Radiative transfer calculations as well as satellite and ground-based retrieval algorithms often use Mie theory to account for atmospheric mineral dust. However, the approximations used in Mie theory are often not appropriate for mineral dust and can lead to inaccuracies in modeling optical properties. Analytic models that are based on Rayleigh theory and account for particle shapes can offer significant advantages when used to model the IR extinction of mineral dust in the accumulation size mode. Here we extend our investigations of the IR optical properties of mineral dust to include samples that have been processed with organic acids. In particular, we aerosolize several individual components of mineral dust with organic compounds that are common in the atmosphere. Through online and offline analysis of the resulting aerosol particles combining Fourier transform infrared (FTIR) extinction spectroscopy, micro-Raman spectroscopy, and scanning electron microscopy, we have identified three distinct outcomes of the interactions, which depend on the nature of the mineral and the organic acid: reactions with segregation of the products within the particle, formation of a uniform coating on the particle, or a formation of external mixture when there is no significant chemical interaction. Analysis of FTIR extinction spectra of the different dust components that have undergone processing shows red shifts of the prominent IR resonance peaks. The extent of the red shift, which varies from 2 to 10 cm−1, depends on the mineral and the nature of the interaction. Spectral simulations showed that the deviation from Mie theory becomes even more pronounced for these processed mineral dust aerosol components.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Mineral dust is one of the major constituents of particulate matter in the atmosphere [Andreae, 1995; D'Almeida et al., 1991]. The radiative effect of atmospheric aerosols strongly influences the global energy balance [Myhre et al., 2003; Satheesh and Krishna Moorthy, 2005], yet there are many uncertainties associated with the role of mineral dust aerosol on climate [Pachauri and Reisinger, 2007]. Mineral dust can also act as cloud condensation nuclei (CCN) or ice nuclei (IN) and as sites for heterogeneous reactions, thereby affecting climate through indirect means.
 Satellite spectral data in the IR region are extensively used in the determination of important properties of both the atmosphere and hydrosphere [Sokolik, 2002]. In order to properly retrieve this information, the radiative effect of atmospheric aerosols must be included in data interpretation [DeSouza-Machado et al., 2006; Klüser et al., 2012]. Data in the mid-IR region may also be used both to indicate the presence of dust and to characterize the specific mineral composition [Ackerman, 1997; Sokolik, 2002].
 Algorithms that are used in processing remote sensing data and in climate modeling calculations often employ Mie theory [Conant et al., 2003; DeSouza-Machado et al., 2006; Moffet and Prather, 2005; Wang et al., 2002]. Mie theory is derived for homogeneous spherical particles [Bohren and Huffman, 1983]. However, it is well known that authentic atmospheric dust particles can exhibit morphologies that greatly deviate from spherical shapes [Dick et al., 1998]. As a consequence, significant errors can arise in the application of Mie theory to such irregularly shaped particles. Shape effects [Hansell et al., 2011] and the associated inaccuracies are a significant source of error in Mie-based calculations of aerosol radiative impact [Kahnert and Nousiainen, 2006; Kahnert et al., 2007]. Earlier studies showed that for clay mineral particles, which have eccentric oblate shape, Mie theory poorly simulates the experimental IR extinction data for fine particles [Hudson et al., 2007; Hudson et al., 2008a]. Models that account for particle shape effects, derived in the Rayleigh regime and valid for particle diameters smaller than the wavelength of light (D < λ), were shown to give a significant improvement over Mie theory. Small clay particles (such as illite, kaolinite, and montmorillonite) are commonly found to have the physical shape of thin flakes and can be successfully modeled as discs in the Rayleigh approximation [Hudson et al., 2008a]. For major nonclay components of dust such as quartz or calcite, a continuous distribution of ellipsoids (CDE) particle model better reproduces the measured IR extinction spectra when compared to Mie theory results [Hudson et al., 2008b]. For authentic, multicomponent mineral dust samples such as Iowa loess and Saharan sand, Mie theory also poorly simulates the measured IR extinction spectra, while the analytic solutions based on these characteristic particle shapes have been shown to give better agreement [Laskina et al., 2012].
 Since important atmospheric and oceanic properties are determined from satellite data in the infrared region, we investigate radiative effects of mineral dust aerosol by measuring and analyzing its infrared extinction spectra. In addition to Fourier transform infrared (FTIR) spectroscopy, we use single-particle micro-Raman spectroscopy for chemical imaging and nondestructive physicochemical characterization of aerosol particles. Chemical imaging of individual particles may be used to directly investigate the heterogeneity of composition within atmospheric aerosol particles.
 The analysis of dust particles from different sources shows that the major minerals present in these particles are clays (illite, kaolinite), quartz, and calcium-rich minerals such as calcite and dolomite [Glaccum and Prospero, 1980; Klaver et al., 2011]. The exact composition of mineral dust aerosol depends on the source region. For example, northern Sahara is abundant in illite, montmorillonite, and carbonates, whereas the southern Sahara and Sahel regions contain more kaolinite [Caquineau et al., 1998]; China loess and Saudi coastal dust are enriched in calcium carbonate [Krueger et al., 2004; Laskin et al., 2005b].
 As aerosols are transported in the atmosphere, they will undergo atmospheric processing. Since mineral dust aerosol surfaces can be highly reactive, aerosol particles may undergo heterogeneous chemical reactions during transport [Tang et al., 2004; Martin et al., 2003; Bauer et al., 2004; Usher et al., 2003]. Besides heterogeneous chemistry, atmospheric processing can also produce coated particles.
 Atmospheric aging of aerosols can alter their physical and chemical properties. These properties include density [Katrib et al., 2005], hygroscopicity [Choi and Chan, 2002; Laskin et al., 2005a, 2005b; Shilling et al., 2006; Ma and He, 2012; Attwood and Greenslade, 2012], CCN activity [Cruz and Pandis, 1998; Abbatt et al., 2005; Gibson et al., 2006a, 2006b], IN activity [Hatch et al., 2008; Baustian et al., 2012], and optical properties [Freedman et al., 2009; Garland et al., 2007; Lu et al., 2011]. Changes in the nucleation properties of mineral dust can have enormous implications for indirect aerosol effects on the global radiative energy balance.
 Mineral dust in the atmosphere is often associated with organic acids due to aging [Lee et al., 2003; Falkovich et al., 2004; Takahama et al., 2010]. For example, a recent field study indicated that 15% of examined particles had an organic coating on a mineral core [Baustian et al., 2012]. Organic acids (both monocarboxylic and dicarboxylic) are important components of atmospheric aerosols [Rogge et al., 1993; Limbeck et al., 2001] with acetic acid being one of the most abundant monocarboxylic acids [Khare et al., 1999; Reiner et al., 1999; Seinfeld and Pandis, 1998]. Acetic acid was observed to be associated with mineral dust aerosol [Wang et al., 2007; Andreae et al., 1988a, 1988b]. Dicarboxylic acids and particularly oxalic acid are among the major water-soluble organic compounds found in the atmosphere [Sorooshian et al., 2006; Limbeck and Puxbaum, 1999; Jacobson et al., 2000]. It was shown that incorporation of dicarboxylic acids into soot aerosol changes its morphology, hygroscopicity, and effective density [Xue et al., 2009a]. In addition, thin coatings of dicarboxylic acids on soot aggregates enhance scattering and absorption of light, which will influence their direct and indirect effects on climate [Xue et al., 2009b]. It is well known that oxalic acid is often internally mixed with mineral dust. Field studies of Asian dust as well as Saharan dust show that oxalic acid is present in these particles [Sullivan and Prather, 2007; Yang et al., 2009; Falkovich et al., 2004]. In addition to small organic acids, macromolecular polyacidic compounds with high molecular weight such as humic-like substances comprise a considerable fraction of organic aerosols [Graber and Rudich, 2006]. These substances are also known to be associated with atmospheric dust [Maria et al., 2004; Falkovich et al., 2004; Russell et al., 2002].
 In this study, we focus on several important components of mineral dust, which were processed through interaction with aqueous solutions of organic acids in the laboratory. Calcite (calcium carbonate) was used as an example of a mineral that is highly reactive with acids [Laskin et al., 2005a, 2005b; Krueger et al., 2004], kaolinite as a clay mineral with a highly eccentric shape factor, and quartz as an example of a less reactive mineral. Acetic and oxalic acids were used to represent abundant monocarboxylic and dicarboxylic acids, and higher molecular weight humic acid sodium salt was also used to coat particles. We first present the spectral characterization of the resulting aerosol particles using IR and Raman spectroscopies, and then we examine the shape and spatial distribution of chemical species in the processed particles using scanning electron microscopy (SEM) images and micro-Raman mapping. Finally, we investigate the accuracy of extinction models based on Mie theory and Rayleigh model solutions for characteristic particle shapes.
2 Experimental Part
2.1 Sources of Materials and Sample Preparation
 Calcium carbonate (CaCO3, 98%) was purchased from OMYACARB, quartz (SiO2, 100%) was purchased from Strem Chemicals, and kaolinite (KGa-1b, low-defect, Washington County, Georgia, USA) was purchased from The Source Clays Repository. Calcium carbonate and kaolinite were used as received; quartz was ground with mortar and pestle prior to use. For the analysis of pure minerals, suspensions of 1 g of mineral were prepared in 20 mL of Optima water (Fisher Scientific). The particles were suspended before the experiment by ultrasonication of the mixture for 15 min. Oxalic acid (H2C2O4, 99.999%) and humic acid sodium salt (technical grade) were purchased from Sigma Aldrich, and glacial acetic acid (C2H4O2, 99.7%) was purchased from Alfa Aesar. Solutions were prepared by dissolving the corresponding acid/salt in Optima water. Table S1 in the supporting information shows the amounts of sample used, as well as weight percent (wt %) of the resulting solutions. For the study of mineral particles reacted with acids, 1 g of mineral (calcite, quartz or kaolinite) was suspended in 20 mL of acid solution. Molar ratios of the resulting suspensions are contained in Table S1 (where applicable). As can be seen from Table S1, the amount of acid varies from 0.03 to 0.67 moles of acid per mole of mineral, which corresponds to the mass fraction of acids between 0.04 and 0.30. During analysis of samples collected in numerous field campaigns in 2004–2008, estimated organic mass fractions for dust particles were 0.3 ± 0.2 [Takahama et al., 2010]. Therefore, mixtures with higher acid content used in this study represent average mass fraction of acids associated with mineral dust aerosol. However, it should be noted that the molar ratio for a given suspension may not represent the ratio between substances in the aerosolized samples obtained from these suspensions. The resulting mixtures were allowed to undergo physical and/or chemical transformations for 1.5 h under ultrasonication and then were left overnight. The samples were re-suspended for 15 min by ultrasonication prior to being used in an experiment. Although cloud processing of mineral dust aerosol is a result of aqueous phase chemistry, it should be noted that the procedures used here were not meant to exactly simulate processes that occur in the atmosphere but, rather, the goal was to better understand how heterogeneous interactions and coatings impact the optical properties of mineral dust aerosol.
 Calcium oxalate monohydrate (CaC2O4•H2O, 99.9985%) was purchased from Alfa Aesar and calcium acetate monohydrate (Ca(CH3CO2)2•H2O, 99.0%) was purchased from Sigma Aldrich. One gram of calcium acetate and 1 g of calcium oxalate were used to prepare a solution/suspension in 20 mL of Optima water. Analysis of these materials was performed for reference and further comparison with mixtures of calcite with oxalic/acetic acid.
 Aerosols were formed using a commercial atomizer (TSI Inc., Model 3076). The aerosol flow of 1.5 Lpm was passed through multiple diffusion dryers (TSI Inc., Model 3062) to achieve a final relative humidity of <5%. The size range of the resulting aerosol population is approximately 10 nm to 2.5 µm (volume-equivalent diameter) as determined from online sampling with a scanning mobility particle sizer (TSI Inc., Model 3936) and an aerodynamic particle sizer (TSI Inc., Model 3321).
2.2 FTIR Extinction Spectra
 Infrared extinction spectra of the aerosol were measured in a long-pass (≈1 m) flow cell using an FTIR spectrometer (Thermo Nicolet, Nexus Model 670) with an external MCT-A detector. Spectra were acquired in the range of 740–4000 cm−1 with 8 cm−1 resolution by averaging 465 scans. Although relative humidity was maintained below 5% during the experiment, weak signals from gas-phase water remained in the spectra and have been subtracted. The standard deviation in the peak position for the main resonance peak of multiple (three to five) measurements is ~2 cm−1.
2.3 Micro-Raman Spectroscopy
 Particles were collected on substrates for Raman analysis from the aerosol flow. A polished quartz disc (1″ × 1/16″, Ted Pella product #16001–1) or 1/4″ × 1/4″ square of Ti foil (0.002″ thick, ESPI Metals) was attached to a microscope slide and placed in the particle flow at the exit of the IR extinction cell for 15 min. Raman spectroscopy was performed using a high-performance dispersive Raman spectrometer (Thermo Fisher Scientific, Nicolet Almega XR). The spectrometer is equipped with an Olympus optical microscope with 10X, 20X, 50X, and 100X magnification lenses. In the experiments described in this paper, the objective lens with 100X magnification was used. Raman spectra were generally recorded in the range of 100–4000 cm−1, or in narrower range with high (4.5 cm−1) or low (15 cm−1) resolution, depending on the sample. Raman scattering was performed using a laser operating at 532 nm; the laser spot size focused on the sample was 0.6 µm. Two exposures of 15 s each were averaged to obtain the resulting spectrum. The laser power was adjusted to avoid sample damage, as was verified by comparing particle images before and after spectrum collection. Raman maps of individual particles were collected by rastering the laser focal spot and gathering point-by-point spectral data with a step size of 0.6 µm.
2.4 Scanning Electron Microscopy
 Particles were collected directly from the aerosol stream onto aluminum stubs for scanning electron microscopy (SEM) analysis, similar to the sample collection method used for Raman spectroscopy (see above). SEM images were collected using a Hitachi S-4800 Scanning Electron Microscope, with a 3000 kV accelerating voltage and a 3.2 mm working distance.
2.5 Model Simulations
 Infrared extinction spectra were simulated using models based on Mie and Rayleigh theories. Details of these models have been described elsewhere [Hudson et al., 2008a, 2008b; Hudson et al., 2011; Laskina et al., 2012]. Briefly, these theories require the particle size distribution and material optical constants as inputs in order to simulate extinction spectra. In this study, extinction spectra were simulated for pure minerals (calcite, quartz, and kaolinite) utilizing the corresponding measured particle size distribution and literature optical constants. Specifically, optical constant data from Querry  were used for kaolinite, from Steyer  for quartz, and from Orofino et al.  for calcite. Size distributions were measured using a scanning mobility particle sizer and an aerodynamic particle sizer operating in tandem. These sizing methods were combined to yield a single size distribution as a function of volume-equivalent diameter [Khlystov et al., 2004]. Simulated IR extinction spectra of minerals were then compared with experimentally measured IR spectra of samples that had been processed with different organic compounds. In the Rayleigh regime, analytic solutions have been derived for characteristic particles shapes [Bohren and Huffman, 1983]. Here we apply the same particle shape models that were shown earlier to most accurately simulate extinction spectra of the dust samples used in this study (i.e., a CDE model for quartz and calcite, and a disc model for kaolinite [Hudson et al., 2008a, 2008b]). Quantitative comparison of the model and experimental results is used to assess the effects of heterogeneous chemical interactions on the quality of the spectral simulations.
3.1 FTIR Extinction Spectra
 The left column in both Figures 1 and 2 shows IR extinction spectra of the minerals, mixtures, and reference materials used in this study. Table 1 gives the list of IR peaks with corresponding descriptions and assignments. In particular, IR spectra of pure calcite, pure calcium oxalate, and calcite aerosolized from solutions containing oxalic acid in different concentrations are shown on the left of Figure 1a. Addition of oxalic acid to calcite gives rise to peaks at 1631 cm−1 and 1323 cm−1 in addition to the CO32− asymmetric stretching mode of calcium carbonate at 1461 cm−1. These two peaks correspond to the in-phase (1631 cm−1) and out-of-phase (1323 cm−1) COO− stretching modes of calcium oxalate, respectively [Trpkovska et al., 2002]. Therefore, the IR data give evidence that a reaction occurs between calcite and oxalic acid via the following chemical reaction pathway:
Table 1. Positions and Assignments of the Experimental Bands Observed in the IR and Raman Spectra of Samples Used in This Study
 It is important to note that Ma and He  did not observe a reaction between oxalic acid and calcite even at 1:1 molar ratio; however, this study and our previous work [Gierlus et al., 2012] clearly indicate that this reaction does occur. Such a discrepancy is likely to be associated with differences in experimental conditions and aerosol generation methods. In the study of Ma and He , externally mixed particles were prepared by grinding different components together that then were placed in a flow reactor. In Gierlus et al. , as well as in this study, internally mixed particles were generated by a constant output atomizer from an aqueous suspension of calcite and oxalic acid solution. In addition, in the current study, the suspension was allowed to undergo physical and/or chemical transformations as discussed in the experimental section.
 Figure S1a depicts an expanded view of the carbonate region (1350–1550 cm−1) for the set of spectra described in Figure 1. The addition of oxalic acid and subsequent formation of calcium oxalate cause a clear red shift of the CO32− vibrational mode at 1461 cm−1. The red line in the plot of Figure 3a represents the shift in the band peak position as a function of oxalic acid concentration. At the maximum concentration of oxalic acid used in this study, the carbonate peak is red shifted by 9 cm−1, as summarized in Table 2.
Table 2. Shifts in IR Resonance Peak Positions Resulting From Interactions of Minerals with Organic Acids and Shifts in Peak Positions for Minerals and Coated/Reacted Minerals Comparing to Theoretical Calculationsa
Peak Position (cm−1)
Red Shift (cm−1)
Δ Analytic Solution
Standard deviation is based on three to five measurements. Positive/negative number denotes blue/red shift of simulation relative to experiment. CDE analytic model was used for calcite and quartz; disc model was used for kaolinite.
 Similar results were obtained for calcite aerosolized from solutions of acetic acid, as shown on the left of Figure 1b. The spectra reveal the appearance and gradual increase of a band at 1567 cm−1 which is assigned to the asymmetric C–O stretching vibration [Frost et al., 2007] in calcium acetate, which is formed according to:
 Spectra of the extended carbonate stretching region in Figure S1b show that the CO32− peak gradually shifts to lower wave numbers as the acetic acid concentration is increased. The shift is represented by the blue line in Figure 3a. As can be seen from the figure and as summarized in Table 2, the red shift reaches 8 cm−1 for the highest concentration of acetic acid used in our study.
 When calcite was added to solutions of humic acid sodium salt, barely any changes were observed in the IR spectra (Figure 1c, left). Only at the highest concentration of humic acid (1.5 wt %) does a weak signal appear at 1593 cm−1 that can be attributed to the aromatic C=C double bonds conjugated with C=O and/or COO− [Sakellariadou, 2006] that are typical of humic materials. The results suggest that there is little interaction and no chemical transformations associated with the addition of calcium carbonate to humic acid solutions. However, when investigating the carbonate region (1350–1550 cm−1) (Figure S1c), it is obvious that CO32− peak also experiences a red shift. As can be seen in the data of Figure 3a (green line) and as summarized in Table 2, the shift is lower than was observed for interactions between calcite and the small organic acids. The maximum shift is only 4 cm−1 compared to 8 and 9 cm−1 for acetic and oxalic acids, respectively.
 IR spectra of quartz aerosol from oxalic acid solutions show spectral signatures for oxalic acid in addition to the characteristic Si–O stretching vibration at 1125 cm−1 (Figure 2a, left). The full list of peak assignments for both quartz and oxalic acid can be found in Table 1. The IR spectral results suggest that quartz does not experience any chemical changes when exposed to oxalic acid solutions. While the Si–O stretching mode appears to be red shifted from 1125 cm−1 in pure quartz to 1122 cm−1 for quartz atomized from a 2 wt % solution of oxalic acid, as illustrated in Figures S1d and 3b (red line), the cumulative 3 cm−1 shift is close to the experimental uncertainty (±2 cm−1). Similar result is observed for quartz aerosolized from a solution of acetic acid. The shift in the peak position is within experimental error. Since no significant changes in IR spectra of quartz mixed with acetic acid are observed, its further analysis was not performed.
 The addition of humic material to quartz and kaolinite gives results similar to those observed for calcite. In addition to the spectral signatures of the respective minerals, bands associated with the humic material are also observed in the aerosol spectra, as can be seen on the left of Figures 2b (quartz) and 2c (kaolinite) (see Table 1 for assignments). While the IR spectra show no evidence for chemical transformations, Figures S1e and S1f indicate that the Si–O peak is red shifted as a result of the interaction with humic acid. Figures S1e and 3b show that the shift is quite significant for quartz (green line), reaching a maximum of 10 cm−1. However, for kaolinite (Figures S1f and 3c), the shift is much lower, only 2 cm−1, which, again, is close to the limit of the experimental uncertainty in the peak position.
 Analysis of kaolinite aerosolized from a solution of acetic and oxalic acid does not show any significant changes in Si–O peak position similar to quartz. The peaks are shifted by less than 2 cm−1 for kaolinite with 2 wt % acetic and oxalic acid. Both of the shifts are within experimental uncertainty that suggests that there is limited interaction between kaolinite and these organic acids. It is expected that such a little change in the measured extinction would be observed if there is only monolayer adsorption of organic acids on kaolinite surface. Further analysis was not performed for these samples.
 A full list of the observed red shifts of the prominent IR resonance peaks for each mineral (CO32− band in the case of calcite; Si–O stretch for quartz and kaolinite) that result from the interactions with organic acids can be found in Table 2.
3.2 Micro-Raman Spectroscopy
 Results of Raman spectroscopy are shown on the right column of Figures 1 and 2 and a full list of the corresponding mode assignments can be found in Table 1. In general, the Raman data support the conclusions drawn from the IR spectroscopic analysis. In particular, Raman spectra shown on the right of Figure 1a give evidence for a reaction between calcite and oxalic acid as, in addition to the strong CO32− stretching vibration at 1087 cm−1 and a weaker CO32− bending mode at 713 cm−1 due to calcite [White, 2009], new bands appear at 1491, 1464, and 897 cm−1 that can be assigned to the C=O stretching band of calcium oxalate monohydrate (1491 cm−1), and C–O (1464 cm−1) and C–C (897 cm−1) stretching modes that are associated with calcium oxalate formation [Frost et al., 2003].
 Similarly, for calcite following the reaction with acetic acid, calcium acetate was detected in the resulting spectra (Figure 1b, right), namely, a strong band at 2933 cm−1 due to the symmetric stretching vibration of the C–H bonds due to the acetic acid methyl group. The full list of the vibrational modes and their assignments can be found in Table 1.
 Raman results for quartz particles aerosolized from solutions of oxalic acid are shown on the right of Figure 2a. As can be seen, even at the highest concentration of oxalic acid used in this study (2 wt %), there was no evidence for oxalic acid in the single particle Raman spectra, nor were any other vibrational bands observed that might be evidence for chemical changes taking place in the quartz particles.
 The Raman spectra of calcite (Figure 1c, right), quartz (Figure 2b, right), and kaolinite (Figure 2c, right) from humic acid solutions also support the conclusions of the IR experiments, namely, that humic material does not cause chemical transformations in these minerals. However, spectral signatures of humic acid sodium salt are present at 1578 cm−1 and 1383 cm−1 due to polycondensed aromatic moieties typically present in humic material [Leyton et al., 2008]. These results suggest that humic material is present as a coating on these particles.
 Raman mapping is a valuable tool that can be used to determine how various chemical components are spatially distributed within a single particle. First, we used micro-Raman spectroscopy for single particle imaging and mapped the location of the minerals used in this study (calcite, quartz, and kaolinite). The two-dimensional intensity profile of one of the resonance lines present in the Raman spectrum can be used to determine the spatial distribution of the corresponding chemical species within the particle. Additional information about particle homogeneity (or heterogeneity) can be gained through a comparison of the chemical and optical images. As can be seen from Figure S2, the optical images overlap very well with Raman chemical distribution images for the pure minerals (calcite, quartz, and kaolinite).
 Raman images of calcium carbonate particles that were aerosolized from a 1.5 wt % solution of oxalic acid are shown in Figure 4a. The spectral map of the 1087 cm−1 peak, attributed to CO32− stretching vibration illustrates the distribution of calcium carbonate within the particle and is shown in red. The intensity of the 1464 cm−1 peak, which is due to C–O stretch of calcium oxalate and thus illustrates the distribution of the oxalate product, is highlighted in blue. Note that the distribution of calcium oxalate on the carbonate core is uneven. The areas of carbonate and oxalate do not overlap, suggesting that the carbonate and oxalate phases are separated within the particle.
 A similar result was obtained for calcite reacted with acetic acid. Figure 4b shows that the distribution of calcium acetate, illustrated as a spectral map of the C–H stretching mode of CH3 at 2933 cm−1 and shown in blue, does not overlap with the map obtained using the carbonate peak at 1087 cm−1 (shown in red). This suggests that the calcium acetate product is segregated and does not evenly coat the entire particle.
 In contrast to the above observations, humic material does not show evidence for such segregation from the mineral core. As can be seen in Figures 4c–4e, the spectral map of the humic material (shown in blue) overlaps with that of the underlying mineral (shown in red), suggesting that humic material forms a coating over the entire particle.
 Raman mapping of quartz aerosolized from oxalic acid solutions was not performed since the Raman spectra of these mixed particles do not show the presence of oxalic acid associated with the quartz.
3.3 Scanning Electron Microscopy
 SEM images can provide information concerning particle shape that is useful for spectral simulations of the IR extinction data. As shown above, the prominent IR resonance peaks of the minerals under study are commonly red shifted upon interaction with organic acids, which might be indicative of a change in the particle shape. It has previously been shown that nonspherical particles exhibit a red shift in IR resonance features when compared to spheres [Hansell et al., 2011]. Therefore, based on our IR analysis, it is reasonable to conjecture that the mineral particles might undergo a change in shape toward more extreme, sharp-edged particles upon interacting with organic acids. However, examination of SEM images does not support this conclusion. The SEM images shown in Figure 5 reveal that treatment of the various minerals with acid introduces more smoothness to the particle contours. This is especially noticeable for humic-coated quartz (Figure 5f) which is more spherical in shape compared to the pure mineral (Figure 5e).
 To quantitatively estimate the change in particles shape, circularity was determined according to
where A is the area of the particle and P the perimeter of the particle. Circularity of 1 indicates a spherical particle. Upon analysis of SEM images, it was determined that circularity changes from 0.74 ± 0.02 for bare minerals to 0.86 ± 0.07 for processed particles that show evidence for either reaction or coating. This indicates that processed particles change toward a more spherical shape (although still deviate from being perfectly spherical).
 It is also important to note that it is still unclear how the particle shape obtained from a two-dimensional analysis of SEM images relates to the particle shape as determined by fitting the particle optical properties [Nousiainen et al., 2011; Meland et al., 2010]. Some studies show that the best fit shape distribution for modeling optical properties does not necessarily correlate with the actual physical shape of the particle [Merikallio et al., 2011].
 Nevertheless, SEM images of particles collected from the aerosol flow stream from a suspension of quartz in oxalic acid solution (Figure 5g) can be used to draw a rather straightforward conclusion. Unlike the other samples studied, quartz particles and oxalic acid particles (rods) are externally mixed. The images also explain why no spectral signatures due to oxalic acid are observed in the Raman spectra of quartz particles, as discussed above.
3.4 Model Simulations
 Extinction spectra for calcite, quartz, and kaolinite were simulated as described in earlier work [Hudson et al., 2008a, 2008b; Hudson et al., 2011; Laskina et al., 2012]. Simulations were based on Mie theory and on analytic solutions derived in the Rayleigh regime for characteristic particle shapes, to model the extinction spectra of the mineral aerosol. For a simple comparison, the simulated IR resonance peak position of the pure mineral was compared with the experimentally determined peak position before and after processing with acid. It is important to note that all spectra were normalized in advance, so we only compare the peak position and do not take the intensity of the peak into consideration. The results of the model simulations are summarized in Table 2.
 As shown in Table 2, the observed IR resonance line peak positions for the unprocessed minerals are routinely red shifted from the Mie theory prediction by values ranging from 8 to 30 cm−1. However, following interaction with the organic acids, the experimental spectral resonances are found to be further red shift away from the Mie theory prediction. Therefore, Mie theory continues to perform poorly for particles that have undergone chemical or physical processing, including particles that are coated with reactive or unreactive compounds. We also note that the analytic solutions for characteristic particle shapes give somewhat better agreement for the IR resonance peak positions of the bare minerals. Since the shifts associated with chemical processing are relatively small (< ~15 cm−1), these Rayleigh model results continue to fit the data for the processed particles somewhat better in that Rayleigh results are closer to experimental values than Mie theory results.
 In order to evaluate the effect of coatings on extinction spectra of the processed minerals, we also used effective medium theories including the Maxwell-Garnett approximation, Bruggeman's model, and core-shell Mie theory [Bohren and Huffman, 1983]. Since optical constants for the coating materials used in this study are unavailable, ammonium sulfate, calcium sulfate, diesel soot, and diethyl phthalate were used as models to represent a range of coating materials: inorganic (ammonium sulfate), organic (diethyl phthalate), and carbonaceous (diesel soot). Wavelength-dependent complex indexes of refraction for these materials are published in a report by Querry . Optical constants for coated particles were calculated using a Maxwell-Garnett approximation and Bruggeman's model. Mie and Rayleigh theories for characteristic particle shapes (CDE model for quartz and calcite, and disc model for kaolinite) were then used to determine the resulting extinction spectra. In addition to these methods, core-shell Mie theory was used. In calculations involving effective medium theories, the volume fraction of inclusions of 20% was chosen, while in coated Mie theory calculations, the coating thickness was set to 10 nm that corresponds to 20% volume fraction for average diameter of minerals D ~ 300 nm. Takahama et al.  showed that organic coating may vary significantly on mineral dust particles. It was estimated that coating thicknesses of carboxylic acid on inorganic cores range between <30 nm (or even less) and 600 nm, but accounted for <0.01 to 0.98 of the particle volume fraction; therefore, the amount of inclusions chosen in this study is in agreement with field observations. The results of the calculations are shown in Tables S2–S4. As can be seen, in general, the materials used as model coatings and the spectral simulation methods used here can show red shifts of the IR resonance peaks in the range of 1–12 cm−1; however, the direction and magnitude of the shift depend on the particular method and, more importantly, the specific coating material. This suggests that it may be possible to predict the experimentally observed red shift of the minerals coated with organic acids by using core-shell Mie or effective medium theories; however, a quantitative analysis would require accurate optical constants for the particular coating material.
4.1 Calcite With Organic Acids
 IR and Raman spectra indicate that the presence of oxalic or acetic acid leads to reaction with the calcium carbonate and resultant formation of calcium oxalate or calcium acetate. It is interesting that these reactions do not form a uniform coating on the particle surface but, rather, form particles with carbonate and oxalate/acetate segregated within individual particles, as can be seen in the Raman maps. Still, it is unclear if this segregation occurred originally during the reaction in solution, or if the segregation is a consequence of the particle drying process.
 Heterogeneous reactions that take place on a particle surface may lead to significant changes in the physical and chemical properties of the particle. For example, hygroscopicity of the particle may be altered, which can cause differences in hygroscopic growth, and, more importantly, it can change the CCN activity of particles. Chemical properties may also change upon atmospheric processing. A particle can gain affinity to specific trace gases, which will then be more likely to adsorb on the particle surface. Finally, these transformations may lead to changes in the aerosol optical properties. Investigation of the carbonate region (1350–1550 cm−1) in the IR spectra for samples studied here shows that position of CO32− peak shifts to the red upon heterogeneous processing by 8–9 cm−1. The changes associated with atmospheric aging can exacerbate the errors associated with Mie theory modeling of the optical properties, which could adversely affect the processing of satellite and surface-based remote sensing data using algorithms based on Mie theory.
4.2 Calcite, Quartz, and Kaolinite With Humic Material
 Humic material forms a coating on the surface of particles, as supported by an analysis of the IR and Raman spectra where spectral signatures of both the mineral and humic acid are present. This is especially clear in the single particle micro-Raman data which show the presence of both the humic acid and the mineral in the same area of the particle. Raman spectral maps give evidence for a uniform humic material coating since the mineral and humic acid maps greatly overlap. Although the chemical composition of the core may not be altered, the formation of a humic coating can still lead to changes in the physical properties of these aerosols. IR extinction spectra of aerosols that were obtained from particles aged in solutions of humic acid show a red shift of the main resonance line from 2 to 10 cm−1 from the unreacted mineral.
 As mentioned above, while these changes may be caused by alterations of the particle shape, our imaging studies suggest something other than particle shape is responsible for the observed spectral shifts. SEM images show that a humic coating leads to a slightly smoother surface and makes the particle appear more spherical. However, the observed red shifts worsen the agreement with Mie theory rather than improve it, as seen from the data in Table 2. Of course, simple Mie theory assumes particles that are not only spherical but also chemically homogeneous, and its application may not be appropriate for particles that have undergone atmospheric processing, as this work suggests.
4.3 Quartz With Oxalic Acid
 Investigation of data obtained for quartz aerosolized from a solution of oxalic acid suggests that quartz and oxalic acid form an external mixture. This conclusion is supported by micro-Raman spectroscopy and SEM images. The fact that there are no spectral features associated with oxalic acid in Raman spectral maps of the processed particle suggests that there is no reactions with the oxalic acid and no acid coating is formed on the quartz particle surface. Additionally, SEM images of samples collected from the aerosol stream definitely show two types of particles: ellipsoidal quartz particles and rod-like oxalic acid particles (Figure 5g). The quartz Si–O stretching mode is slightly red shifted in these external mixtures. However, the maximum concentration of oxalic acid used in this study (2 wt %) causes only a slight red shift (3 cm−1) that is near the limit of experimental error.
 We have incorporated and integrated multiple techniques (Fourier transform infrared extinction spectroscopy, micro-Raman spectroscopy, and scanning electron microscopy) to study the effect of multiphase reactions between model components of mineral dust aerosol and representative organic compounds, with a focus on organic acids. Calcite, quartz, and kaolinite were aerosolized from suspensions of oxalic, acetic, and humic acids with varying concentrations. FTIR and Raman spectroscopy were performed to gain chemical information about the resulting particles. Raman mapping was used to investigate the spatial distribution of chemical species within a single particle. Scanning electron microscopy was performed to examine the changes in particles shape upon processing. Finally, spectral simulations were performed to predict the extinction spectra of minerals, and then these modeled spectra were compared with experimentally measured IR data of processed aerosols. Several important conclusions can be drawn from an examination of the cumulative data.
 Interactions of components of mineral dust with humic and small carboxylic acids may lead to chemical reactions that form heterogeneous particles with phase segregation (calcite with small organic acids). Other acids (humic) can form a uniform coating around the mineral core. Finally, a simple external mixture may be formed, as was observed for quartz with oxalic acid.
 Previous work by our group has shown that spectral simulations based on analytic solutions derived in the Rayleigh regime, which account for particle shape effects, predict the IR resonance line peak position and band shape better than Mie theory for common mineral particles [Hudson et al., 2008a; Hudson et al., 2008b; Laskina et al., 2012]. Processing of mineral dust components by exposure to acids causes a red shift of the major resonance IR peak (CO32− in calcite, or Si–O in quartz and kaolinite). As noted above, it may be possible to explain such shifts using simulations based on effective medium theories and core-shell Mie theory; however, this would require detailed knowledge of the nature of coating and its complex refractive index. On the other hand, the spectral line shifts associated with chemical processing appear to be less significant than the line shifts associated with particle shape effects. As a consequence, the analytic solutions for the resonance spectral line profiles derived in the Rayleigh regime for different characteristic particle shapes continue to fit the data for the processed particles much better than models based on Mie theory. Some narrowband satellite sensors operating in the infrared region have channels as narrow as 35 cm−1. As can be seen in Table 2, errors associated with Mie theory in some cases are close to the width of these satellite channels. Therefore, the use of Mie theory can adversely affect the processing of data received from satellites and for predicting optical properties of mineral dust aerosols. The results of these analyses are summarized in pictorial form in Figure 6.
 This research was supported by the National Science Foundation under grant AGS096824. The views, opinions, and findings contained in this work are those of the authors and do not necessarily reflect the views of the National Science Foundation.