Chemical ionization mass spectrometry utilizing ammonium ions (NH + 4 CIMS) for measurements of organic compounds in the atmosphere

. We describe the characterization and ﬁeld deployment of chemical ionization mass spectrometry (CIMS) using a recently developed focusing ion-molecule reactor (FIMR) and ammonium–water cluster (NH + 4 · H 2 O) as the reagent ion (denoted as NH + 4 CIMS). We show that NH + 4 · H 2 O is a highly versatile reagent ion for measurements of a wide range of oxygenated organic compounds. The major product ion


Introduction
Quantifying atmospheric volatile organic compounds (VOCs) and their oxidation products is critical for understanding the formation of ozone (O 3 ) and organic aerosol (OA).However, this objective has been a long-standing challenge because of the sheer number and significant chemical complexity of organic compounds in the atmosphere Published by Copernicus Publications on behalf of the European Geosciences Union.
L. Xu et al.: Development and deployment of NH 4 + CIMS (Goldstein and Galbally, 2007).Chemical ionization mass spectrometry (CIMS) is a widely used and rapidly developing technique to characterize atmospheric trace gases.The advantages of CIMS include fast time response, high selectivity and sensitivity, and detection linearity over a wide range of analyte mixing ratios.In CIMS, the analytes are ionized via ion-molecule reactions with a reagent ion, which is soft and largely preserves the identity of the analytes.
The detection capability of CIMS depends on the selection of reagent ions, which are sensitive to different classes of organics.The commonly employed reagent ions include H 3 O + to detect reduced and small functionalized VOCs (de Gouw and Warneke, 2007), I − to detect inorganics and polar and acidic organics (B.H. Lee et al., 2014;Robinson et al., 2022), CF 3 O − to detect organic peroxides and other multifunctional organics (Crounse et al., 2006;Xu et al., 2020), SF − 6 to detect organic acids (Nah et al., 2018), NO − 3 to detect highly oxygenated molecules (Ehn et al., 2014), and protonated amines to detect reactive radicals (Berndt et al., 2018).Exploring novel reagent ions is an active research area to expand the detection capability of CIMS and to provide precise measurements of atmospheric species with high sensitivity.These efforts enable a comprehensive description of the complex mixture of atmospheric organic compounds.
One ionization scheme under active development utilizes the ammonium ion (NH + 4 ) chemistry.Several recent studies have demonstrated its capability to detect a range of oxygenated organic compounds, including alcohols, aldehydes, ketones, and even the short-lived peroxy radicals (RO 2 ) (Blake et al., 2006;Lindinger et al., 1998;Canaval et al., 2019;Hansel et al., 2018;Müller et al., 2020;Zaytsev et al., 2019;Berndt et al., 2018;Khare et al., 2022).One reason NH + 4 chemistry is attractive is that it detects oxygenated organic compounds in the positive mass spectrometer mode, in contrast to existing reagent ions (i.e., I − , CF 3 O − , and NO − 3 ) which are operated in negative mode.This offers the potential to rapidly switch between NH + 4 and H 3 O + within the same instrument to detect both oxygenated and reduced organic compounds, respectively, without substantial alteration of the electric fields in the mass spectrometer.Zaytsev et al. (2019) and Müller et al. (2020) demonstrated the feasibility of such rapid switching in laboratory conditions.The application of NH + 4 CIMS in recent studies has largely focused on laboratory studies (Berndt et al., 2018;Zaytsev et al., 2019), but its deployment in field measurements and intercomparison with other analytical instruments are scarce (Khare et al., 2022).
The instrument design, including the ion source and the ion-molecule reactor (IMR), differs between studies.Hansel et al. (2018) applied NH + 4 ion chemistry in a PTR3 instrument (Breitenlechner et al., 2017) (i.e., NH + 4 -PTR3) and detected peroxy radicals and other products from cyclohexene ozonolysis with sensitivities up to 2.8 × 10 4 cps ppbv −1 in a free-jet flow system.Using a similar instrument, Zaytsev et al. (2019) calibrated 16 compounds, with a maximum sensitivity of 8.9 × 10 4 cps ppbv −1 for decanone.In both studies, the major reagent ion is NH + 4 •H 2 O, generated in a corona discharge ion source from a mixture of NH 3 and H 2 O gas.Later, Müller et al. (2020) developed a method to produce NH + 4 using a mixture of water vapor and nitrogen in a hollow cathode glow discharge ion source, which is used in PTR-MS instruments with a traditional drift tube design that includes extraction plates between the hollow cathode ion source and drift tube.Canaval et al. (2019) used a selective reagent ionization time-of-flight mass spectrometer (SRI-ToF-MS) to produce NH + 4 via reaction of He + and gas NH 3 .Different instrument designs affect the distribution of reagent ions (i.e., NH + 4 vs.NH + 4 • H 2 O vs. NH + 4 • NH 3 ), detection efficiency, and sensitivity.
In this study, we describe the performance of NH + 4 CIMS using a Tofwerk Vocus long time-of-flight mass spectrometer (Krechmer et al., 2018).We investigate the impacts of instrument conditions on the distribution of reagent ions and the instrumental sensitivities of 60 analytes from several chemical functional classes.Building upon extensive calibrations, we explore the dependence of sensitivity on the ion-molecule reaction rate constant and the binding energy of the analyte-NH + 4 cluster, aiming to derive a relationship to approximate the sensitivity of analytes for which no calibration standards exist.Further, this instrument was deployed during the RE-CAP campaign (Re-Evaluating the Chemistry of Air Pollutants in California) in Pasadena, California, during the summer of 2021.The instrument performance is further evaluated by comparison to several co-located mass spectrometers.

Instrument description
The instrument in this work is based on the Tofwerk Vocus, which utilizes a new ion source, a focusing ion-molecule reactor (FIMR), and a long time-of-flight mass spectrometer (LToF).A detailed description of the Vocus can be found in Krechmer et al. (2018).Here we briefly summarize the generation of reagent ions and instrument operation conditions.
The chemical ionization gas entering the ion source is produced by mixing NH 3 and H 2 O from two streams: a 20 sccm flow of water vapor from the headspace of a liquid water reservoir (denoted as H 2 O flow) and an additional 1 sccm from the headspace of a reservoir containing 0.5 % (vol %) ammonium hydroxide water solution (denoted as NH 3 flow, which contains both NH 3 and H 2 O).The ion source consists of two conical surfaces with a voltage gradient.A plasma is produced between the conical surfaces, which primarily ionizes water molecules, producing H 3 O + .The discharge current is regulated at 2.0 mA.Because NH 3 has a larger proton affinity than H 2 O, the proton transfer reaction (Reaction R1)  and H 2 O, n = 0, 1, 2) and H 3 O + • (H 2 O) n , are generated from the ion source and can potentially serve as reagent ions.
The reagent gas flow pushes the ions into the FIMR where they subsequently react with analytes.Sample air enters the FIMR through a 25 mm long PEEK capillary (ID 0.18 mm).The sample flow rate is ∼ 100 sccm at FIMR pressure of 3 mbar in this study.The FIMR is a 100 mm long glass tube with an inner diameter of 10 mm.A quadrupole radio frequency (RF) field is applied to the FIMR to collimate ions into a narrow beam, significantly enhancing the sensitivity (Krechmer et al., 2018).The FIMR conditions, including temperature, pressure, drift voltage, and the ratio of NH 3 to H 2 O into the ion source, all control the degree of cluster-ion formation, the distribution of reagent ions, and ultimately the sensitivity, as will be discussed in Sect.3.2 and 3.3.Ions from the FIMR travel through a big segmented quadrupole (BSQ).The BSQ serves as a high-pass band filter to reduce the signal intensity of reagent ions while simultaneously guiding ions into the time-of-flight mass spectrometer.As a result of this filtering, the observed distribution of reagent ions is not the same as the actual distribution in the FIMR (Krechmer et al., 2018).After the BSQ, the ions travel through the primary beam region and are eventually detected by the long time-of-flight mass spectrometer with a mass resolution (full width at half maximum, FWHM) up to 8000 at m/Q 100.The extraction frequency of the ToF is set at 17.5 kHz.

Laboratory characterization
We calibrate the instrumental sensitivities (cps ppbv −1 ) of 60 organic compounds (Table 1) using two methods, standard gas cylinders (SGC) and a home-built liquid calibration unit using either water or hexane as solvent (LCU-W and LCU-H), as described in Sect.S1.1 in the Supplement.We find minimal dependence of sensitivity on sample relative humidity (RH) as shown in Fig. S1 in the Supplement, consistent with Khare et al. (2022) and observations made when running the Vocus in H 3 O + mode (Krechmer et al., 2018).This is mainly because a large amount of water vapor (20 sccm) is deliberately added to the FIMR.As an example, the water amount in a 100 sccm ambient sample under 25 • C and 100 % RH is only 15 % of the added 20 sccm water va-por to the FIMR, assuming no water vapor loss in both processes.The instrument background is determined by passing ambient air through a platinum catalytic converter heated to 400 • C. The detection limit is defined as 3 standard deviations of measurement background for 1 s integration time.
During transport, ions get lost in the BSQ, in the ion guides, and in the extraction region of the ToF.We quantify the mass-dependent transmission efficiency relative to the reagent ion NH + 4 • H 2 O by introducing a series of compounds spanning a range of molecular weight (32-370 m/Q) in a large enough quantity to deplete the fraction of reagent ions by ∼ 20 %-30 % (Huey et al., 1995;Heinritzi et al., 2016).The ratio of the increase in the product ions to the decrease in the reagent ion indicates the relative transmission efficiency between these two masses.A detailed derivation can be found in Sect.S1.2.
We have performed laboratory tests and measured the product distribution of 60 organic compounds.The product ions are identified by sampling the headspace of a small vial containing pure analyte.Distance is kept between the instrument inlet and the vial to keep analyte concentration low.Ions correlating with the parent ion (NH + 4 • A) with r 2 larger than 0.95 and accounting for larger than 1 % of the parent ion signal are considered to be product ions from the analyte.The distribution of product ions depends on the distribution of reagent ions.In this test, we maintain the ratio between 5 and 20.Under this condition, the ion chemistry of H 3 O + • (H 2 O) n is negligible.To probe the stability of product ions, we performed voltage scanning tests following the procedure outlined in Lopez-Hilfiker et al. (2016) and Zaytsev et al. (2019).In brief, we vary the voltage gradient ( V ) between FIMR back and skimmer while keeping the voltage gradient between FIMR front and back constant.A larger V increases the collisional energy, causes stronger collision-induced dissociation of product ions, and tends to decrease the signal of product ions.We define V 50 as the voltage gradient at which the parent ion NH + 4 •A signal drops to half of the maximum signal.This V 50 represents the electric field required to break each NH + 4 • A and is therefore related to the binding energy of NH + 4 • A. Further, V 50 is converted to the kinetic energy of NH + 4 •A in the center of mass (i.e., KE cm,50 ) using a parameterization of mass-dependent ion mobility (Zaytsev et al., 2019, and details in the Sect.S1.3).KE cm,50 is a measure of the NH + 4 • A stability.

Field deployment
The NH + 4 CIMS was deployed during the RECAP campaign (Re-Evaluating the Chemistry of Air Pollutants in California) in Pasadena, California, from August-September, 2021.The ground sampling site is located on the campus of the California Institute of Technology, which is only one block away from the original sampling site during the 2010 CalNex study (Ryerson et al., 2013) on a tower 10 m above the ground.The instrument was operated to sample gas phase from 10 to 19 August.Later, the instrument was coupled to a Vocus Inlet for Aerosol (VIA) to automatically switch sampling between gas and particle phases.This study will focus on the gas-phase sampling period.Co-located instruments of relevance to this study include a proton-transfer-reaction mass spectrometer (PTR-MS) (Yuan et al., 2016;de Gouw and Warneke, 2007) and a gas chromatography mass spectrometer (GC-MS) (Lerner et al., 2017).A CF 3 O − chemical ionization mass spectrometer (CF 3 O − CIMS) (Crounse et al., 2006;Allen et al., 2022) was deployed at a different site on campus, which is ∼ 800 m away from the NH + 4 CIMS.
3 Instrument performance

Overview of ion chemistry
The target primary reagent ion is NH + 4 • H 2 O, which ionizes analytes (A) primarily via ligand-switching reactions (Reaction R5) to form product ion NH + 4 • A. As analogous to proton affinity, we define NH + 4 affinity as the negative of the enthalpy change in the reaction between NH + 4 and an analyte.If an analyte has a larger NH + 4 affinity than H 2 O, Reaction (R5) is exothermic and will occur at a rate close to the collision limit when the difference in NH + 4 affinity is sufficiently large (Adams et al., 2003).Otherwise, the ligandswitching reaction is endothermic.The energy imparted via the drift voltage could aid the endothermic reaction to overcome the energy barrier, but the instrument sensitivity in these instances is expected to be low.Besides the target pri- • NH 3 are also expected to be different, as the NH 3 has a larger NH + 4 affinity than H 2 O (i.e., 108 vs. 86 kJ mol −1 , NIST Chemistry WebBook).Therefore, the presence of multiple reagent ions will complicate the ionization chemistry and the interpretation of the mass spectra.To avoid such complication, the instrument conditions need to be carefully optimized to ensure NH + 4 • H 2 O exists as the dominant ion reacting with analytes.

Modeling the distribution of reagent ions
The distribution of the reagent ions is controlled by several factors, including the FIMR reduced electric field (E/N ), temperature (T ), pressure (P ), the H 2 O mixing ratio (χ H 2 O ), and the ratio of NH 3 to H 2 O (NH 3 /H 2 O).Many of these factors are interdependent -e.g., the E/N depends on pressure and temperature.To unravel the influences of these factors on the distribution of reagent ions, we develop a kinetic model.The model includes a series of reactions between two ions (NH + 4 and H 3 O + ) and two neutral molecules (NH 3 and H 2 O).Clusters containing up to three molecules are considered, which leads to a total of 14 different ion clusters (Fig. S4).The ion-molecule cluster reaction rate constant (i.e., forward reaction with k forward ) is calculated using the parameterization in Su (1994), assuming the reaction proceeds at the collision limit.The reaction rate constant of the declustering reaction (i.e., reverse reaction with k reverse ) is calculated using k forward and the equilibrium constant K eq .k reverse for Reaction (R6), for example, is expressed by Eq. ( 1), where M 0 represents the number density (cm −3 ) under standard condition and K eq represents the reaction equilibrium constant.K eq is calculated using Eq. ( 2), where H 0 and S 0 represent the enthalpy and entropy changes of the reaction at standard condition, respectively (Table S1  (3) (de Gouw et al., 2003), where k B is the Boltzmann constant, m I + , m A , and m buffer are the masses of the ion I + , the neutral analyte A, and the buffer gas, respectively, and the ν d is the drift velocity of ion IA + .ν d is calculated using Eq. ( 4), where µ 0 is the reduced mobility of IA + calculated based on the parameterization in Steiner et al. (2014), P and T are the FIMR pressure and temperature, respectively, and E is the electric field strength across the FIMR. (3) The influences of different FIMR conditions (i.e., E/N , T , P , χ H 2 O , and NH 3 /H 2 O) on the distribution of reagent ions are intertwined.To visualize their impacts, we first conduct simulations covering wide ranges of all five factors to locate the condition yielding the largest fraction of NH . The optimized condition is E/N = 60 Td (Townsend), T = 330 K, P = 5 mbar, χ H 2 O = 0.25, and NH 3 /H 2 O = 0.1 %.Then, we conduct simulations using the optimal condition as a start point and vary one factor at a time while holding the other four constant to investigate the impact of each factor on the distribution.
The simulation results are shown in Fig. 1. Figure 1a shows that the reduced electric field (E/N ) strongly impacts the distribution of reagent ions.• NH 3 .Evaluation of the kinetic simulation results by experimental observations is desirable but challenging.One challenge is that the distribution of reagent ions cannot be measured because the BSQ serves as a high-pass band filter which reduces the signal intensity of reagent ions.Another challenge is that voltages in the ion transfer region between the drift tube and the mass analyzer can change the distribution of reagent ions, which causes the measured distribution to be different from that in the FIMR (Krechmer et al., 2018;Breitenlechner et al., 2022;Yuan et al., 2016).Overall, the simulation results illustrate the controlling effects of FIMR conditions on the distribution of reagent ions.The determination of FIMR conditions is eventually based on experimental calibration of instrumental sensitivity, which can be guided by the modeled distribution of reagent ions, as discussed in the next section.

Dependence of sensitivities on FIMR conditions
While the above section modeled the dependence of the distribution of reagent ions on FIMR conditions, in this section we experimentally evaluate the dependence of analyte sensitivities on FIMR conditions.The analyte sensitivity depends not only on the distribution of reagent ions, but also other factors, including the number density of analytes in the FIMR, ion-molecular reaction time, stability of the product ion, and the transmission efficiency of product ions, as discussed below.Similar to the analysis in kinetic modeling, we experimentally vary one factor while holding the others constant.
Figure 2a shows the impacts of E/N on sensitivities of representative analytes.The E/N is varied by ramping the FIMR front voltage from 100 to 600 V, while holding the FIMR back voltage at 5 V.Under FIMR pressure and temperature of 3 mbar and 313 K, respectively, the E/N ranges from 13 to 83 Td.The dependence of sensitivities on E/N follows a similar trend of the modeled distribution of NH + 4 • H 2 O (Fig. 1a).The sensitivities initially increase with increasing E/N , partly because of more reagent ion 4 , so less NH + 4 • H 2 O causes a decrease in sensitivities.Besides changing the distribution of the reagent ions, changing E/N influences the sensitivity via other mechanisms, including the extent of declustering of NH + 4 •A and the focusing effect of ions in the FIMR.Krechmer et al. (2018) show that the higher E/N better focuses ions to the central axis of the reactor and increases the sensitivity.This may ex- plain the uptick in sensitivities when E/N increases from 80 to 90 Td, which is not observed in the modeled NH + 4 • H 2 O. Overall, the observed dependence of sensitivities on E/N is a superposition of at least three effects: focusing effects and the extent of declustering of both reagent ions and product ions.
The effects of FIMR pressure on sensitivities are shown in Fig. 2b.The sensitivities exhibit a non-monotonic dependence on FIMR pressure, in a similar manner as the reagent ion NH + 4 • H 2 O does (Fig. 1b), suggesting the pressuredependent sensitivities are related to the pressure-dependent distribution of reagent ions.In addition, higher pressure increases the number density of analyte molecules in the FIMR, which tends to increase the sensitivity.However, this effect is smaller than the effect of changing reagent ion on sensitivities, as Fig. 2b shows that the sensitivities decrease with increasing pressure beyond 3 mbar.
The effects of FIMR temperature on sensitivities are shown in Fig. 2c.Among the seven compounds tested here, the sensitivities of six oxygenated compounds exhibit a negative dependence on the temperature between 310 and 370 K.The reduced VOC, isoprene, exhibits a positive dependence.Similar to isoprene, α-pinene sensitivity also increases with temperature in the 303-350 K window as recently reported in Khare et al. (2022).Here we examine the opposite trends of temperature-dependent sensitivity between acetone and α-pinene because their NH + 4 affinities are available in the literature (Sect.S4).α-Pinene has an NH + 4 affinity smaller than that of H 2 O (i.e., 75 vs.86 kJ mol −1 from Canaval et al., 2019), resulting in the ligand-switching reaction between α-pinene and NH + 4 • H 2 O being endothermic.Therefore, the reaction is promoted under higher temperature, which enhances the sensitivity.In contrast, the ligand-switching reaction between acetone and NH  acetone has a larger NH + 4 affinity than H 2 O (i.e., 110 vs. 86 kJ mol −1 from Canaval et al., 2019).For exothermic reactions ( H is negative), higher temperature leads to smaller K eq (Eq.2), a smaller k forward /k reverse ratio, and hence lower sensitivity.To better understand the temperature-dependent sensitivities, we add the reversible reactions of acetone and α-pinene with NH + 4 • H 2 O to the kinetic model depicted in Fig. S4 and simulate the dependence of their sensitivities on temperature.As shown in Fig. S6, the model can reproduce the observed dependence of their sensitivities on temperature.The NH + 4 affinity of isoprene is not available, but it is expected to be even smaller than α-pinene, given that the isoprene sensitivity is 10 times smaller than that of αpinene.Thus, the reaction between isoprene and NH + 4 • H 2 O is likely also endothermic, causing the increasing sensitivity with higher temperature as shown in Fig. 2c.
The effects of the NH 3 /H 2 O ratio on sensitivities are experimentally tested by simultaneously varying the flow rates of NH 3 and H 2 O, while keeping the total flow rate constant.Because the NH 3 flow is a mixture of NH  1e).Taking acetone as an example, its NH + 4 affinity (110 kJ mol −1 ) is higher than that of H 2 O (86 kJ mol −1 ) but close to that of NH 3 (108 kJ mol −1 ).As a result, the ligand-switching reac- • H 2 O ratio change over time owing to the aging effects within the solution that supplies NH 3 .In the current approach to supply the chemical ionization gas, the NH 3 /H 2 O ratio is controlled by the combination of the concentration of ammonium hydroxide aqueous solution and flow rates from the water and ammonium hydroxide reservoirs.Because NH 3 is more volatile than H 2 O, the concentration of the ammonium hydroxide water solution decrease over time, resulting in a decreasing trend of NH 3 /H 2 O over a timescale of weeks.In addition, the temperature variation of the ammonium hydroxide water solution changes the partitioning of NH 3 and hence the NH 3 /H 2 O ratio.One approach to compensate for the NH 3 loss is to adjust the flow rate from the ammonium hydroxide reservoir to maintain a relatively constant NH The optimal FIMR conditions should be explored collectively and systematically.The optimal condition for our instrument is FIMR drift voltage 55 Td, 3 mbar, 40 • C, 1 sccm from 0.5 % ammonium hydroxide aqueous solution, and 20 sccm water vapor.A temperature value that is slightly higher than ambient temperature is chosen for control purposes.

Product distributions from the ion-molecule reactions
The desired reagent ion is NH + 4 • H 2 O and the desired ionmolecule reaction is the ligand-switching reaction between NH + 4 •H 2 O and analyte A, which produces cluster NH + 4 •A as the parent ion (Reaction R5).However, the presence of several reagent ions in the FIMR and the declustering of NH + 4 •A in the electric field induce a variety of reactions and cause complex product distributions.Besides the target parent ion NH + 4 • A, we observe the protonated product (AH + ), analyte clusters (NH , and fragmentation products.The potential ion-molecule reactions and product ions can be generally expressed by Reactions (R7) and (R8).
Figure 4 shows the product distributions for all tested analytes grouped by their chemical class.The analyte sensitivities are represented by the circle size in the figure.Among all classes, acids, ketones, and nitriles have the most desirable product distribution, in which the fraction of parent ion NH + 4 • A in all product ions (denoted as f NH + 4 •A ) is more than 90 %, with the exception of acetic acid.For 2-octanone and 2-nonanone, NH + 4 • A is the sole product ion.For the alcohols, the product distribution is diverse.2-Propanol and 2butanol have fragmentation products (NH + 4 • A − 2H), which account for ∼ 5 % of the total products, but the fragmentation mechanism is unclear.For the aldehydes, the NH + 4 • A generally accounts for more than 80 % of total product ions.The f NH + 4 •A tends to increase with larger molecules, for example when comparing a homologous series of aldehydes (pentanal, hexanal, heptanal, octanal, and nonanal).Four monoterpenes studied here produce a significant amount of protonated product (AH + ), which is comparable to that of NH + 4 • A. The causes of the product distributions of four monoterpenes are possibly explained by their proton affinity and NH + 4 affinity.Three monoterpenes including α-pinene, β-pinene, and camphene have smaller NH + 4 affinities than H 2 O (Table S2).Thus, their ligand-switching reactions with NH + 4 • H 2 O are endothermic and the production of NH + 4 • A is likely aided by the energetic collision energy imparted by the drift voltage.These three monoterpenes have higher proton affinity than NH 3 (Table S2) so that NH + 4 • A can undergo internal proton transfer to produce AH + • NH 3 , which breaks in the electric field and produces AH + .In contrast to the above three monoterpenes, limonene has larger NH + 4 affinity than H 2 O and smaller proton affinity than NH 3 (Table S2).Thus, the ligand-switching reaction with NH •NH 3 are not observed for any compound.Overall, the product distribution is complicated and caution is required in quantification.

Constraining the sensitivity
Because of a lack of calibration standards, the NH + 4 CIMS sensitivities to the majority of routinely detected multifunctional organic compounds in the atmosphere are not quantifiable.We attempt to constrain the sensitivity by building upon the extensive calibration of organic compounds from various chemical classes in this study.The observed instrument sen- Using this integral to represent the product ion formation is only valid when the ion-molecule reaction is in the kinetic-limited regime.In the thermodynamic regime, both forward and reverse ion-molecule reactions need to be considered.
= parent ion formation × transmission efficiency ( 5) Under a constant instrumental condition, the [NH + 4 • H 2 O] and reaction time are fixed.The sensitivity of an analyte is determined by f NH + 4 •A , k, and TE.Among these three factors, f NH + 4 •A and k are more uncertain than TE.The value of k for exothermic ligand-switching reactions is close to the collisional limit (Adams et al., 2003), which can be calculated according to Su (1994) using the dipole moment and polarizability of the analyte (Table S3).f NH + 4 •A can be experimentally measured, and it is close to 1 for multifunctional organic compounds, as discussed in Sect.tics, is difficult to quantify.We assume the overall TE is represented by the product of m Q -dependent TE (denoted asTS m Q ) and binding-energy-dependent TE (denoted as TE B ) (Eq. 6).TS m Q represents the transmission efficiency through BSQ, the extraction region of the ToF, and other processes that are dependent on m Q .TS m Q is experimentally quantified as described in Sect.2.2.TE B accounts for the ion loss via collisioninduced dissociation caused by energy imparted by electric fields.TE B depends on the binding energy of the parent ion, as the parent ion with stronger bonds between analyte and NH + 4 have a larger chance to survive the electric fields and hence a larger TE B .The binding energy of NH + 4 • A is experimentally probed from the voltage scanning tests and is represented by the kinetic energy of NH + 4 • A in the center of mass (i.e., KE cm,50 ) (Sect.2.2).In this way, TE B is related to a measurable parameter KE cm,50 .The mathematical relationship between TE B and KE cm,50 , TE B = f (KE cm,50 ), is the final component to constrain the sensitivity.
We utilize the extensive calibration of 60 compounds from diverse chemical classes to derive the relationship between TE B and KE cm,50 .By rearranging Eqs. ( 5) and ( 6) and representing ] dt as a constant C, TE B can be expressed as Eq. ( 7), where S corr represents the sensitivity corrected for f NH + 4 •A , k, and TE m Q .Using Eq. ( 7), the relationship between TE B and KE cm,50 can be obtained through plotting S corr against KE cm,50 .As shown in Fig. 5, S corr exhibits a positive dependence on KE cm,50 .The relationship between S corr and KE cm,50 of the majority of compounds can be reasonably described using a Hill equation.Analytes with small KE cm,50 (i.e., < 0.15 eV) have very low sensitivity because of declustering of NH + 4 • A in the electric fields.As KE cm,50 increases, the sensitivity increases.This is because NH + 4 • A with a stronger bond between A and NH + 4 is more likely to survive the imparted energy from electric fields and hence more likely to be detected.When KE cm,50 exceeds a threshold (i.e., 0.35 eV), NH + 4 • A does not decluster in the electric field and is detected with maximum S corr .The maximum S corr is constrained using 2-hexanone here, but calibrations of analytes with KE cm,50 larger than 0.35 eV are warranted to constrain the maximum S corr .Such analytes tend to be large oxygenated organic compounds with low volatility, making their calibrations challenging.
A similar relationship between S corr and KE cm,50 has been reported in Zaytsev et al. (2019), who used NH + 4 -PTR3 and explored the relationship between the sensitivity and KE cm,50 for 16 compounds, 9 of which are ketones.Unlike this study, Zaytsev et al. (2019) did not normalize the sensitivity to the ion-molecule collision rate constant k.This is reasonable as the ion-molecule reaction time in NH + 4 -PTR is ∼ 3 ms, about 15 times longer than that in our instrument.The long reaction time results in an equilibrium between cluster formation and fragmentation in the IMR for many analytes.In this thermodynamic regime, the product ion formation is proportional to the equilibrium constant of Reaction (R6) (Iyer et al., 2016;Robinson et al., 2022) so that normalizing the sensitivity in NH + 4 -PTR3 by the equilibrium constant may improve the relationship between sensitivity and KE cm,50 in Zaytsev et al. (2019).
In this study, the relationship between S corr and KE cm,50 is largely defined by monofunctional organic compounds, but we anticipate that this relationship applies to organic compounds containing at least one functional group that binds strongly with NH + 4 , such as C=O, -OH, and nitrile.For example, five multifunctional compounds studied here (i.e., ethylene glycol, 1,3-propanediol, hydroxyacetone, 2,3butanedione, and 2,3-pentanedione) are well described by the fitted Hill equation.Because the fitted Hill equation does not apply to monocarboxylic acids, for reasons discussed later, the applicability of the relationship to multifunctional organic acids is uncertain and warrants future investigation.Moreover, we compare several structural isomers with monofunctional groups, including acetone vs. propanal, MACR vs. MVK, C5-C9 mono-aldehyde vs. mono-ketones.Despite the difference in S corr between isomers, their S corr and KE cm,50 follow the same relationship.
The relationship between S corr and KE cm,50 depicted in Fig. 5 provides an effective approach to estimate the sensitivity of the NH + 4 CIMS to a suite of oxygenated organic compounds.The KE cm,50 can be calculated from the voltage scan tests.TE m Q can be experimentally quantified following the procedure in Sect.2.2.f NH + 4 •A is unknown, but it is close to 1 for multifunctional organic compounds, as discussed in Sect.3.4.The k is also unknown, but it can either be calculated (Su, 1994) or reasonably estimated based on the molecular mass, elemental composition, and functional group (Sekimoto et al., 2017).k is generally on the order of 10 −9 cm 3 molec.−1 s −1 (Fig. S7).Finally, based on the four abovementioned parameters, the sensitivity can be estimated.
The observed relationship between S corr and KE cm,50 in Fig. 5 has limitations.First, it is only applicable to analytes for which the ligand-switching reaction with NH + 4 • H 2 O is exothermic.This arises from approximating the ionmolecule reaction rate constant (k) in S corr using the collisional limiting rate constant.This approximation is not valid for endothermic reactions, which occur at a slower rate.This likely explains why several compounds, including monocarboxylic acids, some monoterpenes, reduced aromatics, isoprene, and 2-methylfuran, are outliers in Fig. 5.For example, the NH + 4 affinity of acetic acid is estimated to be lower than H 2 O (Sect.S5).Two monoterpenes, limonene and α-pinene, do not follow the fitted line, but the behaviors of monoterpenes are more complicated.The calculated NH + 4 affinities of β-pinene and camphene are smaller than that of H 2 O (Table S2), causing their ligand-switching reactions to be endothermic, but they fall on the fitted Hill equation.In contrast, limonene has larger NH + 4 affinity than H 2 O, but it is lower than the fitted line.The reason for such different behavior is unknown but might be related to their structural difference.For example, β-pinene and camphene have an external C=C bond connected to the six-member ring, but α-pinene and limonene do not.Another limitation is that KE cm,50 , which is calculated from V 50 based on voltage scan, may not be a proper proxy for NH + 4 affinity for some analytes.For example, α-pinene has similar NH + 4 affinity as β-pinene and camphene (Canaval et al., 2019), but the voltage scan test shows that α-pinene has a larger KE cm,50 than the other two (Fig. 5).Another exception is that isoprene and 2-methylfuran are expected to have small NH + 4 affinity considering their low sensitivities, but their KE cm,50 is the highest among all analytes studied here.Similar "false positive" behavior (i.e., large KE cm,50 or binding energy, but low sensitivity) is also observed in the I − CIMS (Iyer et al., 2016).We suspect the voltage scanning affects not only the collisional energy of the NH + 4 • A, but also the ion-molecule chemistry or ion transmission via some unknown mechanisms.In the voltage scan, the FIMR front voltage is increased simultaneously with FIMR back voltage to keep the upstream voltage gradient constant.It is generally assumed that the absolute voltages do not affect the ion-molecule chemistry and transmission, as long as the voltage gradient is constant, but this assumption may not be valid.For example, in the voltage scan, we observe that the signal of reagent ion becomes noisy when the FIMR front voltage (450 V) is close to the ion source voltage (440 V), suggesting that the FIMR front voltage affects the ion transmission from the ion source into the FIMR.

Comparison of sensitivities between instruments
In this section, we compare the sensitivities of our NH Figure 6 shows the sensitivity ratio of a selected instrument (S i ) to the NOAA NH + 4 CIMS (S NH + 4 CIMS ) for a number of analytes grouped by their chemical class.Khare NH + 4 CIMS used the same ion source and IMR as NOAA NH + 4 CIMS, but the sensitivities are generally lower than NOAA NH + 4 CIMS by a factor of 5.In particular, the ethylene glycol sensitivity is lower by a factor of 100.The lower sensitivity in Khare et al. (2022) is likely because they used a higher NH 3 /H 2 O ratio than this study.Khare et al. (2022) used 1 sccm vapor from a 1 % ammonium hydroxide solu-tion, while this study used 1 sccm vapor from a 0.5 % solution.As discussed in Sect.3.2, a larger NH 3 /H 2 O ratio leads to a larger fraction of NH + 4 •NH 3 in the total reagent ions and hence reduced sensitivity for most analytes (Fig. 3).The sensitivities in Khare et al. (2022)  The sensitivity ratio of NOAA H 3 O + CIMS to NOAA NH + 4 CIMS spans a wide range from 1 to 10 4 .In general, the sensitivity ratio anti-correlates with the sensitivity of NOAA NH + 4 CIMS within each chemical class.This trend is the most evident for aromatics.For example, for reduced aromatics, the sensitivities of which are smaller than 2 cps ppbv −1 in the NOAA NH + 4 CIMS, their sensitivities are 10 3 higher in the NOAA H 3 O + CIMS.However, for oxygenated aromatics, such as benzaldehyde and furfural, the sensitivities of which are on the order of 10 3 cps ppbv  (Pagonis et al., 2019;Yuan et al., 2017;Sekimoto et al., 2017), which simplifies the interpretation of the mass spectra.
Using the same NH + 4 • H 2 O chemistry, the PTR3 sensitivities are overall 20 times higher than those of NOAA NH + 4 CIMS.This difference is mainly due to different designs of the IMR and ion source.The PTR3 utilized a tripole electrode as the IMR (Breitenlechner et al., 2017).This design enables the IMR to be operated at 60 mbar and 3 ms reaction time (Zaytsev et al., 2019), which are much higher than 3 mbar and 0.2 ms in NOAA NH + 4 CIMS, and leads to enhanced sensitivities.The NOAA NH + 4 CIMS utilizes a lowpressure discharge ion source, which generates more ions than the corona discharge ion source in the PTR3.This compensates for the effects of the lower IMR pressure and short reaction on sensitivity to some extent.The combined influences of ion source, IMR pressure, and reaction time result in the difference in sensitivities between NOAA NH + 4 CIMS and PTR3 NH + 4 CIMS.Despite lower sensitivities, one advantage of the NOAA NH + 4 CIMS is that its sensitivities have a much smaller dependence on the sample relative humidity that the PTR3 NH + 4 CIMS (Zaytsev et al., 2019).

Field deployment
The NH + 4 CIMS was deployed during the RECAP campaign in Pasadena, California, in August-September, 2021.Measurements presented in this section were made from 10 to 19 August when the instrument continuously sampled the gas phase.), and the ion-molecule reaction rate constant (k, 10 −9 cm 3 molec.s −1 ), as defined in Eq. ( 7).The solid line represents a fitting of analytes using a Hill equation.S corr = 1350/(1 + (0.267/KE cm ) 11 ).The dashed line represents a linear fitting for analytes with KE cm between 0.2 and 0.3 eV in a similar fashion done in Zaytsev et al. (2019).Organic acids, naphthalene, isoprene, 2-methyl furan, limonene, α-pinene, and styrene are excluded from both fittings.The ellipses represent the uncertainty range.

Measurement capability
Figure S8 uses a mass defect plot to illustrate the measurement capability of NH + 4 CIMS.In the RECAP campaign, a total of 288 ions have signals above the detection limit.Half of the ions have the formula C x H y N 1 O z (reagent ion included in the formula).These ions mostly represent the non-nitrogen-containing oxygenated organics cluster with NH + 4 or NH + 4 • H 2 O.A total of 70 ions have the formula C x H y N 2 O z , which likely represent nitrogen-containing compounds.This assignment is supported by the analysis of product distribution (Sect.3.4), which shows the product ion contains at most one nitrogen from the reagent ion.A total of 40 out of 288 ions have the formula C x H y O z , which likely represent analytes clustering with H + • (H 2 O) n (n = 0, 1, 2).

Instrument intercomparison
The co-located instruments in the RECAP campaign enable the evaluation of the field performance of NH + 4 CIMS.In this section, we compare the measurements of several important atmospheric species from different chemical classes by three NOAA mass spectrometers (i.e., NH + 4 CIMS, H 3 O + CIMS, GC-MS) and the Caltech CF 3 O − CIMS.For com-pounds that are commercially available, we calibrate the instrumental sensitivity and compare the mixing ratio.For multifunctional oxygenated organics that do not have calibration standards, raw signals are compared.If multiple isomers exist for a parent ion and if these isomers are quantified by GC-MS, we apply the GC-MS resolved isomer ratio and the sensitivities of individual isomers to convert the raw cps of the parent ion to the summed mixing ratio of all isomers for NH + 4 CIMS (Sect.S6).
To account for instrument variability, the ion signals are typically normalized to the changing reagent ion signals.However, previous studies using Vocus in H 3 O + and NH + 4 • H 2 O chemistry did not normalize the signals to reagent ions (Krechmer et al., 2018;Khare et al., 2022) because the BSQ serves as a high-pass band filter and substantially reduces the signal intensity of reagent ions.In this study, we find that without normalization, the comparisons between NH + 4 CIMS and GC-MS exhibit a significant difference between day and night (Fig. S9a and b), which is consistent with the diurnal trend of reagent ion NH + 4 • H 2 O (Fig. S9c).Normalization to the reagent ion signal largely eliminates this difference.In light of this observation, we normalize the ion signals to that + CIMS

Carbonyls
Figure 7c shows the time series of acetone measured by NH + 4 CIMS, H 3 O + CIMS, and GC-MS.In NH + 4 CIMS, we attribute the NH + 4 •C 3 H 6 O solely to acetone and ignore the con-tribution from its structural isomer propanal because GC-MS shows that the propanal concentration is much lower than acetone and because the NH + 4 CIMS sensitivity to acetone is 10 times larger than propanal (1247 vs. 103 cps ppbv −1 ).Acetone concentrations measured by the three instruments

Figure 1 .
Figure 1.The dependence of modeled distribution of reagent ions on FIMR conditions.(a) E/N ; (b) P ; (c) T ; (d) H 2 O mixing ratio; (e) NH 3 /H 2 O.In each panel, the other four factors are held constant at the following conditions: E/N = 60 Td, P = 5 mbar, T = 330 K, H 2 O mixing ratio = 0.25, NH 3 /H 2 O = 0.1 %.Because the impacts of these factors are intertwined, each panel will change if the other four factors are at different values, as shown in an example in Fig. S5.

Figure 2 .
Figure 2. Dependence of instrument sensitivities of representative species on FIMR conditions.(a) E/N ; (b) P ; (c) T .The range of E/N in panel (a) is obtained by varying the drift voltage while maintaining the P and T at 3 mbar and 313 K, respectively.Analytes with sensitivities lower than 50 cps ppbv −1 are shown in dashed lines.The parent ion NH + 4 • A is used to quantify the sensitivity.
3 and H 2 O, the accurate flow rate of NH 3 is unknown.To approximate the NH 3 /H 2 O ratio, we use the observed ratio of NH + 4 •H 2 O/H 3 O + •H 2 O and NH + 4 •NH 3 /NH + 4 •H 2 O because these three ions have similar transmission efficiency and their relative abundance directly depends on the NH 3 /H 2 O ratio. Figure 3 shows the dependence of sensitivities of nearly 50 analytes on the NH + 4 • H 2 O/H 3 O + • H 2 O.For the majority of compounds, their sensitivities initially increase with NH + 4 • H 2 O/H 3 O + • H 2 O and then show a decreasing trend.This trend is caused by the fact that the initial increase in NH 3 /H 2 O favors the formation of NH + 4 • H 2 O and hence higher sensitivity, but high NH 3 /H 2 O produces more NH + 4 • NH 3 clusters, leading to reduced sensitivity (Fig.

Figure 3 .
Figure 3.The effects of reagent ion distribution on sensitivities of various organic species.The sensitivity of each analyte is normalized to the corresponding maximum value.Only analytes with sensitivity larger than 50 cps ppbv −1 are shown here.

Figure 4 .
Figure 4.The product distributions of analytes in the NH + 4 CIMS.The analytes are grouped by chemical class.Within each class, the analytes are sorted by increasing molecular weight.The distributions are obtained under the condition that the ratio of NH+ 4 • H 2 O to H 3 O + • H 2 Ois between 5 and 20.The product ion labeled "other" includes charge transfer products (e.g., C 6 H 6 O + for phenol) and fragmentation products (e.g., C 5 H 12 N + for pentanal).The product distribution of benzene is not shown because the signals of its product ions are too low to be reliably fitted.The circles are scaled to the square root of the analyte sensitivity.

Figure 5 .
Figure 5. Relationship between S corr and KE cm,50 .S corr represents the sensitivity (cps ppbv −1 ) corrected for the fraction of parent ion in all product ions (f NH + 4 •A ), m/Q-dependent transmission efficiency (TE m/Q), and the ion-molecule reaction rate constant (k, 10 −9 cm 3 molec.s −1 ), as defined in Eq. (7).The solid line represents a fitting of analytes using a Hill equation.S corr = 1350/(1 + (0.267/KE cm ) 11 ).The dashed line represents a linear fitting for analytes with KE cm between 0.2 and 0.3 eV in a similar fashion done inZaytsev et al. (2019).Organic acids, naphthalene, isoprene, 2-methyl furan, limonene, α-pinene, and styrene are excluded from both fittings.The ellipses represent the uncertainty range.
NA: not available.
When the E/N is below 40 Td, H 3 O + •(H 2 O) 3 is the dominant ion because the electric field is too weak to decluster.When the E/N is above 80 Td, NH + 4 is dominant because the electric field results in strong declustering and because NH 3 has higher proton affinity than H 2 O.Only within a narrow E/N window (50-65 Td) is the target reagent ion NH + 4 • H 2 O the most abundant ion.Within this window, several other ions also exist, including NH + , as smaller P and larger T result in larger E/N .As a result, f NH + 4 •H 2 O also exhibits a non-monotonic dependence on the FIMR P and T .The impact of H 2 O mixing ratio in the FIMR (χ H 2 O ) on the distribution is shown in Fig. 1d.The f NH + 4 •H 2 O initially increases with the χ H 2 O , reaches a maximum when χ H 2 O is roughly 0.16-0.18,andthendecreases with increasing χ H 2 O .This trend is because low χ H 2 O limits the supply of H 2 O to cluster with NH + 4 and high χ H 2 O favors the formation of larger clusters.To illustrate, Fig. 1d shows that as χ H 2 O increases, the fraction of smaller clusters (i.e., NH + 4 •H 2 O) decreases, but the fraction of larger clusters (i.e., NH + 4 • (H 2 O) 2 and NH + 4 • (H 2 O) 3 ) increases.Lastly, the NH 3 /H 2 O ratio has a strong impact on the cluster ion distribution (Fig.1e).A low NH 3 /H 2 O ratio (< 0.2 %) results in an insufficient supply of NH + + 4 • H 2 O is exothermic and the internal proton transfer reaction is thermodynamically unfavorable.For limonene, the C 10 H 16 H + is likely produced from the declustering of NH + 4 •C 10 H 16 in the electric fields.The energy released from the exothermic reaction together with that imparted via the drift voltage could even break NH + 4 • limonene into fragments C 6 H + 9 , C 7 H + 11 , and C 6 H 12 N + .As a result, limonene produces ∼20 % of fragmentation products.For reduced aromatics (toluene, oxylene, m-xylene, 1,2,4-TMB, and p-cymene), AH + is the dominant product and NH + 4•A is negligible.The product distributions of reduced aromatics are puzzling because these analytes have lower proton affinity than NH 3 .Since their sensitivities are < 2 cps ppbv −1 , it is not recommended to −1 in NOAA NH + 4 CIMS, two instruments have similar sensitivities.Therefore, H 3 O + chemistry is more suitable to quantify reduced VOCs and small oxygenated VOCs (e.g., acetic acid, methanol, acetaldehyde) than NH + 4 •H 2 O chemistry.NH + 4 •H 2 O chemistry is better for quantifying larger oxygenated VOCs because it causes less fragmentation than the H 3 O + chemistry