Ultrahigh Mass Resolution and Accurate Mass Measurements as a

A program developed in-house was used to determine exact mass differences between peaks in the monomer, dimer, and trimer mass range to identify poten...
0 downloads 0 Views 294KB Size
Anal. Chem. 2007, 79, 4074-4082

Ultrahigh Mass Resolution and Accurate Mass Measurements as a Tool To Characterize Oligomers in Secondary Organic Aerosols Alain Reinhardt,† Christian Emmenegger,† Bertran Gerrits,‡ Christian Panse,‡ Josef Dommen,# Urs Baltensperger,# Renato Zenobi,† and Markus Kalberer*,†

Department of Chemistry and Applied Biosciences, ETH Zurich, 8093 Zurich, Switzerland, Functional Genomics Center Zurich, UZH/ETH Zurich, Winterthurerstrasse 190, 8057 Zurich, Switzerland, and Laboratory of Atmospheric Chemistry, Paul Scherrer Institut, 5232 Villigen PSI, Switzerland

Organic aerosols are a major fraction, often more than 50%, of the total atmospheric aerosol mass. The chemical composition of the total organic aerosol mass is poorly understood, although hundreds of compounds have been identified in the literature. High molecular weight compounds have recently gained much attention because this class of compounds potentially represents a major fraction of the unexplained organic aerosol mass. Here we analyze secondary organic aerosols, generated in a smog chamber from r-pinene ozonolysis with ultra-high-resolution Fourier transform ion cyclotron resonance mass spectrometry (FTICR-MS). About 450 compounds are detected in the mass range of m/z 200-700. The mass spectrum is clearly divided into a low molecular weight range (monomer) and a high molecular weight range, where dimers and trimers are distinguishable. Using the Kendrick mass analysis, the elemental composition of about 60% of all peaks could be determined throughout the whole mass range. Most compounds have high O:C ratios between 0.4 and 0.6. Small compounds (i.e., monomers) have a higher maximum O:C ratio than dimers and trimers, suggesting that condensation reactions with, for example, the loss of water are important in the oligomer formation process. A program developed in-house was used to determine exact mass differences between peaks in the monomer, dimer, and trimer mass range to identify potential monomer building blocks, which form the co-oligomers observed in the mass spectrum. A majority of the peaks measured in the low mass region of the spectrum (m/z < 300) is also found in the calculated results. For the first time the elemental composition of the majority of peaks over a wide mass range was determined using advanced data analysis methods for the analysis of ultra-high-resolution MS data. Possible oligomer formation mechanisms in secondary organic aerosols were investigated. Organic aerosols are either emitted directly into the atmosphere as primary aerosols or form in the atmosphere as secondary aerosols from gaseous biogenic or anthropogenic * To whom correspondence should be addressed. Phone: +4144632 2929. Fax: +41446321292. E-mail: [email protected] † ETH Zurich. ‡ UZH/ETH Zurich. # Paul Scherrer Institut.

4074 Analytical Chemistry, Vol. 79, No. 11, June 1, 2007

precursors. Often 30-70% of the ambient aerosols is organic material, but only a minor fraction of the total organic mass has been identified. A better knowledge of this large organic aerosol fraction is crucial to better understand the influence of aerosols in such diverse processes as cloud formation or their effects on human health. In recent years high molecular weight compounds or oligomers have been detected by several authors who analyzed laboratorygenerated secondary organic aerosols (SOA). In most studies mass spectrometric methods were used such as (matrix-assisted) laser desorption/ionization mass spectrometry ((MA)LDI-MS),1-3 electrosprayionization(ESI-)MS3-8 oraerosolmassspectrometry.9-10 In the most recent years electrospray ionization mass spectrometry was increasingly used for the characterization of complex environmental samples containing high molecular weight compounds such as fulvic and humic acids or organic aerosols, because this technique often allows for a mass spectrometric measurement of large compounds without fragmentation.3,5,7,11-13 Bulk experiments with SOA surrogates showed similar results of high molecular weight compound formation.14,15 Suggested formation pathways of these oligomers involve mostly addition or condensation reactions of carbonyls,1,16 which are abundant oxidation products of organic compounds in an oxidizing atmo(1) Kalberer, M.; Paulsen, D.; Sax, M.; Steinbacher, M.; Dommen, J.; Fisseha, R.; Prevot, A. S. H.; Frankevich, V.; Zenobi, R.; Baltensperger, U. Science 2004, 303, 1659-1662. (2) Kalberer, M.; Sax, M.; Samburova, V. Environ. Sci. Technol. 2006, 40, 59175922. (3) Tolocka, M.; Jang, M.; Ginter, J.; Cox, F.; Kamens, R.; Johnston, M. Environ. Sci. Technol. 2004, 38, 1428-1434. (4) Dreyfus, M. A.; Tolocka, M. P.; Dodds, S. M.; Dykins, J.; Johnston, M. V. J. Phys. Chem. A 2005, 109, 6242-6248. (5) Iinuma, Y.; Boege, O.; Gnauk, T.; Herrmann, H. Atmos. Environ. 2004, 38, 761-773. (6) Iinuma, Y.; Boge, O.; Miao, Y.; Sierau, B.; Gnauk, T.; Herrmann, H. Faraday Disc. 2005, 130, 279-294. (7) Gao, S.; Keywood, M.; Ng, N. L.; Surratt, J.; Varutbangkul, V.; Bahreini, R.; Flagan, R. C.; Seinfeld, J. H. J. Phys. Chem. A 2004, 108, 10147-10164. (8) Gao, S.; Ng, N. L.; Keywood, M.; Varutbangkul, V.; Bahreini, R.; Nenes, A.; He, J. W.; Yoo, K. Y.; Beauchamp, J. L.; Hodyss, R. P.; Flagan, R. C.; Seinfeld, J. H. Environ. Sci. Technol. 2004, 38, 6582-6589. (9) Liggio, J.; Li, S.-M.; McLaren, R. Environ. Sci. Technol. 2005, 39, 15321541. (10) Gross, D. S.; Gaelli, M. E.; Kalberer, M.; Prevot, A.; Dommen, J.; Alfarra, M. R.; Duplissy, J.; Gaeggeler, K.; Gascho, A.; Metzger, A.; Baltensperger, U. Anal. Chem. 2006, 78, 2130-2137. 10.1021/ac062425v CCC: $37.00

© 2007 American Chemical Society Published on Web 04/06/2007

sphere, esterification,17 or the formation of large organic peroxides.18-20 In order to obtain unambiguous assignments of elemental compositions of high molecular weight compounds, which is the first step toward a compound identification, very high mass resolution and highly accurate mass measurements are necessary. Conventional mass spectrometers like quadrupole time-of-flight (Q-ToF) do not have enough resolution power (typically 1000020000) to distinguish and differentiate all the peaks present in samples as complex as SOA. In contrast, Fourier transform ion cyclotron resonance mass spectrometers (FTICR-MS) allow analyses with sub-ppm accuracy and a resolution of several 100000. Kalberer et al.1 and Tolocka et al.3 performed the first ultra-highresolution mass spectrometry analyses of SOA samples from trimethylbenzene and R-pinene, respectively, and presented exact mass results for a few peaks in the mass spectrum supporting the above-mentioned oligomer formation processes. Reemtsma et al.21 recently presented FTICR-MS results from ambient aerosol samples showing that high molecular weight compounds in organic aerosols are often sulfated and/or nitrated, similar to results suggested earlier by others.22,23 In this study FTICR-MS is used for a detailed analysis of monomers and oligomers in the mass range of 200-700 Da in SOA formed in R-pinene ozonolysis experiments performed in a smog chamber. We introduce several analysis methods into the field of atmospheric sciences, e.g., the Kendrick mass analysis and the Van Krevelen diagram, which are helpful tools to categorize and structure the large amount of data obtained from ultra-high-resolution mass analyses.24-25 Both methods have been successfully used for the characterization of hundreds of individual compounds in crude oil and in humic acid samples.25-26 Using these analysis methods we determined the sum formula of about 60% of the peaks throughout the whole mass range. Most compounds have a high O:C ratio in the range of 0.5 (with values up to 0.8). A visualization of the complex mass spectra to find, (11) Leenheer, J. A.; Rostad, C. E.; Gates, P. M.; Furlong, E. T.; Ferrer, I. Anal. Chem. 2001, 73, 1461- 1471. (12) Reemtsma, T.; These, A.; Springer, A.; Linscheid, M. Environ. Sci. Technol. 2006, 40, 5839-5845. (13) Kujawinski, E. B.; Hatcher, P. G.; Freitas, M. A. Anal. Chem. 2002, 74, 413-419. (14) Limbeck A.; Kulmala, M.; Puxbaum, H. Geophys. Res. Lett. 2003, 30, doi: 10.1029/ 2003GL017738. (15) Hoffer, A.; Kiss, G.; Blazso´, M.; Gelencse´r, A. Geophys. Res. Lett. 2004, 31, L06115, doi:10.1029/2003GL018962. (16) Jang, M.; Czoschke, N. M.; Lee, S.; Kamens, R. M. Science 2002, 298, 814817. (17) Surratt, J. D.; Murphy, S. M.; Kroll, J. H.; Ng, N. L.; Hildebrandt, L.; Sorooshian, A.; Szmigielski, R.; Vermeylen, R.; Maenhaut, W.; Claeys, M.; Flagan, R. C.; Seinfeld, J. H. J. Phys. Chem. A 2006, 110, 9665-9690. (18) Docherty, K. S.; Wu, W.; Lim, Y. B.; Ziemann, P. J. Environ. Sci. Technol. 2005, 39, 4049-4059. (19) Ziemann, P. J. Faraday Discuss. 2005, 130, 469-490. (20) Kroll, J. H.; Ng, N. L.; Murphy, S. M.; Flagan, R. C.; Seinfeld, J. H. Environ. Sci. Technol. 2006, 40, 1869-1877. (21) Reemtsma, T.; These, A.; Venkatachari, P.; Xia, X.; Hopke, P. K.; Springer, A.; Linscheid, M. Anal. Chem. 2006, 78, 8299-8304. (22) Gao, S.; Surratt, J. D.; Knipping, E. M.; Edgerton, E. S.; Shahgholi, M.; Seinfeld, J. H. J. Geophys. Res. 2006, 111, doi:10.1029/2005JD006601. (23) Romero, F.; Oehme, M. J. Atmos. Chem. 2005, 52, 283-294. (24) Kim, S.; Kramer, R. W.; Hatcher, P. G. Anal. Chem. 2003, 75, 5336-5344. (25) Hughey, C. A.; Hendrickson, C. L.; Rodgers, R. P.; Marshall, A. G. Anal. Chem. 2001, 73, 4676-4681. (26) Stenson, A. C.; Marshall, A. G.; Cooper W. T. Anal. Chem. 2003, 75, 12751284.

for example, repetitive structures or changes in the O:C ratios of series of peaks revealed that condensation reactions such as esterification are important for the oligomer formation. In addition, we developed in-house a program to determine the sum formula of 83 monomer units out of which the oligomers are mainly composed. The aim of this paper is to show the potential of ultrahigh-resolution and accurate mass measurements and to show how advanced data analysis can help in characterizing the highly complex organic mixtures present in SOA. EXPERIMENTAL SECTION Aerosol Production and Extraction. SOA particles were generated in the indoor smog chamber at the Paul Scherrer Institute (PSI), which consists of a 27-m3 transparent Teflon bag suspended in a temperature-controlled housing. The chamber and routine equipment for the analysis of gaseous compounds and aerosol particles are described in detail by Paulsen et al.27 R-Pinene SOA was produced in dark ozonolysis experiments without seed particles, NOx, water vapor (relative humidity m/z 200. High Resolution and Accurate Mass Measurements. High resolution and accurate mass measurements are closely related to each other because a high mass accuracy depends, among (27) Paulsen, D.; Dommen, J.; Kalberer, M.; Prevot, A.; Richter, R.; Sax, M.; Steinbacher, M.; Weingartner, E.; Baltensperger, U. Environ. Sci. Technol. 2005, 39, 2668-2678. (28) Alfarra, M. R.; Paulsen, D.; Gysel, M.; Garforth, A. A.; Dommen, J.; Prevot, A.; Worsnop, D. R.; Baltensperger, U.; Coe, H. Atmos. Chem. Phys. 2006, 6, 5279-5293.

Analytical Chemistry, Vol. 79, No. 11, June 1, 2007

4075

other factors, on sufficiently resolved peaks. However, these two terms should not be confused, as performing a measurement at high resolution alone does not automatically imply a high mass accuracy. All FTICR-MS measurements were carried out on a 7 T LTQFT mass spectrometer (Thermo Electron Bremen, Germany) equipped with a Nanomate electrospray source in positive ion mode (Advion Biosciences, Hethersett Norwich Norfolk, UK). Spray, tube lens, and capillary voltages were optimized to allow for ideal spray conditions. Mass accuracy calibration was performed using a sodium formate solution immediately before the measurements according to manufacturers instruction. The rootmean-square error was typically better than 2 ppm. The automatic gain control (AGC) target settings for the allowed number of ions in the mass analyzer were set to 105 for full FT MS spectral acquisition range. Spectra were acquired at a resolution of 400000 at m/z 400. Typically data was acquired for 1 min, over which the spectra were summed. The mass tolerance in the Xcalibur software (Thermo Electron, San Jose, CA) was set to 2.5 ppm for the data analysis. RESULTS AND DISCUSSION In earlier studies we analyzed SOA samples from various precursors with (MA)LDI-MS.1,2 For those measurements particles were sampled on impactor plates and analyzed without any sample preparation. The experimental details are given elsewhere.2 A MALDI mass spectrum of R-pinene SOA, acquired in positive MS mode, is shown in Figure 1a (upper panel). In an earlier study a comparison with surrogate standards showed that most of the peaks in (MA)LDI-MS are likely cation adducts.1 The lower panel in Figure 1a shows the extracted fraction of R-pinene SOA measured with ESI-FTICR-MS (also positive mode). The overall spectrum of R-pinene SOA is similar for both methods although the number of peaks measured with MALDI-MS is larger. The repetitive structure with ∆m/z 14-18 is visible in both spectra, and their maxima are measured at the same m/z. Often the 2-3 highest peaks have very similar intensities as can be seen in Figure 1a (see insert), and therefore it is often difficult to determine whether the ∆m of the repetitive unit is 14, 16, or 18. The mass spectrometer measurements were carefully examined in order to exclude (or detect) measurement artifacts. Because the ionization processes in MALDI- and ESI-MS are very different, such artifacts can be examined by comparing mass spectra obtained by these two methods. The most abundant artifacts include ion fragmentation, adduct formation, or wrong assignments of multiply charged ions. ESI is generally considered as a very soft ionization method and fragmentation is thought to be minimal, but adduct formation is often seen in ESI-MS. In contrast, fragmentation is sometimes a problem in MALDI-MS, which is a somewhat less soft ionization method. Comparing the two R-pinene SOA mass spectra in Figure 1a acquired with MALDI- and ESI-MS we conclude that the spectra agree well with each other and therefore fragmentation is likely not a major problem for these SOA samples (see insert in Figure 1a). Further possible artifacts in ESI-MS are the formation of noncovalent adducts between analytes during the droplet evaporation and ionization processes (leading to larger m/z ions) and wrong assignments of multiply charged ions. Nonspecific and 4076 Analytical Chemistry, Vol. 79, No. 11, June 1, 2007

Figure 1. (a) Comparison MALDI-MS (top) and ESI-FTICR-MS (bottom) spectra of R-pinene SOA. Both analysis methods show similar results as illustrated by the insert. (b) Comparison of a Q-ToF-MS (blue) and a FTICR-MS (red) of R-pinene SOA in the mass range of 411.00-411.29. The peak at m/z 411.16 measured with the Q-ToF-MS with a resolution of 5200 appears to be due to only one compound. The FTICR-MS measurement with a resolution of 400′000 reveals that actually three peaks with very similar m/z are present in this mass range.

noncovalent complexes can be formed as a result of the shrinking solvent droplets of the ESI spray. Generally, such effects are less pronounced using a nanoESI source compared to conventional ESI sprays due to much smaller initial droplet diameters and the faster ion formation processes in nanoESI.29 The Nanomate system was used here to minimize this potential artifact. To check for the possibility of noncovalent complexes, in-source fragmentation experiments were performed by increasing the source CID voltage. Higher source CID voltages accelerate the analyte ions more rapidly in the interface region resulting in more energetic collisions with the desolvation gas and therefore promote dissociation of the weak noncovalent bonds. Throughout the applied voltage range no shifts in the SOA mass spectra to lower masses were observed, indicating that noncovalent complexes are not a major problem of the samples analyzed here. It is possible that highly oxidized analytes are more easily ionized with electrospray ionization than less oxidized compounds. The O:C ratio deter(29) Karas, M.; Bahr, U.; Dulcks, T. Fresenius J. Anal. Chem. 2000, 366, 669676.

Figure 2. FTICR mass spectrum of R-pinene SOA generated in an ozonolysis experiment in a smog chamber.

mined here (see below) compares well with average bulk data measured recently by Tolocka et al.,30 indicating that such a bias is likely not a major problem for the samples analyzed here and that the compounds analyzed here are major components in R-pinene SOA. The charge state of an ion in a FTICR spectrum can be evaluated by determining the mass difference of a peak with its corresponding 13C isotope signal m/z. For singly charged ions, the mass difference between these two peaks is ∆m/z 1.00335, whereas for doubly and triply charged ions 1/2 and 1/3 of this value would be expected, respectively. All major peaks in the ESIFTICR spectra show mass differences of ∆m/z 1.00335 between 12C and 13C isotopomers, confirming their single charge state. Figure 1b shows a small part of the mass spectra (at m/z 411.00-411.29) of the same R-pinene SOA sample extract measured with ESI-Q-ToF-MS (quadrupole time-of-flight MS) and ESIFTICR-MS, illustrating very well the difference in resolving power between these two mass spectrometers. The resolving power of the Q-ToF-MS is around 5000 to 10000 compared to 400000 of the FTICR-MS. Figure 1b clearly demonstrates that the resolving power of the FTICR is necessary to obtain reliable information about the compounds present in the complex SOA sample. Estimating the elemental composition of the peak at measured m/z 411.16 with the Q-ToF-MS would have resulted in a misleading interpretation because the peak actually consists of three compounds with different elemental compositions as shown by the FTICR-MS measurement. Figure 2 shows a mass spectrum obtained with the FTICRMS (positive mode) from a SOA sample collected 5.8 to 7.3 h after the start of the R-pinene ozonolysis reaction in the smog chamber. Up to 450 peaks are resolved in the mass range between m/z 200-700. Three broad groupings of peaks are observed: the monomer range below m/z 300, the dimer range at m/z 300540, and the trimer range at m/z 540-700. Although it is possible that small dimers have m/z < 300 and that in the dimer and trimer range higher oligomers are present, this terminology is used throughout the manuscript for clarity. Most of the intense peaks have odd nominal masses, which is in agreement with protonated or cationized ions containing only 12C, 1H, and 16O (and Na+) but no 13C. The less intense even mass peaks are mostly isotope peaks containing one 13C atom. (30) Tolocka, M. P.; Heaton, K. J.; Dreyfus, M. A.; Wang, S.; Zordan, C. A.; Saul, T. D.; Johnston, M. V. Environ. Sci. Technol. 2006, 40, 1843-1848.

The broad groupings of peaks are typical for low-molecular weight copolymers31,32 suggesting that co-oligomerization reactions are taking place between a limited number of aerosol components. These measurements are in agreement with the study of Tolocka et al.,3 who also analyzed R-pinene SOA with ESI-MS and detected three groupings of peaks in the mass range of m/z 300-850 (likely dimers, trimers, and tetramers). One of the main goals of the ultra-high-resolution and accurate mass measurements is to assign elemental compositions for all monomers and oligomers measured between m/z 200-700. For the further analysis only peaks with an intensity larger than three times the variability of the background signal were considered, corresponding to about 450 peaks in the mass range m/z 200700. As mentioned above, even m/z peaks mostly represent 13Cisotope signals. Therefore, only peaks with an odd nominal mass (about 360 peaks) were further considered for analysis. The accurate and ultra-high-resolution FTICR-MS measurements allow for an unambiguous assignment of the elemental composition for each m/z in the lower mass range (up to about m/z 400) within the error limits of the mass accuracy of the instrument. Up to m/z 400 the elemental compositions were assigned with the mass calculator software Xcalibur (Thermo Electron, San Jose, CA) using a mass tolerance of ( 2.5 ppm. The following elements and number of atoms of each element were considered in the estimation of the elemental composition of each peak: 12C (0-60 atoms), 1H (0-120 atoms), 16O (0-60 atoms), 23Na (0-1 atoms). More than 95% of all elemental compositions determined contain one Na atom, i.e., cationization is the most important ionization process for SOA compounds. Because essentially no NOy was measured in the smog chamber (as described above), nitrogen was not considered in the calculations of the elemental composition of the SOA components. For most peaks above about m/z 400, more than one elemental composition is theoretically possible and an unambiguous assignment of the elemental composition of these peaks is not possible. Since most oligomer peaks have m/z > 400, other methods are needed to obtain information about their elemental composition, such as the Kendrick mass analysis. Kendrick Mass Scale. The Kendrick mass scale converts the exact mass of CH2 from 14.01565 (IUPAC mass scale) to 14.00000.

KM ) IUPAC massmeasured ×

14.00000 [14.00535 ]

(1)

where KM is the (exact) Kendrick mass of a peak. Thus, homologous series (namely, compounds with the same number of heteroatoms, rings, and double bonds but different numbers of CH2 groups) differ by exactly 14.000000 Da in the Kendrick mass scale and most importantly they all have an identical mass defect. The Kendrick mass defect (KMD) is defined as nominal Kendrick mass - Kendrick mass (KM).25,33

KMD ) nominal KM - KM

(2)

where the nominal KM is the integer of the KM. For compounds composed of only C, H, and O atoms, a small mass defect indicates (31) Nielen, M. W. F. Mass Spectrom. Rev. 1999, 18, 309-344. (32) Zoller, D. L.; Johnston, M. V. Macromolecules 2000, 33, 1664-1670. (33) Kendrick, E. Anal. Chem. 1963, 13, 2146-2154.

Analytical Chemistry, Vol. 79, No. 11, June 1, 2007

4077

Figure 3. (a) Plot of Kendrick mass defect vs nominal Kendrick mass for all peaks in the FTICR mass spectrum in positive mode. Each data point corresponds to a peak in the mass spectrum. (b) Kendrick mass defect vs nominal Kendrick mass for odd-mass ions with z* ) -6. Horizontal lines connect compounds with an equal elemental composition differing only by a CH2 group (see text for details). Letters refer to compounds listed in Table 1.

a small number of oxygens and a large defect a large number of oxygen atoms in the molecule. Figure 3a displays the nominal KM as a function of the KMD of all 360 odd m/z peaks between m/z 200-700. Each data point corresponds to a peak in the FTICR spectrum. The three groups of peaks visible in Figure 3a correspond to the monomer, dimer, and trimer groupings in the mass spectrum (see Figure 2). The increasing KMD with increasing KM immediately visualizes that larger compounds contain a larger number of O atoms than smaller compounds. However, the vertical shift of the trend lines through the three groupings of peaks to lower KMD indicates that the larger compounds have a lower O:C ratio (see below). Compounds within a homologous CH2-series appear in horizontal lines with nominal mass differences of ∆m14 (i.e., CH2) in the Kendrick plot, because they have an identical KMD. However, the large number of peaks present in the SOA samples makes the identification of such homologous series difficult. Therefore, it is advantageous to presort the ions based on another, independent parameter. Hsu et al.34 suggested to subdivide the peaks in 4078

Analytical Chemistry, Vol. 79, No. 11, June 1, 2007

the Kendrick mass plot according to their z* value, which is definded as

z* ) (modulus[nominal KM/14]) - 14

(3)

The modulus in eq 3 is the remainder of the division of the nominal KM by 14. Values for z* can vary between -1 and -14. Ions of identical KMD and z* values can differ from each other only in the number of CH2 groups.26 Figure 3b shows, as an example, all peaks with z* ) -6. Several homologous series differing only by a CH2 group (with up to seven compounds) are now clearly visible. Two compounds with the same nominal KM but a KMD difference of 0.0364 have the same elemental composition except that a CH4 unit is exchanged by an oxygen atom.26 These two features (homologous series and specific ∆ KMD) allow for an assignment of the elemental composition of peaks with m/z much higher than what (34) Hsu, C. S.; Qian, K. N.; Chen, Y. N. C. Anal. Chim. Acta 1992, 264, 7989.

Table 1. Exact IUPAC Masses (measured), Calculated Kendrick Masses, and Elemental Composition Deduced from the Kendrick Mass Plot in Figure 3b. The 17 Peaks Listed Here Belong to Three Homologous Series with 8, 9, and 10 Oxygen Atoms, Respectivelya IUPAC m/z (measured)

peak marked in Figure 3b

Kendrick mass

no. of oxygen atoms

elemental composition

339.10508 353.12074 367.13644 381.15213 395.16777 409.18346 423.19911 395.13143 409.14706 423.16273 437.17840 451.19420 437.14205 451.15777 535.25186 563.28310 577.29826

A

337.72028 351.72029 365.72034 379.72038 393.72037 407.72041 421.72041 393.68407 407.68405 421.68407 435.68409 449.68424 435.64770 449.64785 533.64804 561.64798 575.64749

8 8 8 8 8 8 8 9 9 9 9 9 10 10 10 10 10

C14H20O8Na C15H22O8Na C16H24O8Na C17H26O8Na C18H28O8Na C19H30O8Na C20H32O8Na C17H24O9Na C18H26O9Na C19H28O9Na C20H30O9Na C21H32O9Na C19H26O10Na C20H28O10Na C26H40O10Na C28H44O10Na C29H46O10Na

B C

D E

F

The elemental composition of peaks m/z < 400 was calculated with the XCalibur software, peaks with m/z > 400 were deduced using the Kendrick Mass Analysis (see text for details). a

is possible from the direct measurement (only m/z < 400), i.e., throughout the entire mass spectrum. The oxygen content varies from 5 up to 11 oxygen atoms per molecule, as indicated in Figure 3b. Using this analysis an elemental composition for about 60% of all peaks could be unambiguously assigned including all highintensity peaks. Table 1 lists the peaks with 8, 9, and 10 oxygen atoms displayed in Figure 3b, exemplifying this method. Three of the seven exact masses reported by Tolocka et al.3 were also found in this study (i.e., m/z 347.183, 361.198 and 375.178 corresponding to C18H28O5Na, C19H30O5Na and C19H28O6Na). Whereas Tolocka et al.3 only assigned an elemental composition for seven peaks in the mass spectrum in the range of m/z 347-375, we identified here the elemental composition of about 60% (i.e., about 210) of all considered peaks in the mass spectrum between m/z 200-700. Van Krevelen Diagram. The Van Krevelen plot is a graphical method developed originally to study problems connected with the composition and structure of coal.35 Figure 4a shows a twodimensional Van Krevelen plot with the hydrogen to carbon ratio plotted versus the oxygen to carbon ratio for all 360 odd mass compounds in the MS between m/z 200-700. Most of the

Figure 4. (a) Van Krevelen plot for R-pinene SOA. For each peak in the mass spectrum the O:C and H:C ratio is given. (b) O:C ratio as a function of m/z for all compounds. Dimers and trimers have a lower maximum O:C ratios than monomers suggesting condensation reactions as possible formation pathways of SOA oligomers as indicated with the dotted lines. Compounds with the same number of carbon and oxygen atoms are connected with solid lines.

Analytical Chemistry, Vol. 79, No. 11, June 1, 2007

4079

Figure 5. (a) The MonomerHunter found the exact mass difference ∆m 232.095 (C10H16O6) 52 times in the FTICR mass spectrum. Eighty three exact mass differences with masses between 120 and 250 connecting 10-69 pairs of peaks were identified by the MonomerHunter. (b) Van Krevelen plot of all monomers calculated by the MonomerHunter. Monomers e10 carbon atoms or g1 oxygen atoms (i.e., chemically meaningful) are marked red, and those consistent with peaks measured in the FTICR-MS have filled symbols.

compounds have O:C ratios of 0.4-0.6 and a H:C ratio of 1.41.7. Most compounds have a H:C ratio lower than 1.6, which is the H:C ratio of R-pinene, indicating oxidation reactions. Considering the oxidizing conditions in the smog chamber, this is an expected result. However, about 30% of the compounds have a higher H:C ratio than R-pinene, which could be formed, e.g., in hydration reactions resulting in the formation of acetals or hemiacetals. O:C ratios agree well with a recent study by Tolocka et al.30 where an average O:C ratio of 0.4-0.6 for R-pinene SOA bulk samples was determined. 4080 Analytical Chemistry, Vol. 79, No. 11, June 1, 2007

Compounds along the dotted lines in Figure 4a are potentially linked with reactions such as (de)hydrogenation, (de)hydration, (de)alkylation, or oxidation as indicated in Figure 4a. However, structural relationships cannot be inferred from these graphs alone. Figure 4b shows the O:C ratio of all compounds as a function of their mass. The vertical lines indicate the monomer, dimer, and trimer groupings in the MS. The maximum O:C ratio is highest for monomers and lowest for the trimers (see dotted lines

Table 2. Masses of All Peaks Detected in the Monomer Region of the Mass Spectrum and Mass Differences Calculated by the MonomerHunter That Match the Measured Results m/z measured (m/z 200-300)

elemental compositiona

deviation measured, calcd (ppm)

found in MHb

205.122316 206.125656 207.099171 209.078421 211.057615 222.594354 223.094091 223.130909 224.136224 225.073406 225.109771 227.052557 227.089017 229.068253 237.073351 239.088984 240.092294 241.068241 241.104676 243.083941 250.128536 253.068266 255.083888 256.087253 257.099543 267.096776 269.063230 271.078880 273.094465 287.073802

-c -c (C10H16O3)d (C9H14O4) d (C8H12O5) d -c C10H16O4 C11H20O3 -c C9H14O5 C10H18O4 (C8H12O6) d C9H16O5 C8H14O6 C10H14O5 C10H16O5 -c C9H14O6 C10H18O5 C9H16O6 -c C10H14O6 C10H16O6 -c C10H18O6 -c C10H14O8 C10H16O7 C10H18O7 C10H16O8

2.61 2.66 2.97 2.40 0.46 2.16 2.25 2.64 2.31 2.41 2.28 2.33 2.34 2.14 2.12 2.13 2.23 2.19 1.82 1.81 2.03 1.68

no no yes yes yes no yes no no yes yes yes yes no yes yes no yes yes yes no yes yes no yes no no yes no no

number of pairs found by MHb

61 69 52 64 28 20 19 23 12 49 66 30 39 17 39 52 14 12

found in MHb + H2O no no yes yes yes no yes no no yes yes yes yes yes no yes no yes yes yes no no yes no yes no no yes yes yes

number of pairs found by MHb

possible compound

44 54 22

hydroxypinonealdehyde pinic acid hydroxynorpinic acid

49

hydroxypinonic acid

53 61 21 69 52 41 34 64 28 49 66 39 52 14

a Each compound is measured as Na-adduct, which is not mentioned in the sum formula. b MH: MonomerHunter. c No elemental composition within 3.0 ppm. d Calculated masses of sum formula deviating 2.5-3.0 ppm from the measured masses are given in brackets.

in Figure 4b). This indicates that condensation reactions involving, e.g., the loss of a water molecule as in esterification reactions are likely responsible for the oligomerization reactions linking the compounds of the three groupings of peaks. Lines connecting compounds with equal numbers of oxygen and carbon atoms are also shown in Figure 4b. Peaks in the monomer range are mostly C8-C10 compounds, whereas the most intensive peaks in the dimer range are C17-C20 compounds, i.e., likely combinations of C8-C10 monomers. Within one grouping the O:C ratio increases, i.e., dimers around m/z 500 have on average a higher O:C ratio than at m/z 350. This might be due to the incorporation of monomers with a different O:C ratio into the co-oligomers or/ and an ongoing oxidation of these co-oligomers once they are formed in the aerosol particle. Although Van Krevelen plots do not allow to investigate directly specific reactions of SOA components, general trends of reaction schemes can be deduced. The MonomerHunter. The second main goal of this study was the investigation of repetitive masses (i.e., “monomers”) connecting the monomer with the dimer grouping and the dimer with the trimer groupings to learn more about the structure and formation processes of SOA oligomers. A program developed in-house (called “MonomerHunter” hereafter) using AWK language36 computes all ∆m in a given mass range for an input peak list, followed by the computation of a histogram of all mass differences higher than a given threshold. For the computation of the ∆m, values were rounded to the fourth decimal place. The result is ordered according to the frequency

of occurrence of all mass differences. In the second phase, the algorithm performs a backtracking step. According to a ranking given by the user, e.g., the most frequent ∆m, the program computes peaks having the same distance. Figures 5a shows a typical result of the MonomerHunter analyzing a FTICR mass spectrum. The exact mass difference ∆m 232.0950 connects 52 pairs of peaks in the R-pinene SOA spectrum. This mass is unambiguously assigned to the elemental composition C10H16O6. The MonomerHunter identified 83 exact mass differences that correspond to a specific elemental composition (containing only C, H, and O atoms) within 2.5 ppm accuracy in the mass range of 120-250 and which connect at least 10 pairs of peaks. Only for one mass difference calculated by the MonomerHunter no elemental composition (containing only C, H, and O atoms) could be assigned. The identified monomers contain 6-12 carbon atoms and 0-7 oxygen atoms. Because the MonomerHunter is a purely mathematical procedure, and considering that R-pinene (the parent hydrocarbon in the ozonolysis smog chamber experiment) has only 10 carbon atoms, only those 72 out of the 83 “monomers” which have e10 carbon or g1 oxygen atoms are chemically meaningful (i.e., 87%). Calculated monomers with g10 carbon atoms could be small dimers rather than monomers. This demonstrates that the mathematical procedure (35) Krevelen, D. W. Fuel 1950, 29, 269-284. (36) Aho, A. V.; Kernighan, B. W.; Weinberger, P. J.; Longman, A.-W. AddisonWesley, Amsterdam, 1988, ISBN 0-201-07981-X.

Analytical Chemistry, Vol. 79, No. 11, June 1, 2007

4081

used by the MonomerHunter results in a large fraction of chemically meaningful monomer compounds. The results found by the MonomerHunter can be compared with the 30 peaks measured in the monomer mass region of the FTICR spectrum (m/z < 300, see Figure 2). Assuming either a condensation reaction with the loss of a water molecule (e.g., formation of an ester) or an addition reaction (e.g., formation of a peroxide18-19) 18 and 17 monomers, respectively, of the 83 monomers that are suggested by the MonomerHunter are also found in the measured mass spectrum (see Table 2). There are several potential reasons for the existence of monomers suggested by the MonomerHunter that are not found in the mass spectrum as monomers: They may be chemically not meaningful, or they have very low ionization efficiencies, or they react in another reaction than addition or condensation with loss of water (which were considered here) to form a dimer or trimer or they are so efficiently incorporated into oligomers that they are not present anymore as monomers. In addition, some pairs of peaks identified by the MonomerHunter might be connected rather due to fragmentation of the larger peak than due to the addition of a monomer to the smaller peak. However, fragmentation is likely not a major problem for the samples analyzed here as described above. Figure 5b shows a van Krevelen plot of all 83 monomers. Monomers with e10 carbon or g1 oxygen atoms (i.e., chemically meaningful) are marked with red symbols and those that are found in the monomer mass region of the mass spectrum are shown with full symbols. The majority of the compounds, monomers as well as oligomers, have high O:C ratios, often above 0.5 (Table 2, Figures 4a, 4b, 5b). The elemental composition of only a few of these compounds matches those of known oxidation products of R-pinene (e.g., pinic acid, see Table 2), whereas most elemental compositions determined here have higher O:C ratios than known R-pinene oxidation products. Exact mass determinations and elemental compositions deduced from such analyses do generally not allow determining directly the structure of a compound. Therefore we do not further speculate on the structural identity of the compounds analyzed here, because a high number of possible structures could be deduced from the elemental compositions determined in this study, making the choice of a specific structure rather arbitrary. Exact mass measurements allow for determining the elemental composition of hundreds of unknown compounds in SOA. Combining this technique with other methods will lead to the structures of SOA oligomers.

4082

Analytical Chemistry, Vol. 79, No. 11, June 1, 2007

CONCLUSIONS Secondary organic aerosol from R-pinene was generated in a smog chamber and analyzed with ESI-FTICR-MS. About 450 peaks in a mass range between m/z 200-700 were detected. The mass spectrum is clearly divided in a range with low-molecular weight compounds (m/z < 300) and a high molecular weight (or oligomer) range (300 < m/z < 700). This ultra-high-resolution and accurate mass measurement technique allows to assign unambiguously the elemental composition for peaks with m/z < 400. Above m/z 400 the elemental composition of about 60% of all peaks could be deduced using the Kendrick mass analysis. Van Krevelen diagrams show that most of the compounds in R-pinene SOA have high O:C ratios in the range of 0.4-0.6. Monomers have on average a higher O:C ratio than dimers and trimers, suggesting a condensation reaction leading from the monomers to the small co-oligomers. Evidence for acetal formation and esterification reactions forming the SOA oligomers were found. A program developed in-house, MonomerHunter, was used to calculate the abundance of exact mass differences throughout the spectrum to find potential monomers. Eighty three exact mass differences in the range of m/z 120-250 (i.e., monomers) were found in the mass spectrum, which connect between 10 and 69 pairs of peaks. A majority of the peaks measured in the low mass region of the spectrum (m/z < 300) is also found in the calculated results of the MonomerHunter, indicating that this mathematical procedure gives mostly chemically meaningful results. For the first time ultra-high-resolution MS data from aerosol samples were analyzed for a wide mass range (m/z 200-700) with advanced data analysis methods to examine the elemental composition of monomers and oligomers and to investigate possible oligomer formation mechanisms in secondary organic aerosols. ACKNOWLEDGMENT The help of Jonathan Duplissy, Astrid Gascho, and Axel Metzger during the smog chamber experiments is greatly acknowledged as well as the Functional Genomics Center for providing their infrastructure. This work was supported by ETH Grant TH-14/04-2, the Swiss National Science Foundation (No. 200020-108095), and the EC project POLYSOA (Polymers in Secondary Organic Aerosols).

Received for review December 22, 2006. Accepted March 2, 2007. AC062425V