Anal. Chem. 1998, 70, 1979-1985

Structure Elucidation of 2,4-Dinitrophenylhydrazone Derivatives of Carbonyl Compounds in Ambient Air by HPLC/MS and Multiple MS/MS Using Atmospheric Chemical Ionization in the Negative Ion Mode Stephan Ko 1 lliker and Michael Oehme*

Organic Analytical Chemistry, University of Basle, Neuhausstrasse 31, CH-4057 Basle, Switzerland Christian Dye

Norwegian Institute for Air Research, P.O. Box 100, N-2007 Kjeller, Norway

Ion trap multiple fragmentation mass spectrometry (MSn) combined with high-performance liquid chromatography (HPLC) has been used for the structure elucidation and identification of 2,4-dinitrophenylhydrazone derivatives of carbonyl compounds in ambient air samples. Atmospheric pressure chemical ionization in the negative ion mode was the most suitable detection method. Different measures are described to decrease the MS background originating from the HPLC system. Low-picogram quantities were detectable in extracted mass chromatograms generated from full-scan records. Fragment ions produced by MS/MS allowed identification of substructures of the carbonyls. Detailed fragmentation paths were studied by MS3 to MS4 using reference compounds. A fragmentation scheme was established which enabled a structure confirmation and identification with 1-10 ng by HPLC/MS/MS. The identification of a compound coeluting with n-pentanal-DNPH and of a dimerization byproduct are given as examples. Volatile organic compounds (VOCs) and nitrogen oxides (NOx) are main precursors that are involved in the formation of ozone in the lower troposphere. However, they are not reactive enough to generate ozone directly. Carbonyl compounds such as aldehydes and ketones are formed as important intermediates by reaction of VOCs with ozone and OH radicals. During the past two decades, different analysis methods have been developed for carbonyl compounds in ambient air. The most frequently used procedure is based on chemical derivatization with 2,4-dinitrophenylhydrazine during sampling. Impingers,1,2 adsorption cartridges,3-5 and passive samplers6 have been employed. The formed 2,4-dinitrophenylhydrazones (DNPHs) are usually deter(1) Selim, S. J. Chromatogr. 1977, 136, 271. (2) Tuss, H.; Neitzert, V.; Seiler, W.; Neeb, R. Fresenius Z. Anal. Chem. 1982, 312, 613. (3) Kuwata, K.; Uebori, M.; Yamasaki, H.; Kuge, Y.; Kiso, Y. Anal. Chem. 1983, 55, 2013. S0003-2700(97)00945-1 CCC: $15.00 Published on Web 04/02/1998

© 1998 American Chemical Society

mined by HPLC/UV detection. Air monitoring programs in the United States7,8 as well as in Europe9 apply the DNPH method for the determination of C1-C5 aldehydes and ketones. The risk for HPLC coelutions of carbonyl compounds increases with the number of possible isomers. Already the separation of a standard mixture containing 10 different C4 carbonyl compounds is very difficult to achieve with conventional HPLC columns.10 An interference-free determination of carbonyls in air extracts is even more demanding due to the possible presence of interferences from the derivatization procedure11 as well as derivatized hydroxycarbonyls and dicarbonyls.12 Furthermore, the formation of syn/ anti DNPH isomers can also complicate the HPLC separation.13 Usually, the routine analysis of ambient air samples by HPLC/ UV detection covers only a selection of C1-C5 carbonyls. However, air extracts also contain other not yet identified carbonyl compounds which can be important for the understanding of oxidation mechanisms in the atmosphere. Particularly, C6-C20 carbonyls might contain specific information about precursors and emission sources. Attempts were made to identify carbonyl compounds by UV diode-array spectra of reference compounds. However, this requires undisturbed HPLC signals, and UV spectra only allow differentiation between aliphatic, aromatic, and dicar(4) Guenier, J. P.; Simon, P.; Delcourt, J.; Didierjean, M. F.; Lefevre, C.; Muller, J. Chromatographia 1985, 18, 137. (5) Lipari, F.; Swarin, S. J. Environ. Sci. Technol. 1985, 19, 70. (6) Levin, J. O.; Lindahl, R.; Heeremans, C.; van Osten, K. Analyst 1996, 121, 1273. (7) Purdue, L. J.; Dayton, D. P.; Rice, J.; Bursey, J. Technical assistance for sampling and analysis of ozone precursors; Atmospheric Research and Exposure Assessment Laboratory, U.S. Environmental Protection Agency, Research Triangle Park, NC, October 31, 1991. (8) Grosjean, E.; Grosjean, D.; Fraser, M. P.; Cass, G. R. Environ. Sci. Technol. 1996, 30, 2687. (9) EMEP/CCC. Manual for sampling and chemical analysis; Norwegian Institute for Air Research, Kjeller, Norway, 1992. (10) Po ¨tter, W.; Karst, U. Anal. Chem. 1996, 68, 3354. (11) Karst, U.; Binding, N.; Camman, K.; Witting, U. Fresenius J. Anal. Chem. 1993, 345, 48. (12) Grosjean, E.; Grosjean, D. Int. J. Environ. Anal. Chem. 1995, 61, 47. (13) Binding, N.; Mu ¨ ller, W.; Witting, U. Fresenius J. Anal. Chem. 1996, 356, 315.

Analytical Chemistry, Vol. 70, No. 9, May 1, 1998 1979

bonylic structures.10 Chemical ionization mass spectrometry together with a particle beam interface was also used to determine molecular weights of DNPH carbonyls.8 However, even electron ionization mass spectra of isomers are often identical.10 This work combines HPLC separation of carbonyl-DPNH with atmospheric pressure chemical ionization MSn in the negative ion mode. It shows that HPLC/MS with recently developed quadrupole ion trap mass spectrometers allow us to identify and/or confirm fully or partially the structure of DNPH carbonyls in ambient air samples. The system allowed for accumulation of selected ions in the trap and carry out multiple fragmentation reactions (also called multiple MS/MS (MSn)). This made an on-line structure elucidation possible with quantities of ∼1 ng. Mass chromatograms extracted from full-scan records allowed detection of 10-11-g amounts. However, this required a careful selection of solvents and HPLC column qualities. Detailed information is given. The mass spectra from 30 reference carbonyl compounds were used to develop a structure elucidation scheme. It allows us to distinguish between mono- and dicarbonyl-DNPHs as well as aldehydes and ketones. In addition, aromatic, saturated, unsaturated branched, or straight-chain structures can be differentiated. Even for closely related structures, the MS/MS spectra showed significant differences. At least 70 carbonyl compounds were found in ambient air samples including C2-C6 dicarbonyls, C1-C7 hydroxycarbonyls, and oxo acids. Saturated aldehydes and ketones up to C15 could be detected without sample preconcentration. A simple enrichment step allowed us to extend the range to C25. The possibilities and limitations of the developed structure elucidation scheme are demonstrated by two examples. EXPERIMENTAL SECTION Chemicals. Standard solutions of the following carbonylDNPHs were obtained as a reference solution in acetonitrile (15 µg of each carbonyl/mL) obtained from Supelco. Methanal, ethanal, propanone, propenal, propanal, 2-butenal, 2-methylpropenal, 2-butanone, n-butanal, 2-methylpropanal, n-pentanal (valeraldehyde), 3-methylbutanal, n-hexanal, benzaldehyde, otolualdehyde, m-tolualdehyde, p-tolualdehyde, 2,5-dimethylbenzaldehyde, and 2-methylpropenal. The DNPH derivatives of the following carbonyls were synthesized: trans-2-pentenal, trans2-hexenal, salicylaldehyde, furfural, glyoxal, methylglyoxal, isopropyl methyl ketone, isobutyl methyl ketone, 3-pentanone, 2-hexanone, and acetophenone. Acetonitrile, grade “190 far UV”, was purchased from Romil (Cambridge, England) and tetrahydrofuran (THF) stabilized with 250 ppm 2,6-di-tert-butyl-p-cresol was from Fluka (Buchs, Switzerland). The THF was purified by distillation over NaOH. HPLC water with a total organic carbon content of <2 ppb was obtained from an Elgastat maxima HPLC (Elga Ltd., Bucks, England) water purification unit. Analytical grade ammonium acetate and sodium hydroxide were from Merck (Darmstadt, Germany). Helium and nitrogen of 99.999% purity were used (Carbagas, Switzerland). Glassware. Glassware was soaked in a cleaning solution (2.5% (v/v) Deconex 22 LIQ, Borer Chemie, Switzerland) for two days, neutralized (Deconex 27PF), washed with a complexing agent (Deconex 20 NS), and rinsed six times with HPLC water. 1980 Analytical Chemistry, Vol. 70, No. 9, May 1, 1998

Afterward the equipment was heated for 2 h at 450 °C and rinsed twice with HPLC water. Sampling and Sample Pretreatment. Silica cartridges coated with 2,4-dinitrophenylhydrazine and hydrochloric acid were purchased from Waters (No. 37500). A volume of 750 L of ambient air was sampled at a flow rate of 2 L/h. An ozone scrubber was mounted in front as described in ref 14. The formed DNPH derivatives were eluted with 3 mL of acetonitrile. A 20-µL aliquot of the extract was buffered with 20 µL of a 1 M ammonium acetate solution. Carbonyls were identified in seven air samples obtained from the air monitoring program EMEP (European Monitoring and Evaluation Program).15 The rather large sampling volume allowed direct study of their composition by HPLC/MS, MS/MS, and MSn. One sample was preconcentrated as follows to identify carbonyls of higher molecular weight which are present at much lower concentration. A 100-µL sample of the acetonitrile extract was transferred to a 250-µL vial. After evaporation to dryness by a gentle nitrogen flow, 10 µL of acetonitrile and 10 µL of a 1 M ammonium acetate solution were added and shaken. Due to the high salt content, two phases were formed. The organic layer contained mainly >C4 DNPH carbonyls, dicarbonyls, and some unidentified compounds. The aqueous layer consisted of hydrochloric acid, ammonium acetate, nonreacted DNPH, and the main quantity of C1-C3 DNPH carbonyls. Instrumentation. A quadrupole ion trap system (LCQ, Finnigan MAT, San Jose, CA) was used in the atmospheric pressure chemical ionization mode. Negative ions were detected (APCI(-)). The MS parameter optimization was carried with a solution of acetone-DNPH (∼5 pg/µL)and a flow rate of 500 µL/ min. The signal abundance of m/z 237 [M - H]- was maximized as follows. The optimum temperature of the APCI heater was 450 °C for 3-mm-i.d. columns (flow rate, 560 µL/min) and 550600 °C for 4-mm-i.d. columns (flow rate, 1 mL/min). The best signal-to-noise (S/N) ratio was obtained at a discharge ionization current of 5 µA and the highest signal abundance at 20 µA. Other optimized parameters were as follows: sheath gas flow, 25 arbitrary units (all flow rates); heated capillary temperature, 150 °C (all flow rates); voltages (optimized by the autotune program), heated capillary voltage, +30.5 V; offset voltage, octapole 1, +9 V); interoctapole lens voltage, +75 V; offset voltage, octapole 2, +19 V; octapole rf amplitude, 900 V. Structural information was obtained by further fragmentation using MS/MS and MSn. For MS/MS two techniques can be used. The focusing octapoles allow the transfer of additional energy to the formed molecular ions resulting in a collision-induced decay (CID) at low pressure (octapole CID). Alternatively, a selected ion stored in the trap can be fragmented by transferring energy via the rf field (trap CID). This process can be repeated by again storing one of the generated fragment ions. Trap CID gave more information for structure elucidation than octapole CID. Therefore, this technique was chosen. A collision energy was selected that allowed a sufficient fragmentation of the [M - H]- or the most abundant fragment ion. Low-nanogram amounts were enough to record MS3 to MS4 fragmentation patterns. (14) Waters SEP-PAK DNPH-SILICA cartridge, care and use manual; 037506TP, rev. 1, Waters, Milford, MA 01757, 1994; p A8. (15) Solberg, S.; Dye, C.; Schmidbauer, N.; Herzog, A.; Gehrig, R. J. Atmos. Chem. 1996, 25, 33.

A binary gradient HPLC pump (Rheos 4000, Flux Instruments, Basel, Switzerland) with a dead volume of 70 µL was employed for HPLC/MS. The LCQ injection valve (10-µL loop volume) was used. To eliminate memory effects after injections of >50 ng of DNPH derivative, the needle inlet of the valve had to be removed and cleaned with acetonitrile in an ultrasonic bath (5 min) followed by repetitive valve switching and acetonitrile flushing. HPLC Separation. Fast separations were carried out on a Micra NPS C18 column (33-mm length × 4.6-mm i.d.) filled with 1.5-µm solid core particles. Air samples were separated on a Macherey-Nagel Nucleosil 100, 5-µm C18 HD column (250-mm length × 3-mm i.d., 20% carbon content). Before use, the columns were flushed with acetonitrile for 20 min and equilibrated with the applied mobile phase for 15 min. Solvents were degassed with helium for 5 min. Separations were carried out at 20 °C using the following binary mobile-phase gradients: Micra NPS 1.5-µm column; flow rate, 1 mL/min; from 15% solvent A to 90% A within 40 min (solvent A, 5% freshly distilled THF in water; solvent B, 100% acetonitrile). Macherey Nagel Nucleosil 100, 5-µm HD column; flow rate, 560 µL/min; from 50% acetonitrile/water to 100% acetonitrile within 37.5 min, then 22.5 min isocratic. RESULTS AND DISCUSSION Sample Pretreatment. The derivatization process requires that DNPH-coated cartridges contain either phosphoric3 or hydrochloric acid.14 The acidic extracts cause a slow hydrolysis of the reversed-phase column. The formed polysiloxane fragments (mainly m/z 239) are detectable by APCI(-) and generate a high background. Addition of an ammonium acetate buffer prior to analysis eliminated this problem. Sample preconcentration by a gentle flow of nitrogen caused memory effects and an increased background due to deposits of nonreacted DNPH and acetone-DNPH on the APCI discharge needle. The additional extraction step (see Experimental Section) solved this problem. HPLC Column Selection. Fifteen commercially available C18 phases, one C8 phase, and one cyano phase obtained from six different manufacturers were tested for selectivity, signal tailing, and MS background. Performance criteria were the separation of the C4 carbonyls crotonaldehyde, methacrolein, 2-butanone, and n-butyraldehyde and of o-, m-, and p-tolualdehyde. Most C18 columns had a comparable selectivity. However, only Nucleosil 100, 5-µm C18 HD was able to separate all tolualdehydes and the four C4 carbonyls with a valley of maximum 30-40% (see Figure 1). This phase had the highest organic carbon content, which is probably the reason for the different separation behavior. In addition, the MS background between the columns varied by a factor of 104. Both column types applied in this work showed a low bleeding. Ionization Conditions and Parameter Optimization. Electrospray ionization (ESI) in the positive and negative ion modes showed insufficient detection limits due to strong fragmentation. Even a quantity of 250 ng gave a very weak quasi-molecular ion. The same was valid for APCI in the positive ion mode, which required at least 50 ng. APCI(-) generated mass spectra with only the [M - H]- ion present. A total of 100 pg was sufficient to obtain S/N ratios of ∼20:1 in the HPLC base peak chromatogram (scan range 200-1000 u). All quantities are based on the

Figure 1. (A) HPLC/MS base peak chromatogram obtained by APCI(-) of a mixture of DNPH carbonyls (mass range, 200-500 u). The following carbonyl derivatives (1 ng) were injected: (1) formaldehyde, (2) acetaldehyde, (3) acetone, (4) propionaldehyde, (5) 2-butenal, (6) methacrolein, (7) butanone, (8) n-butanal, (9) benzaldehyde, (10) isopentanal, (11) n-pentanal, (12) o-tolualdehyde, (13) m-tolualdehyde, (14) p-tolualdehyde, (15) n-hexanal, and (16) 2,5dimethylbenzaldehyde. A Nucleosil 100, 5-µm C18 HD column was used (see Experimental Section). (B) Mass chromatogram of m/z 237 ([M - H]-) obtained by full-scan APCI(-) of 10 pg of (3) and (4). Table 1. Molecular Mass Series of DNPH Carbonyl Derivatives with Different Substructures mass of [M - H]- ion

substructure

209 + n × 14 225 + n × 14 235 + n × 14 285 + n × 14 301 + n × 14 417 + n × 14 433 + n × 14

saturated, not alicyclic saturated hydroxycarbonyls or acids one double bond or saturated ring aromatic phenolic dicarbonyls, R-hydroxycarbonyls oxo acids

carbonyl amount of DNPH derivative. A total of 10 pg of carbonylDNPH (1 pg/µL; 10 µL injected) gave a S/N ratio of 3:1-10:1 in [M - H]- mass chromatograms (scan range 200-500 u; see Figure 1). Therefore, APCI(-) was selected for all further studies. Mass Spectra of Reference Carbonyl DNHP Derivatives. The mass spectra of 30 reference carbonyl-DNPHs (see Experimental Section) obtained by APCI(-) showed only the [M - H]ion (recorded mass range, 150-500 u). No ion adducts were observed. Table 1 summarizes the [M - H]- masses of carbonyl homologues with different substructures. Dicarbonyl- and R-hydroxycarbonyl-DNPHs could not be differentiated since they react to the same DNPH derivative. Ion trap CID mass spectra were also recorded. The fragment ions obtained for different structures and their relative abundances Analytical Chemistry, Vol. 70, No. 9, May 1, 1998

1981

Table 2. Fragment Ions in the MS/MS Mass Spectra of Different Structure Classes of Carbonyl DNPHs Obtained by Trap CID fragment ion (m/z)a 120 152 153 163 167 179 182 191 [M - H] - 30 [M - H] - 31 [M - H] - 33 [M - H] - 35 [M - H] - 45 [M - H] - 46 [M - H] - 47 [M - H] - 72 [M - H] - 77 [M - H] - 93 [M - H] - 164 [M - H] - 180 [M - H] - 183 a

aldehyde, saturated unbranched branched xx xx xx xxx xx xx xx xx xx -

xx xx xx xxx xx x xx xx xx -

aldehyde double bond in R-position

aldehyde, aromatic

salicylaldehyde

furfurylaldehyde

ketone saturated

acetophenone

dicarbonyls

xx xxx x xxx xxx xx xx var xx x xxx xxx xxx x -

xx xxx xx xxx var xx -

x xx xx xxx x xxx xx -

xx xx xx xxx xx -

x xxx x xx xxx var xx -

x x x xxx xx x x -

x xxx x x x x x

Relative abundance: -, <2%; x, 2-10%; xx, 10-50%; xxx, >50%; var, variable, but contains additional information.

Figure 2. Ion trap MS/MS spectra of the [M - H]- ion (m/z 265) of different C5-carbonyl-DNPH: (A) n-pentanal-DNPH, (B) isopentanalDNPH, (C) 3-pentanone-DNPH, and (D) methyl isopropyl ketone-DNPH. Some structure significant ions are marked.

are listed in Table 2. Figure 2 shows that the CID spectra of isomers are quite different. Neutral fragment losses and characteristic ions in the medium mass-range allow us to distinguish them. Structure Elucidation by MS/MS. A structure elucidation scheme for DNPH carbonyls was established by interpreting 30 MS/MS spectra and 80 MSn spectra from 11 selected compounds. 1982 Analytical Chemistry, Vol. 70, No. 9, May 1, 1998

A summary is shown in Figure 3. Substructures could be identified as described below. (1) Identification of DNPH Derivatives. The 2,4-dinitrophenylhydrazone rest gives characteristic fragment ions at m/z 120, 122, 152, and 179. Their intensities can be rather low, as in the spectra of dicarbonyl-DNPHs. The fragmentation sequence leading to m/z 152 and 122 was derived by multiple MS/MS

Figure 3. Structure elucidation scheme for the identification of different substructures in carbonyl-DNPH by APCI(-). A selection of structure significant ions is given.

Figure 4. Fragmentation sequences and structures of fragment ions identified by APCI(-) MSn experiments: (A) formation of ions m/z 152 and 122; (B) formation of ion m/z 163; (C) structure of the [M - H]- ion of a dimerization product formed during derivatization.

experiments as [M - H] f [M - H - 30] f m/z 152 f m/z 122 f m/z 94 (see Figure 4a). (2) Differentiation between Aldehydes, Ketones, and Dicarbonyls. A strong m/z 163 daughter ion is typical for aldehyde-DNPHs (see Figure 2). It is the base ion for aliphatic aldehydes and the relative intensity is >50% for aromatic aldehydes. This ion is either absent or its relative abundance is <2% in mass spectra of ketone-DPNHs. Aldehyde-DNPHs show also an intensive m/z 179 ion. Dicarbonyl-DNPHs are characterized by a low abundance of m/z 163 and 179, a strong m/z 182 ion, and almost no other fragments between m/z 100 and 220. The MSn spectra of six aldehydes revealed that m/z 163 is stabilized by the formation of an additional pyrazole ring after loss of the nitro group in the 2-position (see Figure 4b).

(3) Distinction between r-Saturated and Unsaturated Aldehydes. R-unsaturated aldehydes show a typical loss of 47 u [HNO2] from the [M - H]- ion. R-saturated aldehydes cleave off neutrals with mass 30 [NO], 45 [NO2 + H] and 46 [NO2]. Loss of HNO2 is probably favored by a cyclization reaction which stabilizes the formed fragment. The substitution pattern at the double bond can also be determined. A rest R on the β-C atom (HCOCHdCHR) results in the ion series [M - H - 31]-, [M - H - 33]- and [M - H 35]- (example 2-butenal). A missing substitution in β-position gives only [M - H - 31]- and [M - H - 35]- (example 2-methylpropenal). [M - H - 33]- is probably formed by a rearrangement reaction corresponding to the Fischer indole synthesis (loss of H2CdNH). This assumption is supported by Analytical Chemistry, Vol. 70, No. 9, May 1, 1998

1983

Figure 5. APCI(-) base peak chromatogram of a preconcentrated ambient air sample (mass range, 200-500 u). The elution ranges of carbonyls with the same C number are shown.

the simultaneous appearance of [M - H - NH3]- indicating the formation of indole-like structures. (4) Differentiation between Branched and Straight-Chain Aldehydes/Ketones. Because of different reaction rates in the atmosphere, a differentiation between branched and straight-chain saturated aldehydes/ketones is very important. They are difficult to separate by HPLC. Their retention times are often almost identical.12 Neither UV detection nor single-stage HPLC/MS are able to differentiate them.10 However, the APCI(-) MS/MS mass spectra of n-butanal/isobutanal, 2-butenal/2-methylpropenal and n-/isopentanal (see Figure 2) are clearly distinguishable. The limited number of compounds studied so far, did not allow to interpret the observed differences in detail. The spectra of six saturated DNPH ketones indicated also that branched and straightchain structures can be distinguished (see Figure 2). Mass spectra of more structures are needed to draw further conclusions. (5) Identification of Aromatic and Unsaturated Aldehydes. A double bond in a position R to an aldehyde group can be identified by formation of m/z 167, [M - H - 77]- and [M - H - 72]- (see propenal, 2-butenal, methylpropenal, trans-2-pentenal, and 2-hexenal). These masses are missing for aromatic aldehydes which show an abundant [M - H - 164]- ion. Only orthosubstitution forms an intense [M - H - 93]- fragment. Application to Air Samples. The structure elucidation scheme was applied to real air samples using HPLC/MS, MS/ MS, and MS3. In addition, the presence of minor and trace constituents was studied in the concentrated sample extract (see Experimental Section). At the trace level, numerous coelutions between major and minor carbonyls were observed (see Figure 5). For example, the HPLC signal of n-pentanal (16.70 min, Figure 5) contained an abundant m/z 299 [M - H]- ion. 1984 Analytical Chemistry, Vol. 70, No. 9, May 1, 1998

Other minor coeluting compounds formed m/z 275, 277, 415, 433, and 461 (C2 and C4 oxo acids?) as well as m/z 445 and 459 (C4 and C5 dicarbonyls). The coelutions were not observable with UV detection. The major coeluting compound was identified as follows. Mass m/z 299 indicated an aromatic carbonyl (see Table 1). Possible structures were o-tolualdehyde, m-tolualdehyde, p-tolualdehyde, acetophenone, and 2-phenylethanal. Since m/z 163 was not present in the MS/MS spectrum, the compound had to be a ketone. An aliphatic ketone could be excluded by the fragmentation pattern (see Table 2), leaving acetophenone as the only remaining candidate. Its presence was confirmed by a reference standard showing the same MS/MS spectrum and retention time. Occasionally, the chromatograms of air samples showed a large signal in the C2-C3 carbonyl range. Its presence was also reported by Grosjean et al.8 who called the compound U3. They proposed a hydroxycarbonyl as structure based on the UV spectrum and retention time. Therefore, different hydroxycarbonyls were synthesized.16 However, none of them showed a comparable UV spectrum. The APCI(-) spectrum of the disturbing compound showed a [M - H]- ion of m/z 363 which was unexpectedly high. The MS/MS spectrum contained [M - H NO]- (m/z 333) and [M - H - NO2]- (m/z 317) as major fragments and unusually few ions in the low-mass range. The masses for a DNPH group were also present but not abundant (m/z 109, 122, 152, and 182). The MS3 spectrum of m/z 317 showed twice the loss of NO and NO2. The molecular weight and the fragmentation pattern indicated as structure a reaction product between 2,4-dinitrophenylhydrazone and 2,4-dinitro(16) Dye, C. Norwegian Institute for Air Research, N-2007 Kjeller, Norway, June 1997.

chlorobenzene (see Figure 4C). The latter is formed as a derivatization artifact when high levels of nitrogen dioxide are present in the air and HCl is used as acid.10 CONCLUSION APCI(-) mass spectrometry is the most suitable ionization technique for DNPH carbonyl derivatives. Only [M - H]- ions are formed. A 1-10-ng sample is sufficient to record MS/MS to MS4 mass spectra during HPLC separations. DNPH carbonyls could be identified by MS/MS. The MS/MS spectra allowed differentiation between aldehydes and ketones, straight-chain and branched structures, and unsaturated and aromatic carbonyls. In this way, the structure of coeluting DNPH carbonyls and of reaction byproducts could be identified in ambient air extracts.

The elimination of contamination sources allowed to detection lowpicogram quantities in ambient air samples. ACKNOWLEDGMENT We thank Dr. Herbert Abele for valuable discussions and test columns. The help of Leo Merz and Benny Galliker with the synthesis of some DNPH carbonyls is greatly appreciated. The financial support by the Swiss Science Foundation under Project 2100-046665.96 and a grant from the Chemical Industry in Basel (S.K.) is gratefully acknowledged. Received for review August 26, 1997. Accepted February 19, 1998. AC9709458

Analytical Chemistry, Vol. 70, No. 9, May 1, 1998

1985