Light-absorbing “humic-like” compounds of secondary origin have been consistently reported in partly inorganic aerosols and in fog waters but their formation could not be explained until now. In this work, we demonstrate that amino acid- and ammonium sulfate-catalyzed reactions in water and ionic solutions produce compounds of identical molecular and optical properties and account well for the quantities found in atmospheric particles. For typical aerosol concentrations of amino acids or ammonium sulfate the rate constants of reaction are found to be identical to the one in concentrated sulfuric acid (10–15 M), clearly demonstrating the efficiency of these catalysts. Our results also show that these reactions should be common in aqueous and ionic aerosols, as confirmed by the observations, and significantly impact their absorption index. In particular, previous radiative calculations indicate that they should substantially reduce current estimates of the cooling contribution of sulfate aerosols on climate.
 Atmospheric particles transparent to light in the near UV and visible (300–800 nm) have a cooling contribution to Earth's climate, clouds and sulfate aerosols being the major part of it [Foster et al., 2007]. Atmospheric aerosols are however rarely pure. Soot, a material strongly absorbing light, is present in sulfate particles resulting from combustion processes. Single-particle analyses have also established that sulfate aerosols in the boundary layer, free troposphere, and even the stratosphere are internally mixed with organic compounds [Takami et al., 2005; Tolocka et al., 2005; Li et al., 2003; Murphy et al., 1998; Cziczo et al., 2004; Kojima et al., 2004], including oxygenated and, most likely, carbonyl ones [Takami et al., 2005; Tolocka et al., 2005]. Carbonyl compounds are also routinely extracted from aerosols in water, indicating their possible presence in the inorganic fraction of the particles. The ubiquitous presence and quantity of organic compounds in inorganic aerosols shows that these compounds are present from the formation of the particles (co-nucleation or coalescence) rather than transferred from the gas phase. But while most organic compounds do not absorb light in the near UV and visible, polyconjugated “humic-like” substances absorbing up to 500 nm have been consistently found in the water-soluble fraction of aerosols containing large inorganic fractions [Zappoli et al., 1999; Krivacsy et al., 2001; Decesari et al., 2001; Kiss et al., 2002; Duarte et al., 2005], and in fog droplets [Decesari et al., 2001; Kiss et al., 2001]. While some humic substances found in atmospheric particles can be of primary origin, biomass burning for instance, the present work focuses specifically on those thought to be secondary, i.e. produced by in situ formation processes, as indicated by their presence in fine particles. Such formation processes have however not been identified until now, making it difficult to quantify their role on the optical properties of the aerosols and to determine the range of aerosols in which they take place.
 In this work, aldol condensation reactions have been studied in water and ionic solutions to investigate their potential for producing secondary “humic-like” compounds in aerosols. The optical and molecular properties of the reaction products have been compared with those reported for compounds found in aerosols. Kinetics experiments were performed to determine if these reactions could also account quantitatively for the observations and if, generally, they could significantly affect the absorption index of aerosols. Finally, for a broader prospective on the significance of these processes for Earth's radiative balance, results from previous radiative calculations have been used to estimate the effect of these reactions on the global radiative forcing of sulfate aerosols.
 See auxiliary material for details. These experiments focused on acetaldehyde (10−3–0.5 M) as a simple model for carbonyl compounds [Córdova et al., 2002b; Tanaka et al., 2004] and were performed in deionized water, rainwater, ammonium sulfate 20–40 wt %, and sodium chloride 0.1–4 M. The catalytic efficiency of five of the most abundant amino acids found aerosols was studied: glycine, alanine, serine, arginine, and proline (0–0.3 M).
3. Results and Discussion
 In all solutions crotonaldehyde, the dimer of acetaldehyde absorbing at 226 nm, was formed (Figure 1a). In the absence of catalyst this was the only light-absorbing product obtained. But its formation was slow and the absorption band is too far in the UV to affect the optical properties of aerosols. But in ammonium sulfate solutions and all the solutions where amino acids were present, products absorbing at 272 and 320 nm also appeared (Figure 1a). These products were identified as 2,4-hexadienal, the trimer of acetaldehyde, and 2,4,6-octatrienal, the tetramer by comparison with UV spectra of these compounds (Figure 1) and by High Resolution Mass Spectrometry (HRMS) (Figure 2). Note that for spectral comparison with octatrienal, only octatrienoic acid methyl ester was available. With arginine, products absorbing at 464 nm were also observed. Because absorption bands further in the visible indicate larger number of conjugated bonds, this compound is expected to be the pentamer of acetaldehyde (2,4,6,8-decatetraenal). The extension of the absorption bands of all these oligomers into the visible resulted in strong coloration of the solutions from yellow to black.
 The secondary polyconjugated compounds found in atmospheric aerosols were not completely identified but their general molecular properties and absorption spectra compared them with humic substances. These properties were a ratio H/C ∼ 1.4 [Kiss et al., 2002], molecular weights between 200 and 600 g mol−1 with a maximum at 250 g mol−1 [Kiss et al., 2001], and the presence of olefinic (-CH = CH-) bonds [Krivacsy et al., 2001; Decesari et al., 2001]. These characteristics are more typical of small fulvic acids than of humic ones and are also very consistent with aldol condensation products such as those obtained in our experiments. For instance, for the three products identified above H/C = 1.5 − 1.25. Generally, such H/C ratios are characteristic of linear polyconjugated compounds rather than of conjugated aromatic or polyaromatic ones for which H/C < 1. Molecular weights between 200 and 600 g mol−1 are also very consistent with oligomers of carbonyl compounds between 50 and 150 g mol−1 while most humic and fulvic acids have molecular weight in the range 103–106 g mol−1. As a confirmation, studies of the molecular structure of small natural fulvic acids concluded that those with H/C > 1.2 have linear polyconjugated structures [Reemtsma and These, 2005; These and Reemtsma, 2005] identical to aldol condensation products (Figure 1b). Aldol condensation is therefore a possible formation process for the secondary “fulvic acids” in aerosols, and perhaps also in other environments.
 One property that aldol condensation would not account for, however, is the acidic character of the compounds found in aerosols. Aldol condensation products such as those presented above have a terminal aldehydic group, likely to be further oxidized into an acid group in aerosols. In addition, the molecular structure proposed previously for the small fulvic acids containing most of the acidic groups seems to be an oxidation product of the polyunsaturated one (Figure 1b). But this proposed polyacid structure does not contain conjugated double bonds (only one isolated double bond) and therefore would not absorb light beyond 300 nm. A possible conclusion is therefore that small fulvic acids such as those in aerosols might be a mixture of polyconjugated compounds that absorb light and polyacidic ones that do not (or less), the former being produced by aldol condensation, and the later being their oxidation products by processes that remain to be determined.
 For a quantitative estimate of the role of these reactions in aerosols, rate constants of reactions were determined. A complete kinetic and mechanistic study of the amino-acid catalyzed aldol condensation of acetaldehyde is presented elsewhere (B. Nozière and A. Córdova, A kinetic and mechanistic study of the amino acid-catalyzed aldol condensation of acetaldehyde in water and ionic solutions, submitted to Journal of Physical Chemistry A, 2007, hereinafter referred to as Nozière and Córdova, submitted manuscript, 2007). For the amino acid concentrations presented here and, more generally, those relevant to natural environments, the kinetics was first order in carbonyl concentration. In water and sodium chloride solutions and for all amino acids except arginine, the rate constant, kI (s−1), increased proportionally with amino acid concentration (Figure 3). Catalysis by arginine displayed a more complex kinetic behavior (Nozière and Córdova, submitted manuscript, 2007) but resulted in very large rate constants at small concentrations. Among the first-order catalysts, glycine and proline had larger efficiencies than alanine and serine. All rate constants increased by a factor 2 in sodium chloride 4 M compared to water, but were unchanged when replacing deionized water by rainwater. Assuming a concentration of glycine in aerosol of 20 mM, corresponding to the 20 pmol m−3 reported in PM2.5 [Zhang and Anastasio, 2003; Matsumoto and Uematsu, 2005] and an aerosol specific volume of 10−12 (m3/m3), the rate constant for aldol condensation in water with this catalyst would be 2 × 10−7 s−1. This is equivalent to the rate constant measured previously for the same reaction in sulfuric acid 80% wt (14 M) [Nozière and Esteve, 2007], clearly demonstrating the efficiency of amino acids as catalysts. But with only 50 mM of arginine in water a rate constant of 5 × 10−6 s−1 was measured, equivalent to the one in sulfuric acid 85% wt (15 M) [Nozière and Esteve, 2007] and indicating the potential importance of this specific amino acid in aerosols and other natural environments.
 Another important result of this work is that ammonium sulfate alone was found for the first time to catalyze aldol condensation (B. Nozière and A. Córdova, Novel catalyst for aldol condensation reactions, patent pending 2007). A study of this type of catalyst and the variations of the rate constants with ammonium sulfate concentration is also presented elsewhere (B. Nozière et al., Carbon-carbon bond formation in atmospheric aerosols and other natural environments: Inorganic salts catalysis, manuscript in preparation, 2007). In ammonium sulfate 40 wt % solutions, the rate constant of reaction was about 10−6 s−1, independently of the presence of amino acids. This is equivalent to the rate constant in sulfuric acid 78% (13 M) [Nozière and Esteve, 2007].
 With these rate constants the fraction of carbonyl compounds converted into light-absorbing oligomers over the aerosol lifetime and the importance of these reactions for the absorption index of the aerosols can be estimated. Note that, as emphasized in introduction, carbonyl compounds are considered to be present in the particles from their formation, and not to transfer from the gas. Potential gas-particle transfer limitations are therefore irrelevant to the present discussion. After a typical aerosol residence time of 4 days, about 15% of the carbonyl compounds would be converted in oligomers in aqueous particles containing amino acids such as glycine (more if they contain arginine or concentrated salt solutions), and about 30% in ammonium sulfate 40 wt % particles, even in the absence of any amino acid. These proportions compare well with the ∼25% of polyconjugated compounds relative to total organics reported in fine aerosols [Zappoli et al., 1999; Krivacsy et al., 2001; Kiss et al., 2002], assuming 50–80% of carbonyls in the total organic content. Absolute concentrations for the oligomers can also be estimated by assuming an initial concentration of carbonyl compounds of 0.7 μg m−3 in sulfate aerosols, as reported above the boundary layer [Maria et al., 2002], an aerosol specific volume of 10−12 and an average molecular weight of 100 g mol−1. Taking into account the number of monomers, the concentration of tetramer after 4 days would thus be 0.2 M in aqueous aerosols containing amino acids, and 0.5 M in ammonium sulfate 40 wt % aerosols. Using the absorption cross-section for octatrienal, σ320 = 0.25 M−1 [Blout and Fields, 1948], leads thus to an absorption index of 0.05 for the aqueous aerosols, and 0.13 for the sulfate ones, which are optically equivalent to a content of 8 and 22% of soot, respectively. Aldol condensation could thus have a significant impact on the absorption index of inorganic aerosols.
 The concentrations of polyconjugated compounds reported in fine aerosols would lead to even larger estimates: 0.5 μg m−3 [Zappoli et al., 1999; Krivacsy et al., 2001; Kiss et al., 2002] would correspond to 2 M for a molecular weight of 250 g mol−1, and lead to a much larger absorption index (∼0.5). Using the absorption cross-section of pure octatrienal is however likely to overestimate the absorption of aerosol “humic-like” substances since, as discussed above, these substances might be a mixture of absorbing and non-absorbing compounds. But, in any case, both our experiments and the quantities of polyconjugated compounds found in aerosols agree that these compounds could have a significant impact on the absorption index of a wide range of aqueous and ionic aerosols.
 For a broader prospective on the role of these reactions, we have reported the absorption index calculated above for sulfate aerosols into previous estimates of the global radiative forcing for these aerosols. The effects of internally mixed absorbing material, such as soot, on the global forcing of sulfate aerosols have been studied with different radiative codes and atmospheric models [Haywood and Shine, 1995; Myhre et al., 1998]. These radiative calculations showed that internally mixed absorbing material cancels out the forcing of sulfate aerosols proportionally to its content. Thus, depending on the study, a soot content of 7.5% increased the forcing from −0.36 W m−2 to −0.16 W m−2 [Haywood and Shine, 1995] or +0.1 Wm−2 [Myhre et al., 1998]. The absorption index estimated above for ammonium sulfate particles, or the equivalent 22% soot content, would thus correspond to a dramatic increase the global forcing of sulfate aerosols by +0.24 to +0.9 Wm−2. Although these results are only rough estimates awaiting more detailed descriptions of the absorption index with wavelength and of the carbonyl and ammonium sulfate content of aerosols, they clearly show that the aldol condensation of carbonyl compounds is likely to have a significant effect on the global forcing of sulfate aerosols.
 In conclusion, this work shows for the first time that aldol condensation can take place in aqueous and ionic atmospheric particles containing carbonyl compounds and explains both qualitatively and quantitatively the secondary formation of “humic-like” substances. Unlike OH-initiated radical reactions considered as the main fate for organic material in aerosols, the reactions presented in this work takes place in the dark. This work also shows that they can take place in a wide range of aerosols and are likely have significant effects on their chemical composition and optical properties. This would affect important aspects such as their forcing on climate, contribution to visibility, and the observation of the atmosphere with satellite instruments. These effects would have been overlooked until now, or mistakenly attributed to soot. Molecular structures proposed for natural fulvic substances indicates that aldol condensation could also participate in the formation of these substances in other natural environments than aerosols, and could thus have a central importance in the geochemical cycles of organic carbon.
 B.N. (atmospheric chemistry and kinetics) acknowledges support of a Marie Curie Chair from the European Commission. A.C. (reaction mechanisms and amino acid catalysis) acknowledges support from the Swedish National Research Council and The Royal Swedish Academy of Sciences.