Insights into the Chemistry of Iodine New Particle Formation: The Role of Iodine Oxides and the Source of Iodic Acid.

Iodine chemistry is an important driver of new particle formation in the marine and polar boundary layers. There are, however, conflicting views about how iodine gas-to-particle conversion proceeds. Laboratory studies indicate that the photooxidation of iodine produces iodine oxides (IxOy), which are well-known particle precursors. By contrast, nitrate anion chemical ionization mass spectrometry (CIMS) observations in field and environmental chamber studies have been interpreted as evidence of a dominant role of iodic acid (HIO3) in iodine-driven particle formation. Here, we report flow tube laboratory experiments that solve these discrepancies by showing that both IxOy and HIO3 are involved in atmospheric new particle formation. I2Oy molecules (y = 2, 3, and 4) react with nitrate core ions to generate mass spectra similar to those obtained by CIMS, including the iodate anion. Iodine pentoxide (I2O5) produced by photolysis of higher-order IxOy is hydrolyzed, likely by the water dimer, to yield HIO3, which also contributes to the iodate anion signal. We estimate that ∼50% of the iodate anion signals observed by nitrate CIMS under atmospheric water vapor concentrations originate from I2Oy. Under such conditions, iodine-containing clusters and particles are formed by aggregation of I2Oy and HIO3, while under dry laboratory conditions, particle formation is driven exclusively by I2Oy. An updated mechanism for iodine gas-to-particle conversion is provided. Furthermore, we propose that a key iodine reservoir species such as iodine nitrate, which we observe as a product of the reaction between iodine oxides and the nitrate anion, can also be detected by CIMS in the atmosphere.

Introduction

Iodine gas-to-particle conversion is a fast process known since the early laboratory studies of iodine chemistry and spectroscopy.[1−3] The nucleation rates of iodine oxide particles (IOPs) recently measured in the Cosmics Leaving Outdoor Droplets (CLOUD) chamber at the European Organization for Nuclear Research (CERN) suggest that this particle formation pathway can be competitive with sulfuric acid nucleation in pristine environments.[4] In fact, atmospheric IOP particle formation unrelated to H2SO4 was observed for the first time in Mace Head (Ireland), a mid-latitude coastal location where tidal pool algae are exposed periodically to the atmosphere, resulting in strong biogenic emissions of iodine-bearing molecules that are photo-oxidized leading to low tide-day time particle “bursts”.[5,6] Since then, there has been some debate about the potential climatic relevance of this phenomenon[7] because iodine has been shown to be ubiquitous in the marine boundary layer (MBL).[8−10] Although the atmospheric concentrations of gas-phase iodine species in the remote MBL are generally in the parts per trillion (ppt) range, new field observations in the Arctic demonstrate frequent new particle formation episodes triggered by iodine with little contribution from H2SO4.[11] Hence, a regional influence of IOPs on cloud formation and properties over the polar oceans has been suggested, which could potentially accelerate sea ice melting.[4] This could be exacerbated if the emissions of iodine from the ocean to the atmosphere are actually increasing, as indicated by Arctic and Alpine ice core measurements.[12,13] Model efforts directed to evaluating the atmospheric radiative impact of IOPs are needed, but to do that, a feasible chemical mechanism connecting iodine emissions and gas-to-particle conversion is required.

Photolysis of iodine-bearing molecular precursors such as HOI, I2, CH3I, CH2I2, and so forth in the presence of ozone leads to the formation of iodine monoxide (IO), which has been observed in the MBL and in the polar regions,[7] as well as in the free troposphere[14] and lower stratosphere.[15] Iodine dioxide (OIO) is a product of the IO self-reaction[16] that has also been observed in the MBL.[17] IO and OIO undergo rapid recombination reactions to generate higher-order iodine oxides (IO),[18] which eventually form an ultrafine aerosol of I2O5 composition when formed in a dry environment.[19] The composition of atmospheric IOPs is known to be iodic acid (HOIO2, hereafter HIO3 for simplicity), which is the hydrated form of I2O5.[20] HIO3 has been detected in IOPs by photoionization mass spectrometry (PIMS).[21,22]

Recent chemical ionization mass spectrometry (CIMS) measurements confirm that IOPs consist almost entirely of HIO3 but have otherwise challenged the knowledge on gas-to-particle conversion summarized above.[4,23] CIMS[24] has revolutionized the detection of trace atmospheric constituents (e.g., H2SO4[25]) thanks to its extremely high sensitivity, soft ionization, and selective detection and has opened a new era beyond spectroscopic detection of atoms and simple molecules. The development of improved inlets, ionization sources, and atmospheric pressure interfaces has also enabled the detection of elusive gas-phase species, amongst which are iodine-containing molecules. CIMS field observations of the iodate anion (IO3–) have been interpreted by Sipilä et al.[23] as a signature of HIO3 from an analogy with the detection of H2SO4 as HSO4–[25] and based on ab initio proton affinities of NO3– and IO3–

The dominance of the IO3– signal over that of other ions that can be linked to iodine oxides led Sipilä et al. to propose HIO3 as a major iodine-bearing molecule in the atmosphere. Reported HIO3 mixing ratios at Mace Head are comparable to or even higher than IO mixing ratios measured by laser-induced fluorescence.[26] Sipilä et al. also reported the observation of the HIO3 dimer detected as HIO3·IO3– and a mass peak progression that would be consistent with a nucleation mechanism where a cluster takes one HIO3 and upon addition of a second HIO3 sheds a water molecule. This mechanism has been amended recently considering the strong influence of instrumental settings on the observed mass spectra and currently also invokes iodous acid (HOIO, hereafter HIO2) to explain the observed mass peaks.[4] Concurrent CIMS measurements with different ionization sources appear to support the existence of gas-phase HIO2 and HIO3 in the CLOUD experiments[4,27,28] and by extension in the atmosphere.

There is however a major unknown about gas-phase HIO3: how does it form? The CIMS IO3– signal has been observed in the absence of HO in laboratory flow tube experiments[23] and in CLOUD,[4,28] although the only known thermochemically feasible route from I2 photooxidation to iodic acid is the recombination reaction[29]

It has then been postulated that HIO3 could be generated by a composite reaction involving I, O3, and H2O or by reactions between iodine oxides and water.[4,23] However, atomic iodine and H2O form a very weakly bound complex that would not live long enough to react with atmospheric O3 (assuming no barriers in that reaction), and elementary reactions of iodine oxides with H2O generating HIO3 are endothermic or exhibit barriers, according to high-level electronic structure calculations.[22,30,31] Hydrolysis of I2O by the water dimer has only been explored theoretically for I2O5,[32] although to date it is unclear whether this species actually forms in the gas phase to play a role in IOP formation.[4,19,22,28] Moreover, in our previous work, we were unable to detect gas-phase HIO3 by near-threshold PIMS at 11.6 eV, while we did detect it in the particle phase after pyrolysis of IOPs formed in a flow tube in the presence of water vapor.[22] Gas-phase reactions between iodine species and H2O are, according to these experiments, slower than ∼10–19 cm3 molecule–1 s–1. In contrast, in our work, we demonstrated that iodine oxides (IO) readily form molecular clusters whose dry composition tends asymptotically to I2O5 (whose hydrated form is HIO3). We then proposed that the IOP formation mechanism that was commonly accepted before the CIMS observations still holds, i.e. IOPs are formed from IO, and the resulting I2O5 particles hydrate to form HIO3 in the particle phase.[22] As a rebuttal to this conclusion, it has been argued that all laboratory studies on IOP formation have not been performed under atmospherically relevant conditions,[4] implying that the iodine concentration in those studies was high enough for iodine oxides to drive IOP formation through dipole–dipole enhanced second-order chemistry. In principle, it is conceivable that under the low iodine and high water mixing ratios (ppt and %, respectively) typical of the lower atmosphere, a hypothetical reaction with a low rate constant between an iodine species and water vapor could proceed at a faster rate than the recombination of iodine oxides at ppt levels. There could even be a situation where both mechanisms could be competitive, and interestingly, CIMS also detects atmospheric IO in the form of IO·NO3– or IO·Br–, although these signals are uncalibrated.[4,28]

However, another possible explanation for the apparent contradiction between CIMS and PIMS gas-phase measurements is that the ions observed by CIMS may be generated, at least in part, by ion–molecule reactions between the reagent ion and iodine oxides. Our ab initio calculations indicated that different reactions between IO with x = 2 and NO3–, Br–, CH3COO–, and H3O+ are exothermic and can potentially generate some of the ions and cluster ions that have been attributed to HIO3, in particular IO3– and HIO3·NO3– in the nitrate anion CIMS.[22,33] For example,

If this was the case, the observations of these ions in the field using CIMS should be reinterpreted as being representative of both ambient I2O and HIO3. Moreover, this would call into question the need of invoking a gas-phase species of uncertain origin such as HIO3 to interpret signals that can be explained by other species whose formation is thermochemically unhindered. Hence, there is a clear need to carry out laboratory work on ion–molecule reactions that play a role in the different ionization schemes used by CIMS instruments.

Here, we present results from flow tube-mass spectrometry experiments performed to investigate the products of IO ion–molecule reactions in the nitrate CIMS. Our results confirm our theoretical prediction that the IO3– anion (m/z = 175) and the HIO3·NO3– anion (m/z = 238), which we interpret as HNO3·IO3–, are generated from reactions between IO and nitrate core ions. This implies that these ions cannot be exclusively attributed to ambient HIO3 and that the CIMS field observations need to be reinterpreted. We also identify the source of ambient HIO3. Finally, we observe a strong signal at m/z = 251, which corresponds to the ion cluster IONO2·NO3–,[34] where iodine nitrate (IONO2) is formed in part as the coproduct of the iodate core anion in reaction . Hence, we propose that field CIMS instruments that have reported this signal[11] may inadvertently have detected for the first time the key atmospheric iodine reservoir IONO2 (Saiz-Lopez et al., 2012), for which a detection technique has not been developed to date.

Experimental Section

The interaction between NO3– and IO has been investigated by using the flowing afterglow technique, which we have used in the past to determine metal cation–electron recombination rate constants.[35,36] Experiments are carried out in a Y-shaped 3.75 cm in diameter CF-flanged flow tube coupled to a quadrupole mass spectrometer (Hiden HPR 60). A schematic diagram of the apparatus is shown in Figure . A 200 W microwave (MW) discharge on He generates electrons (109 to 1010 cm–3[35]), which are then carried by the He flow into the flow tube. A smaller flow of Ar (∼10% of the He flow) is added to quench excited He metastables generated in the MW plasma. The MW cavity is placed at 90° with respect to the flow tube to avoid irradiating the gas mixture with UV light emitted by the plasma. Once in the flow tube, the thermal electrons attach to O2 added through a side port, forming O2–, which further reacts with HNO3 added downstream of the O2 port to produce NO3– and nitrate core ions[37] with nearly 100% yield.[38] The total flow through the NO3– branch is typically 2–3 slm, and the pressure is kept around 3 Torr.

Figure 1

Flowing afterglow-fast flow tube experimental setup for ion–molecule reactions. The IO branch could be operated at the same pressure as the NO3– branch or at a higher pressure by inserting a pin-holed flange. P indicates pressure heads. Detection of negative ions was performed using a quadrupole mass spectrometer.

In the IO branch of the flow tube, a flow of He (300–500 sccm) carrying I2 and O3 is continuously irradiated with white light from a 75 W Xe lamp (Photon Technology International) through a quartz view port. In a previous study, we used this setup to generate IO, which were detected by PIMS.[22] An excess of I2 (1012 to 1013 molecule cm–3) removes on a ms time scale any OH generated by photolysis of O3 in the presence of residual or added water. The system can be operated in two pressure regimes. In the first one, IO are generated at the same pressure as NO3– (3 Torr) and the two flows are simply merged at the junction of the two branches. The residence time of the gas mixture in the IO branch is 80–140 ms. In the second regime, a flange with a 1 mm pinhole is inserted upstream of the flow tube junction to raise the pressure up to 26 Torr, increasing the residence time to about 1.7 s. In both configurations, IO (∼1012 cm–3) are generated well in excess of the concentration of NO3– core ions (<107 cm–3), and the pressure in the ion–molecule reaction region remains 3 Torr. The flows from the two branches are allowed to mix, and after a contact time of 12–21 ms, the gas is sampled through a skimmer cone with a 200 μm pinhole by the quadrupole mass spectrometer in a negative ion mode.

A roots blower (BOC Edwards, EH500A) backed with a rotary pump (BOC Edwards, E2M80) draws the gas down the flow tube. Flows are set using calibrated mass flow controllers (MKS), and the pressure is monitored using 10 and 1000 Torr calibrated capacitance manometers (MKS Baratron). The experiments are performed with CP grade He (BOC, 99.999%, [H2O] < 2 ppm) and N5 grade O2 (BOC, 99.999%, [H2O] < 1 ppm). Ozone is produced online by a corona discharge (EASELEC, ELO3G) of pure O2 at 1 bar. In some experiments, water vapor (deionized) is entrained in the flow tube by passing the carrier flow through a bubbler. Liquid HNO3 (Sigma-Aldrich, 99.5%) was stored in a glass finger container with 1/4″ connections in order to transfer it to a glass vacuum line equipped with 10 L glass bulbs. HNO3 is in equilibrium with NO2, which was removed by adding a few drops of H2SO4 (J.T. Baker, >51%). The glass finger was subsequently pumped for a few minutes before HNO3 vapor (vapor pressure of 30 Torr at 295 K) was released into the vacuum line in order to make up a diluted mixture in He (2%).

Data were acquired in the form of mass spectra in a negative ion mode usually in the range between 50 and 500 amu. Positive ion and neutral (electron impact ionization) mass spectra were also acquired for characterization of the flowing afterglow. Mass spectra were taken at 0.1 amu steps (10 accumulations). In some experiments, the signal of a set of selected peaks was followed in time to observe variations when changing the experimental conditions.

Electronic structure calculations were carried out to support the interpretation of the experimental data. The stationary points on the potential energy surfaces (PES) of selected reactions were first determined using the hybrid density functional/Hartree–Fock B3LYP method from within the Gaussian 16 suite of programs,[39] combined with the standard 6-311+G(2d,p) triple-ζ basis set for O, N, and H, together with an all-electron basis set for I which was designed for G2 level calculations.[40] This basis set may be described as a supplemented (15s12p6d)/[10s9p4d] 6-311G basis, the [5211111111,411111111,3111] contraction scheme being supplemented by diffused s and p functions, together with d and f polarization functions. Following geometry optimizations and determination of vibrational frequencies and (harmonic) zero-point energies, the energies of the stationary points relative to the reactants were obtained. Higher quality calculations of the relative energies of the reactants and products were made using the B3LYP functional and the significantly larger aug-cc-pVQZ basis set.[41] For I, the aug-cc-pVQZ basis set of Peterson el al.[42] was used. The accuracy of the reaction enthalpies calculated with this method is estimated here to be around ±20 kJ mol–1. A better accuracy may be expected for a large basis set such as aug-cc-pVQZ, but spin–orbit effects are not included, so this is likely a safe estimate. In a limited number of cases, fixed point CCSD(T) energy calculations have been carried out using the geometries optimized at the B3LYP/gen level (i.e., with the “G2” basis set).

Results

Dry Experiments

Mass spectra recorded in the absence and presence of IO without added water are shown in Figure . These experiments were run after pumping down the system to a few mTorr without having added any water prior to the observations. From mass spectrometric residual gas analysis (RGA) using electron impact ionization with and without adding water (e.g., Figure S1c), an upper limit to the water concentration in the IO flow tube of 2 × 1013 molecule cm–3 is estimated (i.e., 4 orders of magnitude lower than atmospheric concentrations).

Figure 2

Mass spectrum of iodine oxide ions and iodine oxide-nitrate cluster ions (black line). Iodine oxides formed at 3 Torr after 137 ms and without addition of water vapor to the gas flow, prior to the ion–molecule reactions. Iodine-nitrate ions formed after 12 ms of the reaction time between the two gas flows. The spectrum of the nitrate core ion source (no IO) is also shown for comparison (red line). Note the logarithmic vertical scale.

Table lists the mass peaks shown in Figure (“dry”) with the corresponding ion assignment and the proposed parent molecule. In these experiments, the pressure in the IO branch was the same as in the NO3– branch (3 Torr). In the absence of iodine oxides, the spectra show the expected peak progression of nitrate core ion peaks at m/z = 62 (NO3–), m/z = 125 (HNO3·NO3–), m/z = 188 ((HNO3)2·NO3–), and m/z = 251 ((HNO3)3·NO3–). The relative signal at m/z = 62 and m/z = 125 peak is determined by pressure and residence time of the gas in the flow tube, with higher pressure and slower flow promoting (HNO3)·NO3– (Figure S2a,b).

Table 1

Observed Peaks and Intensities, Dependence on Light and Humidity, and Assigned Parent Molecules

this work								CIMS literature
peaka	anion	m/z	Intb	no O3c	dark	dry	H2Od	parente	FTf	ECg	Fh
127	I–	126.9	5–6	yes	yes	yes		I2
143	IO–	142.9	2–3	no	yes	yes		I2	yes
145	H2O·I–	144.9	2–3	no	yes	no		I2
163	(H2O)2·I–	162.9	2–3	no	yes	yes		I2
175	IO3–	174.9	4–5	no	yes	yes	↑	I2Oy; y=2–5; HIO3	yes	yes	yes
190	HNO3·I–	189.9	3–4	yes	yes	yes	↑	I2
205	IO·NO3–	204.9	3–4	yes	yes	yes	*	IO		yes
221	OIO·NO3–	220.9	4–5	yes	yes	yes	*	OIO	p	yes	yes
222	HNO3·IO2–; HIO2·NO3–	221.9	3–4	yes	yes	yes	*	I2O2; HIO2		yes	yes
238	HNO3·IO3–; HIO3·NO3–	237.9	4–5	no	yes	yes	↑	I2O3; HIO3	p	yes	yes
251	IONO2·NO3–i	250.9	5–6	yes	yes	yes	*	I2O3	p	yes	yes
254	I2–	253.8	3–4	yes	yes	yes	↓	I2
267	OIONO2·NO3–	266.9	4–5	yes	yes	yes	*	I2O4	p	yes	yes
283	O2IONO2·NO3–	282.9	3–4	no	yes	yes	↓	I2O5	p	no	yes
285	(HNO3)2·IO2–; HIO2·(HNO3)·NO3–	284.9	3–4	yes	yes	yes	↑	I2O2; HIO2	p	yes	yes
301	(HNO3)2·IO3–; HIO3·(HNO3)·NO3–	300.9	3–4	no	yes	yes	↑	I2Oy; y=2,3; HIO3	p	yes	yes
314	IONO2·HNO3·NO3–	313.9	3–4	yes	yes	yes	*	I2O3	p
316	I2·NO3–	315.8	2–3	yes	yes	yes	↔	I2
330	OIONO2·HNO3·NO3–	329.9	3–4	yes	yes	yes	*	I2O4	p
334	IO2·IO3–	333.8	2–3	no	no	no	↑	HIO3·OIO	p
346	O2IONO2·HNO3·NO3–	345.9	2–3	yes	yes	yes	*	I2O5
348	I2O2·NO3–	347.8	3–4	no	yes	yes	↔	I2O2		yes
351	HIO3·IO3–	350.8	1–2	no	no	no	↑	(HIO3)2	yes
364	I2O3·NO3–	363.8	3–4	no	yes	yes	↔	I2O3	p	yes
366	I2O2·H2O·NO3–	365.8	1–2	no	no	no	↑	I2O2·H2O
380	I2O4·NO3–j	379.8	2–3	no				I2O4	p	yes	yes
381	I3–	380.7	5–6	yes	yes	yes	*	I2
396	I2O5·NO3–	395.8	3–4	no	no	yes	↓	I2O5	yes	yes	yes
398	I2O4·H2O·NO3–; H2I2O5·NO3–	397.8	2–3	no	yes	no	↑	I2O4·H2O; H2I2O5	p	yes	yes
411	I2O2·HNO3·NO3–	410.8	2–3	no	yes	yes	↔	I2O2	p
427	I2O3·HNO3·NO3–	426.8	2–3	no	yes	yes	↔	I2O3	p	yes
440	(IONO2)2·NO3–;IONO2·(HNO3)3·NO3–	439.8	2–3	no	yes	yes	*	I2O3
442	OIO·O2IONO2·NO3–	441.8	2–3	no	no	yes	↓	OIO; I2O5
443	I2O4·HNO3·NO3–	442.75	2–3	yes	yes	yes	↔	I2O4	p	yes	yes
456	OIONO2·(HNO3)3·NO3–	455.9	2–3	no	yes	yes	↔	I2O4
461	H2I2O5·HNO3·NO3–; I3O5–	460.75	2–3	no	no	yes	↓	H2I2O5; HIO3·I2O2	p	yes
477	(HIO3)2·HNO3·NO3–; I3O6–	476.75	1–2	no	no	yes	*	HIO3; HIO3·I2O3	p
488	OIONO2·O2IONO2·NO3–	487.8	2–3	no	no	yes	↓	I2O4 and I2O5
493	I3O7–	493.7	1–2	no	no	no	↑	HIO3·I2O4	p

Integer mass (number of neutrons + number of protons).

Average peak intensity logarithmic range (x–y indicates the signal between 10 and 10).

Indicate if the anion signal is above the detection limit without O3 in the dark and without adding H2O.

Indicates the effect of adding H2O on the photolytic signal of each anion after correcting for the effect of H2O on the nitrate core ions: increase (↑), decrease (↓), no change (↔), and unclear (*).

Refers to neutral molecules from the IO flow tube that originate in the observed ion.

Flow tube CIMS: Sipilä et al. 2016 (Figure S4). “Yes” indicates positive detection. Since no table is provided in the original paper, the figure has been digitized; “p” indicates possible detection (i.e., there is a mass in the mass defect plot very close to the mass in the first column of the present table).

Environmental Chamber CIMS: He et al. 2021, Table S2 and Figure S4.

Field CIMS: Baccarini et al. 2020, Table S1.

Overlaps with (HNO3)3·NO3–.

Overlaps with I3–, but it can be observed by subtraction of mass spectra.

Addition of molecular iodine to the flow results in a substantial decrease in the nitrate core ion peaks (Figure S3a) and concurrent appearance of new mass peaks. Peaks at m/z = 127, 254, and 381 indicate the presence of I–, I2–, and I3–, respectively. The latter is a prominent signal that has also been observed in iodine-based CIMS.[34] The peak at m/z = 251 increases by 2 orders of magnitude, and we identify it now as the halogen-bonded complex IONO2·NO3– observed in previous CIMS work when I2 and NO3– are present in sampled air.[34] Other minor masses observed are m/z = 205 (IO·NO3), m/z = 221 (OIO·NO3–), m/z = 222 (HNO3·IO2–), m/z = 254 (I2–), m/z = 267 (OIONO2·NO3–), m/z = 314 (IONO2·HNO3·NO3–), m/z = 316 (I2·NO3–), m/z = 440 ((IONO2)2·NO3– or IONO2·(HNO3)3·NO3–), and m/z = 443 (I2O4·HNO3·NO3–). The oxidation of I2 is not photochemical but caused by surface chemistry following I2 deposition on the wall downstream of the ionization region (note that the gas-phase reaction NO3– + I2 → IONO2 + I– is endothermic using evaluated enthalpies of formation[43,44]).

When iodine oxides are made by adding ozone to the flow, additional peaks of iodine-containing ions emerge, and most peaks that had appeared in the presence of I2 (Figure S3a) increase substantially (Figure S3b). Irradiation with the Xe lamp beam enhances the signals by a factor of 1.5–2.5 (Figure S4a,c), except for I3–, which decreases by ∼5%. This means that IO are generated in this system both by a dark reaction between I2 and O3 and by gas-phase photochemistry[22] within a residence time of tens to hundreds of milliseconds. The gas-phase reaction between I2 and O3 is slow,[18,45] which means that additional wall chemistry is taking place in this system. The flow is not turbulent (Reynold numbers are low), but radial diffusion is favored by relatively low pressures and by the use of He as a carrier gas. This dark source of IO helps to pinpoint species generated exclusively by photochemistry.

The new masses that appear in the mass spectra when IO are made by ozone and/or irradiation are m/z = 175 (IO3–), m/z = 238 (HNO3·IO3–), m/z = 301 ((HNO3)2·IO3–), m/z = 283 (O2IONO2·NO3–), m/z = 348 (I2O2·NO3–), m/z = 364 ((HNO3)3·IO3– and I2O3·NO3–), m/z = 396 (I2O5·NO3–), m/z = 411 (I2O2·HNO3·NO3–), and m/z = 427 (I2O3·HNO3·NO3–) (Figure S3b). Of the three iodate core ion peaks, the most prominent one is generally HNO3·IO3–. The I2O5·NO3– anion is only generated in the presence of light (Figure S4a,c). Other minor peaks are detected at higher m/z (see Figure S4b,d and Table ).

Decreasing the ozone or the iodine concentrations results in the reduction of all these ions and also of IONO2·NO3–, which shows the same behavior as the (HNO3)·IO3– ions on top of its background signal (see time traces in Figure S5). By contrast, I3– increases with lower ozone and with a higher I2 concentration and can be used as a proxy for I2. Reducing the reaction time by injecting the ozone flow further downstream results in reduction of most signals (Figure S4) both for the dark and the photolytic source. It should be noted that because I2, O3, and IO are in excess over the available charged species, variations of the conditions in the IO flow tube may also change the available charge and the relative concentrations of the nitrate core ions. For example, adding more I2 may reduce the (HNO3)·NO3– ions available for reaction with IO (Figure S3a shows that the (HNO3)·NO3– signals decrease when I2 is added). Also, a higher pressure or a slower flow in the ion source flow tube promotes the (HNO3)·NO3– ions versus NO3–, and in the ion–molecule reaction region, clustering of ions and molecules is favored over dissociation. A longer residence time may, on the other hand, enhance reactive and diffusive loss of ions. When the two branches of the experiment are at the same pressure, all these effects overlap in the observed mass spectra. Thus, the observed changes in the (HNO3)·IO3– or IO·NO3– signals may not only result from varying IO but also from varying (HNO3)·NO3–. This is illustrated in Figure S2, which shows mass spectra for two experiments where IO form under the same conditions but the flow through the ion source differs by a factor of two. A slower flow enhances the signals of the heavier ions, reduces the signals of the (HNO3)·IO3– ions, and also changes the signal ratios between the latter.

Keeping the IO branch of the flow tube behind a pin-holed wall (Figure ) has several advantages, which include the ability of changing pressure in the IO formation region without affecting pressure in the ion source and avoiding illumination of the ion–molecule reaction volume. Moreover, in the higher-pressure experiments (26 Torr), IO were mostly generated by gas-phase photochemistry (e.g., a five to ten times more photolytic HNO3·IO3– signal than from the dark reaction, compare Figure a,c) owing to enhanced I2 photolysis (∼30%) resulting from the longer residence time (1.7 s) and to reduced wall interaction as a result of slower molecular diffusion at a higher pressure. In the 26 Torr experiments, the flows through the iodine trap and the ozone generator were reduced to maintain a similar concentration of IO as in the 3 Torr experiments to avoid build-up of particles that could block the pinhole.[18] The ion–molecule reaction products in both experiments are the same (same peaks in Figures a,c), but the signals of the iodine-containing anions are smaller relative to the nitrate core ion signals in the 26 Torr experiments (signals are shown normalized to the NO3– signal in Figure ) for similar contact time in the ion–molecule reaction region, suggesting a different distribution of products in the IO flow tube. Regarding the photolytic signals in the higher-pressure experiments, 30% photolysis of I2 results in 7% less background IONO2·NO3– in the experiments with light and hence the negative peak in the difference spectrum at m/z = 251 (Figure d).

Figure 3

Mass spectra of iodine oxide ions, where iodine oxides were generated at 3 Torr (a,b) or at 26 Torr (c,d). Panels a and c show the raw spectra obtained in the dark (black lines) and by irradiating the tube axially with white light (red lines). Panels b and d show the photolytic signal, that is, the difference between the signals recorded with and without light.

Wet Experiments

Similar to our results above, the first observation in a laboratory setting of IO3– by nitrate CIMS analysis of an I2 + O3 mixture took place without actively adding water to the flow tube.[23] Interpretation of IO3– as HIO3 requires a source of hydrogen atoms. Hence, in the absence of HO, the formation of HIO3 was explained by Sipilä et al.[23] as the result of a very fast reaction between I2, O3, and water degassed from the walls of the flow tube ([H2O] < 8 × 1015 molecule cm–3). Subsequent experiments were conducted where increasing water vapor concentrations up to 4 × 1016 molecule cm–3 were added to the flow tube. This resulted in a factor of two increase of the raw (not charge-normalized) IO3– signal, which was seen as a confirmation of the need of water to form HIO3.[23]

In order to investigate the effect of water in our system, the IO carrier gas was humidified by passing it through a bubbler containing deionized water, at the same pressure as the flow tube (i.e. the bubbler is downstream of the carrier gas flow controller). The water vapor concentration in the IO branch at 3 Torr is estimated from the pressure variation to be ∼8 × 1015 molecule cm–3. The minimum water concentration in these experiments, where water was turned on and off several times, is estimated from the ratios of the H2O·NO3– ion cluster signal, and found to be 1 order of magnitude higher than in the “dry” experiments. The estimated concentration of water vapor at 26 Torr is ∼2.5 × 1017 molecule cm–3, corresponding to the atmospheric water vapor concentration for RH = 33% at 760 Torr and 25 °C. Addition of water to the ion–molecule reaction volume ([H2O] ∼ 1 × 1015 molecule cm–3 after dilution by the larger flow that passes through the ion source) results in a general increase of the nitrate core ion signals, as shown in Figure S1. The NO3– and HNO3·NO3– signals increase by a factor of ∼2. A possible explanation of this observation is that water slows down anion–cation neutralization by forming clusters with negative and positive ions (Figure S1a,b). Another possibility is that water deposition passivates the inner surfaces in the ion–molecule reaction volume, reducing the wall loss of anions.

Mass spectra obtained with and without water at 3 and 26 Torr are shown in Figure S6. The contribution of the dark reaction has been removed from these spectra, and only photolytic signals are shown. Addition of water enhances the iodate core ion signals by a factor of ∼3 in both experiments, while the I2O5·NO3– and O2IONO2·NO3– signals reduce upon addition of water. Figure shows that scaling the IO3– and HNO3·IO3– signals with measured NO3– and HNO3·NO3– enhancement factors in the presence of water (equivalent to the usual normalization to the available charge performed in CIMS measurements) significantly reduces the difference between the dry and wet observations. After correction, the iodate core ion signals in the presence of water are still up to two times higher, both in the 3 Torr and the 26 Torr experiments. This may be an indication of formation of HIO3 followed by R.

Figure 4

Water dependence of nitrate core anions and selected iodine oxide anions for two experiments at 3 and 26 Torr. Panels a and b show, respectively, the ratios between the NO3– and HNO3·NO3– signals (i.e., the integrated area under a mass peak) measured with (shaded blue) and without water. Panels c and d show the IO3– and HNO3·IO3– photolytic signals obtained from the raw spectra (black squares) and corrected with the nitrate core ion ratios in panels a and b, respectively. Panels e and f show the same as panels c and d for I2O3·NO3– and I2O3·HNO3·NO3–.

To complete this picture, we include in Figure the corresponding I2O3·NO3– and I2O3·HNO3·NO3– measurements, which after correction show no difference with the values under dry conditions. Similarly, the I2O2·NO3–, and I2O4·HNO3·NO3– measurements in the presence of water remain close to the dry values after applying the corresponding scaling factor (Figure ). This means that water does not remove IO (y = 2–4). The only I2O–related signal that is significantly reduced by water systematically is that of the I2O5·NO3– anion (Figure d), whose parent neutral is I2O5. The decrease of the I2O5·NO3– signal and the increase of the IO3– signal upon addition of water suggest that the loss of I2O5 results in the formation of HIO3. This is supported by the lack of increase of the iodate core ion signals in the absence of light (Figure S7a,c), where I2O5 does not form (Figure S7f), but other I2O do.

Figure 5

Water dependence of IO·(HNO3)·NO3– photolytic signals for two experiments at 3 and 26 Torr: I2O2·NO3– (panel a), I2O3·NO3– (panel b), I2O4·HNO3·NO3– (panel c), and I2O5·NO3– (panel d). Black squares: signals obtained by integrating the corresponding mass peaks. Red squares: signals corrected with the NO3– ratios with/without water shown in Figure a,b.

There are other important observations in our experiments regarding the molecular clusters that have been proposed as the initial steps in the oxyacid-driven IOP nucleation mechanism. With light and in the presence of water, we observe a small peak at m/z = 351 that could be attributed to the HIO3 dimer.[4,23] There are also other peaks that appear with light and added water that may be related to clusters formed by addition of HIO3 to iodine oxides (m/z = 334, m/z = 477, and m/z = 494, see Table ). In particular, the peak at m/z = 398 (HIO2·HIO3·NO3– or I2O4·H2O·NO3–) only appears in the presence of water.

Discussion

Interpretation of Mass Spectra Obtained without Added Water Vapor

Some of the masses listed in Table (m/z = 205, 221, 348, 364, and 380) result from clustering between well-known iodine oxides[18,46,47] and nitrate ions in the ion–molecule reaction volume and have been reported in previous CIMS studies[4,11]

The observation of I2O5 in the form of I2O5·NO3– is somewhat surprising since gas-phase I2O5 was not unambiguously observed by PIMS under similar conditions.[18,22] This mass is observed both at 3 and 26 Torr only if the mixture is irradiated (Figure S7) and is not formed from the dark I2 + O3 reaction as is the case for the other I2O, which indicates that I2O5 is a gas-phase photolysis product of a higher-order iodine oxide such as I3O7.[48] We note that I3O (n = 5–7) have been previously observed both by PIMS as I3O+[18,22] and by nitrate CIMS as I3O·NO3– (m/z > 500 amu).[23]

Three prominent iodine-containing ions are IO3– (m/z = 175), HNO3·IO3– (m/z = 238), and (HNO3)2·IO3– (m/z = 301). These masses have been previously observed with nitrate CIMS instruments[4,23] and have been interpreted as products of ion–molecule reactions between HIO3 and (HNO3)·NO3– (n = 0–2) reaction in the instrument inlet. Any OH generated by UV photolysis of O3 in the presence of water in our experiments is scavenged by I2 and therefore cannot generate HIO3 via reaction . This leaves water as the only other possible reagent. For water concentrations as low as those in the “dry” experiments at 3 Torr ([H2O] < 2 × 1013 cm3) and a reaction time of 130 ms in the IO flow tube, the rate constant of any hypothetical gas-phase mechanism forming HIO3 from water plus I (+O3), IO, OIO, or I2O (y = 2–4) where the reaction with water is rate limiting would have an effective rate constant of k ≥ 4 × 10–13 cm3 molecule–1 s–1. This is clearly at odds with the upper limits to the effective rate constants of reactions between atomic iodine (+O3) or iodine oxides and water, forming HIO3, which were found to be lower than ∼10–19 cm3 molecule–1 s–1.[22] HIO3 could also be formed by hydrolysis of IO on the surfaces of the flow tube, although no HIO3 from the gas phase or surface chemistry was observed by PIMS in the same system. Furthermore, Born–Oppenheimer molecular dynamics simulations indicate that I2O reactions at the air–water interface do not take place.[22] Therefore, it is likely that masses 175, 238, and 301 result from ion–molecule reactions between iodine oxides, which are detected in our system both by PIMS and CIMS, and nitrate core ions:

Ab initio enthalpies of reactions I2O + NO3– were reported in our previous publication (Supporting Information of Gómez Martín et al.,[22]). These were calculated at B3LYP/6-311+G(2d,p) level of theory with the iodine basis set mentioned above[40] and validated with evaluated thermochemical data. Higher level of theory calculations (CCSD(T)/aug-cc-pVTZ + LANL2DZ//M06-2X/aug-cc-pVDZ + LANL2DZ) confirmed that the reaction , facilitated by the formation of a IO3–IONO2 halogen-bonded adduct, is exothermic and barrierless.[33] This is not too surprising, considering that halogen bonding has been found to play an important role in the iodine CIMS when used for detecting HNO3.[34] Here, we have revisited our previous calculations[22] and extended them to reactions , R3.2, R3.3, and R7 using the larger aug-cc-pVQZ basis set (see Methods). We have confirmed at this level of theory that no barriers exist in the PESs of reactions R3.1 and R7 (Figures S9 and S10, respectively). The PES of reaction in Figure shows that a similar mechanism to R3.1 operates when the nitrate core ion is involved, followed by transfer of the HNO3 to the IO3– end of the adduct over a submerged barrier. The geometries and molecular parameters of the species involved in the PESs of R3.1, R3.2, and R7 are provided in the Supporting Information.

Figure 6

Potential energy for reaction at the B3LYP/aug-cc-pVQZ level of theory (see Table S3 for further details).

It is also plausible that HNO3 adds to iodate core ions to form (HNO3)·IO3– with an increasing number of HNO3 ligands

In fact, the experiments in Figure S2 show that reducing the residence time in the ion–molecule reaction region enhances NO3– relative to (HNO3)·NO3– (n = 1, 2), while the (HNO3)·IO3– (n = 1, 2) ions increase, which suggest that R8 is also a source of (HNO3)·IO3– (n = 2, 3) in our system, besides R3.2 and R3.3.

From the discussion above, it follows that masses 175 (IO3–), 238 (HNO3·IO3–), and 364 (I2O3·NO3–) may be sampling the same parent molecule. Figure S5 shows that these signals change in the same manner when the ozone concentration is doubled, which suggests that they indeed have common parent neutral molecules. Also, HNO3·IO3– and I2O3·NO3– are higher relative to IO3– when the pressure is increased in the ion–molecule reaction volume, which is a result of enhanced ion–molecule clustering. Figure S4 indicates that the ratio of the IO3– signal to the I2O3·NO3– signal remains constant when changing the residence time of the gas in the IO flow tube. These observations rule out the identification of mass 175 as a product of a reaction of IO with water deposited on the reactor walls (i.e., HIO3).

The peaks at m/z = 222 and m/z = 285, which are minor in our experiments, were interpreted in previous CIMS work as resulting from ion–molecule reactions between HIO2 and (HNO3)·NO3– (n = 0–1) in the instrument inlet and as a proof of the presence of HIO2 in the sampled air. However, the m/z = 285 peak (HNO3·IO2–) may also originate from

Reaction is essentially thermoneutral at the B3LYP/aug-cc-pVQZ level, with an accuracy of ±20 kJ mol–1. Higher level calculations are needed to determine whether this reaction is actually exothermic or not.

An important observation is the presence in the mass spectra of peaks at m/z = 251, m/z = 267, and m/z = 283, which have also been observed previously by nitrate CIMS,[4,11,23] although no interpretation was given to them. These masses can be identified as the ion clusters IONO2·NO3–, OIONO2·NO3–, and O2IONO2·NO3–. We have seen that the m/z = 251 signal appears simply by adding I2 to the ion–molecule reaction zone, in line with the CIMS observations of Ganske et al.[34] However, this signal also tracks the iodate core ion signals (Figure S5), which means that part of it is associated with the neutral chemistry in the IO flow tube. In fact, iodine nitrate, IONO2, is a product of reaction , OIONO2 is a product of reaction , and O2IONO2 is a product of an analogous reaction of I2O5 and NO3–. Other nitrate core ion clusters of IONO2 and OIONO2 are also observed at m/z = 314 and m/z = 330, respectively. The interpretation of m/z = 251 as evidence of IONO2 not only brings closure to the proposed interpretation of the (HNO3)·IO3– CIMS signals in the dry experiments but also implies that it may be possible to use this signal to monitor IONO2 in the field.

The flow tube employed in this work is not suitable for studying the kinetics of IO formation (to that end, the nitrate core ions should be in excess over iodine oxides). However, it can be seen that a longer residence time in the lower-pressure experiments enhances all the iodine-containing ions, indicating a general growth stage of the parent molecules (Figure S4). By contrast, in the higher-pressure, longer residence time experiments (Figure ), the concentration of the parent higher-order oxides is higher relative to IO and OIO, which indicates higher concentrations of iodine oxides and faster second-order chemistry.

Interpretation of Mass Spectra Obtained with Added Water Vapor: the Source of HIO3

Addition of water in the presence of light results in:

the removal of ion signals associated with I2O5 (as previously observed by Sipilä et al.[23])

the increase in the iodate core ion signals (factor of ∼2 higher for the highest water concentration relative to the “dry” experiments) and

the appearance of other ions that can be assigned to neutral IO. HIO3 adducts (also observed in previous nitrate CIMS studies,[4,11,23] see Table )

These changes do not occur in the dark, where I2O (y = 2–4) but no I2O5 are formed. In addition, OIO and I2O (y = 2–4) are not removed by water. Hence, these observations suggest that I2O5 reacts with water to generate HIO3. The reaction between I2O5 and H2O is precluded by a large barrier in the PES,[30] but recent ab initio calculations at the CCSD(T)//M06-2X/aug-ccpVTZ-PP + ECP28 level[32] indicate that hydrolysis of I2O5 by the water dimer is feasible

Reaction proceeds over a submerged barrier (−15.1 kJ mol–1). The complete process likely involves dissociation of the (HIO3)2·H2O complex, considering the exothermicity of reaction where we have used the bond energy of the HIO3·H2O complex[49] computed at a similar level of theory than that used for reaction .

By contrast, our equivalent CCSD(T) calculations show that a second water molecule does not sufficiently reduce the height of the barrier of I2O3 + H2O PES (32 kJ mol–1 for one water molecule[22] and 16 kJ mol–1 for the water dimer). This barrier is similar at lower levels of theory employed. Regarding I2O4 + (H2O)2, our B3LYP/6-311+G(2d,p) calculations indicate that a complex bound by 48 kJ mol–1 forms first and then rearranges over a submerged barrier (−44 kJ mol–1) to give

Dissociation of H2I2O5 to I2O4·H2O + H2O is endothermic by 89 kJ mol–1 and requires some rearrangement, so a barrier may be expected as well. This suggests that the peak at m/z = 398 corresponds in fact to H2I2O5·NO3–. The I2O4·H2O adduct formed directly from hydration of I2O4 is bound by 53 kJ mol–1[50] and could also contribute to the signal at m/z = 398 in the high [H2O] experiments. Note however that the available I2O4 ion tracer (m/z = 443, I2O4·HNO3·NO3–) does not disappear by adding water (Figure c), which indicates that R12 is much slower than (R10 and R11).

The peak at m/z = 398 has also been interpreted as HIO2·HIO3·NO3– and considered as evidence of the first HIO2–HIO3 neutral cluster.[4] The proposed HIO2 ion tracers (m/z = 222 and m/z = 285) appear in the absence of water, suggesting that they are formed by R6 or other reactions involving IO. Their dependence on water is not completely consistent across different measurements. The signal at m/z = 285 (Figure S8c) generally increases when water is added. Reaction would be a possible source of HIO2 in the presence of water. Hence, we cannot rule out that the peak at m/z = 398 is also representative of HIO2·HIO3·NO3–. We note nevertheless that the I2O4 concentration is expected to be significantly larger than that of HIO2, and hence it is more likely to contribute to clustering. Larger clusters with m/z > 500 amu (outside our instrumental range) reported in the CLOUD experiments[4] can also be explained by addition of I2O4 to pre-existing clusters (see Table ). It has been argued that the concentration of I2O4 in the CLOUD experiments was only 1% of that of HIO3 based on the comparison of anion signals. However, it is likely that the I2O4·NO3– and I2O4·HNO3·NO3– ion signals underestimate the I2O4 concentration and that part of I2O4 is actually observed as IO3–, as discussed above.

Table 2

Updated Mechanism of Iodine Gas-to-particle Conversion

chemistry	references and notes
I + O3 → IO + O2	evaluated kinetic and photochemical data for modeling of tropospheric iodine chemistry.[51]
IO + IO → I + OIO → I2O2
IO + OIO ↔ I2O3
OIO + OIO ↔ I2O4
I2O2 + OIO → I2O3 + IO	the aggregation and dissociation rate constants of I2Oy + I2Oz reactions were calculated with the master equation solver MESMER using CCSD(T)//MP2/aug-cc-pVTZ energies, but the complete PES of these reactions was not explored.[50] PIMS observations indicate that I3Oy (y = 4–7) molecules form rather than adducts with four iodine atoms.[18,22] I2O3 was found to be very strongly bound and chemically stable to form weakly bound aggregates; hence, its fate remains unclear. The rate constants of some reactions involving I2Oy (y = 2–4) generating I3Oy (y = 4–7) were estimated by numerical modeling of IxOy time traces obtained in flow tube experiments with PIMS detection.[22] These semiquantitative estimates obtained from a tentative mechanism show that the rate constants of IxOy aggregation reactions are close to the collision number. Analogous reactions of I2O5, not considered in previous work because this molecule was not detected, are now included in this table.
I2O2 + I2O2 → I2O3 + I2O
I2O4 + OIO → I3O6
I2O4 + I2O4 → I3O6 + OIO → I3O7 + IO
I2O4 + I2O5 → I3O7 + OIO
I3O6 + I2O3 ↔ I5O9
I3O6 + I2O4 ↔ I5O10
I3O6 + I2O5 ↔ I5O11
I3O7 + I2O3 ↔ I5O10
I3O7 + I2O4 ↔ I5O11
I3O7 + I2O5 ↔ I5O12
I3O7 + I3O7 → I5O12 + OIO
I2O4 + (H2O)2 → H2I2O5 + H2O	H2I5O2 has been observed in previous work using nitrate CIMS, and it is also observed in the present work.
→ HIO3–H2O + HIO2	possible source of HIO2.
I2O5 + (H2O)2 → HIO3 + HIO3·H2O	source of HIO3. The PES of this reaction has been reported.[32,49]
HIO3 + HIO2 ↔ H2I2O5	theoretical estimates of the forward and reverse rate constants of the HIO3 + HIO3 and of HIO3 + I2O4 aggregation reactions have been reported.[22] The I2Oy·HIO3 adducts have been observed in the CLOUD chamber experiments using nitrate CIMS. They are also observed in the present work (m/z < 500 amu).
HIO3 + HIO3 ↔ (HIO3)2
HIO3 + OIO ↔ OIO·HIO3
HIO3 + I2O2 ↔ I2O2·HIO3
HIO3 + I2O3 ↔ I2O3·HIO3
HIO3 + I2O4 ↔ I2O4·HIO3
HIO3 + I2O5 ↔ I2O5·HIO3
I2O4 + H2O·HIO3 ↔ I2O4·H2O·HIO3	the (I2O4)n·H2O·(HIO3)m adducts have been observed in the CLOUD chamber experiments using nitrate CIMS[4] as anions with m/z > 500 amu (outside the mass range in the present work). The nucleation mechanism proceeds by addition of HIO3 and I2O4 to pre-existing molecular clusters.
H2I2O5 + HIO3 ↔ H2I2O5·HIO3
H2I2O5 + I2O4 ↔ H2I2O5·I2O4
H2I2O5 + H2I2O5 ↔ (H2I2O5)2
I2O4 + I2O4 ↔ (I2O4)2
I2O4·HIO3+ I2O4 ↔ (I2O4)2·HIO3
H2I2O5·I2O4 + HIO3 ↔ H2I2O5·I2O4·HIO3
H2I2O5·HIO3 + I2O4 ↔ H2I2O5·I2O4·HIO3
H2I2O5·HIO3 + HIO3 ↔ H2I2O5·(HIO3)2

Comparison to PIMS Laboratory Experiments

In our previous work using PIMS[22] with the same IO source, we did not detect either I2O5 or HIO3 in the gas phase, and we did not observe cations that could be attributed to IO·HIO3 adducts. We argued that if there was a competition between clustering reactions of iodine oxides forming higher-order IO and a fast reaction between iodine or iodine oxides and water-forming HIO3, there would have been a dramatic reduction in the IO-containing ions and a population of oxoacid clusters would have emerged. However, we observed only a limited reduction in the IO signals and no reaction products when water was added. Water changed the composition of the particles to HIO3.[22]

Our present nitrate CIMS experiments indicate that this competition likely occurs between I2O5–IO clustering and slow hydrolysis by the water dimer.[32] The concentration of IO in our flow tube is high (∼1010 to 1012 cm–3) compared to atmospheric conditions (107 to 109 cm–3). For low water concentrations, I2O5 and HIO3 are mainly removed by clustering with IO, and low concentrations of HIO3 and I2O·HIO3 clusters exist, which may be too low to be detectable by PIMS in experiments with the same time scale as in the present ones. By contrast, in environmental chamber studies under MBL conditions, it is likely that even a slow water reaction with I2O5 dominates over clustering with IO, such that I2O·HIO3 and HIO3 molecular clusters drive particle formation.

An important observation of the PIMS experiments is that particle formation is more intense when water is not added. This implies that IO clusters form particles faster than I2O·HIO3 and HIO3 clusters. Since the rate of formation of HIO3 likely depends on [H2O][2] this may have important atmospheric consequences for IOP formation in different environments.

Comparison to Bromide CIMS Environmental Chamber Measurements

The signals observed by nitrate CIMS at m/z = 175 (IO3–), m/z = 238 (HNO3·IO3–), and m/z = 301 ((HNO3)2·IO3–) appear for very low water concentrations where iodine oxides are formed but not HIO3. These masses are also generated in the dark when I2O5 (the most likely precursor of HIO3) is not made. Water vapor does not remove I2O (y = 2, 3, and 4), but it does remove I2O5. At the same time, atmospheric water concentrations result in an increase in the m/z = 175 and m/z = 238 signals by a factor of 2 compared to dry conditions. Hence, the IO3– core anions observed by CIMS are likely both products of the reaction of NO3– with I2O (y = 2, 3, and 4) and with HIO3 in the instrument inlet and can be interpreted as the sum of iodine oxides I2O (y = 2, 3, and 4) and HIO3 present in the sampled air. A similar argument may apply to the IO3– signal observed with a bromide CIMS in the CLOUD experiments[28] since reactions between bromide ions and I2O are also exothermic, for example[22]

In contrast, the HIO3·Br– signal observed in the same experiments cannot result frombecause this reaction is precluded by a barrier of 20 kJ mol–1, according to our quantum calculations at the B3LYP/aug-cc-pVQZ level. Hence the HIO3·Br– anion appears to be a genuine HIO3 tracer.

Atmospheric Implications

The IOP formation mechanism proposed in our previous work[22] can now be updated by adding the source of I2O5 and HIO3 and the two molecular cluster formation pathways (Table ). Further experimental and theoretical work is required to investigate the photolysis products of higher-order iodine oxides, the specific fate of I2O2 and I2O3, and the rate constants of the IO, HIO3, and IO·HIO3 clustering reactions.

CIMS observations should help in better constraining atmospheric iodine models since the most relevant species (IO, OIO, IO, and HIO3) can be detected with this technique with high sensitivity. Laboratory and chamber experiments using spectroscopic instrumentation should be conducted in order to calibrate the CIMS signals of these key species. Comparison between bromide and nitrate CIMS observations of iodate core ions may help in quantifying the fraction of the signal of these ions that can be attributed to I2O and HIO3 under different atmospherically relevant conditions. Our 26 Torr experiments, where almost all I2O5 is depleted when atmospherically relevant water concentrations are added, indicate that ∼50% of the IO3– and HIO3– signals observed by CIMS correspond to I2O (y = 2–4).

The observation of the signal at mass 251 in our experiments is also particularly relevant for the CIMS observations in the context of atmospheric chemistry. We have interpreted this signal as IONO2·NO3–, where IONO2 is a product of the reaction between I2O3 and NO3–. Formed in the atmosphere through the recombination of IO and NO2, IONO2 is also a key iodine reservoir and a carrier of iodine toward the aerosol phase in polluted and semi-polluted regions. To our knowledge, no measurements of this compound have been reported to date, and in fact, no in situ technique has been developed to detect it, in contrast to, for example, ClONO2.[52] Baccarini et al.[11] observed a strong signal at m/z = 251 using nitrate CIMS, which was attributed to the O6N2I– anion but not explicitly to IONO2·NO3–. We propose that this was possibly the first measurement of IONO2 reported in the literature. Further experiments should determine the relative contribution to that signal of ambient IONO2 and IONO2 formed in the CIMS inlet from I2O3 + NO3–.

Our previous results using PIMS indicated that clustering of iodine oxides leads to particle formation. Water is not required to form nucleating molecules, which has implications for where in the atmosphere IOP formation can take place. Since IOP formation is not limited by water abundance, it can occur in the polar MBL, as observed,[11] and perhaps also in the upper troposphere. Most other new particle formation processes (e.g., sulfuric acid, ammonia) depend directly or indirectly on the presence of water. A particle mechanism that does not depend on water may significantly contribute to, even dominate, total new particle formation in water-limited regions, even with small amounts of iodine. This may be the reason why iodine is the dominant nucleating species in the high Arctic.[11] Note that, in addition, water-limited regions will generally be associated with lower pre-existing aerosol loadings, thereby increasing the survival chance of any newly formed iodine particle. A recent experimental study indicates that the transition between the dry and humid IOP formation mechanisms occurs at around 20% RH.[53]

Conclusions

Our flow tube experiments reveal that the iodate core ion signals measured by nitrate CIMS are contributed both by I2O and HIO3 neutral molecules. They also indicate a plausible photolytic and water-dependent source of HIO3, which is consistent with the coexistence of iodine oxides and oxoacids in nitrate CIMS spectra obtained under MBL conditions, as well as with PIMS laboratory observations with typically higher iodine oxide concentrations. In addition, they show that the formation of HIO3 under high water and low iodine concentrations leads to the formation of I2O·HIO3 clusters, which are the likely precursors of iodine particles in the MBL. Under dry conditions, IO clusters lead to different, faster nucleation. These results fill the gaps in the mechanism that connects inorganic and organic iodine emissions and IOPs, which greatly facilitates the implementation of iodine chemistry and iodine-driven nucleation in atmospheric models. This should eventually enable the radiative forcing of IOPs to be computed for the first time.