Measurement of acrolein and 1,3-butadiene in a single puff of cigarette smoke using lead-salt tunable diode laser infrared spectroscopy

Spectrochimica Acta Part A 67 (2007) 16–24

Review

Measurement of acrolein and 1,3-butadiene in a single puff of cigarette smoke using lead-salt tunable diode laser infrared spectroscopy W. Dave Thweatt a , Charles N. Harward Sr. b , Milton E. Parrish c,∗ a

Philip Morris USA Post Graduate Research Program, 4201 Commerce Road, Richmond, VA 23234, USA b Nottoway Scientific Consulting Corporation, P.O. Box 125, Nottoway, VA 23955, USA c Philip Morris USA, Research Center, 4201 Commerce Road, Richmond, VA 23234, USA Received 19 April 2006; received in revised form 30 August 2006; accepted 15 October 2006

Abstract Acrolein and 1,3-butadiene in cigarette smoke generally are measured using two separate analytical methods, a carbonyl derivative HPLC method for acrolein and a volatile organic compound (VOC) GC/MS method for 1,3-butadiene. However, a single analytical method having improved sensitivity and real-time per puff measurement will offer more specific information for evaluating experimental carbon filtered cigarettes designed to reduce the smoke deliveries of these constituents. This paper describes an infrared technique using two lead-salt tunable diode lasers (TDLs) operating with liquid nitrogen cooling with emissions at 958.8 cm−1 and 891.0 cm−1 respectively for the simultaneous measurement of acrolein and 1,3-butadiene, respectively, in each puff of mainstream cigarette smoke in real time. The dual TDL system uses a 3.1 l volume, 100 m astigmatic multiple pass absorption gas cell. Quantitation is based on a spectral fit that uses previously determined infrared molecular line parameters generated in our laboratory, including line positions, line strengths and nitrogen-broadened half-widths for these species. Since acrolein and ethylene absorption lines overlap and 1,3-butadiene, ethylene and propylene absorption lines overlap, the per puff deliveries of ethylene and propylene were determined since their overlapping absorption lines must be taken into account by the spectral fit. The acrolein and 1,3-butadiene total cigarette deliveries for the 1R5F Kentucky Reference cigarette were in agreement with the HPLC and GC/MS methods, respectively. The limit of detection (LOD) for 1,3-butadiene and acrolein was 4 ng/puff and 24 ng/puff, respectively, which is more than adequate to determine at which puff they break through the carbon filter. The retention and breakthrough behavior for the two primary smoke constituents depend on the cigarette design and characteristics of the carbon filter being evaluated. © 2006 Elsevier B.V. All rights reserved. Keywords: 1,3-Butadiene; Acrolein; Carbon filters; Cigarette smoke; Ethylene; Propylene; Mid-infrared; Tunable diode laser

Contents 1. 2.

3.

4.

∗

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1. Instrumentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2. Gas sampling system and cigarette model description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1. High-resolution mid-infrared spectra of cigarette smoke . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2. Comparison of total cigarette deliveries for different cigarette models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3. Comparison of per puff cigarette deliveries for different cigarette models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4. Comparison of TDL results with other analytical methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Corresponding author. Tel.: +1 804 274 3490; fax: +1 804 274 2886. E-mail address: [email protected] (M.E. Parrish).

1386-1425/$ – see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.saa.2006.10.034

17 17 17 18 19 19 20 21 22 23 23 23

W.D. Thweatt et al. / Spectrochimica Acta Part A 67 (2007) 16–24

1. Introduction Cigarette smoke is a complex aerosol matrix consisting of thousands of constituents, with some distributed only in the gas phase, some distributed both in the gas and in the particulate phases, and others found only in the particulate phase [1]. Beginning in the late 1950s, considerable attention was focused on developing analytical methods for measuring selected constituents in cigarette smoke, including acrolein and 1,3-butadiene [2–6]. Reductions of both acrolein and 1,3butadiene in mainstream (MS) cigarette smoke are of interest in the development of potentially reduced exposure products (PREPs) [7–10]. Cigarette filters containing absorbents, such as carbon, have been shown to retain gas phase volatile constituents, including acrolein [11,12], but also will retain desirable flavor components, resulting in a smoke that some view as less sensorially satisfactory [13,14]. Generally, two separate analytical methods are needed for determining 1,3-butadiene and acrolein deliveries in cigarette smoke, as referenced by Counts et al. [9,10], that require different sample trapping, extraction, and/or derivatization techniques, chromatographic separation schemes and detection. Carbonyl-containing smoke constituent molecules such as acrolein, are derivatized with 2,4-dinitrophenylhydrazine and analyzed by HPLC. Due to the reactivity and thermal instability of some smoke constituents, derivatization can have limitations with selectivity and instability of reactive species due to the long reaction times required which often produce unstable derivatives [15]. Deliveries of volatile organic compounds (VOC), such as 1,3-butadiene, are determined by collecting the smoke with an impinger trap in either ice water or dry ice/isopropanol, then analyzing a sample of the collected organics with a GC/MS. While the carbonyl method employs one cigarette per determination, the VOC method involves smoking ten cigarettes, collecting and measuring all of the smoke, then dividing the total deliveries by 10 to obtain a per cigarette value. Generally, carbon filtered cigarettes yield smoke having lower deliveries of acrolein and 1,3-butadiene, approaching the detection limits of the HPLC and GC/MS analytical methods. While the existing analytical methods discussed above are appropriate for comparing smoke deliveries among different cigarette brands, quantitative methods and smoke sampling systems are required that can measure constituents during each puff in order to obtain a better understanding of what is occurring during the formation of the smoke. Such methods should neither alter the smoke during sampling, nor interfere with the combustion, pyrolysis, pyrosynthesis, and distillation processes during analysis. A single real time method, that captures ten or more measurements per second for both 1,3-butadiene and acrolein as the smoke leaves the cigarette filter, offers an advantage over chromatography-based per cigarette analysis methods for developing PREPs. However, the detection limit of this technique must be below 50 ng/puff in order to compare prototype models with very low deliveries of these two smoke constituents. In addition, by not using a Cambridge filter pad to separate the particulate phase from the gas phase prior to the analysis, the uncertainty introduced by reactions between the various con-

17

stituents in the particulate matter on the filter pad and reactions between gas-phase and particulate-phase constituents would be eliminated. Tunable infrared laser differential absorption spectroscopy (TILDAS) has been used for years for detecting combustion product molecules with resolvable ro-vibrational structure in the mid-IR spectral region [16]. TILDAS with lead-salt tunable diode lasers (TDLs) offers the advantage of high sensitivity, selectivity, and a fast time response relative to the previously described chromatographic methods. Due to the very narrow line width of the infrared (IR) laser source (∼0.0008 cm−1 ) and the ability to tune over several tenths of a wavenumber by rapidly ramping the current, light gases with the unique ro-vibrational lines can be detected. Furthermore, due to the small size of the aerosol particles in cigarette smoke (100–400 nm) [17,18] compared to the wavelength of the infrared radiation in the region of interest (∼10,000 nm), the IR laser light is not scattered by the aerosol particles. Finally, since the laser scans over such a small frequency range, larger molecules, which might absorb light in the same region, simply attenuate the beam due to unresolvable rotational structure and consequently do not adversely affect the measurement of the analyte of interest. The application of IR lead-salt TDL spectroscopy for the measurement of formaldehyde in single puffs of cigarette smoke was reported in 2000 [19] using a small dead volume impaction device in place of a Cambridge filter pad [20] since the pad retains ∼30% formaldehyde during smoking [1]. A more direct method of sampling cigarette smoke using our dual lead-salt TDL instrument has been reported [21], and this approach was applied to measure both acrolein and 1,3-butadiene in a single puff. 2. Experimental 2.1. Instrumentation The dual-channel TDL spectrometer system shown in the photograph in Fig. 1 has been described previously [21,22]. It was designed and built for Philip Morris USA by Aerodyne Research, Inc. (ARI, Billerica, MA) for monitoring gaseous constituents in single puffs of MS cigarette smoke. A liquid nitrogen Dewar (IR Labs, Tucson, AZ) houses two lead-salt TDLs (Laser Components, Hudson, NH) and four mercurycadmium-telluride (MCT) detectors (Kolmar Technologies, Inc, Newbury, MA). The two mid-infrared laser beams at 958.8 cm−1 and 891.0 cm−1 referenced as channels A and B, respectively, are collimated into a single astigmatic multiple pass absorption gas cell of Herriott design with a volume of 3.1 l and a path length of 100 m (ARI, Billerica, MA). Acrolein and ethylene concentrations are monitored using the 958 cm−1 TDL, while 1,3-butadiene, ethylene, and propylene are monitored using the 890 cm−1 TDL. All of these species are present in MS cigarette smoke. Ethylene was used as a reference gas since it has strong absorption lines in both spectral regions and overlaps the absorption of the constituents of interest. The signals were collected and the spectra were fit using the method previously described [22] except that the HIgh-resolution TRANsmission (HITRAN) molecule absorption database [23] does not include data on the

18

W.D. Thweatt et al. / Spectrochimica Acta Part A 67 (2007) 16–24

Fig. 1. This figure shows a photograph of the dual TDL laser system. The goldcolored Dewar in the center holds the TDLs as well as the four MCT detectors at liquid N2 temperature. There are positions for four TDLs on each side, but only one per side can be used at a time. There are two astigmatic Herriott multiple pass gas cells that can be used. The cell shown is a 36 m pathlength, 0.3 l cell volume with a 0.15 s response time for time-resolved studies. The other cell (not shown) is a 100 m pathlength, 3.1 l cell volume with a 2 s response time used in this study for increased sensitivity.

spectra of acrolein, 1,3-butadiene, or propylene. In order to measure the concentrations of these three molecules, results from previous spectroscopic studies were used [24,25]. The TDL Wintel software from ARI (Billerica, MA) was used to control the TDL system and collect the spectral data at a rate of 10 Hz. Both lasers were ramped over a small frequency (wavenumber) range by changing the current over 37.84 mA at a rate of 3284.3 Hz for the acrolein channel, and over 21.98 mA at a rate of 2580.6 Hz for the 1,3-butadiene channel. While the

acrolein spectra collected spanned 0.25 cm−1 with 328 individual spectra averaged per collected spectrum, only 0.112 cm−1 of the scan was used in the spectral analysis. Likewise, while the 1,3-butadiene spectra collected spanned 0.796 cm−1 with 256 individual spectra averaged per collected spectrum, only 0.445 cm−1 of the scan was used in the spectral analysis. TDL Wintel utilizes a frequency locking process to keep the frequency stable over time. For both the acrolein and 1,3-butadiene channels, ethylene was used as the frequency reference gas. The TDL Wintel utilizes a non-linear least squares fitting procedure with a polynomial baseline. The procedure fits Voigt line shapes calculated from selected line positions and additional molecular line parameters determined previously in our laboratory for 1,3-butadiene and propylene [24], acrolein [25], and from the HITRAN database for ethylene [23]. The non-linear least squares fitting procedure takes into account the non-linear tuning rate of the TDL, the zero light level, the temperature and pressure in the cell, and the nitrogen dilution of the sampled smoke, typically a factor of 4. In order to optimize both the efficiency of the nonlinear fitting algorithm and the limit of detection of the constituents of interest, the pressure in the multiple pass gas cell was maintained at 20.3 Torr by a pressure control system using a throttle valve located at the exit of the multiple pass gas cell. The temperature of the gas cell was maintained at 22 ± 1 ◦ C. 2.2. Gas sampling system and cigarette model description The smoke injection system has been described previously [21] and a schematic is shown in Fig. 2. The sampling system is computer controlled using valves and mass flow controllers

Fig. 2. The schematic diagram shows the smoke sampling system.

W.D. Thweatt et al. / Spectrochimica Acta Part A 67 (2007) 16–24

19

Table 1 Descriptions of the 1R5F Kentucky Reference and three prototype carbon filtered cigarettes are listed Properties

1R5F

Carbon A

Carbon B

Carbon C

Cigarette total RTD (mm of H2 O) Cigarette length (mm) Cigarette circumference (mm) Paper permeability (CU) Tobacco weight (g) Tipping paper length (mm) Filter ventilation (%) Filter length (mm) FTC butt length (mm) FTC tar (Filtrona) (mg/cigarette) FTC TPM (Filtona) (mg/cigarette) FTC Puffs/cigarette (Filtrona) FTC CO (Filtrona) (mg/cigarette) FTIR TPM (mg/cigarette) FTIR dynamic burn time (min) FTIR (puffs/cigarette) FTIR CO (mg/cigarette)

123 83.8 25.0 23 0.538 32 70 26.7 35 2.0 2.4 7.0 2.6 2.3 6.7 7.1 3.3

96 83.0 24.86 35 0.526 38 48 35 41 5.4 6.4 6.4 5.7 6.8 6.6 7.1 6.7

99 83.0 24.82 38 0.532 38 47 35 41 5.9 6.9 6.5 6.0 6.7 6.3 6.9 6.8

97 83.0 24.85 32 0.529 38 50 35 41 5.4 6.4 6.6 5.7 5.6 6.9 7.4 6.3

The physical parameters of the prototype cigarettes were designed to be similar.

interfaced to a sensitive pressure feedback system. The critical flow orifice controls the flow rate into the multiple pass gas cell at 4 l min−1 . When the cigarette is not being puffed, this flow is provided by 2.95 l min−1 from the dilution flow meter and 1.05 l min−1 from the by-pass flow meter. When a puff is taken, the by-pass flow is turned off and the 1.05 l min−1 is pulled through the cigarette rod. Thus the smoke is diluted by a factor of ∼4. Since the valves controlling the flow are not located between the cigarette and the sampling orifice, the possibility of accumulation of tars on these valves that may alter the smoke matrix prior to analysis is eliminated. The dilution of the smoke also decreases the rate at which tar accumulates in the cell and reduces the frequency of cleaning the cell. The flow rates are saved during the data collection so that the dilution can be taken into account in the post-processing of the smoke constituent concentration data. The Kentucky Reference 1R5F [26] and three different carbon filtered cigarette models were analyzed for deliveries of acrolein, 1,3-butadiene, ethylene, and propylene. The 1R5F reference cigarette (70% filter ventilation) was selected for analysis because the smoke contained low levels of acrolein and 1,3butadiene, which was desirable to evaluate the precision of the TDL per puff method. The 1R5F was analyzed each day during testing to verify that the instrument was performing properly. Also, the 1R5F total MS cigarette delivery data obtained for acrolein and 1,3-butadiene using the TDL per puff method could be compared to the results from three other laboratories at Philip Morris USA which used the VOC GC/MS method for 1,3-butadiene and the carbonyl HPLC method for acrolein. The three models differed only in the type of carbon tested, labeled A, B, and C, and were designed with the same tobacco blends, ∼48% filter ventilation, and 180 mg of carbon inserted into an 8 mm space in the cellulose acetate (CA) filter. Seven cigarettes with each type of carbon and eleven 1R5F cigarettes were analyzed. Prior to sampling, the cigarettes were conditioned for 24 h in a conditioned laboratory at 72 ± 2 ◦ F

(22 ± 1 ◦ C) and 60 ± 2% relative humidity. The cigarettes were smoked under the same conditions. Lighting was performed using a yellow flame butane lighter unless stated otherwise. The sampling system took a 35 ml square wave puff of 2 s duration with an interval of 60 s between the start of each puff. The 1R5F and carbon filtered cigarette models typically delivered eight puffs. Table 1 describes physical and smoke chemistry parameters for the 1R5F Kentucky Reference and three prototype carbon filtered cigarettes tested in this study. The physical parameters of the prototype cigarettes were designed to be similar. The MS smoke deliveries were determined using Federal Trade Commission (FTC) procedures discussed in detail [9] and also using an FTIR method described previously [27,28]. The FTIR method sampled the MS cigarette smoke by passing the whole smoke through a glass fiber Cambridge filter [20]. The smoke fraction that is collected on the filter pad is defined as the particulate phase and is referred to as total particulate matter (TPM) and the fraction that passes through the pad is defined as gas phase. The FTIR method used in research investigations delivered similar TPM data as the FTC procedures for the reference and all three experimental cigarettes. Puff counts, (TPM), tar (TPM minus nicotine and water) and CO deliveries were similar for the experimental cigarettes. The 1R5F delivered less tar at 2.4 mg/cigarette relative to the experimental cigarettes, where their tar deliveries ranged from 5.4 mg/cigarette to 5.9 mg/cigarette. 3. Results and discussion 3.1. High-resolution mid-infrared spectra of cigarette smoke Typical spectra of a puff of MS cigarette smoke (puff number one of 1R5F cigarette) for channels A and B and their fitted spectra as calculated in TDL Wintel are shown in Fig. 3. The ethylene lines appear sharper in channel B than in channel A because

20

W.D. Thweatt et al. / Spectrochimica Acta Part A 67 (2007) 16–24

Fig. 3. This figure is a screen shot from TDL Wintel of spectra collected from the first puff from a 1R5F Kentucky Reference cigarette. On the left is the spectrum from TDL A at 958 cm−1 , and on the right is the spectrum from TDL B at 890–891 cm−1 . Acrolein and ethylene spectra are seen with TDL A, and 1,3-butadiene, propylene, and ethylene are seen with TDL B. The green traces are the actual spectra, the blue traces are the simulated spectra.

the frequency range for channel B needed to be increased in order to include a propylene band located ∼0.4 cm−1 lower in frequency than the dominant 1,3-butadiene band. Inclusion of this propylene band aids in ensuring the correctness of the nonlinear fit of the measured spectrum with the fitted spectrum because the major 1,3-butadiene band overlaps another propylene band. The fitted spectra (blue traces) accurately reproduces the observed spectra (green traces) in the MS cigarette smoke matrices tested. In most cases, the laser energy is slightly attenuated by other smoke constituents which absorb in the 958 cm−1 and 891 cm−1 regions, but have unresolvable rotational structure (i.e. the molecules are “too big” to be seen using high-resolution infrared spectroscopy). Fig. 4 shows overlapped average 1,3-butadiene concentration profiles for 2 s puffs from each of the 1R5F cigarettes and three carbon filtered cigarette models analyzed. The signal response is measurable for 4 s due to the 3.1 l multiple pass cell volume and the flow rate through the cell. The concentrations of the smoke constituents of interest are measured for a total of 12 s, starting at the time when the puff begins. The signal response is integrated over the 12 s in order to calculate the amount (mass) delivered during that particular puff. Perhaps the most obvious characteristic of this figure is the observation that later puffs taken on the carbon-filtered cigarettes deliver more 1,3-butadiene than the earlier puffs. The error bars represent 1σ, and are shown for every fifth data point. The reported error bars are due to cigarette-to-cigarette variations rather than uncertainty in measurement (i.e., noise). The ninth puff from the 1R5F Kentucky Reference cigarette shows no error bars because only one of the eleven 1R5F cigarettes delivered a ninth puff.

3.2. Comparison of total cigarette deliveries for different cigarette models Many cigarettes are designed such that the paper which surrounds the filter is perforated to dilute the cigarette smoke. For example, the smoke exiting the filter of the 1R5F cigarette is

Fig. 4. This figure shows averaged real-time per-puff concentration profiles and 1σ error bars (shown for every 5th point) of 1,3-butadiene for 1R5F Kentucky Reference cigarettes (red traces), Carbon A model cigarettes (green traces), Carbon B model cigarettes (blue traces), and Carbon C model cigarettes (black traces).

W.D. Thweatt et al. / Spectrochimica Acta Part A 67 (2007) 16–24

21

Table 2 Averaged per-cigarette deliveries (␮g/cigarette) and standard deviations of acrolein, 1,3-butadiene, ethylene and propylene for 1R5F (n = 11) and the three carbon filtered cigarette models (n = 7 for each model) Units in ␮g/cigarette

Acrolein

1R5F Carbon A Carbon B Carbon C

15 22 6 1.6

± ± ± ±

2.5 4 2 0.9

1,3-Butadiene 12 11 3.2 1.3

± ± ± ±

2 3 1.1 0.9

Ethylene 62 91 78 69

± ± ± ±

10 17 10 14

Propylene 70 76 33 19

± ± ± ±

9 13 6 5

diluted by 70% with air that is pulled through the ventilation holes in the filter when a puff is taken. The carbon filtered cigarette models analyzed in this work were all designed to have 45% ventilation. By maintaining the same level of filter ventilation and altering the filter materials in the cigarette models, direct comparison of the effect on the smoke constituent deliveries can be obtained. The carbon A cigarette model was used as a control for comparing the smoke deliveries for the carbon B and carbon C cigarette models. The 1R5F cigarette was used to compare the results from the TDL method to those from the VOC GC/MS and carbonyl HPLC methods. The average total cigarette deliveries for acrolein, 1,3butadiene, ethylene, and propylene for 1R5F (n = 11) and the three carbon filtered cigarette models (n = 7 each model) are given in Table 2. The 1R5F cigarette and the carbon A cigarette model delivered similar total amounts of propylene (70 ± 9 ␮g and 76 ± 13 ␮g, respectively) and 1,3-butadiene (12 ± 2 ␮g and 11 ± 3 ␮g, respectively), but 1R5F delivered less acrolein (15 ± 3 ␮g and 22 ± 4 ␮g, respectively) and ethylene (62 ± 10 ␮g and 91 ± 17 ␮g, respectively) than carbon A cigarettes. With respect to carbon A cigarettes, carbon B reduced acrolein, 1,3-butadiene, propylene, and ethylene deliveries by 73%, 71%, 57%, and 14%, respectively. Carbon C deliveries compared to carbon A deliveries for these same constituents were reduced by 93%, 88%, 75%, and 24%, respectively. None of the carbon filtered cigarette models delivered less ethylene than the 1R5F cigarettes. It is clear from these data that ethylene deliveries are much more sensitive to differences in filter ventilation than to differences in types of carbon used in the filter. The deliveries of other constituents, especially acrolein, are sensitive to filter ventilation as well, but more sensitive to the type of carbon used in the filter [29].

Fig. 5. This figure shows the averaged per puff deliveries and 1σ error bars of 1,3-butadiene for 1R5F Kentucky Reference cigarettes (red bars), Carbon A model cigarettes (green bars), Carbon B model cigarettes (blue bars), and Carbon C model cigarettes (black bars).

delivering more acrolein and propylene than 1R5F cigarettes starting at the fourth puff, delivering more 1,3-butadiene starting at the fifth puff, and delivering less ethylene than 1R5F only in the first puff (i.e. the lighting puff). These observations are consistent with the conclusion that filter ventilation is more efficient at reducing smoke constituent levels than is carbon A after carbon A’s filtration abilities deteriorate. Carbon B models tend to show breakthrough of acrolein, 1,3-butadiene, and propylene one puff later than carbon A models (puffs 4, 4, and 3, respectively), and it delivers less ethylene than the carbon A models. Once acrolein, 1,3-butadiene, or propylene becomes detectible in a puff from carbon B models, the deliveries do

3.3. Comparison of per puff cigarette deliveries for different cigarette models Figs. 5–8 show per puff deliveries of the four different smoke constituents measured in this study. These figures show a “breakthrough” effect for all constituents except ethylene. Acrolein and 1,3-butadiene appear in measurable levels on average during puff three for the carbon A models, propylene is measurable during puff two on average, and ethylene is visible in significant quantities in every puff. In all four cigarette types, deliveries of the constituents from carbon A models increased for each puff,

Fig. 6. This figure shows the averaged per puff deliveries of acrolein and 1σ error bars for 1R5F Kentucky Reference cigarettes (red bars), Carbon A model cigarettes (green bars), Carbon B model cigarettes (blue bars), and Carbon C model cigarettes (black bars).

22

W.D. Thweatt et al. / Spectrochimica Acta Part A 67 (2007) 16–24

Fig. 7. This figure shows the averaged puff-by-puff deliveries of propylene and 1σ error bars for 1R5F Kentucky Reference cigarettes (red bars), Carbon A model cigarettes (green bars), Carbon B model cigarettes (blue bars), and Carbon C model cigarettes (black bars).

not increase from puff to puff as rapidly as the deliveries of these constituents increase from the carbon A models. Carbon C delivers measurable amounts of acrolein and 1,3-butadiene at puff 5 or 6, propylene at puff three or four, and delivers less ethylene than carbon models A and B. Furthermore, carbon C cigarette model filters do not lose efficacy to the extent that the other two carbon models lose their efficacy in the latter puffs. Propylene was observed to track the breakthrough behavior of both acrolein and 1,3-butadiene for each carbon filtered cigarette model. Further evaluations on a wide range of carbon materials confirmed this observation to be valid. This suggests that mea-

suring propylene in the smoke may be useful for predicting a carbon’s ability to retain acrolein and 1,3-butadiene for experimental carbon filtered cigarettes without having to measure these two constituents. The lighting puff deliveries of ethylene, propylene, and 1,3butadiene for the 1R5F cigarettes are much higher than for the second puff. In the cases of ethylene and 1,3-butadiene, the first puff is the highest-delivery puff. Roughly 25% of the 1,3butadiene delivery from 1R5F Kentucky Reference cigarette model is delivered in the first puff. Therefore, simply reducing the first-puff delivery of 1,3-butadiene reduces the total 1,3-butadiene delivery by 25%. Acrolein does not show this behavior, suggesting different mechanisms for acrolein and 1,3butadiene formation in cigarette smoke. Experiments were performed to determine if using a Borgwaldt electric lighter affected the lighting puff deliveries differently than using a yellow flame lighter. This information is important since smoke data generated using other analytical methods were obtained using the Borgwaldt electric lighter. The choice of lighters did not affect the first puff deliveries for 1,3-butadiene, acrolein, and propylene. Only ethylene showed a difference, with the electric lighter giving half the amount obtained when using the yellow flame lighter. 3.4. Comparison of TDL results with other analytical methods A comparison of the TDL technique with the chromatographic-based methods from three separate laboratories is shown in Table 3. These laboratories used the VOC method for 1,3-butadiene and the carbonyl method for acrolein. The agreement between the laboratories’ results for acrolein and 1,3-butadiene per-cigarette deliveries from 1R5F Kentucky Reference cigarettes and TDL technique’s results shows the efficacy of the TDL approach for measuring per-cigarette deliveries of these smoke constituents. Since the per-cigarette deliveries of acrolein and 1,3-butadiene as measured using TDL spectroscopy are sums of the individual puff deliveries, one must conclude that the per puff data measured using TDL spectroscopy are valid as well. Since the TDL technique delivers per puff information using individual cigarettes, rather than an average per-cigarette delivery measured from smoke collected from 5 or 10 cigarettes, this technique can be useful Table 3 Comparison of the averaged per-cigarette deliveries (␮g/cigarette) and standard deviations of acrolein and 1,3-butadiene for 1R5F using the TDL method and the acrolein carbonyl derivative HPLC method (1 cigarette per determination × 3 replicates = 3 cigarettes) and 1,3-butadiene VOC GC/MS (10 cigarettes per determination × 3 replicates = 30 cigarettes) method performed by three Philip Morris USA laboratories

Fig. 8. This figure shows the averaged per puff deliveries of ethylene f and 1σ error bars or 1R5F Kentucky Reference cigarettes (red bars), Carbon A model cigarettes (green bars), Carbon B model cigarettes (blue bars), and Carbon C model cigarettes (black bars).

Units in ␮g/cigarette

Acrolein

TDL Lab I Lab II Lab III

15 12 15 13

± ± ± ±

2.5 1 1.5 2

1,3-Butadiene 12 12 12 12

± ± ± ±

2 1 4 3

W.D. Thweatt et al. / Spectrochimica Acta Part A 67 (2007) 16–24

in correlating sensory data with smoke constituent delivery in an effort to produce a cigarette that delivers significantly lower levels of these smoke constituents and also is acceptable to the consumer. The relatively low per puff 1,3-butadiene and acrolein deliveries for Carbon C can be measured because of the excellent detection limit of this TDL method. The peak-to-peak noise of the instrument response between puffs during smoking for 1,3-butadiene and acrolein corresponded to 0.02 ppmV and 0.1 ppmV, respectively. Therefore, the limit of detection (LOD), calculated at 3σ and based on a puff volume of 35 ml, is 4 ng/puff and 24 ng/puff, respectively. The detection capability of the TDL method is an improvement compared to a non-quantitative, comparative GC/MS method [30]. The main advantage of the GC/MS approach is that many more smoke constituents can be monitored simultaneously in one experiment than possible using TDL spectroscopy. In that study, cigarette filters were evaluated that contained less carbon than the experimental carbon filtered cigarettes discussed herein. In that GC/MS investigation reported in 2002, the acrolein and 1,3-butadiene per puff profiles showed no breakthrough, even after eight puffs were taken. The advantage of this TDL method is that breakthrough could be measured for both constituents using experimental cigarettes of similar construction but with significantly more carbon in the filter, resulting in a smoke having even less acrolein and 1,3butadiene than the carbon filtered cigarettes used in the GC/MS evaluation. 4. Conclusions A dual TDL spectroscopy system was used successfully to measure 1,3-butadiene and acrolein in a single 35 ml puff of MS cigarette smoke, with a limit of detection, calculated at 3σ, of 4 ng/puff and 24 ng/puff, respectively. This detection limit is sufficient to determine breakthrough of propylene, acrolein, and 1,3-butadiene for prototype cigarette models with filters containing different carbon materials. Propylene was observed to track the breakthrough behavior of both acrolein and 1,3-butadiene for each carbon filtered cigarette model. Further evaluations on a wide range of carbon materials confirmed this observation to be valid. This suggests that measuring propylene in the smoke may be useful for predicting a carbon’s ability to retain acrolein and 1,3-butadiene for experimental carbon filtered cigarettes without having to measure these two constituents. The lighting puff deliveries of ethylene, propylene, and 1,3-butadiene for the 1R5F cigarettes are much higher than for the second puff. The lighting puff delivery of 1,3-butadiene represents ∼25% of the total 1,3-butadiene delivery for the 1R5F cigarettes. Acrolein does not show this behavior, suggesting different mechanisms for acrolein and 1,3-butadiene formation in MS cigarette smoke. All carbon-filtered cigarette models retained the higher 1,3-butadiene delivery in the lighting puff. The acrolein and 1,3butadiene first puff deliveries were not affected by using either a yellow flame butane or electric lighter. The total cigarette MS deliveries of 1,3-butadiene and acrolein are similar to the deliveries measured using the VOC GC/MS and carbonyl derivative HPLC methods, respectively.

23

Acknowledgements The authors would like to extend our appreciation to Dave Nelson of Aerodyne Research Inc. for computer and software discussions and Jay Fournier and Peter Lipowicz of Philip Morris USA management for their support of the high resolution infrared spectroscopy facility. References [1] R.R. Baker, in: D.L. Davis, M.T. Nielson (Eds.), Smoke Chemistry, TOBACCO Production, Chemistry and Technology, Blackwell Science, London, 1999, p. 398 (Chapter 12). [2] Tobacco and Tobacco Smoke, Studies in Experimental Carcinogenesis, in: Ernest L. Wynder, D. Hoffmann (Eds.), Academic Press, New York, 1967, 730 pp. [3] R.M. Irby Jr., E.S. Harlow, Tobacco Sci. 3 (1959) 52. [4] J.R. Newsome, V. Norman, C.H. Keith, Tobacco Sci. 9 (1965) 102. [5] V. Norman, J.R. Newsome, C.H. Keith, Tobacco Sci. 12 (1968) 216. [6] R.B. Seligman, F.E. Resnik, A.E. O’Keeffe, J.C. Holmes, F.A. Morrell, D.P. Murrill, F.L. Gager Jr., Tobacco Sci. 1 (1957) 124. [7] Clearing the Smoke: Assessing the Science Base for Tobacco Harm Reduction, Institute of Medicine, in: K. Stratton, P. Shetty, R. Wallace, S. Bondurant (Eds.), National Academy Press, Washington, D.C., 2001, 636 pp. [8] A. Rodgman, C.R. Green, Beit. Zur Tabakforsch. Int. 20 (8) (2003) 481. [9] M.E. Counts, M.J. Morton, S.W. Laffoon, R.H. Cox, P.J. Lipowicz, Reg. Toxicol. Pharmacol. 41 (2005) 185. [10] M.E. Counts, F.S. Hsu, S.W. Laffoon, R.W. Dwyer, R.H. Cox, Reg. Toxicol. Pharmacol. 39 (2004) 111. [11] J.T. Williamson, D.R. Allman, Beit. Zur Tabakforsch. Int. 2 (6) (1965) 163. [12] C.J. Kensler, S.P. Batista, N. Engl. J. Med. 269 (22) (1963) 1161. [13] Z. Yang, J.G. Nepomuceno, B. Taylor, Flavored Carbon Useful as Filtering Material of Smoking Article, U.S. Patent Publication WO 03/071886, Int. filing date (January 31, 2003). [14] L.L. Xue, K.B. Koller, Q. Gao, High Efficiency Cigarette Filters Having Shaped Microcavity Fibers Impregnated with Adsorbent or Absorbent Materials, U.S. Patent 6,584,979 (July 1, 2003). [15] A. Hakansson, K. Stromberg, J. Pedersen, J.O. Olsson, Chemosphere 44 (5) (2001) 1243. [16] M. Zahniser, D. Nelson, C. Kolb, in: K. Kohse-Hoinghaus, J. Jeffries (Eds.), Applied Combustion Diagnostics, Taylor and Francis, New York, 2002 (Chapter 26). [17] S.G. Poulopoulos, D.P. Samaras, C.J. Philippopoulos, Atmos. Environ. 35 (26) (2001) 4399. [18] M.J. Rusyniak, D.B. Kane, P.J. Lipowicz, Particle size distribution of mainstream cigarette smoke. Poster presented at the American Association for Aerosol Research, Atlanta, GA (October 4–8, 2004). [19] M.E. Parrish, C.N. Harward, Appl. Spectrosc. 54 (11) (2000) 1665. [20] W.B. Wartman Jr., E.C. Cogbill, E.S. Harlow, Anal. Chem. 31 (1959) 1705. [21] S. Plunkett, M. Parrish, K. Shafer, D. Nelson, J.B. McManus, J.L. Jimenez, M. Zahniser, SPIE Proc. 3758 (1999) 212. [22] S. Plunkett, M. Parrish, K. Shafer, D. Nelson, J. Shorter, M. Zahniser, Vib. Spectrosc. 27 (2001) 53. [23] L. Rothman, C. Rinsland, A. Goldman, S. Massie, D. Edwards, J.-M. Flaud, A. Perrin, C. Camy-Peyret, V. Dana, J.-Y. Mandin, J. Schroeder, A. McCann, R. Gamache, R. Wattson, K. Yoshino, K. Chance, K. Jucks, L. Brown, V. Nemtchinov, P. Varansasi, J. Quant. Spectrosc. Radiat. Transf. 60 (1998) 665. [24] C.N. Harward Sr., R.E. Baren, M.E. Parrish, Spectrochim. Acta Part A 60 (2004) 3421. [25] C.N. Harward Sr., W.D. Thweatt, M.E. Parrish, Poster (B-2) Presented at Fifth International Conference on Tunable Diode Laser Spectroscopy, Florence, Italy, July 11–15, 2005.

24

W.D. Thweatt et al. / Spectrochimica Acta Part A 67 (2007) 16–24

[26] H. Davies, A. Vaught, The Reference Cigarette. Kentucky Tobacco Research & Development Center, University of Kentucky, Lexington, KY, USA, January 2003. [27] M. Parrish, J. Lyons-Hart, K. Shafer, Vib. Spectrosc. 27 (2001) 29. [28] Q. Shi, D.D. Nelson, J.B. McManus, M.S. Zahniser, M.E. Parrish, R.E. Baren, K.H. Shafer, Anal. Chem. 75 (19) (2003) 5180.

[29] L. Xue, J. Nepomuceno, S. Zhuang, T. Sherwood, J. Paine, J. Fournier, C. Thomas, K. Koller, L. Yu, Activated Carbon Fiber Cigarette Filter, U.S. Patent Publication WO 03/086116, Int. filing date (April 11, 2003). [30] L. Xue, C.E. Thomas, K.B. Koller, Beit. Zur Tabakforsch. Int. 20 (4) (2002) 251.