Introduction
The widespread use of agricultural pesticides has resulted in the
observation of many such compounds in soil and water samples. While the
ecological implications of these observations have been the focus of much
research, little work has focused on their atmospheric chemistry, despite
the fact that many of these pesticides are transported through the
atmosphere (Bedos et al., 2006; Coscollà et al., 2008; LeNoir et al.,
1999; Majewski et al., 2014; Peck A.M, 2005; Rice et al., 2002; Sauret et
al., 2000; Tabor, 1965; White et al., 2006). The chemical fate of compounds
in the atmosphere – including oxidation, gas–particle partitioning, and
surface deposition – ultimately controls their chemical identities and
concentrations, and thus their impact on ecosystems and human health. Global
atmospheric concentrations of pesticides are typically considered to be
small, yet the local concentrations near point sources can be large enough
to result in pesticide drift to neighboring farms, negative impacts on
pollinator populations, and substantial occupational exposure of
agricultural workers (Choi et al., 2013; Harris et al., 2010).
Trifluralin, atrazine, permethrin, and metolachlor are among the top 20
most frequently used pesticides in the last decade (Todd and Suter II, 2016). These compounds are thought to be much less toxic, less
carcinogenic, and less likely to bio-accumulate than the organochlorine
pesticides deemed persistent organic pollutants (POPs) (DeWit et al.,
2004; Clausen et al., 1974; Loomis et al., 2015; Oehme and Ottar, 1984; Shen
et al., 2005). However, the environmental and health impacts of current-use
pesticides and chemical parameters controlling their atmospheric fate are
not well understood. These pesticides have been studied with regard to
deposition and volatilization (Glotfelty et al., 1989; Rice et al.,
2002), but the investigation of surface–atmosphere fluxes is limited by
currently available instrumentation. In order to fully understand transport,
concentrations, and chemical behavior in the atmosphere, in situ measurements of
these pesticides need to be fast (< 1 h), sensitive, and selective
in both the gas and particle phases (Farmer and Jimenez, 2010).
Volatilization of these current-use pesticides from agricultural soils to
the atmosphere removes up to 27 % of the applied pesticide (26.5,
12.4, and 7.5 % for trifluralin, metolachlor, and atrazine,
respectively) (Rice et al., 2002). The resulting atmospheric
concentrations vary, and these current-use pesticides have been detected
hundreds of meters to kilometers away from application sites in both gas and
particle phases (Coscollà et al., 2010, 2008;
LeNoir et al., 1999). Atrazine, trifluralin, and metolachlor have been
observed near application areas in concentrations ranging from < 1 ng m-3 to as high as 61 µg m-3, and in urban and remote
locations that are far from sources with concentrations < 2 ng m-3 (Foreman et al., 2000; Majewski et al., 2014; Peck A.M, 2005;
Bedos et al., 2006; Coscollà et al., 2010).
For detection of pesticides in the atmosphere, air samples are typically
collected on solid-phase micro-extraction (SPME) fibers or other adsorbent
materials with sampling times of 2 h–1 week (Bedos et al., 2006;
Glotfelty et al., 1989; LeNoir et al., 1999; Majewski et al., 2014; Peck A.M, 2005). These solid adsorbents are analyzed by offline techniques,
typically gas chromatography coupled to mass spectrometry (GC–MS) or
electron capture detection (GC–ECD) (Bedos et al., 2006; Coscollà et
al., 2010; Foreman et al., 2000; Glotfelty et al., 1989; LeNoir et al.,
1999; Majewski et al., 2014; Peck A.M, 2005; Rice et al., 2002). These
measurement approaches are adequate for offline quantitation of airborne
pesticides, with detection limits for trifluralin ranging from 1.3 pg m-3
(GC–MS) to 0.4 µg m-3 (GC–ECD) with sample collection times of
24 and 2 h, respectively (Peck A.M, 2005; Bedos et al., 2006).
These techniques have proven successful in quantifying atmospheric pesticide
concentrations, although offline analysis introduces steps that can alter a
compound's structure and reduce sampling efficiency (Coscollà et al.,
2010; LeNoir et al., 1999; Peck A.M, 2005; Rice et al., 2002), and it is
inadequate for rapid ambient measurement. Rapid measurements are necessary
for (i) observing pesticide drift in real time to understand meteorological
effects; (ii) directly measuring volatilization and surface–atmosphere
fluxes by eddy covariance or other micrometeorological approaches; (iii) determining whether agricultural workers are exposed to low concentrations
over a long period of time or high concentrations over a short period of
time, and thus for identifying activities that can be targeted to reduce
exposure; and (iv) laboratory smog chamber or flow reactor measurements in
which oxidation chemistry is typically observed on timescales of minutes.
Studies of oxidation chemistry require such rapid measurement as atmospheric
lifetimes of pesticides can be short due to reaction with OH radicals.
Atkinson et al. (1999) reported rapid trifluralin reaction with the OH radical
(> 1×10-10 cm3 molecules-1 s-1) and
photolysis rates on the order of minutes, while the half-life for atrazine
plus OH is 2.6 h for a global average radical concentration of 1×106 molecules cm-3. As an
example of the need for rapid detection for indoor exposure estimates, Vesin
et al. showed that pesticides must be measured rapidly (< 1 h) due
to high emission variation from electronic vaporizers (Vesin et al., 2013).
Here, we investigate the use of chemical ionization mass spectrometry for
real-time in situ atmospheric pesticide measurement.
Chemical ionization mass spectrometry (CIMS) has been previously explored
for pesticide detection. Dougherty et al. detected aromatic chlorinated
pesticides, including DDT and DDE (dichlorodiphenyltrichloroethane and
dichlorodiphenyldichloroethylene, respectively), using isobutane reagent
ions in positive and negative mode (Dougherty et al., 1975).
Tannenbaum employed chloride as a reagent ion to detect the chlorinated
pesticide Aldrin (Tannenbaum et al., 1975). More recently, Vesin
et al. showed that high-sensitivity proton transfer reaction mass
spectrometers can be calibrated to measure indoor concentrations in the
0.5–600 ppbv range of four pyrethroid pesticides with limits of
detection (LODs) of 50 pptv with 1 s time resolution
(Vesin et al., 2012). However, these studies are limited
by the use of a quadrupole mass spectrometer, which only has unit mass
resolution (m/Δm ∼ 1000) and thus limits the
selectivity of the measurement. As pesticides are typically quite large with
molecular weights of 200–500 Da, a separation step before analysis of
ambient air is often required to eliminate other isobaric molecules that may
act as interferences. CIMS is increasingly used to measure trace gas species
in the atmosphere because of high sensitivities, resolution, and
selectivities of the different reagent ions employed (Bertram et al.,
2011; Crounse et al., 2006; Lee et al., 2014; Lopez-Hilfiker et al., 2014;
Nowak et al., 2007; Brophy and Farmer, 2015; Wentzell et al., 2013). Chemical ionization
can be achieved by positive or negative ions, including hydronium, acetate,
iodide, and nitrate. CIMS coupled with high-resolution time-of-flight mass
spectrometry (ToF-MS) is a viable detection technique for atmospheric
pesticides because of its fast time resolution (1–10 Hz), field-portable
design, high mass resolution (m/Δm 4000–6000), and mass accuracy
(< 20 ppm). These features result in measurements of numerous
compounds in a complex atmospheric matrix with no need for pre-separation.
The elemental composition of analyte ions can be determined for a broad
range of m/z ratios (typically 0–1000) (Aljawhary et al., 2013; Ehn et
al., 2014; Lee et al., 2014; Brophy, 2015; de Gouw and Warneke, 2007). Iodide is
an obvious target reagent ion for pesticide CIMS, as iodide has been used to
measure oxidized nitrogen and halogenated species – including N2O5,
ClNO2, and ClNO3 – and more recently semi-volatile organic
compounds, particularly organic acids (Huey et al., 1995; Kercher et al.,
2009; Lee et al., 2014). Pesticides often contain one or multiple of these
previously detected functional groups, suggesting that iodide is an
appropriate reagent ion for their detection.
In this paper, we explore the potential of iodide ToF-CIMS to detect and
quantify four current-use, semi-volatile pesticides: atrazine, metolachlor,
permethrin, and trifluralin. We present calibrations using heated injections
into an iodide CIMS and demonstrate that these compounds can be detected at
atmospheric and laboratory relevant concentrations with fast (seconds–minutes) time resolution.
Experimental method
Chemicals
Standard solutions of trifluralin in acetonitrile (98 ng µL-1 ± 5 %; Sigma Aldrich) and metolachlor in acetonitrile (103 ng µL-1 ± 5 %;
Sigma Aldrich), atrazine in methyl tert-butyl ether (MTBE)
(1032 µg mL-1 ± 12; SupelCo, Bellefonte, PA, USA), and permethrin in
acetone (999 ± 26 µg mL-1; SPEX CertiPrep, Metuchen, NJ, USA) were
used in this study. Solvent choice was dictated by commercial availability
of standards.
High-resolution time-of-flight chemical ionization mass spectrometer
(ToF-CIMS)
The ToF-CIMS (Tofwerk AG, Switzerland, and Aerodyne Research, Inc.,
Billerica, MA, USA) and iodide (I- ) ionization scheme used herein are
described elsewhere (Lee et al., 2014; Brophy, 2015). Briefly, our iodide
ToF-CIMS has five primary components: the ion molecule reactor (IMR), two
radio frequency (RF)-only quadrupoles, an ion lens focusing region, and a time-of-flight mass
analyzer (m/Δm ∼ 4000) with a pair of microchannel
plate detectors. Sample air is continuously drawn into the IMR at 1.9 sLpm
(standard Liter per minute), where the sample interacts with iodide reagent
ions. I- is generated by flowing ultra-high-purity (UHP) N2
(99.999 %, AirGas) over a CH3I permeation device; the N2 carries
gaseous CH3I into a 210Po source to produce I- reagent ions
(Slusher et al., 2004). Iodide is typically thought to
ionize neutral species (M) through a ligand exchange reaction with an
iodide–water adduct (Reaction R1) (Slusher et al., 2004).
[I⚫H2O]-+M→[I⚫M]-+H2O
However, deprotonated species have also been observed in ambient
measurements (Brophy, 2015), though it remains unclear
whether these species are deprotonated in the initial ionization step or
declustered during transmission to the ToF detector.
Heated pesticide injections and calibrations
As most pesticide compounds are commercially available as liquids or solids,
calibration of these compounds in the iodide ToF-CIMS necessitates
quantitative conversion of solutions to the gas phase. We developed a heated
injection system (Fig. S1 in the Supplement) based on work by Lee et al. (2014). We
placed a new 2 µm pore Teflon filter (Pall Life Sciences) for each
experiment onto an in-line filter holder (Advantec MFS, Dublin, CA, USA)
connected by 13 cm of unreactive 1/4 in. (OD) PEEK tubing to
the IMR. Pesticide solutions were then injected as liquid samples onto the
filter. The filter holder was connected to a four-way stainless-steel union tee
(Swagelok). A septum was placed in the second port directly opposite the
filter, and a third port, upstream of the filter holder, was connected to a
dry UHP zero-air flow controlled by two 2000 sccm (standard cubic centimeter
per minute) mass flow controllers (MKS, Andover, MA, USA). The air flow was
directed through a 1/2 in. stainless-steel tube packed with cleaned
steel wool and heated to 200 ∘C by a resistive heating wire on
the outside of the tube with a PID temperature controller (Omega, Stamford,
CT, USA) attached to a thermocouple located between the wire and tube near the
exit of the tube. The heated zero air passed over the filter to volatilize
the liquid sample to the gas phase before entering the iodide ToF-CIMS. The
fourth port was open to the room to allow the zero air to overflow the
system, exhausting to ambient pressure. The zero-air flow was always greater
than the inflow of the iodide ToF-CIMS in order to maintain constant
pressure in the IMR and a known flow rate over the pesticide-containing
filter.
For each set of experiments, we injected known volumes (1–6 µL) of
solvents (blanks) and commercial standard solutions through the septum onto
the filter with a 10 µL syringe (Hamilton, Reno, NV, USA). Following each
injection, mass spectra time series (Fig. 1) were allowed to return to the
same signals as zero air before the next injection was made as can be seen
in Fig. 1b. Following an injection, pesticide-related peaks rapidly
increased and then decayed in 30–120 min. Iodide ToF-CIMS data were
collected at 1 Hz.
Sample data of injections of trifluralin (a, 1 µL) and
sequential metolachlor (b, 2 and 4 µL). The observed mass
spectrometer signals are shown as 1s data time series. Trifluralin is
observed at m/z 462.01, indicative of clustering between the iodide reagent
ion, while metolachlor is detected at m/z 337.98, a cluster of a molecular
fragment and iodide.
Data analysis
In order to identify peaks in the mass spectrum that changed from blank
injections during the pesticide injection experiments, we calculated the
signal-to-noise ratio (S/N) for every nominal m/z peak in the mass spectrum.
Nominal masses with S/N>3 during the injection period were
identified as potential signals from the pesticide samples, and the
high-resolution mass spectra were fit at those m/z ratios to identify the
elemental composition of each ion contributing to the signal (Tofware 2.4.3,
Fig. S2) (Brophy, 2015). The mass spectral peaks
identified during the injections corresponded to iodide–molecule adducts
([I⚫ M]-, trifluralin and atrazine) or iodide–molecular
fragment adducts ([I⚫ F]-, permethrin and metolachlor). The
isotopic patterns for each peak were used to verify our identification of
elemental compositions based on the natural abundance of isotopes.
We normalize the signal of each pesticide ion by a ratio of the reagent ion
signal (defined as the sum of the signals of I- plus its water adduct
[I⚫ H2O]-) during the background signal to the reagent
signal at each second of the pesticide injection. This normalization
accounts for changes in the reagent ion concentration, and thus ion–molecule
collision rates and overall ionization rates, and allows for comparisons
across different CIMS instruments with different reagent ion count rates and
standardization within a single instrument as the ion source ages. The
normalization assumes that variations in the ionization efficiency from
temperature or pressure changes are adequately captured by the normalization
of the summed reagent ions, although the humidity dependence suggests that
mechanisms may not be simple, and the assumption should be tested for field
conditions.
The start of the injection/desorption period is obvious in the data (Fig. 1) with a sharp initial increase in signal above the background (zero-air
signal) count rate; we define the end of the injection/desorption period as
the time at which the mass spectral pesticide signal has returned to within
5 % of the background count rate. Typical desorption periods for
injections were 10 min–1 h for the trifluralin and metolachlor injections
and 1–3 h for the atrazine and permethrin injections. Extended tailing
(0.5–2 h) indicates slow volatilization of the liquid sample to the gas
phase. The background count rate is determined from a 20–40 min average
and standard deviation (σ) of signals detected from UHP zero air, and it is subtracted
from the pesticide mass spectral signals described in the subsequent
analysis. We use the time series of each pesticide-relevant mass
spectrometric peak to develop a calibration curve by assuming that the
total integrated signal at a given m/z ratio observed during the injection
and subsequent desorption is directly proportional to the known mass of
pesticide injected on, and assumed to be completely volatilized from, the
filter. Thus, the signal collected at each 1 Hz data point is taken to
represent the fractional mass of the pesticide standard injection. That is,
if 5 % of a single injection's background-subtracted mass spectral signal
is observed in 1 s, that signal corresponds to 5 % of the calibrant
mass injected on the filter. As the flow rate is constant, this fractional
mass can be converted into a mixing ratio (parts per billion by volume,
ppbv, liters of pesticide per 109 liters of air), and each 1 Hz
data point provides an observed signal for a given mixing ratio. Each
injection peak can thus be used to construct a calibration curve and derive
the instrument's sensitivity to the analyte of interest at a given
high-resolution m/z ratio. Multiple injections allow for the calculation of an
average sensitivity for each analyte and were used to determine average
limits of detection (LODs, S/N=3) from the standard deviation
of the average background count rate of the blanks. This calibration
approach assumes that all of the standard solution deposited on the filter
is volatilized and that the instrument response is linear over the
concentration range of each injection/desorption period.
We note that this calculation differs from the calibration approach
described in Lee et al., in which the total summed signal for an injection
is divided by the number of molecules injected and then converted to a
mixing ratio using the instrument flow rate at 1 s (Lee
et al., 2014). While that calculation is specific for calibrating collected
aerosol samples that are subsequently desorbed from a filter surface, as
described by Lee et al. for the Filter Inlet for Gas and AEROsols (FIGAERO)
(Lopez-Hilfiker et al., 2014), it does not
capture the variation in concentration that occurs on the fast (seconds)
timescale of gas-phase variation and measurement, and it would result in an
increasing sensitivity with decreased mass spectral averaging times. The
calibration approach described herein is thus specific for gas-phase calibrations
using an injection/desorption technique.
Calibration technique comparison
To validate this approach of gas-phase calibrations by solution injection
and thermal desorption, we compared well-characterized HR-ToF-CIMS
calibrations of formic acid from a home-built permeation tube with
injections of a formic acid standard solution (2–4 µL of 10 ng µL-1 of formic acid in acetone) on the injection/desorption
calibration setup described herein. However, as the high-performance liquid chromatography (HPLC)-grade solvents
contain trace quantities of formic acid, identical volume injections of
acetone solvent were necessary to identify and subtract formic signal of the
solvent blank from each standard solution injection/desorption period.
Relative humidity tests
Due to the ligand-switch mechanism described above, analyte detection by
iodide ToF-CIMS is expected to vary with ambient relative humidity (RH)
(Kercher et al., 2009; Lee et al., 2014). As field measurements are a
desired outcome in the development of a real-time pesticide detector, we
investigated the sensitivity of trifluralin and metolachlor with replicate
injections of trifluralin (2.4 µL of 98 ng µL-1 solution) and
metolachlor (3 µL of 103 ng µL-1 solution) over an RH range of
0–80 % (Fig. S5). The RH of zero air was controlled by bubbling zero
air (0–2000 sccm) through water (HPLC grade, Sigma Aldrich) and diluting
with dry zero air (2000–0 sccm) prior to entering the heated tube and
injection region of the calibration apparatus described above. The RH of
zero air entering the heated tube was detected by a RH probe/transmitter
(Omega HX71, Stamford, CT, USA).
Results and discussion
Comparison of calibration techniques
Calibration comparison between the conventional formic acid
permeation tube calibration method (black open circles) and formic acid
injections using the method described herein. Error in the formic acid
concentration is represented by the horizontal bars, while error on the y axis is the standard deviation of the signal. The average calibration
from
the injection method is presented as the dashed red line, and the shaded red
area is the standard deviation of the injection sensitivities.
The results of the formic acid permeation tube and injection calibrations
are presented in Fig. 2. The injection method produces an average
sensitivity to formic acid of 3.8 ± 0.4 normalized counts s-1 pptv-1, while the permeation tube calibration produces a
sensitivity of 3.8 ± 0.2 normalized counts s-1 pptv-1.
The injection method signal is more variable than the permeation tube,
likely due to the large formic acid background in acetone; uncertainties in
the syringe volume; and larger, more variable background formic acid in the
zero air due to thermal decomposition of species to formic acid in the
heated stainless-steel tube. Error in the concentration of formic acid from
the permeation tube arise from uncertainties in the mass loss rate in the
home-built permeation oven. The formic acid injections are much shorter
(< 1 min) than the pesticide injections, indicative of its higher
vapor pressure than the pesticides, but the data analysis is identical. The
injection calibration method for CIMS does prove to be sufficient for
calibration of lower-volatility compounds and can be used for the pesticides
presented herein, for which calibration by permeation tubes is impossible.
Pesticide calibrations
The calibration technique described above is a dynamic gas-phase calibration
approach for low-volatility compounds that are otherwise challenging to
quantitatively convert to the gas phase. While target analytes must be
soluble in non-reactive solvents that do not substantially interfere with
instrument background or reagent ion concentrations, this versatile
technique is a viable alternative to permeation tubes, which require 5–15 mL
of pure liquid analyte and can take weeks to months for equilibration and
mass loss analysis. Thus, this approach enables calibration of semi-volatile
and intermediate-volatility compounds that diffuse through typical Teflon
permeation tubes too slowly, or not at all, for detectable mixing ratios and
for measurable mass loss (an essential component in determining permeation
rates).
Characteristics of the pesticides studied and figures of merit
using the iodide ToF-CIMS.
Trifluralin
Atrazine
Metolachlor
Permethrin
Pesticide class
Dinitroaniline
Triazine
Chloroacetanilide
Pyrethroid
Use
Herbicide
Herbicide
Herbicide
Herbicide & insecticide
Vapor pressure (bar)
6×10-8
4×10-10
2×10-8
2×10-11
Chemical formula
C13H16F3N3O4
C8H14ClN5
C15H22ClNO2
C21H20Cl2O3
Ion detected
I-⚫ C13H16F3N3O4
I-⚫ C8H14ClN5
I-⚫ C11H14ClNO (fragment)
I-⚫ C8H10Cl2O2 (fragment)
m/z of ion detected
462.01
341.99
337.98
334.91
Standard concentration
98 ± 4.9 ng µL-1
1032 ± 12 ng µL-1
103 ± 5.2 ng µL-1
999 ± 26 ng µL-1
Solvent
Acetonitrile
Methyl tert-butyl ether
Acetonitrile
Acetone
Boiling point in solutiona (∘C)
87
53
81
55
Injection volumes for calibration (µL)
1,2,3,4
1.4, 2.8, 4.5, 6
1,2,4,6
0.9,1.8,4
Sensitivity (ncps ppbv-1)
180 ± 40b
75 ± 19
38 ± 6
100 ± 40
LOD (pptv)c
50 ± 30
120 ± 20
110 ± 20
150 ± 80
a From manufacturer's data.
b Error reported as standard error of the average sensitivities.
c pptv= parts per trillion by volume
Calibration curves of pesticide solutions on the iodide ToF-CIMS,
metolachlor (a), trifluralin (b), atrazine (c), and permethrin (d). Red
lines represent the signals from single injections as a function of the
calculated gas-phase mixing ratio. The average sensitivity (black line) is
derived from each set of calibration curves. The error in sensitivity is
calculated as the standard error of the average sensitivity for each
pesticide.
The iodide ToF-CIMS detected the four pesticides in the gas phase with
sufficient sensitivity for laboratory experiments and certain field settings
(Table 1, Fig. 3). Figure 3 shows the calibration curves (red) generated
by the single injections per the calculations described above, with the
average calibration curve (black) of the four pesticides studied. In the UHP
zero-air carrier gas, the average background count rates for the analytes in
synthetic air are very low, between 1 and 7 ncps for the four m/z ratios. LODs
for gas-phase atrazine, trifluralin, metolachlor, and permethrin are
120 ± 20, 50 ± 30, 110 ± 20, and 150 ± 80 pptv,
which correspond to concentrations of 0.56, 0.37, 0.67, 1.1 µg m-3, respectively, and are reported in Table 1. These concentrations
are potentially useful for particle-phase measurements by the iodide
ToF-CIMS, where gases are captured on a thermal denuder and particles are
subsequently volatilized using either a heated inlet or filter system
(Aljawhary et al., 2013; Lopez-Hilfiker et al.,
2014). Trifluralin has been detected in a number of field studies with
average ambient gas-phase concentrations ranging from 0.228 to 1.93 ng m-3
and detected concentrations as low as 0.0013 ng m-3 (Coscollà et
al., 2010; Peck A.M, 2005). Trifluralin volatilization measured the day of
application was as high as 61 µg m-3, decreasing by an order of
magnitude the next day (Bedos et al., 2006).
Metolachlor average gas-phase concentrations are 0.37–12.74 ng m-3
with the lowest reported concentration of 0.0059 ng m-3 (Peck A.M,
2005; Sadiki and Poissant, 2008). Similarly, atrazine average gas-phase
concentrations covered a similar order of magnitudes (0.0018–8 ng m-3)
(Yao et al., 2007; Peck A.M, 2005). The LODs of the iodide
ToF-CIMS suggest that this instrument is appropriate for real-time ambient
measurements made near agricultural targets several days after application,
but not in remote locations.
While the resolution for the ToF-CIMS might be limited at larger m/z ratios
during ambient measurements, these pesticides provide one particular
advantage for detection by mass spectrometry: the presence of halogen and
other heteroatoms such as nitrogen, sulfur, and phosphorus results in
detected ions that have distinct isotopic signatures. The fitting procedures
used in the Tofware software package allow for confirmation of peak identity
not only by the exact mass of the peak fit but also by the fit of
isotope-containing peaks. Such fitting has allowed for measurement of trace
compounds in complex environmental and laboratory samples (e.g.,
Lopez-Hilfiker et al., 2014; Lee et al., 2014; Aljawhary et al., 2013; Ehn
et al., 2014). For example, Fig. S6 shows potential
interferences at the expected m/z ratios of the four target pesticides based on
previous field campaigns relative to the signal for 1 ppbv of each
pesticide. Potential interferences for metolachlor are minor; potential
interferences for trifluralin, atrazine, and permethrin are more substantial,
but as none of the interfering peaks hold identical halogens, the pesticide
isotopes at higher masses can be used to validate the observation (e.g.,
Fig. S2). However, we acknowledge that mass resolution will be a limiting
factor for field measurements of these three pesticides that are far from
agrochemical sources.
CIMS has been used in atmospheric chamber experiments to explore oxidation
reactions and mechanisms with starting precursor concentrations between
1 and 100 ppbv (Paulot et al., 2009; Ehn et al., 2014; Wyche et al.,
2007). The iodide ToF-CIMS is thus more than suitable for chamber and
laboratory experiments of pesticide oxidation chemistry and kinetics. Due to
known relative humidity effects on iodide CIMS sensitivities, trifluralin,
and metolachlor were measured at multiple relative humidities (Fig. S5).
The observed pesticide sensitivities decreased by 70 and 59 % at
0 and 80 % RH for trifluralin and metolachlor, respectively. Only small
changes (8 %) were observed in the background signal and noise (5–30 %)
between 0 and 80 % RH. These changes in sensitivity caused the LOD to
increase from 108 (50) pptv at 0 % RH to 421 (110) pptv at
80 % RH for metolachlor (trifluralin). The decrease in instrument
sensitivity with increased relative humidity suggests ionization occurs
through a clustering reaction with bare iodide reagent ions, rather than a
ligand exchange reaction. Therefore, relative humidity effects on iodide
CIMS sensitivity necessitate inclusion of RH measurements with ambient
field measurements of pesticides.
Comparison of LODsa (µg m-3) between iodide ToF-CIMS and previous
pesticide measurements.
Reference
Instrument
Approach
Phase
LOD (µg m-3)
Collection time
Atrazine
Trifluralin
Metolachlor
Permethrin
LeNoir et al. (1999)
GC–CIMSb
Offline
Gas
n/ac
1.6×10-6
n/a
n/a
> 8 h
Foreman et al. (2000)
GC–MS-SIMd
Offline
Gas and particle
6×10-6
1×10-6
1.2×10-5
2.9×10-5
4 h; 5 min h-1
Peck et al. (2005)
GC–MS-SIM
Offline
Gas
9.8×10-6
1.3×10-6
5.9×10-6
n/a
24 h
Bedos et al. (2006)
GC–MS-SIM GC–ECD
Offline
Gas
n/a
0.004; 0.4
n/a
n/a
2–10 h
This work
Iodide ToF-CIMS
in situ
Gas
0.56
0.37
0.67
1.1
(No collection); 1 s sampling
a Each work reported different method of calculation for LOD.
b Methane chemical ionization mass spectrometry.
c Pesticide was not studied.
d Selected ion monitoring.
Table 2 compares the iodide ToF-CIMS to previous measurement techniques of
the four pesticides. Our work, to the best of our knowledge, is the first
online detection and quantification of atrazine, trifluralin, metolachlor,
and permethrin. Unlike previous work shown in Table 2, the iodide ToF-CIMS
detected the pesticides in situ with no collection, extraction, and separation on a
gas or liquid chromatograph, enabling rapid (1 Hz) detection. The LODs we
report are comparable to those reported by Vesin et al. for online, in situ measurement
of transfluthrin, a pyrethroid compound that is structurally similar to
permethrin (Vesin et al., 2013). However, LODs calculated
in this work are larger than other techniques and could be improved by
longer averaging of sampling time (1–5 min). No significant relationship
between decay time of the injection and vapor pressure was found. While
iodide ToF-CIMS has been typically used for the measurement and
quantification of semi-volatile CxHyOz or small oxidized
halogen compounds (Cl2, BrO) (Liao et al., 2014; Huey et al., 1995),
atrazine (C8H14ClN5) is a triazine-derived compound with
multiple amine groups and a chloride. This suggests that iodide ToF-CIMS
might be appropriate for detecting other triazine or organic halide
compounds (Liao et al., 2014).
Fragmented pesticides
While trifluralin and atrazine were detected as quasi-molecular ions with
the parent molecule clustered with iodide reagent ions, metolachlor and
permethrin were detected as iodide adducts of molecular fragments.
Fragmentation is a well-known phenomenon in CIMS but is not typically
thought to dominate mass spectra in atmospheric measurements
(Lee et al., 2014). However, the intact permethrin–iodide adduct
was not detected (m/z 518.19). Instead, the dichloro-allyl-cyclopropyl acid
fragment is detected clustered with iodide ([I⚫ C8H10ClO2]-, m/z 334.91) following fragmentation at the
ester bond. This is consistent with previous experiments of pyrethroid
compounds similar to permethrin using electron impact ionization that showed
fragmentation at the ester bond (Vesin et al., 2013), the
same bond at which the fragmentation occurs in this study. Similarly, the
intact metolachlor molecule was not detected as a quasi-molecular ion
([I⚫ M]-, i.e., as an iodide cluster with the parent molecule
at m/z 410.79) during either room temperature or heated injections, but instead
as a fragment clustered to iodide ([I⚫
C11H14ClNO]-, at m/z 337.98). The observation is consistent with
fragmentation at the bond between nitrogen and the second carbon of the
methoxypropane group, although we did not observe the corresponding smaller
fragment (C4H8O) as either a bare ion or iodide adduct. We note
that none of the four pesticides' fragments or molecular ions were observed
unbound to iodide reagent ions, as is occasionally observed for some
oxidized organic compounds in other iodide ToF-CIMS instruments (Lee et al., 2014;
Brophy, 2015).
Assumptions of injection calibrations
The injection calibration approach makes four assumptions: (1) volatilization of the pesticide solutions does not cause thermal
dissociation or other chemistry of the analyte prior to ionization, (2) the
sensitivity of the instrument to the detected pesticide ions is linear
across the mixing ratio range created during each injection, (3) complete
volatilization of the pesticide after injection on the filter occurs within
the integration time, and (4) negligible analyte loss to connection between
the filter and the CIMS. We tested the first assumption of negligible
thermal chemistry during the volatilization step by varying the temperature
of the heated air and testing nitrogen as a carrier gas for the
calibrations. We injected the permethrin solutions at two different
temperatures, 23 ∘C (unheated) and 200 ∘C (heated), and
metolachlor solution at three different temperatures: 23, 100, and
200 ∘C. The mass spectra were identical in the unheated and heated
injections, albeit over substantially longer time frames, with permethrin
requiring 6–9 h to return to baseline in the unheated experiment versus
150 min or less for the heated system. Metolachlor sensitivity decreased
substantially during the injections at 23 and 100 ∘C,
due to the inability of the pesticide to volatilize. Permethrin sensitivity
decreased 70 % during the room temperature injection; therefore injections
at 200 ∘C were pursued. The iodide–molecule fragment adduct was
the sole ion observed by the mass spectrometer at both temperatures for
permethrin and metolachlor, while the iodide–molecule adduct was not
observed. Thus, there is no evidence that the permethrin and metolachlor
fragments are produced during the volatilization step; they are thus likely
generated in the ion–molecule reaction chamber during ionization. Oxidation
of the pesticide standards by O2 in zero air could occur to suppress
observed concentrations and thus sensitivity; we note that the sensitivity
of iodide ToF-CIMS to trifluralin increased 38 % when UHP nitrogen was
used as the carrier gas and no ambient O2 was present in the
calibration system or mass spectrometer. However, we also note that
ion–molecule reactions are altered in the absence of ambient O2 and
thus use UHP zero air for all calibrations described herein
(Brophy and Farmer, 2016).
We test the second assumption, linearity in instrument response, by
examining the sensitivities for different volumes and concentrations for
each pesticide standard. These different volumes and concentrations reach
different mixing ratio ranges: a nonlinear detection response would result
in a systematic shift in observed sensitivities as the mixing ratios reached
larger ranges. However, the sensitivities of trifluralin, metolachlor, and
atrazine are normally distributed around the mean with no observable
systematic bias given the concentration ranges (Fig. S3). Thus, the
assumption of linear instrument response is justified for these three
pesticides in the concentration range used. For example, trifluralin
injection volumes of 2 µL (mixing ratio range of 0–10 ppbv)
provided the same mean sensitivity (180 ± 20 ncps ppbv-1) as larger
injection volumes (e.g., 4 µL injection; mixing ratio ranging from 0 to 60ppbv
gave a mean sensitivity of 160 ± 30 ncps ppbv-1) within
the error. This observation is consistent with calibrations of small acids,
including formic acid, acetic, propionic, and nitric acids, which have
previously shown linear calibrations using identical iodide ToF-CIMS (Lee
et al., 2014; Aljawhary et al., 2013; Kercher et al., 2009). Permethrin
standards are less clear but do not show a consistent trend between
sensitivity and injection volume (Figs. S3, S4). To test the third
assumption of complete volatilization, we first note that the baseline
signal of the detected m/z returned to within 5 % of the pre-injection
value within 30–120 min, suggesting that volatilization was complete
before subsequent injections. Further, replicate injections at multiple
volumes are normally distributed for atrazine, metolachlor, and trifluralin,
suggesting that uncertainties are random, while incomplete and inconsistent
volatilization would likely produce unpredictable error and thus
non-Gaussian distributions. Permethrin has a lower sensitivity for the 4 µL injections than for the 0.8 or 2 µL injections, consistent
with incomplete volatilization at the higher (> 15 ppbv)
concentration. Finally, to mitigate loss of analyte between the filter and
instrument, we use the shortest possible piece of unreactive PEEK tube to
connect the filter to the iodide ToF-CIMS entrance.