A representative lignin-type component from biomass burning aerosol has been shown to react with OH radicals in model cloud water yielding colored organic species. In this paper we investigated the chemical properties of the complex reaction products formed from 3,5-dihydroxybenzoic acid. The reaction was followed by UV-VIS spectrophotometry, liquid chromatography, electrospray-mass spectrometry, thermally assisted hydrolysis and methylation-gas chromatography/mass spectrometry and a thermal method. This paper provides experimental proofs that actually larger molecular weight species are formed in the aqueous phase by free radical oligomerization. The features observed by all analytical techniques closely resemble those found for natural humic acids and HULIS found in rural and biomass burning aerosol. Therefore such processes are assumed to produce the ubiquitous humic-like substances (HULIS) in atmospheric aerosol. Since these species show intense absorbance in the lower visible to UV range, they might also be important in atmospheric absorption of solar radiation.
 It is now widely established that humic-like substances (HULIS) are ubiquitous constituents of continental fine aerosol. Recent studies confirm that they are among the most abundant organic species in the aqueous extracts of rural, urban and biomass burning aerosol [Havers et al., 1998; Zappoli et al., 1999; Gelencsér et al., 2000b; Kiss et al., 2002; Mayol-Bracero et al., 2002]. Such abundance may imply–in accordance with a recent hypothesis [Gelencsér et al., 2002]–that such compounds are secondary aerosol constituents which may form in the troposphere from a vast array of precursors emitted by various sources, both biogenic and anthropogenic. In-situ formation of HULIS is not self-evident since humification processes in the soil are assumed to take years and require special microbial environment which is not available aloft. Recently there are three hypotheses with experimental support to explain the atmospheric formation of HULIS. It has been shown that volatile aldehydes produced by photochemical reactions can undergo polymerization and hemiacetal/acetal formation in the presence of sulfuric acid particles as catalyst [Jang and Kamens, 2001]. Other experiments have revealed that isoprenoid or terpenoid hydrocarbons can directly yield HULIS in the presence of sulfuric acid catalyst [Limbeck et al., 2003]. Recently we have also presented experimental evidences proving that aromatic hydroxy-acids which have been found in continental fine aerosol can react with hydroxyl radicals in model solution of cloud water yielding colored organic species [Gelencsér et al., 2003]. We also suggested that the complex array of products of such reactions from a multitude of similar precursors may lead to HULIS found in continental aerosol. In this paper the chemical structure and properties of the reaction products (hereinafter referred to as synthetic HULIS) formed from 3,5-dihydroxybenzoic acid in model cloud water are exploited with modern analytical techniques and compared to those found in natural humic matter as well as in continental fine aerosol.
 We used Fenton-reactions to produce OH radicals in the model solution, adhering to concentration ratios of the reagents which were reported to be prevalent in cloud water. The rationale behind the selection of reagent concentrations is detailed in Gelencsér et al. .
 UV-VIS spectrophotometry was performed with a Jasco UV-VIS 630 instrument. The spectra of the solution containing 2 · 10−3M 3,5-dihydroxybenzoic acid were recorded in a cell of 0.5 cm length.
 Liquid chromatographic separations were carried out on a C18 column (Waters Nova-Pack 3.9 mm × 150 mm, 4 μm) with UV-VIS detection at wavelengths of 425 and 210 nm. In the time range of 0–8 minutes phosphate buffer containing 3% acetonitrile at pH = 2 was used as the eluent, followed with linear gradient elution to 100% acetonitrile to 28 min. The flow rate was 1 ml min−1, and the column was thermostated at 35°C. The initial concentrations (t = 0) of 3,5-dihydroxybenzoic acid, H2O2, and Fe3+ were 2 · 10−3, 10−2 and 5 · 10−7 M, respectively, and the solution was injected after a reaction time of 50 h.
 Mass spectrometry was performed with a Micromass Quattro II tandem quadrupole mass spectrometer equipped with an atmospheric pressure ion source used in the electrospray mode. The mass spectra were recorded from m/z = 40 to 1000 in negative ion mode applying cone voltage of 40V. The initial concentrations (t = 0) of 3,5-dihydroxybenzoic acid, H2O2, and Fe3+ were 2 · 10−4, 10−3 and 5 · 10−7 M, respectively. After reaction time of 3 days the solution was freeze dried, then taken up with isopropanol:water = 1:9 mixture.
 Thermally assisted hydrolysis and methylation-gas chromatography/mass spectrometry (THM-GC/MS) was performed at 400°C for 20 sec in a Pyroprobe 2000 pyrolyzer (Chemical Data System) interfaced to a gas chromatograph (Agilent 6890) coupled with a mass selective detector (Agilent 5973) operating in electron impact mode (EI) at 70 eV. Further details of the method can be found elsewhere [Gelencsér et al., 2000b]. The initial concentrations (t = 0) of 3,5-dihydroxybenzoic acid, H2O2, and Fe3+ were 2 · 10−3, 10−2 and 5 · 10−7 M. After reaction times of 1 day and 7 days the solutions were concentrated to dryness at 40°C at reduced pressure, then taken up with methanol and again evaporated to dryness. 1 μl of tetramethylammonium hydroxyde (15 w% in water, from Fluka, Switzerland) has been poured onto the dry sample of 50 μg before introducing into the pyrolyzer.
 Thermal profiling was performed with a TOC analyzer (Zellweger Analytics, Astro 2100). More details about the operating principle of the instrument can be found elsewhere [Gelencsér et al., 2000a]. The same samples as above were analyzed. Before analysis the aqueous solution made from the dried sample was applied onto pre-baked quartz filters then it was dried again in a desiccator at room temperature for 12 hours.
3. Results and Discussion
3.1. UV-VIS Spectrophotometry of Synthetic HULIS
 The UV-VIS absorption spectra of the 3,5-dihydroxybenzoic acid, and of its reaction products after 25h and 50h reaction time are shown in Figure 1. The absorption of the original 3,5-dihydroxybenzoic acid is comparable with the absorption of the blank at wavelengths above 350 nm, whereas the solution of its reaction products absorbs up to about 650 nm. The featureless spectrum after 50h reaction time implies that the rate of conversion is close to completion since the absorption maximum at 304 nm has disappeared. It can be stated that the free radical reaction produces compounds with delocalized electron systems. The absorption characteristics of the compounds are very similar to those found in the aqueous extract of continental fine aerosol. The ratio of absorbances at wavelengths of 350 and 450 nm measured in urban and rural aerosol extracts were found to be 5.5 and 5.7, respectively [Havers et al., 1998]. These values were considered to be indicative of the low molecular weight of the HULIS. In the model solution the same parameter decreases from 7.1 to 6.0 as the reaction progresses, corresponding to an increase in molecular weight, largely up to the level of atmospheric HULIS.
3.2. Molecular Weight Distribution of the Synthetic HULIS
 The molecular weight distribution of the reaction products was studied by mass spectrometry. Figures 2a and 2b show the mass spectrum of the 3,5-dihydroxybenzoic acid and that of the products of the reaction in solution containing 3,5-dihydroxybenzoic acid, H2O2 and Fe3+, respectively. In the solution containing 3,5-dihydroxybenzoic acid only no change can be observed in the spectrum even after storage for one day at room temperature. The dominant peaks are the deprotonated molecule of the 3,5-dihydroxybenzoic acid (m/z 153) and a fragment at m/z 109 originating from the neutral loss of CO2 from the carboxyl group. An ion with m/z 169 and its fragment at m/z 125 can also be observed. On the other hand, in the spectrum of the reaction products (Figure 2b) a number of ions appear at larger m/z ratios, forming a semi-continuous distribution of ions up to about m/z 500. The relative contributions of the deprotonated 3,5-dihydroxybenzoic acid and its fragment become almost negligible. The charge state and fragmentation of atmospheric HULIS in ESI-MS have recently been studied by Kiss et al.  who have concluded that singly charged ions dominated in the mass spectra of HULIS though the presence of fragments and/or doubly or triply charged ions could not be excluded due to the unit resolution of the mass spectrometer. If it is assumed to be the case in these experiments (using the same instrumentation and settings) it can be stated that the reaction products consist of a large number of species of different molecular weights well below 1000 Daltons. This finding is analogous to those found for the HULIS isolated from urban and rural aerosol [Havers et al., 1998; Kiss et al., 2003]. Consequently the reaction itself may be oligomerization rather than polymerization. In atmospheric multiphase reactions, a vast number of precursors may participate in similar free radical oligomerization reactions yielding an array of species of a continuous molecular weight distribution.
 The course of the reaction was followed also by liquid chromatography. Figure 3 shows the chromatograms of the products of the reaction between 3,5-dihydroxybenzoic acid, H2O2 and Fe3+ at concentrations of 2 · 10−3 M, 10−2 M, 5 · 10−7 M, respectively, detected at the wavelength of 210 nm, after a reaction time of 50 hours. It is evident that several new compounds are formed after a reaction time of 50 hours at the expense of the 3,5-dihydroxybenzoic acid. The yield of the reaction is about 90% as estimated by the reduction of the concentration of the aromatic precursor. The polarity of the reaction products cover a wide range from the highly polar species at the front and less polar unresolved mixture of compounds at the rear of the chromatogram. In the chromatograms of the same solutions with detection at the wavelength of 425 nm (not shown) both highly polar and less polar compounds were found.
 The THM-GC/MS total ion chromatograms of the reaction products revealed that after 1 day reaction time 3,5-dimethoxybenzoic acid methyl ester was still an important peak in the chromatogram, but several other compounds were also detected. Among them a few aromatic species were identified, such as 1,2-dimethoxybenzene, 1,2,3-trimethoxybenzene. Interestingly, some saturated and unsaturated aliphatic dicarboxylic acid dimethyl esters were also found, such as butanedioic, 2-butenedioic, 2-pentenedioic, methylbutanedioic, methylenebutanedioic, 2,3-dimethylbutanedioic acid and ethylbutanedioic acid dimethyl ester. The relative amounts of these low molecular weight aliphatic dicarboxylic acid dimethylesters released by THM increased considerably after 7 days. It should be noted that from the above mentioned compounds the 3,5-dihydroxybenzoic acid (MW = 154), butanedioic acid (MW = 118), 2-butenedioic acid (MW = 116) (in their native forms) show an intense signal in the ESI mass spectra at m/z 153, 117, 115 respectively. The ion in the ESI mass spectra at m/z 169 can be attributed to the 3,4,5-trihydroxybenzoic acid (MW = 170), which appeared after 7 day reaction time. Since it can be explicitly inferred from mass spectrometric and liquid chromatographic analyses that larger molecular weight species formed with extensive delocalized electron systems in the reaction, it is unlikely that such aliphatic moieties could be standalone products of free radical reactions in solution. Nor can these species be thermal decomposition products of aromatic structures. A possible explanation is that condensed and partially oxidized (e.g., quinone-like) phenolic structures are crosslinked with short-chain aliphatic bridges which form by the oxidative cleavage of the phenolic ring. Low molecular weight aliphatic dicarboxylic acid dimethyl esters are common constituents of the THM products of natural humic matter [Lehtonen et al., 2000] as well as of continental fine aerosol [Gelencsér et al., 2000b; Subbalakshmi et al., 2000], so the structural similarity to such compounds is again confirmed.
Figure 4 shows the thermal profiles of the 3,5-dihydroxybenzoic acid and the reaction products formed after 1 day and 7 days in solution. The thermal profiles show that 3,5-dihydroxybenzoic acid can be easily oxidized, and after 1 day more refractory substances are formed. The ratio of the refractory species to the easily oxidizable compounds (estimated from integrated peak areas) increases up to 5:1 after a reaction time of 7 days, implying that over 80% of the organic compounds have been transformed into higher molecular weight refractory species. The striking similarity between the thermal profiles of the reaction products and those of rural fine aerosol is worthy of note [Gelencsér et al., 2000a].
 The UV-VIS spectra of the products of reaction of 3,5-dihydroxybenzoic acid with OH radicals generated in Fenton-type reactions are qualitatively similar to those found for humic and fulvic acids as well as for the aqueous extract of rural fine aerosol. Since the ratios of the initial concentrations of the reactants were similar to those typically prevalent in cloud droplets in a polluted continental scenario, it can well be assumed that such reactions do produce light absorbing organic matter in cloud droplets. Although in the experiments the reaction time was longer than the typical atmospheric lifetime of a cloud, the reactions likely resume in subsequent cloud cycles. In addition, we hypothesize that similar reactions also proceed in hydrated aerosols. Mass spectrometry revealed that the process is free radical oligomerization producing an array of species. The distribution of ions in the mass spectrum was similar to those found for humic substances and HULIS in continental fine aerosol. Liquid chromatographic studies confirmed that the light absorbing compounds formed in the solution covered a wide range of polarity. The reaction studied is by no means intended to truly represent multiphase processes occurring in the atmosphere, it is only a highly simplified approach to demonstrate the potential of such reactions to yield complex larger molecular weight products from even a single aromatic precursor. Nevertheless the results may imply that similar multiphase reactions from a large number of potential precursors can produce HULIS in the atmosphere and may account for its ubiquitous presence in continental fine aerosol. In addition, the significance of the light-absorption by HULIS is yet another issue which calls for further studies.
 The authors are grateful to the Hungarian Scientific Foundation (OTKA Project No T043578, F029610) and the NKFP Grant No. 3/005-2001 for their financial support, and the Institute of Applied Environmental Research at Stockholm University for the possibility of running ESI-MS experiments.