Anal. Chem. 2002, 74, 3492-3497
Picogram Quantitation of Polycyclic Aromatic Hydrocarbons Adsorbed on Aerosol Particles by Two-Step Laser Mass Spectrometry M. Kalberer, B. D. Morrical, M. Sax, and R. Zenobi*
Department of Chemistry, Swiss Federal Institute of Technology (ETH) Zu¨rich-Ho¨nggerberg, 8093 Zu¨rich, Switzerland
Polycyclic aromatic hydrocarbons (PAHs) are emitted into the atmosphere mostly by anthropogenic combustion sources. Because of their carcinogenic and mutagenic properties, PAHs are often analyzed in air quality measurements. Atmospheric concentrations of PAHs, typically in the nanograms-per-cubic-meter range, require significant effort for sample collection and processing when conventional methods such as gas chromatography/mass spectrometry (GC/MS) or liquid chromatography/mass spectrometry are used. In contrast, two-step laser mass spectrometry (L2MS) is highly sensitive and selective for PAHs and requires almost no sample preparation. Here, we present for the first time a method based on L2MS to quantify PAHs adsorbed on aerosol particles collected on a filter. Linear ranges for quantitation were determined for five different PAHs in the mass range of 178-276 Da (i.e., phenanthrene, pyrene, chrysene, benzo[e]pyrene, benzo[ghi]perylene) covering more than 2 orders of magnitude with detection limits between 50 and 300 pg of a single PAH on a whole filter sample. A quantitative comparison with GC/MS was performed using model aerosols consisting of benzo[e]pyrene adsorbed on inorganic salt aerosol particles. On average, 25% less benzo[e]pyrene was determined with GC/MS than with L2MS, with a variability between the two methods of (68%. The general lower amount measured with GC/MS is attributed to losses during the sample preparation for the GC/MS measurements. Polycyclic aromatic hydrocarbons (PAHs) present in the atmosphere have attracted attention for several decades because of the highly carcinogenic and mutagenic properties of some of these compounds (see Finnlayson-Pitts and Pitts for a comprehensive review).1 PAHs are mostly formed by combustion processes and emitted into the atmosphere by a large variety of anthropogenic and biogenic sources, such as diesel and gasoline cars, biomass fires, or cigarettes. PAHs are distributed in the atmosphere between the gas phase and the aerosol phase (solid or liquid). The distribution is mostly governed by the vapor pressure of the compound, and therefore, small 2- and 3-ring PAHs * To whom correspondence should be addressed. E-mail: zenobi@ org.chem.ethz.ch. (1) Finlayson-Pitts, B. J.; Pitts, J. N., Jr. Chemistry of the Upper and Lower Atmosphere: Theory, Experiments, and Applications; Academic Press: San Diego, 2000.
3492 Analytical Chemistry, Vol. 74, No. 14, July 15, 2002
are predominantly found in the gas phase, whereas 4- and 5-ring PAHs are mostly found in the aerosol phase.2,3 PAHs are usually analyzed by chromatographic methods, such as gas or liquid chromatography, which often require a timeconsuming sample preparation. Because of low concentrations of PAHs in the ambient atmosphere (in the nanograms-per-cubicmeter range), these analytical methods typically require the collection of tens to hundreds of cubic meters of air4-7 in order to get detectable signals. This, however, greatly limits the time resolution at which PAHs can be monitored in the atmosphere. Such long sampling times also enhance the possibility of severe sampling artifacts, such as gas adsorption on the filter material, blow off, or heterogeneous reactions of compounds collected on the filter. These artifacts usually scale with the volume of sampled air and sampling time so that a short sampling time is preferable. Two-step laser mass spectrometry (L2MS) offers significant advantages, because it has been shown to be orders of magnitude more sensitive for PAHs than, for example, GC/MS. L2MS has been used for several years to analyze aromatic compounds, such as polycylic aromatic compounds, in complex matrixes, such as rocks, sediments, aerosols, or extraterrestrial material.8-13 Haefliger et al.11 showed that in an urban environment, measurements with a time resolution of 15 min (equal to 0.15-0.3 m3 of sampled air) can be achieved with L2MS. Thus, highly dynamic features of PAH emissions became visible, which other methods could not detect because of the much longer sampling times needed. In addition, L2MS requires almost no sample preparation, (2) Fraser, M. P.; Cass, G.R.; Simoneit, B. R. T.; Rasmussen, R. A. Environ. Sci. Technol. 1998, 32, 1760-1770. (3) Allen, J. O.; Dookeran, N. M.; Taghizadeh, K.; Lafleur, A. L.; Smith, K. A.; Sarofim, A. F. Environ. Sci. Technol. 1997, 31, 2064-2070. (4) Pio, C.; Alves, C.; Duarte, A. Atmos. Environ. 2001, 35, 389-340. (5) Alves, C. A.; Pio, C. A.; Duarte, A. C. Environ. Sci. Technol. 2000, 34, 42874293. (6) Hildemann, L. M.; Mazurek, M. A.; Cass, G. R.; Simoneit, B. R. T. Aerosol Sci. Technol. 1994, 20, 303-317. (7) Rogge, W. F.; Mazurek, M. A.; Hildemann, L. M.; Cass, G. R.; Simoneit, B. R. T. Atmos. Environ. 1993, 27, 1309-1330. (8) Mahajan, T. B.; Plows, F. L.; Gillette, J. S.; Zare, R. N.; Logan, G. A. J. Am. Soc. Mass Spectrosc. 2001, 12, 989-1001. (9) Ghosh, U.; Gillette, J. S.; Luthy, R. G.; Zare, R. N. Environ. Sci. Technol. 2000, 34, 1729-1736. (10) Gillette, J. S.; Luthy, R. G.; Clemett, S. J.; Zare, R. N. Environ. Sci. Technol. 1999, 33, 1185-1192. (11) Haefliger, O. P.; Bucheli, T. D.; Zenobi, R. Environ. Sci. Technol. 2000, 34, 2184-2189. (12) Dale, M. J.; Downs, O. H. J.; Costello, K. F.; Wright, S. J.; Langridgesmith, P. R. R.; Cape, J. N. Environ. Pollut. 1995, 89, 123-129. (13) Zenobi, R.; Philippoz, J. M.; Buseck, P. R.; Zare, R. N. Science 1989, 246, 1026-1029. 10.1021/ac011233r CCC: $22.00
© 2002 American Chemical Society Published on Web 06/18/2002
Figure 1. Schematic setup of the PAH generating apparatus: (a) synthetic air input; (b), (c) gas flows regulated using a critical orifice and mass flow controllers, respectively; (d) generation of (NH4)2SO2 and NaCl aerosols using a nebulizer (0.5% aqueous solution); (e) diffusion dryer with silica gel; (f) round-bottom flask with 10 mg PAH; (g) heated transfer line; (h) aerosol-PAH mixing chamber; (i) diffusion denuder with activated char coal; (j) aerosol filter; and (k) to pump (flow rate: 2 L/min).
allowing measurements to be performed within minutes and minimizing potential sample contamination. A disadvantage of the L2MS method is that until now, quantitative measurements were not possible, and only the relative abundance of different PAHs could be measured. In this paper, we present a method to carry out quantitative L2MS measurements with the addition of an internal standard by electrospray deposition. EXPERIMENTAL SECTION L2MS Measurement. The measurements described here were performed with our home-built L2MS system, which has been described previously,14 so that only a short description will be given. The sample introduction device of the instrument is optimized for filter samples, allowing a quick mounting and introduction of the sample into the ion source region of the timeof-flight MS (TOFMS). A CO2 laser (Alltec 853 MS, Lu¨beck, Germany) with a wavelength of 10.6 µm was focused to a spot of ∼150 µm in diameter, causing desorption of the sample material from the filter substrate. With a delay of 10 µs, an optical parametric oscillator laser (MOPO-730D10, Spectra Physics, Mountain View, CA) was fired to ionize the desorbed material at an ionization wavelength of 266 nm. After a measurement, the filter was rotated by 5° so that another measurement could be performed on a fresh spot. The CO2 laser desorption spot was on the center axis of the instrument, whereas the sample holder that rotated the filter was mounted off-axis. A total of 72 measurements (72 × 5° ) 360°) on a ring with 9-mm diameter could be acquired from one filter. Usually spectra from one filter were averaged to account for inhomogeneities in the sample. Generation of Model Aerosol Particles. L2MS measurements were compared with GC/MS measurements using model aerosols generated in a flow tube system. Niessner described a similar setup to generate PAH-coated aerosol particles for heterogeneous reaction studies.15 The salt aerosol particles were generated by nebulizing a (NH4)2SO4 and NaCl aqueous solutions with N2 (see Figure 1). After drying the particles in a diffusion denuder (14) Voumard, P.; Zhan, Q.; Zenobi, R. Rev. Sci. Instrum. 1993, 64, 2215-2220. (15) Niessner, R. Sci. Total Environ. 1984, 36, 535-562.
filled with silica gel (relative humidity downstream from the diffusion denuder ∼ 5%), the aerosol particles were mixed with a small gas stream containing a gaseous PAH, in the present study benzo[e]pyrene. An ∼10-mg portion of benzo[e]pyrene was placed in a 50-mL three-neck round-bottom flask that was heated in an oil bath to 90 °C. The evaporated benzo[e]pyrene was carried in a N2 stream through heated tubing to a glass mixing tube, where it was combined with the salt aerosol. This results in a sudden cooling of the gaseous PAH flow and an adsorption of benzo[e]pyrene onto the salt aerosol particles. After the mixing chamber, the aerosol flow passes a second diffusion denuder filled with activated charcoal to remove any benzo[e]pyrene remaining in the gas phase. The aerosol particles were then sampled on a Teflon-coated quartz fiber filter of 15-mm diameter. For the L2MS measurements, the samples were treated as described below; for the GC/MS measurements, the filters were extracted in 0.5 mL of toluene for 30 min in a sonication bath. After extraction and internal standard addition, the solvent volume was reduced to ∼100 µL by a gentle flow of N2. A 1-µL portion was then injected into the GC/MS (HP 5890/HP 5971), which was equipped with a Optima delta-6 column (dimensions, 30 m × 0.25 mm, Macherey-Nagel, Oensingen, Switzerland). The GC oven temperature was held at 50 °C for 0.1 min, then ramped at a rate of 15 °C/min to 300 °C, where it was held for 15 min. The injector was held at 300 °C during the whole experiment. The mass spectra were acquired in single ion monitoring mode, that is, only m/z 212 and 252 were measured, corresponding to the molecular ion mass of the internal standard and benzo[e]pyrene, respectively. RESULTS Quantitative L2MS Method. To obtain quantitative results, a known amount of an internal standard was added to the aerosol sample. A compound has to fulfill several requirements to be suitable as an internal standard for L2MS: It should not affect the analysis of the sample. The compound has to be ionizable at the wavelength chosen, in our case, 266 nm. It has to have a sufficiently low vapor pressure that it does not evaporate from the sample during the measurement time of ∼5 min and a pressure of ∼(1-2) × 10-6 Torr. The compound should have a molecular mass for which the interference with sample compounds is minimal. Earlier work16 has shown that in a large set of emission sources, including gasoline and Diesel cars, residential oil, gas, and wood heating systems, open wood fires, and environmental tobacco smoke, as well as from ambient aerosol samples, the L2MS spectrum shows a remarkably consistent gap between m/z 211 and 214, with little to no ion signal. Di-isopropylnaphthalene (DIPN), with a mass of 212 Da, a compound that fulfills all the above requirements, was chosen as the internal standard for the quantitative PAH analysis. Another critical requirement is that the internal standard has to be distributed on the sample very evenly in order to get quantifiable results. Since the desorption spot of the CO2desorption laser has a diameter of ∼150 µm, the spatial variability of the deposited internal standard has to be smaller than the diameter of the desorption spot. Furthermore, the internal standard must be deposited in solid form onto the sample. A liquid layer would be difficult to maintain (16) Haefliger, O. P.; Bucheli, T. D.; Zenobi, R. Environ. Sci. Technol. 2000, 34, 2178-2183.
Analytical Chemistry, Vol. 74, No. 14, July 15, 2002
3493
Figure 2. Schematic representation of the electrospray coating apparatus. A 100-µL portion of the internal standard solution is sprayed onto the filter by an electrospray needle (a) (100 µm i.d.; applied potential, 3.8 kV). The methanol spray droplets (b) are so small that almost all solvent is evaporated when the droplets reach the sample (c). The filter holder is spinning (d) to ensure an even thickness of internal standard layer on top of the aerosol sample.
in the vacuum during the measurement, and it could also partially dissolve the sample material and, thus, disturb its spatially uniform deposition on the filter, which would make a quantitative measurement impossible. By depositing the internal standard with an electrospray method onto the filter, these constraints for the addition of an internal standard could be fulfilled. An aerosol-loaded filter of 15-mm diameter (Teflon-coated quartz fiber filter, Pall Gelman, Ann Arbor, MI) was mounted onto a spin-coating motor (see Figure 2). A stainless steel electrospray needle with a 100-µm i.d. (Agilent, Palo Alto, CA) was positioned ∼9 mm above the filter and was connected to a syringe containing the internal standard solution (typically a methanol solution of 1 µg DIPN/mL methanol). By means of a syringe pump, 100 µL of internal standard solution was sprayed onto the sample at a flow rate of 10 µL min-1 while a potential of 3.8 kV was applied to the electrospray needle. Flow rate, the needle tip-filter distance, and the applied voltage were the three main parameters that had to be adjusted for an optimal spray. The quality of the spray could
be judged visually by the slight coloration of the sprayed area on the filter and filter holder. Spray conditions were chosen such that the sprayed area was ∼50% larger than the filter area (and thus, only ∼50% of the sprayed material was deposited onto the filter) in order to avoid uneven deposition patterns that occurred at the edge of the spray cone, thus affecting the measurement. An occasional cleaning of the needle in a sonication bath was the only maintenance necessary. 2,5-Dihydroxy benzoic acid (DHB) (6 mg/mL) was added to the internal standard solution. This allowed for verification of the deposited amount of internal standard by weighing the sample before and after the addition of the internal standard. The deposited mass was ∼300 µg in order to achieve a measurement precision of 1%. Assuming an even deposition on the filter, this results in a film thickness of 1.5 µm. This is a lower limit, since DHB most likely forms an amorphous layer of crystallites on the sample. Earlier studies showed that the desorbed volume of CO2laser shots with intensities used here (5 mJ/shot) is a crater ∼3050 µm deep.17,18 Thus, a thicker DHB-internal standard layer should not prevent the underlying aerosol sample from being efficiently desorbed. To verify that no such interference was occurring, a test was conducted in which two layers of DHB solutions were sprayed on top of each other onto a filter, with each layer containing a different PAH. The L2MS measurement of this two-layer sample showed that the PAH of the underlying layer was visible with the same relative signal intensities as compared to a control experiment where both PAHs were embedded into one layer. After the internal standard deposition by electrospray, the sample was transferred immediately to the L2MS, and the measurement could be performed without further preparation steps. The total sample preparation an measurement time was ∼15 min. Calibration of L2MS. A calibration curve for quantitative determination of PAHs in aerosol samples was obtained by measuring a mixture of varying concentrations of five different PAHs in the mass range of 178-276 Da (i.e., phenanthrene, pyrene, chrysene, benzo[e]pyrene, benzo[g,h,i]perylene) together with a fixed concentration of the internal standard (see Figure 2). The solutions were deposited onto the filters by electrospray as described above. With samples onto which only DIPN was
Figure 3. L2MS mass spectrum of a standard mixture showing all five PAHs (1 ng of each compound/filter) and the internal standard (DIPN), as well as one signal from DHB and the filter material: (a) dihydroxy benzoic acid (DHB) and decarboxylated DHB, m/z 154 and m/z 110, respectively; (b) phenanthrene, m/z 178; (c) pyrene, m/z 202; (d) diisopropyl naphthalene (DIPN), m/z 212; (e) chrysene, m/z 228; (f) benzo[e]pyrene, m/z 252; (g) benzo[ghi]perylene, m/z 276; (*) filter material, m/z 226. 3494
Analytical Chemistry, Vol. 74, No. 14, July 15, 2002
Figure 4. (a) Signal intensity of pyrene (9) and internal standard, diisopropyl naphthalene (+), as a function of UV laser energy. In a range below 30 µJ/UV laser pulse, the signal intensity increases linearly with laser energy. Above that energy (indicated by the dotted line), fragmentation of the PAHs occurs, resulting in a nonlinear behavior of the signal intensity (small symbols). (b) The ratio of pyrene/internal standard is constant in the range below 30 µJ. Above that value, this constant ratio is no longer valid, as indicated by the equations of the linear regressions. Thus, only measurements below 30 µJ/UV laser pulse were included in the data analysis.
deposited, it was verified that the internal standard did not fragment and, thus, did not contribute to artifacts on real samples. Note that mass m/z 154 is due to DHB, which is ∼105 times more abundant in the samples than the PAHs, but only very poorly ionized. A peak at m/z 226 was always observed, even after thorough cleaning of clean blank filters in a sonication bath with different solvents. Thus, this peak is probably due to filter material of the Teflon coated quartz fiber filter. The pulse energy of the UV ionization laser had to be kept in a range between 10 and 30 µJ. At higher pulse energies, significant fragmentation of the PAHs was observed, whereas the degree of fragmentation was different for the five different PAHs investigated (Figure 4a). Thus, outside this range, the ratios between the PAHs and DIPN were no longer constant with increasing UV laser energy, as can be seen in Figure 4b. Between 10 and 30 µJ, (17) Morrical, B. D.; Zenobi, R. Atmos. Environ. 2002, 36, 801-811. (18) Haefliger, O. P.; Zenobi, R. Rev. Sci. Instrum. 1998, 69, 1828-1832.
the ratio of PAH and internal standard is constant, as can be seen from the linear regression shown in Figure 4b, whereas this is no more valid for UV laser energies >30 µJ. Shot-to-shot variability of the ionization laser was considerable (20-50%), with a mean intensity of 20 µJ. On average, 15-20% of the measurements were outside the defined range. Mass spectra obtained with a UV laser energy outside that range were not included in the data analysis. In contrast, the desorption laser was very stable, with shot-toshot variability 2 orders of magnitude (Figure 5). The total amount of PAHs deposited on the filters ranged between 40 and 0.2 ng for the calibration curves shown. The limit of detection (LOD) for the five PAHs was between 50 and 300 pg (Table 1). The LOD was calculated as three times the standard deviation of the background ion signal at the masses corresponding to the five PAHs, measured on blank Analytical Chemistry, Vol. 74, No. 14, July 15, 2002
3495
Table 1. Limits of Detections (LOD) for Five PAHs Quantified with L2MS
compd phenanthrene pyrene chrysene benzo[e]pyrene benzo[ghi]perylene
LOD, based on practical 95% confidence absolute LODa LODc interval (ng/filter)b (pg/filter)b (amol/shot) 103 54 309 195 171
0.8 0.7 1.2 1.9 0.4
230 110 540 310 250
r2 d 0.988 0.990 0.962 0.992 0.997
a Calculated as three times standard deviation of blank filter signal (see text). b Filter area ) 1.77 cm2. c Theoretical LOD, estimated for a single desorption shot (see text). d Correlation coefficient of a linear regression, as shown in Figure 5.
Figure 5. Calibration curve for phenanthrene, pyrene, chrysene, benoz[e]pyrene, and benzo[ghi]perylene in the range of 40-0.2 ng/ filter. Solid lines are linear regression fits, and dotted lines are equal to 95% confidence intervals.
filters. The LOD based on the 95% confidence interval was between 400 and 1900 pg (Table 1). The five PAHs could be measured with a high linearity (r 2 between 0.96 and 0.99) over 2 orders of magnitude (Table 1). The highest detection limit was determined for chrysene (m/z 228), mainly as a result of the large signal present from the filter material at m/z 226 (see Figure 3) causing some interference. As can be seen from Figure 5, the values for the most heavily loaded filter (i.e., 40 ng/filter) showed significant scattering. The signal intensities of the analyte at such high concentrations were on the order of several volts, possibly causing near saturation of the detection electronics. The signal intensity 3496 Analytical Chemistry, Vol. 74, No. 14, July 15, 2002
of the internal standard, however, was the same for all calibration samples (i.e., several hundred mV) and was not affected by any detector limitations; therefore, the signal intensity ratio of analyte to internal standard for highly concentrated samples is subject to a larger scatter than at lower concentrations. Earlier work showed18 that one single desorption spot is ∼150 µm in diameter, which is ∼1/2500 of the total filter area. Taking this into account, the absolute detection limit for a single desorption shot is between 0.02 and 0.12 pg, equal to ∼110-540 amol (Table 1). The signal from one single desorption spot was always visible; thus, future work will focus on minimizing the deposition area of the aerosol on the filter to optimize the ratio of desorbed material to total collected material. Comparison of L2MS with GC/MS Using a Model Aerosol. The quantification procedure for L2MS described above was tested with artificially generated aerosol and compared with GC/MS measurements. The test aerosol was a mixed inorganic/organic aerosol with an inner core of dry (NH4)2SO4 or NaCl, respectively, and an outer layer of benzo[e]pyrene. These artificial aerosols served as models for particles in the ambient atmosphere that are often a mixture of organic and inorganic compounds. Aerosols were generated and collected using a flow tube setup described above. For these test experiments, a total aerosol mass between 150 and 2000 µg was sampled. The total amount of aerosol could be adjusted by the salt concentration in the nebulizer solution, and the fraction of benzo[e]pyrene in the aerosol was mainly controlled by the oil bath temperature and the flow rate through the PAH three-neck flask (see Figure 1). The internal standard solution was then sprayed onto the aerosol samples, as described above, and the amount of benzo[e]pyrene collected was quantified with L2MS. To compare the quantitative results of the L2MS with GC/ MS, the same filters were used for both measurements. First, the L2MS measurement was performed. The total amount of benzo[e]pyrene on the sample did not significantly decrease during the L2MS measurement, since only ∼3% of the total filter area was used for desorption in the L2MS measurements, and benzo[e]pyrene has a low enough vapor pressure that only a negligible amount evaporated during the L2MS measurement, even at 10-6 Torr. After the L2MS measurement, the filters were extracted and measured with GC/MS, as described above. Experiments to determine the extraction efficiency for the GC/MS sample preparation were performed by measuring the filter before and
after the sonication extraction with L2MS. An extraction efficiency of 85% for benzo[e]pyrene was achieved. The measurements showed that the linear ranges of the two techniques hardly overlap; that is, the upper limit for the L2MS is close to the detection limit of the GC/MS. Thus, the detection limit of benzo[e]pyrene measured with L2MS is ∼100 times lower as compared to GC/MS. In addition, techniques developed only recently, such as FT fluorescence microscopy with LODs in the range of 5-10 ng/cm2 filter area for PAHs collected on filter samples,20 are ∼10 times less sensitive than L2MS (Table 1). In absolute values, the amount quantified on six samples with GC/MS was on average 75% of the L2MS measurement, however, with a rather high variability ((68%). The lower values determined by GC/MS might be explained by incomplete extraction from the filter and losses in the sample processing for the GC/MS measurement, which are not included in the extraction efficiency experiments (e.g., losses of benzo[e]pyrene during solvent evaporation). The rather large variability between the two methods might be due to the small overlap of the linear ranges of the two methods. Larger errors occur at the high end of the L2MS calibration curves (see Figure 5), and on the other hand, measurements close to the detection limit of the GC/MS contributed to the variability between the two measurements. How(19) Morrical, B. D.; Zenobi, R. Int. J. Environ. Anal. Chem., in press. (20) Fisher, M.; Bulatov, V.; Hassan, S.; Schechter, I. Anal. Chem. 1998, 70, 2409-2414.
ever, considering the completely different techniques and the trace amounts analyzed, the agreement between the two methods is reasonable. The high sensitivity of L2MS will allow quantitative PAH measurements to be performed with a high time resolution in the ambient atmosphere. PAHs are typically present in urban environments in the low nanograms-per-cubic-meter or subnanogramsper-cubic-meter range. Thus, sampling of 0.5-1 m3 of air in a polluted environment will be sufficient for L2MS, allowing the tracking of highly dynamic situations in the atmosphere. Concentration variations of PAHs on a short time scale are likely to occur at locations near sources, such as streets, where PAH concentrations can vary rapidly with traffic density. This method will also be advantageous at remote locations using reasonably short sampling times.19 Short sampling times have the advantage of reducing sampling artifacts, such as gas adsorption on filter materials, which often scale with the sampled air volume and sampling time. Thus, a hundred times smaller sampling volume (compared to what is required for a GC/MS measurement) will also reduce such problems drastically.
Received for review December 4, 2001. Accepted April 5, 2002. AC011233R
Analytical Chemistry, Vol. 74, No. 14, July 15, 2002
3497
