Gas/Particle Partitioning Constants of Nicotine, Selected Toxicants, and Flavor Chemicals in Solutions of 50/50 Propylene Glycol/Glycerol As Used in Electronic Cigarettes

Abstract

For an electronic cigarette (e-cigarette) aerosol with known total particulate matter concentration (TPM, μg/m³), predictions of the fractions of some compound i in the gas and particle phases (f_g,i and f_p,i) at equilibrium can be made based on K_p,i (m³/μg), the compound-dependent gas/particle partitioning equilibrium constant. f_g,i and f_p,i affect the modes and locations of deposition in the respiratory tract. K_p,i depends inversely on (1) the pure compound liquid vapor pressure ($p_{\mathrm{L},i}^{\mathrm{o}}$), (2) mole fraction activity coefficient (ζ_i) in the absorbing liquid, and (3) mean molecular weight of the absorbing liquid ($\overline{\text{MW}}$). K_p,i values were measured at 20 °C for 32 compounds as spiked into simulated e-cigarette liquids prepared as 50/50 mixtures (by weight) of propylene glycol (PG) and glycerol (GL). K_p,i values at 37 °C were estimated. The 32 compounds were nicotine (in free-base form), seven toxicants (propanal, acetone, hydroxyacetone, benzene, toluene, p-xylene, and ethylbenzene), and 24 flavor chemicals (2,3-pentanedione (“acetyl propionyl”), isobutyl acetate, ethyl butyrate, butyl butyrate, isoamyl acetate, 2,3-dimethylpyrazine, 3-methyl-1-butanol, limonene, 2,3,5-trimethylpyrazine, p-cymene, benzaldehyde, (Z)-3-hexen-1-ol, menthol, 2-acetylpyrrole, benzyl alcohol, methyl salicylate, cinnamaldehyde, methyl anthranilate, (+)-aromadendrene, cinnamyl alcohol, methyl cinnamate, maltol, ethyl maltol, and coumarin). The measured log K_p,i values were found to be generally correlated with literature values of log $p_{\mathrm{L},i}^{\mathrm{o}}$; the scatter is caused by variation in ζ_i between ~1 and ~1000. K_p measurements were attempted, but values were not reported for acetaldehyde, 2,3-butanedione (diacetyl), vanillin, and ethyl vanillin. Acetaldehyde was found to form significant amounts of its cyclic trimer and cyclic tetramer; for diacetyl, the evidence suggested significant amounts of reaction products, possibly hemiketals and ketals with PG/GL, and for vanillin and ethyl vanillin, the K_p values are large and accordingly more difficult to measure. f_g values are calculated using a range of K_p and TPM values.

Gaphical Abstract

1. INTRODUCTION

Aerosols created by electronic cigarettes (e-cigarettes), conventional cigarettes, cigars, cannabis cigarettes, and “heat-not-burn” tobacco and cannabis products are all collections of small liquid droplets surrounded by a gas phase. Each chemical i in the aerosol (e.g., formaldehyde, acetaldehyde, acrolein, benzene, nicotine, etc.) that possesses some volatility (i.e., a nonzero vapor pressure) will partition between the gas and the droplet phases.¹ The compound-dependent equilibrium gas/particle coefficient K_p,i (m³/μg) is defined as²

$$K_{\mathrm{p},i}({\mathrm{m}}^3∕\mu \mathrm{g})=\frac{c_{\mathrm{p},i}(\mu \mathrm{g}∕\mu \mathrm{g})}{c_{\mathrm{g},i}(\mu \mathrm{g}∕{\mathrm{m}}^3)}$$ (1)

where c_p,i (μg/μg) is the concentration in the particle phase and c_g,i (μg/m³) is the concentration in the gas phase. Even under nonequilibrium conditions, the partition coefficient still plays a fundamental role as it determines the liquid- and air-side concentration gradients that drive mass transfer and equilibration (e.g., see Liss and Slater³). Here, we focus on K_p,i values; their use in mass transfer calculations for inhaled aerosols that are highly physically dynamic (coagulation, deposition, etc.) is far beyond the scope of this work.

For gas/liquid partitioning²

$$K_{\mathrm{p},i}({\mathrm{m}}^3∕\mu \mathrm{g})=\frac{RT}{{10}^6\overline{\text{MW}}{\zeta }_{i}p{}_{\mathrm{L},i}^{\mathrm{o}}}$$ (2)

where R is the gas constant (8.2 × 10⁻⁵ m³ atm mol⁻¹ K⁻¹), T is temperature (K), $\overline{\text{MW}}$ (g/mol) is the mol-average molecular weight of the absorbing liquid phase computed for a sample of that phase, namely [total g]/[total mols]), ζ_i (dimensionless, always > 0) is the mol-fraction-scale activity coefficient in the liquid (L) phase, and $p_{\mathrm{L},i}^{\mathrm{o}}$ is the vapor pressure (atm) of pure liquid i at temperature T. For a compound that is a solid (when pure) at the temperature of interest, $p_{\mathrm{L},i}^{\mathrm{o}}$ is the subcooled liquid vapor pressure of the pure compound, because in a liquid aerosol droplet, i is in a liquid state. K_p,i values can be predicted using eq 2 with known/estimated values of $p_{\mathrm{L},i}^{\mathrm{o}}$, ζ_i, and $\overline{\text{MW}}$.² Time scales for reaching gas/liquid equilibrium in aerosols have been studied.⁴

K_p,i values together with the total particulate matter (TPM, μ/m³) concentration for the aerosol determine the equilibrium fraction of each compound i in the gas (f_g,i) and particle (f_p,i) phases,^1–5 which affect deposition in the respiratory tract (RT).⁵ For the overall aerosol volume, assuming that the volume of the suspended TPM phase is negligible, then c_g,i gives a good approximation of the gas-associated mass concentration (μg/m³ of total aerosol). The mass concentration (μg/m³ of total aerosol) in the TPM is K_p,iTPM. Thus, ^1,2,5

$$f_{\mathrm{g},i}=\frac{c_{\mathrm{g},i}}{c_{\mathrm{g},i}+c_{\mathrm{p},i}\text{TPM}}=\frac{1}{1+K_{\mathrm{p},i}\text{TPM}}$$ (3)

$$f_{\mathrm{p},i}=\frac{c_{\mathrm{p},i}\text{TPM}}{c_{\mathrm{g},i}+c_{\mathrm{p},i}\text{TPM}}=\frac{K_{\mathrm{p},i}\text{TPM}}{1+K_{\mathrm{p},i}\text{TPM}}$$ (4)

$$f_{\mathrm{g},i}+f_{\mathrm{p},i}=1$$ (5)

$$\frac{\text{mass in gas phase}}{\text{mass in PM}}=\frac{c_{\mathrm{g},i}}{c_{\mathrm{p},i}\text{TPM}}=\frac{1}{K_{\mathrm{p},i}\text{TPM}}$$ (6)

$$\frac{\text{mass in PM}}{\text{mass in gas phase}}=\frac{c_{\mathrm{p},i}\text{TPM}}{c_{\mathrm{g},i}}=K_{\mathrm{p},i}\text{TPM}$$ (7)

Being a pure-chemical physical property, $p_{\mathrm{L},i}^{\mathrm{o}}$ is known for many compounds. Propylene glycol (PG) and glycerol (GL) are the dominant chemicals in the “e-liquids” used in electronic cigarettes and thus are the dominant chemicals in the aerosol droplets formed in the vaping process. ζ_i values in PG/GL mixtures are generally not known. The ζ_i can be near unity when i is polar, i.e., like PG and GL. For increasingly nonpolar i, ζ_i will become increasingly larger than 1. Regarding MW, the MW values for PG and GL are 76.09 and 92.09 g/mol, respectively; the densities at 20 °C are 1.04 and 1.26 g/mL, respectively. Thus, for an e-liquid that is 50/50 by volume PG/GL, then $\overline{\text{MW}}\approx 84\mathrm{g}∕\text{mol}$; for an e-liquid that is 50/50 by weight PG/GL, then $\overline{\text{MW}}=83.3\mathrm{g}∕\text{mol}$. Nonzero concentrations of nicotine, flavor chemicals, other additives such as acids and water will all affect $\overline{\text{MW}}$. For a 50/50 by volume mixture of PG/GL that has been spiked with nicotine to a level of 24 mg/mL, then $\overline{\text{MW}}\approx 85\mathrm{g}∕\text{mol}$. For a 50/50 by volume mixture of PG/GL that has absorbed 10% by volume water (as from the laboratory environment or the interior of the RT), then $\overline{\text{MW}}\approx 64\mathrm{g}∕\text{mol}$; by itself, such a change will tend to increase K_p values by ~30% and will in most cases be counteracted by increases in ζ_i Here, we measured K_p values for 32 compounds as spiked into 50/50 by weight PG/GL with $\overline{\text{MW}}$ remaining at ~83.3 g/mol. The 32 compounds included nicotine, seven toxicant compounds, and 24 flavor chemicals. Though not studied directly here, PG, GL, and water will also follow eqs 1–7).

2. MATERIALS AND METHODS

Vapor Pressure.

Values of log $p_{\mathrm{L},i}^{\mathrm{o}}$ at 20 and 37 °C (human body temperature) for 32 compounds studied here are given in Table 1. Twenty-four of the 32 compounds are liquids (L) at 20 °C; 28 are liquids at 37 °C. For those that are liquids at 20 °C and/or 37 °C (T = 293.15 and 310.15 K), log $p_{\mathrm{L},i}^{\mathrm{o}}$ was based on published temperature-dependent functionalities. The same applies analogously to log $p_{\mathrm{S},i}^{\mathrm{o}}$ for compounds that are solids (S) at 37 °C and/or 20 °C; to obtain $p_{\mathrm{L},i}^{\mathrm{o}}$ at 37 and/or 20 °C, the enthalpy of fusion ΔH_fus,i (kJ/mol) was assumed constant in the application of¹¹

$$\text{ln}{p}_{\mathrm{L},i}^{\mathrm{o}}=\text{ln}{p}_{\mathrm{S},i}^{\mathrm{o}}+\frac{\Delta H_{\text{fus},i}}{R{T}_{\mathrm{m},i}}\left(\frac{T_{\mathrm{m},i}}{T}-1\right)$$ (8)

where T_m,i (K) = melting point of i at 1 atm (because of the general steepness of S/L phase boundaries in P vs T phase diagrams, T_m,i is essentially the same as the triple point temperature) and R = 0.00831 kJ mol⁻¹ K^-1. Because ΔH_fus,i > 0, when T < T_nm,i then $p_{\mathrm{L},i}^{\mathrm{o}}>p_{\mathrm{S},i}^{\mathrm{o}}$. The values of ΔH_fus,i used with eq 8 are given in Table 1.

Table 1.

Log K_p and Related Values for 32 Compounds^a

compound	group	abbr.	melting point (°C)	log $p_{\mathrm{L}}^{\mathrm{o}}$ at 20 °C (atm)	log Kp (av, N = 3)b(m3/μg, 20 °C) (measured)	SD of log Kp (N = 3)b (m3/μg, 20 °C) (measured)	ζ (20 °C)	log $p_{\mathrm{L}}^{\mathrm{o}}$ at 37 °C (atm)a	log [ζ $p_{\mathrm{L}}^{\mathrm{o}}$] at 37 °C	τ20→37	log Kp(m3/μg, 37 °C) (extrapolated)
benzene	1	Ben	5.5	−1.00c	−10.02	0.09	30	−0.67	0.81	0.49	−10.32
toluene	1	Tol	−95	−1.54c	−9.81	0.08	65	−1.17	0.64	0.45	−10.15
isobutyl acetate	1	IsoBAc	−99	−1.71c	−9.76	0.21	84	−1.32	0.60	0.44	−10.12
acetone*	1	Act	−95	−0.61c	−9.75	0.07	6.6	−0.31	0.52	0.52	−10.03
ethylbenzene	1	EtBen	−95	−2.03c	−9.64	0.08	130	−1.62	0.51	0.41	−10.03
p-xylene	1	pXyl	13	−2.07c	−9.63	0.09	150	−1.65	0.51	0.40	−10.03
limonene	1	Lim	−74	−2.85d	−9.59	0.06	780	−2.37	0.53	0.35	−10.04
ethyl butyrate	1	EtBut	−93	−1.76c	−9.50	0.08	53	−1.37	0.35	0.43	−9.87
p-cymene	1	pCym	−68	−2.87c	−9.29	0.11	410	−2.38	0.24	0.34	−9.75
isoamyl acetate	3	IsoAmAc	−78	−2.23c	−9.23	0.10	84	−1.79	0.14	0.38	−9.65
propanal*	2	Pad	−81	−0.47e	−9.15	0.01	1.2	−0.17	−0.09	0.54	−9.43
butyl butyrate	1	BuBut	−92	−2.76f	−8.96	0.04	150	−2.32	−0.14	0.39	−9.37
2,3-pentanedione*	2	AcPr	−52	−1.70g	−8.80	0.02	9.2	−1.27	−0.31	0.40	−9.20
hydroxyacetone*	1	Hact	−17	−2.38h	−8.00	0.26	7.0	−1.97	−1.13	0.41	−8.39
3-methyl-1-butanol	3	MeBuol	−117	−2.58e	−7.91	0.25	8.9	−2.00	−1.05	0.28	−8.46
(+)-aromadendrene	3	Arom	6.8	−4.66i	−7.86	NA	960	−4.20	−1.22	0.37	−8.30
benzaldehyde*	3	Bzad	−26	−2.91c	−7.66	NA	11	−2.45	−1.43	0.37	−8.08
2,3-dimethylpyrazine	3	DiMePy	12	−2.50j	−7.50	0.06	2.9	−1.99	−1.52	0.32	−7.99
(Z)-3-hexen-1-ol	3	Z3H1ol	−61	−3.00k	−7.23	0.09	5.0	−2.56	−1.86	0.38	−7.66
methyl salicylate	3	MeSal	−7.5	−4.16c	−7.11	0.19	53	−3.50	−1.77	0.23	−7.74
2,3,5-trimethylpyrazine	3	TriMePy	<−10	−2.81j	−7.07	NA	2.2	−2.28	−1.94	0.31	−7.58
menthol	3	Men	37	−4.05c	−6.55	0.07	11	−3.46	−2.41	0.27	−7.11
methyl cinnamate	3	MeCinn	36	−4.67c	−6.27	0.02	25	−4.06	−2.66	0.26	−6.86
cinnamaldehyde*	3	Cad	−7.5	−4.64c	−6.14	0.04	17	−4.04	−2.80	0.26	−6.72
benzyl alcohol	3	Bzol	−15	−4.07c	−6.04	0.02	3.8	−3.41	−2.84	0.23	−6.68
2-acetylpyrrole	3	AcPyr	~90	−3.31l	−5.93	0.01	0.50	−2.65	−2.95	0.23	−6.56
methyl anthranilate	3	MeAnth	24	−4.63c	−5.77	0.01	7.3	−4.03	−3.16	0.26	−6.35
nicotine	3	Nic	−79	−4.49m	−5.37 (=Kp,fb)	NA	2.1	−3.89	−3.57	0.27	−5.95 (=Kp,fb)
coumarin	3	Cou	71	−5.64l	−5.17	0.05	19	−4.99	−3.72	0.23	−5.80
ethyl maltol	3	EtMalt	~87	−6.13l	−5.10	0.05	48	−5.43	−3.75	0.21	−5.77
maltol	3	Malt	162	−5.50l	−5.07	0.04	11	−4.80	−3.77	0.21	−5.75
cinnamyl alcohol	3	Cinnol	33	−4.50c	−4.88	0.02	0.70	−3.91	−4.07 av:	0.27 0.34	−5.45

^aFor aldehydes and ketones (marked with *), log K_p values may be too large by ~0.1 due to artifacts caused by polymerization, hemiacetal formation, acetal formation, hemiketal formation, and/or ketal formation in the PG/GL liquid phase; log K_p (m³/μg) values were measured at 20 °C in 50/50 (by weight) propylene glycol/glycerol. Corresponding activity coefficient values ζ were calculated by eq 2 using $p_{\mathrm{L}}^{\mathrm{o}}$ at 20 °C and $\overline{\text{MW}}=83.3\mathrm{g}∕\text{mol}$. Values of $p_{\mathrm{L}}^{\mathrm{o}}$ at 20 and 37 °C were used to obtain K_p values extrapolated to 37 °C by eq 9.

^bN = 3 unless NA (not available)is given for the standard deviation for the measured log K_p at 20 °C, in which case N = 1.

^cAntoine eq as given at https://webbook.nist.gov (accessed June 29, 2018).

^dAntoine eq from Diaz et al.⁶

^eAntoine eq as given in Towler and Sinnott.⁷

^f$p_{\mathrm{L}}^{\mathrm{o}}$ at 25 °C from http://www.thegoodscentscompany.com corrected to 20 °C using ΔH_vap for other butyrates.

^gAntoine eq for $p_{\mathrm{L}}^{\mathrm{o}}$ as given in Soni et al.⁸

^hAntoine eq for $p_{\mathrm{L}}^{\mathrm{o}}$ as given in Petitjeanet al.⁹

ⁱ$p_{\mathrm{L}}^{\mathrm{o}}$ at 25 °C from http://www.thegoodscentscompany.com (accessed June 29, 2018). $p_{\mathrm{L}}^{\mathrm{o}}$ at 20 °C calculated using ΔH_vap from https://www.chemeo.com/cid/10-581-5/Aromadendrene.pdf (accessed June 29, 2018).

^j$p_{\mathrm{L}}^{\mathrm{o}}$ at 25 °C from http://www.thegoodscentscompany.com (accessed June 29, 2018). $p_{\mathrm{L}}^{\mathrm{o}}$ at 20 °C calculated using ΔH_vap from https://webbook.nist.gov (accessed June 29, 2018).

^k$p_{\mathrm{L}}^{\mathrm{o}}$ at 25 °C from http://www.thegoodscentscompany.com/, accessed June 29, 2018. $p_{\mathrm{L}}^{\mathrm{o}}$ at 20 °C calculated using ΔH_vap from https://www.chemeo.com/cid/18-655-5/3-Hexen-1-ol,%20(Z)-.pdf, accessed June 29, 2018.

^l$p_{\mathrm{L}}^{\mathrm{o}}$ at 25 °C from Pubchem, Open Chemistry Database, https://pubchem.ncbi.nlm.nih.gov/, accessed June 29, 2018. $p_{\mathrm{L}}^{\mathrm{o}}$ calculated using eq 8 with ΔH_fus (kJ/mol) as follows: 2-Acetylpyrrole, 14.08; coumarin, 19.14 kJ/mol; vanillin, 22.40 kJ/mol; ethyl vanillin, 23.10 kJ/mol, all from https://webbook.nist.gov, accessed June 29, 2018. Experimental ΔH_fus values were not found for maltol and ethyl maltol, so ΔH_fus = 19.68 kJ/mol was used for both.

^mPankow.¹⁰

K_p,i Measurements.

The 32 compounds were divided into groups “1”, “2”, and “3” based on rough initial estimations of expected K_p (~low, ~medium, and ~high log K_p, respectively). Four additional compounds were also studied, but the K_p,i values are not reported because of complications due to (1) formation of a cyclic trimer and cyclic tetramer (acetaldehyde), (2) likely formation of hemiketals and ketals with PG/GL (diacetyl), and (3) low volatility (vanillin and ethyl vanillin). All measurements were carried out at 20 °C using 50/50 by weight mixtures of PG (MW = 76.09 g/mol) and GL (MW = 92.09 g/mol). Various standard stock mixtures (SSMs) were prepared as mixtures of initially pure (neat) compounds (no solvent was used) for spiking the PG/GL solution. For group 1, standard stock mixture 1 (SSM-1) was prepared; for group 2, standard stock mixture 2 (SSM- 2) was prepared. The concentrations in SSM-1 and SSM-2 ranged from 30 to 300 μg/μL. For measurement of the K_p,i values for the 13 compounds in groups 1 and 2, 8 μL of SSM-1 and 50 μL of SSM-2 were spiked into each of three replicate 5 mL volumes of 50/50 by weight PG/GL in glass 155 mL crimp top (Teflon-faced septum) bottles. For each of the 13 compounds, the final solution-phase concentration was in the range 0.06–2.1 μg/μL; higher concentrations were used for compounds that were less volatile. Several minutes of inclined (~20°) rotation was used to mix the liquid followed by a 24 h headspace equilibration period. This was followed by sampling of 50 μL aliquots of headspace gas (5 replicates) with a 100 μL gastight syringe (Hamilton Company Inc., Reno, Nevada). Injection occurred into the injection port (235 °C, 1-to-5 split) of an Agilent 7890A gas chromatograph (GC, Santa Clara, CA). The GC was interfaced to an Agilent 5975C mass spectrometer (MS) operated in electron impact (70 eV) ionization mode. The fused silica capillary GC column was 30 m long with 0.25 mm i.d. and a 1.4 μm film thickness coating of Rxi-624Sil MS (Restek Inc., Bellefonte, PA). The GC temperature program was 40 °C for 3.5 min, ramp to 100 °C at 12 °C/min, ramp to 250 °C at 15 °C/min, hold at 250 °C for 1 min, and then 10 °C/min to 230 °C. For calibration, 50 μL volumes of gas standards of the 13 compounds were also analyzed by GC/MS. The standards were prepared by injecting 2–20 μL of SSM-1 and SSM-2 into a 2 L glass static dilution bottle (Kimble Chase, Vineland, NJ); the bottle was held at 40 °C for 30 min and then at 20 °C for 2 h; on the basis of prior experience with gas standards,¹² it was assumed that 100% of each of the 13 compounds evaporated to the gas phase in the bottle.

For the 19 compounds in group 3, a 50 μL headspace gas sample did not provide sufficient analyte: most of the group 3 compounds exhibit relatively large K_p values, making for relatively low c_g values (see eq 1). Two replicate 60 mL glass “VOA” vials (Thermo Scientific Inc., Texas) were charged with 25 mL of 50/50 PG/GL. For each replicate, 20–230 mg of each group 3 compound was then weighed into the PG/GL, giving final concentrations in the PG/GL between 0.78 and 9.23 μg/μL. The vials were sealed with 0.125 in. thick PTFE-faced silicone septa, stirred for ~30 min with a Teflon-coated stirbar, covered with aluminum foil, then equilibrated for 24 h. The headspace was then sampled at 10 mL/min as follows. The inlet end of a 10 cm length of 0.53 mm i.d. uncoated fused silica capillary tubing (Restek, Bellefonte, PA) passed through the septum and terminated in the headspace of the vial. The outlet end of the tubing was connected to an adsorption/thermal desorption (ATD) gas sampling cartridge (Camsco Inc., Houston, TX) packed with 100 mg of 35/60 mesh Tenax TA followed by 200 mg of 60/80 mesh Carbograph 1 TD. The sample flow was drawn by a syringe pump (Model NE-1010, New Era Pump Systems Inc., Farmingdale, NY) equipped with a customized 300 mL syringe. N₂ gas at 10 mL/min was allowed to enter the vial through the septum via a 30 cm length of 0.53 mm i.d. uncoated fused silica capillary tubing (Restek, Bellefonte, PA) that ended below the level of the PG/GL, thereby creating very small bubbles of N₂. No evidence was observed in the data of errors due to any bubble-created aerosol droplets (i.e., no evidence of biased-low K_p,i values). A tee valve positioned between the cartridge and the syringe pump was used to periodically exhaust gas from the syringe, allowing verification at that time of the sample volume. For each vial, an initial 200 mL of headspace was discarded and not sampled with an ATD cartridge. Three 200 mL samples were then obtained with three cartridges followed by five 100 mL samples obtained with five cartridges, which was then followed by five 50 mL replicates with five cartridges. (The range of sample volumes ensured that all compounds could be quantitated within reliable portions of their calibrations.) After sampling, a 5 min flow of N₂ at 40 mL/min was passed into the inlet end of each cartridge to remove oxygen. The cartridges were loaded into a TurboMatrix 650 ATD unit (PerkinElmer, Waltham, MA). Prior to desorption, the unit automatically loaded each cartridge with 24 ng of fluorobenzene as an internal standard. Each cartridge was thermally desorbed at 285 °C for 10 min (backflush direction) with the TurboMatrix 650 interfaced to the same GC column and GC/MS as described above. In the TurboMatrix 650, 40 mL/min of the desorption stream was focused on a Tenax-TA intermediate focusing trap (IFT) set at —10 °C. The remaining 25 mL/min was discarded as the “inlet split”. Upon completion of ATD cartridge desorption, the IFT was thermally desorbed at 295 °C for 4 min at 25 psi He pressure and a flow of ~24 mL/min with ~4 mL/min proceeding onto the column via a 225 °C transfer line (~20 mL/min was discarded by the TurboMatrix 650). Data acquisition began when the IFT desorption was started. The GC temperature program was 40 °C for 2 min, ramped to 100 °C at 10 °C/min, ramped to 280 °C at 12 °C/min, and then programmed to 230 °C at 10 °C/min. For multipoint calibration for each analyte, 4 μL aliquots of standards in isopropyl alcohol were spiked into the inlet ends of clean cartridges to obtain mass amounts of 2–1400 ng per analyte; the final concentrations in the PG/GL solution were verified by taking aliquots of 25, 50, and 100 μL, diluting in isopropyl alcohol, and analyzing by GC/MS.

Replicate blank cartridges were prepared using gas sample volumes of 200 mL drawn through the experimental setup with no PG/GL solution as yet placed in the vial.

An additional series of measurements was carried out with nicotine alone using the experimental setup for group 3 compounds. The series was conducted to ensure that the K_p value reported is that for free-base nicotine (K_p,fb), i.e., unaffected by any significant degree of protonation. (Protonation will greatly alter the aerosol K_p value for total nicotine according to^5,13,14 K_p = K_p,fb/α_fb, where α_fb is the fraction of the nicotine in the particle phase that is in the free-base form; values of α_fb for selected e-liquids were reported by Duell et al.¹⁵)

Here, 25 mL of 50/50 (by weight) PG/GL was spiked with 103 mg of nicotine to give a level of 4.12 μg/μL. Fifteen milliliters of ammonia gas was then bubbled into the solution (1/1 nicotine/ammonia mol ratio). A small stir bar mixed the solution for 24 h. Sampling of the headspace with bubbling of the PG/GL liquid proceeded as above. Eight ATD cartridge samples were collected and analyzed as above: four replicates at 200 mL and four replicates at 100 mL.

3. RESULTS

The log K_p,i values measured at 20 °C are given in Table 1. Also given are the corresponding values of ζ_i at 20 °C calculated using eq 2 with the log $p_{\mathrm{L},i}^{\mathrm{o}}$ values given and with $\overline{\text{MW}}=83.3\mathrm{g}∕\text{mol}$ for 50/50 PG/GL. The accuracies of the calculated ζ_i values are subject to errors in the log $p_{\mathrm{L},i}^{\mathrm{o}}$. Furthermore, as noted above for acetaldehyde and diacetyl, aldehyde and ketone compounds are subject to reactions at the carbonyl group: detailed studies as to whether any of the aldehyde and ketone compounds in Table 1 (marked with *) were affected by such reactions were not carried out here. Any such effect would cause the reported K_p value to be somewhat too high. Overall, the K_p values for the compounds without an * are considered most reliable.

As noted elsewhere,¹⁰ from eq 2, assuming that ζ_i is approximately independent of temperature,

$$K_{\mathrm{p},i}(310.15\mathrm{K})=\left[\frac{p_{\mathrm{L},i}^{\mathrm{o}}(293.15)}{p_{\mathrm{L},i}^{\mathrm{o}}(310.15)}\times \frac{310.15}{293.15}\right]\times K_{\mathrm{p},i}(293.15\mathrm{K})$$ (9)

$$\text{If}{\tau }_{20\to 37}\equiv \left[\frac{p_{\mathrm{L},i}^{\mathrm{o}}(293.15)}{p_{\mathrm{L},i}^{\mathrm{o}}(310.15)}\times \frac{310.15}{293.15}\right]$$ (10)

$$K_{\mathrm{p},i}(310.15\mathrm{K})={\tau }_{20\to 37}{K}_{\mathrm{p},i}(293.15\mathrm{K})$$ (11)

Values of log K_p,i, at 37 °C (normal human body temperature) calculated by eq 7 are given in Table 1. The average value of τ_20→37 for the Table 1 compounds is 0.32, so on average in 50/50 PG/GL, the values of K_p,i were found to decrease by a factor of ~3 as temperature increases from 20 to 37 °C.

Figure 1 is a plot for the measured log K_p,i values vs log $p_{\mathrm{L},i}^{\mathrm{o}}$ at 20 °C. Dashed lines are included for ζ = 1, 10, 100, and 1000 with $\overline{\text{MW}}=83.3\mathrm{g}∕\text{mol}$. Because PG and GL are polar solvents, the points for the most polar molecules plot near the ζ = 1 line (propanal, acetone, and nicotine). For the molecules that are increasingly hydrophobic, the calculated ζ_i values increase. For comparisons with ζ values found in the literature for solutes in glycol-type solvents, for benzene at 25 °C Opris¹⁶ measured ζ = 14.0 in 1,3-propanediol (an isomer of PG) and ζ = 44.2 in ethylene glycol (EG) (both EG and GL have one −OH group per carbon). The average is ζ = (14.2 + 44.2)/2 = 29. Here, for benzene in 50/50 PG/GL, we obtained ζ = 30 at 20 °C. Similar general agreement is shown in Table 2 for toluene and ethylbenzene. For larger nonpolar compounds such as limonene, p-cymene, and aromadendrene in 50/50 PG/GL, we found ζ = 780, 410, and 960, respectively.

Figure 1.

Log K_p vs log $p_{\mathrm{L}}^{\mathrm{o}}$ at 20 °C for 32 compounds partitioning to 50/50 (w/w) propylene glycol/glycerol (PG/GL). Mini correlation lines shown for related compounds. * for aldehydes and ketones; ● for all other compounds.

Table 2.

Comparison of ζ Values for Benzene, Toluene, and Ethyl Benzene as Obtained by Opris¹⁶ and in This Work

	Opris16 at 25 °C			this work at 20 °C
compound	ζ in 1,3-PDa	ζ in EGb	ζ average	ζ in 50/50 PGc/GLd
benzene	14.0	44.2	29	30
toluene	23.5	82.1	53	65
ethylbenzene	36.2	141	89	130

^a1,3-PD = 1,3-propanediol.

^bEG = ethylene glycol.

^cPG = propylene glycol.

^dGL = glycerol.

In log K_p,i vs log $p_{\mathrm{L},i}^{\mathrm{o}}$ plots, it is well-known that data points for structurally related compounds are in general much more highly correlated than observed for a larger, multicompound-class data set as with alkanes and polycylic aromatic hydrocarbons (PAHs) partitioning to tobacco smoke PM.¹⁷ This type of result is observed in Figure 1, wherein the points for p-cymene, limonene, p-xylene, ethylbenzene, toluene, and benzene exhibt their own mini correlation line.

4. DISCUSSION AND CONCLUSIONS

Within the K_p,i expression, ζ_i occurs with $p_{\mathrm{L},i}^{\mathrm{o}}$ as the product ζ_i$p_{\mathrm{L},i}^{\mathrm{o}}$. Assuming all ζ_i = 1, Pankow¹ provided a plot of log f_g,i vs log $p_{\mathrm{L},i}^{\mathrm{o}}$ with lines for various TPM values between 10⁷ and 10⁹ μg/m³. Here, because we have now obtained information regarding ζ_i values for compounds of interest in PG/GL, Figure 2 gives curves (at equilibrium) at 37 °C for log f_g,i vs $\log ({\zeta }_i{p}_{\mathrm{L},i}^{\mathrm{o}})$ with values for selected compounds of the latter marked on the upper x-axis. At equilibrium, for benzene, toluene, and limonene, 90% or more of the compounds will be in the gas phase even at TPM = 10⁹ μg/m³. For cinnamyl alcohol, maltol, and free-base nicotine, at equilibrium, 90% or more of the compounds will be in the particle phase even at TPM = 10⁷ μg/m³. Upon inhalation, deposition of particles and gaseous compounds will begin. For a gas/particle system initially at equilibrium, the deposition of a compound as a gas will move the system away from equilibrium and moreover enhance further volatilization from the inhaled particles that remain suspended. For compounds with even modest volatility, as perhaps with K_p < 10⁻⁹ μg/m³, a significant fraction of the overall gas + particle inhaled amount may deposit to respiratory tract surfaces even when particle deposition is inefficient.⁵ The K_p,i values obtained here will be useful in aerosol inhalation/deposition modeling efforts that seek to address such deposition questions: even when bulk gas/bulk liquid equilibrium is not present, they can be used to describe conditions directly at the gas/particle interface where the nonequilbrium mass transfer is occurring.

Figure 2.

f_g,i vs $\log [{\zeta }_i{p}_{\mathrm{L},i}^{\mathrm{o}}]$ at 37 °C for partitioning to 50/50 (w/w) propylene glycol/glycerol PG/GL) and a range of TPM (μg/m³) values.

ABBREVIATIONS

ATD

adsorption/thermal desorption

e-cigarette

electronic cigarette

e-liquid

electronic cigarette liquid

fb

free base

GC

gas chromatograph/chromatography

EG

ethylene glycol

GL

glycerol

IFT

intermediate focusing trap

PD

1,3-propanediol

PG

propylene glycol

SSM

standard stock mixture

TPM

total particulate matter

MS

mass spectrometer/spectrometry

MW

molecular weight

VOA

volatile organics analysis

w/w

weight/weight