Herbicides are widely used in agriculture globally. In the U.S. alone, an estimated 260 Gg of herbicides were applied in 2019. 1 In addition to the herbicide itself, herbicide formulations typically contain supplemental chemicals to modify the herbicide's properties. 2 For example, glyphosate, the most used herbicide in the U.S., 3 is often formulated with an amine counterion. In 2017, 55% of glyphosate applied to soybeans, cotton, and wheat was applied with an amine (i.e., isopropylamine (IPA), dimethylamine (DMA)). 4 These amines are included in formulations at an equimolar concentration to increase the solubility of glyphosate, as well as other herbicides (i.e., 2,4-dichlorophenoxyacetic acid (2,4-D)), by deprotonating the acidic herbicide to form a salt. 5,6 Beyond solubility, semi-volatile herbicides like 2,4-D and dicamba (3,6-dichloro-2methoxybenzoic acid) are formulated as amine salts to reduce their vapor pressure and thereby prevent off-target damage to other vegetation resulting from herbicide drift. [7][8][9] Concerns about the drift of these two herbicides have been amplified since the release of genetically modified (GM) crops that tolerate 2,4-D and dicamba in 2014-2015. 2 Although the postemergent use of 2,4-D and dicamba on GM tolerant crops has been considered imperative to overcome the reduced effectiveness of glyphosate, 10,11 this practice has been jeopardized by widespread damage to non-tolerant crops associated with their drift. 2 Drift of these herbicides may also impact yields of other fruits and vegetables, [12][13][14][15] function of neighboring ecosystems, 16 and the emergence of resistant weeds. 17 While additional processes including spray drift can contribute to off-target herbicide movement, 18 herbicide volatilization, which can occur for days after application, 19,20 contributes significantly to drift even when amine-salt formulations are used. 2,21 To prevent volatilization, dicamba applications on GM tolerant crops are restricted to formulations that include diglycolamine (DGA) or n,n-bis-(3-aminopropyl)methylamine (BAPMA), which have reduced herbicide volatilization 19,20,[22][23][24][25][26][27] relative to older DMA formulations. 28,22,19,23,24 Unfortunately, herbicide volatilization from GM tolerant crops has remained a persistent challenge despite these advances. 2 Notably, the volatilization of amines themselves has not been investigated, even though the amines included in formulations have higher vapor pressures than the herbicides (SI Table S1). The loss of amines to the atmosphere may promote herbicide volatilization leading to heightened off-target drift damage. In addition, volatilized amines may affect atmospheric processes that impact health and climate. Despite their widespread use, herbicide-amine salt formulations have not been considered alongside other anthropogenic sources of amine pollution to the atmosphere (e.g., animal husbandry, industry). 29 Amine vapors oxidize to form potential carcinogens such as nitrosamines and nitramines. [30][31][32][33][34][35] Both the parent amines and their oxidation products contribute to formation and growth of particles in the atmosphere, 36- 38 which influences the climate [39][40][41] and negatively impacts human health. 42 For example, polystyrene particles modified to have amine functional groups present on their surface have been associated with greater deleterious health outcomes in laboratory animals compared to unmodified polystyrene particles. 43,44 In this study, we experimentally evaluated amine loss from herbicide-amine salt formulations, as well as confirmed that amine vapor was detectable above herbicide-amine solutions. By measuring co-occurring losses of amine and dicamba, we proposed underlying processes that may contribute to volatilization of both components. Integrating available data, we estimated the magnitude of amine input into the environment from herbicide-amine salts relative to other known sources of amine pollution.
Chemicals, suppliers, and glassware cleaning procedures can be found in the Supporting Information (SI Section S1). Due to inhalation hazards, open vessels with amines were handled in a fume hood. Stock solutions of all chemical components were prepared in Milli-Q water, with the exception of 2,4-D free acid that was prepared in ~40/60 (v/v) acetonitrile/water due to low aqueous solubility. 45
Expanding on prior protocols, 24 herbicide-amine residues were dried from a 2.0 mL aqueous solution of amine and herbicide at an equimolar concentration of 123 μM in a 50 mL glass beaker (VWR catalog number 13912-149), resulting in an expected amount of 246 nmoles of dicamba and amine each present (𝑛 𝑥 𝑒𝑥𝑝𝑒𝑐𝑡𝑒𝑑 , where x corresponds to the constituent measured). When the number of moles in solution were measured (𝑛 𝑥 𝑠𝑜𝑙𝑛 ), 2.3 mL of solution was prepared instead of 2.0 mL so that the additional 300 μL aliquot could be taken for derivatization and quantification of the amine. The solutions were dried in a fume hood at room temperature (~22°C) for 24 h before at least two beakers were extracted to determine the moles remaining in the herbicide-amine residue after drying (𝑛 𝑥 0 ℎ,𝑑𝑟𝑖𝑒𝑑 ). To determine moles lost during drying, we calculated the difference between the molar quantity expected initially in solution and the measured value in the dried residue after 24 h (𝑛 𝑥 𝑒𝑥𝑝𝑒𝑐𝑡𝑒𝑑 -𝑛 𝑥 0 ℎ,𝑑𝑟𝑖𝑒𝑑 ). We chose 24 h because it allowed for evaporation of all visible water from the beaker. To measure volatilization from residues at high temperatures, the remaining beakers containing the dried residues (prepared in triplicate) were then placed on hot plates at elevated temperatures for a time (t) before extraction to determine moles remaining (𝑛 𝑥 𝑡,𝑑𝑟𝑖𝑒𝑑 ). To compare losses from the residue phase, we elevated the temperature to 40°C for 48 h before measuring remaining molar quantity to calculate the difference in moles remaining (𝑛 𝑥 0 ℎ,𝑑𝑟𝑖𝑒𝑑 -𝑛 𝑥 48 ℎ,𝑑𝑟𝑖𝑒𝑑 ). We chose 48 h
for most experiments because it allowed for enough volatilization for comparison between the different herbicide-amine salts. To control for any environmental conditions (e.g., relative humidity) that may influence results, all data included in individual figure panels were collected from experiments conducted simultaneously.
Residues were extracted with 2 mL of a 50/50 (v/v) acetonitrile/water mixture (SI Section S2). To improve subsequent analysis, samples containing glyphosate were instead extracted with 2 mL of 100% water (SI Section S3). Each extract was divided into 1 mL for direct quantification of dicamba or 2,4-D (when applicable) and 300 μL for amine and/or glyphosate derivatization. Extracted aliquots were stored at 4°C in 2 mL vials until derivatization or quantification. Amines and glyphosate were derivatized using 9fluorenylmethyloxycarbonyl chloride (FMOC-Cl) based on a prior method 46 (SI Sections S4, S5). Dicamba, 2,4-D, derivatized amines, and derivatized glyphosate were quantified on HPLC-UV (SI Section S6).
IPA was measured in the headspace of 20 mL amber glass vials (Thermo Scientific catalog number B7921VO) containing 100 μL solutions by in situ TD GC-MS (SI Section S7).
Measurements were performed on four sets of six glyphosate-IPA solutions and three sets of five glyphosate-only solutions, with each included constituent diluted in water to 13.2 mM. Vials were sampled individually; one gas phase measurement was taken immediately after sample set preparation and every 20 min thereafter. The IPA GC peak was positively identified with a calibration standard as well as comparison with a library mass spectrum 47 (SI Figure S6).
Herbicide-amine salts are applied as aqueous solutions that subsequently dry to produce residues on leaf or soil surfaces. 2 We first measured changes in the remaining amounts of DMA and 2,4-D when solutions prepared with equimolar concentrations of the two constituents were dried to generate residues at room temperature over 24 h (Figure 1A). While the amount of 2,4-D was unchanged, the remaining moles of DMA decreased by 25±2%. The majority of this loss occurred between 12 and 16 h, which corresponded to complete evaporation of visible water. Consequently, the transition from aqueous solution to dried residue appears to drive DMA volatilization at ambient temperatures possibly due to the change in protonation state of the amine as the aqueous solution dries to a residue.
We expanded our investigation to compare the losses of DMA and IPA from salts prepared with three different herbicides (i.e., dicamba, glyphosate, 2,4-D) corresponding to combinations used in practice 4 (Figure 1B). For all six DMA and IPA salts, we observed that amine losses, which ranged from 90±22 to 219±11 nmoles over the experiment duration (Figure 1B), greatly exceeded losses of the corresponding herbicides (𝑛 ℎ𝑒𝑟𝑏𝑖𝑐𝑖𝑑𝑒 𝑒𝑥𝑝𝑒𝑐𝑡𝑒𝑑 ̶ 𝑛 ℎ𝑒𝑟𝑏𝑖𝑐𝑖𝑑𝑒 48 ℎ,𝑑𝑟𝑖𝑒𝑑 <23 nmoles, SI Figure S7). Losses of amines from all salts occurred during both the drying period (𝑛 𝑥 𝑒𝑥𝑝𝑒𝑐𝑡𝑒𝑑 -𝑛 𝑥 0 ℎ,𝑑𝑟𝑖𝑒𝑑 ) and from the residue phase at elevated temperature (𝑛 𝑥 0 ℎ,𝑑𝑟𝑖𝑒𝑑 -𝑛 𝑥 48 ℎ,𝑑𝑟𝑖𝑒𝑑 ). DMA was consistently lost to a greater extent than IPA, likely corresponding to ~3-fold higher vapor pressure of DMA relative to IPA (SI Table S1). While amine losses tended to be moderately higher from glyphosate salts than dicamba and 2,4-D salts, amine losses were significant from all combinations regardless of herbicide type indicating that similar phenomena may drive volatilization in all cases.
Using in situ thermal desorption gas chromatography-mass spectrometry (TD GC-MS), we confirmed that amine volatilization contributed to amine loss by measuring the presence of IPA in the headspace above glyphosate-IPA salt solutions concentrated ~100-fold to reflect conditions during drying (Figure 1C). Though IPA was consistently detected in the headspace (confirmed by library comparison), 47 we observed a highly variable amount of IPA. Additional analysis at 6 time points spanning a 100 min period indicated that the amine flux from these concentrated solutions was highly dynamic (SI Section S9), likely contributing to variability in the measured amounts of IPA among samples because our method only permitted sampling for 5 min for each 20 min cycle rather than continuous measurement. Additional variability may also result from minor differences in sensitivities between instrument collection channels (SI Section S9). To exclude interference due to thermal decomposition of glyphosate into amine products, 48,49 we also analyzed solutions prepared without IPA (Figure 1C). We observed that no IPA was detected with the exception of the first control sample analyzed (Rep.1), which we attributed to minor carryover of IPA from analysis of the IPA-containing samples (SI Section S9). Consequently, volatilized IPA from the glyphosate-IPA salt solution was the dominant contribution to the IPA measured in the headspace of the IPA-containing samples (Figure 1C).
After demonstrating that amines volatilize from all herbicide-amine salts tested (Figure 1B), we next aimed to elucidate key factors that may control amine volatilization from herbicide-amine salts (e.g., vapor pressure, amine-dicamba interactions). We investigated these factors by testing experimental variables including temperature, experiment duration, and the initial ratio of amine to herbicide in solution. In these experiments, we monitored the loss of the herbicide dicamba alongside amines, which also tested the effect of amine loss on dicamba volatilization.
One important parameter is temperature, which we hypothesized would increase the vapor pressure of the molecules, as well as alter additional factors (e.g., protonation state) [50][51][52][53][54] particularly in the dried residue phase. Elevated temperature consistently increased the loss of both DMA (Figure 2A) and IPA (Figure 2B) from residues over 48 h. Consistent with previous work, 24 dicamba losses were significant at higher temperatures (i.e., 60°C or 80°C). However, the lower loss of dicamba relative to the amines, particularly at moderate temperatures (i.e., 40°C or 60°C), suggests that at least some of the amines are lost from the residue as neutral molecules rather than dicamba-amine ion pairs and that the extents of proton transfer from dicamba to the amines are possibly incomplete in our residues.
We expanded the amines studied to include DGA (Figure 2C) and BAPMA (Figure 2D), which are used in specific formulations intended to prevent dicamba volatilization. 18 Unlike DMA and IPA, neither DGA nor BAPMA were lost to measurable extents during drying (SI Section S2), possibly due to their lower vapor pressures compared to DMA and IPA (SI Table S1). DGA's low pure phase vapor pressure is likely due to the intermolecular hydrogen bonds between alcohol and amine groups that are present in ethanolamines, 55 which increase the enthalpy of vaporization. 56 We expect that the additional hydrogen bonding groups present on DGA compared to DMA and IPA have a similar impact in our residues, which limits the volatilization of DGA. Like DMA and IPA, DGA loss increased consistently with increasing temperature; however, dicamba loss was only measurable at the highest temperature (i.e., 80°C). Surprisingly, among the four amines, BAPMA loss was highest at the lower temperatures (20°C and 40°C) but did not increase further at higher temperatures (Figure 2D).
Dicamba loss from the dicamba-BAPMA salt residue was also uniquely absent at all temperatures possibly due to the two additional amine functional groups present on BAPMA.
At selected temperatures, we extended the experiment duration to determine if amine loss resulted in greater dicamba loss over longer times (Figure 3A-C). We observed that DMA loss during drying was two-fold higher than in other experiments in this study (e.g., Figure 1B), which we attributed to seasonal variations in laboratory ambient conditions (e.g., relative humidity) that were controlled by exclusively comparing data collected simultaneously. We observed that DMA loss from the final residue at both 40°C and 60°C continued until at least 96 h, resulting in ~80% loss at 40°C (Figure 3A) and near-complete loss at 60°C (Figure 3B).
Despite extensive DMA loss at both temperatures, dicamba loss was much greater at 60°C, approaching near-completion though lagging DMA loss. In contrast, dicamba loss at 60°C was lower in the presence of DGA, which also exhibited reduced loss relative to DMA at the same temperature (Figure 3C). Consequently, dicamba loss from the residue phase at longer times was influenced by the temperature, but also the identity and remaining amount of the amine.
To gain further insight into the influence of amine amount on dicamba loss, we investigated how the molar ratio of amine to dicamba affects the loss of each component during and after drying. To examine co-occurring losses of amine (i.e., DMA) and dicamba during drying, we reduced the amount of DMA in the initial solution while holding dicamba concentration constant (Figure 3D). At lower DMA/dicamba ratios, we observed, for the first time in this study, dicamba loss during drying at room temperature. In contrast, DMA loss during drying decreased at lower ratios and was not measurable loss from solutions prepared with 0.25/1 𝑛 𝐷𝑀𝐴 𝑒𝑥𝑝𝑒𝑐𝑡𝑒𝑑 /𝑛 𝑑𝑖𝑐𝑎𝑚𝑏𝑎 𝑒𝑥𝑝𝑒𝑐𝑡𝑒𝑑 (Figure 3D). We hypothesize that the reduced DMA loss measured at lower ratios is due to the remaining fraction of DMA being ionic, and therefore less volatile, after undergoing proton transfer with dicamba.
To examine amine and dicamba losses from the dried residue, we selected DGA, which is not lost during drying, so that the initial ratio of DGA/dicamba in the dried residue could be controlled. Reducing the ratio of 𝑛 𝐷𝑀𝐴 𝑒𝑥𝑝𝑒𝑐𝑡𝑒𝑑 /𝑛 𝑑𝑖𝑐𝑎𝑚𝑏𝑎 𝑒𝑥𝑝𝑒𝑐𝑡𝑒𝑑 in the initial solution from 1/1 to 0.75/1 led to less DGA loss when subjected to elevated temperature (60 °C) after drying (Figure 3E).
However, unlike DMA loss during drying, DGA loss after drying remained measurable even at lower DGA/dicamba ratios (Figure 3E). Consistent with prior results conducted at 40°C, 24 dicamba loss at 60°C increased at lower DGA/dicamba ratios (Figure 3E). Under conditions where the starting amount of DGA was reduced, dicamba volatilization exceeded that of DGA, indicating that dicamba may also be lost as a neutral molecule from our residues. Unlike with the reduced DMA/dicamba residues from which DMA loss was no longer measurable at 0.25/1 𝑛 𝐷𝑀𝐴 𝑒𝑥𝑝𝑒𝑐𝑡𝑒𝑑 /𝑛 𝑑𝑖𝑐𝑎𝑚𝑏𝑎 𝑒𝑥𝑝𝑒𝑐𝑡𝑒𝑑 ratio, we did not reach a ratio where DGA loss was no longer measurable.
This could be due to the changing amount of dicamba present in the residues as it volatilized, as well as the higher temperature (i.e., 60°C versus 20°C used in the DMA experiments) both increasing the volatility of DGA and decreasing the extent of proton transfer to form additional volatile neutral DGA molecules.
In the initial dilute aqueous solutions, we expect that deprotonation of the acidic herbicides (pKa 1.9-2.7, SI Table S1) and protonation of the basic amines (pKa 9.6-10.7, SI Table S1) occur via water-mediated proton transfer, resulting in a pair of non-volatile ions.
The ionic speciation, coupled with other possible factors (e.g., rates of air-water exchange of the neutral molecules), prevents volatilization of both the amine and the herbicide (Figure 1A, 0-12 h). As water evaporates, the solutes change from minor components in the aqueous solution to major components in the dried residue. Although additional characterization of the dried herbicide-amine residue is needed, we anticipate that, depending on the melting point, it may behave analogously to a protic ionic liquid (PIL) or a solid (cocrystal/salt). Both dicamba-DGA and dicamba-BAPMA have been found to be liquids at room temperature, 57 but to our knowledge the phases of the other dicamba-amine residues included in this study have not been reported. However, it should be noted that the characterization of PILs (including dicamba-DGA and dicamba-BAPMA) 57 has been typically carried out in the absence of water, whereas some water likely remains in our herbicide-amine residues 58 (as well as likely dried herbicideamine residues in the field) and may affect molecular speciation and volatilization by altering intermolecular interactions. 59,60 Molecules in PILs will be distributed between their neutral and ionic species according to the extent of proton transfer (i.e., HA + B ↔ A -+ HB + ). 50,61,62 Although the extent of proton transfer in PILs correlates with the difference in the aqueous-phase pKa values of the acid and base (∆𝑝𝐾 𝑎 ), 50,62,63 deviations from aqueous conditions [64][65][66] only permit ∆𝑝𝐾 𝑎 from aqueous species to be used semi-quantitatively to determine the extent of proton transfer in PILs. 67 Complete proton transfer in PILs has been proposed to require ∆𝑝𝐾 𝑎 > 10. 50 Because our herbicide-amine salts have known ∆𝑝𝐾 𝑎 values ranging from 7.7-8.8 (SI Table S2), proton transfer between the herbicide and amine is likely incomplete, resulting in the presence of neutral species with a vapor pressure closer to that of their pure phase. 50,61,62 Volatilization of amines may then occur due to their high vapor pressures (SI Table S1). Although some PILs have been measured in the gas phase as ionic pairs or aggregates 68,69 (particularly for combinations of strong acids and bases with ∆𝑝𝐾 𝑎 ≥ 17.6), 68 we believe this is unlikely to dominate in our herbicide-amine salts due to the relatively low ∆𝑝𝐾 𝑎 values and our observation that amine loss often exceeds herbicide loss (Figure 123).
Upon formation of other PILs, the more volatile component is observed to volatilize alongside the solvent, 58,70 similar to what we observed in our results with DMA and IPA (Figure 1A,B). We hypothesize that DMA loss as the aqueous solution dried (Figure 1A) may be attributed to a fraction of DMA remaining unprotonated that volatilized due to DMA's high vapor pressure (SI Table S1). After the neutral DMA had volatilized (after 16 h), we hypothesize that the remaining DMA was ionic, which prevented further significant volatilization (Figure 1A). We expect that IPA losses during the drying period may result from a similar process (Figure 1B).
At higher temperatures, proton transfer in PILs shifts towards neutral molecules 50,51 that also undergo increased volatilization. At higher temperatures, all four amines volatilize from dicamba-amine salt residues (Figure 2), likely due to the combined temperature-dependent processes of proton transfer and volatilization. At longer times at these elevated temperatures, amine volatilization slows in some cases (e.g., Figure 3A), which we hypothesize may occur when volatile neutral amines are depleted. Similarly, reduced abundance of neutral amines may decrease amine volatilization when the residues are initially prepared at lower amine/dicamba ratios (Figure 3E).
For herbicide-amine residues with higher melting points, a solid (cocrystal/salt) may serve as a better model than a PIL. Similar to PILs, ∆𝑝𝐾 𝑎 values of a salt/cocrystal with a defined crystal structure has been correlated with extent of proton transfer. 71 Unlike PILs, the minimum ∆𝑝𝐾 𝑎 for complete proton transfer to generate an ionic salt is much lower (∆𝑝𝐾 𝑎 >4), 63 indicating that the dried residues prepared in our study, if solid, may behave as ionic salts rather than cocrystals (SI Table S2). Higher temperatures may promote volatilization from salts/cocrystals by multiple factors, including promoting volatilization of molecules or melting the salt/cocrystal to form a PIL if the melting temperature is exceeded.
For organic salts, increased temperature also has been found to shift the proton closer to the acidic molecule [52][53][54] and, in some cases, increase the length of the bond between the acid and base 52,72,73 which both may weaken interactions that may lead to increased volatilization observed at higher temperatures (Figure 2).
While herbicide-amine salt formulations are typically formulated as equimolar solutions, interactions may also involve multiple acidic or basic molecules. For example, some carboxylic acids form hydrogen-bonded dimeric or oligomeric combinations with other carboxylic acids when mixed with amines or other nitrogen containing molecules (i.e., HAx + B ↔ Hx-1Ax -+HB + ), 74,75 which may contribute to lower herbicide loss relative to amines even at elevated temperatures (Figure 1,2). In addition, PILs of dicamba-BAPMA prepared at 2/1 and 3/1 molar ratios of dicamba/BAPMA demonstrated that individual BAPMA molecules were capable of deprotonating multiple dicamba molecules. 57 This unique ability may explain why BAPMA loss at higher temperatures plateaued when the remaining BAPMA reached a ~2/1 dicamba/BAPMA molar ratio (Figure 2D).
To determine the magnitude of herbicide-amine salt applications on four major crops (i.e., soybean, corn, cotton, wheat) that serve as a potential source of amine pollutants into the atmosphere, we integrated Region State survey data from 2017 and 2018 from the U.S.
Department of Agriculture on type of herbicide formulation used 4 with U.S. herbicide use estimates in 2017 3 (SI Section S10). From overall application rate of amines, we provide an estimated upper limit of atmospheric input of amines from this pathway, which may motivate subsequent measurements of amine fluxes in the field.
In the U.S., amine salt formulations comprise slightly more than half of combined glyphosate, 2,4-D, and dicamba applications by mass (Figure 4A). Approximately half of glyphosate is applied with an amine, typically IPA and -to a much lesser extent -DMA (49% and 3% of total applications, respectively). In contrast, almost 90% of dicamba is applied as an amine salt, which is distributed among DMA, DGA, and BAPMA (24%, 45%, and 20% respectively) likely because postemergent application on GM dicamba-tolerant crops require DGA and BAPMA salt formulations specifically. 2 Amine salt formulations for 2,4-D are applied less extensively (44%) due to the prevalence of 2,4-D-ester formulations. 76 Variations in the inclusion of amines alongside specific herbicides affect the application rates of amines on different crops, which may have important regional impacts on amine input into the environment. Although total herbicide applications on soybeans and corn exceed application on cotton by ~6-and 4-fold respectively (Figure 4B), the potential amine input (as nitrogen, N) applied to those crops only exceed those from cotton by 4-and 3-fold (Figure 4C). This difference results largely from the fact that approximately 74% of herbicide applications to cotton contained an amine (typically amine formulations of dicamba and glyphosate), whereas only 48-58% of herbicide applications on the other major crops included an amine counterion (Figure 4B).
Overall, the use of amines in herbicide formulations in the U.S. is estimated to input about 4.0 Gg N into the environment (Figure 4D). The majority is IPA (3.2 Gg N) due to its use with glyphosate, which is applied at a 6-fold higher rate than 2,4-D and dicamba combined (Figure 4A). DMA is second (572 Mg N) due to its inclusion across salt formulations of all three herbicides. Notably, these two amines have the highest vapor pressures (SI Table S1)
and demonstrated the greatest losses in our experiments at ambient temperatures, suggesting their atmospheric flux may be particularly important. Because DGA and BAPMA are exclusively applied with dicamba, which was used at the lowest rate in 2017, their potential inputs are smaller (135 and 162 Mg N, respectively). Though DGA is used more often than BAPMA, BAPMA salt formulations could potentially contribute more nitrogen to the atmosphere due to the three amine groups present per molecule (Figure 4D).
Herein, we consistently measured amine molar losses that are greater than or equal to the herbicide molar losses from the same salts, suggesting that herbicide-amine salt applications are a source of amines to the atmosphere. The measurable losses of DMA and IPA during drying at room temperature indicate that amine losses can occur from these herbicideamine salts at ambient temperatures. Because the amine losses often exceed herbicide losses, we expect that amines are likely to be volatilizing from applications of herbicide-amine salts even when off-target movement of the herbicide is not observed. Though our study is the first to raise amine volatilization as a possible factor contributing to subsequent herbicide volatilization, it is also apparent that extensive loss of some amines (i.e., BAPMA, Figure 2D)
can occur with little impact on dicamba loss, possibly because the amine molecules remaining interact with multiple dicamba molecules to suppress volatilization. 24 We proposed that amine volatilization from herbicide-amine salts was enabled by incomplete proton transfer between the herbicide and the amine, resulting in a fraction of the molecules present as the more volatile neutral species. Recently, some permanently charged organic cations have been included in commercial herbicide formulations (i.e., 2,4-Dcholine) 77 and researched for further use alongside 2,4-D and dicamba. 57,78,79 These permanently charged cations may have reduced volatilization compared to the basic amines considered in this study. 62 Further characterization of herbicide-amine salts (i.e., extent of proton transfer, intermolecular interactions) and of vapor composition (i.e., neutral molecules, aggregates, ion pairs) from herbicide-amine salts will help improve the understanding of volatilization from these multicomponent systems.
Though this study provides the first evidence that herbicide-amine salts may be an important source of atmospheric amines, the magnitude of this source would be better constrained by further investigations, particularly involving field measurements. Laboratory measurements from representative soil and leaf surfaces may also help to constrain the magnitude of amine volatilization relative to other co-occurring processes (e.g., biological uptake and transformation), as well as identify key environmental parameters. Additional complexities in chemical compositions of the salts may also influence amine volatilization. For example, herbicide formulations can be applied in combination, which has been shown to alter herbicide volatilization 20,24,26 and may also impact amine volatilization.
We found that an estimated 4.0 Gg N of amines were applied as glyphosate, dicamba, and 2,4-D salt formulations during 2017 in the U.S. alone. While this number is relatively small compared to the estimated 285 Gg N global flux of methylamines, 29 regional effects may be significant. For example, herbicide-amine salt applications may provide an overlooked source for atmospheric amines in rural environments, where amine-assisted particle nucleation (i.e., new particle formation) is likely to occur at low concentrations of pre-existing particles. 29 Amine volatilization from herbicide-amine salt formulations may also help rectify differences in predictive models that estimate lower concentrations of atmospheric amines than what has been measured particularly in rural environments. 80 Furthermore, our estimate only considers U.S. herbicide applications, which accounts for ~12% of the herbicide use worldwide. 1 Since global use of glyphosate and 2,4-D exceeds U.S. use by more than 5-and 7-fold respectively, amine formulations of these herbicides in particular may be prevalent worldwide. Though use rates are unavailable, both glyphosate and 2,4-D-amine salt formulations are approved for use in countries such as China and Brazil [81][82][83][84] that rank with the U.S. as the top herbicide applicators globally. 1 Accounting for formulation type in addition to total use 1 would help determine the global contribution of herbicide-amine salt formulations to amine pollution in the atmosphere.