Environ. Sci. Technol. 2007, 41, 3843-3849
Effects of Relative Humidity on Chloroacetanilide and Dinitroaniline Herbicide Desorption from Agricultural PM2.5 on Quartz Fiber Filters W E N L I Y A N G A N D B R I T T A . H O L M EÄ N * Environmental Engineering Program, University of Connecticut, Storrs, Connecticut 06269-2037
This study quantified the release of seven relatively polar preemergence herbicides to the gas phase from soilgenerated PM2.5-loaded quartz fiber filters (QFFs) and bare QFF as a function of relative humidity (RH). A 48-hour desorption fraction, F48, was defined to evaluate the relative desorption behavior of herbicides from two families, chloroacetanilide (alachlor, butachlor, metolachlor, and propachlor) and dinitroaniline (pendimethalin, prodiamine, and trifluralin) using temperature- (8 °C) and humidity- (1064% RH) controlled air at a flow rate of 4 L/min. With increasing RH, an increase in F48 by a factor of 2-8 was observed for all herbicides, except metolachlor and butachlor, which showed significantly strong sorption to both sorbents. The conjugate carbonyl oxygen and amide nitrogen in the chloroacetanilide structure enables stronger specific interactions with the sorbents, leading to lower desorption compared to the dinitroaniline herbicides. Desorption of chloroacetanilides decreased in the order propachlor > alachlor > metolachlor ∼ butachlor, and desorption of dinitroanilines decreased in the order trifluralin > pendimethalin > prodiamine. These orders are consistent with the different substituents in the herbicide molecules for each family and their relative tendencies to coordinate with surface moieties as indicated by electron-donating capacity. Henry’s law constant and Abraham’s H-acceptor parameter were found to be useful empirical parameters for describing the F48 desorption behavior for all seven herbicides.
Introduction Concerns about the potential adverse effects of herbicides on the environment and human health have grown steadily because herbicide residues have been detected widely in the air, water, soil, and agricultural produce (1). The atmosphere is now recognized as a major pathway by which herbicides are transported and deposited, both within a short distance of their application sites and far downwind from their sources (2). Processes resulting in herbicide emission to the atmosphere during and after field application include volatilization (3), herbicide-laden soil resuspension during soil cultivation practices (4), and natural wind erosion (5). The partitioning * Corresponding author current affiliation: University of Vermont, School of Engineering, 213B Votey Hall, 33 Colchester Avenue, Burlington, VT 05405; phone: (802) 656-8323; fax: (802) 656-8446; e-mail:
[email protected]. 10.1021/es062692i CCC: $37.00 Published on Web 05/03/2007
2007 American Chemical Society
of herbicides between the gas phase and particle phase strongly influences their atmospheric residence time, transport distance, and transformation rates (6-11). Given the wide range of possible environmental behavior, having fundamental knowledge of the factors affecting gas/particle partitioning is essential for further understanding and controlling the atmospheric behavior of herbicides. Numerous studies have examined gas/particle partitioning of semivolatile organic compounds (12-19) but research has been focused on nonpolar compounds. Relative humidity (RH) has been recognized as an important parameter affecting this partitioning process (11, 20-22). Decreased partitioning to the solid phase has been observed with increasing RH in field and laboratory experiments for polycyclic aromatic hydrocarbons (PAHs) and n-alkanes on QFFs (30% < RH < 70%) (21) and urban particulate matter (UPM) (42% e RH e 95%) (22); and polychlorinated biphenyls (PCBs) and organochlorine compounds on UPM (35% < RH < 95%) (23). For small polar compounds, such as acetone and ethyl ether, increasing RH also reduced their partitioning to quartz (10% e RH e 90%) (20), clay minerals (40% < RH < 80%) (24), and combustion soot (10% < RH < 95%) (25). The effects of RH on compound gas/soil partitioning at the mechanistic level have recently been summarized as follows: (1) RH does not influence absorption of nonpolar compounds into soil organic matter; and (2) adsorption on minerals decreases with increasing RH for both nonpolar and polar compounds due to reduced intermolecular interactions (at RH < 90%) or reduced available surface area for adsorption (at RH > 90%) (11). Large, more polar herbicides typically have multiple functional groups that enable specific herbicide-sorbent interactions. Hence, previous work on gas/particle partitioning of nonpolar compounds and small polar compounds may not be directly applicable for these compounds. For example, higher RH conditions resulted in less sorption on UPM for γ-hexachlorocyclohexane, mecoprop, carbofuran, trifluralin and atrazine, but no such effect was observed for R-hexachlorocyclohexane, hexachlorobenzene, carbaryl, and diuron in the same field study (49% e RH e 70%) (6). Differences in compound polarity could not convincingly explain the observations for these two groups of pesticides, but diurnal variation in RH conditions may have influenced the reported field results. In the current work, chloroacetanilide and dinitroaniline herbicides (both widely used for preemergence control of annual grasses and certain broadleaf weeds (26, 27)) were studied to better quantify the influence of RH on their gas/ particle partitioning. Acetachlor, metolachlor, and alachlor are the most commonly used chloroacetanilides, and trifluralin and pendimethalin are the most popular in-use dinitroanilines (28). Because of their use as preemergence herbicides, they have great potential to exist as both gaseous and particle-bound species in the atmosphere. Thus, this study aims to (1) quantify the effects of RH on polar herbicide gas/particle partitioning; (2) understand how different functional groups affect their partitioning behavior; and (3) evaluate the relationships between herbicide physicochemical properties and their partitioning behavior. In aerosol sampling, filters are usually employed to collect the particle-phase portion of semivolatile organic compounds (SOCs) and QFFs are commonly used because of their low blank levels, despite the documented artifact caused by sorption of gas-phase SOCs to filter surfaces (21, 29, 30). Knowledge of the herbicide-QFF partition coefficient is therefore important to correct for gas/particle partitioning VOL. 41, NO. 11, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3843
FIGURE 1. Experimental desorption apparatus with RH control. artifacts generated during field sampling. It is also well-known that quartz surfaces tend to sorb water (21), Thus RH is expected to affect the magnitude of the QFF-sorption artifact. Other reasons to study partitioning of herbicides to QFFs include the following: (a) quartz is commonly found in atmospheric particle samples; (b) oxide minerals are important constituents of soil and provide a portion of the surface area available for partitioning to soil particles; (c) the fibers of a QFF are nonporous, making adsorption to the external fiber surfaces the only sorption mechanism (21); and (d) QFF can be baked at relatively high temperature to obtain very low blank levels.
Experimental Section Herbicides. Analytical grade standards of selected chloroacetanilide and dinitroaniline herbicides were purchased from Chem Service Incorporated (West Chester, PA) and used without further purification to study the partitioning behavior of herbicides with a range of physicochemical properties. The chemical structures and physicochemical properties of the chloroacetanilides (metolachlor, propachlor, alachlor, and butachlor) and dinitroanilines (pendimethalin, trifluralin, and prodiamine) are listed in Table S1, found in the Supporting Information. PM2.5 Generation from Soil. Bulk soil was collected from the University of Connecticut Plant Science Research Farm and wet sieve particle size analysis indicated that the soil is a sandy clay loam. The soil was air-dried, ground, and sieved by a 200-mesh sieve (75 µm) prior to being resuspended and collected as PM2.5 using a fluidized bed resuspension chamber (4, 31-33). Approximately 2-5 mg of PM2.5 was collected on 47-mm QFFs (Pallflex7202, Pall Laboratory, Ann Arbor, MI) for use in desorption experiments. The QFFs were baked at 450 °C overnight prior to PM2.5 collection. The specific surface area (SSA) determined by N2-BET method was 3.72 ( 0.32 and 2.66 ( 0.01 m2/g for bare and PM2.5-loaded QFF, respectively. Desorption Experiments. An RH-controlled desorption apparatus was constructed to conduct the gas/solid partitioning experiments (Figure 1). Room air was dried by silica 3844
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 11, 2007
gel, filtered by charcoal (Alltech Associates, Deerfield, IL) and HEPA (Pall Corporation, Ann Arbor, MI), and then humidified to the desired RH by controlling the flow through a humidifier (G in Figure 1, MH-070-24P-4, Perma Pure LLC, Toms River, NJ). Humidified air was then split into two identical streams and each gas stream passed through a herbicide-treated sorbent sample (in a 47-mm stainless steel filter holder) at a constant flow rate. Gas-phase herbicides desorbed from the treated sorbents were captured by polyurethane foam plugs (PUFs) in Teflon holders and backup PUF plugs were located downstream to check for breakthrough. This setup was operated using a vacuum pump (Gast, model 71R645-V114-D303X, Benton Harbor, MI) to draw air through the system. Continuous temperature and RH measurements were made with a model TP120 temperature and humidity data logger (Dickson, Addison, IL). Teflon tubing was used throughout the system. Desorption experiments were conducted for 48 h in a temperature-controlled chamber at 8.1 ( 0.2 °C and five different experimental RH conditions: 9.7 ( 0.7%, 24.5 ( 1.5%, 32.9 ( 0.7%, 53.5 ( 1.0%, and 64.1 ( 0.4%. Procedure. The sorbent (PM2.5 loaded on baked QFF or bare baked QFF) was preconditioned for 24 h under the desired experimental temperature and RH prior to being spiked with 500 µL of a 10 ppm acetone solution of the seven herbicides. After air drying for 30 min, the herbicide-treated sorbent samples were then loaded into the filter holders for desorption. Triplicate samples of spiked sorbent were collected to quantify the initial mass of herbicide applied prior to desorption. A continuous airflow of 4 L/min was maintained through the sorbent and PUFs. After the 48-hour desorption period, the PUF and QFF samples were sealed in glass jars and stored at -20 °C until analysis. Analysis Methods. All PUF and QFF samples were extracted using a Spe-ed 2 Supercritical Fluid Extractor (Applied Separations, Allentown, PA). Benfluralin and acetachlor were added to each sample as surrogates before extraction. The extractions were performed with supercritical carbon dioxide with 10% v/v acetone modifier at 2 mL/min at 20 MPa and 50 °C (34). The optimized extraction method
for PUFs was 30 min static mode followed by 90 min dynamic mode, and for QFF samples was 10 min static mode followed by 50 min dynamic mode. Hexane was used as the trap solvent and the collected solutions were concentrated to 1 mL by air evaporation. The PUF plugs were cleaned using Soxhlet extraction with two 250 mL aliquots of acetone, replaced after 24 h, dried, and stored for reuse. All sample extracts were analyzed using an Agilent 6890GC/5973MSD operated in selected ion monitoring mode. The capillary column was a 30 m long, 0.25 mm i.d. DB-5MS with a 0.25 µm film thickness. The GC temperature program was 45 °C for 1 min, 30 °C/min to 130 °C, hold for 3 min, 12 °C/min to 180 °C, and 7 °C/min to 240 °C. QA/QC. Quality control checks routinely performed included system blanks, surrogate recovery, and replicates. No target herbicide was detected in the extraction of QFF and PUF blanks. Acceptable mass balance (ratio of the gasand particle-phase herbicide collected during desorption to the initial spiked amount) ranging from 82.4 to 105.4% was obtained for all experiments. No PUF breakthrough was detected in the desorption experiments. Duplicate experiments were performed for both PM2.5 and QFF desorption at 64% RH. Data Analysis. The gas/particle partitioning behavior of a compound is commonly characterized in two ways: the fraction of a compound sorbed to the particles, φ (23), and the gas/particle partition coefficient, KP (35). In this study, laboratory desorption experiments on herbicide-spiked soilderived particles were conducted to mimic the real-world gas/particle distribution of preemergence herbicides after their field application. The mass of herbicide released from the sorbent to the gas phase, normalized by the initial mass of herbicide in the sorbent (PM2.5-loaded or bare QFF) is defined as the desorption fraction, F, which is a function of desorption time t.
Ft )
MG,t Mini
(1)
where Mini (ng) is the initial mass of herbicide in the sorbent before desorption; MG,t (ng) is the total mass of gas-phase herbicide desorbed from the sorbent over t (hour). Therefore, Ft represents the fractional approach to complete desorption at time t. In the present work, the sorbents were exposed to clean purging air over 48 h in all experiments and F48, i.e., desorbed fraction over 48 h, was used to describe the relative desorption behavior of the study herbicides.
Results and Discussion PM2.5-Loaded QFF Herbicide Desorption. Five nominal RH conditions (10%, 25%, 33%, 54%, and 64%) were examined in the desorption experiments of herbicide-spiked PM2.5loaded QFF. The values of F48 for PM2.5 exhibited a strong dependency on RH (Figure 2). Increased desorption of herbicide from PM2.5 to the gas phase was observed for all seven compounds at higher RH. As RH increased from 10% to 64% for the chloroacetanilides, F48 was enhanced 5.3 and 8.7 times for propachlor and alachlor, respectively, but only by a factor of ∼2 for metolachlor and butachlor, and for the dinitroanilines, F48 was increased by factors of 1.5 to 3.9. Generally, F48 for dinitroaniline desorption from PM2.5-loaded QFF was higher than for the chloroacetanilides under each experimental RH condition. For the chloroacetanilides, F48 for PM2.5-loaded QFF/gas desorption increased in the following order: butachlor e metolachlor < alachlor < propachlor; for the dinitroanilines, the order was as follows: prodiamine < pendimethalin < trifluralin. It is important to note that less than 10% of the applied mass of both metolachlor and butachlor was desorbed from PM2.5-loaded
QFF under all experimental conditionssthis is significantly lower than the fractions measured for all other herbicides. Herbicide Desorption from Bare QFF. Four RH conditions (10%, 33%, 54%, and 64%) were studied in the bare QFF desorption experiments. An increase in herbicide desorption from bare QFF to the gas phase was also observed for all study herbicides (Figure 2) with increasing RH over the range of 10% to 64%. When RH increased from 10% to 64%, F48 values increased 3 and 8 times for propachlor and alachlor, respectively, but only slightly again for metolachlor and butachlor. For dinitroanilines, as RH increased from 10% to 64%, F48 increased by factors of 1.3, 3.4, and 6.1 for trifluralin, pendimethalin, and prodiamine, respectively. These results are consistent with the observations made above for chloroacetanilide and dinitroaniline desorption from PM2.5loaded QFF. For the chloroacetanilides, F48 increased in the following order: butachlor e metolachlor < alachlor < propachlor; for the dinitroanilines, the order was as follows: prodiamine < pendimethalin < trifluralin. The order of F48 for the bare QFF was the same as that observed for the PM2.5loaded QFF desorption, suggesting that herbicide properties were more important than sorbent properties in determining the extent of desorption over 48 h. For the chloroacetanilides, the magnitude of F48 varied with sorbent type (Figure 2a). For example, at 64% RH, F48 for propachlor was about 0.2 higher when desorbing from PM2.5-loaded QFF than from bare QFF. However, for the dinitroanilines, there was no measurable significant difference in F48 between the PM2.5-loaded and bare QFF desorption (Figure 2b). This observation agrees with previous results for metolachlor and pendimethalin desorption from PM2.5loaded and bare QFF (4). Similarly, Mader and Pankow measured lower sorption of PAHs to QFFs previously used to sample suburban air than on bare QFF at the same RH condition (30). In the current study, the smaller specific surface area of PM2.5-loaded QFF (2.66 m2/g) compared to bare QFF (3.72 m2/g) suggests that the generated PM2.5 partially blocked some of the QFF surface sorption sites, subsequently resulting in higher measured desorption fractions for the chloroacetanilide herbicides. Influence of RH on Herbicide Desorption. Under natural conditions, environmental surfaces, including soil and airborne particulate matter, are typically covered by some adsorbed water (11). Studies have shown that hydrophilic mineral surfaces have 1-2 layers of water molecules at 30% ambient RH and water coverage increases to 5-10 molecular layers when RH reaches 90% (11). It is generally agreed that water molecules, because of their polar nature, effectively compete with organic compounds for adsorption sites on hydrophilic surfaces under low surface moisture conditions. However, at RH > 30%, when more than one layer of water is likely already covering the sorbent surfaces, the displacement of organic compounds by water adsorption does not explain the observed continuous increase in herbicide desorption. Thus, an alternative mechanistic interpretation is needed. A multi-mechanism approach proposed by Pennell indicated that sorption of organic vapors to hydrated soil involves four possible mechanisms: (a) adsorption on mineral surfaces, (b) adsorption at the gas-liquid interface, (c) partitioning into soil organic matter (SOM), and (d) dissolution into adsorbed water films (36). However, compound dissolution within the surface water film is not a reasonable mechanism at RH below 90% due to the small dimensions of the water layer compared to the adsorbate size (11). Hence, only mechanisms (a), (b), and (c) likely contributed to the observed RH effect for herbicides in the current study where experimental RH varied from 10 to 64%. The relative importance of these mechanisms to the sorption process should be determined by the properties of the sorbate VOL. 41, NO. 11, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3845
FIGURE 2. PM2.5-loaded (solid circles) and bare QFF (open triangles) desorption fraction at various RH conditions for (a) chloroacetanilide (note the different y-axis scales for butachlor and metolachlor) and (b) dinitroaniline herbicides. Results of duplicate experiments for both PM2.5 and QFF desorption are shown for 64% RH (at 8 °C). and sorbent: vapor pressure (Po), water solubility (SW) and octanol-water partition coefficient (KOW) of the sorbate; and sorbent organic content (fOC), specific surface area (SSA), and water content (or related, ambient RH). Linear best-fit relationships were determined between F48 and RH for the study herbicides both for PM2.5-loaded and bare QFF desorption (F48 ) a‚RH + b; Table S2, Supporting Information). For PM2.5 desorption, coefficients of determination (R 2) were above 0.9 for propachlor, pendimethalin, and prodiamine, but less than 0.7 for butachlor and metolachlor. For bare QFF desorption, the linear relationships between F48 and RH were above 0.8 for all study compounds except butachlor which had a zero best-fit slope. A larger regression slope indicates a higher dependency of F48 on RH. The experimental results showed that, for the 3846
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 11, 2007
chloroacetanilide herbicides, changing RH affected the desorption of propachlor much more than the other compounds, especially metolachlor and butachlor, for which RH had little influence. The desorption of pendimethalin had the greatest dependency on RH among the three dinitroaniline herbicides (Table S2). It is interesting to note that the two compounds with the greatest dependency on RH, propachlor and pendimethalin, are also the herbicides with the lowest molecular weight and highest SW in their respective herbicide families. Influence of Physicochemical Properties on Herbicide Desorption. Vapor pressure is often used to model herbicide transfer from soil to the air (37, 38) and sub-cooled liquid vapor pressure (PoL) has been successfully related to the
TABLE 1. Best-Fit F48-Herbicide Propertya Linear Regression Coefficients (a and b) with Their Standard Errors (in Parentheses) for the Study Herbicides at Each Experimental RH Condition PM2.5-loaded QFF
PLo
10%
24%
33%
54%
64%
10%
33%
54%
64%
a
0.06 (0.08) 0.36 (0.28) 0.09
0.11 (0.10) 0.66 (0.35) 0.18
0.13 (0.10) 0.82 (0.35) 0.25
0.18 (0.12) 1.04 (0.41) 0.32
0.14 (0.12) 0.94 (0.42) 0.20
0.08 (0.10) 0.45 (0.34) 0.12
0.10 (0.11) 0.58 (0.37) 0.14
0.13 (0.13) 0.79 (0.44) 0.17
0.12 (0.13) 0.81 (0.45) 0.14
-0.13 (0.07) 0.90 (0.40) 0.40
-0.19 (0.08) 1.36 (0.47) 0.52
-0.22 (0.08) 1.61 (0.46) 0.60
-0.26 (0.10) 1.91 (0.55) 0.59
-0.26 (0.09) 1.94 (0.51) 0.63
-0.16 (0.09) 1.07 (0.49) 0.41
-0.19 (0.09) 1.36 (0.49) 0.51
-0.24 (0.10) 1.69 (0.60) 0.50
-0.24 (0.10) 1.80 (0.59) 0.53
0.14 (0.02) 0.48 (0.05) 0.92
0.18 (0.03) 0.67 (0.09) 0.85
0.17 (0.05) 0.76 (0.12) 0.72
0.19 (0.07) 0.86 (0.18) 0.58
0.18 (0.07) 0.87 (0.17) 0.58
0.17 (0.03) 0.54 (0.08) 0.86
0.19 (0.03) 0.66 (0.08) 0.89
0.22 (0.04) 0.84 (0.11) 0.83
0.22 (0.05) 0.88 (0.13) 0.78
-1.01 (0.29) 0.98 (0.23) 0.71
-1.40 (0.32) 1.40 (0.26) 0.79
-1.51 (0.33) 1.58 (0.26) 0.81
-1.82 (0.39) 1.89 (0.32) 0.81
-1.76 (0.37) 1.87 (0.30) 0.82
-1.16 (0.42) 1.10 (0.34) 0.61
-1.42 (0.35) 1.39 (0.28) 0.77
-1.90 (0.26) 1.87 (0.21) 0.92
-1.96 (0.21) 1.97 (0.17) 0.95
b R2 KOA
a b R2
KH
a b R2
Σβ2H
bare QFF
RH
a b R2
a Estimated Po, K , and K values at 8 °C and ΣβHvalues for study herbicides are listed in Supporting Information Tables S3, S4, S5, and S6, OA H L 2 respectively.
extent of gas/particle partitioning of nonpolar compounds (39). Compounds with higher PoL tend to partition more to the gas phase from soil (40) or PM (17, 41). However, in the current study the order of F48 could not be explained by this parameter. The poor fits of F48 ) a‚logPoL+ b (R 2 ) 0.09-0.32; Table 1) suggest PoL alone is not a good predictor for desorption of these relatively polar herbicides. Because PoL describes only the intramolecular bonding forces of a sorbate compound, it is not appropriate for characterizing the sorbate-sorbent intermolecular interactions which would be disrupted during the desorption processes. Two-phase partitioning constants, such as octanol-air partition coefficient (KOA), have been found to better parameterize gas/particle partitioning of many nonpolar compounds than did PoL (13, 42, 43). For the PM2.5-loaded and bare QFF desorption of herbicide, the regression, F48 ) a‚ logKOA + b had R 2 between 0.40 and 0.63 (Table 1), a considerable improvement over the regression models for PoL (R 2 0.09-0.32). Adsorption of herbicides likely occurred at the air-water interface and directly to quartz surfaces. Thus, Henry’s law constant KH was examined as a useful parameter to estimate F48 for PM2.5-loaded and bare QFF desorption. The best-fit linear regression of F48 ) a‚logKH + b had R 2 between 0.58 and 0.92 (Table 1), indicating KH was a better predictor than either PoL or KOA for the desorption of these relatively polar herbicides. The regression results also show that correlations between F48 and logKH for bare QFF desorption were better than those for PM2.5-loaded QFF desorption for most of the RH conditions studied. This variation may be explained by differences in sorbent complexity. The ∼29% reduction in measured sorbent SSA indicates the PM2.5 derived from the bulk soil (fOC 1.68%; clay fraction 24%, by hydrometer test) contained components that would “block” QFF surface sites, including soil organic matter, oxides, and clays that contain polar functional groups capable of hydrogen bonding and other more specific interactions with the quartz surface. Herbicide Structure and Desorption Fraction. Herbicides from the same family have their common core structure,
and individual herbicides are distinguished by different substituent functional groups. Structural differences among herbicides in one family affect compound physicochemical properties, and the types and extent of interaction with sorbents, resulting in different desorption behavior. Chloroacetanilide Herbicides. The common structural components of chloroacetanilides are the carbonyl oxygen and amide nitrogen attached to the benzene ring (Table S1). Both atoms have the potential to form hydrogen bonds with electron acceptors, such as hydroxyl and carboxyl groups of soil humic acid, as previously indicated by FT-IR spectra for saturated clay and humic acid systems of chloroacetanilides (26, 44). Hydrogen bonding between the carbonyl oxygen and amide nitrogen of chloroacetanilides with hydroxyl groups on QFF surfaces would inhibit herbicide desorption from the sorbents. The presence of electron-donating substituent groups on the side chains of chloroacetanilides would also increase the electron density on the amide nitrogen and carbonyl oxygen, thus enhancing herbicide-sorbent interactions (26). For the study chloroacetanilides, the capacity of electron donation by the R3 substituents decreases in the following order: butachlor [-CH2-O-(CH2)3-CH3] > metolachlor [-CH(CH3)-CH2-O-CH3] > alachlor [-CH2-OCH3] > propachlor [-CH(CH3)-CH3]. This order coincides with the reverse order of the measured F48 values for PM2.5loaded and bare QFF desorption (propachlor > alachlor > metolachlor g butachlor). In other words, desorption to the gas phase is enhanced when R3 substituents have weaker electron-donating capacity to the amide nitrogen atom that is involved in forming specific interactions with the adsorbent. Dinitroaniline Herbicides. The presence of two nitro groups on the benzene ring in dinitroanilines has been reported as explaining the observed strong molecular bonding to soil organic matter, chiefly via hydrophobic partitioning in water-saturated systems (27, 45-47). Because bare QFF is free of organic matter, sorption of dinitroanilines to QFF should be very weak unless the herbicide functional groups allow hydrogen bonding. This observation is consistent with the desorption fraction data for the herbicides. For the three dinitroaniline herbicides studied, only prodiamine has a NH2 VOL. 41, NO. 11, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3847
group that can act as a hydrogen bond donor with the hydroxyl groups on the QFF surfaces; thus explaining the lower desorption fraction of prodiamine from the QFF and PM2.5 sorbents compared to trifluralin and pendimethalin. Modeling Effects of RH and Herbicide Properties on Herbicide Desorption. For the relatively large and polar herbicides examined in this study, F48 increased with increasing RH and, at a given RH, herbicides tended to remain sorbed to the sorbent surfaces when functional groups capable of hydrogen bonding to the silanol surfaces of QFF were present in the chemical’s structure. Thus, empirical regression models were developed to account for both the sorbent and sorbate properties that control the partitioning behavior of these herbicides. At a given RH condition, herbicide desorption was estimated reasonably well, based on the compound’s KH as well as its hydrogen-bonding basicity (H-acceptor) parameter, ΣβH 2 , estimated from Abraham type equations (ADME Boxes Ver 3.0; http://www.ap-algorithms.com/absolv.htm). In fact, KH and ΣβH 2 (which are linearly related to one another) were better single-parameter predictors of F48 than PoL and KOA for any given RH condition (Table 1). In order to better understand the relative importance of herbicide properties and environmental conditions (i.e., RH) on desorption behavior, ANOVA multiple regression analysis was applied to characterize the influence of KH, ΣβH 2 , and RH on measured F48 values. Best-fit multiple regression models for desorption of the seven herbicides from the two sorbents at 10-64% RH conditions (8 °C) were (all p-values
