Environ. Sci. Technol. 2005, 39, 8728-8735
Secondary Organic Aerosol Formation by Glyoxal Hydration and Oligomer Formation: Humidity Effects and Equilibrium Shifts during Analysis WILLIAM P. HASTINGS, CHARLES A. KOEHLER, EARL L. BAILEY, AND DAVID O. DE HAAN* Chemistry Department, University of San Diego, 5998 Alcala Park, San Diego California 92110
Glyoxal is a significant atmospheric aldehyde formed from both anthropogenic aromatic compounds and biogenic isoprene emissions. The chemical behavior of glyoxal relevant to secondary organic aerosol (SOA) formation and analysis is examined in GC-MS, electrospray ionization (ESI)MS, and particle chamber experiments. Glyoxal oligomers are shown to rapidly decompose to glyoxal in GC injection ports at temperatures g120 °C. Glyoxal dihydrate monomer is dehydrated at temperatures g140 °C during GC analysis but shows only oligomers (n e 7) upon ESI-MS analysis. Thus both of these analytical techniques will cause artifacts in speciation of glyoxal in SOA. In particle chamber experiments, glyoxal (at ∼0.1 Torr) condensed via particlephase reactions when relative humidity levels exceeded a threshold of ∼26%. Both the threshold humidity and particle growth rates (∼0.1 nm/min) are consistent with a recent study performed at glyoxal concentrations 4 orders of magnitude below those used here. This consistency suggests a mechanism where the surface water layer of solidphase aerosol becomes saturated with glyoxal dihydrate monomer, triggering polymerization and the establishment of an organic phase.
Introduction Because of the adverse health effects (1-5) and climactic effects (6) of fine particulate matter, the speciation and formation processes of secondary organic aerosol (SOA) have been the subject of increasing interest (7). The gas/particle (g/p) partitioning of most compounds into a liquid organic layer has been successfully described with a temperaturedependent equilibrium model (8, 9) based on compound vapor pressures and particle-phase activities. However, this model appears to significantly underpredict the condensation of volatile carbonyl-containing compounds, especially aldehydes (10-13). Glyoxal, an R-dicarbonyl implicated in SOA formation (14-17), and directly identified in the particulate phase (18-20), is a secondary pollutant that has been attributed to a dominant ring fragmentation pathway of the oxidation of aromatic precursors (21-23) and to isoprene oxidation (24, 25). Glyoxal is significantly more volatile than most other compounds found in the particulate phase, with * Corresponding author phone: (619) 260-6882; fax: (619) 2602211; e-mail:
[email protected]. 8728
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 22, 2005
a vapor pressure at 20 °C of approximately 18 Torr (26). Investigations of the condensation process of glyoxal are therefore of special interest. Given accurate inputs of vapor pressures and particlephase activity coefficients, equilibrium models will successfully describe g/p partitioning of SOA components provided that, first, the organic layer is indeed a liquid and, second, that the condensed-phase species are correctly identified. The fact that the magnitude of g/p equilibrium model underpredictions depends on the presence of carbonyl functional groups on the species in question (10, 12, 27, 28) suggests misidentification of condensed phase species due to surface chemical reactions. Several condensation mechanisms have been proposed that involve acid catalysis and carbonyl conversion, including polymerization, hydration, acetal formation (14-16, 18, 29), aldol condensation (1416, 18, 29, 30), and peroxyacetal formation (31). However, direct evidence for SOA formation by polymerization of aldehydes (17, 32) and peroxyhemiacetals (33) has been found only recently. At this point, it is clear that knowledge of the mechanisms of aldehyde condensation is inadequate and at the root of the apparent failure of g/p partitioning models. Among aldehydes, glyoxal is notably reactive by two reversible equilibria. First, glyoxal reacts with water to form a hydrate to an extent equaled only by formaldehyde (34).
The overall hydration equilibrium constant, Kh ) [Hyd]/ [Mon], where [Hyd] ) glyoxal monomer dihydrate (“hydrate”) and [Mon] ) aqueous-phase glyoxal monomer (CHOCHO), is 7.22 × 104 (35). Since hydrate formation is so heavily favored, an effective Henry’s law constant, H*, is used to describe the equilibrium between gas-phase (g) glyoxal and aqueous (aq) media
H* )
[Mon]aq + [Hyd] [Mon]g
) H(1 + Kh)
(2)
In seawater, H*glyoxal ) 3.6 × 105 M/atm, the highest of any aldehyde (36). Second, aqueous glyoxal solutions with concentrations in excess of 1 M have been shown to form significant quantities of glyoxal dimers and trimers, which NMR data suggests are linked via five-membered oxalane rings (3739).
The equilibrium constant for dimerization, [D]/[Hyd]2, where [D] ) glyoxal dimer dihydrate (“dimer”), is equal to 0.56 at 25 °C (40). The calculated vapor pressures (41) of the products of reactions 1 and 3 are, respectively, 5 and 10 orders of magnitude less than that of glyoxal. The favorability of these reactions, and the low vapor pressures of the products, may explain observations of glyoxal in cloudwater at micromolar concentrations (42, 43). It seems reasonable that glyoxal would also exist in a hydrophilic SOA particulate phase in hydrate and oligomer forms. However, 10.1021/es050446l CCC: $30.25
2005 American Chemical Society Published on Web 10/12/2005
it has been speculated that extraction and analytical processes disturb these equilibria, produce glyoxal, and thus cause misidentification of the condensed-phase species (20, 44). In this work we examine how common methods of aerosol analysis, such as GC-MS, may affect the form of glyoxal detected. Because of problems encountered in getting the majority of SOA material to elute from GC systems (45), many researchers have recently turned to alternative methods for SOA analysis, such as liquid chromatography-electrospray ionization-mass spectrometry (LC-ESI-MS) (46, 47), ESIMS (44, 46, 47), matrix-assisted laser desorption ionization (MALDI) MS (32, 44), or even capillary electrophoresis-ESIMS (48, 49). Other groups have bypassed the collection and extraction steps and analyzed SOA directly via aerosol mass spectrometers to identify components in the particulate phase (17, 50-55). In this paper, we show how glyoxal equilibria 1 and 3 are distorted by GC-MS and ESI-MS analysis and outline the potential for particulate speciation errors using other techniques. Then, to gain mechanistic information about SOA formation by glyoxal polymerization, we report measurements of the conditions under which high concentrations of glyoxal will condense onto a variety of seed particle types in a humidified particle chamber.
Experimental Section Glyoxal was generated by sampling the headspace of vials containing glyoxal trimer dihydrate (GTD) powder (92%, ICN Biomedicals, Inc.) heated to 60 °C in a sand bath, as shown in the Supporting Information. Even mild heating causes some thermal breakdown of the oligomer, resulting in a vapor pressure due to the formation of glyoxal and water vapor. Since the vapor pressure of GTD is negligible (estimated to be ∼10-11 Torr at 20 °C (41)) and glyoxal hydrate and dimer forms are also more than 5 orders of magnitude less volatile than glyoxal itself, we can be sure that headspace samples consisted chiefly of gas-phase glyoxal and water vapor in air. Glyoxal monomer dihydrate was made by dissolving GTD in deionized water overnight to allow time for trimer hydrolysis. The hydrolysis of glyoxal dimers is known to be highly pH and temperature dependent, occurring at a rate of ∼5 × 10-4 s-1 at a pH of 5 and 25 °C (40). GTD was injected into the GC-MS (HP5890/HP5970) by dissolving the solid in dimethyl sulfoxide (Acros, anhydrous 99.7%). One-microliter liquid injections and 2-mL headspace injections were performed manually. Since one of the objectives of this study was to show the effects of GC analysis conditions on glyoxal equilibria, the injection port temperature was varied between 80 and 280 °C, with the MS transfer line 30 °C hotter than the injector (unless otherwise stated). Initial oven temperatures remained in the range 80-120 °C. ESI-MS experiments were performed using a ThermoFinnigan LCQ Advantage ion trap mass spectrometer, using medical nitrogen (Air Gas Inc.) as sheath gas. Aqueous GTD solutions diluted in acetone (Fisher, certified ACS) were injected by syringe pump infusion at 2.5 µL/min. The transfer capillary temperature ranged from 150 to 200 °C. Particle chamber experiments were conducted indoors at room temperature (20 °C) in ∼300-L Tedlar or Teflon bags, using filtered dry air to fill and rinse the chamber. The experimental setup has been described earlier (56, 57). Seed particles were atomized from aqueous solutions (5 mM ammonium sulfate, J. T. Baker; 1 mM sodium chloride or 1,10-decanediol, Sigma Aldrich; 1 mM NaCl/3 mM H2SO4) and dried in a diffusion dryer. Particles were size-selected by a differential mobility analyzer (TSI, Inc. 3080 DMA) en route to the chamber. (A small amount of doubly charged particles of larger diameter will also pass through the DMA.) Micelle formation in 1,10-decanediol experiments was eliminated by using a heat gun on a small section of the particle line (56). Monodisperse particle size distributions in the chamber
FIGURE 1. GC-MS peak heights for m/z ) 30, the H2CO+ fragment of glyoxal, produced by injections of glyoxal monomer dihydrate (1 g/L aqueous solutions, b) and GTD (1 g/L solutions in dimethyl sulfoxide (DMSO), 2) as a function of GC ip temperature. For glyoxal monomer dihydrate injections, the glyoxal detection limit was ∼10 000 counts (peak height); thus, no glyoxal was observed to elute at ip temperatures below 140 °C. For GTD injections, glyoxal eluted just before the DMSO peak, which caused an elevated baseline and increased the glyoxal detection limit to ∼30 000 counts. Thus, when the ip temperature was set to 100 °C, no glyoxal was observed to elute in any run. (Glyoxal’s base peak at m/z ) 29 is not used in the analysis because of greater background interference.) were monitored at 5-min intervals by scanning mobility particle sizing (TSI, Inc. 3080/3010 condensation particle counter), operated with an aerodynamic particle diameter range of 10-700 nm. After particles equilibrated in the chamber for ∼15 min, glyoxal was added to the chamber by flowing air through a glass bulb containing GTD powder heated to ∼60 °C. (See Supporting Information for evidence supporting this method of glyoxal generation.) Glyoxal additions were quantified by mass loss and were corrected for the accompanying water vapor losses from the solid, assuming that glyoxal and water vapor were generated in proportional amounts corresponding to the molecular formula of GTD. GTD mass losses ranged from 5 to 80 mg per experiment and were assumed to occur at a constant rate, resulting in glyoxal mixing ratios in the range 26-150 ppm. Glyoxal is known to equilibrate with chamber walls; fortunately, this equilibration was found to contribute to a stable gas-phase concentration (17). Total particle volume and number densities in the chamber were corrected for dilution caused by subsequent gas additions, as previously described (56, 57).
Results & Discussion Analytical Chemistry of Glyoxal. The behavior of glyoxal during GC-MS analysis was explored in a number of experiments. When mixtures of gas-phase glyoxal and water vapor were injected, glyoxal was unretained and eluted with normal peak shapes at injection port (ip) temperatures g170 °C. However, at ip temperatures below 120 °C, extremely wide glyoxal peaks were observed that were incompletely eluted even after 15 min (see Supporting Information). Our past studies of GC-reactive compounds indicate that analytes exhibiting several-minute peak widths are involved in chemical reactions in the ip or the head of the column (57). To determine whether glyoxal dihydrate or glyoxal polymers are converted to glyoxal by GC-MS analysis, a series of injections of 1.0 g/L GTD solutions made in water (therefore containing almost entirely glyoxal monomer dihydrate) and in dimethyl sulfoxide (containing the unhydrolyzed trimer) were performed. The mass spectra produced by these injections were identical to when gas-phase glyoxal was injected. The CH2O+ (m/z ) 30) fragment ion abundance as a function of ip temperature is shown in Figure 1 for the two samples. Clearly, glyoxal is rapidly produced from its trimer VOL. 39, NO. 22, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
8729
FIGURE 2. Electrospray ionization/ion trap positive mass spectrum of 9 mM glyoxal in acetone solution containing 1% v/v water. Capillary temperature ) 150 °C, electrospray voltage ) 3530 V. Refer to Table 1 for peak identification. at ip temperatures g120 °C. (Thermal breakdown in the range 60-120 °C is evidently too slow to produce a fast-eluting glyoxal peak. See Supporting Information.) Smaller amounts of glyoxal are produced from glyoxal monomer dihydrate, rising above the detection limit only when the ip reaches 140 °C. The relative ease of glyoxal production from GTD relative to the monomer dihydrate indicates that the heat-induced decomposition of GTD may not occur by a pathway where glyoxal monomer dihydrate is an intermediate. Alternatively, this observation may simply indicate the difficulty of volatilizing glyoxal from an aqueous phase. It therefore appears that glyoxal has been reported in the particulate phase not because it is actually present there, but rather because it is easily produced from nonvolatile glyoxal oligomer species during GC analysis, as suspected (20, 44). Hydroxyl and carbonyl derivatization techniques currently in use are not expected to protect against this thermal breakdown, since the oligomers are bridged by C-O-C ether linkages. Since OH-derivatization might be expected to block dehydration of the hydrate (reaction -1), the oligomer is the form of glyoxal most likely to be present in the particulate phase in studies where glyoxal was detected by GC-MS after chemical derivatization (18-20). The behavior of glyoxal in other mass spectrometer systems was examined via a series of syringe pump injections into an ESI-MS. The result of ESI-MS analysis of a 0.3 M aqueous GTD solution diluted 100-fold in acetone is shown in Figure 2. Further fragmentation of ions in the ion trap during this analysis (“MSn” experiments) showed that the observed ions fragment by repeated loss of 18 and 58, consistent with glyoxal oligomers (see Scheme 1 of the Supporting Information). The assignments of mass spectral peaks seen in Figure 2 are listed in Table 1. It is notable that although each oligomer is a dihydrate, four water molecules can be lost during fragmentation in the mass spectrometer. This fragmentation by water loss requires rupture of the C-H bonds on the outermost carbons, a process that does not occur during polymerization and hydrolysis reactions under ambient conditions (37). Possible structures of a series of observed water-loss fragments from GTD are suggested in the Supporting Information. 8730
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 22, 2005
TABLE 1. Attribution of Mass Spectrometer Fragment Ions to Parent Glyoxal Polymersa glyoxal units (n)
molecular ion (dihydrate)
- OH
- H2O
- 2H2O
- 3H2O
- 4H2O
1 2 3 4 5 6 7
94 152 210 268 326 384 442
77 135 193 251 309 367 425
76 134 192 250 308 366 424
58 116 174 232 290 348 406
na 98 156 214 272 330 388
na 80 138 196 254 312 370
a Boldface highlights peaks observed at >3% relative abundance in mass spectrum of 0.3 M aqueous solution of GTD diluted 1:100 in acetone (Figure 2); na, not applicable.
In contrast with the theoretical prediction of glyoxal monomer being the dominant form present (see Supporting Information), Figure 2 shows clear evidence of glyoxal oligomers up to n ) 7 and no signal for glyoxal (m/z ) 56, 58) or monomer hydrate (expected at m/z ) 86, 87, or 94). Since the level of observed polymerization exceeds equilibrium levels in the analyte solution, we conclude that the observed glyoxal polymers must either be formed or preferentially ionized during the ESI process. Polymerization is reasonable, since the concentration of glyoxal hydrate would quickly increase as the solvents, acetone and water, evaporate from the droplets exiting the electrospray nozzle. Under such conditions, equilibrium 3 would be expected to shift to the right if the reaction is rapid (58). Alternatively, preferential ionization will occur either if glyoxal polymers partition to the droplet surface or if glyoxal monomers volatilize from the droplets (59, 60). Since it is reasonable that these effects would also occur at lower glyoxal concentrations, ESI-MS studies of glyoxal-containing aerosol will generally overstate the degree of oligomer formation in the condensed phase. Gao et al. (46) recently reported on ESI polymerization of C6-C8 monomers containing ketone and carboxylic acid functional groups. Some dimers (positive ion mode) and minimal amounts of unstable adduct ions (negative ion mode) were observed. Assuming that gem-diol reactions are
FIGURE 3. Experiment 040729. SMPS-measured aerosol parameters of monodisperse ammonium sulfate seed particles during additions of water vapor and 42 mg of glyoxal: O, average volume per particle (scaled up by 1 × 105, in µm3); 2, mode particle diameter (in nm); ×, number densities (scaled down by a factor of 100, in cm-3); 9, total particle volumes (scaled up by a factor of 100, in µm3/cm3). Number densities and total particle volumes are corrected for dilution caused by glyoxal and water vapor addition. Glyoxal addition times are indicated by horizontal arrows and vertical dotted lines. Water vapor was added after glyoxal. Relative humidity levels shown are interpolated between hygrometer measurements, shown as crosses. the operative mechanism of oligomer formation in each case, these observations of poor oligomer formation are consistent with the known behavior of the functional groups employed in their experiment. Ketones are much less likely to form gem-diols by hydration than aldehydes, while carboxylic acids do not form gem-diols at all. Since the reactivity of aldehydes toward gem-diol formation lies between that of ketones and of glyoxal, we can predict that ESI analysis of aldehydes in organic aerosol will result in appreciable oligomer formation by this mechanism, intermediate between that of glyoxal reported in this study and that reported for ketones and acids by Gao et al. (46). In other aerosol speciation studies, MALDI was used to introduce particulate material from R-pinene ozonolysis (44) and 1,3,5-trimethylbenzene (TMB) photooxidation (32) into a mass spectrometer. In the R-pinene study, this resulted in oligomer mass spectra that were described as similar to ESI results. (No comparison was made in the second study.) On one hand, this could indicate that oligomers formed from these materials are not affected by either analytical process, and thus look the same in both cases. On the other hand, both methods could be disturbing oligomer equilibria to similar extents. Aldehydes are known to be significant products of R-pinene/ozone oxidation (61) and are likely, as stated above, to form oligomers favored by ESI analysis. While MALDI itself is not expected to cause significant oligomer formation, the sample drying processes typically associated with MALDI could do so. Attention to the effects of MALDI sample preparation on oligomer formation equilibria is warranted. An Aerodyne aerosol mass spectrometer used in a recent study of aerosol particles formed by glyoxal condensation (17) produced excellent direct evidence for particulate-phase glyoxal dimers and trimers. This instrument uses heat to volatilize collected aerosol particles inside the mass spectrometer, under high vacuum. Although our data show that glyoxal oligomers cannot be volatilized by heat at or above atmospheric pressure without thermal breakdown, the work of Liggio et al. (17) demonstrates that it is possible to volatilize glyoxal oligomers up to n ) 3 under high vacuum. However, the possibility that larger oligomers may have been present in their study, but underwent thermal breakdown during attempted volatilization, cannot be ruled out.
In general, the phase change required for analyzing condensed phase material by mass spectroscopy, and the equilibrium disturbance that this inevitably causes, limits the utility of MS-based techniques in speciating glyoxal oligomers. Clearly, different analytical techniques suitable for trace molecular analysis directly in the condensed phase must be brought to bear on the problem of speciation of reactive hydrates and oligomers in organic aerosol. SMPS Characterization of the Conditions Necessary for Glyoxal Condensation. The condensation of glyoxal was observed as a function of measured humidity levels in a series of particle chamber experiments using monodisperse seed particles of varied composition. Data from an experiment with ammonium sulfate seed particles, where glyoxal monomer and water vapor were added simultaneously to the chamber, is shown in Figure 3. These data are not corrected for wall losses, which create strong downward trends in particle number density and total particle volume. Wall losses do not, however, affect mode particle diameter or average volume per particle for the monodisperse populations used in this work. Particle growth is minimal for the first 150 min of the experiment, as seen by a shift in the mode of the particle diameter of only 1 nm, equal to the resolution of the SMPS technique. At 160 min, particle growth rates noticeably increase. This point corresponds to ∼27% RH and a calculated glyoxal partial pressure of 0.11 Torr, based on evaporative mass loss. Defining the saturation level of glyoxal as Sgly ) pgly/p°gly, where pgly is the glyoxal partial pressure and p°gly is the vapor pressure of glyoxal at a given temperature, Sgly is estimated to be only 0.63% at 160 min. A similar growth initiation is shown on sodium chloride seeds in Figure 4. Although the exact point of initiation is difficult to pinpoint, it is certainly underway by the time humidity levels reach 28% RH and pgly ) 0.15 Torr. In this experiment, the addition of dry air to lower the relative humidity at 240 min resulted in cessation of particle growth. It can be seen in Figure 4 that the growth of particles by uptake of glyoxal and water vapor is not reversible when humidity declines as low as 14% RH. Although reactions 1 and 3 are known to be reversible with heat, drying alone does not appear to be able to cause the dehydration of glyoxal hydrate into volatile glyoxal monomer (reaction -1). Beginning at 250 min, the particle chamber equilibrated for 2 h. As water vapor desorbed from the walls, the humidity again VOL. 39, NO. 22, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
8731
FIGURE 4. Experiment 040727. Average volume per particle (scaled up by 1 × 106, in µm3), mode particle diameter (in nm), and number densities (scaled down by a factor of 10, in cm-3) of monodisperse sodium chloride seed particles during additions of 50 mg of glyoxal, water vapor (70-240 min), and dry air (240-250 min). Symbols (and dilution corrections) are identical to those in Figure 3. Glyoxal addition times are indicated by horizontal arrows and vertical dotted lines. The glyoxal saturation level at 170 min, including dilution, is 0.81% (0.15 Torr).
TABLE 2. Particle Growth Experiments Involving Glyoxal and Water Vapor expt
% RH
pgly (Torr)
seed type
growth rate (nm min-1)
accommodation coefficient, r
040729a 040727a 040730a 040722b 040728ac 040728bd 040611e 040609e 040205g 040121h 040120g
trigger: 27 trigger: 26 trigger: 25 44 37 35
