Environ. Sci. Technol. 2009, 43, 2753–2759
Measurement of Atmospheric Hydroxyacetone, Glycolaldehyde, and Formaldehyde X I A N L I A N G Z H O U , * ,†,‡ G U H U A N G , ‡ KEVIN CIVEROLO,§ AND JAMES SCHWAB| Wadsworth Center, New York State Department of Health, School of Public Health and Atmospheric Science Research Center, State University of New York at Albany, Albany, New York, and New York State Department of Environmental Conservation, Albany, New York
Received October 28, 2008. Revised manuscript received February 12, 2009. Accepted February 16, 2009.
A method has been modified and optimized for the measurements of hydroxyacetone as well as formaldehyde and glycolaldehyde, based on aqueous scrubbing using a coil sampler followed by DNPH derivatization and HPLC analysis. Derivatization equilibrium and kinetics were studied to optimize the hydroxyacetone-DNPH derivative yield. It was found that the low sensitivity of hydroxyacetone by this method is due to a relatively small equilibrium constant for the hydroxyacetone-DNPH derivatization reaction, and thus it can be improved by increasing DNPH reagent concentration. In a medium containing 500 µM DNPH and 50 mM HCl, the derivatization reaches equilibrium within 30 min. An online reagent purification procedure using a DNPH-saturated Sep-Pak C-18 cartridge effectively removed hydrazone impurities in the DNPH reagent solution, and a sample preconcentration procedure using a C-18 guard column greatly enhanced the sensitivity and lowered the detection limits. The lower detection limits of the system under optimized conditions are 30, 9, and 36 pptv for hydroxyacetone, glycolaldehyde, and formaldehyde, respectively, with a sampling/ analysis cycle time of 30 min. The method has been successfully deployed at a rural site in Pinnacle State Park in Addison, NY, for a 5 week period during the summer of 1998. The ambient concentration means (medians) were 372 (332), 301 (323), and 2040 (2030) pptv for hydroxyacetone, glycolaldehyde, and formaldehyde, respectively. A late-afternoon maximum and an early morning minimum were observed in the diurnal concentration distributions of all three carbonyl compounds. Good correlations among the three carbonyl compounds suggest that they originated from a common source, i.e., photochemical oxidation of biogenic hydrocarbons. Formaldehyde photolysis accounted for about 23% of the total radical photoproduction, whereas contributions from hydroxyacetone and glycolaldehyde photolysis were insignificant because of the much slower photolysis and lower concentrations of these compounds.
* Corresponding author e-mail:
[email protected]. † Wadsworth Center, New York State Department of Health. ‡ School of Public Health, State University of New York at Albany. § New York State Department of Environmental Conservation. | Atmospheric Science Research Center, State University of New York at Albany. 10.1021/es803025g CCC: $40.75
Published on Web 03/09/2009
2009 American Chemical Society
Introduction Carbonyl compounds in the atmosphere are mainly produced by oxidation of hydrocarbons (1). Because of their photolytic reactivity in the solar spectrum of the lower troposphere, carbonyl compounds play a critically important role in the production of atmospheric free radicals and photooxidants (2). Identification and quantification of atmospheric carbonyl compounds are necessary to the understanding of hydrocarbon oxidation mechanisms, photooxidant formation potential, and free radical dynamics. Isoprene is the principal nonmethane volatile organic compound emitted from vegetation (2). Its oxidation initiated by OH produces a suite of carbonyl compounds, including methyl vinyl ketone, methacrolein, and formaldehyde as the first-generation products, and R-oxygenated carbonyls as second- and third-generation products such as glycolaldehyde, glyoxal, methylglyoxal, and hydroxyacetone (1, 3-6). These carbonyl compounds have been positively identified and quantified in field studies (7-15). Several techniques have been developed for the measurements of these highly water-soluble R-oxygenated carbonyl compounds. Several spectroscopic techniques have been recently reported for the measurement of glyoxal, including differential optical absorption spectroscopy (DOAS) (16), incoherent broadband cavity enhanced absorption spectroscopy (IBBCEAS) (17), and laser-induced phosphorescence (LIP) (18). Proton-transfer reaction mass spectrometry (PTR-MS) has also been reported useful in identifying carbonyl compounds, including hydroxyacetone (19, 20). However, more common and versatile methods usually involve air sampling followed by derivatization with a chemical reagent, such as 2,4-dinitrophenylhydrazine (DNPH) (10, 11) and O-(2,3,4,5,6-pentafluorobenzyl) hydroxylamine (PFBHA) (12-15, 21), and separation and detection by a chromatographic technique, such as HPLC and GC/MS. These techniques are capable of measuring multiple compounds at the same time. The first systematic identification and quantitation of R-oxygenated carbonyls in the ambient air was achieved by a method that combined aqueous scrubbing of these highly water-soluble species using a glass coil sampler with HPLC separation and UV detection of their DNPH derivatives (10). This technique is capable of measuring formaldehyde, glycolaldehyde, glyoxal, methylglyoxal, glyoxylic acid, and pyruvic acid simultaneously, with a detection limit of 10-20 pptv. It has been successfully deployed in several field campaignsonbothgroundandaircraftplatforms(10,11,22,23). However, this technique exhibited much lower sensitivity for hydroxyacetone, by 1-2 orders of magnitude under the experimental conditions deployed (10). The lower sensitivity is due in part to the lower reactivity of hydroxyacetone toward the DNPH reagent as compared to other carbonyl compounds such as formaldehyde, and in part due to its peak broadening and tailing on chromatograms. PFBHA-derivatization and GC/MS analysis, coupled with sample collection using an impinger, a mist chamber or a denuder, has been developed to identify and quantify these multioxygenated carbonyls compounds both in chamber studies (21) and in the field (12-15). Although this technique is powerful for identifying and quantifying R-oxygenated carbonyls, it involves batch sampling followed by a tedious procedure of derivatization and solvent extraction, making it difficult to automate and thus unsuitable for routine field measurement. VOL. 43, NO. 8, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2753
In this paper, we describe a modified DNPH derivatization-HPLC method (10), optimized for the measurement of atmospheric hydroxyacetone. We also present an extensive field data set of hydroxyacetone, formaldehyde, and glycolaldehyde, obtained by this technique at a rural site in New York State, with significant discussion on their atmospheric chemistry.
Experimental Section Reagents. 2,4-Dinitrophenylhydrazine (DNPH) with 99+% purity was purchased from Radian. Reagent-grade chemicals, including HCl (36 wt % in water), hydroxyacetone (90 wt % in water), formaldehyde (37.2 wt % in water), glycolaldehyde (dimer), glyoxal (40 wt % in water), and methylglyoxal (40 wt % in water) were from Aldrich. HPLC-grade acetonitrile (AcN) and methanol (MeOH) were from Burdick & Jackson or J.T. Baker. The SEP-PAK C-18 cartridge for DNPH reagent purification was from Waters Associates. The C-18 reversed phase packing material (30-40 µm) for the preconcentration cartridge was from Upchurch Scientific. Water was purified with a Millipore Milli-Q water system, with resistivity g18 MΩ-cm. Sample Collection. A 28-turn coil sampler (22) was used to collect hydroxyacetone, glycolaldehyde, and formaldehyde with a scrubbing solution containing 500 µM DNPH and 50 mM HCl. The ambient air is pulled through the air inlet of the coil sampler by a vacuum pump (model 7055-60, Vacuubrand), whereas the scrubbing solution is delivered by a multichannel peristaltic pump (IPC, Ismatec) at a flow rate of 0.24 mL min-1. The flow rate for gas sampling is 2.0 standard liter per min (SLM) controlled by a mass flow controller (Cole-Parmer 5 SLM). During sample scrubbing in the coil sampler, interferences from the atmospheric O3 and SO2 are unimportant because of the slow reaction kinetics and short gas-liquid contact time (10). HPLC Analysis. The hydrazone derivatives were separated on a C-18 reverse-phase HPLC column and detected by a UV-visible absorbance detector. The HPLC system consisted of a Hitachi model L-7100 gradient pump, a Valco electrically actuated 10-port auto injection valve, a C-18 reversed phase column (Microsorb-MV, Rainin) with a guard column (Upchurch), and a Dynamax UV D II wavelength-programmable UV-visible absorbance detector. The hydrazone derivatives of formaldehyde, glycolaldehyde, and hydroxyacetone were detected at a wavelength of 375 nm, and those of glyoxal and methylglyoxal at 410 nm. System Automation. An automated measurement system is illustrated in Figure S1 in the Supporting Information. A coil sampler was installed at a desirable sampling location, and ambient air was pulled through a pinhole of -0.1 cm in diameter on the wall of a short pierce of 1/8′′-OD Teflon tubing connecting the scrubbing solution line and the sampling coil. The “surface-less” inlet of the sampler minimizes the potential loss of these water-soluble species on inlet wall surface. The carbonyl compounds were continuously scrubbed from the ambient air by the DNPH scrubbing solution. A multichannel peristaltic pump continuously delivered the DNPH scrubbing solution through the sampling coil and back into a derivatization coil of 8 mL volume to allow 30 min derivatization reaction. The derivatized solution was then alternately loaded onto one of the two preconcentration C-18 columns, and injected into the HPLC system for analysis every 30 min. A PC-based HPLC data system (PeakSimple, SRI) controlled sample injection and acquired chromatographic data. Both the derivatization coil and the HPLC column were kept in a water bath controlled at 25 °C to compensate for the diurnal temperature variation during field measurements. 2754
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 8, 2009
Results and Discussion Collection Efficiency. Carbonyl collection efficiency (β) was measured by connecting 2 identical 28-turn coil samplers in series β ) (1 - C2/C1)
(E1)
where C1 and C2 were the carbonyl concentrations collected by the first and second samplers, respectively. Under experimental conditions, i.e., at an air flow rate of 2.0 L min-1, a liquid flow rate of 0.24 mL min-1, using a scrubbing solution containing 500 µM DNPH and 50 mM HCl, and at ∼22 °C and 1000 mbar, the collection efficiency of this sampler was determined to be quantitative (100 ( 2%) for hydroxyacetone and glycolaldehyde, and (73 ( 2%) for formaldehyde, which is similar to ∼70% previously reported by a similar 28-turn coil sampler under the similar sampling conditions (22). We also observed an increase in formaldehyde collection efficiency with scrubbing solution acidity, from 66% in 5 mM HCl, to 73% in 50 mM HCl and 77% in 100 mM HCl, as a result of acid-catalyzed hydration of formaldehyde to form more water soluble gem-diol. The ambient concentration of a carbonyl compound p (in part per trillion by volume, pptv) can be obtained from its concentration C (nM) in the scrubbing solution as follows p ) CRTFl /(βFg)
(E2)
where Fl is scrubbing solution flow rate (in mL min-1), Fg the sampling gas flow rate (in SLM), R the gas constant, and T the absolute temperature. Derivatization Reaction Equilibrium. To examine if carbonyl-DNPH derivatization yield is improved by increasing the DNPH concentration, we first studied the DNPHcarbonyl (>CdO) reaction equilibrium in formation of a hydrazone derivative (DNPHzone)
The equilibrium constant is expressed as K)
[DNPHzone] [DNPH][Carbonyl]
(E3)
which can be rearranged into the following equation when the DNPH concentration is in a large excess compared to the carbonyl concentration: 1 1 1 + ) [DNPHzone] [Carbonyl]t K[Carbonyl]t[DNPH] (E4) where [Carbonyl]t denotes the total concentration of the carbonyl compound () [Carbonyl] + [DNPHzone]). Linear relationships existed between the reciprocal hydrazone responses and the reciprocal DNPH concentration for both hydroxyacetone and formaldehyde (Figure 1). According to eq E4, the value of K for hydroxyacetone or formaldehyde can be estimated from the ratio of the Y-intercept to the slope. At a room temperature of 22 ( 1 °C, the K value is 1.2 × 103 M-1 for hydroxyacetone, which is about the 40 times smaller than that of 4.3 × 104 M-1 for formaldehyde. The low equilibrium constant for hydroxyacetone explains the low sensitivity of DNPH/HPLC method for hydroxyacetone reported by earlier authors (10). In a 100 µM DNPH reaction solution, only ∼11% of hydroxyacetone is converted into hydroxyacetone-hydrazone derivative, compared to ∼81% for formaldehyde. Measurement sensi-
FIGURE 1. Reciprocal hydrazone response (Response-1) as a function of reciprocal DNPH concentration ([DNPH]-1), hydroxyacetone in open cycles and formaldehyde in crosses. Each line represents the least-squares fit for the data (N ) 8). Derivatization conditions: reaction for 4 h, 5 mM HCl and 20% ACN in water at room temperature.
FIGURE 2. Detailed derivatization kinetics of hydroxyacetone (circles), glycolaldehyde (squares), and formaldehyde (crosses) at room temperature. In the inserted figure, the Y axis is the difference between derivative responses at equilibrium and at time t, in the log scale. The figure combines results of four separate experiments with different reaction time intervals. tivity for hydroxyacetone can thus be significantly improved by using a sufficiently high concentration of DNPH reagent, e.g., the hydrazone yield for hydroxyacetone increases by a factor of ∼3.5 to 38% in a 500 µM DNPH derivatization solution. Kinetics of the Derivatization Reaction. To optimize derivatization conditions, we investigated detailed derivatization kinetics over a range of acidities. The DNPH-carbonyl reactions are acid-catalyzed. The rate of hydroxyacetoneDNPH derivatization increased with increasing reaction acidity; derivative yield reached ∼70% of the equilibrium value in a 10 mM HCl medium and g99% of the equilibrium value in a 50 mM HCl medium in 17 min. The hydrazone response as a function of reaction time for hydroxyacetone, glycolaldehyde and formaldehyde is illustrated in Figure 2. The pseudo-first-order rate constants were estimated by the logarithm curve fitting of the relative peak response and the reaction time (the inserted figure), to be 0. 43 min-1 for hydroxyacetone, 0.15 min-1 for formaldehyde, and 0.29 min-1 for glycolaldehyde, respectively, in a 500 µM DNPH solution containing 50 mM HCl at room temperature. The pseudofirst-order rate constants were also determined to be 0.099
min-1 for glyoxal and 0.087 min-1 for methylglyoxal. Because of the acid catalysis, the measured DNPH derivatization rates in this study are higher than those reported previously at lower acidity, i.e., 0.049, 0.039, 0.028, 0.033, and 0.01 for hydroxyacetone, formaldehyde, glycolaldehyde, glyoxal, and methylglyoxal, respectively, in a reaction medium containing 4 mM acid (10). Within a 30 min reaction time at room temperature, the derivatization yields reach g99% of the equilibrium maximum values for hydroxyacetone, formaldehyde and glycolaldehyde, 95% for glyoxal, and 93% for methylglyoxal. The fact that hydroxyacetone has the highest reaction rate constant among this group of carbonyl compounds suggests that the low sensitivity of the DNPH derivatization/HPLC method for hydroxyacetone is due not to the kinetic factor but rather to its small equilibrium constant. Acetonitrile (AcN) was used for the preparation of DNPH reagent solution, and its effect on the reaction rate and derivative yield was also studied. The presence of AcN in the reaction solution (10-20% by volume) does not have any significant effect on the reaction rate or the derivative yield; it does minimize the adsorption of hydrazone derivatives on the wall surface of a derivatization coil or a sample container. In field measurements described later, the derivatization reactions occurred in a 500 µM DNPH solution containing 50 mM HCl and 10% AcN at 25 °C for 30 min. Preconcentration of Analytes. To further improve the detection sensitivity, we took a preconcentration step before sample injection into the HPLC system. Two guard columns (Upchurch) packed with C-18 material were used to replace the sample loops on a 10-port autoinjection valve of the HPLC system. Because the C-18 preconcentration column preferably retains more hydrophobic hydrazone derivatives over DNPH, a whole sample of the derivatized carbonyls can be preconcentrated online. The efficiency of preconcentration and the sample breakthrough threshold on the column were examined at various levels of AcN. Figure 3a illustrates the effect of AcN content on the preconcentration of the hydroxyacetone derivative: the sample breakthrough threshold decreases with increasing AcN content. At 15-20% AcN in the sample, the detected response of the hydroxyacetone derivative was linear only when the sampling volume was e3.5 mL. At 10% AcN content, no breakthrough occurred for hydroxyacetone up to a sampling volume of 7.5 mL. Sample breakthrough thresholds on the C-18 preconcentration column were also examined for all five carbonyl compounds at 10% AcN (Figure 3b). The detected responses of all carbonyl derivatives except glycolaldehyde were linear up to 7 mL of sampling volume. The breakthrough thresholds of the derivatives increase in the order of glycolaldehyde < hydroxyacetone < formaldehyde < glyoxal < methylglyoxal, reflecting the increasing hydrophobicity of the hydrazone derivatives. In the field measurement application, a sample of e5 mL containing 10% AcN was collected, so that the derivatives of all five carbonyl compounds would be quantitatively retained on the C-18 preconcentration column. HPLC Separation and Analysis. The gradient elution consisting of water, AcN and MeOH (gradient profile see Table S1 in the Supporting Information) offered good HPLC separation for all five carbonyl compounds of interest on a Rainin Microsorb-MV 150 × 4.6 mm, 5 µm C-18 column (for the chromatogram, see Figure S2 in the Supporting Information). An unidentified compound was eluted before glycolaldehyde and its peak size increased with aging of the DNPH reagent, and caused interference. Therefore, it was recommended that the DNPH reagent be freshly prepared for field measurement and used within 2 days if glycolaldehyde is to be quantified. Calibration using aqueous standard solutions demonstrated good linearity for the detection of five carbonyl VOL. 43, NO. 8, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2755
FIGURE 3. (a) Preconcentration of hydroxyacetone derivative in sample solution containing 10% (crosses), 15% (squares), and 20% acetonitrile (circles); (b) preconcentration for five carbonyl derivatives in the sample solution containing 10% of ACN, 500 µM DNPH, ∼0.3 µM hydroxyacetone (squares), formaldehyde (circles), and glycolaldehyde (diamonds), and ∼0.15 µM glyoxal (crosses) and methylglyoxal (triangles). Panel (a) combines results of six separate experiments and panel (b) two separate experiments. compounds, in the concentration ranges of 0-2 µM for formaldehyde and 0-350 nM for hydroxyacetone, glycolaldehyde, glyoxal, and methylglyoxal, with the correlation coefficients r2 g 0.998. During field campaigns, the instrument was calibrated by a set of standard solutions for each batch of DNPH scrubbing solution. The working standard solutions were prepared by spiking the stock carbonyl aqueous standards into known volume of DNPH scrubbing solution for derivatization. The working standards were then delivered by the peristaltic pump into the derivatization coil and the HPLC system, with the coil sampler bypassed. One prederivatized carbonyl standard was injected every day as a quality control standard to monitor the system consistency. Detection Limit and Reagent Blank. Nine subsets of DNPH solutions containing low levels of carbonyl (i.e., 10 nM glycolaldehyde, glyoxal, and methylglyoxal; 30 nM hydroxyacetone; and 100 nM formaldehyde) were analyzed to determine the detection limits, defined as three times the standard deviation of measurement signals. The detection limits were determined to be 1 nM for glyoxal and methylglyoxal, 3 nM for glycolaldehyde, 10 nM for hydroxyacetone, and 12 nM for formaldehyde. The detection limit for formaldehyde is higher than those for glycolaldehyde and hydroxyacetone because of the higher reagent blank of formaldehyde. Under typical sampling conditions, i.e., a scrubbing solution with a flow rate of 0.24 mL min-1 and a sampling gas flow rate of 2.00 L min-1, the corresponding gas phase detection limits are equivalent to 9 pptv for glycolaldehyde, 30 pptv for hydroxyacetone, and 36 pptv for formaldehyde. The automated system is not suitable for the measurement of glyoxal and methylglyoxal, because their hydrazone derivatives tend to adsorb to the derivatization coil wall surface (10). The analytical blank for all carbonyl compounds, except for formaldehyde, remained low, i.e., below or slightly over the detection limit, throughout the method development 2756
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 8, 2009
and in field studies. However, the reagent blank for formaldehyde has been found to increase with the age of reagent solution during some periods of urban measurements, by a factor of 5 to as high as ∼100 nM in a week, probably because of contamination from high ambient formaldehyde concentrations. To minimize the reagent blank, particularly for formaldehyde, two DNPH-saturated Sep-Pak C-18 cartridges (Waters) were connected in series at the outlet of the DNPH reagent solution reservoir. The DNPH-saturated cartridges selectively removed ∼90% of hydrophobic hydrazone impurities while allowing the DNPH to pass through. The effectiveness of reagent purification by the DNPH-saturated Sep-Pak C-18 cartridges lasted for more than 2 weeks. Field Measurements. The automated system has been used to measure water soluble carbonyl compounds at a rural site in Pinnacle State Park in Addison, NY, from June 10 to July 14, 1998. Site descriptions have been given elsewhere (24). Statistical summary of the measurement results is shown in Table S2 of the Supporting Information. The mean formaldehyde concentration was 2038 pptv, which is at the lower end of the mean values of 2000-4000 pptv in other rural regions where isoprene oxidation was found to be the major formaldehyde source (11, 25, 26), but higher than ∼1300 pptv at a more remote mountain site (27, 28). It should be pointed out that reliable glycolaldehyde data were only collected during the first 2 weeks. During the remaining period of the campaign, glycolaldehyde data are not used because of significant interference from the broadened DNPH peak from the aged reagent solution. The mean glycolaldehyde value of 301 pptv was similar to the means of 210 and 260 pptv at a site in rural Georgia during 1991 and 1992 summer campaigns (11). It was lower than means of 690 pptv measured above a ponderosa pine plantation in California (13), but significantly higher than the values of 53 pptv at a forested site (14) and 22 pptv at a semirural site (15) in Japan, and ∼100 pptv at sites downwind of Berlin, Germany (12). Our mean hydroxyacetone concentration is 372 pptv, similar to 420 pptv measured above a ponderosa pine plantation in California (13), but was lower than the daytime mean value of 900 pptv at a NARSTO site at Brookhaven National Laboratory, Upton, NY (29), and higher than the values from 26 to 150 pptv measured in Japan (14, 15) and in Germany (12). Two days of continuous measurements of hydroxyacetone, glycolaldehyde, and formaldehyde at Pinnacle site during June 1998 are illustrated in Figure 4, along with oxides of nitrogen (NOx), total reactive nitrogen (NOy) and solar radiation. Strong diurnal variations occurred in the concentrations of the three carbonyl compounds, with maxima in the daytime, i.e., in the rain-free, late afternoon of June 24, and in the early afternoon of June 23 just before the onset of rain. The NOy also demonstrated a significant diurnal trend, with a maximum in the midmorning, due to the vertical mixing that brought down the entrained pollutants from aloft. The significant offset of the concentration peaks suggests that the precursors of these carbonyl compounds were biogenic VOCs produced locally, rather than the anthropogenic VOC associated with transported NOy. NOx is a local pollution indicator at this rural site. The absence of obvious correlation between NOx and carbonyl compounds suggests that local anthropogenic pollution was not a significant source for the three carbonyl compounds. On June 23, carbonyl concentrations increased during the first half of the day when plenty of sunlight was available, from 200, 250, and 1500 pptv in the early morning to about 660, 450, and 3000 pptv in the early afternoon, for hydroxyacetone, glycolaldehyde and formaldehyde, respectively. They then dropped sharply in the midafternoon during a thunderstorm, to 150, 210, and 1000 pptv. The rapid decreases were due to the wet scavenging of these water-soluble compounds by
FIGURE 4. Field measurements from June 23-25, 1998 at Pinnacle State Park, NY: (a) hydroxyacetone (open triangles), glycolaldehyde (open squares), and formaldehyde (open circles); (b) NOy (crosses), NOx (solid line), temperature (open diamonds), and Eppley UV (dashed line). The short horizontal bar in the upper part of (b) denotes a rain period. the rain, the enhanced dry deposition to the wet ground surface, and also the change of air masses during the thunderstorm event. More pronounced diurnal variations in carbonyl concentrations were apparent on June 24, a mostly sunny day; steady increases from 50, 150, and 500 pptv in the morning to 700, 500, and 4700 pptv in the late afternoon, were observed for hydroxyacetone, glycolaldehyde, and formaldehyde, respectively. The NOx level, however, remained relatively low throughout the day,
