A novel method to quantify the emission and conversion of VOCs in the smoking of electronic cigarettes

An analytical technique was developed for the quantitation of volatile organic compounds (VOC) in three different forms of electronic cigarette (EC): solution, vapor, and aerosol. Through the application of the mass change tracking (MCT) approach, the consumed amount of the solution was measured to track the conversion of targets between the different phases. The concentration of aerosol plus vapor (A&V) decreased exponentially (559 to 129 g m−3) with increasing puff velocity (0.05 to 1 L min−1). A strong correlation existed between sampling volume and consumed solution mass (R2 = 0.9972 ± 0.0021 (n = 4)). In the EC solution, acetic acid was considerably high (25.8 μg mL−1), along with trace quantities of some VOCs (methyl ethyl ketone, toluene, propionic acid, and i-butyric acid: 0.24 ± 0.15 μg mL−1 (n = 4)). In the aerosol samples, many VOCs (n-butyraldehyde, n-butyl acetate, benzene, xylene, styrene, n-valeric acid, and n-hexanoic acid) were newly produced (138 ± 250 μg m−3). In general, the solution-to-aerosol (S/A) conversion was significant: e.g., 1,540% for i-butyric acid. The emission rates of all targets computed based on their mass in aerosol/ consumed solution (ng mL−1) were from 30.1 (p-xylene) to 398 (methyl ethyl ketone), while those of carboxyls were much higher from 166 (acetic acid) to 5,850 (i-butyric acid).

The use of electronic cigarettes (EC) is prevalent in Korea and other countries 1,2 . ECs are marketed as less harmful alternatives to tobacco smoking 3 . However, according to recent studies, many types of hazardous compounds (i.e., carcinogenic compounds) are emitted from EC smoke, some in significantly large quantities. For instance, some volatile organic compounds (VOC) like formaldehyde (FA) were generated in considerably large quantities upon puffing, as the EC solution was aerosolized via heating 4,5 . Therefore, information concerning the conversion of EC-related pollutants needs to be accurately assessed in order to better describe their potential damage to human health. To this end, studies that acquire or establish an analytical technique for the reliable quantitation of pollutants contained in EC solution, as well as those released directly to or generated newly by its smoking, should be prioritized.
In general, smoke samples of conventional tobacco cigarettes are analyzed using official methods such as those of the ISO and Canadian intense 6 . In the case of ECs, although some analysis methods have been proposed, they are poorly validated for general application to the quantitation of all different types of EC samples (i.e., liquid, vapor, and aerosol). Therefore, many researchers are investigating optimal methods for both the collection and analysis of EC samples. In general, EC refill solutions are analyzed using gas chromatography (GC) or liquid chromatography (LC) after the pretreatment of the EC solution (such as derivatization) or through dilution with an appropriate solvent(s) [7][8][9][10] . For EC vapor (or aerosol) samples, the analytical approaches that are comparable to the official method for conventional cigarettes have commonly been employed 11 . For example, EC aerosols can be absorbed on a solvent (or Scientific RepoRts | 5:16383 | DOI: 10.1038/srep16383 collected on a glass filter) and subjected to extraction in a certain medium. The extracted proportion is then analyzed using GC or LC systems 11 . However, there are large differences in the two cigarette types not only between the mechanisms of smoke generation (i.e., sensitized control of puff for EC vs. conventional cigarette), but also in the measurement methods used to determine their compositions. More specifically, in the case of ECs, the solvent effect must be considered, as the collected samples contain excessive quantities of propylene glycol (PG) and/or vegetable glycerin (VG). Unless treated properly, those solvents disrupt the detection of targets in the GC-based analysis. The direct application of conventional methods to EC smoke samples is unlikely to be feasible in many respects. Therefore, analytical methods should be developed to effectively accommodate the physicochemical characteristics of both the major and trace components of EC so that the generation mechanisms of those samples should be reflected properly in their quantitation.
The EC aerosols are generated from the EC refill solution in the electronically heated cartomizer (atomizer + cartridge). As the solvent of the EC liquid used primarily consists of PG and/or VG, it is important to consider two key factors: the solvent effect of these major components and the consumption rate of the EC solution. First, the presence of massive quantities of solvent (like PG and/or VG) in the EC solution disrupt the detection of target compounds, since they remain after thermal desorption 12,13 . In particular, the quantitative analysis of some target compounds with physicochemical properties similar to those of the solvent is hindered due to interference from chromatographic separation. Therefore, an analytical method that can effectively resolve this solvent effect (unlike tobacco cigarette) needs to be established. Second, the ECs can generate new compounds (e.g., formaldehyde, acetaldehyde, and acrolein) that did or did not exist in the original solution; they are generally produced via oxidation of the EC components (PG and VG) through heating 4,5 . The formation of those pollutants directly reflects the consumption of EC solution. Likewise, quantification of mass changes before and after the use of EC is more useful than only recording of the puffing conditions (e.g., puffing velocity, puffing duration, number of puffing, etc). This concept was thus defined as mass change tracking (MCT) approach. As a result, the mass balance between the amount of EC solution consumed by puffing and the amount of EC smoke caught up on the ST can be established to accurately assess mass transfer between different EC phases.
In the quantitative stage of this study, a total of 24 target VOCs in EC samples (either as aerosol phase or refill solution: refer to Table S1 in Supplementary Information. Note that acronyms are generally used for all target compounds throughout the text (e.g., benzene = B)) were treated identically not only during the sample collection (by a sorbent tube (ST)) but also during the instrumental detection (by the same GC-mass spectrometry (MS) system equipped with a thermal desorber (TD)) 14 . As this combination of ST-TD-GC-MS approach allowed us considerably reduce sampling volume (e.g., EC liquid <3 μ L and aerosol <10 mL), it helped us eliminate or effectively reduce the solvent effect between different sample types -between liquid and vapor phase samples. In addition, the solvent effect can be reduced further with the aid of purging 15 . The extracted ion chromatogram (EIC) and selected ion monitoring (SIM) modes of the MS system were applied to quantify the target VOCs at trace quantities 16 . The emission rates (ER) of the VOCs from the EC were calculated by simultaneously considering the consumption rate of the EC solution and the puff conditions of the EC smoke based on the optimal analytical approaches designed in this study.

Results and Discussion
Measurement of the sampling mass of the EC smoke (A&V) using the mass change tracking (MCT) method. In this study, the amount of EC solution consumed by puffing was measured based on the MCT approach, which allowed for the comparison of the consumed solution mass with the collected mass of the EC smoke on the ST. The difference between the two mass terms was almost negligible, at less than 3%. For this reason, the consumed amount of refill solution was used as a key index in the calculation of VOC emission rate from the EC device with varying puff conditions ( Table 1).
The V E concentration emitted from the EC was significantly low, at a value less than 1 g m −3 , regardless of the puff conditions (total sample volume (0.01 to 2 L) and puff velocity (0.05 to 1 L min −1 ). In addition, the total amount (mg) of V E collected on the sampler did not exhibit significant correlation with puff volume (L) (mean R 2 ± SD = 0.2872 ± 0.2773 (n = 4)). As a result, when air was swept through the EC device filled with EC solution (without puffing; no operation of EC device), the amount of vaporized EC (V E ) could be ignored in terms of mass. In contrast, the collection of A E was accompanied by considerable changes in the mass quantities at all of the different puff conditions (puff velocity and volume). The emission mass of the A E gradually increased with an increase in puff volume at a constant puff velocity (correlation (R 2 ) between A E mass vs. puff volume = mean 0.9972 ± 0.0021 (n = 4)). However, the concentrations of the A E samples decreased logarithmically with an increase in puff velocity ((1) A E concentration (puff velocity) = 559 g m −3 (0.05 L min −1 ) to 129 g m −3 (1 L min −1 ) and (2) R 2 = 0.9888) (Fig. 1). The solvent of the EC solution consists of PG and VG. Because the boiling points of PG and VG were relatively high (188.2 and 290 °C, respectively), their conversion rates (solution to A&V) did not increase proportionally with the increases in the puff velocity.
The results of the environmental EC samples (Exp. 1). In this study, a total of 22 target VOCs in three types of EC samples (the EC solution (S E ) and smoke (V E and A E )) were analyzed using the Scientific RepoRts | 5:16383 | DOI: 10.1038/srep16383 ST-based method. In addition, acetic acid (ACA) was also quantified because of its abundance in all three types of EC samples. The quantitation of ACA was conducted using a linear regression analysis between carbon number and response factor (RF, ng −1 ) of seven of the target carboxyl compounds [17][18][19] . In the S E analysis, five of the target VOCs (MEK, T, ACA, PPA, and IBA) were detected among all of the selected targets (Table S2). Except for ACA, the detected VOCs (n = 4) had low concentrations in the range of 0.059 (T) to 0.379 μ g mL −1 (IBA) (Eq. (1)).

[B] Emission trends of the V E and A E samples in terms of consumed amount of EC solution
Order Puff velocity (L min −1 )   However, ACA was considerably high at 25.8 μ g mL −1 . In the case of V E , the VOCs (n = 10) with relatively low molecular weights (or high volatility) were also (or newly) detected (BA, MEK, BuAc, B, T, p,m,o-X, S, and ACA). The average concentration of the 10 VOCs detected from the V E sample was 22.5 ± 51.8 μ g m −3 (Eqn (2)). In the V E sampling stage, the detection of light VOCs was expected due to their high volatility. In the A E sample, all of the target VOCs detected from the S E and V E were present in greatly enhanced values relative to the other sample types (S E and V E ) studied concurrently.
Therefore, the recovery of many of the VOCs in the S E sample commonly exceeded 100% (232% for MEK: 0.40 μ g mL −1 to 1,540% for IBA: 5.85 μ g mL −1 ) (Eqs. (3) and (4)) ( Fig. 2). As a result, we confirmed that the VOC concentrations increased, especially in the case of carboxyl compounds; the solution-to-aerosol (S/A) conversion rates in these situations were 637% (PPA), 1,540% (IBA), and 646% (ACA)). In addition, two carboxyl compounds (VLA and HXA) that were not detected in the S E were observed in the A E with values of 0.31 μ g mL −1 and 1.51 μ g mL −1 , respectively (Eq. (3)). The results of the V E analysis were compared with those of the A E based on the detected VOC mass (μ g) per sample volume of V E or A E (L) (Eqs. (2) and (5)).  The recoveries of the six aromatic compounds in the A E also increased considerably to 579% (o-X) −1,020% (B) (Eq. (6)): mean concentration ± SD = 57.5 ± 63.9 μ g m −3 (n = 6) (Eq. (5)) (Fig. 2). The recoveries of BA, MEK, and BuAc in the A E were 5,140%, 2,440%, and 1,300%, respectively, which were substantially larger than those of the V E (Eq. (6)). The carboxyl compounds, PPA, IBA, VLA, and HXA, were not detected from the V E but were detected in the A E (mean concentration ± SD = 1,310 ± 1,250 μ g m −3 (n = 4) (Eq. (5)). In the V E , ACA was measured at a moderately higher concentration of 169 μ g m −3 (Eq. (5)) compared to the other simultaneously detected VOCs. The concentration of ACA increased significantly to 8.72E + 4 μ g m −3 in the A E (Eq. (5)), more than 500 times higher than that in the V E .
Overall, the concentrations of most of the target VOCs increased considerably. It was striking to find that several of the carboxyl compounds were produced in large quantity through the conversion of S E to A E . In addition, the VOCs with a relatively low boiling point (e.g., BP < 150 °C) were easily vaporized from the EC device filled with S E when air was swept through the EC device without a puff. The changes in the VOC levels between the different sample phases can be examined thoroughly due to the availability of the data from these spiking samples. The analysis of the S S samples showed that the VOC recoveries were fairly variable depending on the functional groups present. For this comparison, the calibration results obtained by the L-WS were used as the basis for the recovery estimates. Accordingly, 13 of the target VOCs (except carboxyl (n = 7) and phenol compounds (n = 2)) had considerably lower recovery, with a mean of 38.7 ± 10.4% (n = 13) (Eq. (7)).
Detected VOC mass in S g Spiked VOC mass in S g 100 7 S S S This low recovery of the 13 VOCs might have reflected the significant interfering effects of the EC solvents (PG or VG) or the unstable conditions of their mixtures in these solvents. In contrast, a nearly full recovery was achieved for the six carboxyl and two phenol compounds (mean recovery ± SD (n = 8) = 98.4 ± 4.17%) (Eq. (7)). The vaporization efficiencies of the VOCs (from S S to V S ) were distinguished at least partially based on the physicochemical properties of each of the compounds. For example, the concentrations of the two carboxyl compounds (PPA and IBA) in the S S were higher than those of the six aromatic compounds (mean concentration of S S : (1) two carboxyls = 10.5 ± 0.22 μ g mL −1 and (2) six aromatics = 2.23 ± 0.22 μ g mL −1 ) (Eq. (1)). However, in the V S , the aromatic compounds (n = 6) had noticeably higher concentrations, with a mean value of 386 ± 60.7 μ g m −3 (n = 6), than those of the PPA and IBA with a mean value of 14.8 ± 0.17 μ g m −3 (Eq. (2)). When the S S was heated to produce the A S sample, the emission patterns of the VOCs were clearly distinguished between the relatively light VOCs (BA, IA, VA, MEK, MIBK, BuAc, i-BuAl, B, T, p,m,o-X, and S (n = 13)) and the heavy VOCs (PPA, IBA, IVA, VLA, HXA, HPA, o-C, and m-C (n = 8)). In the case of the eight heavy VOCs (carboxyl and phenol compounds), their concentrations were similar between the A S and S S , with a recovery of 96.6 ± 5.36% (n = 8) (calculated by Eq. (4)). The distribution of the heavy VOCs (n = 8) was comparable between the S S and A S . In contrast, the concentrations of the light VOCs (n = 12), except for the i-BuAl, increased when the S S was converted into A S via puffing (246% (MEK) to 1,070% (T) (Eqn (4))). Therefore, most of those light VOCs can be very effectively converted from S S to A S in comparison to the heavy VOCs, such as carboxyl and phenol.
Emission rates of the VOCs in the EC aerosol from the EC device. In this study, the emission rates of the VOCs from the EC were calculated in terms of 'VOC mass released per puff duration' in a direct connection to the consumption rate of the EC solution (S E ). First, the consumption amount of the S E was measured using the MCT approach at a puff velocity of 0.05 L min −1 , while the VOCs aerosolized from the S E were quantified using the ST/TD-GC-MS system. In addition, as the consumption rates of the S E were measured at varying puff velocities of 0.2, 0.5, and 1 L min −1 , the emission rates of the VOCs from the EC device with other puff flow rates (0.2, 0.5, and 1 L min −1 ) were also calculated using the results of the consumption trend of S E at each puff velocity ( At the puff flow rate of 0.05 L min −1 , six aromatic compounds and three VOCs (BA, MEK, and BuAc) had low mean emission rates of 3.64 ± 3.48 ng min −1 (n = 9) (Eq. (8)). In contrast, the emission rates of the five carboxyl compounds (PPA, IBA, VLA, HXA, and ACA) were relatively high (range of 7.89 (VLA) to 4,250 ng min −1 ( ACA)) (Eq. (8)). The emission of the A E sample from the EC device was evaluated depending on puff velocity (0.2, 0.5, and 1 L min −1 ) using the MCT approach. Based on these results, the emission rates of the VOCs from the EC device could be extrapolated at a puff velocity of 1 L min −1 , which was comparable to the recommended puff velocity used for testing conventional cigarettes 6 . The concentration levels of the VOCs in the EC aerosols measured by McAuley, et al. 11 were quite comparable to those in our study. That previous study measured the VOCs in the EC aerosols emitted from EC solutions (n = 4) with tobacco flavor. To collect the EC aerosol samples, the EC solutions were aerosolized using an EC device equipped with a single cigarette smoking machine. Then, the EC aerosols were collected on the ST via a polyethylene bag. According to McAuley, et al. 11 , the emission concentrations of the VOCs were 219 ± 160 μ g m −3 for BA, 180 ± 725 μ g m −3 for VA, 28.2 μ g m −3 for B, 244 ± 209 μ g m −3 for T, and 107 ± 26.5 μ g m −3 for p/m-X. It should be noted that our VOC results were 66.3 μ g m −3 for BA, 87.5 μ g m −3 for B, 176 μ g m −3 for T, 15.7 μ g m −3 for p-X, 26.5 μ g m −3 for m-X, 18.8 μ g m −3 for o-X, and 20.8 μ g m −3 for S. In the case of a conventional cigarette, the emission rates of BTSX (mass per cigarette) generally ranged from 85 (p/m-X) to 1,100 (T) μ g cigarette −1 20-25 . If it is assumed that it takes 20 puffs to consume a conventional cigarette, the average emission rate of the VOCs (mass per puff) was 16.3 ± 16.4 μ g puff −1 (n = 6). If these results were compared with the data in this study, the EC counterparts of this study were significantly lower, with a mean of 0.43 ± 0.48 ng puff −1 (Eq. (9)).

Conclusions
An electronic cigarette (EC) is a device designed to imitate tobacco cigarettes without combusting tobacco; therefore, the smoke (aerosol and vapor) is generated from the EC solution through electronic heating via a cartomizer. Therefore, it is important to assess the amount or exposure level of the hazardous

Order
Puff velocity (L min −1 ):  compounds generated by EC puffing. In this study, VOCs in three types of EC samples (solution, vapor, and aerosol) were collected on the ST and analyzed using the same TD-GC-MS system to allow a parallel comparison. In addition, the consumption rates of the EC solution (S E ) were evaluated with respect to puff conditions, so that the emission rates (VOC mass per puff duration) of the VOCs were compared between diverse criteria (per puff or per solution consumed).
In the analysis of the S E , MEK, T, PPA, and IBA were detected in low concentrations (mean 0.24 μ g mL −1 ), while ACA was observed at a relatively higher concentration of 25.8 μ g mL −1 . The seven VOCs with relatively high volatility (BA, BuAc, B, p-X, m-X, o-X, and S) were also detected from the EC vapor sample (V E ) without a puff. In the EC aerosol sample (A E ), all of the VOCs detected from the S E and V E were observed at much larger quantities. In addition, two carboxyl compounds (VLA and HXA), which were not detected from the S E and V E , were also produced. The concentrations of all of the VOCs detected from the A E sample were considerably increased compared to those of the S E (2.32 (MEK), by 15.4 times (IBA)), or the V E (5.79 (o-X), by 515 times (ACA)). If the A E sample was produced and inhaled at 1 L min −1 , then approximately 20 ng of benzene (B) for one minute puff duration can be inhaled. In addition, the carboxyl compounds, such as the PPA, IBA, VLA and HXA, were also inhaled with a relatively high concentration of approximately 300 ng for one minute puff duration.
In this study, we developed an analytical method for the accurate quantitation of all types of EC samples (solution, vapor, and aerosol) based on ST sampling and TD-GC-MS analysis. Using the quantitative results of the EC samples, the emission rates of VOCs from an EC device were calculated based on the MCT approach in direct association with the puff conditions. The versatile applicability of this proposed approach was determined to be sufficiently efficient to accurately assess the emission properties of VOCs from all different types of EC samples. In addition, this method was sufficiently effective enough to    estimate the emission factor through the direct quantitation of the diverse target analytes (transferred or converted) among the different EC phases with the aid of MCT approach.

Materials and Methods
Experimental outline. In this study, an experimental method for the accurate analysis of three types of EC samples (solution, vapor, and aerosol) was developed using the ST/TD-GC-MS system. The changes in concentration of target analytes in the EC samples were measured using the MCT approach with varying puff conditions of puff velocity, puff duration, and puff volume. As the consumption rate of the EC smoke was measured simultaneously at each of the smoke sampling stages, the VOC emission rates (or concentration) were presented in a number of ways using the detected mass of the target (in ng (or μ g)) against per puff (mass per puff (ng puff −1 ), per puff duration (ng min −1 ), per air (or gas) collected (ng m −3 (air)), or per solution consumed (ng mL −1 (solution)). Therefore, our approach allows for the parallel comparison of all of the different concepts and in associated unites.
The calibration and quality assurance (QA) data for the target VOCs were obtained by analyzing their liquid working standard with the ST/TD-GC-MS system (Table S3). The three types of EC samples (solution, vapor, and aerosol) were then quantified using these calibration results. For this experiment, we used the EC device (Korea) and an EC solution without nicotine (Korea) ( Table S4). Our experiments were then conducted in two different stages (Table 3). In Exp. 1, three different forms of the environmental EC samples (subscript of E) were collected and analyzed: (1) (refill) solution (S E ), (2) vapor (V E ) collected during the interpuff period (without puffing), and (3) aerosol (A E ) plus V E collected simultaneously due to puffing (It should be noted that aerosol alone cannot be sampled as both aerosol and vapor are released during puffing. The aerosol data were hence obtained by subtracting the vapor fraction from the A&V concentration). In Exp. 2, a spiked EC solution was prepared by adding the liquid working standard of the target VOCs into the S E . As all of the types of EC samples generally contained little or none of most of the target VOCs, their partitioning patterns were analyzed in detail using these spiked samples (with a subscript of S). The spiked EC solution (S S ) and spiked A&V (A S and V S ) samples were then analyzed using the system introduced above. Based on the analysis of the spiked samples, we sought explanations for the generation and/or conversion patterns of the various targets in EC solution and in EC smoke. A detailed description of the liquid working standard preparation for quantitation of EC samples (Table S5), instrumental system (Table S6), and the sampling approaches of three types of EC samples ( Figure S1) is provide in Supplementary Information (SI). In addition, the representative chromatograms of carboxyl compounds with different MS detection mods (total ion chromatogram (TIC) and selected the spectrum (SIM)) are shown in Figure S2.