On the Deposition of Volatiles and Semivolatiles from Cigarette Smoke

1020

Chem. Res. Toxicol. 2004, 17, 1020-1037

On the Deposition of Volatiles and Semivolatiles from Cigarette Smoke Aerosols: Relative Rates of Transfer of Nicotine and Ammonia from Particles to the Gas Phase Jeffrey I. Seeman,*,† Peter J. Lipowicz,*,‡ Jean-Jacques Piade´,*,§ Laurent Poget,§ Edward B. Sanders,§ James P. Snyder,*,| and Clarence G. Trowbridge*,| SaddlePoint Frontiers, 12001 Bollingbrook Place, Richmond, Virginia 23236-3218, Philip Morris USA Research Center, 4201 Commerce Road, Richmond, Virginia 23234, Philip Morris International, Research and Development, c/o Philip Morris Products S.A., CH 2000, Quai Jeanenaud 56, 2000-Neuchaˆ tel, Switzerland, and Department of Chemistry, Emory University, Atlanta, Georgia 30322 Received July 2, 2003

The hypothesis that elevated levels of ammonia-releasing compounds in tobacco and ammonia in mainstream (MS) smoke increase the rate and amount of nicotine evaporation from the particles of MS smoke aerosol was examined by kinetic modeling and experiments with MS cigarette smoke. Computational simulation of a kinetic mechanism describing volatile loss of nicotine, ammonia, and acetic acid from an aqueous solution was used to compute the timedependent concentration of all species in the model. Because of the high volatility of ammonia relative to that of nicotine, variation over a wide range of initial ammonia concentration had no significant effect upon the rate of loss of nicotine from the model system. The effects of a variation in the volatile loss rate constant for ammonia and for the acid were examined. The simulations show that ammonia is lost from the model solution at a greater rate than nicotine and acid, and the loss of volatile acid has a significant role in the rate and amount of nicotine loss. Simulations with a model system undergoing a continuous steady addition of ammonia showed that high rates of ammonia addition could significantly increase the rate of nicotine volatile loss from the model solution. A series of smoking experiments was performed using blended cigarettes connected to a denuder tube. Deposition of smoke constituents can occur directly from the gas phase and by the deposition of smoke aerosol particles themselves. As nicotine exists >99% in the particle phase of MS smoke, in the absence of particle deposition, denuder tube deposition of nicotine occurs via the evaporation-deposition pathway. Solanesol, a nonvolatile tobacco and smoke terpene, was used to quantify the amount of particle deposition onto the denuder tube. The amount of ammonia deposited on the denuder tube was an order of magnitude greater than that of nicotine, showing that ammonia evaporates from the MS smoke particles much faster than does nicotine. The experimental results were supported and explained by the aqueous model simulations. Included in these experiments are cigarettes that differ in their MS smoke ammonia content by a factor of ca. five. However, an increased amount of MS smoke ammonia does not increase the rate of nicotine loss from the particles. The combined results support the conclusion that ammonia in mainstream smoke has little effect, if any, upon the rate and amount of nicotine evaporation from MS smoke particles.

I. Introduction Mainstream tobacco smoke is an aerosol composed of gases and heterogeneous particles whose physical and chemical composition can change with time and whose components can interconvert from one phase to the other (1-4). Greater than 99% of the nicotine in mainstream (MS)1 smoke aerosol is in the particles (1, 5). During the past few years, several research groups have considered the possibility that the presence of ammonia-releasing * To whom correspondence should be addressed. E-mail: jiseeman@ yahoo.com (J.I.S.); [email protected] (J.-J.P.); peter.j. [email protected] (P.L.); [email protected] (J.P.S.); [email protected] (C.G.T.). † SaddlePoint Frontiers. ‡ Philip Morris USA Research Center. § Philip Morris International. | Emory University.

compounds in tobacco and any resulting ammonia in MS smoke could affect the volatility of MS smoke nicotine (6-12). Ammonia, highly volatile, is present in tobacco 1 Abbreviations: ETS, environmental tobacco smoke, the total smoke aerosol (particles and gases) that is found in a room, resulting from the combination of sidestream smoke and exhaled mainstream smoke; FTC, the U.S. Federal Trade Commission; IM #16, the Industry Monitor #16 cigarette; ISTD, internal standard; ISO, International Organization for Standardization; ML, Marlboro Lights cigarette used in these studies, manufactured in Switzerland; MS smoke, mainstream tobacco smoke, the smoke stream issuing from the mouth end of a cigarette smoke upon puffing; PTFE, poly(tetrafluoroethylene); SI, Supporting Information; SS smoke, sidestream smoke, the smoke stream that issues, not from the mouth end of a cigarette, but rather from the burning end, including through the paper or filter; “Tar”, the mainstream smoke total particulate matter minus its nicotine and water content, obtained using the FTC smoking method; TEA, triethylamine; TPM, total particulate matter, the portion of mainstream smoke that is collected on a Cambridge filter in the machine smoking FTC and ISO methods.

10.1021/tx0300333 CCC: $27.50 © 2004 American Chemical Society Published on Web 07/29/2004

Deposition from Cigarette Smoke Aerosols

Chem. Res. Toxicol., Vol. 17, No. 8, 2004 1021

Figure 1. Illustration of the effect of bases and acids on the evaporation of compounds from aerosol particles, assuming the presence of bases modeled only by ammonia and organic acids modeled by RCO2H. Nicotine represents the semivolatile amines and ammonia the volatile amines in tobacco smoke. Irreversible deposition is represented at the far right of the figure. The simulation model assumes that only neutral species can evaporate, that acids and bases are mobile and can react within the particles, that the diprotonated form 3 is present in negligible amounts, and that evaporation is irreversible.

(13) and also formed during the smoking process (1). A review summarizes the scientific literature related to this topic as of mid-2000 (14). Pankow and co-workers proposed that tobacco companies add ammonia to their cigarettes to increase MS smoke ammonia, that this MS smoke ammonia converts particulate phase protonated nicotine (2) to nonprotonated nicotine (1), that a greater concentration of nonprotonated nicotine in the smoke particles increases the rate and amount of nicotine evaporation from particles to the gas phase, and that an increased amount of nonprotonated nicotine in MS smoke increases nicotine’s bioavailablity to the smoker (6, 7). These authors trapped environmental tobacco smoke (ETS) on a filter. They then purged nicotine from the filter with a continuous flow of ammonia-water vapor with a nitrogen carrier. By assuming that their ETS experimental data was applicable to MS smoke aerosol, Pankow et al. concluded that the ammonia in MS smoke could increase nicotine’s concentration in the gas phase of MS smoke by a factor of over 100 (6).

The Pankow et al. studies (6, 7) considered neither the volatility of ammonia, which is much greater than that of nicotine, nor the volatilization of acids from MS smoke particles, as shown in Figure 1. The cover of the November 2001 issue of this journal, provided by Pankow et al. (7, 15), incorporated the volatilization of only nicotine from smoke particles. In fact, the ability of any acid or base in MS smoke aerosol particles to alter nicotine volatility will be limited by the volatility of these acids and bases. Ingebrethsen et al. performed denuder tube deposition experiments on cigarette smoke generated under a number of different experimental conditions and

considered ammonia volatility and the effect of MS smoke carbon dioxide (8). To provide new information about factors affecting nicotine’s volatility in cigarette MS smoke, a mathematical simulation was designed and a new experimental investigation of the problem was executed. In its present state, the simulation model is an aqueous solution containing nicotine, ammonia, and acetic acid (or formic acid), all of which are explicitly recognized as volatile. Simulation of the model is achieved by representing all conjugate acid-base reactions and component volatile loss as a set of simultaneous rate equations. In the second part of this investigation, a specially designed denuder tube technique was developed to allow quantification of both ammonia and nicotine loss from cigarette mainstream smoke. The kinetic simulations show that the high rate of loss of ammonia and the influence of acetic acid volatility prevents increased initial levels of ammonia from having a significant effect upon the rate of loss of nicotine from the model aqueous solution. The experimental results, conducted directly on mainstream tobacco smoke aerosol, show that ammonia is lost from MS smoke about 10 times faster than nicotine. The modeling and experimental parts of the joint investigation both support the conclusion that a base as volatile as ammonia is lost quickly from the system, and that even high initial concentrations of ammonia will have no appreciable affect upon the nicotine loss rate from fresh MS smoke. A fundamentally different condition, that being a continuous steady addition of ammonia into the model solution [or a continuous addition of ammonia through a thin film of smoke precipitate, as in the Pankow et al. experiments (6)] was also considered. A modified simulation model shows that a continuous steady addition of ammonia significantly increases the rate of evaporation of nicotine.

II. Experimental Approach Figure 1 illustrates the volatilization of neutral (i.e., noncharged) aerosol particle constituents to the gas phase and their subsequent irreversible deposition onto biological tissue, e.g., onto buccal cavity tissue or respiratory

1022 Chem. Res. Toxicol., Vol. 17, No. 8, 2004

Seeman et al.

Table 1. Simulation Modela,b conjugate acid ionization AH ) H+ + AH+‚A-/AH ) Ka ) k1/k2 AH f gas phase, rate constant kvah NH4+ ) H+ + NH3, H+‚NH3/NH4+ ) Kn ) k3/k4 NH3 f gas phase, rate constant kvn RH+ ) H+ + R H+‚R/RH+ ) Kr ) k7/k8 R f gas phase, rate constant kvr HOH ) H+ + OH-, H+‚OH- ) Kw ) k5/k6

associated rate equations acetic acid, AH

dAH/dt ) -k1AH + k2A-‚H+ - kvahAH dA-/dt ) k1AH - k2A-H+

ammonia, NH3

nicotine, RH+

dNH4+/dt ) -k3NH4+ + k4NH3‚H+ dNH3/dt ) -dNH4+/dt - kvnNH3 dRH+/dt ) -k7RH+ + k8RH+ dR/dt ) -dRH+/dt - kvrR

water, HOH

dOH-/dt ) k5 - k6H+OHdH+/dt ) dOH-/dt + dA-/dt - dNH+/dt - dRH+/dt

conservation of charge, used for initial equilibrium state calculations only H+ + NH4+ + RH+ ) A- + OHa The volatile loss rate constants for conjugate acid acetate, conjugate base ammonia, and conjugate base nicotine are k vah, kvn, and kvr, respectively. The concentration of water (55.56 M) is included in the dissociation rate constant k5. b The values of the various rate constants in Figure 1 but not included in the model equal zero.

system tissue. This illustration contrasts sharply with the Pankow presentation (6, 7, 15) in that ammonia and the acid(s) now are included. A simplistic view of Figure 1 would suggest that ammonia in the particle would alter the distribution of nicotine, favoring the nonprotonated 1 (semivolatile, i.e., low but measurable volatility) at the expense of nicotine’s singly charged (nonvolatile) form 2, thereby increasing the amount and rate of nicotine volatilization and ultimately increasing nicotine bioavailability (7, 16). Importantly, the efficacy with which acids or bases can affect nicotine volatility will depend on the volatility of those components compared to the volatility of nicotine itself. The first step in this work was the development and analysis of a simulation model for a volatilizationirreversible loss system. As stated above, literature analyses had not included ammonia or any of the known organic acids in the MS smoke aerosol particles (6-8, 15). The model simulation was to evaluate the extent to which acids and bases form a “linked or coupled acidbase system”. Simply put, to what extent (i) do acids in smoke particles affect the evaporation of the bases and vice versa and (ii) do the relative volatilities of the acids and bases, including ammonia, affect nicotine’s volatilization. The model was also designed to reveal the extent to which the loss of compounds would be affected by varying the initial concentrations of all the components. In this regard, one literature experimental study examined the continuous addition of exogenous ammonia to trapped particulate matter from environmental tobacco smoke and from mainstream tobacco smoke (6). The simulation model was likewise designed to allow for the evaluation of a continuous steady source of exogenous ammonia. The simulation model is described by Figure 1 and the set of simultaneous rate equations shown in Table 1. The model includes nicotine, ammonia, acetic acid (or formic acid), and their conjugate salts in an aqueous solution using the relevant physical chemical parameters available in the literature. Irreversible deposition onto either a biological tissue or a denuder tube was simulated. The simulation output included tables of time:concentration values for each of the model’s components. Nicotine in tobacco and smoke can exist in a diprotonated form 3 (17), but the simulation model follows literature precedent (6,

7, 16) in considering only the monoprotonated form 2 and the nonprotonated 1 as the concentration of 3 is significant in dilute aqueous solutions only below pH 4. The organic acids were modeled by acetic acid, the major acid in MS smoke (bp 118 °C), and by formic acid, the second most prominent acid in MS smoke (bp 100 °C, present in ca. 25-50 wt % relative to acetic acid). Numerous other acids are present in far lower concentrations (18-21). The sensitivity of ammonia and nicotine to volatile loss due to varied volatile loss rate constant and pKa of the acidic component were also examined.

While water is a major component of tobacco total particulate matter (TPM) (1), smoke aerosol is not a dilute aqueous solution. The lack of lipids in the test model could impact the relative rates of volatilization of the species in the model. The aqueous assumption has been used in the past by other workers in simplifying the complexities of smoke chemistry (6, 8, 22). The lack of accounting for lipophilicity may make the model a “worst-case scenario”, given that the rate of evaporation of ammonia from a more lipid-like medium is most likely greater than from an aqueous solution. The second phase of this work was the experimental examination of the relative rates of evaporation of nicotine and ammonia from mainstream tobacco smoke. Three distinct types of deposition pathways are possible: (i) direct deposition of ammonia and nicotine originally in the gas phase; (ii) direct deposition of MS smoke particles; and (iii) volatilization of substrates from the particles and subsequent deposition. As the hypothesis being examined deals exclusively with the third pathway, a smoking apparatus was designed that would allow the quantification of irreversible deposition of nicotine and ammonia from the gas phase while simultaneously being able to account and correct for particle


deposition. Pathway i is actually coupled mechanistically with pathway iii, in that gases in an aerosol can redeposit onto particles and gases can evaporate from the particles (5). Denuder tube technology was used in this work. A novel combined cigarette-denuder tube system was developed that is an improvement over earlier designs for the analysis of MS cigarette smoke (7, 23-28). Control experiments were performed (with and without the denuder tube), demonstrating that the incorporation of a denuder tube did not affect the overall smoke chemistry of the two cigarettes examined. Further, a new theory that underpins the use of denuder tube analyses for nicotine in tobacco smoke was recently developed by two of us [P.J.L. and J.-J.P.] (5). In discussing the experimental portion of this work, deposition of gas-phase smoke constituents onto the denuder tube is typically discussed rather than evaporation of substances from the particles. In these denuder tube experiments, deposition is indicative of but not identical to evaporation. First, it is deposition that is experimentally measured, not evaporation. Second, the amount of particle deposition was quantified by determining the amount of deposition of a substance that, because of its nonvolatility at room temperature, can deposit only by particle deposition. Careful experimentation resulted in very low particle deposition. Hence, the measured deposition is indicative of evaporation. Third, as discussed in section V.A, any ammonia in the gas phase does not directly affect nicotine’s evaporation from the particles. The results and conclusions of this work are transferable to other U.S. blended commercial cigarette brands for a number of reasons, discussed in greater detail below. The mainstream smoke from two blended cigarettes was examined, having a 5-fold variation in concentration of MS smoke ammonia. In both cases, gas-phase ammonia deposition was much greater than nicotine deposition. The underlying physical properties of ammonia and nicotine dictate their relative volatilities from tobacco smoke particles from all cigarettes. The acid/base chemistry of cigarette smoke is known to be rather uniform among standard U.S. commercial cigarettes (14, 29, 30).

III. Experimental Section A. Experimental DesignsGeneral Considerations. The object of the experimental portion of this work is to characterize the relative rates of loss of ammonia and nicotine from tobacco smoke aerosol particles (as distinctly different from trapped particulate matter or aged smoke) and determine if ammonia in MS smoke enhances nicotine’s volatilization rate. Denuder tubes have been used expeditiously to evaluate nicotine’s volatilization from smoke particles and are known to be able to separate the gas phase from particles in evolving aerosols (5, 7, 8, 23, 24, 28, 31-33). When an aerosol is passed under laminar flow conditions through a denuder tube, in theory the compounds in the gas phase can reach and adhere to the walls of the tube, while the residual particles travel through and exit the tube because they have a much lower diffusion coefficient (5, 23, 24). By this approach, substances in the gas phase or those evaporating into the gas phase can be quantified relative to the initial total mass of the aerosol. In practice, both gasphase molecules and aerosol particles can deposit. The latter can happen when the aerosol passes through the denuder tube via nonlaminar flow, i.e., if sufficient turbulence is present. The deposition of particles on the walls may substantially alter the results, and the extent of particle deposition has to be measured. The experimental design required the following: (a) simultaneous quantification of deposition of ammonia, nicotine, and

Chem. Res. Toxicol., Vol. 17, No. 8, 2004 1023 a molecular marker for particle deposition from MS smoke particles; (b) minimizing particle deposition; (c) quantification of the relative amounts of particle vs gas-phase deposition; (d) shielding of the MS smoke from the sidestream smoke, because much more ammonia is present in the SS smoke than in the MS smoke, the ratio being in the range of 40-170:1 (1); (e) avoiding smoke dilution to any meaningful extent; (f) sampling before the aerosol could react, degrade, evaporate, or age; (g) using the standard FTC and ISO machine smoking protocols, i.e., one 2-s 35-mL puff every 60 s (34); and (h) using either commercial or reference cigarettes, the latter designed to be similar to U.S. commercial blended cigarettes (35). The cigarettes should have a range of MS smoke ammonia levels. To measure the extent of particle deposition when examining tobacco smoke aerosol, solanesol (4) deposition can be quantified. Because of solanesol’s extremely low volatility (it is a C45-terpene alcohol), it is present only in the particulate matter of MS smoke (1, 36, 37). Any solanesol found on the denuder tube must therefore result from particle phase deposition. The contribution of particulate matter deposition to the total collected amounts of nicotine was assessed by determining the ratio of the analytically determined solanesol amounts on the denuder tube to the total amounts trapped at the denuder tube exit.

With regard to requirements g and h, it is well known that machine-smoking protocols do not represent how smokers smoke (38-41). However, the FTC and ISO procedures are useful for comparative purposes among different cigarettes and between different laboratories. These protocols have been used worldwide for years to examine the effect of physical and chemical modifications on tobacco smoke chemistry (34, 42-45). In addition, to ensure the applicability of the results of smoking studies, experiments are best performed within the composition and configuration range of commercial cigarettes. If not, extrapolations of the data back to commercial cigarettes may be in error. The experimental design parameters included the following: the length of the denuder tube, the number of segments the denuder tube was divided into following the MS smoke deposition, the flow rate through the cigarette, the design of the denuder tube flow system and its connection to the cigarette, and the analytical methods used to quantify the relevant smoke constituents. The flow through the denuder had to remain laminar, and the flow had to be unaffected by switching between the two smoking conditions (puffing and static smoldering of the cigarette). Any discontinuity or variation in the flow entering the denuder would result in turbulence within the denuder tube. To meet these requirements, the cigarette was smoked in a reverse configuration, i.e., by increasing the pressure at the burning end of the cigarette relative to that prevailing at its filter end rather than by pulling a vacuum at the mouth end as is done by a smoking machine or by human smoking. In this fashion, the puff of smoke exits the cigarette butt into the open, so that it is immediately captured within a sheath of laboratory air by the open end of a vertical denuder operating at constant flow and placed ca. 2 mm below the cigarette tip. The length of the denuder tube was in excess, by about a factor of 2, of what is required to capture particle-free nicotine vapor (5). B. Positive Pressure Smoking Apparatus and Denuder Assembly. The reverse configuration positive-smoking apparatus was designed to smoke cigarettes as close as possible to standard ISO 3402, 3308 conditions (44, 45). The main differences between the experimental design and standard smoking conditions were that (a) puffs were generated by forcing air into the cigarette instead of drawing it with a syringe, and (b) the exact conditions of air velocity around the cigarettes


Seeman et al.

Figure 3. Pressure profile measured during a puff in chamber C in Figure 2 as a function of time for the 1.6, 2.0, and 2.5 s puff times. Closing the P1 valve forced 1 L/min of air through the cigarette and generated a puff. This figure demonstrates the effect of varying the time that valve [V in Figure 2] is open. The tailing is due to the inertia caused by the volume of the cigarette enclosure and does not match the sine shape specified by the ISO method. The time was set so that the number of puffs and MS tar and nicotine yields obtained from the cigarettes smoked with this system was approximately the same as during the ISO smoking procedure. In the denuder experiments performed herein, the 1.6 s puff time was used.

Figure 2. Denuder tube assembly interfaced with a positive pressure, reverse flow smoking system. See section III.B for a description of its operation. (specified according to ISO at 170-230 mm/s) could only be tentatively ensured. A standard smoking port was located inside a sealed 200 mL glass chamber [C] (see Figure 2). Since the sidestream smoke is a major source of ammonia that could be taken up by the cigarette during the puff, the chamber was continuously flushed at 3 L/min with humidified laboratory air injected through an automated mass flow controller (Tecan gas diluter) [A] to ensure a constant flow. The air was humidified by passing it through an impinger containing deionized water cooled to 16 °C in a thermostatic bath (HAAKE G). The humidity was checked with a dew point sensor (Vaisala; HMP 31 UT) prior to each experiment and was close to 60% RH since the laboratory air was kept at 23 ( 1 °C. The air was extracted from the chamber using an assembly of two constant flow pumps (SKC; Air check sampler). The first pump [P1] was set to 1.0 L/min and connected to the extraction line via a three-way valve [V] controlled by a timer and an electronic actuator. The second pump [P2] was set to approximately 2 L/min; the exact flow was adjusted to maintain a 30-60 Pa negative pressure in the test chamber. This slight negative pressure ensured that small amounts of smoke were not continuously injected into the denuder during smolder between the puffs. The P2 pump was connected to the air extraction line via a restriction imparting a 30 mbar pressure drop to dampen pressure perturbations during puffs and to improve the control of pump flow. Depending on the three-way valve [V] status, the pump flow was either connected to (status-1, smolder) or isolated from (status-2, puffing) the extraction line. During the puff, a flow of 1 L/min was diverted to the cigarette while the remaining 2 L/min flushed the chamber [C]. The exact flow rate through the cigarette during the puff was set at exactly 1 L/min in continuous puffing mode (status-2) by tuning the [P2] pump (using an unlighted cigarette seated in the smoking port of the chamber to ensure the exit-chamber pressure drop). The timer actuating the valve [V] was programmed to switch every minute from status-1 to status-2 for the defined puff time and then back to status-1. The process was continuously monitored by measuring the pressure inside

the chamber (Almemo; pressure sensor FD A602-M1K, with Delphin Message Medana electronic recorder). Typical profiles are depicted in Figure 3, showing the pressure evolution during three tested puff timings. To a first approximation, the pressure profile is proportional to the flow through the cigarette. The pressure profiles are thus representative of the puff profiles, and the profiles shown in Figure 3 demonstrate the kind of adjustment that can be made on the profile by varying the time that valve [V] is open to status-2. The exact puff volumes during smoking could not be measured because the instrumentation used is reliable only for gases, not aerosols. The duration of the puffing status was adjusted to match the puff counts and nicotine and tar yields from a cigarette under the experimental conditions to the one observed under standard smoking conditions; the 1.6 s setting was used in these experiments. If the last part of the tailing in the 1.6 s puff time curve in Figure 3 is discarded and the baseline adjusted, the base width of a sineshaped curve that would best match the observed signal is ca. 2.8 s. This study is about evaporation from unaltered MS smoke as its gas phase is constantly depleted and not about MS smoke formation yields themselves. Therefore, it is unlikely that a small change in puff duration could change the composition or size distribution of the smoke to such an extent that the relative volatilization rates of ammonia and nicotine from the smoke particles would be significantly altered. The fact that essentially the same results are found for two cigarettes having significantly different MS smoke ammonia levels supports this assumption. The denuder tubes were prepared from 150 cm glass tubes of 7.8 mm i.d. × 12 mm o.d. The tubes were filled with an alkaline cleaning solution (Deconex 11.2%) and kept overnight, and then they were washed successively with deionized water (Type II deionized water, Millipore; Helix-3) and ethanol (Fluka puriss. p.a). The tubes were dried and stored under a gentle stream of pure nitrogen to avoid any ammonia contamination. The oxalic acid coating solution was prepared daily; 5 g of anhydrous oxalic acid (Fluka puriss. p.a.) was dissolved at 60 °C in 100 mL (w/v) of dry 2-propanol (Fluka; puriss; absolute; over molecular sieve). The open end of the 150 cm section of the tube was placed in the 5% hot coating solution, and by suction from a 100 mL syringe, the tube was filled to its top and then allowed to drain back slowly into the coating solution container. The denuder was dried for 20 min under a stream of 700 mL/min pure nitrogen. A proper coating had a homogeneous, milky appearance when dry. Both ends of the tube were immediately sealed with stainless steel cap connectors equipped with Teflon ferrules (Swagelok). The coating level was ap-

Deposition from Cigarette Smoke Aerosols proximately 0.2 mg/cm (2 µmol H+/cm for monoprotic equivalence), as determined by rinsing the tubes with deionized water and subsequent titration of the oxalic acid. It was critical to make the flow through the denuder insensitive to the puffing/nonpuffing conditions and to permit switching from one to the other without the use of valves and without loss of MS smoke. This was achieved as follows: The tip of the cigarette filter extended through the airtight smoking port outside the chamber. The front side of the denuder was positioned 1-2 mm below the filter end. A slightly greater amount of air was pulled through the denuder than what came from the cigarette (1.2 and 1.0 L/min, respectively). This allowed air flow to progressively be replaced by the smoke stream during a puff without changing the gas flow rate through the denuder tube. To ensure that ambient air could not be sucked into the denuder, a glass jacket 62 mm × 26 mm i.d. flushed with “synthetic” air at 4 L/min was used to cover both the top end part of the denuder and the cigarette mouth end. To trap the whole smoke exiting the denuder tube, the denuder outlet was connected to a filter holder enclosing a 44 mm Cambridge filter followed by two 25-mL impingers (open stem) connected in series. The sampling flow through the denuder was ensured by a constant flow pump (SKC; Air check sampler) set at the lowest flow rate compatible with the complete capture of the mainstream smoke, ca. 1.2 L/min for the Swiss-manufactured Marlboro Lights (ML) cigarette. The exact flow rate depended on cigarette ventilation (46). Ventilated cigarettes have an air intake in the filter, and smoke exiting the mouth end of the filter will be more dense near the axis of the cigarette and surrounded by a narrow sheath of air coming from the vent holes (47). Consequently, the smoke capture was easier when cigarettes with higher ventilation levels were used. Prior to the start of the experiment, a 10-cm section was cut from each end of the denuder; these two sections were capped (Swagelok cap connectors) and used as blanks. The remaining 130-cm section of the tube was installed into the smoking assembly. A conditioned and marked cigarette was lit outside the chamber and quickly reinserted into the smoking port inside the chamber. As soon as the butt-mark was reached and the last puff completed, the pump sampling through the denuder was turned off and the cigarette butt removed. No fractional puffs were used. The smoking chamber was cleaned successively with ethanol, hot water, deionized water, and ethanol. The chamber was then dried under a stream of nitrogen. This procedure was repeated for each cigarette smoked. A maximum of 10 individual cigarettes were smoked into the same denuder tube, but the Cambridge filter and the front impinger solution were replaced, respectively, by a new Cambridge filter and a volume of fresh impinger solution after five cigarettes. The filter protecting the pump extracting the sidestream smoke from the chamber was exchanged and discarded every five cigarettes. C. Smoke Constituent Analyses. (1) Analytical Methods. The effluents from the denuder tube were analyzed for nicotine, ammonia, and solanesol (4, to quantify particle deposition). The impinger/Cambridge filter extract solutions were analyzed separately either for nicotine and ammonia or for nicotine and solanesol. Ammonia, nicotine, and solanesol could not be extracted simultaneously from the Cambridge filters without compromising the precision of the analytical results. Ammonia quantification was performed using suppressed ion chromatography (IC) following the method of Hodge et al. (48), using a HP-1090 HPLC, cation self-regenerating Suppressor Dionex Conductivity Detector II, 1.7% temperature compensation, 30 µs output range, with 25 µL injections onto an IonPac precolumn/column CS12/CG 12 Dionex, 1 mL/min flow rate with 19 mM methanesulfonic acid (Fluka; Chemika; puriss.) in an isocratic mode at ambient temperature; autosuppression, Dionex CSRX-1 in recycle mode. Nicotine was determined using the ASTM method (49) with nicotine-N′-CD3 (5) (50) as the internal standard (ISTD) for nicotine. Quantification was performed using an Agilent 6880

Chem. Res. Toxicol., Vol. 17, No. 8, 2004 1025 gas chromatograph, with an Agilent MSD 5873 mass-selective detector heated at 280 °C with transfer line heated at 300 °C, electronic impact ionization at 70 eV. Selective detection for nicotine was determined using m/z 84, 133, and 162 amu; selective detection ions for nicotine-N′-CD3 were three amu higher each than for nicotine. A DB-5 MS column (J&W Scientific, 30 m × 0.24 mm × 1 µm) was used with helium at a 1.3 mL/min flow rate. The injector was equipped with a splitless liner (Agilent 5181-3316), 150 kPa held during injection, then decreased to 50 kPa after 30 s. The injector temperature was 235 °C, injection volume 2 µL, and the oven temperature program as follows: 70 °C for 1 min and then increased at 25 °C/min to 280 °C with a 3 min hold.

Solanesol quantification was performed using a recently developed ASTM method (51) using cholesterol as the ISTD. A Hewlett-Packard HP 1090 high-performance gas chromatograph was used with UV detection at 205 nm and a Deltabond ODS column (Keystone Scientific, 255-204-3-CPG), 250 mm length, 3 mm i.d., 5 µm size, 300 A pore. The mobile phase was acetonitrile (HPLC grade) containing 5% v/v methanol (HPLC grade) run isocratically, 0.5 mL/min, with a 25 µL injection volume. (2) Nicotine, Ammonia, and Solanesol Extraction from Denuder Tubes. After the last cigarette was smoked, both ends of the denuder were tightly capped using Swagelok cap connectors. The denuder was cut into sections using a glass knife. Each section was tightly capped for storage prior to extraction of the smoke constituents from the denuder tube walls. In some instances herein, quantities are reported per 10 cm of denuder tube length. In those cases, the amount of reagent was adjusted proportionally for the specific denuder tube length. The denuder segments were filled with (a) 2 mL of an aqueous solution containing 5% oxalic acid (Fluka, anhydrous puriss. p.a.) and the ISTD for ammonia (LiCl: Fluka, puriss. p.a.; 4.3 µg/mL in water); (b) 2 mL of methylene chloride (Fluka, puriss. p.a.); (c) 100 µL of ISTD for solanesol (cholesterol: Fluka, Biokema; 1.5 mg/mL, and R-tocopherol as antioxidant: Fluka, Biokema; 400 µg/mL, in methanol); and (d)100 µL of the ISTD for nicotine (nicotine-N ′-CD3 (5): Cambridge Isotope Laboratories; 50 µg/mL in methanol). Then the tube was tightly capped and shaken. For ammonia determination, a 300-µL aliquot of the aqueous phase was collected and analyzed by IC. For nicotine determination, the extraction was performed by adding 200 µL of 10 N sodium hydroxide (40 g/L; Fluka; puriss. p. a. in deionized water) and 1 × 250 µL of triethylamine (TEA; Fluka; puriss. p.a.; 1.2 mg/mL in methylene chloride) into the tube and shaking it, keeping the tube tightly capped. TEA is necessary to avoid loss of nicotine on glassware surfaces during sample preparation and to enhance conditions for the GC analysis of nicotine. Both aqueous and organic phases were transferred into a 50 mL separatory funnel, and the organic phase was separated in a graduated flask. The tube was rinsed by repeating shaking with fresh portions of a solution of 1 × 250 µL of triethylamine (TEA; Fluka; puriss. p.a.; 1.2 mg/mL in methylene chloride) in 2 mL of methylene chloride. This rinsing solution was used to complete the extraction of the aqueous solution remaining in the separatory funnel. Both organic extracts were collected together for analysis without preconcentration. For solanesol determination, 500 µL of extract was evaporated to dryness in a 1 mL glass vial at 30 °C under slight nitrogen current and recovered in 500 µL of methanol for chromatographic analysis. (3) Storage of Cambridge Filters and Impinger Solutions. The Cambridge filters were stored with both impinger solutions in a 40-mL glass-stoppered bottle (or with a single Cambridge filter in case of a 10-cigarette experiment). Impingers


Seeman et al.

Figure 4. pH values of aqueous extracts of the TPM from MS, SS, and environmental tobacco smoke samples of Merit filter cigarettes (7 mg of nominal “tar” yield; for MS smoke, 2-3 puffs collected; for SS smoke, 1-2 puffs collected; see text for ETS collection details). These results indicate that both acids and bases volatilize irreversibly from mainstream tobacco smoke particles. See text for additional details. and tube fittings were rinsed with 10 mL of fresh impinger solution (see below), and the extract and impinger solutions were combined and stored with the filters. For the simultaneous determination of nicotine and solanesol, the impingers were filled with 10 mL of methanol. For the simultaneous determination of nicotine and ammonia, the impingers were filled with 10 mL of 0.5% aqueous oxalic acid. (4) Ammonia and Nicotine Extraction from the Cambridge Filter. Internal standards for nicotine (1 mL of a methanolic solution of 500 µg/mL nicotine-N ′-CD3, Cambridge Isotope Laboratories) and ammonia (30 µL of a deionized water solution of 4.3 mg/mL LiCl) were added to the filters and combined impinger/wash solutions, and the resulting mixtures were extracted for 30 min in an ultrasonic bath. A 1.0 mL aliquot of the resultant extract was placed in a 2 mL polypropylene autosampler vial for subsequent ammonia determination by IC. Sulfuric acid (2 mL of 20%, 200 g/L in deionized water) was then added to the flask, and nicotine was further extracted from the filters for 30 min in the ultrasonic bath. This extract was transferred into a 150 mL separatory funnel. To complete nicotine recovery, two additional extractions with water (20 mL) and sulfuric acid (2 mL of 20%) were performed in the ultrasonic bath, and the combined extracts were neutralized with 10 N NaOH. Nicotine was extracted twice with two individual portions of 20 mL of methylene chloride. The organic extracts were combined for analysis, and the aqueous phase was discarded. For the nicotine determination, 100 µL of extract was mixed with 800 µL of methylene chloride and 100 µL of TEA solution in a 2 mL autosampler vial. (5) Solanesol and Nicotine Extraction from the Cambridge Filter. Nicotine and solanesol were extracted from the Cambridge filter as follows. The internal standards for solanesol [10 mL of a methanolic solution of 1.5 mg/mL of cholesterol (Fluka; Biokema) containing 400 µg/mL of R-tocopherol (Fluka; Biokema) added as an antioxidant] and nicotine (1 mL of a methanolic solution of 500 µg/mL of nicotine-N′-CD3) were added to the flask containing the filters and impinger solutions and washes, and the resultant mixtures were extracted for 30 min in an ultrasonic bath. For the solanesol determination, 100 µL of extract was evaporated to dryness in a 2 mL autosampler vial at 30 °C under a gentle flow of nitrogen and recovered in 1 mL of methanol. For nicotine determination, 100 µL of extract

was mixed with 800 µL of methylene chloride and 100 µL of TEA solution in a 2 mL autosampler vial. D. Determination of pH of Aqueous Extracts from TPM of Mainstream Smoke, Sidestream Smoke, and Environmental Tobacco Smoke. Merit filter cigarettes (7 mg of FTC “tar” yield) were chosen for these determinations. TPM was collected on PTFE filters (diameter, 47 mm; pore size, 2.0 µm; pH ) 6.3), with a maximum loading capacity of ca. 4 mg of particulate matter. PTFE filters were used because, in our hands, aqueous extraction of unloaded Cambridge filters showed a pH shift to pH 9. MS and SS smoke were generated with a four-port smoking machine (Borgwaldt SM R04) under standard conditions (35 mL/puff, 2 s/puff, 1 puff/min). For MS smoke, 2-3 puffs were collected per filter. For SS smoke, 1-2 puffs were collected per filter, using the fishtail chimney system (sample flow, 2.5 L/min) (52). ETS was generated by three smokers, each smoking two cigarettes in a sealed 22-m3 room (53). ETS respirable suspended particulate matter (RSP) concentration in the chamber was ca. 3000 µg/m3, with a sample flow of 10 L/min for a 1-h duration. The filters were extracted with potassium chloride solution (0.1 mol/L). The pH measurements were performed after 10 min of extraction. Nicotine determination was performed by extracting the TPM-loaded PTFE filters with 2-propanol. Analysis was achieved by gas chromatography with a nitrogen-phosphorus detector. The pH data of aqueous extracts of smoke (MS smoke, SS smoke, and ETS) are shown in Figure 4. This high concentration of ETS in the room was chosen for convenience in the sampling of suspended particulate matter and is not representative of real-life conditions. The intent of the measurement was to address a wide range of particle lifetime and evaporative conditions and to assess the relative molar volatilization of both acids and bases in MS smoke.

IV. Model Simulation Studies A. Model Structure. The computational model is limited to an examination of the role of volatile acids and bases upon apparent nicotine volatility in smoke aerosol particles. The highly simplified system is an aqueous solution of ammonia, acetic acid or formic acid, and



Table 2. pKa and Rate Constant Valuesa,b species acetic acid (AH) ammonia (NH4+) water nicotine (RH+)

pKa

dissociation (s-1)

association (M-1 s-1)

volatile loss rate constantc (s-1)

4.76 9.23 14 8.0

k1 7.8 × k3 24.6 k5 1.4 × 10-3 (M s-1) k7 1 × 102

k2 4.5 × k4 4.3 × 1010 k6 1.4 × 1011 k8 1 × 1010

kvah 9.36 kvn 1536

105

a

1010

kvr 1

b

Definitions of the various ki in this table are found in Table 1. Ionic rate constant values: acetic acid, Eigen, M., and Schoen, J. (1955) Z. Elektrochem. 59, 483-494; ammonia, Emerson, M. T., Gruenwald, E., and Kromhout, R. A. (1960) J. Chem. Phys. 33, 547-555; water, Eigen, M., and DeMaeyer, L. (1955) Z. Elektrochem. 59, 986-993. The corresponding rate constants for nicotine have not been experimentally determined. The value of kvr was estimated and compared with the experimental concentration:time values for nicotine deposition in the denuder tube runs. The rate constant values used here are estimates based upon the values for ammonia and for imidazole (Eigen, M., Hammes, G. G., and Kustin, K. (1960) J. Am. Chem. Soc. 82, 3482-3483) and other nitrogen compounds cited in the preceding references and in Eigen, M., and DeMaeyer, L. (1963) In Technique of Organic Chemistry, vol 8, part 2, chapter 18, Interscience, New York. The forward rate constant for water has been multiplied by the concentration of water, 55.56 M. c Absolute values of the volatile loss rate constants are arbitrary. Their relative values are estimated from vapor pressure, gas solubility measurements, and gas diffusivity; see Part A, Supporting Information.

nicotine. The goal has been to simulate by chemical kinetics the manner in which a solution of weak acids and bases can mutually affect their volatile loss from solution. Assumptions made in the construction of the model will be described as they arise. The model is limited to one narrow but fundamental aspect of a dynamic chemical system; it does not purport to represent the entire complexity of events in the life of a smoke aerosol particle. The chemical transformations and volatile loss steps are shown pictorially in Figure 1 and in detail in Tables 1 and 2. The model is expressed mathematically by the set of simultaneous rate equations defined as shown in Table 1, where it can be seen there are nine species linked by hydrogen ion in the mechanism. The three solute components can each be present in two conjugate acidbase forms within the pH range 5.5-6.5 corresponding to the pH range of aqueous extracts of MS smoke of commercial blended cigarettes (14, 30). This pH range is an indication of the relative molar concentrations of water-soluble acids and bases extractable from smoke total particulate matter and as such is a reasonable starting point for these simulations. The volatile loss of each component is linked to the others because hydrogen ion is involved, along with water, in all of the component ionization reactions. The model mechanism is constructed to reflect the fact that smoke aerosol particles are the sole source of all compounds that can affect nicotine loss. The amount of water in the aerosol particles was assumed to reach a steady state in the initial state of the system. Acetic acid is the most abundant acid in MS smoke, followed by formic acid (25-50 wt %) and far lesser amounts of other carboxylic acids (see Supporting Information (SI), Table SI-1, for a listing of organic acids and their relative amounts in MS smoke) (19, 20, 54-57). Acetic and formic acids are volatile enough to undergo significant loss from the smoke particle. Each component added to the model introduces three rate constants: two associated with forward and reverse ionization and one for the volatile loss. Among the acids most likely to have a significant effect on the progress of the simulation, experimentally determined values for these parameters have been reported only for acetic acid. Acetic acid was therefore used to represent all volatile acids in the model in most of these simulations, although a series of simulations using formic acid were performed as well. Simulation trials were performed to examine the sensitivity of the model to variation of the volatility and forward and reverse ionization rate constants for formic and acetic acids, as discussed below.

The initial state of the model aerosol particle was prepared by “adding” the conjugate uncharged form of each component to water. All charged species are therefore formed by ionization or protonation of the components, including solvent water. The initial set of species concentrations was calculated under the assumption of an ideal solution. Conventional chemical kinetics practice was followed by writing the rate equations in terms of concentrations. At time zero in the simulation, the volatile loss steps are started and the full set of rate equations in the second column of Table 1 are required to represent the system as it changes with time. Values used for equilibrium and rate constants used in the model are listed in Table 2. The procedure followed to estimate the set of first-order volatile loss rate constants in the last column of Table 2 is presented in the Supporting Information, Part A. The rate constants for the forward and reverse nicotine ionization steps have not been evaluated experimentally. Subject to the requirement that the ratio of the forward to reverse rate constants in an elementary process must equal the thermodynamic dissociation constant, values for the two nicotine rate constants were estimated that are representative of organic nitrogen bases (references cited in Table 2). The step in which each uncharged conjugate form is lost from the solution was represented as an apparent first-order process. Loss of volatiles to the gas phase was assumed to be effectively irreversible. This approximation would apply if the gas phase is sufficiently dilute in volatiles. Researchers have estimated the percentage of nicotine in its various protonated forms in smoke particles by using either the pH of an aqueous extract of smoke (22, 58-60) or by studying the concentration of nicotine in the headspace over MS smoke particulate matter deposited onto the walls of a Teflon bag (16). The relationship between the pH of such a solution and hydrogen-ion activity in the interior of a smoke aerosol particle and “effective pH of particulate matter” is at present unknown. Nonetheless, the relative amounts of the conjugate acid-base states of nicotine and of all ionizable components are controlled by hydrogen activity, whatever the environment. The relative values for the volatile loss rate constants in Table 2 are assumed to be proportional to the corresponding relative volatility estimates based upon vapor pressure measurements and activity coefficients of the compounds in dilute aqueous solution. The relative volatilities were in turn estimated using partition coefficients in aqueous solution and the gas diffusion coefficients in air (Part A, Supporting Information). Although loss of each volatile species is represented in the model


as an apparent first-order process, the observed time course of component loss does not, in general, appear as first order because of the mechanistic interdependence of all linked rate processes in the model mechanism. No attempt was made to include surface area effects upon the rates of volatile loss. The model in Table 1 was modified in a separate series of simulations to include a special case reported in the literature (6) in which external ammonia is added to the component mixture at a steady rate during the volatile loss. The program MathCad (MathSoft, Inc.) was used to calculate the initial solution composition for each simulation. The set of rate equations in Table 1 were converted to symbolic representation (61) and solved by numerical integration using the program VisSim (Visual Solutions, Inc.). The loss of volatiles was then started, defining time zero. The subsequent time course of the system was completely determined by the rate equations in Table 1, the parameter values in Table 2, and the choice of initial component concentrations. The integrity of the calculation was monitored by testing continually for conservation of mass and net charge. The program produced a record of all species concentrations and rates of change as the simulation proceeded. The accumulated simulation records were subsequently pooled to provide the analysis in the following section. B. Analysis of Simulation Results. (1) Time-Dependent Composition of the System as It Responds To Variation in Initial Ammonia Concentration. All of the model simulations presented herein have an initial nicotine concentration of 0.4 M. The concentration of nicotine in the solution (i.e., the simulated aerosol particle phase) is based upon a weight ratio in MS smoke of 1 mg of nicotine/16 mg of total particulate matter (45, 62-64), with an estimate of aerosol particle density of 1 g/cm3. Experimentation with variable amounts of acetic acid showed that it is necessary to start with 0.65 M total acetate to extend the initial pH range as low as 5. Initial total ammonia varies from 0.01 to 0.58 M to provide an initial pH ranging from 5.0 to 8.3, which includes “pH” values attributed to sidestream, mainstream, and environmental tobacco smoke estimated by a variety of experimental methods (65, 66); see also Figure 4. Figure 5 is an example of a simulation whose initial pH is 6. The figure shows the fraction of the initial total amount of each component remaining in the solution as time passes. The general character of this plot is similar across the entire range of initial ammonia concentrations (0.01-0.58 M): almost all ammonia and about one-half of total acetate is lost in the first 3-4 s of the simulation. Much less nicotine is lost in the same time interval. The insert in Figure 5 shows the change in pH during the early part of the same simulation. The time required for complete loss of ammonia coincides with the end of the initial rapid change in pH. Volatile loss of conjugate base ammonia and conjugate acid acetate during that period causes a rise in pH to near 6.9. About 55% of initial acetate and 5% of total initial nicotine are lost in the same time interval. Figure 6A,B shows the time-dependent concentrations of acetate and ammonium ions and both conjugate forms of nicotine for an initial solution pH of 8.3, the highest value used in the simulations. Figure 6C shows the accompanying change in pH. The rapid volatile loss of ammonia within the first few tenths of a second causes

Seeman et al.

Figure 5. Fraction of each component remaining after the start of volatile loss using the simulation model. Initial conditions: pH 6, with total acetate, ammonia, and nicotine of 0.65, 0.22, and 0.40 M, respectively. The inset shows the variation of pH with time during the first 5 s of simulation; thereafter, the pH remains constant near 6.9 while the remaining nicotine and acetate are lost from the solution.

the early steep drop in pH. Acetate-ion concentration is only slightly affected in this same interval because pH . pKa for acetic acid. Therefore, there is very little volatile conjugate acid acetate present. The very early (less than 0.3 s) rapid drop in pH converts most conjugate base nicotine in the solution to its nonvolatile monoprotonated form 2, with only a small volatile loss to the gas phase. Between 0.4 and 2.5 s, the pH in Figure 6C rises as total acetate begins to decrease, caused by volatile loss of its conjugate acid. Little nicotine is lost within the same time interval, as shown by the nearly horizontal record for both nicotine conjugate forms. During this period, acetate is lost more rapidly than nicotine and the solution pH approaches 6.9 asymptotically. At about 3 s, the concentration of ammonium ion has fallen to near zero and the concentrations of acetate ion and protonated nicotine have converged to equal concentrations. During the same time interval, the concentration of conjugate base nicotine increases with rising pH to 0.02 M. When the concentrations of acetate and nicotine ions became equal, 58% of acetate and 9.5% of nicotine have been lost from the system; 95% of nicotine remaining is in the conjugate acid form. The total concentration of nicotine is slightly greater that that of acetate. Figure 6 shows that although a high concentration of ammonia increases the initial equilibrium pH of the solution, rapid volatile loss of conjugate base ammonia sharply limits its effect upon the rate of loss of the other components. After 3 s, the system becomes equivalent to a solution of nicotine acetate, a slightly volatile buffer at this pH, analogous to ammonium acetate. Once this state is reached, the subsequent rate of nicotine loss is independent of the initial ammonia content in the solution. The effect of initial ammonia concentration upon volatile loss of nicotine is summarized in Figure 7 for the full range of initial conditions studied. The pooled simulations are presented as the time required to reach fraction nicotine lost (0.3, 0.7, and 0.9) from the liquid phase, plotted vs the initial mole ratio of ammonia/



Figure 7. Time required for selected fractional loss of total nicotine as it varies with increasing total initial ammonia. The mole ratio of total acetate to nicotine is constant at 1.625. The traces for 0.3, 0.7, and 0.9 fractional nicotine loss are denoted, respectively, by open diamonds, squares, and triangles. The plots show that the rate of nicotine loss is not significantly affected by variation of initial ammonia concentration in the model simulations. The slight increases in loss times at the left of the figure arise from the mechanistic linkage among acetate, ammonia, and nicotine volatility. See Figures SI-1 and SI-2 (Supporting Information) for corresponding information on the volatile loss times for acetate and ammonia, respectively.

Figure 6. (A, B) Variation of species concentrations within the first 5 s after the start of volatile loss, using the simulation model. Initial composition is 0.65 M acetate, 0.40 M nicotine, and 0.48 M ammonia, pH 8.3 (the highest initial pH in the series). The traces represent the molar concentrations of nonprotonated nicotine, protonated nicotine, ammonium ion, and acetate ion. Note that there is little early loss of total nicotine, but the nicotine conjugate base is rapidly converted to protonated nicotine as the ammonia evaporates in the first few seconds. This is also reflected by the nearly constant pH value after ca. 2.5 s. (C) Accompanying change in pH as a function of time.

nicotine (at constant total acetate). Model simulations were made for fraction depleted at 0.1 intervals ranging from 0.1 to 0.9, but for clarity of presentation, only 0.3, 0.7, and 0.9 depletion fractions are shown in Figure 7. The corresponding figures for acetate and ammonia are in Part B of the Supporting Information. The nearly horizontal behavior of the curves in Figure 7 shows that the time required to reach a specified fractional nicotine volatile loss from the simulated aerosol particles is but slightly affected by very large increases in the amount

of ammonia initially present in the model solution. The slight increase in time required to reach a particular fraction of lost nicotine arises from changes in the linked relative rates of loss of all three volatile components at low initial ammonia concentrations while maintaining constant initial concentrations of acetate and nicotine. Volatile loss of any component will affect the loss of the other two because they are linked via hydrogen ion in the reaction mechanism (Figure 1). The effect of variation of the volatile loss rate constant for ammonia was also examined (see Part C, Supporting Information). A reduction of the ammonia volatile loss rate constant within a plausible range does not appreciably alter the rate of nicotine loss from the simulation model provided that the ammonia in the system is limited to only that which was initially present. (2) Time-Dependent Composition of the System as It Responds To Variation in Choice of Acid Used in the Model and Acids’ Volatility Rate Constants. The sensitivity of the evaporative loss of nicotine and ammonia to the acetic acid volatile loss rate constant was tested by varying values of kvah from zero (i.e., a completely nonvolatile acid) to 20/s, the latter somewhat over twice the estimate for acetic acid. This is discussed in detail in the Supporting Information, Part D. Increasing acid volatility increases the loss rate of both ammonia and nicotine. In all cases, the ammonia is lost far faster than is nicotine, though the difference becomes greater as the acid volatile loss rate constant increases. Because the volatility of acetic acid lies within the range for several tobacco smoke carboxylic acids and because acetic acid is, by weight and by mole fraction, by far the major smoke acid (see section IV.A), its use as a representative volatile acid appears reasonable.


Different acids in MS smoke not only have varying volatility, but they have different acid properties, typically described by their pKa values (see Table SI-1). Formic acid is the second most prominent smoke acid, by weight and especially by mole fraction. With each of these acids acting as the sole representative acid in the simulation, increasing the volatile loss rate constant increased the rate of acid and nicotine volatile loss. Formic acid caused a slower loss of acid and nicotine than did acetic acid. The rate of ammonia loss was also affected but ammonia was lost much more quickly that the other components. Significant amounts of nonvolatile acids will cause an increasingly acidic solution as time passes, slowing nicotine loss. The significance of each acid in a mixture will depend on the amount present. From this, it appears that the principal effect of using acetic acid as the sole representative acid in the simulation model is likely to be a faster rate of nicotine loss than would result if a mixture of acids were used. Additional details are discussed in the Supporting Information, Part D. (3) Time-Dependent Composition of the System as It Responds to a Steady Addition of Ammonia to the Particle, i.e., to the Solution Phase. Several literature procedures used for sampling tobacco smoke, including the denuder tube experiments reported herein, draw the tobacco smoke aerosol through a variety of trapping devices for subsequent analysis (24, 30, 65, 67). In a recent publication (6), environmental tobacco smoke (ETS) was trapped on a filter pad and the resulting thin film on the filter was subsequently purged with a gas stream. This gas stream was humidified nitrogen containing varying concentrations of ammonia. The authors assumed in their analyses that the filter pad could be regarded as an aqueous system at equilibrium (6). This is reasonable because the nitrogen stream is likely to reach near saturation with respect to water vapor. The aqueous approximation is made in the model in Table 1 and by Ingebrethsen (8). The rate of desorption of nicotine from the filter was used by Pankow et al. to estimate a gas-particle equilibrium distribution of nicotine in the ETS aerosol that was originally passed through the filter pad (6, 67) and by extension, to calculate the maximum increase in volatility of the nicotine from the ETS particles (6). To simulate the steady addition of ammonia, the model in Table 1 was modified to include a step in which ammonia was added at a constant rate. This modification appears as an additional additive term in the rate equation dNH3/dt, expressible as moles of ammonia added per liter per second. Simulations with the modified model start with the steady addition of ammonia at the same time volatile loss begins. Figure 8 shows the results of one such simulation in which external ammonia is added at a rate of 0.15 M/s to the same initial state represented by Figure 5. For the conditions in Figure 8, acetate loss is slowed and nicotine loss is increased relative to the simulations in Figures 5-7. The fraction of ammonia reaches a near steady state level after about 5 s in Figure 8. Simulations of this modified model at higher rates of ammonia addition can cause ammonia concentration to exceed that of the initial equilibrium state. The asymptotic state of the system is an aqueous solution of ammonia, all acetate and nicotine having been irreversibly purged. In agreement with the published experimental results (6, 7), the “filter pad model” simulations show that sustained elevated ammonia levels can increase the rate of nicotine loss from the system.

Seeman et al.

Figure 8. Effect of steady ammonia absorption from surroundings into the model solution. The simulation used for Figures 5-7 has been modified to model the experimental studies of Pankow et al. (6). Ammonia is added at a constant rate of 0.15 M s-1. The three traces in the main body show the fraction of total original nicotine (light dashed line), acetate (solid line), and ammonia (dark dashed line) remaining as time passes, starting from the same initial equilibrium state (pH 6) used for Figure 5. The inset shows the corresponding variation of pH with time. Ammonia concentration in the system decreases to a near steady-state level, while nicotine and acetate are depleted by volatile loss. For the conditions shown, the rate loss of acetate is decreased and that of nicotine is increased, caused by the higher pH relative to that in Figure 5. The detailed character of the curves is strongly dependent upon the relative rates of ammonia addition and loss. The final state of this model simulation is a solution in which the only components are ammonia and water.

There are a number of fundamental reasons why the desorption method with trapped tobacco smoke and ammonia in the desorption gases (6) is not a suitable representation of volatile loss from newly formed tobacco smoke aerosol particles: (i) The filter desorption experimental procedure incorporated the addition of ammonia at a constant rate from an unlimited source, quite different than MS smoke formation. (ii) Deducing an equilibrium partition coefficient Kp from rate measurements is questionable. (iii) The trapped, aged smoke particulate matter was likely in the form of a “thin film” on a filter or pad. MS smoke is, of course, a dynamic aerosol. (iv) The desorption technique is based on a theoretical treatment (68) for nonreacting desorbing systems. In the experiments, ammonia from the desorption gases reacts with nicotine salts, e.g., 2 and possibly diprotonated nicotine salts 3, on the filter (6). Because the theoretical basis of the desorption method specifically excludes its application to reactive systems (68), the calculated values of percent nicotine in the gas phase using this procedure may not be accurate. As Pankow et al. based their conclusions regarding MS smoke on an apparent increase in percent nicotine in the gas phase in the presence of a continuous influx of ammonia in a reacting desorbing system of environmental tobacco smoke deposited thin film, the validity of the Pankow et al. conclusions (6, 7) is suspect. Other criticisms of the use of these Pankow et al. conclusions have appeared in the literature (69). (4) Experimental Evidence for Evaporation of Acids from both MS Smoke Particles and Environmental Tobacco Smoke Particles. Total particular matter (TPM) of tobacco smoke is collected on a Cam-



Table 3. Comparative Smoking Studies Using the Marlboro Lights Cigarette with Both a Borgwaldt Smoking Machine under Modified ISO Conditions and the Reverse Flow Smoking Setup without the Denuder Tubea smoking procedure single port Borgwaldt smoking machine

reverse flow smoking without denuder tube

test

ammonia (mg/cig), n)3

nicotine (mg/cig), n)3

solanesol (mg/cig), n)3

ammonia (mg/cig), n)1

nicotine (mg/cig), n)2

solanesol (mg/cig), n)2

1 2 3 average

0.0035 0.0034 0.0042 0.0037

0.93 0.80 0.96 0.90

0.192 0.202 0.236 0.210

0.0038 n/d n/d 0.0038

0.66 0.86 n/d 0.76

0.177 0.170 n/d 0.174

a In both sets of experiments, Cambridge filters were used to trap the total particulate matter (i.e., for the nicotine and solanesol analyses). For ammonia, an impaction trap containing aqueous hydrochloric acid was used.

bridge filter pad placed downstream of a puffing cigarette using the FTC (and ISO) smoking paradigms (34, 45). In this work, PTFE filters were used because an unloaded Cambridge filter led to a pH shift to pH 9. Figure 4 provides the pH values of aqueous extracts of TPM of a series of MS, SS, and ETS samples from a filter cigarette having an FTC “tar” yield of 7 mg. To confirm that the trapping capacity of the filter pads was not overwhelmed, pH values were measured for a series of smoke experiments in which varying amounts of TPM were collected on the filters. As can be seen from Figure 4, the pH values of aqueous extracts of MS smoke, SS smoke, and ETS are all independent of the amount of TPM collected. The pH measurements may provide an indication of the relative molar concentrations of water-extractable acids and bases trapped on the filter pads. For these cigarettes, the pH of the aqueous extracts of MS smoke and ETS TPM are essentially the same, slightly acidic, while that of SS smoke is slightly basic. The basic pH of aqueous extracts of SS smoke is likely due to the high concentration of SS smoke ammonia. The SS/MS ratio of ammonia is 40-170:1 (1). There is very little nicotine in the TPM of ETS, consistent with the well-known observation that >95% of nicotine in ETS is in the gas phase (1, 70). Certainly >95% of the far more volatile ammonia in ETS is also in the gas phase. Indeed, the simulation model predicts that ammonia, nicotine, acetic acid, and other volatile and semivolatile components of ETS smoke particles will volatilize and transfer to the gas phase of ETS. The pH of an aqueous extract of ETS TPM is almost neutral (pH 6.4). These three experimental observations, i.e., the near neutral pH of aqueous extracts of ETS TPM, the similarity of the pH of aqueous extracts of MS smoke and of ETS, and the near total evaporative loss of nicotine and other bases from ETS particles, suggest that both acids and bases have evaporated from the ETS particles, as predicted by the model simulations.

V. Experimental Results and Discussion A. Denuder Tube Results and Discussion. A “firstgeneration” set of experiments were performed by Lewis et al. (23, 24), Mariner and Frost (33), by initial studies in the laboratories of one of us [J.-J.P.] (25-27), and in one set of the recently published experiments by Ingebrethsen et al. (8). These experiments were focused on tobacco smoke aerosol using either a “continuous puff” or puffs at flow rates through a denuder tube considerably lower than the flow rates utilized in the standard smoking methods. This decreased flow rate was thought to be required to ensure a laminar flow and because of practical limitations on the tube length. In addition, these

studies are unrealistic models of actual cigarette smoking, as smokers inhale individual puffs at higher flow rates. To achieve the objectives of this work and to meet the requirements listed in section III.A, the smoking system shown in Figure 2 was developed. Two sets of experiments were performed (i) to assess if the positive pressure smoking system affected MS smoke deliveries of the three compounds of interest (ammonia, nicotine, and solanesol), and (ii) to determine if there was any contamination of the MS smoke aerosol by sidestream smoke ammonia. First, the 6 mg “tar” Swiss-manufactured Marlboro Lights cigarette was smoked under standard ISO conditions using a Borgwaldt RM1 single-port smoking machine inside a ventilation hood with the air velocity around the cigarette at 200 mm/s at 22 °C. Second, the same cigarette was smoked using the system shown in Figure 2 with the sole exception that the denuder tube was omitted. As can be seen in Table 3, the positive-pressure smoking system did not substantially alter the MS smoke yields of nicotine, ammonia, or solanesol. While only a single determination of ammonia was performed in the reverse flow experiment, the value of ammonia observed was almost identical to that found using the normal flow Borgwaldt machine, and these values were essentially the same as those found during runs 1 and 2 described immediately below. Hence, the system was judged to be capable of providing fresh, undiluted MS smoke to the denuder tube. Following completion of the design stage, five smoking “runs” were performed: run 1 (smoking 3 cigarettes), run 2 (smoking 10 cigarettes) (see Table 4), and runs 3-5 (smoking 10 cigarettes in each run) discussed below. In runs 1 and 2, ammonia, nicotine, and solanesol on the denuder tube were each quantified in four different denuder tube sections: 0-10, 10-40, 40-70, and 70120 cm. Nicotine and solanesol were quantified on the Cambridge pad in run 1, and ammonia and nicotine were quantified on the Cambridge pad in run 2. An extremely low percentage of solanesol was deposited on the tube (95% of the ammonia and nicotine that deposited on the tube in runs 1 and 2 was due to gas-phase deposition rather than particle deposition. The cumulative amounts of gas-phase deposition of ammonia and nicotine are shown in Figure 9A,B. The apparent loss of nicotine by gas-phase evaporation from the particles is rather constant throughout the range of ammonia depletion, as shown by the nearly linear


Seeman et al.

Table 4. Determination of Ammonia, Solanesol, and Nicotine from the 6 mg “tar” Swiss-manufactured Marlboro Lights Cigarettes Using the Denuder System particle phase denuder deposition (mg/cig) deposition (mg/cig) tube run segment ammonia nicotine solanesol ammonia nicotine

e

gas-phase deposition (mg/cig)

cumulative gas-phase deposition (mg/cig)

cumulative gas-phase deposition (% of total)

ammonia nicotine

ammonia

nicotine

ammonia

nicotine

1a

0-10 10-40 40-70 70-120 filter total

0.00207 0.00207 0.00063 0.00047 n/de n/d

0.0084 0.0072 0.0046 0.0074 0.6300 0.6576

0.00008 0.00007

On the Deposition of Volatiles and Semivolatiles from Cigarette Smoke

Recommend Documents