Correlation gas chromatography and two-dimensional volatility basis methods to predict gas-particle partitioning for e-cigarette aerosols

Abstract E-cigarette aerosols contain a complex mixture of harmful and potentially harmful chemicals. Once released into the environment, they evolve and become new sources of indoor air pollutants that could pose a significant threat to both users and non-users. However, current understanding of the physicochemical properties of e-cigarette aerosol constituents that govern gas-particle partitioning in the atmosphere is limited, making it difficult to estimate the health risks associated with exposure. Here, we used correlation gas chromatography (C-GC) and two-dimensional volatility basis set (2D-VBS) methods to determine the vapor pressures and volatility for commonly reported toxic and irritating e-cigarette aerosol constituents. The vapor pressures of target compounds at 298 K were estimated from the Antoine-type linear relationship between the vapor pressure of reference standards and their retention times. Our C-GC results showed an overall positive correlation (R = 0.84) with estimates using the EPI (Estimation Programs Interface) Suite. The volatility calculated by 2D-VBS correlates well with the calculated vapor pressure from both C-GC (R = 0.82) and EPI Suite (R = 0.85). The volatility distribution also indicated fresh e-cigarette aerosol constituents are mainly more volatile organic compounds. Our case study revealed that low-vapor-pressure compounds (e.g., σ-dodecalactone, γ-decalactone, and maltol) become enriched in the e-cigarette aerosols within 2 h following vaping emissions. Overall, these findings demonstrate the applicability of the C-GC and 2D-VBS methods for determining the physiochemical properties of e-cigarette aerosol constituents, which can aid in assessing the dynamic chemical composition of e-cigarette aerosols and exposures to vaping emissions in indoor environments. Copyright © 2024 American Association for Aerosol Research Graphical Abstract


Introduction
The use of e-cigarettes and vaping products has become a major public health concern due to the increasing number of reports of severe lung diseases and deaths linked to their use (Gordon et al. 2022;Creamer et al. 2019).According to the National Health Interview Survey in 2021, an estimated 11.1 million adults in the United States were active vapers CONTACT Ying-Hsuan Lin ying-hsuan.lin@ucr.eduEnvironmental Sciences, University of California Riverside, 900 University Ave., Riverside, CA 92521, USA.
Supplemental data for this article can be accessed online at https://doi.org/10.1080/02786826.2024.2326547.(Cornelius et al. 2023).With the growing popularity of e-cigarettes and the lack of regulation regarding their emissions, they continue to evolve and become a new source of indoor air pollutants.Passive exposure to vaping aerosols may pose a significant threat to non-users, particularly vulnerable populations (Giroud et al. 2015;Ballb� e et al. 2014).Commonly, the e-liquid ingredients in e-cigarettes include glycerol, propylene glycol, nicotine, and different flavors.Through the vaping processes, e-liquids are aerosolized and inhaled by the users, whereas exhaled e-cigarette aerosols can linger in the air (Schripp et al. 2013).The size of the aerosols can determine the dose of aerosols deposited onto different regions of the respiratory tract (Su et al. 2021), while the chemical composition of the aerosols is a crucial factor in determining their fate and transport in the atmospheric environment and toxicity.The combination of these characteristics is crucial for determining the health risks associated with vaping aerosols (Paur et al. 2011;Schmid et al. 2009;Stoeger et al. 2006).However, current knowledge about the size distribution and chemical composition of e-cigarette aerosols and the differences in exposure between active and passive vaping remains incomplete.
Vapor pressure is one of the important factors governing the fate and transport of chemical compounds in the atmosphere.Compounds with low vapor pressure are more likely to partition onto the particle phase of airborne aerosols (Lumiaro et al. 2021;Xiao and Wania 2003).Thus, it is critical to understand the vapor pressure of vaping aerosols in order to estimate their impacts on indoor air quality and human exposure.However, determining vapor pressure for numerous aerosol constituents to achieve a comprehensive assessment is challenging, and it is worth noting that the reported values of vapor pressure for many compounds are frequently estimated in silico, such as through quantum chemistry, regression, or machine learning-based modeling for molecular descriptors coupled to quantitative structure-property relationships (Zang et al. 2017;Gomis et al. 2015;Bhhatarai et al. 2011;Yan and Gasteiger 2003).However, estimates generated in silico have varying degrees of error, the average absolute prediction error of vapor pressure on log(p/(Pa)) ranged from 0.105 to 0.285 for different modeling (Felmlee, Morris, and Mager 2012;Dearden 2003).Even errors can be observed between variable software packages and versions (Mansouri et al. 2016;US EPA 2012).This is because potential errors in the predictions may arise due to inaccuracies in the molecular descriptors or limitations in the underlying mathematical models.In addition, there can be variations between the predicted vapor pressure and the actual vapor pressure observed in real-world conditions.Factors such as temperature, pressure, and interactions with other compounds can all alter molecular behavior, which may not be fully captured by the computational models.As a result, experimental validation is still required to verify the accuracy of these computational predictions.
Experimentally, the sub-cooled liquid vapor pressure or vaporization enthalpies (D l g H m (T)) of volatile or semi-volatile organic compounds can be estimated by gas chromatography (GC) based on the correlation of their retention times (RT) (Vuong et al. 2022;Brommer et al. 2014;Hinckley et al. 1990).The slope and intercept indicated the ratio of vaporization enthalpies ( where t x and t ref are the retention time of analytes and reference compounds. Then, the vapor pressure at 298K ðp GC, 298K ) can be obtained from: The GC-RT method has been widely used to determine the vapor pressure of non-polar or semi-polar compounds, such as polycyclic aromatic hydrocarbons, organochlorine pesticides, perfluorinated compounds, and polychlorinated biphenyls, by using nonpolar GC columns (Okeme et al. 2020;Vuong et al. 2020;Brommer et al. 2014;Meylan and Howard 2005;Shen and Wania 2005;Lei et al. 2002Lei et al. , 2004;;Hinckley et al. 1990).However, the application of the GC-RT method to polar compounds is somewhat limited since the complex functional groups of polar compounds influence their vapor pressure to some extent (Wilson, Gobble, and Chickos 2015;Gobble, Chickos, and Verevkin 2014).A modified approach, correlation gas chromatography (C-GC), has been reported to provide improved correlations, especially for polar compounds (Gobble and Chickos 2015).The application of the C-GC method relies on the linear relationship between adjusted retention time (t a ) and both temperature and vapor pressure, and estimates the enthalpy of transfer of the analyte from the column to the vapor at the mean temperature, D trn H a ðT m Þ: The adjusted retention time of each analyte (t a ) under 298K is estimated by assuming the linear relationship between ln(t 0 /t a, t 0 ¼ 60s) and 1/T, where the R is the gas constant (8.3145 Then, the vapor pressure of each analyte is determined by the linear relationship between ln(t 0 /t a ) and ln(p/p 0, p 0 ¼ 101325 Pa) among reference compounds.Compared to the traditional GC-RT method, the application of the C-GC method is less demanding on the selection of reference compounds but more dependent on the linear relationship between temperature and the retention time of the compound itself.
Compared to the experimental approaches mentioned above that require the explicit molecular information of individual organic aerosol constituents, the two-dimensional volatility basis set (2D-VBS) (Donahue et al. 2006(Donahue et al. , 2011) ) approach has been recently developed to parameterize the dynamic organic aerosol (OA) growth and evolution by lumping organic species with similar volatility together into "bins" in order to reduce the complexity of the numerous constituents of organic aerosols.Within this framework, the volatility bins are constrained by the saturation mass concentration (C � ) (Wang et al. 2020) and the oxygen content in bulk organic aerosol composition to approximate the extent of oxygenation using the oxygen to carbon (O:C) ratios (Donahue et al. 2011), regardless of the availability of their molecular identities.The 2D-VBS is particularly useful to characterize the volatility distribution of a variety of organic compounds in both gas and particle phases, and these compounds are grouped into the following categories: volatile (VOC), intermediate volatility (IVOC), semi-volatile organic compounds (SVOC), low volatility (LVOC), and extremely low volatility (ELVOC).In addition, 2D-VBS has been widely used in recent studies to investigate the properties of OAs (Ng et al. 2010), highly oxygenated molecules (Bianchi et al. 2019), and organic oxygenated molecules (Tian et al. 2023).
In this study, we aim to establish a fundamental dataset of the physicochemical properties of e-cigarette aerosol constituents using the C-GC and 2D-VBS methods.Specifically, we aim to utilize the vapor pressure and volatility of constituents in e-cigarette aerosols as important parameters to illustrate the dynamic changes in chemical compositions of e-cigarette aerosols with respect to time and equilibrium with the simulated indoor environment, which can be linked to the differential exposure profiles and their potential health effects.For this reason, after the estimation of vapor pressure and volatility, we conducted a case study using Puff Bar Grape TM to compare the fractional contributions of chemical species with different vapor pressures and investigate the changes in e-cigarette aerosol composition governed by gas-particle partitioning.A list of 30 chemicals that have been commonly detected in e-cigarette aerosols in prior studies (Sassano et al. 2018) and known as toxins or irritants was selected to estimate their vapor pressure and volatility.Although carbonyls such as formaldehyde, acetaldehyde, and acrolein have also been detected in e-cigarette aerosols (Ranpara et al. 2023), due to the analytical challenges, the carbonyls of interest cannot be detected by GC-MS directly without functional group derivatizations.As a result, in the current study, we only selected a set of analytes that are directly detectable by GC-MS (Hua et al. 2019;Omaiye et al. 2019) in the C-GC analysis.To ensure the reproducibility of C-GC results, we used freshly prepared mixtures of commercially available chemical standards at a concentration of 20 ng/mL for each target analyte that exceeded the GC/MS detection limit.The vapor pressure derived from the C-GC technique was compared to the model-estimated vapor pressure from the EPI Suite.Moreover, we applied the 2D-VBS method to determine the volatility of those fresh vaping aerosols to investigate the correlation between volatility and vapor pressure.This study contributes to an improved understanding of the dynamic nature of e-cigarette aerosol composition, which can serve as a foundation for future studies to assess their exposure and potential health effects.

Chemicals
In this work, 30 major constituents in e-cigarettes, which are concentrated and commonly reported as toxic and irritating, were studied (Sassano et al. 2018).Target compounds were separated into three categories based on their functional groups, including hydrocarbons & alcohols, carbonyls, alkoxy carbonyls & organic acids.The compound name, formula, molecular weight (MW), CAS number, boiling point, and the EPI suite estimated vapor pressure of target compounds are presented in Table 1.

Instrumentation
The C-GC analysis was performed on an Rtx-VMS GC column (30 m � 0.25 mm i.d., 1.40 lm, cat #19915) installed in an Agilent Technologies 6890 N GC equipped with a 5975 C inert mass selective detector.A solvent delay of 6 min was included in the GC runs.The temperatures of the injection inlet (250 � C), the transfer line (280 � C), the ion source (230 � C), and the quadrupole (150 � C) were kept constant.According to the vapor pressure of the target compounds, the GC oven temperatures ranged from T ¼ (318 to 348 K), (338 K to 368 K), (348 to 388 K), (388 to 418 K), (418 to 433 K), and (433 to 468 K) at T ¼ 5 K intervals.The details of each run are provided in Tables S1 to S18 in supporting information (SI).Three replicates were conducted for each group, the final retention time is the average value among three runs.The flow rate of helium carrier gas was 2 mL min −1 for all runs.Compounds were identified using the NIST 2008 mass spectral database.The detailed procedures for the operation of GC/MS can be found in previous publications (Canchola et al. 2022b(Canchola et al. , 2022a;;Chen et al. 2019).

The procedure
The schematic of the whole procedure of the C-GC method is shown in Figure 1.The adjusted retention time (t a ) of the target compound on the column is determined by the original retention times of the analyte (t) and solvent (t s ), assuming the solvent is unretained by the column, which can represent the time required for the analyte to travel through the system dead volume.
RTs for solvent in each run are provided in Figure S1 in SI.To estimate the RT of each analyte at 298K, the linear relationship between ln(t 0 /t a , where t 0 ¼ 60s) and 1/T was utilized.All plots were characterized by correlation coefficients (Pearson's R > 0.99).Meanwhile, the correlation of the vapor pressure (ln(p/p 0 ), p 0 ¼ 101325 Pa) with the retention time (ln(t 0 /t a )) of the reference standards at temperature T ¼ 298K was observed to be linear.Thus, by using the slope and intercept of ln(p/p 0 ) and ln(t 0 /t a ), and the values of ln(t 0 /t a ) of the target compounds at the same temperature, their vapor pressures can be determined.

Vapor pressure reported in the literature
The compounds listed in Table 2 were used as vapor pressure standards.The Antoine equation (Thomson 1945), which is a semi-empirical equation that describes the relationship between vapor pressure and temperature, is shown below; A, B, and C are chemical-specific constants (Table 2).

Estimation of vapor pressure by the EPI Suite
The  (Mackay et al. 1982).The modified Grain method is better suited for estimating solid analytes because it can use an experimental melting point to adjust the vapor pressure from a supercooled liquid to a solid, whereas, for our target compounds, which are mostly liquid, the mean of the Grain method and the Antoine method is the recommended model (US EPA 2012).

Estimation of volatility by 2D-VBS
The volatilities of our target analytes were estimated based on their carbon and oxygen contents (Donahue et al. 2011).The functional groups of target analytes are mainly hydroxyl (-OH), carbonyl (-C ¼ O), alkoxy carbonyl, or carboxyl (-C(O)OH-) groups.Therefore, the saturation mass concentrations of those OOMs at room temperature (300 K) were estimated using the equation shown below: where nC, nO, and nN are the numbers of carbon, oxygen, and nitrogen in each molecule, respectively, and the carbon-carbon interaction term (bC) ¼ 0.475, oxygen-oxygen interaction term (bO) ¼ 2.3, and carbon-oxygen nonideality (bCO) ¼ −0.3, which are empirical values established in prior studies (Donahue et al. 2011).The final volatility was adjusted to 298K, and the details are presented in SI.

Case study
To evaluate the capability of using estimated vapor pressure or volatility to predict the changes in the chemical composition of e-cigarette aerosols through gas-particle partitioning, a case study was conducted using a Puff Bar Grape TM to generate e-cigarette aerosols in a 2-m 3 fluorinated ethylene propylene film chamber.The composition of e-liquids for Puff Bar Grape TM has been previously reported (Omaiye et al. 2022), which is mainly composed of PG/VG (50/50), nicotine, coolants, and flavor chemicals (e.g., ethyl maltol, menthol, and methyl anthranilate).The chamber was filled with zero air prior to each experiment, with temperature ranging from 18 to 22 � C and relative humidity set around 5%.Twenty puffs of Puff Bar Grape e-cigarette aerosols were injected into the chamber using an e-cigarette puffing machine (CSM-eSTEP, CH Technologies, Inc., Westwood, NJ, USA) with a flow rate pressure of 20 psi.The puff cycle followed the CORESTA recommended puffing topography (Tayyarah 2015), with a puff period of 3 s, a puff interval of 30 s, and around 55 mL of puff volume during each cycle under a temperature is around 116.6 � C (Ranpara et al. 2023).Both online and offline measurements were performed to characterize the changes in aerosol composition and size distribution throughout the experimental course.The particle concentrations and size distributions of e-cigarette aerosols were monitored by a scanning electrical mobility spectrometer (SEMS, Brechtel Manufacturing Inc.).A total of six aerosol samples were collected on the polytetrafluoroethylene membrane filters (Zefluor, Pall Laboratory, 25 mm, 1.0 lm pore size) at a flow rate of 10 L min −1 for 25 mins at three different time points during the experimental time course to capture the evolution of aerosol composition: (1) at the beginning, (2) 1 h post-injection, and (3) 2 h post-injection.It should be noted that the equilibrated e-cigarette aerosols were not directly sampled from the output of the e-cigarette device, but directly sampled from the chamber without a dilution or makeup airflow.The aerosol sampling volume was 0.25 m 3 (10 LPM � 25 mins ¼ 250 L ¼ 0.25 m 3 ).Considering the collapsible nature and the size of our FEP chamber (2 m 3 ), it was anticipated that the dilution effects during aerosol sampling would be minimal.
Offline techniques, including GC/EI-MS and an iodide-adduct time-of-flight chemical ion mass spectrometer coupled with a Filter Inlet for Gases and AEROsols system (FIGAERO-ToF-CIMS, Aerodyne Research Inc.), were used to analyze the molecular composition of collected aerosol samples.Three filter samples were extracted with 10 mL of methanol in an ultrasonic bath for 60 mins.The methanol extracts were then evaporated to dryness under a gentle stream of nitrogen.The dried residue was re-dissolved in 100 mL of methanol and injected into the GC-MS for analysis.The other three filters were directly analyzed with FIGAERO-ToF-CIMS.The temperature program for FIGAERO thermal desorption was set as follows: (1) temperature ramping from room temperature (�25 � C) to 200 � C in 15 min (11.6 � C min −1 ), (2) a 10-min soaking period (200 � C) to allow signals to return to background levels, and (3) a cooling period to decrease the temperature from 200 � C to 25 � C in 10 min.Other details about the instrumental setups and operation conditions for GC/EI-MS and FIGAERO-ToF-CIMS have been described in previous studies (Canchola et al. 2022b;Chen et al. 2019Chen et al. , 2022)).

Data analysis
The NIST 2008 mass spectral database was used in the identification of analytes in the C-GC method and e-cig aerosols in the case study.Compounds with a probability >50% and match factor scores >800 were considered good matches.The CIMS data were analyzed using the Tofware software package (v3.2.5) developed by Tofwerk and Aerodyne Research Inc. Pearson's correlation coefficients were calculated to determine the correlation between each method.

Estimated vapor pressure under 298K
A series of C-GC experiments were performed to determine the retention time of all analytes under different temperatures.Average values of ln(t 0 /t a ) for duplicate runs were first measured as a function of temperature in a T/K ¼ 30-40 range at T/K ¼ 5 intervals.Figure S2 shows the adjusted retention time (t a ) of each e-cigarette vaping aerosol under different temperatures, and values of t 0 /t a were averaged and logarithmic as ln((t 0 /t a ) avg ).The linear regression relationship of each analyte is well-established, with a R over 0.99, and the retention time of each analyte under 298 K was estimated from the resulting slopes and intercepts.The vapor pressures of reference compounds under 298 K from EPI and the temperature-dependent vapor pressure calculated from the literature are shown in Figure 2. The vapor pressures under 298 K are very close to these two approaches (Figure 2a).Therefore, the vapor pressure of analytes was calculated from the linear equation (Equation 8) between the EPI Suite-derived vapor pressure and the retention time of reference compounds (Figure 2b).

Comparison of C-GC and EPI Suite estimated vapor pressures
To further evaluate the accuracy of the calculated vapor pressure from the C-GC method, values of vapor pressure were predicted by the MPBPWIN program within the EPI Suite TM (version 4.11) (US EPA 2012).The vapor pressure estimated from our C-GC measurements (ln(p/p 0 )) and the predicted values using the EPI Suite (ln(p lit /p 0 )) are shown in Figure 3.
In Figure 3, ln(p/p 0 ) obtained from the C-GC method exhibits a strong correlation with ln(p lit /p 0 ) predicted by the EPI Suite.The R value between these two variables is 0.84, indicating the applicability of the C-GC method to e-cigarette vaping aerosols.Meanwhile, the analytes were separated into different groups based on their functional groups, as shown in Figure 3.For more volatile analytes, such as hydrocarbons and alcohols, there is a higher level of agreement in vapor pressure values between the two methods, as indicated by their alignment along the 1:1 line.In contrast, the C-GC method tends to overestimate the estimated values for less volatile compounds, like carbonyls and alkoxy carbonyls with lower vapor pressure, deviating from the 1:1 line to a relatively greater extent.The variation in estimated vapor pressure of less volatile analytes between these two approaches can be interpreted in two ways: (1) underestimation of vapor pressure in the EPI Suite, or (2) overestimation of the calculated vapor pressure by the C-GC method.For the first possibility, the model results are based on measured vapor pressure at a range of temperatures; for those analytes with relatively low vapor pressure and higher boiling temperatures, estimated vapor pressures under 298 K would show greater deviations.In addition, in this model, MW is a more important predictive variable than the chemical structure, which plays a crucial role in molecular interactions.As a result, the vapor pressure of compounds containing carbonyl groups may be  underestimated more than that of hydrocarbons or alcohols.While in the C-GC method, the estimated vapor pressure is based on the retention time of each targeted species, which is influenced by both their MW and structures; hence, the deviation in the C-GC method would be relatively lower than that in the estimated model to some extent.For the second possibility, a previous study pointed out that the GC-RT-related method overestimates p 298 for polar compounds, which is mainly caused by the greater variation of infinite dilution activity coefficients (c) among polar compounds than nonpolar compounds (Hinckley et al. 1990).Therefore, it is important to take into consideration the chemical nature of the targeted species when using the GC-based method for accurate results.
From above, the good correlation and consistency between two approaches in estimating vapor pressure imply that the C-GC method is applicable to e-cigarette aerosol constituents.Because the major components of vaping aerosols are polar compounds, including carbonyls and alkoxy carbonyls, the C-GC method can help reduce the discrepancy in the underestimation of polar compounds by using modeling estimates.

Comparison of GC-RT-related vapor pressure and VBS-related volatility
In this work, the volatility (i.e., saturation mass concentration) of e-cigarette aerosol constituents was also estimated using the 2D-VBS method to verify both calculated and estimated vapor pressure from different methods.The volatility of those aerosols can help determine their partitioning between gas and particle phases and is the key factor influencing their atmospheric lifetimes and contributions to secondary organic aerosol (SOA) formation.Instead of direct emission, SOA is formed from chemical processing of VOCs, either in the gas phase or in the particle phase (Kroll et al. 2015;Kleindienst et al. 2012;Shakya and Griffin 2010), which accounts for 60% of OA in a global scale (Hallquist et al. 2009;Kanakidou et al. 2005).
The 30 selected analytes mainly contain 1-3 oxygen atoms, which fall in the bins of SVOCs and IVOCs (Figure 4).Considering the oxygen contents of molecules and the effects of functional groups on volatility, hydrocarbons and alcohols mainly contain 0-1 oxygen and contribute to IVOCs, while carbonyls and alkoxy carbonyls are composed of both IVOCs and SVOCs with 1-3 oxygen atoms.The volatility of the e-cigarette constituents in fresh vaping emissions is relatively high, making it difficult to condense unless OA loadings are large, in which case absorptive partitioning can promote SVOC condensation (Williams et al. 2010;Schauer et al. 1999).However, in recent studies, IVOCs have been proposed to be an important source of SOA, and the estimated SOA from IVOCs is even 5 times that produced from single-ring aromatics, the dominant anthropogenic SOA precursors (Zhao et al. 2014).Moreover, those I/SVOCs with carbonyl functional groups, like flavorings and cooking emissions, are considered the main contributors to SOA formation (Yu et al. 2022).Therefore, the potential contribution of fresh vaping to SOA formation cannot be ignored, and the environmental aging processes may transform e-cigarette fresh vaping aerosols to become more oxygenated and enrich their contribution to SOA formation.
The relationship between volatility and vapor pressure as determined by two different methods is shown in Figure S3.The R obtained from the EPI suite estimates with volatility and GC calculators with volatility are 0.85 and 0.82, respectively.Correlations between volatility and vapor pressure were separated into three groups based on their functional groups, which are plotted in Figure 5.An opposite trend was observed between oxygen number and both volatility and vapor pressure.In both the EPI model and the C-GC calculation, it is interesting to note that the correlations between volatility and vapor pressure are stronger for substances containing C ¼ O groups.The R values for hydrocarbons/alcohols in the EPI Suite model and the C-GC method are 0.72 and 0.58, respectively.Furthermore, the R values range between 0.78 and 0.96 for carbonyls, while alkoxy carbonyls range between 0.85 and 0.92.The correlation between volatility derived from the VBS method and vapor pressure calculated or estimated from the two approaches is relatively weak among less oxidized compounds.Because this 2D-VBS equation is primarily used to estimate the volatility of oxidized molecules in the atmosphere, i.e., OAs and OOMs, the estimation of unoxidized or less oxidized compounds in the 2D-VBS parameterization is mainly influenced by their carbon numbers, resulting in a certain degree of error in the volatility estimation of those compounds (Donahue et al. 2011).Thus, its applicability to alcohols and hydrocarbons is limited.Meanwhile, the structures are overlooked (i.e., they are not parameterized) in both the 2D-VBS method and the EPI Suite model, which may explain why these two approaches have a relatively stronger correlation than the 2D-VBS and C-GC methods.

Gas-particle partitioning of e-cigarette aerosol constituents
The case study sought to investigate how the vapor pressure and volatility of e-cigarette aerosol constituents modulate the aerosol composition through gas-particle partitioning.The fresh e-cigarette aerosols contain inhalable particles (i.e., liquid droplets) in emissions along with volatile chemicals.Upon injecting the fresh e-cigarette aerosols into the 2-m 3 FEP chamber filled with zero air, a series of aerosol samples were collected using three separate filters at distinct time intervals.Since there were no oxidants in the chamber, the chemical process, such as oxidation or fragmentation, had no effect on changing the chemical composition of aerosols following injection.
The subsequent GC/EI-MS analysis allowed for the identification and quantification of six target compounds across a range of vapor pressures, which were meticulously chosen to facilitate a comparison of their relative abundance after a 2-h partitioning period.These findings are presented in Figure 6.Within Figure 6a-c, a consistent decrease in the proportions of three compounds with relatively high vapor pressures, including benzaldehyde, benzyl alcohol, and methyl anthranilate, was evident throughout the 2-h partitioning duration.It is important to note that the signals were initially standardized based on the aerosol mass concentrations measured by SEMS, followed by a further normalization using the highest signal recorded among the three distinct time points.In the case of r-dodecalactone, c-decalactone, and maltol, which have lower vapor pressures (Figure 6 df), the opposite trend was observed.These findings suggest that compounds with lower vapor pressures tend to remain in the aerosol phase during the partitioning process, despite the possibility of slight underestimation due to potential wall losses, and that vapor pressure can be used to predict changes in the chemical composition of e-cigarette aerosols.The FIGAERO-ToF-CIMS (Figures 6g and h) also showed similar trends that were linked to volatility.During the 2-h partitioning period, the concentrations of compounds with higher volatility and comprising 3 oxygen atoms (C 4 H 8 O 3 or C 4 H 10 O 3 ) consistently decreased.On the other hand, compounds with 4-6 oxygen atoms (C 4 H 6 O 4 , C 5 H 10 O 5 , and C 6 H 10 O 6 ) exhibited an incremental rise in their proportions during the partitioning process.This suggests that volatility predicted by the bulk aerosol composition is also useful to predict the contribution of compounds to aerosol formation, even if the molecular structures of individual constituents are not fully elucidated.Overall, these observations offer crucial insights into the interplay between vapor pressure, volatility, and the partitioning of compounds during the formation of aerosols from vaping emissions.Estimating vapor pressure with the C-GC approach and volatility with the 2D-VBS method can facilitate a better understanding of the contributions of compounds to aerosol formation and their impact on indoor air quality.Our case study results also support our conjecture: species with lower vapor pressure have a higher contribution after long-term exposure.Therefore, for the dominant species with higher vapor pressure in fresh e-cig aerosols, the influence on primary users is more significant, while the species dominating after vaping for 1 and 2 h may be correlated with the effects of secondhand exposures to bystanders.

Potential limitations
While the application of these methods can facilitate the estimation of physiochemical properties compared to conventional direct measurements, some limitations of this study should be noted.First, the empirical coefficients determined in previous 2D-VBS studies were classified mainly based on the sources of VOC precursors: anthropogenic and biogenic.There are two sets of empirical coefficients available: one equation is applied to estimate the volatility of anthropogenic oxidation products, which mainly contain carbonyl groups and hydroxyl groups, while another equation considers the appearance of hydroperoxyl groups in most biogenic compounds, such as oxidation products of terpenes.In this work, our target analytes in e-cigarette aerosols mainly contain hydroxyl and carbonyl functional groups.Therefore, the empirical coefficients for anthropogenic products were used to estimate the volatility.However, because e-cigarette aerosol formation mechanisms differ from those of aerosols derived from the atmospheric oxidation of VOC precursors, using 2D-VBS to estimate the volatility of compounds formed during vaping processes may result in a relatively higher variation in actual volatility.Second, the GC-based method is influenced to some extent by interactions between the stationary phase of the GC column and the target compounds.Although, when compared to traditional GC-RT studies, the C-GC method weighs the enthalpy of analyte transfer from the column to the vapor more than the reference compounds, the column selection still has some influence on the calculated vapor pressure, especially for vaping aerosol constituents with complex functional groups (Gobble and Chickos 2015).Furthermore, the chemical composition could directly influence the gas-particle coefficient and the partitioning efficiency.In addition, the loss of e-cigarette aerosols to the chamber wall may introduce some uncertainties to the final fraction of each compound.Given that particle wall loss rates are known to be size-and time-dependent (Wang et al. 2018), it is unclear if certain e-cigarette aerosol constituents may preferentially condense on either smaller or larger particles.Furthermore, the gas-particle partitioning coefficient can be influenced by several factors, such as the aerosol composition and the activity coefficient of compounds inside aerosols (Pankow 1994(Pankow , 2007)).Therefore, the vapor pressure alone is insufficient to fully represent the actual partitioning process that occurs during vaping.Lastly, because our major focus was to investigate the influence of volatility and vapor pressure on partitioning and their impacts on the change of e-cigarette aerosol composition over time in our case study, we did not vary the environmental conditions.We fully acknowledge that further research is required to investigate the possible influence of different parameters in the tested conditions.

Conclusion
Taken together, the vapor pressures measured by the C-GC method showed a strong correlation with the estimated value from the EPI Suite, with the R value of approximately 0.84.However, the two methods differed in their estimation of the vapor pressure of compounds with relatively low vapor pressure.The C-GC method tended to overestimate compared to the EPI Suite method.This deviation may be attributed to the different principles underlying the modeling approach and the GC method.The C-GC calculation considers the structure of analytes to a greater extent, whereas this aspect is overlooked in the modeling approach.Hence, the application of the C-GC method can assist in minimizing the underestimation discrepancy of polar compounds in e-cigarette vaping aerosols that may occur with modeling estimates, since the main constituents of vaping aerosols are polar compounds, such as carbonyls and alkoxy carbonyls.Furthermore, we utilized the 2D-VBS method to estimate the volatility of fresh e-cigarette aerosols.We compared this approach with two other methods and found the R value of over 0.8.The volatility and vapor pressure exhibited a consistent distribution, with higher values observed for hydrocarbons and alcohols in comparison to carbonyls.While the volatility of fresh vaping aerosols may be relatively high and fall within the IVOC and SVOC categories, the potential impact on SOA formation and indoor environments cannot be disregarded.
Overall, this study demonstrates the applicability of the C-GC and 2D-VBS methods for determining the vapor pressure and volatility distribution of e-cigarette aerosol constituents.Our case study revealed that lowvapor-pressure compounds (e.g., r-dodecalactone, c-decalactone, and maltol) become enriched in the e-cigarette aerosols within 2 h following vaping emissions, indicating that changes in the chemical composition of e-cigarette aerosols over time in the indoor environment may lead to differential exposures and potential health risks to primary users and bystanders, especially for secondhand exposure conditions.The findings of this study lay the groundwork for future investigations into the dynamic chemical composition, exposure profiles, and differential toxicity of vaping aerosols, which can aid in assessing the potential impact of these e-cigarette emissions on indoor environments and human health.

Supporting information
Temperature adjustment of volatility (Section S1); Retention times of target analytes at different temperatures (Table S1-S18); The retention time of methanol at a given temperature range (45 -195 � C) (Figure S1); The retention time ln(t0/ta) of target compounds under variable temperature (Figure S2); Correlations between the volatility derived by the VBS method and vapor pressure estimated from (a) the EPI suite model and (b) the correlation GC method; Figure S3).

Figure 1 .
Figure 1.The schematic of the whole procedure of C-GC method.

Figure 2 .
Figure 2. The vapor pressures of the target analytes: (a) plot of ln(p/p 0 ) versus T/K for vapor pressures reported in previous studies (line) and ln (p/p 0 ) under 298 K estimated from EPI Suite (marker); (b) the correlation between ln(p/p 0 ) and ln(t 0 /t a ).

Figure 3 .
Figure 3. Correlations between calculated vapor pressure from the C-GC method and estimated vapor pressure from the EPI Suite.

Figure 4 .
Figure 4.The volatility distributions of the target e-cigarette aerosol constituents are grouped into different functional groups.nC means number of carbon atoms, nO means number of oxygen atoms, IVOC means intermediate volatile organic compounds and SVOC means semi-volatile organic compounds.

Figure 5 .
Figure 5. Correlations between volatility and vapor pressure of hydrocarbons/alcohols (a and d, circle), carbonyls (b and e, triangle) and alkoxy carbonyls (d and f, square) among 2D-VBS and C-GC (a-c), and the EPI Suite approaches (d-f).

Table 1 .
The compound name, formula, molecular weight, CAS, boiling point, and vapor pressure of target compounds.
bVapor pressure at 25 � C from the EPI suite.

Table 2 .
Constants for vapor pressure calculation of standard compounds.