Design and characterization of a smog chamber for studying gas-phase chemical mechanisms and aerosol formation

We describe here characterization of a new stateof-the-art smog chamber facility for studying atmospheric gas-phase and aerosol chemistry. The chamber consists of a 30 m3 fluorinated ethylene propylene (FEP) Teflon film reactor housed in a temperature-controlled enclosure equipped with black lamps as the light source. Temperature can be set in the range from −10 to 40 C at accuracy of ±1 C as measured by eight temperature sensors inside the enclosure and one just inside the reactor. Matrix air can be purified with non-methane hydrocarbons (NMHCs) < 0.5 ppb, NOx/O3/carbonyls < 1 ppb and particles < 1 cm−3. The photolysis rate of NO2 is adjustable between 0 and 0.49 min−1. At 298 K under dry conditions, the average wall loss rates of NO, NO2 and O3 were measured to be 1.41 × 10−4 min−1, 1.39×10−4 min−1 and 1.31×10−4 min−1, respectively, and the particle number wall loss rate was measured to be 0.17 h−1. Auxiliary mechanisms of this chamber are determined and included in the Master Chemical Mechanism to evaluate and model propene–NOx–air irradiation experiments. The results indicate that this new smog chamber can provide high-quality data for mechanism evaluation. Results of α-pinene dark ozonolysis experiments revealed secondary organic aerosol (SOA) yields comparable to those from other chamber studies, and the two-product model gives a good fit for the yield data obtained in this work. Characterization experiments demonstrate that our Guangzhou Institute of Geochemistry, Chinese Academy Sciences (GIG-CAS), smog chamber facility can be used to provide valuable data for gas-phase chemistry and secondary aerosol formation.


Introduction
Smog chambers provide a controlled environment to study the formation and the evolution of specific compounds of interest by isolating the influence of emissions, meteorology and mixing effects.Smog chambers were initially constructed for developing and evaluating atmospheric gasphase chemical mechanisms or models for predicting secondary pollutants (Akimoto et al., 1979;Carter et al., 1982;Jeffries et al., 1982Jeffries et al., , 1985)).In the mid-1980s, Seinfeld and colleagues at the California Institute of Technology in Pasadena developed a 65 m 3 outdoor chamber made of fluorinated ethylene propylene (FEP) Teflon film to study the aerosol formation from gas-phase precursors such as aromatic and biogenic hydrocarbons (Leone et al., 1985;Stern et al., 1987).In the subsequent three decades, outdoor and indoor chambers have been widely used to study formation of secondary pollutants such as ozone (Hess et al., 1992;Simonaitis et al., 1997;Carter, 2000;Dodge, 2000) and secondary organic aerosols (SOA) (Odum et al., 1996(Odum et al., , 1997;;Griffin et al., 1999;Martín-Reviejo and Wirtz, 2005; Published by Copernicus Publications on behalf of the European Geosciences Union.Paulsen et al., 2005;Rollins et al., 2009) and evolution of SOA (Donahue et al., 2012).Although the aims of these smog chambers are similar, their designs and capacities vary widely, displaying larger differences in factors such as sizes, reactor wall materials and light sources.
Outdoor chambers are unique in the availability of natural sunlight.Many simulation experiments on SOA formation have been carried out in large outdoor chambers (Leone et al., 1985;Stern et al., 1987;Pandis et al., 1991;Johnson et al., 2004;Martín-Reviejo and Wirtz, 2005;Rollins et al., 2009).However, diurnal variations of the actinic flux and temperature make it difficult to model experimental data and to reproduce the experiments.Indoor chambers instead can provide precise control of temperature and humidity, but may suffer from the difference between the artificial light spectrum and solar spectrum, resulting in different rates of some photolysis reactions between the artificial and natural systems (Takekawa et al., 2003;Carter et al., 2005;Paulsen et al., 2005).Nonetheless, artificial light allows the experiments to be repeatable under the same irradiation conditions.Smog chambers are, however, not without any limitations or uncertainties in chamber wall effects when evaluating the mechanisms or models (Carter et al., 1982;Carter and Lurmann, 1991;Dodge, 2000).For example, off-gassing of NO x and other species from chamber walls may introduce contaminations into the background gas and affect the utility of the data.Similarly, chamber wall effects may lead to large uncertainties when evaluating experiments at low concentrations.Large volume reactors with small surface-to-volume ratios can minimize the wall effects and the wall loss of particles and gas-phase species.Moreover, larger volume allows experiments of longer durations to be conducted and more instruments to be used.Considering these concerns, Tobias and Ziemann (1999) used a 7000 L Teflon environmental chamber to generate organic aerosols through gas-phase reactions of 1-tetradecene and ozone.Cocker et al. (2001a) developed dual 28 m 3 indoor chambers to study the mechanisms of aerosol formation.Recently, to minimize reactor effects in the studies of volatile organic compounds (VOCs) reactivity and SOA formation, Carter et al. (2005) constructed a stateof-the-art indoor chamber facility with two collapsible 90 m 3 FEP Teflon film reactors.
In China, some small-volume smog chambers have been developed to study gas-phase kinetic mechanisms since 1980s (Wang et al., 1995;Ren et al., 2005;Xu et al., 2006).Wu et al. (2007) constructed a 2 m 3 , precisely temperaturecontrolled indoor smog chamber made of FEP Teflon film to study the SOA formation.However, the small volumes of these smog chambers impart the disadvantage of relatively large wall effects and also make it difficult to do experiments of long durations.At present, ozone and fine particles (PM 2.5 ) have become serious air quality problems in China (Chan and Yao, 2008;Q. Zhang et al., 2012).As ozone and a large portion of PM 2.5 components are secondary from gaseous precursors under atmospheric gas-phase and/or multi-phase processes, it is imperatively necessary to set up further quality smog chambers in China for the deep understanding of complex air pollution, particularly in China's megacities.
This paper describes a new state-of-the-art indoor smog chamber facility established at the Guangzhou Institute of Geochemistry, Chinese Academy Sciences (GIG-CAS).This GIG-CAS chamber facility is designed to study formation mechanisms of ozone and SOA as well as the evolution of SOA; to evaluate the mechanisms, particularly under low-NO x and VOC conditions; and to serve as a platform for evaluating the performance of newly developed gas or particle monitors.A series of initial characterization experiments have been carried out and are discussed in this paper.

Facility
The 8.5 m × 4.0 m × 3.5 m thermally insulated enclosure is housed in a 14.0 m × 6.0 m × 8.5 m laboratory in the first floor and has been equipped with an array of continuous gas-phase and aerosol-phase monitors.Situated directly over the enclosure on the second floor are offline VOCs/semi-VOCs/ions/anions samplers and analyzers.The list of equipment is shown in Table 1.Inside the enclosure are a 30 m 3 reactor made of FEP Teflon film (FEP 100, Type 200A; DuPont, USA) with dimensions of 5.0 m × 3.0 m × 2.0 m and two banks of 60 W 1.22 m black lamps.A schematic of the GIG-CAS smog chamber is shown in Fig. 1a.

Enclosure
The inner walls of the enclosure are covered with reflective and polished stainless steel sheets to obtain a maximum and homogenous light intensity.The floor is covered with less reflective but more durable stainless steel sheets.For indoor Teflon chambers, air conditioning systems are widely used to control the temperature within the tropospheric temperature range, which is roughly from −60 to 40 • C (Takekawa et al., 2003;Carter et al., 2005;Paulsen et al., 2005;Wu et al., 2007).The enclosure temperature of our chamber is controlled by three cooling units (total power 40 kW), which are located outside the laboratory next to the enclosure.The cooling air is distributed uniformly and constantly in three different ducts through the enclosure false ceiling made of perforated reflective aluminum sheets, and is returned to the cooling units through a porous, highly efficient 3.0 m × 0.2 m × 0.6 m filter that contains activated charcoal inside to clean the enclosure air.A heater is also installed inside each duct.Eight thermocouples (Fig. 1a) are placed approximately 1 m above the housing floor between the enclosure and the reactor walls.The enclosure temperature is controlled by an electronic control system through adjusting the power of cooling units and heaters according to the difference between the average temperature of the eight thermocouples and the set temperature.

Teflon reactor
Chamber reactor walls can be made of Teflon film (Cocker et al., 2001a;Carter et al., 2005), Pyrex (Doussin et al., 1997), quartz (Barnes et al., 1987), aluminum (Akimoto et al., 1979) or stainless steel (Wang et al., 2011).Except for the AIDA chamber (Saathoff et al., 2003), most larger smog chambers on the order of tens of cubic meters were made of Teflon film for technical and economical reasons (Cocker et al., 2001a;Carter et al., 2005;Pandis et al., 1991;Johnson et al., 2004;Martín-Reviejo and Wirtz, 2005;Rollins et al., 2009).The reactor wall material of this chamber is 2 mil (54 µm) FEP Teflon film that is transparent, chemically inert and UV permeable.The Teflon film is flexible enough to avoid altering the pressure inside the reactor during air extraction.The Teflon reactor is self-made by a heat sealing laminator.All seams on the reactor are reinforced by a polyester film tape with a silicone adhesive (polyester tape 8403; 3M, USA).The reactor wall is more than 1 m away from the lights to avoid the heating on the surface of the Teflon film.The reactor is mounted inside the enclosure with a fixed bottom stainless steel frame and a movable top stainless steel frame.
The top frame is lifted and lowered by a mechanical step motor that has power-off protection.During experiments, as the volume decreases due to sampling, leaks and permeation, the top frame is lowered slowly to maintain a differential positive pressure between the inside of the reactor and the enclosure, thereby reducing the contamination of the enclosure air.Siemens QBM 66.201 (Siemens AG, Germany) is used to measure the differential pressure between the reactor and the enclosure with an accuracy of ±3 Pa.When the reactor volume decreases to 1/3 of its maximum value, the experiment (typically about 10 h, depending on the numbers of instruments taking samples from the chamber) is terminated.
When lowering the top frame of the reactor during wall loss evaluation experiments, the total particle number concentration (N t ) changed smoothly from an initial number concentration N 0 and the slope of ln (N 0 /N t ) versus t remained constant, suggesting that the movement of the chamber will not affect the aerosol wall loss rates.Four Teflon ports are installed inside the reactor.One port located in the middle of the chamber floor has 12 holes of 0.635 cm in diameter.This port is used for injection of purified air, sample and seed aerosols.A second port next to the instruments, with 13 holes of 0.635 cm in diameter, is used for sampling.Since the residence time of air inside the sampling lines is within seconds, the influence of sampling process on the homogeneity of air reaching the online instruments is negligible.The other two ports, each with four holes of 0.635 cm in diameter, are also used for sampling.Furthermore, an air blower is connected to the reactor through a 10 cm solenoid valve controlled by computer, providing a pumping flow rate of 1 m 3 min −1 to evacuate the reactor.
Two three-wing stainless steel fans coated with Teflon are installed at the bottom inside the reactor to provide sufficient mixing of the gas species and particles.No detectable nonmethane hydrocarbons (NMHCs) are emitted from the fans.Rotating speed of the fans can be varied to a maximum of 1400 rpm by adjusting the input power.source.One bank of 80 lamps is mounted on one enclosure wall, and another bank of 55 lamps is mounted near the opposite enclosure wall on an aluminum alloy frame.The artificial light produced by the black lamps gives a good representation of the ground-level solar light spectrum in the low-wavelength region, but they do not emit in the longer wavelength regions (> 400 nm) that photolyze some species, such as peroxy radicals or methylglyoxal in the range of 450-550 nm (Cocker et al., 2001a).However, because of the low-cost and efficient UV irradiation, black lamps are frequently used in environment chamber experiments (Cocker et al., 2001a;Carter et al., 2005;Hynes et al., 2005;Wu et al., 2007).The 135 black lamps are divided into 4 separately controlled groups; therefore the light intensity can be regulated to different levels.During the experiments, a J NO 2 filter radiometer (Metcon GmbH, Germany) is used to monitor the light intensity inside the enclosure.

Air purification system
As matrix gas for simulation experiments and carrier gas for reactants, purified dry air is supplied by passing compressed house air through a condenser and a Thermo zero-air supply (Model 1160; Thermo Scientific, USA).The maximum flow rate of purified air is 45 L min −1 .The purified dry air includes < 0.5 ppb NMHCs, < 1 ppb NO x , O 3 and carbonyl compounds, and no detectable particles.A separate source of purified air with < 5 ppb NMHCs (mainly ethane and propane) is also used for high-concentration precursor (hundreds of parts per billion) experiments.Compressed indoor air is forced through a combustion chamber filled with Hopcalite and a series of bed scrubbers containing activated charcoal, Purafil, Hopcalite and allochroic silica gel, followed by a PTFE filter to provide this source of purified air with a maximum flow rate of 200 L min −1 .The schematic of these two air purification systems is shown in Fig. 1b.Before each experiment the reactor is evacuated and filled with purified dry air at least five times, then the reactor is flushed with purified dry air at a flow rate of 100 L min −1 for at least 48 h until no residual hydrocarbons, O 3 , NO x or particles are detected in the reactor to avoid carry-over problems from day-to-day experiments.When the reactor is not in use, it is continuously flushed with purified dry air.

Injection system
Gaseous reactants are injected using gas-tight syringes through a septum installed in one port of a union tee that is connected to a FEP Teflon line, and then flushed by purified dry air or nitrogen into the reactor.For the introduction of liquid reactants, known volumes are measured and injected by means of microliter syringes through a heating system with an injection port similar to that used in gas chromatography.The temperature of the injection port is adjustable according to boiling points of the injected reactants.Ozone is generated by a commercial ozone generator (VMUS-4; Azco Industries Ltd, Canada) with pure oxygen as the feed gas to prevent the generation of NO x .Ozone introduced into the reactor is controlled by the generating time and the flow rate.
Humidification is achieved by vaporizing Milli-Q ultrapure water contained in a 0.5 L Florence flask and the water vapor is flushed with purified dry air into the reactor.Relative humidity inside the reactor can be varied from < 5 to 80 %.The humidification process does not introduce detectable hydrocarbons or particles into the reactor.
Seed particles are generated by an atomizer (ATM-220; Topas GmbH, Germany) and pass through a diffusion dryer (DDU-570; Topas GmbH, Germany) to remove water and a neutralizer (TSI 3080; TSI Incorporated, USA) to eliminate the charge before introducing into the reactor.

Instrumentation
The array of gas-phase and aerosol-phase instruments equipped with the chamber facility is briefly described in Table 1.
Ozone is measured by an EC9810 ozone analyzer (Ecotech, Australia), which is calibrated weekly using a Thermo Scientific Model 146i multi-gas calibrator.The detection limit and accuracy of the ozone instrument are 0.5 ppb and ±0.5 %, respectively.An EC9841T chemiluminescence analyzer (Ecotech, Australia) is used to measure NO and NO 2 .The NO x instrument is calibrated weekly using a certified cylinder of NO.The detection limit and accuracy of the NO x instrument are 50 ppt and ±0.5 %, respectively.
VOCs inside the reactor are measured both offline and online.Offline measurement are performed by using a model 7100 preconcentrator (Entech Instruments Inc., USA) coupled with an Agilent 5973N gas chromatography-mass selective detector/flame ionization detector/electron capture detector (GC-MSD/FID/ECD; Agilent Technologies, USA).During simulation, air inside the reactor is sampled about every 15 min with evacuated 2 L stainless steel canisters and simultaneously analyzed with the three detectors of this selfmodified system for a wide spectrum of VOCs.This analytical system is well established, and a detailed description of the method can be found elsewhere (Wang and Wu, 2008;Y. L. Zhang et al., 2010Y. L. Zhang et al., , 2012Y. L. Zhang et al., , 2013)).Carbonyl compounds are analyzed offline by high-performance liquid chromatography (HPLC -HP1200; Agilent Technologies, USA) coupled to UV detection at 360 nm after being collected by drawing air through a Sep-Pak DNPH-Silica cartridge (Waters Corporation, USA) with a sampling pump (Thomas, USA) every 15 min.A detailed description of the method can be found elsewhere (Tang et al., 2003).
Online monitoring of parent NMHCs such as propene, αpinene, and some aromatic hydrocarbons as well as their oxidation products are also available with a proton-transferreaction time-of-flight mass spectrometer (PTR-TOF-MS; Ionicon Analytik GmbH, Austria).Detailed descriptions of the PTR-TOF-MS technique can be found elsewhere (Lindinger et al., 1998;Jordan et al., 2009).At present the accuracies for some species by the PTR-TOF-MS are still not comparable to that by offline methods: for example, PTR-MS quantification of HCHO is highly influenced by the humidity (Vlasenko et al., 2010).Therefore in this work we only report our VOCs results by offline techniques, and further calibration of the PTR-TOF-MS is still under way.
Particle number concentrations and size distributions are obtained using a scanning mobility particle sizer equipped with a differential mobility analyzer (TSI 3081; TSI Incorporated, USA) and a condensation particle counter (TSI 3775; TSI Incorporated, USA).Flow rates of sheath and aerosol flow are 3.0 and 0.3 L min −1 , respectively, allowing for a size distribution scan ranging from 14 to 700 nm within 135 s.The accuracy of the particle number concentration is ±10 %.
A high-resolution time-of-flight aerosol mass spectrometer (HR-TOF-MS; Aerodyne Research Incorporated, USA) is used to measure the chemical compositions and evolutions of submicron aerosols (PM 1 ) (Jayne et al., 2000;DeCarlo et al., 2006).The HR-TOF-MS is also able to determine the average element ratios of organics, like H / C, O / C, and N / C (Aiken et al., 2007(Aiken et al., , 2008)).The instrument is operated in the high-sensitivity V mode and high-resolution W mode alternatively.The HR-TOF-MS is calibrated using 300 nm monodisperse ammonium nitrate particles.

Temperature control and its homogeneity
In the GIG-CAS chamber, three temperature control units and two Teflon-coated fans are used to provide a homogeneous and stable temperature inside the reactor.Figure 2a shows evolutions of the average temperatures as measured by the eight thermocouples (T1-T8, Fig. 1a) and the temperature inside the reactor (T r , Fig. 1b) measured by Siemens QFM2160 (Siemens AG, Germany) under black light irradiation when temperatures were set to 40, 15 and −10 • C, respectively.After the initial increase or decrease, the average temperatures stabilized after about 1.5 h, 20 min and 40 min

Mixing
Ozone is chosen as a tracer to test the gas-phase mixing time inside the reactor.The injection line was in the middle of the reactor near a fan.Ozone was generated at a rate of 0.8 L min −1 for 10 s and was injected into the reactor in three pulses.While higher rotation speed of fans provides better mixing inside the enclosure, wall loss rates were found to increase with rotation speed of the fan (Saathoff et al., 2003;Bloss et al., 2005;Carter et al., 2005;Wang et al., 2011).To make a balance between lower wall loss and better mixing, we finally settled on a fan rotation speed of 700 rpm.At this speed, ozone can be well mixed in about 120 s after each injection (Fig. 3).Compared to the experiment's duration, which may be several hours, this mixing time is very short.Tests with other compounds, like α-pinene, also revealed fairly good mixing within 1 min.

Dilution
Dilution may occur due to small leaks or high-volume sampling.Dilution rate may vary with the number and type of instruments taking samples from the reactor.In each experiment, a low-reactive compound such as SF 6 or CH 3 CN was injected as a tracer for dilution.Dilution was not detectable within the uncertainty of the instrument.

Light spectrum and intensity
The light spectrum emitted by black lamps is measured by EPP2000CXR-50 concave grating spectrometer (StellarNet Inc, USA) and is shown in Fig. 4. The black lamps produce an irradiation over the range from 340 to 400 nm with peak intensity at 369 nm.For the wavelength above 400 nm, essentially no intensity is produced except for some small peaks.In the first 8 months of operation, no appreciable variation of the light spectrum was observed.
Light intensity is represented by the photolysis rate of NO 2 , which is estimated by a steady-state actinometry.The operation process includes injecting NO 2 into the reactor, irradiating it and continuously measuring the concentrations of NO, NO 2 and O 3 .The photolysis rate of NO 2 , J NO 2 , is estimated according to the equation where [NO], [O 3 ] and [NO 2 ] represent concentrations (molecule cm −3 ) and k NO+O 3 (Atkinson et al., 2004) is the rate constant of ozone and NO reaction.A series of NO 2 actinometry experiments were carried out.A NO 2 photolysis rate of 0.49 ± 0.01 min −1 at full light intensity was obtained (corrected for the reaction of ozone and NO in the sampling lines).This value is a little lower than 0.58 min −1 calculated by TUV/NCAR model for Guangzhou on 22 June at noon.The four separately controlled groups of black lamps allow the NO 2 photolysis rate to be varied from 0 to 0.49 min −1 .

Wall loss of gases
Chamber wall effects include off-gassing of NO x and other reactive species, chamber radical sources and gas losses to the walls (Carter et al., 1982).They may have an impact on gas-phase reactivity and secondary organic aerosol formation.In early times of smog chambers, Grosjean (1985) performed a series of experiments to evaluate the wall loss rates of inorganic and organic species.
In the GIG-CAS smog chamber, wall loss rates of NO, NO 2 , O 3 and propene were evaluated by injecting a certain concentration of these gases and continuous monitoring their decay in the dark.Wall loss rates are obtained by  treating the wall loss as a first-order process.The average wall loss rates of NO, NO 2 and O 3 were 1.41 × 10 −4 min −1 , 1.39×10 −4 min −1 and 1.31×10 −4 min −1 , respectively.They are all within the range of reported values of other chamber facilities (Table 2).For propene, Wu et al. (2007) calculated a wall loss rate of 1.1 × 10 −7 s −1 in a 2 m 3 Teflon smog chamber.While in our GIG-CAS smog chamber, no wall loss of propene was observed within the uncertainty of GC-FID.Unlike NO, NO 2 and O 3 that might be lost to the reactor walls due to heterogeneous reactions, wall loss rate of propene was relatively small and not detectable.

Particle wall loss
Particle deposition onto the wall of the reactor is believed to be influenced by diffusion and the charged wall.Particle wall loss rate is proportional to the particle concentration and depends on the particle size.The particle number-weighted wall loss rate is described by first-order kinetics as where N (d p , t) is the particle number concentration, d p is the diameter of the particle and K N (d p ) is the particle number loss coefficient (Cocker et al., 2001a).The K N (d p ) values can be estimated from the particle number concentration versus time data in any experiment when no new particle is formed.Ammonium sulfate ((NH 4 ) 2 SO 4 , AS) seed particles are introduced into the reactor to evaluate the production of water-soluble inorganic particles and the particle number wall loss.AS solution amounting to 0.5 mol L −1 was atomized with a flow rate of 0.5 L min −1 for 20 min.As shown in Fig. 5, AS aerosols exhibit a median diameter of 150 nm within a few minutes after the injection and the median diameter slowly increases to 200 nm due to coagulation of smaller particles and probably higher wall loss rate of smaller particles.For aerosols with d p of 60, 100, 200, 300 and 400 nm, K N was determined to be 0.41, 0.23, 0.12, 0.08 and 0.10 h −1 , respectively.Like the study by Takekawa et al. (2003), our results also suggest that smaller particles deposit more easily onto the reactor walls.For aerosols with d p = 300 nm, the wall loss rate of 0.08 h −1 is close to but lower than that of 0.10 h −1 reported for the TCRDL chamber (Takekawa et al., 2003).The particle number wall loss rate is determined to be 0.17 h −1 , which means a lifetime of 5.9 h for particles.The result is comparable to results reported for other smog chambers (Table 3).

Propene-NO x experiments
The propene-NO x irradiation system was widely used as a reference system (Carter et al., 2005;Hynes et al., 2005) to evaluate the ability of a chamber to test the mechanisms for single organic-NO x irradiation systems.A set of four propene-NO x irradiation runs were carried out under dry conditions with a controlled temperature of 298.3 ± 0.5 K.The initial experimental conditions are listed in Table 4.The initial propene concentrations varied from 669 to 967 ppb and the initial propene / NO x ratios ranged from 1.8 to 2.6.
A near-explicit mechanism for propene from the Master Chemical Mechanism version 3.2 (Saunders et al., 2003) was used to simulate the propene-NO x irradiation experiments.Auxiliary mechanisms and relevant parameters used in the model are listed in Table 5.The parameters were determined by simulating clean air, low NO x -air, CO-NO x -air and CO-air irradiation experiments.Reaction rate coefficients of the N 2 O 5 hydrolysis to adsorbed HNO 3 , the photolysis of adsorbed HNO 3 to OH and NO 2 , and the wall loss of HNO 3 were assumed similar to those used by Hynes et al. (2005).
The reaction rate coefficient of the NO 2 dark, heterogeneous reaction NO 2 → 0.5 HONO + 0.5 wHNO 3 was determined to be 2.32 × 10 −6 s −1 from NO 2 dark decay experiments.
As mentioned above, the ozone and NO wall loss rates were determined to be 2.19 × 10 −6 s −1 (1.31 × 10 −4 min −1 ) and 2.34×10 −6 s −1 (1.41×10 −4 min −1 ), respectively.The reaction rate coefficients of light-induced production of OH and off-gassing of NO 2 from Teflon walls (Carter and Lurmann, 1991;Bloss et al., 2005) were determined to be 0.005 J NO 2 by simulating the experiments.The initial HONO concentration was varied from 0 to 5 ppb in order to give best fits to the experimental concentration profiles.
Figure 6 shows a comparison of observed and simulated concentration-time profiles of a propene-NO x irradiation system.The initial HONO concentration was adjusted to be 5 ppb in order to give the best fit for this experiment.Good agreements are obtained for most of the compounds such as propene, O 3 , NO, HCHO and CH 3 CHO.The peak NO 2 concentration is overpredicted; this phenomenon was also observed by Hynes et al. (2005) for the propene-NO x photolysis experiments or by Bloss et al. (2005) and Zádor et al. (2005) for ethene-NO x photolysis experiments.For better simulation, some unknown mechanisms involving NO 2 sinks most likely need to be included.The model shows a faster sink for NO 2 in the last 1 h.O 3 is overpredicted at the end of the experiment with a relative deviation of 4.8 %.The quantity ([O 3 ] − [NO]) is used to evaluate the model performance (Carter et al., 2005;Pinho et al., 2006).This quantity is defined as  (1.17-3.45)× 10 Varied from 0-5 ppb * wHNO 3 represents adsorbed HNO 3 on the wall; similarly, wNO and wO 3 represent adsorbed NO and O 3 , respectively.) have been calculated to vary from −2.2 to 23.7 % at the end of the experiments, which are within the values of ±25 % reported by Carter et al. (2005) when simulating VOC-NO x systems.The model results mentioned above illustrate that the GIG-CAS smog chamber can provide valuable data for mechanism evaluation.

α-pinene ozonolysis SOA yield
A series of experiments of α-pinene ozonolysis in the dark were carried out to evaluate the chamber facility in studying SOA formation chemistry.This reaction has been widely studied and numerous data are readily found in the literature (Hoffmann et al., 1997;Griffin et al., 1999;Cocker et al., 2001b;Saathoff et al., 2009;Wang et al., 2011).Methods developed by Pankow (1994a, b) and Odum et al. (1996) are used to analyze the data.Briefly, SOA yield, Y , is defined as follows: where ROG is the mass concentration of reactive organic gas (ROG) reacted and M 0 is the total mass concentration of organic aerosols formed.Y is a function of M 0 and the relation is described as where K om,i and α i are the mass-based absorption equilibrium partitioning coefficient and stoichiometric coefficient of product i, respectively, and M 0 is the total mass concentration of organic material.Odum et al. (1996) found that a two-product model could fit the yield data well if appropriate values for α 1 , α 2 , K om,1 , K om,2 were chosen.Five experiments of dark ozonolysis of α-pinene were carried out under dry conditions near 295 K in the absence of OH scavenger and seed particles.The initial experimental conditions and results are listed in Table 6.Initial mixing ratios of α-pinene varied from 13 to 87 ppb.An aerosol density of 1 g cm −3 was assumed to convert the volume concentration into the mass concentration for the calculation of aerosol yields.This assumed aerosol density is lower than that of 1.3 g cm −3 by Bahreini et al. (2005) or Alfarra et al. (2006) for α-pinene SOA.However, Wirtz and Martin-Reviejo (2003) and Wang et al. (2011) both adopted a value of 1.0 g cm −3 .Here we used the unit density for all the experiments when comparing our results with those from previous studies.Particles wall losses have to be accounted for to accurately quantify the SOA formation.The approach involving measuring the size-dependent wall loss rate of dry inert ammonium sulfate particles is not suitable for experiments in which nucleation occurs (Keywood et al., 2004).Here, we use the same method as Pathak et al. (2007) in that we calculate the total mass wall loss rate and apply this rate to the entire experiment.Comparison of yield data of this work and previous studies (Hoffmann et al., 1997;Griffin et al., 1999;Cocker et al., 2001b;Saathoff et al., 2009;Wang et al., 2011) is shown in Fig. 7.Only the yields obtained for the dark ozonolysis of α-pinene under dry conditions in the absence of OH scavenger and seed particles were considered.The two-product model gives a good fit for the yield data obtained in this work.The appropriate values for α 1 , α 2 , K om,1 and K om,2 are 0.189, 0.486, 0.0958 and 0.00218, respectively.Only the data of Saathoff et al. (2009) are slightly higher than the yield curve obtained in this work, probably owing to their assumed aerosol density of 1.25 g cm −3 when calculating aerosol mass concentrations.Most yield data of other studies are lower than the yields in this work for the same amount of SOA generated.This may be attributed to the influence of temperature because SOA yield is demonstrated to inversely depend on temperature (Saathoff et al., 2009).

Conclusions
We have built the largest indoor smog chamber in China, and initial characterization experiments described in this paper demonstrate that our GIG-CAS smog chamber facility can be used to provide valuable data for gas-phase chemistry and secondary aerosol formation.The chamber has been shown to exhibit good temperature homogeneity and mixing efficiency.Observed relative lower wall loss rates of gas species and particles reflect the long lifetime of these species and small wall effects.Results of propene-NO x -air irradiation experiments illustrate its utility for evaluating gas-phase chemical mechanisms.Furthermore, the good reproducibility of α-pinene ozonolysis experiments and the good agreement with previous studies demonstrate its ability for the study of secondary organic aerosol formation.

A
total of 135 black lamps (1.2 m long, 60 W Philips/10R BL; Royal Dutch Philips Electronics Ltd, the Netherlands) are arranged in two banks as the light www.atmos-meas-tech.net/7

Fig. 1 .
Fig. 1.(a) Schematic of the GIG-CAS smog chamber facility.T1-T8 are thermocouples in the enclosure and Tr is temperature sensor inside the reactor.(b) Schematic of two air purification systems.Route 1 represents the procedures of the Thermo zero air supply system; Route 2 represents the procedures of the larger flow rate air purification system.
Fig. 2. (a) Evolutions of the average temperature of nine thermocouples, eight of them 700 outside the reactor and one inside the reactor (See Fig. 1a), under black light irradiation 701 condition.Error bars are the standard deviations of the nine temperatures.The solid lines 702 represent set temperatures.(b) Average temperature values measured by the nine 703 thermocouples over a 6 h period under black light irradiation condition.Error bars show the 704 temperature variability in the 6 h period.705 706 707

Fig. 3 .
Fig. 3. Concentration-time plot of ozone after each injection.The two mixing fans rotated at a speed of 700 rpm in the experiments.

Fig. 3 .
Fig. 3. Concentration-time plot of ozone after each injection.The two mixing fans rotated at a speed of 700 rpm in the experiments.
where[NO]  0 and [O 3 ] 0 are the concentrations of NO and O 3 at the beginning of the experiment and [NO] final , [O 3 ] final are those at the end.([O 3 ] − [NO]) represents the amounts of NO oxidized and O 3 formed in the experiments, and also gives an indication of the biases in simulating O 3 formation.In the four experiments, the prediction biases of ([O 3 ] − [NO]

Fig. 7 .
Fig. 7. Comparison of yield data obtained for α-pinene ozonolysis experiments in GIG-CAS chamber with other chamber facilities.All experiments are carried out under dry condition.Blue line is the best fit two-product model for the data set of GIG-CAS.

Table 1 .
Overview of instruments.
a NA = not applicable.b Whichever is greater.c For total number concentration.d For organics.

Table 2 .
Summary of wall loss rates of gas species in GIG-CAS chamber and comparison with other chamber facilities.

Table 3 .
Comparison of particle wall loss rates in different smog chambers.

Table 4 .
Summary of initial conditions for propene-NO x -air irradiation experiments.

Table 5 .
Auxiliary mechanisms for chamber-dependent reactions.

Table 6 .
Summary of initial conditions and results for α-pinene ozonolysis experiments.