Atomization of High-Viscosity Fluids for Aromatherapy Using Micro-heaters for Heterogeneous Bubble Nucleation

The development of a novel lead-free microelectromechanical-system (MEMS)-based atomizer using the principle of thermal bubble actuation is presented. It is a low-cost, lead-free design that is environmentally friendly and harmless to humans. It has been tested to be applicable over a wide range of fluid viscosities, ranging from 1 cP (e.g., water) to 200 cP (e.g., oil-like fluid) at room temperature, a range that is difficult to achieve using ordinary atomizers. The results demonstrate that the average power consumption of the atomizer is approximately 1 W with an atomization rate of 0.1 to 0.3 mg of deionized (DI) water per cycle. The relationships between the micro-heater track width and the track gap, the size of the micro-cavities and the nucleation energy were studied to obtain an optimal atomizer design. The particle image velocimetry (PIV) results indicate that the diameter of the ejected droplets ranges from 30 to 90 μm with a speed of 20 to 340 mm/s. In addition, different modes of spraying are reported for the first time. It is envisioned that the successful development of this MEMS-based atomizing technology will revolutionize the existing market for atomizers and could also benefit different industries, particularly in applications involving viscous fluids.


Results
Atomizer Design and Its Working Principle. The lead-free design of the atomizer is shown in Fig. 1(a,b).
It consists of three key components: a nozzle plate with hundreds of micro-sized orifices, a spacer, and a microheater plate with a line heater. The micro-fabricated nozzle plate, as shown in Fig. 1(c,d), is fabricated using a nickel-cobalt (Ni-Co) alloy. The orifice shown is a cone-shaped structure whose diameter is approximately 20 μ m on the bottom side and 95 μ m on the top side. Along with 20 μ m orifices, 15,25,40, 100 and 150 μ m orifices were fabricated for use in the experiments. The center of the nozzle plate is intentionally slightly curved inward to increase the rate at which the working fluid can be refilled from the refilling chambers to the firing chamber by capillary action. Additionally, a spacer and a micro-heater plate are used to form a firing chamber and to nucleate bubbles, respectively. The spacer consists of polydimethylsiloxane (PDMS) of various thicknesses, ranging from 50 μ m to 500 μ m. (Unless otherwise specified, a thickness of 250 μ m was used in experiments.) The micro-heater substrate consists of ROGERS 4003 material and has a thickness of 250 μ m. The micro-heater consists of a 35-μ mthick layer of copper and a 0.5-μ m-thick layer of immersion-deposited gold. As shown in Fig. 1(e), many micro-cavities are fabricated on the surface of the micro-heater, with diameters ranging from 0.2 to 5 μ m. These cavities serve as nucleation sites for bubble generation. As noted in the discussion section, the minimum heater temperature required for bubble nucleation is between 107 and 285 °C.
The atomizer, as shown in Fig. 1(b), can be divided into 2 chambers: firing and refilling chambers. A schematic diagram illustrating the working principle of a thermal bubble actuator is shown in Fig. 2. Both chambers are initially filled with a working fluid. During operation, the fluid near the surface of the micro-heater is superheated by applying a short current pulse. Heat is spontaneously transferred by convection to the fluid and increases the temperature of the fluid, leading to bubble growth. This instant growth of bubbles is also known as explosive boiling 33,34 . The bubbles act as an actuator to push the fluid out of the orifices. We call this process a "firing event". In this manner, a capillary force is initiated by the negative pressure developed between the two chambers. Thus, the firing chamber is refilled with fluid from the refilling chambers and is subsequently ready for the next firing event. Similarly, the empty refilling chamber is refilled through a giant liquid reservoir.

Heat Transfer Model of the Micro-heater.
A lumped heat transfer model is developed to better understand the thermal response of the micro-heater. In general, heat lost through conduction and convection is considered in analysing the heat transfer mechanisms of the structure. In our case, the input energy for onset nucleation is lost through convection to the surrounding fluid medium and conduction to the heater plate, as shown in Fig. 3.
By applying an energy balance equation for heat generation and transfer on the micro-heater based on Fig. 3, the heat transfer can be modeled as follows 35,36 : where E is the energy consumed by the heater, ρ is the density of the heater, c is the specific heat capacity of the heater, w is the width of the filament, l is the length of the heater, b h and b p are the thicknesses of the heater and heater plate, respectively, h fluid is the convection heat transfer coefficient of the fluid used, k p is the thermal conductivity of the plate, t is the pulse width, T nl is the nucleation temperature which can be calculated using Eq. (2), T nozzle is the temperature of the nozzle and T ∞ is the ambient temperature. The ROGERS 4003 plate has a thermal conductivity of 0.71 Wm −1 K −1 and a thickness of 250 μ m. Equation (1) is derived based on the following assumptions. First, the heat transfer through the side of the micro-heater is considered due to the length of the filament. Second, heat radiation is neglected. Finally, the micro-heater has a uniform temperature distribution along the heater, and the thermal coefficient of resistance is negligible. The energy consumed by the heater is expected to increase with increases in the area of the heater. Bubble Nucleation on the Micro-heater. As mentioned in the earlier section, bubbles are generated through heterogeneous nucleation of micro-cavities on our micro-heaters. Equation (2) below, which is given by Griffith and Wallis 37 , shows the relationship between heater temperature and radius of curvature, r, of the meniscus in a cavity.
where T s is the saturation temperature, T w is the wall temperature, σ is the surface tension, L is the latent heat of vaporization (2,258 kJ/kg for water), and v g and v l are the specific volumes for air (0.77 m 3 /kg) and liquid (0.001 m 3 /kg), respectively. It is derived through Young-Laplace and Clausius-Clapeyron relation by assuming the contact angle between the liquid and solid is 90° 37 . According to the conditions for heterogeneous nucleation 28-32 , a cavity must have trapped gas to be an active nucleation site. If the radius of curvature of the meniscus in the cavity is equal or greater than the critical nucleation radius r * , a bubble can be formed. This critical nucleation radius r * , also known as the minimum cavity mouth radius for nucleation to occur, is obtained by Eq. (2) when a wall temperature is given. Similarly, a critical nucleation temperature ⁎ T w is obtained when the radius of curvature is given. In our design, assuming the radius of curvature of meniscus is the same as the radius of cavity, the calculated critical nucleation temperature ⁎ T w ranges from 107 to 285 °C. The minimum energy required for onset bubble nucleation can then be estimated by combining Eqs (1) and (2). As shown in Fig. 4(a), the shaded region is the minimum energy required for bubble generation for micro-cavities with diameters between 0.2 μ m (upper dotted line) and 5 μ m (lower dotted line).

Nucleation Energy of Different Heater Designs.
A matrix of micro-heaters was designed to study their bubble nucleation energy consumption. The heaters were designed to have a similar resistance of 0.2 Ω but with different combinations of track/filament width and length, as shown in Table 1. There is a 20-30% reduction in the actual track width due to variations in process tolerance. However, the difference did not affect the comparison of the micro-heater's performance.  In this experiment, the electrical energy consumed by bubble onset nucleation was calculated based on the duty cycle selected for the micro-heater; the voltage supplied was fixed. Therefore, the energy absorbed can be calculated as follows: where V is the voltage supplied to the heater, t is the pulse width and R h is the resistance of the micro-heater. The resistance of the heaters was measured at room temperature; it is expected to vary with temperature. The fluid used was deionized (DI) water. Both the period and voltage supplied were fixed, whereas the pulse width was slowly increased to observe the nucleation of bubble under motion analysis microscopy (VW-6000). To ensure better observation of the heater surface, no nozzle plate was used in this experiment. The number of droplets ejected per cycle by the heaters in Table 1 was estimated to identify the optimal design. In the experiment, a power of 0.8 W was supplied to the heaters described above with a period of 1 s and a pulse width of 100 ms. The nozzle plate was a 7.5-mm-diameter region of 40-μ m-diameter orifices. Each measurement was collected for 120 s. Ten measurements were taken for each heater design. Using Eqs (6) and (7), the estimated  number of droplets ejected per cycle of each heater is plotted in Fig. 4(b). The method of applying Eqs (6) and (7) is discussed in the "number of droplets ejected per cycle" part. Figure 4(b) shows that the heater with the smallest track width and area has the largest number of droplets ejected under the same power configuration. Therefore, the optimum design of the heater is deduced to be a large or sufficient area formed by a mesh of narrow heater track, as shown in Fig. 4(c). Additionally, Fig. 4(c) shows the images of the onset of bubble nucleation under pulse heating recorded by a motion analysis microscope (VW6000, Keyence). The bubble diameter ranges from 100 to 250 μ m. The bubbles shrank and remained on the surface of the heater after nucleation and were capable of immediate growth during the next pulse.
Note: For all the results below, they are based on heater design A with the resistance reduced to 0.11 Ω .

V-I Characteristics of the Atomizer. A series of experiments was conducted to determine the relation-
ship between the instantaneous power consumption and voltage supplied as well as the surface temperature of the nozzle plate. The medium used was DI water. We used a fixed heater design (heater design A) and duty cycle (20%) and varied the voltage supply to obtain the surface temperature of the nozzle plate and the current, I. The instantaneous power consumed, P i , was calculated as follows: The period was 1 s, and the pulse width used was 200 ms; the micro-heater resistance was 0.11 Ω. The surface temperature of the nozzle plate was monitored with an infrared camera (FLIR SC660). Throughout the experiment, the surface temperature of the nozzle plate ranged from 37 to 64 °C. The linear I-V characteristics of the heater are shown in Fig. 5. The temperature increases with increases in the current and voltage. Furthermore, the amount of atomization increases with increasing applied voltage. However, a voltage of 2.28 V and above results in the burning of the heater. The main cause is the overheating of the heater surface, as a phase change occurs (i.e., bubble nucleation) when the fluid reaches the metastable state. In this case, there is little to no fluid acting as a coolant covering the surface of the heater, resulting in poor heat transfer. Thus, when the voltage applied increases to a certain limit, the pulse width must be reduced to prevent excessive heating of the heater. In addition, experiments indicated that a duty cycle of 50% and above leads to the burning of micro-heater. Thus, a balance in pulse width and voltage supply must be achieved, as they are correlated.
Speed, Envelope and Size of Droplets. Different sets of experiments were conducted using particle image velocimetry (PIV) to investigate the effect of power and orifice size on the droplets ejected. In this section, a nozzle plate in which the area with orifices has a diameter of 7.5 mm was used, and the additional conditions were as follows: DI water, a period of 1 s, and a pulse width of 200 ms. Figure 6(a) shows the effect of orifice size on the speed and envelope of the droplets, whereas Fig. 6(b) shows the effect of power on the speed and envelope of the droplets. The color scale in Fig. 6(a) indicates a range of speed from 20 mm/s (dark blue) to 340 mm/s (red). Figure 6(a) illustrates that the number of droplets having a high speed (i.e., the total area of the red region) increases with increases in the diameter of the orifices. This phenomenon can be explained by considering the energy dissipated when liquids go through an orifice. When a unit volume of fluid is ejected from the firing chamber through the orifices, the pressure head of the liquid is converted into kinetic energy and surface energy of the ejected droplets as follows 38 : where P is the pressure head of the liquid, ρ T is the density of the fluid at certain temperature, v is the speed of the droplets, σ is the liquid surface tension, D is the diameter of the orifice, μ is the fluid viscosity, l is the thickness of the orifice and f is the minor loss. By analysing Eq. (5), it can be shown that the surface energy due to the surface tension and viscosity is larger when the diameter of the orifices is smaller. Therefore, the speed of the droplets decreases, assuming that the minor loss is negligible and the other variables are kept constant. The decrease in the speed further affects the envelope of atomization. This finding is supported by Fig. 6(a), which shows that the envelope of droplets increases (i.e., the colored contour in the middle) with increases in the diameter of the orifices. Therefore, by modifying the size of orifices, the speed and envelope of droplets can be controlled. Figure 6(b) shows the variation of speed with different power supplied to the micro-heater. The orifice diameter used was 40 μ m. The color scale of Fig. 6(b) ranges from 10 mm/s (dark blue) to 210 mm/s (red). The area of droplets with high speed and the envelope of atomization both increase with increases in the power supplied. Based on the bubble nucleation criteria given by Griffith and Wallis 37 , the solid-liquid interface temperature is expected to increase with increasing power supplied, leading to increased activation of surface cavities and an increase in the bubble nucleation rates. Therefore, more bubbles act as actuators to pump the fluid.
Additionally, the size of the ejected droplets was measured. The mean droplet diameter is 60 μ m, with the majority of droplet diameters ranging from 30 to 90 μ m in all of the aforementioned cases.  Previous results demonstrate that for each current pulse, hundreds to thousands of water droplets with estimated sizes ranging from 30 to 90 μ m were ejected. The number of droplets ejected per cycle can be calculated by considering the weight change of the atomizer and the weight of a single droplet. The weight of a single droplet can be approximated as follows: where m droplet is the mass of one droplet, ρ T is the density of the liquid at a certain temperature, and d is the diameter of the droplet. Because the experiments were performed at room temperature, the density of DI water was taken as 998.20 kg/m 3 , whereas the average droplet diameter was taken to be 60 μ m based on the PIV results. This droplet size is also supported by determining the average terminal velocity using a technique known as particle tracking. The terminal velocity of targeted droplets could be calculated by labelling a fixed travelling distance in the different images. The approximate average terminal velocity of the droplets obtained was 100 mm/s, which indicated an average diameter of 60 μ m for the droplets according to Eric R. Lee 39 . Furthermore, the number of droplets ejected per cycle can be calculated as follows: droplet atomizer droplet where n droplet is the number of droplets ejected per cycle and m atomizer is the change in mass of the atomizer per cycle. The period used was 1 s, and the fluid was DI water. The average powers supplied for both atomizers with different nozzle plates were maintained to be within ± 0.005 W of each other for each pulse width. Ten measurements were taken for each data point. The number of droplets ejected are calculated by using Eqs (6) and (7) and plotted in Fig. 7(a). The rate of atomization ranged from 0.1 to 0.3 mg per cycle for the micro-heater designed to have a 2 mm 2 area. The results show that an increase in pulse width increases the number of droplets ejected due to additional bubbles pumping the fluids with increasing interface temperature. Furthermore, a larger orifice area results in more droplets ejected per cycle for pulse widths lower than 350 ms. At a pulse width of 400 ms, the number of droplets ejected by both atomizers reaches a peak of approximately 2,400 droplets as shown in Fig. 7(a), which is equivalent to approximately 0.3 mg of DI water. This maximum amount of ejected droplets is due to the limited volume of the firing chamber, which is approximately 0.37 mm 3 , capable of storing up to approximately 0.37 mg of DI water. The firing chamber volume is estimated by considering the cross-sectional model in Fig. 7(b) and multiplying the area of the trapezium (0.12 mm 2 ) by the circumference of the micro-heater (3.1 mm, assuming that the area of the heater is circular). Thus, the number of droplets ejected per cycle can be adjusted by calibrating the pulse width, and the firing chamber volume can be increased to maximize the amount of droplets ejected.  Fig. 8(a). As shown, the surface tension and viscosity of the fluids decrease with increasing temperature. The instantaneous sharp increase of the temperature in the firing chamber leads to an extremely low viscosity. Therefore, less energy is viscously dissipated and the fluids can be pumped out by the bubbles generated 42 . In this experiment, the atomizer was assembled with a nozzle plate with a 7.5-mm-diameter area of 40 μ m orifices. The heater "ON" time was 200 ms with a period of 4 s and a power of 1.05 W for both liquids. The images were captured with the VW6000 motion analysis microscope. The sequence of images of a firing event is shown in Fig. 8(b,c). We note here that, based on Eq. (5), the velocity of the droplets theoretically should be higher when the viscosity of the fluid is smaller. However, as shown in Fig. 8(a), the viscosity of both fluids decreases to a similar value when the temperature increases. Thus, there is no observable difference in the resulting spray.

Discussion
For heterogeneous nucleation, it is found that the nucleation energy highly depends on cavity size. Based on Eq.
(2), a larger radius of curvature of trapped gas in the cavity indicates that a smaller nucleation temperature is needed and thus less energy is required. In order to accommodate a large trapped gas, the size of the cavity must be equal or larger than the trapped gas. Therefore, it can be deduced that a larger cavity tends to have a lower nucleation temperature assuming that the radius of curvature of trapped gas is the same as the radius of cavity. As indicated by the results of S. Witharana et al. 43 , the bubble nucleation temperature decreases when the diameter of the cavity increases. The results of Griffith and Wallis 37 also showed that a smaller cavity requires a higher nucleation temperature as well. Qi and Klausner 44 have also observed that the wall superheat is larger when size of cavity is smaller. However, the gas entrapment mechanism by Bankoff 30 suggested that a cavity is filled with liquid if its cone angle is larger than the contact angle of the liquid on the surface. Hence, relatively large cavities would be filled with liquid and could not be active nucleation sites 28 . The measured energy curve obtained by Eq. (3) in Fig. 4(a) is not a linear curve; it would be a linear straight line if Eq. (1) were used with the only variable being the surface area (i.e., w times l). With the aid of Fig. 9, the larger track width corresponds to a higher possibility of having more large cavities on average. Thus, the energy required by heater design A with a track width of 80 μ m is closer to the estimation energy line of 0.2-μ m-diameter cavities, whereas heater design C with a 140 μ m track width is closer to the estimation energy line of 5-μ m-diameter cavities. Moreover, Fig. 4(a) illustrates that when the heater area is small, the size of the cavity has a lesser effect on the energy required compared to large heater area (i.e., 0.2 J difference between 0.4 and 5 μ m cavities for a 2 mm 2 heater vs. an approximately 0.6 J difference for a 6.3 mm 2 heater). In our case, because the cavities are randomly distributed across the heater, the effect is unpredictable and is expected to be greatly reduced. Also, Fig. 4(a) shows that the energy required for bubble nucleation is lower when the heating area is smaller. Therefore, to reduce the unpredictable effect of cavities and to achieve a high energy efficient design, a narrow heater track is preferable for us because a narrow track results in a smaller heater area among the same resistance heaters. Apart from having a narrow micro-heater, there should be a large enough heating area for bubble nucleation in order to achieve a desired amount of atomization. Thus, an optimum design would be a micro-heater with a narrow width but long length. The characteristics of the resulting spray (i.e., size and velocity) were studied. As reported, there are two main factors that affect the resulting spray: the initial disturbances at the liquid-gas interface and a mechanism that leads to the growth of the disturbances, leading to the breakup of the liquid flow 2 . However, in our case, there is another factor that influences the characteristics of the spray, namely, the ratio of the thickness of the firing chamber b s (i.e., the spacer thickness) to the size of the bubbles. This factor, which leads to different spraying modes, as shown in Fig. 10, can be classified into 3 modes: pure droplet (i.e., b s > 250 μ m), mixed (i.e., 50 < b s < 250 μ m) and vapor (i.e., b s < 50 μ m) modes. A clear division cannot yet be defined for the thickness boundaries of different modes due to the finite sampling size. According to our measurements, as shown in Fig. 4(c), the bubble size ranges from 100 to 250 μ m. For 50-μ m-thick spacers, the sprays are faint but noticeable with vapor trails and negligible amount of droplets (vapor mode). The formed bubbles are constrained and prone to bursting to eject the hot steam content, which would appear as vapor trails upon cooling. For 150-and 200-μ m-thick spacers, the sprays contain a mixture of both vapor trails and droplets. For 250-and 350-μ m-thick spacers, the sprays are mainly droplets, and for 400-μ m-thick spacers, no noticeable spray can be observed. When the firing chamber thickness becomes significantly large compared to the bubble size, the volumetric change (i.e., force) created by the bubbles may not be sufficiently large to produce sprays. In addition, the sprays for 250-μ m-thick spacers are stable with a negligible amount of vapor trails, whereas for 350-μ m-thick spacers, the heaters are easily overloaded and burnt when providing the same noticeable sprays. Similarly, for spacer thicknesses of 400 μ m and above, the heaters are burnt after a few cycles of pulses, possibly due to the defective cooling effect over a region with large unburst bubbles. To gain a better understanding of the impact of bubble size and the thickness of the firing chamber, a bubble size of 250 μ m is considered based on the trend shown above. When the ratio of the thickness of the firing chamber to the size of the bubble is slightly less 1, vapor trails are observed, whereas the stability of the heater is low when the ratio is considerably larger than 1. These observations are summarized in Table 2. Vapor mode spraying should be minimized or avoided, as it is uncontrollable with the current configuration and operational parameters. Therefore, a thickness of 250 μ m and above is preferred (a ratio of unity and above).
Besides, there are some problems that should to be addressed before the device discussed in this paper could be commercialised. First, there may be overheating problem of the micro-heater that will lead to failure of the device. For example, a duty cycle of more than 50% will lead to excessive burning of the heater because there is insufficient time for liquid refilling and micro-heater cooling. Therefore, it is critical to ensure that the firing chamber is completely refilled in time and/or adopt a protective circuit that prevents overheating. Moreover, according to the previous investigation of inkjet devices, it was found that thin film micro-heaters suffers from mechanical damage due to cavitation 45 . This is because extreme high pressure is exerted on the heater's surface, during nucleation and collapse of bubbles, and would lead to cavitation. Although heterogeneous nucleation occurs in our atomizer, which is less damaging than homogeneous nucleation of the inkjet devices, the accumulation effect of the damages might still cause the breakdown of the micro-heater. The conventional method of using a protection layer such as Tantalum (Ta) cannot be applied in our case as Ta might cause irritation to human subjects. Therefore, a possible solution is to use thicker and/or hard materials to serve as the micro-heaters. In addition, it is reported that thin film micro-heaters may also suffer from the accumulation of fluid residues (kogation) which degrades their performance. Similar problems might occur in our heaters if the chemical composition of the target fluid does produce residues after being heated. Other than adjusting the fluid formulation, it is suggested that residues can be removed by rapid and low heating boiling 46 . This may be a cleaning mechanism that has to be considered by our team in the future.

Conclusion
This study presents a lead-free atomizer that was designed based on thermal bubble nucleation under pulse heating. It consists of a nozzle plate, a heater plate with micro-heater lines and a spacer. Different experiments were carried out to study the optimized design, power consumption, atomization rate, size, speed and envelope of the droplets ejected. Based on the current design, the device has an average power consumption of approximately 1 W with an atomization rate of 0.1-0.3 mg DI water per cycle, whereas the surface temperature of the nozzle plate ranges from 35 to 65 °C. The speed, size and envelope of droplets ejected were obtained through PIV inspection. The speed of the droplets was determined to be 20-340 mm/s with diameters of 30-90 μ m. Additionally, the mode of spraying was determined by the ratio of the thickness of the firing chamber to the bubble size. Finally, our atomizer was shown to operate with a wide range of fluid viscosities, up to 200 cP, at room temperature. The development of this novel lead-free MEMS-based atomizer will benefit the healthcare and well-being industries.

Method
The pulse width-modulated (PWM) driving circuit consists of a pulse generator, an n-channel MOSFET transistor (IRFZ44N), a high-current power source and an oscilloscope. The driving circuit is shown in Fig. 11(a). The principle of operation of atomization is based on the superheating effect. For every cycle, a short-duration current pulse (one to a few hundred milliseconds) with a peak current of 1-5A is applied to the micro-heater to superheat the fluid. Thermal bubbles are then generated at defects, micro-cavities and/or corners of the micro-heater. Consequently, the thermal bubbles pump the firing chamber's fluid out of the micro-orifices. The period of the driving signal can be adjusted such that the duty cycle lies between 10-50%. However, any duty cycles higher than 50% induce boiling of the fluid, which increases the average temperature of the chamber.
In addition, a standard PIV (2D2C) technique was used to measure the speed and size of the droplets ejected in a plane. In the experiment, standard PIV was employed to measure the speed of droplets in a particular vertical plane using a high-speed camera (Phantom V641, 4-megapixel sensors and 2,560 × 1,600 pixels resolution) together with a Nikon Nikkor 90 mm f/1.2 lens and a dual-beam laser (Litron LDY304-PIV, Nd:YLF), which is non-intrusive for the experiment. The laser was equipped with cylindrical lenses to form a laser sheet perpendicular to the high-speed camera, and the trigger rate was fixed at 400 Hz in single frame mode. A Dantec PIV system was used to acquire and process the images.