Montreal is considered a relatively lowly air-contaminated city, with an average annual PM₁₀ (particles with an aerodynamical diameter <10 µm) concentration of 16 μg m⁻³ (World Health Organization 2016), lower than the 20 μg m⁻³ guidelines set by the WHO. Still, the city exhibits local variations at several monitoring stations and records discrete episodes of air pollutions with concentrations exceeding the mean 50 μg m⁻³ 24-h guidelines (Boulet and Melancon, 2012; Boulet and Melancon, 2013). PM₁₀ aerosols were sampled over a 1-year period in 2015 by the RSQA (Réseau de Surveillance de la Qualité de l’Air) in Montreal (45°N 73°W, Canada) and its vicinity, in collaboration with the Ville de Montréal. Two monitoring stations (13 and 98) disseminated onto Montreal Island were selected for their specific environmental conditions (Figure 1). Station 13, referred as “Drummond,” is located downtown and represents the urban background, whereas Station 98 referred as “Sainte-Anne de Bellevue” is located at the westernmost end of the island, in a semi-rural environment under dominant southwest-northeast blowing winds, and thus represents a station less impacted by local anthropogenic atmospheric emissions (Boulet and Melancon, 2012). Station 13 is thus expected to collect aerosols of both local and more distant sources, i.e., emitted locally and brought to Montreal, compared to station 98 where aerosols are expected to mostly come from outside the city. Our sampling strategy was thus designed to allow comparing a station mostly affected by local anthropogenic emissions to a station where aerosols are expected to have a remote origin (i.e., aerosols transported to Montreal). At this second station, aerosols are expected to at least partially derive from atmospheric secondary processes. In parallel, major pollutant gases, including ozone (O₃), sulfur dioxide (SO₂), nitrogen dioxide (NO₂), as well as PM_2.5 hourly concentrations were continuously acquired from the Réseau de Surveillance de la Qualité de l’Air (RSQA) for both stations (http://ville.montreal.qc.ca/, last access: 02 January 2020). We calculated daily mean concentrations for each compound in order to fit with the 24 h sampling period of the aerosols.

PM₁₀ samples and blank samples were also weekly collected by the RSQA on pre-combusted quartz filters using a high volume PM₁₀ size selective inlet, with an average flow of 1.13 m³ min⁻¹ for a period of 24 h (pumped air volume of ∼1627 m³ per filter). Within the sample set, four samples were selected and analyzed each month during an entire year, with two samples collected at station 13 and two others at station 98. Samples selected at each station were collected on the same day during the first 2 weeks of each month. Blank filters were analyzed randomly for each site, and yielded Hg concentrations under the analytical detection limit (<0.1 ppb).

Concentrations of selected soluble inorganic species (Na²⁺, K⁺, Ca²⁺, Mg²⁺, NO₃ ⁻, SO₄ ²⁻, Cl⁻) were measured using a Dionex^®ICS-90 Ion Chromatography system, after extraction from a 3 cm × 3 cm filter piece with 30 ml of Milli-Q water, following the method described by Paris et al. (2010). Detection limits for these ion species were usually in the order of 5 µg L⁻¹, i.e., 0.1 ng m⁻³ considering our sampling and extraction protocols. Hg concentrations were measured by cold vapor atomic fluorescence spectroscopy (CVAFS, Tekran 2500) after reducing an aliquot of the pre-concentrated Hg^II (i.e., acid trap solution) into Hg⁰ using SnCl₂. The Hg⁰ was then collected on a gold-coated bead trap and analyzed by CVAFS (Huang et al., 2015).

Aerosol filters were combusted using the dual-stage protocol described in Huang et al. (2015) to concentrate Hg. Briefly, each filter was introduced into a quartz tube that then underwent two successive combustions first at 950°C, followed by a combustion at 1000°C for a total time of 3.5 h to decompose the Hg^II present under the form of Hg_p into vapor Hg⁰ (Sun et al., 2013). The combustion products, Hg⁰ (and other compounds if any), were purged using Hg-free O₂, and bubbled through a 5 ml HNO₃-HCl-H₂O mixture (2:4:9) acid trap (Sun et al., 2013; Huang et al., 2015). The generated solution was then transferred into a pre-cleaned glass bottle. 50 µl of 0.2 M BrCl were then added to fully convert Hg⁰ into Hg^II.

Hg^II was then converted back into Hg⁰ by reacting with SnCl₂ and injected into a multi-collector inductively coupled plasma mass spectrometer (MC-ICPMS; Neptune plus) simultaneously with Tl, which was used as an internal standard to correct for the instrumental mass bias (Blum et al., 2007; Yin et al., 2016). A high concentration of Tl (20 ppb) was injected with each sample to prevent the formation of Hg hybrids during the analysis (Yin et al., 2016). The faraday cups were positioned to collect ¹⁹⁸Hg (L3), ¹⁹⁹Hg (L2), ²⁰⁰Hg (L1), ²⁰¹Hg (C), ²⁰²Hg (H1), ²⁰³Tl (H2), and ²⁰⁵Tl (H3). Hg multi-isotopic compositions were then determined by standard bracketing using the Hg NIST 3133 international standard. The Hg isotopic compositions are expressed as (Blum et al., 2007): δ x Hg = [ ( H x g / H 198 g ) sample ( H x g / H 198 g ) std − 1 ] × 1000 (1) Where x = 199, 200, 201, 202, 204, and “std” is the NIST SRM 3133 Hg international standard. In general, the Hg multiple isotopic ratios are related to each other according to their mass, called the mass-dependent fractionation (MDF), which is expressed as follows (Young et al., 2002; Dauphas et al., 2016): α y = ( α 202 ) yβ (2) Where ^yα is either ¹⁹⁹α, ²⁰⁰α, ²⁰¹α or ²⁰⁴α and ^yβ is either ¹⁹⁹β, ²⁰⁰β, ²⁰¹β or ²⁰⁴β. The ^3yβ -exponent describes the relative fractionation of ^yHg/¹⁹⁸Hg compared to ²⁰²Hg/¹⁹⁸Hg where ¹⁹⁹β, ²⁰⁰β, ²⁰¹β and ²⁰⁴β-values are respectively 0.252, 0.502, 0.752 and 1.493 under equilibrium (Blum et al., 2007). The α-notation corresponds to the different isotopic fractionation factors between ¹⁹⁸Hg and any of the other isotopes. For the oxidation of Hg⁰ into Hg^II, α is defined as follows (expressed here for the oxidation of Hg⁰ into Hg^II): α Hg II − Hg 0 x = ( H x g H 198 g ) Hg II ( H x g H 198 g ) Hg 0 = δ H x g Hg II 1000 + 1 δ H x g Hg 0 1000 + 1

Any deviation of the Hg isotopic ratios from MDF is defined as the mass-independent fractionation (MIF) and is represented by the “capital delta” notation (Δ^×Hg, in ‰) defined following (Farquhar and Wing, 2003; Blum et al., 2007): Δ 199 Hg = 1000 × [ ( l n ( { δ H 199 g 1000 } + 1 ) ) − 0.2520 ( l n ( { δ H 202 g 1000 } + 1 ) ) ] (3) Δ 200 Hg = 1000 × [ ( l n ( { δ H 200 g 1000 } + 1 ) ) − 0.5024 ( l n ( { δ H 202 g 1000 } + 1 ) ) ] (4) Δ 201 Hg = 1000 × [ ( l n ( { δ H 200 g 1000 } + 1 ) ) − 0.7520 ( l n ( { δ H 202 g 1000 } + 1 ) ) ] (5)

The NIST 3177 standard was also regularly analyzed with concentrations matching those of the aerosol samples (i.e., 2 ppb) to control the instrument stability and to guarantee the measurement quality (Geng et al., 2018). Repeated analyses (n = 22) of the NIST 3177 standard yielded δ²⁰²Hg = −0.52 ± 0.03‰ (2σ), Δ¹⁹⁹Hg = −0.02 ± 0.05‰ (2σ), Δ²⁰⁰ Hg = 0.01 ± 0.03‰ (2σ), Δ²⁰¹ Hg = −0.01 ± 0.02‰ (2σ) relative to NIST 3133, consistent with previous reported values (Wang et al., 2015; Chen et al., 2016; Sun et al., 2016; Yuan et al., 2018; Fu et al., 2019; Zhang et al., 2020). A Chinese loamy sand CRM024, used as a second certified reference material, with a Hg concentration of 0.71 ppm was also analyzed (n = 8). The result yielded a recovery of 104 ± 7% after pre-concentration by dual-stage combustion, and δ²⁰²Hg = −1.43 ± 0.08‰ (2σ), Δ¹⁹⁹Hg = 0.03 ± 0.02‰ (2σ), Δ²⁰⁰ Hg = −0.00 ± 0.02‰ (2σ), Δ²⁰¹ Hg = 0.00 ± 0.01‰ (2σ), in good agreement with the Hg isotopic compositions reported by Huang et al. (2015).

To investigate potential relationships between air masses and the aerosol Hg multi-isotopic compositions, we modelled daily 72 h back-trajectories at a 10 m agl height using HYSPLIT (Hybrid Single Particles Lagrangien Integrated Trajectory) for each sample. The model used NCEP-NCAR reanalysis data fields, using a 2.5-degree latitude-longitude global grid with a time resolution of 6 h obtained from the Air Resources Laboratory (ARL). Back-trajectories were ultimately incorporated into a map generated by GMT (Generic mapping tools).

The significance of each correlation was calculated using a Spearman Correlation, as this non-parametrical test does not carry any assumption about the data distribution and is not sensitive to outliers.

The both low odd and even-MIF, close to 0‰, coupled to the absence of Hg isotopic seasonality at station 13, suggest that the factor controlling the Hg multi-isotopic compositions is constant and perennial during 2015. This contrasts with the seasonality in the Hg isotopic compositions previously reported in urban areas that has been demonstrated to reflect varying contributions from anthropogenic emission sources (Huang et al., 2016; Huang et al., 2018). Still, considering the geographical location of station 13, we hypothesize that the corresponding Hg isotopic compositions may be largely impacted by anthropogenic emissions. In fact, the low Hg-MIF, in particular the Δ¹⁹⁹Hg close to 0‰, and the negative δ²⁰²Hg measured at this station are characterized by values that are consistent and within the range of variations reported for anthropogenic emissions (Biswas et al., 2008, Sun et al., 2014, Yin et al., 2014, Wang et al., 2015, Das et al., 2016, Huang et al., 2016, Zheng et al., 2016, Wang T. et al., 2017; Wang T. et al., 2017, Huang et al., 2018, Yuan et al., 2018, Zhang et al., 2020). Moreover, we observe that the Δ¹⁹⁹Hg/Δ²⁰¹Hg ratio, a proxy for identifying processes triggering Hg-MIF (Bergquist et al., 2007; Zheng et al., 2009; Sherman et al., 2010; Sun et al., 2016) is characterized by a yearly value close to 0.95 at station 13. A Δ¹⁹⁹Hg/Δ²⁰¹Hg ratio ranging from 1 to 1.3 is generally explained as the result of photoreduction (Bergquist et al., 2007; Zheng et al., 2009). As seen, Supplementary Figure S1 reports a compilation of primary aerosols emitted by various emission sources and shows that anthropogenic PBM is characterized by a Δ¹⁹⁹Hg/Δ²⁰¹Hg ratio of 1.075, close to the 0.95 Δ¹⁹⁹Hg/Δ²⁰¹Hg ratio we observe at station 13 (Figure 3), implying probably a dominance of anthropogenic emissions while the implication of photoreduction cannot be rejected. Without further data, determining the respective contributions of primary (mixing) and secondary (photoreduction) Hg is difficult at this point. Moreover, as both mechanisms do not yield specific Δ²⁰⁰Hg variation in the Δ²⁰¹Hg/Δ²⁰⁰Hg and Δ¹⁹⁹Hg/Δ²⁰⁰Hg diagrams (Figure 4), even-MIF cannot be used to distinguish sources from process effects (Bergquist et al., 2007; Zheng et al., 2009). The identification of the different sources involved is not further discussed here as it is beyond the scope of our study. However, in the case of a dominance of anthropogenic emission, we suggest that the anthropogenic sources and their respective contributions are not expected to vary over time, as demonstrated by the absence of Hg isotope seasonality at station 13.

Unlike at station 13, samples collected at station 98 display a seasonality in their Hg isotopic compositions, with positive peaks for both odd- and even-MIF in the summer, with values up to 0.78 and 0.15‰, respectively. As the range of Hg concentrations at stations 98 and 13 are similar (i.e., varying from ∼30 to 1800 ng m⁻³), this suggests that, in addition to anthropogenic inputs, the Hg isotopic compositions of aerosols at station 98 may be modified by secondary processes.

The annual seasonality for the δ²⁰²Hg and odd-MIF we observe at station 98 presents similarities with the one reported by Huang et al. (2018) in the urban area of Xiamen (China) where the authors measured low δ²⁰²Hg (i.e., down to −4‰) in winter and high Δ¹⁹⁹Hg, up to 0.7‰ in summer. This is also in agreement with the Δ¹⁹⁹Hg seasonality reported by Huang et al. (2016) in the urban area of Beijing that also presents the highest values in summer but differences in the δ²⁰²Hg seasonality with the corresponding highest values measured during the winter. Finally, our findings are also consistent with the seasonality reported by Fu et al. (2019) in a forested site with high Δ¹⁹⁹Hg-values, up to 0.82‰, in the summer. The authors explained the seasonality by either varying contributions from Hg sources, in particular local anthropogenic ones, or by the long-distance transportation of atmospheric Hg (Huang et al., 2018; Fu et al., 2019). In order to better constrain the potential mechanisms triggering the seasonal variations of Hg isotope MIF, we elected to separate our results obtained at station 98 into two periods, the summer (i.e., high Hg-MIF) and the remaining seasons (i.e., spring, autumn and winter, thereafter referred as RS). This discrimination is based on the fact that PBM samples during these two periods present distinct Hg isotopic compositions, with a mean Δ¹⁹⁹Hg/Δ²⁰¹Hg ratio of 1.63 (R ² = 0.84, p-value < 0.01) for RS aerosols and of 1 (R ² = 0.89, p-value = 0.02) for summer ones.

Our study demonstrates that the current scheme for atmospheric Hg that only considers anthropogenic emissions, oxidation by halogen atoms and photoreduction cannot account for the whole range of Hg multi-isotopic compositions we measured in atmospheric PBM in Montreal. Based on the strong correlations we observe between Δ²⁰⁰Hg and the ozone concentrations, we suggest that Hg oxidation involving ozone might represent the missing oxidation pathway. Our results also demonstrate that an approach coupling Hg isotopic compositions and chemistry discriminates the different oxidation mechanisms controlling the Hg atmospheric budget. Indeed, as both odd and even-MIF Hg isotopic compositions cannot be explained by halogen atom oxidation pathways, nor anthropogenic emissions, nor stratospheric inputs, this demonstrates the necessity that another oxidation pathways is involved. In that perspective, integrating the Hg oxidation by O₃ in the current atmospheric models would help better elucidate the Hg biogeochemical cycle, both in distinct environments and during different seasons, in particular in industrialized urban areas where high O₃ concentrations resulting from the photochemical oxidation of VOCs and CO in the presence of nitrogen oxides (NO_x) have been reported. These urban precursors (i.e., VOCs, CO, and NO_x) concentrations are higher in industrialized urban areas compared to sub-rural and urban ones (Jia et al., 2008). They are mainly emitted by industrial activities, power plants and road traffic (Li et al., 2017; Zheng et al., 2018; Wang et al., 2019), in particular in emerging countries such as China that has now become a hot spot for urban pollution by ozone (Wang et al., 2017; Lu et al., 2018). Considering that O₃ oxidation would be more important in regions characterized by high O₃ concentrations, Hg^II and PBM concentrations should be expected to be enhanced in these regions due to this oxidation pathways for Hg coming from non-point sources (Rutter et al., 2008), affecting Hg^II deposition rates and ultimately Hg^II fluxes at a global scale. This increase would be thus different from the increase of Hg^II and PBM observed due to an artifact of sampling as shown by Lynam et al. (2005). Our hypothesis is furthermore supported by the fact that Hg⁰ oxidation by O₃ can exist in the presence of aerosols and is enhanced in areas experiencing high NO₂ concentrations (Hong et al., 2016; Gencarelli et al., 2017; Travnikov et al., 2017). As several model studies have shown discrepancies between observed and simulated Hg^II concentrations, this arises the urgent need to take both the O₃ and halogen atom oxidation pathways into account in future atmospheric models (Wang et al., 2018).

Overall, our study demonstrates that the Hg multiple isotopic compositions provide a reliable complementary proxy of the atmosphere oxidant capacity and of its chemistry that could help improve our understanding of the processes involved (Fu et al., 2019; Huang et al., 2019). Still, in order to validate our hypothesis and its significant role in the global Hg cycle, further experimental studies are needed to better understand the Hg isotopic fractionations associated to the Hg⁰ oxidation by ozone.