References

Sulfur isotope analysis of atmospheric aerosols is a well established tool for identifying sources of sulfur in the atmosphere, estimating emission factors, and tracing the spread of sulfur from anthropogenic sources through ecosystems. Conventional gas mass spectrometry averages the isotopic compositions of several different types of sulfur aerosol particles, and therefore masks the individual isotopic signatures. In contrast, the new single particle technique presented here determines the isotopic signature of the individual particles. Primary aerosol particles retain the original isotopic signature of their source. The isotopic composition of secondary sulfates depends on the isotopic composition of precursor SO 2 and the oxidation process. The fractionation with respect to the source SO 2 is poorly characterized. In the absence of conclusive laboratory experiments, we consider the kinetic fractionation of −9‰ during the gas phase oxidation of SO 2 by OH as suggested by Saltzman et al. (1983) and Tanaka et al. (1994) to be the most reasonable estimate for the isotope fractionation during gas phase oxidation of SO 2 (α hom =0.991) and the equilibrium fractionation for the uptake of SO 2 (g) into the aqueous phase and the dissociation to HSO 3 - of +16.5‰ measured by Eriksen (1972a) to be the best approximation for the fractionation during oxidation in the aqueous phase (α het =1.0165). The sulfur isotope ratio of secondary sulfate particles can therefore be used to identify the oxidation pathway by which this sulfate was formed. However, the fraction of heterogeneous and homogeneous oxidation pathway calculated is very sensitive to the isotope fractionation assumed for both pathways. With the new single particle technique, different types of primary and secondary sulfates were first identified based on their chemical composition, and then their individual isotopic signature was measured separately. Our samples were collected in Mainz, Germany, in an urban environment. Secondary sulfates (ammonium sulfate, gypsum, mixed sulfates) and coatings on silicates or organic aerosol dominated sulfate loadings in our samples. Comparison of the chemical and isotopic composition of secondary sulfates showed that the isotopic composition was homogeneous, independent of the chemical composition. This is typical for particles that derive from in-cloud processing. The isotopic composition of the source SO 2 of secondary sulfates was calculated based on the isotopic composition of particles with known oxidation pathway and showed a strong dependence on wind direction. The contribution of heterogeneous oxidation to the formation of secondary sulfate was highly variable (35%–75%) on day-to-day basis and depended on meteorological conditions.


Introduction
Particulate air pollution has been a severe problem since the onset of urbanization.Research has shown a clear connection between particulate air pollution and daily mortality (Spix et al., 1993;Pope et al., 1995;Daniels et al., 2000).EU regulations (Guideline 1999/30EG) limit the airborne particulate matter (PM 10 ) to a daily average of 50 µg m −3 .This limit is exceeded frequently at urban air quality monitoring stations, and legislators are planning to decrease these limits even further.Therefore, severe cuts in urban background aerosol concentrations will become necessary, and in order to devise effective control strategies, a quantitative assessment of sources is required.The sulfur isotopic signature of the source SO 2 is changed during homogeneous (gas phase) and heterogeneous (aqueous phase) oxidation.Provided that the isotopic composition of source SO 2 is known and no water-soluble primary sulfate acted as cloud condensation nuclei, the relative contribution of condensations of gaseous sulfuric acid onto the droplet and heterogeneous oxidation in the droplet can be calculated.The isotopic composition of source SO 2 can be estimated from particles that derive from gas to particle conversion, such as fine mode ammonium sulfate.al., 2004;Hueglin et al., 2005;Sillanpää et al., 2006;Beekmann et al., 2007) has shown that a significant portion of PM 10 consists of secondary aerosol formed by the condensation of gaseous precursors.Sulfur dioxide, the gaseous precursor of sulfate aerosol, is released as a result of anthropogenic activity (fossil fuel and biomass burning, 60-100 Tg a −1 ; all values expressed as mass of sulfur) and from natural sources (volcanic gases and dimethyl sulfide (DMS), 20-60 Tg a −1 ) (Penner et al., 2001).In central Europe, stationary sources account for ca.90% of all sulfur dioxide emissions (Lövblad et al., 2004).
Since the 1980s, the emission of SO 2 decreased drastically (∼90%) in Germany, resulting in a 90% reduction of ambient SO 2 concentrations.However, these drastic cuts in ambient SO 2 concentrations did not correspond to a similar decrease in SO 2− 4 concentrations (only ∼70% decrease).For some countries, e.g., France and the Czech Republic, observed discrepancies were even greater (∼80% decrease in SO 2 and only 50% in SO 2− 4 concentrations).The same holds for areas close to sources (i.e., urban areas; Lövblad et al., 2004).This nonlinear response of particulate sulfate concentrations to emission reductions has been widely noticed all over Europe (Irwin et al., 2002;Larssen et al., 2003;Hunova et al., 2004;Klein et al., 2004;Lövblad et al., 2004;Fowler et al., 2005).Possible explanations are changes in oxidation patterns, deposition rates, or long range transport.Sulfur isotope ratios can be used to elucidate oxidation pathways and identify sources of sulfur in the atmosphere, and this combined information can help in understanding possible reasons for the nonlinear behavior.
In this study, we examine the chemical and isotopic composition of individual aerosol particles collected in Mainz, Germany, using the Cameca NanoSIMS 50 ion microprobe to elucidate sources and oxidation processes of sulfur in the urban and regional atmosphere.(n( 34 S)/n( 32 S)) VCDT =0.044163 (Ding et al., 2001).Primary sulfate particles, such as sea salt, mineral dust, fly ash or industrial dust are directly emitted with sulfur in the form of SO 2− 4 .Therefore, the isotopic composition of primary sulfate particles can be interpreted directly as a source signature.Five particle types dominate primary particles: biological particles, mineral dust, industrial dust, resuspended road dust and fly ash.Sulfur in plant tissue mostly reflects the isotopic composition of the atmospheric input (dry and wet deposition), unless other sources such as artificial fertilizer or local geology dominate the sulfur input into soil (Krouse and Grinenko, 1991;Gebauer et al., 1994;Novak et al., 2000Novak et al., , 2001aNovak et al., , 2005b;;Zhao et al., 2003;Bol et al., 2005).The most common sources of sulfate in mineral dust are marine evaporites.The isotopic composition depends on the geological age of the deposit and δ 34 S varies between +10‰ and +30‰.It is impossible to distinguish industrial dust emitted during the processing of natural minerals (stone dressing, cement industry, mining of mineral fertilizer) from the isotopic composition of the deposit being industrially exploited.The largest deposits exploited in Germany have δ 34 S of ∼10‰ (Zechstein).The isotopic composition of fly ash depends on the technology applied during coal combustion.The isotopic composition of fly ash depends on the technology applied during coal combustion.For power plants using flue gas desulphurization, the δ 34 S of the fly ash is generally closer to the δ 34 S of the coal and more positive than the δ 34 S of the SO 2 emitted during the same combustion process.The isotopic composition of re-suspended road dust is expected to lie somewhere between that of primary minerals and atmospheric dry and wet deposition, which can form coatings on particles.In continental Europe the contribution of sea salt (δ 34 S=(20.7±0.3)‰;Krouse and Grinenko, 1991) and nss-sulfate produced by the oxidation of DMS (δ 34 S=+14‰ to +22‰; Calhoun et al., 1991;McArdle and Liss, 1995;Patris et al., 2000a, b) to the sulfur budget is negligible compared to anthropogenic emissions.In winter, the contribution of sea salt to aerosol loadings is easily overestimated due to re-suspension of road salt.
At first sight, the fractionation calculated by Leung et al. (2001) agrees well with measurements of stratospheric sulfate (Castleman et al., 1974).The data of Castleman et al. (1974) seemed to indicate that during the oxidation of SO 2 to sulfate in the stratosphere following the Mt.Agung eruption, Rayleigh fractionation occurred with 34 S being enriched in sulfate and SO 2 depleted in 34 S.However, recent research has shown that following the volcanic eruption a separation of the sulfur into two reservoirs carrying a mass independent isotope fractionation with opposing signs took place (Baroni et al., 2007) and UV induced photo oxidation has been suggested to explain the mass independent signature.Since the oxidation of SO 2 by OH is not responsible for the mass independent signature observed in the sulfate during the time period in question, this reaction is not the only reaction dominating the isotopic signature of the sulfate.Therefore, the simple Rayleigh fractionation during oxidation of SO 2 by OH proposed by Leung (2001) can no longer be used to interpret the dataset.
Currently, the best way to estimate the fractionation of the gas phase and aqueous phase oxidation is to look at seasonal trends in the isotopic composition of simultaneously collected SO 2 and SO 2− 4 and to evaluate the equation (δ 34 S SO42− −δ 34 S SO2 )=(1 The contribution of the gas phase oxidation (f hom ) varies from 0% (nighttime, arctic winter) to roughly 60% (noon/early afternoon on a bright summer day) of the total sulfate formed.Therefore, during winter more sulfate should be formed through oxidation in the aqueous phase, while during summer the importance of gas phase oxidation by OH should increase.The seasonal trends of the isotopic composition of simultaneously collected SO 2 and sulfate allow an estimate of the direction of the isotopic fractionation involved in both pathways.It has been observed that during summer months (more gas phase oxidation) the difference in the δ 34 S of SO 2 and sulfate (δ 34 S SO42− −δ 34 S SO2 ) is generally lower than during winter months (more aqueous phase oxidation) (Saltzmann et al., 1983;Mukai et al., 2001;Kawamura et al., 2001;Tichomirowa, unpublished data).Occasionally, the sulfate is depleted in 34 S compared to the SO 2 during summer months.The same holds for the comparison of the isotopic composition of throughfall (wet deposition of sulfate plus SO 2 from dry deposition on the leaves) and bulk precipitation (Groschekova et al., 1998;Novak et al., 2000;Zhang et al., 1998) with the throughfall occasionally showing a higher δ 34 S than the bulk precipitation at the same site during summer months.Even when only one of the species (SO 2− 4 or SO 2 ) was collected the seasonality encountered is similar for most sites in the northern hemisphere.In winter the δ 34 S of bulk sulfate increases, compared to the summer values at the same site (Caron et al., 1986;Niagru et al., 1987;Ohizumi et al., 1997;Alewell et al., 2000;Ohizumi et al., 2001).On the contrary, the δ 34 S of the remaining SO 2 during winter is typically lower than during summer (Novak et al., 2001).This is in line with an enrichment of the heavier isotope in the sulfate due to the increased importance of the aqueous phase oxidation in the winter months and depletion of the SO 2 .These seasonal trends support α het >1 for the aqueous oxidation pathway.The fact that δ 34 S SO42− −δ 34 S SO2 and δ 34 S(bulk precipitation)−δ 34 S(throughfall) is sometimes negative during summer months supports α hom <1 for the gas phase oxidation.
In the absence of any conclusive laboratory experiments the seasonality of the sulfur isotopic composition is the best way to estimate the direction of the isotopic fractionation during gas phase oxidation and aqueous phase oxidation.The numbers associated with both processes are far from certain.For the heterogeneous oxidation pathway only the fractionation during the uptake of SO 2 into the aqueous phase and the dissociation to HSO − 3 has been determined, that too under equilibrium conditions (α het =1.0165; Eriksen, 1972a, b).The effect of the terminating reactions such as oxidation by H 2 O 2 , O 3 , and metal catalyzed oxidation by O 2 , to name just the three most important ones, has never been properly assessed and equilibrium is typically not reached under atmospheric conditions.The fractionation of the gas phase oxidation pathway (α hom ) has not been determined experimentally at all.The absence of laboratory experiments that include the net effect of the reaction for both the formation of sulfate in the gas phase as well as for the aqueous phase oxidation makes the data interpretation a challenging task.Based on current scientific understanding, we consider the kinetic fractionation during the gas phase oxidation of SO 2 by OH as suggested by Saltzman et al. (1983) and Tanaka et al. (1994) to be the most reasonable estimate for the isotope fractionation during gas phase oxidation of SO 2 (α hom =0.991) and the equilibrium fractionation for the uptake and dissociation measured by Eriksen (1972a) the best approximation for the fractionation during oxidation in the aqueous phase (α het =1.0165).
In order to attribute SO 2 emissions to their source, the isotopic composition of the SO 2 sources must be known.Until the application of more advanced technology, the sulfur isotopic composition of SO 2 emitted during combustion  Buzek et al., 1991;Krouse and Grinenko, 1991;Querol et al., 2000;Bericnik-Vrbovsek et al., 2002).However, the introduction of flue gas desulphurization technology changed this relationship.Before the introduction of this technology, Pichlmayer et al. ( 1998) reported an isotopic composition of δ 34 S SO42− =+6‰, similar to that of the coal (δ 34 S coal =∼+8‰) for emissions from coal burning in Poland.In contrast, δ 34 S of SO 2 emissions from a Polish power plant employing flue gas desulphurization technology is 13‰ more negative than the coal used in the combustion processes (δ 34 S coal =∼+8‰, δ 34 S SO2 emissions =∼−5‰; Table 1: Derda and Chmilewski, 2003).As a result of the widespread use of flue gas desulphurization, the isotopic composition of the fuel can no longer be used as an indicator of the source signature of anthropogenic SO 2 .Instead, the isotopic composition of gaseous emissions needs to be characterized directly at the source.2).Fields and gardens are located to the west, while the city of Mainz and the urban Rhine-Main area are located to the east of the sampling site (Fig. 2).A municipal garbage combustion plant emitting ∼25 mg of SO 2 per m 3 of flue gas is located 4 km north of the sampling site.Industrial activity is located mainly along the Rhine River to the north and east of the sampling site.Several measurement stations monitor the air quality in the city, including meteorological data as well as measurements of SO 2 , O 3 , PM 2.5 , PM 10 and soot, to which our data can be compared (Fig. 2; Table 3; Landesamt für Umwelt, 2005).Samples were collected on gold coated 47-mm-diameter Nuclepore ® polycarbonate filters with 0.4 µm pore sizes.The filters were placed in a stacked filter unit operated with one stage only.The start and stop time of each individual sample are specified in Table 2.After sample collection, the filters were placed in individual Petri-slides, wrapped in aluminum foil and stored in a desiccator.
Backward trajectories were calculated using the vertical motion model in the HYSPLIT4 (HYbrid Single-Particle Lagrangian Integrated Trajectory) program (Draxler and Hess, 1998;Draxler and Rolph, 2003) with the FNL meteorological database at NOAA Air Resources Laboratory's web server (Draxler and Rolph, 2003).Back trajectory calculations were started 10 m above ground level (Fig. 3).

Classification of particles based on chemical composition
Prior to ion microprobe analysis, the samples were characterized by scanning electron microscopy (LEO 1530 FESEM) operating at an accelerating voltage of 10 keV, equipped with were sampled at predefined equidistant spots.Whenever the predefined spots were located within a particle, the particle was counted and its size and chemical composition were measured.Typically more than 500 particles of each sample were examined at a magnification of 6000x.Based on the 2-D surface area of each particle was measured by counting the number of pixels it occupied in the digital secondary electron image and converted to µm 2 (pixel size at 6000x=111 nm).
From the 2-D surface area of the particle, the particle equivalent diameter was calculated.The equivalent diameter is the diameter of a spherical particle occupying the same area as the analyzed particle.Only particles with an area >80 pixels were considered for sizing to ensure good accuracy for the estimated equivalent diameter (Gwaze et al., 2006).The size cut off is 1 µm (pixel size 111 nm).In order to retrieve the volume and mass of particles, the height of the particles was ascertained.Particles typically lie on their flat side.Therefore, the height of larger particles was much less than the 2-D diameter.Based on manual analysis of numerous particles, the typical height was determined to be half the 2-D diameter for particles 1 µm<x<5 µm.The average height of particles >5 µm did not exceed 2 µm.
The approximate composition of each particle was estimated based on an EDX analysis of seven of the following elements: C, N, Na, Mg, Si, P, S, Cl, K, Ca and Fe.The energy windows were chosen for each sample individually, based on the elements with the highest abundance in the sample.The X-ray spectra were acquired for predefined equidistant spots.The acquisition time was fixed at 2 s.
Sampling regular spots is an established method to quantify the phase composition of samples with randomly distributed particles (Amelinckx et al., 1998).To avoid multiple sampling of the same particle, the distance between the spots has to be greater than the Feret's diameter of the largest particle.Whenever this criteria is fulfilled, the probability of acquiring an EDX spectrum of a particle of particular size and chemical composition is directly proportional to the total filter area covered with particles of that size and chemical composition and, therefore, to the number of the particles.This method allows fast quantification of the abundance of different particle types in the samples.To ensure that the distance between the spots is larger than the Feret's diameter of the particles the grid chosen for EDX data acquisition and image analysis was varied according to the particle size range investigated.The grid chosen for data analysis was 10 µm for particles with a Feret's diameter <10 µm, 20 µm for particles with a Feret's diameter between 10 µm and 19 µm and 50 µm for particles ≥20 µm in diameter.
The background contribution of the empty filter to the EDX spectrum of individual particles was estimated for each sample and energy window separately using the upper (Q u ) and lower (Q l ) quartile values of the raw signals of that energy window by applying robust statistics as , which is equivalent to a 3 sigma outlier limit (Stoyan, 1998).The background signal was then subtracted from the particle signal.
Chemical signals of particles below the detection limit of the image analysis (<1 µm) were frequent.Numerous particles >1 µm were only identified by image analysis (based on the contrast of the SEM image).EDX analysis of these particles did not show any signal for the chosen energy channels.For the other particles, after background correction, the X-ray intensities were normalized to the sum of intensities detected for the particle.The relative intensities for the major elements detected were used as a proxy for the particle composition.Particles were classified into different groups based on their chemical composition and on the characteristics of different particle types observed in other studies (Xhoffer et al., 1991;Ebert et al., 2000;Mamane et al., 2001;Li et al., 2003;Sobanska et al., 2003;Ro et al., 2004;Niemi et al., 2005).As the main objective of this research is the analysis of sulfur isotope ratios, particles that contained sulfate were treated separately (see Sect. 4.1).Each particle chosen for sulfur isotope analysis was documented individually with a picture taken before and after analysis along with a full x-ray spectrum.Particles identified as ammonium sulfate based on the spectrum acquired during the automatic run were only documented after NanoSIMS analysis, because damage by the electron beam can alter their isotopic composition (Winterholler et al., 2008).

Isotope analysis of individual particles with the Cameca NanoSIMS 50
The sulfur isotope measurements were done with the Cameca NanoSIMS 50 ion microprobe at the Max Planck Institute for Chemistry in Mainz (Hoppe et al., 2005;Gröner and Hoppe, 2006;Hoppe, 2006).The high lateral resolution (<100 nm for Cs + primary ions) coupled with a high transmission of secondary ions for isotope measurements of the light-to-intermediate-mass elements and multi-collection capabilities (up to 5 isotopes can be analyzed simultaneously) make this instrument the only one capable of analyzing sulfur isotope ratios on individual aerosol particles (Winterholler et al., 2006(Winterholler et al., , 2008)).
The data in this study were obtained in multi-collection detector mode by sputtering the sample with a ∼1 pA Cs + primary ion beam focused onto a spot of ∼100 nm diameter.
The primary ion beam was scanned over 2×2 µm 2 around the center of individual grains.Each analysis consisted of integration of secondary ion signals over 1200 cycles of 1 s each, preceded by 600 s of pre-sputtering.Energy centering was used to compensate for charging.Secondary ions of 16 O −32 S − , 33 S − , 34 S − and 36 S − were simultaneously detected in five electron multipliers at high mass resolution.The detector dead time is 36 ns and the S − count rates were corrected accordingly.Low-energy secondary ions were collected at a mass resolution sufficient to separate 33 S − from the 32 SH − interference.The energy slit was set at a bandpass of ∼20 eV and the transmission was set at ∼15-20% (specific setting of entrance, aperture, and energy slits).Here, we concentrate on the measured 34 S/ 32 S ratios because, due to the low isotopic abundances of 33 S and 36 S, the resulting errors of 33 S/ 32 S and 36 S/ 32 S ratios in single particles are large.The grain size and matrix dependence of the instrumental mass fractionation (IMF) were corrected based on the equivalent diameter and chemical composition measured for the respective particle in the SEM according to the method described in Winterholler et al. (2008).The necessity to correct for the size of the particles is caused by charging.Since the size determination of particles in the SEM is very accurate, this is a simple and straightforward correction, which is relevant mainly for coarse mode particles.Matrix dependent instrumental mass fractionation occurs during sputtering and ionization.Winterholler et al. (2008) found a linear relationship between the ionic radius of the cation (i.e., the chemistry) and the matrix specific instrumental mass fractionation for different sulfate salts.Riciputi et al. (1998) showed that the IMF of fine grained mixed samples, which contain two phases on a spatial scale smaller than the primary ion beam, can be accurately corrected using coarse grained standards of the individual phases.The instrumental mass fractionation relative to BaSO 4 has been established for most sulfates relevant for atmospheric research (Winterholler et al., 2008).Correction of pure sulfate particles and "internally mixed" particles in which the sulfate containing phases are clearly separated such as the aged sea salt particle in Fig. 5 containing sodium chloride, gypsum and sodium sulfate as separate phases, is straightforward.This particle is a classical example of a particle that is "internally mixed" from an aerosol point of view, but "externally mixed", i.e., separated into distinct components on the spatial scale relevant for IMF correction in the NanoSIMS.IMF correction of particles contain several cations in the same sulfate (e.g.glauberite, Na 2 Ca(SO 4 ) 2 ) is difficult.For such particles (part of the particles in group 6, mixed sulfates, <5% of the total particles) we calculated the IMF of the mixture based on the chemical composition of the respective particle as a linear mixture of the IMF of the pure salts of the major cations.
The instrumental mass fractionation for each session was determined using two BaSO 4 standards (IAEA SO-5 and SO-6, Isotope Hydrology Laboratory of the International Atomic Energy Agency, Vienna, Austria).Individual particles of Table 4. Instrumental mass fractionation factors for 34 S/ 32 S measured with the NanoSIMS and average diameter of the standard particles on which instrumental mass fractionation was determined.When the instrumental mass fractionation is determined on particles pressed into gold substrate, no grain size correction is necessary.BaSO 4 true is the calibrated isotope ratio of BaSO 4 based on delta values of 0.5‰ for IAEA SO-5 and 34.2‰ for IAEA SO-6 and a n( 34 S)/n( 32 S)) VCDT =0.044163 (Ding et al., 2001).BaSO 4 SIMS is the measured N ( 34 S)/N ( 32 S))-ratio measured by SIMS.both standards were put on two gold coated Nuclepore filters with the help of a micromanipulator and were analyzed along with the samples (Table 4, Fig. 4).The uncertainty of isotope measurements on individual aerosol particles is comparatively large due to a poor grain-to-grain reproducibility observed on standards (Winterholler et al., 2008).The grain-to-grain reproducibility of standard grains with identical chemical and isotopic composition is typically around 5‰ for micron-sized grains and between 2 and 5‰ for submicron-sized grains.This grain-to-grain reproducibility or residual error (σ R ) is determined by subtracting the average counting statistical error σ P ,m from the standard deviation σ of the measurements performed on the BaSO 4 standard (σ R = σ 2 −σ 2 P ,m ).While calculating the total measurement error (σ T ) of a measurement on an individual grain the residual error (σ R ) is added to the counting statistical error (σ P ) of the individual grain (σ T = σ 2 R +σ 2 P ).However, the accuracy observed upon averaging measurements of several grains is typically 2‰ for standards (Winterholler et al., 2008) as well as for sea salt aerosol particles (Winterholler et al., 2006).

Chemical classification of aerosol particles
The approximate chemical composition of each particle was derived from the EDX spectra and used to group particles into 10 groups.Oxygen and carbon were present in the filter background and were, therefore, excluded from data analysis.Table 5 lists the semi-quantitative chemical composition of each group.Typical particles and EDX spectra of all groups except Group 9 (other particles) and Group 10 (identified by image analysis only) are shown in Fig. 5.   Sea salt particles (Group 1) were recognized by high intensities of sodium and chlorine.Sea salt particles age in the atmosphere by reaction with H 2 SO 4 , SO 2 , HNO 3 and other nitrogen components giving rise to Cl depletion and sulfate/nitrate formation (Sievering et al., 1991;Mamane and Gottlieb, 1992;Zhuang et al., 1999;Laskin et al., 2003;Hoffman et al., 2004;Hwang and Ro, 2006;Saul et al., 2006).Aged sea salt particles were treated separately (Group 2).These particles typically contained >7.5% of sulfur and, therefore, significant amounts of non-sea-salt sulfate (nsss).
Silicon bearing particles (SiO 2 >6%) with or without variable amounts of Na, Ca, K, Mg and Fe were considered to be quartz, clay or aluminosilicate.Silicon bearing particles can be of natural (mineral dust, erosion of soil) as well as of anthropogenic origin (fly-ash).Both particle types were grouped into the same group (Group 3) during automated analysis but treated separately during isotope analysis.Almost all atmospheric particles can obtain a sulfur coating by condensation of SO 2 and/or H 2 SO 4 .Some mineral dust particles even react with sulfuric acid (Krueger et al., 2005).Silicate particles with sulfur coating were treated separately (Group 3a).In a similar manner, silicates (Si>6%) that acquired a nitrate coating (N>6%) during atmospheric processing, or were mixed with sea salt (Cl>6%) were assigned a separate group (Group 3b).S-only particles, i.e., particles that showed no significant signal for elements other than S (S>95%) were considered to be secondary sulfates formed from gaseous SO 2 , i.e., sulfuric acid or ammonium (bi)sulfate (Group 4).As oxygen was not analyzed, S was considered to be SO 2− 4 except if it was associated with iron (FeS 2 ).Unfortunately, gold interferes with sulfur in the EDX spectrum, making high background correction necessary.Small S-only particles were, therefore, missed by single particle analysis.This missing fine mode ammonium sulfate was quantified during bulk analysis of Atmos.Chem.Phys., 8, 2008 www.atmos-chem-phys.net/8/7217/2008/Fig. 5. SEM images and typical EDX spectra for all particle groups (except 9 and 10) and images of selected primary particle.the aerosol samples.Ammonium sulfate/sulfuric acid particles (Group 4) derive from gas to particle conversion (Group 4a) and/or in-cloud processing (Group 4b).Ammonium sulfate particles that went through in-cloud processing were assigned to Group 4b based on the following three criteria.Firstly, ammonium sulfate in the form of dried droplets was assigned to this group.Secondly, coarse mode ammonium sulfate particles (2.5-15 µm) were considered to be formed by in-cloud processing based on their large size and spherical shape.Thirdly, ammonium sulfate particles <2.5 µm were assigned to this group if their isotopic composition agreed within the analytical uncertainty with that of other secondary particles in the respective sample that were known to have been homogenized by in-cloud processing (i.e., Group 3a, 5 and 6).Calcium sulfate particles were identified by the absence of all elements other than Ca and S in the EDX spectrum (Group 5).As oxygen was not analyzed, S was considered to be SO 2− 4 .Primary gypsum particles have natural (soil, mineral dust, fractional crystallization of sea salt) as well as anthropogenic sources (flue gas desulphurization, metal and cement industry and road dust) (Hoornaert et al., 1996;Li et al., 2003).Reactions between sulfuric acid and CaCO 3 or Ca-feldspars can result in the formation of secondary gypsum (Foner and Ganor, 1992) on coarse mode particles.Cloud processing leads to the formation of secondary gypsum in the form of large needles (Fig. 6, Sample 8) or fine particles (Fig. 5, Gypsum) (Andreae et al., 1986).
All particles containing sulfur that could not be grouped into any of the above groups were referred to as mixed sulfates (Group 6).This group included sulfate particles with more than one cation.The most frequent particles were particles with Na and Ca or K and Ca as cations.Other parti-cles in this group included sodium sulfate and potassium sulfate.Sulfide minerals (FeS 2 ) did not contribute significantly to any of our samples and were excluded from NanoSIMS analyses.
Particles with a relative intensities of Ca or Ca+Mg higher than 90% (Group 7) were considered to be CaCO 3 , as oxygen and carbon were not analyzed.The sources of these particles are soil erosion and construction activities (McGee et al., 2003), limestone mining (Lei et al., 2004), cement production (Abdul-Wahab et al., 2005), flue gas desulphurization, glass and fertilizer production and metal industries (Hoornaert et al., 2003).In Mainz and Wiesbaden there are two cement production facilities, north and east of the sampling site.Limestone is mined south east and lime malm brick north of our sampling location (Fig. 2).Glass as well as fertilizer producing industries are located in Mainz, northeast of the sampling site.
Particles containing Fe>90% but no Cl, Si or S were considered to be iron oxides or oxyhydroxides, all of which are soil minerals (Group 8).
All particles that could not be classified into any of the above mentioned groups were grouped together (Group 9).These were secondary aerosol particles for which no sulfur was detected.Some were nitrates and phosphates, while for others Na, K and/or Ca were detected, but no anions.These particles might be oxides or hydroxides.Moreover, particles for which only one element was above the detection limit were assigned to this group.
Particles identified by image analysis only, but without any detectable EDX signal, (Group 10) included pollen grains and other biological particles, soot and secondary aerosol.Particles without characteristic EDX signal <1 µm were not analyzed.

Chemical composition of aerosol samples
The contribution of the different particle types for the different samples (Fig. 7) collected in August 2005, are as follows: Sample 1 (Figs.6 and 7) was characterized by dried droplets and thin films often with secondary crystals in the fine mode <3 µm (Group 4, 5, 6, 9 and 10: 95%).In the coarse mode, biological particles (Group 10: 38%), and mineral dust (Group 3, 3a, 3b, 7, 8 and 9: ∼50%) were present.Dried droplets, thin films and secondary crystals within droplets were assigned to Group 4, 5, 6, 9 or 10 depending on their chemical composition, mainly for the sake of instrumental mass fractionation correction during sulfur isotope analysis.However, these distinctions can be somewhat arbitrary.Different crystals formed by fractional crystallization from a droplet on the filter may require separate instrumental mass fractionation correction.Nevertheless, they impacted on the filter as one (liquid) particle.Vester (2006) assigned all these particles to one group termed "complex secondary aerosol".
Samples 4, 5 and 7 showed the highest contribution of mineral dust to both fine and coarse mode particle loadings.
The aerosol mass calculated from single particle analyses was compared to measurements by the state agency for environmental protection (Landesamt für Umwelt, Wasserwirtschaft und Gewerbeausicht Rheinland-Pfalz).PM 10 is measured at several sites in Mainz by Beta-Absorption, and has been reportedly corrected to be consistent with the standard procedure DIN EN 12341 (1998), which is a gravimetric analysis at (50±5)% relative humidity and (20±1) • C after 48 h conditioning.PM 2.5 and soot are monitored only at one site (Table 3).The difference PM 2.5−10 estimated from our data agreed within ±10% with PM 2.5−10 reported for the monitoring station Mainz Zitadelle for those samples where Mainz Zitadelle is located upwind of our sampling site (Sample 4 and 5) and within ±30% for other wind directions (Table 3), while PM 1−2.5 calculated from the single particle analysis under ultra high vacuum conditions was approximately 80% lower than PM 2.5 at 50% relative humidity at the station upwind from our sampling site (Table 3).There are two reasons for the PM 1−2.5 estimated by single particle analysis being lower than the bulk measurements.Firstly, the automated procedure chosen for characterizing the aerosol focused on identifying sulfates for sulfur isotope analysis and missed particles <1 µm without characteristic EDX signal, such as secondary organic aerosol and soot particles, which were not relevant for this study.Secondly, PM 2.5 at 50% RH contains water (10%-30%; Hueglin et al., 2005) which is absent under the ultra high vacuum conditions during SEM analysis.

Isotopic composition of different types of sulfate aerosol particles and bulk samples
Chemical analysis of aerosol collected in Mainz led to the identification of six groups of sulfate-containing particles.The contribution of each of these groups to the sulfate content of each sample was calculated based on results from single particle and bulk analyses (Table 6).
The isotopic composition of each group was measured by NanoSIMS (Table 6).Details of all analyses are listed in When comparing the isotopic composition of chemically different groups of secondary sulfates within the same sample, a sticking feature is, that for five out of six samples, the isotopic composition of secondary gypsum (Group 5), mixed sulfate particles (Group 6), sulfur coatings on silicates (Group 3a) and ammonium sulfate (Group 4b) agree with each other within the analytical uncertainty.Thus, irrespective of the chemical composition, precursor SO 2 and oxidation process that might have lead to the formation of different secondary aerosol particles, all secondary particles in the same sample show a uniform isotopic signature.This is only possible if all of these particles were formed from droplets that had been isotopically homogenized by frequent in-cloud processing.The weighted averages of particles from Groups 2, 3a, 4b, 5 and 6 are δ 34 S=(19±3)‰, δ 34 S=(19±3)‰, δ 34 S=(4±2)‰, δ 34 S=(15±1)‰, and δ 34 S=(8±3)‰ for Samples 1, 2, 5, 7 and 8, respectively.Sample 4, for which the isotopic composition of different secondary particles differed, was collected on a day with low relative humidity (Sample 4; Table 2).In Sample 4 the isotopic composition of sulfur coatings on silicates (Group 3a and 4b, δ 34 S=(1±1)‰) differed from the isotopic composition of secondary gypsum and mixed sulfate particles (Groups 5 and 6, δ 34 S=(10±2)‰).There are two explanations why the δ 34 S of sulfur coatings on silicates and ammonium sulfate was lower than that of other particles.Firstly, the contribution of heterogeneous oxidation to the formation of these particles might be lower.Secondly, different precursor SO 2 might have lead to the formation of these coatings.
Ammonium sulfate/sulfuric acid particles (Group 4) derive from gas to particle conversion (Group 4a) and/or incloud processing (Group 4b).The isotopic composition of ammonium sulfates derived from gas to particle conversion and in cloud processing typically differed by 18‰.For particles <2.5 µm the fraction formed by gas to particle conversion (typically 65% of the total mass of Group 4) was established based on the number of ammonium sulfate particles in this size range assigned to Group 4a and Group 4b, respectively.Only the isotopic composition of particles deriving from gas to particle conversion was used to estimate the isotopic composition of source SO 2 (Fig. 1).The contribution of ammonium sulfate particles deriving from gas to particle conversion to the total sulfate mass in the sample was high (∼40% of the total sulfate mass) only for Sample 2. This sample was collected on a day with low relative humidity.Typically only <20% of total sulfate mass was found in particles deriving from gas to particle conversion only.The rest was homogenized by in cloud processing (60%-95%) or was present in the form of primary sulfates.As the lifetime of SO 2 with respect to oxidation by OH is at the order of 10 days, the amount of sulfur found in particles presumably formed by gas to particle conversion might look a bit high at first sight.However, except for Sample 7, samples were collected on dry and sunny days, favoring gas-phase processes.
Table 6.δ 34 S values of different particle types in different samples.The semi-quantitative chemical composition was characterized by EDX.Primary and secondary gypsum particles and silicates and fly ash were distinguished based on particle morphology during manual SEM analysis.The δ 34 S of individual particles was measured by NanoSIMS.Errors are 1σ and include the standard deviation of the isotopic composition caused by the presence of different oxidation pathways/different sources in separate particles within the same particle group (i.e., the error of the weighted mean is multiplied by √ (χ 2 ) for χ 2 >1) and, therefore, includes the natural variability of the sample.The error of an individual analysis is typically 7‰ due to inherent limitations in the grain-to-grain reproducibility and the counting statistical limitations imposed by small grains.Errors <7‰ indicate a very low natural variability between different particles in the same group.Errors >7‰ indicate large differences between different particles in the same group, e.g., pollen grains in Sample 1. f SO42− : fraction of total sulfate mass contributed by the respective group.Group #2 = aged sea salt, group #3a = coatings on silicate particles, group #4a = ammonium sulfate/sulfuric acid particles from gas to particle conversion, group #4b = ammonium sulfate/sulfuric acid particles from in-cloud processing, group #5 = gypsum, group #6 = other sulfates/mixed sulfates.

Group
Sample 1 Sample 2 Sample 4 Sample 5 Sample 7 Sample 8 Gypsum particles can be of either primary or secondary origin.Primary gypsum particles were typically coarse mode particles (Fig. 5, Group 5).The isotopic composition of primary gypsum particles was δ 34 S=(17±2)‰ for Sample 7 and particles were associated with silicates, suggesting soil minerals as the origin of primary gypsum in this sample.The 34 S/ 32 S ratio agrees with the isotopic composition expected for soil minerals.The isotopic composition of fly ash (δ 34 S nsss =(25±5)‰) (Fig. 8, Table 6) from a northwestern wind direction indicated that the isotopic composition of gypsum formed in the fumes of this emission source cannot be distinguished from natural sources, such as fractional crystallization of sea salt or soil minerals.Primary Ca-phosphate (Sample 5) with an isotopic composition of δ 34 S=(23±5)‰ most likely originated from fertilizer production located north east of our sampling site.Unaltered sea salt (Group 1) particles were absent in our samples.Particles classified into this group by automated single particle analyses always showed reactions with sulfuric acid and formation of nss-sulfate upon visual inspection (e.g., the particle shown in Fig. 5, aged sea salt).The contribution by these particles to the total sulfate mass of the individual samples was minor (0%-5%).The isotopic composition of aged sea salt particles was measured as δ 34 S=(12±7)‰ (Table 6).The isotopic composition measured on biological particles averages the isotopic composition of plant sulfur and fine mode particles or coatings on the surface of the particles, and ranged from δ 34 S=(9±10)‰ to (26±5)‰.
The bulk isotopic composition of each sample was calculated based on the isotopic composition of each group and the fraction that it contributed to the total sulfate mass: and the error of the calculated bulk composition is Missing measurements were taken as 0 with an error of ±20‰.This error spans the full range of values expected for anthropogenic emissions.The gypsum needle associated with the fly ash was sputtered away completely, while the rest of the fly ash was resistant enough to survive analysis.
The δ 34 S value of bulk sulfate in air masses reaching Mainz from the north-western direction (Sample 2, 7 and 8: δ 34 S nsss =(13±1)‰) was higher than that of bulk sulfate in air masses reaching Mainz from an eastern direction (Samples 4 and 5: δ 34 S nsss =(5±2)‰).Sample 1 was collected on a sunny day, in the aftermath of rainfall that occurred in the previous night, and has the lowest particle and sulfate loadings.This sample likely represents local sulfur sources and has an isotopic composition of δ 34 S nsss =(10±2)‰.

Isotopic composition of source SO 2
The isotopic composition of secondary sulfates depends on two factors -the isotopic composition of the source SO 2 , and the oxidation process responsible for oxidizing SO 2 to SO 2− 4 .In order to unambiguously interpret the measurements, one of these two factors needs to be known, i.e., for interpreting sulfur isotope data of secondary sulfate in terms of the source composition of the SO 2 , the oxidation process needs to be known, or, alternatively, to understand the oxidation process the source composition has to be identified first.
Since in our case, both source composition and oxidation pathway are unknown, we have to make the assumption that we have correctly identified those fine mode ammonium sulfate particles that derived from gas to particle conversion as opposed to ammonium sulfate form by in-cloud processing (Table 6, Fig. 1).This assumption is justified, because our single particle data shows, that secondary sulfates that went through in-cloud processing are isotopically homogenized irrespective of their chemical composition (see Sect. 4.2).The large differences (∼18‰) observed between the isotopic composition of most fine mode ammonium sulfate particles (65% of fine mode ammonium sulfate; Group 4a in Table 6) and all other secondary sulfate particles including coarse mode ammonium sulfate particles (Group 2, 3a, 4b, 5 and 6 in Table 6) is strong evidence for the fact that they were formed by different atmospheric processes.Note that particles <2.5 µm with an isotopic composition that agreed within errors with the isotopic composition of other secondary particles homogenized by in-cloud processing (∼35% of the particles in this size range) were excluded from Group 4a and assigned to Group 4b.
The OH lifetime of SO 2 is of the order of 10 days, which means that more distant sources might contribute to the H 2 SO 4 (g).However, due to aqueous phase oxidation and SO 2 deposition the overall lifetime of SO 2 is on the order of two days (Fig. 9) and the isotopic composition of precursor  SO 2 can only be calculated for ammonium sulfate/sulfuric acid particles that have not been isotopically homogenized by inclusion into no-precipitating clouds.Therefore, the isotopic composition calculated for the precursor SO 2 is most likely influenced by local and regional sources.
The highest local SO 2 concentration was always observed at the measurement site located inside the city (Goetheplatz), northeast of our sampling site, pointing towards the existence of SO 2 sources inside the city.Previous research in Antwerp and Munich showed that the isotopic composition of SO 2 at an urban site is controlled by local sources rather than long range transport (Torfs et al., 1997;Mayer et al., 1995a).
Sample 1 has the lowest sulfate content of all samples.This sample was collected from 2 August 16:00 UT+2 to 3 August 15:00 UT+2.As it rained from 1 August into the early morning hours of 2 August the air was very clean.However, sunny conditions prevailed during most of 2 August and on 3 August, favoring gas to particle conversion.The isotopic composition calculated for the source SO 2 of this sample was δ 34 S=(−7±6)‰.
The isotopic composition for SO 2 reaching Mainz from northern direction was δ 34 S=(11±2)‰.The source of these emissions is unknown, but a municipal waste incineration plant is located north of our sampling site.Nevertheless, more distant sources such as large stationary sources in the Ruhr area, over which the back trajectories of the samples passed must also be considered.Over all the isotopic composition of SO 2 reaching the sampling site shows a clear dependence on wind direction (Fig. 10).

Contribution of homogeneous and heterogeneous oxidation to secondary sulfate formation in different types of aerosol particles
As the objective of this work is to understand the formation process of secondary sulfate aerosol, primary sulfate and secondary sulfate must be treated separately.Non-soluble primary particles such as primary gypsum particles (mineral dust, industrial dust and fly ash) and plant fragments are identified based on morphology and sulfur in such particles is neglected for the present analysis.These particles are typically not water soluble and, therefore, not internally mixed with secondary sulfate.The main source of water soluble primary sulfate is sea salt.The K/Na-ratio helps to evaluate sources of sodium in the urban atmosphere and the K/Na-ratio even of those particles termed aged sea salt is clearly higher than marine values, while e.g. for mixed sulfates the K/Na-ratio is similar to the K/Na-ratio produced by refuse incineration and car exhaust (Ooki et al., 2002).This and the low chlorine concentration of the samples indicate that the contribution of sea salt to our samples is very minor.Therefore, correction for sea salt sulfate is not necessary for particles of Group 3a, 4a, 4b, 5 and 6.
The contribution of heterogeneous oxidation to the formation of secondary sulfate in particles of Group 3a, 4a, 4b, 5 and 6 was calculated according to the formula f i,het = δ 34 S i − δ 34 S 4a / 0.0257 • 1 + δ 34 S 4a .(4) for each of the groups separately.
The error is The isotopic composition of fine mode ammonium sulfate particles that have not been homogenized by in-cloud processing (Group 4a) is considered to represent the isotopic composition of particles derived from homogeneous oxidation only.Assuming a fractionation in the δ 34 S of +16.5‰ with respect to the source SO 2 for the heterogeneous oxidation pathway and −9‰ with respect to the source SO 2 for the homogeneous oxidation pathway, the maximum difference between the two pathways is 25.7‰.The fraction of heterogeneous and homogeneous oxidation pathway calculated by this formula is very sensitive to the isotope fractionation assumed for both pathways.Any change of these numbers due to new experimental evidence will affect the fraction of the homogeneous and heterogeneous oxidation pathway calculated.The contribution of heterogeneous oxidation to the total secondary sulfate was based on the contribution of heterogeneous oxidation to the secondary sulfate on the individual group e.g.Group 5, and the fraction that each group contributed to the total nss-sulfate content of the sample The error of the estimate is As particles in Group 4a derive from homogeneous oxidation only, σ 4a,het is 0 by definition.
Including Group 4a, the contribution of heterogeneous oxidation to the formation of secondary sulfates was (102±26)%, (42±24)%, (60±15)%, (71±28)% and (44±16)% for Sample 1, Sample 2, Sample 4, Sample 5 and Sample 7, respectively.The highest contribution of heterogeneous oxidation was found in Sample 1, which was collected following a rainfall event and in which most particles were collected in the form of droplets, and Sample 5.The lowest contribution of heterogeneous oxidation was observed for a sample collected on a day when the daytime relative humidity was low (∼36%, Sample 2, Table 2).
In order to establish whether the nonlinear response to emission reductions coincided with a change in the relative contribution of the homogeneous and heterogeneous oxidation pathways to the formation of secondary sulfates, we compare our data to previously reported results.Current atmospheric chemistry models suggest that 24-56% of precursor SO 2 is removed by dry and wet deposition before oxidation and only 42-82% of precursor SO 2 is oxidized (Penner et al., 2001, Fig. 9;Alexander et al., 2008).Of the SO 2− 4 formed by oxidation of SO 2 , 64-90% is formed by aqueous oxidation and 10-36% by homogeneous oxidation.The contribution of heterogeneous oxidation to the formation of secondary sulfates can be estimated by simultaneous measurements of the isotopic composition of SO 2 and SO 2− 4 (Tanaka et al., 1994).Querol et al. (2000) measured the S isotope fractionation between the SO 2 emitted by a coal fired power plant in Spain and the SO 2− 4 derived from the oxidation of the SO 2 .In the plume the difference increased to 2.8‰.Numerous other observations in the 1970s and 1980s by Krouse and Grinenko (1991) showed that the oxidation of SO 2 is associated with an average 34 S/ 32 S fractionation of about +3‰ (range: −5.1‰-12.5‰).This implies a typical contribution of heterogeneous oxidation to the conversion of SO 2 to SO 2− 4 of ∼46% under ambient atmospheric conditions.In Central Europe, Mayer et al. (1995a) found no difference between the isotopic composition of SO 2 and sulfate in bulk precipitation in 1989 (∼35% heterogeneous oxidation), while Novak et al. (2001b) found an average difference of 4.1‰ between the isotopic composition of SO 2 and SO 2− 4 at several sites in the Czech Republic (averaged over the years 1992 to 1997) pointing towards ∼50% contribution of heterogeneous oxidation to the formation of sulfate.
The average differences between the δ 34 S values of SO 2 and SO 2− 4 observed in all these previous studies were lower than the average differences between SO 2 and SO 2− 4 in our samples collected in August 2005 ((6.7±2.4)‰).However, our results compare well with recent measurements by Tichomirowa et al. (2007), who found an average difference ranging from 4.3‰ to 12.7‰ between the isotopic composition of SO 2 and aerosol samples (termed dust) at two sites in Saxony (in 1997Saxony (in /1998)).Both results support an increase in the contribution of heterogeneous oxidation to the formation of sulfate from a typical contribution of ∼46% in the 1970s and 1980s to approximately 60-85% in recent years.Nevertheless, the fraction of aqueous phase oxidation estimated by the study of sulfur isotope ratios is at the lower end of the contribution of aqueous phase oxidation estimated by atmospheric chemistry models.
There are two possible reasons why such an increase in the efficiency of the heterogeneous oxidation pathway is the most likely explanation for this shift in the relative contribution of both oxidation pathways.Firstly, the oxidation of SO 2 by ozone is strongly pH dependent and emission reductions of all major acidifying compounds have lead to a decrease in the acidity of cloud droplets and precipitation from pH 4.4 in the early 1980s to pH 4.9 from 2000 to 2004 at all EMEP measurement stations in Germany (Klein et al., 2004).Oxidation by H 2 O 2 with an oxidation rate of 10 −8 Ms −1 dominates aqueous phase oxidation both at pH 4.4 and pH 4.9.However, the oxidation rate by O 3 increases by one order of magnitude from 10 −10 Ms −1 to 10 −9 Ms −1 when the pH changes from 4.4 to 4.9 for 0.2 ppb=[SO 2 (g)], 46 ppb=[O 3 (g)] and 0.6 ppb=[H 2 O 2 ] ( Lee and Thiemens, 2001).Consider the competitive rates of gas phase oxidation by OH and aqueous phase oxidation as a whole (i.e.oxidation by H 2 O 2 , O 3 and metal catalyzed oxidation by O 2 ) the overall importance of aqueous phase oxidation increases.Secondly, median ozone concentrations have increased during the aforementioned period (  al., 2004).Therefore, the nonlinear response of particulate sulfate concentration to emission reductions is not only caused by a shift from an oxidant limited system towards more complete oxidation closer to sources due to lower sulfur dioxide emissions, but also to a shift towards a higher fraction of heterogeneous oxidation.In fact, our results suggest that SO 2 emission reductions coupled with rising ozone concentrations lead to an increase in the oxidation capacity of the urban atmosphere.

Conclusions
The results of this study show that, despite limitations in precision, the NanoSIMS technique is a novel and useful tool for the isotope analysis of individual atmospheric particles, enabling us to compare the chemical and isotopic composition of individual aerosol particles.Given the range of S-isotopic ratios in aerosol bulk samples, the achievable precision and accuracy of a few per mil for the measurement of the 34 S/ 32 S ratio in individual aerosol particles is sufficient to investigate physical and chemical processes related to aerosol formation and transport.We found that the isotopic composition of sulfate and SO 2 at our site depended mainly on wind direction, suggesting a dependence on local sources.Different types of secondary sulfate particles were usually isotopically homogeneous, irrespective of chemical composition, except on days with extremely low relative humidity.
The contribution of heterogeneous oxidation to the formation of secondary sulfates was estimated to be typically around 60% and showed a dependence on meteorology.The comparison of our data to previous results in Central Europe (Mayer et al., 1995a;Novak et al., 2001b;Tichomirowa et al., 2007) indicated that the estimated contribution of heterogeneous oxidation to the formation of sulfate has increased from around 50% in the early 1990s to ca. 60-70% in 2005.The fraction of heterogeneous and homogeneous oxidation pathway calculated from the average differences between the δ 34 S values of SO 2 (g) and SO 2− 4 is very sensitive to the isotope fractionation assumed for both pathways.However, a change of these constants would affect all the datasets in a similar manner.The shift in the relative contribution of the two major oxidation pathways coincided with a strong decrease of SO 2 emissions, and might be partially responsible for the weaker response of urban PM 2.5 concentrations to the drastic decrease in the emission of gaseous precursors.
Future studies of the mass independent oxygen isotope fractionation of sulfate particles could confirm whether changes in the contribution of ozone to sulfate formation are taking place.

Fig
Fig.1.The sulfur isotopic signature of the source SO 2 is changed during homogeneous (gas phase) and heterogeneous (aqueous phase) oxidation.Provided that the isotopic composition of source SO 2 is known and no water-soluble primary sulfate acted as cloud condensation nuclei, the relative contribution of condensations of gaseous sulfuric acid onto the droplet and heterogeneous oxidation in the droplet can be calculated.The isotopic composition of source SO 2 can be estimated from particles that derive from gas to particle conversion, such as fine mode ammonium sulfate.

2
Isotope chemistry of natural and anthropogenic sulfur in continental Europe Sulfur isotope ratios are expressed in delta notation defined according to the equation given below (VCDT: Vienna Canyon Diablo Troilite) δ 34 S = δ 34 S VCDT = n 34 S /n 32 S Sample n 34 S /n 32 S VCDT − 1 (1)

Figure 4 .
Figure 4. BaSO4 standard grain illustrating the analytical procedure.Particles are documentedwith help of the SEM before (A) and after SIMS analysis (B).SEM conditions: EHT 10 keV, WD 9 mm, scale bar 2 µm.SIMS conditions: Field of view 2 µm x 2 µm, simultaneous collection of 16 O -, 32 S -, 33 S -, 34 S -and 36 S -, Cs + primary ions, 1 pA primary current, 100 nm beam diameter.The black square on SEM image B is the area were the filter material was sputtered away during analysis and indicates the exact position of the SIMS measurement field.

Fig. 4 .
Fig. 4. BaSO 4 standard grain illustrating the analytical procedure.Particles are documented with help of the SEM before (A) and after SIMS analysis (B).SEM conditions: EHT 10 keV, WD 9 mm, scale bar 2 µm.SIMS conditions: Field of view 2 µm×2 µm, simultaneous collection of 16 O − , 32 S − , 33 S − , 34 S − and 36 S − , Cs + primary ions, 1 pA primary current, 100 nm beam diameter.The black square on SEM image (B) is the area were the filter material was sputtered away during analysis and indicates the exact position of the SIMS measurement field.

Fig. 7 .
Fig. 7. Sample composition in % of total particle number (Na) calculated from single particle analysis in the SEM.Results are given for three size ranges: (a) Particles below the detection limit of the image analysis (<1 µm), (b) particles in the size range 1-3 µm and (c) coarse mode particles >3 µm.Fine mode ammonium sulfate is usually underestimated by single particle analysis.Particle numbers are estimated based on the bulk analysis of the respective samples.

Fig. 8 .
Fig. 8. Fly ash particles before and after SEM analysis.Working distance 11 mm, EHT 10 keV.The black square on the right SEM image is the area was the filter material was sputtered away during analysis and indicates the exact position of the measurement field.The gypsum needle associated with the fly ash was sputtered away completely, while the rest of the fly ash was resistant enough to survive analysis.

Fig. 9 .
Fig. 9. Relative contribution of different oxidation pathways and precursor deposition to SO 2 removal.The annual average of 12 global 3-D models (Penner et al., 2001) is compared with the relative contribution of gas phase and aqueous phase oxidation on mostly sunny August days in Mainz.Approximately 3% of SO 2− 4

Fig. 10 .
Fig. 10.Dependence of δ 34 S of source SO 2 on the wind direction.

Table 1 .
δ 34 S values of coal, oil, slag fly ash and SO 2 emissions of power plants in Europe.

Table 3 .
Comparison of PM 10 and PM 2.5 calculated from single particle analysis with PM 10 , PM 2.5 and SO 2 or soot (all reported in µg m −3 ) at several measurement stations in Mainz.Data for the measurement stations in Mainz were downloaded from http://www.luft-rlp.de.

Table 5 .
Average semi-quantitative composition of different particle groups in weight %.Group 1 sea salt is missing, as no particles of this group were encountered in samples collected in Mainz.N a : Number of particles analyzed in the respective Group.Element concentrations are semi-quantitative.The total counts of the EDX spectrum were normalized to 100% after background correction and used to estimate the semi-quantitative composition of the particles.Oxygen and carbon are present in the filter substrate and were not analyzed.S is assumed to be SO 2− 4 and Si is assumed to be SiO 2 .Cation concentrations close to 100% indicate carbonates, oxides or hydroxides.N.d.= not detected, n.a.not analysed.