Real-time measurement of sodium in single aerosol particles by flame emission: laboratory characterization

A # ame emission aerosol sodium detector (ASD) has been developed to study the distribution of seasalt in individual marine aerosol droplets. The instrument detects sodium via D-line emission in a fuel-rich, laminar, hydrogen/oxygen/nitrogen # ame. Laboratory studies with synthetic monodisperse aerosols were carried out in order to characterize the sensitivity, precision, and linearity of the technique. Experiments were also carried out with aerosols generated from mixed salt solutions and seawater in order to determine whether ionic or other matrix e ! ects lead to interference. The ASD has a linear response function for NaCl aerosol particles from 100 nm to 2.0 (cid:1) m in diameter. The precision of sodium mass measurements is on the order of $ 3% standard error on replicate measurements, with a quantitative response to the sodium content of a single aerosol particle that is independent of the chemical composition of the particle, i.e. anions, cations, seawater. No interferences were found with major ions in seawater and common atmospheric aerosols. These experiments demonstrate a detection limit equivalent to a 100 nm diameter dry 100% NaCl aerosol. (cid:1)


Introduction
Seasalt aerosol particles are an important component of marine air a!ecting both physical processes, such as light scattering and cloud droplet growth, and chemical process involving the uptake and release of reactive gases (Livingston & Finlayson-Pitts, 1991;Chameides & Stelson, 1992;Sievering et al., 1992;Keene et al., 1998). Seasalt particles are produced at the ocean surface by the bursting of entrained air bubbles produced by the action of wind (Woodcock, 1953;Blanchard & Woodcock, 1957, 1980Blanchard, 1983;Monahan, Spiel, & Davidson, 1986;Gong, Barrie, Blanchet, & Spacek, 1998). Mechanical generation typically produces particles ranging from 0.1 to 100 m in diameter. The production rate of sub-micron primary seasalt particles is poorly known. Although non-seasalt (nss) sulfate particles have traditionally been considered to dominate this size range based on early measurements (for example, Hobbs, 1971), recent indirect evidence has suggested that sub-micron seasalt production may, in fact, be signi"cant (O'Dowd & Smith, 1993;O'Dowd, Smith, Consterdine, & Lowe, 1997). More recently, single aerosol mass spectrometer measurements at Cape Grim, Tasmania, showed that nearly all particles larger than 0.13 m in diameter contained seasalt, and seasalt was a signi"cant fraction of the CCN (Murphy et al., 1998a,b). To resolve the relative contribution of seasalt aerosol particles to number and mass distributions in the sub-micron size range, more direct quantitative measurements of sodium levels in ambient single aerosol particles under a wide range of wind speeds and local conditions are clearly needed.
One approach to measuring sodium quantitatively is #ame spectrometry. This technique has been extensively used in the past to analyze nebulized solutions of sodium salts in both laminar and turbulent #ames (Alkemade & Herrmann, 1979a). Numerous experimental studies have demonstrated a linear relationship between the integrated intensity of absorption, #uorescence and emission and the concentration of ground state atoms, provided that the optical density, i.e. atomic concentration, is low (Zeegers, Smith, & Winefordner, 1968). At high concentrations, the relationship is no longer linear due to self-absorption e!ects in optically thick #ames (Zeegers et al., 1968). Other causes of non-linear behavior include incomplete desolvation and volatilization of the nebulized droplets in the #ame, and incomplete dissociation of the salts to give free sodium atoms (Alkemade & Herrmann, 1979b,c).
A variety of #ames have been used in analytical #ame spectroscopy studies, ranging from acetylene/air to hydrogen #ames (Alkemade & Herrmann, 1979a). A hotter #ame reduces potential problems with incomplete volatilization and dissociation. Early studies showed that sodium emission in a hydrogen #ame is practically the same for all sodium halides and other sodium salts (Alkemade & Herrmann, 1979a). More recently, many researchers have studied the chemistry of sodium in hydrogen/air #ames of varying composition in great detail (e.g. Hynes, Steinberg, & Scho"eld, 1984). These studies all utilized steady-state #ame concentrations resulting from the large numbers of droplets produced by nebulizing solutions.
There have been very limited measurements of sodium in #ames from single aerosol particles. One early device counted sodium-containing atmospheric aerosol particles when they entered a hydrogen #ame (Soudain, 1951;Vonnegut & Neubauer, 1953). This system used turbulent #ames on two burners for large and small particles. Aerosol particles were entrained with the air #ow into both #ames, with transmission e$ciencies less than 10%. This system was subject to large #uctuations in background signal from the turbulent #ames, masking signals from small particles.
The system overall had very low sensitivity, with an estimated detection limit of 1 m diameter particles (Vonnegut & Neubauer, 1953). This apparatus was not calibrated to give a quantitative response, but simply used as a sodium counter.
A later variation of the salt counter incorporated an acetylene pre-mixed laminar #at #ame to ensure all particles would traverse the same temperature pro"le (Radke & Hobbs, 1969). In this instrument, atmospheric aerosol particles were also sampled by entrainment into the air #ow drawn in by the burner at a rate that ensured proper burning. Since particles were introduced into the #ame over the entire burner head region and would most likely not be imaged evenly by the detection optics, reproducibility was probably poor. An improved version of this sodium #ame spectrometer was later #own over the Paci"c ocean (Hobbs, 1971). Sodium particles in the range 0.06}0.68 m diameter of equivalent dry NaCl particles were measured, with very low sodium particle counts above 0.23 m. No details on instrument calibration, precision or design changes to the 1969 version were given.
A sodium-particle counter was most recently deployed on several #ights o! the east coast of the US in 1993 (Hegg, Hobbs, Ferek, & Waggoner, 1995). This system used a #ame photometric detector with a reducing hydrogen/air #ame, coupled to a pulse height analyser, to measure sea salt size distributions between 0.1 and 5 m in diameter. The instrument was calibrated with NaCl particles generated by an atomizer/electrostatic classi"er, but precision and transmission e$ciencies are not reported. Seasalt particle concentrations in this study ranged from 0.2 to 2 cm\, comprising just a few percent of the sub-micron aerosol particles.
We have developed an improved aerosol sodium detector (ASD), which is designed to detect and quantify sodium in single aerosol particles from 100 nm to 2.0 m with high precision, via thermal emission in a hydrogen/air #ame. In this paper, we describe the design and operational principles of the instrument and present the results of laboratory tests with monodisperse aerosols to characterize its performance in terms of linearity of response, precision, sensitivity and potential chemical interferences. Some preliminary measurements on ambient aerosols from marine air are also presented.

The aerosol sodium detector (ASD)
The basic principles of operation of the ASD are the volatilization of aerosol particles in a hightemperature #ame, decomposition of the sodium salts to give sodium atoms, and detection of the emission at 589.0 (D line) and 589.6 nm (D line) from thermally excited sodium atoms. The ASD consists of the following components: (1) an aerosol sampling and injection system to introduce aerosol particles into the #ame, (2) a pre-mixed laminar hydrogen/air #ame for volatilization of the aerosol and atomization of sodium salts, and (3) a PMT, sodium "lter and associated electronics for data acquisition. A schematic of the ASD is shown in Fig. 1 and each of its components is discussed in more detail below. Unlike the earlier versions of sodium aerosol #ame spectrometers (Soudain, 1951;Vonnegut & Neubauer, 1953;Radke & Hobbs, 1969), aerosol particles in the ASD are introduced reproducibly into the center of the laminar #at #ame through the aerosol inlet capillary. All sampled aerosol particles traverse the same region imaged by the optical detectors, giving high precision sodium emission measurements.

Aerosol sampling
Aerosols are sampled by drawing air at 1 l min\ through a system of electronically controlled on/o! valves into a copper collection tube 53 cm long by (9.5 mm) OD. An early version of the ASD sampled air directly into the burner induced by pumping on a sealed chamber surrounding the burner head. This proved unsuccessful due to instabilities in the aerosol #ow. In the "nal con"guration of the ASD, the collected sample is periodically injected into the #ame front through the aerosol inlet capillary under pressure with "ltered laboratory compressed air (0.25 l min\). During a sampling cycle, the compressed air bypasses the sampling tube and enters the #ame front directly (Fig. 1). Valve switching is controlled and monitored by the data acquisition program. To minimize particle losses by impaction, the aerosol #ow path is unobstructed and sharp bends in tubing were avoided. Particle transmission e$ciencies through the sampling/injection system are estimated to be '95% for particles below 2 m in diameter based on calculated particle settling velocities and residence times for the #ow rates used in this system (Reynolds number"147; Stokes number"0.084; residence time"3 s; settling velocity for 2 m particle"0.0261 cm s\).

Burner
The ASD burner consists of a machined aluminum body incorporating a stainless-steel honeycomb head 0.5 (1.27 cm) in diameter that can be easily removed for cleaning. A simpli"ed schematic of the burner is shown in Fig. 2. The #ame gas #ow rates used in the laboratory study were 3.0 l min\ hydrogen, 5.8 l min\ air, and 3.4 l min\ nitrogen, giving a stable fuel-rich #ame with a temperature &1600 K. A fuel-rich hydrogen/air #ame minimizes the formation of NaOH and NaO which result in a net loss of elemental sodium from the system (Hynes et al., 1984). The colorless laminar #ame is &7 cm in height, with a 2 mm high bright blue zone at its base. A nitrogen sheath #ow of 2.2 l min\ around the honeycomb head minimizes entrainment of aerosol particles from the room air. All gases are "ltered with HEPA capsule "lters. The aerosols are introduced at a #ow rate of 0.25 l min\ into a well-de"ned central region of the #ame through  a capillary tube. The linear velocities of the aerosol and pre-mixed #ame gas #ows below the burner head are matched to ensure laminar #ow conditions and minimize particle losses from turbulence.

Signal detection and data acquisition
The light emitted from the thermally excited sodium atoms at 589.0/589.6 nm (P and P Na doublet) is imaged by a telescope (consisting of iris, mask, two lenses, a sodium interference "lter) and detected with a 1P28 photomultiplier tube (Hamamatsu). A 10 nm broadband "lter was used for the linearity and precision experiments discussed below in section 3.1. For all subsequent experiments a 0.1 nm narrow line "lter was used, reducing sodium signal by a factor of approximately 3.4. The narrow line "lter reduced the sodium signal by a factor of approximately 3.4 relative to the broad-band "lter. The telescope is used to image a roughly cylindrical volume approximately 6 mm in diameter, which intersects the #ame (approximately 12.6 mm diameter), 2 cm above the burner head. The response of the instrument to a given mass of sodium varied during the course of this study because of changes in iris diameter and PMT alignment. Each individual experiment was carried out under a single set of conditions, but it is not meaningful to compare the absolute magniture of the response between di!erent experiments. The PMT signal is passed to an adjustable gain, high-input impedance ampli"er, and recorded using a PC-based oscilloscope board. Data acquisition is triggered by the emission signal. The duration of a sodium signal from a single particle is approximately 2 ms. Fig. 3 is a plot of the signal acquired from a single 1.25 m diameter NaCl aerosol. System`backgrounda signals are collected by running the aerosol inlet #ow to the #ame through a "lter and setting the delay generator to internally trigger.
Data analysis consists of integrating the area of the sodium emission peak relative to the baseline PMT signal. These peak areas are normalized by the gain of the signal ampli"er. Aerosol acquisitions with multiple aerosol events in a single shot are discarded to avoid errors in the calculated areas. As a measure of the precision of the emission signal, the standard error of the mean peak area for a measurement of 100 monodisperse aerosol particles is typically $3% (2 ).

Calibration
The ASD is calibrated with monodisperse aerosols of known sizes, produced by a vibrating ori"ce aerosol generator (VOAG; John, 1993). Aqueous salt solutions ranging from 0.2 to 4;10\ M are forced through a 20 m platinum ori"ce by a nitrogen backing pressure of 40 Psi (240 kPa). All sodium salts are standard purity from Aldrich. Salts are dried for at least 24 h at 1003C and cooled in a vacuum dessicator prior to solution preparation in deionized MilliQ water. The ori"ce is vibrated by a piezo electrode with a square wave pulse (frequency"272.5 kHz; amplitude"30 V) to break up the liquid stream, producing large monodisperse droplets. The droplets are entrained in an air #ow of 1.75 l min\ to prevent coagulation. The droplets are then dried in a column with dry "ltered compressed air (10 l min\), and sized with a commercial aerodynamic particle sizer (TSI 3310A; Baron, Mazunder, & Cheng, 1993).
The initial size of the droplet as a geometric diameter produced by the VOAG can be calculated from (John, 1993) where f is the frequency of ori"ce vibration (s), Q is the liquid #ow rate (cm s\) and D is the droplet diameter (cm). For the current set of operating conditions, the frequency is 272.5 kHz, and the #ow rate was measured at 0.4 cm min\, giving an initial geometric droplet diameter of 37 m. The "nal size of the particles generated in these experiments varied from 0.8 to 2 m in diameter, depending on the composition (i.e. density of the salt) and initial concentration of the salt solutions used. The sodium mass contained in a single aerosol is calculated from the initial sodium concentration, the initial droplet volume and the molar mass of sodium. In the experiments described here, sodium levels ranged from 0.004 to 0.24 pg in a single aerosol particle. The accuracy of the sodium calibration is dependent on the calculation of the initial droplet volume. The largest source of uncertainty in the calculated initial droplet size lies in the liquid #ow-rate measurement, giving a potential systematic error of $10% in droplet volume.

Field setup for ambient measurements
In April 2000 during the Shoreline Environment Aerosol Study (SEAS), ambient aerosols were sampled through an aerosol intake system on the tower at the University of Hawaii's meteorological research site at Bellows air force base on the southeast coast of Oahu, Hawaii. The aerosol intake was located at 12 m above sea level and consisted of a (1.27 cm) OD goose-neck stainless-steel tube facing into the prevailing winds. The inlet was connected to (9.52 mm) OD black, high-carbon-content plastic tubing (TSI) run vertically down to the ASD. A pump drew outside air through the inlet tube at a rate of 11 l min\ (Reynolds number" 1219; residence time"10 s; Stokes number"0.18, calculated for a 4.8 m particle * 50% cut point at 0.47 for a round tube inlet). Inside the laboratory, a smaller #ow (0.4 l min\) was iso-kinetically drawn o! from the main #ow through a (3.18 mm) OD stainless-steel capillary into a 122 cm long, (6.35 mm) diameter PVC Na"on drying tube (counter-air drying #ow"0.8 l min\; Reynolds number"267; residence time"0.24 s; Stokes number"0.34). This ran directly to the intake of the aerosol injection system. Measurements with an aerodynamic particle sizer (APS; TSI) demonstrated '95% transmission of all aerosol particles (2 m in diameter through the sampling system. For these ambient aerosol experiments, a smaller, 0.25 (6.35 mm) OD burner head and correspondingly lower #ame gas and aerosol injection #ows were used (H "1.23 l min\; N "0.6 l min\; air"3.06 l min\; sheath"1.02 l min\; aerosol"0.117 l min\). For the "eld experiment, a small chamber #ushed by "ltered air at 11 l min\ was "tted around the burner head and detection optics to minimize entrainment of laboratory aerosol particles into the #ame.

Linearity of response/precision
In order to assess the linearity of the response of the ASD, a series of solutions of di!erent NaCl concentrations were introduced into the VOAG. NaCl concentrations ranged from 0.12 to 3.08;10\ M NaCl, giving aerosol particles in the 0.9}2 m diameter size range after conditioning in the drying column (aerodynamic diameters measured by the APS). The sodium content of these particles ranged from 0.006 to 0.18 pg. These dried particles are probably of irregular geometry and may consist of crystal aggregates. No attempt was made to measure sodium in larger aerosols due to the low system transmission e$ciency above 2 m. Fig. 4 shows emission signals as a function of sodium content for single aerosols. Emission signals were linear (r"0.998), increasing with increasing sodium concentration as expected. The linearity of the response suggests that: (1) volatilization of the aerosols is complete, and (2) the #ame remains optically thin (i.e. no evidence of self-absorption).
Standard errors for each point (corresponding to a &100 particle data set) are shown as y-error bars in Fig. 4. Errors vary from 1.8 to 3.6%, with a mean value of 2.3%, demonstrating excellent Fig. 4. Emission vs. sodium content for single monodisperse NaCl aerosol particles. Each point represents the mean of the integrated emission signal for approximately 100 particles. Error bars represent the standard error for sodium emission and $10% for sodium content/particle. short-term precision. However, there were apparent departures from linearity that exceeded the uncertainty in the individual data sets. We believe these re#ect variations in #ame temperature resulting from drift in the #ow rates of the #ame gases, which were not mass #ow controlled during these experiments. The response of the ASD would be expected to vary with #ame temperature as a result of the temperature dependence of the Maxwell}Boltzmann distribution of sodium atoms in the P and P excited states in the #ame. To further probe the linearity of the system response, this experiment was repeated with a series of NaCl/KCl mixtures, ranging from 100% down to 4% v/v NaCl. In all cases, the total chloride concentration of the solutions was 1.65;10\ M, and the aerodynamic diameters of the aerosol particles measured by the APS were 1.3$0.2 m. Fig. 5 shows emission as a function of sodium content in single mixed NaCl/KCl aerosol particles. Again, the response function is linear within the #ame temperature uncertainty from 0.004 to 0.10 pg Na in a single aerosol particle.

Chemical interferences
Although the atomic emission from sodium is quite speci"c, the chemical composition of the non-sodium components of the aerosol could a!ect the response of the ASD in a number of ways. For example, chemical composition might a!ect the rate of volatilization and decomposition of the sodium salts, and other aerosol constituents could a!ect the #ame chemistry in such a way as to alter the abundance of atomic sodium relative to non-emitting forms such as sodium oxide. To determine if the sodium response of the ASD depends only on the amount of sodium present and not on the nature of the other chemicals present in the aerosol particle, the emission was measured for a series of sodium salts with anions that are important constituents of seawater and/or seasalt aerosols (NaBr, NaNO , Na SO and Na CO ). Salt concentrations for these experiments were 1, 2, 3, and 4;10\ M, giving monodisperse aerosol particles with sodium contents ranging from Fig. 6. Emission signals vs. sodium content (in pg) for single aerosol particles containing univalent sodium salts and seawater. Error bars are not shown for ease of viewing, but are &3% for sodium emission and $10% for sodium content/particle (as in Figs. 4 and 5).  5. Emission signals vs. sodium content (in pg) for mixed NaCl/KCl single aerosol particles. Error bars are &3% standard error for sodium emission and $10% for sodium content/particle. 0.06 to 0.24 pg, and 1}2 m in diameter. Fig. 6 shows the results for NaBr and NaNO . NaCl is included for purposes of comparison. All three salts show a linear response with very similar sodium emission signals for single aerosol particles with the same sodium content, i.e. the sodium response is independent of the nature of the anion. Divalent sodium salts also show a linear response for sodium. This is shown in Fig. 7a and b for Na SO and Na CO , respectively. Again, this linear response indicates that there are no problems with complete volatilization and atomization of these salts in the #ame. Since seawater largely consists of NaCl and MgSO , a mixture of these two salts was tested and also gave a linear Fig. 7. Emission signals vs. sodium content (in pg) for divalent sodium salts in single aerosol particles Error bars are not shown, but are on the order of &3% standard error for sodium emission and $10% for sodium content/particle (as in Figs. 4 and 5). response, with no apparent salt e!ects (Fig. 8). Experiments were also carried out with seawater (Fig. 6). Gulf of Mexico seawater (collected in November 1999 on board the R/V Calanus) was "ltered through a 0.2 m "lter under vacuum to remove particulates. The "ltrate was diluted with MilliQ water to give a set of three solutions with sodium concentrations of 1, 2 and 3;10\ M, based on an initial concentration of 10.775 g Na kg\ seawater. The sodium response of the single aerosol seawater particles is also linear, and identical to that from the other sodium salts. Although seawater is a complex mixture of many salts and trace organic and inorganic constituents, the results shown in Fig. 6 indicate that none of these constituents signi"cantly interfere with sodium emission under these conditions. Fig. 9. Emission signal vs. sodium content of a single aerosol particle (in pg) for the high gain PMT. Error bars shown are standard errors for emission signals and $10% for Na mass/particle. The highest and lowest points correspond to NaCl particles of 429 and 300 nm diameter (DMA) respectively (RH &25%).

Low-level detection limits
To determine the low-level detection limit of the ASD, Fig. 9 shows a plot of sodium emission signals as a function of sodium content for sub-micron monodisperse NaCl aerosols generated by the VOAG using a high-gain PMT (Hamamtsu, R3896). NaCl solutions ranged from 1.4;10\ to 0.58;10\ M NaCl, i.e. a factor of 10}100 times lower sodium concentration than used in the linearity/chemical interference studies. The NaCl aerosol particles produced by the VOAG ranged from 300 to 429 nm in diameter after conditioning in the drying column, as measured by a Di!erential Mobility Analyzer (TSI 3010;Yeh, 1993). For the most dilute solutions, measured particle sizes were larger than sizes calculated for dry salt particles on the basis of the VOAG size and salt concentration. This di!erence is attributed to impurities introduced into the salt solutions by glassware and tubing used to deliver solution to the VOAG. The signal : background ratio for the lowest calibration point was &10 : 1 (based on the measured integrated emission signal vs. the average intercept of the sodium calibration curves extrapolated to a sodium concentration of zero). The lowest sodium level measured in this study would correspond to a &100 nm diameter, completely dry, pure NaCl aerosol particle (1;10\ pg Na). This is the detection limit of the ASD in its current con"guration.

Ambient aerosols
An ASD was deployed in April 2000 during the SEAS intensive at the University of Hawaii's coastal meteorological research site (Bellows air force base, Hawaii). At this sampling site, winds were consistently out of the East over the ocean at speeds from 6 to 8 m s\ and RH&70}80%, with an o!-shore reef break contributing locally to seasalt aerosol production (Tony Clark, University of Hawaii, private communication). An example of typical results from this "rst "eld  Fig. 10. Winds were easterly at 6 m s\, with a relative humidity of approximately 75%. Emission signals were converted to sodium content (in pg) with a four-point NaCl calibration curve. Data are reported as equivalent diameters of dry NaCl aerosol particles.
The maximum in sodium-containing particles is observed between 200 and 800 nm, with a concentration of approximately 1.2 particles cm\ in the entire interval (100 nm}2.0 m) sensed by the ASD. It should be noted that the particle transmission e$ciency through the ASD burner is not included in this estimation, and the actual sodium-containing particle concentration may be higher than given here. The peak maximum in the sodium-containing particles is in contrast to the accumulation mode peak maximum of &100}200 nm measured over the same time period at the same site by the University of Hawaii, with total particle concentrations of 100}150 particles cm\ (Tony Clark, private communication). These results suggest that sub-micron seasalt aerosol particles compose just a few percent of the sub-micron aerosol number distribution under the conditions at this site. This is in accord with the low levels of sub-micron sodium-containing particles measured by Hobbs (1971) in the Paci"c Ocean o! the west coast of the USA at surface wind-speeds of 4 m s\. These "eld results will be described in detail in a subsequent paper.

Summary
A single-particle sodium detector (aerosol sodium detector; ASD) has been developed based on thermal sodium emission in hydrogen/air #ames. Laboratory experiments with synthetic monodisperse aerosols demonstrate that the ASD has: E a linear response function over the size range of interest (0.1}2.0 m diameter); E precision on the order of $&3% standard error on replicate measurements; E a quantitative response to the sodium content of a single aerosol particle independent of the chemical composition of the particle, i.e. anions, cations, seawater; E a lower detection limit of &100 nm equivalent diameter 100% dry NaCl aerosol.
The ASD can provide quantitative near real-time measurements of sodium in single aerosol particles in the sub-micron size range. The ASD will allow us to directly measure the relative importance of seasalt aerosol particles in this size range under a wide range of ambient conditions and to test current models for seasalt aerosol production as a function of wind speed.