Gas-Wall Partitioning of Oxygenated Organic Compounds: Measurements, Structure–Activity Relationships, and Correlation with Gas Chromatographic Retention Factor

Abstract

Gas-wall partitioning of 50 oxygenated organic compounds was investigated by using gas chromatography to monitor time-dependent gas-phase concentrations of authentic standards added to a large Teflon environmental chamber. Compounds included C₈–C₁₄ monofunctional ketones and alcohols, C₅–C₉ monoacids, and C₄–C₁₀ diols with linear and cyclic structures. Measured time constants for reaching gas-wall partitioning equilibrium ranged from ∼10 to 100 min with an average value of ∼30 min and exhibited no obvious trend with compound structure, whereas the extent of equilibrium partitioning to the walls ranged from ∼0 to 100% and increased with increasing carbon number and with functional group composition in the order ketones < alcohols < monoacids < diols. When results were modeled using an approach analogous to one commonly used to describe absorptive gas-particle partitioning in terms of compound vapor pressure and aerosol mass loading it was determined that the absorptive properties of the Teflon film walls were equivalent to 2–36 mg m⁻³ of liquid organic aerosol particles. These results, when combined with those obtained in previous studies, indicate that most multifunctional products formed from the oxidation of atmospherically important hydrocarbons including isoprene, monoterpenes, aromatics, and alkanes have the potential to undergo significant partitioning to the walls of Teflon chambers and thus be lost from further chemical reaction and secondary organic aerosol formation as well as from gas and particle analyses. Two approaches for estimating equilibrium gas-wall partitioning in such studies are presented: one is a structure–activity relationship based on the absorptive gas-wall partitioning model and the other involves the use of observed correlations between gas-wall partitioning and compound retention on a gas chromatographic column.

Copyright 2015 American Association for Aerosol Research

1. INTRODUCTION

Environmental chambers are often used as reaction vessels in laboratory studies of gas and aerosol atmospheric chemistry. These chambers are most commonly constructed with clear fluorinated ethylene propylene (FEP) Teflon film (Hallquist et al. 2009), which provides chemically inert containment for reactions as well as optical transparency so that reactants can be exposed to ultraviolet light. Teflon-walled environmental chambers are effective for qualitative studies and for studies that require quantitation of volatile organic compounds (VOCs), but difficulties can arise in quantitative studies when lower volatility organic products participate in the chemistry of interest. The primary problem is that compounds with sufficiently low volatility can partition to Teflon chamber walls, thus removing from the gas phase compounds that might otherwise undergo further oxidation or participate in secondary organic aerosol (SOA) formation. These lost compounds will also be unavailable for sampling, and thus chamber concentrations and measured product yields will be lower than the true values. Matsunaga and Ziemann (2010) used gas chromatography to monitor gas-wall partitioning of homologous series of n-alkanes, 1-alkenes, 2-ketones, and 2-alcohols added to a FEP Teflon chamber and observed that the fraction of compound that partitioned to chamber walls increased with increasing carbon number and the addition of functional groups, that the absorptive properties of the Teflon walls were equivalent to that of ∼2 to 20 mg m⁻³ of liquid organic aerosol particles, that partitioning occurred by reversible absorption into the walls in which equilibrium was reached on timescales ranging from ∼10 to 60 min, and that partitioning did not depend significantly on the age of the Teflon chamber. Loza et al. (2010, 2014) and Zhang et al. (2015) have instead employed real-time mass spectrometry to monitor the loss of monofunctional and multifunctional organic compounds to the walls of Teflon chambers after injection and mixing of standard compounds or in situ formation via photooxidation of VOCs. Whereas they made similar observations regarding reversibility and the effects of chamber age, they measured longer gas-wall partitioning timescales and a larger range of equivalent absorbing organic aerosol mass concentrations for the walls. Recent modeling studies have also explored the interplay between gas-phase reactions, gas-particle partitioning, and gas-wall partitioning in SOA formation in chambers (Matsunaga and Ziemann 2010; McVay et al. 2014), while others have attempted to correct measured yields of SOA and individual reaction products for the effects of gas-wall partitioning. For example, Zhang et al. (2014) showed that losses of gaseous VOC reaction products to the walls of Teflon chambers lead to underestimates in measured SOA yields by up to a factor of 4, which helps to explain observed discrepancies between measured and modeled atmospheric SOA mass concentrations, whereas Yeh and Ziemann (2014a) showed that measured yields of alkyl nitrates formed from OH radical-initiated reactions of n-alkanes could be corrected for significant wall losses using a structure–activity relationship developed from measurements of gas-wall partitioning of alkyl nitrate standards and thereby achieved good agreement with model-predicted yields.

As a result of these recent studies, it has become apparent that knowledge of gas-wall partitioning of organic compounds in Teflon chambers is essential for accurate product quantitation and for modeling the chemistry of these compounds and their role in SOA formation in chamber reactions. Unfortunately, direct partitioning measurements are tedious and sometimes impractical. They usually either require compound standards—which may be commercially unavailable and impractical to synthesize—to be quantitatively added to the chamber and sampled over time, or in situ formation followed by identification and monitoring using real-time mass spectrometric analysis. Structure–activity relationships or other partitioning estimation methods, such as chromatographic retention, may provide attractive alternatives to direct measurement. To that end, we have measured the gas-wall partitioning of 50 oxygenated organic compounds containing one or more carbonyl, hydroxyl, or carboxyl moiety, and combined these with previous measurements to develop structure–activity relationships for hydrocarbons and oxygenated compounds. Additionally, gas-wall partitioning is shown to correlate with the gas chromatographic retention factors of the compounds studied, thus providing an additional means for estimating partitioning without the necessity of direct partitioning measurements.

2. EXPERIMENTAL METHODS

2.1. Chemicals

Ethyl acetate was HPLC grade purchased from Fisher Scientific. 2-Octanol (99%), 2-octanone (98%), 2-nonanol (97%), 2-nonanone (99%), 2-decanol (96%), 2-decanone (98+%), 5-decanol (99%), 1-undecanol (98%), 2-undecanol (99%), 3-undecanol (97%), 3-undecanone (98%), 4-undecanol (98+%), 4-undecanone (97%), 5-undecanol (98%), 5-undecanone (98%), 6-undecanol (97%), 6-undecanone (98%), 2-dodecanone (98%), 3-dodecanol (97%), 3-dodecanone (98%), 4-dodecanol (98%), 4-dodecanone (98+%), 5-dodecanol (97%), 5-dodecanone (98+%), 6-dodecanol (97%), 6-dodecanone (98%), cyclododecanol (98%), 2-tridecanol (99%), 2-tridecanone (98%), 2-tetradecanol (96+%), 2-tetradecanone (98%), and 7-tetradecanol (96+%) were purchased from ChemSampCo. 1,2-Butanediol (98%), 1,2-pentanediol (96%), 1,4-pentanediol (99%), 1,5-pentanediol (96%), 1,2-hexanediol (98%), 1,6-hexanediol (97%), 1,2-octanediol (98%), 1,8-octanediol (98%), 1,10-decanediol (99%), 1,2-decanediol (98%), 2-undecanone (99%), 1-dodecanol (98%), 2-dodecanol (99%), and 1,3,5,-trimethylbenzene (98%) were purchased from Sigma Aldrich. Pentanoic, heptanoic, octanoic, and nonanoic acid were quantitation grade purchased from PolyScience Corporation.

2.2. Environmental Chamber Experiments

Experiments were conducted in a 8.2 ± 0.4 m³ Teflon FEP environmental chamber filled with clean, dry air (<5 ppbv hydrocarbons, <0.1% RH) at room temperature (24–26°C) and ambient pressure (740 torr). The chamber volume was determined many months prior to these experiments by adding known amounts of trans-2-butene and then measuring concentrations using gas chromatography with flame ionization detection (GC-FID), similar to the method described below. Subsequent experiments in which known amounts of various VOCs were added and then analyzed prior to reaction confirmed the volume within the uncertainty due to filling (Yeh and Ziemann 2014b). It should also be noted that although this chamber had been used for many previous experiments, this should not have affected the results since it has been shown that gas-wall partitioning is essentially the same in new and used Teflon chambers (Matsunaga and Ziemann, 2010; Zhang et al. 2015).

Solutions containing one to seven compounds (chosen on the basis of chromatographic resolution) and a reference standard of 1,3,5-trimethylbenzene (TMB) dissolved in ethyl acetate were added to the environmental chamber. The protocols for compound addition and sampling and the precautions taken to eliminate compound losses during either process are described in detail in Matsunaga and Ziemann (2010). Compounds were added by evaporating the compound/TMB solution from a 300 cm³ glass bulb into a stream of nitrogen. The bulb had a 6 cm long outlet tube and by using careful and extended heating over the entire surface of the bulb it was guaranteed that the entire solution entered the chamber. The time required to add a compound/TMB solution to the chamber was ∼5 min. A Teflon-coated fan was turned on for 1 min after adding the solution to achieve rapid mixing. The resulting chamber concentrations based on the amount of compound added and the chamber volume were 18–30 ppbv for each compound. A TSI 3081 scanning mobility particle sizer (SMPS) with a model 3772 condensation particle counter was used to monitor the possible formation of particles during the experiments. In most cases low concentrations of particles were formed by homogeneous nucleation when heated compounds were added to the cooler chamber air, but the particles then quickly evaporated as they mixed in the chamber and were not detected a few minutes later. Their disappearance was much too fast to be due to particle deposition to the walls, which typically occurs in this chamber at a rate of ∼15% h⁻¹. The 18–30 ppbv concentrations used here were all well below saturation concentrations, with the highest saturation ratio being ∼0.3 for 1,10-decanediol and all other compounds being orders of magnitude lower. This was consistent with the observed rapid evaporation of particles, and guaranteed that compounds partitioned to the walls due to absorption into the Teflon film and not because they reached saturation with respect to the condensed phase.

Concentrations of compounds in the chamber were monitored over time by sampling air through 30 cm of stainless steel tubing and then into a glass tube packed with Tenax TA solid adsorbent and fritted with pesticide-grade glass wool. Prior to sampling, chamber air was drawn through the stainless steel tubing for 20 min at 250 cm³ min⁻¹ using a mass flow controller to allow the walls of the tube to equilibrate with the organic compounds of interest. This process limits the time at which the first sample can be collected to 20 min after adding compounds to the chamber. Samples were then collected for 10 min at the same flow rate at 20–30, 80–90, and 140–150 min after compound addition, which provided sufficiently complete time profiles for curve fitting purposes. After sampling, the Tenax cartridge was transferred to the inlet of an Agilent 5890 gas chromatograph equipped with a 30 m × 0.53 mm Agilent DB-5 column with 1.5 µm film thickness and a flame ionization detector. Analytes were thermally desorbed from the Tenax cartridge and eluted on an 8°C min⁻¹ gradient. The chamber concentration of each compound was expressed as a ratio of compound peak area relative to that of the TMB reference standard. The ratio at time t = 0 was determined by injecting 0.4 µL of the sample mixture into the Tenax cartridge and performing thermal desorption GC-FID analysis identical to that used for the chamber samples. TMB was chosen as a reference standard because it exhibited good GC retention and peak shape, and has been shown to undergo negligible gas-wall partitioning (Aschmann et al. 2013). By expressing compound chamber concentration as a ratio of the peak areas of compound/TMB reference standard, measurement errors due to variations in chamber and sample volumes were minimized.

3. RESULTS AND DISCUSSION

3.1. Gas-Wall Partitioning Time Constants and Equilibrium

The results of gas-wall partitioning measurements for the 50 investigated compounds are given in Table S1 in the online Supplemental Information (SI), where [OC_g/OC_T]_t is the ratio of the concentration of organic compound in the gas phase, OC_g, relative to the total concentration of organic compound in the chamber (amount of compound added divided by chamber volume), OC_T, measured at time t = 25, 85, and 145 min, [OC_g/OC_T]_eq is the value of the ratio at equilibrium, and τ_gw (min) is the time constant for gas-wall equilibration. The value of [OC_g/OC_T]_t was set to 1 at t = 0 min after all compound(s) had been added to the chamber, which usually took ∼5 min, and the chamber was mixed with the fan for 1 min. Values of [OC_g/OC_T]_eq and τ_gw (min) were obtained by fitting time profiles of [OC_g/OC_T]_t to the equation(1) which describes the change in compound concentration ratio from an initial value of 1 to the equilibrium value due to first-order loss of compound to the walls. In order to avoid artifacts due to measurement and curve fitting uncertainties, values of τ_gw determined for three compounds for which the initial concentration changed by less than 3% were considered unreliable and were discarded and values of [OC_g/OC_T]_eq were set equal to 1. It was also assumed that if ([OC_g/OC_T]_t=25 – [OC_g/OC_T]_eq)/(1 – [OC_g/OC_T]_eq) = exp(–25/τ_gw) was <0.05, then the value of τ_gw was <8 min. Since ([OC_g/OC_T]_t=25 – [OC_g/OC_T]_eq)/(1 – [OC_g/OC_T]_eq) decreases from 1 initially to 0 at equilibrium, if the value was <0.05 at 25 min then the compound should have essentially already been in gas-wall partitioning equilibrium when the first sample was taken.

3.1.1. Gas-Wall Partitioning Time Constants

The measured time profiles of [OC_g/OC_T]_t for 2-ketones, 2-alcohols, monoacids, and 1,2-diols are shown in Figure 1. The average values of τ_gw measured for these compound classes were approximately 49, 14, 10, and 12 min, respectively, and the isomers not shown in the figure behaved similarly (Table S1; see online SI). The average value of τ_gw for all 50 compounds was 26 ± 23 min with a range from <8 to 85 min. These time constants are similar to average values of 60, 16, <8, 17, and 95 min we measured previously for n-alkanes, 1-alkenes, 2-ketones, 2-alcohols (Matsunaga and Ziemann 2010), and alkyl nitrates (Yeh and Ziemann 2014a), respectively. For all the compound classes we have investigated, which include hydrocarbons and monofunctional and difunctional compounds, the range of values of τ_gw is ∼10–100 min with τ_gw less than ∼30 min for most compounds. Measurements made on isomers with similar structures or on the same compounds in different chambers and at different times indicate that measured values are usually reproducible to within a factor of ∼2 to 3 or less, so it appears that the range of measured time constants is probably due to a combination of measurement uncertainty, chamber conditions, and compound structure.

FIG. 1. Fraction of organic compound in the gas phase, [OC_g/OC_T]_t, measured at different times after adding (a) 2-ketones, (b) 2-alcohols, (c) monoacids, and (d) 1,2-diols to the chamber. The solid lines are least-squares fits of the time profiles to the equation [OC_g/OC_T]_t = [OC_g/OC_T]_eq + (1 – [OC_g/OC_T]_eq)exp(–t/τ_gw).

We further note that in a previous experiment (Matsunaga and Ziemann 2010) in which 2-ketones were added to the chamber and their concentrations were monitored for 7 h, although equilibrium was essentially reached in ∼25 min this was followed by a very slow decrease in compound concentrations at a rate of ∼1% h⁻¹. This corresponds to a value of τ_gw ∼100 h, which is approaching the values of ∼200 h or more that are predicted for Fickian diffusion (Frisch 1980) of small organic molecules such as acetic acid using permeability data provided by the Teflon film manufacturer (DuPont; http://www.americandurafilm.com/film-distribution/teflon-fep-film/). One interpretation of these observations is that gas-wall partitioning involves rapid uptake and equilibration of an organic compound on a timescale of ∼10 to 100 min due to perturbation and subsequent relaxation of the Teflon polymer structure, followed by much slower continuous uptake on a timescale of ∼10 to 100 h and greater that occurs as the polymers slowly relax under stress caused by absorbed organic compounds. This mechanism is similar to one proposed by Meares (1958) to explain results of experiments in which first rapid transient and then slow continuous changes in the permeability of allyl chloride in polyvinyl acetate films were observed on timescales similar to these.

The values of τ_gw ∼10–100 min measured here are similar to the value of ∼30 min calculated using the model of McMurry and Grosjean (1985) for a turbulently mixed chamber of either 4 or 60 m³ volume in which the wall accommodation coefficient is greater than ∼10⁻⁵. Under these conditions wall loss is limited by turbulent mixing and molecular diffusion through the boundary layer at the wall and is insensitive to the value of the wall accommodation coefficient, which therefore cannot be determined from our measurements. Values of τ_gw ∼10–100 min are also consistent with the optimum value of ∼70 min determined by Zhang et al. (2014) by fitting measured time profiles of SOA formed in a chamber from photooxidation of a variety of VOCs, but are much smaller than the values of ∼10 to 100 h measured by Zhang et al. (2015) for 25 multifunctional compounds formed in similar photooxidation experiments. It is not obvious why our results differ by so much from those of Zhang et al. (2015), although there were differences in experimental apparatus and methods. For example, Zhang et al. (2015) measured wall loss by using real-time mass spectrometric analysis to monitor relative changes in concentrations of compounds after they had been formed by ∼0.2–7 h of photooxidation, the chamber was 28 m³ and was mixed only by convection without use of a fan, compounds were sampled through Teflon tubing, and the equivalent absorbing organic aerosol mass concentration of the Teflon film was determined using partitioning theory and changes in compound concentrations observed when the chamber walls were heated to induce evaporation and shift gas-wall partitioning equilibrium.

With regards to the possible effects of using a fan to mix the chamber for 1 min after adding chemicals, it appears that this had no significant effect on our results. Support for this conclusion comes from experiments in which we have measured similar values of τ_gw for compounds formed in situ by photooxidation long after the fan was turned off and its effects on convection had disappeared. For example, in a series of experiments in which n-alkanes, methyl nitrite, and NO were mixed for 1 min with a fan and then ∼50 min later blacklights were turned on for 3 min to form OH radicals (Yeh and Ziemann 2014a), the time constants for alkyl nitrate reaction products to reach gas-wall partitioning equilibrium were similar to the values of ∼20 to 140 min measured when synthesized alkyl nitrate standards were added to the chamber. The time profiles of [OC_g/OC_T]_t measured for two C₁₂ and C₁₄ alkyl nitrate isomers are shown in Figure 2. If gas-wall partitioning equilibrium time constants were, for example, ∼10–100 h (Zhang et al. 2015) instead of ∼20 to 140 min, then the profiles for compounds formed in situ would be nearly flat. This comparison also demonstrates that it is important to know the initial concentration of compound prior to wall loss and that gas-wall partitioning measurements need to be continued out to times that are sufficiently long that the fraction of compound in the gas phase at equilibrium can be determined with reasonable confidence, since both these values affect the accuracy of the curve fitting used to determine the time constant. For example, if the alkyl nitrate time profiles were fit by using the first measurement at t = 25 min as the t = 0 point and an assumed first-order wall loss in which all of the compound eventually partitioned to the chamber walls, then the estimated time constants would be increased to ∼2–6 h.

FIG. 2. Fraction of organic compound in the gas phase, [OC_g/OC_T]_t, measured at different times after adding the C₁₂ and C₁₄ alkyl nitrate isomers 2- and 6-nitrooxydodecane (C₁₂ – 2N and C₁₂ – 6N) and 2- and 7-nitrooxytetradecane (C₁₄ – 2N and C₁₄ – 7N) to the chamber or forming them in situ. The open symbols correspond to synthesized alkyl nitrate standards added to the chamber using the same methods as for the other compounds described in the present study, whereas the closed symbols correspond to alkyl nitrates formed in situ by the reaction of n-dodecane and n-tetradecane with OH radicals in the presence of NO. For the purpose of comparing time profiles, the values of OC_T for the compounds formed in situ were chosen so that the value of [OC_g/OC_T]_t at 25 min was equal to that of the standard at 30 min. The solid lines are least-squares fits of the time profiles of the standards to the equation [OC_g/OC_T]_t = [OC_g/OC_T]_eq + (1 – [OC_g/OC_T]_eq)exp(–t/τ_gw). The data used to create this plot were taken from Yeh and Ziemann (2014a).

3.1.2. Gas-Wall Partitioning Equilibrium

The values of [OC_g/OC_T]_eq measured for linear n-alkanes, 1-alkenes, 2-ketones, 2-alcohols, 2-nitrates, monoacids, 1,2-diols, and α,ω-diols (hydroxyl groups on each of the terminal carbon atoms) for a range of carbon numbers are shown in Figure 3. The values for n-alkanes and 1-alkenes were taken from Matsunaga and Ziemann (2010), the values for alkyl nitrates were taken from Yeh and Ziemann (2014a), and the other values were measured in this study. Values of [OC_g/OC_T]_eq covered essentially the full range of possible values from <0.02 (the limit of detection) to 1.00. Values of [OC_g/OC_T]_eq <0.02 were obtained for 1,8-octanediol and 1,10-decanediol, which partitioned to the walls sufficiently rapidly that their gas-phase concentrations were below the limit of detection in all samples. Values of [OC_g/OC_T]_eq measured for 2-ketones and 2-alcohols were only slightly lower than those measured by Matsunaga and Ziemann (2010) in a similar chamber using the same methods.

FIG. 3. Fraction of organic compound in the gas phase at equilibrium, [OC_g/OC_T]_eq, measured for linear n-alkanes, 1-alkenes, 2-ketones, 2-alcohols, 2-nitrates, monoacids, 1,2-diols, and α,ω-diols for a range of carbon numbers. Values for n-alkanes and 1-alkenes were taken from Matsunaga and Ziemann (2010), values for alkyl nitrates were taken from Yeh and Ziemann (2014a), and all other values were measured in this study.

As observed previously (Matsunaga and Ziemann 2010), gas-wall partitioning depended on compound carbon number and functional groups. The fraction of compound that partitioned to the walls at equilibrium increased with increasing carbon number and for compounds with the same carbon number increased in the order n-alkane < 1-alkene < ketone < alcohol < nitrate < monoacid < diol. Although a significant isomer effect was observed previously for alkyl nitrates, in which partitioning to the walls increased as the functional group moved from the middle to the end of the molecule (Yeh and Ziemann 2014a), for the ketones and alcohols studied here the effect was relatively minor. Whereas primary alcohols did partition to the walls to a greater degree than the other isomers, partitioning of secondary alcohols and ketones did not correlate significantly with isomer structure. The greater effect of isomer structure observed for alkyl nitrates may be due to the larger size of the nitrate group compared to hydroxyl and carbonyl groups, which could affect transport through the Teflon polymer film. Unlike the monoalcohols, partitioning by diols exhibited a pronounced dependence on the location of the hydroxyl groups such that α,ω-diols underwent substantially higher partitioning to the walls than 1,2- and 1,4-diols. For example, measured amounts of the C₈ and C₁₀ α,ω-diols 1,8-octanediol and 1,10-decanediol were both below the limit of detection, whereas the amounts of 1,2-octanediol and 1,2-decanediol in the gas phase at equilibrium were 49% and 8%, respectively. It is also worth noting that for the C₅ diols 1,2-pentanediol, 1,4-pentanediol, and 1,5-pentanediol the fraction of compound that partitioned to the walls increased from 9% to 20% to 54%. This is also the order of decreasing compound vapor pressure, since vapor pressure decreases as functional groups move from the middle to the ends of the molecule and hydrogen bonding between adjacent functional groups can increase compound vapor pressure (Compernolle et al. 2011). These observations appear to confirm that for compounds in a particular class, whether they are monofunctional or multifunctional, the fraction that partitions to the walls increases with decreasing vapor pressure. Although our previous studies on monofunctional compounds were consistent with this conclusion, prior to the present study it also seemed possible that partitioning to the walls might instead increase with proximity of the functional group to the end of the molecule if the corresponding increase in the length of the carbon chain made it easier for molecules to diffuse into the holes between polymer chains in the Teflon film. If this were the case, however, then one would expect that 1,2-diols would partition to the walls more than α,ω-diols, which is the opposite of what was observed. With regards to other possible effects of compound geometry on gas-wall partitioning, the results obtained here for cyclic compounds, which one would expect to diffuse less easily between polymer chains than linear compounds, were inconclusive. Whereas the fraction of cyclododecanol that partitioned to the walls (38%) was approximately twice as much as for linear secondary dodecanol isomers (20%), the fraction of cyclododecanone that partitioned to the walls (7%) was approximately half as much as for linear dodecanone isomers (16%).

3.2. Gas-Wall Partitioning Model and Structure–Activity Relationship

It was shown previously (Matsunaga and Ziemann 2010; Yeh and Ziemann 2014a) that gas-wall partitioning of organic compounds to Teflon chamber walls can be modeled by analogy to gas-particle partitioning, which is generally described using the theory developed by Pankow (1994). According to this model the ratio [OC_w/OC_g]_eq, where OC_w = OC_T – OC_g is the concentration of organic compound that partitions to the walls (amount of compound lost to the walls divided by chamber volume), is given by the equation(2) where C_w is the equivalent absorbing organic aerosol mass concentration of the Teflon film, M_w is the mean molecular mass of the Teflon film, γ_w is the compound activity coefficient in the Teflon film, R is the gas constant, T is temperature, and P^o is the pure compound vapor pressure. Thus, when [OC_w/OC_g]_eq is plotted against RT/P^o, a straight line is expected with the slope being equal to C_w/M_wγ_w. This inverse relationship between the fraction of compound that partitions to the walls and compound vapor pressure is similar to what has been observed for Fickian diffusion of smaller compounds in FEP Teflon, where the permeability (mol m⁻¹ d⁻¹ Pa⁻¹) follows the order acetic acid > ethanol > acetone > hexane, which is in the order of decreasing compound vapor pressure (Matsunaga and Ziemann 2010). Here, vapor pressures at 298 K were calculated using the group contribution methods SIMPOL.1 (Pankow and Asher 2008) with coefficients that have been updated based on an improved data fitting procedure (K. Barsanti, personal communication) and EVAPORATION (Compernolle et al. 2011). For the compounds investigated here the updated SIMPOL.1 coefficients give values that are within ∼30% of those calculated with the original coefficients, which are reported at 293 K, and the major differences between the SIMPOL.1 and EVAPORATION methods are that the latter includes the effects of functional group position on a molecule and non-additive contributions of multiple functional groups. The vapor pressures calculated for 2-ketones, 2-alcohols, monoacids, and 1,2-diols are given in Table S2 in the online SI and plots of [OC_w/OC_g]_eq against RT/P^o are shown in Figure 4. The slopes of the lines determined for 2-alcohols and monoacids were similar for the two methods, but differed by a factor of ∼3 for the 2-ketones and 1,2-diols, with SIMPOL.1 yielding the larger slope for 2-ketones and smaller slope for 1,2-diols. The y-intercepts for all plots were close to 0. Because of the relatively large range of slopes, the inclusion of other isomers would not significantly affect the results. The values of C_w were calculated from the slopes of the lines (units of µmol m⁻³ in Figure 4) and the equation C_w = slope × M_wγ_w, assuming for convenience (as in Matsunaga and Ziemann 2010) that the Teflon film had absorptive properties similar to typical organic aerosol particles such that γ_w = 1 (an ideal solution) and M_w = 200 g mol⁻¹ (Seinfeld et al. 2001). The values of C_w for 2-ketones, 2-alcohols, monoacids, and 1,2-diols were 8 and 24, 3 and 2, 30 and 36, and 24 and 8 mg m⁻³ for vapor pressures calculated using EVAPORATION and SIMPOL.1, respectively. These values are in the same range as those determined previously (Matsunaga and Ziemann 2010; Yeh and Ziemann 2014a) for n-alkanes, 1-alkenes, 2-ketones, 2-alcohols, and 2-alkyl nitrates of 2, 4, 10, 24, and 36 mg m⁻³ using vapor pressures estimated by empirical methods that gave values similar to those calculated using EVAPORATION and SIMPOL.1. The values of C_w determined here and in the previous study were 8, 10, and 24 mg m⁻³ for 2-ketones and 2, 3, and 24 mg m⁻³ for 2-alcohols, respectively.

FIG. 4. Relationship between measured values of the ratio of organic compound in the chamber walls and in the gas phase at equilibrium, [OC_w/OC_g]_eq, and calculated values of RT/P^o for (a) 2-ketones, (b) 2-alcohols, (c) monoacids, and (d) 1,2-diols. Values of P^o were calculated at 298 K using SIMPOL.1 (Pankow and Asher 2008) with updated coefficients (K. Barsanti, personal communication) and EVAPORATION (Compernolle et al. 2011) group contribution methods. The solid lines are linear least-squares fits to the data and the listed slopes have units of µmol m⁻³.

Combining results of all the measurements we have made previously (Matsunaga and Ziemann 2010; Yeh and Ziemann 2014a) with those presented here, C_w ranges from 2 to 36 mg m⁻³, with average values of 3 ± 1 mg m⁻³ for hydrocarbons and 20 ± 10 mg m⁻³ for oxygenated compounds, where the stated uncertainty is one standard deviation. Although functional groups appear to increase C_w by about an order of magnitude compared to values for hydrocarbons, probably due to effects of functional groups on the activity coefficients for compounds absorbed in Teflon, which are here assumed to be unity, because of uncertainties in gas-wall partitioning measurements and calculated vapor pressures used to determine these values it is not clear if there is a relationship between C_w and specific functional groups. On the basis of these results, it appears that for many modeling applications the use of Equation (2)(2) with C_w = 3 × 10⁻⁶ kg m⁻³ for hydrocarbons and 20 × 10⁻⁶ kg m⁻³ for oxygenated compounds, γ_w = 1, M_w = 0.2 kg mol⁻¹, R = 8.21 × 10⁻⁵ m³ atm K⁻¹ mol⁻¹, and P^o (atm) calculated at T (K) using either the SIMPOL.1 (Pankow and Asher 2008) or EVAPORATION (Compernolle et al. 2011) group contribution methods should provide a reasonable structure–activity relationship for estimating equilibrium gas-wall partitioning of organic compounds in Teflon chambers. These results differ significantly from those of Zhang et al. (2015), who determined values of C_w that varied by six orders of magnitude, from 4 × 10⁻⁴ to 3 × 10² mg m⁻³, and increased with increasing compound vapor pressure. As noted above, the method they used to determine those values involved the use of partitioning theory and changes in compound concentrations observed when the chamber walls were heated to induce evaporation and shift gas-wall partitioning equilibrium.

3.3. Gas-Wall Partitioning and GC Retention Factor Correlation

In gas chromatography, retention of a compound on the column is governed by partitioning of the compound between the gas phase and liquid organic phase that coats the column. This partitioning process is sufficiently similar to the one responsible for gas-wall partitioning of organic compounds in Teflon film chambers to suggest that retention of compounds on a GC column might correlate with [OC_w/OC_g]_eq values and thus provide a simple means for estimating gas-wall partitioning via GC analysis. This approach seems especially promising given that gas chromatography has been successfully used to estimate vapor pressures for a variety of compounds by correlating vapor pressure to retention time (Fischer et al. 1992; Donovan 1996; Luxenhofer et al. 1996; Lei et al. 1999), and it was shown above in Equation (2)(2) that [OC_w/OC_g]_eq is inversely proportional to P^o. The DB-5 GC column (Agilent Technologies) used here is a cross-linked/surface bonded 5% phenyl, 95% methylpolysiloxane capillary column that has been used previously to estimate vapor pressures of polychlorinated napthalenes (Lei et al. 1999), and so is expected to provide reasonable correlations between GC retention time and [OC_w/OC_g]_eq. GC retention factors calculated as RF = (compound retention time/unretained compound retention time) – 1 were determined from liquid injections of all compounds with the earliest eluting baseline perturbation used for the retention time of an unretained compound (Snyder et al. 1997). Results are given in Table S3 in the online SI.

Plots of log[OC_w/OC_g]_eq against GC retention factor for all compounds and the linear least-squares regression lines for monofunctional linear alcohols, ketones, and monoacids, and difunctional 1,2-diols, and α,ω-diols are shown in Figure 5. There is a strong positive correlation between the extent of partitioning to the walls and retention factor for each class of compound, with all r² values being >0.97. Furthermore, the slopes of the regression lines all fall within a narrow range from 0.177 to 0.200, which is a reflection of the fact that within any compound class the retention factor increases nearly linearly with increasing carbon number. This is because logP^o of a compound decreases by about a factor of 0.44 to 0.50 per CH₂ unit (Pankow and Asher 2008; Compernolle et al. 2011) and within a particular compound class the retention factor increases linearly with decreasing logP^o (the basis for GC vapor pressure measurement techniques). It is worth noting that the data point for pentanoic acid, the smallest monoacid studied, was not used in the correlation for monoacids because its inclusion leads to a significantly steeper slope that is probably caused by dimer formation that is known to occur at room temperature for monoacids with carbon number ≤5 (Compernolle et al. 2011). These low volatility dimers will partition to the chamber walls to a greater extent than monomers, but because they dissociate at the elevated temperatures encountered in the GC their presence will not be reflected in the retention factor, thus leading to an artifact in the correlation. The good linear fits shown in Figure 5a and b indicate that, for a particular class of compounds, gas-wall partitioning can be estimated using retention factors measured on a GC equipped with a DB-5 column. With such a correlation, one may only need to directly measure gas-wall partitioning for a few model “calibrant” compounds and then use those compounds as retention standards for the GC analysis. The estimation of gas-wall partitioning using retention factors is rapid and insensitive to compound purity. Furthermore, as long as compounds will pass through the GC column and can be identified, for example by mass spectrometry, then even if standards of chamber reaction products are commercially unavailable and impractical to synthesize, GC retention times could provide useful estimates of gas-wall partitioning.

FIG. 5. Relationship between measured values of the logarithm of the ratio of organic compound in the chamber walls and in the gas phase at equilibrium, log[OC_w/OC_g]_eq, and GC retention factor for (a) monofunctional, (b) difunctional, and (c) all compounds. The solid lines are linear least-squares fits to data for linear alcohols, ketones, and monoacids in (a) and 1,2-diols and α,ω-diols in (b). The solid lines in (c) were obtained by averaging the slopes and y-intercepts of the regression lines of monofunctional and difunctional compounds in (a) and (b), and the dashed lines were drawn 0.5 log units above and below the solid lines.

Unfortunately, although the slopes of the regression lines are similar, the y-intercepts shown in Figure 5a and b differ significantly across compound classes, ranging from −3.99 to −1.42, and thus the retention factor correlations break down when comparing partitioning behavior of compounds with different functional groups. Nonetheless, it appears that one might be able to obtain reasonable estimates of gas-wall partitioning by using two regression lines: one for monofunctional compounds and one for difunctional compounds, obtained by averaging the slopes and y-intercepts of the regression lines of the individual monofunctional and difunctional compounds. The resulting equations are log[OC_w/OC_g]_eq = 0.182 × RF – 3.45 and log[OC_w/OC_g]_eq = 0.190 × RF – 1.80 for the monofunctional and difunctional compounds, respectively. These are plotted as the two solid lines in Figure 5c. For each of the two lines the associated data points fall within relatively narrow bands whose borders are about 0.5 log units above and below the solid regression lines and are designated in Figure 5c by dashed lines. The values of log[OC_w/OC_g]_eq calculated from a measured retention factor and one of these solid lines should therefore be uncertain by about ±0.5, which corresponds to factors of approximately 0.3 and 3 for lower and upper bounds on [OC_w/OC_g]_eq. For example, for values of log[OC_w/OC_g]_eq = −1, 0, or 1, which covers most of the range of gas-wall partitioning, the corresponding values of [OC_w/OC_g]_eq will be 0.1 (0.03, 0.3), 1 (0.3, 3), and 10 (3, 30), which correspond to [OC_g/OC_T]_eq = 0.9 (0.8, 1.0), 0.5 (0.3, 0.8), and 0.1 (0.03, 0.3), where the values have been rounded to one significant figure and those in parentheses correspond to the lower and upper bounds. These may be acceptable uncertainties for many experimental and modeling applications. Furthermore, assuming this approach is valid and the two equations given above can be used to estimate [OC_w/OC_g]_eq values for monofunctional and difunctional compounds from measured retention factors, regardless of the identity of the functional groups, then a general equation for estimating [OC_w/OC_g]_eq values for compounds containing N functional groups might be log[OC_w/OC_g]_eq = (0.186 × RF) + (1.65 × N) – 5.10, which was obtained by using the average of the slopes of the lines for monofunctional and difunctional compounds and assuming that the shift in the y-intercept of the line per functional group is 1.65, the difference between the y-intercepts of the lines for monofunctional and difunctional compounds. One should be careful in using such extrapolations, however, since the effects of functional groups on absorption of compounds into Teflon may not be additive for more highly functionalized compounds.

4. CONCLUSIONS

The results of this study can be combined with those reported previously by Matsuanga and Ziemann (2010) and Yeh and Ziemann (2014a) to obtain a reasonably self-consistent description of gas-wall partitioning of organic compounds in a Teflon film chamber. The process appears to occur by relatively rapid absorptive uptake and equilibration of the organic compound on a timescale of ∼10 to 100 min due to perturbation and subsequent relaxation of the Teflon polymer structure, followed by much slower continuous uptake on a timescale of ∼100 h that occurs as the polymers slowly relax under stress caused by absorbed organic compounds. The observed timescales for rapid uptake and equilibration are consistent with those calculated using a model developed by McMurry and Grosjean (1985) to describe a turbulently mixed chamber in which the wall accommodation coefficient is greater than ∼10⁻⁵ and thus wall loss is limited by turbulent mixing and molecular diffusion through the boundary layer at the wall. Measurements also indicate that gas-wall partitioning is reversible and follows a Henry's law relationship, and thus can be modeled by analogy to gas-particle partitioning. For all the compounds measured here and in our previous studies, which include hydrocarbons (n-alkanes and 1-alkenes) and oxygenated compounds (ketones, alcohols, alkyl nitrates, monoacids, and diols), the average equivalent absorbing organic aerosol mass concentration of the Teflon film walls was 3 ± 1 mg m⁻³ for hydrocarbons and 20 ± 10 mg m⁻³ for oxygenated compounds. The differences in the mean values for the two compound classes is probably due to differences in activity coefficients for absorption in the Teflon film, whereas the range of values is probably due to uncertainties in gas-wall partitioning measurements and calculated vapor pressures used to determine these values. This relatively narrow range of values indicates that for the purpose of modeling equilibrium gas-wall partitioning of hydrocarbons and oxygenated compounds it should be reasonable to use the structure–activity relationship presented here with assumed values of 3 and 20 mg m⁻³, respectively, and compound vapor pressures calculated using either the SIMPOL.1 (Pankow and Asher 2008) or EVAPORATION (Compernolle et al. 2011) group contribution method. Because the monofunctional and difunctional compounds used to develop the relationships presented here are less functionalized than many SOA components, however, extrapolation of these results to wall losses of more highly oxidized compounds should be done with caution. Additionally, gas-wall partitioning may be affected by factors not yet studied, such as humidity, chamber geometry, wall materials other than FEP Teflon, and mixing conditions.

Unfortunately, because our studies have identified no clear relationship between compound structure and the timescale for gas-wall partitioning, one needs to assume a value for this quantity, probably in the range from ∼10 to 100 min with a reasonable average of ∼30 min. For example, Zhang et al. (2014) assumed a value of 10 mg m⁻³ and then determined an optimum value of ∼70 min for the timescale for gas-wall partitioning by fitting measured time profiles of SOA formed in chamber reactions from photooxidation of a variety of VOCs.

In the present study, we have also explored the use of empirical correlations for obtaining compound-specific equilibrium gas-wall partitioning parameters as an alternative to using the approach recommended above. In particular, we have compared the gas-wall partitioning of organic compounds with their GC retention factors measured with a DB-5 column that is commonly used for analysis of oxygenated compounds. Unfortunately, the results indicate that this approach is unsuitable for developing a single correlation across compound classes, due to the different compound selectivity exhibited by DB-5 when compared to the FEP Teflon film. The use of GC columns with DB-210 (100% trifluoropropyl methyl polysiloxane) stationary phase may better approximate the chemical selectivity and partitioning characteristics of the FEP Teflon film, or a FEP Teflon capillary column could perhaps be substituted for the silica-based GC columns. In the latter case, the lack of a bonded liquid stationary phase would likely require the GC oven to be operated at sub-ambient temperatures to ensure adequate compound retention. Alternatively, although a single correlation suitable for predicting gas-wall partitioning based on compound retention could not be obtained with the DB-5 column, it appears that correlations obtained when monofunctional and difunctional compounds are treated separately may be sufficient for this purpose. Additional studies should be conducted to verify these results with other classes of multifunctional compounds, but, if valid, correlations obtained using a few commercially available compounds with good chromatographic properties could be routinely employed to estimate gas-wall partitioning.

Finally, it is important to note that these results indicate that gas-wall partitioning has the potential to significantly affect the gas- and particle-phase concentrations of many of the products formed in chamber studies of the oxidation of atmospherically important hydrocarbons such as isoprene, monoterpenes, monoaromatics, and alkanes. These hydrocarbons all react with OH radicals and in some cases with NO₃ radicals and O₃ (Atkinson and Arey 2003), with a large fraction of the first-generation products being multifunctional compounds containing various combinations of hydroxyl, carbonyl, nitrate, hydroperoxy, and carboxyl groups (Capouet et al. 2008; Paulot et al. 2009; Ziemann and Atkinson 2012). In the case of multifunctional alcohols, for example, it was shown here that the fraction of compound that partitioned to the walls was ∼10% and ∼20% for C₅ 1,2- and 1,4-diols, and ∼90% and ∼100% for C₁₀ 1,2- and α,ω-diols, respectively. These C₅ and C₁₀ difunctional compounds have structures similar to many first-generation products formed from the oxidation of isoprene (C₅ diene) and monoterpenes (C₁₀ alkenes), respectively, although substitution of different functional groups will likely reduce (carbonyl), enhance (nitrate and carboxyl), or minimally change (hydroperoxy) partitioning to the walls compared to hydroxyl groups. Similar products are formed from reactions of monoaromatics, which typically have carbon numbers in the range C₇–C₁₁ (Gentner et al. 2012) and from alkanes ≥C₅. For all these hydrocarbons there are also reaction pathways that either lead to fragmentation of the carbon backbone or further addition of functional groups, which will usually reduce and enhance partitioning to the walls, respectively. Furthermore, it has been shown through measurements and modeling studies (Kroll et al. 2007; Matsunaga and Ziemann 2010; McVay et al. 2014; Zhang et al. 2014) that when aerosol particles are added as seed or formed in a chamber reaction that gas-particle partitioning will compete with gas-wall partitioning and thus reduce the loss of gaseous compounds to the walls.

SUPPLEMENTAL MATERIAL

Supplemental data for this article can be accessed on the publisher's website.

ACKNOWLEDGMENTS

Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation (NSF).

Funding

This material is based on work supported by the National Science Foundation under Grants AGS-1219508 and AGS-1420007.