Environ. Sci. Technol. 1990, 2 4 , 898-902
Collection of Formaldehyde from Clean Air for Carbon Isotopic Analysis Brian J. Johnson* and George A. Dawson
Institute of Atmospheric Physics, University of Arizona, Tucson, Arizona 8572 1
A method has been developed for collecting milligram quantities of formaldehyde from clean air (sub-ppb levels) onto dry, sulfito-treated glass substrates. Spectrometric and chromatographic characterizations of the surface were performed to establish conditions for optimal collection. Efficiency of collection for formaldehyde is dependent on the fractional oxidation of the filter surface, as well as on ambient humidity; the rate of oxidation is cation dependent. Under favorable conditions, collection efficiencies of over 90% (based on breakthrough in a multiple-filter stack) are routinely obtained. Selectivity of collection is high, and the possibility of artifact formation is minimal. Preliminary 13Cdata for two remote sampling sites have been obtained. Introduction Formaldehyde is a ubiquitous component of the background atmosphere, present at levels ranging from -0.2 to 0.8 parts per billion by volume (ppb) (1-3). Photochemical oxidation of methane has been considered the primary source ( 4 , 5 )and is often the only source included in photochemical models of the global atmosphere. However, there may be a substantial contribution from the oxidation of non-methane hydrocarbons (NMHC) from the continents (6, 7). Concentration data seem to support this but due to the nonlinearity of the photochemidea (8,9), ical processes involved, such data are only qualitative in nature. Quantitative determination of the relative fluxes of carbon through formaldehyde from methane and NMHC sources are, however, in principle obtainable by isotopic analysis. In clean air where green plant emissions are the dominant source of NMHC, the 13C composition of this source should be characteristic of the plants, and very different from the 13Cof methane emitted to the air (10,11). A determination of the 13Ccontent of atmospheric formaldehyde should therefore quantify the relative magnitude of the two sources once adequate account has been taken of possible fractionation effects (12, 13). To make such a determination, it is necessary to collect large quantities of formaldehyde carbon (- 100 pg of C) present in the atmosphere at sub-ppb levels and convert it to pure C02. To minimize isotopic fractionation effects, collection should be as near to quantitative as possible. It is clear that extremely large volumes of air must be processed in the chosen collection scheme, and for reasonable time resolution this means high flow rates. Cryogenic trapping was investigated and discarded; input air could not be dried without affecting formaldehyde, while, without drying, resulting solutions were too dilute. Cryogenic collection was also insufficientlyspecific. Instead, chemical adsorbents were considered. An ideal adsorbent would be stable, inorganic (to avoid carbon contamination), selective, and unlikely to produce artifact formaldehyde from the other organics in the air. This latter restriction and the requirement that flow rates be high strongly favor a dry surface collection me-
* Present address: Department of Chemistry, University of Nevada, Las Vegas, NV 89154. 898
Environ. Sci. Technol., Vol. 24, No. 6, 1990
thod, since wet collection can be accompanied by artifact formation (14). One of the few classes of compounds that has the potential to meet most of these stringent requirements is the family of sulfito salts, i.e., SO?-, HS03-, and S2052-; their major failing is stability against oxidation. Aqueous reactions between S(1V) compounds and carbonyls have long been known, although reliable kinetic data for the reaction with formaldehyde have appeared in the literature only fairly recently (15). The reaction product, hydroxymethanesulfonate, is formed in the presence of either sulfite or bisulfite, reacting faster with sulfite but being destabilized in the presence of base. The net result is that the maximum rate of reaction is at pH 7, although the complex has maximum stability at pH 4. No significant reaction between formaldehyde and the relatively weak nucleophile disulfite is expected. Dry sulfito surfaces have previously been used in active (16) and passive (17) sampling modes for collection of formaldehyde in indoor environments (ppm to sub-ppm levels). However, the methodology for collection by sulfito surfaces of bulk quantities of formaldehyde at background atmospheric levels has not been previously reported. Chemical characterization of the sulfito surface as to species present, evolution during active sampling, chemical selectivity, and the effect of chemical factors on collection efficiency have similarly been neglected. This paper describes the production and characterization of sulfito-impregnated filters that were finally used for bulk formaldehyde collection.
Experimental Section Materials. Solutions of the alkali-metal bisulfites with S02(aq)were made by dissolving the reagent-grade salts, LiOH, NaHSO,, KOH, Rb2C03,or CsOH, in HzO and thoroughly saturating with SO2gas (99.98%, Matheson) at room temperature. The SO2 was scrubbed through a 0.7-m column of activated charcoal to remove gaseous CS2 and OCS, which apparently were present in sufficient quantities to interfere with subsequent reactions (see below). Glass-fiber filters (1200 CO, Pallflex Corp., binderless) as received from the manufacturer contained a significant amount of organic residue. The carbon blank was reduced by digesting the filters in an acidic solution of 3.5% K2S20B(Aldrich) before washing and drying. Quartz filters (2500 QAOT, Pallflex Corp.), to be used as prefilters, were untreated. The hydrocarbons isoprene, a-pinene, @-pinene(Fluka), and ethylene (10.8 ppm in nitrogen, Scott Specialty Gases) were used without further purification. Preparation of Sulfito Filters. Sulfito-treated filters were prepared by saturating two dry glass-fiber filters (pretreated as above) with SO,-saturated 0.5 M bisulfite solution (usually sodium or potassium), squeezing out excess solution between two Teflon sheets, and drying the filters in a microwave oven under a flow of scrubbed SO2. This procedure has been shown to minimize oxidation; S(1V) to S(V1) ratios of better than 5:l were typically obtained, as established by ion chromatographic analysis (see below). To monitor S(IV)concentration and oxidation on the filters, a standard area was punched out and added to deionized water and the solution analyzed immediately.
0013-936X/90/0924-0898$02.50/0
0 1990 American Chemical Society
/ /
ertlFi-/&/
holder -Untreated
quartz
pref ilter
/,---S20: Treated glass filters (2per unit, 6 total)
Slainless steel hival sampler
-Motor
Figure 1. High-volume collector assembly.
Without special deoxygenation of eluent and solvent but with preconditioning, oxidation was a few percent and constant. Analytical Methods. The ion chromatograph (Dionex QIC) used an HPIC-AS4 (strong anion) column and conductivity detection with continuous background suppression. The flow of eluent (2.25 X M Na2C03and 2.80 X M NaHCOJ was 1.2 mL/min; sample loop volume was 250 pL. The pH of solutions was measured by using an Orion Ionalyzer Model 407A pH meter with Orion 91-03 micro (6 mm diameter) glass electrode (internal reference). The meter was first standardized with buffers of pH 4.0 and 7.0. Determination of formaldehyde on filters and in aqueous extracts was performed by the chromotropic acid (CTA) method (18, 19). In the presence of H2S04,two CTA molecules form a complex with a molecule of formaldehyde that absorbs strongly at 580 nm; the method is quite selective and is not affected by large amounts of HS03-/ S032-. Confirmation of the method was achieved by an HPLC method (see below). The carbonyls formaldehyde, acetaldehyde, and acetone were determined by reacting them with excess (2,4-dinitropheny1)hydrazine (2,4-DNPH) and analyzing by HPLC (I). Conditions were as follows: column, C8; eluent, 60% methanol, 40% water; flow rate, 1.5 mL/min; detection, UV-visible (240 nm). Strips of sulfito-treated filter were analyzed directly by transmission IR spectroscopy (Perkin-Elmer 983) and X-ray diffraction (Siemens D-500, Cu K a source, 15" I 2e I 550). Stable carbon isotope ratios were determined at the University of Arizona's Laboratory of Isotope Geochemistry on a modified VG 602 mass spectrometer using the McKinney-Nier arrangement (20). Results are reported as 613C vs the PDB standard (21). Surface Characterization. 1. Sampling. Sampling trials were conducted with the experimental setup shown in Figure 1. The arrangement of multiple treated filters allowed for greater collection efficiency and its evaluation via formaldehyde breakthrough. Both sodium- and potassium-treated filters were used in characterization experiments. Untreated quartz prefilters were used to remove particles from the sampled air stream because of their high efficiency and inert character. Although slightly basic, filters examined in our laboratory have not shown a tendency to oxidize formaldehyde to formate to any
appreciable extent at atmospheric concentrations; also, measured formaldehyde concentrations on the quartz filters after sampling runs were typically less than 5 % of the total formaldehyde collected. 2. Surface Oxidation during Active Sampling. In the initial sampling experiments, sodium-treated filters were used exclusively. These experiments established that the efficiency of collection of formaldehyde was high, but erratic. An effect of high relative humidities was noted; high humidity appeared to accelerate oxidation of S(1V) to sulfate, resulting in decreased collection efficiency. To quantitatively examine this oxidation, a time profile of the state of the surface during collection was run. Sampling, following the procedure given above, was performed at Steward Observatory on Mt. Lemmon, 50 km north of Tucson at 2800-m elevatioa. Weather conditions were damp and cool (- 15 "C); some light rain fell between (but not during) sampling runs. At selected intervals, a 3.2-cm2 section was punched through the three treated filters with a die; these samples were stored in order in Petri dishes for later analysis. The hole in the remaining filters was covered with masking tape and the sampling continued. The filter sections were analyzed by cutting out a small strip of filter, placing it in water, and analyzing by ion chromatography; the pH of a similar but more concentrated solution was also determined. The ratio of the heights of the S(1V) and SO:- peaks was used to monitor the oxidation process, while solution pH was used primarily to determine the bisulfite to sulfite ratio. 3. Cation Effects. The sodium-treated filters were subject to relatively rapid surface changes (see below), so other cations were investigated to see if they imparted greater stability against atmospheric oxidation. A rapid screening process was necessary so that changing ambient conditions could be neglected. An aerosol sprayer was calibrated as to volume delivered and was used to spray SO2-saturated0.5 M bisulfite solution (containingLi+,Na+, K+, Rb', or Cs+ as the cation) onto a quartz filter mounted on a running high-volume sampler. About 15 mL of solution was sprayed onto each filter; the sampler was run for 4 min to allow complete drying. The pH and bisulfite to sulfate ratio of those filters in water were then determined by methods described above. Selectivity of the Collection Procedure. 1. Versus Non-Formaldehyde Carbonyls. While subsequent workup of exposed samples (which involves a wet chemical oxidation to C02;see below) is highly selective in isolating formaldehyde carbon in preference to other types of carbon, it was deemed desirable to characterize the selectivity of collection against compounds of similar chemical properties and atmospheric concentration. Acetaldehyde and acetone were selected as model compounds due to their structural similarity to formaldehyde and their known presence in the atmosphere (22). Exposed filters were extracted with water to isolate and concentrate the sample, and carbonyls were then distilled from the solution at boiling point into an iced receiver containing dilute 2,4DNPH solution. This solution was finally analyzed by HPLC. Prior standard additions showed that recovery of the species via this procedure was 90% for acetone, 50% for acetaldehyde, and virtually 0% for formaldehyde. 2. Formic and Acetic Acids. It was particularly important to demonstrate that formic acid was not collected, because it is one of the few known interferents in the subsequent workup procedure. The analytical methodology used did not discriminate between formic acid and other small carboxylic acids (especially acetic acid), so the concentrations obtained actually represent the sum of Envlron. Sci. Technol., Vol. 24,
No. 6, 1990 899
these species present. To analyze the sampled filters for these acids, it was necessary to separate them from the concentrated salt matrix. An aqueous extract of a sampled filter was evaporated to dryness and then extracted with portions of methanol. The methanol extract was then evaporated to dryness under a flow of He and mild heat. The orange-yellow residue was reconstituted in 2 mL of H 2 0 (some of the solid was insoluble), and this solution was analyzed by ion chromatography (IC). 3. Other Forms of Carbon. In an attempt to determine if other forms of carbon had been collected, an aqueous extract of a filter was prepared. Both IR and UV-visible spectra were obtained for the solution. A temperature-programmed GC run of 1pL of the solution was performed on a PB-1 capillary column with an FID detector. (Temperature program was 35 OC for 1 2 min, then 5 OC/min until 220 OC, and then 6 min until the end of the run.) 4. Artifact Formation by Reactive Hydrocarbons. There remains the possibility that non-formaldehyde organics could adsorb on the front filter and be oxidized by free radicals or other species to create artifact formaldehyde. To check for such an occurrence, the most likely organics need to be identified. From the standpoint of reactivity and atmospheric abundance, the non-methane hydrocarbons ethylene, isoprene, a-pinene, and 6-pinene seemed to be good representative choices. For this check, a device was constructed that fitted over the sampling filters, dividing the surface into four quadrants of presumably equal flow. Air for each quadrant entered via a turbulent constriction, at which point the hydrocarbon of interest could be added to the flow. Quadrants 2 and 4 sampled atmospheric air; quadrants 1 and 3 sampled the same air plus an appropriate amount of hydrocarbon. The compounds were added at a controlled flow rate (Tylan Model RO-28 electronic flow controller). Suitable concentrations of a-pinene and ppinene were obtained by bubbling N2 (UHP grade) into the liquids at 0 "C; for isoprene, cooling to -78 "C was necessary to sufficiently reduce the vapor pressure. Ethylene was available premixed in Nz(10.8 ppm, Scott Specialty Gases). The vapor pressure over the first three substances was calculated at the appropriate temperature (23). Based on this calculation, the flow rate for the appropriate hydrocarbon was adjusted so as to be present in the sampled flow stream at -10 ppb; for ethylene, concentration was limited to 5 ppb due to limitations in maximum flow.
Results Surface Characterization. 1. Composition of Unsampled Surfaces. Since the infrared absorption spectra for HS0,- and S20j2-are totally different (i.e., C,, vs C, symmetry), infrared spectroscopy held promise as a method to establish the surface composition. The spectrum was, however, obscured below 1600 cm-' by Si-0 bonds in the substrate. Both sodium and potassium filters showed a strong absorption feature at 3450 cm-', assigned to water adsorbed on the surface. X-ray diffraction spectra showed the predominant S(1V) species present was disulfite; agreement between observed and library spacings for Na2S205and K2S20j(20 peaks) was excellent. The locations of superfluous peaks in the filter spectra were consistent with the presence of small amounts of sulfate, but there was no evidence of bulk bisulfite. Analysis of aqueous extracts of the filters by IC also showed that -10% of the S(1V) was oxidized to sulfate, with the potassium filters exhibiting less oxidation than the sodium filters. 900
Environ. Sci. Technol., Vol. 24, No. 6, 1990
5.01
0 x
Second filter Third filler
+
-0.4
lbl
A 44
Hours sampled
Flgure 2. Changes in a sampled Na2S205surface as a function of time: (a) pH, (b) S(1V) to S(V1) ratio.
2. Surface Changes Due to Active Sampling. Figure 2a shows the pH changes that occurred during a typical sampling experiment; Figure 2b shows the corresponding S(1V) to sulfate ratio of aqueous extracts of the filters. A mechanism for surface oxidation that is consistent with its formaldehyde collection properties and with known chemistry was developed to explain these results, The large jump in pH during the first few hours of sampling indicates a loss of SO2 from the surface, presumably via a two-step process: SzOj2- + HzO
2HSO,-
-
F!
2HSO3-
S032-+ SO2 + H 2 0
(1)
(2)
Obviously, these reactions would be favored by our sampling method since the reactants have large surface area, are exposed to large volumes of air, and product SO2 is removed continually. After the initial rise, pH gradually fell again. This change is consistent with oxidation of sulfite, as formed in eq 2, to sulfate (i.e., a weak base is transformed to a pH-neutral species):
-
S032-+ '/202 S042-
(3)
The pH declines more slowly than it rose, because it is now close to a pH buffer region at pH pK2 = 7.2. In some experiments, the S(1V) was virtually completely oxidized to sulfate, with the pH reaching a minimum value of 4.5. The formaldehyde collection efficiency was adversely impacted well before the surface was completely oxidized, however. 3. Effect of Cation on Oxidation Rate. With the goal of slowing down the processes mentioned in the previous section, and thereby maintaining the high collection efficiency of the surface for a longer period of time, the study of cation effects was undertaken. In Figure 3a, the pH values are plotted against alkali-metal rank. The trend toward lower pH values as one moves down the periodic table is unmistakable, and the magnitude of the effect is impressive. This trend has been mentioned previously in the literature (24) but not quantified. It is here assumed that the predominant effect is loss of SO2, in which case the trend reflects the strengths of the bases (Le., Cs+ > Rb+ > K+ > Na+ > Li+) and their subsequent greater affinity
Table 11. Effect of Reactive Hydrocarbons on Formaldehyde Collection
(a)
It
sample
I
41((liiB1 Li'
Na'
K'
Rb'
concn, wg/mL
Dates Sampled: 11/09-10/87 a-pinene 1 0.51 a-pinene 2 0.37 f 0.06 @-pinene1 0.42 0-pinene 2 0.45 f 0.02 0.40 f 0.21 no HC 1 no HC 2 0.43 f 0.06
CS'
Dates Sampled: 11/10-11/87 isoprene 1 0.23 isoprene 2 0.31 ethylene 1 0.29 f 0.03 ethylene 0.26 f 0.02 no HC 1 0.25 f 0.05 no HC 2 0.26 f 0.02
'11
I
I
ti[
Li' Na' K' Rb' Cs' Figure 3. (a) pH and (b) S(IV) to S(V1) ratio of rapidly dried filter surfaces, ranked by cation.
Table I. Formaldehyde Collection in a Sampled K2S2O6 Filter Stack collected
collected
filter noso
HCHO,pg
filter n0.O
HCHO,pg
1 2
315 126 C30
4 5 6