Introduction
Recent commercialization and packaging of time-of-flight chemical ionization
mass spectrometers (TOF-CIMS) into field-deployable packages by Aerodyne
Research Inc. and Tofwerk AG has led to the widespread use of these
instruments (Aljawhary et al., 2013; Bertram et al., 2011; Brophy and Farmer, 2015; Chhabra et al.,
2015; Ehn et al., 2010, 2011, 2014; Faust et al., 2016; Friedman et al.,
2016; Jokinen et al., 2012; Junninen et al., 2010; Krechmer et al., 2015;
Lee et al., 2014; Lopez-Hilfiker et al., 2016, 2015, 2014; Mohr et al.,
2013; Sipilä et al., 2015; Yatavelli et al., 2012, 2014; Zhao et al.,
2014). Any chemical ionization (CI) source, or more generally any
near-atmospheric pressure ion source, can be installed on the front end of
the mass spectrometer providing a flexible TOF instrument platform. The
design and operation of the ion source affects the sensitivity of the
instrument, but the fundamental ion chemistry is the key consideration to
designing a CI source that is both sensitive and selective. Thus, the
selection of an appropriate reagent ion for detecting the compound, or class
of compounds, of interest is important (Huey, 2007). The ions
observed in the TOF mass spectrum do not necessarily represent the
distribution of ions generated in the ion source due to collisional
dissociation (Bertram et al., 2011) and mass-dependent transmission
effects (Heinritzi et al., 2016). Collisional dissociation simplifies
the observed mass spectrum and has a long history of use dating back to the
original developments of tropospheric CIMS measurements (Eisele,
1986). Controlling the extent of collisional dissociation can be used to
investigate the ion-neutral chemistry occurring in the ion source. The
TOF-CIMS uses a tunable multistate atmospheric pressure interface (API) that
can eliminate or transmit clusters, but the operational details of this
interface have not been investigated with systematic rigor.
TOF-CIMS represents a distinct departure from traditional quadrupole CIMS
methodologies in which specific species are targeted for quantification.
TOF-CIMS collects a continuous mass spectrum at high (< 1 Hz)
acquisition rates, whereas quadrupole detectors collect a limited number of
ions due to limitations on sensitivity and time resolution due to duty cycle
effects. Additionally, the high-resolution TOF-CIMS (HR-TOF-CIMS) enables the
assignment of a molecular formula to every observed mass peak. These two
features of the HR-TOF-CIMS provide an opportunity to examine CI ion
chemistry. Moreover, users can identify and observe the temporal behavior of
compounds that have not previously been known to exist or calibrated in a
non-targeted approach (Ehn et al., 2014). These features are,
essentially, added benefits of TOF-CIMS over quadrupole CIMS instruments;
however, we note that there is no substitute for authentic standard
calibrations despite the alluring benefits of using TOF mass spectrometers.
Quadrupole systems contrast this non-targeted approach, but calibrations
remain as important for TOFs as quadrupole systems. Ideally, quadrupole CI
is deployed with the intent of measuring specific species with readily
available authentic calibration standards and well-characterized
interferences. Calibrations are typically conducted for a limited number of
compounds, but interferences are difficult to address until they are
identified through instrument intercomparisons and careful study. Mass scans
are often conducted using quadrupole-based CIMS to examine the temporal
behavior of uncalibrated species, but these results are more difficult to
understand compared to the mass spectrum acquired by TOF-CIMS because of the
lack of elemental information present in high-resolution data. The recent
identification of the decomposition of isoprene hydroxy-hydroperoxides
(ISOPOOH) to methyl vinyl ketone (MVK) and methacrolein (MACR) in both gas
chromatograph instruments and proton-transfer reaction mass spectrometers
(PTR-MS) highlights this challenge as both techniques have a long history of
MVK and MACR measurements (Rivera-Rios et al., 2014).
Both TOF and quadrupole detectors remain subject to misinterpretation of the
mass spectrum in the absence of complex interferences. Quadrupole systems
with unit mass resolution can suffer from attributing the signal from a
single mass to charge ratio (m/z) to a single species and potentially miss
isobaric interferences at the same nominal, or unit, mass. Recent
intercomparisons between co-located quadrupole PTR-MS and time-of-flight
PTR-MS instruments highlight the power of high-resolution analysis in the
identification of multiple overlapping peaks (Warneke et al., 2015).
HR-TOF systems can separate closely spaced peaks, but knowing the actual
identity of ion signals from their exact mass and extracting the
high-resolution information remains challenging (Cubison
and Jimenez, 2015; Stark et al., 2015). Improving the knowledge of the ion
chemistry, ionization mechanisms, and instrument performance is paramount to
correctly interpreting the mass spectrum because CI relies on selectively
ionizing specific compounds or classes of compounds. Quantitative and
qualitative non-targeted analysis in the complex chemical space of the
atmosphere using HR-TOF-CIMS necessitates characterization of the chemistry
occurring within the ion source and the instrument's subsequent control over
the transmission, clustering, and fragmentation of those ions.
Acetate CIMS, originally termed negative-ion proton-transfer chemical
ionization, is conventionally thought to selectively ionize carboxylic acids
and some inorganic acids by proton abstraction (Reaction R1) (Veres
et al., 2008). Other compounds, such as nitrated phenols, are detectable
with acetate CIMS due to their gas-phase acidity relative to the acetate ion
(Mohr et al., 2013). However, acetate can also form adducts with
levoglucosan, which are detected as [levoglucosan + acetate]-
clusters and not deprotonated due to their low gas-phase acidity relative to
the acetate ion (Reactions R2–R3) (Zhao et al., 2014). Similarly, isoprene
epoxydiols (IEPOX) and ISOPOOH have also been reported to cluster with
acetate (Budisulistiorini et al., 2015). Perhaps due to this wide
array of potential analytes, acetate CIMS is extensively applied to TOF-CIMS
platforms under a variety of experimental configurations
(Aljawhary et al., 2013; Brophy and Farmer, 2015; Budisulistiorini et al., 2015;
Chhabra et al., 2015; Lopez-Hilfiker et al., 2014, 2015; Mohr et al., 2013;
Wentzell et al., 2013; Yatavelli et al., 2012, 2014; Zhao et al., 2014).
Using an acetate CI source coupled to the Tofwerk API, Bertram et al. (2001)
demonstrated that a distribution of acetate clusters exists but can be
collisionally dissociated during their transfer through the API by applying
stronger electric fields across the ion optics (2011). Potential
ion–molecule reactions occurring between the reagent ion and analyte (R–H)
are thus as follows:
proton abstractionCH3C(O)O-+R-H→CH3C(O)OH+R-cluster reactionCH3C(O)O-+R-H→[CH3C(O)O+R-H]-.
We note, however, that two other types of reactions may be occurring:
ligand exchange[CH3C(O)O+R-H]-+R′-H→[CH3C(O)O+R′-H]-+CH3C(O)OHdeclustering via collisional dissociation[CH3C(O)OH+R]-+ECDC→CH3C(O)OH+R-.
While rare, fragmentation reactions are also known to occur within CIMS
instrumentation:
R+ECDC→r1-+r2.
In light of recent studies detecting nitrated phenols as deprotonated
products (Mohr et al., 2013) and detecting levoglucosan
(Zhao et al., 2014) and IEPOX/ISOPOOH (Budisulistiorini et
al., 2015) as acetate clusters, we suggest that these reaction should be
more generalized to include other molecules with various functional groups
and non-acidic protons. Reactions R1 and R2 have been reported in the
literature for acetate CIMS assuming that carboxylic acids are detected
(Bertram et al., 2011). Ligand exchange reactions (Reaction R3) have not
directly been identified to occur with acetate CIMS, but the chemistry
appears to be very similar to iodide adduct CIMS, where [I + H2O]-
reacts in a ligand exchange reaction with some analyte (X)
to produce [I + X]- (Lee et al., 2014). The declustering
reaction (Reaction R4) is implicitly discussed by Bertram et al. (2011). Lastly, fragmentation
in PTR-MS instruments is known to be an extensive feature. A study of
atmospheric sesquiterpenes (C15H24) identifies seven fragment ions
where the specific sesquiterpene will contribute to each fragment to a
different extent due to structural differences (Kim et al., 2009).
Acetate CIMS can be contrasted to iodide adduct CIMS, another rapidly
developing chemical ionization method being applied to TOF-CIMS platforms
(Aljawhary et al., 2013; Friedman et al., 2016; Lee et al., 2014; Lopez-Hilfiker et al., 2014, 2016;
Zhao et al., 2014). Iodide adduct CIMS predominantly form iodide adducts
with neutral species due to its high electronegativity; iodide is not
expected to substantially abstract protons or transfer electrons
(Iyer et al., 2016). Once ion-neutral clusters are formed, the
ion optics of the mass spectrometer must efficiently transmit these clusters
to the mass analyzer. The lack of proton abstraction or charge transfer
allows this CIMS method to be operated in a cluster mode because the iodide
ion holds the vast majority of the negative charge. Thus, the dominant
clustering mechanism involves iodide. The dominant clustering mechanism with
acetate CIMS involves the acetate ion, but the prevalence of proton
abstraction produces stable anions that will also undergo clustering
reactions.
The Tofwerk API enables users to control and vary the extent of collisional
dissociation, allowing for more representative descriptions of ion source
chemistry. The Tofwerk API consists of two segmented radio-frequency-only (RF) quadrupoles:
the Short Segmented Quadruple (SSQ) and the Big Segmented Quadrupole
(BSQ). These components are housed in two differentially pumped vacuum
chambers and contain various skimmers and entrance plates (Fig. 1).
Between the entrance of the API and the last skimmer after the BSQ, there
are nine individually controllable voltage components and the two RF-only
segmented quadrupoles, making the task of optimally tuning the API a serious
undertaking. This task is made more complex by the realization that
instrument resolution, ion transmission efficiency, and extent of
collisional dissociation are all interrelated.
Schematic of the Tofwerk atmospheric pressure interface (API) showing where
the IMR mounts on the API, the short segmented quadrupole (SSQ), the big
segmented quadrupole (BSQ), and the primary beam (PB) region.
![]()
Previous studies have examined these components with regard to controlling
and/or understanding the transmission of ions and clusters to the TOF
(Bertram et al., 2011; Heinritzi et al., 2016). The original
description of the Tofwerk API uses no ion source and describes ambient ions
and ion-neutral clusters in the atmosphere (Junninen et al., 2010).
The authors characterize the ion transmission efficiency of the API-TOF
tuned to transmit clusters using an electrospray source emitting
tetraheptylammonium bromide. Ion transmission efficiency is defined as the
fraction of ions at the inlet or in the ion source that make it to the
detector. No attempt to systematically characterize or optimize the API
components is presented, although the authors note that ion transmission
efficiency is strongly dependent on the voltage settings in the API. A
comparison of methods for experimentally determining mass-dependent ion
transmission efficiency has also been reported, but no evaluation of the
voltage settings, their relationships, and their effect on clustering or
transmission efficiency is reported (Heinritzi et al., 2016).
The application of the Tofwerk API to a C-TOF configured as an acetate CI
instrument provides some more insight to understanding the relationships
between various API components as they relate to cluster transmission and
collisional dissociation (Bertram et al., 2011). Here, the authors
suggest that collisional dissociation of ion-neutral clusters occurs between
the exit of the SSQ and the entrance of the BSQ vacuum stage. This claim
appears to be in slight contrast to the recent results describing the use of
voltage scanning to determine instrument sensitives with the iodide reagent
ion (Lopez-Hilfiker et al., 2016). The authors scan the API voltages
to increase the difference between the BSQ front and the last skimmer after
the SSQ and lens skimmer (Fig. 1 and Table 1). Ion transmission efficiency
is maintained by floating all components upstream of the last skimmer more
negative as the voltage difference between these two components is
increased. Systematic floating of API components to change the voltage
difference between two components maintains the electric field strengths
between all other components. This approach also prevents changing the
axial-electric field across the RF-only segmented quadrupoles; changing the
axial-electric field will result in changes in ion transmission efficacy
which must be avoided so that a mass spectrum collected under one voltage
setting is comparable to results collected using a different voltage
configuration.
Component relationships defined by adjacent components in the API
(see Fig. 1 for API schematic).
Component
Component A
Component B
relation
1
Q1 EP
SSQ front
2
SSQ front
SSQ back
3
SSQ back
Lens skimmer
4
Lens skimmer
Skimmer
5*
Skimmer
BSQ front
6
BSQ front
BSQ Back
7
BSQ back
Skimmer 2
* Component relation used by Lopez-Hilfiker et al. (2016).
Previous work comparing the iodide adduct, acetate, and water cluster CIMS
methodologies using a HR-TOF-CIMS highlights the need for significant
characterization of collisional dissociation in the API (Aljawhary
et al., 2013). The authors tune a HR-TOF-CIMS to a “weak-field mode” for
iodide adduct and water cluster CIMS operation. A “strong-field mode” is
used while operating in acetate mode. Comparing the negative ion mode
voltage configurations under strong-field and weak-field operation presented
in the supplementary information shows numerous voltage relationships that
may lead to subtle differences in relative ion transmission efficiency. This
problem is not unique, and authors rarely publish exact voltage
configurations as the exact voltages needed to tune the API will vary across
instruments. The lack of careful study when configuring the API is obvious
in the available HR-TOF-CIMS literature using acetate CI where reported
[acetate + acetic acid] to acetate ratios (referred to here after as the
acetate cluster ratio) vary by orders of magnitude (Brophy and Farmer,
2015). For example, Bertram et al. (2011) report an acetate cluster ratio of 2.6×10-3, Mohr et al. (2013) report an acetate cluster ratio of 0.07, Brophy and
Farmer (2015) report an acetate cluster ratio of 0.02, and Chhabra et al. (2015) report
an acetate cluster ratio of 0.2.
We present a comprehensive characterization of the Tofwerk API. This
characterization of the API allows for a detailed investigation of the
acetate ionization mechanisms and the impact of controlling for these
mechanisms with collisional dissociation in the ion transfer optics on
sensitivity, detection limits, selectivity, and mass spectral ambiguity with
the general aim of non-targeted analysis. We show that the majority, if not
all, ion-neutral chemistry occurs in the ion molecule reactor (IMR) where
incoming sample air mixes with the output of the ion source. Lastly, we
provide insight on configuring these HR-TOF-CIMS systems for non-targeted
analysis and the detection of clusters.
Methods
Instrument description and chemical ionization source
The HR-TOF-CIMS (Tofwerk AG and Aerodyne Research, Inc.) is described
extensively in the literature (Bertram et al., 2011; Brophy and Farmer, 2015;
DeCarlo et al., 2006; Jokinen et al., 2012; Junninen et al., 2010; Lee et
al., 2014). The instrument described herein is operated in the negative ion
mode with acetate reagent ions. The configuration is described in detail by
Brophy and Farmer (2015). Notable differences include the use of a larger
SH-112 single scroll pump (Agilent Technologies, Inc.) backing the IMR, a
custom-built quartz glass reservoir with metal to quartz fittings for holding
the reagent precursor, and the use of the standard IMR critical orifice for
sampling from atmospheric pressure at 1900 sccm. Mass spectra are acquired
at an extraction frequency of 25.0 kHz and pre-averaged to 1 s mass spectra
over a mass range from 2 to 494 m/z using an analog to digital converter
(ADQ1600 SP Devices). Instrument resolution is > 5000 for peaks above
∼ 100 m/z. A TofDaq recorder is used to configure TOF acquisition
parameters (Tofwerk AG, ToFDaq Version 1.97) and record mass spectra.
The quartz glass reagent reservoir is filled with acetic anhydride
(Sigma-Aldrich, ≥ 99 % ReagentPlus Grade) and the headspace is
continually flushed with ∼ 10 sccm ultrahigh purity (UHP) N2 from a cryogenic dewar
(Airgas, Inc.) using a 50 sccm mass flow controller (MKS Instruments, Inc.
1179A) and analog controller (MKS Instruments, Inc. 247D). All connections
to and from the reagent reservoir are made with instrument grade 1/8′′
outer diameter, 0.055′′ inner diameter stainless steel instrument-grade
lines (Restek, Inc.) and stainless steel Swagelok fittings and ferrules. The
saturated headspace is mixed with a diluting UHP N2 flow
(∼ 2000 sccm) controlled with a stainless steel needle valve
(Swagelok). Approximately 900 sccm of the N2–acetic anhydride mixture
is subsampled through a critical orifice (O'Keefe Controls) into a Po-210
ionizer (NRD Static Controls LLC). A short, 1/4′′ outer diameter piece of
PEEK tubing (Vici Metronics) separates and electrically isolates the
stainless steel line from the Po-210 ionizer that is directly threaded into
the IMR body. The glass reservoir, stainless steel lines, and Po-210 ionizer
are held at 40 ∘C with a PID (proportional–integral–derivative) temperature controller and heating tape
(Omega Engineering, Inc.). The IMR is held at 50 ∘C using the
temperature controls on board the HR-TOF-CIMS. IMR and SSQ pressures are held
constant at 100 and 2 mbar, respectively.
Experimental setup
Gas-phase standards of formic (CH2O2), propionic
(C3H6O2), butyric (C4H8O2), methacrylic
(C4H6O2), nitric (HNO3), and hydrochloric acid (HCl) are
generated using permeation tubes (KIN-TEK Laboratories, Inc.) and custom-built permeation ovens. The ovens are continually flushed with UHP N2 at
a constant flow of ∼ 50 sccm and resistively headed to a constant
temperature using PID temperature controllers (Omega Engineering, Inc.). The
permeation rate of each species is determined by monitoring the mass loss of
the permeation tubes over the course of months. Ultra zero grade air (Airgas
Inc.) is mixed with the output of the permeation tubes to create single-component alkanoic acid standards in clean air. The same source of ultra zero
grade air is also used for instrument zeros and humidified air.
Humidification of zero air is accomplished by passing the air through a
series of large-volume glass, custom-built water bubblers filled with LCMS
grade water (Sigma-Aldrich). The humidity system operates at a constant flow
and varies the relative humidity (RH) using two mass flow controllers (MKS
Instruments, Inc. 1179A) and a PID loop controlling the RH from an inline RH
sensor (Omega Engineering HX71-V1). The sum of two mass flow controllers is
held constant by the PID control loop and the ratio of the flows are changed
to produce a large range of humidified air (0–90 %). All additional
flows are controlled with mass flow controllers (MKS Instruments, Inc.
1179A). The gas sample flow is sent to the HR-TOF-CIMS through a
polychlorotrifluoroethylene 3-way solenoid valve with a 3 mm orifice
(NResearch, Inc) positioned upstream from the humidified air system. The
total humidified flow is set below the sampling flow of the HR-TOF-CIMS
(2 standard L min-1) and either zero air or a calibration mixture is
subsampled through the 3-way valve to make up the remaining sample flow
(Fig. S1 in the Supplement).
All components of this system are automated to allow for comprehensive
calibrations of the six authentic acid standards under different instrument
settings and different RH conditions. LabVIEW scripts (LabVIEW 2014 Version
14.0f1, National Instruments, Inc.) control the gas flows using
predetermined sets of flow rates, humidity settings, and instrument voltage
configurations. Multiple data acquisition devices (Labjack Inc, U12) are
implemented to record all flows, RH sensor output, and valve states. The
HR-TOF-CIMS is controlled using the Tofwerk Application Programming Interface
(Tofwerk AG, Version 1.97) from within the LabVIEW environment. All data
streams read by the data acquisition devices are logged to the Tofwerk HDF
files along with the HR-TOF-CIMS data.
Two general modes of operation exist for this experimental setup: full
calibration mode and voltage scanning mode. Briefly, operating in full
calibration mode produces one background-subtracted multipoint calibration
curve at each specified RH setting. Next, LabVIEW changes the instrument
voltage settings and repeats the experiment. One file is created for each
instrument zero and calibration step in order to simplify data processing by
averaging entire files of a fixed length. Voltage scanning mode utilizes the
same flow system but maintains all the flows while switching instrument
voltages. Again, a separate data file is created for each voltage
configuration.
Data analysis
Post-processing is performed in Igor Pro (WaveMetrics Inc, Version 6.3.7.2)
running Tofware (Tofwerk AG, Aerodyne Research, Inc. Version 2.5.3). Tofware
is used to process, fit, and then extract HR-TOF data and auxiliary data
generated from the experimental setup. Once the integrated high-resolution
time series are extracted, scripts developed in Igor Pro process all of the
experimental data to produce calibration curve summaries and statistics. TOF
duty cycle corrections are made within Tofware at m/z 59 for all data
collected (Drewnick et al., 2005). Mass calibration is conducted using a
three-parameter fit available within Tofware using O2-,
35Cl-, 37Cl-, CHO2-, NO2-,
C2H3O2-, and NO3- as mass calibration peaks.
Additionally, C4H7O4-, the [acetic acid + acetate]
cluster, is also included for mass calibration when operating in clustering
modes where sufficient signal from this species is detected.
Calibration experiments are normalized by the ratio of the total ion signal
at each calibration step relative to the total ion signal in zero air.
Traditionally, normalization is conducted using the acetate reagent ion.
Under declustered settings, acetate accounts for most of the total ion signal
(> 80 %). These calibration experiments are complicated by voltage
scanning of the API. Clustered settings retain the acetate clusters, which
can contribute more of the total ion signal than acetate. Thus, the use of
total ion signal is appropriate to maintain consistency in normalization
procedures across a large range of clustering conditions. Experiments where
only zero air is used are normalized by simply dividing the individual ion
signals by the total ion signal and expressing the observed signals as a
fraction of total signal.
Top panel: an ion of interest is normalized to the total ion signal
and plotted against the voltage difference for some component relation
(component relation 3 is shown here). The black circles show the portion of
the curve used to average the signal of the ion during operation under weak
(clustered) electric field strength, with an inset box-and-whisker plot
representing the clustered-average. Blue circles show the portion of the
curve used to average the signal of the ion during operation under strong
(declustered) electric field strength, with an inset box-and-whisker plot
representing the declustered-average. The dV50 value obtained from the
nonlinear least-squares sigmoidal fit is also displayed. Bottom: the
correlation scatter plots of the ion of interest with acetate and the [acetic
acid + acetate] cluster. Linear regression produces an r2 correlation
coefficient for the ion of interest vs. acetate (used as the model for a
declustered-deprotonated species) and the ion of interest vs. [acetic acid + acetate].
These two correlation coefficients are summed and used as
criteria for including or excluding declustering scans.
![]()
Voltage set determination with Thuner
Exploration of the API component relations provides additional insight to the
operation of these complex instruments. Very large sets of voltages
(> 40 000 configurations) are produced using Thuner (Tofwerk AG, Version
1.9.11.0), a design of experiment optimization software produced by Tofwerk.
Thuner enables the user to establish relationships between various API
components and set performance targets (resolution, sensitivity, peak shape).
These component relations are used to set voltages in the API and optimize
voltage settings based on the performance target criteria. For example, to
optimize the instrument in a clustered mode (i.e., maintaining low electric
field strengths through the API), the voltage difference between the last skimmer
of the SSQ and the BSQ front can be set to a small range (0–2 V), allowing
Thuner to test the impact of tuning each region of the API on desired
parameters. Increasing the voltage difference between the skimmer and the BSQ
front (2–4, 4–6 V, etc.) moves the instrument stepwise from a cluster
transmitting regime to a declustering regime. The SSQ and BSQ RF frequency
and amplitude are held constant at 2.65 and 4.26 MHz with an amplitude of
200 and 400 V, respectively.
Maintaining instrument and sample stability is essential during these
experiments, particularly when comprehensive (> 7 days) Thuner experiments
are conducted. For this reason, the instrument samples ultra zero air
throughout the entire Thuner experiment. Further details from these
experiments are presented in the Supplement (Sect. S2). Ultimately, a single
voltage configuration at a voltage difference of 1 V between the last
skimmer and the BSQ front are chosen by filtering for sensitivity, resolution,
and the acetate cluster ratio using the Thuner XML output and various scripts
written in Igor Pro. This voltage starting point is used to create all the
voltage set point files for the voltage scanning experiments (Table S1).
Seven component relations are defined between adjacent components (Fig. 1 and
Table 1). The component nearest the TOF is held constant, while all components
upstream are floated together including the IMR body itself.
Voltage scanning and cluster detection
We use nonlinear least-squares sigmoidal regression following the work by
Lopez-Hilfiker et al. to describe declustering voltage scans and determine
the characteristic voltage (dV50) at half signal maximum (Lopez-Hilfiker
et al., 2016). The work by Lopez-Hilfiker et al. (2016) focuses on the
sensitivities for iodide ion-neutral clusters and only examines declustering
scans of species initially clustered with iodide that fall apart upon
increasing the electric field strength between two components. In contrast,
acetate CIMS produces both ion-neutral clusters with acetate (and other
negatively charged ions) and deprotonated-declustered ions. This more complex
case means that ion signals may either increase or decrease as the electric
field strength increases during a voltage scan. As such, we introduce another
parameter to further describe the behavior of ions detected in the
high-resolution mass spectrum: positive and negative change. Figure 2 details
the fitting procedure and quantification of this change. A stable region at
low-voltage differences (low electric filed strength, high cluster
transmission) is averaged and compared to a stable region at high-voltage
differences (high electric field strength, low cluster transmission) using a
Student's t test. If the null hypothesis (the two populations are the same)
is rejected, then the percent change is calculated. This allows for the
dV50 to be converted into a positive or negative number.
A representative background mass spectrum obtained by overflowing the IMR
with zero air is shown at three voltage differences (component relation 5).
Both the log-scale mass spectrum (left column) and linear-scale (right
column) mass spectrum are displayed. Dominant peaks related to the reagent
ion chemistry are labeled.
![]()
Lopez-Hilfiker et al. (2016) filter their fits based on the criterion that if
the mean square residual is > 10 %, the fit is rejected. This is
insufficient for our purposes of characterizing this more complex set of
voltage scans. Instead, we first reject unreasonable dV50 values where
the calculated dV50 is greater than twice the scan range. Next, we use
the output of the Student's t test to remove voltage scans with start and
end points that likely belong to the same population (3 s certainty).
Lastly, linear least-squares cross correlation is performed on the
high-resolution time series to identify correlation with the acetate reagent
ion and the [acetate + acetic acid]- cluster. The acetate reagent
ion follows an increasing sigmoidal shape due to the declustering of various
acetate (or other anion)-containing clusters and thus acts as a model shape
for deprotonated-declustered species. The opposite is true for the
[acetate + acetic acid]- cluster: the decluster scan follows a
decreasing sigmoidal shape characteristic of clusters breaking apart. The sum
of the correlation coefficients of the species of interest vs. acetate and
the species of interest vs. the [acetate + acetic acid]- provides a
final cutoff (rsum2≥1.5). Most sigmoidal fits that remain
follow anticipated declustering or clustering shapes.
Results
Typical mass spectrum and voltage scanning
The overall effect of voltage scanning on the observed mass spectrum using
acetate CIMS is partially described in previous work characterizing the
application of the Tofwerk API with a C-TOF-CIMS (Bertram et al., 2011). Our
use of a HR-TOF-CIMS enables further identification of dominant peaks in the
mass spectrum and a more comprehensive analysis of tuning effects and
ionization chemistry. Figure 3 shows both the log-scale mass spectrum and
linear-scale mass spectrum collected while flowing ultra zero air into the
inlet and changing the voltage difference between the skimmer and BSQ front
(component relation 5). The mass spectrum collected under high electric field
strength (dV = 12) is dominated by the acetate reagent ion (m/z 59.014,
80.1 % total signal) with a small contribution from O2-
(3.2 % total signal), CHO2- (2.4 % total signal), and
C2HO4- (4.7 % total signal). Decreasing the voltage
difference by half (dV = 6) decreases the acetate reagent ion
contribution (67.8 % total signal) while enabling the appearance of the
[acetic acid + acetate]- cluster (m/z 119.035, 7.9 % total
signal) and contributing a small amount of C2H4O4-
(1.85 % total signal) and [acetate + C2H3O5]-
(m/z 166.012, 2.1 % total signal). The most clustered settings
(dV = 2) completely change the reagent ion distribution. The acetate
reagent ion decreases significantly (28 % total signal) and both the
[acetic acid + acetate]- cluster and
[acetate + C2H3O5]- cluster increase drastically
(35.2 and 19.3 %, respectively).
Voltage scan results conducted between the seven component relations in the API.
Acetate (red dots) and the [acetic acid + acetate] cluster (black dots)
are plotted on the left axis. Fitted dV50 values for acetate (red
trace) and the [acetic acid + acetate] cluster (black trace) are included
along with the least-squares sigmoidal fit. The acetate ratio (gray dots) is
plotted on the right axis as a log scale. Left column: components housed in
the vacuum chamber containing the SSQ (2 mbar). Right column: components
housed in the vacuum chamber containing the BSQ (0.013 mbar).
![]()
The appearance of C2H3O5 clustered with acetate is
unanticipated. We suggest this cluster is a radical fragment from the reagent
precursor (acetic anhydride) that rapidly undergoes auto-oxidation with two
O2 molecules (Sect. S3). Overflowing the IMR with UHP N2 and
eliminating O2 completely removes this observed cluster. Bertram et
al. (2011) observe even more clustering than herein, with the higher order cluster
[(acetic acid)2+ acetate]- nearly equal to the acetate ion.
Another interesting feature of the mass spectrum presented by Bertram et al.
is the presence of a peak at m/z 166, potentially corresponding to the
[acetate + C2H3O5]- cluster. Despite observing a large
abundance of the [(acetic acid)2+ acetate]- cluster,
[acetate + C2H3O5]- remains quite small and may point
to other operational differences such as the amount of acetic anhydride added
to the Po-210 ionizer. The effect of the amount of acetic anhydride remains
an unanswered question in the literature, but we present preliminary experiments
addressing this variable in the Supplement (Sect. S4).
API component relations
Voltage scans and cluster control have been discussed in terms of the voltage
difference between the skimmer and the BSQ front (component relation 5), but
numerous other component relations exist that may be suitable for controlling
collisional dissociation. To address other component relations, dry ultra
zero air is flowed into the instrument inlet, and acetate and the first
cluster, [acetic acid + acetate]-, are examined as model compounds
for deprotonated-declustered species and cluster species, respectively (see
Sect. 2.5). Figure 4 shows the voltage scans for adjacent components in the
API. It is apparent that while the dV50 for acetate and the [acetic
acid + acetate]- cluster is extremely similar for any given set of
component voltages, the dV50 for these species varies substantially
across components. This is due to differences in both component spacing and
pressure in the two regions of the API (BSQ, SSQ), resulting in different
electric field strengths at the same voltage difference. The [acetic acid + acetate]- to
acetate ratio is also displayed in Fig. 4. This ratio is reported in a number of acetate CIMS publications (discussed in the
Introduction) and provides a direct comparison between instruments to
describe the extent of clustering. Converting the applied voltage differences
into units of Townsend (Td) (i.e., electric field strength, E, normalized to
the number density, N) shows orders of magnitude variability in the fitted
dV50 values for acetate and the [acetic acid + acetate]-
cluster. Thus, the E/N formulation fails to explain the collisional
dissociation energies between various components in this system.
If the voltage difference between adjacent components is set with a voltage
difference of 0 V, ion flow through these components is controlled by fluid
mechanics alone, and a decrease in ion transmission efficiency is observed.
Thus, there is a lower limit to how gently one can transmit ions through
these components while maintaining an electric field and high ion
transmission efficiency. Deviations from the sigmoidal fit are observed at
higher voltage differences for the axial voltage component on the BSQ
(component relation 6). This field is applied between the BSQ back and BSQ
front, but this deviation is attributable to ion transmission effects through
the BSQ. This feature does not appear with the SSQ (component relation 2)
because sufficiently high voltages needed to complete the curve could not be
achieved due to voltage limits applied to the API to prevent electrical
discharge. Another interesting feature is observed when scanning the second
skimmer, located after the BSQ, and the BSQ back (component relation 7).
Here, the cluster never reaches zero and the acetate signal remains
correspondingly low in comparison to other components.
The effect of water vapor on various reagent ions is shown under two voltage
settings, dV = 2 (clustered) and dV = 20 (declustered), at component relation
5.
![]()
The exit of the SSQ to the lens skimmer (component relation 3) provides a
promising region for cluster control compared to the choice of components
used in previous studies (component relation 5) (Lopez-Hilfiker et al.,
2016). The dV50 of the [iodide + water] cluster cannot be resolved
because sufficiently weak electric fields cannot be generated. Similarly, the
dV50 for the [iodide + formic acid] cluster can only be partially
resolved. Scanning the BSQ front and skimmer relation (component relation 5)
for the relevant acetate ions and ion-neutral clusters results in the
identical problem described by Lopez-Hilfiker et al. (2016) when attempting
to resolve the entire sigmoidal curve: we are unable to generate sufficiently
low electric fields needed to transmit weakly bound clusters. In contrast,
the exit of the SSQ to lens skimmer (component 3) better allows us to
quantify stable cluster and decluster regions; this greatly improves our
ability to detect clusters during non-targeted voltage switching experiments
because there are stable regions that can be defined as clustered and
declustered (see Sects. 2.5 and 3.6).
Acetate CIMS ion chemistry
Operating the HR-TOF-CIMS with acetate reagent ions in a clustering mode
provides a more representative view of the ion-neutral chemistry occurring in
the IMR than the declustered mode. One interesting observation is that
despite the relatively high pressure in the SSQ region (2 mbar) ion-neutral
clusters do not appear to form in this region. One can attribute all the
ion-neutral clustering chemistry to either reactions in the IMR or cluster
condensation during the jet expansion from the IMR into the SSQ. This is
inferred because clustering can be controlled between the SSQ entrance plate
and the SSQ front (component relation 1). After passing through this region,
the ions must make it through the entire length of the SSQ and subsequent
skimmers, making up most of the residence time through this region.
RH effects on the reagent ions are investigated while operating the
HR-TOF-CIMS in both cluster mode (component relation 5, dV = 2) and
declustered mode (component relation 5, dV = 20) (Fig. 5). Using dry
ultra zero air, the abundances of acetate, [acetate + acetic acid]-
cluster, and [acetate + C2H3O5]- radical cluster are
quite similar. Upon the addition of water, these abundances drastically
change with the appearance of an [acetate + water]- cluster. The
[acetate + water]- cluster competes with the other clusters while
the sum total of acetate, [acetate + acetic acid]-,
[acetate + C2H3O5]-, and [acetate + water]-
remains unchanged. We note that the [acetate + water]- cluster is
observed to increase under the highly declustered settings (component relation
5, dV = 20), although it only makes up a small fraction of the total
signal (∼ 0.01 % total ion signal) compared to the clustered
settings (∼ 17% total ion signal) at the highest RH (80 % RH).
Although Veres et al. (2008) note that a collisional dissociation chamber is
important to “dissociate weakly bound cluster ions such as
CH3C(O)O-(H2O)n”, neither the
[acetate + C2H3O5]- nor [acetate + water]-
clusters have been directly identified in previous studies. This may be due
to most acetate CIMS experiments being run under relatively declustered
settings. The large abundance of acetate clusters observed in this study
suggests that instead of a single reagent ion, chemical ionization in acetate
CIMS is controlled by a distribution of reagent ion-neutral clusters that
vary with RH and the concentration of the acetic anhydride precursor. This is an observation
consistent with Bertram et al. (2011), in which their comparison between
the observed mass spectra under low and high electric field strengths leads
to the realization that numerous acetate clusters exist and are involved in
ionizing reactions.
The sensitivity to propionic acid and related clusters is plotted against
the voltage difference applied between the skimmer and BSQ front (component
relation 5) in units of normalized counts per second per ppb (ncps ppb-1). The
points are colored by the calculated water vapor content in the IMR
corresponding to changing the relative humidity from 0 to 80 % under
laboratory conditions.
![]()
The propionic acid data used in Fig. 6 at a dV = 2 are replotted as a
function of water vapor content in the IMR. Top: the calculated values for the
limit of detection for propionic-acid-related ions (S/N=3, 1 s
integration). Middle: the limit of detection (S/N=3, 1 s integration)
relative to the limit of detection under dry conditions. Bottom: the change
in sensitivity relative to the sensitivity under dry conditions.
![]()
Comprehensive calibrations
Calibrations of six acid standards exhibit similar RH and voltage dependences
for both the deprotonated-declustered ions and ion-neutral clusters (Fig. 6,
Sect. S5). All voltage scans are conducted between the skimmer and the BSQ
front (component relation 5). Propionic acid is exemplary of the behavior of
the carboxylic acids. In cluster mode, the dominant ions are the
deprotonated-declustered conjugate base of the acid (R-), the acetate
cluster [acetate + R–H]-, the water cluster
[H2O + R–H]-, and the self-cluster [R–H + R]-. The
self-cluster is not observed for methacrylic acid (Sect. S5, Fig. S9). The
observation of all these clusters is suppressed by operating the instrument
in a declustered mode. We reiterate that while the detection of these
clusters can be eliminated in the ion transfer optics, the cluster chemistry
is still occurring in the IMR. Operation in clustered mode produces linear
calibration curves for all species presented. Here, we define the sensitivity
as the analytical sensitivity of the instrument; the sensitivity is the
observed response to a known concentration, which is a function of both ion
chemistry and instrument operational parameters. In clustered mode, the most
sensitive ion for the four carboxylic acids is the acetate cluster, although
the sensitivities for this cluster rapidly decrease as the electric field
strength is increased and the clusters are broken up. Increasing the electric
field strength simultaneously increases the calculated sensitivity of all the
acids at their deprotonated-declustered mass to a point where either ion
transmission effects or fragmentation begin to occur and lower the
sensitivity. The most sensitive ion for the two inorganic acids in cluster
mode (hydrochloric acid and nitric acid) is the deprotonated-declustered ion.
The [nitrate + nitric acid]- self cluster also shows high
sensitivity for nitric acid.
Voltage scanning experiments for a variety of potential fragment ions. Top:
component relation 5. Bottom: component relation 3.
![]()
The RH dependence of these clusters proceeds in the same manner as the
reagent clusters. Increasing water vapor concentration in the IMR (or RH in
the sample line) decreases the [acetate + R–H]- cluster by forming
the associated water cluster. Thus, the water cluster and acetate cluster
have opposite RH effects. The same effect is observed for the self-cluster.
The deprotonated-declustered ion is more difficult to reconcile. Under dry
conditions, the deprotonated-declustered ion exhibits the lowest sensitivity.
Increasing the water vapor content leads to an initial increase in
sensitivity, followed by a suppression of sensitivity at the highest water
vapor content.
These effects can be clearly observed by examining a single voltage set
corresponding to a vertical slice of Fig. 6. Figure 7 shows this picture at
dV = 2 and examines the change in sensitivity as a function of IMR water
content. Here, the sensitivity to water vapor is normalized to the
sensitivity under dry conditions following the work by Lee et al. (2014). The
change in sensitivity to the [propanoate + H2O]- cluster
increases 149-fold while the sensitivity of the self-cluster and acetate
cluster decrease by a factor of 5 and 1.6, respectively. For all analyte
acids, the sensitivity of the deprotonated-declustered ions changes by, at
most, a factor of 2 as a function of water vapor concentration, indicating a
robust signal.
Acetate CIMS measurements are characterized by high background count rates
which affect the limit of detection (LOD). The LOD is calculated for all
calibration curves (Sect. S5.3). The LODs of propionic acid ions detected in
cluster mode are plotted in Fig. 7. The high sensitivity to the
[acetate + R–H]- clusters leads to a very low 1 s LOD (S/N=3)
below 10 ppt for the alkanoic acids detected as clusters. The same acids
have higher LODs operating in declustered mode ∼ 50–100 ppt, examining
the deprotonated-declustered ions. The lower sensitivity towards the
inorganic acids produces much higher LODs. Deprotonated-declustered nitric
acid and hydrochloric acid have 1 s LODs (S/N=3) of 2–4 ppb and
< 8 ppb, respectively.
These low LODs for the [acetate + R–H]- carboxylic acid clusters are
attributable to the lower background count rates observed at higher m/z and
the higher sensitivity of the [acetate + R–H]- carboxylic acid
clusters. Increasing the voltage difference decreases the sensitivity to
these clusters; therefore we calculate very high LODs. Under normal field
operating conditions, these clusters will not contribute substantially to the
observed mass spectrum. This greatly improves our ability to identify and
extract molecular information by decreasing mass spectral complexity;
artifacts like water clusters and self- or cross-clustering reactions are
eliminated. The Supplement (Sect. S5.3) provides a more detailed presentation
of the LODs for all calibrated species.
Evidence of fragmentation
Molecular fragmentation can occur at high electric field strengths. Specific
ions observed in the mass spectrum enable investigation of the voltages at
which fragmentation onsets. We investigate fragmentation between the SSQ back
and the lens skimmer (component relation 3) and between the skimmer and BSQ
front (component relation 5). We identify at least six ions (O2-,
C2HO-, C2H3O-, HO2-, C3H3-,
C3H3O-) with very high dV50 values and molecular formulae
consistent with fragmentation (Fig. 8). Fragmentation must be considered when
configuring a CIMS experiment to avoid destroying compounds of interest.
Scanning component relation 3 exhibits a clear fragmentation curve, with most
dV50 values occurring at dV = 40. This is about twice as large as
the dV50 values observed for the [acetate + acetic acid]-
cluster. The sigmoidal fits for component relation 5 are less obvious, but
calculated dV50 values for the fragment ions are nearly 4 times as
large as observed for the [acetate + acetic acid]- cluster. The
possibility of molecular fragmentation may explain some of the observed
decreases in sensitivity at the higher dV values for the
deprotonated-declustered ions (Fig. 6). It remains challenging to separate
the fragmentation effect from ion transmission effects that may dominate
under very high electric field strengths (dV > 40 V component relation
3, dV > 15 V component relation 5).
Mass defect plots from scanning component relation 3 during the α-pinene PAM
chamber experiment. (a) All species (642 ions) are plotted as open circles
and colored by the sum correlation coefficient (gray color-scale).
Hydrocarbons (599 ions) are plotted as solid circles and colored by the
calculated oxygen to carbon ratio (rainbow color scale). Both the open
circles and solid circles are sized by the percent change during the
scanning experiment. Small circles are decreasing as a function of
increasing electric field strength while large circles are increasing.
(b) Hydrocarbons that meet the scanning criteria (334 ions) are plotted as solid
circles colored by their oxygen to carbon ratio (rainbow color-scale); these
are also sized by the percent change during the scanning experiment.
(c) Left: ions (94 non-clusters) which meet the scanning criteria and increase
during a voltage scan (+ dV50) are plotted as solid circles (Cyan
color scale). Middle: ions which meet the scanning criteria (334 ions) are
colored by dV50 (cyan-magenta color-scale). Right: ions (262 clusters)
which meet the scanning criteria and decrease during a voltage scan
(- dV50) are plotted as solid circles (magenta color scale).
![]()
Bulk properties calculated in Tofware during the α-pinene PAM
chamber experiment are plotted as a function of dV (component relation 3).
![]()
Complex sample declustering and implications for ambient atmospheric
data
We use a potential aerosol mass (PAM) chamber to create a complex mixture of
oxidized organic species in high concentrations from the oxidation of α-pinene by OH. The purpose of this experiment is not to make any claims
about α-pinene + OH chemistry but instead to generate a complex
mixture of oxygenated organic molecules that is not currently possible to
obtain through authentic standards. The configuration of the PAM chamber is
described in detail by Friedman et al. (2016). Briefly, a flow of
∼ 65 ppbv α-pinene is oxidized by 2.8×106 molecules cm-3 OH (or ∼ 1.9 days equivalent OH exposure)
in a 13.1 L PAM chamber in ∼ 2 min. Declustering scans between the
SSQ back and lens skimmer (component relation 3) changes the observed mass
spectrum and is best observed in mass defect space (Fig. 9). The dV50
values obtained for the clustered ions show an increasing trend with
increasing mass, consistent with Lopez-Hilifiker et al. (2016)'s observation
that large multifunctional nitrates and large oxygenated species exhibit high
dV50 values.
Alkanoic acid species scanning results during authentic standard API
scanning (left column) and PAM chamber scanning (right column). dV50
values are reported for the deprotonated-declustered ions (increasing signal
with increasing dV) and for the acetate-clustered ions (decreasing signal
with increasing dV).
Standard scans (dV50)
PAM scans (dV50)
Species
Primary ions
Acetate cluster
Primary ion
Acetate cluster
(appearance)
(disappearance)
(appearance)
(disappearance)
Formic acid (CH(O)OH)
23.02±0.9
22.62±0.4
24.00±0.1
22.86±0.03
Propionic acid (C2H5C(O)OH)
24.56±0.2
25.02±0.9
30.15±0.8
25.37±0.7
Butyric acid (C3H7C(O)OH)
25.65±0.2
26.09±0.1
29.67±0.6
25.81±0.3
Pentanoic acid (C4H9C(O)OH)
NA
NA
30.15±0.8
25.37±0.71
Hexanoic acid (C5H11C(O)OH)
NA
NA
33.27±3.3
(23.97±1.0)1
Acetate reagent (CH3C(O)OH)
23.87±0.03
24.05±0.02
24.03±0.05
24.25±0.03
Individual alkanoic acid scans obtained during the scanning PAM chamber
experiment (left) and [alkanoic acid + acetate] cluster scans (right).
![]()
Bulk descriptive values are calculated by Tofware using the ion signal
intensity to weight the contribution of each individual ion to the total
signal (Fig. 10). This approach is frequently conducted with the HR-TOF-CIMS,
either without correcting for differences in sensitivity (Friedman et al.,
2016), or by applying the sensitivity of one species (typically formic acid) to
every species (Chhabra et al., 2015). The main finding is that the average
oxygen to carbon ratio (O : C), hydrogen to carbon ratio (H : C),
oxidation state, and carbon number (number of carbons) all change significantly as
a function of applied voltage difference. The average number of carbons per ion
decreases by ∼ 1.7 carbon atoms per ion at the most declustered
voltages. This is consistent with primarily removing or declustering acetate-containing clusters; other species are simultaneously declustered, but
acetate is the most abundant ion and thus comprises the dominant cluster. The
average oxidation state decreases with declustering voltage with three
distinct regions: low (0–18 V), intermediate (18–30 V), and high
(> 30 V) dV values. Little declustering is observed at low dV. The
observed average oxidation state remains stable and is consistent with the
behavior of other bulk metrics. Acetate clusters and other ion-neutral
clusters dissociate in the intermediate dV range, causing the steep change
observed in average oxidation state. Calculated bulk oxidation state
continues to decrease at high dV values and is either due to the onset of
fragmentation or the continued dissociation of strongly bound clusters.
dV50 values obtained by scanning component relation 3 using authentic
standards are compared to the dV50 values of the same ions observed
during the declustering PAM experiment (Table 2). The dV50 values
determined for the disappearance of acetate clusters using authentic
standards in zero air are similar to the dV50 values obtained during the
PAM scans (Fig. 11). However, the PAM declustering scan consistently shows
larger dV50 values for the deprotonated-declustered ions of all alkanoic
acids. We present two hypotheses to account for this observation as follows. (1) The
complex mix of species produces clusters not only with acetate, but also with
other abundant ions. This is consistent with the observation of self-clusters
and clusters with background ions during the single-component comprehensive
calibrations (Sect. 3.4). If these clusters are more strongly bound than the
acetate-containing clusters, then the destruction of the more strongly bound
clusters will continue as the voltage difference increases and lead to the
observed increase in dV50 values. (2) The difference in dV50 for
deprotonated-declustered ions between the standards and PAM mixture may be
the result of fragmentation of multifunctional oxygenated hydrocarbons that
are decomposing to ions that are isobaric with the alkanoic acids.
The shape of the declustered-deprotonated ions during the PAM declustering
scan is different from the behavior of these species during single-component
declustering scans in zero air. When individual authentic standards are
added to zero air and declustering depletes the acetate-carboxylic acid
cluster, the corresponding deprotonated-declustered ion ceases to change. In
contrast, PAM declustering scans show a continually increasing signal for the
C3–C5 alkanoic acids with declustering. The signal intensity of the C3–C5
alkanoic acids during the PAM experiment is quite low in comparison to
formic acid. Thus, the amount of fragmentation or declustering from strongly
bound clusters must be substantial to actually observe this effect for
formic acid.
Discussion: acetate CIMS
Acetate CIMS ionizes analytes by both proton abstraction (Reaction R1) and
ion-neutral clustering reactions (Reacions R2–R3). Detected ions are
observed as deprotonated-declustered ions because of the collisional
dissociation that occurs during the transfer of the ions from the ion source
to the mass analyzer (Reaction R4). The original development of this method
by Veres et al. (2008) does not investigate the importance of clustering in
the ion source due to the use of a quadrupole mass spectrometer, limited mass
scan range, and a collisional dissociation chamber (CDC). The idea that
acetate CIMS is selective towards carboxylic acids is true, but the two
ionization pathways (clustering vs. direct proton abstraction) complicate
mass spectral interpretation and efficient declustering with a CDC is
necessary. Thus, the selectivity of acetate towards acids is really a
function of both ion-neutral chemistry and instrument operation.
We find that the acetate CIMS reagent ions and reagent ion clusters behave
similarly to the detected species in both clustering behavior and effects of
API declustering. The observed clustering behavior of the reagent ions with
water (Fig. 5) explains the sensitivity dependence on RH (Figs. 6–7,
Sect. S4). During calibration, the analyte-containing clusters are shifting
in abundance as a function of water vapor concentration, leading to
differences in collisional dissociation efficiency and proton-abstraction
efficiency. This is inconsistent with previous quadrupole acetate CIMS
experiments that indicate no humidity dependence for formic acid (Veres et
al., 2008). However, the ions most susceptible to humidity effects are the
ion-neutral clusters; these species are rarely detected because of the
operation of the API on the HR-TOF-CIMS in a declustered mode and the use of
a CDC on quadrupole instruments. As such, the cluster distribution in the
Veres et al. (2008) study may be completely different than the cluster
distribution observed here, making comparison between these instruments and
relative humidity effects nebulous. Collisional dissociation both simplifies
the observed mass spectrum and eliminates the observation of acetate-containing ion-neutral clusters. Effective collisional dissociation is the
key to predominantly detecting proton-abstraction reaction products and
maintaining the level of selectivity desired with chemical ionization.
Ambient detection of IEPOX and ISOPOOH using acetate clusters
(Budisulistiorini et al., 2015) will likely suffer from severe humidity
effects leading to large changes in sensitivity.
Similar humidity dependences are observed with iodide adduct CIMS (Lee et
al., 2014). Acetate CIMS may be simpler because the sensitivities for the
deprotonated-declustered ions follow approximately the same trend at a given
voltage configuration in the API. In contrast, species clustered with iodide
exhibit different RH dependences in both magnitude and shape for iodide
adduct CIMS. We observe similarly complex RH dependencies in acetate and
iodide reagent ions when run in a clustering mode (Fig. 7). The observation
of carboxylic acids clustering with water and other ions has been observed
using quadrupole instruments with atmospheric pressure ion sources (Viidanoja
et al., 1998).
The role of water on acetate CIMS chemistry remains difficult to reconcile.
Propionic acid sensitivities are the lowest under dry and very wet conditions
(Fig. 6), but other trends exist for other deprotonated-declustered acids
(Sect. S5.1). The formation of the water clusters, acetate clusters, and self-clusters
shows identical RH dependence for all the calibration compounds: the addition of
water shifts the cluster distribution as water is incorporated into acetate
clusters. Additionally, normalization methods described herein do not
eliminate the relative humidity dependence. Normalization of acetate CIMS
data remains a challenge because the information about the cluster
distribution is lost when collisional dissociation is sufficiently high that
we observe only declustered, deprotonated ions. Relative-humidity-dependent
calibrations may be the most direct method for rigorously addressing the
water interference.
Controlling for clustering reactions by operating the API on the HR-TOF-CIMS
under declustered settings is obvious, but the API voltage configurations do
not exist as a binary system of clustered and declustered operation,
making the choice of voltages a balancing act. The data presented herein
indicate that operating with an acetate cluster ratio of ∼1×10-2 is sufficient to eliminate the contribution of all clusters with
carboxylic acids directly investigated here under ambient conditions. This
corresponds to a dV at component relation 5 of ∼ 15 V and a dV at
component relation 3 of ∼ 35 V. However, these exact voltage
differences may be instrument-dependent, making the acetate cluster ratio an
important operational parameter that should be reported when using acetate
CIMS as suggested in previous publications from this group (Brophy and
Farmer, 2015). The effect of acetate precursor and acetic anhydride concentrations on sensitivity warrants further study. Acetic anhydride is a
difficult reagent precursor to work with because it is difficult to quantify
the mass entering the Po-210 ionizer. It is even more difficult to constrain
how efficiently acetate is produced from the α-particle flux, and
comparison between instruments and ion sources remains nebulous.
Acetate CIMS requires significant declustering for ambient atmospheric
measurement. The sensitivity to propionic acid detected as propanoate is
maximized for component relation 5 at a dV of 10–12 V (Fig. 6), although
the LOD and relative contribution of each cluster vs. the
deprotonated-declustered species remains surprisingly high at these voltages
(see Sect. S5 for additional compounds, LODs, and sensitivity ratios).
Increasing the dV at component relation 5 causes the relative contribution of
each cluster to drop and decreases the sensitivity of the
deprotonated-declustered ions (Sect. S5.2). However, these high-voltage
differences lead to the formation of potential fragment ions at low m/z
(Fig. 8). We note that the average bulk parameters (O : C, H : C, number
of carbons, oxidation state) continually change as a function of applied voltage
difference (Fig. 10). The experiments calculating bulk parameters scan
component relation 5, and distinct changes appear at a dV of 30 V. This may
support the hypothesis that fragmentation is occurring because at
dV = 30 V, the average H : C increases while the average O : C
decreases, consistent with multifunctional oxygenate fragmentation.
Increasing the voltage difference between adjacent components decreases the
probability of transmitting a cluster through the API, although the clusters
will be detectable at sufficiently high concentrations as shown by the
calculated LODs (Sect. S5.3).
Chhabra et al. (2015) present a method to account for clustering, or adduct
formation, in a study of α-pinene and naphthalene oxidation products
using a PAM chamber and the acetate HR-TOF-CIMS. However, this method
underestimates the complexity of the clustering problem by assuming that
clustering reactions proceed only via adducts between acetate and a
non-clustered ion (R–H) forming acetate clusters, [R–H + acetate]-.
The acetate-cluster mechanism is the most dominant clustering mechanism for
the carboxylic acids studied in this work because there is a very high
concentration of acetate ions relative to any other species in the IMR.
Nitric acid, however, provides one exception to this hypothesis. The [nitric
acid + acetate]- cluster and deprotonated declustered nitrate ion
are less sensitive (i.e., contribute less to the mass spectrum total signal)
relative to the [nitrate + nitric acid] self-cluster under all voltage
differences and RH conditions (Sect. S5.2, Fig. S17). Thus, this approach
neglects self-clustering, water-clustering, and other cross-species
clustering reactions which occur, albeit to a lesser extent than acetate
clustering, in the system described herein. We evaluate the Chhabra et al.
methodology for accounting for acetate clusters (Sect. S6), and find that
propionic acid clustering with acetate is underestimated by a factor of
15.5–26 depending on RH. Formic acid is accurately addressed under dry
conditions, but is underestimated by a factor of 15 with the addition of
water. Butyric acid is underestimated by a factor of 17–26, and methacrylic
acid is underestimated by a factor of 5–12 depending on RH. The findings
described herein further emphasize the importance of accounting for RH
dependences of the reagent ion and thus cluster distribution.
We note two challenges in quantifying the impact of clustering on observed
bulk properties or mass. (1) The presence of self-clusters and clusters
formed with other ions present in the background spectrum during single-component calibrations suggests that complex mixtures will be impacted by
clustering from other species; for example, ambient formic acid may form
formate ions that cluster with other carboxylic acids. In situ standard
additions are one approach for identifying this problem of secondary chemical
ionization. (2) RH changes the ratio of clustered analyte to
deprotonated-declustered analyte (Fig. 7, Sect. S5.2), further complicating
quantification in ambient field measurements. However, controlling cluster
interference in observed mass spectra by collisional dissociation is a
straightforward approach to the complexity of the acetate CIMS. Other
formulations proposed in the literature may oversimplify this problem.
Quantification of complex mixtures with the acetate CIMS is a complex problem.
Clustering is a key mechanism for abstracting protons from carboxylic acids.
Proton-abstracted declustered ions are predominantly observed if clusters are
collisionally dissociated during transmission from the IMR to the detector.
This suggests that some combination of both cluster binding energy and gas-phase acidity control the extent to which the analyte species retains a
charge upon declustering. The prevalence of cross-clustering reactions also
demonstrates that secondary ion chemistry is occurring to an appreciable
extent. The challenge of quantification when sensitivity varies by both
analyte and RH may be further complicated by IMR design and ion transmission
through the ion optics. With all these considerations, it is remarkable that
such good agreement has been found between acetate CIMS measurements and
aerosol mass spectrometer data, suggesting that despite the complexities and
unknowns, the acetate CIMS captures an important fraction of the gas-phase
chemistry relevant to secondary organic aerosol production and evolution
(Aljawhary et al., 2013; Chhabra et al., 2015; Lopez-Hilfiker et al., 2016).
Conclusions
Non-targeted analysis using HR-TOF-CIMS with no pre-separation is
challenging, but remains a promising technique to understand atmospherically
relevant species at low (< 1 ppb) concentrations. We characterize
numerous operational parameters using authentic standard calibrations that
drastically improve our ability to understand and interpret the acetate CIMS
mass spectrum. Tuning the HR-TOF-CIMS to a declustered mode in which the
acetate ratio is ≤1×10-2 eliminates the clusters formed in the
IMR. Further, we investigate the efficiency of declustering by applying the
voltage scanning tools described herein to a complex mixture of α-pinene oxidation products. These tools provide a convenient approach to
identify whether alkanoic acid signals, for example, are solely due to the
organic acid, or also the product of clustering or fragmentation reactions.
Iodide adduct and nitrate adduct CIMS may also benefit from routinely
operating in a voltage scanning mode for non-targeted analysis. The iodide
CIMS mass spectrum contains a poorly understood region that is separated in
mass defect space from the iodide-cluster region by the “iodide valley”
(Lee et al., 2014). This region is thought to contain peroxy acids
(R-C(O)OOH) which appear as carboxylic acids upon increasing the applied
voltage difference in the API. Thus, under normal iodide adduct CIMS
operation, species in this region will exist as a complex mixture of
ion-neutral clusters without iodide. Upon declustering, the iodide adducts
will fall apart along with any of the non-iodide-containing ion-neutral
clusters observed in the more positive mass defect region. This would
provide an additional set of information that can be compared to the results
obtained in a clustered mode where only the iodide-containing clusters are
evaluated. Lastly, if the lessons learned here about acetate CIMS apply to
deprotonated-declustered anions in general, one may decrease the RH
dependence observed with the iodide adducts by operating in a declustered
mode and examining declustered species.
The API characterization presented herein may impact the analysis of
atmospheric ions and new-particle formation under both ambient and
laboratory conditions, such as the Cosmics Leaving OUtdoor Droplets (CLOUD)
facility. Recent publications detailing CLOUD chamber measurements show
stable clusters containing up to 17 sulfuric acid molecules clustered with
other species (Schobesberger et al., 2015). The authors note that water is
absent from most observed clusters due to evaporation inside the API-TOF, and
that other species may also fragment (Olenius et al., 2013). The literature
surrounding the API-TOF further acknowledges that declustering inside the
instrument is poorly understood, and that fragmentation is highly related to
instrument settings (Ehn et al., 2011; Junninen et al., 2010; Olenius et al.,
2013). The scanning procedures presented herein may be of particular use to
API-TOF instruments, in determining both the strength of these clusters and
the API control/bias on observed cluster size and composition.
The observed mass spectrum acquired using the acetate CIMS is the combined result
of CI occurring in the IMR and declustering occurring throughout the
instrument. Ignoring clustering will result in either an over- or
underestimation of the average H : C ratio, O : C ratio, average
oxidation state, and average number of carbons depending on the extent of
clustering. Clustering is efficiently controlled using API component
relations, and clusters can be identified using nonlinear least-squares
sigmoidal regression and dV50 detection. The techniques and
considerations described herein will be relevant for a wide variety of
API-TOF users.