Effects of Ammonia Mitigation on Secondary Organic Aerosol and Ammonium Nitrate Particle Formation in Photochemical Reacted Gasoline Vehicle Exhausts

Abstract

Gaseous air pollutants emitted primarily by anthropogenic sources form secondary products through photochemical reactions, complicating the regulatory analysis of anthropogenic emissions in the atmosphere. We used an environmental chassis dynamometer and a photochemical smog chamber to conduct a parameter sensitivity experiment to investigate the formation of secondary products from a gasoline passenger car. To simulate the mitigation of ammonia emissions from gasoline vehicle exhausts assuming future emission controls and to allow photochemical oxidation and aging of the vehicle exhaust, ammonia was selectively removed by a series of five denuders installed between the vehicle and photochemical smog chamber. Overall, there were no differences in the formation of secondary organic aerosols and ozone with or without ammonia mitigation. However, the potential for ammonium nitrate particle formation was significantly reduced with ammonia mitigation. In addition, ammonia mitigation resulted in increased aerosol acidity due to nitric acid in the gas phase not being neutralized by ammonia and condensing onto the liquid particle phase, indicating a potentially important secondary effect associated with ammonia mitigation. Thus, we provide new insights into the effects of ammonia mitigation on secondary emissions from gasoline vehicle exhaust and into a potentially useful experimental approach for determining primary and secondary emissions.

1. Introduction

Vehicle exhaust emissions are a major source of gaseous and particulate matter (PM) in the urban atmosphere. The recent regulatory landscape and electrification of vehicles have led to decreases in primary PM emissions released directly into the atmosphere via vehicle exhaust, such as elemental carbon (EC), primary organic aerosols (POAs), volatile organic compounds (VOCs), and nitrogen oxides (NO_X) (e.g., [1]). However, there are still concerns about high ozone concentrations in susceptible areas due to transboundary pollution and localized high ozone (O₃) concentration formation due to the reaction of VOCs and NO_X (e.g., [2]). Additionally, several studies have shown that secondary PM emissions, such as secondary organic aerosols (SOAs), are formed in the atmosphere by photochemical reactions of VOCs and NO_X contained in vehicle exhaust [3,4]. Because SOAs produced from vehicle exhaust can be more toxic than SOAs produced from VOCs from other sources or from POAs produced from vehicle exhaust [5,6], a better understanding of the properties of SOAs produced from vehicle exhaust is needed.

Studies using radiocarbon (¹⁴C) isotope measurements to investigate the origin of SOA, secondary organic carbon (SOC) in PM_2.5 at 25 sites around the world, have shown that SOC from fossil fuel contributes 12–54% to the total SOC emitted to the atmosphere (e.g., [7]). However, such approaches using field-based atmospheric observations are limited in their ability to accurately estimate origin-based contributions compared to model-based calculations. Indeed, although SOAs formed from vehicle exhaust are an important fossil fuel-derived atmospheric pollutant, their relative contribution to the total SOAs emitted varies considerably among studies. In modeling studies, gasoline vehicles are reported to account for only 56–79% of SOAs formed in the United States as a whole [8] but ≥90% in Southern California [9,10]. The relative contributions of diesel and gasoline vehicles to total vehicle SOAs are also unclear. For example, diesel vehicles are reported to contribute significantly to SOAs in the atmosphere in the United States (around 90% of the total vehicle SOA) [11] and in London [12]. However, in a laboratory-based study, gasoline vehicles produced SOA, but diesel vehicles equipped with diesel particulate filters did not, suggesting that gasoline passenger cars make the major contribution to the SOAs produced by vehicles in urban areas [13]. Thus, to be able to obtain accurate estimates of the contribution to environmental pollution of different sources of secondary products such as SOA, a deeper understanding of the characteristics of the secondary products from vehicle exhaust is needed.

The photochemical smog chamber is a useful tool for examining the formation and evolution of air pollutants under controlled conditions, as well as for parameterizing atmospheric processes and revealing their underlying mechanisms. However, generating realistic vehicle emissions under controlled and reproducible conditions also requires specialized test equipment (e.g., chassis dynamometers) that is usually not available in large stationary smog chambers. To overcome these facility limitations, mobile smog chambers have been used to evaluate automotive emissions and the photochemical reactions they undergo in the atmosphere [13,14,15,16,17]. However, it is clear that the potential for secondary particle formation varies greatly, depending on the experimental parameters, including vehicle type (gasoline or diesel, year) [13,14,15,18,19,20], the use of emission control technologies [21,22], driving conditions [23,24], and fuel type [25,26,27,28]. In addition, these studies have focused on elucidating the mechanisms underlying SOA formation and have not evaluated the impacts of the formation of high concentrations of O₃ from the reaction of VOCs and NO_X, which is a recent issue in the field of transboundary pollution (e.g., [2]). A further limitation of these studies has been that nitrous acid was used as the source of OH radicals in the experiments (e.g., [15,16]), which precluded evaluation of the formation of nitrate-derived particles.

Evaluating the impacts of vehicle emissions on air quality requires consideration not only of photochemical reactions involving VOCs and NO_X but also those involving ammonia (NH₃), which is a highly reactive alkaline inorganic gas that can negatively impact the environment [29,30,31]. Therefore, quantification of NH₃ emissions from vehicle exhaust is crucial for assessing air quality impacts and developing effective control strategies. The National Emission Ceilings Directive 2001/81/EC, the Gothenburg Protocol under the United Nations Convention on Long-Range Transboundary Air Pollution, and the Integrated Pollution Prevention and Control Directive (2008/1/EC) all aim to reduce emissions of NH₃ [30]. As the only alkaline inorganic gas with high reactivity, NH₃ is known to contribute to the formation of ammonium nitrate (NH₄NO₃) particles in automotive emissions [17,19,22,25,26,28]. However, since there is an overall paucity of knowledge regarding the role of NH₃ in the formation of SOAS in automotive emissions, more detailed investigations of its potential impacts are needed.

As already mentioned, secondary pollution is caused by the formation of secondary particles and O₃. Previous studies of vehicle emissions based on photochemical smog chamber experiments have been limited in their assessment of both secondary particles and O₃. Therefore, studies are needed to evaluate the O₃ formation for modern vehicles in this study compared to previous studies. Here, we used an environmental chassis dynamometer and a photochemical smog chamber to conduct a parameter sensitivity experiment to investigate the formation of secondary pollutants, both secondary particles and ozone, in exhaust from a direct-injection gasoline passenger car with or without NH₃ emissions mitigation.

2. Materials and Methods

2.1. Vehicle Chassis Dynamometer Experiments

Vehicle tests were performed in an environmental chamber with controlled temperature and with or without humidity (23 °C with 50% relative humidity control, 0 °C without relative humidity control, or −7 °C without relative humidity control) at the Japan Automobile Research Institute (JARI). The temperature and humidity settings were based on existing protocols that are widely used in vehicle exhaust testing to ensure that vehicle testing with realistic ambient temperature conditions can be performed. The ambient temperatures of 23 °C and −7 °C were taken from the Worldwide Light Duty Test Cycle (WLTC) with cold engine start. In the WLTC, humidity control is used only at 23 °C. The ambient temperature of 0 °C corresponds to the upper threshold of the Real Driving Emission measurement of the Euro 6 regulation.

The chassis dynamometer test facility within JARI complies with the World harmonized Light duty vehicle Test Procedure (WLTP) for type approval testing. The chamber contained a four-wheel drive chassis dynamometer for environmental-type experimental testing (Meiden, Tokyo, Japan) on which the test vehicle was placed. The exhaust pipe of the test vehicle was connected to a dilution tunnel (DLS-ONE-D; Horiba, Ltd., Kyoto, Japan) and a constant volume sampler (12 m³ min⁻¹; CVC-ONE-MV-HE(ESU); Horiba, Ltd., Kyoto, Japan). Exhaust gases were diluted an average of 24 times with high-efficiency particulate air (HEPA). Background air was filtered through an activated carbon, and a HEPA filter was maintained at 298 K and 50% relative humidity. Carbon monoxide, carbon dioxide, NO_X, total hydrocarbons, and non-methane hydrocarbons (NMHC) were sampled near the end of the dilution tunnel, and their concentrations were measured with an exhaust gas analyzer (MEXA7200LE; Horiba, Ltd., Kyoto, Japan). EC and primary organic carbon (OC) were collected from the dilution tunnel onto a quartz filter (Pallflex, 2500QAT-UP, 47 φ; Pall Corp., Port Washington, NY, USA). The EC and OC were determined by a thermal–optical carbon analyzer (model 2001; Desert Research Institute, Reno, NV, USA) using the IMPROVE protocol [32].

A small, direct-injection, two-wheel drive, gasoline passenger car of a size that is currently popular in Japan was tested (vehicle mass: 1340 kg, displacement: approximately 1.5 L, mileage: 23,192 km, model year: 2020, regulatory compliance year: 2018). The average life expectancy of Japanese passenger cars in 2022 is reported to be 13.84 years, i.e., passenger cars driven in the 2020 model year and reflecting the modern market [33]. Cold-start operational tests were performed based on the WLTC of operation. For these cold-start experiments, the vehicle was warmed up by the same test cycle prior to the experiment and then left for approximately 23.5 h. The cold-start experiment was performed once per day.

2.2. Photochemical Smog Chamber

2.2.1. Facility

A mobile photochemical smog chamber (e.g., [13,14,15,16,17,34]), designed for the evaluation of secondary particle formation potential from automotive exhaust gases, was installed next to the environmental chassis dynamometer. The photochemical smog chamber comprised a fixed frame (2.08 m height × 2.08 m width × 2.66 m depth) fitted with casters, a reactor bag, and a light source. The reactor bag was made from a transparent, chemically inert, ultraviolet light (UV)–permeable, 54-μm thick, fluorinated ethylene propylene (FEP) Teflon film. When fully filled with sample gas, the volume of the bag was 7.5 m³. The bag was fixed inside its own aluminum frame (1.5 m height × 2 m width × 2 m depth), and the frame was attached to the fixed frame of the smog chamber with a polytetrafluoroethylene gasket between the two frames such that the reactive gas contact area was not in contact with the aluminum frame. A reflector made of treated stainless steel (SUS304) was placed on the underside of the fixed frame, and 80 UV lamps (40 W, UVA340+; Q-Lab Crop., Westlake, OH, USA) were installed on the reflector with a sheet of FEP Teflon placed between the lamps and the reactor bag to prevent direct contact. The light transmission through FEP at wavelengths in the range 290–800 nm is reported to be >90% [35,36]. The UV lamps were cooled by push–pull ventilation of the room air through 16 fans, with the air being exhausted by 16 other fans.

2.2.2. Light Source

To trigger photolysis within the vehicle exhaust, particularly that of NO₂, which occurs at wavelengths <420 nm, the lights installed under the photochemical smog chamber were used to irradiate the reactor bag. A portable light spectrometer (USB2000 UV-VIS; Ocean Optics, Inc., Orlando, FL, USA) was used to characterize the irradiance spectrum reaching the inside of the reactor bag (Figure 1). The irradiance peaked at 350 nm, which was within the range of peak UV irradiances used in previously reported indoor chamber-based studies (340–370 nm). Compared to the sunlight spectrum observed during the summer season in Tokyo (Figure 1), the installed lamps were found to be reproducing mainly UV radiation between 295 and 320 nm. Furthermore, UV light sources have a different visible light spectrum at wavelengths >380 nm compared with sunlight (Figure 1), which significantly reduces the rate of photolysis reactions affected by longer wavelength light, such as photolysis of NO₃ radicals [37]. Therefore, ideally, a light that produces a realistic spectrum that reproduces daytime environmental conditions should be used, but if not, the spectral difference between natural sunlight and the light used in the experiment should be taken into account when interpreting the results. Consequently, the reaction in a mobile photochemical smog chamber is generally evaluated in terms of photochemical reaction performance based on the photolysis rate constant of NO₂, which is related to OH radical generation, as described in the next paragraph.

The photolysis rate constant of NO₂ can be used to characterize irradiation intensity. In previous studies [37,38,39,40], the photolysis rate constant has often been calculated from the photolysis rate ($J_{{NO}_2}$) of NO₂ and steady-state concentrations of NO_X and O₃ [37,38,40]. However, in the present study, the $J_{{NO}_2}$ of NO₂ was measured, and the photolysis rates of several species important in atmospheric photochemistry were calculated; the J_NO2 values were obtained by NO_X and O₃ analyzers connected to the photochemical smog chamber while under UV illumination from steady-state NO_X-O₃ concentrations. Using the NO₂ photolysis intensity determined from the zenith angle [37,38], our light source at 23 °C (photolysis rate: 0.43 min⁻¹) was comparable to an average morning during the summer solstice in Tokyo (0.46 min⁻¹, N 35°41′45″, E 139°45′8″) and at 0 °C (photolysis rate: 0.31 min⁻¹) to the transit time when the sun is due south at the winter solstice in Tokyo (0.33 min⁻¹). Maximum values in previous studies typically ranged from 0.12 to 0.54 min⁻¹ [15,37,38], with the present values being close to the median value of the other studies. At −7 °C (photolysis rate: 0.2 min⁻¹), our light source was comparable to an average morning during the summer solstice in Sapporo, a city in the far north of Japan (0.16 min⁻¹, N 43°3’51”, E 141°30′49″).

OH exposure is a factor that affects the concentration of SOAs and SOA carbon/oxygen ratios (e.g., [19]). OH exposure was determined by using Equation (1):

$$\mathrm{O}\mathrm{H}\mathrm{e}\mathrm{x}\mathrm{p}\mathrm{o}\mathrm{s}\mathrm{u}\mathrm{r}\mathrm{e}=-{\displaystyle \frac{1}{k_{\mathrm{O}\mathrm{H}}}}\left(\ln {\displaystyle \frac{{\left[\mathrm{V}\mathrm{O}\mathrm{C}\right]}_{\mathrm{F}inal}}{{\left[\mathrm{V}\mathrm{O}\mathrm{C}\right]}_{\mathrm{I}nitial}}}\right)$$

(1)

where OH exposure (molecules cm⁻³ h⁻¹) was calculated based on the average decay rate of toluene as measured by a gas chromatograph equipped with a flame ionization detector (GC-NMHC analyzer, GL Sciences Inc., Saitama, Japan). The average OH reactivity rate coefficient (k_OH) of toluene (5.63 × 10⁻¹² cm³ molecules⁻¹ s⁻¹ at 298 K (25 °C) [41]) and measurement errors in toluene were potential sources of uncertainty in the estimation of OH exposure.

2.2.3. Instrumentation

A series of instruments was used to characterize the gas- and particle-phase emissions in the reactor bag. Particle number distribution was measured with a scanning mobility particle sizer (classifier model 3080 and condensation particle counter model 3750; TSI, Inc., Shoreview, MO, USA). Particle mass was measured as a 1-h value with a beta-ray absorption PM densitometer (PM712; Kimoto, Osaka, Japan). Secondary particle formation was monitored either by a soot-particle time-of-flight aerosol mass spectrometer or an aerosol chemical speciation monitor (Aerodyne, Inc., Billerica, MA, USA). Secondary particles were quantified by collecting a volume of 1 m³ aerosol sample in the reaction bag on a quartz filter (Pallflex, 2500QAT-UP, 47 φ; Pall Corp., Port Washington, NY, USA) after the 5-h reaction. The concentrations of nitrate and ammonium ions collected on the quartz filter were determined by ion chromatography (Dionex Integrion; Thermo Fisher Scientific Inc., Waltham, MA, USA). The concentration of organic carbon (OC) and EC collected on the quartz filters were determined by a thermal–optical carbon analyzer (model 2001; Desert Research Institute, Reno, NV, USA) using the IMPROVE protocol.

Gas-phase organic species in a reaction bag were monitored using a proton transfer reaction mass spectrometer (PTR-TOF 8000; Ionicon, Innsbruck, Austria). The gas-phase organic species were collected into a 5-L vinyl alcohol polymer bag (Smart Bag PA-AAK-5; GL Sciences Inc., Saitama, Japan) both before and after the photochemical reaction, and the concentrations of non-methane hydrocarbons with carbon numbers between 2 and 12 were determined by a gas chromatograph equipped with a flame ionization detector (model GC-NMHC, GL Sciences Inc., Tokyo, Japan). Gas monitors were used to determine the concentrations of CO₂, NO_X-NH₃, and O₃ (Models 410i, 17i, and 49i, respectively; Thermo Fisher Scientific Inc., Waltham, MA, USA); the monitors were zeroed daily and calibrated at least weekly.

The NOx-NH₃ gas analyzer used in this study uses a chemiluminescence detector for NO to make measurements. The air sample enters the reaction chamber and reacts with O₃ produced by a generator inside the instrument; the reaction of NO with O₃ produces luminescent radiation that is directly proportional to the NO concentration. Once inside the instrument, the air sample is passed first through a molybdenum converter heated to 325 °C to convert NO₂ to NO and then through a stainless-steel converter heated to 750 °C to convert NH₃ to NO. The detector measures NO, then NO and NO₂, and finally NO, NO₂, and NH₃. The conversion efficiency of NO₂ and NH₃ through the converter is corrected by a calibration procedure. Molybdenum converters are also known to convert peroxyacetyl nitrate, nitrous acid (HONO), nitric acid (HNO₃), and organic nitrates formed after the reaction of VOCs with NO [42,43,44]. NO₂ interferences are generally lower in concentration than NO₂ and photochemical smog chamber experiments can obtain effective NO₂ concentrations under pre-reaction conditions [45]. The NO_X-NH₃ gas analyzer used in this study has been widely used to measure NH₃ concentrations in the atmosphere [45]. The results of the present NO_X and NH₃ measurements are described in detail in Section 3.1., Primary Gas and Particle Emissions, where the ratios of NO to NO_X were 82 ± 3%, 87 ± 5%, and 95 ± 2% at 23 °C, 0 °C, and −7 °C, respectively, and the ratio of NO₂ (approximately equal to NO_X minus NO) to NO_X was small; NO₂ measurements can provide valid NO₂ concentrations under conditions prior to photochemical reactions. Since the measurement of NO₂ in emissions is beyond the scope of this study, the nitrogen oxide types are not distinguished and are expressed as the NO_X concentrations in this study. Regarding NH₃, no NO_X was detected when NH₃ was removed by the denuder, which provides a reasonable basis upon which the present study can show the effect of NH₃ reduction prior to the photochemical reaction.

The concentration of HNO₃ gas and the aerosol acidity (H⁺(aq)), i.e., the concentration of free hydrogen ions in the aqueous liquid phase of the particles, were obtained by calculation due to facility limitations. The concentration of HNO₃ gas was approximated from O₃ and HNO₃ concentration curves for the photochemical smog chamber obtained by using the Statewide Air Pollution Research Center chemical reaction model [46]. Aerosol acidity (H⁺(aq)) was determined by using the ISORROPIA thermodynamic equilibrium model [47].

2.2.4. Experimental Procedures

Figure 2 shows a summary of the process of injecting the exhaust gas into the photochemical smog chamber, performing photochemical oxidation and aging, measuring the sample gas, and cleaning the chamber after the experiment.

Figure 3a shows a photograph of the experimental chamber with the test vehicle placed on the environmental chassis dynamometer. The vehicle exhaust was connected to the photochemical smog chamber via an ejector diluter without (base scenario) or with a denuder line connected between the ejector diluter and photochemical smog chamber. The exhaust gas was injected by the ejector diluter (dilution ratio: 12 times; DI-1000; Dekati Ltd., Kangasala, Finland) into the reactor bag of the portable smog chamber at a rate of 4.2 L min⁻¹. To supply the dilution gas with a constant flow rate into the reactor bag, clean air was supplied to the ejector diluter at 50 L min⁻¹ under the control of a mass flow controller (MQV0050; Azbil Corp., Tokyo, Japan). During a 30-min WLTC, 1.625 m³ of dilution gas was introduced into the reactor bag; therefore, before injecting the exhaust gas into the photochemical smog chamber, suction was used to reduce the volume of the reactor bag by approximately 2 m³. As an additional hydroxyl radical (OH) source when adding the exhaust, H₂O₂ (0.25 mL, 30% v/v, 5 ppm equivalent in the reactor in our experiments) was also injected into the reactor bag via the make-up air supplied at 5 L min⁻¹ [17,22,48]. The air supplied to the reactor bag was dry, but the exhaust gas contained moisture; at 23 °C, the relative humidity inside the reactor bag was <13%, at 0 °C <34%, and at −7 °C <43%.

Clean air was generated by an oil-free scroll compressor equipped with a membrane air dryer (SLP-221CD; Anest-Iwata Corp., Kanagawa, Japan); the dehumidified, compressed air was passed through a manual air dryer (model 4001; CKD, Aichi, Japan), an oxidation catalyst heated to 350 °C, Purafil chemical adsorbents (Purafil and Purafil Puracarb AM; Purafil Inc., Doraville, GA, USA) [37], activated carbon, a molecular sieve, and finally through a HEPA filter.

The difference between the pressure in the reactor bag and atmospheric pressure was monitored by a differential pressure gauge (GC62; Nagano Keiki Co. Ltd., Tokyo, Japan), and the mass flow controller automatically stopped the air supply when the pressure was ≥5 Pa, thus ensuring a constant reactor bag volume. After receiving the exhaust gas, the reactor bag was allowed to stand for 15 min to allow mixing. To allow determination of the gas concentrations in the exhaust prior to the photochemical reaction, 5 L of sample gas was collected in a Smart Bag PA and an aldehyde cartridge (InertSep mini AERO DNPH-HR; GL Sciences Inc., Saitama, Japan) over a 20-min period.

To obtain NH₃-free exhaust gas, NH₃ was selectively removed by installing a series of five stainless steel concentric tube denuders (DN-315; Sunset Laboratory Inc., Portland, OR, USA) (Figure 3) between the ejector dilutor and the photochemical smog chamber. The inner wall of the denuder was impregnated with 10% malic acid in ethanol, which was allowed to dry before use.

UV light irradiation was started at about 2 h after the exhaust gas from one WLTC run was started to be injected into the photochemical smog chamber. The irradiation duration was 5 h, which corresponds to the average daily solar irradiation duration in Japan. During the UV irradiation, the instruments described in Section 2.2.3 collected samples of gas from the reactor bag and continuously measured their concentrations. The volume of the reactor bag was maintained by introducing clean air as a make-up gas at 5 L min⁻¹; 25% dilution was achieved during the 5-h photochemical reaction.

Immediately after the end of the photoreaction period, 5 L of sample gas was collected in a Smart Bag PA and an aldehyde cartridge over a 20-min period. The sample gas was collected in the Smart Bag PA while mixing at 20 mL min⁻¹ with 10 ppm nitric oxide and in the aldehyde cartridge with a potassium iodide cartridge in series in front to avoid reactions with the O₃ collection device and analyzer, respectively. Particles in the reaction bag after the 5-h photochemical reaction were collected on a quartz filter with a volume of 1.00 m³ at a flow rate of 50 L min⁻¹ by a pump with mass flow control, and secondary particles were quantified by measuring the concentrations of ionic components (NO₃⁻, NH₄⁺), OC, and EC.

2.2.5. Data Analysis

Emission factors (EF_c) for the aerosols and gases detected in the reactor bag were calculated using the following equation and are reported as mass per mass of fuel burned (mg kg-fuel⁻¹):

$${\mathrm{E}\mathrm{F}}_{\mathrm{C}}={\displaystyle \frac{\left[\mathrm{C}\right]}{\left[{\mathrm{C}\mathrm{O}}_2\right]}}\times {\mathrm{E}\mathrm{F}}_{{\mathrm{C}\mathrm{O}}_2}$$

(2)

where [C] is the background- and dilution-corrected concentration of the gas or aerosol compositions in mg m⁻³, [CO₂] is the background- and dilution-corrected concentration of CO₂ in the chamber in g m⁻³; and EF_CO2 is the emission factor of CO₂ measured by a gas analyzer with the dilution tunnel shown in Section 2.1 in g kg-fuel⁻¹. The emission factors for SOAs (secondary aerosol), NH₄NO₃, and NH₃ (primary gas) were determined from the concentrations in the smog chamber. The emission factors for EC and primary organic carbon (POC) (primary aerosols) and for NOx and NMHC (primary gases) were determined from the concentrations in the dilution tunnel.

To quantify O₃ and HNO₃ gas formation in the smog chamber, we corrected the rate of dilution air introduced into the smog chamber (0.04 h⁻¹). To quantify secondary organic aerosol and NH₄NO₃ particle formation in the smog chamber, we corrected the loss of aerosol particles to the reactor bag walls. Briefly, aerosol particle loss to the bag walls was treated as a first-order process with rate constants determined from decay measurements of inert tracer species (black carbon or sulfate seed) [49]. An aerosol particle wall loss rate constant without dilution was calculated using black carbon that is not lost because of reaction but only decays in concentration through dilution or wall loss, as measured on a microAeth black carbon monitor (model MA350; AethLabs, San Francisco, CA, USA) [50]. Smog chambers with nearly spherical surface-to-volume ratios have the lowest aerosol particle wall loss rate constants [51], and compared to previous studies (e.g., 0.46 to 0.66 h⁻¹, e.g., [52]), the aerosol particle wall loss rate constant in the present study (0.12 h⁻¹) was smaller.

For the quantification of SOA, the most common measurement, secondary organic carbon (SOC), was quantified as organic carbon (OC) [53,54,55], which was collected on a quartz fiber filter and quantified by a thermal–optical carbon analyzer. The conversion of OC, where only carbon was quantified, to organic aerosol, which also contains carbon, hydrogen, and oxygen, was performed by multiplying OC by a constant conversion factor, the organic mass-to-organic carbon (OM/OC) ratio, to estimate the total amount of SOAs and POAs [53] in Equation (3):

$${\mathrm{E}\mathrm{F}}_{\mathrm{S}\mathrm{O}\mathrm{A}}={\mathrm{E}\mathrm{F}}_{\mathrm{S}\mathrm{O}\mathrm{C}}\times {\mathrm{O}\mathrm{M}/\mathrm{O}\mathrm{C}}_{\mathrm{S}\mathrm{O}\mathrm{A}\_\mathrm{T}}-{\mathrm{E}\mathrm{F}}_{\mathrm{P}\mathrm{O}\mathrm{C}}\times {\mathrm{O}\mathrm{M}/\mathrm{O}\mathrm{C}}_{\mathrm{P}\mathrm{O}\mathrm{A}\_\mathrm{T}}$$

(3)

where OM/OC_{SOA_T} is the OM/OC ratio of OA containing mainly SOAs and some POAs in the chamber after the photochemical reaction, as observed by soot-particle time-of-flight aerosol mass spectrometry, which is a standard high-resolution time-of-flight aerosol mass spectrometer (HR-ToF-AMS) used to analyze elements in organic compounds [54,55,56], coupled to a diode-pumped, Nd:YAG, intracavity, 1064-nm infrared laser vaporizer. The OM/OC ratios obtained were 1.8 at −7 °C, 2.0 at 0 °C, and 2.2 at 23 °C. OM/OC_{POA_T} is the OM/OC ratio including only POA, also observed by HR-ToF-AMS, and the value was 1.2 at all three temperatures. Our OM/OC ratio values are consistent with the OM/OC ratio of 1.2 for POAs reported for gasoline vehicles (1.2) [55], 2.0 reported from smog chamber experiments [55], and the range of 1.7 ± 0.5 for the Southern California atmosphere [56]. Further investigations are needed to assess the suitability of the OM/OC ratios used in our experiment, and this study only uses it as an empirical coefficient.

3. Results and Discussion

3.1. Primary Gas and Particle Emissions

Table 1. Comparison of emission results from the present and previous studies, showing formation potentials for primary gases (NMHC, NO_X, NH₃) and particles (EC and POA), a secondary gas (O₃) and particles (SOAs and NH₄NO₃), and OH radical exposure.

Source	Condition	Temperature	NMHC	NOx	NH3	EC	POA	SOA	NH4NO3	O3	SOA Yields	OH Exposure
		[°C]	[mg kg-fuel−1]								[−]	[Molecules cm−3 h−1]
This study *1	Base case	23	161	97	25	2.3	1.1	46	189	1015	0.731	6.2 × 107
		23	164	112	28	2.0	1.0	54	228	1091	0.833	6.0 × 107
	NH3 denuder	23	187	116	4	2.5	1.1	63	2	1093	0.866	6.3 × 107
	Base case	0	562	119	25	3.6	1.5	74	82	587	0.597	3.3 × 107
		0	521	119	40	13.4	2.5	73	66	519	0.503	3.7 × 107
	NH3 denuder	0	546	109	0.6	7.1	1.9	79	3	585	0.729	3.0 × 107
	Base case	–7	763	92	23	11.0	2.3	36	47	215	0.372	1.9 × 107
		–7	1210	97	73	20.6	2.7	86	52	262	0.593	2.2 × 107
	NH3 denuder	–7	1196	89	0.8	14.5	2.6	43	19	234	0.390	1.7 × 107
Ref.a *2	NEDC *3	22	1676–2064	123–131	94–102	0–0.4	0–0.4	1.3–2	N.A. *4	N.A.	0.07–0.7	0.1 × 107
		–7	5362–5778	384–408	111	4.5–5.1	1–4.9	13.4–38.2	N.A.	N.A.	0.18–0.35	0.1 × 107
Ref. *5		25	1750 ± 210	890 ± 140	N.A.	11.2–20.0	25.8–40.5	344–347	N.A.	N.A.	N.A.	1.2 × 107
Ref. *6	CSC70	23	4–360	34–352	4–223	0.08–62	N.A.	0–2154	0–5300	N.A.	N.A.	0.2–0.9 × 107
Ref. *7	CVS-75	22	283–819	224–2702	N.A.	1.2–10	0.1–11	28–66	65–153	N.A.	N.A.	2.5 × 108
	NEDC	22	283–3104	224–2702	N.A.	0.9–12.5	0.1–9.8	12–66	20–135	N.A.	N.A.	2.5 × 108
Ref. *8	Pre–LEV	25	2680–34,980	5440–19,640	N.A.	2.4–44.4	0.5–136.3	26.9–108	N.A.	N.A.	N.A.	0.5 × 107
	LEV I	25	230–17,820	1250–21,970	N.A.	0.8–122.2	0.1–106.2	10.1–131	N.A.	N.A.	N.A.	0.5 × 107
	LEV II	25	80–740	80–630	N.A.	1.1–35.2	0.3–4.8	1.6–54.6	N.A.	N.A.	N.A.	0.5 × 107
Ref. *9	NEDC	25	N.A.	N.A.	N.A.	N.A.	0.001–0.12	1–44	N.A.	N.A.	0.038–0.17	0.5 × 107

*¹ Driving conditions: Worldwide harmonized light-duty driving Test Cycle. The vehicle used in the present study was tested as an off-cycle vehicle because Japanese regulations do not include extra high. *² Adapted from reference [25]; to facilitate comparison, values were converted from g-CO₂ kg-fuel⁻¹ to mg kg-fuel⁻¹. *³ NEDC: New European Driving Cycle. *⁴ N.A.: Not available. *⁵ Adapted from reference [15]. *⁶ Adapted from reference [28]; to facilitate comparison, values were converted from mg km⁻¹ to mg kg-fuel⁻¹. *⁷ Adapted from reference [57]. *⁸ Adapted from reference [16]. *⁹ Adapted from reference [20].

shows a comparison of emission results from the present study versus previously published data. For the gasoline vehicle tested under the WLTC in the present study, the emission factors at 23 °C for NMHC (161–187 mg kg-fuel⁻¹) and NO_X (97–116 mg kg-fuel⁻¹) were generally lower than those reported from previous studies; NHMC and NO_X emissions in previous studies were in rather large ranges of 4–34,980 mg kg-fuel⁻¹ and 34–21,970 mg kg-fuel⁻¹, respectively [15,16,25,28,57].

In general, gasoline vehicle emissions contribute more to final pollutant emissions during cold starts in cold environments than at room temperature [16,22,23]. This is because the engine and drive-related systems take time to stabilize, and the after-treatment system needs time to reach its optimum activation temperature. Several previous studies have highlighted the need to consider the effects of low-temperature environments when evaluating emissions and air pollution [16,22]. The present emission factors for the low-temperature environments were 521–562 mg kg-fuel⁻¹ for NMHC and, 107–119 mg kg-fuel⁻¹ for NO_X at 0 °C, and 763–1210 mg kg-fuel⁻¹ for NMHC and 89–97 mg kg-fuel⁻¹ for NO_X at −7 °C. At these low temperatures, the NMHC emissions exceeded those detected under the 23 °C environment, whereas there was no marked change in NO_X emissions. These emission levels were similar to those observed for gasoline vehicle emissions in a previous study where NMHC emissions were in the range of 1676–2064 mg kg-fuel⁻¹ at 22 °C and 5362–5778 mg kg-fuel⁻¹ at −7 °C, and NO_X emissions were in the range of 123–131 mg kg-fuel⁻¹ at 22 °C and 384–408 mg kg-fuel⁻¹ at −7 °C [25].

Secondary pollution is suggested to occur when NH₃ reacts with nitrogen oxides in the atmosphere to form NH₄NO₃ [22,25,26,28,57]. For the gasoline vehicle used in the present study, the emission factor for NH₃ was in the range 25–28, 25–40, and 23–73 mg kg-fuel⁻¹ at 23 °C, 0 °C, and −7 °C, respectively; thus, NH₃ emissions tended to be greatest at −7 °C. Previous studies have reported NH₃ emissions with considerable variation, ranging from 4 to 223 mg kg-fuel⁻¹ at 23 °C [25,28]. The NH₃ in gasoline vehicle exhaust is produced by a mechanism involving NO and H₂; H₂ is produced by the water–gas shift reaction between CO and water or by fluid reforming of hydrocarbons [30]. Consequently, the NH₃ emissions in gasoline-vehicle emissions typically increase as tailpipe CO concentrations increase [30]. Regarding the emission factors of NH₃ and CO in our study, the average NH₃ emission factors (n = 2) were 26.4, 32.6, and 48.3 mg kg-fuel⁻¹ at 23 °C, 0 °C, and −7 °C, respectively. The mean emission factors for CO (n = 2) were 10.3, 15.7, and 27.8 mg kg-fuel⁻¹ at 23 °C, 0 °C, and −7 °C, respectively. There was a strong correlation between the emission factors of NH₃ and CO at these three temperatures in our study (correlation coefficient r = 0.9997, n = 3). This suggests that the differences in NH₃ emissions between the previous studies and ours may be due to the difference in CO emissions occurring in the three-way catalyst in the tailpipe.

In interpreting these results, it is important to consider the causes of the large variations in the reported values. Emission factors contributing to these large variations include regulatory age [13,14,15,18], engine maintenance history, and the type of after-treatment equipment installed on the vehicle [21,22]; thus, it is important to note the likelihood of individual vehicle effects when interpreting the results.

3.2. Primary and Photochemical Reacted Exhaust

Figure 4 shows the PM composition (sum of primary emissions and secondary formation potentials as emissions) in the diluted exhaust gas after it was subjected to UV irradiation for 5 h. The graph also summarizes the primary emissions of NMHC, NO_X, and NH₃ (taken from Table 1) and the secondary formation potentials of SOA, NH₄NO₃, aerosol acidity, and O₃, in order to allow comparison of overall emissions, including primary emissions and secondary formation potentials.

The repeatability of the particle mass composition measurements (n = 2) at 23 °C was 9.3% for EC, 5.6% for POA, 11.0% for SOA, 13.4% for NH₄NO₃, and 5.1% for O₃; at 0 °C it was 37.7% for EC, 72.0% for POA, 7.6% for SOA, 24.9% for NH₄NO₃, and 30.0% for O₃; and at −7 °C it was 42.8% for EC, 9.9% for POA, 58.2% for SOA, 7.7% for NH₄NO₃, and 14.1% for O₃.

The emission percentages of the particle mass composition varied slightly with temperature, with the secondary particles—NH₄NO₃ (79.6% at 23 °C, 58.8% at 0 °C, and 38.3% at −7 °C) and SOAs (19.2% at 23 °C, 35.7% at 0 °C, and 47.5% at −7 °C)—comprising 79.6%, 58.8%, and 38.3% of the particles, respectively. The emission percentages of the total primary particles—EC, POA, and POA—were 0.8%, 4.0%, and 12.3% at 23 °C, 0 °C, and −7 °C, respectively, and 0.4%, 1.5%, and 1.9% at 23 °C, 0 °C, and −7 °C, respectively.

The PM composition after UV irradiation varied with temperature. Among the primary particles, EC accounted for 0.8%, 5.4%, and 12.3% of the total PM composition at 23 °C, 0 °C, and −7 °C, respectively, and POAs accounted for 0.4%, 1.3%, and 1.9%, respectively. Among the secondary particles, NH₄NO₃ accounted for 79.6%, 46.7%, and 38.3% of the total PM composition at the three temperatures, and SOAs accounted for 19.2%, 46.6%, and 47.5%, respectively. Thus, the majority of aerosols remaining after the photochemical reaction were secondary aerosols of NH₄NO₃ and SOA. Overall, the primary particles accounted for only 1.2%, 6.7%, and 14.2% of the total PM at 23 °C, 0 °C, and −7 °C, respectively.

Recent emission controls targeting primary particles have generally been based on worst-case scenarios, such as under conditions of very low temperature (e.g., −7 °C, e.g., [58]). However, the present findings show that when secondary particles formed by atmospheric photochemical reactions are considered, the worst-case scenario for total PM emissions is actually at 23 °C and that under this scenario, the PM emissions are dominated by secondary particles. In addition, our results shown in Figure 4 indicate that NH₃ mitigation resulted in significant aerosol acidity formation in the low-temperature environment. Further studies are needed to elucidate the impacts of vehicle type and humidity conditions on the secondary particle composition of vehicle exhaust.

3.3. Effects of Ammonia Mitigation

3.3.1. NH₄NO₃ Particle Formation

As shown in Figure 4, NH₃ mitigation by using a series of five NH₃ denuders had a marked effect on the amount of secondary NH₄NO₃ particles formed during UV irradiation. Without NH₃ mitigation, NH₄NO₃ particles were emitted at 208 ± 28, 74 ± 12, and 49 ± 4 mg kg-fuel⁻¹ at 23 °C, 0 °C, and −7 °C, respectively, whereas with NH₃ mitigation the emissions were 2, 3, and 19 mg kg-fuel⁻¹, respectively, which was a considerable reduction. Thus, the NH₃ removal efficiency of the series of denuders was 85% at 23 °C, 98% at 0 °C, and 98% at −7 °C, and the reduction of NH₄NO₃ particle formation was 99%, 96%, and 61% at 23 °C, 0 °C, and −7 °C, respectively. These findings show that although NH₃ mitigation was effective at reducing NH₄NO₃ particle formation, the reduction was not necessarily linear with environmental temperature. In addition, although we do not yet have a complete explanation for the discrepancy between the collection efficiency of the denuders and the reduction rate of NH₄NO₃ particle formation, we consider that it is likely due to the rather complicated chemical equilibrium of NH₄NO₃ particle formation, as described below.

NH₄NO₃ is formed when nitric acid (produced by the oxidation of NO_X) reacts homogeneously with gaseous NH₃. It has been noted that the secondary formation of NH₄NO₃ from gasoline vehicle emissions is due to the presence of NH₃ and NO_X in the emissions [59]. It is also known that nitrate radicals (NO₃) and dinitrogen pentoxide (N₂O₅) are involved in the formation of nitric acid (HNO₃) gas from NO_X [60]. The rate of photolysis of NO₃ radicals by visible light (wavelength 420–690 nm) is about 10 times that of nitrogen dioxide; therefore, the atmospheric NO₃ radical concentration is very low during the daytime (i.e., under conditions involving visible light) [61].

$${\mathrm{N}\mathrm{O}}_3·+h\nu (\lambda 420-690\mathrm{n}\mathrm{m})\to \mathrm{N}\mathrm{O}+{\mathrm{O}}_2$$

(4)

$${\mathrm{N}\mathrm{O}}_3·+h\nu (\lambda 420-690\mathrm{n}\mathrm{m})\to {\mathrm{N}\mathrm{O}}_2+\mathrm{O}\left({}_{}{}^3\mathrm{P}\right)·$$

(5)

As demonstrated in Figure 1, the light spectrum of the mobile photochemical smog chamber differs from that of sunlight in that there is less of the spectrum above the long daytime wavelength >380 nm, resulting in slower photolysis of NO₃ radicals. Recently, however, it has been found that marked amounts of nitrate are present during the evening and morning twilight hours and occasionally during the day when light levels are low [62]. Studies assessing the importance of this nitrate radical generation during the day have also been published [63,64]. During the night, NO₃ radical is produced by the reaction of NO₂ with O₃; NO₃ radical reacts with NO₂ to produce N₂O₅, and the produced N₂O₅ then reacts with liquid water droplets on aerosol particles to produce HNO₃:

$${\mathrm{N}}_2{\mathrm{O}}_5+{\mathrm{H}}_2\mathrm{O}\to 2\mathrm{H}\mathrm{N}{\mathrm{O}}_3$$

(6)

When HNO₃ is present in the atmosphere, it tends to react with basic species such as NH₃ gas. The neutralization reaction between NH₃(g) and HNO₃ gas (HNO₃(g)) to form NH₄NO₃ particles (NH₄NO₃(p)) is reversible and is considered the main source of particulate nitric acid aerosols (NH₄NO₃(p)) in urban air [65]. The formation of NH₄NO₃ particles was confirmed in this study (Figure 4), implying that HNO₃ gas is produced according to Equation (6). Consequently, the mobile photochemical smog chamber used in the present study, which uses UV light sources, reproduces not only the daytime response but also that during the twilight hours from evening to morning and the daytime when light levels are low. The present approach, therefore, will be useful for the evaluation of the worst-case scenario of secondary pollutant formation that includes both mechanisms. The reaction for the formation of NH₄NO₃ is as follows:

$$\mathrm{N}{\mathrm{H}}_3\left(\mathrm{g}\right)+\mathrm{H}\mathrm{N}{\mathrm{O}}_3\left(\mathrm{g}\right)\leftrightarrows \mathrm{N}{\mathrm{H}}_4\mathrm{N}{\mathrm{O}}_3\left(\mathrm{p}\right)$$

(7)

The equilibrium constant for the reaction in Equation (7) depends on the gas concentration, relative humidity, and temperature [66,67,68]. The formation of particulate NH₄NO₃ is enhanced under conditions of high gas concentration, high relative humidity, and low temperature [67]. Aqueous ammonium nitrate exhibits temperature dependence, and the amount of particulate ammonium nitrate is determined from the amount above the equilibrium concentration of HNO₃ and available NH₃. Aqueous NH₄NO₃ also exhibits a temperature dependence, and the amount of particulate NH₄NO₃ is determined from the concentration of HNO₃ and NH₃ above the equilibrium of Equation (7). As shown in Figure 4, only NH₄NO₃ particles tended to be reduced by selective NH₃ mitigation; the reason is that the equilibrium reaction is not established due to the elimination of NH₃ gas on the left side of Equation (7).

Figure 5 shows contour plots of the emission factors of NH₄NO₃ particles formed in the equilibrium reaction in Equation (7) versus the emission factors of HNO₃ and NH₃ gases, as determined in the present study. The contour plots were calculated in ISORROPIA [47] using relative humidity and the concentrations of HNO₃ and NH₃ precursor gases in the reaction bag. The NH₃ gas was the initial concentration before the photochemical reaction, and the HNO₃ gas was the calculated concentration after the reaction (see Section 2.2.5). The concentrations of NH₄NO₃ particles, NH₃ gas, and HNO₃ gas were each converted to an emission factor, which was then used to create the contour plots shown in Figure 5. The contour plots do not show significant differences, indicating that the influence of water on the process in the presence of NH₃ did not play a very important role in the wide range of NH₃ and HNO₃ concentrations obtained in our study with respect to changes in humidity and temperature. The deliquescence relative humidity (DRH) of NH₄NO₃ particles is 61.8% at 298 K and increases exponentially with decreasing temperature [69]. The relative humidities listed in the present experimental procedures (see Section 2.2.4) are lower than the DRH of NH₄NO₃ particles. The implication here is that the process of NH₃ gas uptake into the deliquescent droplet particles does not contribute markedly to the gas-to-particle equilibrium in Equation (7).

The equilibrium constant for the reaction in Equation (6) shows that the formation of NH₄NO₃ particles is less temperature-dependent when the emissions of HNO₃ and NH₃ are sufficiently high. In the present study, when NH₃ mitigation was used, the concentration of NH₃ was reduced (i.e., the plots in Figure 5 were shifted toward the y-axis), and thus the formation of NH₄NO₃ particles was reduced at all three temperature conditions (i.e., 23 °C, 0 °C, and −7 °C) compared to the base scenario without NH₃ mitigation. Compared to results at 23 °C, the HNO₃ concentration produced from NO_X emissions at 0 °C and −7 °C tended to decrease (i.e., the plots in Figure 5 were shifted toward the x-axis) because of a slowing of the photochemical reaction.

In the measurement concentrations of NH₄NO₃ particles, a denuder is sometimes used in filter sampling to reduce interference from measurement artifacts such as adsorption or evaporation of ammonia or nitric acid gas. In the present study, although evaporation of NH₄NO₃ particles was a concern (see Equation (7)), we did not perform any special sampling. This was because the purpose of the study was to examine the relative reduction of NH₄NO₃ particles with the reduction of NH₃ gas, not to obtain exact measurements of NH₄NO₃ particle production. In Figure 5, the measured value of the NH₄NO₃ particles agrees with the calculated value if the color of the plot is the same as the background contour plots. Not all plots matched the calculated values, tending to be slightly overestimated at 23 °C and slightly underestimated at 0 °C and −7 °C. In this study, the concentration of HNO₃ gas was not directly measured, and NH₃ gas is generally quite difficult to accurately measure due to its sticky nature. Therefore, the consistency between the calculated and observed NH₄NO₃ values may be attributed to experiments based on limited resources for measuring the precursor gas NH₃ and calculating HNO₃. Further research should clarify the consistency with the calculated and observed NH₄NO₃ values based on highly sensitive and accurate measurements of the precursor gases. Regardless, our experiments indicated that selective NH₃ mitigation using NH₃ denuders tended to reduce NH₄NO₃ particles relative to those without NH₃ mitigation.

Since the HNO₃ concentration was lower at lower temperatures, the trend of increasing aerosol acidity due to NH₃ removal also tended to be less at lower temperatures (Figure 4). Given that the aerosol formation potential was evaluated under dry conditions in the present study, further studies are needed to evaluate changes in the NH₄NO₃ particle formation potential and aerosol acidity due to NH₃ removal in relation to humidity.

It has been reported that the increase in NH₄NO₃ mass after the photochemical reaction of gasoline vehicle exhaust is due more to the presence of NH₃ than to a reduction in NO_X emissions [22,59]. Gasoline particulate filters (GPFs) with catalysts, an after-treatment device for gasoline vehicles intended to reduce primary PM emissions, are reported to reduce NO_X emissions from the tailpipe by as little as 16.6% [22] or as much as 87.6% [59], and it has been reported previously that more NH₄NO₃ was produced in a photochemical smog chamber experiment with GPFs than without [22]. Together, these previous studies suggest that NH₃ emissions may contribute significantly to the formation of secondary inorganic aerosols, primarily in the form of NH₄NO₃. NH₃ can also be produced in three-way catalysts from NO_X emitted from the engine, and H₂ is produced through the water–gas shift reaction and steam reforming of hydrocarbons [70,71]. Such NH₃ is known to pass through GPF systems or be oxidized to N₂O, NO_X, or N₂ [70,71]. To address this, three-way catalysts are usually coated with precious metals such as Pt, Rh, or Pd on a ceramic or metal substrate. In general, Rh reduces NO_X, whereas Pd or Pt oxidizes CO and CH₄ emissions [72]. The composition of the catalytically active metal, the air/fuel ratio, and the operating temperature all play important roles in the formation of NH₃, which itself is a factor in the secondary formation of NH₄NO₃ particles and of N₂O, a global warming potential. Also, catalysts with Pd/Rh or Pt/Rh as active metals produce NH₃ [73,74] and N₂O [75]. In the present study, we did not evaluate the differences in the formation potential of NH₄NO₃ particles relative to differences in NH₃ emissions. However, it is reasonable to assume that the differences in the ratio of NH₄NO₃ particles to overall PM between the previous studies [22,59] and our study were most likely the cause of the difference in the amount of NO_X and NH₃ in the tailpipe detected in the present study.

3.3.2. Aerosol Acidity

Fewer total PM emissions (sum of primary and secondary particles) were observed with NH₃ mitigation compared to without (Figure 4). These lower total PM emissions are attributed to the fact that with NH₃ mitigation, the nitric acid gas produced by the oxidation of NO_X reacts uniformly with the NH₃ gas to neutralize it and produce fewer NH₄NO₃ particles; with NH₃ removal, the nitric acid gas condenses onto aerosol particles (aq: particles in the liquid phase), increasing acidity (i.e., increasing the net of Equations (8) and (9)):

$$\mathrm{H}\mathrm{N}{\mathrm{O}}_3\left(\mathrm{g}\right)\leftrightarrows {\mathrm{H}}^{+}\left(\mathrm{a}\mathrm{q}\right)+{\mathrm{N}{\mathrm{O}}_3}^{-}\left(\mathrm{a}\mathrm{q}\right)$$

(8)

$$\mathrm{H}\mathrm{N}{\mathrm{O}}_3\left(\mathrm{g}\right)\leftrightarrows \mathrm{H}\mathrm{N}{\mathrm{O}}_3\left(\mathrm{a}\mathrm{q}\right)$$

(9)

The aerosol acidity showed an increasing trend with NH₃ mitigation (red circle plot in the middle of Figure 4), but no change in nitrate gas formation was obtained (blue circle plot in the middle of Figure 4). The decrease of NH₄NO₃ and the increase of aerosol acidity confirmed that no neutralization reaction between nitrate and NH₃ gases occurred. Less HNO₃ gas was produced with decreasing temperature due to a decrease of OH exposure (6.1 × 10⁷ at 23 °C, 3.3 × 10⁷ at 0 °C, and 1.9 × 10⁷ molecules cm⁻³ h⁻¹ at −7 °C), indicating a slowdown in the progress of the atmospheric oxidation reaction. Although there was a small amount of NH₃ gas remaining at 23 °C with 85% NH₃ gas removal, it was neutralized such that there was no marked increase in aerosol acidity. These findings suggest that a small amount of residual NH₃ gas can be neutralized but that excessive NH₃ mitigation promotes acidification. Further studies are needed to assess the human health effects of such increased aerosol acidity, as well as the effects on air quality, rainfall, soil, and vegetation.

3.3.3. SOA Formation

No marked difference in SOA formation potential was observed with NH₃ mitigation (Figure 4). Without NH₃ mitigation, SOAs were emitted at 50.2 ± 5, 73.8 ± 1, and 61.1 ± 36 mg kg-fuel⁻¹ at 23 °C, 0 °C, and −7 °C, respectively, whereas with NH₃ mitigation it was emitted at 63.3, 78.9, and 42.5 mg kg-fuel⁻¹, respectively. The effect of NH₃ on SOA formation has been demonstrated by previous photochemical smog chamber experiments and model analyses, which have shown that ammonium salts formed by the reaction of NH₃ with organic acids in SOAs derived from styrene and α-pinene cause an increase of SOA formation [76] and that NH₃ competes with aldehydes to reduce the yield of secondary ozonides, which decreases SOA formation [77,78]. Although the organic acids and ozonide in photochemically reacted gasoline vehicle exhaust were not quantified in the present study, it is unlikely that they would have an effect on a reaction dependent on the presence of NH₃. It is reasonable to assume that the differences in the values of SOA formation due to NH₃ mitigation obtained in this study were due to experimental variability, given the limited number of experiments.

The trend of SOA formation can be quantified in terms of effective SOA yields (Y), defined as the measured SOA mass divided by the mass of SOA precursors reacted. Since SOA yields vary widely among VOC components [79,80], only a portion of NMHC emissions are SOA precursors. Because of the limited number of laboratory studies available in the literature, SOA production data are not available for all precursors. Although SOA yields remain a subject of debate [79,80], presenting the data as SOA yields accounts for differences in SOA production across experiments. Our estimates of SOA yields (0.372–0.866) shown in Table 1 and Figure 6a varied but were comparable to previously reported values (0.07–0.9, e.g., [25]). SOA yields (Y) have been shown to be a function of SOA concentration (M_o) according to a classical model [81,82,83], and the relationship is described as follows:

$$\mathrm{Y}={\mathrm{M}}_{\mathrm{o}}\sum \left({\displaystyle \frac{{\mathsf{\alpha }}_i{\mathsf{{\rm K}}}_{\mathrm{o}\mathrm{m},i}}{1+{\mathsf{{\rm K}}}_{\mathrm{o}\mathrm{m},i}{\mathrm{M}}_{\mathrm{o}}}}\right)$$

(10)

where K_om,i and i are the mass-based gas-particle equilibrium partition coefficient and stoichiometric coefficient of product i, respectively, and M_o is the total mass concentration of organic matter (μg m⁻³). “M_o” is a common notation in previous studies, and it was obtained in this study by multiplying the dilution- and particle loss-corrected OC concentrations observed in the reaction bag after 5 h of photochemical reaction by the OM/OC_{SOA_T} ratio used in Equation (3). Our effective SOA yield estimates shown in Figure 6a varied considerably, but they plotted in Figure 6 roughly backward and forwards on the SOA yield curves for each environmental temperature (23 °C, 0 °C, and −7 °C). The higher SOA yields at higher OA concentrations reflect the partitioning of the semi-volatile oxidation products in the SOAs into the particulate phase at higher OA concentrations.

The scatter plots in Figure 6b compare the emission factor of the SOA production potential with that of the VOCs consumed in the reaction (ΔNMHC). The straight lines in Figure 6b are the regression lines of SOA vs. ΔNMHC and, therefore, indicate the expected SOA yield. At 23 °C and −7 °C, the plots lie mostly on the regression lines, indicating that reasonable experimental results for SOA yield were obtained; at 0 °C, the plots deviate from the straight line, indicating that SOAs and NMHC were overestimated and underestimated, respectively. At 0 °C, the relative standard deviation (i.e., the scatter) of the three experiments, including NH₃ reduction, was 4% for SOAs and 15% for NHHC, suggesting that the measurements were not significantly different, and the experiments were well reproduced. Figure 6c, which shows a scatter plot of the emission factors of the SOA production potential for the NMHC emission factors, also shows that the scatter at 0 °C was minimal. Comparing SOA yields with OH exposure (Figure 6d), the SOA yields at 0 °C showed some variation, while the OH exposure remained somewhat constant. At 23 °C and −7 °C, the OH exposure, and SOA yields were less variable than at 0 °C; at −7 °C, the OH exposure was more variable than the SOA yield, whereas at 23 °C the OH exposure and SOA yield were both variable to the same degree. Although the reproducibility of the SOA measurements at 0 °C was good, the results suggest that SOA yields varied due to measurement error, although this is not relevant to understanding the rationale for the SOA yields obtained; further systematic, large-scale studies will be needed to settle the SOA yield debate.

With NH₃ mitigation, the obtained SOA yields may be judged as deviating somewhat from the SOA yield curve obtained for the 0 °C condition; however, based on the relationship between the NMHC and SOA formation potentials (Figure 6c), it is reasonable to interpret this as simply the variation obtained from the series of experiments. The relationship between SOA yield and temperature remains a subject of debate, with reports of both higher [13] and lower [25] SOA yields under low-temperature conditions for SOAs produced from gasoline vehicle exhaust. Considering a gas–particle equilibrium based on the concept of effective evaporation enthalpy of a liquid becoming a gas (e.g., [84,85]), the higher SOA yield at low temperatures is considered to be a natural phenomenon in which less volatile gases condense into the particle phase, leading to more particle formation.

Our estimated yield of SOAs at 23 °C was higher than previously reported values (lower SOA yields of 0.07–0.7 with lower OH exposure of 0.1–1.5 × 10⁷ molecules cm⁻³ h⁻¹ in Table 1). We attribute this difference to a higher OH exposure (higher SOA yields of 0.731–0.866 with higher OH exposure of 6.0–6.3 × 10⁷ molecules cm⁻³ h⁻¹), and many studies support a higher SOA yield with higher OH exposure (e.g., [21]). In the present study, we found that the lower the temperature, the lower the OH exposure. We believe that the lower SOA yield at lower temperatures is due to the lower SOA formation as a result of the slowing of the oxidation process (Figure 6d). Consequently, our data shown in Figure 6c indicate that the higher NMHC emissions at low temperatures (0 °C and −7 °C), which are often used as worst-case scenarios for atmospheric environmental policymaking, did not lead to the greatest SOA formation potential. They also suggest that the reduction of NH₃ did not lead to a reduction in SOA formation. However, we would like to emphasize that the actual experimental data suggest that the yield of SOAs formed from gasoline vehicle emissions is also highly dependent on environmental temperature conditions.

3.3.4. O₃ Formation

No marked differences in O₃ emissions were observed with NH₃ mitigation (Figure 4). Without NH₃ mitigation, O₃ was emitted at 1053 ± 53, 553 ± 48, and 239 ± 34 mg kg-fuel⁻¹ at 23 °C, 0 °C, and −7 °C, respectively, whereas with NH₃ mitigation it was emitted at 1093, 585, and 234 mg kg-fuel⁻¹, respectively. The effect of NH₃ on O₃ formation can be evaluated through compound-limited photochemical smog chamber experiments; however, few such studies exist in the literature. In one study, a photochemical smog chamber was used to investigate the effect of NH₃ on secondary aerosol formation by photooxidation of toluene and NO_X under different O₃ formation regimes, and the study showed that although NH₃ concentration does not affect O₃ formation, it does affect secondary particle formation and composition [86]. Thus, our results indicate that O₃ is only formed by the reaction cycle between VOCs and NO_X, and NH₃ concentration has no marked effect on O₃.

The O₃ formation potential (OFP) index, which quantifies the relative impact of individual VOCs on O₃ formation, has been widely used to help develop cost-effective ground-level ozone pollution control strategies [87,88]. For a given VOC or VOC mixture, OFP is determined by using the maximum incremental reactivity (MIR) index [87,88,89,90,91,92,93,94,95,96,97,98]. The MIR index is defined as the gram of change in O₃ per gram of VOC. The index was developed using the Statewide Air Pollution Research Center (SAPRC) chemical reaction model, which was built on a semi-explicit chemical mechanism [89,90,97,98]. Generally, the calculation of the OFP index uses the MIR index developed under high NO_X conditions, thus limiting the O₃-forming region to conditions of limited VOC concentration or at least conditions where VOC and NO_X mixing is limited [91]. For most studies of gasoline vapor emissions and gasoline vehicle emissions with low NO_X emissions (e.g., [88,92,93,94,95,96]), however, the OFP index has been derived by multiplying the MIR index by observed VOC emission factors. Thus, the OFP index (mg-O₃ kg-fuel⁻¹) of emitted NMHC can be calculated using the MIR index [97,98] and Equation (11):

$$\mathrm{O}\mathrm{F}\mathrm{P}\mathrm{i}\mathrm{n}\mathrm{d}\mathrm{e}\mathrm{x}=\sum {\mathrm{M}\mathrm{I}\mathrm{R}}_i\times {\mathrm{C}}_i$$

(11)

where MIR_i is the maximum incremental reactivity of VOC composition i (mg-O₃ mg-VOC⁻¹), and C_i is the emission factor [mg kg-fuel⁻¹] of VOC composition i (VOC type: alkanes, alkenes, aromatics, aldehydes).

Figure 7 shows a comparison of the OFP index (calculated using the VOC concentration in the reaction bag) and O₃ formation potentials as emissions (measured in the reaction bag); the percentage contributions of alkenes and aromatics to the OFP are also indicated. In general, alkenes, aromatics, and aldehydes contribute more to a higher MIR index [96]. In the present study, alkenes contributed 31%, 35%, and 33% to the OFP index at 23 °C, 0 °C, and −7 °C, respectively; aromatics contributed 40%, 43%, and 47%, respectively; and aldehydes contributed 12%, 4.3%, and 1.7%, respectively. The distribution of the alkenes and aromatics did not change significantly with ambient temperature. There was almost no change in the contribution of the VOC categories to the OFP index due to the removal of NH₃. The calculated OFP index results are interpreted as being representative of the relative O₃ formation potential as emissions from different fuel compositions. They do not suggest the possibility of changing ozone concentrations in urban areas [94]. In the present study, the OFP index did not agree with the O₃ formation potential; assessment by the OFP index generally requires the use of detailed atmospheric chemistry models that account for many important additional factors (such as local meteorology and all sources of ozone precursors) [94]. Because the present study was too small in scope (i.e., single vehicle and fuel type), we could not conclude that the observations conclusively explained the performance of the technologies considered. However, the present observations contribute to our understanding of the potential for changes in the composition of vehicle emissions to have a positive effect on the suppression of atmospheric ozone formation. That is, our findings emphasize that the ratios of VOCs contributing to the OFP index are largely independent of ambient temperature and the presence of NH₃ mitigation.

4. Conclusions

The formation potentials of secondary particles and O₃ from gasoline vehicle exhaust were examined at different temperatures with or without NH₃ mitigation. Among the total PM emitted, which included that produced by photochemical oxidation reactions, POAs, and EC accounted for only a small fraction, whereas the contribution of the secondary particles NH₄NO₃ and SOAs was dominant. The yield of SOAs was lower at lower temperatures. In a parameter sensitivity analysis, using a denuder to selectively reduce the concentration of NH₃ gas in the vehicle exhaust was found to have a marked effect on reducing the formation of NH₄NO₃ but not of SOAs or O₃. Increased aerosol acidity was also observed with NH₃ mitigation. Overall, the present study highlights the importance of using photochemical smog chamber experiments to gain an informed understanding of the potential toxic effects and atmospheric and environmental impacts of vehicle emissions when implementing source control measures such as NH₃ emission limits.