The C-ToF-AMS was calibrated with monodisperse ammonium nitrate and sulfate before, during, and after the campaign in order to estimate relative ionization efficiencies (RIE) for nitrate and sulfate. The resulting calibration values can be found in Table .

A value of 0.5 for the collection efficiency (CE) was used for all flights . The main fraction of the measured aerosol mass consists of organic matter (see Sect. ), which has no clear effect on the CE. Furthermore, large amounts of nitrate that would lead to a higher CE were not encountered during the campaign .

Detection limits (DLs) were derived as 3 times the standard deviation from the background signal according to . Here, a time-dependent cubic spline function was used to determine a detection limit for each data point. This function was developed at the MPIC and is introduced and explained in more detail in a PhD thesis . For every data point of the background signal a third-order polynomial (cubic) function is calculated through the four neighboring points (two before and two after) while omitting the actual point. Applying this method, all trends from the background signal are excluded and just the short-term noise remains. A quantity R is introduced in the algorithm and characterizes the statistical spread of the noise. R is defined by the squares of the deviation between the omitted center point and the cubic function along a moving window. To relate this R to the standard deviation (σ) a proportionality factor of 18/35 is needed . The exact derivation of the proportionality factor will not be explained here, but qualitatively it accounts for the fact that not only are the points affected by noise but also the cubic function itself. Thus, R is larger than the standard deviation (σ). This calculation also provides a continuous σ from which the DL can be derived using Eq. (). DL=3⋅2⋅σ Averaged over a single flight, the detection limit is around 0.11 µg m-3 for organics, 0.02 µg m-3 for nitrate, 0.03 µg m-3 for sulfate, and 0.09 µg m-3 for ammonium. It should be noted here that ammonium can experience additional uncertainty due to interferences with water in the fragmentation table . A vertical profile of the averaged detection limits for all species can be found in the Supplement (see Fig. S1).

The oxidation state of the organic aerosol indicates the degree of photochemical aging. It can be determined using the correlation between the ratio of the signal at m/z 44 (mostly CO2+) to the total organic signal, f44, and the ratio of the signal at m/z 43 (mostly C2H3O+) to the total organic signal, f43 . These two ions are important tools to identify the photochemical aging of organic aerosol components in the atmosphere. Organic aerosol can be grouped into oxygenated organic aerosol (OOA) and hydrocarbon-like organic aerosol (HOA), whereby OOA is further classified into low-volatility OOA (LV-OOA) and semi-volatile OOA (SV-OOA) e.g.,. The difference between these two subcomponents is represented by the two ions at m/z 43 and m/z 44. These two ions, or more specifically f43 and f44, change during photochemical aging of organic aerosol, leading to higher f44 and lower f43 values for LV-OOA than SV-OOA. With increasing photochemical aging, organic aerosols of different origins become more similar in terms of chemistry . We use the ratio R44/43 (f44 divided by f43) to show the effect of photochemical aging on organic aerosol particles dependent on the altitude.

The tracer ion at m/z 82 (C5H6O+) is attributed to methylfuran and was identified to be related to IEPOX-SOA . Due to thermal vaporization and subsequent electron impact ionization in the C-ToF-AMS, isoprene photooxidation products in the aerosol are decomposed and lead to an increased signal at m/z 82 . A strong correlation is found with the ion at m/z 53, which corresponds to C4H5+ . Calculating the ratio of the signal at m/z 82 to the whole organic signal leads to the fraction f82. This fraction is identified as a tracer just for IEPOX-SOA, not for other isoprene-derived SOA from different reaction pathways e.g.,. A study by found background values of fC5H6O+ (ratio of C5H6O+ to the total organic signal) for different regions worldwide. For urban and biomass-burning-influenced regions an averaged background value of 1.7±0.1 ‰ is found. Areas strongly impacted by monoterpene emissions show higher averaged background values of 3.1±0.6 ‰. For ambient organic aerosol particles that are influenced by enhanced isoprene emissions and no extensive NO emissions with the presence of HO2, values of 6.5±2.2 ‰ have been reported . This enhanced value indicates a strong IEPOX-SOA influence. All the mentioned values are derived from high-resolution ToF-AMS (HR-ToF-AMS) data. For C-ToF-AMS data with unit mass resolution, the tracer f82 can be used, although interferences from ions other than C5H6O+ at m/z 82 are possible. Empirical parameterizations relating f82 and f44 for unit mass resolution are given by Appendix A, which will be used here for the further analysis.

Figure shows the normalized, measured mass spectrum for a period with high f82 during flight AC13 (for the flight notation, see ) conducted on 19 September 2014 at an altitude of 12.6 km. The averaging time for this spectrum was 1 min (15:54–15:55 UTC). The two peaks for determining IEPOX-SOA, m/z 82 and m/z 53, are labeled and clearly visible in the spectrum. For m/z 82 and m/z 53 the same RIE as for organics is applied . The calculated f82 from this spectrum is 27.4 ‰.

A method for the estimation of the mass concentration of IEPOX-SOA is reported by ; see Eq. (). For this calculation the mass concentration at m/z 82 [C(m/z82)], total organic mass concentration [C(Org)], a reference f82 value for IEPOX-SOA (f82IEPOXSOA), and a background value (f82Bg) are taken into account. The reference value for f82IEPOXSOA is set to 22 ‰ . The background value f82Bg can be determined with an empirical equation and depends on the influence of urban, biomass burning, or strong monoterpene emissions . For our data set we assume strong monoterpene emission influence and use Eq. () to determine f82Bg according to Appendix A. C(IEPOX-SOA)=C(m/z82)-C(Org)⋅f82Bgf82IEPOXSOA-f82Bgf82Bg=0.0077-0.019⋅f44 In Eq. (), C(m/z82) and C(Org) designate mass concentrations in units of µg m-3.

Several studies show that qualitative measurements and quantification of organic nitrates are of major interest e.g.,. Organic nitrates are decomposed during the evaporation and/or ionization processes in the C-ToF-AMS and, therefore, are divided into an organic and a nitrate signal . The following use of RONO2 refers to the nitrate content of organic nitrates, as the organic content cannot be estimated with the described methods. Due to observations during the measurement campaign, the analysis of organic nitrates is described as one part of this study. Four methods are applied to find out whether organic nitrates have been present during the measurements.

A first estimation of organic nitrates can be derived from the ratio of the nitrate-related ions at m/z 30 (NO+) and m/z 46 (NO2+). The signal at m/z 30 is mostly from NO+, but also the organic ion CH2O+ can contribute with a small amount . Such interferences at m/z 30 with CH2O+ are corrected in the evaluation software by the fragmentation table , but it is not possible to distinguish unambiguously between the NO+ and the CH2O+ ions with a C-ToF-AMS. The signal at m/z 46 is usually dominated by NO2+ ions . As organic interferences on the mass spectral signals at m/z 30 (interference from CH2O+) and m/z 46 (interference from CH2O2+) can occur in environments with high biogenic contribution and/or small nitrate concentrations, a correction according to was applied. The correction of both signals at m/z 30 and 46 is achieved by using correlated organic signals at m/z 29, 42, 43, and/or 45 derived by high-resolution measurements. The organic signals at m/z 29 (CHO+) and m/z 45 (CHO2+) are closest to those affected by the interference and used for the correction here. Equations () and () give the individual correction for the nitrate signal at m/z 30 and 46, respectively. The correction for NO+ includes the total signal at m/z 30, the default fragmentation correction from the air signal , and a correction coefficient that depends on the m/z used for the correction (Ai). As for m/z 30 the correlated organic signal at m/z 29 is used here, and the organic signal at m/z 29 (Org29) needs to be taken into account as well as the contribution of the isotopes of organic CO. For the correction of the nitrate fraction at m/z 46, a term which includes a correlation coefficient Bi and the organic signal at m/z 45 is subtracted from the signal at m/z 46. The correction coefficient Ai is in this case 0.215, and Bi is 0.127 (see Supplement to ). In the organic signal at m/z 28, 29, 30, and 45, the relative ionization efficiency (RIEOrg) is already applied and needs to be reversed for the correction of the nitrate signal. Nitrate fraction atm/z30:NO+=m/z30-0.0000136⋅m/z28-Ai⋅(Org29-0.011⋅Org28)⋅RIEOrg-Org30⋅RIEOrg.Nitrate fraction atm/z46:NO2+=m/z46-Bi⋅Org45⋅RIEOrg. The total nitrate signal is then calculated by adding both fractions. The final nitrate mass concentrations were reduced by 0.045 µg m-3 (STP), on average corresponding to an averaged reduction of 39 % of the initial nitrate mass concentrations. A comparison of the initial and finalized nitrate mass concentrations can be found in the supplement (see Fig. S2).

The ratio of NO+ to NO2+ is different for inorganic ammonium nitrate and organic nitrate. The ratio for inorganic nitrate is known from the ionization efficiency calibration with pure ammonium nitrate. For our instrument its value lies between 1.49 and 1.56 with a mean and standard deviation value of 1.52±0.03 and is derived from calibration measurements during the campaign. For organic nitrates the literature presents a range of ratios of NO+ to NO2+ that are higher than the ratio for inorganic nitrates and lie between 5 and 12.5 e.g.,.

The meteorological situation during the ACRIDICON-CHUVA campaign was quite similar for all days. Convection was dominating the daily weather and affecting every flight. The invariance of the meteorological situation is also visible in the temperature profile (see Fig. a), which barely shows any deviation. An overview of some flight details is provided in Table S1 in the Supplement.

The boundary layer (BL) height was determined with the help of the ambient temperature (Tambient), the virtual potential temperature (Θv), relative humidity with respect to water RHw, and aerosol number concentration for particle diameters larger than 20 nm (Nd>20nm) (Fig. ). During the whole campaign the vertical profiles of Tambient and Θv showed almost no deviation, indicating stable conditions for the measurement period. The daily evolution of the BL height was noticeable during the flights, which lasted typically 7 h. The maximum heights of the BL varied between 1.2 and 2.3 km, depending on the flight time. The highest BL heights were measured later in the afternoon. Over the whole measurement period, a mean height of the BL was found to be at 1.7±0.2 km (Fig. , horizontal dashed line; see also Fig. S3). For the binned vertical profile shown in Fig.  the BL height is not so obvious due to smoothing effects while averaging over all flights. The observed BL height is consistent with previous studies in the Amazonian dry season . Above the BL is the convective cloud layer, which reached altitudes of about 4–5 km during the campaign . The thermal tropopause was at an altitude of 16.9±0.6 km (mean and standard deviation) . Therefore, all flights of the ACRIDICON-CHUVA campaign were performed in the troposphere.

Figure shows the vertical profiles of median relative humidity with respect to water (RHw) and median aerosol number concentration for particle diameters larger than 20 nm (Nd>20nm). RHw decreases above the BL, showing a minimum with constant median values between 5 and 9 km. At higher altitudes, RHw increases again with altitude. The median values of aerosol number concentration are constant in the BL, decrease at middle altitudes with a minimum at 4 km, and rise strongly at altitudes above 7 km. This increase was interpreted as evidence for new particle formation at altitudes above 7 km by .

In Fig. , the vertical profiles of median mass concentrations of organics, nitrate, sulfate, ammonium, and black carbon given in µg m-3 (STP) are shown, with STP calculated for T=300 K and p=995 hPa from calibration measurements at the ground.

At all altitudes, the main fraction of the submicron particulate mass consists of organic matter. The highest aerosol mass concentration, in terms of total aerosol mass as well as median values for all species, is observed at lower altitudes between 0.1 and 4.5 km. This includes the BL (see Sect. ). An exception from that is nitrate, showing maximum median values at altitudes above 10 km. At middle altitudes (between 5 and 8 km) the mass concentrations of all shown species decrease rapidly. A different behavior of the species is observed at high altitudes between 8 and 14 km. The mass concentrations increase again with increasing altitude for organics and nitrate. For this altitude range, nitrate shows the highest median values. Although the median values decrease between 13 and 14 km, the 75th percentile range still indicates that high mass concentrations were encountered. In Sect.  the increase in nitrate mass concentration is discussed with respect to the potential existence of organic nitrates. In contrast to this, the mass concentrations of sulfate and black carbon are highest at lower altitudes, decline above 4.5 km, and stay constant at middle and high altitudes. Just between 13 and 14 km, sulfate median values show a slight increase again. At lower altitudes, the mass concentrations of ammonium are highest, above 4.5 km decreasing, and stay constantly low for the rest of the altitude range.

The vertical distribution of aerosol mass concentration allows the classification into three different regions. These are the lower troposphere (LT), ranging from 0.1 to 4.5 km; the middle troposphere (MT), covering altitudes between 4.5 and 8 km; and the upper troposphere (UT), which includes altitudes between 8 and 14 km (see Fig. ).

As described in Sect. , we calculated the correlation between f43 and f44. Figure a presents the data from all flights for the two different altitude regimes, LT (0.1 to 4 km) and UT (8 to 14.4 km), averaged over 2 min. The dashed lines represent the region where previous boundary layer aerosol data lie . The arrow illustrates the direction (upper left corner) in which data points are “moving” when photochemical aging occurs. The two different colors indicate the altitude dependency. The dark green markers represent data sampled in the LT, whereas the light green markers show data from the UT. Although there is an overlapping region of both, there is a difference between LT and UT. Most of the data sampled in the LT are located towards the upper left corner of the triangle, meaning that they are more oxidized. In comparison to this, data from the UT show different properties. Lower f44 and at the same time increased f43 values show a lower oxidation level of the organic aerosol. Also presented in Fig. a are the median values with the interquartile ranges for LT and UT, respectively. The median value for the UT has a higher f43 and a lower f44 than that for the LT. That means organic aerosol measured in the UT is significantly less photooxidized than in the LT. Thus, the organic aerosol particles in LT and UT are different from each other. Some of the LT organic aerosol may be transported to higher altitudes, but most of the organic aerosol measured in the UT must have a different source that is not located in the BL or in the LT. The possibility that substantial amounts of aerosol are transported from the BL into the UT has been ruled out by the study of , based on the absence of detectable amounts of black carbon in the UT (also see Fig. ) and other differences in the properties of aerosol in the LT and UT.

Previous field measurements that have been performed in the Amazon allow a comparison of the presented data set with measurements taken at the ground at two different stations (T3 and T0t). Station T3 is an open field ca. 70 km west of Manaus and has frequent pollution influence from this city, whereas T0t is in a near-pristine rainforest ca. 60 km northwest of Manaus. Thus, they provide anthropogenically influenced or natural measurement conditions, respectively. For further information on the research stations see and .

The data were collected during the AMAZE-08 and the GoAmazon2014/5 campaigns during the wet season (AMAZE-08: February–March 2008; GoAmazon2014/5: February–March 2014) and represent ground measurements . Figure b illustrates the median and interquartile ranges for the different data sets. Median values from the GoAmazon2014/5 campaign are similar to the median values derived from the ACRIDICON-CHUVA campaign sampled in the LT, showing that the organic aerosol is oxidized. In comparison, the AMAZE-08 campaign data differ, showing that the organic aerosol during AMAZE-08 was less oxidized. Our data for the UT differ from both these data sets, indicating again a different source for organic aerosol than in the LT. It should be mentioned here that there is a significant variability of m/z 44 (and f44) among different AMS instruments such that no quantitative comparison can be done among the different data sets shown in Fig.  .

Figure shows the vertical profiles of median and interquartile ranges for R44/43, f82, calculated IEPOX-SOA mass concentration, and relative contribution of IEPOX-SOA to the organic mass concentration, RIEPOX-SOA/Org, respectively. Panel (a) illustrates the changes of R44/43 with altitude. In the BL, R44/43 is constant, demonstrating that this layer is well mixed. With increasing altitudes this ratio decreases significantly to much lower values than in the LT. The lowest median values are observed in the UT.

This raises the question about the source of the observed organics in the UT. There are three different possibilities. First, horizontal long-range transport and a subsequent mixing of these air masses with convectively lofted air could occur. Second, particles from the near or distant BL might be transported aloft. A third possibility would be in situ secondary organic aerosol formation in the UT.

A horizontal long-range transport of air masses can be excluded due to the less photochemically aged organics in the UT. The organic aerosol particles in the UT show a lower R44/43 ratio, meaning that they did not experience much photooxidation, as it would be expected from aerosol influenced by long-range transport.

The second possibility, representing the fast convective vertical transport of boundary layer particles, can also be ruled out. The aerosol number concentrations differ considerably (see Fig. d) between the LT and UT, with strongly increased values in the UT. Furthermore, the sulfate and black carbon aerosol mass concentrations show the highest values in the LT and decrease at higher altitudes (see Fig. c and d). In the case of a fast convective vertical transport of boundary layer particles, the sulfate and black carbon aerosol mass concentrations would show similar values in the LT and UT. The decrease at altitudes above 4.5 km indicates that the aerosol particles have been efficiently removed (e.g., scavenging) during vertical transport .

Air mass trajectories were calculated using the FLEXPART model. The trajectories are calculated along the flight tracks starting every minute and calculated backwards for 10 days, providing hourly information on the location of each trajectory. The FLEXPART model is not able to resolve convective transport (see Fig. S4) for the ACRIDICON-CHUVA campaign. Nevertheless, the origin of the trajectories that are released in the LT (<4 km) differs from the origin of the trajectories released in the UT (>8 km) (see Fig. S5). The trajectories released in the LT have their origin also in the LT and show almost no interaction with higher air masses. Most of the trajectories come from the Atlantic Ocean and the southern part of South America. In contrast to this, the trajectories released above 8 km have their origin mainly above the Pacific Ocean and circulate at high altitudes above South America. Just a minor part of the trajectories originate from the eastern direction, coming from the Atlantic Ocean and/or Africa. Interactions with air masses at lower altitudes are rare; most prominent is the lifting at the Andes mountains.

This leads to the conclusion that the third possibility, in situ SOA formation with subsequent growth of the aerosol particles to large enough sizes that they can be detected by the C-ToF-AMS, is the dominant process in the UT.

Another indication supporting this is the size information of aerosol particles with diameters between 90 and 600 nm. Figure  shows the vertical profile of the median and the mode of the binned size distributions measured with the UHSAS-A (panel a). It should be noted here that the lowest cutoff of the considered size range of the UHSAS-A is at 90 nm. Accordingly, the displayed mode diameters are confined by this lower limit. Also, the displayed size distribution medians are affected by the size range limits and should only be interpreted in this context. Whereas in the LT the median and the mode are at diameters around 150 nm (median) and 130 nm (mode), respectively, both the median and the mode are shifted towards smaller diameters with increasing altitude. The lowest value of the median is reached at altitudes above 4 km and (apparently) remains constant. The color code in the vertical profile refers to the size distributions for the three different altitude regions in panel (b) of Fig. . Shown are the median and interquartile range of the size distributions. The size distribution in the LT shows a maximum at 130 nm. The size distributions in the MT and UT are shifted towards smaller diameters, and it is clearly seen that the highest concentrations of small particles (around 90 nm) are found in the UT.

In the following we present evidence that the formation of IEPOX-SOA in the UT can partly explain this observation.

Figure presents the vertical profile of the IEPOX-SOA tracer f82 (panel b) and the calculated median background values f82Bg (see Eq. ). The median background values are quite constant between 4 ‰ and 5 ‰ over the whole altitude range with small interquartile ranges (see Fig. b, vertical dashed line with interquartile ranges in grey). This indicates that continuous emissions and processing of isoprene tend to build an ubiquitous background level up to 14 km. In the LT, f82 shows constant median values around 5 ‰, slightly increasing in the interquartile ranges in the upper part of the LT. These values are similar to or higher than the calculated median background values f82Bg, suggesting a strong IEPOX-SOA influence (see Sect. ). In the UT, the values of f82 are again quite constant over the altitude range, but with increased median values around 8 ‰. The values in the UT lie above the background values and are even higher than in the LT, where an influence of IEPOX-SOA is observed. This indicates that IEPOX-SOA can have an important impact also in the UT.

Although the characteristics of the MT are not the focus here due to the overall low organic mass concentrations, it should be mentioned that the highest median values of f82 were observed in the MT between 4.5 and 8 km. The interquartile ranges are extended here, but the median values reach up to 12 ‰.

To the best of our knowledge, f82 data from the tropical upper troposphere have not yet been reported in the literature. Aircraft measurements above the Amazon rainforest were reported by , but these data are restricted to altitudes below 5 km, corresponding to the definition of the LT in this study. In , two cases are presented and show background (4 ‰) and increased values (9 ‰) of f82. The highest values of f82 were found on top of the boundary layer, decreasing with increasing altitudes. These values are similar to the data from the LT presented here.

Figure c depicts the vertical profile of the median IEPOX-SOA mass concentrations, which are calculated using Eq. (). The lowest values can be found in the LT and MT. However, in the UT, especially at altitudes above 10 km, a strong increase in IEPOX-SOA is observed. Here IEPOX-SOA contributes up to 20 % of the organic mass concentration in the UT (see Fig. d). The highest contribution of IEPOX-SOA to the organic aerosol mass is observed in the MT, whereas in the LT up to 10 % can be attributed to IEPOX-SOA. This suggests that in the MT as well as in the UT IEPOX-SOA formation can occur.

The correlation of f82 with R44/43 is presented in Fig.  using data averaged over 2 min. Again, the two different green colors refer to the LT (dark green) and the UT (light green). In the LT, high R44/43 but low f82 values are observed. The f82 values are similar to reported ones that describe a strong IEPOX-SOA influence (see Sect. ). Interestingly, the values with low R44/43, indicating less photooxidized organic aerosol, are correlated with even higher f82.

This suggests that IEPOX-SOA is contributing to in situ SOA formation in the UT. In general, SOA formation can occur either through new particle formation with subsequent growth or by condensation or reactive uptake on preexisting particles without NPF. However, at the time that NPF occurs, the aerosol particles would be too small to be measurable with the C-ToF-AMS, implying that growth of these newly formed particles is necessary.

The observed enhanced NO mixing ratios in the UT would likely change the reaction pathway of isoprene to a non-IEPOX route and subsequently no IEPOX-SOA formation would occur. Based on the observed IEPOX-SOA in the UT, the oxidation of isoprene to IEPOX must occur before reaching these high altitudes.

In previous laboratory studies it was found that acidic aerosol is needed for reactive uptake or condensation of gaseous IEPOX onto particles. In the laboratory, the acidic conditions are often realized by using sulfate seed particles e.g.,. Also in field studies a correlation between IEPOX-SOA and sulfate aerosol was proposed e.g.,.

Figure shows the scatter plot of IEPOX-SOA against nitrate (panel a) and sulfate (panel b) mass concentrations for LT and UT, respectively. The data are averaged over 2 min. In the LT, no correlation between IEPOX-SOA and nitrate is found. However, in the UT a stronger correlation between IEPOX-SOA and nitrate can be seen (see Fig. a). For sulfate, the correlation is very low both in the LT and UT (see Fig. b). This indicates that sulfate might not be necessary for the formation of IEPOX-SOA, but nitrate could be an important, possibly sufficient component in the UT.

Although the correlation between IEPOX-SOA and nitrate is weak, it may indicate that not only sulfate, but also nitrate can provide the acidic conditions for the partitioning of IEPOX to IEPOX-SOA. For our data, taking only the inorganic species (nitrate and sulfate) into account for acidity calculations, the aerosol is mainly neutralized (see Fig. S6). Although there is a tendency in the UT above 10 km that the measured ammonium is not sufficient to neutralize the inorganic species, a quantitative statement cannot be made as the values fall below or close to the DL. The presence of organosulfates and organonitrates could also affect the acidity calculations. As for the organonitrates, the data are already corrected; for organosulfates such a similar correction is not possible with data from a C-ToF-AMS as there are no different fragmentation patterns between inorganic and organic sulfates . The partitioning of IEPOX on nitrate aerosol results in organic nitrates. The formation and the estimated amount of organic nitrates are discussed in Sect. .

The vertical profiles of NO and NOy mixing ratios are shown in Fig. a and b. The mixing ratios of both species are highest in the LT and decrease towards the MT before reaching increased values in the UT again. In the LT, increased NOx mixing ratios arise from anthropogenic emissions coming from Manaus and other pollution sources. A likely source for NOx in the UT is the production by lightning . Due to the relatively low O3 values in the UT, NOx exists mainly in the form of NO. After conversion to NOy (mainly HNO3), NOx can act here also as a source for (organic) nitrate aerosol and explain the increase in nitrate aerosol mass concentration (see Sect. ).

As already mentioned in Sect. , the detected ammonium in the presented data is mainly sufficient enough to neutralize the aerosol. There is a tendency that the aerosol is not fully neutralized above 10 km. However, organics can also react with inorganic species forming organic nitrates and sulfates. In the following, the four approaches to estimate the presence of organic nitrates as described in Sect.  are discussed.

A first estimation of the particulate nitrate content from organic nitrates is the ratio of NO+ to NO2+. From calibration measurements with ammonium nitrate during the campaign this ratio is known and was found to be in the range between 1.49 and 1.56 with a mean and standard deviation value of 1.52±0.03.

Figure  shows the scatter plot of the corrected NO+ and NO2+ for the LT (dark blue) and the UT (light blue). The linear fit curves for the LT and UT have an intercept of 0, proving that the applied correction is essential. Nevertheless, it has to be noted that the correction is based on correlations between different m/z signals derived from measurements at low altitudes ; thus, the application to UT data bears uncertainties, because the conditions (especially temperature) are different. However, high-resolution AMS measurements at these altitudes are currently not available.

The linear fit for the LT data shows a higher slope than that derived from calibrations with ammonium nitrate. The linear fit for the UT shows an even higher slope and some data points are still significantly above the fitted ratio between NO+ and NO2+. This can be seen as a first hint that organic nitrates might be observed, especially in the UT.

Vertical profiles of median values of m/z 30 and m/z 46 and of the ratio NO+ to NO2+ are depicted in Fig. c and d, respectively. During the whole vertical profile the two lines in panel (c) behave similarly, except for the altitude range between 2 and 6 km, where the distance between them becomes smaller. Compared to the values derived from calibrations with ammonium nitrate, the measured ratios of NO+ to NO2+ during the flights are much higher and the median values range between 2 and 5 (panel d). As described in Sect. , higher NO+ to NO2+ ratios are linked to organic nitrates and the observed ratio profile can be seen as a first evidence for the presence of organic nitrates.

The second estimation provides a range with a lower and upper limit of nitrate mass concentration of organic nitrates according to Eqs. () and () (Sect. ). Figure a shows the estimated lower and upper limits as a vertical profile. In the LT and MT the derived values are below or around zero such that a presence of organic nitrates is unlikely. However, with increasing altitude, the lower and upper limits are also increasing. Especially at altitudes higher than 10 km, both parameters are above zero and the presence of organic nitrates becomes likely. The weakness of this method lies in its dependence on ammonium measurements (see Eqs.  and ). Interferences of water in the fragmentation table can lead to a biased estimation of ammonium concentrations . Biased ammonium concentrations can be one reason for the derived negative values for upper and lower limits of pRONO2.

The third approach uses Eq. () to estimate the amount of nitrate mass concentration of organic nitrates. As described in , a fixed value of 10 for RRONO2 is used here (see Sect. ). In Fig. b the calculated particulate nitrate mass concentration of organic nitrates (pRONO2Farmer) is presented. The measured nitrate mass concentration is also shown. The detection limit is set to 0.1 µg m-3, which was reported by as a conservative approach. Calculated pRONO2Farmer values are quite similar to the measured nitrate in the vertical profile, although they are slightly smaller, especially in the UT. Values for pRONO2Farmer in the LT and MT are below the detection limit. Interestingly, at altitudes above 10 km the derived values are partly above the detection limit. This points again to the presence of organic nitrates, especially in the UT at altitudes above 10 km, but the presence of organic nitrates at altitudes below 10 km cannot be excluded.

The fourth approach is based on Eq. () and addresses an estimation of the ratio of organic nitrates from the ratio for inorganic nitrates determined from the usual calibration measurements (see Sect. ). Using our value of Rcal=1.53±0.03 results in RRONO2=3.44±0.05 (mean and standard deviation). This value is lower than the ratios reported above from standards and also lower than the ratio that is used for the estimation of and . The vertical profile of pRONO2Fry is shown in Fig. c, again with the measured nitrate mass concentration in the back. The vertical profile of pRONO2Fry shows similar values to the measured nitrate. At altitudes above 10 km the highest values are reached, suggesting the likely presence of organic nitrates. Some median values of pRONO2Fry are slightly higher than the measured nitrate concentration, but this lies in the range of uncertainty of the data.

In Fig. , a comparison between the measured nitrate mass concentration and the calculated pRONO2 according to and (panel a) and (panel b) is presented. The data are divided into LT (dark blue) and UT (light blue) and averaged over 2 min. The 1 : 1 line implies that all nitrate is present as organic nitrate with a relative fraction of 100 %. The slope of the linear regression that lies below the 1 : 1 line describes the relative content of nitrate that is present as organic nitrate less than 100 %.

As a result of the scatter plot shown in Fig. , the estimation according to and could explain around 65 % of the measured nitrate in the UT and 35 % in the LT with organic nitrates (see Fig. a). For the comparison with , the whole amount of measured nitrate in the UT and 62 % of the measured nitrate in the LT could be explained by organic nitrates (see Fig. b). The slight overestimation for the UT lies in the range of uncertainty of the data. As a qualitative result both comparisons show that in the UT a higher organic nitrate fraction is observed than in the LT.

The quantification of the nitrate content of organic nitrates remains difficult, but a qualitative statement can be made. Methods one, three, and four agree with each other that at all altitudes organic nitrates are likely present. In the LT and MT, the calculated values lie below the detection limit introduced by . However, in the UT above 10 km the presence of organic nitrates is supported by the results of all four methods.

Organic sulfates can also be formed in the absence of sufficient ammonium. reported that organic sulfates cannot be quantified using C-ToF-AMS data due to a missing tracer ion. Organic sulfates would lead to similar fragmentation patterns as inorganic sulfates . Facing the similar mass concentration of nitrate and sulfate, and especially the increase in nitrate in the UT, the presence of organic nitrates likely plays a similar or even larger role than organic sulfates.

The study by about formation and stability of organic nitrates and sulfates reports that nitrate and sulfates have similar kinetic properties regarding the reaction with tertiary epoxides. According to their study, organic nitrates have shorter lifetimes than organic sulfates and are stable only for short time periods (a few days) before they undergo substitution reactions of nitrate by sulfate or water. Unfortunately, no statement of the temperature dependency of the reactions is given . The temperatures measured in the UT are low (210–240 K) and could slow down chemical reaction processes and possibly shift the reactions in favor of organic nitrates for a longer time period. In field experiments it was shown that low temperatures shift the chemistry to the formation of organic nitrates . The condensation of organic nitrates may be important for the growth of newly formed particles in the atmosphere . Also the phase state of the particles may be important for condensational growth. Over the Amazon basin, SOA particles are predicted to be solid at altitudes above 5 km .