Atmospheric Chemistry in a Box or a Bag

Abstract

Environmental chambers have proven to be essential for atmospheric photochemistry research. This historical perspective summarizes chamber research characterizing smog. Experiments with volatile organic compounds (VOCs)-nitrogen oxides (NO_x) have characterized O₃ and aerosol chemistry. These led to the creation and evaluation of complex reaction mechanisms adopted for various applications. Gas-phase photochemistry was initiated and developed using chamber studies. Post-1950s study of photochemical aerosols began using smog chambers. Much of the knowledge about the chemistry of secondary organic aerosols (SOA) derives from chamber studies complemented with specially designed atmospheric studies. Two major findings emerge from post-1990s SOA experiments: (1) photochemical SOAs hypothetically involve hydrocarbons and oxygenates with carbon numbers of 2, and (2) SOA evolves via more than one generation of reactions as condensed material exchanges with the vapor phase during “aging”. These elements combine with multiphase chemistry to yield mechanisms for aerosols. Smog chambers, like all simulators, are limited representations of the atmosphere. Translation to the atmosphere is complicated by constraints in reaction times, container interactions, influence of precursor injections, and background species. Interpretation of kinetics requires integration into atmospheric models addressing the combined effects of precursor emissions, surface exchange, hydrometeor interactions, air motion and sunlight.

1. Introduction

Atmospheric processes are a blend of meteorological phenomena, air chemistry and interactions with the hydro-, pedo- and biosphere. Natural chemical processes are further complicated by the addition of anthropogenic gas and particle sources emitted into the atmosphere. Contemporary reductionist philosophy seeks separation of these components for a simplifying strategy to characterize parts of widely different elements of atmospheric phenomena in time and space. Research strategies have applied separate, fundamental results of physics, chemistry and biology to construct conceptual models leading to current understanding of the atmospheric chemical dynamics. An adjunct to these applications includes laboratory simulation of different processes believed to be relevant to the atmosphere [1]. Contemporary simulation technology has recognized the spatial and temporal scale of atmospheric processes from the microscale, including chemical kinetics and cloud micro-physics, to planetary scale fluid dynamics of rotating stratified fluids.

Simulation of processes indirectly relevant to atmospheric chemistry began with laboratory experiments of J. Tyndall in 1868 [2] followed with C.T.R. Wilson’s 1897 cloud chamber studies [3]. Tyndall experimented with particle production in irradiated gases in a tubular reactor. The gases exposed to light revealed the potential for photochemical processes, perhaps the first hint of secondary aerosol production possible from irradiated trace species in air. Wilson’s experiments elucidated the nature of homogeneous nucleation in supersaturated vapors and found the mediating effects lowering supersaturation that were required for vapor condensation by ions. These experiments partially formed the basis for several atmospheric studies of ions, nuclei formation and cloud formation.

Physicochemical phenomena in the atmosphere depend on the interactions of processes involving the atmosphere’s energy balance and distribution, fluid dynamics, cloud physics and chemistry, and underlying (heterogeneous) surface exchange. Expanding knowledge of photochemical processes in the atmosphere called attention to the potential for experimentally separating chemical processes from these other processes. The genesis of simulating tropospheric photochemistry began with suspicions of the presence of oxidants (ozone) in the Los Angeles atmosphere [4]. In the 1950s, Haagen-Smit’s [5] famous 2-liter flask experiments surfaced confirming the presence of ozone (O₃) in smog, and empirically exploring the role of nitrogen oxides (NO_x = NO + NO₂) and hydrocarbons (HC) in oxidant chemistry. These studies aroused the regulatory concern of both the local pollution control authorities and industrial or motor vehicles representatives leading early on to the creation of a cooperative research entity in 1954, the Air Pollution Foundation. Renzetti’s review [6] describes a succession of early laboratory experiments following up on Haagen-Smit’s experiments. Subsequent studies of the smog photochemistry expanded in chambers up to 50 liters volume. In 1961 a major, sustained US commitment to gas-phase smog chamber studies was implemented at the University of California Riverside (UCR) Statewide Air Pollution Research Center (SAPRC, later renamed APRC). This initiative fostered the design and development of major smog chamber facilities.

A complementary concern from urban observations of smog was the presence of suspended particles leading to severe visibility impairment and the potential for exacerbation of ischemic or specific respiratory disease [7]. Particle concentrations following the diurnal changes in photochemical smog pointed to a chemical link distinct from smoke or dust emissions. This conjecture was verified early on in laboratory experiments of Renzetti and colleagues [8,9,10], who irradiated auto exhaust in a reactor with and without sulfur dioxide (SO₂). They found that SO₂ was readily oxidized by photochemical oxidants with water to form sulfuric acid along with organic material. Earlier, Mader et al. [11] found organic particles in filtered smoggy Los Angeles air, along with sulfate. The presence of sulfur oxides in polluted air and nucleation forming H₂SO₄ or sulfate salts in the 1950s followed a long history of knowledge dating back centuries [12].

Atmospheric measurements showing photochemical phenomena motivated investigators to begin studying aerosol production in laboratory photochemical systems after the 1960s [13,14,15,16,17]. Photochemical aerosol studies initially focused on sulfate [18,19,20,21]. Later, simulation experiments on organic carbon particles began to appear, for example, [22,23,24,25]. Most of these experiments were conducted in batch photochemical reactors or environmental (smog) chambers that were operated at various institutions.

Following the photochemical research of the 1950s [26], the mechanisms responsible for oxidant formation was hypothesized to involve the O₃-NO_x inorganic cycle often attributed to Chapman [27] combined with free radical interactions of reactive nitrogen (NO_y) and non-methane hydrocarbons (part of volatile organic compounds--VOCs). The free radical chemistry, especially for hydroxyl radical (OH), was determined to be critical in chain reactions leading to maintaining oxidant concentrations beyond the Chapman cycle.

Conceptually, the rapidly evolving post-1960s photochemical kinetic studies fostered the interest in using smog chambers as a means of study of atmospheric photochemistry. Simulating microscale processes relevant to the atmosphere included the complex chemistry of trace gases and aerosol particles in the air [28,29]. In the 1960s, the studies of gas-phase and aerosol chemistry paralleled one another. Gas-phase photochemical processes involving NO_x and VOC relevant to urban conditions took precedence through the 1980s with combined bench scale experiments and smog chamber research actively pursued at the at SAPRC, for example, and later at the University of North Carolina, Carnegie Mellon University and the US Environmental Protection Agency (EPA). Experiments for aerosol photochemistry took place using indoor or outdoor chambers at SAPRC (later the UCR Center for Environmental Research and Technology, CE-CERT), the California Institute of Technology and Carnegie Mellon University. These paralleled similar research at the University of North Carolina, and in Europe and Asia [30,31,32,33,34,35].

Smog chamber experiments have complemented laboratory chemical kinetics research and enabled the verification of various smog chemistry mechanisms for air quality models, as well as for the exploration of source emission controls on photochemistry in different environmental conditions [28,36]. Aerosol chamber studies have advanced the knowledge of secondary particle production from gaseous species in different environments [37].

The goal of this narrative is to first summarize the technical capabilities of environmental simulation chambers required for experimentation, and other studies that have qualified chamber results relevant to the atmosphere. A survey is given of chamber facilities operated throughout the early 21st century. To illustrate the importance of chamber research, a summary is provided for some of the many contemporary contributions to knowledge of smog chemistry resulting from these experiments. Atmospheric relevance of smog chambers for kinetics studies and as simulators end the narrative. The details of research leading to various gas-phase photochemical-aerosol mechanisms are left to the wisdom derived from many reviews in the literature exemplified in

Table 1. Some reviews of studies relevant to atmospheric chemistry (2000–2019).

Review	Year	Content
Finlayson-Pitts and Pitts [28]	2000	Textbook-comprehensive review of gas and aerosol chemistry relevant to the atmosphere through the 1990s.
Dodge [36]	2000	Atmospheric chemistry of O3 with analysis of smog chambers and their results.
Atkinson [38]	2000	Photochemical mechanism review with relevance to atmospheric chemistry.
George et al. [39]	2005	Comprehensive review of heterogeneous atmospheric chemistry
Seinfeld and Pandis (3rd ed.) [29]	2006	Textbook-comprehensive gas and aerosol chemistry with summary of meteorological and climate relevance through early 2000s.
Kroll and Seinfeld [40]	2008	Review of photochemical aerosol formation with reference to smog chamber studies through mid- 2000s.
Hallquist et al. [37]	2009	Review of secondary organic aerosol (SOA) formation including photochemical processes.
Carlton et al., [41]	2009	Secondary organic aerosols (SOA) from isoprene oxidation.
Lim et al. [42]	2010	Aqueous chemistry and formation of SOA.
Stockwell et al. [43]	2012	Chemical mechanisms and models for O3.
Glasius and Goldstein [44]	2016	Review of contemporary atmospheric organic chemistry status and future emphasizing SOA.
Bianchi et al. [45]	2019	Review of autoxidation and highly oxygenated organic molecules (HOM) in atmospheric chemistry.

.

2. Requirements for Chamber Technology

Simulation of atmospheric chemical processes involves separation of an array of interactions that result in ambient concentrations of gaseous and particulate species. These interactions involve phenomena that range of time scales from a few seconds to several days, and spatial scales from less than a cm to km’s. They include rapid homogeneous or heterogeneous reaction pathways superimposed on a “background” of ever-present trace species concentrations, the intermittent presence of aerosols and hydrometeor clouds, contaminant emissions, dilution via mixing of air with different properties. Simulation schemes obviously apply to conditions favoring extremely small time and space constraints. They are designed to isolate “pure” chemical phenomena from the many other phenomena taking place in the atmosphere and the atmospheric boundaries. This approach asserts that such a separation leads to descriptions from building blocks relevant to the atmosphere; in so doing, they advance our understanding of the atmosphere, and enable a predictive capability from integrated process modeling (see also Section 4).

The study of atmospheric photochemistry derives from observations and early experiments of Haagen-Smit and colleagues, and scientists affiliated with the Los Angeles Air Pollution Control District. These studies focused on the NO_x-VOC system, air pollutant agents from a variety of sources. Early on, a decision was made to separate the gas-phase processes from aerosol processes even though they were obviously observationally coupled in the atmosphere. The separation of gas and particle processes in the 1950s evidently followed from a conclusion that visibility and haze were connected to smoke sources distinct from photochemical reactions. By the 1970s the coupling between haze and gas-phase photochemistry was widely recognized as an issue in characterizing smog.

Observations of the atmospheric NO_x-VOC photochemical cycle showed that simulation would require several hours to trace the chemical reactions to some end point past a maximum oxidant concentration. This precluded most laboratory designs that focused on kinetics on a scale of minutes or less even though flow reactors with short gas residence times 100–250 seconds have been used to obtain chemical information relevant to the atmosphere [46,47,48,49].

The value of chamber simulations lies with the ability to measure under controlled conditions the complex chemical kinetics taking place in as much detail as instrumentation will permit. Since the 1950s, gas-phase measurement technologies have improved dramatically for extracted samples as well as continuous in situ observations. Advances in measurement methods, including spectroscopy and chromatography, have supplanted early use of wet chemical methods and Haagen-Smit’s [5] rubber cracking rate for ozone (O₃) concentration changes. Continuous aerosol particle measurement capability has followed gas-phase technology closely in recent decades, complementing extractive methods of substrate sampling and batch chemical analysis.

2.1. Ideal Design Considerations

A high-fidelity simulation of atmospheric photochemistry requires that several conditions be considered. These include:

• Initial and endpoint concentrations of reactive species should approach those observed in the atmosphere.
• Initial input of gas mixtures (or aerosols) should be well-defined so that spurious reactions caused by reactant contamination are controlled entering the vessel.
• The light source for irradiation should simulate sunlight, or at a minimum, the light wavelengths known to be key for excitation of airborne species.
• The duration of an experiment should approach at least most of a day or a span of several hours.
• Air in the chambers should be well mixed to minimize concentration gradients in the chamber.
• Documentation and control of environmental variables like pressure, humidity and temperature should be assured or variability at least documented.
• Loss of reactants, degassing or byproducts of surface reactions at the vessel walls should be accounted for as an influence on the measured reactant-product profiles.
• Instrumentation for characterizing reactions should be as complete as practical from speciated VOCs and NO_y (reactive nitrogen species) to oxidants, and particle physicochemical characteristics or other stable products and inference of free radical or other intermediate species.

The designers of simulators have responded to these requirements with approaches adopting batch reactors (chambers) or flow reactors with nominally inert surfaces which are transparent to light below ~<450 nm wavelength. Chambers have been located either indoors as part laboratory capabilities or outdoors. They have been constructed of pyrex (borosilicate glass), quartz, stainless steel or Teflon films that are transparent to light to sunlight, particularly in the range of 200–450 nm wavelength, or artificial lighting from sources that simulate sunlight. Finlayson-Pitts and Pitts [28] summarize the pros and cons of these materials including their transparency to radiation and wall effects discussed below.

Simulation chambers have been deployed outdoors or indoors using light sources that include sunlight, black light from filtered mercury lamps, sun lamps, and xenon or argon arc lamps. Light intensities varying with configuration require measurement inside the chambers to document the wavelength coverage of the source, and require periodic verification of intensity compared with sunlight. Some of the indoor chambers include temperature and relative humidity (RH) control, and vacuum capability to expand the range of experiments accessible in the chamber. Outdoor chambers generally exclude temperature and RH control. Some chambers are capable of injecting water vapor as a means of RH variation during an experiment. Chambers have been operated as static vessels or dynamic systems with gas flow to simulate fresh reactant injection during the aging of the irradiated mixture.

The reactant and product concentrations need to be constrained in experiments such that the reactant gases are in the ppb-ppm range and O₃ maxima are less than ~400 ppb and NO_y should be similarly constrained. Free radical concentrations or stable intermediates in chambers will be far lower, challenging investigators to measure them either directly or inferring them through tracers. If particle formation is investigated, concentrations of 1–100 µg/m³ should be present as products. Desired measurements include continuous observation of total number concentration of particles and particle size distributions.

Concentrations of gases injected into chambers are historically higher than current urban atmospheric conditions, but are similar to observations in Los Angeles in the 1970s as an upper limit. The input gases for simulation chambers are blends containing pure NO_x or mixtures of NO_x and VOC with super-purified air to define the starting point of a sequence of reactions. Injection of gases can include outdoor air, source emissions such as internal combustion engines, or purified gas streams to investigate reactions associated with single VOC species or mixes that are similar to those from sources. Chambers are designed with the means of purifying inlet gases, including air, to remove NO_x and VOC contamination to less than ppb levels.

Concentrations of NO_x and VOC reactants are varied to achieve different VOC/NO_x ratios in excess of 2. This ratio is well known to be an important parameter for establishing conditions near ambient, and limits of conditions of VOC or NO_x sensitivity [28].

Studies of secondary organic aerosol (SOA) formation involve injection of specific VOC or mixtures of VOCs, beyond those usually adopted for oxidant formation. Also, the capability for adding “seed” aerosols for catalyzing new particle formation, or atmospherically representative particles to investigate the heterogeneous growth of particles. Other injection or generation capabilities include adjustment for humidity and “background or baseline” aerosols. During or after injection of the gas mixture needs to be well-mixed by fans or other methods to insure no gradients exist in the reactor volume.

Chambers are operated over time periods of a few hours to more than ten hours in various runs, simulating the early morning build-up of smog chemistry through midday and later. The later periods beyond the O₃ maximum are problematic in the potential ambiguities of various reaction pathways occurring with declining sunlight and hourly “aging” of the gas or aerosol mixture. Multiday experiments with dilution have been attempted in outdoor chambers [50], but such experiments are subject to cumulative increasing ambiguities from wall effects and potential air stratification without mixing assistance.

Because of their importance for ensuring the fidelity of homogeneous chemistry believed to represent the atmosphere, two essential features need to be addressed: light sources and wall surface effects. Leighton and later Finlayson-Pitts and Pitts [26,28] discussed the solar radiation spectrum in detail and the actinometric range of interest for tropospheric O₃ chemistry. Finlayson-Pitts and Pitts, and others [28,51,52] summarized the various light sources used in smog chambers, including filtered UV lamps, sunlight, black light and xenon arc lamps. Light sources have different irradiation intensity distributions that constrain their simulation of the solar spectrum. For example, black light lamps are widely used because they are inexpensive and have a good UV cutoff at ~300nm. However, they do not represent the solar light intensity well in the visible regime, which is potentially important for photolysis [53]. Darnall et al. [54] noted the differences expected from constant and diurnal varying light. Light sources have been studied extensively to obtain chamber-specific data with light conditions to constrain the gas-phase reaction progress to take account of these variables.

Wall effects are of major concern for in tracing particle forming reactions in chambers. These effects not only involve the biases established from gas-phase NO_x-VOC-oxidant reactions, but also interactions involving volatile and non-volatile products in the formation, coagulation or evaporation of aerosol particles, as well as physical removal of particles by deposition. For gas-phase systems, wall effects have been studied in detail to determine ways for correcting the observations for such effects. Dodge et al. [55] and Bufalini et al. [56] discussed the implications of contamination in chambers as a concern. Lonneman et al. [57] and Kelly et al. [58] investigated VOC contamination from Teflon films. Grosjean [59] noted wall losses as an uncertainty in chamber operations, and Yarwood et al. [60] examined NO_x wall losses as part of more recent chamber experiments. Trump et al. [61] constructed a model for chamber losses. Empirically, the uncertainties associated with wall interactions can be addressed by baking out chambers [62], or treatment with dilute concentrations of O₃.

Particularly concerning for atmospheric simulation is the potential for free radical interactions from chamber walls, especially the HONO disassociation to give a spurious OH supply. Carter et al. [63] hypothesized a release of free radicals from chamber walls with temperature and irradiation. Carter et al. [64,65] investigated this question further. Their studies were complemented with work of Besemer and Nieboer [66] and Sasamaki and Akimoto [67]. Rohrer et al. [68] noted the release of HONO from chamber walls as a continuing issue from the use of chambers. Iraci et al. [69] proposed a series of experiments that could add more information about gas-phase wall interactions.

As a practical measure, the influence of wall effects of deposition and adsorption–desorption of NO, NO₂ and O₃, or release of free radicals (e.g., OH from HONO reactions is said to be reduced by minimizing the surface to volume ratio of the chamber.). Empirical adjustments can be made to account for wall effects in gas-phase studies [61,66]. Alternatively, investigators have designs employing low surface to volume ratios to minimize the influence of walls on the photochemistry.

Experiments of photochemical aerosol formation add complexity of particle interactions with chamber walls along with gas interactions. Aerosols formed in chambers can be lost by diffusion to the wall, or sedimentation of larger particles. Loss of charged particles to the walls from electrostatic attraction also has been documented [70,71]. Vapor and aerosol loss in secondary aerosol formation was discussed early on [72]. More recent studies of organic vapor-particle loss on Teflon surfaces have reported the potential for low bias in particle yields [73,74,75,76,77,78]. For low-volatility or semi-volatile organics, wall loss is complicated by mass transfer and reactions on seed or SOA surfaces, which can be accounted for empirically [79]. With the difficulty in comparing chambers and their results for wall-loss of organics, Brune [80] has proposed an empirical “index” for interrelating chamber conditions. Pierce et al. [81] proposed a theoretically-based model for loss from wall effects, coagulation and condensation-evaporation in chambers. Sunol et al. [82] presented a comprehensive model for characterizing secondary organic aerosol (SOA) formation in chambers, including accounting for wall effects.

2.2. Some Chamber Facilities

The design of smog chambers has taken into account the ideal requirements above in various ways to create practical containments for experimentation. The methodology adopted for design and conduct of experiments are discussed [17]. Since the 1950s, there are several smog chambers that have been constructed for both homogeneous gas-phase reactions and photochemical aerosol studies. These facilities have served a range of experiments for major efforts to develop well-documented data bases for mechanism studies and evaluation. The chambers available before 2000 for gas-phase studies are listed in Dodge [36]. The listing includes examples of both indoor and outdoor facilities with varying volumes and media, NO_x levels and VOC selected for simulating polluted air. Dodge noted more than 1000 individual runs performed at her listed facilities have been available for developing or evaluating gas-phase oxidant forming mechanisms.

After the 1970s, smog chambers also were used to follow up earlier investigations of photochemical aerosol formation that were assumed to be relevant to the atmosphere. The post-1980s focused much more on secondary organic aerosols (SOAs) rather than the inorganic species, sulfate and nitrate.

Table 2. Some Smog Chambers Used for Photochemical Gas-Aerosol Studies (updated table from Hallquist et al. [37]).

Location	Type	Volume (m3)	Material	Temperature (°K)	Reference a
California Inst. of Technology	Indoor photoreactor black light	28 (dual)	Teflon (FEP)	290–303	Cocker [86]; Ng [87]
California Inst. of Technology	Outdoor photoreactor	65	FEP	ambient	Leone et al. [88]
Carnegie Mellon Univ.	Indoor photoreactor black light	10	Teflon (PTFE/FEP)	288–313	Stanier [89]; Robinson et al. [90]
Forschungazentrum Julich, Germany (SAPHIR)	Outdoor photoreactor	270	FEP	ambient	Rohrer et al. [68]
Forschungazentrum Julich, Germany	Dark chamber	250	PTFE/FEP	ambient	Mentel et al. [91]; Saathoff et al. [92]
Forschungazentrum Karlsruhe, Germany (AIDA)	Dark chamber (clouds)	4–84	Metal	183–323	Saathoff et al. [93]; Jonsson et al. [94]
Fundacion Centro de Estudios Ambientales del Mediterranean Spain (EUPHORE)	Outdoor photoreactor	200	FEP	ambient	Rodenas et al. [95]; Klotz et al. [96]
Leibniz Institute for Tropospheric Research Germany	Indoor photoreactor black light	19	FEP	289–308	Iinuma et al. [97]
Paul Scherrer Insitute Switzerland	Indoor photoreactor xenon arc	27	FEP	290–298	PSI [30]; Paulsen et al. [98]
University College Cork Ireland	Indoor photoreactor black light	6.5	FEP	293–305	Temime et al. [99]; Healy et al. [100]
University of Manchester UK	Indoor photoreactor Halogen/xenon arc	18	FEP	288–313	Centre for Atmos. Sci. [101]
University of California Riverside (CECERT) US	Indoor photoreacto xenon arc/black light	90 (dual)	FEP	278–323	Carter et al. [85]; Song et al. [102]
University of California Riverside (APRC) US	Photoreactors (several indoor and outdoor) xenon arc/black light	6–8	PTFE/FEP	ambient	Dodge [36]; Tobias and Ziemann [103]
University of North Carolina US	Outdoor photoreactor	120; 137 (dual); 150 (dual)	FEP	ambient	Jeffries [104,105,106]; Lee et al. [107]
US EPA	Indoor photoreactor Black light	14.5	FEP/TFE	293-298	Edney [108]
CSIRO	Outdoor photoreactor	20 (dual)	FEP	ambient	Wiegand [109]
CSIRO Energy Technol Australia	Indoor photoreactor black light	18	FEP	ambient	Hynes et al. [110];
CSIRO Australia	Indoor photoreactor black light	~24	FEP	ambient	White et al. [52]
CNRS-ICARE (HELIOS), France	Outdoor	90	FEP	ambient	Ren et al. [111]
Nat’l Inst. Environ. Studies Japan	Indoor photoreactor xenon arc	6	PFA	--	Akimoto et al. [62]
Nat’l Inst. Environ. Studies Japan	Indoor photoreactor	18	FEP	ambient	Sato et al. [112]
Tsinghua University China	Indoor photoreactor	2	FEP	283–333	Wu et al. [113]
Key Lab. Organic Geochem. Guangzhou China	Indoor photoreactor black light	30	FEP	283–313	Wang et al., [34]
Key Lab. of Atmos. Inst. Comp. and Optical Radiation China	Indoor photoreactor	0.83	FEP	298	Hu et al. [114]
Kyungpook Natl Univ. Korea	Indoor photoreactor black light	7	FEP	291–306	Babar et al. [35]
Korea Institute of Science and Technol.	Indoor photoreactor black light	5.8 (dual)	FEP	ambient	Bae et al. [31]; Moon et al. [115]

a: References include example studies identified [37]; added references that describe the photoreactors from recent literature.

gives a list of chambers for this application, which are located all over the world. These obviously could also be used for gas-phase kinetics studies as well. Table 2 includes the institution location, the volume of the container, light source, indoor-outdoor configuration, material of construction, operating temperature and references to their descriptions. They range in volume from less than 1 m³ to nearly 300 m³. Some of the facilities have dual (side-by-side) chambers that permit parallel experiments with differing input conditions.

Aerosol experiments have included not only NO_x- oxidant-related VOC mixtures to produce oxidant, but have added a variety of aerosol-forming VOC (single and mixtures [8,83,84]) and SO_x as SO₂− sulfuric acid to seek knowledge of the aerosol-forming kinetics expected in the atmosphere.

An example dual chamber facility constructed at the University of California Riverside is shown in Figure 1 [85].

The photograph in Figure 2 is an example of another indoor facility at the Paul Scherrer Institute (PSI) in Switzerland. This chamber is part of cooperative research programs in Europe (e.g., EUROPHORE [95] or EUROCHAMP [116]). Special applications of chambers also have been reported, including portable bag experiments [117] to attempt to isolate the chemistry controlling oxidations of hydrocarbons, or characterize the reactivity of source emissions [118]. Exploration of in situ biogenic emissions from plants and interactions of plant VOCs with oxidant chemistry [119,120,121]. Other applications have used photochemical reaction products to investigate toxicity to humans [106,122]. Smog chambers have also been used to investigate photochemical pesticide degradation. Many pesticides contain chlorine, so that degradation can release this element to the O₃ cycle [53,123,124].

A second essential element for smog chamber research is the instruments for measurement of gases and aerosol particles. Measurements are sampled for external measurement or sampled in situ. Generic gas-phase experiments have relied on measurement of conventional O₃, NO and NO₂ concentrations and speciation for VOCs including hydrocarbons, oxygenates and organic nitrogen compounds (e.g., peroxyacetyl nitrate—PAN). Methods for measurement of these gases are described [28].

Smog chamber experiments provide a means for identifying precursor composition, intermediate species or organic products using different measurement techniques. These include gas chromatograph-mass spectrometer (GC-MS) characterization of VOCs, and instruments such as Fourier transform infrared spectroscopy (FTIR) or differential optical absorption spectrometry (DOAS). High-sensitivity options for VOC determination also include proton transfer spectrometry (PTS) [125]. Contemporary instruments for continuous monitoring include The Hantzsch instrument [126] for formaldehyde, and long path absorption photometer (LOPAP) for HONO [127]. Peroxide or hydroperoxide species can be determined using chemical ionization spectroscopy (CIMS) [128]. Use of high-performance liquid chromatography (HPLC) with detection using peridase [129] could also be adapted for hydroperoxide determination in smog chambers. Radical species such as OH and HO₂ can be determined using fluorescence assay with gas expansion (FAGE) [130].

Aerosol experiments in chambers add complexity to measurement options. In addition to the gas-phase measurements, particle chemistry needs to be characterized, including physical properties such size distributions and chemical composition [131]. The range of particle size found in experiments (<0.01 µm–~5 µm diameter) has required the use of not only condensation nuclei counters (<~0.05 um), but also differential mass analyzers (0.05 µm–~5 µm). Chemical composition of particles uses extracted samples obtained by filters for laboratory analysis, including inorganic and organic components by ion chromatography (IC), GC-MS, or aerosol mass spectrometry (AMS). Peroxides in particles have also been measured by LOPAP [132].

Organic species from reactants to condensed phase products require a variety of techniques including sampling for derivatization of species and subsequent laboratory analysis. Some of these adopt improvements in mass spectroscopy [133] including so-called “soft” ionization methods for large organic species (atmospheric pressure chemical ionization, API-MS [134], or electrostatic ionization, ESI- MS [135]). High-resolution time-of-flight mass spectrometers-HR T-of-F AMS have been applied to measure both gas-phase and condensed species atmospheric or smog chamber application (e.g, [136,137]). Gross et al. [138] used this instrument to measure continuously oligomers from SOA formation.

Quite apart from measurement capability, issues of sampling organic species from smog in recent years have been addressed It is widely known from ambient air sampling that reactive gas adsorption on tubing needs to be minimized and particles need to be sampled with attention to loss of volatile components of inorganic and organic species [139]. In the same way, sampling from chambers can bias measurements from tube wall losses. The wall loss from Teflon films applies also to various kinds of sample tubing, which in turn can bias low estimates of particle formation yields [140]. To facilitate combined sampling for gases and particles, a filter inlet (FIGAERO) has been adopted for mass spectrometric, chromatographic and other sensor applications. This sampler should make combined gas and particle characterization more efficient [136].

Aside from chemical kinetics evaluation, smog chambers have been used studies and documentation of instrument intercomparisons and interferences. Examples of these include NO₂, HO₂ and OH [141,142,143].

2.3. Limitations of Laboratory Chemical Simulation

There are obvious limitations to the smog chamber characterization of atmospherically relevant chemistry. The first are internal in nature, as discussed above, including simulation of sunlight from artificial light sources, completeness of mixing and chamber surface reactions. The second are external in nature concerning the coupling of chemical processes to other atmospheric phenomena. Internal limitations beyond surface reactions or reactant contamination are summarized next; external ambiguities are discussed below and later in Section 4. Finlayson-Pitts and Pitts [28] noted that the issue of wall effects is believed to be moot by some investigators. This is rationalized in that the relevance to the atmosphere is moderated by the fact that the atmosphere is adjacent to underlying surfaces. Photochemical processes along with aerosols take place with a wide range of surface area per unit volume of air near the ground (having widely varying properties).

Uncertainty introduced in translation from controlled laboratory experiments using defined or known gas mixtures vs. irradiation of urban or rural air is not necessarily well-established; experiments have elucidated this question with dual chambers or interpretation of smog evolution compared with oxidant forming mechanisms. In a sense, a “proof” of the relevance of chamber studies results from a correspondence with atmospheric concentrations and composition over durations including diurnal and longer term averages in urban or rural air (see also Section 4). The same heuristic argument is true for the influence of semi-continuous injections of reactants during experiments. The fact that, at least qualitatively, the diurnal changes in reactants and products are similar to those in the atmosphere is interpreted that this emission injection is accounted for in chemical transport models (see also Section 4).

Uncertainty in the translation of smog chamber experiments to current atmospheric conditions is also related to the latter ambient concentration levels. Emission control of precursor gases over the years has reduced urban concentrations substantially, and rural concentrations are generally much lower than current urban levels. Dodge [36], for example, noted this issue in 2000 and recommended experiments for low NO_x and VOC concentrations and low VOC/NO_x ratios paralleling contemporary conditions.

Another question of smog chamber experiments and the atmosphere is the presence of a “background” of photochemical reactants and products to begin each sunlight cycle. Most chamber experiments focus on highly purified air introduced with known pure VOC and NO gases, or in the case of aerosols pure salts or acids as seed species. In principle, the background issue could be dealt with to first order by conducting gas-phase experiments with a range of reactant concentrations and VOC/NO_x ratios; this precludes an ability to simulate a background of partially reacted species, including free radicals, some of which may be unidentified. The latter may be moot, since these species in aged air are likely to be very short-lived and not relevant to a background residual of a day or more, some of which mixes down to the ground from the upper atmospheric boundary layer or free troposphere.

An illustration of the lifetimes expected for various VOC exposed to OH, NO₃ radicals or O₃ range from minutes to years (e.g., Table 1 from [38]), showing that background photochemistry of VOCs could be factor over multiple time scales. An early attempt to deal with the evolution of smog over a sequence of time scales to a few days was reported [50]. They used a dual outdoor chamber to simulate multiday degradation by a sequence of daily dilution in the chamber. Their results suggest a qualitative diurnal repetition of O₃ forming reactions, but wall effects are likely to cumulatively affect the product concentrations.

The residence time for aerosols in the atmosphere varies from hours to several days depending on particle size and surface deposition rates or cloud or rainout [29]. For background perturbations associated with photochemical aerosol formation, the superposition of photochemical processes on known existing aerosol populations ranging from dust and carbonaceous material combined with inorganic salts poses a challenge for interpretation of the laboratory results to tropospheric aerosols. Nevertheless, the experiments using acid or metal oxide seeds simulates a real-world condition for catalyzing photochemical aerosol formation.

Another possible way to deal with background is to investigate the photochemical reactions of complex mixtures derived from combustion processes for comparison with the so-called “pure” systems. Chamber experiments of this kind have included reactions of internal combustion and diesel engine exhaust, as well as biomass burning [8,77,144,145,146,147,148,149,150,151,152].

Smog chamber experiments generally address “dry” (RH < 70%) gas-phase or seed-catalyzed chemistry. These processes do not account for the potential of heterogeneous or condensed phase chemistry known to take place in the atmosphere. The role of clouds or fog for introducing uncertainty from light variation is exemplified by daily variation in sunlight for outdoor chambers. Perturbation in sunlight from cloud or haze intervention presumably acts to suppress photochemical aerosol formation and could be investigated as part of an experimental protocol. Reactions from microscale interactions with moist particles add complexity to irradiated chemical processing. Hallquist et al. [37] comment on cloud interactions that include the influence of RH on gas-particle partitioning, changes in water content with change in RH of droplets, or ice nucleation with organic content. This list does not address the potential for photochemical oxidant absorption and reaction with organics in droplets [153,154]. Since the Hallquist et al. review, studies using humidity-controlled smog chambers and bulk laboratory methods have been reported [41,42].

3. Knowledge Gained from Chambers

Even with its limitations, chamber simulation of photochemical processes has contributed in many ways to advancing knowledge about the atmosphere. Smog chambers constructed since the 1950s have gone well beyond qualitative characterization of oxidant formation in the presence of VOC and NO_x precursors. A large number of gas-phase kinetic simulations complementing other laboratory techniques have produced quantitative kinetic data for conditions near the atmosphere. Dodge [36] and Finlayson-Pitts and Pitts [28] have reviewed the kinetic and mechanistic studies published through the late 1990s. Smog chamber experiments have provided fertile ground for advancing knowledge of photochemical aerosol formation as well.

3.1. Atmospheric Photochemical Processes

A major contribution of smog chamber data for gas-phase processes involves their use in verification and evaluation of mechanisms relevant to the atmosphere. A large body of kinetic data for oxidation reactions was collected relating to VOCs and NO_y species since the 1950s [38,155,156,157]. These have provided the ingredients constructing detailed mechanisms for oxidant production [158,159,160,161,162]. Potentially relevant reactions could involve more than 20,000 reaction pathways with thousands of trace species. Jenkin et al. [161] and Saunders et al. [162] constructed a gas-phase mechanism (Master Chemical Mechanism--MCM v3) involving NO_y and 120 VOCs which translated to 7000 reactions with 2500 species.

Detailed mechanisms required too many computer resources for most atmospheric modeling applications. In practice, compressed or condensed mechanisms were adopted that evolved from highly parameterized schemes for O₃ chemistry in the 1970s to more specific descriptions after the 1980s. Examples of compressed mechanisms include the SAPRC 07 (CS07A) model, or the RADM chemical models [162,163]. Both detailed and truncated mechanisms have been evaluated with smog chamber data. The detailed mechanism experiments added to the advanced understanding of the complex chemistry. Experiments with truncated mechanisms provided knowledge of the potential inaccuracies involved in constraining the oxidation chemistry.

In the US, three “streamlined” chemical models have been developed since the late 1970s. These are: (a) Carbon Bond (CB; [164,165,166]); (b) Regional Acid Deposition Model—RADM [163] (also the RADM derivative, Regional Atmospheric Chemistry Mechanism—ACM [167]); and (c) SAPRC series [158,159]. The Carbon Bond mechanism is perhaps most commonly used in the US regulatory environment, and has evolved through several modifications since their first versions; e.g., CB05 [166].

According to Dodge [36], the sequence of reactions involving NO_y, O₃, OH, HO₂, H₂O₂ and CO is essentially the same except for the reaction of OH and NO₂. The models adopt parameterized chemistry for VOC, particularly biogenic compounds and aromatics whose kinetic degradation mechanisms have been adjusted from smog chamber experiments to estimate O₃ concentrations. Dodge’s Table 1 [36] lists over a thousand smog chamber runs exploring VOC-NO_x systems that have been archived and have been used for guiding the fidelity of compressed model elements. Several hundred runs have been recorded and used to evaluate oxidant forming mechanisms since the 1970s.

An important conceptual outgrowth of application of chamber experiments that resulted in the NO_x-VOC-O₃ relationship traces back to Haagen-Smit’s research, and the reports of Romanovsky et al. [15] and Hamming et al. [14]. This relationship was represented from chamber experiments as the widely cited Haagen-Smit “mountain” diagram, shown, for example, in Figure 3. The diagram shows the two idealized regimes of initial NO_x sensitivity and VOC sensitivity vs. the maximum O₃ expected nominally during late morning to mid-day hours. The NO_x sensitive regime is in the high initial VOC/NO_x regime to the left of the line of slope of VOC/NO_x ~8:1. The VOC sensitive regime of relatively low initial VOC/NO_x is to the right of the ridge line. These regions are separated by isopleth contours that show maximum O₃ formation with VOC/NO_x ratio. The idealized relationship on Figure 3 was demonstrated early on from a selection of experiments with different VOC/NO_x ratios. The relationship was calculated using the empirical kinetic modeling approach (EKMA) chemical model [168]. The mapping in the figure established a qualitative basis for precursor management that would estimate the average maximum expected O₃ concentration. This approach evolved from simple, empirical (mainly VOC) roll-back schemes for O₃ precursor reduction [169].

The current regulatory practices have evolved to adapt models for diurnal change in precursors using chemical transport models accounting for daily oxidant chemistry with precursor emissions and meteorological conditions for air transport and mixing [170]. Compressed kinetics calculations found in several air quality models, including the urban airshed (chemical transport) model (UAM) [171], the regional, RADM, “engineering” version [172] and versions of community multiple (spatial and temporal) scale chemical transport model (CMAQ) [173]. This portfolio of models is said to approximate by computer the “real” lower troposphere (see also Section 4).

3.1.1. Mechanism evaluation

Evaluation of photochemical mechanisms ideally would involve comparisons between the mechanistic calculations and smog chamber measurements of as many reactant and product species as practical for as long as possible. In most cases the evaluation is constrained for a given irradiation to concentration changes in a few hours for O₃, NO, NO₂, VOC species, products like formaldehyde or peroxyacetylnitrate (PAN) [174,175]. Recent evaluations can also address a variety of intermediate products including free radicals and peroxides.

A typical time series for smog chamber chemistry of a single aromatic compound over a period of a few hours is shown in the panels of Figure 4 with four different mechanism calculations. The upper panels follow O₃, NO, NO₂ and PAN; the lower panel shows the decline of o-xylene concentrations in the chamber. Here NO decreases exponentially after 0800 LDT, followed with a growth in NO₂ concentration to a peak, and an O₃ peak later in the run. Ozone then declined after ~1500 LDT. PAN concentration peaks at approximately noon.

For comparison, the time variation from kinetics mechanisms is calculated and compared with the smog chamber runs. In this case, the CB mechanisms tend to under predict the O₃ concentration, including the maximum and SAPRC and MCM over predict the measurements for O₃; MCM is relatively poor for estimating the maximum O₃ concentration in this experiment. The model predictions for NO and NO₂ also show variable departure from the measurements. Comparison between the smog chamber observations and the different model chemistry has led to selective constraints on principal reaction pathways and rate constants. These results combined with similar testing with different chamber data facilitated improvements in mechanisms for simulating photochemistry in the atmosphere.

In another example, Carter and Lurmann [177] discuss a systematic evaluation of the SAPRC-90 detailed mechanism [158] with smog chamber data. One approach in their study is a test to see how well the mechanism predicts O₃ concentrations for a range of different VOCs. Carter’s scatter plot in Figure 5 compares chamber measured O₃ produced (O₃-NO) after a period of time (5 h) with model calculated values [178]. The graph indicates the model calculation gives (O₃-NO) production to within approximately 30% for a range of single VOC chamber experiments. This result is found even though the reactivity of the chosen VOCs varies widely.

3.1.2. Relative Reactivity

An important result from smog chamber experiments concerns the influence of differences in reactivity of different VOC species. The reactivity of VOCs relates to the rate and amount of O₃ or major oxidation products formed in the reaction sequence. VOC reactivity also affects reactions on overall free radical concentrations and the rate of NO_y removal. Reactivity has been defined in different ways [178]. One simple measure is the reaction of OH on VOCs measured in terms of the OH radical rate constant (kOH). Another is the incremental reactivity, defined in terms of the change in O₃ formation caused by the addition of a small amount of a VOC to the emissions in an O₃ formation event; maximum incremental reactivity has been used as a useful index for mechanistic analyses [178].

There is a multitude of anthropogenic VOCs present in the atmosphere (e.g., alkenes, alkanes, aromatics and oxygenates). To characterize the influence of VOCs, initial practical approach was to combine species together using non- methane hydrocarbon (NMHC) concentration common to ambient air. The next step in complexity was to separate the VOC by the major ambient contributors by concentration. Then increased attention could be given to the reactivity of the individual species as potential contributors to the chemical process. The smog chamber experiments with various VOC species provided information about the capability of a mechanism to estimate reactivity. Then the correspondence of VOC reactivity differences in the atmosphere could be estimated from application of the mechanism in a chemical transport model.

Smog chamber experiments can give insight about the production of O₃ from a variety of single VOCs as suggested in Figure 5. Alternatively, reactivity can be tested for a mixture by adding a VOC to the mixture and recording a change in O₃ production. If adding a VOC to a mixture has sufficiently large effect on reducing free radical levels, then the addition may cause less O₃ to be produced than in its absence. Conversely, if the VOC added increases radical levels then more O₃ would be produced than in the VOC’s absence. Reactions of VOCs also increase rates of NO_y removal, for example, by forming alkyl nitrates, inhibiting ozone production when NO_y levels are low [179]. With NO_y present, this VOC “sensitivity” is a practical consideration for managing atmospheric O₃ production, especially for conditions of high VOC/NO_x concentrations, or NO_x sensitive conditions. When NO_x concentrations are sufficiently low, O₃ concentrations are very sensitive to NO_x and less sensitive to VOC as reflected ideally in Figure 3.

Smog chamber experiments with NO_x-VOC combinations are well-suited to test mechanisms used to calculate reactivity scales. Based on reactivity comparisons, scaling with incremental reactivity has been used to improve the SAPRC model results with experimental data in different gas-phase kinetic mechanisms [158,159].

3.2. Photochemical Aerosols

In this section, the gas-phase smog chemistry is extended to include photochemical aerosol formation. Contemporary smog chamber experiments have followed the initial work between the 1950s and 1990. Chamber experiments have proven to be an important asset in developing knowledge about atmosphere relevant particle chemistry. They provide a means to: (a) establish how much particle mass is formed compared with reactants lost to gas-phase products and container walls (b) obtain yield data for parameterized SOA mechanisms for use in chemical transport or airshed models; (c) understand the link between gas-phase and particle chemistry, including physical and chemical properties of SOA; and (d) evaluate chemical mechanisms to predict SOA production rates for semi-volatile and low volatility species.

Recognition of aerosol formation from irradiated, atmospherically relevant gas mixtures in smog chamber experiments was noted after the 1950s as discussed earlier [8,18,21]. These experiments inferred that there existed a clear linkage of oxidant chemistry and aerosol formation in smog instead of alternative sources such as smoke plumes. The experiments focused on sulfuric acid formation even though organics were present in the reacting mixture. Workers also recognized that particle nitrate could be formed through a pathway of nitric acid from the oxidant cycle [180]. Like their inorganic counterparts, organic aerosols are produced in the atmosphere as SOA. In the troposphere, high carbon number VOC precursors generally exist in the ppt-ppb level even in urban conditions.

Studies of the yield of photochemical reactions to produce low volatility material suggested that insufficient high carbon number vapors were present to explain major amounts of apparent photochemical SOA in smog, except for cyclic olefins and aromatics found in gasoline and other fuels. Grosjean and Friedlander [25] reported smog chamber measurements for some cyclic olefins that supported development of an aerosol-forming mechanism. Reactions analogous to those involving aromatics or cyclic VOC oxidize biogenic vapors. Isoprene, a relatively low-molecular-weight species, was thought to form products (low in carbon number) non-condensable at atmospheric concentrations. However, higher molecular weight species such as the terpenes (e.g., pinenes) could produce photochemically derived condensable products in ambient conditions where the reactants are present in concentrations less than ppb.

Smog chamber experiments have enabled the study of mechanisms relevant to photochemical aerosol--well beyond the early experiments of Los Angeles smog. Atmospheric chemical reactions of VOCs contribute to their degradation to final endpoints of CO and CO₂. Intermediate stages of degradation by mass are mainly short-lived gas phase and condensed products. The photochemical formation of organic particles from carbon number species >C₇ was established after the 1970s both in smog chamber experiments and in the field [23,25,181,182,183,184,185,186,187]. The main thermodynamic criterion for aerosol formation was a low vapor pressure for products that would partition significantly into the condensed phase. The time series for particle production in chambers coupled with O₃ as an index for oxidation; the increase in O₃ concentration parallel with aerosol production are noted in comparison between the gas-phase product-time profiles in Figure 4. The relationship between O₃ (or odd oxygen) and SOA concentrations also has been documented in ambient air, e.g., in the southeastern US [188] and in Mexico City [189].

A contemporary picture of the linkages between O₃ photochemistry and photochemical aerosols is shown in Figure 6. For inorganic acid species, sulfuric acid and nitric acid, the photochemistry derives from oxidation via OH or H₂O₂ or O₃. Secondary organic particle (SOA) formation follows from radical reactions or ozonolysis and depends on producing low vapor pressure (low volatility) species from higher volatility VOC vapors, partitioning semi-volatile product between the vapor and condensed phase. Reactions follow from an initial oxidation step generating products containing one or more oxygenated functional groups. Early on, one of the atmospheric pathways identified was an O₃ attack on unsaturated VOC through formation of Criegee intermediates and ozonides [190] followed hypothetically by formation of oxygenated species of reduced vapor pressure [29]. Particle formation as a part of gas-phase oxidation reactions has been identified with OH, RO₂, RO and NO₃ radicals, and heterogeneous reactions on existing particle surfaces [29].

Smog chamber and other reactor studies have indicated the presence of highly oxygenated material in photochemical aerosols than expected from oxidant attacks on large VOC molecules. As indicated in

Table 3. Vapor pressure change of an organic compound with addition of a common functional group, based on group contributions calculated by Pankow and Asher [192]. (Reproduced from Kroll and Seinfeld [40] with permission from Elsevier).

Functional Group	Structure	Change in Vapor Pressure (298 °K) a
Ketone	-C(O)-	0.10
Aldehyde	-C(O)H	0.085
Hydroxyl	-OH	5.7 × 10−3
Hydroperoxyl	-OOH	2.5 × 10−3
Nitrate	ONO2	6.8 × 10−3
Carboxylic acid	-C(O)OH	3.1 × 10−4
Peroxyacid	-C(O)OOH	3.2 × 10−3
Acyl peroxynitrate	C(O)OONO2	2.7 × 10−3
Extra carbon	-CH2-	0.35 b

a: Multiplier with hydrocarbon root. b: For comparison between changes in polarity by addition of a functional group and changes in size of the carbon skeleton.

addition of oxygenated functional groups to VOCs dramatically reduce vapor pressure, changing the partition between vapor and condensed phase. Vapor pressure reduction in smog chamber experiments has a direct effect on particle yields [192].

The yields of aerosol mass produced from photochemical reactions of VOCs (change in mass of aerosols/change in mass of VOC) relate to the amount of aerosol mass produced (Mo µg/m³). Yield curves for smog chamber reactions of C6-C9 species are illustrated in Figure 7. These curves are non-linear nature but indicate that the larger the yield or the larger the change in VOC, the larger the expected amount of aerosol relative to total VOCs present [193]. For single VOC species, the yield Mo relationship is qualitatively similar for many species, but the yield curves vary beyond simply a carbon number depending on the nature of the VOC vapor. At the turn of the 20th–21st century, two “discoveries” opened the door to an aggressive research program in photochemical aerosol formation. The first was the finding from smog chamber experiments that isoprene, a C5 unsaturated hydrocarbon, could form aerosols partly as a consequence of heterogeneous reactions with an acid catalyst [194,195] or other seed material [196,197,198]. “Seeding” may also be related to acid gases present in irradiated gas producing SOA [199]. The second finding derived in part from smog chamber and other experiments, indicated that aerosol “aging” in atmospheric conditions such that reactions modulated with a sequence of oxidation or cleavage reactions, and condensation to evaporation of semi-volatile species in successive chemical generations [40,89,200]

In the case of SOA formation from isoprene, Paulot et al. [201] and Surratt et al. [202] identified key intermediates in the oxidation process, including IEPOX (isoprene epoxydiols) and MPAN (methyl-acryloylperoxynitrate) that have also been found in the atmosphere. These intermediates play an important role in successive generation of oxygenated semi-volatile species from isoprene. They also open a pathway for organosulfate aerosol production from isoprene. Where isoprene is rich in VOC concentrations, organosulfate appears in the atmosphere as a fraction of the organic particles present [203].

Illustrated in Figure 8 are smog chamber results for aerosol production from isoprene oxidation in two NO_x regimes. In both cases, particles measured by volume are formed after NO and isoprene are lost to the mixture. In the low NO_x case, particle concentration reaches a maximum after 4 h with estimated >50% loss of isoprene, and weak O₃ production of a few ppb. In the high NO_x case, a substantial amount of O₃ is produced, peaking at 3 h. Particle volume grows with O₃, but levels at ~3 h. Kroll et al. [194] interpret the photochemical pathways in the NO_x regimes in terms of the OH attack on the unsaturated carbon bond of isoprene producing degradation products from peroxy radical chemistry, including reduced volatility oxygenated species.

The SOA chemistry of isoprene and other VOCs involves oxidized products some of which lead to the formation of polymeric material, oligomers or humic-like substances [40,204], whose production pathways vary with NO_x concentration. The aging of semi-volatile SOA is seen to involve a series of reaction “generations” with increasing oxygenation. “Aging” of semi-volatile VOCs by multiple generation chemical processes is potentially important on the order of a day timescale [83,205,206].

Most kinetics experiments and derived mechanisms rely on the first generation of reactions and not subsequent processes. The evolution of particle composition during aging is hypothesized to involve organic species over a wide range molecular weight and oxygen content. The potential for interactions between high and low carbon number species created difficulty in seeking detailed mechanisms for aerosols to complement the gas-phase mechanisms. Two rules of thumb seem to hold for SOA from photochemical reactions--volatility and partitioning vapor pressure reduction, and increasing O/C ratio in particles with aging. The Donahue et al. [207] and Chacon-Madrid et al. [208] modeling approach embodies partitioning, and defines a volatility basis set (VBS) that involves saturation vapor concentration and the O/C ratio compared with the H/C ratio.

A third finding in recent years resurrects an important discussion of aqueous reactions to form sulfate [209]. As noted earlier, the oxidation of SO₂ to form sulfuric acid is well known to involve aqueous reactions of OH, H₂O₂ and O₃; Nitrate formation is influenced by equilibrium considerations involving HNO₃, NH₃ (or other cations) and H₂O [210]. As noted above, heterogeneous processes on particles exist beyond catalyzed reactions [37]. The formation of SOA from aqueous reactions in the 1980s [153] has surfaced in recent studies. Evidence has emerged for photochemical (and non-photochemical) SOA production through aqueous, catalyzed reactions of water soluble organics, including glyoxal and hemiacetal, provided the medium for forming low-volatility compounds that are potentially embodied in wet, existing particles at high relative humidity, or clouds and fog [42,211,212,213].

Laboratory study of aqueous reactions producing SOAs in the presence of hydrometeors is problematic for conventional smog chambers. However, smog chamber studies at varying humidities have added insight into SOA characterization, since generated particles are hygroscopic and contain major amounts of water at relative humidity above ~70% [214,215]. Reactions on wet aerosols or hydrometeors are mediated by organic surface coatings that complicate the liquid-gas interface [216,217].

SOA Thermo-kinetic Models

Most photochemical aerosol models to date have relied on the application of smog chamber measurements for yield and a measure of VOC reduction, combined with thermodynamic estimates of phase partitioning. Formulation of chemical kinetic mechanisms for formation condensable species has lagged in development because of uncertainties in multi-generations of reaction pathways and products.

The assessment and interpretation of photochemical SOA production [218] guides researchers constructing kinetic models integrating gas-phase reactions and condensed phase processes. The aerosol dynamics involves not only chemical reactions, but also the physicochemical shaping the particle size-composition distribution. Kinetic modeling of aerosol dynamics traces back to the 1970s. For example, Middleton and Brock [219] reported a simulation that accounted for particle growth, coagulation, condensation-evaporation and nucleation, but without chemical production, per se. Examples of kinetic models for photochemical aerosol formation potentially applicable to atmospheric processes and supported with smog chamber data include those of [84,220,221,222,223]. Jathar et al. [224] have reported a three-dimensional kinetics model for use in air quality modeling. Current kinetic models used in CMAQ include CB05, RACM 2, SAPRC 07 and AERO-AE06 [225]. Vernecek et al. [226] reported the analysis of a preliminary version of the SAPRC-16 mechanism in CMAQ airshed modeling.

Continued progress to develop an aerosol kinetic model coupled with gas-phase chemistry called GECKO-A is an international collaborative enterprise [227,228]. This effort intends to synthesize available laboratory and field measurements with current kinetics mechanisms and models to produce a community-based model analogous to MCM. As part of this research, increased attention is likely to feed back to laboratory investigators that chamber studies are restricted to concentrations too high for relevance to contemporary ambient conditions. Autoxidation reactions [45,229], for example, may come into play in semi-volatile VOC oxidation to form highly oxidized molecules.

Evaluation of SOA models using smog chamber data follows from a history of evaluations of gas-phase mechanisms. Unlike the gas-phase mechanisms, however, SOA experiments can only rely on certain coarse comparisons. First, their reproducibility of results from chamber to chamber, and from semi-empirical model comparisons. Chamber experiments need to include not only the gas-phase measurements of temporal reactant and product change, but also aerosol properties such as nuclei concentration, mass or volume concentration and size distributions or AMS observations. Intercomparisons and evaluation studies have been reported that improve existing kinetics codes, for example [81,230,231].

4. The Atmosphere—Judging the Simulation Quality

As noted above, the gas-phase and aerosol mechanisms have been added to computer models to simulate the real atmosphere. The chemical components are tested for their relevance to atmospheric conditions in different ways. The relevance of gas-phase chemistry is well established through extensive and sustained measurements of ozone, its precursors and intermediate products, including the reactive nitrogen cycle and certain oxygenated organic compounds [28,43]. The presence of sulfate and nitrate and ammonium in atmospheric aerosols is also thoroughly documented in the literature [29,232,233].

The relevance of photochemical SOA has relied on four different routes: (a) establishing the amount and temporal variation of SOA present relative to primary carbon sources, (b) measurement of tracer compounds that are expected from photochemical processes, (c) tracing daily time series of gases and particles across the earth’s surface, and (d) comparison of chemical transport (airshed) models with ground and aloft observations for gas-phase species, organic carbon and SOA.

The first goal establishes the apparent amount of SOA present relative to contributions from primary sources such as fuel combustion. The SOA variation by season would hypothetically associate photochemical aerosol presence with a seasonal maximum with O₃. At least two methods exist for this determination. The first uses the difference between black and organic carbon ratios (OC/BC) from primary emissions (or OC/CO) vs. ambient values from which the secondary OC can be determined [188,234]. The second method derives from measurements using an AMS [235]. These observations have enabled the characterization of OC in terms of the ratios hydrocarbon rich (HOA) to reflect hydrocarbon- like species and oxygen rich organic aerosol (OOA) to identify with oxidized secondary species [236]. The OC/BC ratio methods and the AMS establish that SOA is a major component of OC in atmospheric aerosols, both in urban and rural-remote conditions.

The chemical tracer method provides direct evidence for SOA chemistry from identification of products from smog chamber studies, supporting their relevance to atmospheric processes. Tracer species for photochemical reactions, and their temporal variation represent key specific evidence of mechamisms. Organic tracers have identified characteristic molecular species with low volatility present in aerosols in special field campaigns. Species include certain oxygenated gases and condensed, or semi-volatile species such as the pinonic acid or pinonaldehyde from pinene oxidation, glyoxals, or isoprene epoxydiols or methyl tetrols from isoprene oxidation [108,237,238,239]. Organosulfates like C₅H₁₂O₆S as intermediates also have been identified as tracers for biogenic SOA from smog chamber experiments [240,241]. While this approach has proven to be useful for some SOA chamber-based reactions, the method is constrained by sampling and analytical limitations for identifying and quantifying complex organic products expected in the first generation of oxidation and beyond [242].

As an alternative, the SOA contribution can be estimated from the difference between particles from primary OC sources and ambient species concentrations. Zheng et al. [243] report an example of this approach. Their analysis of urban conditions in the southeastern US identifies wood burning, transportation and meat cooking (and other, including SOA) to account for ~ 89% of the fine particle mass concentration, but the amount attributed to SOA and unknown sources was not resolved.

The third approach involves measurements of photochemical indicators in time and space in comparison with temporal measurements in smog chambers. These comparisons have been done for several modeling schemes. Conceptually the Lagrangian approach offers an analogy with smog chamber experiments. The USEPA developed a design for Lagrangian studies in the 1970s [244]. An experimental design was conceived in the 1973 Los Angeles Reactive Pollutant Program (LARPP), which used instrumented helicopters following the motion of tetroons for air mass tracking. Feigley et al. [245,246] conducted smog chamber experiments that were aimed at creating a comparative basis for measurements of O₃, NO_x and VOCs in the study conducted over Los Angeles in late summer and fall of 1973.

A similar Lagrangian-like experiment was conducted about the same time in Los Angeles [247]. This experiment involved the use of a blimp platform for O₃, NO and NO₂ instrumentation, and two indicators of particle production. The blimp was chosen for a capability to approximately mimic travel with an air mass across the Los Angeles Basin, which corresponds to a smog chamber run. The results of the flight are shown in Figure 9. The flight originated in relatively clean air near the Pacific Ocean; the blimp traveled in smog with the winds at a height of 150–300 m mslsl inland across the city. The portion of the flight that follows a track with air mass movement is between time 0 and 160 min.

The experiment showed the rapid rise of O₃ and NO_x early in the flight, followed by peaking of O₃ mixing ratio after about 100 min. Both condensation nuclei concentrations--CN (~<0.05 µm diameter) and the particle light scattering coefficient (b_sc) (mainly related to 0.1–5 µm diameter) depend on primary particle sources and secondary chemical production. The results of this experiment and parallel data for vertical concentration profiles demonstrate the complexity of interacting atmospheric processes over Los Angeles, far beyond the simulations in smog chambers. Notable among these are meteorological features involving the winds, thermal stratification, and transport differences with height [247,248].

The results of the blimp experiment indicated qualitatively the O₃ production and NO_x peaking took place as expected. In addition, evidence early in the experiment suggested link between light scattering and O₃. The results support the presence of photochemical processing for both gases and a portion of particles. These in turn support the relevance of smog chamber experiments as simulators for atmospheric processes.

The fourth approach involves the application of chemical transport models as simulators embodying approximations of known atmospheric processes and the air–surface exchange (including emissions) occur continuously at various temporal and spatial scales [249]. The models most useful for mechanism relevance are those that have a time and space resolution of hours and a km or less. These contain gas-phase photochemical mechanisms and aerosol components that evolved mainly after the 1980s. The mechanisms are evaluated for fidelity of simulating the atmosphere with a combination of ambient measurements at the surface and aloft. The comparisons are supported with determination of the veracity of emissions data and meteorological observations. The chemical transport modeling can be extended to global averages over time and space constraints relevant to long-term phenomena; these are not discussed here, but are accessed in other venues [250,251,252].

Measurements for model testing exist from long-term averages from air monitoring of O₃, NO_y, VOCs, and particle mass concentration and composition. In addition, there are data from several experiments including major efforts characterizing O₃ photochemistry [253] and sulfate and nitrate [254,255]. Particulate carbon in the presence of reactive gases also has been studied extensively in various projects [256,257,258].

Model performance compared with the observations as a reference has been reported as part of evolving model development and applications. Ideally comparisons are most useful for models with temporal resolution of an hour or less and spatial resolution of approximately a kilometer. Examples of deterministic model testing with photochemical mechanisms include those reported for CMAQ [259,260,261,262,263]. As an example for a 2002 Canadian exercise using a model similar to RADM [170], annual average O₃ concentrations model estimates are 15–35% of observations. Typically, for average atmospheric sulfate, models are within 27–50%; performance is worse for nitrate 44%, and 85% for organic carbon. This difference between gas comparisons reflects in part the extensive experience with gas-phase models compared with other photochemical species. Other analyses have suggested that the volatility-based models underestimate measured SOA concentrations [40,260]. Another example, a study of Volkamer et al. [264] in Mexico City, showed that reactive VOCs from anthropogenic sources produced much greater amounts of SOA than calculated from a volatility-based model.

Results of the kind noted indicate that the gas-phase and inorganic aerosol mechanisms adopted for chemical transport models support the photochemistry derived in part from smog chambers. For SOA, the aerosol chemistry in models needs additional work to give results expected from ambient measurements. The model and field campaigns provide a stimulus for continuing research on photochemical aerosols through linkages with smog chamber and other operations in collaborative programs like FIXCIT [265] and GECKO-A [227].

5. Summary

Construction and use of laboratory models of processes believed to represent the Earth’s atmosphere has been a major tool for atmospheric research. Simulation of atmospheric photochemical kinetics began in the 1950s. These experiments have relied heavily on environmental (smog) chambers, first for gas-phase studies and later for characterization of aerosol particle production. Smog chamber studies concentrate on oxidant production from reactions of NO_x and VOCs and sometimes sulfur oxides. Much of the effort has been devoted to the VOC participation in the chemistry as free radicals and other species. These have led to an understanding of the role of VOC reactivity in O₃ formation. The results of literally thousands of experimental runs have provided not only insight into reactions relevant to the atmosphere, but quantitative evaluation of mechanisms created from basic kinetics data.

Smog chamber experiments are essential in the knowledge about secondary aerosol production related to the atmosphere. The results of many experiments have characterized particle formation in terms of volatility of oxidation products. Low-volatility species from high-carbon-number VOC at the ppb level readily form aerosol particles. Formation of particles from higher-volatility VOC species depends on oxidation pathways hypothesized to involve multi-generation of steps. Important ratios for describing photochemical aerosols are O/C and H/C.

Mechanisms derived from smog chamber experiments are constrained by the ability to interpret results for relevance to the atmosphere. Chamber constraints include: (a) chamber wall effects for both gas-phase and aerosol processes, (b) reaction time limits to a day or less, (c) intermittent variations in solar radiation that modulate the photochemical rates, and (d) inability to deal with the atmospheric background of aged gases and particles, which affect the mechanistic interpretation in poorly known ways.

Photochemical mechanisms developed from theory and experiments are adapted in chemical transport models for numerically simulating the atmosphere. The results to date from various intercomparisons between atmospheric measurements and models indicate that gas-phase and inorganic particle photochemistry simulates well parts of atmospheric chemistry. However, mechanism development and simulation of SOA formation in the atmosphere remains incomplete.

The reviews listed in Table 1 give several recommendations for continued research on photochemical processes, including the use of smog chamber experiments and collaborative studies in the atmosphere. Among the recommendations are:

--

Focus on multiday gas and particle chemistry, including temperature and humidity interactions.

--

Conduct chamber and other experiments at concentrations approaching current levels atmospheric levels in rural and remote areas.

--

Strengthen aerosol models with kinetic mechanisms for condensed products.

--

Identify in more detail intermediate products including oxygenated species of gas-phase and particle reactions and identify added atmospheric tracer species for confirming laboratory mechanisms.

--

Examine and evaluate the potential importance of background conditions in the atmosphere as they may interact with the controlled experiments.

--

Give more attention to photochemistry of the free troposphere as a function of altitude, including hydrometeor interactions.