Deployment and evaluation of an NH₄⁺∕ H₃O⁺ reagent ion switching chemical ionization mass spectrometer for the detection of reduced and oxygenated gas-phase organic compounds

Reactive organic carbon (ROC) is diverse in its speciation, functionalization, and volatility, with varying implications for ozone production and secondary organic aerosol formation and growth. Chemical ionization mass spectrometry (CIMS) approaches can provide in situ ROC observations, and the CIMS reagent ion controls the detectable ROC species. To expand the range of detectable ROC, we describe a method for switching between the reagent ions ${\mathrm{NH}}_{\mathrm{4}}^{+}$ and H₃O⁺ in a Vocus chemical ionization time-of-flight mass spectrometer (Vocus-CI-ToFMS). We describe optimization of ion–molecule reactor conditions for both reagent ions, at the same temperature, and compare the ability of ${\mathrm{NH}}_{\mathrm{4}}^{+}$ and H₃O⁺ to detect a variety of volatile organic compounds (VOCs) and semi-volatile and intermediate-volatility organic compounds (SVOCs and IVOCs), including oxygenates and organic sulfur compounds. Sensitivities are comparable to other similar instruments (up to ∼5 counts s⁻¹ ${\mathrm{ppt}}_{\mathrm{v}}^{-\mathrm{1}}$), with detection limits on the order of 1–10 s of ppt_v (1 s integration time). We report a method for characterizing and filtering periods of hysteresis following each reagent ion switch and compare use of reagent ions, persistent ambient ions, and a deuterated internal standard for diagnosing this hysteresis. We deploy ${\mathrm{NH}}_{\mathrm{4}}^{+}/{\mathrm{H}}_{\mathrm{3}}{\mathrm{O}}^{+}$ reagent ion switching in a rural pine forest in central Colorado, US, and use our ambient measurements to compare the capabilities of ${\mathrm{NH}}_{\mathrm{4}}^{+}$ and H₃O⁺ in the same instrument, without interferences from variation in instrument and inlet designs. We find that H₃O⁺ optimally detects reduced ROC species with high volatility, while ${\mathrm{NH}}_{\mathrm{4}}^{+}$ improves detection of functionalized ROC compounds, including organic nitrates and oxygenated SVOCs and IVOCs that are readily fragmented by H₃O⁺.

1 Introduction

Tropospheric aerosol formation, oxidant reactivity, and ozone production are driven by the molecularly diverse pool of atmospheric reactive organic carbon (ROC; all organic species excluding methane) (Heald and Kroll, 2020). Speciation of atmospheric ROC is an ongoing analytical challenge (e.g., Goldstein and Galbally, 2007; Hunter et al., 2017), especially at time resolutions relevant to atmospheric mixing and chemistry. While reduced volatile organic compounds (VOCs, with saturation vapor concentration, C^*, $>\mathrm{3}\times {\mathrm{10}}^{\mathrm{6}}$ µg m⁻³) are an important fraction of ROC, functionalized species with lower volatility (SVOCs and IVOCs with C^* between 0.3 and 3×10⁶ µg m⁻³) are major contributors to ozone production and aerosol formation (e.g., Xu et al., 2021; Heald and Kroll, 2020; Bianchi et al., 2019; Donahue et al., 2011). For example, near-comprehensive measurements of ROC at a forested site showed that SVOCs and IVOCs contribute approximately one-third of ⋅OH reactivity and potential secondary organic aerosol production (Hunter et al., 2017). Further, semi-volatile and oxygenated VOCs contribute to marine secondary aerosol formation (Burkart et al., 2017; Mungall et al., 2017; Croft et al., 2019, 2021), and oxygenated species such as furans contribute significantly to ⋅OH reactivity and aerosol production in wildfire plumes (Xu et al., 2021). In many urban environments, volatile chemical products and other classes of IVOCs make a growing contribution to aerosol and ozone production (Coggon et al., 2021; Zhao et al., 2014b). However, owing to limitations in analytical techniques, as well as partitioning to inlet and instrument surfaces (Deming et al., 2019; Pagonis et al., 2019), oxygenated and otherwise functionalized SVOCs and IVOCs are often unmeasured.

Chemical ionization mass spectrometry (CIMS) represents a family of analytical techniques applied to detect and characterize organic and inorganic trace gases in whole air at high time resolution (e.g., Zhang et al., 2023; Yuan et al., 2017; Huey, 2007). The choice of reagent ion determines the scope of the measurement in terms of ROC functionality and chemistry, while instrument construction and design impacts the range of detectable species in terms of volatility and reactivity (e.g., Riva et al., 2019; Krechmer et al., 2018). A range of reagent ions are in common use and are selective toward specific ROC classes. Oxygenated, multi-functional organic gases can be detected as negative ions using iodide (I⁻ or $({\mathrm{H}}_{\mathrm{2}}\mathrm{O}{)}_{\mathrm{n}}\cdot {\mathrm{I}}^{-}$) (Lee et al., 2018, 2014), acetate (${\mathrm{CH}}_{\mathrm{3}}{\mathrm{O}}_{\mathrm{2}}^{-}$) (Brophy and Farmer, 2015; Roberts et al., 2010), CF₃O⁻ (Crounse et al., 2013, 2006), sulfur hexafluoride (${\mathrm{SF}}_{\mathrm{6}}^{-}$) (Nah et al., 2018; Huey, 2007), and nitrate (${\mathrm{NO}}_{\mathrm{3}}^{-}$) and bromide (Br⁻) (Rissanen et al., 2019; Bianchi et al., 2019) reagent ions. Highly oxygenated organic species (with $C^{*}<\mathrm{0.3}\times {\mathrm{10}}^{\mathrm{6}}$ µg m⁻³), together with low-volatility inorganic species (e.g., H₂SO₄), can be detected with nitrate and bromide ionization at ambient pressure (Bianchi et al., 2019; Rissanen et al., 2019; Riva et al., 2019). CF₃O⁻ effectively detects organic peroxides (Crounse et al., 2013, 2006), while I⁻, Br⁻, ${\mathrm{NO}}_{\mathrm{3}}^{-}$, and ${\mathrm{SF}}_{\mathrm{6}}^{-}$ detect a range of polar and acidic gases (Riva et al., 2019; Lee et al., 2018, 2014). Reduced VOCs, small oxygenated VOCs (e.g., methanol, ethanol, acetone, acetaldehyde), and reduced sulfur compounds (e.g., dimethyl sulfide and methanethiol) are readily detected as positive ions via proton transfer with hydronium (H₃O⁺) reagent ions (e.g., Kilgour et al., 2022; Pagonis et al., 2019; Krechmer et al., 2018; Yuan et al., 2017). NO⁺ and ${\mathrm{O}}_{\mathrm{2}}^{+}$ allow detection of reduced VOCs, with proton affinities below that of water, that are generally not detectable with H₃O⁺ (Jordan et al., 2009; Smith and Spanel, 2005). While H₃O⁺ can detect functionalized VOCs, fragmentation is common and complicates the interpretation of mass spectra from complex samples (e.g., Coggon et al., 2024; Kilgour et al., 2024; Li et al., 2021; Pagonis et al., 2019; Yuan et al., 2017). To overcome some of the limitations induced by H₃O⁺ ionization, fast separation techniques have been coupled to proton transfer instruments (Claflin et al., 2021; Stockwell et al., 2021; Vermeuel et al., 2023b; Coggon et al., 2024; Kilgour et al., 2024), and other positive-polarity reagent ions have been applied to functionalized ROC. Benzene (${\mathrm{C}}_{\mathrm{6}}{\mathrm{H}}_{\mathrm{6}}^{+}$) reagent ions detect dimethyl sulfide, monoterpenes, and sesquiterpenes with reduced fragmentation and higher selectivity compared to H₃O⁺ (Kim et al., 2016). Water clusters (i.e., (H₂O)_nH⁺) can detect a small subset of species detected by H₃O⁺, such as dimethyl sulfide, with high selectivity (Blomquist et al., 2010). An array of oxygenated, multi-functional compounds in the intermediate to semi-volatile range can be detected using ammonium reagent ions (e.g., Xu et al., 2022; Khare et al., 2022; Muller et al., 2020; Hansel et al., 2018), which provide some overlap in the fractions of ROC detected by negative-polarity reagent ions such as I⁻ and CF₃O⁻.

Compared to other positive-polarity reagent ions, ammonium (${\mathrm{NH}}_{\mathrm{4}}^{+}$) adduction ionization is selective toward a wider range of multi-functional oxygenated compounds, such as carbonyls, alcohols, ethers, furans, and siloxanes (Xu et al., 2022; Khare et al., 2022; Muller et al., 2020; Zhang et al., 2019; Zaytsev et al., 2019a; Berndt et al., 2019). Ambient observations have recently shown that ${\mathrm{NH}}_{\mathrm{4}}^{+}$ ionization can detect organic nitrates (Xu et al., 2022), while laboratory studies have demonstrated detection of organic peroxides (Zhou et al., 2018) and peroxy radicals (Hansel et al., 2018). Selected ion flow tube mass spectrometry (SIFT-MS) studies show that ${\mathrm{NH}}_{\mathrm{4}}^{+}$ ions form the strongest associations with carbonyl groups, relative to other oxygenates (e.g., alcohols and ethers) (Adams 2003). However, the conditions under which reagent ions form and ion–molecule reactions occur determine the dominant reagent ion and ionization mechanism, which in turn controls the scope of detectable compounds and associated sensitivity. Possible reagent ions include ${\mathrm{NH}}_{\mathrm{4}}\cdot {\mathrm{X}}_{\mathrm{n}}^{+}$ (where X= H₂O or NH₃ and n=0, 1, 2, etc.). In practice, multiple reagent ions can be present, with ${\mathrm{NH}}_{\mathrm{4}}\cdot {\mathrm{H}}_{\mathrm{2}}{\mathrm{O}}^{+}$ providing optimal sensitivity to oxygenated compounds (Xu et al., 2022). Reactions with neutral analytes occur through ligand switching (Reaction 1), where the evaporation of X promotes softer adduct formation compared to ${\mathrm{NH}}_{\mathrm{4}}^{+}$ alone (i.e., n=0) (Xu et al., 2022; Canaval et al., 2019; Adams et al., 2003).

$$\begin{array}{}\text{(1)}& {\mathrm{NH}}_{\mathrm{4}}\cdot {\mathrm{X}}^{+}+\mathrm{A}\to \mathrm{A}\cdot {\mathrm{NH}}_{\mathrm{4}}^{+}+\mathrm{X}\end{array}$$

Given an analyte, A, with larger ${\mathrm{NH}}_{\mathrm{4}}^{+}$ affinity (i.e., the negative enthalpy of the reaction: ${\mathrm{NH}}_{\mathrm{4}}^{+}+\mathrm{A}\to \mathrm{A}\cdot {\mathrm{NH}}_{\mathrm{4}}^{+}$; e.g., Xu et al. (2022)) than X, the ionization reaction (1) is exothermic (Adams et al., 2003). Therefore, ${\mathrm{NH}}_{\mathrm{4}}\cdot {\mathrm{X}}^{+}$ ligand switching reactions will proceed efficiently at or near the collision limit, with little importance of the reverse reaction unless reaction timescales are long or the reaction is endothermic or only slightly exothermic (Xu et al., 2022; Zaytsev et al., 2019a). Reaction (1) is exothermic for the majority of oxygenates and multi-functional compounds (Xu et al., 2022; Adams et al., 2003; Canaval et al., 2019; Zaytsev et al., 2019a) (Table S3, Hunter and Lias, 2023; Meot-Ner (Mautner) and Lias, 2023); however, ion–molecule reactor (IMR) conditions must be selected to promote pure ion chemistry, optimize sensitivity, and minimize fragmentation.

Many CIMS reagent ions provide access to complementary fractions of ambient ROC, and combining multiple reagent ions in a single CIMS instrument can provide several advantages. First, reagent ion switching can maintain the benefits of selectivity afforded by specific reagent ions while expanding the number of detectable compounds. For example, H₃O⁺ and ${\mathrm{NO}}^{+}/{\mathrm{O}}_{\mathrm{2}}^{+}$ are combined in proton transfer reaction (PTR) and SIFT-MS instruments to expand detection capabilities to a broad range of alkanes, alkenes, aromatics, and some oxygenated species (e.g., Agarwal et al., 2014; Smith and Spanel, 2005). Further, the combination of nitrate and bromide reagent ions allows detection of a wide range of highly oxygenated molecules along with hydroperoxyl radicals, iodine compounds, and sulfuric acid (He et al., 2023; Rissanen et al., 2019). Iodide, acetate, and water cluster reagent ions have been combined in laboratory studies, using repeated experiments rather than active reagent switching, to expand the range of detectable ROC (Aljawhary et al., 2013) and inform development of reaction mechanisms (e.g., Zhao et al., 2014a). Thus, combining reagent ions can decrease the potentially large number of instruments required to characterize a broad range of ROC classes (e.g., Heald and Kroll, 2020). Second, combining reagent ions in a single instrument allows for direct comparison between the fractions of ROC detected by each chemical ionization reagent (e.g., Zaytsev et al., 2019a). For example, rapid switching between iodide and the acid-selective acetate reagent ion informs detection of organic acids with iodide (e.g., Brophy and Farmer, 2015). Switching between ${\mathrm{NH}}_{\mathrm{4}}^{+}$ and H₃O⁺ has benefits for measuring both reduced VOCs and their early generation oxidation products (Zaytsev et al., 2019a, b), while also allowing a direct comparison between the subsets of ROC detected by each reagent ion without the complications associated with differing instrument and inlet design (e.g., Riva et al., 2019).

Ambient atmospheric observations with ${\mathrm{NH}}_{\mathrm{4}}^{+}$ adduct ionization CIMS have focused primarily on urban environments, where ${\mathrm{NH}}_{\mathrm{4}}^{+}$ ion chemistry allows detection of oxygenated VOCs from volatile chemical products (Xu et al., 2022; Khare et al., 2022). ${\mathrm{NH}}_{\mathrm{4}}^{+}/{\mathrm{H}}_{\mathrm{3}}{\mathrm{O}}^{+}$ reagent ion switching has so far been limited to laboratory experiments demonstrating feasibility of switching (Muller et al., 2020) and application following laboratory oxidation of VOCs and oxygenated VOCs (Zaytsev et al., 2019a, b). We characterize ${\mathrm{NH}}_{\mathrm{4}}^{+}$/H₃O⁺ reagent ion switching using a Vocus chemical ionization time-of-flight mass spectrometer (Vocus-CI-ToFMS) using both laboratory standards and deployment at a rural forested site. We describe selection of ideal IMR conditions for ${\mathrm{NH}}_{\mathrm{4}}^{+}/{\mathrm{H}}_{\mathrm{3}}{\mathrm{O}}^{+}$ reagent ion switching, at the same temperature, with a focus on sensitivity, fragmentation, and prominence of competing ionization pathways. Using ambient reagent ion switching data, we describe an approach to filter periods of impure ion chemistry, and once filtered, ambient observations allow us to directly compare the fractions of ambient ROC detected by H₃O⁺ and ${\mathrm{NH}}_{\mathrm{4}}^{+}$. Our observations demonstrate that ${\mathrm{NH}}_{\mathrm{4}}^{+}$ is able to detect oxygenated and multi-functional biogenic ROC with both reduced fragmentation and higher selectivity compared to H₃O⁺, illustrating a highly complementary set of CIMS reagent ions.

2 Methods and field site description

2.1 Instrument description

The Vocus-CI-ToFMS (Vocus S, Tofwerk AG and Aerodyne Research Inc.) is described in detail elsewhere (Krechmer et al., 2018), with a brief description supplied here. Two features differentiate the Vocus-CI-ToFMS from other chemical ionization or proton transfer reaction time-of-flight mass spectrometers. First, the Vocus-CI-ToFMS is equipped with a focusing ion–molecule reactor (fIMR) which consists of a radio-frequency-only quadrupole oriented around a 10 cm long resistive glass tube (Krechmer et al., 2018). The fIMR focuses ions toward the centerline, reducing ion losses to the walls and promoting ion transmission into a quadrupole high-pass mass filter (big segmented quadrupole, BSQ). Second, polyether ether ketone (PEEK) tubing is used to establish flow restriction between ambient pressure and the fIMR. The use of PEEK at the instrument inlet reduces interactions between sampled air and more absorptive surfaces which impact transmission of SVOCs and IVOCs into the fIMR (Deming et al., 2019). These modifications to the Vocus-CI-ToFMS improve the ability to detect both reduced and oxidized ROC (Riva et al., 2019). The Vocus-CI-ToFMS used in this study has a mass resolving power of ∼ 5000 m dm⁻¹, a mass range of ∼ 50–500 $m/z$, and a 25 kHz ToF extraction frequency and is equipped with a multi-port reagent ion injection, current-regulated discharge ion source. Details about instrument voltages are available in Table S1 in the Supplement.

When using H₃O⁺ ionization, we inject 20 cm³ min⁻¹ (STP) from the headspace above ultra-high-purity water (MilliporeSigma, OmniSolv LC-MS) under vacuum into the discharge ion source. When switching to ${\mathrm{NH}}_{\mathrm{4}}^{+}$ ionization, we further add a flow from the headspace above a ∼ 1 w/w % solution of ammonium hydroxide (Oakwood Products Inc., trace metals grade) in water to the ion source. Additionally, when switching between reagent ions, the voltages and pressure in the fIMR and the ion optics are adjusted to complement each reagent ion, taking into account sensitivity, fragmentation, and purity of ionization chemistry; this is discussed in Sect. 3.1. A change between ionization modes results in hysteresis where the ion chemistry is impure. The filtering of hysteretic periods is discussed in Sect. 3.4.

2.2 Reactor pressure and voltage gradient

The fIMR collision energy can be controlled in part by adjusting the conditions that impact the velocity, free path, and thermal energy of ions: axial voltage gradient, pressure, and temperature. The temperature must remain constant during reagent ion switching to allow for switching on 15 min timescales. This restricts control of collisional energy to adjustments of the fIMR voltage gradient and pressure. To understand the impact that these parameters have on ion chemistry, we introduce a constant flow of dilute calibration standard while systematically changing the fIMR voltage gradient and pressure. We change the fIMR pressure in 0.1 mbar increments and hold it constant while we increase the fIMR front voltage by 10 V steps. With ${\mathrm{NH}}_{\mathrm{4}}^{+}$, we characterized from 2.5 to 3.5 mbar and from 45 to 65 V cm⁻¹ (60–120 townsends (Td)). For H₃O⁺, we characterized from 1.5 to 2.5 mbar and from 45 to 65 V cm⁻¹ (80–200 Td).

2.3 Sensitivity, detection limit, and fragmentation with standards

We calibrated 23 analytes from multi-component standardized gas cylinders (Apel-Riemer Environmental Inc.) to report sensitivities (counts s⁻¹ ${\mathrm{ppt}}_{\mathrm{v}}^{-\mathrm{1}}$) and detection limits (3σ of the background with 1 s integration). The 23 analytes come from three separate multi-component cylinders where the composition was selected to avoid interferences from fragments (Table S2). Backgrounds were obtained using a zero air generator (Sabio model 1001). We investigate fragmentation of molecular ions using single-component samples of trans-2-hexen-1-ol (96.5 %, Acros Organics, lot A0340603), β-cyclocitral (92.3 %, Thermo Fisher, lot 10237632), 2-hexenal (97.5 %, Oakwood Products Inc., lot 098868J07I), 2-hexanone (100 %, Oakwood Products Inc., lot 098350R22K), and 2-methyl-3-buten-2-ol (99.5 %, Oakwood Products Inc., lot 051281K14H). To calculate a molecular ion fraction, we average 15 s of 2 Hz data and fit peaks corresponding to molecular ions (i.e., $\mathrm{A}\cdot {\mathrm{H}}^{+}$ for H₃O⁺ and $\mathrm{A}\cdot {\mathrm{NH}}_{\mathrm{4}}^{+}$ for ${\mathrm{NH}}_{\mathrm{4}}^{+}$), we identify fragments, and clusters then divide the molecular ion signal by the sum of all related peaks. The fIMR conditions for H₃O⁺ ionization during these experiments were 2.2 mbar and 60 °C, with a voltage gradient of 67.5 V cm⁻¹ (140 Td) and a BSQ amplitude of 270 V. Using a 60 °C reaction chamber with H₃O⁺ is lower than commonly reported in the literature (∼ 80–100 °C) (e.g., Vermeuel et al., 2023a; Coggon et al., 2024); this choice arises from fIMR temperature constraints for ${\mathrm{NH}}_{\mathrm{4}}^{+}$ (Xu et al., 2022) and is discussed in more detail in Sect. 3.1. For ${\mathrm{NH}}_{\mathrm{4}}^{+}$ ionization the fIMR settings were 3.1 mbar and 60 °C, with a voltage gradient of 60 V cm⁻¹ (90 Td) and a BSQ amplitude of 250 V. The BSQ frequency was 2.2 MHz, and the fIMR amplitude and frequency were 500 V and 1.6 MHz, respectively (Table S1).

2.4 Signal response to ambient relative humidity

To test the effect of relative humidity on sensitivity, we varied the ratio of wet and dry flows (controlled with mass flow controllers, MKS Instruments model 1179C Mass-Flo) to achieve a range of relative humidities. The relative humidity was measured inline (Omega Engineering model HX71-V1). Downstream of the relative humidity measurement 10 cm³ min⁻¹ (STP) of a certified gas standard (Apel-Riemer Environmental Inc. and Airgas for dimethyl sulfide) was added to the humidified flow before being introduced into the Vocus-CI-ToFMS. Relative humidity ranged between 15 % and 85 % during the experiments. Measurements alternated between elevated relative humidity and dry conditions, such that each measurement at elevated humidity could be directly compared to a dry (0 % RH) measurement immediately before.

2.5 Observations at the Manitou Experimental Forest Observatory

We deployed a Vocus-CI-ToFMS at the Manitou Experimental Forest Observatory (MEFO) from 3 to 24 September 2021. MEFO is a rural ponderosa pine forest at middle elevation (∼ 2300 m), located ∼ 40 km northwest of Colorado Springs and ∼ 70 km southwest of Denver (39.1006° N, 105.0942° W). The ROC composition at this site is well characterized, with emissions dominated by local biogenic sources (Hunter et al., 2017; Vermeuel et al., 2023a; Riches et al., 2024; Link et al., 2024). A full description of the field site can be found in Ortega et al. (2014). The Vocus-CI-ToFMS sample inlet was ∼ 4 m of 0.25 in. outer diameter perfluoroalkoxy (PFA) tubing situated ∼ 4 m above ground. The inlet flow was ∼ 3.8 L min⁻¹ (i.e., 2.9 L min⁻¹ (STP)) pulled by a flow-restricted bypass pump, resulting in a laminar flow inlet (Reynolds number of ∼ 1150) and corresponding to a residence time of ∼ 0.7 s. The inlet likely produced wall loss of oxygenated ROC, and while the extent was not quantified, minimizing the inlet inner diameter and maximizing the flow rate, while maintaining laminar flow, serve to minimize inlet losses and tubing delays (Pagonis et al., 2017). The Vocus-CI-ToFMS sub-sampled 93–100 cm³ min⁻¹ (i.e., 71–77 cm³ min⁻¹ (STP)) perpendicular to the main inlet flow, which helps prevent ambient aerosol clogging the capillary inlet interface compared to a linear sub-sampling assembly (Jensen et al., 2023). We performed bi-hourly, 3 min instrument backgrounds with ultra zero air (Airgas, UZA grade) followed by a 1 min, single-point calibration with a certified calibrant mixture (Apel-Riemer Environmental Inc.) for both reagent ions.

The Vocus-CI-ToFMS switched between ${\mathrm{NH}}_{\mathrm{4}}^{+}$ and H₃O⁺ ionization on 15 min time intervals during the deployment in MEFO. The fIMR conditions for H₃O⁺ ionization were 2.5 mbar and 60 °C, with a voltage gradient of 62 V cm⁻¹ and a BSQ amplitude of 350 V. The fIMR conditions for ${\mathrm{NH}}_{\mathrm{4}}^{+}$ ionization were 3.1 mbar and 60 °C, with a voltage gradient of 65 V cm⁻¹ and a BSQ amplitude of 250 V. The fIMR settings correspond to $E/N$ values of 114 and 96 Td for H₃O⁺ and ${\mathrm{NH}}_{\mathrm{4}}^{+}$, respectively. The BSQ frequency was 2.2 MHz, and the fIMR amplitude and frequency were 450 V and 1.3 MHz, respectively (Table S1). The fIMR parameters for both reagent ions were informed by experiments detailed in Sect. 3.1.

2.6 ARTofMELT expedition on icebreaker (I/B) Oden

We deployed the Vocus-CI-ToFMS aboard the Swedish I/B Oden from 7 May to 15 June 2023 as part of the “Atmospheric rivers and the onset of ice melt” (ARTofMELT) measurement expedition. The cruise on I/B Oden took place within the pack ice and marginal ice zone between Svalbard and Greenland in the Fram Strait. The Vocus-CI-ToFMS switched between ${\mathrm{NH}}_{\mathrm{4}}^{+}$ and H₃O⁺ ionization on 15 min time intervals during the expedition. The Vocus-CI-ToFMS was mounted to the floor of a sea container on I/B Oden's fourth deck using metal plates and five high-deflection vibration isolation feet (Barry Controls model 2K2-BA-90) and was mounted to an open wall using two wire rope isolators (Enidine model WR6-850-10-E).

Ambient air was sampled through a 0.375 in. outer diameter PFA tube ∼15 m in length. The entire length of tubing was insulated and heated to 30 °C in three separately controlled sections. The Vocus-CI-ToFMS inlet flow was driven by a vacuum pump (Agilent IDP-7 dry scroll) regulated by a mass flow controller (15 L min⁻¹ (STP)). A deuterated internal standard (containing dimethylsulfide-d₃, acetone-d₆, 2-hexanone-d₄, and mesitylene-d₁₂) was injected near the top of the Vocus-CI-ToFMS inlet through a 0.125 in. outer diameter PFA tube at a rate of 5 cm³ min⁻¹ (STP) into the total inlet flow of 15 L min⁻¹ (STP), yielding ∼333 ppt_v from the nominally 1 ppm standardized cylinder (Apel-Riemer Environmental Inc.). The inlet assembly was affixed to a metal pipe extending from the top of the sea container toward the bow of the ship at a ∼40° angle ∼3 m above the top of the container or 25 m above the water/ice surface. The use of internal standards during this expedition allows us to analyze reagent ion hysteresis (Sect. 3.4) using a persistent and known set of ions. The fIMR conditions for H₃O⁺ ionization during ARTofMELT were 2.2 mbar, 60 °C, a voltage gradient of 67.5 V cm⁻¹ (140 Td), and a 270 V BSQ amplitude. For ${\mathrm{NH}}_{\mathrm{4}}^{+}$ ionization the fIMR settings were 3.1 mbar, 60 °C, a 60 V cm⁻¹ (90 Td) voltage gradient, and a 250 V BSQ amplitude. The BSQ frequency was 2.2 MHz, and the fIMR amplitude and frequency were 500 V and 1.6 MHz, respectively (Table S1). We reserve further analysis of the ARTofMELT data set for future work.

2.7 Data analysis

Raw mass spectral data were collected with Acquility (version 2.3.18) and TofDaq (version 1.99) (Tofwerk AG) and processed in Tofware (version 3.2.5, Tofwerk AG and Aerodyne Research Inc.). For MEFO data, the time resolution was pre-averaged from 1 to 0.1 Hz. For ARTofMELT data, the time resolution was pre-averaged from 2 to 1 Hz. All data were mass-calibrated, baseline-subtracted, and peak-fit in Tofware. Time-integrated high-resolution ion signals were exported for further analysis in Python (version 3.9.12). Responses in the $E/N$ scans in Sect. 3.1 were interpolated using a linear interpolation on a triangular grid (using matplotlib.tri.LinearTriInterpolator). C^* values are estimated using EPI Suite (US EPA, 2023).

Figure 1 (a–d) Contour plots of fIMR pressure and voltage gradient scans with a constant concentration of analyte (10 ppb_v) introduced into the Vocus-CI-ToFMS. Normalized signal intensity for (a) methyl ethyl ketone measured with ${\mathrm{NH}}_{\mathrm{4}}^{+}$ (${\mathrm{C}}_{\mathrm{4}}{\mathrm{H}}_{\mathrm{8}}\mathrm{O}\cdot {\mathrm{NH}}_{\mathrm{4}}^{+}$) and (b) α-pinene measured with H₃O⁺ (${\mathrm{C}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{16}}\cdot {\mathrm{H}}^{+}$). (c) Fractional contribution of the ${\mathrm{NH}}_{\mathrm{4}}^{+}$ molecular ion (${\mathrm{C}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{16}}\cdot {\mathrm{NH}}_{\mathrm{4}}^{+}$) to the total α-pinene signal (i.e., the sum of the proton transfer product (${\mathrm{C}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{16}}\cdot {\mathrm{H}}^{+}$) and the molecular ion). (d) Fractional contribution of the H₃O⁺ molecular ion (${\mathrm{C}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{16}}\cdot {\mathrm{H}}^{+}$) to the total α-pinene signal (i.e., the sum of the α-pinene fragment (${\mathrm{C}}_{\mathrm{6}}{\mathrm{H}}_{\mathrm{8}}\cdot {\mathrm{H}}^{+}$) and the molecular ion). Contour plots of calculated $E/N$ values over the scanned space for both (e) ${\mathrm{NH}}_{\mathrm{4}}^{+}$ and (f) H₃O⁺.

3 Results and discussion

3.1 Ion–molecule reactor pressure and voltage gradient

H₃O⁺ and ${\mathrm{NH}}_{\mathrm{4}}^{+}$ ionization operate optimally at differing combinations of fIMR pressure, voltage gradient, and temperature (Xu et al., 2022; Gouw and Warneke, 2007), all of which impact the reduced electric field ($E/N$) of the fIMR:

$$\begin{array}{}\text{(1)}& E/N={\displaystyle \frac{T\times \mathrm{\Delta }V\times k_{\mathrm{B}}}{l_{\mathrm{imr}}\times p}},\end{array}$$

where T is the temperature (K), ΔV is the difference between the front and back voltage (V), l_imr is the fIMR length (m), k_B is the Boltzmann constant (J K⁻¹), and p is pressure (Pa or J m⁻³). $E/N$ has units of townsends (Td, 1 Td =1$\times {\mathrm{10}}^{-\mathrm{17}}$ V cm²) and describes ion velocity and collisional energy. High $E/N$ values promote increased fragmentation and reduced clustering, while low $E/N$ values promote cluster formation and reduced fragmentation. We analyzed relative sensitivity, fragmentation, and prevalence of ionization pathways while varying the fIMR pressure and voltage gradient with a constant temperature of 60°C (Fig. 1) to inform our selection of fIMR settings. Because fIMR temperature takes tens of minutes to stabilize, we selected a constant 60 °C fIMR temperature to promote ${\mathrm{NH}}_{\mathrm{4}}^{+}\cdot {\mathrm{H}}_{\mathrm{2}}\mathrm{O}$ clusters in ${\mathrm{NH}}_{\mathrm{4}}^{+}$ ionization. This is lower than most H₃O⁺ fIMR temperatures (∼ 80–100 °C); however, the effect of lower fIMR temperature on the reagent ion distribution can be mitigated by adjusting other fIMR settings.

Selecting H₃O⁺ ionization fIMR parameters requires balancing fragmentation and sensitivity. We observe a large increase (> 60 % at 2.4 mbar) in the sensitivity to the molecular ion (${\mathrm{C}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{16}}\cdot {\mathrm{H}}^{+}$) with increased fIMR voltage gradient (Fig. 1b), which arises from three main factors. First, reduced residence time with increased voltage gradient (163 or 113 µs at 45 or 65 V cm⁻¹, respectively, with pressure and temperature of 2.4 mbar and 333.15 K) could increase ion transmission through the fIMR. Second, at low $E/N$ protonated water clusters contribute to the ionization of α-pinene. The production of protonated water clusters is evident from the reduced benzene sensitivity at lower $E/N$ (Fig. A1) (Gouw and Warneke, 2007). Water clusters have a higher proton affinity (i.e., the negative enthalpy of the reaction: ${\mathrm{H}}^{+}+\mathrm{A}\to \mathrm{A}\cdot {\mathrm{H}}^{+}$) compared to water (Hunter and Lias, 1998), and α-pinene has a higher proton affinity than both water and the first water cluster (i.e., (H₂O)₂), making ionization reactions with both exothermic. In contrast, benzene has a proton affinity higher than water but lower than the first cluster, making the ionization reaction of benzene with (H₂O)₂H⁺ endothermic and unlikely. Therefore, the formation of water clusters will reduce the sensitivity to benzene while increasing the sensitivity to α-pinene. Third, the known H₃O⁺ α-pinene fragment, ${\mathrm{C}}_{\mathrm{6}}{\mathrm{H}}_{\mathrm{8}}\cdot {\mathrm{H}}^{+}$, has a larger contribution to the total α-pinene signal at higher $E/N$ (Fig. 1d), which suggests that part of the increased sensitivity to ${\mathrm{C}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{16}}\cdot {\mathrm{H}}^{+}$ at higher pressures is also attributable to reduced loss to fragmentation. The combined responses of transmission, fragmentation, and proton affinity to changes in voltage and pressure with H₃O⁺ ionization result in a sensitivity that does not follow $E/N$ directly (Fig. 1b, 1f).

Selecting ${\mathrm{NH}}_{\mathrm{4}}^{+}$ ionization fIMR parameters requires balancing between signal intensity and purity of ion chemistry. Similar to H₃O⁺, with ${\mathrm{NH}}_{\mathrm{4}}^{+}$ we observe increased signal intensity of the methyl ethyl ketone molecular ion (${\mathrm{C}}_{\mathrm{4}}{\mathrm{H}}_{\mathrm{8}}\mathrm{O}\cdot {\mathrm{NH}}_{\mathrm{4}}^{+}$) at higher voltage gradients (Fig. 1a). In contrast to α-pinene detection with H₃O⁺, sensitivity to ${\mathrm{C}}_{\mathrm{4}}{\mathrm{H}}_{\mathrm{8}}\mathrm{O}\cdot {\mathrm{NH}}_{\mathrm{4}}^{+}$ is not impacted by fragmentation and changing reagent ion proton affinity across the range in voltage gradient. The sensitivity to the ${\mathrm{C}}_{\mathrm{4}}{\mathrm{H}}_{\mathrm{8}}\mathrm{O}\cdot {\mathrm{NH}}_{\mathrm{4}}^{+}$ ion is highest at high $E/N$, but the change in sensitivity is mostly dependent on the voltage gradient and is less impacted by the fIMR pressure (Fig. 1a). We observe a similar trend for other oxygenated ROC species (Fig. A2). The vertical contours in the ${\mathrm{NH}}_{\mathrm{4}}^{+}$ sensitivity contrasted with the L-shaped contouring in the H₃O⁺ sensitivity (Fig. 1a, b) support the three-factor dependence for H₃O⁺ on transmission, fragmentation, and proton affinity and point to transmission as the major factor impacting sensitivity with ${\mathrm{NH}}_{\mathrm{4}}^{+}$ ionization. However, at high $E/N$ we observe impurities in the ionization chemistry (i.e., proton transfer products ($\mathrm{A}\cdot {\mathrm{H}}^{+}$) occurring for α-pinene) with ${\mathrm{NH}}_{\mathrm{4}}^{+}$ ionization (Fig. 1c). Proton transfer ionization under ${\mathrm{NH}}_{\mathrm{4}}^{+}$ is undesirable because it provides lower selectivity, leads to higher fragmentation rates compared to the ligand switching mechanism, and complicates interpretation of the mass spectrum. α-Pinene has a lower ammonium affinity than H₂O (Canaval et al., 2019), which makes the ligand switching reaction (1) endothermic and thus dependent on increased collisional energy at higher voltage gradients (Xu et al., 2022). We observe a larger contribution of the proton transfer product at higher $E/N$, which is consistent with electric-field-induced production of ${\mathrm{C}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{16}}\cdot {\mathrm{H}}^{+}$ through internal proton transfer (Xu et al., 2022). Alternatively, it is also possible that ${\mathrm{C}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{16}}\cdot {\mathrm{H}}^{+}$ production is enhanced by declustering of the ${\mathrm{NH}}_{\mathrm{4}}^{+}\cdot {\mathrm{H}}_{\mathrm{2}}\mathrm{O}$ ions to form ${\mathrm{NH}}_{\mathrm{4}}^{+}$, which is more likely to undergo proton transfer reactions directly, without the need for internal proton transfer, due to the lower proton affinity of NH₃ compared to α-pinene (Canaval et al., 2019). Regardless of the mechanism, formation of proton transfer products is ideally avoided, and we find that their production is minimized at low $E/N$ (Fig. 1c). We note that the formation of secondary clusters (i.e., ${\mathrm{NH}}_{\mathrm{4}}\cdot {\mathrm{H}}_{\mathrm{2}}\mathrm{O}\cdot {\mathrm{A}}^{+}$ and ${\mathrm{NH}}_{\mathrm{4}}\cdot {\mathrm{NH}}_{\mathrm{3}}\cdot {\mathrm{A}}^{+}$) is negligible over the entire investigated $E/N$ space (Figs. S1, S2). The sensitivity to methyl ethyl ketone and prevalence of undesirable reaction products (${\mathrm{C}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{16}}\cdot {\mathrm{H}}^{+}$) are optimal under opposing conditions in the voltage-gradient–pressure space, such that optimal fIMR parameters for ${\mathrm{NH}}_{\mathrm{4}}^{+}$ require a compromise between sensitivity and purity of ionization chemistry.

Figure 2 Sensitivities (b) and detection limits (a, log y axis) for the Vocus-CI-ToFMS with ${\mathrm{NH}}_{\mathrm{4}}^{+}$ ionization (orange) and H₃O⁺ ionization (blue) for 23 analytes from standardized gas cylinders, grouped by functional group/compound type. Sensitivities and detection limits (DLs) are calculated for the molecular ion only (i.e., $\mathrm{A}\cdot {\mathrm{NH}}_{\mathrm{4}}^{+}$ for ${\mathrm{NH}}_{\mathrm{4}}^{+}$ or $\mathrm{A}\cdot {\mathrm{H}}^{+}$ for H₃O⁺), and mass spectral fragments are not included. DLs are calculated as 3σ over a 600 s background at 1 Hz.

Figure 3 Molecular ion fraction and the contribution of various non-molecular ions for five analytes using H₃O⁺ ionization (blue/left) and ${\mathrm{NH}}_{\mathrm{4}}^{+}$ ionization (orange/right). The contribution of the molecular ion is highlighted with a scatter plot, and uncertainty bars are the standard deviation across 30 mass spectra at 2 Hz. Data represented in this figure are shown in Table S6.

3.2 Sensitivities, detection limits, and fragmentation

Direct calibrations demonstrate the selectivity of ${\mathrm{NH}}_{\mathrm{4}}^{+}$ relative to H₃O⁺ ionization for a range of carbonyls, hydrocarbons, alcohols, and organic sulfur compounds (Fig. 2). We observe that H₃O⁺ is capable of detecting nearly every compound in this set of analytes from standardized gas cylinders at the molecular ion, which demonstrates the utility of H₃O⁺ as a general reagent ion that allows for detection of reduced and some oxidized species. In contrast, ${\mathrm{NH}}_{\mathrm{4}}^{+}$ ionization is more selective toward oxygenates, including saturated and unsaturated ketones, unsaturated aldehydes, and the multi-functional 1-methoxy-2-propanol. ${\mathrm{NH}}_{\mathrm{4}}^{+}$'s selectivity toward oxygenates demonstrates its utility for expanding the range of compounds detectable with a single instrument, as well as for supporting the identification of molecular ions and fragments detected simultaneously with H₃O⁺. In addition, ${\mathrm{NH}}_{\mathrm{4}}^{+}$ does not detect aromatics, small alkenes, and reduced sulfur compounds that H₃O⁺ detects well, demonstrating the complementary nature of these reagent ions. While Fig. 2 suggests that H₃O⁺ detects dimethyl sulfoxide (DMSO), an oxidation product of dimethyl sulfide (Barnes et al., 2006), with a detection limit (DL) of 16 ppt_v, this DL is optimistic because DMSO peak separation is hindered by isobaric ions of protonated benzene and a protonated water cluster of acetic acid. We therefore expect that DMSO cannot be detected with H₃O⁺ in the Vocus-CI-ToFMS at concentrations relevant to the marine boundary layer (i.e., <100 ppt_v, Putaud et al., 1999; Sciare et al., 2000; Legrand et al., 2001; and Nowak et al., 2001).

For the compounds detected with both ionization modes, sensitivities and detection limits for H₃O⁺ and ${\mathrm{NH}}_{\mathrm{4}}^{+}$ are in the same order of magnitude (Figs. 2, A3). ${\mathrm{NH}}_{\mathrm{4}}^{+}$ detects the subset of ketones and the unsaturated aldehydes shown in Fig. 2 with a lower or similar DL to H₃O⁺. 1-methoxy-2-propanol, trans-3-hexenol, and D5-siloxane suggest that ${\mathrm{NH}}_{\mathrm{4}}^{+}$ has a greater ability than H₃O⁺ to detect oxygenated and functionalized compounds, but this is not broadly apparent across the families of compounds we calibrated directly (Fig. 2). This likely arises because the analytes shown in Fig. 2 are limited to compounds amendable to gas cylinder calibration and are therefore biased toward VOCs, SVOCs, and IVOCs with minimal oxygenation and relatively high volatility (i.e., C^* > 9×10⁴ µg m⁻³; Table S4). Despite the compromises in the fIMR temperature made to allow for the switching system (Sect. 3.1) the sensitivities for ${\mathrm{NH}}_{\mathrm{4}}^{+}$ ionization reported here are similar to the sensitivities reported in recent ${\mathrm{NH}}_{\mathrm{4}}^{+}$ literature (Khare et al., 2022; Xu et al., 2022) (Table S5).

Compared to H₃O⁺ ionization, ${\mathrm{NH}}_{\mathrm{4}}^{+}$ ionization reduces molecular ion fragmentation for functionalized compounds (Fig. 3). We use a molecular ion fraction (Fig. 3; ratio of the molecular ion signal to the total signal from the molecular ion, fragments, and clusters) to analyze the contributions of molecular ions and mass spectral fragments from both ${\mathrm{NH}}_{\mathrm{4}}^{+}$ and H₃O⁺ ionization for a series of analytes complementary to those calibrated with standardized gas cylinders (Fig. 2). Alcohols fragment substantially using both reagent ions, with H₃O⁺ and trans-2-hexenol fragments almost completely away from the molecular ion resulting in a near-zero molecular ion fraction. This is consistent with the negligible sensitivity to the similarly structured trans-3-hexenol molecular ion (Fig. 2) with H₃O⁺ owing to fragmentation (e.g., Pagonis et al., 2019). In contrast, ${\mathrm{NH}}_{\mathrm{4}}^{+}$ ionization detects trans-2-hexenol with a molecular ion fraction of 0.67. For the tertiary alcohol, 2-methyl-3-buten-2-ol (2,3,2-MBO), we observe substantial fragmentation with both ionization modes but a higher molecular ion fraction under ${\mathrm{NH}}_{\mathrm{4}}^{+}$ (i.e., 0.31 with ${\mathrm{NH}}_{\mathrm{4}}^{+}$ and 0.19 with H₃O⁺). The ketone and aldehydes sampled only fragmented substantially under H₃O⁺ ionization, while ${\mathrm{NH}}_{\mathrm{4}}^{+}$ retains the molecular structure, leading to a high molecular ion fraction. This is consistent with the lower ${\mathrm{NH}}_{\mathrm{4}}^{+}$ detection limit for the majority of ketones we examined (with the exceptions of acetone and 2-octanone) and the aldehyde trans-2-hexenal (Fig. 2). Our observations suggest that reduced fragmentation has a larger impact on detection capability of the two reagent ions for more highly oxidized compounds with multiple functional groups. This is observed for 1-methoxy-2-propanol, which is readily detected with ${\mathrm{NH}}_{\mathrm{4}}^{+}$ but not with H₃O⁺; the detection of oxidized ROC is discussed further in Sect. 3.5. Overall, these observations demonstrate the high selectivity of ${\mathrm{NH}}_{\mathrm{4}}^{+}$ ionization for oxygenates and the benefits of reduced fragmentation with ${\mathrm{NH}}_{\mathrm{4}}^{+}$. Coupling the detection of reduced ROC and organic sulfur from H₃O⁺ with the detection of oxygenates from ${\mathrm{NH}}_{\mathrm{4}}^{+}$ expands the fraction of atmospheric ROC that we can detect with a single instrument.

Figure 4 Signal dependence on sample relative humidity for ${\mathrm{NH}}_{\mathrm{4}}^{+}$ (a, orange) and H₃O⁺ (b, blue). Measurements were made at relative humidities of 15 %, 30 %, 50 %, 70 %, and 85 %; points are offset from these values for visibility. Ethanol, benzene, m-xylene, and 1,2,4-trimethylbenzene are excluded for ${\mathrm{NH}}_{\mathrm{4}}^{+}$ due to low signal. Acetonitrile and ethanol are omitted for H₃O⁺ due to low transmission through the BSQ. Methyl ethyl ketone is omitted for H₃O⁺ due to interference of the reagent ion cluster $({\mathrm{H}}_{\mathrm{2}}\mathrm{O}{)}_{\mathrm{3}}\cdot {\mathrm{H}}^{+}$. The error bars represent propagated relative deviations in dry and humidified signals.

3.3 Impact of sample relative humidity

Previous studies have characterized the significant humidity dependence of sensitivity in various CIMS instruments to understand and correct for changing ambient humidity (e.g., Warneke et al., 2001; Gouw and Warneke, 2007; Kari et al., 2018; Zaytsev et al., 2019a). Humidity-driven changes in reagent ion chemistry, and therefore sensitivity, are generally small in the Vocus-CI-ToFMS due to the large flow of water vapor (i.e., 20 cm³ min⁻¹ (STP)) injected into the ion source (Krechmer et al., 2018; Khare et al., 2022). Varying sample humidity with constant analyte concentration demonstrates low humidity dependence with both ${\mathrm{NH}}_{\mathrm{4}}^{+}$ and H₃O⁺ ionization across a range of reduced and oxygenated ROC (Fig. 4). We note an approximately 10 % increase in the ${\mathrm{NH}}_{\mathrm{4}}^{+}$ sensitivity to nitriles and oxygenates, while alkene sensitivities remain unchanged up to 85 % RH. We also observe a slight (5 %–10 %) increase in sensitivity with humidity for oxygenated species with H₃O⁺, while alkene sensitivities are less affected. The low humidity dependence of the Vocus-CI-ToFMS has been demonstrated previously for H₃O⁺ for a variety of analytes (Krechmer et al., 2018; Kilgour et al., 2022; Li et al., 2024) and for a select number of small oxygenates, alkenes, and acetonitrile with ${\mathrm{NH}}_{\mathrm{4}}^{+}$ (Khare et al., 2022; Xu et al., 2022). We demonstrate the low dependence of sensitivity on sample humidity with ${\mathrm{NH}}_{\mathrm{4}}^{+}$ ionization under different instrumental conditions and for a selection of analytes including oxygenated alkenes and siloxanes (Fig. 4).

Figure 5 Ion signal after a reagent ion switch for ${\mathrm{NH}}_{\mathrm{4}}^{+}$ (a, b) and H₃O⁺ (c, d) in the MEFO data, showing ${\mathrm{NH}}_{\mathrm{4}}\cdot {\mathrm{H}}_{\mathrm{2}}{\mathrm{O}}^{+}$ ions (a, c) and ${\mathrm{C}}_{\mathrm{3}}{\mathrm{H}}_{\mathrm{6}}\mathrm{O}\cdot {\mathrm{NH}}_{\mathrm{4}}^{+}$ (b, d). We grouped ion signals by the time after a switch and normalized the mean of each group by the maximum, and normalized means were fit with a bi-exponential function. The derivative of the fit (δ_n) is displayed on the right axes (purple traces) and is used as a metric to filter reagent ion hysteresis. A summary of the amount of data removed as a function of the selected threshold for these ions is shown in Fig. S4.

Figure 6 Ion signal after a reagent ion switch for ${\mathrm{NH}}_{\mathrm{4}}^{+}$ (a, b) and H₃O⁺ (c, d) in the ARTofMELT data, showing ${\mathrm{NH}}_{\mathrm{4}}\cdot {\mathrm{H}}_{\mathrm{2}}{\mathrm{O}}^{+}$ ions (a, c) and ${\mathrm{C}}_{\mathrm{6}}{\mathrm{H}}_{\mathrm{8}}{\mathrm{d}}_{\mathrm{4}}\mathrm{O}\cdot {\mathrm{NH}}_{\mathrm{4}}^{+}$ (b) and ${\mathrm{C}}_{\mathrm{6}}{\mathrm{H}}_{\mathrm{8}}{\mathrm{d}}_{\mathrm{4}}\mathrm{O}\cdot {\mathrm{H}}^{+}$ (d) internal standard ions. We grouped ion signals by the time after a switch and normalized the mean of each group by the maximum, and normalized means were fit with a bi-exponential function. The derivative of the fit (δ_n) is displayed on the right axes (purple traces) and is used as a metric to filter reagent ion hysteresis. A summary of the amount of data removed as a function of the selected threshold for these ions is shown in Fig. S5.

3.4 Removal of reagent ion hysteresis from switching

Reagent ion chemistry does not stabilize immediately upon switching between ${\mathrm{NH}}_{\mathrm{4}}^{+}$ and H₃O⁺ ionization. This reagent ion switch requires adjustment of instrument conditions that impact ion chemistry; these include the reagents introduced into the ion source, the fIMR pressure and voltage gradient, and downstream ion optic voltages (Table S1). However, instrument conditions for each reagent ion (Sect. 3.1) are such that analyte detection through the alternate ionization pathway is possible with both ${\mathrm{NH}}_{\mathrm{4}}^{+}$ and H₃O⁺ ionization (e.g., Zaytsev et al., 2019a). This is in contrast to some other reagent ion pairs (e.g., ${\mathrm{CH}}_{\mathrm{3}}{\mathrm{O}}_{\mathrm{2}}^{-}$ and I⁻) where instrument conditions differ drastically, and so hysteresis is not observed (e.g., Brophy and Farmer, 2015). As a result, when ${\mathrm{NH}}_{\mathrm{4}}^{+}$ and H₃O⁺ are paired in a single instrument, we observe a distinct transitional period of reagent ion hysteresis following each reagent ion switch. The hysteretic period arises from (1) changes in ion transmission due to instrument conditions, such as ion optics and fIMR settings, which are fast (one to tens of seconds), and (2) changes in the reagent ion speciation due to the presence or absence of NH₃(g), which is slower (tens of seconds to minutes). Periods of hysteresis must be characterized and removed to ensure stable and consistent measurements without drifting sensitivities over the course of each 15 min measurement period. We accomplish this by monitoring ion stability over a large number of repeated switches.

We compare the utility of three ion types as markers to quantify the timescale of reagent ion switching hysteresis: ${\mathrm{NH}}_{\mathrm{4}}^{+}$ reagent ion signal, a persistent ambient ${\mathrm{NH}}_{\mathrm{4}}^{+}$ adduct ion, and known and persistent ${\mathrm{NH}}_{\mathrm{4}}^{+}$ adduct or proton transfer molecular ions from an internal standard infused in the sampling inlet. We selected ions primarily measured with ${\mathrm{NH}}_{\mathrm{4}}^{+}$ ionization because the influence of NH₃(g) is observed under both H₃O⁺ and ${\mathrm{NH}}_{\mathrm{4}}^{+}$ ionization modes, whereas the influence of H₂O(g) reagent ions is not observed in ${\mathrm{NH}}_{\mathrm{4}}^{+}$ mode (Fig. S3). This arises because NH₃(g) has a higher proton affinity than H₂O(g) (Hunter and Lias, 1998), which causes any NH₃(g) present in the ion source and fIMR to readily form ${\mathrm{NH}}_{\mathrm{4}}^{+}$ or ${\mathrm{NH}}_{\mathrm{4}}\cdot {\mathrm{H}}_{\mathrm{2}}{\mathrm{O}}^{+}$ at the expense of H₃O⁺ formation.

3.4.1 Filtering hysteresis with reagent and persistent ambient ions

In the absence of a known and persistent signal from an internal standard to diagnose reagent ion hysteresis, we compare the use of ${\mathrm{NH}}_{\mathrm{4}}\cdot {\mathrm{H}}_{\mathrm{2}}{\mathrm{O}}^{+}$ (i.e., the prominent ammonium reagent ion signal) and ${\mathrm{C}}_{\mathrm{3}}{\mathrm{H}}_{\mathrm{6}}\mathrm{O}\cdot {\mathrm{NH}}_{\mathrm{4}}^{+}$ (i.e., a ubiquitous analyte-ammonium molecular ion) under both ionization methods. Both ${\mathrm{NH}}_{\mathrm{4}}\cdot {\mathrm{H}}_{\mathrm{2}}{\mathrm{O}}^{+}$ and ${\mathrm{C}}_{\mathrm{3}}{\mathrm{H}}_{\mathrm{6}}\mathrm{O}\cdot {\mathrm{NH}}_{\mathrm{4}}^{+}$ decay after switching to H₃O⁺ ionization (Fig. 5c, d; NH₃(g) depletion in the fIMR) and intensify after switching to ${\mathrm{NH}}_{\mathrm{4}}^{+}$ ionization (Fig. 5a, b; NH₃(g) accumulation in the fIMR). We grouped a total of 558 h of ambient ${\mathrm{NH}}_{\mathrm{4}}^{+}$/H₃O⁺ 15 min reagent ion switching observations from MEFO (Sect. 2.5) by time after a reagent ion switch. We normalized the mean signal (in 10 s intervals starting at 5 s into a switch) to the maximum and fit the normalized data with a bi-exponential function (Fig. 5). The bi-exponential function describes changes in both instrument conditions (fast) and equilibration of NH₃(g) in the ion source and fIMR that drives reagent ion chemical speciation (slow). We use the derivative of this decay function (δ_n) to quantify a normalized rate of change in the ion signal as a function of the time after a switch (purple lines in Fig. 5a–d). We use δ_n to set a threshold for filtering hysteresis, removing data before δ_n reaches the set threshold. For all ions in Fig. 5, δ_n changes rapidly in the first ∼ 100 s after a reagent ion switch and slowly approaches but does not reach zero on the measurement timescale (900 s) (Fig. S4), likely due to the timescale for complete NH₃(g) equilibration with instrument surfaces.

Monitoring ${\mathrm{NH}}_{\mathrm{4}}\cdot {\mathrm{H}}_{\mathrm{2}}{\mathrm{O}}^{+}$ δ_n has the benefit of being directly related to the abundance of reagent ion; however, both the decay of ${\mathrm{NH}}_{\mathrm{4}}\cdot {\mathrm{H}}_{\mathrm{2}}{\mathrm{O}}^{+}$ in H₃O⁺ ionization mode and its initial increase in ${\mathrm{NH}}_{\mathrm{4}}^{+}$ ionization mode are driven largely by changes in the BSQ mass range. We are able to avoid the impacts of changing BSQ mass range by monitoring an ion with higher $m/z$. Additionally, using an analyte for filtering reagent ion hysteresis means that we are using a direct measurement of the formation of analyte ions for ${\mathrm{NH}}_{\mathrm{4}}^{+}$ ionization and a direct measurement of contamination from other reagent ion chemistry with H₃O⁺ ionization. Therefore, in the absence of an internal standard (Sect. 3.4.2), we use the persistent ambient ion ${\mathrm{C}}_{\mathrm{3}}{\mathrm{H}}_{\mathrm{6}}\mathrm{O}\cdot {\mathrm{NH}}_{\mathrm{4}}^{+}$ to monitor hysteresis. This approach brings two major complications: (1) variable contributions of isomers with potentially disparate sensitivities (e.g., acetone and propionaldehyde) and (2) potentially variable ambient concentrations. A switch-by-switch analysis of hysteresis from MEFO (available as Fig. S6) shows that the 0.05 % s⁻¹ δ_n cutoffs for ${\mathrm{C}}_{\mathrm{3}}{\mathrm{H}}_{\mathrm{6}}\mathrm{O}\cdot {\mathrm{NH}}_{\mathrm{4}}^{+}$ calculated in Fig. 5b and d do not capture the majority of the switch-by-switch cutoffs (37 % for ${\mathrm{NH}}_{\mathrm{4}}^{+}$ and 39 % for H₃O⁺). Therefore, if a persistent ambient ion is used to diagnose hysteresis timescales, this should be done on a switch-by-switch basis. This variability may be associated with ambient variations in the ${\mathrm{C}}_{\mathrm{3}}{\mathrm{H}}_{\mathrm{6}}\mathrm{O}\cdot {\mathrm{NH}}_{\mathrm{4}}^{+}$ signal which can be avoided by applying our method described in Fig. 5 to a persistent and known signal from an internal standard (Sect. 3.4.2).

The choice of δ_n threshold represents a compromise between ion chemistry stability and data loss. For both reagent and analyte ions, the amount of data removed becomes very sensitive to a small decrease in the δ_n threshold below ∼0.05 % s⁻¹ (Fig. S4). During the deployment in MEFO, a 0.05 % s⁻¹ threshold applied to ${\mathrm{C}}_{\mathrm{3}}{\mathrm{H}}_{\mathrm{6}}\mathrm{O}\cdot {\mathrm{NH}}_{\mathrm{4}}^{+}$ results in the loss of ∼185 s (Fig. 5d) and ∼75 s (Fig. 5b) of data per switch with H₃O⁺ and ${\mathrm{NH}}_{\mathrm{4}}^{+}$ ionization, respectively. Optimizing ion chemistry stability while preserving data coverage results in the loss of ∼260 s (75 s for ${\mathrm{NH}}_{\mathrm{4}}^{+}$ and 185 s for H₃O⁺) of data on a reagent switching full cycle (1800 s), corresponding to ∼86 % data retention for 15 min switching intervals.

3.4.2 Filtering hysteresis with reagent and internal standard ions

When an internal standard signal is available, as in the ARTofMELT expedition (Sect. 2.6), reagent ion hysteresis can be more reliably monitored using known unique and persistent molecular ions. We applied the δ_n thresholding method to a 2-week period from the ARTofMELT data set (from 17 to 31 May 2023). We quantify the timescale of reagent ion hysteresis by monitoring the internal standard signal of 2-hexanone-d₄ as ${\mathrm{C}}_{\mathrm{6}}{\mathrm{H}}_{\mathrm{8}}{\mathrm{d}}_{\mathrm{4}}\mathrm{O}\cdot {\mathrm{NH}}_{\mathrm{4}}^{+}$ with ${\mathrm{NH}}_{\mathrm{4}}^{+}$ ionization (Fig. 6b) and as ${\mathrm{C}}_{\mathrm{6}}{\mathrm{H}}_{\mathrm{8}}{\mathrm{d}}_{\mathrm{4}}\mathrm{O}\cdot {\mathrm{H}}^{+}$ with H₃O⁺ ionization (Fig. 6d). For direct comparison to Sect. 3.4.1, we also use the reagent ion ${\mathrm{NH}}_{\mathrm{4}}\cdot {\mathrm{H}}_{\mathrm{2}}{\mathrm{O}}^{+}$ under both ${\mathrm{NH}}_{\mathrm{4}}^{+}$ (Fig. 6a) and H₃O⁺ (Fig. 6c). Notably, in this marine environment the ambient ${\mathrm{C}}_{\mathrm{3}}{\mathrm{H}}_{\mathrm{6}}\mathrm{O}\cdot {\mathrm{NH}}_{\mathrm{4}}^{+}$ signal is highly variable, precluding its use for filtering reagent ion hysteresis (Fig. S7), further motivating the use of an internal standard to diagnose reagent ion hysteresis. A 0.05 % s⁻¹ threshold applied to 2-hexanone-d₄ results in the loss of 34 s from ${\mathrm{NH}}_{\mathrm{4}}^{+}$ ionization after a switch (Fig. 6b) and 168 s from H₃O⁺ ionization (Fig. 6d). This results in ∼19 % of H₃O⁺ data being removed and ∼ 4 % of ${\mathrm{NH}}_{\mathrm{4}}^{+}$ ionization from 15 min switching, or ∼89 % total data retention over a full (1800 s) switching cycle. It is worth noting that under both ionization modes, the hysteresis timescale for the ${\mathrm{NH}}_{\mathrm{4}}\cdot {\mathrm{H}}_{\mathrm{2}}{\mathrm{O}}^{+}$ is longer (Figs. 6a, c, S8) than the 2-hexanone-d₄ internal standard ion. This suggests that a conservative approach to monitoring ion chemistry could be to use the reagent ions for establishing hysteresis timescales. However, the impacts of the BSQ on reagent ions raise concerns of how representative the reagent ion signal is of true composition in the fIMR (Krechmer et al., 2018; Khare et al., 2022; Xu et al., 2022). A switch-by-switch analysis of variation in the hysteresis timescale (available as Fig. S8) demonstrates that the cutoffs calculated in Fig. 6 for 2-hexanone-d₄ capture the majority of variability in the hysteresis timescale for independent switches (75 % for ${\mathrm{NH}}_{\mathrm{4}}^{+}$ and 72 % for H₃O⁺).

Our results using an internal standard for reagent ion hysteresis filtering are qualitatively similar to the analysis above (Sect. 3.4.1) using a persistent ambient ion, though fewer data are removed from ${\mathrm{NH}}_{\mathrm{4}}^{+}$ ionization. Rather than resulting from use of an ambient or internal standard ion, this difference in hysteresis time is likely the result of having more similar BSQ settings between the two ionization modes during the ARTofMELT campaign (Table S1). The smaller change in BSQ settings results in a faster change in instrument conditions impacting ion transmission. The approach we describe here can be applied easily to other instruments and at different instrument conditions (e.g., fIMR temperature and pressure) to best balance the need for both measurement stability and data coverage. While the use of an internal standard signal is ideal for diagnosing reagent ion switching hysteresis, the choice of both product molecular ions and rate-of-change threshold must be optimized for each application and sampling environment. Overall, our results are qualitatively similar to hysteresis timescales suggested by Zaytsev et al. (2019a) for ${\mathrm{NH}}_{\mathrm{4}}^{+}$ and H₃O⁺ switching in an Ionicon PTR3: ∼120 s and ∼60 s of data removed after switching to H₃O⁺ and ${\mathrm{NH}}_{\mathrm{4}}^{+}$ ionization, respectively.

Figure 7 Campaign mean mass defect plots for ${\mathrm{NH}}_{\mathrm{4}}^{+}$ and H₃O⁺ ionization from deployment in MEFO. The ${\mathrm{NH}}_{\mathrm{4}}^{+}$ mass spectrum is displayed as a left half circle and the H₃O⁺ is displayed as a right half circle. Points are sized by the average signal across the campaign divided by the detection limit signal (DL: 3σ of campaign zero air background). The reagent ion masses have been removed from the ion molecular mass. The top 100 ions in terms of signal-to-DL ratio for both reagent ions have been colored according to their molecular formulae with periods of reagent ion hysteresis removed (Sect. 3.4). A total of 725 ions are shown, and selected ions and their signal-to-DL ratios are displayed in a bar chart format in Fig. S9.

Figure 8 Selected campaign average high-resolution mass spectra from MEFO for (a, b) monoterpene (C₁₀H₁₆) and (c, d) monoterpene oxygenate (C₁₀H₁₆O₃) molecular ions detected with ${\mathrm{NH}}_{\mathrm{4}}^{+}$ (a, c, orange) and H₃O⁺ (b, d, blue).

3.5 Reagent ion comparison from ambient measurements in MEFO

We compare the capabilities of ${\mathrm{NH}}_{\mathrm{4}}^{+}$ and H₃O⁺ reagent ions in a single instrument using ambient observations from MEFO (Sect. 2.5). Switching between H₃O⁺ and ${\mathrm{NH}}_{\mathrm{4}}^{+}$ on a 15 min timescale over the 21 d deployment allows us to directly compare the two reagent ions in a predominantly biogenic environment (Hunter et al., 2017; Vermeuel et al., 2023a; Riches et al., 2024; Link et al., 2024). A single-instrument approach avoids inlet and instrument design influences on detection that would otherwise complicate a direct reagent ion comparison (e.g., Riva et al., 2019). Previously Zaytsev et al. (2019a) used a switching ${\mathrm{NH}}_{\mathrm{4}}^{+}/{\mathrm{H}}_{\mathrm{3}}{\mathrm{O}}^{+}$ Ionicon PTR3 with a modified helical tripole reaction chamber to measure products from the ⋅OH-initiated oxidation of 3-methylcatechol. This chamber study demonstrated the sets of compounds detected by each reagent ion and concluded that ${\mathrm{NH}}_{\mathrm{4}}^{+}$ is able to detect larger, more functionalized molecules, while H₃O⁺ is able to detect smaller organic molecules (Zaytsev et al., 2019a). To facilitate a direct and quantitative reagent ion comparison, we use the ratio of campaign average ambient signal to detection limit signal (i.e., 3σ of the background) for each ion as a measure of the signal-to-noise ratio (Fig. 7). With the assumption that ambient concentrations measured with each reagent ion over the campaign mean are equivalent, a higher signal-to-DL ratio also implies a lower detection limit. This analysis allows us to evaluate the relative capability of each reagent ion without direct calibrations for multifunctional biogenic organic compounds (e.g., Hunter et al., 2017; Vermeuel et al., 2023a; Link et al., 2024).

${\mathrm{NH}}_{\mathrm{4}}^{+}$ ionization detects oxygen-containing species with a higher signal-to-noise ratio than H₃O⁺. At MEFO, four series of C_xH_yO_z ions dominate our mass spectrum (highlighted in Fig. 7) – C₁₀H₁₆O_n, C₁₀H₁₄O_n, C₁₀H₁₈O_n, and C₅H₈O_n – which represent a mixture of biogenic terpenoid compounds and their early generation oxidation products. The C₁₀H₁₆O_n and C₁₀H₁₄O_n series suggest a mixture of primary emissions, such as citral (C₁₀H₁₆O; C^* = 1×10⁶ µg m⁻³) and thymol or carvone (C₁₀H₁₄O; C^* = 1–9×10⁵ µg m⁻³) (McKinney et al., 2011; Kaser et al., 2013; Vermeuel et al., 2023a), and oxidation products of other terpenoids. H₃O⁺ detects the reduced C₁₀H₁₄ species with higher signal-to-noise ratio compared to ${\mathrm{NH}}_{\mathrm{4}}^{+}$, with signal-to-DL ratios of 10.5 and 6.00, respectively. Similarly, H₃O⁺ detects C₁₀H₁₆ with a signal-to-DL ratio of 23.5, compared to 14.3 for ${\mathrm{NH}}_{\mathrm{4}}^{+}$. Following this C₁₀H₁₆O_n series, the n=1 ion is detected with a signal-to-DL ratio of 6.31 with H₃O⁺ and 10.8 with ${\mathrm{NH}}_{\mathrm{4}}^{+}$. The tendency toward increased signal-to-noise ratio with oxygenation for ${\mathrm{NH}}_{\mathrm{4}}^{+}$ ionization continues in the C₁₀H_14,16O_n series up to C₁₀H_14,16O₃ (with C^* between 4×10² and 9×10³ µg m⁻³, Table S7). The C₁₀H₁₈O_n series lacks a C₁₀H₁₈ ion, and the distribution in the x–y scatter between the C₁₀H₁₆ and C₁₀H₁₈O peaks is bimodal (Fig. S10), which suggests multiple paths to form C₁₀H₁₈O ions. These paths are likely (1) primary emissions of C₁₀H₁₈O terpenoid compounds with similar emission profiles to monoterpenes and (2) water clusters formed with monoterpenes (${\mathrm{C}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{16}}\cdot {\mathrm{H}}_{\mathrm{2}}\mathrm{O}\cdot {\mathrm{H}}^{+}$ and ${\mathrm{C}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{16}}\cdot {\mathrm{H}}_{\mathrm{2}}\mathrm{O}\cdot {\mathrm{NH}}_{\mathrm{4}}^{+}$). The C₅H₈O_n series represents a combination of fragments, primary emissions, and oxidation products. The C₅H₈ ion is likely a mixture of isoprene and fragments from larger oxygenates and 2,3,2-MBO with both reagent ions (e.g., Kilgour et al., 2024). Within the C₅H₈O_n series, the C₅H₈ signal-to-DL ratio is larger than expected with ${\mathrm{NH}}_{\mathrm{4}}^{+}$ ionization; we suspect that this is due partly to fragmentation of other compounds into the C₅H₈ mass. The ${\mathrm{NH}}_{\mathrm{4}}^{+}$ sensitivity to isoprene is very low; however, 2,3,2-MBO fragments substantially into C₅H₈ through dehydration of the tertiary alcohol group (Fig. 3). Fragmentation into C₅H₈ combined with a low background (Figs. A3, S11) leads to a very large signal-to-DL ratio (Fig. 7). The C₅H₈O₃ and C₅H₈O₄ peaks are likely oxidation products of isoprene and 2,3,2-MBO, while C₅H₈O₂ is likely an isoprene oxidation product (Saunders et al., 2003; Jenkin et al., 2015).

${\mathrm{NH}}_{\mathrm{4}}^{+}$ can detect organic nitrates that easily fragment with H₃O⁺ (Aoki et al., 2007; Duncianu et al., 2017) and so often go undetected in ambient measurements with H₃O⁺ ionization (Fig. 7). Organic nitrates ionized with H₃O⁺ fragment to form nitric acid (HNO₃) or nitronium ions (${\mathrm{NO}}_{\mathrm{2}}^{+}$), where the loss of HNO₃ results in fragmentation into the masses for other oxygenates (Aoki et al., 2007). The two predominant series of organic nitrate ions (C₁₀H₁₅NO_n and C₁₀H₁₇NO_n, with C^* down to tens of µg m⁻³; Table S7) are generally detected with a higher signal-to-DL ratio with ${\mathrm{NH}}_{\mathrm{4}}^{+}$ ionization compared to H₃O⁺ ionization, if the ion is detected with H₃O⁺ at all (Fig. 7). The exception is C₁₀H₁₅NO₃, which is detected at a higher signal-to-DL ratio (2.13) with H₃O⁺ ionization compared to ${\mathrm{NH}}_{\mathrm{4}}^{+}$ (0.774); though this ion may arise from dehydration of hydroxy nitrates (i.e., C₁₀H₁₇NO₄-H₂O), and H₃O⁺ is fragmenting larger organic nitrates into the C₁₀H₁₅NO₃ ion. The nitrates we observe are potentially a mixture of carbonyl, hydroxy, and peroxy nitrates derived from the oxidation of monoterpenes (C₁₀H₁₆) and potentially other terpenoid (${\mathrm{C}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{16}}\mathrm{O}/{\mathrm{C}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{14}}\mathrm{O}$) precursors (Table S8, Fry et al., 2013; Jenkin et al., 2015; Faxon et al., 2018; Bates et al., 2022). Additionally, C₅H₁₁NO₅ and C₅H₉NO₅ are detected using both reagent ions but both with a higher signal-to-DL ratio with ${\mathrm{NH}}_{\mathrm{4}}^{+}$. C₅H₁₁NO₅ is likely a nitrate from the oxidation 2,3,2-MBO, while both 2,3,2-MBO and isoprene could form the C₅H₉NO₅ ion (Link et al., 2024).

${\mathrm{NH}}_{\mathrm{4}}^{+}$ ionization's ability to detect oxygenated compounds with a higher signal-to-noise ratio and lower detection limits than H₃O⁺ arises from two main factors. First, ${\mathrm{NH}}_{\mathrm{4}}^{+}$ is a softer ionization method compared to H₃O⁺, resulting in less molecular ion fragmentation (Sect. 3.2). This is evident in our ambient data for the known H₃O⁺ monoterpene fragment, ${\mathrm{C}}_{\mathrm{6}}{\mathrm{H}}_{\mathrm{8}}\cdot {\mathrm{H}}^{+}$, compared to the analogous fragment with ${\mathrm{NH}}_{\mathrm{4}}^{+}$ ionization, ${\mathrm{C}}_{\mathrm{6}}{\mathrm{H}}_{\mathrm{8}}\cdot {\mathrm{NH}}_{\mathrm{4}}^{+}$. The fragment is present in ${\mathrm{NH}}_{\mathrm{4}}^{+}$ ionization mode at a 1 : 10 fragment-to-molecular-ion ratio, compared to a 1 : 1 ratio with H₃O⁺ under our fIMR conditions (Fig. S10). Second, ${\mathrm{NH}}_{\mathrm{4}}^{+}$ has higher selectivity toward oxygenates compared to H₃O⁺ (Sect. 3.2). Our ambient reagent ion switching observations further demonstrate this selectivity (Figs. 8, A3). The monoterpenes (C₁₀H₁₆) are easily distinguished from isobaric ions with both H₃O⁺ and ${\mathrm{NH}}_{\mathrm{4}}^{+}$, but oxygenates (C₁₀H₁₆O₃) have multiple isobaric interferences with H₃O⁺ (Fig. 8). Both higher selectivity and reduced fragmentation contribute to fewer isobaric ions with ${\mathrm{NH}}_{\mathrm{4}}^{+}$. While our observations demonstrate the utility of ${\mathrm{NH}}_{\mathrm{4}}^{+}$ for detecting oxidized species that H₃O⁺ ionization struggles to detect (e.g., Yuan et al., 2017; Riva et al., 2019; Pagonis et al., 2019; Coggon et al., 2024), the extent of fragmentation for specific compounds is difficult to diagnose in ambient and complex laboratory mass spectra. Overall, our ambient reagent ion comparison demonstrates quantitatively that ${\mathrm{NH}}_{\mathrm{4}}^{+}$ is complementary to H₃O⁺, and together these two reagent ions allow improved detection and identification of a range of biogenic reactive organic carbon compounds and their early generation oxidation products.

4 Conclusions

To expand the range of ROC detectable with a single chemical ionization instrument, we present an approach to combine two positive reagent ions, ${\mathrm{NH}}_{\mathrm{4}}^{+}$ and H₃O⁺, in a Vocus-CI-ToFMS. To accommodate the need for a constant ion–molecule reactor temperature during switching, we apply an $E/N$ space scanning approach to select fIMR conditions compatible with both reagent ions. We characterize the ability of ${\mathrm{NH}}_{\mathrm{4}}^{+}$ and H₃O⁺ to detect a range of reduced and oxygenated VOCs, SVOCs, and IVOCs through analysis of laboratory standards and find that H₃O⁺ detects reduced species well and fragments functionalized oxygenates away from the molecular ion, while ${\mathrm{NH}}_{\mathrm{4}}^{+}$ retains the molecular ion and allows for improved detection of oxygenates. We find that fragmentation generally correlated with $E/N$, while sensitivity is impacted by a combination of ion transmission, competing ionization pathways, and molecular ion fragmentation. To diagnose and quantify the timescales for reagent ion switching hysteresis, we compare the use of three ion types: ${\mathrm{NH}}_{\mathrm{4}}^{+}$ reagent ions, a persistent ambient ${\mathrm{NH}}_{\mathrm{4}}^{+}$ adduct ion, and ${\mathrm{NH}}_{\mathrm{4}}^{+}$ adduct or proton transfer molecular ions from an internal standard infused in the sampling inlet. Reagent ion signal variability at each switch is driven largely by changes in ion transmission, so it is less representative of ion chemistry, while monitoring a product ion is more directly related to ionization reactions taking place in the fIMR. An internal standard signal provides the ideal means to monitor reagent ion hysteresis with a known and persistent product ion; however, persistent ambient ions and internal standard product ions can produce similar rates of data retention (∼86 %–89 % data retention across a full 1800 s switching cycle with a 0.05 % s⁻¹ rate-of-change threshold). We deploy our ${\mathrm{NH}}_{\mathrm{4}}^{+}/{\mathrm{H}}_{\mathrm{3}}{\mathrm{O}}^{+}$ reagent ion switching Vocus-CI-ToFMS during a 3-week period at a rural pine forest (Manitou Experimental Forest Observatory) to facilitate a direct and quantitative reagent ion comparison. Our ambient observations demonstrate that ${\mathrm{NH}}_{\mathrm{4}}^{+}$ detects oxygenated ROC with a higher signal-to-noise ratio and lower DL, including organic nitrates that H₃O⁺ does not detect, while H₃O⁺ detects reduced species that are undetectable with ${\mathrm{NH}}_{\mathrm{4}}^{+}$. ${\mathrm{NH}}_{\mathrm{4}}^{+}/{\mathrm{H}}_{\mathrm{3}}{\mathrm{O}}^{+}$ reagent ion switching takes advantage of the complementary nature of the two reagent ions to expand the range of ROC detectable with a single instrument.

Appendix A

Figure A1 Normalized signal intensity for benzene measured with H₃O⁺ (${\mathrm{C}}_{\mathrm{6}}{\mathrm{H}}_{\mathrm{6}}\cdot {\mathrm{H}}^{+}$) from pressure–voltage-gradient scans.

Figure A2 Normalized signal intensity for camphor (a), 2-octanone (b), and trans-3-hexenol (c) measured with ${\mathrm{NH}}_{\mathrm{4}}^{+}$ (${\mathrm{C}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{16}}\mathrm{O}\cdot {\mathrm{NH}}_{\mathrm{4}}^{+}$, ${\mathrm{C}}_{\mathrm{8}}{\mathrm{H}}_{\mathrm{16}}\mathrm{O}\cdot {\mathrm{NH}}_{\mathrm{4}}^{+}$, and ${\mathrm{C}}_{\mathrm{6}}{\mathrm{H}}_{\mathrm{12}}\mathrm{O}\cdot {\mathrm{NH}}_{\mathrm{4}}^{+}$, respectively).

Figure A3 Mass-dependent campaign average detection limit signal for ions detected with ${\mathrm{NH}}_{\mathrm{4}}^{+}$ (orange) and H₃O⁺ (blue). The reagent ion masses have been removed from ion exact masses.