Vibrational Sum Frequency Generation Spectroscopy of Secondary

Secondary organic material (SOM) was produced in a flow tube from α-pinene ozonolysis, and collected particles were analyzed spectroscopically via a ...
0 downloads 0 Views 1MB Size
Download PDF
Article pubs.acs.org/JPCA

Vibrational Sum Frequency Generation Spectroscopy of Secondary Organic Material Produced by Condensational Growth from α‑Pinene Ozonolysis Mona Shrestha,†,† Yue Zhang,‡,† Carlena J. Ebben,† Scot T. Martin,‡,* and Franz M. Geiger†,* †

Department of Chemistry, Northwestern University, 2145 Sheridan Road, Evanston, Illinois 60208, United States School of Engineering and Applied Sciences & Department of Earth and Planetary Sciences, Harvard University, 29 Oxford Street, Cambridge, Massachusetts 02138, United States

‡

S Supporting Information *

ABSTRACT: Secondary organic material (SOM) was produced in a flow tube from α-pinene ozonolysis, and collected particles were analyzed spectroscopically via a nonlinear coherent vibrational spectroscopic technique, namely sum frequency generation (SFG). The SOM precursor α-pinene was injected into the flow tube reactor at concentrations ranging from 0.125 ± 0.01 ppm to 100 ± 3 ppm. The oxidant ozone was varied from 0.15 ± 0.02 to 194 ± 2 ppm. The residence time was 38 ± 1 s. The integrated particle number concentrations, studied using a scanning mobility particle sizer (SMPS), varied from no particles produced up to (1.26 ± 0.02) × 107 cm−3 for the matrix of reaction conditions. The mode diameters of the aerosols increased from 7.7 nm (geometric standard deviation (gsd), 1.0) all the way to 333.8 nm (gsd, 1.9). The corresponding volume concentrations were as high as (3.0 ± 0.1) × 1014 nm3 cm−3. The size distributions indicated access to different particle growth stages, namely condensation, coagulation, or combination of both, depending on reaction conditions. For filter collection and subsequent spectral analysis, reaction conditions were selected that gave a mode diameter of 63 ± 3 nm and 93 ± 3 nm, respectively, and an associated mass concentration of 12 ± 2 μg m−3 and (1.2 ± 0.1) × 103 μg m−3 for an assumed density of 1200 kg m−3. Teflon filters loaded with 24 ng to 20 μg of SOM were analyzed by SFG. The SFG spectra obtained from particles formed under condensational and coagulative growth conditions were found to be quite similar, indicating that the distribution of SFG-active C−H oscillators is similar for particles prepared under both conditions. The spectral features of these flow-tube particles agreed with those prepared in an earlier study that employed the Harvard Environmental Chamber. The SFG intensity was found to increase linearly with the number of particles, consistent with what is expected from SFG signal production from particles, while it decreased at higher mass loadings of 10 and 20 μg, consistent with the notion that SFG probes the top surface of the SOM material following the complete coverage of the filter. The linear increase in SFG intensity with particle density also supports the notion that the average number of SFG active oscillators per particle is constant for a given particle size, that the particles are present on the collection filters in a random array, and that the particles are not coalesced. The limit of detection of SFG intensity was established as 24 ng of mass on the filter, corresponding to a calculated density of about 100 particles in the laser spot. As established herein, the technique is applicable for detecting low particle number or mass concentrations in ambient air. The related implication is that SFG is useful for short collection times and would therefore provide increased temporal resolution in a locally evolving atmospheric environment.

I. INTRODUCTION

among other factors. The particle phase is said to be constituted by secondary organic material (SOM). The particles can affect climate, human health, and visibility.4,8 Secondary organic material is estimated to account for about 50% of the total fine particle (PM2.5) mass concentration on an annual average4,9 and 30−70% of the mass concentration of submicrometer particles.10 However, the production mecha-

Terrestrial plants emit a variety of volatile organic compounds such as isoprene, monoterpenes, and sesquiterpenes into the atmosphere at levels that far exceed anthropogenic nonmethane hydrocarbon emissions.1−5 The oxidation of these species can form products having decreasing volatilities as the oxygen-tocarbon ratio increases. The interaction of these species results in the formation of a so-called secondary organic aerosol,6,7 where aerosol is defined as a mix of gas-phase and particlephase molecules that partition between the two phases depending on local conditions of temperature and dilution, © 2013 American Chemical Society

Received: May 22, 2013 Revised: July 17, 2013 Published: July 22, 2013 8427

dx.doi.org/10.1021/jp405065d | J. Phys. Chem. A 2013, 117, 8427−8436

The Journal of Physical Chemistry A

Article

Figure 1. Schematic diagram of the apparatus, including precursor generation, flow tube reactor, and particle analysis. An illustration of the gas connections and flows in the reactor is shown in the lower part of the figure.

II. EXPERIMENTAL SECTION 1. Flow Tube for Particle Production. The flow tube used in this work consisted of three parts (Figure 1). The first part included the production and dilution of gases, controlled using four mass flow controllers (MKS, M100B model). A flow of air from an AADCO 737 pure air generator passed through an ozone generator (Jelight, model 1000), producing 200−500 ppm of ozone at flow rates ranging from 0.3 to 3 sLpm. Ozone concentrations were controlled and adjusted to the desired levels by dilution in dry air at various flow rates. α-Pinene was introduced into the flow tube via an additional flow of pure air at 0.50 sLpm passed through a syringe injector system (CHEMYX, Fusion Touch 200 model) that was heated to 135 ± 1 °C. The syringe contained a 1:1 mixture of (+)- and (−)-α-pinene (Sigma-Aldrich, ≥99% purity, 97% enantiomeric excess) in excess 2-butanol (Sigma-Aldrich, ≥99.5% purity) at a dilution ratio of 1:624,13,14,21 which was used as an OH scavenger to ensure that ozonolysis was the only reaction occurring inside the flow tube. The concentration of α-pinene in the flow was adjusted by changing the syringe injection rate. A control experiment carried out using pure 2-butanol under the same ozone concentration led to no particle production. The ozone and α-pinene concentrations reported here are relevant for the conditions at the inlet of the flow tube. The second part of the setup comprised the flow tube itself, which was made of glass with an inner diameter of 48.2 mm and a length of 1.30 m. The flow tube was operated slightly above ambient pressure, around 25 °C, in the laminar flow regime (Reynolds number of 9.4 ± 0.5), and with a residence time of 38 ± 1 s.22 As shown in Figure 1, the ozone and αpinene flow inlets were arranged perpendicular to each other to induce turbulence at the injection point to promote rapid mixing. The mixing time for ozone and α-pinene are estimated from calculations to be less than 2 s, which is very small compared to their residence time in the flow tube. The flow tube was equipped with an ozone sensor (Ecosensors, UV-100) and a temperature sensor (National Instrument, USB-TC01) at

nisms, as well as the physicochemical properties of the secondary organic material, remain insufficiently characterized and understood for quantitative modeling at the level needed for accurate predictions of effects on climate, human health, and visibility. This knowledge gap has motivated many research efforts.11−20 Herein, we prepare submicrometer particles of SOM by ozonolysis of an important atmospheric monoterpene species (viz. α-pinene) in a flow tube. The flow tube enables the rapid synthesis of aerosol particle populations for a wide range of particle number and mass concentrations. Given that the particle population is ultimately derived from gas-phase precursors, the diameter and mass distributions of the population are directly coupled to processes of particle growth. Through the analysis of size distributions prepared for a matrix of 20 different reaction conditions, we determine when the size distribution is mainly dominated by processes of nucleation and condensation or alternatively mainly by coagulation. For the case of condensation, particle growth necessarily involves processes occurring at the particle/gas interface, a region that is challenging to access directly and with molecular specificity. Here, we characterize the surface-localized species of SOM using vibrational sum-frequency generation (SFG), which is a label-free molecularly specific laser-spectroscopic technique that is nondestructive and does not require any sample manipulation other than particle collection. As we demonstrate in the present study, sub-100 nm sized particles can be analyzed directly on filters or impactor plates at ambient temperature and pressure. The technique is sensitive enough to detect particle densities on filters as low as 10 million per cm2 within a few minutes (or 100 per laser spot), as established herein, and is thus applicable for the detection of low particle number or mass concentrations in ambient air. The related consequence is that SFG is useful for short collection times and would therefore provide increased temporal resolution for studying atmospheric aerosol particles. 8428

dx.doi.org/10.1021/jp405065d | J. Phys. Chem. A 2013, 117, 8427−8436

The Journal of Physical Chemistry A

Article

Table 1. Number Concentrations (cm−3), Mass Concentrations (μg m−3), Mode Diameter (nm), and Geometric Diameter Standard Deviation (gsd) of Particle Populations Produced by α-Pinene Ozonolysis. A Material Density of 1200 kg m−3 Was Used for Conversion of Volume Concentrations to Mass Concentrations α-pinene (ppm) 0.125

1

10

100

a

no. concn mass concn mode diameter gsd no. concn mass concn mode diameter gsd no. concn mass concn mode diameter gsd no. concn mass concn mode diameter gsd

O3 (ppm) 0.15 ± 0.02

O3 (ppm) 0.9 ± 0.1

0 0 N/A N/A 0 0 N/A N/A (2 ± 2) × 101 0a 8±9 1.0 (4.4 ± 0.3) × 105 35 ± 3 48 ± 2 1.4

(1 ± 1) × 10 (3 ± 5) × 10−2 22 ± 4 1.2 (3.1 ± 0.9) × 102 (9 ± 3) × 10−3 33 ± 7 1.3 (4.0 ± 0.2) × 105 (1.6 ± 0.2) × 102 81 ± 2 1.4 (8.3 ± 0.3) × 105 (8.6 ± 0.1) × 102 88 ± 5 1.6 2

O3 (ppm) 5.7 ± 0.2 (1.0 ± 0.6) 15 ± 5 60 ± 5 1.3 (1.5 ± 0.2) 61 ± 9 86 ± 6 1.4 (6.0 ± 0.7) (2.5 ± 0.2) 147 ± 9 1.4 (8.3 ± 0.4) (1.3 ± 0.1) 134 ± 8 1.5

× 10

5

× 105

× 105 × 103

× 106 × 104

O3 (ppm) 43 ± 1

O3 (ppm) 194 ± 2

(4.4 ± 0.6) × 10 11 ± 3 35 ± 3 1.3 (5.5 ± 0.2) × 105 (52 ± 0.1) × 102 84 ± 3 1.5 (6.3 ± 0.7) × 105 (1.19 ± 0.02) × 104 245 ± 38 1.4 (9.1 ± 0.2) × 106 (1.6 ± 0.04) × 105 262 ± 12 1.7

(3.2 ± 0.2) × 105 20 ± 2 34 ± 2 1.5 (5.8 ± 0.4) × 105 (66 ± 0.1) × 102 85 ± 19 1.7 (1.8 ± 0.2) × 106 (1.57 ± 0.02) × 104 155 ± 5 1.5 (1.3 ± 0.02) × 107 (4.0 ± 0.1) × 105 334 ± 4 1.9

5

Although particles were present, the mass concentration was below the detection limit.

Table 2. Summary of Experimental Conditions and Observations, As Described in the Literature and for the Present Work author Tolocka et al.13

Jonsson et al.36

Gao et al.14

Winkler et al.57

this work

α-pinene concn (ppm)

ozone concn (ppm)

residence time (s)

particle number concn (cm−3)

particle mass concn (μg m−3)

83 11 136 136 43 0.019 0.024 0.023 43 43 43 43 90 90 90 0.15 to 194

1 1 1 1 0.2 1.18 1.11 1.17 0.24 0.3 0.45 0.55 N/A N/A N/A 0.125 to 100

3 22 3 22 23 242 242 242 23 23 23 23 30 30 30 38

6.50 × 105 2.10 × 106 4.00 × 106 5.90 × 106 1.50 × 106 1.20 × 104 6.00 × 103 1.00 × 104 3.90 × 106 9.30 × 105 2.70 × 106 3.70 × 106 1.00 × 106 2.00 × 107 3.00 × 108 3 × 102 to 1 × 107

15 95 36 400 3.5 2 2 0.6 14 49 187 279 N/A N/A N/A 10 to 1 × 104

mode diameter (nm) 22 43 30 50 34 N/A N/A N/A 37 44 48 49 10 20 30 Variable, maximum of 350

and number concentrations were collected. The filters were subsequently analyzed by SFG. 2. Sum Frequency Generation (SFG) of SOM Particles. SFG is a coherent nonlinear technique in which two beams with different frequencies, typically in the infrared and the visible regions, mix in a noncentrosymmetric medium, such as at a surface or an interface, and generate a signal oscillating at the sum of the input frequencies.25,26 The tunable broadband infrared beam used in the SFG experiments for this study was generated with a regeneratively amplified femtosecond laser system and an optical parametric amplifier, which have been described in detail in previous publications.27−33 Here, we used the ssp polarization combination to probe the components of vibrational transitions that are oriented mainly parallel to the plane of incidence. The three-letter combinations (s for perpendicular and p for parallel to the plane of incidence) represent the polarizations of SFG, visible, and IR beams respectively. The ssp-polarized SFG spectra reported here were generally associated with the most signal intensity of the

the inlet. The temperature was 25 ± 1 °C over the course of the several hours required to perform individual experiments. The total pressure inside the flow tube, measured using a baratron (Omega PX409), ranged from 1.00 to 1.01 atm, depending on the day. The third part of the setup consisted of an outlet that was split between sample collection and measurements of number− diameter distribution. The particles were collected on Teflon filters (Sartorius Stedium, 47 mm diameter, part no. 11807-47N, 200 nm pore size, collection efficiency of 95%) at a sampling rate of 1.82 sLpm. Particle number−diameter distributions were measured using a scanning mobility particle sizer (SMPS, TSI, 3934 model).23 Integrated number concentrations were measured using a condensation particle counter (CPC, TSI Incorporated, 3022 model) up to a detection limit of 107 particles cm−3. To calculate the mass concentration of collected particles on the filters, a material density of 1200 kg m−3 was used.24 By varying the collection time and the reaction conditions (Table 1), filter samples of variable mass loadings 8429

dx.doi.org/10.1021/jp405065d | J. Phys. Chem. A 2013, 117, 8427−8436

The Journal of Physical Chemistry A

Article

distinct parts of the flow tube (see further section III.3). The reaction conditions described here produced particle concentrations that were stable over the course of several hours (e.g., mean value of 3.30 × 105 cm−3 and standard deviation of 0.13 × 105 cm−3; cf. Figure 2A).

polarization combinations employed in our studies, including those that probe components of vibrational transitions that are oriented mainly perpendicular to the plane of incidence. To present an internal reflection geometry, an optical window was pressed against each filter sample collected from the flow tube.20,31 The sample was moved so that at least three spots were analyzed per sample. The SFG spectra were recorded at the same incident up-converter pulse energy. The spectra were then calibrated to a polystyrene frequency standard, normalized to the nonresonant SFG spectrum from gold to account for the line shape of the incident IR pulse, and also to the standard deviation in the nonresonant frequency region of 3100−3200 cm−1. Particle samples were stored in a freezer at −12 °C in order to minimize compositional changes that could occur with the long-term storage. SFG spectra shown in the Supporting Information have negligible differences, outside the signal intensity variations that we characterize in detail in the following section, for freshly prepared particle samples as opposed to those stored in the freezer for over a year.

III. RESULTS AND DISCUSSION 1. Regimes of Particle Growth. Several α-pinene oxidation studies exist that focus on SOM particle composition and mass yield (Table 2).12−15,21,34,35 The current study explores ozone concentrations that are lower than those published in the literature, except for the work of Jonsson et al.,36 but employs residence times in the flow tube that are similar to those other studies. Here, we hypothesize that the lower concentrations employed in the present study will be useful for constraining the modeling of nucleation and growth processes that accompany particle production by α-pinene ozonolysis. To this end, we explored a matrix of reaction conditions, summarized in Table 1, which shows that the number and mass concentrations of the particles produced increased with α-pinene and ozone concentrations. Assuming that the ozonolysis of α-pinene is overall secondorder,37 the half-life of the limiting reagent can readily be computed for pseudo-first-order conditions. From this calculation, we estimate the extent of the reaction as the ratio of the half-life of α-pinene or ozone to the flow tube residence time (Table 3). For an ozone concentration equal to or greater than 43 ± 1 ppm, the half-life of α-pinene was calculated to be 10 times smaller than the residence time, indicating α-pinene reacted to completion near the entrance of the flow tube. For these conditions, nucleation of new particles and growth by coagulation of the particles are expected to occur in spatially

Figure 2. (A) Time series of number concentration of a particle population produced from 0.125 ± 0.01 ppm of α-pinene and 43 ± 1 ppm of ozone. (B) Number-diameter distributions of particle population produced from 0.125 ± 0.01 ppm, 1.0 ± 0.03 ppm, and 10.0 ± 0.3 ppm α-pinene. Residence time of 38 ± 1 s and ozone concentration of 43 ± 1 ppm.

Figure 2B shows number−diameter distributions for particles produced using α-pinene precursor concentrations ranging from 0.125 ± 0.01 to 10.0 ± 0.3 ppm and a fixed ozone concentration of 43 ± 1 ppm. As shown in Table 1, these conditions produced (4.4 ± 0.6) to (6.3 ± 0.7) × 105 particles cm−3 and mass concentrations of 101 to 104 μg m−3, respectively, depending on reaction conditions. Given that coagulation imposes an upper limit of 105 to 106 particles cm−3 under the conditions relevant here,1 coagulation processes can be taken to dominate the number−diameter distribution of this particle population. For monodisperse particles and a residence time of 38 ± 1 s,22 particle collisionsand thus coagulation are estimated to contribute more than 90% of the observed diameter growth for an ozone concentration of 43 ± 1 ppm and an α-pinene precursor concentration of 10.0 ppm (Figure 2B). As summarized in Table 1, the mode diameter and the width increased for higher α-pinene and ozone precursor concentrations under these conditions. One purpose of this study was to use SFG to probe surface chemistry related to condensational growth processes. Therefore, the reaction conditions of ozone and α-pinene precursor concentrations were optimized (e.g., matrix of experiments in Table 1) for particle sampling under conditions in which the number−diameter distribution was determined dominantly by condensation rather than coagulation. Survey experiments showed that the α-pinene concentration producing the largest particle diameters while avoiding a significant extent of coagulation was 0.125 ppm for excess ozone. As a specific example, Table 4 shows that an α-pinene concentration of 0.125 ± 0.01 ppm and an ozone concentration of 51 ± 1 ppm

Table 3. Ratio of the Limiting-Reactant Half-Life to Flow Tube Average Residence Time for the Matrix of Experimental Conditions in This Studya α-pinene (ppm)

O3 (ppm) 0.15 ± 0.02

O3 (ppm) 0.9 ± 0.1

O3 (ppm) 5.7 ± 0.2

O3 (ppm) 43 ± 1

O3 (ppm) 194 ± 2

0.125 1.00 10.0 100

N/A 8.5 0.85 0.085

9.5 N/A 0.85 0.085

1.5 1.5 N/A 0.085

0.20 0.20 0.20 N/A

0.044 0.044 0.044 N/A

The limiting reactant can be ozone or α-pinene depending on which was in excess. Entries of “N/A” indicate that the reaction conditions were pseudo-first-order neither in ozone nor in α-pinene. A bimolecular rate constant of 8.70 × 10−17 cm3 molecules−1 s−1 was used.58 A residence time of 38 ± 1 s was used.

a

8430

dx.doi.org/10.1021/jp405065d | J. Phys. Chem. A 2013, 117, 8427−8436

The Journal of Physical Chemistry A

Article

Table 4. Number Concentrations (cm−3), Mass Concentrations (μg m−3), Mode Diameter (nm), and Geometric Standard Deviation of the Particle Populations Collected by Filter Sampling and Subsequently Characterized by SFG (Figures 4 to 6)a no. concn (cm−3) mass concn (μg m−3) mode diameter (nm) gsd

α-pinene (ppm) 0.125

α-pinene (ppm) 1.0

(9.5 ± 0.7) × 104 12 ± 2 63 ± 3 1.29

(1.1 ± 0.1) × 106 (1.2 ± 0.1) × 103 93 ± 3 1.49

increased. For instance, for a residence time of 17 ± 0.5 s, the number concentration was (8.6 ± 0.5) × 104 cm−3. For a residence time of 38 s, the number concentration increased to (2.56 ± 0.07) × 105 cm−3. The implication of these results is that particles continuously nucleated and grew by condensation as the dominant process, rather than coagulatively. The particles sampled on filters for SFG analysis were prepared using these conditions. For comparison, we also studied particles formed under coagulative growth conditions by SFG. 2. Surface Vibrational Spectroscopy. Previously reported20,31 ssp-polarized SFG spectra of the equilibrium roomtemperature vapor phase of (+)-α-pinene vapor in contact with a fused silica window and of the surfaces of SOM particles prepared from (+)-α-pinene ozonolysis at the Harvard Environmental Chamber (HEC) are shown as spectra A and B in Figure 4. Spectrum 4A specifically was obtained from α-

The ozone concentration used for each experiment was 51 ± 1 ppm. A material density of 1200 kg m−3 was used for conversion of volume concentrations to mass concentrations.

a

produced a number concentration of (9.5 ± 0.7) × 104 cm−3 and a mode diameter of 63 ± 3 nm for the particle population. As an approximate estimate of the relative importance of coagulation, a monodisperse particle population of these properties can be considered. For this population, there is a calculated coagulation coefficient K of 2 × 10−9 cm3 s−1.22 The corresponding characteristic time for coagulation τ is estimated as 3 h, which is 2 orders of magnitude longer than the flow tube residence time of 38 ± 1 s. Even so, given that the particle population was not monodisperse, the calculated value of τ represents a lower-limit estimate. As an experimental demonstration of condensational growth for our reaction conditions, an experiment with a longitudinally movable particle sampler was conducted using 50 ± 1 ppm of ozone and 0.125 ± 0.01 ppm of α-pinene. The position of the particle sampler inside the flow tube allowed measurements of number−diameter distributions for residence times ranging from 3 ± 0.2 to 38 ± 1 s. The measured distributions are shown in Figure 3. For a residence time of 3 s, no particles were detected by the SMPS. For longer residence times, the mode diameter of the particle population increased from less than 10 nm to more than 50 nm. The number concentration also

Figure 4. ssp-Polarized vibrational spectra of (A) (+)-α-pinene vapor in contact with a fused silica window, (B) SOM prepared from the ozonolysis of (+)-α-pinene in the HEC, (C) SOM prepared from the ozonolysis of (+)-α-pinene in the flow tube, and (D) SOM prepared from the ozonolysis of a 50/50 mix of (+)- and (−)-α-pinene in the flow tube. In addition to the data workup described in the paper, the spectra were normalized to their maximum intensities.

pinene vapor adsorbed on the window at the vapor/window interface maintained at room temperature. As a clarification, this experiment was conducted by mounting a window on a Teflon cell and then injecting about 0.3 mL of liquid α-pinene into the cell such that the liquid did not contact the window. The void space inside the cell then fills up with the equilibrium vapor pressure of α-pinene.31 The signal collected at this vapor/window interface is shown in spectrum 4A. Figure 4 also presents SFG spectra of SOM prepared from the ozonolysis of (+)-α-pinene (spectrum C) and 50/50 pinene mix of the two enantiomers (spectrum D) in the flow tube under conditions of coagulative growth (1 ppm of α-pinene and 51 ppm of ozone). The spectra are remarkably similar to those obtained from particles collected at a field site in Southern Finland during the summer of 2010.20,31,38 The SFG spectra shown in Figure 4 feature a dominant peak near 2950 cm−1, separated by around 10−20 cm−1 for vapor phase in comparison to the particles, which is generally associated with asymmetric CH 3 stretches and Fermi resonances in published SFG spectra of common hydrocarbon systems.39−45 This result is consistent with the presence of the three methyl groups that are tightly interlocked in the bridged cyclic structure of α-pinene. Wilson assigned the peak near 2950 cm−1 as a methyl asymmetric stretching mode, while the smaller peak near 2880 cm−1 was putatively assigned as a methyl symmetric stretch.46 Given the complexity of the species, though, we expect the assignments to be somewhat less

Figure 3. Number−diameter distributions of particle populations depending on flow tube residence time from 3 to 38 s. The number concentration of each population increased, as follows: 1.69 × 10−1, 7.50 × 103, 8.58 × 104, 2.00 × 105, 2.33 × 105, and 2.56 × 105 particles cm−3 for residence times of 3, 10, 17, 25, 32, and 38 s. The shaded regions represent one standard deviation in the observed number− diameter distributions across 8−12 replicates. 8431

dx.doi.org/10.1021/jp405065d | J. Phys. Chem. A 2013, 117, 8427−8436

The Journal of Physical Chemistry A

Article

SFG signal intensity was acquired from filter sample containing 3 μg of particle material, and the average SFG spectrum of that sample is shown in Figure 5A. The filter sample having the

straightforward. We are currently working on a spectroscopic assignment of the vibrational features produced by α-pinene that is based on isotope-labeling, computational approaches, and vibrational spectroscopy. In the absence of such a study and given the rigid structure of α-pinene in which the relative position of the methyl oscillators is quite inflexible, we speculate here that the strong SFG response near 2950 cm−1 may be attributable to strongly coupled coherences involving Fermi resonances and asymmetric C−H stretches. Indeed, αpinene and SOM prepared from it produce some of the strongest SFG signal intensities we have recorded in our laboratory from strongly ordered molecular systems, yielding signal responses as strong as those obtained from well-ordered self-assembled alkyl silanes on fused silica.27,29,47 Whatever the detailed molecular origin of the strong SFG responses obtained from the particle material discussed here in Figure 4, we observe that the ssp-polarized SFG spectra of SOM prepared at the HEC and in the flow tube appear similar, irrespective of the pinene mix or the way the particles were prepared. For our setup, the SFG signal originates mainly from the particle/air interface and not from internal interfaces. Specifically, the similarity of the SFG signal intensities obtained under similar signal collection conditions for α-pinene-derived SOM from the flow tube (for which no seed particles were present for the synthesis) to the mixed-phase particles prepared in earlier work in the HEC31 (for which SOM was deposited on ammonium sulfate seed particles) suggests that internal interfaces did not contribute significantly to the SFG intensity. Comparable filter loadings (upper limits of 20 μg vs 285 μg) and similar particle sizes (about 93 nm) were used for the present flow tube and 2011 chamber studies.31,48 Control studies carried out as part of this present work show that a filter containing α-pinene-derived SOM from the flow tube produced higher SFG signal intensity than SOM transferred from the same Teflon filter onto a fused silica window when probed in external reflection geometry (see Supporting Information). The weak SFG signal in the latter case may be due to the small quantity of material transferred from the Teflon filter onto the fused silica window. Finally, we note that we conduct our SFG experiments in internal reflection geometry. Since the Teflon filter surface is not nearly as reflective as that of the optical windows used in our experimental setup, the SFG signal contribution from internal interfaces between SOA particle material and the Teflon fibers of the supporting filter are expected to be minor. 3. Limit of Detection Analysis and Condensational versus Coagulative Growth Conditions. Having verified that particles synthesized in the flow tube produce SFG responses that are comparable to those produced by particles synthesized at the HEC, we prepared filter samples having particle mass loadings ranging from 24 ng to 20 μg to evaluate the SFG responses from particle material prepared under condensational vs coagulative growth conditions. This study then allowed us to assess the limit of detection for particle analysis by SFG. The experiments were carried out in two sets. The first set consisted of α-pinene and ozone concentrations of 1.0 ± 0.03 ppm and 51 ± 1 ppm, respectively, which produced mainly coagulatively grown particles of a mode diameter of 93 nm and a geometric standard deviations of 1.49 (Table 4). The filter collection time for these particles ranged from 15 s to 10 min. The incident IR power was focused on the C−H asymmetric stretching region between 2940 and 2950 cm−1. The highest

Figure 5. Average ssp-polarized SFG spectra of SOM sampled at filter mass loadings of (A) 3 μg, (B) 0.5 μg, and (C) 24 ng. Also shown for comparison are average spectra for (D) blank Teflon filters and (E) a clean silica window.

lowest particle mass loading (0.5 μg) produced the lowest signal intensity (Figure 5B). Even so, the SFG signal intensity was sufficient to determine that the limit of detection had not been reached. The fact that strong SFG signal intensities can be observed within minutes from filter samples containing submicrogram amounts of SOM highlights the high sensitivity of SFG for the analysis of the particles prepared in the flow tube. Control experiments using a HEPA filter to remove the SOA particles from the chamber outflow while sending the vapor phase over collection filters resulted in no appreciable SFG signal from the collection filter.48 We conclude that the high volatility of the SOA gas phase species prevents them from binding irreversibly to the Teflon filter. As the 1.0 ppm α-pinene concentration utilized in this study did not allow for the preparation of samples with mass loadings below 0.5 μg, the α-pinene concentration was lowered to 0.125 ppm. As discussed above, this second set of reaction conditions resulted in condensational growth and lower concentrations. However, it also led to a smaller particle mode diameter (63 nm with geometric standard deviations of 1.29). To probe the condensationally vs coagulatively grown particle material by SFG, we had to take into account that the smaller particles, which were present at significantly smaller mass loadings (see Table 4), produced less SFG signal intensity than the larger particles. We recorded the SFG spectra of the smaller particles for 5 min, as compared to the 2 min used for the particles of larger mode diameter. To account for this difference in spectral acquisition times and instrument response functions, we determined the intensity ratio of ssp-polarized SFG spectra of α-pinene vapor in contact with a fused silica window collected using spectral acquisition times of 2 and 5 min. We then scaled the intensities of the SFG spectra of the collected particles for which SFG spectra were collected over 5 min by this experimentally determined ratio of 0.5 to compare them to the SFG intensities obtained from the spectra acquired for 2 min. Spectrum C in Figure 5, albeit less intense, shows no new additional features. In fact, the SFG spectra obtained from particles grown under conditions where coagulation dominated over condensational growth are quite similar, indicating that the distribution of SFG-active C−H oscillators is similar for particles prepared under both conditions. Regarding a limit of detection, Figure 5C shows that even the lowest mass loading of 24 ng, which was collected from the flow 8432

dx.doi.org/10.1021/jp405065d | J. Phys. Chem. A 2013, 117, 8427−8436

The Journal of Physical Chemistry A

Article

tube for only a few minutes, produced SFG response. With 24 ng of SOM distributed over the 47 mm-wide Teflon filter, we calculate that, within our assumptions, there would be 10 femtograms in the 30 μm laser spot generated in our SFG setup. As in any limit of detection study, interference from background signals needs to be assessed. As shown in Figure 5D, weak SFG signal intensities were produced on occasion by blank Teflon filters that had not been exposed to outflow from the flow tube. This residual SFG signal should not be impurities on the fused silica window that was used to press the filters against the sample stage because the window itself did not produce any SFG response (Figure 5E). Instead, the residual SFG signal intensity obtained from a minority of the bare Teflon filters used in this work could be due to C−H impurities present in Teflon itself, as indicated by 1H and 13C NMR spectra provided in the Supporting Information. These impurities, which may be those listed as “extractables with isopropyl alcohol < 1%” in the manufacturer product brochure, do not significantly affect our analysis. The average SFG signal intensity for the blank filters shown in Figure 5D is much lower than that of filter samples of low loadings (see Figure 5 panels B and C). Teflon filters from other manufacturers will be explored in future studies. 4. Nature of SFG Signal Generation from Randomly Arranged Particles Consisting of α-Pinene Derived Secondary Organic Material (SOM). Along with the limit of detection study, we investigated the dependence of the SFG intensity at the dominant peak around 2950 cm−1 on the number of particles in the 30 μm laser spot. This number was estimated for each mass loading by firstly finding the total number of particles on the filter, which was calculated by multiplying the volume of the gas passing through the filter (computed based on the sampling time and the collection flow rate) with the number concentrations of aerosol particles obtained from condensation particle counter (CPC) and assuming filter collection efficiency of 95%. The number of particles in the laser spot was then found by multiplying the number of particles per area of the filter with the total area of the laser spot. At least three spots on each filter containing a given mass loading were probed by SFG. In Figure 6 panels A and B, the average SFG intensity around 2950 cm−1 obtained for each particle sample is displayed as a function of average mass on filter (top x-axis) and number of particles per area of laser spot (bottom axis), calculated as discussed above. The individual SFG signal intensities, which are higher than the average response produced by three blank filters except for one measurement, are reported in the Supporting Information. Error bars in Figure 6 represent the standard deviation among the individual measurements and account for the randomness in the sample loading, including the 2−5% uncertainty associated with the collection of very low mass loadings. Figure 6A shows that the SFG signal intensity around 2950 cm−1 increases linearly up to a mass loading of 3 μg and then decreases slightly with higher loadings for particles with mode diameter of 93 ± 3 nm. The reduction we observe in the SFG signal intensities obtained for mass loadings exceeding 3 μg suggests that multiple light scattering processes occur, as discussed by Roke and co-workers in the context of decreased SFG signal intensities observed for high particle densities.49 The same authors also reported that the intensity of sum frequency scattering light obtained from spherical particles depends linearly on the particle density under conditions of low

Figure 6. Dependence of ssp-polarized SFG intensity on (bottom xaxis) the particle number density in the laser spot and (top x-axis) the filter mass loading (calculated from SMPS). Results are shown for particle populations having mode diameters of (A) 93 nm and (B) 63 nm (Table 4). Lines are fits to the data for a power function of the form y = a + bxn. Error bars denote the standard deviation among individual measurements of each particle sample. The dashed line denotes the signal intensity obtained from blank filters. Top and bottom axes are not proportional.

particle concentrations. This situation is similar to what had been reported earlier by Eisenthal and co-workers for second harmonic generation from microparticles.50,51 The linear dependence of the SFG intensity on particle density can be rationalized by realizing that the random distribution of the particles on the filter (or in the solution in the work by Eisenthal or Roke) should randomize the phase relationship among the SFG E-field produced by each particle, provided that the particles are separated over distances approaching or exceeding the SFG coherence length. As we show in the Supporting Information, the average particle−particle distance ranges from roughly 3 to 50 particle diameters for the samples studied here, satisfying, in principle, the condition for which the SFG coherences vanish. We note here that in contrast to the particle ensemble case, the SFG signals from each individual particle should be produced coherently from the SFG-active oscillators on it, provided that the oscillators are subject to nonzero molecular orientation distributions and that the strength of the incident E-fields is invariant over the relevant molecular length scales.50,51 We report herein experimental evidence to support some of the theoretical considerations discussed above for the particle samples studied in this work. Specifically, fitting a power function to the data shown in Figure 6A returns a power of 1.0 ± 0.4 for mass loadings below 3 μg, corresponding to calculated particle densities below 1200 particles in the 30 μm laser spot. For this fit, we included the signal intensity from the blank 8433

dx.doi.org/10.1021/jp405065d | J. Phys. Chem. A 2013, 117, 8427−8436

The Journal of Physical Chemistry A

Article

those obtained from particles collected at a field site in Southern Finland.20,31 The SFG spectra show one dominant peak near 2950 cm−1 for both condensational and coagulative growth regimes. The SFG signal intensity increases linearly with particle density in the laser spot, which suggests that the average number of SFG active oscillators per particle is constant for a given particle size, that the particles are arranged randomly on the Teflon filters, and that the particles are not coalesced. The lowest mass loading investigated here was 24 ng of SOM on the filter, corresponding to a calculated mass loading of 10 femtograms in the 30 μm-sized laser spot or approximately 100 particles for the size distributions of this study. This mass loading is adequately detected by SFG at collection times of a few minutes. This finding indicates that the high sensitivity of SFG spectroscopy for detection of SOM on filter samples can be exploited for fast, quantitative sampling approaches that might be needed to understand particle formation processes during nucleation events in the natural environment under conditions that are currently difficult to sample with other nondestructive, ambient analysis methods, such as infrared52−54 or Raman55 spectroscopy. Applications of SFG-based methodologies can thus provide for increased time resolution for the investigation of atmospheric particles. The limit of detection is likely improvable by collecting aerosol particles such that they are present in the form of an ordered array, for which coherent SFG signals could interfere constructively, as was discussed in the theoretical work of Xu and Zhang.56 In addition, mass measurements using a quartz crystal microbalance during filter collection are suggested to address some of the uncertainties associated with the quantification of the absolute number of particles on the collection filters at very low mass concentrations. Further experiments are suggested to use isotope labeling studies and computational work to assign the spectral response of α-pinene and SOM particles produced from it. Finally, the flow tube approach presented here will enable future studies, that, when combined with advanced organic synthesis, will allow us to follow (1) the interaction of isotope-labeled α-pinene and its oxidation products during their interaction with SOM prepared from wild-type (unlabeled) α-pinene and (2) the production of SOM from first- and second-generation precursors in order to “fast-forward” through particle formation mechanisms.

filter, taken to have one particle in the laser spot in order to avoid the singularity in the power function that would be produced by a “zero” x-value entry. The linear increase in the SFG signal intensity with particle density in the laser spot then supports the notion that the average number of SFG active oscillators per particle is constant for a given particle size, that the particles are present on the collection filters in a random array, and that the particles are not coalesced (i.e., mass is proportional to the particle count). However, we also find some exceptions from these general conclusions: Figure 6B illustrates that the SFG signal intensity for the filter samples containing the smaller (63 nm), condensationally grown particles increases continually, just as in the case of the larger coagulatively grown particles, for the mass loading studied here (24 ng to 0.24 μg). However, unlike for the larger particles, fitting a power function to the data set obtained from the smaller particles returns a power of 0.7 ± 0.3, clearly less linear than the value obtained for particles of 93 ± 3 nm. This departure from a linear power function could be attributed to (1) a breakdown of the assumptions made in the calculation of the number of particles in the laser spot at these very low mass loadings, (2) particle coalescence (however unlikely given the large interparticle distances), and/or (3) an increasing importance of wall-loss under the conditions of the experiment. Mass measurements using a quartz crystal microbalance during filter collection would help to address some of these uncertainties.

IV. CONCLUSIONS AND IMPLICATIONS FOR UNDERSTANDING ATMOSPHERIC AEROSOL PARTICLES Particle populations of secondary organic material were produced in a flow tube by the ozonolysis of α-pinene at various precursor concentrations. The flow tube approach presented here allows us to prepare particle populations having number−diameter distributions controlled either by condensation or coagulation. In the context of aerosol particle surface science, the conditions of SOM particle growth are particularly important. For condensational growth, processes such as physisorption and heterogeneous chemical reactions can dominate the growth processes. Recent work11 shows that at 30% relative humidity, particles prepared from the watersoluble component extracted from α-pinene-derived SOM have a high enough viscosity that they shatter upon being poked with a needle. Furthermore, the mixing time of an organic species inside a 100 nm sized particle would be at least 106 s under those conditions, suggesting that reactions inside the particle bulk are of limited importance. The relative humidity applied in our present study (