# Experimental research on surface acoustic wave microfluidic atomization for drug delivery

### Effect of input Power on aerosol size distribution

Figure 5 shows a high-speed sequence diagram of the continuous atomization behavior of salbutamol solution with a SAW device at a frequency of 30 MHz and different input powers. One end of the paper was connected with a sink, and the other was flat on the substrate surface. The entire atomization process mainly includes three stages: (a) the initial transient state at the beginning of excitation, (b) meniscus formation and deformation, (c) steady atomization process. In the initial stage, due to the siphoning effect at the front end of the paper, the liquid was absorbed from the sink to the substrate surface, and the liquid film began to accumulate. Simultaneously, the thickness of the liquid film (σt) was similar to the thickness of paper, which was much larger than the acoustic wavelength in the liquid (λf ≈50 μm). Once the signal generator was started, acoustic energy coupling into the liquid in the direction of the Rayleigh angle drives the generation of Eckart streaming. In addition to the resulting acoustic streaming within the liquid, the acoustic radiation pressure also imparts a force at the interface, that together with the momentum transfer to the interface due to the Eckart streaming, exerts a body force on the liquid whose vertical component causes the interface into a sharp axisymmetric cone, forming multiple small crests on the free surface, and the threads were elongated and pinched off to produce droplets ejection, as shown in Fig. 5a1–a4. In particular, the jetting was formed in the inertial dominated regime when σt >  > λf . In addition, the duration of this stage has a great relationship with the siphon rate of paper and the input power, and it will be significantly shortened as the liquid film dimensions decreases or the input power increases12. As the jetting continues, the liquid film dimensions decrease. When σt was closed to or less than λf, capillary force dominated the regime compared to the inertial force. At sufficient input power, Schlichting streaming (boundary layer streaming) drives fluid motion within the boundary layer adjacent to the substrate through viscous shear, pulling a thin film out of paper towards the SAW irradiation, as shown in Fig. 5a6. The area pulled out by Schlichting streaming was called the meniscus area. During this process, the capillary wave destabilizes at the meniscus and subsequently breaks up into aerosols. As the atomization progressed, the liquid was continuously consumed, and negative pressure was generated at the meniscus so that the liquid was continuously siphoned from the sink through the paper strip to the surface of the substrate for the atomization process. When the flow rate in the paper matched with the atomization rate, the volume of the free liquid in the meniscus area remained constant, and a steady and continuous aerosol mist was generated on the surface of the meniscus, as shown in Fig. 5a9–a12. We also found that many satellite droplet jets were always mixed during stable atomization at this input power, as shown in the Fig. 5a11.

Figure 6 shows the aerosol size distribution of salbutamol solution atomized under different input powers. It was observed that the total aerosol size measured results have a large distribution span, ranging from a few microns to one or two hundred microns, mainly including three peaks, which were concentrated around 3 μm, 10 μm, and 1000 μm. These peaks depend on the meniscus area formed at the tip of the paper strip. Large droplets of 1000 μm were produced from the satellite droplet ejection. It can be seen that under the action of 4.17 W input power, the most significant volume contribution was from the largest droplets. Interestingly, as the input power increased, the liquid was consumed rapidly due to the intense atomization, resulting in a thin front-running film head of the paper, as shown in Fig. 5b1–b4. This description was utterly consistent with the volume fraction changes of peak 3 in Fig. 6. Specifically, the volume fraction of particle size above 100 μm gradually decreased from 78.25% to 26.31% when the input power was increased from 4.17 W to 6.62 W. From the above results, it can be concluded that there is competition between jetting and atomization in the stage of the stable atomization process, which depends on the input power. Note that at low power, the required acoustic energy was dissipated by the accompanying jetting portion during atomization, resulting in a significantly weakened Eckart streaming and thus a significant drop in capillary wavenumber, which leads to the rise of both capillary wavelength and aerosol size. Therefore, it was essential to suppress these large droplets. The insights from the high-speed flow visualization studies revealed that this was possible by setting the input power above 4.17 W, thus allowing atomization to occur from a relatively thin film, forming a dense, monodisperse aerosol. Note that the first peak of 3 μm was produced from capillary waves due to Schlicting streaming, and the 10 μm order droplets were produced from capillary waves due to Eckart streaming effect, which is a steady vertical flow due to viscous attenuation. Both peaks originate from a thin front-running film of the meniscus, the ratio of these two peaks directly affects the volume fraction of droplets with diameter below 5 μm, which will determine whether a atomizer device is acceptable for asthma inhalation therapy (for this application, at least 30% is usually required). Table 1 shows the data of aerosol droplet size under different input power. From the distributions, three parameters were chosen to characterize the aerosol size distribution, Dv10, Dv50 and D32. The former two parameters represent the 10%, 50% volume percentiles, The parameter D32 indicates that the Sauter mean diameter was the mean particle size related to the dose delivery efficiency when the particles were droplets of unit density12,31. In order to better understand the effect of input power on aerosol size, the total SAW power(PSAW)entering the liquid was calculated, and the related calculation process can refer to our previous study27.

Note that the SAW power rise increased aerosol droplets size, this phenomenon can be explained by ultrasonic atomization theory, that is, increasing PSAW directly caused an increase of energy absorbed by the liquid. This will increase the amplitude of the capillary wave generated on the surface of the liquid film, increasing aerosol droplet size and droplet velocity after atomization. Additionlly, increasing PSAW drives stronger Eckart streaming effect in the film in the SAW propagation direction, leads to an increase in the acoustic body force, therefore producing shorter films, that, in turn, lead to larger droplets. Collins et al. suggested that thin film geometries determine the ejected droplet size, and given by 35:

$$D\sim \frac{{\gamma H^{2} }}{{\mu L^{2} }}\frac{{We^{2/3} }}{f}$$

(4)

where H and L were characteristic height and length scales of film, H/L was close to the paper strip thickness, We = ρL(u0)2/εγ was an acoustic Weber number, where u0 was the vibration velocity of the SAW substrate. Figure 7 shows the experimentally measured average droplet size data under different SAW powers for each peak (i.e., peak 1 and peak 2).

It can be seen that under the SAW power of 5.16 W, the linear average diameter Dv10 was 2.74 μm, and the mean droplet size Dv50 was 8.58 μm, respectively. Similarly, the volume of the liquid film in the meniscus area remained constant. Note that as the PSAW increases, the data points of peak 1 basically maintain a constant trend, which was directly reflected in the diameter changes of Dv10. Therefore, Schlicting streaming has not been enhanced during this change. A simple power function can be obtained to describe the relationship between Peak 2 data points and PSAW, i.e., y = 3.44PSAW 0.656. It was indicated that the peak 2 data increased linearly according to the PSAW 2/3 trend. Since PSAW ~ ρu02, implying that We ~ PSAW. Therefore, the mean droplet size produced by SAW atomization Dv50 ~ PSAW 2/3, consistent with the results in Table 1. Particularly, at higher power, the film length was approaching the limit due to the increase of atomization intensity, and the influence of further increased power on droplet size gradually decreased.

### Effect of device frequency on aerosol size distribution

Figure 8 shows the aerosol size distribution results of salbutamol solution atomized under different device frequencies at 6.62 W input power. To obtain accurate results, each frequency value was tested three times, and the results were averaged. The mean diameters Dv50 for the three analyzed frequencies were 10.66 μm, 5.13 μm, and 3.02 μm, respectively. With the increase of device frequency, the mean diameter of aerosol droplets produced by atomization decreases gradually, consistent with previous studies 39. For a frequency of 30 MHz, the prominent peak was around 10.16 μm, and the secondary peak was observed in measurements of about 2.88 μm. As the input frequency increases to 60 MHz, the initial double peaks merge into a single one located at 4.8 μm, at the same time, a third peak located above 100 μm increases. Interestingly, this change can be explained that by the streaming Reynolds number. The streaming Reynolds number is expressed as follows46:

$${\text{Re}}_{s} = \frac{\rho \nu R}{\mu }$$

(5)

where is the streaming velocity inside the liquid (ca. 10–1–10–2 m/s, measured by using particle image velocimetry (PIV) method, can refer to our previous study 19), R is the characteristic height of liquid film (ca. 10–3 m), μ is the liquid viscosity (salbutamol solution, ca. 10–3 Pa.s), ρ is the density of the liquid. According to the previous literature46, the mechanism that allows the cascade of subharmonics from high excitation frequencies to low capillary wave frequencies was the formation of turbulence in the acoustic streaming induced induced by surface acoustic waves within the fluid bulk. When the value of Res is above the critical value of ~ 102, acoustic streaming is turbulent, which predicts the transition of turbulent flow to subharmonic and period-doubling cascades, leading to result in chaotic flow behavior. For salbutamol solution, ρ = 1.03 × 103 kg/m3μ = 1 × 10–3 Pa.s, In the case of frequency of 30 MHz and input power of 6.62 W, the streaming velocity v was 180 mm/s; Substituting these parameter values into Eq. (5), then the predicted value for the streaming Reynolds number was 185. This also shows that the turbulent acoustic streaming induces bulk flow within the liquid and drives the generation of capillary waves through shear before atomization occurs. At this time, the Eckart streaming effect becomes dominant, resulting the most significant volume contribution of the middle peak. Note that for a 90 MHz device, the streaming velocity was reduced to 50 mm/s, since the higher the frequency, the less SAW energy entering the liquid. In this case, the predicted value for the Reynolds number can be obtained to be 50, the acoustic streaming failed to reach the given threshold of turbulence. Also, as the SAW amplitude decreases, the capillary wave response transforms from a low-frequency broadband cascade to a single peak at fSAW, which directly leads to a gradual increase in the proportion of peak 1. This is consistent with previous studies (Qi et al., 2008, Blamey.J et al., 2013). From the above results, the Schlicting streaming gradually dominates in the high-frequency atomization process. However, the proportion of peak 3 increases slightly due to the ejection of much finer droplets from the film. For a frequency of 60 MHz, with an input power of 2.09 W, atomization was often challenging to initiate even though significant capillary waves and surface vibrations can be observed on the liquid film surface, as shown in Fig. 9a. As the input power increases to 6.62 W, at 1000 ms, the generated mists were more intensive than at low power, as shown in Fig. 9b. The generated mist appeared much finer, compared to those at 30 MHz, when comparing Fig. 9b with those in Fig. 5b. Specifically, for a frequency of 60 MHz, under a power of 6.62 W, the liquid temperature at the paper edge was 85 ℃, while the temperature of the aerosol produced was only 35 ℃. This is essential for the protection of drug properties. For a more detailed experimental description of the thermal effect during liquid atomization process, please refer to our previous study27.

The insights from the high-speed flow visualization studies revealed that the minimum threshold power values required for continuous paper atomization at each frequency were 2.09 W, 2.63 W, and 3.31 W, respectively. For a 6.62 W input power, the atomization rates of water at 30 MHz, 60 MHz, and 90 MHz were 6.35 μl/s, 2.31 μl/s, and 1.33 μl/s respectively. For detailed experimental data of atomization rate, refer to our previous study27. It has to be remarked that, although increasing the device's frequency can effectively reduce the size of aerosol droplets, the rate at which liquid is atomized is also decreased. Thus, there was an operation trade-off. In order to obtain optimal 1–5 μm aerosol droplets size for deep lung penetration and maximum drug delivery, a higher device frequency was favored. On the other hand, it was ideal that SAW devices administer a given volume of drug in the shortest possible time. These two parts were contradictory, and a balance must be found, which will be confirmed by measurements with the lung model later on.

### Effect of the liquid flow rate on aerosol size distribution

For an existing device and a specific liquid to be atomized, none of these parameters can be changed in situ, that is, during device operation. Therefore, for a given setup, liquid flow rate and input power were two parameters that can control the aerosol size in real-time. To analyze the aerosol droplet size distribution variation with the liquid flow rates, measurements were obtained atomizing with constant input power to eliminate the input power dependence. During the atomization process, there is a critical flow rate. When the liquid flow rate marginally above the critical flow rate, the substrate surface can be completely wetted and atomized to avoid parasitic heating of the substrate. Refer to the ultrasonic atomization theory, the critical liquid flow rate Qcrit was calculated as follows40

$$We = \frac{{fQ_{crit} \rho }}{\gamma }$$

(6)

where We is the acoustic Weber numbers, f is SAW frequency, γ is the surface tension, ρ is the liquid density. Note that the critical flow Qcrit is independent of the SAW power and is related to the threshold at which atomization begins to occur. The atomization process will discontinue when the liquid flow rate is less than the critical flow rate Qcrit. The increase in the temperature of the substrate causes the liquid to boil, which changes the local refractivity of the particle size analyzer, so that accurate detection results cannot be obtained. This part will be explained in detail later. According to previous studies 35, atomization begins when the Weber number approaches a critical value of 1. That is, there is sufficient inertial stress to overcome the surface tension of the liquid. It was proposed that by equating the inertial forcing to stabilizing capillary stresses i.e. by taking We equal to unit, that is

$$Q_{crit} = \frac{\gamma }{f\rho }$$

(7)

This is the minimum critical flow rate above which the film's sufficient thickness is available to cover the working areas all time. Thus, the energy was utilized to the maximum. For a frequency of 30 MHz, substituting the parameter values of salbutamol solution into Eq. (7), the predicted value for the critical liquid flow rate Qcrit was 43 μl/min. The upper bound of the maximum flow rate was closely related to the atomization rate. When the maximum flow rate was more significant than the atomization rate, the liquid film began to accumulate and no longer meet the film condition, i.e., the conditions for atomization. For a 6.62 W input power in a 30 MHz device, the atomization rate was about 381 μl/min. Figure 10 shows the aerosol size distribution for different liquid flow rates. Note that, when the flow rate was increased from 70 to 400 μl/min, the Sauter mean diameters D32 appear to decrease first and then increase. It can be seen that at the flow rate of 70 μl/min and 90 μl/min, the volume contribution of peak 3 was 74.59% and 43.37%, respectively. Associated with an increase in the flow rate, the volume contribution of this part gradually decreased. It has to be remarked that if the flow rate was too low for sufficient thickness of a film to be maintained on the substrate surface, intermittently, the working surface gets exposed to air. Thus, a low flow rate will render the insufficient supply of fluid and the discontinuous atomization state. This was also concluded in earlier studies of Wink et al.43,44. Under the action of 6.62 W input power, at this point, it must be noted that the temperature of the substrate exceeds 150 °C. At this point, the liquid produced large droplets in the form of rapid boiling. The significant volume contribution of peak 3 origins came from this, which causes a change in the local refractivity in the aerosol.

When the flow rate reached 110 μl/min, the atomization was stable, and the aerosol size distribution presented a multi-peak mode. At this time, the volume contribution with aerosol droplet size greater than 100 μm accounts for 9.94%. When the flow rate reached 400 μl/min, the liquid accumulated on the SAW propagation path, and then the atomization stopped. Similarly, note that the dominant inertia regime and the ejection occurred. Figure 11a shows the effect of flow rates and input power on aerosol droplet size. According to experimental measurements, we found that the Sauter mean diameters can be effectively changed by adjusting the input power and liquid flow rate to change the film conditions. For a 5.26 W input power with a 90 μl/min flow rate, the Sauter mean diameter D32 was 8.5 μm. Interestingly, note that the combination of input power and flow rates can define three different atomization regimes, i.e., discontinuous atomization zone, stable atomization zone, and liquid film accumulation area, as shown in Fig. 11b. When the liquid flow rate exceed the atomization rate, results in fluid accumulation, accompanied by capillary wave disturbance and and no atomization occurs. Low flow rates will make the atomization process discontinuous as the atomization rate far exceeds the rate of fluid supplied to the substrate. Significant heating was observed in the case of a discontinuous atomization regime, especially using high SAW power. Since heat generation was representative of underutilized mechanical energy, it will cause local temperature rise, device damage and affect the SAW device's overall efficiency47,48. Therefore, the best aerosol droplet generation can be obtained when atomization behavior is located in the stable continuous atomization zone.

### Dose measurements and comparison

Figure 12a shows the UV absorption curve of the standard salbutamol solution. Note that the salbutamol solution mainly absorbs UV light 287 nm. As the concentration increases, the absorption peak gradually increases, and the UV absorption of salbutamol has a linear fitting relationship with the concentration, as shown in Fig. 12b. We atomized the salbutamol solution and collected aerosols in each stage of the glass impinger. By comparing the measured UV absorption against the calibration curve, we can get the salbutamol concentration in each area in the glass impinger.

In inhalation therapy, the aerodynamic behavior of the droplets (controlled by Stokes' law) was crucial as an important indicator to measure the quality of inhaled drugs. The data recorded represent the physical diameters of the droplets can be converted into equivalent aerodynamic diameters using the following relationship12

$$D_{a} = \left( {\frac{\rho }{{\rho_{0} }}} \right)^{1/2} D$$

(7)

where Da and D are the aerodynamic and the physically measured diameters, ρ is the density of the solution, and ρ0 ≡1 g/cm3. In order to evaluate the effect of SAW atomization technology on the deposition of salbutamol in each area of the glass impinger, an experimental study was carried out based on the current physical aerosol particle size research results. It was worth noting that the proportion of drug particles entering the lung area during the experiment was called the acceptable particle dose (FPD). This parameter will determine whether an atomizer is suitable for asthma inhalation therapy and delivery efficiency. Table 2 shows the deposition of salbutamol atomized inhalation in each area with a frequency of 30 MHz, an input power of 5.26 W, and a flow rate of 90 μl/min. Each set of experiments was repeated three times.

It can be seen from Table 2 that the deposition rates of oral cavity A, throat B, and lung C were 53%, 9%, and 38%, respectively. Experimental results also revealed that the emitted dose generated by SAW atomization accounts for 70% of the theoretical dose. The remaining (< 30%) drug loss was mainly due to thermal effect precipitation on the device substrate, which is in complete agreement with the literature results12. Increasing the input power can increase the atomization rate, but it will raise the aerodynamic size of the drug, enhance the thermal effect, and then reduce the efficiency of the SAW drug delivery device. Figure 13 shows the deposition rate of the lung area at 5.26 W input power with different device frequency versus different flow rates. Note that under the action of a frequency of 30 MHz, the deposition rate of the lung area gradually decreases with the increase of flow rate. When the flow rate reaches 270 μl/min, it exceeds the atomization rate of 266 μl/min under 5.26 W input power27. At this time, the deposition rate of salbutamol particles in the lung area was only 8.1%. This was consistent with the accumulated volume fraction of the aerosol size less than 6.4 μm measured by the laser diffractometer at the same input power, as shown in Fig. 8, which verifies the accuracy of the experimental scheme. It should be noted that the flow rate of 90 μl/min was already the minimum flow rate under the action of this power. If the supplied rate continues to decrease, it will lead to the discontinuous state of atomization. Simultaneously, reducing the flow rate will increase the atomization time, which will cause various discomfort for patients, so it was challenging to meet the requirements of drug atomization inhalation. At a frequency of 60 MHz and flow rate of 90 μl/min, the deposition rate reaches 57%. In this case, the thermal effect of the device was reduced, the corresponding SAW administration efficiency was improved, and more than 80% of the drug could be delivered into the respiratory system. If the device frequency was increased to 90 MHz, the atomization rate was only about 60 μl/min at the same power, which was challenging to meet the requirements of atomization inhalation. However, increasing the atomization rate requires increasing the input power, which was accompanied by a series of problems such as further heating the device and the increase of aerosol drug particle size, so this part of the experiment did not continue. Therefore, we can conclude that the device should be operated at a frequency of 60 MHz, an input power of 5.26 W, and a flow rate of 90 μl/min, such that the dose ratios across the various criteria were maximized.

As SAW excites the liquid and destabilizes the liquid–air interfacial boundary equilibrium, the liquid spreads into a thin film as a precursor to the start of atomization. A thin film liquid is more accessible to atomize than a drop-shaped liquid because less viscous energy will be lost, and more energy can reach the surface of the liquid to produce capillary waves. In previous research26,35, the mist size strongly depends on the thickness and volume of the liquid on the SAW device. However, due to the limitations of the paper itself, it was unrealistic to use a smaller thickness of paper for the flow. Therefore, the fluid supply via on-chip polymeric microchannels prepared by photolithography directly on the chip surface was a good choice. These related processes have been introduced in Sec.IIC above. Figure 14 shows the high-speed camera captures images of SAW atomized salbutamol solution based on SU-8 microchannel fluid supply at supply at a frequency of 60 MHz, an input power of 5.26 W and flow rate of 90 μl/min. We can identify the dynamically stable fluid film at the exit of the microchannel at the boundary of the acoustic beam. Note that the salbutamol solution was rapidly actuated and atomized under the excitation of SAW, and we found that the ejection of large satellite droplets was not observed within 10 s. Figure 15 shows the lung deposition rates obtained by atomizing salbutamol solution using a Omron C25s medical atomizer and a SAW atomizer. Note that the SAW atomized salbutamol solution based on SU-8 microchannel fluid supply with a flow rate of 90 μl/min and input power of 5.26 W. It can be seen that in the inhalation experiment of the Omron C25s medical atomizer, the deposition rate in the lung area was 46%, which was much higher than the deposition efficiency of traditional DPI (about 19%) and MDI (about 29%)49. Note that compared with Omron medical atomizer, the deposition rate of using a SAW atomizer with a frequency of 60 MHz was significantly improved (P < 0.01), the deposition rate can reach 75%. Simultaneously, there was no statistically significant difference in lung deposition rate between Omron medical atomizer and SAW atomizer with a frequency of 30 MHz (P > 0.05). From the standpoint of dosage alone, the results show that SAW atomization technology based on microchannel fluid supply was a strong competitor of drug delivery technology on the market. Additionally, the unique advantages of SAW devices, such as small size, low cost, low power consumption, and no noise, coupled with accurately designed and dedicated electronics as well as signal solutions, solve the miniaturization issue of radiofrequency signal sources, which makes SAW atomization technology can become an alternative to these conventional atomizers. Moreover, the drive circuit for SAW devices has been developed, and only three pieces of 3.7 V lithium batteries are needed to supply power, which shows the potential of this device for portable applications. For portable circuit-driven atomization details, please refer to the supplemental material video.