E-cigarette Aerosol Collection Using Converging and Straight Tubing Sections: Physical Mechanisms

Abstract

Hypothesis

Identification and quantification of harmful chemicals in e-cigarette aerosol requires collecting the aerosolized e-liquid for chemical analysis. In 2016, Olmedo at al. empirically developed a simple method for aerosol collection by directing the aerosol through a sequence of alternating straight and converging tubing sections, which drain the recovered e-liquid into a collection vial. The tubing system geometry and flow conditions promote inertial impaction of aerosolized e-liquid on tube walls, where it deposits and flows into the collection vial.

Experiments

We use high-speed optical imaging to visualize aerosol transport in proxies of the collection system. We also determined collection efficiencies of various configurations of the collection system.

Findings

A turbulent jet emerges from converging conical sections and impinges onto the wall of downstream tubing sections, resulting in inertial impaction and deposition of the aerosol. For inertial impaction to occur the tip radius of the converging section must be small enough for a jet to be formed and the sequence of tubing sections must be curved in a polygon-like manner such that the jet emerging from a converging section impinges on the downstream tube wall. The collection efficiency is significantly smaller without such curvature.

1. INTRODUCTION

Electronic cigarettes (e-cigarettes) generate aerosols by heating up liquid solutions of propylene glycol (PG) and vegetable glycerin (VG) (e-liquids) that may contain nicotine and flavors.¹ Despite the increasing popularity of e-cigarettes, much is unknown about the adverse human health effects of e-cigarette use.² Adequate characterization of the health risks of e-cigarette use requires analyzing not only the chemical composition of the unused e-liquid but, more importantly, of the aerosolized e-liquid inhaled by the user. The composition of the aerosol is different from that of the unused e-liquid, because new chemicals are formed during device usage (e.g. aldehydes^3–6) and chemicals are released from the e-cigarette device itself to the aerosol (e.g. metals^7–9).

Chemical characterization of e-cigarette aerosol requires the ability to collect it. Sampling techniques that collect the aerosol on filters,^10–13 in impingers,^14–16 and in sorbent tubes^10,17–19 can be used, but they require the aerosol to be diluted into or later extracted from a collection medium. In contrast, Olmedo et al.²⁰ developed a method that avoids this extraction step, thus simplifying sample processing and preventing possible contamination of the aerosol sample. In their method, aerosol collected from an e-cigarette is directed through a tubing system consisting of a sequence of “aerosol collection units.” Each collection unit consists of a converging tubing section (pipette tip) connected to a straight tubing section, where the tip radius of the converging section is significantly smaller than the radius of the straight section. Within the tubing system, the recovered aerosol forms bulk e-liquid phase and flows into a collection vial for later chemical analysis.

Olmedo et al.’s method has been used in various studies for different types of e-cigarettes^8,9,21,22, including open-system devices (e.g., “Mods”), which allow users to replenish the e-liquid, change heating coils, and adjust power output, as well as closed-system devices (e.g., “Pods”), which do not offer these features. Typically, about 1 g of aerosol is collected for chemical analysis, and aerosol recovery efficiencies range between 70 and 90%. However, as the method was developed heuristically, there is no clear understanding on the number of collection units recommended in different cases. For example, Zhao et al.⁹ used one collection unit for “Mod” devices but three collection units for “Pod” devices.

The objectives of this work are (1) to explain the physical mechanisms that are responsible for the effectiveness of the e-cigarette aerosol collection method empirically developed by Olmedo et al.²⁰ and (2) to examine the role of the geometry in the collection system, that is, to evaluate the importance of the relative angle between the converging and straight tubing sections. To accomplish these objectives, we used proxies of the original collection device, in which we replaced a round straight tubing section by a square one, to visualize aerosol transport at high speed. We interpreted our observations based on first principles of fluid mechanics and aerosol science. Moreover, we performed supporting experiments using variations of the original tubing system, in which all tubing sections had circular cross-section, to elucidate the effects of the number of collection units and alignment of tubing sections on the overall aerosol collection efficiency.

2. BACKGROUND: ORIGINAL AEROSOL COLLECTION SYSTEM

The system used for actual/real aerosol collection consists of a variable number of aerosol collection units (Figure 1 and Figure 2a). All the tubing sections used in these systems have a circular cross-section. Each aerosol collection unit begins with a 250-μL pipette tip (Supersilk, Labcon. Petaluma, CA, USA) connected downstream to a Tygon tubing with uniform cross-section. The narrow end of the pipette tip has an inner tip radius R_tip = 0.28 mm and the Tygon tubing has an inner radius R_Tygon = 0.75 mm (S3 E-3603, Saint-Gobain Corporation, France) and is approximately 2 cm long. The upstream end of the first collection unit is connected to peristaltic pump tubing (Masterflex L/S 15, Vernon Hills, IL, USA) with inner radius R_per = 2.4 mm and an approximate length L_per = 30 cm. The peristaltic pump tubing is guided through the head of a peristaltic pump (Masterflex L/S Digital Drive, 600 RPM with Masterflex L/S Easy-Load II Head) and then connected to the mouthpiece of an e-cigarette. The first collection unit can be connected to additional identical collection units. The original method uses four aerosol collection units.²⁰ This original design was later modified⁸ to use a different number of aerosol collection units depending on e-cigarette type (Figure 1). The pipette tips of the additional units are cut down to a 2-cm length (leaving the narrow tip portion intact) so that they can be connected to the Tygon tubing (which has a smaller diameter than the peristaltic pump tubing). The Tygon tubing of the last collection unit drains the recovered aerosol into a 1.5-mL centrifuge tube for later chemical analysis.

Figure 1:

Sketches of collection systems for aerosol from a “Mod” and “Pod” e-cigarette. Green rectangles indicate imaged tubing regions.

Figure 2:

Depiction of the six different geometric scenarios we examined. The two scenarios on top used tubing materials of circular cross-section and were used to measure aerosol collection efficiencies. In the four scenarios at the bottom, one round straight tubing section was replaced by a square one for aerosol visualization. Virtual vertical planes were added to illustrate the downward or backward curvature of the polygonal tubing systems. Note that for Setups a and c, the number of aerosol collection units ranged from 1 to 3.

Note that a pipette tip can be tightly connected to an adjacent, slightly flexible Tygon tubing section, even without aligning their axes. As indicated in Figure 1, this feature introduces a polygonal-like downward curvature of the entire train of aerosol collection units. In fact, the peristaltic pump tubing emerges nearly horizontal (~25° angle) from the peristaltic pump head and the final tubing section finishes with an approximately vertical orientation in the collection tube.

3. METHODS

3.1. Collection Systems used in the Visualization Experiments

We imaged the aerosol in the straight tubing section of one of the aerosol collection units, also capturing the outlet of the upstream pipette tip and the inlet of the downstream pipette tip. To avoid optical distortions due to light refraction, we replaced the straight Tygon tubing section of the selected aerosol collection unit by a square glass tubing section, i.e., a square cross-section borosilicate glass straight tubing section (3 mm inside side length, BST-3-50, Friedrich & Dimmock, Inc., Millville, NJ, USA). Pipette tips identical to those used in the original collection system are inserted on both ends of the square glass tube. We sealed both connections by wrapping flexible laboratory film (Parafilm M, American National Can, Chicago, IL, USA) around them.

For aerosol collection from a “Mod” device, imaging was performed in the straight tubing section of its single aerosol collection unit⁹ (corresponding to the top schematic drawing in Figure 1 but replacing the circular tube with the square glass tubing section as mentioned). To mimic the actual aerosol collection system, the upstream, uncut pipette tip pointed slightly toward the upper wall of the square glass tubing section (see “Original Visualization System” in Figure 2c). For better process understanding, we also imaged aerosol collection using the following three modifications of the “Original Visualization System:”

1. “Aligned Visualization System” (Figure 2d). In order to examine the effect of the relative angle between the converging and straight tubing sections we considered a case in which they are both aligned in the flow direction.

2. “Rotated Visualization System” (Figure 2e). In an attempt to characterize the threedimensional distribution of collected aerosol, we substituted the natural polygonal downward curvature of the tubing by a backward bend (away from the imaging system).

3. “Larger Tip Visualization System” (Figure 2f). In order to evaluate the importance of using a converging flow channel (pipette tip), we increased the radius R_tip of the pipette tip from 0.28 mm to 1 mm by cutting away a portion of its narrow conical end (thereby reducing its total length).

Imaging was performed only during the first puff, because aerosol deposition on the tube walls hindered view into the tubing section during subsequent puffs.

For aerosol collection from a “Pod” device, imaging was performed in separate experiments in either the first or the last aerosol collection unit (corresponding to the bottom schematic drawing in Figure 1 but using a square glass tubing, as discussed). Due to the lower aerosol mass per puff of ”Pod” devices when compared to “Mods,” we imaged the aerosol for 41 puffs until aerosol deposition on the tube walls significantly obstructed the view. (For actual aerosol collection from “Pods”, up to 300 puffs might be needed to collect 0.5 g of recovered aerosol.)

For each configuration, imaging experiments were performed at least in duplicate.

3.2. Collection System used for Collection Efficiency Quantification

To elucidate the effects of various features of the aerosol collection system on the overall aerosol collection efficiency, we examined the collection efficiency of four variations of the “real” collection systems, which we define here as those using tubing sections of circular cross-section:

1. We determined how the aerosol collection efficiency of the “Original System” (Figure 2a) depends on the number of collection units (n = 1 to 3). The bottom schematic drawing in Figure 1 illustrates a system for n = 3.

2. We measured collection efficiency in an “Aligned System” (Figure 2b), in which we connected the pipette tip to a 13-cm long downstream Tygon tubing section, where the axes of the tip and the Tygon tubing were aligned at the junction point. The relatively long tubing section was gradually bent downward such that collected aerosol could drain into a vertically oriented collection vial. See Figure B in Supplementary Material for a photo.

For each of the four collection device variations, triplicate experiments were performed together with the Pod e-cigarette. Like for real aerosol collection from Pods, we ran the experiments until about 0.5 mL of aerosol accumulated in the collection vial. The puff time was T_puff = 4 s, and the puff period was 15 s.

3.3. Puff topography

As described in Hilpert et al.²³ we used an Arduino™ micro-controller to control both a solenoid that pushed the activation button of a “Mod” device (i-Stick 25) and a relay that short-circuited two pins on a remote control port to start the peristaltic pump. This setup allows programming an arbitrary puff topography. In this work, we puffed the “Mod” device only once, because the aerosol deposited on the tube walls during the first puff obstructed the view on the aerosol during subsequent puffs. We activated the Mod device for a puff time T_puff = 3 s. The peristaltic pump was operated for 4 s, i.e. longer than the puff time T_puff, in order to remove most of the aerosol present in the tubing. The flow rate was Q = 1 L/min. The e-cigarette was operated at a power of 60 W with a manufacturer supplied HW2 coil. For the e-liquid, we used Pale Whale Vixen’s Kiss consisting of a 20% propylene glycol (PG) / 80% vegetable glycerin (VG) mixture with zero nicotine content, a flavor we currently use in rodent exposure experiments.²³

We used a “Pod” device (JUUL, purchased in 2017) with a Virginia Tobacco flavored e-liquid consisting of an about 30% PG / 70% VG mixture²⁴ with 5% by weight nicotine content. Activation of this device was triggered by the vacuum created by the peristaltic pump, which as in the case of the Mod device, was operated for 4 s at a flow rate Q = 1 L/min.

3.4. Imaging

Imaging was performed with a high-speed camera (Basler A504kc).^25,26 A 20-27 mm zoom lens (Schneider-Kreuznach, Germany) was used together with two 10-mm macro extension tubes (Schneider-Kreuznach, Germany). The camera was operated in free-running mode yielding a frame rate of 86 fps at a resolution of 1280 × 450 pixels (full resolution: 1280 × 1024). Images were transferred to a PC using a frame grabber card (National Instruments PCIe-1429), and captured images saved to a movie file using LabView software (version 9.01, 32-bit).

To reduce light reflections, the collection system was secured to a vertical backboard covered by light-absorbing black duveytyne. This setup was illuminated from the front with a high-intensity LED light (Cygolite, Metro Plus 650). Due to the small depth of field of the camera/lens system and the relatively small imaged region of interest (~3 by 20 mm), we needed the ability to adjust finely the optical system. We therefore mounted the backboard onto a vertical optical translation stage, and the camera onto a twodimensional optical translation stage that allowed moving the camera forward and backward, and left and right. Figure 3 shows the experimental setup.

Figure 3:

Photo of experimental setup for imaging aerosol in the collection system.

4. RESULTS

4.1. Visualization of Aerosol Collection from a Mod E-cigarette

In Figure 4a, we first present results for the “Original Visualization System” which resembles the actual collection device. In this case, the tubing system is curved downward. In Fig. 4a.iii, a jet carrying aerosol particles (aerosol jet) is observed as it is ejected from the pipette tip into the downstream straight tubing section. This aerosol jet is aligned with the axis of the pipette tip and directed toward the upper wall of the straight tubing section, due to the relative angle between the pipette tip and the downstream tubing section. As the jet hits the wall of the straight tubing section, aerosol droplets appear on the glass wall. The accumulation starts at the position on the wall that is directly in front of the pipette tip. With time, the deposited droplets form a stream of bulk e-liquid that flows toward the next tubing section (Fig. 4a.iv). Note that aerosol droplets already appear on the wall during an earlier phase of aerosol injection (Fig. 4a.ii), before the jet can be clearly observed. Large e-liquid droplets can be observed in the pipette tip, and these droplets eventually flow from the tip into the straight tubing section (Fig. 4a.iv), suggesting additional collection in the converging section. After puff completion, a substantial amount of stagnant bulk e-liquid is present at the top of the straight tubing section.

Figure 4:

Four selected snapshots (indicated by roman numbers in red boxes) for aerosol collection from a “Mod” device for four different collection scenarios. (a) System representative of the actual aerosol collection unit where the pipette tip points toward the upper wall of the straight tubing section. (b) Modification where the collection unit is rotated such that the pipette tip points toward the front wall of the straight tubing section. (c) Modification where the collection tubing system is aligned. (d) Modification in which a pipette tip with larger radius is used. The first snapshot i always shows the unit before aerosol injection.

A similar picture emerges for the “Rotated Visualization System,” when the tubing system is curved backward and away from the camera (Figure 4b). The aerosol jet ejected from the pipette tip into the downstream straight tubing section is in this case directed toward the front wall of the square glass tube (Fig. 4b.iii). Like for the “Original Visualization System,” e-liquid accumulation starts at the position on the wall toward which the axis of the upstream pipette tip points. At the wall, the aerosol forms bulk e-liquid (Fig. 4b.iii) eventually creating a continuous stream in the direction of flow imposed by the pump (Fig. 4b.iv). Analogous to the previous case, aerosol droplets can be observed on the front wall before the jet is visible (Fig. 4b.ii). E-liquid droplets are again present in the pipette tip, and these droplets flow into the straight tubing section (Fig. 4b.iv). After puff completion, a substantial amount of bulk e-liquid is present on the front wall of the straight tubing section.

A different picture emerges in the “Aligned Visualization System,” when the axes of the pipette tip and the square glass tube are aligned (Figure 4c). There is no direct aerosol impaction on the tube wall due to lack of an impinging jet. Rather, a relatively uniform aerosol cloud develops in the entire tubing section (Fig. 4c.ii and iii). A jet emanating from the pipette tip is barely visible (Fig. 4c.iii), and only for a short period of time. Only small amounts of bulk e-liquid accumulate on the walls of the straight tubing section. Toward the end of the puff, e-liquid flows from the upstream pipette tip into the straight tubing section (Fig. 4c.iv). After the puff is completed, a substantially smaller amount of e-liquid accumulated in the straight tubing section compared to the previous two cases.

Finally, results for a larger tip opening are shown in Figure 4d. No jet can be discerned in this case. Instead, the aerosol leaves the pipette tip in a diffusive manner occupying the entire cross-section of the straight tubing section (Fig. 4d.ii and iii). After completion of the puff, most of the aerosol has left the straight tube, and relatively small aerosol droplets are present on the walls of the entire straight tubing section (Fig. 4d.iv). Overall, less e-liquid was deposited and collected compared to the “Original Visualization System.”

4.2. Visualization of Aerosol Collection from a Pod E-cigarette

Figure 5a shows results obtained in the first collection unit of the “Original Visualization System” resembling the actual collection device used with the “Pod” e-cigarette. In this case, an aerosol jet cannot be observed. During the 1^st puff, droplets form on the upper wall in the straight tubing section. Analogous to the collection in Mod devices, this accumulation starts at the position on the wall directly in front of the pipette tip. Then, these droplets form a continuous e-liquid film (Fig. 5a.3.i) which eventually exceeds 3 mm in length (Fig. 5a.3.ii). Around the film, disconnected droplets can be observed. These droplets are engulfed by the film during the 5^th puff. Droplets also appear in the upstream pipette tip and move towards the straight tubing section (Fig. 5a.3.ii and 5a.5.i). Many puffs later, during the 41^st puff, bulk e-liquid (a drop >1.5 mm high and >6 mm long) has been formed at the bottom of the straight tubing section and flows upstream (Fig. 5a.41.i and ii). From the recorded movies, we determined the distance by which the advancing contact angle of the drop traveled in ~200 frames and estimate that the droplet moved with a speed of ~0.6 mm/s. Moreover, a large droplet (~1 mm long) left the upstream pipette tip but remains attached to the outside surface of the tip (Fig. 5a.41.i).

Figure 5:

Seven selected snapshots for aerosol collection from a “Pod” e-cigarette in (a) the first collection unit and (b) the last unit. Red boxes: arabic numbers indicate puff number, and roman numbers the snapshot number.

Figure 5b shows results for the last aerosol collection unit. Like in the first unit, an aerosol jet cannot be observed. During the 1^st and 2^nd puff, droplets can be observed on the tube walls (Fig. 5b.1.i) that move downstream (Fig. 5b.2.i). During the 3^rd puff, a film becomes visible at the upper section of the unit (Fig. 5b.3.i). This film then fuses with other deposited droplets, signifying the beginning of an e-liquid stream (Fig. 5b.3.ii). Finally during the 5^th puff, an e-liquid stream emerges and travels downstream (Fig 5b.i). Many puffs later, during the 41^st puff, large droplets have emerged that cover the glass wall (Fig. 5b.41.i and ii). A relatively large amount of bulk e-liquid (>6 mm long) sits at the bottom of the straight tubing section and flows upstream. After the puff, the flow ceases. Additionally, an e-liquid stream emanates from the pipette tip and then travels downstream (Fig. 5b.41.ii).

4.3. Aerosol Collection Efficiencies

Figure 6 shows the results of experiments in which we determined the aerosol collection efficiency of the Pod device as a function the number of aerosol collection units used. The mean aerosol collection efficiency (of the triplicate experiments) increased with the number of collection units (n = 1, 2 and 3). However, the three efficiencies did not differ significantly at a significance level of 0.05. We also compared the results with the “Aligned System” and observed that the efficiency is reduced by about half compared to the non-aligned systems, independent of the number of units. The amount of aerosol observed to leave the collection system through the air escape channel (see Figure 1) appeared to be far larger for the aligned system than for the three non-aligned systems.

Figure 6:

Average aerosol collection efficiency for the Pod device depending on number of aerosol collection units used. Error bars indicate standard deviations. The “0” for the # of collection units indicates that the pipette tip was aligned with the beginning of the downstream Tygon section, suggesting no aerosol collection through jet impaction on tube walls.

5. DISCUSSION

Main observation.

Our visualization experiments shed light on the physical processes that are responsible for Olmedo et al.’s simple and efficient method for aerosol collection from e-cigarettes. We found that inertial impaction of aerosol jets, which form at the tip of the pipettes, causes significant deposition and accumulation of the aerosol droplets on the walls of the straight tubing. This impaction process requires the axes of the pipette tip not to be aligned with the straight tubing section. Next we discuss the physics underlying various processes involved in aerosol collection.

Aerosol jet.

Our experiments show the formation of an aerosol jet as the flow emerges from a converging portion of the collection system created by a pipette tip. For the “Original” and “Rotated Visualization Systems,” this jet is directed toward one of the flat walls of the square glass tube where bulk e-liquid is observed to form. Even though aerosol droplets do not necessarily follow fluid flow streamlines,^27,28 they provide a clear indication of the formation of a circular jet aligned with the axis of the pipette tip. The geometry of the pipette tip connected to a straight tubing section is similar to that of a sudden expansion in a circular tube, a flow system that exhibits flow separation and a central jet emerging from the entrance to the sudden expansion, with annular recirculation zones between the core jet and the tube wall downstream of the expansion.^29,30 Several studies have shown that the flow becomes unstable at relatively small Reynolds numbers, leading to vortex shedding and eventually turbulent flow.^29,30 We estimate the Reynolds number in our system based on the tip size, Re = ρ_fv D_h / μ where ρ_f is the fluid density, μ is the dynamic viscosity, D_h is the hydraulic diameter, and v = Q / (π$R_{\mathit{tip}}^2$) is the average flow velocity. In a first approximation, we use the density ρ_f and viscosity μ of aerosol-free air at a temperature of 20°C: μ = 1.825 × 10⁻⁵ kg/(m-s) and ρ_f = 1.204 kg/m³. With the tip radius R_tip = 0.28 mm and the flow rate Q = 1 L/min, we obtain Re = 2 ρ_f Q / (π R_tip μ) = 2,500. This Re value suggests that the jet resulting from the pipette tip and the sudden expansion is probably turbulent.^29–33

We conjecture that a central jet is always present in the experiments with the “original” and “rotated” collection systems but not always observable, likely due to an aerosol density that in some cases is below the level that can be detected by our camera. Moreover, we believe the formation of this central jet is responsible for the accumulation of bulk e-liquid at the walls of the downstream tubing section as we discuss next.

Jet impaction.

The collection of aerosol droplets as the jet impinges upon the tubing wall can be explained by aerosol impaction theory, which has been developed to explain the retention of aerosol particles by so-called impactors. In particular, in inertial impactors the flow carrying the particles is accelerated through a nozzle that creates a jet. The jet then impinges on a collector plate located at a small distance in front of the nozzle. As the air flow is diverted away from the collector plate, large particles deviate from the turning flow due to particle inertia and impact on the collector plate. The effect of inertia compared to viscous effects is characterized by the Stokes number, $St=\frac{{\rho }_{p}vC{D}_p^2/(18\mu )}{R_{noz}}$ where ρ_p is particle density, v is the mean velocity in the nozzle, R_noz is the radius of the nozzle, and C is the Cunningham slip correction factor which has a nonlinear dependence on particle size D_p.³⁴ According to aerosol impactor theory developed for solid aerosol particles,³⁴ a significant fraction of aerosol particles is deposited in a round impactor if the Stokes number exceeds a critical value St ≈ 0.25. We assume that particle impaction due the impingement of the e-cigarette aerosol jet on the tube walls of our collection system is comparable to deposition in a round impactor with nozzle radius R_noz = R_tip. This assumption hinges on the fact that the two systems have similar cross-sectional areas through which the majority of the flow is funneled. Also, assuming that impactor theory developed for solid particles can be applied to liquid e-cigarette aerosol droplets, we can determine, from the critical Stokes number of 0.25, a cutoff particle size, above which particles are retained by the impactor. We solved the optimization problem for cutoff particle size (which is a nonlinear problem as C depends nonlinearly on D_p) using the generalized reduced gradient algorithm,³⁵ obtaining a cutoff size D_p = 490 nm and C = 1.4.

Aerosol collection efficiencies.

For an aerosol collection unit to retain a significant percentage of the aerosol droplets, the size of the incoming aerosol droplets needs to exceed the cutoff size of 490 nm. E-cigarette aerosol is size-distributed,^13,36–38 and the distribution can be described by the mass median diameter (MMD), above or below which 50% of the aerosol mass is contained. The MMD of the generated aerosol depends on several factors, including power output of the e-cigarette and the PG/VG ratio.^39,40 While these dependencies are important, Son et al.³⁹ showed that the MMD of aerosol generated by a Mod device at three different powers outputs, ranging from 6.4 to 31.3 W, was approximately 3 μm, and did not significantly depend on power output. Since our estimated MMD of 3 μm is substantially larger than the cutoff size of 490 nm, a single aerosol collection unit should retain at least 50% of the e-cigarette aerosol mass. This is consistent with our visualization experiments, the about 80% aerosol recovery we measured for “Original Systems” (Figure 6), and previously reported efficiencies⁹ ranging between 70 and 90%.

Surprisingly, we measured a considerable aerosol recovery efficiency close to 80% for collection from the Pod device using only a single aerosol collection unit (n = 1), even though typically three to four units are used for Pods.^9,41 Given that the observed increase in collection efficiency using n = 2, 3 was not significant, it is not clear why more aerosol collection units have been used for Pods in previous studies. However, we offer the following explanation. Pods produce much less aerosol than Mods. For instance, a JUUL device releases 1.4±0.4 mg of aerosol per puff for a 3-s puff time,⁴² whereas the Mod device we used operated at a power setting of 60 W can generate 50 mg per puff. Thus, collecting 0.5 mL of aerosol from a Pod can take one hour or more. Certain users of the collection method may therefore consider that using multiple units is justified as it can reduce, even though only slightly, the duration of the experiments.²¹ However, it is possible that using too many units could decrease the collection efficiency as then more e-liquid might get trapped in the tubing sections.

The method, originally developed with four collection units,²⁰ can be changed by users according to their needs. This may be necessary because different devices and e-liquids can produce aerosols that differ in particle size distribution and number density,³⁹ factors that affect aerosol recovery. Specifically, particle size directly affects aerosol impaction and recovery according to impactor theory, as discussed above. Particle number density indirectly affects collection efficiency, because it promotes droplet coalescence^43–45 which in our collection system is enhanced by the turbulent shear flow generated in the aerosol jet.⁴⁶ Coalescence in turn increases droplet size and thus the recovery efficiency.

Importance of a narrow tip.

Aerosol collection was significantly less efficient for a larger tip opening, likely because no fluid jet was formed that directed aerosol onto the tubing wall where it would deposit and form bulk e-liquid. Aerosol still deposited on the tube walls in the form of separated sessile e-liquid droplets that covered the tube walls rather uniformly. Deposition can be attributed to the turbulent flow field that was generated in the straight tubing section, as evidenced by the spatially non-uniform and ever-changing cloudiness in the straight tubing section. Turbulence can lead to enhanced aerosol deposition as demonstrated for ash deposition during turbulent pipe flow.⁴⁷

Importance of a non-aligned pipette and tubing section.

It is important that the axis of the converging tubing section is not aligned with the axis of the straight tube, even though an aerosol jet is still formed in the aligned case. That jet, however, in the case of the aligned system directs aerosol along the center of the straight tube, not promoting jet impaction onto the wall of the straight tube. We note that the collection system proposed by Olmedo et al. naturally ensures that the axes of converging and straight tubing sections are not aligned. Specifically, the nearly vertically oriented collection vial at the downstream end of the collection system curves the entire tubing train in a polygon-like manner. The visualization experiments are also supported by the experiments measuring the aerosol collection efficiencies in “real” collection systems without the observational square tubing section (Figure 6). In particular, the “Aligned System” (Figures 2b and B) yielded only about one half of the recovery efficiency of the other three systems with non-aligned tubing sections, again suggesting that inertial impaction due to non-alignment is crucial for aerosol recovery. Finally, non-alignment may also help reduce the potential rebounding of impinging aerosol droplets, because droplet impact onto inclined surfaces promotes partial or complete droplet rebounding (when compared to approach to the surface in the normal direction).^48,49

Role of gravity.

The orientation of the angle between the converging and straight tubing section does not affect aerosol deposition substantially. We found that when the pipette tip is pointing toward the bottom wall of the straight tubing section, e-liquid forms on the bottom wall (Figure A in Supplementary Material). To explain these observations, we calculated the sedimentation efficiency in the jet, η_G = (ρ_P — ρ_f) g $D_{\mathit{P}}^2$ / (18 μ v) where g is the gravitational acceleration.^28,50 The sedimentation efficiency is the ratio of the gravitational settling velocity and the average fluid velocity in the tip/jet v. In our application for aerosol droplets with the MMD, this ratio is very small, η_G = 4×10⁻⁶, suggesting that gravity forces do not noticeably affect aerosol transport including the direction of the aerosol jet, which is consistent with our visualization experiments.

However, the recovered aerosol is significantly affected by gravity once it forms large droplets after a sufficient number of puffs, as evidenced by the large droplet sitting at the bottom in the Pod collection device after 41 puffs. To explain this observation, we calculated the Bond number, Bo = (ρ_p — ρ_f) g L² / γ where L is the length of the droplet, and γ is the surface tension. We assumed γ = 63.4 mN/m, the surface tension of pure glycerol at 20°C.⁵¹ For the droplet from the Pod with L = 6 mm, we then obtain a relatively high Bond number, Bo = 6, suggesting that gravity significantly affects the droplet shape and can cause the droplet to accumulate at or flow to the bottom of the tubing sections.

Flow of recovered aerosol.

Aerosol that is recovered in the real collection devices is carried toward the collection vial by the aerosol-laden air flow imposed by the peristaltic pump. This flow can be characterized by the capillary number: Ca = μ ν / γ where we define the velocity v to be the cross-sectional average velocity in the Tygon tubing during a puff. Therefore Ca = μ Q / (γ π $R_{\mathit{Tygon}}^2$) = 0.003. This number is relatively high and indicates that viscous forces dominate interfacial forces during the flow of the recovered aerosol.⁵²

Size distributed aerosol.

Our theoretical analysis of aerosol deposition accounts for the fact that freshly emitted e-cigarette aerosol has a non-uniform droplet size distribution^13,36,37 by basing the analysis on the MMD of the emitted aerosol. This analysis allowed us to explain the high aerosol collection efficiency of the Olmedo et al.²⁰ method. It has, however, been speculated that chemical composition of e-cigarette aerosol droplets depends on their size.¹³ Therefore, the composition of the collected aerosol may not be entirely representative of all aerosol, because smaller droplets are collected less efficiently by inertial impaction. Future work should therefore examine not only the bulk aerosol collection efficiency but also the droplet size dependence of the collection efficiency.

Limitations.

The mismatch in cross-sections between the pipette tips and the observational square cross-section glass tube prevented collected e-liquid from flowing into downstream collection units or the collection vial, a process that is observable by the bare eye in the real collection system. We believe this mismatch is also the reason why we observed upstream flow of bulk e-liquid for the Pod device (Figure 5a) as accumulated e-liquid could not flow downstream. Thus, we could not determine aerosol recoveries for the four visualization systems; however, our experiments measuring aerosol recovery in four systems using only tubing materials of the real aerosol collection setup support our conclusions.

Moreover, it is possible that aerosol droplets behave differently in the observational and real collection systems because of different surface roughness, interfacial tensions, and contact angle. These properties affect the processes that droplets impinging on an initially dry surface (in our case the walls of the straight tubing section) might undergo: deposition, spreading, splashing and rebounding.⁴⁹ These processes are also affected by subsequent wetting of the surface.⁵³ From an aerosol collection perspective, rebounding is the least desirable process. However, from the high measured aerosol collection efficiencies we and others measured using real collections systems, we do not expect droplet rebounding to be significant in real collection systems. It is possible that rebounding was more important in our visualization systems; however, we did not observe any indication of such behavior.

6. CONCLUSIONS

The primary objective of our work was to identify the physical processes responsible for the empirical method developed by Olmedo et al.²⁰ for collecting aerosol from e-cigarettes. Our novel visualization experiments clearly illustrate that collection relies on the formation of an aerosol jet in the converging tubing section created by a pipette tip and the subsequent impaction of this jet onto the walls of the downstream straight tubing section. Such impaction requires the sequence of tubing sections to be curved in a polygon-like manner. A properly designed aerosol collection system, based on inertial impaction, should ensure that the Stokes number St in an aerosol collection unit exceeds a critical value of 0.25 for a significant fraction of the droplets. Turbulent coalescence of the aerosol droplets may also play an important role, allowing for more efficient aerosol deposition due to increased droplet size.

Our experiments with real collection systems suggest that a single collection unit is sufficient to obtain high recovery efficiency from both “Pods” and “Mods” devices. However, using more units may be convenient if other factors such a collection time and specific e-cigarette/e-liquid characteristics are considered.

In future work, we intend to use the physical understanding discussed here to optimize the collection system, e.g., by determining the optimal relative angle between the conical and straight tubing sections and the optimal number of collection units. It also seems possible to construct collection devices using 3D printing^54–57 which has also been used to study droplet impact on solid surfaces,⁵⁸ a process occurring in our collection device. We expect optimal system parameters to depend on a number of variables including aerosol droplet size distribution, number density of aerosol particles, puff time T_puff, puff flow rate Q, and e-liquid viscosity.

Additional work is required to compare our aerosol collection method to other collection devices that also use inertial impaction to collect aerosol for chemical, biological, radiological, and explosives characterization.^11,59–63 However, these devices have often been designed for flow rates much higher than the inhalation flow rate of Q = 1 L/min used here. In addition, as mentioned before, the collected aerosol has typically to be removed from an impacted surface for chemical analysis, while in our method collected aerosol flows into a vial for direct chemical analysis, avoiding possible contamination.

The aerosol collection technique studied here could also be used to recover aerosolized liquid generated by other devices such as ultrasonic nebulizers^64,65 and vibrating orifice aerosol generators.⁶⁵ In fact, the device studied here could be beneficial for the biological, chemical and physical characterization of aerosolized liquids generated by medical metered-dose inhalers and nebulizers.^66–69