higher than that under R_CN ≥ 6.7 (27%). Under low R_CN, the initial concentration of NO₂ was larger than that under high R_CN, which generated higher HONO concentrations and OH radical concentrations. This is a main reason for larger consumption of acetaldehyde under low R_CN.

Figure 3.  Variations of proportion (γ) of CH₃CHO consumed in the reactions with R_CN.

In order to further study the formation of O₃ from CH₃CHO, the incremental reactivity (IR) of acetaldehyde is calculated using the following formula (Carter et al., 1995):

where ${\rm{\Delta }}\left({{{\rm{O}}_{\rm{3}}}{\rm{ - NO}}} \right)_{t{\rm{,test}}}^{}$ is ([O₃]_t−[NO]_t)−([O₃]₀−[NO]₀) measured at time t from the start of an experiment where the test VOC is added, ${\rm{\Delta }}\left({{{\rm{O}}_{\rm{3}}}{\rm{ - NO}}} \right)_{t{\rm{,blank}}}^{}$ is the corresponding value from the experiment where the test VOC is not present, and [VOC]₀ is the amount of test VOC added. Taking Expt. 2 as an example, we see that IR shows a strongly increasing trend with reaction time (Fig. 4a) until the end of the experiment (6 h). Similar changes in IR from other experiments are also seen, which is probably due to the presence of PAN as the reservoir of NO_x in the reaction, leading to the time lag to reach the maximum O₃ concentration. From the comparison of Expt. 2 (RH = 10.9%) with Expt. 10 (RH = 78.0%), it is found that RH has little effect on IR during the experiment, and the IR at the end of the experiment (6 h) in Expt. 2 is about 5% larger than that in Expt. 10. Similarly, for Expts. 8 and 14, RH generates a 12% difference in IR between these two experiments (Fig. 4a). However, the comparison of Expts. 2 (R_CN = 2.0) and 8 (R_CN = 13.7) shows that R_CN has an effect on IR, and the IR at 6 h (IR_6h) from Expt. 2 is about 1.8 times that from Expt. 8. The variation of IR_6h under different R_CN is shown in Fig. 4b. This indicates that under different RH conditions, the IR_6h is larger under low R_CN than under high R_CN. Under high R_CN, the low concentration of NO₂ is not conducive to the formation of ozone, which is consistent with the small amount of CH₃CHO consumed. In addition, the IR_6h under high RH is lower than that under low RH, which is consistent with the variation of O₃ concentrations under different RH conditions. The wall loss rate of O₃ increases with increasing RH, which leads to a decrease in IR. Besides, the wall loss rate of NO₂ also increases with the increase in RH, which further reduces the O₃ formation under high R_CN, leading to great reduction of the IR of acetaldehyde.

Figure 4.  Variations of experiment-based incremental reactivity (IR) of acetaldehyde with time (a), and IR_6h with R_CN (b).

MCM v3.3.1 was used to simulate the concentrations of NO, NO_y, CH₃CHO and O₃ in Expts. 1−14. The simulation from the original MCM mechanism and relevant auxiliary reactions (mainly wall losses) is named Run1 below.

Compared with the measured O₃ concentrations, the simulated results from Run1 are underestimated under different RH conditions. Taking Expts. 2 and 10 as examples (Fig. 5), we see that the differences in the O₃ concentration between the simulations and experiments increase with the reaction time. As a result, the simulated O₃ concentrations at the end of experiments (6 h) are 29% and 49% lower than the measured concentrations in Expts. 2 and 10, respectively. Further analysis shows that the maximum difference between the simulated and measured O₃ concentrations at the end of experiments appears to be at the inflection point of the curve that describes variation of O₃ concentrations with R_CN (Figs. 6a and 6b). Meanwhile, it should be noted from Figs. 6a and 6b that under both low and high RHs, the R_CN values at the inflection point of O₃ concentrations at 6 h differ greatly between the experiments and simulations. For Expts. 9−14 with high RH, the experimental O₃ concentration at 6 h reaches a maximum at R_CN = 3.0 (Expt. 11), while the simulated O₃ concentration reaches the maximum at R_CN = 7.7 (Expt. 13). Thus, the simulated O₃ concentration at 6 h for the R_CN of 3.0 (Expt. 11) is 49% lower than the observed value, while for the R_CN of 7.7 (Expt. 13) the simulated value is only 17% lower than the observed one. Similarly, for Expts. 1−8 with low RH, the experimental O₃ concentration at 6 h reaches a maximum at R_CN = 3.4 (Expt. 3), while the simulated value reaches the maximum at R_CN = 4.6 (Expt. 4), leading to the maximum difference between the simulated and measured O₃ concentrations appearing at R_CN = 3.4 (Expt. 3).

Figure 5.  Variations of simulated and measured concentrations of NO, NO_y, O₃ and CH₃CHO (Symbol in the figure represents experimental observation values, and solid line represents simulated values by MCM v3.3.1).

Figure 6.  Comparison of simulated and measured concentrations of O₃, NO_y, CH₃CHO and IR_6h at the end of experiments from different cases (Run1: original MCM v3.3.1; Run2: MCM v3.3.1 with R1; Run3: MCM v3.3.1 with R2; Run4: MCM v3.3.1 with R1 and R2).

It can be concluded that when R_CN ≤ 3.4, the difference in the O₃ concentration at 6 h between simulations and experiments in the high-RH experiments (48% ± 1%) is larger than that in the low-RH experiments (29% ± 6%), while when R_CN ≥ 4.6, this difference in the high-RH experiments (23%±5%) is smaller than that in the low-RH experiments (29% ± 1%).

Run1 simulates well the variation of the experimental NO concentration, although the simulated concentration is higher than the measured value, which is consistent with the underestimation of O₃ in Run1. Meanwhile, Run1 simulates well the variation of NO_y with time (Fig. 5), with the largest difference between the simulated and measured values mainly appearing at the end of the experiments (6 h) as a whole, as shown in Figs. 6c and 6d. The differences in the NO_y concentration at 6 h between the simulations and experiments are generally less than 20% except for Expt. 9 (41%) and Expt. 10 (38%). Under high-RH conditions, the effect of wall losses on NO_y was probably large, and at the same time, the NO_y concentration decreased with increasing R_CN. Combination of high RH and low R_CN generates the large difference in NO_y for Expts. 9 and 10.

The difference between the measured and simulated CH₃CHO concentrations is relatively smaller than that for other chemical components, with the largest difference appearing at the end of Expt. 1, in which the simulated result is 31% lower than the measured result (Figs. 6e and 6f). Nevertheless, under low- and high-RH conditions, the differences between the simulated and measured CH₃CHO concentrations at the end of the experiments are 12% ± 10% and 9% ± 7% on average, respectively. In addition, in the low-RH experiments (Expts. 1−8), the simulated acetaldehyde concentration at 6 h gradually varies from underestimate to overestimate, while in the high-RH experiments (Expts. 9−14), all Run1-simulated results are slightly larger than the experimental results. The simulation of CH₃CHO can be affected by the simulation of NO_y. For example, when Run1-simulated results underestimate the experimental concentrations of NO_y under low R_CN, the reaction between OH radical and acetaldehyde is generally enhanced, which leads to the underestimation of the acetaldehyde concentration, whereas under high R_CN, the overestimation of NO_y can promote the formation of OH radicals, which also can lead to the underestimation of acetaldehyde.

It is known that the accuracy of Run1-simulated results for different species is different in the photochemical reaction of acetaldehyde, and the accuracy of simulated results is closely related to the reactant concentration and RH. Therefore, for the verification of the accuracy of different mechanisms for the photochemical reaction of acetaldehyde, the influence of reaction conditions should be fully considered.

The simulated IR value of acetaldehyde at the end of experiments (6 h) by MCM v3.3.1 is lower than the experiment-based value (Figs. 6g and 6h), with the underestimated values being 28% ± 5% in Expts. 1−8 and 32% ± 12% in Expts. 9−14, respectively. In the low-RH experiments (Expts. 1−8) both the simulated and experiment-based IR_6h values decrease with increasing R_CN, while in the high-RH experiments (Expts. 9−14), the experiment-based IR_6h varies non-monotonically with increasing R_CN, but the simulated IR_6h varies steadily as a whole. It is known that high RH accentuates the difference in IR_6h between MCM and experiments. This further demonstrates that the mechanism of the photochemistry of acetaldehyde in MCM v3.3.1 probably needs to be improved.

In the simulation results from Run1, the NO_y concentration includes the PAN concentration, which can account for about 80% of the NO_y concentration at the end of the experiments. Because there are some differences in the NO_y concentration between the Run1-simulated and experimental results, it is considered that the reaction process of PAN is probably important for improving the simulation.

The removal process of PAN in the real atmospheric environment mainly includes its reaction with OH radicals, pyrolysis process and photolysis process (R1) (Seinfeld and Pandis, 2006). It is considered that the atmospheric loss rate of PAN by photolysis is greater than that by its reaction with OH radicals, and when altitude is > 7 km from the ground, the loss rate of PAN by photolysis is even larger than that by pyrolysis (Talukdar et al., 1995).

Many researchers have investigated the photolysis process of PAN (Libuda and Zabel, 1995; Talukdar et al., 1995; Flowers et al., 2007). With the further understanding of the photolysis process of PAN, the influence of photolysis processes on the loss of PAN has been gradually realized in the photochemical reaction of acetaldehyde. In the earlier version (SAPRC-99) of the SAPRC mechanism, the photolysis process of PAN (R1) was ignored. This process was included for the first time in version SAPRC-07 (Carter, 2010), and it is also used in the latest version of SAPRC-11 (Carter and Heo, 2013). The photolysis rate constant adopted was k₁ = 6.12 × 10⁻⁵ cm³ molecule⁻¹ s⁻¹ (300 K). Nevertheless, only the reaction of PAN with OH radicals and its pyrolysis have been included in MCM v3.3.1, while the photolysis of PAN has not yet been included.

In order to determine the influence of R1 on the photochemical reaction of acetaldehyde in this study, R1 was added to original MCM v3.3.1, and k₁ = 6.12 × 10⁻⁵ cm³ molecule⁻¹ s⁻¹ was also used in the simulation, but the influence of temperature difference between different experiments is ignored. The smog chamber experiments in this study were re-simulated, which is named Run2 below.

The simulated O₃ concentrations at time t = 6 h from Run2 are shown in Figs. 6a and 6b. Compared with Run1, the simulated O₃ concentrations at 6 h in Expts. 1−8 and Expts. 9−14 in Run2 increased by 4%−10% and 5%−8%, respectively. Also, the increased amount for R_CN ≤ 3.4 (8.3% ± 2.1% for Expts. 1−8 and 7.7% ± 0.2% for Expts. 9−14) is larger than that for R_CN ≥ 4.6 (5.1% ± 0.9% and 6.0% ± 0.9%). In Expts. 1−8 and Expts. 9−14, the average differences between the Run2-simulated and measured NO_y concentrations at 6 h are 3% ± 9% and 19% ± 17%, respectively, which are close to those between the Run1-simulated and measured results (4% ± 9% and 20% ± 17%) (Figs. 6c and 6d). Similarly, the average differences between the Run2-simulated and measured CH₃CHO concentrations at 6 h are 4% ± 16% and 11% ± 6% for Expts. 1−8 and Expts. 9−14, which are also close to those between Run1 and experimental results (5% ± 15% and 10% ± 6%) (Figs. 6e and 6f). Because of the competitive reactions of PAN and CH₃CHO with OH radicals, adding R1 to MCM will lead to the increase of OH radical concentrations, which will slightly increase the amount of CH₃CHO consumed in the reactions and promote the formation of O₃. For example, compared with the simulation results for Expt. 2 in Run1, the OH radical concentration at 6 h in Run2 increased by 9% in Run2, and the amount of the consumed CH₃CHO increased by 3%, leading to a 10% increase in the O₃ concentration at 6 h.

In recent years, some studies have found that the photolysis reaction (R2) of nitric acid or nitrate adsorbed on the surface of natural and artificial materials may be an important source of HONO in the atmospheric environment, which is 1−3 orders of magnitude higher than that in the gas or liquid phase (Baergen and Donaldson, 2013; Ye et al., 2016, 2019). In the simulation of Run1 and Run2, the photolysis process of HNO₃ adsorbed on the reactor surface was treated the same as that in the gas phase, which may lead to a difference between the simulated and measured results. Thus, reaction R2 was added to the original MCM v3.3.1 (without R1) and then the simulation was re-conducted (Run3). In order to determine the maximum effect of reaction R2 on the results, the photolysis rate constant used for R2 was taken to be 3 orders of magnitude higher than that in the gas phase.

Compared with Run1, under low-RH conditions, the simulated O₃ concentrations at 6 h for R_CN ≤ 3.4 in Run3 increased by 4%−8%, while for R_CN ≥ 4.6, they increased by only 1%−3%. Nevertheless, under high-RH conditions, the simulated O₃ concentrations at 6 h for R_CN ≤ 3.0 in Run3 increased by 5%−13%, but for R_CN ≥ 6.7, they increased by only 1%−2%. Compared with Run1, the Run3-simulated NO_y concentrations at 6 h only increased by 8% and 6% on average in Expts. 1−8 and Expts. 9−14, respectively. The Run3-simulated CH₃CHO concentrations at 6 h are 6% ± 19% and 9% ± 6% higher than the measured results in Expts. 1−8 and Expts. 9−14, respectively, which are also close to those between Run1-simulated results and experiments (5% ± 15% and 10% ± 6% in Expts. 1−8 and Expts. 9−14) (Figs. 6e and 6f). Therefore, compared with the Run1-simulated results, the results of NO_y and O₃ are improved to some extent under different reaction conditions.

As discussed above, compared with the Run1-simulated results, the increase of O₃ in Run2 was generally larger than that in Run3, which may be due to the effect of HNO₃ concentration on the reactor surface. Because the effect of R2 on the reaction increases with the increase of the HNO₃ concentration on the reactor surface, in the early stage of the reaction, the low content of HNO₃ on the reactor surface led to the small effect of R2 on the reaction. Taking Expt. 2 as an example, compared with the simulation results in Run1, the OH radical concentration at 3 h increased by 11% in Run3, while it increased by 28% at 6 h. Under high R_CN, the content of HNO₃ on the reactor surface is lower than that under low R_CN. Thus, after R2 was added into the MCM, the increase for O₃ under high R_CN is smaller than that under low R_CN.

Because both reactions R1 and R2 can affect the conversion of NO_x, there may be interactions between R1 and R2 in the reaction. In order to determine the influence of the interaction between R1 and R2 on the simulated results, both R1 and R2 were added into the original MCM v3.3.1 for re-simulation (Run4) (Fig. 6). Under low-RH conditions (Expts. 1−8), the simulated O₃ concentrations at 6 h in Run4 increased by 14%−23% as compared with those in Run1 for R_CN ≤ 3.4, while they displayed an increase of 5%−10% for R_CN ≥ 4.6 (Fig. 6a). Similarly, under high-RH conditions (Expts. 9−14), compared with Run1, the simulated O₃ concentrations at 6 h in Run4 increased by 13%−21% and 6%−9% for R_CN ≤ 3.0 and R_CN ≥ 6.7, respectively (Fig. 6b). On the whole, the differences in O₃ concentrations at 6 h between simulations and experiments are greatly reduced from 24%−35% and 17%−49% in Run1 to 6%−26% and 10%−42% in Run4 for low- and high-RH experiments, respectively. This demonstrates that the inclusion of both R1 and R2 in MCM can greatly improve the simulation of O₃ by 13%−74%, relative to original MCM. Meanwhile, compared with the Run1-simulated IR_6h, Run4-simulated results of IR_6h increased by 12% ± 8% and 14% ± 7% in Expts. 1−8 and Expts. 9−14, respectively (Figs. 6g and 6h).

It is known from the simulated O₃ concentrations that the combined effect of R1 and R2 is clearly greater than that of R1 or R2 alone and is nearly equal to the total effect of the sum of R1 and R2 in some experiments. Taking Expt. 2 as an example, compared with Run1, the addition of R1 into original MCM (Run2) reduced the concentration of PAN at 6 h by 1.2%, but increased the concentration of HNO₃ on the reactor surface at 6 h by 4.0%, which increased the O₃ concentration at 6 h by 10.4%. After adding R2 into original MCM (Run3), the HNO₃ on the reactor surface at 6 h reduces by 69.1%, but the PAN concentration at 6 h increased by 15.8%, leading to the increase of the O₃ concentration at 6 h by 8.4%. After adding R2 and R3 into original MCM (Run4), the HNO₃ on the reactor surface at 6 h reduced by 67.9%, but the PAN concentration at 6 h increased by 15.0%. As a result, the O₃ concentration at 6 h increased by 19.7%. Therefore, when R1 and R2 exist at the same time, their effects on the O₃ concentration are close to the total effect of the sum of R1 and R2.

In order to see whether MCM can accurately determine the inflection point of variation of the O₃ concentrations at 6 h with R_CN, the MCM mechanisms used in Runs1−4 above were used to simulate the O₃ concentrations at 6 h (CH₃CHO = 600 ppb). Under the conditions of RH = 12%, the R_CN at the inflexion points in Runs1−4 was found to be 2.4, 2.2, 1.8 and 2.1, respectively (Fig. 7). Obviously, the simulation that is closest to the fitting result in Expts. 1−8 (R_CN = 3.2) is Run1, followed by Run2, Run4 and Run3. Under the condition of RH = 78%, the R_CN at the inflexion points in Runs1−4 is 4.8, 4.6, 4.1 and 4.0, respectively, while it is 2.8 from the fitting result in Expts. 9−14. Therefore, the modified MCM can improve the simulation of R_CN at the inflection point under high-RH conditions while it can worsen the simulation under low-RH conditions. In addition, the simulated results in Runs1−4 also show that the effects of RH on the O₃ concentrations under high R_CN are larger than those under low R_CN, which is in agreement with experimental results.

Figure 7.  Variations of simulated O₃ concentrations at 6 h by MCM mechanism with R_CN (CH₃CHO = 600 ppb, T = 300 K; a: RH = 12%; b: RH = 78%; Run1: original MCM v3.3.1; Run2: MCM v3.3.1 with R1; Run3: MCM v3.3.1 with R2; Run4: MCM v3.3.1 with R1 and R2).

The ozone concentration at 6 h is generally used to represent the actual ozone formation ability of VOC under typical sunlight irradiations, while the maximum ozone concentration is used to represent the ozone formation potential (Carter et al., 1993). The modified MCM with R1 and R2 was used to simulate the O₃ concentrations generated by 6 h of photochemical reactions of acetaldehyde under different concentrations of CH₃CHO and NO₂. The results are shown in Fig. 8a and Table 2. It can be estimated that in Fig. 8a, the R_CN value at the ridge of the ozone isolines is about 1.7, which indicates that in the regime of R_CN < 1.7, the formation process of O₃ by acetaldehyde photochemistry is sensitive to CH₃CHO, while in the regime of R_CN > 1.7, O₃ is under the NO_x -sensitive condition. In Expts. 9−14, the R_CN value at the inflection point of the O₃ concentrations at 6 h is about 2.8 (Fig. 2). This demonstrates that the modified MCM still underestimates the experiment-based R_CN by 39%.

Figure 8.  Isogram of simulated O₃ concentrations after 6 h-photochemical reactions (RH = 70%, J[NO₂] = 0.56 min⁻¹, T = 300 K; (a) acetaldehyde as a single VOC; (b) acetaldehyde plus other 11 species of VOCs: m/p-xylene = 1.6 ppb, propylene = 1.2 ppb, ethylene = 1.5 ppb, isoprene = 1.0 ppb, toluene = 2.4 ppb, 1-butene = 0.5 ppb, o-xylene = 0.6 ppb, isopentane = 2.2 ppb, 1,2,4-thrimethylbenzene = 0.3 ppb, n-butane = 2.7 ppb). The dashed line indicates the transition from CH₃CHO-sensitive to NO_x-sensitive conditions.

It is known from the simulation results that under different scenarios the contribution of acetaldehyde to ozone is very different (Fig. 8a). For example, under the scenario of NO₂ = 50−200 ppb, CH₃CHO = 20−200 ppb, acetaldehyde can contribute 16−416 ppb O₃, whereas under the scenario of NO₂ = 30−50 ppb, CH₃CHO = 10−20 ppb, acetaldehyde can contribute 18−55 ppb O₃, which indicates that the contribution of acetaldehyde to ozone can still increase with the concurrent decrease of NO₂ and acetaldehyde concentrations. However, under the scenario of NO₂ ≤ 10 ppb, CH₃CHO ≤ 2 ppb, acetaldehyde can only contribute 1−7 ppb O₃.

In the actual atmospheric environment, there are many other kinds of VOCs that can affect the photochemical reaction of acetaldehyde and NO_x. For example, under the conditions of CH₃CHO = 2 ppb and NO₂ = 40 ppb, it is calculated that the O₃ concentration after 6-h reaction is 17 ppb. When the background air contains 1.6 ppb m-xylene, which was measured in the actual urban environment (Li et al., 2020), it can be calculated that the O₃ concentrations after 6-h reaction are 20 ppb and 24 ppb under the CH₃CHO concentrations of 0 and 2 ppb (NO₂ = 40 ppb), respectively, indicating that the ozone formation potential of m-xylene is larger than that of acetaldehyde, and that the existence of m-xylene greatly reduces the contribution of acetaldehyde to ozone. In order to analyze the sensitivity of O₃ formation to the acetaldehyde concentration in the actual atmospheric environment, we selected 11 species of VOCs with large contributions to the O₃ concentration in urban areas in the study of Li et al.(2020) as the background condition for the photochemical reaction of acetaldehyde. The simulated results are shown in Fig. 8b and Table 2. It is found that when the acetaldehyde concentration is much lower than the background VOCs (e.g., CH₃CHO ≤ 2), because the reactions of background VOCs with OH radicals are dominant relative to acetaldehyde, background VOCs will weaken the contribution of acetaldehyde to O₃. When CH₃CHO = 10−20 ppb, NO₂ = 30−50 ppb, because the reaction of acetaldehyde with OH radicals is dominant or comparable relative to the background VOCs, and O₃ formation is mainly VOC-limited, background VOCs will enhance the contribution of acetaldehyde to ozone. When the acetaldehyde concentration is much larger than those of background VOCs, the effect of background VOCs on the photochemical reaction of acetaldehyde depends on the actual ratio of VOCs to NO₂. Indeed, in the actual environment, since the acetaldehyde concentration is much lower than those of background VOCs, it is considered that other VOCs will weaken the O₃ formation by the photochemical reaction of acetaldehyde in the actual environment.

The IR values of acetaldehyde under different reaction conditions (RH = 70%, T = 300 K) were simulated by using the modified MCM and Eq. (1) in this study. Taking the O₃ concentrations at 6 h, the IR_6h values were computed, and the results are shown in Fig. 9a. In addition, the maximum incremental reactivity (MIR) was also simulated according to the maximum O₃ concentration under different scenarios (Table 2). The simulated results indicate that the MIR of acetaldehyde from the modified MCM is 4.0 ppb ppb⁻¹, which is 23% lower than the MIR value computed by Carter (2010) using the SAPRC-07 mechanism (RH = 50%, T = 300 K).

Figure 9.  Isogram of simulated IR at 6 h (RH = 70%, J[NO₂] = 0.56 min⁻¹, T = 300 K; (a) acetaldehyde as a single VOC; (b) acetaldehyde plus other 11 species of VOCs: m/p-xylene = 1.6 ppb, propylene = 1.2 ppb, ethylene = 1.5 ppb, isoprene = 1.0 ppb, toluene = 2.4 ppb, 1-butene = 0.5 ppb, o-xylene = 0.6 ppb, isopentane = 2.2 ppb, 1,2,4-thrimethylbenzene = 0.3 ppb, n-butane = 2.7 ppb).

Similarly, to illustrate the effects of other VOCs on the IR value of acetaldehyde in the actual environment, the 11 species of VOCs were also added into the simulation of MCM as the background condition (Fig. 9b). It is found that the variation of IR_6h is similar to the variation of O₃ concentration before and after adding background VOCs. The IR_6h value of acetaldehyde will decrease under CH₃CHO ≤ 2 ppb and NO₂ ≤ 10 ppb, but it will increase when CH₃CHO = 10−20 ppb and NO₂ = 30−50 ppb. Nevertheless, when CH₃CHO = 20−200 ppb and NO₂ = 50−200 ppb, the variation of IR_6h depends on the actual ratio of VOCs to NO₂. Therefore, when the IR value is used to estimate the O₃ formation by acetaldehyde, the IR value should be selected according to the actual environmental conditions.

It is known from the MIR of acetaldehyde that acetaldehyde has an intermediate ozone formation potential. As estimated by Carter (2010), some species have high MIR values, such as methylglyoxal (22.2 ppb ppb⁻¹) and m/p-xylene (14.6/18.1 ppb ppb⁻¹), while some species have low MIR, such as benzene (0.68 ppb ppb⁻¹) and acetylene (0.27 ppb ppb⁻¹). If the difference in IR_6h between our simulation and experiments is taken into account, our MIR of acetaldehyde is underestimated by 25%. Thus, actual MIR of acetaldehyde should be 5.0 ppb ppb⁻¹, which is close to the value proposed by Carter (2010) (5.2 ppb ppb⁻¹). In this way, the MIR of acetaldehyde is probably between that of formaldehyde (4.5 ppb ppb⁻¹) and isobutene (6.2 ppb ppb⁻¹).

To estimate the role of acetaldehyde in the formation of O₃ in terms of MIR, the acetaldehyde concentration in the atmosphere is required. In Beijing of China, the concentration of acetaldehyde can reach about 5 ppb (Gu et al, 2019). In Sao Paulo megacity of Brazil, the concentration of acetaldehyde can reach about 10 ppb (Dominutti et al., 2020), while in the port megacity of Istanbul of Turkey, it can even reach about 15 ppb (Thera et al., 2019). Therefore, the contribution of acetaldehyde to ozone formation potential can be 25−75 ppb according to the MIR of acetaldehyde, which is an important precursor of ozone.

The photochemical reaction processes of CH₃CHO-NO_x have been studied in detail by smog chamber experiments and simulated by the MCM mechanism. The experiment results show that the R_CN has a significant effect on the concentrations of O₃ in the reaction process. In the low-RH experiments (RH = 11.6%±1.1%), when R_CN < 3.2, the O₃ concentration at 6 h increases with the increase of R_CN, while it decreases with the increase of R_CN when R_CN > 3.2. Similarly, the R_CN at the inflexion point of the change of O₃ concentration at 6 h is 2.8 in high RH experiments (RH = 78.8%±1.0%). Under low R_CN (< 3), RH has no significant effect on the O₃ concentrations, but under high R_CN (> 3), the O₃ concentrations at 6 h in the high-RH experiments were 16% lower than those in the low-RH experiments on average. Original MCM v3.3.1-simulated results of O₃ concentrations at 6 h are much smaller than those from the low- and high-RH experiments. After adding both the photolysis process of PAN and the photolysis process of HNO₃ on the reactor surface into the original MCM, the difference between the simulated O₃ concentrations by MCM and the experimental results at the end of experiments reduced from 24%−35% and 17%−49% to 6%−26% and 10%−42% under low- and high-RH experiments, respectively. The current MCM mechanism of acetaldehyde underestimates the ozone generation potential of acetaldehyde, and the simulated MIR of acetaldehyde with the modified MCM mechanism is 4.0 ppb ppb^–1 without considering the effects of other VOCs. When other 11 typical VOCs are included into the simulation of ozone as a background condition, the MIR value of acetaldehyde increased to 6.4 ppb ppb^–1. This demonstrates that acetaldehyde is probably an important precursor for ozone in the troposphere.

Acknowledgements. This work was supported by the National Key R&D Program of China (2017YFC0210005), the National Natural Science Foundation of China (Nos. 41875163, 41875166 and 41375129).