Fabrication of piezoelectric thickness mode transducers via direct current sputtering of plate electrodes on lithium niobate is described. Additionally, reliable operation is achieved with a transducer holder and fluid supply system and characterization is demonstrated via impedance analysis, laser doppler vibrometry, high-speed imaging, and droplet size distribution using laser scattering.
We present a technique to fabricate simple thickness mode piezoelectric devices using lithium niobate (LN). Such devices have been shown to atomize liquid more efficiently, in terms of flow rate per power input, than those that rely on Rayleigh waves and other modes of vibration in LN or lead zirconate titanate (PZT). The complete device is composed of a transducer, a transducer holder, and a fluid supply system. The fundamentals of acoustic liquid atomization are not well known, so techniques to characterize the devices and to study the phenomena are also described. Laser Doppler vibrometry (LDV) provides vibration information essential in comparing acoustic transducers and, in this case, indicates whether a device will perform well in thickness vibration. It can also be used to find the resonance frequency of the device, though this information is obtained more quickly via impedance analysis. Continuous fluid atomization, as an example application, requires careful fluid flow control, and we present such a method with high-speed imaging and droplet size distribution measurements via laser scattering.
Ultrasound atomization has been studied for almost a century and although there are many applications, there are limitations in understanding the underlying physics. The first description of the phenomenon was made by Wood and Loomis in 19271, and since then there have been developments in the field for applications ranging from delivering aerosolized pharmaceutical fluids2 to fuel injection3. Although the phenomenon works well in these applications, the underlying physics is not well understood4,5,6.
A key limitation in the field of ultrasonic atomization is the choice of material used, lead zirconate titanate (PZT), a hysteretic material prone to heating7 and lead contamination with elemental lead available from the inter-grain boundaries8,9. Grain size and mechanical and electronic properties of grain boundaries also limit the frequency at which PZT can operate10. By contrast, lithium niobate is both lead-free and exhibits no hysteresis11, and can be used to atomize fluids an order of magnitude more efficiently than commercial atomizers12. The traditional cut of lithium niobate used for operation in the thickness mode is the 36-degree Y-rotated cut, but the 127.86-degree Y-rotated, X-propagating cut (128YX), typically used for surface acoustic wave generation, has been shown to have a higher surface displacement amplitude in comparison with the 36-degree cut13 when operated in resonance and low loss. It has also been shown that thickness mode operation offers an order of magnitude improvement in atomizer efficiency over other modes of vibration13, even when using LN.
The resonance frequency of a piezoelectric device operating in the thickness mode is governed by its thickness t: the wavelength λ = 2t/n where n = 1, 2,… is the number of anti-nodes. For a 500 µm thick substrate, this corresponds to a wavelength of 1 mm for the fundamental mode, which can then be used to calculate the fundamental resonance frequency, f = v/λ if the wave speed, v, is known. The speed of sound through the thickness of 128YX LN is approximately 7,000 m/s, and so f = 7 MHz. Unlike other forms of vibration, particularly surface-bound modes, it is straightforward to excite higher-order thickness mode harmonics to much higher frequencies, here to 250 MHz or more, though only the odd-numbered modes may be excited by uniform electric fields14. Consequently, the second harmonic (n = 2) near 14 MHz cannot be excited, but the third harmonic at 21 MHz (n = 3) can. Fabrication of efficient thickness mode devices requires depositing electrodes onto opposing faces of the transducer. We use direct current (DC) sputtering to accomplish this, but electron-beam deposition and other methods could be used. Impedance analysis is useful to characterize the devices, particularly in finding the resonance frequencies and electromechanical coupling at these frequencies. Laser Doppler vibrometry (LDV) is useful to determine the output vibration amplitude and velocity without contact or calibration15, and, via scanning, the LDV provides the spatial distribution of surface deformation, revealing the mode of vibration associated with a given frequency. Finally, for the purposes of studying atomization and fluid dynamics, high-speed imaging can be employed as a technique to study the development of capillary waves on the surface of a sessile drop16,17. In atomization, like many other acoustofluidic phenomena, small droplets are produced at a rapid rate, over 1 kHz in a given location, too quickly for high-speed cameras to observe with sufficient fidelity and field of view to provide useful information over a sufficiently large droplet sample size. Laser scattering may be used for this purpose, passing the droplets through an expanded laser beam to (Mie) scatter some of the light in reflection and refraction to produce a characteristic signal that may be used to statistically estimate the droplet size distribution.
It is straightforward to fabricate piezoelectric thickness mode transducers, but the techniques required in device and atomization characterization have not been clearly stated in the literature to date, hampering progress in the discipline. In order for a thickness mode transducer to be effective in an atomization device, it must be mechanically isolated so that its vibration is not damped and it must have a continuous fluid supply with a flow rate equal to the atomization rate so that neither desiccation nor flooding occur. These two practical considerations have not been thoroughly covered in the literature because their solutions are the result of engineering techniques rather than pure scientific novelty, but they are nonetheless critical to studying the phenomenon. We present a transducer holder assembly and a liquid wicking system as solutions. This protocol offers a systematic approach to atomizer fabrication and characterization for facilitating further research in fundamental physics and myriad applications.
1. Thickness mode transducer fabrication via DC sputtering
2. Making electrical and mechanical contact with the transducer
NOTE: Several methods are described below (steps 2.1−2.4), and it is highlighted later in the protocol which method is most appropriate for each subsequent step.
3. Resonance frequency identification via impedance analysis
4. Vibration characterization via LDV
5. Fluid supply
6. Dynamics observation via high-speed imaging
7. Droplet size measurement via laser scattering analysis
Thickness mode piezoelectric devices were fabricated from 128YX lithium niobate. Figure 1 shows a complete assembly to hold the transducer in place with a custom transducer holder used with the passive fluid delivery system developed for continuous atomization. The characterization steps for these devices include determination of the resonant frequency and harmonics using an impedance analyzer (Figure 2). The fundamental frequency of the devices was found to be close to 7 MHz using the technique described in this protocol, as predicted by the thickness of the substrate. Further characterization of substrate vibration was performed using noncontact laser Doppler vibrometer measurements. These measurements determine the magnitude of displacement of the substrate and is usually in the nm range (Figure 3). Continuous atomization is essential to enable practical applications of thickness mode devices, and this has been demonstrated by developing a passive fluid delivery system to the substrate. Finally, two techniques were described to observe droplet vibration and atomization dynamics by performing high-speed imaging and by measuring droplet size distribution as shown in Figure 4 and Figure 5.
Figure 1: The whole assembly of a custom transducer holder. (A) The positions of the transducer holder and the fluid supply assembly are each controlled with articulating arms such that the tip of the wick is just in contact with the edge of the transducer. Inset (B) reveals nature of the electrical and mechanical contact with the transducer electrodes. Inset (C) reveals the nature of the contact between the transducer edge and the fluid wick. Please click here to view a larger version of this figure.
Figure 2: The real s11 scattering parameter values measured over a range of 1−25 MHz for a 127.86° YX lithium niobate device, indicating the presence of a resonance peak at approximately 7 MHz. Please click here to view a larger version of this figure.
Figure 3: A multi-carrier, FFT scan with 5 averages at each point was performed over 9 by 9 scan points defined in a 0.6 by 0.6 mm area in the frequency range 5−25 MHz. The reported displacement is the maximum displacement averaged over all points. The fundamental thickness mode for 0.5 mm thick LN can be seen at 7 MHz, and a weaker second harmonic is present at ~21 MHz. Notice there are multiple narrow peaks at each resonance due to interference with lateral modes. Multi-carrier scans spread the voltage input, so the displacement here is not an accurate measure of the performance of the device. For such a measurement, it is recommended to perform a single-frequency scan at the resonance frequency and with application relevant voltages. For example, this 10 mm x 5 mm thickness mode transducer produces a 5 nm max amplitude at 45 Vpp when driven at 6.93 MHz. Please click here to view a larger version of this figure.
Figure 4: Onset of capillary waves on a 2 µL water drop is indicated by an 8,000 fps video of the fluid interface; the drop is driven by a thickness mode transducer driven at 6.9 MHz, showing the significant time difference between the hydrodynamic response and the acoustic excitation. Please click here to view a larger version of this figure.
Figure 5: Droplet size distribution is typically measured as a volume fraction versus the droplet diameter, here comparing (A) a commercial nebulizer and (B) an LN thickness mode device, both using water. Please click here to view a larger version of this figure.
Supplemental Figure 1: A comparison of the impedance analysis spectra for the same transducer with two different forms of electrical contact (pogo-plate, pogo-pogo, and transducer holder) shows significant differences in s11 scattering parameter values. Please click here to download this file.
Movie 1: LDV vibration mode of 5 mm x 5 mm square transducer. Please click here to view this video. (Right-click to download.)
Movie 2: LDV vibration modes of 3 mm x 10 mm transducer. These are close approximations to thickness modes without the presence of significant lateral modes. Please click here to view this video. (Right-click to download.)
The dimensions and aspect ratio of a transducer affects the vibration modes it produces. Because the lateral dimensions are finite, there are always lateral modes in addition to the desired thickness modes. The above LDV methods can be used to determine dominant modes in the desired frequency range for a given transducer. A square with dimensions below 10 mm typically gives a close approximation to a thickness mode. Three by ten millimeter rectangles also work well. Movie 1 and Movie 2 show LDV area scans of the square and the 3 mm x 10 mm transducers indicating that they are close to the thickness mode. These have been empirically determined rather than selected by simulation and design, though such methods could be used to find ideal lateral dimensions.
The method of electrical and mechanical contact with the transducer also affects the vibrations it produces since these are the boundary conditions to which the piezoelectric plate is subject. We have included an impedance spectrum for three measurement techniques: pogo-plate, pogo-pogo, and transducer holder as a comparison in Supplemental Figure 1. Clearly, the resonance peak locations are not changed in this case by our choices of contact. We do note that mechanical contact between the transducer and a plate surface dampens vibrations making atomization less efficient. Pogo-plate contact is used in the case of LDV measurements, because it is the simplest way to get a flat, stationary surface on which to focus the laser.
The fluid supply assembly described here relies on capillary action and gravity to passively resupply the transducer with a thin film of water as it is atomized away. The vibration of the transducer produces an acoustowetting effect that can be enough to create a thin film and avoid flooding, but in some cases a hydrophilic treatment will be necessary on the transducer surface. If continuous atomization is not achieved, this is the most likely route to resolving the problem.
Measurements were performed with an ultra-high frequency vibrometer (Table of Materials) here, but other LDVs may be used. Electrical contact can be made by soldering a wire to each face of the transducer, though the solder can significantly alter the resonance frequencies and modes of the transducer. Another technique is to place the transducer on a metal base and use “pogo” spring contact probes pressed into contact on the top face of the piezoelectric transducer element while it sits flat upon the stage, useful when a large area has to be scanned. Accurate measurement of the resonance frequencies is important to efficiently operate the transducer and maximize energy throughput to mechanical motion at these frequencies. A frequency scan using the LDV provides this information, but requires a long time, on the order of tens of min. An impedance analyzer can determine the resonance frequencies much more quickly, often less than a minute. However, unlike the LDV, the impedance-based measurement does not provide information on the vibration amplitude at the resonance frequencies, which is important in determining fluid atomization off the surface of the transducer.
Though vibration of the substrate occurs in the 10−100 MHz regime, the dynamics of fluids in contact with the substrate occur at far slower time scales. For example, capillary waves on the surface of a sessile drop are observable at 8,000 fps, assuming that the spatial resolution of the camera can distinguish the amplitude of a wave crest and that the wave frequency of interest is below 2,000 Hz. The camera arrangement described above images transmitted light and thus is good for observing the outline of objects that transmit light differently than air. If insufficient, a reflected or fluorescent light arrangement may be required. The exposure time for each frame decreases as the frame rate is increased so the light intensity must be increased accordingly. The objective lens should be chosen based on the length scale of the phenomenon under study, but the above protocol will work with any commonly available magnification. As an example, Figure 4 was obtained with the above high-speed video method. The contrast at the drop interface would allow these frames to be segmented in software (ImageJ and MATLAB) so that the interface dynamics may be tracked over time.
In the droplet sizing equipment used in this protocol (Table of Materials), the laser optics and scattering detectors are relatively standard but the software is proprietary and complex. In addition to Mie theory, multiple scattering events make droplet size and enumeration calculations much more difficult. Mie theory assumes that most photons are scattered only one time, but when droplets are densely spaced, i.e., the spacing between droplets is not much larger than the droplets themselves, and the spray plum covers a suefficiently large area, then this assumption fails18. As an example of troubleshooting results from this instrument, consider Figure 5. Notice that the 0.5 mm diameter peak appears in both distributions. The commercial nebulizer is known to produce monodisperse droplets near 10 µm, so the larger peak is likely either a false result due to the large amount of multi-scattering events or agglomeration of smaller droplets within the spray. This implies that the large peak in the thickness mode distribution may also be a false result. This can be directly verified by high-speed video: such large droplets would be readily visible, but they are not observed in this case.
Laser scattering particle size analysis can also be difficult when the scattering signal becomes weak. This is typically due to a low atomization rate or when part of the spray does not pass through the laser path. A weak vacuum may be used to draw the complete atomized mist through the expanded laser beam of the equipment in cases where it would otherwise escape measurement. For even greater control of spray conditions a humidity chamber can be installed around the laser beam path, but this is not required.
The authors have nothing to disclose.
The authors are grateful to the University of California and the NANO3 facility at UC San Diego for provision of funds and facilities in support of this work. This work was performed in part at the San Diego Nanotechnology Infrastructure (SDNI) of UCSD, a member of the National Nanotechnology Coordinated Infrastructure, which is supported by the National Science Foundation (Grant ECCS−1542148). The work presented here was generously supported by a research grant from the W.M. Keck Foundation. The authors are also grateful for the support of this work by the Office of Naval Research (via Grant 12368098).
Amplifier | Amplifier Research, Souderton, PA, USA | 5U1000 | |
Articulating arm | Fisso, Zurich, Switzerland | ||
CF4 Objective | Edmund Optics, Barrington, NJ, USA | Objective used for high speed imaging | |
Dicing saw | Disco, Tokyo, Japan | Disco Automatic Dicing Saw 3220 | |
Fiber Fragrance Diffuser Wick | Weihai Industry Co., Ltd., Weihai, Shandong, China | https://www.weihaisz.com/Fiber-Fragrance-Diffuser-Wick_p216.html | |
High Speed Camera | Photron, San Diego, USA | Fastcam Mini | |
Laser Doppler Vibrometer | Polytec, Waldbronn, Germany | UHF120 | Non-contact laser doppler vibrometer |
Laser Scattering Droplet size measurement system | Malvern Panalytical, Malvern, UK | STP5315 | |
Lithium niobate substrate | PMOptics,Burlington, MA, USA | PWLN-431232 | 4” double-side polished 0.5 mm thick 128°Y-rotated cut lithium niobate |
Luer-lock syringes | Becton Dickingson, New Jersey, USA | ||
Nano3 cleanroom facility | UCSD, La Jolla, CA, USA | Fabrication process is performed in it. | |
Network Analyzer | Keysight Technologies, Santa Rosa, CA, USA | 5061B | |
Oscilloscope | Keysight Technologies, Santa Rosa, CA, USA | InfiniiVision 2000 X-Series | |
PSV Acquistion Software | Polytec, Waldbronn, Germany | Version 9.4 | LDV Software |
PSV Presentation Software | Polytec, Waldbronn, Germany | Version 9.4 | LDV Software |
Signal generator | NF Corporation, Yokohama, Japan | WF1967 multifunction generator | |
Single Post Connector | DigiKey, Thief River Falls, MN | ED1179-ND | |
Sputter deposition | Denton Vacuum, NJ, USA | Denton 18 | Denton Discovery 18 Sputter System |
Surface Mount Spring Contacts | DigiKey, Thief River Falls, MN | 70AAJ-2-M0GCT-ND | |
Teflon wafer dipper | ShapeMaster, Ogden, IL, USA | SM4WD1 | Wafer Dipper 4" |
XYZ Stage | Thor Labs, Newton, New Jersey, USA | MT3 | Optical table stages |