Introduction
The photochemical oxidation of volatile organic compounds (VOCs) in the
atmosphere generates a wide array of multifunctional organic molecules which
contribute to the formation of secondary organic aerosol (SOA), hydroxyl
radical sources and sinks, and the cycling and fate of reactive nitrogen.
Determination of the identities of these organics, and their abundance in
the atmosphere, has remained an analytical challenge because of the inherent
complexity of the chemical system, which involves a multitude of precursors
and significantly more oxidation products (Bertram et al., 2009;
Goldstein and Galbally, 2007). Chemical ionization mass spectrometry (CIMS)
has become increasingly utilized for the measurement of these types of
compounds
(Bertram
et al., 2011; Brophy and Farmer, 2015; Fortner et al., 2004; Hearn and
Smith, 2004; Holzinger et al., 2010; Huey et al., 1995; Jokinen et al.,
2012; Jordan et al., 2009; Lee et al., 2014; Lopez-Hilfiker et al., 2014;
Slusher, 2004; Veres et al., 2008, 2010; Yatavelli et al., 2012; You et al.,
2014; Yu and Lee, 2012). Typically, a specific reagent ion is generated
using a radioactive ion source, X-rays, or corona discharge, and then mixed
with ambient air for a fixed time. Ion–molecule reactions then lead to the
formation of product ions which are separated and counted with a mass
spectrometer. Common ion–molecule reaction mechanisms include ligand
switching (adduct formation), reactive electron transfer, or proton
transfer/abstraction. The benefits of CIMS include linearity,
reproducibility, sensitivity with some degree of selectivity, and high time
resolution without sample preparation or handling. General disadvantages of
CIMS include a lack of isomer or isobaric separation and thus structural
information without the coupling of addition separation dimensions, and the
range of potential sensitivities which require calibration with authentic
standards.
Recently, chemical ionization has been coupled to field deployable
time-of-flight mass spectrometers (ToF-MS) such as the Tofwerk AG
high-resolution version, commonly referred to as the HRToF-CIMS
(Aljawhary et al., 2013; Huey et al.,
1995; Jokinen et al., 2012; Junninen et al., 2010; Lee et al., 2014;
Lopez-Hilfiker et al., 2014; Yatavelli et al., 2012). As a result, hundreds
of oxidized organic compounds are now routinely detected in ambient air or
photo-oxidation experiments in the laboratory with a single instrument. A
major limitation of these instruments thus far is that calibration of the
instrument response to many of the detected ions is impossible, as either
the sheer number of calibrations required is unrealistic, or calibration
standards do not exist.
Herein, we present the maximum sensitivity of an HRToF-CIMS using the
collision limit for iodide adduct chemical ionization which is becoming
widely used by the atmospheric chemistry community
(Aljawhary et al., 2013; Huey et al., 1995; Kercher et al.,
2009; Lee et al., 2014). We also present an ion adduct declustering scanning
procedure which experimentally determines the relative binding energies of
the detected ion adducts and therefore their approximate sensitivity. The
combination of declustering scanning to determine effective binding
enthalpies, which can be compared with theoretical estimates from quantum
mechanical calculations, along with the experimentally determined collision
limit provides an approximate calibration for many compounds in the mass
spectrum which would otherwise be impossible to obtain by traditional
methods.
Iodide ToF-CIMS sensitivity to organics
Iodide adduct chemical ionization mass spectrometry has been described in
detail previously (Huey et al., 1995; Kercher et al., 2009; Lee et al.,
2014). As summarized in Eq. (1), for adduct ionization, there are essentially
two components to the instrument sensitivity that will be specific to a
molecule: (i) the rate at which product ions are formed via reagent
ion–molecule reactions over the fixed interaction time, and (ii) the
transmission of the molecular ion to the detector. In Eq. (1), Si is the
sensitivity observed for reaction time t, kf is the product ion
formation rate constant, [I-] is the concentration of the reagent ions
in the ion molecule region (IMR), and Ti is the ion-specific
transmission efficiency, which depends upon the ion mass-to-charge (m/Q),
net electric field strength of the transfer optics (ε), and the
adduct ion binding energy (Bi).
Si=∫0tkfI-dt×Ti(mQ,Bi)=product ion formation×transmission
A neutral molecule that forms a strongly bound cluster with iodide at the
collision limit should be detected with relatively high sensitivity given
that it will survive transmission through the ion optics which inherently
impart energy to the ions via electric fields. In contrast, a molecule might
form an iodide adduct at the collision limit, but be so weakly bound that it
is not detected due to collision-induced dissociation (“declustering”)
during transit through the vacuum chamber. Thus, knowledge of a cluster's
binding energy and the collision-limited formation rate can provide a means
to further constrain the instrument's sensitivity to a broader range of
compounds it detects, even if standards do not exist. Experimental
constraints on binding energies and collision-limited product ion formation
rates are discussed below.
Collision limit determination of the UW-ToF-CIMS
We have calibrated the iodide adduct ToF-CIMS to many organic and inorganic
molecules including hydroxy-hydroperoxides, multifunctional acids, diols,
triols, tetrols, nitrated aromatics and other oxidized organic molecules in
an effort to constrain the instrument response to a variety of different
functionalities (Lee et al., 2014). As expected, given constraints
imposed by ion–molecule collision frequencies, we empirically find that
there is a “maximum sensitivity”, which for iodide–organic clusters in our
instrument is ∼ 19–22 cps pptv-1 (per million cps of
reagent ion) (see, e.g., Lee et al., 2014). As discussed below, this limit is also
consistent with the experimental ion–molecule collision limit of our
instrument.
Huey et al. (1995) first showed that dinitrogen pentoxide (N2O5) reacts with iodide ions at the collision limit (Huey et al., 1995). We
therefore use this reaction to determine the upper-limit sensitivity of our
instrument (to N2O5) given that the number of product ions
detected cannot exceed those produced by the number of iodide–N2O5
collisions occurring within the interaction time of ions and molecules.
We generate isotopically labeled 15N2O5 by reacting excess
15NO2 (Scott-Marin) with ozone leading to the formation of
15NO3 and 15N2O5 during transit down a Teflon
reaction cell held at 207 kPa above ambient pressure by a glass capillary. The output can be
modeled (Bertram et al., 2009) and has been independently verified by
other techniques such as thermal-dissociation laser-induced fluorescence
or cavity ring-down spectroscopy (Brown et al., 2001;
Day et al., 2002). We use the independently calibrated output concentrations
as inputs into the mass spectrometer and monitor the response.
N2O5 reacts with iodide ions at the collision limit (Huey et
al., 1995) but via two channels (Kercher et al., 2009). One channel is the
formation of an ion–molecule adduct between N2O5 and iodide
I(N2O5)-. This cluster may simply be a stable intermediate on the
way to the lowest energy reaction products NO3- and INO2, but it
is detected as a major product under weak electric field settings (weak
declustering) in the ion optics used to transmit ions through the vacuum
chamber to the mass separation region (Kercher, et al., 2009). The other
channel results in NO3-+ INO2, presumably from the
dissociation of the iodide adduct. Its contribution can be enhanced by
increasing the strength of the electric fields in the atmospheric pressure
interface (APi) of the mass spectrometer (Kercher et al., 2009). In the
work of Huey et al. (1995) only NO3- is observed due presumably to a
combination of low pressure in the ion–molecule reaction drift tube, where
the iodide–N2O5 collision complex might not be stabilized, and
there are strong electric fields in the vacuum chamber. Therefore, to track the
formation of product ions from the reaction of I- with N2O5,
we add the product ion signals from the two detection channels
(NO3- and IN2O5-). An example time series of this
type of experiment is shown in Fig. 1.
An example of collision-limited sensitivity determination of the
ToF-CIMS. Top: the reagent ions and sum of total ions during the addition of
high concentrations of 15N2O5 to the inlet of the ToF-CIMS.
By calibrating the output of our N2O5 source independently (in
this case by the UC, Berkeley TD-LiF instrument, Day et al., 2002), we
are able to derive the collision-limited sensitivity of the instrument by
adding the two detection channels as described in the text (Day et
al., 2002; Huey et al., 1995). As total ion current (TIC) remains constant
during the experiment despite the depletion of reagent ions (I-+ IH2O–) the mass transmission efficiency between 63 and 237 Th is
therefore constant.
As the N2O5 product ions are detected at different mass-to-charge
ratios (63 vs. 237 Th), the absolute count rate of the sum of the two ion
signals could be influenced by mass-dependent ion transmission through the
ion optics of the instrument. We therefore measured the mass-dependent
transmission of our instrument by adding large quantities of known compounds
with varying molecular mass to the ionization region (Huey et al.,
1995). This method assumes that the total number of charges (ions) in the
ionization region remains unchanged over short time periods (controlled by
the activity of the 210Po); therefore, any changes to the total
number of ions measured at the detector is due to the varying efficiency
with which ions having different masses are transmitted through the mass
spectrometer. By measuring the relative change in total ions detected as a
function of mass to charge (m/Q), a linear system of equations can be solved
to derive the transmission efficiency as a function of mass to charge. The
transmission efficiency depends on ion optic settings, primarily the two
quadrupole ion guides which act as band-pass filters. The lower mass cutoff
is most important for our sensitivity determination as 15NO3 (63 Th) is near the low end of the mass transmission window. We tune the
transmission function to be as flat as possible by adjusting the radio
frequency, amplitude and axial voltage gradient along the quadrupole ion
guides. As a result, in the mass range of interest (63–237 Th), the
transmission efficiency is approximately constant in our instrument as
evidenced by the ion closure shown in the top panel of Fig. 1 (top panel)
during N2O5 additions.
Dividing the transmission-efficiency-corrected sum of NO3- and
IN2O5- count rates by the N2O5 concentration (pptv)
sampled, we calculate the total sensitivity to N2O5 to be 22–26 cps ppt-1 per million reagent ions. Given that I- and
N2O5 react at the collision limit, and assuming there are no other
product ions (we detect no others with the ToF-CIMS), then this sensitivity
represents the maximum possible sensitivity for compounds with collision
cross sections similar to N2O5. As noted above, this estimate is
also consistent with an empirically determined upper-limit sensitivity for
organic compounds, in that we have yet to measure a sensitivity above this
value. Some of the organic compounds to which we have calibrated that are
near this collision-limited sensitivity include isoprene-derived 2-methyl
tetrols (19 cps ppt-1), dipentaerythritol (22 cps ppt-1), malonic
acid (19 cps ppt-1) and levoglucosan (20 cps ppt-1).
Distribution iodide adduct binding energies
In our instrument, organic molecules are nearly exclusively detected as
molecular clusters with iodide. However, outside of a few of the simplest
carboxylic acids, very few binding energies of organic compounds with iodide
have been measured or calculated. Binding energies calculated using quantum
chemical methods provide valuable information, but carrying out the
computationally expensive calculations for the 100s of molecular ions
typically identified in our spectra is not feasible, especially when
molecular structure is unknown. Therefore, to constrain the effective
binding energies of the actual multifunctional organics that are measured,
we scan the electric field strength within the transfer optics in real time
while measuring a steady-state distribution of organic compounds. These
scans experimentally determine the electric field strength required to break
apart the iodide–organic adducts, which, in turn, are directly related to
the binding energy of the adduct.
To assess this approach, we used a steady-state atmospheric simulation
chamber at the University of Washington to generate a wide range of oxidized
organics from the reaction of α-pinene in the presence of ozone and
NOx (Lopez-Hilfiker et al., 2015). While sampling this mixture, we
scanned the voltage difference (dV) between the skimmer and the entrance to
the second quadrupole ion guide of the mass spectrometer (see Fig. 2 for
schematic). We call these declustering scans because – by increasing dV – we
systematically increase the collisional energy of the iodide–organic adducts
above our normal operating dV until the adducts dissociate, mostly into
I- and a neutral organic molecule. An example of this type of
experiment is shown in Fig. 3a, where normalized ion count rates are
plotted as a function of dV. During a declustering scan, all potentials
upstream of the second quadrupole are moved together towards more negative
voltages such that the electric field and therefore the declustering
strength is incrementally changed while maintaining a constant gradient
across the quadrupoles to avoid simultaneous changes in the mass
transmission function, which depends upon the axial voltage gradient along
the quadrupole rods (see Fig. 2 schematic).
A schematic of the voltages in the APi region of the Tofwerk
ToF-MS. The region of the mass spectrometer that we conduct declustering
scanning is between the skimmer (orange) and the entrance to the second
quadrupole (red). All voltages upstream of the skimmer are moved
incrementally towards more negative voltages to create a stronger
declustering field while keeping mass transmission effects constant by
keeping the voltage gradients across each quadrupole constant, typically in
steps of -1 V. Iodide adducts which are formed in the ion molecule region
(IMR) interact with the changing electric fields and dissociate based on
their ion–molecule binding energy.
We show the survival of a representative set of iodide adducts as a function
of electric field strength in Fig. 3a. A variety of behaviors are
observed; however, all adducts follow a similar sigmoidal response to dV, as
expected for a threshold-driven process. We find that some adducts (e.g.,
simple monocarboxylic acids, diols) are rapidly dissociated during the
first few voltage steps, while larger multifunctional organics (inferred
from the O / C ratio) tend to survive to higher potentials during the scan
(e.g., multifunctional organic nitrates in Fig. 3a). The observation that
the larger, multifunctional molecules (e.g., C10H15O8N) are
detected with the same efficiency during the first few voltage steps implies
that they are bound to iodide with sufficient binding energy to be
efficiently transmitted by the normal operation of the ion optics in the
mass spectrometer and are therefore likely to be detected at a sensitivity
that depends only on their formation rate, which for lack of a better
constraint we would assume to be at the collision limit.
Effective binding energy. Top: declustering scans of gas-phase
products measured from the reaction of α-pinene in the presence of
ozone and NOx. Some iodide adducts dissociate rapidly during the first
few voltage steps. Multifunctional nitrates and highly oxidized C10
molecules show no dependence on the initial voltage steps before
dissociating. We infer these compounds to be strongly bound to I-, and
therefore likely detected at a high sensitivity. Middle: an example of a
non-linear least-squares fit to the declustering scan is shown for
C10H16O3I-. The raw scan data are shown in blue circles,
the region of optimization for the fit is shown in red, and the extrapolated
scan curve is shown in black, constraining So. Bottom: the results of
fitting declustering scans (1/So) shows a plateauing effect as a
function of dV50. We use dV50 as a measure of the relative binding
energy between compounds. In colored dots, specific molecules are shown
which span the binding enthalpy space.
Even at our weakest electric field settings, many iodide–organic adducts
are partially declustered, and thus the true sigmoidal declustering scan
curve is not observed. In these cases, we calculate an effective maximum
sensitivity by using a custom non-linear least-squares fitting algorithm to
determine the extent to which the ion adduct has been declustered during
transit through the mass spectrometer's atmospheric pressure interface
(APi). The fitting algorithm uses a characteristic sigmoidal shape with
variable amplitude, So, and location of the voltage at half signal
maximum (dV50). If there are isomers or isobaric compounds that
contribute significantly to the ion signal but have different iodide binding
energies, then the declustering scan should not have a sigmoidal shape.
While this information could be useful with a highly resolved dV scan, and
is reflected in the fit error, herein we remove ion adducts with mean square
residual > 10 % from the analysis. So is the relative
signal that would be detected under weaker electric field strengths than we
can operate the instrument (dV < 1 V). So is not a measure of
the actual sensitivity, which is generally unknown, but instead is a measure
of the extent to which declustering during transmission to the detector
affects the actual sensitivity. The dV50 is a measure of relative
binding enthalpy. Iodide adducts that are tightly bound survive to higher
voltage gradients and therefore have a higher dV50 than more weakly
bound adducts.
Pinonic acid (assumed to be measured as the C10H16O3I-
ion) is an example of a compound for which a true sigmoidal curve is not
observed (see Fig. 3b). The fit for pinonic acid implies that the
sensitivity would be enhanced if weaker declustering conditions existed
(e.g., 1.4 = So > 1). The sensitivity of the
instrument to pinonic acid, calibrated by an authentic standard is 15 cps ppt-1. Therefore, if weaker declustering conditions existed in the
instrument, transmission-optimized sensitivity for pinonic acid would be
15×1.4=21 cps ppt-1. This value is near the collision limit
determined by calibration to N2O5, which suggests that the
iodide–pinonic adduct is formed at near the collision limit in the IMR, but
it is partially declustered during transit through the ion optics of the mass
spectrometer, resulting in a lower observed sensitivity during normal
operation settings (e.g., Sobs < So).
In Fig. 3c we show similar voltage scanning fit results for all
iodide–organic adducts (black circles) identified in the mixture produced
from α-pinene ozonolysis in the presence of NOx. We find that
1/So, which is related to the maximum possible transmission for each
compound, plateaus with increasing dV50, suggesting that iodide adducts
with a dV50 ∼ 6 V or higher, are sufficiently bound to
transit the ion optics without significant declustering losses. Adducts
having dV50 > 6 V are composed of highly functionalized
organics (e.g., Fig. 3c: C10H17NO6 (black) and
C20H32O7 (green)), which is consistent with the more strongly
bound iodide adducts generally involving one or more hydrogen bonds from a
polar hydroxy, hydroperoxy, or carboxylic acid group. We also expect that
these compounds are likely formed at near the collision limit as steric
effects are unlikely to significantly limit their formation rate, but this
hypothesis remains to be tested.
Relationship of dV50 to quantum-chemical-derived binding
energies
Figure 4 shows the relationship between iodide adduct binding enthalpies from
quantum chemical calculations, the dV50 values determined from the fits
to the declustering scans (see also Table 1). Assuming the linear
relationship (R2=0.92) between the subset of compounds for which we
have quantum chemical calculations and experimental determinations holds, the
derivation of the binding energy from declustering scans for hundreds of
compounds simultaneously is then possible without explicit knowledge of the
functional groups or molecular geometry, which is required for quantum
calculations. We have shown in a related article that there is a reasonable
relationship between theoretical binding enthalpies and measured sensitivity
(Iyer et al., 2016). Therefore, by constraining the relationship between
quantum calculations and measured scan shape for a subset of compounds, we
can use the measured dV50 to estimate the binding enthalpy, and thus
instrument sensitivity, for compounds that are too computationally intensive
or for which we lack knowledge of molecular structure or cartesian geometries
necessary for optimization. As noted above, the binding enthalpy of an adduct
alone does not necessarily determine overall sensitivity. The rate of adduct
formation and transmission through the mass spectrometer are both important
components of the overall sensitivity.
Compounds used to determine the relationship between dV50%
and binding enthalpy derived from quantum calculations (Iyer et al., 2016) at
the DLPNO-CCSD(T)//PBE-aug-cc-pVTZ-PP level. The relationship is
approximately linear R2=0.9 (see Fig. 4). For details see text.
Compound
Composition
Binding enthalpy
Fit dV50 (V)
(kcal mol-1)
Glycolic acid
C2H4O3
-21.1
4.70
Glyoxylic acid
C2H2O3
-20.8
4.29
Malonic acid
C3H4O4
-27.8
6.21
Formic acid
CH2O2
-23.9
5.80
Acetic acid
C2H4O2
-17.4
4.10
Succinic acid
C4H8O4
-27.6
6.19
Nitric acid
HNO3
-22.2
5.50
Nitrous acid
HONO
-18.7
4.56
The relationship between calculated binding enthalpy and dV50
is shown for compounds which were observed from the oxidation of α-pinene in the presence of NOx and for which quantum calculations have
been performed (see Table 1). The modeled and measured relationship allows
estimation of the binding energy of molecules which are too complex or
computationally intensive to calculate. Also shown in red is a linear least-squares fit to the data (R2=0.9).