Bimitech White Paper (web-version)A novel biomimetic piezoelectric fan for electronics cooling
This paper presents a novel piezoelectric flapping fan. The fan includes a bio-inspired thin wing attached to piezoelectric bending actuator supported through two connecting members. When activated by an AC signal, the piezoelectric actuator deflects up and down, inducing the wings to oscillate around their neutral position, generating a net air flow. The performance of the fan was characterized experimentally using a custom flow measurement system based on pressure difference and a high-speed camera. The cooling capability of the fan was also demonstrated with a fan-heatsink setup. The experimental results show this fan is capable of generating significant flow rate, thus making it a promising candidate for electronics cooling.
Fans are broadly used to move air in active cooling systems. The typical high-performance rotary fans enable heat transfer coefficients that are several times higher than those of passive cooling systems . However, rotary fan bearing lubrication usually dries out after a certain time period leading to fan failure. Due to these reliability issues, rotary fans are not used in applications requiring high life expectancy, such as high-power LED lighting and telecommunication electronics . Piezoelectric fans [3, 4] and synthetic jet cooling devices  have been designed to fill in this gap as they have many advantages over the rotary fan including increased reliability, reduced noise, and lower weight. Piezo electric fan operation is simple: the structure that is attached to the piezoelectric material (which will be exclusively called “the wing” in this study) oscillates around a neutral position, moving air around it. The air flow generated from the oscillation can enhance the rate of convective heat transfer between the moving air and the surfaces of the heated electronic components. There is much potential for piezoelectric fan use for spot cooling in low power applications. However, the present piezoelectric fans generate very little net air flow [6, 7], which greatly limits their usage in these applications. Similarly, synthetic jet devices are quite reliable but suffer from low performance therefore adoption has been limited. Currently there is a lack of cooling devices in the market that offer both high performance and robust reliability. This work presents a novel flapping fan design that addresses both of the discussed limitations of the existing air moving solutions. The fan includes two wings designed to mimic a hummingbird’s structure and flapping motion in order to generate air flow. After millions of years of evolution, hummingbirds have evolved to have a nearly perfect system of flight and aerodynamic efficiency. Unlike any other bird wings, hummingbirds flap their wings in the same manner as insects. The wing rotates around its long axis during transitions between forward and backward strokes, allowing it to generate significant lift on both strokes allowing the bird to hover . Using the power factor as a measure of the aerodynamic efficacy, research has found that most hummingbirds are about 25% more efficient than the best drones designed to date . This suggests that an efficient air mover could be created by mimicking the hummingbird flapping mechanism. The wings of the novel flapping fan are driven by two advanced piezoelectric bending actuators. Fan curve characterization is achieved using a custom-built testing apparatus, while a high-speed camera is used to analyze the wing motion. The cooling capability of the fan is demonstrated by removing heat from a high-power LED light. The temperature of the LED light is characterized using thermocouples with and without the fan. For comparison purposes, the fan curve and thermal characterization are also performed with conventional rotary fans. The results show that the fan is a great candidate for high power electronic cooling applications.
This piezoelectric fan includes two identical thin wings arranged in parallel, as shown in figures 1(a), (b), (c) and (d). Each wing is connected to a piezoelectric bending actuator through two small connecting members. The four ends of the actuators are attached to two plastic legs using four rubber living hinges. This simply supported boundary condition is used to maximize the deflection of the actuators. The wing includes a silk fabric membrane attached to a carbon fiber frame. The frame consists of an elongated beam and multiple ribs branching from the beam.
The dotted lines in figures 1(b) and (c) demonstrate the oscillation extent of the actuators when activated by an AC signal. The piezoelectric actuators drive the wings to oscillate around their neutral positions, generating air flow. Figure 2 shows a picture of the fan. The overall size of the fan is 50x45x140 mm, while the mass of the fan is 19 grams.
A high speed camera (IDT MotionPro Model Y4-S1) was set up to record the fan in motion from the side view. The sampling rate of the camera was set at 4590fps. The fan was illuminated with two beam focused 150W LED lights. The initial results of the high speed images are processed using Open Source image processing software ImageJ ver1.52a. Figure 3 (below) shows a sequence of the wing motion during one flapping cycle. Three consecutive flapping cycles were analyzed and the motion of the wing is found to be repeatable.
The motion of the leading edge and trailing edge of the wing is tracked and plotted in Figure 4 (below) The displacement of the trailing edge from the neutral position is defined as the rotation displacement shown on the primary axis in Figure 4b, and the displacement of the leading edge from the neutral position is called the stroke displacement, shown on the secondary axis. At anytime t, the measured rotation displacement and the stroke displacement of the wing can be fitted as:
Where f is the oscillation frequency, and are the rotation and stroke amplitude, respectively. h is the phase shift. The oscillation frequency of the wing is 65HZ, which is 8% higher than the frequency of the AC power input. Studies [6,7,10] have shown that the piezoelectric fans are most efficient when the first resonant frequency of the wing is set at the frequency of the power input. The difference between the oscillation frequency and the AC power input frequency in this study could be due to the resolution of the high speed images. The rotation amplitude is 13mm, which is 4.2 times larger than the stroke amplitude.
In addition, Figure 4 (below) shows a phase shift between and. Similar shift has been observed in motion of hummingbirds measured in previous literature . Figure 5 shows stroke and rotation angles recorded for several hummingbirds. The kinematic changes associated with the phase shift is accompanied with an increase of speed, which could lead to higher flow rate. Furthermore, the maximum rotation and stroke angle of the piezoelectric fan are 32 degrees. It is significantly smaller than those of the hummingbirds (about 70 degrees shown in figure 5, below). This suggests that there is still a lot of room to further improve performance and efficiency of the fan presented in this paper. For example, the geometrical parameters including the width, length and the shape of the wing, as well as the wing material will be optimized in future studies.
A testing apparatus shown in Figure 6 (below) was built to evaluate the performance of the fan. It was constructed in accordance with AMCA Standard 210-99 . The apparatus includes an air-tight chamber having an opening to receive air flow from the fan. A converging nozzle is connected to the chamber, two differential pressure transducers were used to measure pressure in the chamber and across the nozzle. An auxiliary fan attached at the end was used to adjust the pressure in the air-tight chamber. The volumetric flow rate can be calculated using the following equation :
where C is the nozzle discharge coefficient, is the area of the nozzle throat, d2 is the throat diameter, Dp is the pressure drop through the nozzle, r is the air density, B = d2/d1 is the contraction ratio of the nozzle, and d1 is nozzle inlet diameter.
The fan was operated with multiple AC voltages from 200 Vp-p to 320 Vp-p using a piezo amplifier (model A-304 from A.A. Lab Systems). For comparison purposes, the fan curve characterization was also performed with two 60x60x15 mm rotary fans (Sofaco model sD6015V24MBL) arranged in parallel shown in figure 7a. The fans were attached to an acrylic plate with two 60mm diameter openings, which are 10mm from each other. They are selected for comparison because the height and length of the layout (60mm x 130mm) with the two fans selected spaced at 10mm is close to the footprint of the biomimetic flapping fan presented in this study (50mm x 140mm). Figure 7 (below) shows the rotary fans and the experimental setup.
The fan curves are shown in Figure 8 (below). The pressure and the flow rate are normalized using maximum pressure and maximum flow rate for each experiment to find a general relationship. Both the flapping fan and the rotary fan show a quadratic relationship between the pressure and flow rate. The curve has convex shape for the two rotary fans, whereas the curve for the flapping fan at various voltage shows a concave shape. The flow rate of the flapping fan decreases dramatically with the pressure, and it is more sensitive to static pressure compared to rotary fans. The maximum flow rate of the two rotary fans is about 21.7 CFM, while that of the piezoelectric flapping fan is 27% less, about 15.9 CFM. The maximum pressure head of the rotary fans (9.5 Pa) was found to be about 3x higher than that of the flapping fan (3.37 Pa). The flow rate of the flapping fan measured in this study is more than 10 times higher than the flow rate of a typical piezoelectric fan . It is worth noting that the flow rate measurements presented by Kimber et al  is for a single piezoelectric fan operating at 120 volts. The width of blade in their study is 12.7mm, whereas the wing of the flapping fan is 60mm. Fan arrays with various orientations instead of a single piezoelectric fan has been proposed to increase the air flow [13, 14] but the flow rate was not reported. Further study will be conducted to make a fair comparison between them.
Power consumption of the flapping fan was measured at 120VAC, 60Hz using a wattmeter. Power consumption at other driving voltages was then calculated using the following equation :
Figure 9 (below) shows the power consumption versus driving voltage for the flapping fan. The DC input voltage of the two rotary fans is 24V and their total power consumption is 1.9W (0.95W each).
A high-power LED light was built to demonstrate the cooling capability of the flapping fan as shown in Figures 10(a)(b)(c) below. The light includes a chip-on-board (COB) LED module (Bridgelux Vero 29 BXRC-50C10K1-C-74) attached to an extruded aluminum heat sink. The heat sink size is 150x70x60 mm3. It has eighteen 53-mm-height fins and a 7-mm-thick base. Thermal grease was used between the COB and heat sink to reduce the interfacial thermal resistance. Temperature was monitored using a thermometer and two type-K thermocouples. One is attached to the operating case temperature point on the COB, and the other is attached to a point on the heat sink base which is 35mm from the edges. The COB is driven by a constant current DC power supply (Meanwell HLG-320H-C3500A). The potentiometer on the power supply was adjusted so that the total power consumption of the LED light was 200W.
The temperature measurements are shown in Figure 11 below. The operating case temperature with the flapping fan (T1_flapping) (69o C) is quite close to that measured with the rotary fans (T1_rotary), while the temperature on the heat sink with the flapping fan (T2_flapping) is slightly higher than with the rotary fans (T2_rotary). This result seems counterintuitive as the rotary fans have significantly higher flow rate and pressure head than the flapping fan as shown in Figure 8. However, the flapping and rotary fans appear to have significantly different airflow profiles, which could greatly affect their cooling capability. Rotary fans do not produce airflow at the center region, which is the location of the fan hub. Furthermore, the swirling airflow  from the rotating blades is less effective for heat removal from a heat sink with parallel fins. Future work will characterize the flow field resulting from the piezoelectric fan.
A novel piezoelectric flapping fan was introduced in this paper. The fan was built and its performance was characterized. It is capable of generating a maximum flow rate of 16CFM, while its maximum pressure head is 3.4 Pa. The thermal performance of the fan was evaluated in a high-power LED cooling application. Its performance is similar to two rotary fans with higher flow rate and static pressure. The fan is demonstrated to be a promising candidate for electronics cooling. Fan motion was analyzed using high speed camera. The results reveal that the wing kinematics of the fan has a similar flapping motion to that of hummingbird’s, which could be the contributing factor to the increasing flow rate, comparing with other piezoelectric fans. Furthermore, there is much potential to improve this advanced fan, and more research is required to further understand the flapping motion and resulting air flow as well as to optimize the fan performance.
 S. Lee, Optimum design and selection of heat sinks, Proceedings of 1995 IEEE/CPMT 11th Semiconductor Thermal Measurement and Management Symposium.
 C. JM. Lasance, A. Poppe, Thermal management for LED applications, Volume 2 of Solid State Lighting Technology and Application Series, 2013, Springer
 K. H. Tseng, M. Mochizuki, K. Mashiko, T. Kosakabe, E. Takenaka, K. Yamamoto, and R. Kikutake, Piezo Fan for Thermal Management of Electronics, Fujikura Technical Review, 2010.
 M. Kimber, S. Kazuhiko, K. Nobutaka, S. Kenichi, S. V. Garimella, Quantification of piezoelectric fan flow rate performance and experimental identification of installation effects, 2008 11th Intersociety Conference on Thermal and Thermomechanical Phenomena in Electronic Systems, pp. 471 – 479.
 Synthetic jet cooling device, https://www.ledsmagazine.com/ugc/2010/05/nuventix-introduces-synjet-cooltwist-led-cooler-for-philips-twistable.html
 M. Maaspuro, Piezoelectric oscillating cantilever fan for thermal management of electronics and LEDs — A review, Microelectronics Reliability, Volume 63, August 2016, Pages 342-353, https://doi.org/10.1016/j.microrel.2016.06.008
 A. Hales, X. Jiang, A review of piezoelectric fans for low energy cooling of power electronics, Applied Energy, 215 (2018) pp 321-337
 B. Cheng, B. W. Tobalske, D. R. Powers, T. L. Hedrick, S. M. Wethington, G. Chiu, and X. Deng, Flight mechanics and control of escape manoeuvres in hummingbirds. I. Flight kinematics, Journal of Experimental Biology 2016 219: 3518-3531.
 J. W. Kruyt, E. M. Quicazán-Rubio, G. F. van Heijst, D. L. Altshuler, and D. Lentink, Hummingbird wing efficacy depends on aspect ratio and compares with helicopter rotors, J. R. Soc Interface, 11 https://doi.org/10.1098/rsif.2014.0585
 S.M. Wait, S. Basak, S.V. Garimella, A. Raman, Piezoelectric fans using higher flexural modes for electronic cooling applications. IEEE Transactions on Components and Packaging Technologies, vol. 30, no. 1, pp. 119-128, March 2007, doi: 10.1109/TCAPT. 2007.892084
 ANSI/AMCA Standard 210-99, Laboratory methods of testing fans for aerodynamic performance rating.
 F. M. White, Fluid Mechanics. McGraw-Hill, New York, fourth edition,1999.
 H. K. Ma, H. C. Su, W.F. Luo, Investigation of a piezoelectric fan cooling system with multiple magnetic fans, Sensors and Actuators A: Physical, vol 189, pp 356-363, Jan 2013, https://doi.org/10.1016/j.sna.2012.09.009
 M. Kimber, R. Lonergan, S. V. Garimella, Experimental study of aerodynamic damping in arrays of vibrating cantilevers, Journal of Fluids and Structures, vol 25, Issue 8, pp1334-1347, November 2009, https://doi.org/10.1016/j.jfluidstructs.2009.07.003
 T. Jordan, Z. Ounaies, J. Tripp, and P. Tcheng, Electrical properties and power consideration of a piezoelectric actuator, https://ntrs.nasa.gov/archive/nasa/casi.ntrs. nasa.gov/20040110251.pdf
Piezoelectric Bimorph Actuator Design guidelines
The effect of substrate material and thickness of a bimorph actuator on its performance was studied using the finite element method (FEM). Experiments were also performed and the result was used to validate the FEM model. The FEM result shows that performance is more sensitive to the thickness than the Young’s modulus of the substrate. Material selection for the substrate was also discussed.
Piezoelectric actuators can be categorized based on their displacement generating mechanisms as either stack actuators or benders. A stack actuator usually includes multiple layers of piezo plates stacked on top of each other and generating displacement thanks to longitudinal expansion of each piezo layer when a voltage is applied. A piezo bender usually includes one or more piezo layers attached to a substrate. When voltage is applied, the piezo layer expands in-plane. Since it is mechanically constrained by the substrate, the actuator bends out-of-plane . Piezoelectric bending actuators (benders) have been used in many applications such as sensing, actuating, and energy harvesting for many years. The actuators have many advantages such as nano resolution, millisecond response, and low power consumption. Piezo benders can be classified as: unimorph, birmorph, and multimorph [2, 3] – depending on the number of piezo layers they include. Bimorph benders are the most popular as they are easy to manufacture, bi-directional, and relatively easy to control. A bimorph typically includes two identical piezoelectric layers attached to two sides of a substrate. The attachment is usually achieved using epoxy. The substrate can be made of a variety of different materials such as steel, brass, or fiber composite. Stiffness and thickness of the substrate are two critical design parameters of bimorph benders. In this paper, the effects of these parameters on the performance of a bimorph bender are studied using the design of experiments (DOE) and the finite element method (FEM) . Design guidelines are provided based on the DOE results.
Experiments were performed to generate data points for calibrating the FEM model (next session). A bimorph actuator composed of two PZT layers attached to a steel substrate using expos was fabricated. Its structure and dimensions are shown in Figure 1(a). A photo of the actuator is shown in Figure 1(b).
Performance (displacement and blocking force) of the actuator was evaluated using the setup shown in Figure 2. The actuator was clamped at one end. An AC sine wave was generated using a function generator (Rigol DG1022U) in conjunction with an amplifier (A.A LAB A-301 HS) and supplied to the actuator. Tip displacement was measured using a non-contact laser sensor (LK-031), while blocking force was measured using a force gauge (Mark-10 M5-2).
A ½ symmetry FEM model was built in Ansys to simulate the bimorph as described in the experiment section above. The model contained 3,816 3D elements and 18,485 nodes. To obtain deflection the model was clamped at one end while the other end was free. The same simulation was repeated with the free end constrained in a vertical direction in order to obtain blocking force simulations.
Electro-mechanical behavior of the PZT layer was captured using a simple thermal analogy. Temperature applied to the top PZT layer was calculated using equation 1.
Where d31 is the piezoelectric charge constant, V is the applied voltage, a is the coefficient of thermal expansion, and t is the thickness of the PZT layer. Properties of the materials are shown in table 1 below.
Figure 4 (below) shows the comparison between the simulation and experiment results. As can be seen, the simulation results match the experiment results fairly well. The model was then used for a design of experiment (DOE) study. In the FEM model, the thickness of the substrate layer was varied from 0 to 0.25mm, while its Young’s modulus was varied from 0 to 200 GPa. The other layer in the model remains the same. Displacement and blocking force were obtained from the simulation and compared.
Figures 5(a) and 5(b) below show the relationship between the actuator displacement and blocking force, respectively, versus Young’s modulus of the substrate. Both displacement and blocking force increase rapidly as the Young’s modulus increases from zero. This is because a higher Young’s modulus provides a stronger mechanical interaction between the top and bottom piezo layers. The displacement reaches its peak when the Young’s modulus approaches 2 GPa. Beyond that, the displacement is almost constant (up to 200 GPa). Similarly, the blocking force increases rapidly when the Young’s modulus increases from 0 to 2 GPa, and only slightly increases when the Young’s modulus exceeds 2 GPa. This result suggests that the substrate should be made of a material with a Young’s modulus greater than 2 GPa. Almost all popular materials for substrates such as carbon fiber composite (~200 GPa), brass (~110 GPa), aluminum (~70 GPa), steel (~200 GPa), epoxy (~2-10 GPa), and FR4 (~20 GPa) have a Young’s modulus significantly higher than 2 GPa.
Result at a Young’s modulus of 10 GPa were extracted (Figure 5 below) and replotted as shown Figure 6 to better demonstrate the effects of substrate thickness on displacement and blocking force. This shows a trade-off exists between displacement and blocking force. Displacement decreases while the blocking force increases with the substrate thickness. This suggests that if high displacement is desired, substrate thickness should be minimized. In practice, aluminum, brass or steel shims with a thickness as low as 25 µm are commercially available and can be used to achieve this. In the case of high blocking force being desired, a higher substrate thickness should be selected.
Even though this paper investigates bimorph actuators with a rectangle shape, the design guidelines should also apply to round shaped bimorphs. This work did not study the effects of seal/protective layers (as seen in our PythonTM) on performance of the bimorph. A substrate with non-uniform thickness was not studied either. Both of these would be great topics for future study.
Experiments and FEM simulations were used to study the effect of thickness and Young’s modulus of the substrate of a typical bimorph actuator on its displacement and blocking force. Both displacement and blocking force were found to be insensitive to the Young’s modulus as long as the Young’s modulus is high enough (>2 GPa). Substrate thickness has a strong effect on the displacement and blocking force. When high displacement is the design objective, substrate thickness should be as thin as possible, while a thicker substrate should be selected if higher blocking force is desired. FEM was found to be an effective tool for bimorph actuator design.
 Zuo-Guang Ye et al., Handbook of advanced dielectric, piezoelectric and ferroelectric materials, 2008
 Kenji Uchino, Introduction to Piezoelectric Actuators and Transducers, 2003
 Shuxiang Dong, Review on piezoelectric, ultrasonic, and magnetoelectric actuators, Journal of Advanced Dielectrics, 2012
 Michal Staworko, Modeling and simulation of piezoelectric elements – Comparison of available methods and tools, Mechanics Vol. 27, No. 4, 2008