Primary Radical Effectiveness: Do the Different Chemical Reactivities of Hydroxyl and Chlorine Radicals Matter for Tropospheric Oxidation?

The atmospheric oxidation of organics occurs primarily via reaction cycles involving gas phase radical species, catalysed by nitric oxide (NO), which result in the production of secondary pollutants such as ozone. For these oxidation cycles to occur, they must be initialized by a primary radical, i.e., a radical formed from non-radical precursors. Once formed, these primary radicals can result in the oxidation of organic compounds to produce peroxy radicals that, providing sufficient NO is present, can re-generate “secondary” radicals which can go on to oxidize further organics. Thus, one primary radical can result in the catalytic oxidation of multiple organics. Although the photolysis of ozone in the presence of water vapor to form two hydroxyl (OH) radicals is accepted as the dominant tropospheric primary radical source, multiple other primary radical sources exist and can dominate in certain environments. The chemical reactivity of different radicals to organic and inorganic compounds can be very different, however, and how these differences in radical chemistry impact atmospheric organic oxidation under different atmospheric conditions has not been previously demonstrated. In this work, we use a series of model simulations to investigate the impact of the chemical reactivity of the primary radical on the effectiveness in initializing organic oxidation and thus the production of the secondary pollutant ozone. We compare the chemistries of the OH and atomic chlorine (Cl) radicals and their effectiveness at initializing organic oxidation under different nitrogen oxide and organic concentrations. The OH radical is the dominant tropospheric radical, with both primary and secondary sources. In contrast, Cl has primary sources that show significant spatial heterogeneity throughout the troposphere but is not typically regenerated in catalytic cycles. Both primary OH and Cl can initiate organic oxidation, but this work shows that the relative effectiveness with which they oxidize organics and produce ozone depends on their balance of propagation vs termination reactions which is in turn determined by the chemical environment in which they are produced. In particular, our work shows that in high NOx radical-limited environments, like those found in many urban areas, Cl will be more efficient at oxidizing organics than OH.

Table S1.Reactions added to MCM for the simulation of methane chemistry with OH and Cl.In the OH simulations, where the ClNO2 photolysis products were changed to OH + NO2, there was a need to also modify the reactions of HOCl due to its production via OH + ClNO2.This represented a negligible change to the model chemistry, but was required to ensure there was no Cl in the OH simulations.The three reactions that were changed are given below, with the OH products in parenthesis.

Model validation
Although the simplified model simulations used in this study are by design unrealistic representations of the real atmosphere, due to a desire to keep the VOC chemistry as simple as possible to enable diagnosis of the source of any differences between OH and Cl oxidation, it is our intention to keep the simulations as realistic as possible within these constraints.The following sections provide supporting detail on the model simulations and their sensitivities.
Model constraints and spin-up.
Model simulations were initialised at 4am local time on July 1 st , using the CALNEX ground site (34.140582N, 118.122455W) location for TUV calculations, and a surface albedo of 0.1.Temperature and pressure were fixed at 298 K and 1013 hPa respectively, and water vapour was fixed at 1%.All simulations were initialised with 40 ppb of ozone.In order to represent the impact of Cl vs OH chemistry on NOx within the model a fixed first-order emission of NO was used to obtain the targeted NOx concentrations.This enabled the impact of additional NOx reservoir species in the Cl simulations to be investigated and was considered a more realistic representation than fixing the NO and NO2 concentrations, which would result in changes in the effective NOx emission as NOx reservoir production varied.In order to reduce the spin-up time required to reach a stable NOx concentration, models were initialised with starting concentrations of NO2 equivalent to 50 pptv for the lowest NOx simulations through to 14 ppbv for the highest.This resulted in average model NOx mixing ratios ranging from approximately 0.5 to 50 ppbv across the 100 different NOx levels simulated for each VOC level (see Figure 2 in main text).Figure S1 shows example NO and NO2 profiles from different NOx and CH4 emission simulations.As described in the main text, simulations were performed using single primary VOCs in each simulation (methane, propane or propene) in order to focus on the specific differences in the chemistries, using near explicit mechanisms for both the OH and Cl oxidation pathways.In order to make this simplistic representation of atmospheric VOCs as representative as possible, 10 target VOC concentrations were chosen to cover a realistic range of OH reactivities, ranging from approximately 0.1 s -1 in the lowest VOC simulations (corresponding to minimum mixing ratios of approximately 500 ppbv for CH4, 3 ppbv for C3H8, and 0.07 ppbv for C3H6) to approximately 20 s -1 in the highest VOC simulations (corresponding to maximum mixing ratios of approximately 80 ppmv for CH4, 800 ppbv for C3H8, and 20 ppbv for C3H6).This was deemed a preferential approach to using atmospherically realistic mixing ratios of each of the VOCs as this made it difficult to achieve realistic radical levels and thus ozone production within the simulations.As with NOx, the VOCs were constrained using a starting concentration and a fixed first-order emission, scaled to achieve the targeted OH reactivity equivalent VOC concentrations.Simulations were performed across 10 different VOC levels, for each of the 100 NOx levels.
Simulations were run for 48 hours, with the first day used to generate realistic levels of secondary oxidation products, and the 12-hours centred around local solar noon on the second day used for the analysis presented.Figure S1 shows some example model mixing ratio profiles for O3, NO, NO2, OH, HO2 and formaldehyde, showing that atmospherically relevant values for all these species are achieved within the range of variable space covered by the simulations, despite the simplistic single primary VOC approach.As described in the main text, the only species within the model that was constrained via a fixed diurnal profile was ClNO2.This was done in order to isolate the impact of the different reactions of the OH and Cl chemistries under different NOx and VOC conditions, without the much larger effect of NOx on ClNO2 production, via N2O5 production and uptake.Increasing levels of NOx and O3 will inevitably result in increased N2O5 production and uptake to aerosol, and providing sufficient available particulate chloride this will result in the production of ClNO2.This has been the focus of multiple previous studies (e.g., [5][6][7] ) and is thought to be the primary driver of chlorine chemistry in many coastal urban environments.In this study, however, we are wanting to investigate the impacts of the different reactivity profiles of OH and Cl, not the production mechanisms.For this reason it was decided to use a fixed diurnal profile for ClNO2 (Figure S2), in order to provide direct comparisons between the OH and Cl simulation cases.The diurnal profile chosen was generated by a free running model with parameterised ClNO2 production and tuned to achieve a maximum within the range of the reported ClNO2 concentrations observed during the CALNEX campaign. 8Simulations where the magnitude of this ClNO2 profile were scaled with model NOx emission were explored, however, the simulated Cl chemistry was highly sensitive to this scaling, as would be expected, and normalising for this effect in order to isolate the impact of the chemistry of interest was deemed to add unnecessary complexity compared with the fixed ClNO2 profile across all simulations.In order to represent the physical removal of species from the model (e.g. via deposition or mixing), and prevent the unrealistic build up of oxidation products, a first order loss process was used for all species, with a lifetime of 24 hrs with respect to this loss.As stated in the main text, sensitivity studies were carried out and confirm that our conclusions are not sensitive to this model parameter.Figure S3 shows the simulated ozone for three different methane levels across the full NOx range investigated with the 24 hr lifetime for species to physical removal (solid) and a 12 hr lifetime to physical loss (dashed line).

Atomic chlorine as a fraction of the total radical budget
Although the ClNO2 source of Cl or OH is constant throughout all the model simulations, the magnitude of other radical sources changes significantly as the VOC and NOx concentrations change.This means that throughout the model variable space explored in this work the fraction of the total radicals coming from ClNO2 changes.Figure S4 shows the fraction of the total daytime average OH + Cl production that is Cl in the ClNO2 --> Cl + NO2 simulations, showing a range from 0.1% of the total OH + Cl production in the highest methane and highest NOx simulations to as much as 46% in the lowest methane and highest NOx simulations.This change in the fractional contribution of ClNO2 as a radical source in these simulations makes it difficult to compare the absolute magnitude of the impact of Cl vs OH on ozone production across multiple simulations, such as Figure 2 in the main text, and for this reason our analysis and interpretation focussed on the radical reactivities and in particular the organic reaction fraction.It is worth noting, however, that the largest observed differences in peak ozone between OH and Cl simulations shown in Figure 2 are for the highest methane level that fully transitions into the radical limited regime (orange), for which the fraction of total radical production that comes from ClNO2 is the lowest compared to other simulations in this chemical regime (Figure S4).The production of OH within the model comes from both primary and secondary sources, and the relative magnitudes of these change depending on the NOx and VOC concentrations.Figure S5 shows the magnitudes of the significant radical production reactions within the model for 3 simulations with methane in the middle of the range studied and either low, mid or high NOx, in both the Cl and OH implementations of these simulations.Figure S5A shows that under low-NOx conditions the differences between the OH source strengths are very similar for both the OH and Cl simulations.As NOx increases, and the simulations transition into a more radical limited regime, the OH source from both formaldehyde and ozone photolysis becomes larger in the Cl simulations compared with the analogous OH simulations.This is due to the increased hydrocarbon oxidation from the primary radicals in the Cl simulations, due to the increased radical organic reaction fraction for Cl, resulting in increased production of both formaldehyde and ozone.The secondary OH production from NO + HO2 is also larger in the mid and high NOx Cl simulations than in the analogous OH simulations, again due to increased hydrocarbon oxidation resulting in more radical propagation in the Cl case and thus greater HO2 production.
Figure S5B shows the same data as in Figure S5A but as a stacked plot, showing the change in total radical source across the NOx range investigated.Figure S6 shows a similar breakdown of model radical sources but for the propane simulations, highlighting the negligible impact of larger oxygenates as radical sources within the simulations.

Additional model outputs
The following plots show additional model diagnostics that, although not central to our conclusions, support the evidence presented in the main text.

Figure S2 :
Figure S2: Constrained ClNO2 profile used in all simulations.

Figure S3 .
Figure S3.Comparison of simulations using base conditions and increased deposition showing daily maximum ozone for reactions initiated with Cl (solid lines) and OH (dashed lines) as a function of mean NOx for different levels of CH4 for (a) full range and (b) lower CH4 simulations.

Figure S4 :
Figure S4: Fraction of daytime average model radical (OH + Cl) production rate that comes from ClNO2 photolysis.

Figure S6 :
Figure S6: Radical production in propane (C3H8) simulations for (a) grouped by radical sources for all major radical sources; (b) acetaldehyde and propanal only; and (c) grouped by simulation.

Figure S7 .
Figure S7.Ozone (O3) production for reactions initiated with Cl (solid lines) and OH (dashed lines) as a function of (a) NO emission and (b) mean NOx for different levels of methane (CH4), indicated by different colours.

Figure S8 .
Figure S8.Daily maximum ozone (O3) for reactions initiated with Cl (solid lines) and OH (dashed lines) as a function of NO emission for different levels of methane (CH4), indicated by different colours focused on low NO emission regime.Same data as Figure 2a.

Figure S9 .
Figure S9.Daily maximum ozone (O3) for reactions initiated with Cl (solid lines) and OH (dashed lines) as a function of (a) NO emission and (b) mean NOx for different levels of propane (C3H8), indicated by different colours.

Figure S10
Figure S10 Daily maximum ozone (O3) for reactions initiated with Cl (solid lines) and OH (dashed lines) as a function of (a) NO emission and (b) mean NOx for different levels of propene (C3H6), indicated by different colours.

Figure S11 .
Figure S11.Fraction of radical reactivity attributed to loss versus propagation, shown with total calculated reactivity for (a) OH and (b) Cl during the 2010 CalNex campaign in Los Angeles.Data from Young et al. 9

Table S2 .
Reactions added to MCM (in addition to those listed in TableS1) for the simulation of propene and propene chemistry with OH and Cl.

Table S3 .
Rate coefficients for selected molecules with Cl and OH at 298K.