Characterisation of gaseous iodine species detection using the multi-scheme chemical ionisation inlet 2 with bromide and nitrate chemical ionisation methods

Abstract

The multi-scheme chemical ionisation inlet 1 (MION1) enables rapid switching between the measurement of atmospheric ions without chemical ionisation and neutral molecules using various atmospheric pressure chemical ionisation methods. In this study, we introduce the upgraded version, the multi-scheme chemical ionisation inlet 2 (MION2). The new design incorporates enhanced ion optics, resulting in increased reagent ion concentration, ensuring a robust operation, and enabling the use of multiple chemical ionisation methods with the same ionisation time.

In order to simplify the regular calibration of MION2, we developed an open-source flow reactor chemistry model called MARFORCE. This model enables quantification of the chemical production of sulfuric acid (${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$), hypoiodous acid ($\mathrm{HOI}$), and hydroperoxyl radical (${\mathrm{HO}}_{\mathrm{2}}$). MARFORCE simulates the convection–diffusion–reaction processes occurring within typical cylindrical flow reactors with uniform inner diameters. The model also includes options to simulate chemical processes in the following two scenarios: (1) when two flow reactors with different inner diameters are connected and (2) when two flows are merged into one using a Y-shaped tee, although with reduced accuracy. Furthermore, the chemical mechanism files in the model are compatible with the widely used Master Chemical Mechanism (MCM), allowing for future adaptation to simulate other chemical processes in flow reactors.

Furthermore, we conducted a comprehensive characterisation of the bromide (${\mathrm{Br}}^{-}$) and nitrate (${\mathrm{NO}}_{\mathrm{3}}^{-}$) chemical ionisation methods with different ionisation times. We performed calibration experiments for ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$, $\mathrm{HOI}$, and ${\mathrm{HO}}_{\mathrm{2}}$ by combining gas kinetic experiments with the MARFORCE model. The evaluation of sulfur dioxide (${\mathrm{SO}}_{\mathrm{2}}$), water (${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$), and molecular iodine (${\mathrm{I}}_{\mathrm{2}}$) involved dilution experiments from a gas cylinder (${\mathrm{SO}}_{\mathrm{2}}$), dew point mirror measurements (${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$), and a derivatisation approach combined with a high-performance liquid chromatography quantification (${\mathrm{I}}_{\mathrm{2}}$), respectively.

Our findings indicate that the detection limit is inversely correlated with the fragmentation enthalpy of the analyte–reagent ion (${\mathrm{Br}}^{-}$) cluster. In other words, stronger binding (resulting in a larger fragmentation enthalpy) leads to a lower detection limit. Additionally, a moderately longer ionisation time enhances the detection sensitivity, thereby reducing the detection limit. For instance, when using the ${\mathrm{Br}}^{-}$ chemical ionisation method with a 300 ms ionisation time, the estimated detection limit for ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ is $\mathrm{2.9}\times {\mathrm{10}}^{\mathrm{4}}$ molec. cm${}^{-\mathrm{3}}$. Notably, this detection limit is even superior to that achieved by the widely used Eisele-type chemical ionisation inlet ($\mathrm{7.6}\times {\mathrm{10}}^{\mathrm{4}}$ molec. cm${}^{-\mathrm{3}}$), as revealed by direct comparisons.

While the ${\mathrm{NO}}_{\mathrm{3}}^{-}$ chemical ionisation method remains stable in the presence of high humidity, we have observed that the ${\mathrm{Br}}^{-}$ chemical ionisation method (${\mathrm{Br}}^{-}$–MION2) is significantly affected by the air water content. Higher levels of air water lead to reduced sensitivity for ${\mathrm{HO}}_{\mathrm{2}}$ and ${\mathrm{SO}}_{\mathrm{2}}$ under the examined conditions. However, we have found that a sharp decline in sensitivity for ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$, $\mathrm{HOI}$, and ${\mathrm{I}}_{\mathrm{2}}$ occurs only when the dew point exceeds 0.5–10.5 ${}^{\circ }$C (equivalent to 20 %–40 % RH; calculated at 25 ${}^{\circ }$C throughout this paper). For future studies utilising the atmospheric pressure ${\mathrm{Br}}^{-}$ chemical ionisation method, including ${\mathrm{Br}}^{-}$–MION2, it is crucial to carefully consider the molecular-level effects of humidity. By combining approaches such as the water-insensitive ${\mathrm{NO}}_{\mathrm{3}}^{-}$–MION2 with ${\mathrm{Br}}^{-}$–MION2, MION2 can offer more comprehensive insights into atmospheric composition than what can be achieved by either method alone.

By employing instrument voltage scanning, chemical kinetic experiments, and quantum chemical calculations, we have conclusively established that the presence of iodine oxides does not interfere with the detection of ${\mathrm{HIO}}_{\mathrm{3}}$. Our comprehensive analysis reveals that the ions ${\mathrm{IO}}_{\mathrm{3}}^{-}$, ${\mathrm{HIO}}_{\mathrm{3}}\mathrm{⚫}{\mathrm{NO}}_{\mathrm{3}}^{-}$, and ${\mathrm{HIO}}_{\mathrm{3}}\mathrm{⚫}{\mathrm{Br}}^{-}$, which are detected using the ${\mathrm{Br}}^{-}$ and ${\mathrm{NO}}_{\mathrm{3}}^{-}$ chemical ionisation methods, are primarily, if not exclusively, generated from gaseous ${\mathrm{HIO}}_{\mathrm{3}}$ molecules within atmospherically relevant conditions.

1 Introduction

Chemical ionisation mass spectrometry (CIMS) has been widely used in atmospheric chemistry and aerosol formation studies due to its versatility and high sensitivity in measuring trace level gaseous species (see, e.g., ). With chemical ionisation methods, an analyte is charged either by (1) receiving charge (proton, electron, or ion) from the reagent ion or (2) forming a relatively stable cluster with the reagent ion. Mass spectrometers further measure the charged analyte-containing ions to obtain their molecular information.

Various chemicals have been employed as reagent ions in chemical ionisation methods. The commonly used reagent ions include nitrate (${\mathrm{NO}}_{\mathrm{3}}^{-}$; ), acetate (${\mathrm{C}}_{\mathrm{2}}{\mathrm{H}}_{\mathrm{3}}{\mathrm{O}}_{\mathrm{2}}^{-}$; ), iodide (${\mathrm{I}}^{-}$; ), hydronium (${\mathrm{H}}_{\mathrm{3}}{\mathrm{O}}^{+}$; ), and, sporadically, bromide (${\mathrm{Br}}^{-}$; ) and ammonium (${\mathrm{NH}}_{\mathrm{4}}^{+}$; ). These reagent ions transfer charges to or form clusters with distinct subsets of trace gases. However, the detection of an analyte-containing ion is influenced by its transmission through the mass spectrometer's ion optics, as collision-induced cluster fragmentation can diminish the analyte's signature. Analyte–reagent ion clusters with strong bonds have a higher likelihood of reaching the detector when compared to weakly bonded clusters . Hence, it is crucial to select a chemical ionisation method that preserves the analyte's signature. For example, the ${\mathrm{NO}}_{\mathrm{3}}^{-}$–CIMS has been widely used to detect sulfuric acid (${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$; ) and highly oxygenated organic molecules . ${\mathrm{I}}^{-}$–CIMS is regularly used to detect semi-volatile organic compounds , bromine and chlorine-containing species , and, for example, dinitrogen pentoxide (${\mathrm{N}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{5}}$; ). ${\mathrm{C}}_{\mathrm{2}}{\mathrm{H}}_{\mathrm{3}}{\mathrm{O}}_{\mathrm{2}}^{-}$–CIMS was used to detect small organic acids and highly oxygenated organic compounds . The bromide chemical ionisation method has recently been used to detect species such as ${\mathrm{HO}}_{\mathrm{2}}$ and ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ . The detection of a series of halogenated species by the ${\mathrm{Br}}^{-}$ chemical ionisation method was first demonstrated by . A detailed characterisation of the ${\mathrm{Br}}^{-}$ chemical ionisation method utilising the multi-scheme chemical ionisation inlet 1 (MION1) was presented in several of our earlier studies . Multiple species were successfully calibrated using either analytical methods or inter-instrument comparison, including ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$, ${\mathrm{I}}_{\mathrm{2}}$, ${\mathrm{Cl}}_{\mathrm{2}}$, and $\mathrm{HOI}$ . Among the calibrated species, ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ and ${\mathrm{I}}_{\mathrm{2}}$ were shown to be detected at the collision limit (highest sensitivity). Although ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ has been quantified using standard methods , the quantification of the measured ${\mathrm{I}}_{\mathrm{2}}\mathrm{⚫}{\mathrm{Br}}^{-}$ signal remains challenging. This is primarily contributed by the following two factors: (1) the current ${\mathrm{Br}}^{-}$–MION1 and ${\mathrm{Br}}^{-}$–MION2 have a detection upper limit of a few hundred pptv (parts per trillion by volume) of ${\mathrm{I}}_{\mathrm{2}}$, beyond which the reagent ions become depleted and the measurement is non-linear. (2) On the contrary, spectroscopic and other methods could be limited by their high detection limits and may not be able to detect ${\mathrm{I}}_{\mathrm{2}}$ at appropriate levels. Therefore, the key is to find sensitive methods to quantify gaseous ${\mathrm{I}}_{\mathrm{2}}$ at tens to hundreds of pptv levels.

Ideally, simultaneous measurement of all the mentioned analytes could be achieved by employing their corresponding CIMS methods concurrently in ambient observations or complex laboratory experiments. However, CIMS instruments are costly, and research institutes often face limitations due to the availability of such instrumentation. As an alternative approach, chemical ionisation inlets that have the capability to switch between different reagent ions can be employed. Many switchable systems have been developed previously, such as for proton transfer reaction mass spectrometers and other chemical ionisation mass spectrometers . A common feature of these techniques is using a reduced-pressure ion molecule reaction chamber, thus unavoidably diluting the gas molecules of interest by orders of magnitude. While the detection limit of an instrument is also affected by other factors, it is commonly observed that chemical ionisation inlets operating at reduced pressures have higher limits of detection compared to atmospheric pressure chemical ionisation inlets. For instance, reduced pressure inlets reported detection limits of various organic compounds from sub-pptv to hundreds of pptv levels , while the best-performing atmospheric pressure chemical ionisation inlets regularly detect vapours at ppqv (parts per quadrillion by volume) levels for selected acids and highly oxygenated organic vapours using the same time-of-flight mass spectrometer .

To reduce the detection limit of switchable reagent ion chemical ionisation systems, we developed the MION1 inlet, which allows for fast switching of reagent ion chemistry at atmospheric pressure . This technique has predominantly been employed for the detection of sulfuric acid and halogenated species, using either the ${\mathrm{NO}}_{\mathrm{3}}^{-}$ or ${\mathrm{Br}}^{-}$ chemical ionisation methods . However, there are some remaining issues with the MION1. First, its limit of detection is higher compared to another commonly used atmospheric pressure chemical ionisation inlet referred to as the Eisele inlet . Second, the ionisation times for different chemical ionisation methods have to be different due to the design, which involves aligning and attaching the chemical ionisation units at varying distances on a cylindrical tube. These challenges may restrict its suitability for detecting vapours at extremely low concentrations (e.g. at ${\mathrm{10}}^{\mathrm{5}}$ to ${\mathrm{10}}^{\mathrm{6}}$ molec. cm${}^{-\mathrm{3}}$ or 4 to 40 ppqv) and interpreting the species detected by different chemical ionisation methods.

In this study, we introduce an upgraded version of the multi-scheme chemical ionisation inlet (MION), referred to as MION2, which specifically addresses these issues. We conducted laboratory experiments to characterise the performance of this inlet, using analytical methods and a newly developed open-source kinetic model. As chemical ionisation methods based on halogen anions (such as ${\mathrm{I}}^{-}$) are commonly influenced by the water content in the air , we also systematically examined the impact of air water content on the detection of ${\mathrm{Br}}^{-}$–MION2.

2 Methods

2.1 Characterisation of the MION2 inlet

The ionisation inlet utilised in this study is the upgraded multi-scheme chemical ionisation inlet, MION2, developed by Karsa Ltd. This inlet is specifically designed to enable the measurement of neutral molecules using chemical ionisation methods, while also facilitating the detection of atmospheric ions by disabling chemical ionisation. It offers the capability of rapid switching between two or more chemical ionisation methods, allowing for selective measurement of gaseous species at ambient pressure. Currently, the MION2 inlet supports up to six ion sources.

Due to the geometric limitations of the previous MION1 inlet, the different ionisation sources in MION1 have to employ different ionisation times. The ionisation time is defined by the sample flow rate and the distance between the ion injection port and the instrument pinhole (refer to Fig. ). The improved geometry of the MION2 inlet overcomes this limitation, allowing for the operation of three bipolar ion sources per ionisation time, which are all positioned at the same distance from the pinhole. Additionally, for the longer ionisation time, the length of the connecting pipe between the sources can be adjusted, providing flexibility in modifying the ionisation time.

Figure 1

Schematic of the MION2 inlet illustrating its gas flows and ion paths. The new design increases the primary ion concentration and allows the operation of multiple chemical ionisation methods with the same ionisation time. L1 and L2 refer to the distances between the ion sources and the pinhole of the mass spectrometer. The exhaust flows are connected to two ports in the middle of the ion source.​​​​​​​

[Figure omitted. See PDF]

In this study, we employed the MION2 inlet with two chemical ionisation methods, namely ${\mathrm{NO}}_{\mathrm{3}}^{-}$ and ${\mathrm{Br}}^{-}$, along with two different ionisation times (35 and 300 ms, respectively). This configuration was chosen to investigate the characteristics of the inlet. To facilitate clear referencing, we designate the ion source positioned 3 cm away from the mass spectrometer as tower 1 (T1), while the source located 25 cm away from the mass spectrometer is referred to as tower 2 (T2) throughout this paper (see Fig. ).

Figure  illustrates the conceptual schematic of one of the ion sources, depicting the airflow and ion paths. The entire source is attached to a 24 mm inner diameter tube that is electrically grounded. The sample flow, which is provided by a mass flow controller (MFC) connected to a vacuum pump, is set at a rate of 22.5 standard litres per minute (L min${}^{-\mathrm{1}}$). The target molecules undergo ionisation by reacting with the reagent ions (${\mathrm{NO}}_{\mathrm{3}}^{-}$ or ${\mathrm{Br}}^{-}$).

In this configuration, the ionisation time for the target molecules and charged reagent ions is approximately 35 ms for tower 1 and 300 ms for tower 2. A neutral reagent inflow is introduced, which consists of nitrogen or air enriched with reagent vapour. The reagent vapour is generated by passing nitrogen or air over a liquid reagent (nitric acid, ${\mathrm{HNO}}_{\mathrm{3}}$, or dibromomethane, ${\mathrm{CH}}_{\mathrm{2}}{\mathrm{Br}}_{\mathrm{2}}$, in this study). The resulting mixture is then fed into the ion source, where it is ionised by a soft X-ray source (Hamamatsu L12535; 4.9 keV). The charged reagent ions are guided into the sample flow by an electric field within the ion source. This electric field is generated using concentric stainless-steel electrode plates with orifices of different sizes (ranging from 5–10 mm in diameter), with resistances placed between each pair of plates. Two high voltages (approximately 2500 and 250 V) are used in the inlet. The lower of the two voltages determines whether the reagent ions pass through the final orifice in the deflector electrode, effectively controlling the ionisation process and enabling the rapid switching between ion sources.

In contrast to the MION1 design , which relied on the reagent inflow and exhaust flow to define the source reagent flow, MION2 incorporates an additional purge flow to prevent the sample flow from entering the ion source. The purge flow consists of the same nitrogen or air used to generate the reagent flow. Upon entering the ion source, the purge flow splits into two streams; one stream prevents the sample flow from entering, while the other stream ensures that the neutral reagent does not enter the sample flow.

In MION2, the typical flow rates for the reagent, purge, and exhaust are 10, 100, and 50 standard cubic centimetres per minute (m${}^{\mathrm{3}}$ min${}^{-\mathrm{1}}$), respectively. The reagent concentration in the ion source is estimated to be $\mathrm{2}\times {\mathrm{10}}^{\mathrm{17}}$ cm${}^{-\mathrm{3}}$. This design effectively addresses the challenges faced in MION1, where a compromise had to be made between the risk of contaminating the sample pipe with a neutral reagent or introducing sample air into the ion source, potentially leading to contamination or uncontrolled ion chemistry and resulting in detection biases. In MION2, the water vapour and other contaminants in the sample flow do not have the opportunity to oxidise the surfaces of the electrodes inside the ion source. Such oxidation would result in reduced ion transmission from the ion source to the sample flow. Operational testing during ambient measurements has demonstrated that MION2 exhibits significantly improved stability compared to MION1. For example, recent measurements conducted at a coastal site in Finland involved the uninterrupted operation of MION2 for at least 2 months.

Additionally, the upgraded ion optics inside the ion sources of MION2 have increased the transmission of reagent ions and the observed reagent ion concentration at the mass spectrometer by approximately 1 order of magnitude when compared to MION1. This improvement was achieved by modifying the last electrode within the ion source to minimise the ion residence time and reduce the diffusion losses of ions.

2.2 Experimental set-up

2.2.1 Calibration of inorganic species

The experimental set-up used for characterising the MION2 inlet is illustrated in Fig. . It consists of three main sections, namely the flow reactor section, the MION2 chemical ionisation inlet section, and an atmospheric pressure interface time-of-flight mass spectrometer (APi-TOF; Aerodyne Inc.; ).

The flow reactor section includes a calibration source and several gas feeds. Synthetic air (Woikoski Oy, Finland; purity $\ge$ 99.999 %, with 20.9 % ${\mathrm{O}}_{\mathrm{2}}$), nitrogen (${\mathrm{N}}_{\mathrm{2}}$; Woikoski Oy, Finland; purity $\ge$ 99.999 %), and sulfur dioxide (${\mathrm{SO}}_{\mathrm{2}}$; Air Products, USA; 99.5 % purity) were injected into the system using mass flow controllers connected to standard gas cylinders or tanks. These gases were pre-mixed before reaching the calibration source.

${\mathrm{I}}_{\mathrm{2}}$ was generated either from a homemade permeation tube or a commercial permeation tube (VICI Metronics). A stream of nitrogen (50 m${}^{\mathrm{3}}$ min${}^{-\mathrm{1}}$) was passed over the permeation tube at controlled temperatures ranging from 120 to 140 ${}^{\circ }$C. The temperature of the permeation tubes was regulated using an electronically controlled heating mantle, allowing for adjustable yet stable iodine concentrations. Water vapour (${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$) was controlled by adjusting the flow of nitrogen through a water bubbler, providing a controllable source of humidity.

The calibration source was mainly used to calibrate ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$, ${\mathrm{HO}}_{\mathrm{2}}$, and $\mathrm{HOI}$. The ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ calibration was detailed in , and the $\mathrm{HOI}$ calibration was presented in and . The calibration source in the experimental set-up was constructed using an aluminium box that encloses a $\mathrm{3}/\mathrm{4}$ in. (1.905 cm) quartz tube. The quartz tube was chosen for its high transmission properties for ultraviolet (UV) light emitted from a mercury lamp. Adjacent to the quartz tube, the mercury lamp is housed in an aluminium block that contains a filter-covered hole. The filter used in the aluminium block allows for high transmission of 185 nm light emitted from the lamp. This specific wavelength of light is effective in photolysing ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$ molecules, thus generating OH radicals.

Before conducting the calibration experiment, a mixed flow of nitrogen (${\mathrm{N}}_{\mathrm{2}}$), oxygen (${\mathrm{O}}_{\mathrm{2}}$), ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$, and either ${\mathrm{SO}}_{\mathrm{2}}$ or ${\mathrm{I}}_{\mathrm{2}}$ was continuously passed through the calibration source. This flow ensures that the source is thoroughly flushed with the desired gases and vapours, thereby creating a controlled environment for subsequent calibration measurements. The produced OH radicals from the calibration source then undergo reactions with an excess amount of ${\mathrm{SO}}_{\mathrm{2}}$ or a moderate amount of ${\mathrm{I}}_{\mathrm{2}}$ to produce ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ or $\mathrm{HOI}$ as the final products, respectively. Additionally, the ${\mathrm{HO}}_{\mathrm{2}}$ radical is produced as a byproduct of the ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ calibration process.

To quantify the concentrations of ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$, $\mathrm{HOI}$, and ${\mathrm{HO}}_{\mathrm{2}}$, an open-source Python library based on two-dimensional convection–diffusion–reaction equations was developed (Marine Atmospheric paRticle FORmation and ChEmistry, MARFORCE; ). This library aims to provide a framework for performing similar calibration tasks. Furthermore, it also allows users to simulate and predict concentrations of other chemical species by adopting different chemical reaction schemes. The MARFORCE library can be used as a tool in future research endeavours involving flow reactor chemistry simulations.

The ${\mathrm{SO}}_{\mathrm{2}}$ calibration is straightforward. The ${\mathrm{SO}}_{\mathrm{2}}$ flow from the ${\mathrm{SO}}_{\mathrm{2}}$ gas cylinder was diluted with humidified nitrogen, and the mixed sample was fed into the inlet. The normalised ${\mathrm{SO}}_{\mathrm{2}}\mathrm{⚫}{\mathrm{Br}}^{-}$ signal was further compared with the estimated ${\mathrm{SO}}_{\mathrm{2}}$ concentration to derive a calibration factor.

To calibrate the absolute concentration of ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$, a dew point mirror hygrometer (DewMaster Chilled Mirror Hygrometer; Edgetech) was employed. The dew point mirror hygrometer drew a sample from a branch of the humidified flow before it entered the MION2 inlet tube. By measuring the dew point temperature, the dew point mirror hygrometer provides an accurate and reliable determination of the absolute ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$ concentration in the sample. This calibration method ensures precise measurement of ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$ concentration, which is important for accurate analysis and interpretation of the experimental data.

2.2.2 Development of an iodine source

To calibrate the measured signals of ${\mathrm{I}}_{\mathrm{2}}\mathrm{⚫}{\mathrm{Br}}^{-}$ in ${\mathrm{Br}}^{-}$–MION2, we acquired its stable signals by utilising ${\mathrm{I}}_{\mathrm{2}}$ emitted from a permeation tube, which was regulated at a constant temperature and subjected to a continuous nitrogen stream (50 m${}^{\mathrm{3}}$ min${}^{-\mathrm{1}}$).

The key to this calibration is determining the quantities of ${\mathrm{I}}_{\mathrm{2}}$ emitted from the permeation tube. We have previously calibrated the ${\mathrm{I}}_{\mathrm{2}}$ measurement of ${\mathrm{Br}}^{-}$–MION1 using a cavity-enhanced differential optical absorption spectroscopy (CE-DOAS) instrument , an UV–Vis (visible) spectrophotometer and an inductively coupled plasma mass spectrometer (ICP–MS; ). As none of these instruments was available for this study, we further adapted an alternative method.

The collection of the ${\mathrm{I}}_{\mathrm{2}}$ sample followed exactly the same procedure as described in our previous studies . Briefly, 50 m${}^{\mathrm{3}}$ min${}^{-\mathrm{1}}$ nitrogen carrier gas flow was passed through an ${\mathrm{I}}_{\mathrm{2}}$ permeation tube for 300 min under 120–140 ${}^{\circ }$C. The nitrogen carrier stream containing the released ${\mathrm{I}}_{\mathrm{2}}$ was bubbled through a Schlenk-type impinger charged with 20 mL of hexane kept at $-$70 ${}^{\circ }$C by a dry ice and acetone bath. After completion of the sampling process, the absorption flasks were allowed to warm to the ambient temperature and sealed with a Teflon-coated glass stopper. The solution was stored at 4 ${}^{\circ }$C until further processing.

Inspired by , ${\mathrm{I}}_{\mathrm{2}}$ was converted into a non-volatile and stable derivative, followed by quantification of the latter using gas or liquid chromatography. quantified ${\mathrm{I}}_{\mathrm{2}}$ in aqueous matrices by gas chromatography–mass spectrometry (GC–MS) after ${\mathrm{I}}_{\mathrm{2}}$ reacted with 2,6-dimethylaniline to form the corresponding 4-iodo derivative.

An adaptation of this method was required, as the iodine to be determined was diluted in hexane. Specifically, the iodine derivatisation reaction was conducted directly with the hexane solutions in the presence of an aqueous buffer to reduce the losses associated with a hexane-to-water transfer. To avoid any losses of the volatile ${\mathrm{I}}_{\mathrm{2}}$ through evaporation, the reaction was conducted in hermetically sealed headspace vials, with efficient phase mass transfer being facilitated by vigorous magnetic stirring.

Control of the pH of the buffer was crucial for achieving high derivatisation yields, with pH at 7.00 providing the most favourable level of conversion after 2 h. Attempts to perform the derivatisation reaction under homogeneous conditions in hexane in the presence of a variety of soluble organic bases (e.g. tertiary amines) returned poor yields and led to the formation of several side products, most probably due to iodine oxidation. Experiments using 1.00 mL aliquots of the ${\mathrm{I}}_{\mathrm{2}}$ sample solutions under investigation produced the derivative at the limit of detection, precluding a reliable quantification of the derivative by the reverse-phase high-performance liquid chromatography (RP–HPLC–UV).

To improve the analytical sensitivity, 10 mL aliquots of the iodine sample solution were employed for derivatisation. To boost the sensitivity further, a high volume (15 $\mathrm{\mu }$L) of the concentrated derivatisation solution was injected into the HPLC system. Unfortunately, the hexane in the injection solution and the high injection volume gave rise to retention time instability and peak distortion. Subsequent optimisation of the chromatographic method provided robust reverse-phase chromatographic conditions. Specifically, this was achieved by using relatively weakly eluting isocratic conditions for sample elution, followed by strongly eluting conditions for column cleaning and reconditioning. Using the fully optimised protocol, the derivative could be readily quantified for 0.17 to 11.05 $\mathrm{\mu }$g mL${}^{-\mathrm{1}}$ initial iodine concentrations, with the LOD (limit of detection) and LOQ (limit of quantification) being 0.012 and 0.035 $\mathrm{\mu }$g mL${}^{-\mathrm{1}}$. Using this method, the hexane solution obtained by absorption of iodine from the permeation tube was found to contain 0.26 $\mathrm{\mu }$g iodine per millilitre. Considering a total sample volume of 20 mL, the iodine output rate of the permeation tube under the employed conditions was calculated to be 17.3 ng min${}^{-\mathrm{1}}$.

It is worth noting that the sensitivity of the current method can be further improved by employing more sensitive separation and/or detection techniques, e.g. liquid chromatography–mass spectrometry (LC–MS) or GC–MS.

2.2.3 Humidity dependence of analyte detection

An integral aspect of the characterisation involves investigating the influence of water on the detection of MION2 when employing the bromide chemical ionisation method. As water is essential in the calibration source to generate $\mathrm{OH}$ radicals, which subsequently yield either ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ or $\mathrm{HOI}$, we incorporated an additional dilution flow that merges with the calibration source through a Y-shaped piece (refer to Fig. ). This experimental configuration allows for the alteration of the absolute humidity of the sample, independent of the $\mathrm{OH}$ production rate in the calibration source. During the experiments with varying humidity, the total flows of the dilution part and the flow reactor section were kept constant, while the relative humidity of the dilution flow was varied by mixing different combinations of dry and humidified flows. By employing this approach, we maintained a consistent level of systematic errors arising from the blending of the dilution and sample flows. By comparing the relative signal intensities of analyte-containing ions, we could examine the influence of water on the detection of different analytes (e.g. ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$, $\mathrm{HOI}$ and ${\mathrm{HO}}_{\mathrm{2}}$).

2.2.4 Quartz flow reactor set-up

In order to study the sensitivity of ${\mathrm{Br}}^{-}$–MION2 to other oxidised iodine species, e.g. $\mathrm{IO}$, $\mathrm{OIO}$, ${\mathrm{HIO}}_{\mathrm{3}}$, ${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{3}}$, ${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{4}}$ and ${\mathrm{HIO}}_{\mathrm{2}}$, a quartz flow reactor with an inner diameter of 2.4 cm and a length of 94 cm was used. The residence time inside the quartz tube was 8.5 s. A green LED, with a wavelength of 528 nm, was hung on top and in parallel to the quartz flow reactor to initiate iodine photochemistry. In order to keep the temperature and light uniformity in the quartz flow reactor, the flow reactor was wrapped together with the green LED light by aluminium foil. The schematics of the set-up are shown in Fig. .

2.3 MARFORCE model description

As described above, calibration of ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$, ${\mathrm{HO}}_{\mathrm{2}}$, and $\mathrm{HOI}$ requires a numerical model to simulate the radial diffusion, chemical reactions, and transport in the calibration source and inlet tube. These processes determine the concentration of the analyte and can be simplified into a two-dimensional convection–diffusion–reaction problem. The concept of such a model was illustrated elsewhere , specifically for the calibration of ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$. Our earlier studies also presented a numerical model for $\mathrm{HOI}$ calibration with similar principles, but a simplified iodine chemistry scheme was instead implemented . Nevertheless, neither of these studies made their calibration scripts publicly accessible, and the scripts lack adaptability for different chemistry schemes. Consequently, we have developed an open-source two-dimensional flow reactor model named MARFORCE to address these limitations. MARFORCE is built in Python and hosted on GitHub , allowing free access to interested users. The model comprises two main components, namely (1) the fluid dynamics simulation module and (2) the gas-phase photochemistry module.

2.3.1 Convection–diffusion–reaction equation

The convection–diffusion–reaction equation has been derived and discussed extensively in the literature and is only briefly discussed here.

1$$\begin{array}{rl}{\displaystyle \frac{\partial c_i}{\partial t}}& =D_i\left({\displaystyle \frac{\mathrm{1}}{r}}{\displaystyle \frac{\partial c_i}{\partial r}}+{\displaystyle \frac{{\partial }^{\mathrm{2}}{c}_i}{\partial r^{\mathrm{2}}}}+{\displaystyle \frac{{\partial }^{\mathrm{2}}{c}_i}{\partial z^{\mathrm{2}}}}\right)\\ & -{\displaystyle \frac{\mathrm{2}Q}{\mathit{\pi }{R}^{\mathrm{2}}}}\left(\mathrm{1}-{\displaystyle \frac{r^{\mathrm{2}}}{R^{\mathrm{2}}}}\right){\displaystyle \frac{\partial c_i}{\partial z}}+P,\end{array}$$ where $i$ corresponds to a specific chemical (e.g. ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$), $c_i$ is the concentration, $D_i$ is the diffusion coefficient, $r$ is the distance in the radial direction, $R$ is the radius of the flow reactor, $Q$ is the total flow in the flow reactor, $z$ is the distance in the axial direction, and $P$ shows the production (positive values) or loss (negative values) rate due to chemical reactions. As the flow in tangential direction is symmetrical, the $\frac{\mathrm{1}}{r^{\mathrm{2}}}\frac{{\partial }^{\mathrm{2}}{c}_i}{\partial {\mathit{\theta }}^{\mathrm{2}}}$ term has been ignored.

The diffusion coefficient in the model can be defined in the following three ways: (1) defined manually, using experimental values; (2) calculated by kinetic theory; or (3) calculated based on elemental composition using Fuller's method .

The convection and diffusion processes were validated against a theoretical prediction by . A fixed amount of ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ was set at the first cross section of the MARFORCE simulation, and ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ was further carried to the outlet of a cylinder only by convection and diffusion processes. Comparing the ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ profiles at the outlet yields on average a 0.4 % difference between the MARFORCE model and the theoretical prediction by Fig. . This suggests that the convection and diffusion processes in the MARFORCE model are well simulated.

2.3.2 Gas-phase photochemistry

The photolysis and chemical reactions in the ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$, ${\mathrm{HO}}_{\mathrm{2}}$ and $\mathrm{HOI}$ calibrations can be simulated by a set of differential equations which describe the production and loss of various species. To make the MARFORCE model more versatile, the model was designed to accommodate the input file format from the Master Chemical Mechanism (MCM; ), a near-explicit chemistry mechanism for numerous organic precursors. The scripts used to compile and interpret MCM mechanisms were adapted from . The input file extracted from MCM is reshaped, and the reaction equations, reaction rate coefficients, reactants, products, their indices, and stoichiometric numbers are generated accordingly. The temperature and pressure dependence of reaction rate coefficients are taken into consideration. Finally, differential equations for each species based on its production and loss processes are produced and solved. Additionally, the MARFORCE model leaves an option to set abundant species as constants, so their concentrations are assumed uniform and homogeneous in the flow reactor. These species include, for example, ${\mathrm{O}}_{\mathrm{2}}$, ${\mathrm{N}}_{\mathrm{2}}$, ${\mathrm{SO}}_{\mathrm{2}}$, ${\mathrm{I}}_{\mathrm{2}}$, and ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$ in the ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ and $\mathrm{HOI}$ calibration experiments. With its flexibility, the MARFORCE model can be readily adapted to simulate organic oxidation or any other experiments using a laminar flow reactor.

There are two default chemistry schemes provided in the MARFORCE model, and they are used for the ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ calibration and $\mathrm{HOI}$ calibration, respectively. The reaction rate coefficients utilised in these two schemes are tabulated in Table . The most important procedure of these calibration experiments is to obtain the $\mathrm{OH}$ concentration. The $\mathrm{OH}$ concentration is determined by the photon intensity produced by the calibration source (It-product) and the absolute water content in the air passing through the calibration source. It-product refers to the product of UV light intensity at 185 nm and effective illumination time. In this study, we derived the It-product from the ${\mathrm{N}}_{\mathrm{2}}\mathrm{O}$ experiment, which was conducted under the same conditions as the ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ calibration experiments. The details of the It-product determination can be found in . In brief, the chemical actinometry method was employed, which involves the conversion of ${\mathrm{N}}_{\mathrm{2}}\mathrm{O}$ to ${\mathrm{NO}}_x$ (primarily $\mathrm{NO}$), to determine the light intensity. Since $\mathrm{NO}$ exhibits lower reactivity compared to $\mathrm{OH}$ and can be conveniently measured using commercial ${\mathrm{NO}}_x$ monitors, the It-product can be derived accordingly. Considering that the calibration experiment duration is relatively short (a few hours) compared to the potential lifetime of the mercury lamp, it is reasonable to assume that the attenuation of the It-product over time is negligible.

The $\mathrm{OH}$ initial concentration is further defined as

2$$\left[\mathrm{OH}\right]=\mathrm{It}\times {\mathit{\sigma }}_{{\mathrm{H}}_{\mathrm{2}}\mathrm{O}}\times {\mathrm{\Phi }}_{{\mathrm{H}}_{\mathrm{2}}\mathrm{O}}\times \left[{\mathrm{H}}_{\mathrm{2}}\mathrm{O}\right],$$ where ${\mathit{\sigma }}_{{\mathrm{H}}_{\mathrm{2}}\mathrm{O}}$ is the absorption cross section of water vapour, $\mathrm{7.22}\times {\mathrm{10}}^{-\mathrm{22}}$ cm${}^{\mathrm{2}}$ , and ${\mathrm{\Phi }}_{{\mathrm{H}}_{\mathrm{2}}\mathrm{O}}$ is the quantum yield (unity in this case).

2.3.3 Flow mixture

In addition to its ability to simulate a standard cylindrical flow reactor with uniform size, the MARFORCE model also possesses limited capabilities in two specific conditions, including (1) simulating two interconnected flow reactors with varying sizes; the model is capable of simulating scenarios where two flow reactors of different sizes are connected. Moreover, (2) simulating reactions when a dilution flow is merged with the sample flow through a Y-shaped tee is also possible. These additional features enhance the versatility of the MARFORCE model, allowing for a more comprehensive analysis of complex flow reactor systems.

The first design aims to cope with the different sizes of the chemical ionisation inlet and the calibration source itself. For example, the MION2 inlet utilises a KF25 connector with an inner diameter of 24 mm, while the calibration source utilises a $\mathrm{3}/\mathrm{4}$ in. (1.905 cm) tube with an inner diameter of 15.6 mm. Our model considers an instantaneous transition between the tubes of different sizes; i.e. the chemical distributions at the last cross section of the first cylinder are copied into the first cross section of the second cylinder, while the axial flow speed is adjusted to the cross section of the second cylinder. As this simplification ignores the convective transport of species to the walls at the transition region, it likely gives the concentration upper limit at the pinhole of the mass spectrometer. Since the inner diameter difference between the calibration source and the MION2 inlet is relatively small in this study, we expect that the resulting uncertainty is well within the overall systematic uncertainty of $-$50 %/$+$100 %.

The second design considers that an additional dilution flow is utilised to reduce the sample water content when entering the ${\mathrm{Br}}^{-}$–MION2 inlet. Similarly, we assume an instantaneous transition at the spot where the dilution flow is added. In this case, both the chemical distribution and axial flow speed are changed, since a new branch of flow is added. The simulation is carried out with a two-process procedure, i.e. before and after the dilution. First, we carry out a standard simulation before adding the dilution flow. Once the flow is fully developed and the chemical distribution reaches a steady state in the simulation, the last cross section at the grid right before adding the dilution flow is stored and recalculated into the first cross section of the next simulation. The second simulation is further carried out after considering the dilution flow, together with the changes in chemical distribution and axial flow speed.

It should be noted that the fluid dynamics processes are overly simplified in the second design, and therefore, this option should be used with caution. In this study, this option is necessary only because investigating the detection humidity effect of e.g. ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$, ${\mathrm{HO}}_{\mathrm{2}}$ and $\mathrm{HOI}$ requires adding a dilution flow after the calibration source. In order to estimate the magnitude of error caused by the simplification of fluid dynamics, we carried out experiments comparing calibration results obtained with the first design (straight tube) and the second design (Y piece), and the results are shown in Fig. . We find that the second design additionally introduces a 12 % higher calibration factor in the ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ calibration and a 27 % higher calibration factor in the $\mathrm{HOI}$ calibration when compared with the calibrations using the first design. Therefore, the application of the second design for the purpose of this study is reasonable and does not introduce excess uncertainties. This mainly concerns the ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$, ${\mathrm{HO}}_{\mathrm{2}}$ and $\mathrm{HOI}$ calibration experiments.

2.4 Quantum chemical calculations

Cluster fragmentation enthalpies were calculated using quantum chemical methods. The initial conformational sampling was performed using the Spartan'18 program. The cluster geometry was then optimised using density function theory (DFT) methods at the $\mathit{\omega }$B97X-D/aug-cc-pVTZ-PP level of theory . Iodine and bromine pseudopotential definitions were taken from the Environmental Molecular Sciences Laboratory (EMSL) basis set library . Calculations were carried out using the Gaussian 16 program . An additional coupled-cluster single-point energy correction at the DLPNO-CCSD(T)/def2-QZVPP level of theory was carried out on the lowest energy conformers to refine the DFT calculated enthalpies. The coupled-cluster calculation was performed using the ORCA program version 4.2.1 .

The master equation solver for multi-energy well reactions (MESMER) program was used to investigate the ionisation chemistry of ${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{3}}\mathrm{⚫}{\mathrm{HNO}}_{\mathrm{3}}{\mathrm{NO}}_{\mathrm{3}}^{-}$. For the ${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{3}}\mathrm{⚫}{\mathrm{HNO}}_{\mathrm{3}}{\mathrm{NO}}_{\mathrm{3}}^{-}$ complex, Lennard–Jones potentials of $\mathit{\sigma }$ $=$ 6.5 Å and $\mathit{ϵ}$ $=$ 300 K were used, which are identical to those used previously for similar iodine systems . The MesmerILT method was used with a pre-exponential value of $\mathrm{1.26}\times {\mathrm{10}}^{-\mathrm{9}}$ cm${}^{\mathrm{3}}$ molec.${}^{-\mathrm{1}}$ s${}^{-\mathrm{1}}$, which is equal to the ${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{3}}$ $+$ ${\mathrm{HNO}}_{\mathrm{3}}{\mathrm{NO}}_{\mathrm{3}}^{-}$ collision rate calculated using the average dipole orientation (ADO) method. Varying the collision rate by a factor of 3 has no effect on the MESMER results, indicating that the reported final fragmentation rate coefficients of ${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{3}}\mathrm{⚫}{\mathrm{HNO}}_{\mathrm{3}}{\mathrm{NO}}_{\mathrm{3}}^{-}$ are not sensitive to the accuracy of the computed collision rate.

3 Results and discussion

3.1

Calibration of ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$, $\mathrm{HOI}$, and ${\mathrm{HO}}_{\mathrm{2}}$ using MARFORCE

Gaseous ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ concentration is regularly measured around the globe using the nitrate chemical ionisation method. In this study, a direct ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ calibration has been carried out for the MION2 inlet at tower 1, using either ${\mathrm{Br}}^{-}$ (${\mathrm{Br}}^{-}$–MION2–T1; Fig. ) or ${\mathrm{NO}}_{\mathrm{3}}^{-}$ (${\mathrm{NO}}_{\mathrm{3}}^{-}$–MION2–T1) chemical ionisation methods, and additionally at tower 2 with the ${\mathrm{Br}}^{-}$ (${\mathrm{Br}}^{-}$–MION2–T2; Fig. ) chemical ionisation method. The MARFORCE model is utilised to simulate the evolution of various species at the cross section of the inlet tube, as shown in Fig. . The actual ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ concentrations can be calculated by correlating the count rates, which represent the ratio of the measured ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ signals to primary ions. Subsequently, the predicted ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ concentrations are compared with the measured normalised signals to derive calibration factors (Table ).

Figure 2

MARFORCE simulation results of a ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ calibration experiment. The $x$ axis shows the distance from the UVP pen-ray lamp to the entrance of the chemical ionisation chamber, and the $y$ axis shows the distance in the radial direction. Conditions for the simulation are $R$ $=$ 0.78 cm, $L$ $=$ 26 cm, sample flow $=$ 10.6 L min${}^{-\mathrm{1}}$, $\left[{\mathrm{SO}}_{\mathrm{2}}\right]=\mathrm{5.78}\times {\mathrm{10}}^{\mathrm{13}}$ cm${}^{-\mathrm{3}}$, $\left[{\mathrm{O}}_{\mathrm{2}}\right]=\mathrm{2.42}\times {\mathrm{10}}^{\mathrm{16}}$ cm${}^{-\mathrm{3}}$, and $\left[{\mathrm{H}}_{\mathrm{2}}\mathrm{O}\right]=\mathrm{2.8}\times {\mathrm{10}}^{\mathrm{16}}$ cm${}^{-\mathrm{3}}$.

[Figure omitted. See PDF]

Table 1

Calibration factors and detection limits for selected species measured by the MION2 inlet and Eisele-type inlet. It should be noted that the reported numbers are specific to the experimental conditions and instrument tuning in our experiments. Different instrument tuning can result in different calibration factors and detection limits. Undesired impurities may result in elevated detection limits, despite the calibration factors being the same for the analytes. Therefore, these numbers should not be applied to another study without carrying out the same calibration experiments described in this study.

Species	Calibration factors (MION2)			Detection limit
				MION2 (${\mathrm{Br}}^{-}$)		Eisele inlet (${\mathrm{NO}}_{\mathrm{3}}^{-}$) (160 ms)
	Tower 1 (ionisation time $=$ 35 ms)		Tower 2 (300 ms)	Tower 1	Tower 2
	${\mathrm{NO}}_{\mathrm{3}}^{-}$	${\mathrm{Br}}^{-}$	${\mathrm{Br}}^{-}$	(RH $=$ 3.7 %)	(RH $<$ 0.1 %)	(RH $=$ 5.6 %)
${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$	$\mathrm{1.3}\times {\mathrm{10}}^{\mathrm{10}}$	$\mathrm{8.1}\times {\mathrm{10}}^{\mathrm{9}}$ (RH $=$ 0.2 %–23.3 %)	$\mathrm{9.8}\times {\mathrm{10}}^{\mathrm{8}}$ (RH $=$ 0.3-11.6 %)	$\mathrm{1}\times {{\mathrm{10}}^{\mathrm{5}}}^{\mathrm{a}}$	$\mathrm{2.9}\times {{\mathrm{10}}^{\mathrm{4}}}^{\mathrm{f}}$	$\mathrm{7.6}\times {{\mathrm{10}}^{\mathrm{4}}}^{\mathrm{i}}$
$\mathrm{HOI}$	NA	$\mathrm{1.8}\times {\mathrm{10}}^{\mathrm{10}}$ (RH $=$ 3 %–17 %)	$\mathrm{5.1}\times {\mathrm{10}}^{\mathrm{9}}$ (RH $=$ 3 %–17 %)	$\mathrm{5.9}\times {{\mathrm{10}}^{\mathrm{5}}}^{\mathrm{b}}$	$\mathrm{1.6}\times {{\mathrm{10}}^{\mathrm{5}}}^{\mathrm{g}}$	NA
${\mathrm{HIO}}_{\mathrm{3}}$	NA	NA	NA	$\mathrm{1.3}\times {{\mathrm{10}}^{\mathrm{5}}}^{\mathrm{a}}$	NA	$\mathrm{9}\times {{\mathrm{10}}^{\mathrm{4}}}^{\mathrm{i}}$
${\mathrm{HO}}_{\mathrm{2}}$	NA	$\mathrm{2.8}\times {\mathrm{10}}^{\mathrm{11}}$ (RH $=$ 2.5 %)	$\mathrm{1.2}\times {\mathrm{10}}^{\mathrm{11}}$ (RH $=$ 2.7 %)	$\mathrm{3.3}\times {{\mathrm{10}}^{\mathrm{6}}}^{\mathrm{c}}$ (RH $=$ 2.7 %)	$\mathrm{5.7}\times {{\mathrm{10}}^{\mathrm{5}}}^{\mathrm{h}}$ (RH $=$ 0.3 %)	NA
${\mathrm{SO}}_{\mathrm{2}}$	NA	$\mathrm{2.6}\times {\mathrm{10}}^{\mathrm{16}}$ (RH $=$ 10 %)	$\mathrm{2.1}\times {\mathrm{10}}^{\mathrm{16}}$ (RH $=$ 9.9 %)	$\mathrm{1.8}\times {{\mathrm{10}}^{\mathrm{9}}}^{\mathrm{d}}$ (RH $=$ 0.5 %)	NA	NA
${\mathrm{I}}_{\mathrm{2}}$	NA	$\mathrm{8.2}\times {\mathrm{10}}^{\mathrm{9}}$ (RH $=$ 26 %–37 %)	NA	$\mathrm{3.3}\times {{\mathrm{10}}^{\mathrm{5}}}^{\mathrm{e}}$	NA	NA
$\mathrm{IO}$	NA	NA	NA	$\mathrm{1.6}\times {{\mathrm{10}}^{\mathrm{5}}}^{\mathrm{a}}$	$\mathrm{2.5}\times {{\mathrm{10}}^{\mathrm{4}}}^{\mathrm{f}}$	NA
$\mathrm{OIO}$	NA	NA	NA	$\mathrm{2.0}\times {{\mathrm{10}}^{\mathrm{5}}}^{\mathrm{a}}$	$\mathrm{3.1}\times {{\mathrm{10}}^{\mathrm{4}}}^{\mathrm{f}}$	NA
${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{2}}$	NA	NA	NA	$\mathrm{1.9}\times {{\mathrm{10}}^{\mathrm{5}}}^{\mathrm{a}}$	$\mathrm{3.5}\times {{\mathrm{10}}^{\mathrm{4}}}^{\mathrm{f}}$	NA
${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{3}}$	NA	NA	NA	$\mathrm{1.9}\times {{\mathrm{10}}^{\mathrm{5}}}^{\mathrm{a}}$	$\mathrm{4.2}\times {{\mathrm{10}}^{\mathrm{4}}}^{\mathrm{f}}$	NA
${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{4}}$	NA	NA	NA	$\mathrm{1.9}\times {{\mathrm{10}}^{\mathrm{5}}}^{\mathrm{a}}$	$\mathrm{3.0}\times {{\mathrm{10}}^{\mathrm{4}}}^{\mathrm{f}}$	NA
${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{5}}$	NA	NA	NA	$\mathrm{2}\times {{\mathrm{10}}^{\mathrm{5}}}^{\mathrm{a}}$	$\mathrm{3.7}\times {{\mathrm{10}}^{\mathrm{4}}}^{\mathrm{f}}$	NA

Units are given in molecules per centimeter cubed; NA​​​​​​​ stands for not available. The experiments were conducted at room temperature. The ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ calibration factor is applied to estimate the detection limits of iodine oxides. Since iodine oxides may not be detected at the kinetic limit, their LOD are mere estimations and can be higher than the reported values in this study. The detection limits are estimated with 1 min data and 1 h data collection time. The RH reported in this table is calculated at 25 ${}^{\circ }$C. The calibration factors and LOD have a systematic error of a factor of 2 ($-$50 %/$+$100 %).Calibration factors used for the LOD calculation are as follows: ${}^{\mathrm{a}}$ $\mathrm{8.1}\times {\mathrm{10}}^{\mathrm{9}}$, ${}^{\mathrm{b}}$ $\mathrm{1.8}\times {\mathrm{10}}^{\mathrm{10}}$, ${}^{\mathrm{c}}$ $\mathrm{2.8}\times {\mathrm{10}}^{\mathrm{11}}$, ${}^{\mathrm{d}}$ $\mathrm{1.03}\times {\mathrm{10}}^{\mathrm{14}}$, ${}^{\mathrm{e}}$ $\mathrm{8.2}\times {\mathrm{10}}^{\mathrm{9}}$, ${}^{\mathrm{f}}$ $\mathrm{9.8}\times {\mathrm{10}}^{\mathrm{8}}$, ${}^{\mathrm{g}}$ $\mathrm{5.1}\times {\mathrm{10}}^{\mathrm{9}}$, ${}^{\mathrm{h}}$ $\mathrm{4.1}\times {\mathrm{10}}^{\mathrm{10}}$, and ${}^{\mathrm{i}}$ $\mathrm{3.5}\times {\mathrm{10}}^{\mathrm{9}}$. ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ calibration factor is applied to estimate the detection limits for iodine oxides. Since iodine oxides may not be detected at the kinetic limit, their LOD are mere estimations and can be higher than the reported values in this study. The detection limits are estimated with 1 min data and 1 h data collection time.

The derived calibration factor for ${\mathrm{Br}}^{-}$–MION2–T1 ($\mathrm{8.1}\times {\mathrm{10}}^{\mathrm{9}}$) is approximately 8 times higher than that of ${\mathrm{Br}}^{-}$–MION2–T2 ($\mathrm{9.8}\times {\mathrm{10}}^{\mathrm{8}}$). This observation is consistent with the fact that the ionisation time from tower 2 to the pinhole (around 300 ms) is roughly 8.6 times longer than that of tower 1 (35 ms). A longer ionisation time leads to a greater conversion of ${\mathrm{Br}}^{-}$ and ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}\mathrm{⚫}{\mathrm{Br}}^{-}$ into ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}\mathrm{⚫}{\mathrm{Br}}^{-}$ or ${\mathrm{HSO}}_{\mathrm{4}}^{-}$, resulting in a lower calibration factor.

In the case of ${\mathrm{NO}}_{\mathrm{3}}^{-}$–MION2–T1, it exhibits a similar sensitivity to ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ detection as ${\mathrm{Br}}^{-}$–MION2–T1. This similarity is likely due to the consistent ionisation time (using tower 1) for both methods, since both the ${\mathrm{NO}}_{\mathrm{3}}^{-}$ and ${\mathrm{Br}}^{-}$ chemical ionisation methods measure ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ at the collision limit, as mentioned in previous studies .

reported that the bromide chemical ionisation mass spectrometer (${\mathrm{Br}}^{-}$–CIMS) is capable of detecting ${\mathrm{HO}}_{\mathrm{2}}$ radicals at ambient relevant concentrations. In this study, we calibrated ${\mathrm{HO}}_{\mathrm{2}}$ together with ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$, as ${\mathrm{HO}}_{\mathrm{2}}$ is a byproduct in the chemical production of ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ (see Table ). As the binding of ${\mathrm{HO}}_{\mathrm{2}}$ with ${\mathrm{Br}}^{-}$ is significantly weaker than that of ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ with ${\mathrm{Br}}^{-}$, the collision-induced fragmentation of ${\mathrm{HO}}_{\mathrm{2}}\mathrm{⚫}{\mathrm{Br}}^{-}$ in the ion optics of the mass spectrometer is larger . Additionally, as the humidity effect of ${\mathrm{HO}}_{\mathrm{2}}$ will be shown to be strong in Sect. 3.3, the calibration factor of ${\mathrm{HO}}_{\mathrm{2}}$ has to be derived with respect to a specific humidity level. The derived ${\mathrm{HO}}_{\mathrm{2}}$ calibration factors at 2.5 %–2.7 % RH (25 ${}^{\circ }$C) are $\mathrm{2.8}\times {\mathrm{10}}^{\mathrm{11}}$ and $\mathrm{1.2}\times {\mathrm{10}}^{\mathrm{11}}$, respectively, for ${\mathrm{Br}}^{-}$–MION2–T1 and ${\mathrm{Br}}^{-}$–MION2–T2 (Table ).

The $\mathrm{HOI}$ calibration was also carried out using the ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ calibration source, except that the ${\mathrm{SO}}_{\mathrm{2}}$ source was replaced with an ${\mathrm{I}}_{\mathrm{2}}$ source. As can be seen in Fig.  and Table , the calibration factor for $\mathrm{HOI}$ is roughly 2 times that of ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$. This suggests that $\mathrm{HOI}$ is detected close to the collision limit. It is worth noting that we find the instrument setting affects $\mathrm{HOI}$ detection significantly, since $\mathrm{HOI}$ is not strongly bonded to ${\mathrm{Br}}^{-}$. The preferred fragmentation pathway is $\mathrm{HOI}\mathrm{⚫}{\mathrm{Br}}^{-}$ $\to$ $\mathrm{HOI}$ $+$ ${\mathrm{Br}}^{-}$ (Table ), and thus a fraction of $\mathrm{HOI}\mathrm{⚫}{\mathrm{Br}}^{-}$ dissociates into $\mathrm{HOI}$ and ${\mathrm{Br}}^{-}$ after passing the ion optics of the mass spectrometer. A more fragmentation-oriented setting can result in a higher fraction of $\mathrm{HOI}\mathrm{⚫}{\mathrm{Br}}^{-}$ becoming lost in the ion optics, thus resulting in a higher calibration factor, i.e. lower sensitivity. As an example, in our earlier studies , we used a relatively higher fragmentation settings compared to the one used in this study in an attempt to reduce $({\mathrm{H}}_{\mathrm{2}}\mathrm{O}{)}_{\mathrm{n}}\mathrm{⚫}{\mathrm{Br}}^{-}$ clusters and other water-associated clusters. This experimental set-up led to a calibration factor for $\mathrm{HOI}$ that was 8 times higher than the calibration factor for ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$.

Figure 3

The modelled or calculated vapour concentrations vs. the normalised signals for (a) ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$, (b) HOI, (c) ${\mathrm{SO}}_{\mathrm{2}}$, and (d) ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$. The dashed–dotted, solid, and dashed lines are the linear fits of the results from different inlet modes, namely (1) tower 1 with the ${\mathrm{NO}}_{\mathrm{3}}^{-}$ chemical ionisation method, (2) tower 1 with the ${\mathrm{Br}}^{-}$ chemical ionisation method, and tower 2 with the ${\mathrm{Br}}^{-}$ chemical ionisation method. The slopes of the fitted lines represent the calibration factors, as shown in the legend. The colour bar shows the relative humidity in the calibration experiments.

[Figure omitted. See PDF]

Table 2

Fragmentation enthalpies (the opposite of binding enthalpies) of analytes with the ${\mathrm{Br}}^{-}$. The cluster geometry was optimised at the $\mathit{\omega }$B97X-D/aug-cc-pVTZ-PP level of theory at 298.15 K . The enthalpies were calculated at the DLPNO-CCSD(T)/def2-QZVPP at 298.15 K.

Cluster fragmentation pathway	Fragmentation
	enthalpies
	(kcal mol${}^{-\mathrm{1}}$)
${\mathrm{I}}_{\mathrm{2}}\mathrm{⚫}{\mathrm{Br}}^{-}$ $\to$ ${\mathrm{I}}_{\mathrm{2}}$ $+$ ${\mathrm{Br}}^{-}$	33.3
${\mathrm{I}}_{\mathrm{2}}\mathrm{⚫}{\mathrm{H}}_{\mathrm{2}}{\mathrm{OBr}}^{-}$ $\to$ ${\mathrm{I}}_{\mathrm{2}}\mathrm{⚫}{\mathrm{Br}}^{-}$ $+$ ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$	8.0
$\mathrm{IO}\mathrm{⚫}{\mathrm{Br}}^{-}$ $\to$ $\mathrm{IO}$ $+$ ${\mathrm{Br}}^{-}$	24.5
$\mathrm{IO}\mathrm{⚫}{\mathrm{H}}_{\mathrm{2}}{\mathrm{OBr}}^{-}$ $\to$ $\mathrm{IO}$ $+$ ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}\mathrm{⚫}{\mathrm{Br}}^{-}$	21.3
$\mathrm{IO}\mathrm{⚫}{\mathrm{H}}_{\mathrm{2}}{\mathrm{OBr}}^{-}$ $\to$ $\mathrm{IO}\mathrm{⚫}{\mathrm{Br}}^{-}$ $+$ ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$	9.9
$\mathrm{OIO}\mathrm{⚫}{\mathrm{Br}}^{-}$ $\to$ $\mathrm{OIO}$ $+$ ${\mathrm{Br}}^{-}$	23.2
$\mathrm{OIO}\mathrm{⚫}{\mathrm{H}}_{\mathrm{2}}{\mathrm{OBr}}^{-}$ $\to$ $\mathrm{OIO}$ $+$ ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}\mathrm{⚫}{\mathrm{Br}}^{-}$	22.1
$\mathrm{OIO}\mathrm{⚫}{\mathrm{H}}_{\mathrm{2}}{\mathrm{OBr}}^{-}$ $\to$ $\mathrm{OIO}\mathrm{⚫}{\mathrm{Br}}^{-}$ $+$ ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$	11.9
${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{3}}\mathrm{⚫}{\mathrm{Br}}^{-}$ $\to$ ${\mathrm{IO}}_{\mathrm{3}}^{-}$ $+$ $\mathrm{IBr}$	24.6
${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{4}}\mathrm{⚫}{\mathrm{Br}}^{-}$ $\to$ ${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{4}}$ $+$ ${\mathrm{Br}}^{-}$	42.6
${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{4}}\mathrm{⚫}{\mathrm{H}}_{\mathrm{2}}{\mathrm{OBr}}^{-}$ $\to$ ${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{4}}$ $+$ ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}\mathrm{⚫}{\mathrm{Br}}^{-}$	48.8
${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{4}}\mathrm{⚫}{\mathrm{H}}_{\mathrm{2}}{\mathrm{OBr}}^{-}$ $\to$ ${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{4}}\mathrm{⚫}{\mathrm{Br}}^{-}$ $+$ ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$	10.5
${\mathrm{HIO}}_{\mathrm{3}}\mathrm{⚫}{\mathrm{Br}}^{-}$ $\to$ ${\mathrm{IO}}_{\mathrm{3}}^{-}$ $+$ $\mathrm{HBr}$	29.9${}^{\mathrm{a}}$
${\mathrm{HIO}}_{\mathrm{3}}\mathrm{⚫}{\mathrm{Br}}^{-}$ $\to$ ${\mathrm{HIO}}_{\mathrm{3}}$ $+$ ${\mathrm{Br}}^{-}$	35.7${}^{\mathrm{a}}$
${\mathrm{HIO}}_{\mathrm{3}}\mathrm{⚫}{\mathrm{H}}_{\mathrm{2}}{\mathrm{OBr}}^{-}$ $\to$ ${\mathrm{HIO}}_{\mathrm{3}}$ $+$ ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}\mathrm{⚫}{\mathrm{Br}}^{-}$	33.1
${\mathrm{HIO}}_{\mathrm{3}}\mathrm{⚫}{\mathrm{H}}_{\mathrm{2}}{\mathrm{OBr}}^{-}$ $\to$ ${\mathrm{HIO}}_{\mathrm{3}}\mathrm{⚫}{\mathrm{Br}}^{-}$ $+$ ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$	11.2
${\mathrm{HIO}}_{\mathrm{3}}\mathrm{⚫}{\mathrm{H}}_{\mathrm{2}}{\mathrm{OBr}}^{-}$ $\to$ ${\mathrm{IO}}_{\mathrm{3}}\mathrm{⚫}{\mathrm{H}}_{\mathrm{2}}{\mathrm{O}}^{-}$ $+$ $\mathrm{HBr}$	26.7
${\mathrm{HIO}}_{\mathrm{2}}\mathrm{⚫}{\mathrm{Br}}^{-}$ $\to$ ${\mathrm{HIO}}_{\mathrm{2}}$ $+$ ${\mathrm{Br}}^{-}$	29.2${}^{\mathrm{b}}$
${\mathrm{HIO}}_{\mathrm{2}}\mathrm{⚫}{\mathrm{H}}_{\mathrm{2}}{\mathrm{OBr}}^{-}$ $\to$ ${\mathrm{HIO}}_{\mathrm{2}}$ $+$ ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}\mathrm{⚫}{\mathrm{Br}}^{-}$	15.5
${\mathrm{HIO}}_{\mathrm{2}}\mathrm{⚫}{\mathrm{H}}_{\mathrm{2}}{\mathrm{OBr}}^{-}$ $\to$ ${\mathrm{HIO}}_{\mathrm{2}}\mathrm{⚫}{\mathrm{Br}}^{-}$ $+$ ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$	1.3
${\mathrm{HIO}}_{\mathrm{2}}\mathrm{⚫}{\mathrm{H}}_{\mathrm{2}}{\mathrm{OBr}}^{-}$ $\to$ ${\mathrm{IO}}_{\mathrm{2}}\mathrm{⚫}{\mathrm{H}}_{\mathrm{2}}{\mathrm{O}}^{-}$ $+$ $\mathrm{HBr}$	27.4
$\mathrm{HOI}\mathrm{⚫}{\mathrm{Br}}^{-}$ $\to$ $\mathrm{HOI}$ $+$ ${\mathrm{Br}}^{-}$	26.9${}^{\mathrm{b}}$
$\mathrm{HOI}\mathrm{⚫}{\mathrm{H}}_{\mathrm{2}}{\mathrm{OBr}}^{-}$ $\to$ $\mathrm{HOI}$ $+$ ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}\mathrm{⚫}{\mathrm{Br}}^{-}$	22.9
$\mathrm{HOI}\mathrm{⚫}{\mathrm{H}}_{\mathrm{2}}{\mathrm{OBr}}^{-}$ $\to$ $\mathrm{HOI}\mathrm{⚫}{\mathrm{Br}}^{-}$ $+$ ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$	9.6
$\mathrm{HOI}\mathrm{⚫}{\mathrm{H}}_{\mathrm{2}}{\mathrm{OBr}}^{-}$ $\to$ $\mathrm{IO}\mathrm{⚫}{\mathrm{H}}_{\mathrm{2}}{\mathrm{O}}^{-}$ $+$ $\mathrm{HBr}$	48.4
${\mathrm{H}}_{\mathrm{2}}\mathrm{O}\mathrm{⚫}{\mathrm{Br}}^{-}$ $\to$ ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$ $+$ ${\mathrm{Br}}^{-}$	13.2
${\mathrm{HO}}_{\mathrm{2}}\mathrm{⚫}{\mathrm{Br}}^{-}$ $\to$ ${\mathrm{HO}}_{\mathrm{2}}$ $+$ ${\mathrm{Br}}^{-}$	23.1
${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}\mathrm{⚫}{\mathrm{Br}}^{-}$ $\to$ ${\mathrm{HSO}}_{\mathrm{4}}^{-}$ $+$ $\mathrm{HBr}$	27.9${}^{\mathrm{b}}$
${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}\mathrm{⚫}{\mathrm{H}}_{\mathrm{2}}{\mathrm{OBr}}^{-}$ $\to$ ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ $+$ ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}\mathrm{⚫}{\mathrm{Br}}^{-}$	36.1
${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}\mathrm{⚫}{\mathrm{H}}_{\mathrm{2}}{\mathrm{OBr}}^{-}$ $\to$ ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}\mathrm{⚫}{\mathrm{Br}}^{-}$ $+$ ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$	8.2
${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}\mathrm{⚫}{\mathrm{H}}_{\mathrm{2}}{\mathrm{OBr}}^{-}$ $\to$ ${\mathrm{HSO}}_{\mathrm{4}}\mathrm{⚫}{\mathrm{H}}_{\mathrm{2}}{\mathrm{O}}^{-}$ $+$ $\mathrm{HBr}$	22.0
${\mathrm{SO}}_{\mathrm{2}}\mathrm{⚫}{\mathrm{Br}}^{-}$ $\to$ ${\mathrm{SO}}_{\mathrm{2}}$ $+$ ${\mathrm{Br}}^{-}$	19.4

${}^{\mathrm{a}}$ The fragmentation enthalpy is updated from as a lower-energy ${\mathrm{HIO}}_{\mathrm{3}}\mathrm{⚫}{\mathrm{Br}}^{-}$ cluster geometry, which has an additional $\mathrm{Br}$–$\mathrm{I}$ interaction, has been located in this study​​​​​​​. ${}^{\mathrm{b}}$ Value adopted from .

3.2

Calibration of ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$ and ${\mathrm{SO}}_{\mathrm{2}}$

${\mathrm{H}}_{\mathrm{2}}\mathrm{O}\mathrm{⚫}{\mathrm{Br}}^{-}$ is a regular peak and one of the primary ions measured by the ${\mathrm{Br}}^{-}$–MION2. ${\mathrm{Br}}^{-}$–MION2 is, therefore, able to measure absolute water content if the ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}\mathrm{⚫}{\mathrm{Br}}^{-}$ signal is calibrated against a dew point mirror instrument. Such a calibration has at least two purposes. (1) The calibrated ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}\mathrm{⚫}{\mathrm{Br}}^{-}$ : ${\mathrm{Br}}^{-}$ can be used as an indicator of the fragmentation level of the ${\mathrm{Br}}^{-}$–MION2, and (2) compared to regular relative humidity sensors and dew point mirrors, ${\mathrm{Br}}^{-}$–MION2 exhibits higher sensitivity towards ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$. In this study, the calibration of ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$ was performed for both ${\mathrm{Br}}^{-}$–MION–T1 and ${\mathrm{Br}}^{-}$–MION–T2, as illustrated in Fig. . Interestingly, the calibration factors for both towers did not show significant differences. This can be attributed to the presence of an excess amount of ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$, which establishes a rapid equilibrium with ${\mathrm{Br}}^{-}$ and ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}\mathrm{⚫}{\mathrm{Br}}^{-}$, irrespective of the ionisation time.

As a reasonable binding enthalpy of ${\mathrm{SO}}_{\mathrm{2}}\mathrm{⚫}{\mathrm{Br}}^{-}$ was predicted using quantum chemical calculations (Table ), we continued to check whether the ${\mathrm{Br}}^{-}$–MION2 allows us to detect ${\mathrm{SO}}_{\mathrm{2}}$. A variable amount of ${\mathrm{SO}}_{\mathrm{2}}$ was mixed with a fixed amount of dilution flow at a constant relative humidity (RH $=$ 10 %) which was measured by the ${\mathrm{Br}}^{-}$–MION2. Clear ${\mathrm{SO}}_{\mathrm{2}}\mathrm{⚫}{\mathrm{Br}}^{-}$ was measured, and it increased linearly with the ${\mathrm{SO}}_{\mathrm{2}}$ concentration in the sample flow (Fig. ). However, the calibration factor of ${\mathrm{SO}}_{\mathrm{2}}$ is roughly 6 orders of magnitude higher than that of ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ at 10 % RH. This is consistent with the weaker binding of ${\mathrm{SO}}_{\mathrm{2}}\mathrm{⚫}{\mathrm{Br}}^{-}$ compared with ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}\mathrm{⚫}{\mathrm{Br}}^{-}$. Additionally, ${\mathrm{SO}}_{\mathrm{2}}$ calibration is extremely sensitive to RH changes, as can be seen in Fig. . In this study, the best achieved detection limit was $\mathrm{9.4}\times {\mathrm{10}}^{\mathrm{7}}$ cm${}^{-\mathrm{3}}$ at an RH below 0.1 %. Theoretically, it is possible to enhance the sensitivity even further by reducing the absolute water content.

Figure 4

The effect of humidity on the detection efficiency of (a) $\mathrm{HOI}$, (b) ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$, (c) ${\mathrm{HO}}_{\mathrm{2}}$, (d) ${\mathrm{SO}}_{\mathrm{2}}$, and (e) ${\mathrm{I}}_{\mathrm{2}}$. The measured signals in each set of experiments are normalised by the signal at the lowest RH. Therefore, the normalised signals represent how the increasing RH is affecting the detection limit compared with the initial point. The purple circles and orange triangles show the detection humidity effects of tower 1 and tower 2, respectively. The blue squares refer to the experiments conducted with a dry flow added before the MION2 inlet. The RH is converted from absolute ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$ concentrations at 25 ${}^{\circ }$C. Error bars represent 1 standard deviation.

[Figure omitted. See PDF]

3.3 Detection humidity effect

The measurement sensitivity of the halide anion-based chemical ionisation method was regularly reported to be affected by atmospheric water content . The humidity effect of atmospheric pressure ${\mathrm{Br}}^{-}$–MION2 could be amplified due to the higher water content present in air samples. Although demonstrated that the detection of ${\mathrm{I}}_{\mathrm{2}}$ by ${\mathrm{Br}}^{-}$–MION1 was not affected within a limited humidity variation (40 %–80 % RH at $-$10 ${}^{\circ }$C), characterisation under a wider range of humidity conditions is needed. As the detection humidity effect in this study exclusively refers to the effect of absolute humidity instead of relative humidity, absolute humidity parameters such as the dew or frost point or ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$ molecule concentration are commonly presented together with the relative humidity (at 25 ${}^{\circ }$C, unless indicated otherwise).

In this study, we examine the detection humidity effect of ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$, $\mathrm{HOI}$, ${\mathrm{HO}}_{\mathrm{2}}$, ${\mathrm{SO}}_{\mathrm{2}}$, and ${\mathrm{I}}_{\mathrm{2}}$ with RH from below 1 % to 60 % at 25 ${}^{\circ }$C. Unlike ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$, ${\mathrm{HO}}_{\mathrm{2}}$, and $\mathrm{HOI}$, which require generation from a calibration source, both ${\mathrm{SO}}_{\mathrm{2}}$ and ${\mathrm{I}}_{\mathrm{2}}$ have their own standardised sources. This simplifies their control during the characterisation of the detection humidity effect. Therefore, a straight flow reactor is used to premix the analyte containing air sample to the ${\mathrm{Br}}^{-}$–MION2 (Fig. ). It is worth noting that we do not account for the wall loss of ${\mathrm{SO}}_{\mathrm{2}}$ and ${\mathrm{I}}_{\mathrm{2}}$ in the analysis. The wall loss of ${\mathrm{SO}}_{\mathrm{2}}$ is negligible at the timescale of the calibration processes (a few seconds). Despite ${\mathrm{I}}_{\mathrm{2}}$ vapour being able to both condense on and evaporate from the walls of the flow reactor, equilibrium can be achieved given a sufficient amount of time. In our experiments, it took up to 24 h to reach equilibrium. Once equilibrium is established, the condensation and evaporation of ${\mathrm{I}}_{\mathrm{2}}$ balance each other out, making the estimation of ${\mathrm{I}}_{\mathrm{2}}$ concentration straightforward.

On the other hand, the characterisation of the detection humidity effect of ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$, $\mathrm{HOI}$, and ${\mathrm{HO}}_{\mathrm{2}}$ is more challenging, as the production of these species is nearly proportional to the amount of ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$ passing the calibration source. Therefore, an experimental apparatus was built, which enabled humidifying the air sample after the calibration source without disturbing ${\mathrm{HO}}_x$ production processes in the calibration source (Fig. ).

The results of the humidity characterisation are shown in Fig. . Although only five species were characterised and observed for their distinct humidity sensitivity, a general conclusion can be drawn that applies to essentially all of the species; i.e. an excessive amount of water content leads to a decrease in detection sensitivity. The species with stronger binding with ${\mathrm{Br}}^{-}$ exhibit less sensitivity to changes in humidity (e.g. ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ and ${\mathrm{I}}_{\mathrm{2}}$), while the weakly bonded ones ($\mathrm{HOI}$, ${\mathrm{SO}}_{\mathrm{2}}$, and ${\mathrm{HO}}_{\mathrm{2}}$) are strongly affected. The humidity tolerance of the measured species can be ordered as ${\mathrm{I}}_{\mathrm{2}}$ $>$ $\mathrm{HOI}$ $>$ ${\mathrm{HO}}_{\mathrm{2}}$ $>$ ${\mathrm{SO}}_{\mathrm{2}}$, which is the same order as the strength of their bindings with ${\mathrm{Br}}^{-}$ (Table ).

Interestingly, the detection humidity effect of ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ is observed to be non-linear; i.e. the detection sensitivity of ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ first increases with higher RH but eventually has a sharp drop at around 33 % RH. The enhancement of ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ detection at below ca. 33 % RH could be contributed by two mechanisms. First, the diffusivity of ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ is lower at higher RH . A higher RH, therefore, reduces the wall deposition of ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ in the inlet tube, thus effectively increasing the detected ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$. This is a universal factor that influences all ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ detection techniques with an appreciable sampling line residence time. The second possibility is that, at low RH regime, ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$ enhances ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ detection by offering more modes through which the excess energy of the cluster can dissipate in the formation of ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}\mathrm{⚫}{\mathrm{Br}}^{-}$, thus resulting in a relatively more stable cluster . Regardless of the sources of the detection humidity effect at the low water content regime, the maximum systematic error is measured to be 37 % by comparing the experiment carried out at 2 % RH (frost point of $-$25 ${}^{\circ }$C) and the experiment carried out at 33 % (dew point of 7.6 ${}^{\circ }$C) in Fig. b. Based on our findings, we anticipate that the detection humidity effect of ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ would be moderate when the dew point is below approximately 7.6 ${}^{\circ }$C. However, it is important to exercise caution when conducting measurements under higher absolute humidity conditions.

Additionally, a longer ionisation time by utilising the ${\mathrm{Br}}^{-}$–MION2–T2 results in a stronger detection humidity effect, as shown in Fig. . This phenomenon is the most significant for $\mathrm{HOI}$; i.e. the detection of $\mathrm{HOI}$ is more humidity dependent when using ${\mathrm{Br}}^{-}$–MION2–T2 instead of ${\mathrm{Br}}^{-}$–MION2–T1. This phenomenon also elucidates the curvature observed in the $\mathrm{HOI}$ calibration when employing the MION2–T2 (Fig. ); the diminished detection sensitivity of $\mathrm{HOI}$ counterbalances the augmented $\mathrm{HOI}$ production at elevated water content. Although this effect is difficult to quantify, it practically suggests that the ${\mathrm{Br}}^{-}$ chemical ionisation method should employ a shorter ionisation time (i.e. using the T1) when operating MION2 with multiple chemical ionisation methods.

In summary, we find that the detection of ${\mathrm{Br}}^{-}$–MION2 is strongly affected by air water content. The atmospheric pressure ${\mathrm{Br}}^{-}$ chemical ionisation method is suitable for laboratory experiments where water content is controlled and atmospheric observations in the cryosphere where air water content is low. Nevertheless, the humidity effect should be considered individually for different analytes, and the binding enthalpy between the analyte and ${\mathrm{Br}}^{-}$ is likely a good indicator. As the ${\mathrm{NO}}_{\mathrm{3}}^{-}$–MION2 (or the ${\mathrm{NO}}_{\mathrm{3}}^{-}$ chemical ionisation in general) is known to have minimal detection humidity sensitivity, it is commonly operated together with the ${\mathrm{Br}}^{-}$–MION2. Performing a cross-check of mutually measured species, such as ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$, ${\mathrm{HIO}}_{\mathrm{3}}$, and oxidised organic species, will provide crucial insights into whether and when the detection capability of ${\mathrm{Br}}^{-}$–MION2 is compromised by the water content in the air. In this context, the new design of ${\mathrm{Br}}^{-}$–MION2, which enables three chemical ionisation methods to have the same ionisation time, is essential.

3.4 Attempts to reduce the detection humidity effect

Various approaches were explored to mitigate the detection humidity effect. One commonly used method is to employ a low-pressure chemical ionisation system, which has been successfully implemented in iodide chemical ionisation systems and bromide chemical ionisation systems . However, reducing the relative humidity (RH) of the air sample comes at the expense of reducing the measurement sensitivity for species detected at the collision limit, such as ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$, ${\mathrm{HIO}}_{\mathrm{3}}$, and ${\mathrm{I}}_{\mathrm{2}}$, as the air sample unavoidably undergoes dilution in this process. We estimated previously that the ${\mathrm{Br}}^{-}$–FIGAERO (Filter Inlet for Gases and AEROsols) inlet had a more than 10 times higher detection limit compared to the ${\mathrm{Br}}^{-}$–MION1 inlet . For example, the ${\mathrm{Br}}^{-}$–FIGAERO had a ${\mathrm{HIO}}_{\mathrm{3}}$ detection limit of $\mathrm{5.1}\times {\mathrm{10}}^{\mathrm{6}}$ cm${}^{-\mathrm{3}}$, which struggles to detect atmospheric levels of ${\mathrm{HIO}}_{\mathrm{3}}$ (commonly below ${\mathrm{10}}^{\mathrm{7}}$ cm${}^{-\mathrm{3}}$) . The lower level of the detection limit provided by the ${\mathrm{Br}}^{-}$–MION2 inlet is therefore essential in the detection of iodine species. Another important factor is the reaction of halogen radicals with analytes. Besides halogen anions, halogen radicals can also be produced by chemical ionisation processes. While iodine radical (I$\mathrm{⚫}$) mostly reacts with halogen species and a minimal number of organic species, the bromide radical (Br$\mathrm{⚫}$) reacts with a wider range of organic species, as it has a larger reactivity. Conventional low-pressure systems that involve mixing analytes with reagent gases, such as the FIGAERO inlet, can introduce additional complexities when interpreting mass spectra. As a result, alternative approaches were pursued to effectively reduce the detection humidity effect.

The first method is the dilution method. Instead of measuring the air sample directly, a dry dilution flow was mixed with the air sample at the entrance of the ${\mathrm{Br}}^{-}$–MION2 inlet (see Fig. ). We tested this method for the ${\mathrm{SO}}_{\mathrm{2}}$ detection with an air sample flow of 1.8 L min${}^{-\mathrm{1}}$ and a dilution flow of 20.7 L min${}^{-\mathrm{1}}$ (Fig. ). The $x$ axis for this set of experiments represents humidity in the air sample instead of the humidity after the dilution to compare with the experiments without adding the dilution flow. We observe a significantly reduced detection humidity effect compared to the case without dilution. It is noteworthy that as the air sample was diluted by a factor of 12.5, the detection limit of the instrument is likely enhanced by the same factor. However, since the detection humidity effect for ${\mathrm{SO}}_{\mathrm{2}}$ is significantly higher than other species (e.g. ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$, $\mathrm{HOI}$, and ${\mathrm{I}}_{\mathrm{2}}$), the dilution is still effective for ${\mathrm{SO}}_{\mathrm{2}}$ measurement. For example, no ${\mathrm{SO}}_{\mathrm{2}}\mathrm{⚫}{\mathrm{Br}}^{-}$ signal would not be measured at 40 % RH (25 ${}^{\circ }$C) if the air sample is not diluted but a noticeable signal would be measured if the air sample is diluted. A similar conclusion is likely applicable to other species but with a different optimal humidity cutoff.

The second method is additionally introducing a core-sampling device that uses the air sample as the core flow and a dry synthetic airflow as the sheath flow (Fig. ). This takes advantage of the fact that ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$ diffuses into the sheath flow faster than other analytes with larger molecular weight, thus effectively reducing the RH in the core flow from which the instrument pinhole collects the most sample. Nevertheless, it is important to note that the core-sampling method, while helping to mitigate the detection humidity effect, also leads to a reduction in the ${\mathrm{SO}}_{\mathrm{2}}\mathrm{⚫}{\mathrm{Br}}^{-}$ signal. This is because the ${\mathrm{SO}}_{\mathrm{2}}$ itself is diluted, partially counteracting the benefits of the reduced detection humidity effect.

Various sample-to-sheath flow combinations were tested, as presented in Fig. . The measured ${\mathrm{SO}}_{\mathrm{2}}\mathrm{⚫}{\mathrm{Br}}^{-}$ signal from all sets of experiments was normalised by the experiment with the sample-to-sheath ratio of 21 : 1 at 0.21 % RH (25 ${}^{\circ }$C). The results indicate that reducing the sample-to-sheath ratio effectively alleviates the ${\mathrm{SO}}_{\mathrm{2}}$ detection humidity effect. It is observed that different mixing ratios have only a moderate impact on the measured ${\mathrm{SO}}_{\mathrm{2}}\mathrm{⚫}{\mathrm{Br}}^{-}$ when the ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$ concentration is below ${\mathrm{10}}^{\mathrm{16}}$ cm${}^{-\mathrm{3}}$ (1 % RH), indicating a low-detection humidity effect in such conditions. However, the core-sampling device clearly enhances the ${\mathrm{SO}}_{\mathrm{2}}$ detection efficiency when the ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$ concentration is larger than ${\mathrm{10}}^{\mathrm{16}}$ cm${}^{-\mathrm{3}}$. The sample-to-sheath ratio of 1 : 21 enables the effective detection of ${\mathrm{SO}}_{\mathrm{2}}$ at around $\mathrm{4.5}\times {\mathrm{10}}^{\mathrm{17}}$ cm${}^{-\mathrm{3}}$ (60 % RH) of ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$, while the sample-to-sheath ratio of 21 : 1 is not able to detect ${\mathrm{SO}}_{\mathrm{2}}$ after around $\mathrm{4.3}\times {\mathrm{10}}^{\mathrm{16}}$ cm${}^{-\mathrm{3}}$ of ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$ (6 % RH). Overall, the sample-to-sheath ratio of 1 : 21 is at least 2 orders of magnitude more effective at detecting ${\mathrm{SO}}_{\mathrm{2}}$ when ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$ is greater than $\mathrm{2}\times {\mathrm{10}}^{\mathrm{16}}$ cm${}^{-\mathrm{3}}$. Therefore, the core-sampling method is an effective method for reducing the detection humidity effect of species which are weakly bonded with ${\mathrm{Br}}^{-}$. Despite the reduced detection humidity effect, it is important to note that the sample water content still impacts the detection limit of ${\mathrm{SO}}_{\mathrm{2}}$. Therefore, dedicated experiments need to be conducted to accurately determine the concentration of ${\mathrm{SO}}_{\mathrm{2}}$.

Figure 5

Reducing the detection humidity effect with the core-sampling method (Fig. ). This design takes advantage of the faster diffusion of ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$ than ${\mathrm{SO}}_{\mathrm{2}}$ from the sample flow to the sheath flow and effectively reduces the RH in the sample flow. Various sample-to-sheath ratios were tested at different ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$ concentrations to find the optimal setting. All the data are normalised to the lowest RH data point in the sample-to-sheath $=$ 21 : 1 experiment. Due to experimental constraints, the sample-to-sheath ratios are (a) 3.5 : 18.5, (b) 2 : 20, and (c) 1 : 21, and experiments started only from the second-, third-, and fourth-lowest RH points, respectively. All other experiments collected data in all humidity conditions. The error bars represent the standard deviation of the normalised ${\mathrm{SO}}_{\mathrm{2}}$ signals.

[Figure omitted. See PDF]

3.5 Limit of detection

The limit of detection (LOD) is an essential parameter for a chemical ionisation inlet system. For the convenience of intercomparison, we define the LOD in this study as

3$$\mathrm{LOD}=\mathit{\mu }+\mathrm{3}\times \mathit{\sigma },$$ where $\mathit{\mu }$ is the mean value of 1 h mass spectrometric data with a 1 min time resolution, and $\mathit{\sigma }$ is the standard variation in the same data. Both $\mathit{\mu }$ and $\mathit{\sigma }$ include the experimentally derived calibration factor. The species without direct calibration utilise the calibration factor of ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$; thus the LOD for these species generally represent the lower limit. The LOD is determined by introducing pure nitrogen or synthetic air into the chemical ionisation inlet, where none of the species listed in Table  is expected to be present. It is important to emphasise that this LOD definition is specifically suitable for distinguishing trace gas concentrations from background levels in long-term observations.

The reported LOD can be affected by many factors. Some of these factors are (1) the purity of the reagent source (e.g. ${\mathrm{HNO}}_{\mathrm{3}}$ or ${\mathrm{CH}}_{\mathrm{2}}{\mathrm{Br}}_{\mathrm{2}}$ solution); (2) the purity of the sample air used at the LOD determination experiment; (3) the signal-to-noise (electronic background noise) ratio of the instrument; (4) the fragmentation level (controlled by the tuning of the instrument) of the mass spectrometer; (5) the humidity of the sample air used at the LOD determination experiment (for ${\mathrm{Br}}^{-}$ chemical ionisation method); and (6) different ways of estimating LOD.

Therefore, comparing the LOD derived in this study with earlier studies may not be meaningful. Hence, we additionally compared the ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ LOD of the MION2 inlet with the widely used Eisele-type inlet, which are both attached to the same mass spectrometer (Table ). The direct comparison suggests that the ${\mathrm{Br}}^{-}$–MION2–T1 LOD is roughly 30 % higher than the LOD of the Eisele inlet and is thus a comparable performance. When we increased the ionisation time from 35 ms (${\mathrm{Br}}^{-}$–MION2–T1) to 300 ms (${\mathrm{Br}}^{-}$–MION2–T2), the LOD of ${\mathrm{Br}}^{-}$–MION2 for ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ is further reduced by a factor of 3; thus, ${\mathrm{Br}}^{-}$–MION2–T2 performs better than the Eisele inlet. This suggests that the MION2 inlet can achieve a comparable (${\mathrm{Br}}^{-}$–MION2–T1) or even better (${\mathrm{Br}}^{-}$–MION2–T2) LOD than the Eisele inlet. Additionally, the Eisele-type inlet was regularly shown to have a LOD as low as ${\mathrm{10}}^{\mathrm{4}}$ cm${}^{-\mathrm{3}}$ , so a well-performing mass spectrometer may further reduce the LOD of MION2. Nevertheless, the attained LOD are sufficiently low for atmospheric measurements. The molecules in question typically require concentrations above ${\mathrm{10}}^{\mathrm{6}}$ cm${}^{-\mathrm{3}}$ to exert a significant influence on atmospheric chemistry and aerosol formation.

3.6 Voltage scanning and cluster formation enthalpy

Collision-induced cluster fragmentation is an unavoidable issue which affects the detection of analytes that are weakly bonded with the reagent ion. Since if a charged cluster is loosely bonded, then collisions between the charged clusters and air molecules in the atmospheric pressure interface may break a large portion of the charged clusters apart prior to reaching the detector . Therefore, the charged cluster binding strength is an important factor determining whether an analyte–reagent ion cluster can be measured by the mass spectrometer . showed that the level of collision-induced cluster fragmentation is associated with the voltage differences between the first and second quadrupoles in the atmospheric pressure interface of the mass spectrometer. The voltage difference was shown to be indicative of the fragmentation level of the CIMS, and it positively correlates with the cluster formation enthalpy .

In this study, we carried out voltage scan experiments with the same procedures as described in . Briefly, we kept the voltage differences inside two individual quadruples constant, while changing the voltage difference between these two quadruples to modulate energies in the collision processes, and the results are shown in Fig. . Generally, a higher voltage difference indicates a higher fragmentation level, which in turn results in a lower remaining fraction of charged clusters. Charged clusters that are less sensitive to voltage changes, especially in the low-voltage difference regime are more stable.

Figure 6

Normalised signal remaining vs. the scanning voltage ($\mathrm{\Delta }V$). The normalised signal remaining of each species is normalised by the maximum and minimum values of its values with different $\mathrm{\Delta }V$ (or $\mathrm{d}V$). The $\mathrm{\Delta }V$ describes the voltage difference between the skimmer, and the second quadruple and can be considered an indicator of the softness of the instrument tuning . A higher $\mathrm{\Delta }V$ commonly indicates a more fragmenting setting.

[Figure omitted. See PDF]

A series of iodine oxides and oxoacids is evaluated together with other inorganic species, such as ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$, ${\mathrm{HO}}_{\mathrm{2}}$, ${\mathrm{SO}}_{\mathrm{2}}$, and ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ (Fig. ). Based on the results, we categorise the analytes into the following three categories: (1) analytes which are strongly bonded with ${\mathrm{Br}}^{-}$, (2) analytes which are moderately bonded with ${\mathrm{Br}}^{-}$, and (3) analytes which are weakly bonded with ${\mathrm{Br}}^{-}$. The species ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$, ${\mathrm{HIO}}_{\mathrm{3}}$, ${\mathrm{HIO}}_{\mathrm{2}}$, and ${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{4}}$ can be classified into the first category, since the initial change in voltage difference does not have a significant impact on the normalised signal. This indicates that these species are detected at the collision limit. It is also apparent that ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$, ${\mathrm{HO}}_{\mathrm{2}}$, and ${\mathrm{SO}}_{\mathrm{2}}$ belong to the third category, since a small increase in the voltage difference leads to substantially reduced normalised ratios. Finally, $\mathrm{IO}$, $\mathrm{OIO}$, ${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{3}}$, and $\mathrm{HOI}$ are moderately bonded with ${\mathrm{Br}}^{-}$. These moderately bonded charged clusters can reach a close to collision limit detection if the instrument is softly tuned (the voltage difference is small), but their detection sensitivity can change dramatically if the instrument fragmentation level is high. defined a parameter $\mathrm{\Delta }{V}_{\mathrm{50}}$ ($\mathrm{d}{V}_{\mathrm{50}}$; i.e. the $\mathrm{d}V$ value at half the maximum of the signal remaining) to describe the analyte and reagent ion binding strength. In this study, the $\mathrm{d}{V}_{\mathrm{50}}$ is defined by the following equation: 4$$\mathrm{NSR}={\displaystyle \frac{\mathrm{SR}}{\mathrm{1}+e^{-k\times \left(\mathrm{d}V-\mathrm{d}{V}_{\mathrm{50}}\right)}}}+{\mathrm{SR}}_{max,\mathrm{pred}},$$ where NSR is the normalised signal remaining, SR is the signal remaining, $\mathrm{d}{V}_{\mathrm{50}}$ is the desired fitted value as represented in Fig. 7, and SR${}_{max,\mathrm{pred}}$ is the fitted value that represents the maximum SR when a compound does not undergo fragmentation while passing through the ion optics.

Figure 7

The voltage at which 50 % of analyte-+bromide adducts have dissociated ($\mathrm{d}{V}_{\mathrm{50}}$) vs. the fragmentation free enthalpies of the adducts (Table ).

[Figure omitted. See PDF]

Additionally, the formation of free enthalpies of various charged clusters are calculated using quantum chemical calculations (see Sect. 2) and are compared with $\mathrm{d}{V}_{\mathrm{50}}$, as shown in Fig. . The two sets of parameters, consisting of theoretical predictions and measurements of the binding strength, provide a consistent understanding, as demonstrated in previous studies (; ). In summary, strongly bonded charged clusters exhibit larger fragmentation-free enthalpies, larger $\mathrm{d}{V}_{\mathrm{50}}$ values, and lower calibration factors. Examples of such species include ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}\mathrm{⚫}{\mathrm{Br}}^{-}$ and ${\mathrm{I}}_{\mathrm{2}}\mathrm{⚫}{\mathrm{Br}}^{-}$. On the other hand, weakly bonded charged clusters exhibit opposite properties, including species like ${\mathrm{HO}}_{\mathrm{2}}\mathrm{⚫}{\mathrm{Br}}^{-}$, ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}\mathrm{⚫}{\mathrm{Br}}^{-}$, and ${\mathrm{SO}}_{\mathrm{2}}\mathrm{⚫}{\mathrm{Br}}^{-}$.

3.7

Validation of the iodic acid (${\mathrm{HIO}}_{\mathrm{3}}$) measurement

Oxidised iodine vapours have been shown to influence atmospheric oxidation capacity and particle formation processes . Recent publications have proposed iodine oxoacids as the critical driver for iodine particle formation processes . However, active debate remains concerning the presence of gaseous ${\mathrm{HIO}}_{\mathrm{3}}$ and whether ${\mathrm{HIO}}_{\mathrm{3}}$ plays an important role in atmospheric aerosol nucleation. For example, a recent laboratory study sheds doubt on the existence of gaseous ${\mathrm{HIO}}_{\mathrm{3}}$, as the authors only managed to measure ${\mathrm{HIO}}_{\mathrm{3}}$ in the particle phase, with a photoionisation mass spectrometer, but not in the gas phase. They concluded that the particle-phase ${\mathrm{HIO}}_{\mathrm{3}}$ was formed from higher iodine oxides instead of from gaseous ${\mathrm{HIO}}_{\mathrm{3}}$ . Furthermore, they proposed a hypothesis that the ${\mathrm{IO}}_{\mathrm{3}}^{-}$ signal, previously attributed to gaseous ${\mathrm{HIO}}_{\mathrm{3}}$ measurements using ${\mathrm{NO}}_{\mathrm{3}}^{-}$–CIMS , could also originate from ${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{2}-\mathrm{4}}$ species. Their evidence is primarily the exothermicity of the reactions from ${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{2}-\mathrm{4}}$ $+$ ${\mathrm{NO}}_{\mathrm{3}}^{-}$ to ${\mathrm{IO}}_{\mathrm{3}}^{-}$. However, it should be noted that exothermic reactions do not guarantee that the reactions occur at significant rates, as various transition states and barriers exist in these reactions. Therefore, direct instrument validation is desired. In a more recent study, alternatively used the nitrate chemical ionisation method and detected gaseous ${\mathrm{HIO}}_{\mathrm{3}}$, which is consistent with our earlier studies regarding the existence of gaseous ${\mathrm{HIO}}_{\mathrm{3}}$. The authors suggested that the measured ${\mathrm{HIO}}_{\mathrm{3}}\mathrm{⚫}{\mathrm{NO}}_{\mathrm{3}}^{-}$ ion, previously interpreted as ${\mathrm{HIO}}_{\mathrm{3}}$, could potentially be formed through reactions such as the following:​​​​​​​ R1$${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{3}}\mathrm{⚫}{\mathrm{HNO}}_{\mathrm{3}}{\mathrm{NO}}_{\mathrm{3}}^{-}\to {\mathrm{IONO}}_{\mathrm{2}}+{\mathrm{HIO}}_{\mathrm{3}}\mathrm{⚫}{\mathrm{NO}}_{\mathrm{3}}^{-},$$ due to the reaction being exothermic. This hypothesis is challenged by the fact that the reaction R2$${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{3}}\mathrm{⚫}{\mathrm{HNO}}_{\mathrm{3}}{\mathrm{NO}}_{\mathrm{3}}^{-}\to {\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{3}}\mathrm{⚫}{\mathrm{NO}}_{\mathrm{3}}^{-}+{\mathrm{HNO}}_{\mathrm{3}}$$ is a favoured pathway compared to Reaction (), as shown in Fig. . The authors of additionally estimated that the MESMER-derived overall rate coefficients at 298 K and 1 atm for Reactions () and () are $\mathrm{2.3}\times {\mathrm{10}}^{-\mathrm{12}}$ and $\mathrm{1.26}\times {\mathrm{10}}^{-\mathrm{9}}$ cm${}^{\mathrm{3}}$ molec.${}^{-\mathrm{1}}$ s${}^{-\mathrm{1}}$, respectively. Therefore, the yield of Reaction () is close to unity and cannot affect the ${\mathrm{HIO}}_{\mathrm{3}}$ detection.

Figure 8

Fragmentation pathways of ${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{3}}\mathrm{⚫}{\mathrm{HNO}}_{\mathrm{3}}{\mathrm{NO}}_{\mathrm{3}}^{-}$. The enthalpies are calculated at the DLPNO-CCSD(T)/def2-QZVPP//wB97X-D/aug-cc-pVTZ-PP level of theory.

[Figure omitted. See PDF]

It is essential to highlight that our previous studies and the studies by have consistently concluded that ${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{4}}$ is the predominant form of ${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{y}}$. Fortunately, the gaseous ${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{4}}$ species can be effectively measured using both the ${\mathrm{NO}}_{\mathrm{3}}^{-}$ and ${\mathrm{Br}}^{-}$ chemical ionisation methods. calculated the cluster formation enthalpy of ${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{4}}\mathrm{⚫}{\mathrm{NO}}_{\mathrm{3}}^{-}$ as $-$45.6 kcal mol${}^{-\mathrm{1}}$, which indicates that the ${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{4}}\mathrm{⚫}{\mathrm{NO}}_{\mathrm{3}}^{-}$ cluster is extremely stable. found that the ${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{4}}$ $+$ ${\mathrm{NO}}_{\mathrm{3}}^{-}$ $\to$ products $+$ ${\mathrm{IO}}_{\mathrm{3}}^{-}$ reaction is endothermic thus less likely to occur. The same principle applies to the ${\mathrm{Br}}^{-}$ chemical ionisation method as well. As mentioned earlier in the previous section, voltage scan experiments have shown that the ${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{4}}\mathrm{⚫}{\mathrm{Br}}^{-}$ cluster is the most stable among the clusters investigated (refer to Fig. ). Consequently, ${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{4}}$ is detected at the collision limit using the ${\mathrm{Br}}^{-}$ chemical ionisation method, and it does not fragment into species such as ${\mathrm{IO}}_{\mathrm{3}}^{-}$. Given that the measured concentration of ${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{4}}$ is more than 1 order of magnitude lower than that of ${\mathrm{HIO}}_{\mathrm{3}}$, according to previous studies , it is unlikely that ${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{3}-\mathrm{4}}$ has a significant impact on the detection of ${\mathrm{HIO}}_{\mathrm{3}}$.

Figure 9

The normalised ${\mathrm{IO}}_{\mathrm{3}}^{-}$ signal vs. the normalised signals of (a) ${\mathrm{HIO}}_{\mathrm{3}}\mathrm{⚫}{\mathrm{NO}}_{\mathrm{3}}^{-}$, (b) ${\mathrm{IONO}}_{\mathrm{2}}\mathrm{⚫}{\mathrm{NO}}_{\mathrm{3}}^{-}$, (c) ${\mathrm{HIO}}_{\mathrm{3}}\mathrm{⚫}{\mathrm{Br}}^{-}$, (d) ${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{3}}\mathrm{⚫}{\mathrm{Br}}^{-}$, and (e) ${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{4}}\mathrm{⚫}{\mathrm{Br}}^{-}$. The iodine injection and light intensity were kept constant, but the ${\mathrm{O}}_{\mathrm{3}}$ concentration was varied to modulate the ratio of iodine oxides to oxoacids. Error bars show 1 standard deviation. Notice the different $y$-axis scales.

[Figure omitted. See PDF]

Most importantly, complex and distinct chain reactions lead to the formation of higher iodine oxides and iodine oxoacids. Conducting laboratory experiments with elevated iodine concentrations could inevitably disrupt the balance between iodine oxides and iodine oxoacids. The concentration of iodine monoxide ($\mathrm{IO}$) is widely regarded as being a reliable indicator of atmospheric iodine chemistry intensity and has been shown to influence the ratio of iodine oxides to iodine oxoacids . Leveraging this phenomenon, we executed chemical perturbation experiments by varying the concentration of ozone (${\mathrm{O}}_{\mathrm{3}}$), while maintaining constant concentrations of iodine (${\mathrm{I}}_{\mathrm{2}}$) and light intensity within a laminar flow reactor. The concentrations of $\mathrm{IO}$ were carefully controlled to levels below to a few parts per trillion by volume levels. These experiments were replicated for both ${\mathrm{Br}}^{-}$–MION2–T1 and ${\mathrm{NO}}_{\mathrm{3}}^{-}$–MION2–T1, as depicted in Fig. . The measured ${\mathrm{IO}}_{\mathrm{3}}^{-}$ signal was compared with ${\mathrm{HIO}}_{\mathrm{3}}\mathrm{⚫}{\mathrm{NO}}_{\mathrm{3}}^{-}$ and ${\mathrm{IONO}}_{\mathrm{2}}\mathrm{⚫}{\mathrm{NO}}_{\mathrm{3}}^{-}$ signals from ${\mathrm{NO}}_{\mathrm{3}}^{-}$–MION2–T1 and with ${\mathrm{HIO}}_{\mathrm{3}}\mathrm{⚫}{\mathrm{Br}}^{-}$, ${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{3}}\mathrm{⚫}{\mathrm{Br}}^{-}$, and ${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{4}}\mathrm{⚫}{\mathrm{Br}}^{-}$ signals from the ${\mathrm{Br}}^{-}$–MION2–T1, in order to ascertain the source of ${\mathrm{IO}}_{\mathrm{3}}^{-}$. It is noteworthy that the gaseous signals of ${\mathrm{HIO}}_{\mathrm{3}}$ (${\mathrm{HIO}}_{\mathrm{3}}\mathrm{⚫}{\mathrm{NO}}_{\mathrm{3}}^{-}$ and ${\mathrm{HIO}}_{\mathrm{3}}\mathrm{⚫}{\mathrm{Br}}^{-}$) exhibit a perfectly linear relationship with the ${\mathrm{IO}}_{\mathrm{3}}^{-}$ signals. However, the signals of ${\mathrm{IONO}}_{\mathrm{2}}\mathrm{⚫}{\mathrm{NO}}_{\mathrm{3}}^{-}$, ${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{3}}\mathrm{⚫}{\mathrm{Br}}^{-}$, and ${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{4}}\mathrm{⚫}{\mathrm{Br}}^{-}$ demonstrate a non-linear dependence on ${\mathrm{IO}}_{\mathrm{3}}^{-}$. This observation implies that the primary source of ${\mathrm{IO}}_{\mathrm{3}}^{-}$ is gaseous ${\mathrm{HIO}}_{\mathrm{3}}$, as a non-linear correlation between ${\mathrm{HIO}}_{\mathrm{3}}\mathrm{⚫}{\mathrm{NO}}_{\mathrm{3}}^{-}$, ${\mathrm{HIO}}_{\mathrm{3}}\mathrm{⚫}{\mathrm{Br}}^{-}$, and ${\mathrm{IO}}_{\mathrm{3}}^{-}$ would be expected if ${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{3}-\mathrm{4}}$ significantly contributed to ${\mathrm{IO}}_{\mathrm{3}}^{-}$. Furthermore, if the proposed Reaction () were to occur at a substantial rate, one would anticipate the ${\mathrm{IONO}}_{\mathrm{2}}\mathrm{⚫}{\mathrm{NO}}_{\mathrm{3}}^{-}$ signal to display a linear dependence on ${\mathrm{IO}}_{\mathrm{3}}^{-}$ and ${\mathrm{HIO}}_{\mathrm{3}}\mathrm{⚫}{\mathrm{Br}}^{-}$. However, this is not observed.

Hence, we deduce that the contribution of ${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{3}-\mathrm{4}}$ to the ${\mathrm{IO}}_{\mathrm{3}}^{-}$ and ${\mathrm{HIO}}_{\mathrm{3}}\mathrm{⚫}{\mathrm{NO}}_{\mathrm{3}}^{-}$ signals within the marine boundary layer conditions is improbable. Consistent experimentation with precursor concentrations at ambient levels consistently reveals a notable scarcity of gaseous ${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{4}}$ when compared to ${\mathrm{HIO}}_{\mathrm{3}}$ . Furthermore, simulations of iodine chemistry conducted for the Maïdo Observatory in Réunion, indicate that the collective concentration of ${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{3}}$ and ${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{4}}$ accounts for a mere 1 % of ${\mathrm{HIO}}_{\mathrm{3}}$, thus making it unlikely to exert any substantial influence on ${\mathrm{HIO}}_{\mathrm{3}}$ measurements or the generation of iodine particles in marine boundary layer conditions .

4 Summary and conclusion

In this study, we present an upgraded version of the multi-scheme chemical ionisation inlet known as MION2. It is capable of simultaneously operating in atmospheric ion measurement mode and employing multiple chemical ionisation methods. While the fundamental concept of this inlet remains the same as MION1 , MION2 enhances the operational stability and enables the concurrent use of multiple chemical ionisation methods with the same ionisation time. Additionally, we find that the new version significantly improves performance by effectively focusing reagent ions, resulting in lower limits of detection (LOD).

We further developed a Python open-source flow reactor kinetic model (MARFORCE; see ) to simulate convection–diffusion–reaction equations in cylindrical flow reactors to calibrate gaseous species such as ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$, $\mathrm{HOI}$, and ${\mathrm{HO}}_{\mathrm{2}}$. The model is also compatible with the widely used Master Chemical Mechanism, thus allowing future implementation of other chemical mechanisms.

Furthermore, we undertook a comprehensive characterisation of the MION2 inlet's capabilities for detecting an array of inorganic species, employing both ${\mathrm{Br}}^{-}$ and ${\mathrm{NO}}_{\mathrm{3}}^{-}$ chemical ionisation techniques with distinct ionisation times. By combining the analytical calibration with the MARFORCE model, we quantified the photochemical production of ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$, $\mathrm{HOI}$, and ${\mathrm{HO}}_{\mathrm{2}}$ within a flow reactor. We reveal that the LOD hover around ${\mathrm{10}}^{\mathrm{5}}$ molec. cm${}^{-\mathrm{3}}$ (averaged over a 1 min interval) for species such as ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ and ${\mathrm{HIO}}_{\mathrm{3}}$, with an ionisation time of 35 ms. With a longer ionisation time (300 ms), the LOD for ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ experiences further reduction to $\mathrm{2.9}\times {\mathrm{10}}^{\mathrm{4}}$ molec. cm${}^{-\mathrm{3}}$ (approximately 1 ppqv). Upon direct comparison, the MION2 inlet demonstrates equivalent or superior LOD in contrast to the widely employed Eisele inlet . Thus, this enhanced iteration of the inlet showcases exceptional sensitivity, making it a formidable asset for the precise measurement of trace gases relevant to atmospheric particle formation.

Furthermore, we conducted an assessment of the detection capabilities of ${\mathrm{SO}}_{\mathrm{2}}$ and ${\mathrm{I}}_{\mathrm{2}}$, since they serve as crucial precursors for ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ and ${\mathrm{HIO}}_{\mathrm{3}}$, respectively. We found that the ${\mathrm{Br}}^{-}$–MION2 inlet is capable of detecting ${\mathrm{SO}}_{\mathrm{2}}$ by diluting a gas cylinder containing a known quantity of ${\mathrm{SO}}_{\mathrm{2}}$. In addition to our previously established methods for calibrating gaseous ${\mathrm{I}}_{\mathrm{2}}$ , we successfully employed a derivatisation approach in conjunction with high-performance liquid chromatography to quantify the iodine permeation rate, which was found to be as low as 17.3 ng min${}^{-\mathrm{1}}$. The ${\mathrm{I}}_{\mathrm{2}}$ calibration using the ${\mathrm{Br}}^{-}$–MION2 inlet further confirms that ${\mathrm{I}}_{\mathrm{2}}$ is detected at the collision limit, similar to ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$, and aligns with our previous estimations .

As the ${\mathrm{Br}}^{-}$–MION2 measures ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$ in the form of ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}\mathrm{⚫}{\mathrm{Br}}^{-}$, we quantified the ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$ detection with a dew point mirror instrument by running them side by side. As a large portion of ${\mathrm{Br}}^{-}$ is converted to ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}\mathrm{⚫}{\mathrm{Br}}^{-}$ in the ion molecule reaction chamber, we predicted the fragmentation pathways of analyte–${\mathrm{H}}_{\mathrm{2}}\mathrm{O}\mathrm{⚫}{\mathrm{Br}}^{-}$ clusters using quantum chemical calculations. We show that ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$ evaporates from the analyte–${\mathrm{H}}_{\mathrm{2}}\mathrm{O}\mathrm{⚫}{\mathrm{Br}}^{-}$ clusters when passing the ion optics of our mass spectrometer, due to the weak attachment of ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$ to the charged clusters. However, the chemical signature of the analyte is commonly preserved as the analyte–${\mathrm{Br}}^{-}$ cluster or deprotonated analyte anion.

We have observed that the application of the ${\mathrm{Br}}^{-}$ chemical ionisation method at atmospheric pressure is susceptible to the influence of elevated moisture levels in the ambient air, despite its enhanced detection sensitivity. We find that analytes detected at collision limits, including ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$, ${\mathrm{HIO}}_{\mathrm{3}}$, and ${\mathrm{I}}_{\mathrm{2}}$, experience a noticeable reduction in measurement sensitivity when the dew point rises beyond 0.5–10.5 ${}^{\circ }$C (equivalent to 20 %–40 % RH). Moreover, the detection of analytes with weaker bonds, such as ${\mathrm{HO}}_{\mathrm{2}}$ and ${\mathrm{SO}}_{\mathrm{2}}$, is notably more profoundly impacted by variations in water content, even in situations where the dew point remains below 0 ${}^{\circ }$C. To illustrate, the LOD for ${\mathrm{HO}}_{\mathrm{2}}$ is approximately 1 order of magnitude higher than that of ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ at a relative humidity (RH) of 2.7 %, while the LOD for ${\mathrm{SO}}_{\mathrm{2}}$ surpasses that of ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ by approximately 3 orders of magnitude when the RH falls below 0.1 %. These outcomes show the critical role played by atmospheric water content in the effectiveness of ${\mathrm{Br}}^{-}$ chemical ionisation, particularly for species characterised by weak bonding interactions. Such insights contribute to our understanding of the method's limitations and provide valuable considerations for optimising analytical conditions in future atmospheric studies.

In order to mitigate the impact of humidity on detection, two methods, namely the dilution method and the core-sampling method, were tested in this study. We found that both methods effectively reduce the influence of humidity on detection. By employing these methods, it becomes possible to detect ambient levels of ${\mathrm{SO}}_{\mathrm{2}}$ (below 1 part per billion by volume), even at RH levels of up to 50 %, which would otherwise be challenging. However, it should be noted that the use of these methods inevitably results in sample dilution, thereby affecting the detection of species that are less affected by air water content, such as ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$, $\mathrm{HOI}$, and ${\mathrm{I}}_{\mathrm{2}}$. Therefore, these methods should be employed selectively, when there is a specific objective, such as detecting extremely low levels of ${\mathrm{SO}}_{\mathrm{2}}$, or when the sample's dew point is higher than 10 ${}^{\circ }$C (40 % RH). This implies that atmospheric pressure ${\mathrm{Br}}^{-}$ chemical ionisation is suitable for laboratory experiments with controlled RH and for ambient measurements in relatively cold environments. When interpreting data obtained through the atmospheric pressure ${\mathrm{Br}}^{-}$ chemical ionisation method, it is crucial to carefully account for the influence of water by employing analytical characterisation or predicting fragmentation enthalpy. Despite these considerations, the MION2 inlet, which allows for the concurrent operation of the water-insensitive ${\mathrm{NO}}_{\mathrm{3}}^{-}$ chemical ionisation method and the water-sensitive yet more versatile ${\mathrm{Br}}^{-}$ chemical ionisation method, provides a more comprehensive understanding of atmospheric conditions compared to using either of these methods in isolation.

Finally, we validated the measurement of gaseous ${\mathrm{HIO}}_{\mathrm{3}}$ using both the ${\mathrm{NO}}_{\mathrm{3}}^{-}$ and ${\mathrm{Br}}^{-}$ chemical ionisation methods. The signal of ${\mathrm{HIO}}_{\mathrm{3}}$ typically consists of ${\mathrm{IO}}_{\mathrm{3}}^{-}$ and either ${\mathrm{HIO}}_{\mathrm{3}}\mathrm{⚫}{\mathrm{NO}}_{\mathrm{3}}^{-}$ or ${\mathrm{HIO}}_{\mathrm{3}}\mathrm{⚫}{\mathrm{Br}}^{-}$, depending on the chemical ionisation method employed. Through experimental and theoretical validation, we confirmed that all three ions primarily originate from genuine gaseous iodic acid and that iodine oxides do not contribute to the formation of these ions under marine boundary layer conditions.

Appendix A Table A1

Chemical reactions and the reaction rate coefficients used for ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ and $\mathrm{HOI}$ calibration experiments.

Chemical reactions	Reaction rate coefficients
${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ calibration
1. OH $+$ ${\mathrm{SO}}_{\mathrm{2}}$ $\to$ ${\mathrm{HSO}}_{\mathrm{3}}$	$\mathrm{1.32}\times {\mathrm{10}}^{-\mathrm{12}}\times (\mathrm{temp}/\mathrm{300}{{)}^{-\mathrm{0.7}}}^{\mathrm{a}}$
2. OH $+$ ${\mathrm{HO}}_{\mathrm{2}}$ $\to$ ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$ $+$ ${\mathrm{O}}_{\mathrm{2}}$	$\mathrm{4.8}\times {\mathrm{10}}^{-\mathrm{11}}\times \exp (\mathrm{250}/\mathrm{temp}{)}^{\mathrm{b}}$
3. ${\mathrm{HO}}_{\mathrm{2}}$ $+$ ${\mathrm{HO}}_{\mathrm{2}}$ $\to$ ${\mathrm{H}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{2}}$	$\left(\mathrm{2.2}\times {\mathrm{10}}^{-\mathrm{13}}\times \exp (\mathrm{600}/\mathrm{temp})+\mathrm{1.9}\times {\mathrm{10}}^{-\mathrm{33}}\times M\times \exp (\mathrm{980}/\mathrm{temp})\right)\times \mathrm{KMT}{\mathrm{06}}^{\mathrm{b}}$
4. OH $+$ OH $\to$ ${\mathrm{H}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{2}}$	$\mathrm{2}\times \mathrm{6.9}\times {\mathrm{10}}^{-\mathrm{31}}\times (\mathrm{temp}/\mathrm{300}{)}^{-\mathrm{0.8}}\times p/(\mathrm{1.38}\times {\mathrm{10}}^{-\mathrm{23}})/\mathrm{temp}/{{\mathrm{10}}^{\mathrm{6}}}^{\mathrm{c}}$
5. OH $+$ OH $\to$ ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$	$\mathrm{6.2}\times {\mathrm{10}}^{-\mathrm{14}}\times (\mathrm{temp}/\mathrm{298}{)}^{\mathrm{2.6}}\times \exp (\mathrm{945}/\mathrm{temp}{)}^{\mathrm{b}}$
6. ${\mathrm{HSO}}_{\mathrm{3}}$ $+$ ${\mathrm{O}}_{\mathrm{2}}$ $\to$ ${\mathrm{HO}}_{\mathrm{2}}$ $+$ ${\mathrm{SO}}_{\mathrm{3}}$	$\mathrm{1.3}\times {\mathrm{10}}^{-\mathrm{12}}\times \exp (-\mathrm{330}/\mathrm{temp}{)}^{\mathrm{b}}$
7. ${\mathrm{SO}}_{\mathrm{3}}$ $+$ $\mathrm{2}{\mathrm{H}}_{\mathrm{2}}\mathrm{O}$ $\to$ ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$	$\mathrm{3.9}\times {\mathrm{10}}^{-\mathrm{41}}\times \exp (\mathrm{6830.6}/\mathrm{temp}{)}^{\mathrm{b}}$
HOI calibration
1. IO $+$ IO $\to$ I $+$ I	$\mathrm{0.11}\times \mathrm{5.4}\times {\mathrm{10}}^{-\mathrm{11}}\times \exp (\mathrm{180}/\mathrm{temp}{)}^{\mathrm{d}}$
2. IO $+$ IO $\to$ OIO $+$ I	$\mathrm{0.38}\times \mathrm{5.4}\times {\mathrm{10}}^{-\mathrm{11}}\times \exp (\mathrm{180}/\mathrm{temp}{)}^{\mathrm{d}}$
3. IO $+$ IO $\to$ ${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{2}}$	$\mathrm{0.45}\times \mathrm{5.4}\times {\mathrm{10}}^{-\mathrm{11}}\times \exp (\mathrm{180}/\mathrm{temp}{)}^{\mathrm{d}}$
4. ${\mathrm{I}}_{\mathrm{2}}$ $+$ OH $\to$ HOI $+$ I	$\mathrm{2.1}\times {{\mathrm{10}}^{-\mathrm{10}}}^{\mathrm{e}}$
5. IO $+$ OIO $\to$ ${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{3}}$	$w\mathrm{1}a\times \exp (w\mathrm{2}a\times \mathrm{temp}{)}^{\mathrm{f}}$
6. OIO $+$ OIO $\to$ ${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{4}}$	$w\mathrm{1}b\times \exp (w\mathrm{2}b\times \mathrm{temp}{)}^{\mathrm{f}}$
7. IO $+$ OH $\to$ ${\mathrm{HO}}_{\mathrm{2}}$ $+$ I	$\mathrm{1.0}\times {{\mathrm{10}}^{-\mathrm{10}}}^{\mathrm{g}}$
8. HI $+$ OH $\to$ ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$ $+$ I	$\mathrm{1.6}\times {\mathrm{10}}^{-\mathrm{11}}\times \exp (\mathrm{440}/\mathrm{temp}{)}^{\mathrm{b}}$
9. HOI $+$ OH $\to$ ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$ $+$ IO	$\mathrm{2.0}\times {{\mathrm{10}}^{-\mathrm{13}}}^{\mathrm{h}}$
10. I $+$ ${\mathrm{HO}}_{\mathrm{2}}$ $\to$ HI $+$ ${\mathrm{O}}_{\mathrm{2}}$	$\mathrm{1.47}\times {\mathrm{10}}^{-\mathrm{11}}\times \exp (-\mathrm{1090}/\mathrm{temp}{)}^{\mathrm{i}}$
11. IO $+$ ${\mathrm{HO}}_{\mathrm{2}}$ $\to$ HOI $+$ ${\mathrm{O}}_{\mathrm{2}}$	$\mathrm{1.4}\times {\mathrm{10}}^{-\mathrm{11}}\times \exp (\mathrm{540}/\mathrm{temp}{)}^{\mathrm{b}}$
12. OH $+$ OH $\to$ ${\mathrm{H}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{2}}$	$\mathrm{2}\times \mathrm{6.9}\times {\mathrm{10}}^{-\mathrm{31}}\times (\mathrm{temp}/\mathrm{300}{)}^{-\mathrm{0.8}}\times p/\left(\mathrm{1.38}\times {\mathrm{10}}^{-\mathrm{23}}\right)/\mathrm{temp}/{{\mathrm{10}}^{\mathrm{6}}}^{\mathrm{c}}$
13. OH $+$ OH $\to$ ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$	$\mathrm{6.2}\times {\mathrm{10}}^{-\mathrm{14}}\times (\mathrm{temp}/\mathrm{298}{)}^{\mathrm{2.6}}\times \exp (\mathrm{945}/\mathrm{temp}{)}^{\mathrm{b}}$
14. OH $+$ ${\mathrm{HO}}_{\mathrm{2}}$ $\to$ ${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$ $+$ ${\mathrm{O}}_{\mathrm{2}}$	$\mathrm{4.8}\times {\mathrm{10}}^{-\mathrm{11}}\times \exp (\mathrm{250}/\mathrm{temp}{)}^{\mathrm{b}}$
15. ${\mathrm{HO}}_{\mathrm{2}}$ $+$ ${\mathrm{HO}}_{\mathrm{2}}$ $\to$ ${\mathrm{H}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{2}}$	$\mathrm{2.2}\times {\mathrm{10}}^{-\mathrm{13}}\times \mathrm{KMT}\mathrm{06}\times \exp (\mathrm{600}/\mathrm{temp})+\mathrm{1.9}\times {\mathrm{10}}^{-\mathrm{33}}\times M\times \mathrm{KMT}\mathrm{06}\times \exp (\mathrm{980}/\mathrm{temp}{)}^{\mathrm{b}}$

${}^{\mathrm{a}}$ ; ${}^{\mathrm{b}}$ ; ${}^{\mathrm{c}}$ ; ${}^{\mathrm{d}}$ ; ${}^{\mathrm{e}}$ ; ${}^{\mathrm{f}}$ ; ${}^{\mathrm{g}}$ ; ${}^{\mathrm{h}}$ ; ${}^{\mathrm{i}}$ .$\mathrm{KMT}\mathrm{06}=\mathrm{1}+(\mathrm{1.4}\times {\mathrm{10}}^{-\mathrm{21}}\times \exp (\mathrm{2200}/\mathrm{temp})\times [{\mathrm{H}}_{\mathrm{2}}\mathrm{O}\left]\right)$, and [${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$] is the absolute water concentration. $M$ is the total number of molecules in the atmosphere. $p$ is the pressure. $w\mathrm{1}a=\mathrm{4.7}\times {\mathrm{10}}^{-\mathrm{10}}-\mathrm{1.4}\times {\mathrm{10}}^{-\mathrm{5}}\times \exp (-\mathrm{0.75}\times p/\mathrm{1.62265})+\mathrm{5.51868}\times {\mathrm{10}}^{-\mathrm{10}}\times \exp (-\mathrm{0.75}\times p/\mathrm{199.328})$;$w\mathrm{2}a=-\mathrm{0.00331}-\mathrm{0.00514}\times \exp (-\mathrm{0.75}\times p/\mathrm{325.68711})-\mathrm{0.00444}\times \exp (-\mathrm{0.75}\times p/\mathrm{40.81609})$;$w\mathrm{1}b=\mathrm{1.166}\times {\mathrm{10}}^{-\mathrm{9}}-\mathrm{7.796}\times {\mathrm{10}}^{-\mathrm{10}}\times \exp (-\mathrm{0.75}\times p/\mathrm{22.093})+\mathrm{1.038}\times {\mathrm{10}}^{-\mathrm{9}}\times \exp (-\mathrm{0.75}\times p/\mathrm{568.154})$;$w\mathrm{2}b=-\mathrm{0.00813}-\mathrm{0.00382}\times \exp (-\mathrm{0.75}\times p/\mathrm{45.57591})-\mathrm{0.00643}\times \exp (-\mathrm{0.75}\times p/\mathrm{417.95061})$.

Figure A1

Schematic of an ionisation source of the MION2 inlet.

[Figure omitted. See PDF]

Figure A2

Schematic of a typical calibration experiment connecting the MION2 inlet (internal diameter of 24 mm) with the calibration source (internal diameter of 15.6 mm).

[Figure omitted. See PDF]

Figure A3

Schematic of the set-up for examining the detection humidity effect of ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$, $\mathrm{HOI}$ and ${\mathrm{HO}}_{\mathrm{2}}$.

[Figure omitted. See PDF]

Figure A4

Schematic of the experimental set-up for iodine chemistry experiments to produce higher concentrations of iodine oxides and oxoacids.

[Figure omitted. See PDF]

Figure A5

Comparison of the ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ profiles at the outlet of a flow reactor. Theoretical values are predicted using , and the model results indicate the MARFORCE simulation. In both the theoretical prediction and the MARFORCE model, the tube length is assumed to be 2 m, the inlet flow is set to 10 L min${}^{-\mathrm{1}}$, and the diffusivity of ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ is set to 0.088 cm${}^{\mathrm{2}}$ s${}^{-\mathrm{1}}$.

[Figure omitted. See PDF]

Figure A6

Comparing calibration experiments of (a) ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$ and $\mathrm{HOI}$ with a straight tube (Fig. ) or additionally with a dilution flow (Fig. ). The difference in the calibration factors between the two experimental set-ups is the result of the less accurate representation of fluid dynamics when the dilution flow is added (Fig. ).

[Figure omitted. See PDF]

Figure A7

Schematic of the set-up for the ${\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}}$, HOI, and ${\mathrm{HO}}_{\mathrm{2}}$ calibration experiment with the tower 2. The difference between this set-up and the one shown in Fig.  is that the position of the MION2 tower is changed from tower 1 to tower 2.

[Figure omitted. See PDF]

Figure A8

The configuration of the core-sampling device (Karsa Ltd.), which is used for adjusting the sheath and sample flows. This core-sampling piece features three ports for the dilution flows which pass through a mesh and are further mixed with the sample flow.

[Figure omitted. See PDF]

Code availability

The MARFORCE model is shared through the GitHub repository (https://github.com/momo-catcat/MARFORCE-flowtube, ; 10.5281/zenodo.8318790, ). Other data analysis codes can be requested from the corresponding authors.

Data availability

Data for all figures in the main text and the Appendix​​​​​​​ are available from the Zenodo repository (10.5281/zenodo.8318790, ).

Author contributions

XCH and JS designed and carried out the experiments. JS and XCH wrote the MARFORCE model. JZ wrote the documentation of the MARFORCE model. SI carried out quantum chemical calculations. NMM, MrK, and MMK analysed the molecular iodine samples. JK, PJ, MS, and JM provided technical support. XCH wrote the paper, with contributions from JS, NMM, PJ, and SI. Finally, JK, JS, MR, SI, DRW, and MaK commented on and edited the paper.

Competing interests

Paxton Juuti and Jyri Mikkilä work for Karsa, Ltd., Finland. Juha Kangasluoma works part-time for Karsa, Ltd., Finland. The peer-review process was guided by an independent editor, and the authors also have no other competing interests to declare.

Disclaimer

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Acknowledgements

Miska Olin, Gustaf Lönn, and Heikki Junninen are acknowledged for their helpful discussions and contributions to the MARFORCE model. Simon Patrick O'Meara and Gordon McFiggans are acknowledged for their contributions to the MCM interpreter in the MARFORCE model.

Financial support

This research has been supported by the ACCC Flagship funded by the Research Council of Finland (grant no. 337549), the Academy professorship funded by the Research Council of Finland (grant no. 302958), and the Research Council of Finland (project nos. 331207, 346370, 325656, 316114, 314798, 325647, 341349, 353836, 346373 and 349659). This project has received funding from the European Research Council under the European Union's Horizon 2020 research and innovation programme (grant nos. 742206 and 101002728). The Arena for the gap analysis of the existing Arctic Science Co-Operations (AASCO) has been funded by the Prince Albert Foundation (grant no. 2859). Markku Kulmala has received funding from the Jane and Aatos Erkko Foundation. Markku Kulmala and Xu-Cheng He have received funding from the Jenny and Antti Wihuri Foundation.Open-access funding was provided by the Helsinki University Library.

Review statement

This paper was edited by Anna Novelli and reviewed by four anonymous referees.