Supplement of CRI-HOM: A novel chemical mechanism for simulating highly oxygenated organic molecules (HOMs) in global chemistry–aerosol–climate models

Abstract. We present here results from a new mechanism, CRI-HOM, which we have developed to simulate the formation of highly oxygenated organic molecules (HOMs) from the gas-phase oxidation of α-pinene, one of the most widely emitted biogenic volatile organic compounds (BVOCs) by mass. This concise scheme adds 12 species and 66 reactions to the Common Representative Intermediates (CRI) mechanism v2.2 Reduction 5 and enables the representation of semi-explicit HOM treatment suitable for long-term global chemistry–aerosol–climate modelling, within a comprehensive tropospheric chemical mechanism. The key features of the new mechanism are (i) representation of the autoxidation of peroxy radicals from the hydroxyl radical and ozone initiated reactions of α-pinene, (ii) formation of multiple generations of peroxy radicals, (iii) formation of accretion products (dimers), and (iv) isoprene-driven suppression of accretion product formation, as observed in experiments. The mechanism has been constructed through optimisation against a series of flow tube laboratory experiments. The mechanism predicts a HOM yield of 2 %–4.5 % under conditions of low to moderate NOx, in line with experimental observations, and reproduces qualitatively the decline in HOM yield and concentration at higher NOx levels. The mechanism gives a HOM yield that also increases with temperature, in line with observations, and our mechanism compares favourably to some of the limited observations of [HOM] observed in the boreal forest in Finland and in the southeast USA. The reproduction of isoprene-driven suppression of HOMs is a key step forward as it enables global climate models to capture the interaction between the major BVOC species, along with the potential climatic feedbacks. This suppression is demonstrated when the mechanism is used to simulate atmospheric profiles over the boreal forest and rainforest; different isoprene concentrations result in different [HOM] distributions, illustrating the importance of BVOC interactions in atmospheric composition and climate. Finally particle nucleation rates calculated from [HOM] in present-day and pre-industrial atmospheres suggest that “sulfuric-acid-free” nucleation can compete effectively with other nucleation pathways in the boreal forest, particularly in the pre-industrial period, with important implications for the aerosol budget and radiative forcing.


fitted to data * The rate coefficient for the production of the closed shell and alkoxy radical from reaction of the first generation O3RO2 species, RN26BO2, with RO2m and RO2s was taken to be the average of the rate coefficients of the three actual species (C107O2, C109O2 and C10BO2 using the notation of Molteni et al (2019)), weighted by the branching ratio of their 5 production. The rate coefficients for C107O2, C109O2 and C10BO2 were calculated using the methodology of Jenkin et al (2019a).   Table 3). The model was able to replicate the general trend of increasing C15d with isoprene when the rate coefficients were increased with 20 increasing peroxy radical functionalisation (line marked "Vary"), reproducing observation within experimental error (shaded region). The lines with k=1×10 -11 , 1×10 -12 and 1×10 -13 show model performance when the specified rate coefficient (in units of cm 3 molecules -1 s -1 ) was used for all O3RO2 and OHRO2.   (2019) 30
(1) 35 (3) 40 where [ 8 ], [ ], [ 10 ] and [ 10 ] are the concentrations of O3, OH and the 10-carbon HOMs formed from ozonolysis and OH oxidation respectively, 23.329 is the rate coefficient for the reactions of HOMs with OH, is the HOM condensation sink, is the HOM photolysis frequency and 2 % and 23 are the reaction rate coefficients of α-pinene with O3 and OH respectively. 45

50
Comparison to CRI v2.2 The new mechanism and the CRI v2.2 were run in a box model (Simulation D, Table 3) for 8 days with varying temperature (298 K average, amplitude of 4 K) and emissions of isoprene and α-pinene varying sinusoidally ( Fig S3). Time-independent base NO emissions of 4.7x10 9 molecules m -2 s -1 were used with scaling factors of 1, 3, 10, 30, 100 and 200 employed in a manner consistent with Jenkin et al (2015). Time dependent isoprene emissions reached a maximum of 1.1x10 12 molecules m -55 2 s -1 at 13:00 local time and had an average of 7.1x10 11 molecules m -2 s -1 over the period 06:00 to 18:00, similar to emissions used in Jenkin et al (2015) and Bates et al (2019). Time dependent base α-pinene emissions with a mean of 3.23x10 9 molecules m -2 s -1 and maximum of 5.30x10 9 molecules m -2 s -1 at 1500 hours were applied. Further runs were performed with α-pinene emissions scaled by factors of 10 -3 , 10 -2 , 0.1, 0.2, 0.5, 1, 2, 3 and 5 to investigate the model's performance. Initial conditions of CH4 (1.8 ppm), CO (100 ppb), O3 (20 ppb) and HCHO (300 ppt) were applied. 60 Photolysis frequencies simulating conditions at the equator also varied in the diurnal cycle. The box model simulated an instantaneously well-mixed planetary boundary with mixing with the free troposphere (with same composition of initial conditions) represented by the box height increasing from 250 m at night to 1500 m at midday before collapsing back to 250 m at 2100 hours. 65 The "concentration" of a species was taken to be the mean daytime concentration on the 8 th day, the metric used by Jenkin et al (2015) and Bates et al (2019). The performance of all the HOM mechanisms (HOMTI, HOM6000, HOM9000 and HOM12077) was compared to the CRI v2.2.

70
The HOM mechanisms matched the CRI extremely well for OH, O3, NO, NO2, HO2, α-pinene and isoprene as well as the hydroperoxides and nitrates derived from isoprene, methyl vinyl ketone and methacrolein, and the important SOA precursor isoprene epoxy diol (IEPOX)).

80
The difference between mechanisms is less than ±0.05 ppb. Figure S7 -Absolute and percentage difference in 8 th day daylight mean OH between the CRI v2.2 R5 and the HOM9000 mechanism. The difference between mechanisms is less than ±0.3% for the vast majority of the emissions space with 85 the difference exceeding this only under very high emissions of α-pinene.   Figure S21 -Closed shell species in base mechanism compared to HOMTI mechanism. The lower concentrations of TNCARB26, CARB16 and RTN28NO3 were attributed to the increased competition from the autoxidation pathways in the HOM mechanism. RN18NO3 was significantly lower in the HOM mechanisms (not shown) as discussed in the main text.  145 Table S3 -Species and physical parameters used in the HOM altitude profile modelling. Note that for nucleation calculations, the same input species and parameters were used but all data were monthly means. * The modelled monoterpene concentration was halved to approximate the α-pinene concentration (Rinne et al., 2002) 150

Nucleation Parameterisations
The rates of neutral and ion-induced pure biogenic nucleation (Jn and Jiin respectively) are described by the parameterisations (Kirkby et al (2016)) in Eq. 4 and Eq. 5: 155 In reality, the larger accretion products are likely to be better at nucleating due to their lower volatility and even among 10-160 carbon HOMs, more oxidised species will also be more proficient at new particle formation. The condensation sink for ions was calculated by summing over aerosol modes and (Eq. 6).
Where is the Boltzmann constant, temperature (in Kelvin), =1.2 x 10 -4 m 2 V -1 s -1 , =1.6022 x 10 -19 C, is the wet diameter (in m) of the aerosol mode and the mode's particle concentration (per cm 3 ) (wd and c were taken from UKCA run). 165 The ion loss rate, , was then calculated as the sum of the condensation and nucleation sinks (Eq. 7).
The recombination coefficient, , is given by Eq. 8: Where is the concentration of air in molecules per cm 3 . 170 Where is the rate of ion-pair production in cm -3 s -1 .
The sulphuric acid activation parameterisation used was that developed by Kulmala et al (2006) Where A=2x10 -6 s -1

Changes to CRI v2.2 R5 mechanism
Simple rate coefficients (e.g. kRO2NO) and photolysis frequencies (e.g. J41) were taken from CRI (Jenkin et al., 2008, Jenkin et al., 2019b. Unless otherwise stated, unimolecular rate coefficients have units of s -1 . The peroxy radical pools (RO2b, RO2m, 185 RO2s and RO2) represent the total concentration of peroxy radicals falling within the respective pool. In the mechanism used in modelling, certain reactions were lumped together with product fractions weighted by relative rate coefficients to reduce the total number of reactions. For clarity, reactions have been decomposed below. The autoxidation coefficients provided are those fitted at 297 K. Table S5 shows the expressions for the autoxidation coefficients in the 3 temperature dependent mechanisms. 190 The standard reactions rate coefficients used by the CRI are as follows:

Reactions of RN26BO2
Reaction with HO2 forms hydroperoxide species already in CRI, not a HOM due to insufficient oxygens.

Reactions of O3RO2 with medium and small peroxy radical pools (RO2m and RO2s)
Reaction of RN26BO2 is based on corresponding species in MCM. 31. RN26BO2 = 0.5RN25BO2O2 + 0.5CARB16 + 0.5RN9O2 : 8.13E-13 (RO2s+RO2m) ; 275 Rate coefficient and branching ratios of later generation O3RO2 with medium and small peroxy radical pools taken from Roldin et al (2019). The alkoxy radical produced goes on to react as described earlier in this work.

Reactions of RTN28BO2
Reaction with HO2 forms hydroperoxide species already in CRI, not a HOM due to insufficient oxygens. Rate coefficient of OHRO2 with isoprene-derived peroxy radical from fitting of model to experimental data (Berndt et al, 2018b

Photolysis of HOMs
Photolysis of peroxide linkage and carbonyl linkages were considered using MCM frequencies J41 and J22 respectively. The 345 KPP parameter "SUN" was used in experiments where the photolysis frequency was varied.
Photolysis of peroxide linkage in HOM monomer produces one OH and one alkoxy radical which behaves as previously discussed (50% decomposition, 50% isomerisation). As the extent of oxidation of the HOM is unknown, isomerisation produces second generation peroxy radical by default.
78. C15d = UCARB12 + 0.25RN25BO2O2 + 0.25RTN27BO2O2 + 0.5RN9O2 + 0.5CARB16: J41; Photolysis of carbonyl linkage produces an acyl radical and an alkyl radical which will form peroxy radicals. It is assumed that one of these peroxy radicals is big enough to be considered (2nd generation) O3RO2 or OHRO2 For C15d, one of the two peroxy radicals formed is assumed to be of medium size and produce UCARB12 which isomerisation (as occurs for isoprene-derived peroxy radicals).

HOM loss to OH
All HOM species are lost to OH with same rate coefficient as that for large hydroperoxide RTN28OOH in CRI v2.2 R5. The products, closed shell CRI species CARB10 and CARB15, were chosen under the assumption that the HOM fragments and the sum of CRI indices of the product is close to the CRI index of the peroxy radical which formed the HOM (23-27). The reaction of C15d also produces a product featured in the oxidation pathway of isoprene, UCARB10.   Table 4 Derived in this work* Estimated uncertainty in