3.1. Changes in Observed Size Distributions
 Since PALMS measures the aerodynamic diameter along with the composition of individual particles, size distributions for different types of particles can be investigated by aggregating the size data based on particle composition. Figure 2 shows the size distributions measured by PALMS for sea salt particles in the marine boundary layer, sulfate-organic particles in the free troposphere, and biomass burning particles in the boundary layer over the Colombian jungle sampled from both the ambient and heated inlet flows. These histograms represent raw counts of aerodynamic diameter measured by the PALMS instrument and are not corrected for inlet transmission and size-dependent particle detection efficiency.
Figure 2. Comparison of unheated (black) and heated residual (gray) aerodynamic size distributions for (a) sea salt particles in the marine boundary layer, (b) biomass burning and organic sulfate particles in the free troposphere, and (c) biomass burning and organic sulfate particles in the continental boundary layer over the Colombian jungle. The lower size limit near 0.2 μm is set by the reduced efficiency of sampling and detection of particles smaller than about 0.3 μm diameter by PALMS.
Download figure to PowerPoint
 Aerosol particles in the marine boundary layer were almost always either sea salt or particles entrained from the free troposphere. The size distribution of just the sea salt particles measured by PALMS in the marine boundary layer shows a broad peak centered at 1 μm da (Figure 2a). The size distribution does not exhibit any significant changes when the particles are sampled through the heated inlet, indicating that there was not likely a significant loss of mass from the sea salt particles upon heating.
 More than 90% of the particles sampled in the free troposphere were sulfate-organic mixtures with or without biomass burning influence. The size distribution of sulfate-organic particles sampled in the free troposphere had a mode aerodynamic diameter around 0.4 μm, but with a small tail out to around 0.9 μm (Figure 2b). The falloff in the size distribution at smaller diameters is due to the decrease in sampling efficiency of the PALMS instrument at diameters below 0.35 μm. The size distribution of the heated residual aerosols with sulfate-organic classification had a similar shape, but is shifted to slightly smaller sizes, with a mode diameter around 0.35 μm and the tail extending to only 0.7 μm. Since this is a change in the aerodynamic diameter, it is not necessarily the case that this represents a change in particle geometric diameter, but could also arise from changes to the particle density or shape.
 Particles identified as of biomass burning origin dominated the particle types analyzed in the low-altitude flight segment over the Colombian jungle, accounting for 72% of ambient and 78% of heated residual positive spectra. The size distribution of these particles resembles that of the free tropospheric sulfate-organic particles with a narrowly peaked distribution, but at a slightly smaller mode diameter of 0.35 μm (Figure 2c). The size distribution of the heated residual particles again shows a slight shift to smaller sizes with a mode diameter of 0.3 μm. A second mode in the biomass burning aerosol population is also discernable between 0.8 and 1.4 μm, which becomes more apparent in the heated residual population as these particles are less likely to shrink below the PALMS detection threshold.
3.2. Changes in Particle Sampling Rate
 Evaporation in the thermal denuder causes some particles to shrink to sizes with reduced detection efficiency or below the PALMS detection limit altogether. One measure of this is the decrease in the rate at which particles were sampled between the heated and proximate unheated sampling periods. In the marine boundary layer, where a significant fraction of the particles sampled were large sea salt particles, total particle sampling rates through the heated inlet averaged 65% that of the ambient sampling rates. Most of the evaporated particles contributing to the decrease in sampling rate were not sea salt since the fraction of particles identified as sea salt increased from 34% in ambient samples to 73% in heated samples. In the free troposphere, where the ambient particles sampled were dominated by non-biomass-burning sulfate-organic particles (75% of positive spectra), the heated sampling rates averaged only 15% that of the proximate ambient sampling rates, indicating that a significant fraction of the ambient aerosols were evaporated to sizes smaller than the PALMS detection threshold. In the Colombian jungle boundary layer, the heated sampling rate averaged 33% of the ambient sampling rate. The difference between the free troposphere and Colombian jungle heated detection rates indicates that particles of biomass burning origin are less likely to shrink to below the PALMS detection threshold than other sulfate-organic aerosols.
3.3. Sea Salt Particles in the Marine Boundary Layer
 A significant fraction of the particles sampled in the marine boundary layer were sea salt particles. Figure 3a shows on a log scale the average negative ion spectrum of sea salt particles sampled in the marine boundary layer through the ambient temperature inlet. The spectrum is dominated by chlorine at m/z 35 and 37 (Cl−) and 93 and 95 (NaCl·Cl−), oxygen at m/z 16 (O−) and 17 (OH−), and sulfate at m/z 80 (SO3−) and 96 (SO4−). There are small peaks attributable to organics at m/z 12 (C−), 24 (C2−), 25 (C2H−), and 26 (C2H2−; CN−). m/z 46 is NO2−, a fragment ion from nitrate. This spectrum is typical of that observed from fresh sea salt particles. Since chloride forms negative ions much more efficiently than organics do, the small carbon-containing peaks indicate that a significant minority of the mass of the sea salt aerosol was organic [Middlebrook et al., 1998].
Figure 3. Change in sea salt particle negative spectra. (a) Average mass spectrum of unheated sea salt particles in the marine boundary layer, (b) observed average mass spectrum of heated sea salt particles, and (c) difference in the mass spectra.
Download figure to PowerPoint
 Figure 3b shows the average spectrum of sea salt particles sampled through the heated inlet. This spectrum is dominated by the same peaks and overall exhibits only small changes from the average ambient sample spectrum. To better discern the effect of heated sampling on the aerosol composition, Figure 3c shows the difference spectrum determined by subtracting the average ambient spectrum from the average heated spectrum. Downward bars represent decreases in relative contribution in the heated spectrum while upward bars denote relative increases. The difference spectrum shows that heating the particles led to a decrease in the relative signals from chlorine and sulfate species on the order of a few percent. The loss of chlorine and sulfate, though small, is interesting since it implies that some of the chlorine and sulfate were present in forms that were volatile at 300°C. Dominant species in sea salt, such as sodium chloride, are not volatile at this temperature [Brooks et al., 2002]. The carbonaceous peaks increased in relative size in the heated samples, indicating that a larger fraction of organic carbon survived the heating than did chlorine. Since only a small fraction of the chlorine was likely to have been volatile, this implies that a large fraction of the organics must have survived the heating.
 Negligible differences were observed between the positive spectra of ambient and heated residual sea salt spectra. The dominant positive ions are from nonvolatile species such as Na, Mg, and K.
3.4. Sulfate-Organic Particles in the Free Troposphere
 The dominant particle type detected by PALMS above 6000 m altitude during TC4 was a sulfate-organic internal mixture. Figure 4a shows the average negative ion spectrum of sulfate-organic particles sampled through the unheated inlet line during times proximate to heated samples. Biomass burning particles are not distinguished from other sulfate-organic particles in negative ion spectra since the potassium marker is not detected. The largest peaks in the negative spectra of sulfate-organic particles are the sulfate peaks at m/z 80, 81 (HSO3−), 96, 97 (HSO4−), and 99 (H34SO4−), O− and OH− at m/z 16 and 17, and organics at m/z 12, 24, 25, and 26. The averaged spectrum of sulfate-organic heated residuals (Figure 4b) shows decreases in the sulfate ion signals by more than an order of magnitude and corresponding increases in the organic ion signals. The difference spectrum (Figure 4c) clearly shows the near complete loss of sulfate and increase in the fractional contribution from organic ion signals. Unlike the sea salt spectra above, the oxygen ion signals at m/z 16 and 17 are smaller in the heated sulfate-organic residuals than in the ambient particles.
Figure 4. Change in sulfate-organic particle negative spectra. (a) Average mass spectrum of unheated sulfate-organic particles sampled in the free troposphere, (b) average mass spectrum of heated sulfate-organic residuals, and (c) difference in the mass spectra.
Download figure to PowerPoint
 Based on in-flight comparisons between PALMS and other instruments on earlier missions [Murphy et al., 2006], the relative sulfate and organic mass fractions of these particles can be estimated from the relative intensities of the sulfate and organic negative ion signals. For unheated sulfate-organic aerosols, the average estimated sulfate mass fraction was 0.59 (±0.30), with a relatively flat distribution among individual particles between 0.01 and 0.99. For heated residual particles, the average estimated sulfate mass fraction was 0.002 (±0.002), indicating essentially complete absence of sulfate in the residuals.
 Figure 5a shows the averaged positive ion spectrum of sulfate-organic particles in the free troposphere. In this spectrum the sulfate is visible as SO+ (m/z 48), SO2+ and HSO2+ (m/z 64 and 65), HSO3+ (m/z 81), and H2SO4·H+ (m/z 99). The dominant organic peaks are m/z 12 (C+) and 28 (CO+), with additional contributions from m/z 24 (C2+), 26 (C2H2+), and 29 (CHO+). Organic fragments are also observed at many other masses below m/z 50. m/z 30, the second largest peak in the spectrum, is NO+, an ion produced commonly from nitrate, ammonium, and other organic nitrogen species. Figure 5b shows the average spectrum of sulfate-organic heated residuals. The sulfate peaks have vanished (with the exception of a small amount of m/z 48, which can be attributed to C4+) and the organic portion of the spectrum has simplified somewhat with many of the minor peaks decreasing. The m/z 30 peak decreased from 22% of the total ion current in the ambient spectrum to 1% in the heated spectrum, indicating that most of the nitrogen species have also evaporated. The difference spectrum in Figure 5c shows that there were significant increases in the relative signals at m/z 12 and 24 and smaller increases at m/z 25, 26, and 36, while the CO+ peak at m/z 28 decreased.
Figure 5. Change in sulfate-organic particle positive spectra. (a) Average mass spectrum of unheated sulfate-organic particles sampled in the free troposphere, (b) average mass spectrum of heated sulfate-organic residuals, and (c) difference in the mass spectra.
Download figure to PowerPoint
 The fraction of free tropospheric particles categorized as biomass burning increased significantly when sampling through the heated inlet. Approximately 19% of particles were classified as biomass burning in all proximate unheated mass spectra, whereas in the heated samples this increased to 70% of particles. There are three likely contributions to this observed difference. The first is that biomass burning particles in the free troposphere tended to be slightly larger than other sulfate-organic particles (not shown) and the second is that the typically lower sulfate content of biomass burning particles led to smaller decreases in diameter. These two factors both lead to a larger fraction of biomass burning particles remaining within the PALMS detection size range. A third potential cause is that as particles evaporated, the increase in the relative contribution to the total ion current from small amounts of potassium crossed the threshold used to classify particles as of biomass burning origin, leading to a recategorization. Examination of histograms of the potassium peak size shows that this third reason was less important than changes in particle size.
 Based on the relative sampling rate between heated and unheated inlets, at least 15% of free tropospheric particles in the PALMS size range left measurable residuals. The mode diameter of the detected residuals decreased by about 10% from the unheated aerosols (see Figure 2). Taken at face value, this would indicate that about 11% of the aerosol mass in the PALMS size range survived heating to 300°C. The actual number of surviving particles was probably much higher than the 15% detected, due to the decreasing detection and transmission efficiency of PALMS between 0.3 and 0.2 μm. For example, a 0.3 μm particle that evaporates down to 0.2 μm is less likely to be detected by up to a factor of 10. Accounting for the decrease in detection efficiency, roughly 20 to 40% of the initially detectable particle mass survived heating. Elemental carbon and potassium in the original biomass burning particles cannot account for this much mass. Single particle soot photometer measurements on a different aircraft but in the same time and vicinity during TC4, showed that elemental carbon comprised less than 2% of the mass of accumulation mode particles in the free troposphere (J. R. Spackman, personal communication, 2009). The complete evaporation of sulfate by itself accounts for a loss of about 55% of the particle mass, so a substantial fraction of the organic mass appears to have survived heating.
 The identification of the heated residuals as including a substantial organic carbon mass contribution is further supported by cluster analysis of the mass spectra. Despite some simplification of the mass spectra from many organic peaks toward relatively more C+ and C2+ (Figure 5c), the residual spectra are not the same as those from elemental carbon, which typically contain significant signals from longer carbon chain ions at m/z 48, 60 and 72. None of the heated residual particles were assigned by the cluster analysis to the same category as the soot particles observed in the polluted boundary layer over Costa Rica.
 Additional information about the properties of the sulfate-organic heated residual particles can be gained by comparing the YAG laser scattering signals from the heated residuals with those from unheated sulfate-organic aerosol. Figure 6 shows a plot of the YAG scatter signal versus the aerodynamic diameter for heated residuals and unheated sulfate-organic particles with a sulfate fraction less than 5%. The similarity in the observed scatter signals between all heated organic residuals and unheated organic-rich particles of the same aerodynamic diameter indicates that they have similar optical sizes and therefore a similar density to shape factor ratio. If we assume that the shape of the heated residuals has not changed significantly (i.e., become significantly nonspherical), then we can infer that the density of the residual organic material remained similar to that in the unheated aerosols, in the range of 1.0–1.2 based on calibrations with polystyrene latex spheres [Murphy et al., 2004].
Figure 6. Plot of the YAG scattering height versus aerodynamic diameter for free tropospheric sulfate-organic heated residuals (red) and for unheated sulfate-organic particles with a sulfate mass fraction <0.05 (blue). The large symbols represent the median values within 0.1 μm bins of aerodynamic diameter, and the error bars indicate the upper and lower quartiles. For contrast, the green trace shows the median values and interquartile range for the scattering signal from unheated sulfate-organic particles with a sulfate mass fraction >0.95. Particles with the same aerodynamic diameter but lower densities have larger geometric diameters and hence scatter more light.
Download figure to PowerPoint
3.5. Biomass Burning Particles in the Colombian Jungle Boundary Layer
 Heated samples were acquired in the boundary layer over the jungle in Colombia during one flight on 8 August 2006. Despite this being the rainy season, the particle type classification of acquired positive spectra during the low-altitude leg (<600 m) was dominated by biomass burning aerosols (72% of positive mass spectra). The absolute particle concentration was fairly low (optical particle counter concentrations (Dp > 300 nm) of about 30 cm−3), so the biomass burning particles were probably from scattered small fires, not the large-scale burning characteristic of the Amazon during the dry season.
 The average positive ion spectrum of ambient temperature biomass burning particles (Figure 7a) is similar to that of the free tropospheric sulfate-organic aerosols shown above (Figure 5a), but the sulfate peaks are substantially smaller relative to the organic peaks and the m/z 39 peak from potassium is the largest peak in the spectrum, contributing 39% on average of the total ion current. Potassium ionizes very easily so this peak can arise from small amounts of potassium in a particle. The ratios of the organic peaks also exhibit significant differences, with the organic signal being much more concentrated at m/z 12 in the biomass burning particles than in the free tropospheric sulfate-organic particles.
Figure 7. Change in biomass burning particle positive spectra. (a) Average mass spectrum of unheated biomass burning particles sampled over the jungle in Colombia, (b) average mass spectrum of heated biomass burning residuals, and (c) difference in the mass spectra.
Download figure to PowerPoint
 The average spectrum of biomass burning heated residuals is shown in Figure 7b. Similar to the sulfate-organic particles in the free troposphere, the sulfate signal has decreased by more than an order of magnitude, indicating nearly complete evaporation of sulfate from the particles. The behavior of the organic peaks, however, is different, with only the signal at m/z 12 increasing in the heated residuals. The fraction of particles classified as biomass burning increased slightly in the heated residuals to 78%. Analysis of the shifts in size and particle sampling rates between the heated and unheated inlets shows that, like the free tropospheric particles, it is difficult to explain the surviving mass without a significant contribution from nonelemental organic carbon in the 300°C residuals.