Iron from coal combustion particles dissolves much faster than 1 mineral dust under simulated atmospheric acid conditions 2

. Mineral dust is the largest source of aerosol iron (Fe) to the offshore global ocean, but acidic processing of coal fly 14 ash (CFA) in the atmosphere could be an important source of soluble aerosol Fe. Here, we determined the Fe speciation and 15 dissolution kinetics of CFA from Aberthaw (United Kingdom), Krakow (Poland), and Shandong (China) in solutions which 16 simulate atmospheric acidic processing. In CFA-PM 10 fractions, 8%-21.5% of the total Fe was as hematite and goethite 17 (dithionite extracted Fe), 2%-6.5 % as amorphous Fe (ascorbate extracted Fe), while magnetite (oxalate extracted Fe) varied 18 from 3%-22%. The remaining 50%-87 % of Fe was associated with other Fe-bearing phases, possibly aluminosilicates. High 19 concentrations of ammonium sulfate ((NH 4 ) 2 SO 4 ), often found in wet aerosols, increased Fe solubility of CFA up to 7 times 20 at low pH (2-3). The oxalate

burning, coal combustion, oil combustion, and metal smelting (e.g., Ito et al., 2018;Rathod et al., 2020). Although these 51 sources are only a small fraction of the total Fe in atmospheric particulates, the Fe solubility of pyrogenic sources can be 1-2 52 orders of magnitude higher than in mineral dust (Ito et al., 2021b and references therein), and thus can be important in 53 promoting carbon uptake. However the Fe solubility of pyrogenic sources varies considerably depending on the particular 54 sources with higher values observed for oil combustion and biomass burning than coal combustion sources (Ito et al., 2021b 55 and references therein). 56 Wang et al. (2015) estimated that coal combustion emitted around ~0.9 Tg yr -1 of Fe into the atmosphere (on average for 1960-57 2007), contributing up to ~86% of the total anthropogenic Fe emissions. A more recent study, which has included metal 58 smelting as an atmospheric Fe source, estimated that coal combustion emitted ~0.7 Tg yr -1 of Fe for the year 2010, contributing 59 around 34% of the total anthropogenic Fe atmospheric loading (Rathod et al., 2020). Although the use of coal as a principal 60 example calcium carbonates (CaCO3), lime (CaO), and portlandite (Ca(OH)2). The estimated concentration of H + buffered was 150 used to input the concentration of H + into the E-AIM model. For each experiment, the pH was calculated before adding the 151 CFA samples and at the end of the experiments. The pH of the original solution before adding the samples was estimated from 152 the molar concentrations (mol L −1 ) of H2SO4, H2C2O4 and (NH4)2SO4 used to prepare the solution. The model inputs included 153 the total concentrations of H + (without H2C2O4 contribution), NH4 + , SO4 2and H2C2O4. For the experiment solutions with no 154 (NH4)2SO4, we calculated the final pH by reducing the total H + concentration input into the model to match the pH measured 155 at the end of the experiments. The buffered H + was then estimated from the difference between the original and final H + 156 concentration input into the model. To determine the final pH of the solutions with high ionic strength, the H + concentration 157 input in the model was calculated as the difference between the H + concentration in the original solution and the buffered H + 158 estimated at low ionic strength. 159 For the solution with no (NH4)2SO4, the difference between calculated and measured pH is <7%. Table S1  For the extraction of FeM, the CFA samples were first leached for 2 hours using a citrate-buffered dithionite solution to remove 173 FeD. The residue collected after filtration was then leached for 6 hours in a solution of 0.2 M ammonium oxalate ((NH4)2C2O4) 174 and 0.17 M H2C2O4 at pH 3.2 (Poulton and Canfield, 2005). The Fe extractions were all carried out in the dark at room 175 temperature. The Fe concentration in the filtered extraction solutions was measured using the ferrozine method (Viollier et al., 176 2000) or by inductively coupled plasma optical emission spectrometry (ICP-OES) analysis for the solutions containing high 177 concentration of oxalate. 178 The total Fe content in the samples was determined by microwave digestion in concentrated nitric acid (HNO3) followed by 179 inductively coupled plasma mass spectrometry (ICP-MS) analysis. The recovery of Fe assessed using a standard reference 180 material for urban particulate matter (NIST SRM 1648A) was around 89%. Therefore, the total Fe in the Libyan dust precursor 181 sample could be underestimated somewhat as crystalline aluminium silicate minerals may not be fully digested. 182 The sequential extraction techniques were tested using the Arizona Test Dust (ATD, Power Technology, Inc.). The RSD% 183 obtained for each extract using the ATD was 3% for FeA, 11% for FeD, 12% for FeM (n=7) and 2% for the total Fe (n=3). A 184 summary of the results for the ATD is reported in Table S2. 185 2.4 X-ray absorption near edge structure (XANES) analysis 186 We collected XANES spectra to qualitatively examine the Fe speciation in the CFA samples. The XANES spectra at the Fe 187 K-edge were collected at the Diamond Light Source beamline I18. A Si(111) double-crystal monochromator was used in the 188 experiments. The beam size was 400 µm×400 µm. The XANES spectra were collected from 7000 to 7300 eV at a resolution 189 varying from 0.2 eV for 3 s in proximity to the Fe K-edge (7100-7125 eV) to 5 eV for 1 s from 7100 to 7300 eV. Powder 190 samples were suspended in methanol and deposited on Kapton ® tape. The analysis was repeated three times. We measured the 191 XANES spectra of the CFA-PM10 fractions and mineral standards including hematite, magnetite, and illite. Data were 192 processed using the Athena program, part of the software package Demeter (version 0.9.26) (Ravel and Newville, 2005). 193

Model description 194
This study used the Integrated Massively Parallel Atmospheric Chemical Transport (IMPACT) model (Ito et al., 2021a and 195 references therein). The model simulates the emission, chemistry, transport, and deposition of Fe-containing aerosols and the 196 precursor gases of inorganic and organic acids. The coating of acidic species on the surface of Fe-containing aerosols promotes 197 the release of soluble Fe in the aerosol deliquescent layer and enhances the aerosol Fe solubility (Li et al., 2017). On the other 198 hand, the external mixing of oxalate-rich aerosols with Fe-rich aerosols can suppress the oxalate-promoted Fe dissolution at 199 low concentration of oxalate near the source regions (Ito, 2015). However, the internal mixing of alkaline minerals such as 200 calcium carbonate with Fe-containing dust aerosols can suppress the Fe dissolution (Ito and Feng, 2010). Since CFA particles 201 are co-emitted with acidic species, the transformation of relatively insoluble Fe in coal combustion aerosols into dissolved Fe 202 is generally much faster than that for mineral dust aerosols during their atmospheric lifetime (Ito, 2015;Ito and Shi, 2016). 203 Additionally, the size of CFA particles is substantially smaller than that of mineral dust. Thus, we adopted an observationally 204 constrained parameter for the dry deposition scheme (Emerson et al., 2020) to improve the simulation of dry deposition velocity 205 of fine particles. 206 To improve the accuracy of our simulations of Fe-containing aerosols, we revised the on-line Fe dissolution schemes in the 207 original model (Ito et al., 2021a) in conjunction with a more dynamic range of pH estimates. To apply the Fe dissolution 208 schemes for high ionic strength in aerosols, we used the mean activity coefficient for pH estimate (Pye et al., 2020). Moreover, 209 the dissolution rate was assumed to be dependent of pH for highly acidic solutions (pH < 2) unlike in the former dissolution 210 scheme (Ito, 2015), which allowed us to predict the sensitivity of Fe dissolution to pH lower than 2. 211 To validate the new dissolution scheme, we compared our model results with observations of Fe solubility in PM2.5 aerosol 212 particles over the Bay of Bengal (Bikkina et al., 2020). 213

Fe dissolution kinetics 215
We determined that Krakow ash had the largest buffer capacity, around 0.008 moles of buffered H + per litre, which was related 216 to the content of alkaline minerals in the sample. The buffer capacity of Aberthaw and Shandong ash was ~10 times smaller 217 than that of Krakow ash, around 0.0007 moles of buffered H + per litre. Leaching Krakow ash in 0.005 M H2SO4, the initial 218 concentration of H + was similar to the concentration of the H + buffered. As a result, the solution pH raised from 219 approximatively 2.1 to 2.7 corresponding to a pH change of around 20% (Table S1). For all the other experimental conditions, 220 the pH change was below 12% (Table S1). At the pH conditions used in this study (pH 1-3), acid buffering was fast and likely 221 occurred within the first 1-2 hours. We assumed that the calculated final pH was representative of the solution pH over the 222 duration of the experiments. The leaching experiments were conducted up to 168 h to better capture the dissolution curve in 223 the kinetic model but also considering the tropospheric lifetime of aerosol particles. 224 Dissolved Fe at different time intervals is reported as Fe%, which is the fraction of Fe dissolved to the total Fe content (FeT) 225 in the CFA samples. For all samples, a fast dissolution rate was observed at the beginning of the experiment. In the case of 226 Krakow ash, the dissolution plateau was reached after 2-hour leaching in 0.005 M H2SO4 as sufficient Fe may be dissolved 227 from the highly reactive Fe species to suppress the dissolution of less reactive Fe. For that sample/initial condition the pH 228 increased to 2.7, and no more Fe was dissolved, leading to a total Fe solubility of ~9% over the duration of the experiment (7 229 days) (Fig. 1a). Dissolving Krakow ash in 0.01 M H2SO4 (Fig. 1a), the experiment solution had a final calculated pH of 2.1. 230 The total Fe solubility was 34% at pH 2.1, almost 4 times higher than that at pH 2.7 (in 0.005 M H2SO4). Dissolution of 231 Aberthaw and Shandong ash was slower compared to Krakow ash (Figs. 1b and 2c, respectively). Leaching Aberthaw and 232 Shandong ash in 0.005 M H2SO4 resulted in solutions with a pH of around 2.2. At this pH, the total Fe solubility was 18% for 233 Aberthaw ash and 21% for Shandong ash, which is 9-10 times higher than the total Fe solubility at pH 2.9 (in 0.001 M H2SO4), 234 around 2% for both samples. 235 The experimental treatment of dissolved Fe from Krakow ash in 0.05 H2SO4 solution with 1 M (NH4)2SO4 (Fig. 1a) resulted 236 in a final predicted pH of 2.1. At that pH, the total Fe solubility of Krakow ash increased from 34% with no (NH4)2SO4 to 48% 237 with high (NH4)2SO4 concentration. The total Fe solubility of Krakow ash was around 28% at pH 3.0 with 1 M (NH4)2SO4 238 ( Fig. 1a), 3 times higher than that at pH 2.7 with no (NH4)2SO4. At around pH 2, the total Fe solubility of Aberthaw (Fig. 1b) 239 and Shandong ash (Fig. 1c) increased by around 20% and 30% in the presence of (NH4)2SO4. By contrast, the total Fe solubility 240 at pH 3.1 with 1 M (NH4)2SO4 was 7.5% for Aberthaw ash (Fig. 1b) and 14% for Shandong ash (Fig. 1c), respectively, which 241 was around 4 and 7 times higher than in the experiments carried out at pH 2.9 without (NH4)2SO4. 242 The Fe dissolution of the CFA samples in H2SO4 solutions with 0.01 M H2C2O4 (at around pH 2) is shown in Fig. 2

. The total 243
Fe solubility of Krakow ash at pH 1.9 with 0.01 M H2C2O4 was 61% (Fig. 2a), which was almost 2 times higher than that at 244 pH 2.1 but without H2C2O4 (Fig. 2a). For Aberthaw ash, oxalate contribution to the dissolution process led to a total Fe 245 solubility of 30% at pH 2.0 ( Fig. 2b), which was 70% higher than in the experiment carried out in 0.005 M H2SO4 (~pH 2.2) 246 ( Fig. 2b). Shandong ash dissolution behaviour was not affected by the presence of oxalate (Fig. 2c). 247 We also investigated the effect of high (NH4)2SO4 concentration on oxalate-promoted dissolution. In Fig. 2a, the total Fe 248 solubility of Krakow ash decreased from 61% at pH 1.9 in the presence of oxalate to 54% at pH 2.0 with oxalate and (NH4)2SO4. 249 For Aberthaw ash, the total Fe solubility at pH 2.0 decreased from 30% in the presence of oxalate to 19% after the addition of 250 (NH4)2SO4 (Fig. 2b). 251  Table S3). The highest total Fe solubility was observed at pH 1.0 (~67%). At pH 2.0, the total Fe solubility 255 decreased to 54%, and no substantial variations were observed between pH 2.0 and pH 2.9 (54%-51%). At pH 1.0, the 256 concentration of H + was considerably higher compared to pH 2.0-2.9, leading to a faster dissolution rate. The total 257 concentration of oxalate ions was 1.5-1.6 times higher in the solution at pH 1.0 than at pH 2.0-2.9, which may also contribute 258 to the faster dissolution rate. C2O4 -2 concentration increased with rising pH. Although the concentration of H + was lower at pH 259 2.9 than at pH 2.0, the E-AIM model estimated that C2O4 -2 contributed around 35% of the total oxalate concentration at pH 260 2.9, which was 4.5 times higher than at pH 2.0 (Experiments 3 Table S3). The similar dissolution behaviour at pH 2.0 and pH 261 2.9 conditions may reflect the combination of these two opposite factors, higher concentration of C2O4 -2 but lower 262 concentration of H + at pH 2.9 compared to 2.0. 263 We determined the Fe dissolution behaviour of Krakow ash at pH 1.0 in the presence of oxalate and increasing concentrations 264 of (NH4)2SO4. The ash was leached in H2SO4 solutions with 0.03 M H2C2O4 at pH 1.0, while the concentration of (NH4)2SO4 265 varied from 0 to 1.5 M. In Fig. 4, the total Fe solubility of Krakow ash in the presence of oxalate was 75% at pH 1.0 and 266 decreased to 68% after the addition of 0.5 M (NH4)2SO4. Higher (NH4)2SO4 concentrations did not affect the Fe dissolution 267 behaviour in the presence of oxalate at pH 1.0. 268

Fe speciation 269
The Fe-bearing phases in the CFA samples determined through sequential extractions are shown in Fig. 5c. The Fe speciation 270 in the Libyan dust precursor is added for comparison. Krakow ash had a total Fe (FeT) content of 5.2%, while FeT in Aberthaw 271 and Shandong ash was 3.1% and 1.6% respectively. Amorphous Fe (FeA/FeT) was 6.5% in Krakow ash, 2% in Aberthaw ash, 272

Fe dissolution scheme 293
Based on the laboratory experiments carried out on the CFA samples, we implemented a 3-step dissolution scheme for proton-294 promoted and oxalate-promoted Fe dissolution (Table 1). The Fe dissolution kinetics were described as follows (Ito, 2015): 295 where RFei is the dissolution rate of individual mineral i, ki is the rate constant (moles Fe g −1 s −1 ), a(H + ) is the H + activity in 297 solution, mi represents the empirical reaction order for protons. The function fi (0 ≤ fi ≤1) accounts for the suppression of 298 mineral dissolution by competition for oxalate between surface Fe and dissolved Fe (Ito, 2015): 299 (2) 300 in which, [Fe] is the molar concentration (mol L −1 ) of Fe 3+ dissolved in solution, and [lig] is the molar concentration of ligand 301 (e.g., oxalate). fi was set to 1 for the proton-promoted dissolution. 302 The scheme assumes 3 rate constants "fast", "intermediate" and "slow" for the proton-promoted, and the proton + oxalate-303 promoted dissolution (Table 1). These were obtained by fitting the parameters to our measurements for Krakow ash in H2SO4 304 and (NH4)2SO4 at pH 2-3, with and without oxalate (Experiments 2 and 3 in Table S1), which are shown in Fig. 6 Shi et al., 2015). Similarly, we predicted the dissolution kinetics of Aberthaw ash and Shandong ash (Fig. 7). The dissolution 309 kinetics of Krakow ash were calculated based also on the experimental results at pH 1.0, which is shown in Fig. S2 in 310 comparison with kinetics predicted at pH 2.0 and pH 2.9 conditions. 311 The contribution of the oxalate-promoted dissolution to dissolved Fe was derived as the difference between the estimated 312 dissolution rates for the proton + oxalate-promoted dissolution and the proton-promoted dissolution: 313 The Fe dissolution rates were predicted at a wider range of pH using Eq. (1) and Eq. (3) and the parameters in Table 1: 315 RFe i = RFe i(proton + oxalate) when RFe i(oxalate) < 0 (4) 316 Since RFei(oxalate) is less than 0 at low pH (< 2), this equation applies to highly acidic conditions. As a result, the predicted 317 amount of dissolved Fe was smaller when using the dissolution rate for the proton + oxalate-promoted dissolution, RFei(proton + 318 oxalate), rather than the rate for the proton-promoted dissolution, RFei(proton), at pH < 2. Accordingly, the dissolution rate, RFei, 319 was less dependent on the pH compared to RFei(proton) at highly acidic conditions, possibly due to the competition for the 320 formation of surface complexes. 321 At pH > 2 when oxalate does promote Fe dissolution, the following equation applies: 322

Aerosol Fe solubility over the Bay of Bengal 324
The new dissolution scheme was applied in the IMPACT atmospheric chemistry transport model to predict the For all simulations, the total Fe emissions from anthropogenic combustion sources and biomass burning were estimated using 332 the Fe emission inventory of Ito et al. (2018) including also emissions from the iron and steel industry, whereas Fe emissions 333 from mineral dust sources were dynamically simulated (Ito et al., 2021a). In Test 0, we ran the model without the upgrades of 334 the dissolution scheme discussed in section 2.4, and apply in addition the photoinduced dissolution scheme for both combustion 335 and dust aerosols (Ito, 2015;Ito and Shi, 2016), which was turned off in Test 1, Test 2, and Test 3 due to the lack of laboratory 336 measurements under high ionic strength. To estimate the aerosol pH, we applied a H + activity coefficient of 1 for Test 0, while 337 the mean activity coefficient from Pye et al. (2020) was used for the other tests. The dissolution rate was assumed as pH-338 independent for highly acidic solutions (pH < 2) (Ito, 2015) in Test 0, based on the laboratory measurements in Chen et al. 339 (2012), while no pH threshold was considered in Test 1, Test 2, and Test 3 as the total dissolution (proton + oxalate) was 340 suppressed at pH < 2 from the predicted dissolution rate. 341 In Test 1, we used the new dissolution scheme accounting for the proton-and oxalate-promoted dissolution of Krakow ash 342 for all combustion aerosols in the model ( Table 1). The dissolution kinetics were calculated using the base mineralogy for 343 anthropogenic Fe emissions reported in Table S11  Thus, the extreme value recorded only for PM2.5 on this date may be an outlier. 370 The comparison of Fe solubility using the same total Fe emissions directly represents the effect of the new dissolution scheme 371 on PM2.5. The aerosol Fe solubility measured over the South Bay of Bengal is higher than that over the North Bay of Bengal, 372 respectively 32% ± 11% and 15% ± 7% (Bikkina et al., 2020), and model estimates showed a similar trend (Fig. 9). In Fig. 9  373 and Table S5, the calculated Fe solubilities over the North Bay of Bengal in Test 1 (11% ± 4%), Test 2 (17% ± 5%), and Test 374 3 (17% ± 6%) were in good agreement with observations. The aerosol Fe solubility over the South Bay of Bengal was better 375 captured in Test 1 (30% ± 5%) and Test 3 (37% ± 7%), whereas Test 0 showed higher variability (37% ± 22%). The proton-376 promoted dissolution scheme in Test 2 significantly overestimated the Fe solubility over the Bay of Bengal ( Fig. 9 and Table  377 S5). The aerosol Fe solubility was largely overestimated in all scenarios after 22 January 2009, as open biomass burning 378 sources become dominant ( Fig. 8 and Table S4). 379 The comparison between observations and model predictions of aerosol Fe solubility over the Bay of Bengal is shown in Fig.  380 S3. The agreement between measurements and model predictions was the best in Test 1 and Test 3. These after 6 hours at pH 2 was 6%-10% for Aberthaw and Shandong ash respectively, and 28% for Krakow ash (Fig. 1) Our results showed that high ionic strength has a major impact on dissolution rates of CFA at low pH (i.e., pH 2-3). The Fe 402 solubility of CFA increased by approximatively 20%-40% in the presence of 1 M (NH4)2SO4 at around pH 2 over the duration 403 of the experiments, and by a factor from 3 to 7 at around pH 3 conditions (Fig. 1). At high ionic strength, the activity of ions 404 in solution is reduced, thus, in order to maintain similar pH conditions, the H + concentration has to be increased (Table S1). 405 Although Fe dissolution was primarily controlled by the concentration of H + , the high concentration of sulfate ions could also 406 be an important factor contributing to Fe dissolution, in particular when the concentration of H + in the system was low (e.g., 407 pH 3). Previous research found that the high ability of anions to form soluble complexes with metals can enhance containing 1 M NaCl. The Fe solubility measured after 24 hours varied from 15% to 70% in different CFA (bulk samples) at 418 pH 2 with 1 M NaCl, which was considerably higher than that observed at pH 2 with 1 M NaNO3 (<20%) (Kim et al., 2020). 419 Both studies did not investigate the impact of ionic strength on the dissolution behaviour, i.e., by comparing the dissolution at 420 low and high ionic strength. Note that both studies did not specify how the pH conditions were maintained at pH 2. Here, we 421 considered the most important sources of high ionic strength in aerosol water and simulated Fe dissolution in the presence of 422 (NH4)2SO4 and H2C2O4 under acidic conditions. We emphasize that the pH under high ionic strength here is estimated from a 423 thermodynamic model, similar to those implemented in the IMPACT model. 424 The presence of oxalate enhanced Fe dissolution in Krakow and Aberthaw ash but not in Shandong ash at around pH 2 (Fig.  425 2). The effect of oxalate on the Fe dissolution kinetics has also been studied by Chen and Grassian (2013) at pH 2 (11.6 mM 426 H2C2O4). After 45-hour leaching, the Fe solubility of the certified CFA 2689 increased from 16% in H2SO4 at pH 2 to 44% in 427 H2C2O4 at the same pH (Chen and Grassian, 2013). Therefore, the enhancement in Fe solubility of CFA in the presence of 428 oxalate observed in this study (from no impact in Shandong ash to doubled dissolution in Krakow ash) is lower than the 2. Our results also indicated that high (NH4)2SO4 concentrations suppress oxalate-promoted Fe dissolution of CFA (Fig. 2), which 437 was not considered in previous research. At pH 1.9 in the presence of oxalate, the Fe solubility of Krakow ash decreased by 438 around 10% after the addition of (NH4)2SO4, while the Fe solubility of Aberthaw ash decreased by 35% (Fig. 2). We used the 439 E-AIM model to estimate the concentration of oxalate ions and their activity (Table S3). The pH influences the speciation of 440 H2C2O4 in solution (e.g., Lee et al., 2007). H2C2O4 is the main species below pH 2, whereas HC2O4is dominant between pH 441 2-4. Above pH 4, C2O4 -2 is the principal species. In our experiments, H2C2O4 is mainly as HC2O4at around pH 2 (Experiments 442 Table S3). In the presence of (NH4)2SO4, the activity coefficient of HC2O4was reduced by approximatively 35-38% 443 (Experiments 3 in Table S3). Increasing the ionic strength lowers the activity of the oxalate ions, but at the same time favours 444 the dissociation of the acid. At around pH 2 conditions, the E-AIM model estimated that the activity of C2O4 -2 was reduced by 445 around one order of magnitude in the presence of (NH4)2SO4, while its concentration increased 12-15 times (Experiments 3 in 446 Table S3). The adsorption of anions can reduce oxalate adsorption on the particle surface due to electrostatic repulsion which 447 results in slower release of Fe (Eick et al., 1999). Precipitation of ammonium hydrogen oxalate (NH4HC2O4) can also occur in 448 the system, but this is very soluble and easily re-dissolves forming soluble oxalate species (Lee et al., 2007). We speculate that 449 the high concentration of sulfate ions is likely to be responsible for inhibiting the oxalate-promoted dissolution by reducing 450 oxalate adsorption on the particle surface. At pH 1 in the presence of oxalate, increasing the concentration of (NH4)2SO4 from 451 0.5 M to 1.5 M did not affect the Fe dissolution behaviour of the CFA samples (Fig. 4). As previously discussed, the adsorption 452 of sulfate ions on the particle surface may inhibit oxalate-promoted dissolution. However, once the saturation coverage is 453 reached, increasing the concentration of anions has no further effect on the dissolution rate (Cornell et al., 1976). 454

3-4 in
Fe speciation is an important factor affecting the Fe dissolution behaviour. CFA particles have very different chemical and 455 physical properties depending for example on the nature of coal burned, combustion conditions, cooling process and particle 456 control devices implemented at the power stations (e.g., Blissett and Rowson, 2012;Yao et al., 2015). This is likely the reason 457 why the Fe speciation observed in the CFA samples analysed in this study from different locations varied considerably (Fig.  458   5). In the CFA samples, the Fe dissolution curves for different pH and ionic strengths generally showed the greatest rate of Fe 459 release within the first 2 hours, followed by a slower dissolution, reaching almost a plateau at the end of the experimental run. Finally, the modelled dissolution kinetics obtained using the new dissolution scheme for CFA (Table 1) showed better 481 agreement with laboratory measurements than when using the original scheme (Ito, 2015) (Fig 10). In Figs. 10a-b, we 482 compared the Fe dissolution kinetics of Krakow ash at around pH 2 and 3 with 1 M (NH4)2SO4 calculated using the proton-483 promoted dissolution scheme in Table 1 with the dissolution kinetics calculated at similar pH but using the proton-promoted 484 dissolution scheme for combustion aerosols in Ito (2015) ( Table S6) Table 1 and the dissolution  492 kinetics calculated at similar pH and H2C2O4 concentration but using the scheme in Ito (2015) (i.e., single phase dissolution, 493 see Table S6). The Fe dissolution kinetics predicted using the new dissolution scheme had a much better agreement with 494 measurements. Figure 10e shows the suppression of the oxalate-promoted dissolution at pH 2.0 and high (NH4)2SO4 495 concentrations. At pH 2, the proton-promoted dissolution was comparable to the proton + oxalate-promoted dissolution (Fig.  496 10e), with RFe(oxalate) close to zero (see Eq. 3). At pH 2.9, the proton + oxalate-promoted dissolution was higher than the proton 497 + oxalate-promoted dissolution (Fig. 10f), with RFe(oxalate) > 0 (Eq. 5). 498 Moreover, the new 3-step dissolution scheme better captured the initial fast dissolution of CFA (Fig. 10) Fig. 7, the 504 dissolution kinetics of Aberthaw and Shandong ash calculated using the dissolution rates in Table 1 and the Fe-bearing phases 505 determined in the samples showed a good agreement with measurements. 506

Comparison with mineral dust 507
High ionic strength also impacted the dissolution rates of the Libyan dust precursor sample at low pH (Fig. S4). At around pH 508 2 conditions, the proton-promoted Fe dissolution of Libyan dust was enhanced by ~40% after the addition of (NH4)2SO4. At 509 around pH 2 and with 0.01 M H2C2O4, the Fe solubility of Libyan dust decreased by ~30% in the presence of (NH4)2SO4. 510 Overall, the Fe solubility of Libyan dust was lower compared to that observed in the CFA samples. After 168 hour-leaching 511 at pH 2.1 with 1 M (NH4)2SO4, the Fe solubility of Libyan dust was 7.2% (Fig. S4), which was from around 3 to 7 times lower 512 compared to that of the CFA samples ( Fig. 1). At around pH 2 conditions in the presence of oxalate and high (NH4)2SO4 513 concentration, the Fe solubility of Libyan dust rose to ~13.6% (Fig. S4), which is still 4 times lower than that of Krakow ash 514 and around 1.5 lower than Aberthaw and Shandong ash (Fig. 2). The Fe solubilities of the Libyan dust observed in this study 515 are comparable with those of the Tibesti dust (Tibesti Mountains, Libya, 25.583333N/16.516667E) in Ito and Shi (2016) at 516 similar experimental conditions. 517 The enhanced Fe solubility in CFA compared to mineral dust could be primarily related to the different Fe speciation (Fig. 5). 518 CFA contained more highly reactive Fe and magnetite but less hematite and goethite than mineral dust. 519 Although mineral dust is the largest contribution to aerosol Fe while CFA accounts for only a few percent, atmospheric 520 processing of CFA may result in a larger than expected contribution of bio-accessible Fe deposited to the surface ocean. It is 521 thus important to quantify the amount and nature of CFA in atmospheric particles. 522

Comparison of modelled Fe solubility with field measurements 523
The model results obtained using the new dissolution scheme for the proton + oxalate-promoted dissolution (Table 1) in Test 524 1 and Test 3 provided a better estimate of aerosol Fe solubility over the Bay of Bengal than the other tests ( Figs. 9 and S3). At 525 the same time, the new model improved the agreement of aerosol Fe solubility from Test 0 (68% ± 5%) to Test 1 (35% ± 2%) 526 and Test 3 (47% ± 1%) with the field data (25% ± 3%) but still overestimated it after 22 January 2009, when open biomass 527 burning sources become dominant (Bikkina et al., 2020) as also shown in Fig. 8 and Table S4. This could be due to the 528 unrepresentative Fe speciation used in Test 1 and Test 3 for biomass burning over the Bay of Bengal. To reduce the uncertainty 529 in model predictions, emission inventories could be improved through a comprehensive characterization of Fe species in 530 combustion aerosol particles. 531 The revised model also enabled us to predict sensitivity to a more dynamic range of pH changes, particularly between 532 anthropogenic combustion and biomass burning by the suppression of the oxalate-promoted dissolution at pH lower than 2. In 533 Test 0, the dissolution rate was assumed to be independent from the pH for extremely acidic solutions (pH <2). The results 534 show that the proton-promoted dissolution scheme in Test 2 significantly overestimated aerosol Fe solubility (Figs. 9 and S3), 535 which indicates the suppression of the proton + oxalate-promoted dissolution at pH < 2. In Fig. S5, the model estimates of 536 aerosol Fe solubility over the Bay of Bengal considerably improved in Test 1 (RMSE 11) compared to Test 0 (RMSE 21), but 537 more work is needed to improve size-resolved Fe emission, transport, and deposition. The model results in Test 1 indicate a 538 larger contribution of anthropogenic combustion sources to the atmospheric Fe loading over East Asia (Fig. 11), but a smaller 539 contribution of biomass burning sources downwind from tropical regions (Fig. 12). We demonstrated that the implementation 540 of the new Fe dissolution scheme, including a rapid Fe release at the initial stage and highly acidic conditions, enhanced the 541 model estimates. However, in Test 1, we turned off the photo-reductive dissolution scheme (Ito, 2015), which was based on 542 the laboratory measurements in Chen and Grassian (2013). To determine the photoinduced dissolution kinetics of CFA 543 particles it is necessary to account for the effect of high concentration of (NH4)2SO4 on photo-reductive dissolution rate which 544 should be considered in future research. 545

Data availability statement 546
The new dissolution schemes for the proton-promoted and oxalate-promoted dissolution are reported in Table 1. Table S1 Table S4 and Table S5,  and Aberthaw ash were provided by TJ, while Shandong ash was provided by WL. Soil 5 from Libya was collected by ND. 561 CB prepared the article with contributions from MDK and all the other co-authors. 562

Competing interests 563
The authors declare that they have no conflict of interest.  scheme assumes 3 rate constants "fast", "intermediate" and "slow" for the protonand oxalate-promoted dissolution. The 578 parameters were fitted to our measurements for Krakow ash. 579 b E(pH) = -1.56 × 10 3 × pH + 1.08 × 10 4 . The parameters were fitted to the measurements for soils (Bibi et al., 2014). 580 c m is the reaction order with respect to aqueous phase protons, which was determined by linear regression from our 581 experimental data in the pH range between 2 and 3 for proton-and oxalate-promoted dissolution schemes.  Experiments 1, 3-4 at around pH 2). The data uncertainty was estimated using the error propagation formula.   Table S1). The data uncertainty was estimated using the error propagation formula.  Table S1). The data uncertainty was estimated 608 using the error propagation formula.   Table S1). The Fe dissolution kinetics were predicted using the rate constants in Table 1 Table S6 (Ito, 2015). The proton + oxalate dissolution scheme (Table 1) Table S6 (Ito, 2015). The proton + oxalate dissolution scheme (Table 1)