Effect of dry or wet substrate deposition on the organic volume fraction of core – shell aerosol particles

Understanding the impact of sea spray aerosol (SSA) on the climate and atmosphere requires quantitative knowledge of their chemical composition and mixing states. Furthermore, single-particle measurements are needed to accurately represent large particle-to-particle variability. To quantify the mixing state, the organic volume fraction (OVF), defined as the relative organic volume with respect to the total particle volume, is measured after generating and collecting aerosol particles, often using deposition impactors. In this process, the aerosol streams are either dried or kept wet prior to impacting on solid substrates. However, the atmospheric community has yet to establish how dry versus wet aerosol deposition influences the impacted particle morphologies and mixing states. Here, we apply complementary offline single-particle atomic force microscopy (AFM) and bulk ensemble high-performance liquid chromatography (HPLC) techniques to assess the effects of dry and wet deposition modes on the substrate-deposited aerosol particles’ mixing states. Glucose and NaCl binary mixtures that form core–shell particle morphologies were studied as model systems, and the mixing states were quantified by measuring the OVF of individual particles using AFM and compared to the ensemble measured by HPLC. Dry-deposited single-particle OVF data positively deviated from the bulk HPLC data by up to 60 %, which was attributed to significant spreading of the NaCl core upon impaction with the solid substrate. This led to underestimation of the core volume. This problem was circumvented by (a) performing wet deposition and thus bypassing the effects of the solid core spreading upon impaction and (b) performing a hydration–dehydration cycle on drydeposited particles to restructure the deformed NaCl core. Both approaches produced single-particle OVF values that converge well with the bulk and expected OVF values, validating the methodology. These findings illustrate the importance of awareness in how conventional particle deposition methods may significantly alter the impacted particle morphologies and their mixing states.

. AFM phase images of 1:8 (M) glucose:NaCl particle at varying RH (left -18% RH, right -36% RH). Image from 36% RH shows greater contrast between the core and shell, which aided in OVF quantification. Figure S2. AFM phase image of a sucrose:NaCl particle. Colored regions are masks created in Igor Pro Particle Analysis software, which is transferred onto the 3D height image for volume analysis. The red region indicates the mask used to calculate the total particle volume. The green region indicates the mask used to calculate the core volume. Equation 1 in the main text is used to calculate organic volume fraction. Figure S3. HPLC dry versus wet OVF data for 1:8 (M) glucose:NaCl (purple squares). The dashed black line represents a 1:1 correspondence between the x-and y-axis. The close overlap between the data and expected 1:1 correspondence indicates no significant differences between dry and wet deposition in OVFbulk. Figure S4. Schematic of AFM tip and core particle, with known ℎ and , used to calculate . The core is assumed to be rectangular in shape.

Environmental factors that affect AFM phase imaging
In this study, AFM imaging was performed at a relatively low RH range of 25 -35% to minimize water uptake contribution in organic volume fraction (OVF) measurements, but also still maximize the observed phase separation between the inorganic core and organic shell to facilitate identification of boundaries between two components within a particle. An example of an ideal phase contrast level is shown in Figure S1, where the same particle at 36% RH clearly shows distinct topographical features of the inorganic core, whereas the particle core at 18% RH is not as evident. Here, we define the term "phase bleeding" to describe the lack of distinct inorganic core features that separate it from the organic shell, which is evident on the particle at 18% RH. This phase bleeding likely stems from the relatively high glucose viscosity at this RH, where the smaller difference in viscosity between the core and shell leads to less distinct phase separation. From 18% RH to 36%, a doubling of RH results in two orders of magnitude decrease in the organic component viscosity, with no change in the inorganic component viscosity, which qualitatively aids in better phase separation. 1 NaCl viscosity will be unaffected until reaching a sharp deliquescent point at ~75% RH, at which point a phase transition from solid to liquid results in a many order of magnitude decrease in viscosity. Likewise, the effect of changing temperature would most likely produce similar effects, where the greater difference in viscosity between the core and shell through modulating the temperature could aid in observed phase separation.

Ensemble average OVF comparison between AFM and bulk
To validate our quantitative single particle OVF measurements, a comparison is made to bulk OVF measurements from HPLC, a well-established analytical tool that can measure the concentrations of various compounds in a chemical mixture. In this case, the concentrations of glucose, sodium and chloride were determined in aqueous extracts of filter samples and were field blank subtracted.
Masses of these species were converted to volumes, and used to calculate OVFbulk following Equation S1,

= +
Eq. S1 where is organic volume fraction of bulk, & are masses, and & are pure solute densities of organic and inorganic components, respectively. Specifically, = 1.54 / and = 2.16 / were used for glucose and NaCl, respectively (Lee et al., 2017). Masses and were obtained by converting the concentrations measured by the HPLC using the sample extract volume to determine the moles of analyte and the molar masses of glucose, sodium and chloride to determine the mass.

HPLC measurement propagated analytical uncertainty
Vertical error bars for HPLC measurements, or bulk, in Figures 2, 4 are propagated analytical uncertainty for each measurement. The absolute error was calculated: where e is absolute uncertainty, is standard deviation of field blank, [ ] is sample concentration, and |% | is absolute deviation of spike recovery. When combining errors in propagation, the following was used: where 3 is the absolute uncertainty of summed values, and 1 & 2 are absolute uncertainties of starting values. Specifically, Equation S3 was used to propagate the error when the mass of sodium and chloride were summed and when the volume of sodium chloride and glucose were summed.
To determine the absolute uncertainty in the OVF measurement when division is used, the absolute uncertainty in the volume for glucose and the total volume (glucose+NaCl) was converted to % relative uncertainty: where % is the percent relative uncertainty of OVF, and % & % + are percent relative uncertainty for glucose and glucose+NaCl volumes, respectively.

Expected OVF calculations
Comparisons to single particle and bulk OVF in the main text are made to the expected OVF, which was calculated by measuring the molar mixing ratio of the bulk solution before and after bubbling, and calculating the volume with known densities of individual components: Eq. S7 where ̅̅̅ is sample mean, is population mean, is the corresponding standard deviation, is number of samples, and is sum of squares.
To reject the null hypothesis that dry deposited versus wet deposited OVF data are statistically insignificant, an independent samples t-test (dark orange vs. blue) was performed for the 1:8 (M) glucose:NaCl data. This comparison will be used to calculate the probability of obtaining the observed differences from dry versus wet OVF values out of random chance. Also, to quantify how the sample means from dry versus wet deposited OVF data compared to the expected value, Cohen's d was calculated, which represents how far away the sample mean is with respect to the expected value, but normalized to the distributions shown. Equations for t-test and Cohen's d are in the supporting information (Eq. S6 & S7). First, comparing the dryversus wetdeposited OVF distributions, two tailed t-test confirmed extremely strong statistical significance, t(354) = 32.838, p < 0.0001. Thus, statistical analysis shows that there is less than 0.01% probability of random chance that can explain the differences in the sample means. Therefore, ttest confirmed that dry and wet deposited particles produced statistically different results, which allows us to reject the null hypothesis outlined in the beginning. Also, Cohen's d calculations for the dry (dark orange) and wet (blue) OVF data versus the expected value was 4.142 and -0.419, respectively. Therefore, the wet deposited particles were statistically much closer to the expected value then the dry deposited particles.

Decoupling AFM tip convolution from NaCl spreading
To validate that the observed width of the NaCl particle data is attributed directly to the NaCl particle spreading and not from AFM particle broadening as a result of convolution of the shape of AFM tip and shape of the core, an equation was derived to calculate for the maximum width broadening (Fig. S4), = * tan ( ℎ ) Eq. S9 where is maximum imaging width broadening, is height of the core, and ℎ is half cone angle of the AFM tip. is inherent to scanning probe microscopy techniques such as AFM, that depend on the probe sharpness and shape. For the AFM tips used in this study, ℎ is 20 , and would be up to 36% of the particle height, or roughly maximum of 180 nm for a particle height of 500 nm. Therefore, the observed 300 nm directional spreading shown in Figure   3B cannot be solely attributed to the width broadening due to the shape of AFM probe, and thus at least 120 nm is directly attributable to NaCl spreading as a result of impaction to the solid substrate.
The data set supporting this manuscript is hosted by the UCSD Library Digital Collections (https://doi.org/10.6075/J04M92SF)