What’s the real role of iron-oxides in the optical 1 properties of dust aerosols?

: Iron oxides compounds constitute an important component of mineral dust aerosol. Several 15 previous studies have shown that these minerals are strong absorbers at visible wavelengths and 16 thus that they play a critical role in the overall climate perturbation caused by dust aerosol. When 17 compiling a database of complex refractive indices of possible mineral species of iron-oxides to study their optical properties, we found that uniformly continuous optical constants for a single 19 type of iron-oxides in the wavelength range between 0.2 μm and 50 μm is very scarce and that the 20 use of hematite to represent all molecular or mineral iron-oxides types is a popular hypothesis. 21 However, the crucial problem is that three continuous datasets for complex refractive indices of 22 hematite are employed in climate models, but there are significant differences between them. Thus, 23 the real role of iron-oxides in the optical properties of dust aerosols becomes a key scientific 24 question, and we address this problem by considering different refractive indices, size 25 distributions, and more logical weight fractions and mixing states of hematite. Based on the 26 microscopic observations, a semi-external mixture that employs an external mixture between 27 Fe-aggregates and other minerals and partly internal mixing

For the real part of the refractive index for iron-oxides, there is a reasonable agreement 166 between the hematite and magnetite datasets from the different references ( Figure 1c). Because the 167 real refractive index of hematite shows large fluctuations at wavelengths longer than 18 μm due to 168 anisotropic refraction, the agreement between the different datasets decreases at these wavelengths. 169 For goethite we are aware of only two sets of optical constants: one at visible wavelengths from 170 Bedidi and Cervelle (1993) and the other at IR wavelengths from Glotch and Rogers (2007), but 171 the wavelength gap between these two datasets hampers continuity. Unfortunately, Meland et al. 172 (2011) have checked the former dataset for goethite using simulations according to Mie and 173 T-Matrix theories and show that it may be in error. Nevertheless, we can see that goethite has 174 optical constants similar to hematite. The real refractive index of hematite is larger than that of 175 magnetite at wavelengths less than 2μm, but is smaller between 2 and33 μm (Figures 1c&d). 176 For the imaginary part of the refractive index of iron-oxides, hematite and goethite have 177 different optical properties at short wavelengths, both in terms of magnitude and spectral 178 dependence (Bedidi and Cervelle, 1993). Between 460 and 700 nm the imaginary part of the 179 complex refractive index (representing absorption) of goethite is up to 3 times smaller than that of 180 hematite. As a consequence, the proportions of hematite and goethite in mineral dust can 181 dust. However, the limited and discontinuous refractive indices of goethite have constrained the 183 evaluation of the effects of specific compositions of goethite and hematite to dust optical 184 properties and solar radiation balance over broader wavelength ranges. 185 From Figure 1a, we clearly see that the k values for hematite from QE1985 and from

186
LG1988 show significant differences for wavelengths between 650 nm and 15 μm. These 187 differences are present at visible wavelengths and disappear at ultraviolet wavelengths, but the two 188 datasets have similar trends at UV and visible wavelengths (Figure 1b) Size distribution is another important factor that affects the optical properties of particles. 199 Because Sokolik and Toon (1999) has employed the refractive index dataset for hematite from 200 QE1985 to calculate the radiative properties, we adopt here the same particle size distribution but 201 with the refractive index dataset for hematite from LG1988 to compare our results with Sokolik 202 and Toon (1999). The lognormal number size distribution is applied to dust aerosols: 203 where 0 r is the median radius,  is the geometric standard deviation, and 0 N is the total 205 particle number density of the component in particles per cubic centimeter. 206 In order to compare with the results of Sokolik and Toon (1999), the optical properties of 207 minerals are calculated on the assumption that they have one size mode but varying median radius. 208 The particle size modes are selected as r 0 = 0.5 and 0.7 μm, and σ = 2.0. The size mode with 209 long-lived, long-distance-transport mode of airborne dust (Patterson and Gillette, 1977;Arimoto et 211 al., 1997). The larger r 0 is representative of a particle size mode which occurs near the dust source 212 (Gomes and Gillette, 1993). In reality, the size distribution of dust aerosols can have one or 213 several modes, characterized by a specific composition (Mahowald et al., 2013). have shown that spherical/non-spherical differences only influence the single scattering albedo by 219 less than 1%. Meland et al. (2011) have also shown that moderate departures from spherical shape 220 are relatively unimportant in determining the scattering matrix for particles with high refractive 221 index values, such as hematite. Therefore, we expect the aerosol asphericity to have a negligible 222 impact on our calculated results of optical properties and subsequent calculations using the Mie 223 theory (which assumes a spherical morphology for the dust particles). 224 There are several different computer codes that can be used to compute optical properties 225 for a lognormal particle size distribution. The theoretical light scattering simulations in this paper 226 have used the MieTab software. MieTab uses a FORTRAN code with continued fraction 227 modification produced by W. J. Lentz from the Mie code originally developed by Dave and Center 228 (1968 precision T-Matrix code for a lognormal particle size distribution from Mishchenko et al. (2002). 232 The double precision Lorenz-Mie and T-Matrix codes are available from 233 http://www.giss.nasa.gov/staff/mmishchenko/t_matrix.html. 234 In addition to the wavelength dependent optical constants and the size distribution, the 235 T-Matrix theory also requires assumptions about the particle shape. In this work we use an aspect 236 ratio of 1.000001 to represent a spherical particle shape, because use of an aspect ratio exactly 237 equal to 1 causes computational overflow in some cases. The calculated results from the three 238 codes at different wavelengths and complex refractive indices for the same size distribution are 239 possibility of computational error affecting the interpretation of the calculated optical properties of 241 iron-oxides can be neglected. 242 243

Basic optical properties 247
We focus here on modeling the spectral optical properties of iron-oxides which are needed 248 for climate modeling: the volume extinction coefficient ext  (which is the sum of the scattering 249 coefficient sca  and the absorption coefficient abs  ), the single scattering albedo 0  , and the 250 asymmetry parameter g (a cosine weighted integral of the scattering phase function). This set of 251 parameters allows the calculation of radiation forcing in most climate models. Figure 2 shows 252 calculated optical parameters for hematite (with complex refractive indices from QE1985 and 253 LG1988) and illite with varying median radius at solar and infrared wavelengths. The volume total 254 extinction coefficients ext  have been normalized as  is clearly distinguishable from that for illite at UV, visible and IR wavelengths. One point should 265 be noted: hematite has a lower normalized spectral extinction coefficient than illite at wavelengths 266 less than 1.3 μm, which means that hematite has a weaker optical extinction capacity than illite at 267 these wavelengths. In the IR region, the spectral features of hematite in  For r 0 = 0.5 and 0.7 μm, hematite from QE1985 has g = 0.3 -0.99, g decreasing as λ increases.

295
The magnitudes of g from LG1988 are in the range from 0.2 to 0.99 with a few fluctuations.

296
For r 0 = 0.01 μm, both datasets put g in the range from about 0.15 to 0.38. Thus, the magnitude of 297 g depends significantly on the particle size distribution. reasons for the aggregation and the attachment are not well understood but are likely to be related 307 to interactions of surface charge characteristics between iron-oxides and quartz or clay minerals 308 (Poulton and Canfield, 2005). According to Hinds (1982), the binding mechanisms that hold 309 separate aerosols together in an agglomerate formed in the air include the van der Waals force, the 310 electrostatic force and the surface tension of adsorbed liquid films. As mentioned above, the 311 dispersed nanoparticles of iron-oxides which are attracted to larger dust particles have more 312 prominent optical absorption than aggregated iron-oxides, but the use of size distributions for 313 nanoparticles (such as r 0 = 0.01 μm and σ =2.0) will overestimate the optical absorption of 314 iron-oxides in natural dust aerosol samples. Single particle analysis has also been conducted for detecting the free iron oxides. Fe-rich 380 particles (iron oxides) represented no more than 5% of the particle number in aerosol samples and 381 hematite or goethite were found more often in the fine fraction (Chou et   has the critical implication that the content of goethite measured by absorption spectroscopy is 392 actually the sum of goethite and ferrihydrite. This does not, however, affect the optical 393 calculations due to their optical similarity. Table 3 summarizes the measured ratios of hematite to 394 goethite in global dust aerosol samples and shows higher ratios of Hm/Gt in Asian dust samples 395 compared to African samples. Over the whole world, it is concluded that goethite predominates 396 over hematite with a relative abundance of 50% -75% of iron oxides in dust aerosols. 397 Based on the above reported results, we conclude that the iron-oxides account for 398 approximately half of the mass of elemental Fe and for between 2 and 5 % of the dust mass. Most 399 of them are composed of goethite, representing between 50 and 75 % of the iron oxide mass. 400

Mixing states 402
As free-iron particles are always mixed with other kinds of particle, the condition of the 403 mixture could be important for their ability to scatter and absorb radiation. The 3D structure of 404 iron-oxide particles obtained by tomography reveals that these Fe-rich inclusions are often found 405 at the surface of aluminosilicate particles but that some are also included inside particles ( which represents the clay minerals. As mentioned above, we adopt 0% hematite as the lower limit 419 for the aerosol samples with no free-iron particles, 2.5% hematite for the transported dust aerosol 420 samples, 5% hematite for the locally emitted dust samples and 7.5% hematite for the upper limit. can be modeled by: 433 Huffman, 1998). Detail information about the three methods is given by Sokolik and Toon (1999). 501 We have calculated the single scattering albedo (SSA) of illite-hematite mixtures with different 502 hematite contents using internal mixing according to the above three internal approximations and 503 also using external mixing. 504 The calculated SSA values for illite-hematite mixtures using internal and external mixture 505 the case of external mixing, the SSAs at 405 nm show good agreement for refractive indices from 507 QE1985 and LG1988, but the calculated SSAs at 870 nm for hematite with refractive indices from 508 QE1985 are much smaller than those using LG1988. This is explained by Figure 4b where the two 509 datasets have the same optical scattering and absorbing properties for λ < 0.55 μm but the dataset 510 of QE1985 leads to higher optical absorption for λ > 0.55 μm. The calculated SSAs with the three 511 different internal mixing methods are all much smaller than those for external mixing both at 405 512 nm and 870 nm since the assumption of an external mixture results in less absorption and less 513 wavelength dependence of absorption than does the assumption of an internal mixture for small 514 amounts of hematite. The basic reason for this is due to the extremely high imaginary refractive 515 index for hematite at short wavelengths. For the case of internal mixing, the SSAs from the 516 volume mixing method are smaller than for the other methods. This is due to the averaged 517 imaginary refractive index being larger than for the other two approximations. On the basis of the 518 study of Peterson (1968), only the effective refractive index of the non-metallic part of the dust 519 can be calculated using the volume mixing method. Thus, adopting the volume mixing method to 520 calculate the optical properties of aerosol samples will lead to a smaller SSA (Levoni et