Analytical Methodology for Determination of Organic Aerosol

High-performance liquid chromatography (HPLC) studies of bulk aerosols can ..... This material is available free of charge via the Internet at http://...
0 downloads 0 Views 660KB Size
Article pubs.acs.org/ac

Analytical Methodology for Determination of Organic Aerosol Functional Group Distributions Alicia J. Kalafut-Pettibone and W. Sean McGivern* Chemical Science Division, National Institute of Standards and Technology, 100 Bureau Drive, Stop 8320, Gaithersburg, Maryland 20899, United States S Supporting Information *

ABSTRACT: Secondary organic aerosol (SOA) particles result from the condensation of oxidized volatile organic compounds (VOC) and consist of a complex mixture that is not conducive to complete compositional analysis. We present a simple methodology for determining the quantities of aldehydes and ketones, alcohols, and carboxylic acids in such samples via derivatization and high-performance liquid chromatography (HPLC) with ultraviolet−visible absorption detection and tandem mass spectrometry (MS/MS). Useful concentration ranges are presented with an examination of the specificity of these compounds in multiply substituted systems. Such data are valuable in modeling the formation of laboratory-generated aerosols and in identifying point sources of field-collected aerosol samples. Calibration curves on standard samples, MS/MS transitions, including collisionally induced dissociation (CID) products, and a quantitative examination of the specificity of the derivatization reagents toward multiple functional groups are presented for a series of aliphatic standard samples. These methods are then applied to filter extracts from SOA derived from 1-iodooctane photolysis at 254 nm to demonstrate the methodology on a complex system. Ultimately, this methodology enables the measurement of variations in the chemical nature of the oxygen within an SOA particle, providing a distribution of functional group concentrations.

S

a need for more detailed chemical information call for additional complementary techniques for determining functional group distributions. High-performance liquid chromatography (HPLC) studies of bulk aerosols can provide data on the molecular composition of the oxygenated species in SOA, and numerous research groups have utilized these methods, often coupled to mass spectrometry and in tandem with other techniques.13−17 In general, however, these HPLC methods focus on a single class of compounds, and the extreme complexity of the mixtures in SOA limits the ability to cleanly identify individual compounds. Tandem mass spectrometric detection (MS/MS) can be effective, but ionization of many compounds in SOA is poor, often yielding no significant signal from aldehydes, ketones, and alcohols and only a weak negative [M − H]− signal for carboxylic acids. Numerous groups have applied size-resolved aerosol mass spectrometry via volatilization/electron impact ionization18 to oxidized products in both field19,20 and chamber21−27 samples. However, these size-resolved experiments are necessarily limited in their ability to identify individual compounds or

econdary organic aerosols (SOA) represent a significant fraction of the overall condensed-phase atmospheric carbon load and are formed from gas-phase oxidation of volatile organic compounds (VOC) and subsequent condensation of the oxidation products. These reactions lead to complex mixtures of oxidized organic species, which undergo further transformation through atmospheric aging. The measurement of the composition of SOA, both from laboratory and field samples, is an area of vigorous study. However, the extreme complexity of even the simplest SOA renders a complete chemical analysis of the composition intractable.1 Numerous approaches to functional group analysis have been used to probe aerosol samples. Fourier transform infrared (FTIR) spectroscopy has been utilized extensively to provide a “semiquantitative”2 distribution of functional groups in aerosol samples.2−12 Blando et al. have examined the use of FT-IR to probe functional group distributions in standards, laboratorygenerated particles, and ambient samples.2 They found that FTIR was capable of determining functional group concentrations on aerosol particles as a whole without bias from filtration and extraction. However, these whole-aerosol techniques were observed to be sensitive to sampling geometry (e.g., particle distributions on impactors). In addition, knowledge of the relative absorptivity of the moiety of interest relative to that of an ion used to determine absolute concentration chromatographically is required. Such uncertainties in utilizing FT-IR and This article not subject to U.S. Copyright. Published 2013 by the American Chemical Society

Received: October 3, 2012 Accepted: March 5, 2013 Published: March 5, 2013 3553

dx.doi.org/10.1021/ac3028728 | Anal. Chem. 2013, 85, 3553−3560

Analytical Chemistry

Article

on systems broadly consistent with SOA constituents. (Certain commercial materials and equipment are identified in this paper in order to specify adequately the experimental procedure. In no case does such identification imply recommendation of or endorsement by the National Institute of Standards and Technology, nor does it imply that the material or equipment is necessarily the best available for the purpose.) Derivatization. The derivatization reagents were chosen from the literature and subsequently optimized for coreagent type, relative concentrations of reagents and samples, and temperature using model compounds similar to those found in SOA. Emphasis was placed on finding reagents that would enhance both ESI ionizability and UV absorption response. All reagents and coreagents and their respective concentrations were required to maintain compatibility with standard HPLC columns, show good separation from other reagents and the analytes, and require mild conditions with minimal handling requirements. The reagents and coreagents chosen used are shown in Table 1 with their respective functional group targets.

compound classes. In general, these experiments can provide elemental ratios on nonrefractory particle samples in aerosols, the most important of which for SOA are O/C and H/C.28 In general, derivatization can be used to enhance the chromatographic separability and detectability of these compounds. Kowalewski and Gierczak describe a methodology to probe functional group distributions by derivatizing carbonyls, alcohols, and carboxylic acids to enhance volatility for analysis in a gas chromatograph.29 This technique is effective but requires difficult-to-handle silyl reagents and will ultimately be limited by the vapor pressures of the analytes. Similar gas-phase derivatization studies have been performed by Jaoui et al.,30 Li and Yu,31 and Jang and Kamens32 with similar advantages and disadvantages. Liquid chromatography methods offer the ability to probe low-volatility species, and numerous derivatization methodologies for HPLC analysis of SOA samples have been described in the literature. However, these methods require extremely difficult-to-handle reagents or numerous steps prior to analysis.33−38 We have developed and characterized a simpler methodology to examine the functional group distribution of compounds typically observed from SOA based on derivatizations that have been utilized previously in the biochemical community39,40 and in gas-phase atmospheric studies.41,42 The derivatization reactions, which are specific to aldehydes and ketones (referred to as “carbonyls” in the remainder of this work), carboxylic acids, and alcohols, enhance both the ultraviolet absorption and the electrospray ionization (ESI) of compounds containing these three functional groups. In this work, we utilize three separate derivatization reactions to probe a series of oxygenated aliphatic standard compounds similar to those found in newly formed SOA and characterize their ultraviolet responses. We have chosen singly and doubly substituted standard compounds to determine useful concentration ranges for these species by measuring absorption response linearity and specificity of these derivatization agents toward their target functional groups. In particular, we present relative ultraviolet absorption responses for the derivatization reagents and calibration curves with appropriate limits of detection and linearity ranges. This methodology utilizes acetonitrile (ACN) as a solvent and is conducive to the analysis of ACN filter extracts. Finally, the applicability of these reagents to more complex systems is shown via the analysis of SOA formed from the photolysis of 1iodooctane in air.11

Table 1. Summary of Derivatization Reagents and Targetsa reagent

target

dansylhydrazine (DnsHz) 4-(dimethylamino)benzoyl chloride (DMABC) 2-picolylamine (2-PA)

coreagents (equivalents)

R1(CO)R2 RCHO ROH

trifluoroacetic acid (0.2)

RCOOH

triphenylphosphine (1) dipyridyldisulfide (1)

pyridine (5, 1)

a

Coreagent quantities are provided as molar equivalents of derivatization reagents.

Aldehydes and ketones are analyzed via condensation of the sample with DnsHz to form hydrazones (Scheme 1).43,44,47 A Scheme 1. Ketone/Aldehyde Derivatization To Form Hydrazone with DnsHza



a

CID of the molecular ion results in cleavage of indicated bond and a product ion with m/z 170.

EXPERIMENTAL SECTION Derivatization for various functional groups was accomplished using 5-dimethylaminonaphthalene-1-sulfonyl hydrazine (dansyl hydrazine, DnsHz), 2-aminomethylpyridine (2-picolylamine, 2-PA), and 4-(dimethylamino)benzoyl chloride (DMABC) to derivatize carbonyls, carboxylic acids, and alcohols, respectively.42−46 All derivatizations were performed in ACN through mixing of the analyte with the active derivatization reagent and necessary catalysts/coreagents in a single vial. In all cases, the addition of an aromatic functionality from the derivatization agent provides strong ultraviolet absorption, and the presence of the nitrogen, whether as a heteroatom in the aromatic ring or as part of a dimethylamine moiety, leads to strongly enhanced ionization in the electrospray ionizer. In addition, each reagent is found to show good specificity to its target functional group. In this work, we have elected to focus primarily on octyl derivatives as standard samples in order to minimize interfering signals while testing

mixture of 100 μL of 19 mmol L−1 (5.0 mg mL−1) DnsHz was added to 100 μL of analyte in ACN. In general hydrazone formation can be acid- or base-catalyzed; in the present experiments, we have chosen to add 10 μL of 2.5% (vol) trifluoroacetic acid (TFA) in ACN (0.2 mol equiv) to maintain acidic conditions, leading to complete reaction in 5 min at room temperature. In addition to 2-octanone as a standard, octanal was also used as an analyte and confirmed that aldehydes derivatize equivalently, although a complete set of experiments for the aldehyde including calibration curves and specificity was not performed. Hydrazone formation is reversible in aqueous solutions at neutral pH, particularly in the presence of electron-withdrawing groups.4 We utilize an acetic acid solution as the aqueous HPLC mobile phase to maintain a sufficiently low pH to prevent hydrolysis. No 3554

dx.doi.org/10.1021/ac3028728 | Anal. Chem. 2013, 85, 3553−3560

Analytical Chemistry

Article

were used with a 1 cm guard column in front with the same stationary phase as the main column. A Pinnacle II C18 (Restek Corporation, Bellefonte, PA) chromatography column with 3 μm particles was used for all analyses described in this work. The typical chromatography method was a reversed-phase H2O/organic gradient method ranging from 20% ACN to 90% ACN over 8 min and held for several minutes (varying depending on the amount of derivatizing agent). Injections of 2 or 20 μL were used, depending on the concentration of the standard. The entire detector UV−vis absorption spectrum, 190−800 nm, was recorded for all retention times; however, analyses were typically performed at the wavelength corresponding to the absorption maximum. Mass spectrometric analysis was performed on the output of the diode array detector and was accomplished with an Agilent 6410 triple-quadrupole mass spectrometer. A dual electrospray/ atmospheric pressure chemical ionization source (Agilent) was used; source parameters were optimized to the derivatized standards shown in Table 2. These source parameters were used for all analyses with the corresponding derivatizing reagent.

evidence for the reversible hydrolysis was observed during the chromatographic analysis. Carboxylic acids are analyzed from the amide formed via reaction with 2-PA in the presence of 1.0 mol equiv of two coreagents (Scheme 2): triphenylphosphine (TPP) and Scheme 2. Carboxylic Acid Derivatization To Form Amide with 2-PAa

a

CID results in cleavage of the indicated bond in the molecular ion to yield a product ion with m/z 109.

dipyridyl disulfide (DPDS).46 The formation of the amide is mediated by the oxidation of the TPP to a phosphine oxide and reduction of DPDS to a cyclic thioketone and has been found to be extremely rapid in a range of solvents over a large temperature range.48 To 100 μL of analyte in ACN 50 μL of 19 mmol L−1 solutions of 2-PA (2.1 mg mL−1), DPDS (4.3 mg mL−1), and TPP (5.0 mg mL−1) were added. Reactions were allowed to occur for a minimum of 20 min at room temperature prior to analysis. Alcohols are detected as esters formed from the reaction of DMABC in the presence of pyridine (Pyr) in a slight excess.45 The formation of the ester is facilitated by the formation of an acetylpyridinium ion formed from displacement of chloride ion with pyridine.49 The basic DMABC solution was colorless; addition of pyridine turned the solution a deep yellow. DMABC is moisture-sensitive; the use of undried ACN or lengthy exposure to room air will result in the formation of 4(dimethylamino)benzoic acid crystals and deactivation of the reagent. In practice, individual vials of DMABC solution in dry ACN (dried for several hours over 3 Å molecular sieves in a desiccator) were prepared in a glovebox and sealed. For derivatization, the vials were warmed to room temperature and briefly opened to room air to allow addition of sample and pyridine solutions. With this handling, vials of the derivatizing reagent should be used in approximately 8−12 h. Longer useful life of the reagent may be obtained by adding anhydrous pyridine (1 mol equiv) to a DMABC solution in anhydrous ACN in an inert environment and storing the resulting yellow mixture below 0 °C for direct transfer to the sample vial. Reagent should be removed anaerobically, such as with a Mininert valve (Vici Valco, Houston, TX) and syringe, to prevent deactivation for quantitative work. Typically, 400 μL of sample in ACN was added to 400 μL of 38 mmol L−1(7 mg mL−1) solution in ACN with 6 μL Pyr (5 equiv). The vial was heated to 70 °C for 20 min prior to initiating chromatographic analysis, at which point the solution is colorless. As noted below, these conditions will also produce ∼30% of an anhydride when reacted with a carboxylic acid. Chromatographic Analysis. Liquid chromatography was performed with a Dionex UltiMate 3000 system, consisting of an HPG-3200SD binary pump, autosampler, column oven, and DAD-3000RS diode array ultraviolet−visible (UV-vis) detector with a 10 mm absorption cell. In all experiments, the column oven was held at 40 °C. Complete UV-vis spectra were collected at 10 Hz and were measured with the “wide” slit width setting (yielding a resolution of ∼5 nm). Chromatography columns were 2.1 mm in diameter and 5 cm long and

Table 2. Limiting Quantities for Octyl Derivatizations

a

analyte/deriv. agent

wavelength/ nm

LOD/ nga

LOQ/ ngb

relative abs.

2-octanone/DnsHz 2-octanone/DnsHz 1-octanol/DMABC decanoic acid/2-PA

215 254 308 254

4.3 4.1 0.5 4.1

7.1 5.9 2.5 5.1

17 ± 3 5.9 ± 1.1 6.4 ± 1.2 1

Signal-to-noise ratio >3. bSignal-to-noise ratio >10.

Chemicals. All reactions were performed in HPLC-grade ACN (Aldrich CHROMASOLV Plus, ≥ 99.9%). When necessary for derivatization, ACN was dried over 3 Å molecular sieves (Aldrich). The derivatizing reagents were obtained from Aldrich and used without further purification: DMABC (97% and ≥99.0%), pyridine (≥99.0%), 2-PA (99%), TPP (≥98.5%), DPDS (≥99.0%), DnsHz (98%), and TFA (99%). Standard samples were obtained from Aldrich and used as received: 1-hexanol (99%), 1-octanol (≥99%), 2-octanone (98%), octanoic acid (≥99%), octanedioic acid (98%),1,6hexanediol (99%), 2,5-hexanedione (>96%), 5-hydroxy-2pentanone (≥95%, mixture of monomer and dimer), 7-oxooctanoic acid (≥98%), 10-hydroxydecanoic acid (technical grade, up to 15% lactones), and decanoic acid (≥98%). Liquid chromatography was performed with HPLC-grade acetonitrile and 0.1% (volume) acetic acid (Aldrich) in 18 MΩ water (generated with a Water Pro PS filtration system, Labconco Kansas City, MO). In systems containing a free acid moiety (Table 3), a 50 mmol L−1 ammonium acetate buffer with pH 4.0 was used to improve peak shapes after confirmation of the peak identities with MS/MS using 0.1% acetic acid solution.



RESULTS AND DISCUSSION Chromatography and Detection. The primary challenge in analyzing the present single-reactor derivatization products is the separation of the products of interest from the typically large excess of derivatizing reagents. Although a second solidphase extraction step could remove these larger products, such an analysis adds both complexity and the possibility of product

3555

dx.doi.org/10.1021/ac3028728 | Anal. Chem. 2013, 85, 3553−3560

Analytical Chemistry

Article

Table 3. Specificity of Derivatization Reagentsa active reagent singular std functional groups dual standard R(OH)COOH 10-hydroxydecanoic acid R(OH)CO 5-hydroxy-2-pentanoned R(CO)COOH 7-oxooctanoic acid R(OH)2 1,6-hexanediol R(CO)2 2,5-hexanedione R(COOH)2 octanedioic acid

DnsHz

2-PA

DMABC

2-octanone

decanoic acid

1-octanol/1hexanol

NAb

0.99

>0.4c

0.92

NA

0.68

0.97

1.2

0.69

NA

NA

0.57(0.64)

0.82(0.04)e

NA

NA

NA

0.97(0.0)

0.49(0.15)

Scheme 3. Alcohol Derivatization with DMABC To Form Estera

a

Both H+ and Na+ adducts of the parent mass are observed; CID of both results in cleavage of the indicated bond and an m/z 148 daughter ion. The protonated ion also yields an m/z 166 daughter.

ultraviolet absorption coefficients of the substituted species to permit quantitative correction of measured chromatograms. In determining these limits a 20 μL injection volume was used, although the increased background resulting from tailing on the overloaded derivatizing reagent peaks limited the overall advantage in sensitivity from the larger injection. The wavelengths shown in the table were chosen to be very near the absorption maximum above 200 nm. For the single aromatic ring reagents 2-PA and DMABC, the π → π* B-band transitions near 254 and 308 nm, respectively, dominate the UV−vis spectrum. However, the naphthyl moiety absorbs on the stronger E1 and E2 transitions at wavelengths longer than 200 nm, with the E1 band at 215 nm dominating. We have included both bands in Table 2 in case interferences preclude the use of the stronger E1 band wavelength. The linearity of the ultraviolet response was evaluated with a series of standard 5× dilutions of 1.000 ± 0.009 mL samples delivered with an adjustable volume pipettor (Eppendorf, 1000 μL) to a 5.00 ± 0.02 mL volumetric flask. In all cases, the response of the derivatized sample was linear over at least 2 orders of magnitude and extrapolated smoothly to zero response. Calibration curves and relevant fitting parameters are provided as Supporting Information. Specificity and Functional Group Quantitation. The specificity of these derivatization reagents to their target functional groups was evaluated by reacting each reagent with a mixture of the singly substituted standards (Table 2, with hexanol replacing octanol for better chromatographic separability) at 2.5 mmol L−1 concentrations. The resulting chromatograms (UV absorption) are shown in Figure 1. All

a

Values in parentheses are relative amounts of the doubly derivatized standards to the singly derivatized standards. Values in parentheses are half of the relative amounts of the doubly derivatized standards relative to the singly substituted standards and sum to unity for complete derivatization and linear scaling of the absorption. Concentration uncertainties are estimated to be 5%; however, systematic uncertainties may exceed 50% in some cases (see text). bNot reported because there was no reaction toward either functional group. cPeak overlap precludes more detailed analysis. dMixture of monomer and dimers. e Derivatized product underwent dehydration of OH functionality ([M − 18 + H]+ observed in MS).

loss. The aliphatic standard samples derivatized with DnsHz and DMABC are typically well-separated on a C18 column with the method described above. Samples derivatized with 2-PA, which requires two additional coreagents in excess, showed good chromatographic separation with a 2 μL injection; however, a 20 μL injection showed overlap with the heavily tailing DPDS and the standard 2.5 mmol L−1 octanoic acid sample on a C18 column. A larger diameter column would alleviate this problem; however, we have elected to utilize decanoic acid as the standard compound to eliminate the overlap. In general, higher concentration standards (≥25 μmol L−1) were analyzed with a 2 μL injection, which provided excellent peak shapes and complete separation of all components for all three systems. At lower sample concentrations, with the same amount of derivatization reagent, a 20 μL injection was utilized. All three derivatization agents contain chromophores that provide a strong ultraviolet absorption; when combined with the aliphatic compounds in this study, the absorption coefficients are expected to vary little from those of the parent derivatization reagents. As discussed below, this can be used for approximate quantitation of the relative concentrations of compounds containing the functional groups. The derivatization reagents also provide ESI-ionizable moieties as shown in Schemes 1−3. In most cases, the [M + H]+ molecular ion is dominant, and multiple reaction monitoring (MRM) of ([W + 265 − 18 + H]+ → 170), ([W + 183 − 36 + H]+ → 148), and ([W + 108 − 18 + H]+ → 109), where W is the mass number of the analyte, can be used to differentiate compounds with overlapped peaks for derivatized samples with DnsHz, DMABC, and 2-PA, respectively. Limiting concentrations for the derivatization of the standard octyl compounds are shown in Table 2 with the relative

Figure 1. Chromatograms of derivatized samples of singly substituted standards. 2-PA and DnsHz are 254 nm bsorption signals; DMABC signals are taken at 308 nm.

reagents were strongly specific to their target functional groups with the exception of DMABC, which reacted to near completion with alcohols but showed a partial yield of anhydride from reaction with decanoic acid. Limiting the reaction time of the DMABC to 20 min at 70 °C results in a yield of ∼30% for the anhydride. Increasing the reaction temperature led to a greater proportion of unwanted anhydride, 3556

dx.doi.org/10.1021/ac3028728 | Anal. Chem. 2013, 85, 3553−3560

Analytical Chemistry

Article

and decreasing the temperature to 60 °C resulted in a decreased yield of derivatized samples even after double the reaction time and was deemed unacceptable. Although this additional derivatization is not ideal for many SOA samples due to typically large acid concentrations, DMABC still shows a strong preference for OH and represents a rare simple derivatization reagent for alcohols that can be analyzed without substantial additional sample processing for analysis. Additionally, we evaluated the behavior of the derivatization agents on molecules with more than one functional group. Derivatization was carried out on a series of mixtures each consisting of one singly substituted and one doubly substituted analyte as indexed by the columns and rows of Table 3, respectively. The derivatized mixtures were then analyzed chromatographically to determine their relative UV absorption signals. The results are presented in Table 3. The UV absorption coefficient for the chromophore added as a result of derivatization was assumed to be unchanged due to the presence of the second functionality; the veracity of this assumption is examined in the following section. Different chain lengths were used as shown in the table to ensure good chromatographic separation for the various components. In constructing Table 3, we have attempted to minimize interactions between the functional groups by using widely spaced functionalities. 2-Hydroxyoctanoic acid was also derivatized with 2-PA and DMABC. These results were compared to singly substituted standards decanoic acid and 1-octanol, respectively, to evaluate the effect of closely spaced moieties. In the 2-PA case, the ratio of derivatized products from 2-hydroxyoctanoic acid/decanoic acid was 0.9 ± 0.1 with the α-hydroxy acid adduct showing a broad double peak (vide infra). For DMABC, the ratio of derivatized products of 2hydroxyoctanoic acid/1-octanol was found to be 1.1 ± 0.2. In general, the derivatization reagents showed strong preferences for their intended moieties, yielding the same amount of derivatized product when the doubly substituted sample contained only one of the target functional groups to within ±20%. In the case of 2,5-hexanedione, a dehydration reaction of the singly derivatized product was observed wherein the analyte peak (confirmed to be the [M + H]+ molecular ion peak with MS) was observed to disappear with a first-order rate of 0.50 ± 0.07 h−1 with a concurrent increase in a separate peak (confirmed to be [M − 18 + H]+). The sum of the singly derivatized dione and the dehydrated derivatized dione showed only a very slight increase (6% h−1) over the 50 min span used to obtain the dehydration kinetics. We speculate this dehydration occurs via intramolecular nucleophilic substitution of the other carbonyl carbon but have made no further attempt to characterize the reaction. The doubly derivatized product peak showed no time dependence. In addition, it is believed that the low conversion for DMABC derivatization of 5hydroxypentanone is due to the presence of dimers in the standard that are not effectively derivatized. As indicated in Table 3, analytes containing two of the target moieties (e.g., a diacid derivatized with 2-PA) showed a mixture of singly and doubly derivatized products. Doubly derivatized product signals were scaled by 0.5 under the linear absorption assumption, and the sum of the parenthesized entries in the table and their corresponding partners is expected to be unity in the case of purely linear and complete specific derivatization. In each case products were confirmed with MS. In the case of DMABC, the formation of esters favors the formation of [M + Na]+ ions. Chains containing carbonyls in addition to the

DMABC-derivatized moiety formed sodium adducts, whether the carbonyl moieties were in the underivatized ketones or acids or in a doubly derivatized (diester) species. For distantly spaced functional groups, such as in octanedioic acid or 10hydroxydecanoic acid, [M + Na]+ is the primary ion formed. For more closely spaced moieties, such as in 1,6-hexanediol (doubly derivatized) both [M + Na]+ and [M + H]+ ions are formed. Finally, the α-hydroxy acid 2-hydroxyoctanoic acid (not shown in Table 3) showed no evidence for a sodium adduct, despite showing two distinct retention times, presumably corresponding to derivatization of the hydroxy moiety or the OH in the carboxylic acid. The collisionally induced dissociation (CID) of any [M + H]+ species leads to both m/z 148 and m/z 166; that from [M + Na]+ species leads to only m/z 148. On the basis of these observations we have chosen to recommend m/z 148 as the primary diagnostic peak for identifying CID products of DMABC derivatization, despite the more obvious choice of m/z 166 based on the typical reaction scheme (Scheme 3). The relatively large deviations from unity (up to 20%) found in the table are most likely due to deviations from linearity in the ultraviolet absorption of variously substituted derivatized analytes. In addition, particularly with DMABC but to a lesser extent with DnsHz, the freshness and water content of the standard solutions had an effect on the completeness of the derivatization. Better results would be obtained if all of these reactions were done strictly in an inert environment. However, the emphasis in this work is on ease and speed of analysis at the possible expense of uncertainty. Within such a framework, variations on this order are certainly acceptable, and these compounds provide good overall specificity toward their target functional groups. Quantitation from Ultraviolet Absorption. As discussed previously, the absorption coefficient of a chromophore from the derivatization reaction was assumed to be unaffected by the analyte chain length or presence of other functional groups. The detection wavelengths, shown in Table 2, were chosen to provide maximum signal with minimum interference. Under these conditions, the doubly derivatized products in Table 3 are expected to exhibit twice the molar absorption as the singly derivatized products. HPLC spectra of derivatized hexanoic and decanoic acids are shown in Figure 2. Both of these systems, which are purely aliphatic beyond the derivatization substitution, yield very similar peak sizes at 254 nm, where the absorption derives solely from the aromatic chromophore. Under this assumption, individual concentrations of compounds containing OH, COOH, and CO (i.e., a functional

Figure 2. Chromatograms of hexanoic and decanoic acids derivatized with 2-PA. 3557

dx.doi.org/10.1021/ac3028728 | Anal. Chem. 2013, 85, 3553−3560

Analytical Chemistry

Article

rapidly forms an octyl peroxy radical. As shown in Scheme 4, under these low NOx conditions, the peroxy radicals are

group distribution) can be derived from the base absorption of individual derivatized products and the appropriate relative absorptivity scaling (Table 2). Although the absorption linearity assumption is quite strong, several factors favor its applicability. Group additivity for ultraviolet absorption is effective in estimating molar absorptivity of molecules,50 and chromophores at various locations along a carbon backbone tend to be linearly additive. The linearity assumption can fail dramatically in compounds containing conjugated double bonds or aromatic moieties. The first case is unlikely to pose major problems in SOA analysis. Exposed double bonds will often have undergone reaction during the particle formation and aging, although SOA has been found to include some olefinic compounds, such as oleic acid observed in ambient samples.51 Even in such cases, however, a significant shift in the UV absorption spectrum will be present only in vinylic substitutions that lead to conjugation in the π-system. For reaction with OH, addition across the double bond would be strongly favored over abstracting the vinylic hydrogen atom for O2 to initiate oxidation. The second case involving aromatic compounds could yield drastic deviations in linearity if the derivatization occurred on a functional group bound directly to an aromatic. Substitution of these highly aromatic derivatization reagents will have a large nonlinear impact on the final molar absorptivities and absorption maxima and would lead to a very poor estimate of the functional group distribution. Consideration of the ultraviolet absorption spectrum of the underivatized product may provide evidence that such a nonlinearity is present, such as if an unexpectedly large peak relative to the collected sample mass or a large and unexpected shift in the absorption maximum was observed. Overall, the presence of aromatic species represents an additional challenge that will demand the use of an alternative technique to ensure quantitative results. In the end, the goal of this work is not to provide rigorous absolute measurements but to provide a methodology to enable the rapid measurement of functional group concentrations with greater certainty than previous methods. Even measurements accurate to only a factor of 2× to 3× are improved from previous work and are expected to offer significant value in understanding and modeling oxidized VOC volatility and subsequent SOA formation. Further work to better elucidate the limitations of the methodology in these areas, particularly when derivatizing aromatic substituents, is currently underway. Although this methodology does enhance the separation of many components, the presence of an unresolved complex mixture is still a distinct likelihood. In this case, the enhanced product ionizability and retention time parameterizations provided by the derivatization reactions can provide much greater insight into the nature of this unresolved mixture. In any case, the complex mixture will contribute to the overall decrease in the derivatization reagent concentration and will be included in a total functional group distribution analysis. Application to Synthetic SOA. As a demonstration of the applicability of this methodology to complex particle systems, we have utilized it to probe the functional group distribution in SOA formed from the 254 nm photolysis of iodooctane in a flow cell. The photolytic generation of iodooctane SOA was originally described by Kessler et al. and was studied using aerosol mass spectrometry to determine the O/C and O/H ratios as well as the average number of oxygen atoms per parent molecule in the SOA particle.11 Photolysis of the iodide leads to relatively unreactive iodine atoms and octyl radical, which in air

Scheme 4. Octyl Radical Oxidation to Condensable Productsa

a

Stable major products are shown in bold.

expected to primarily undergo reaction with other peroxy radicals to form stable alcohol and aldehyde products or alkoxy radicals.11 These alkoxy radicals undergo rapid isomerization or H-abstraction by O2 to form aldehyde and HO2 (not shown in the scheme) The alkyl radicals formed from isomerization rapidly react again with O2 to form hydroxy alkyl peroxy radicals, which undergo similar peroxy radical reactions to form 1-hydroxy-4-octanone, 1,4-octanediol, and a hydroxy alkoxy radical, which predominately reacts to form 4-hydroxyoctanal. These reactions lead to a reasonably complex mixture of compounds that is conducive to functional group analysis. We have performed a similar experiment to that of Kessler et al.11 utilizing 12 uncoated fluorescent light bulbs (117 cm, 32 W) for 254 nm photolysis in a quartz laminar flow tube with a length of 183 cm and an inner diameter of 14.0 cm at a total residence time of 3.7 min. Photolysis of iodooctane occurred over the length of the fluorescent bulbs. Under these conditions, iodooctane concentrations of 2−12 μL L−1 in air lead to the formation of SOA products, which were monitored by a scanning differential mobility analyzer (TSI, Inc. model 3080L) and an n-butanol condensation particle counter (TSI, Inc. model 3775) and collected on quartz microfilters (Munktell) for a time sufficient to obtain a few hundred micrograms of sample, 2−24 h. These filters were sonicated in dry HPLC-grade acetonitrile for 1 h, filtered, and analyzed with the present methodology. A typical set of spectra are shown in Figure 3 for the DMABC (hydroxyl moiety) derivatization of the filter extract. Numerous peaks, most of which are well-resolved, are present corresponding to compounds with different OH substitution patterns. In these spectra, DMABC of higher purity (HPLC grade, ≥ 99%) was used to minimize background peaks in the complex chromatogram. A complete analysis of these peaks is beyond the scope of the present work and will be the subject of a future publication. Overlaid on the spectra are the peak integrals. Provided the assumption of linear ultraviolet response is valid, the total integrals of all peaks should remain identical in the blank and sample spectra. The measured total integrals shown in the figure differ by