Final Report Summary - MSAM (Micro Synthetic Jet Actuator Manufacturing)
Executive summary:
In this report the design, manufacturing and characterization of the micro-technology based synthetic jet actuator is described. In the design and optimization work package a Combined Simulation μSJA-model was created. Therefore, the functionality was virtually split into two sub-systems. The Finite Element Method (FEM) was used to determine relevant lumped parameters of a piezoelectric transducer - electromechanical sub-system, which can be used with the lumped element method (LEM). Furthermore, lumped element models for the acoustic sub-system already exist. These were adapted for the μSAM geometry and a complete electro-acoustic simulation was done by bringing together and solving the LE models. Relevant constraints and parameters for the simulation had been defined and determined. Further acoustic investigations identified possible failure sources due to additional eigenmodes. Other failures sources were discussed. The simulation results were compared with optical deflection measurement results of the piezoelectric transducer and with flow measurements of the μSJA. Out of the preliminary simulation results an optimal design had been chosen.
The wafer preparation and wafer bonding work packages manufactured the whole μSJA chip. This includes the silicon manufacturing of the nozzle- and the membrane wafers, the bonding of the membrane- and the nozzle wafers to the µSJA system, the dicing of these bonded μSJA wafers to μSJA chips as well as the bonding of the piezoelectric transducer, (PZT ceramics) to the diced μSJA chips. The most important challenge of the μSJA manufacturing is the integration of the PZT ceramic into the μSJA system because this faces some limitations concerning the whole manufacturing- as well as the integration concept. Therefore two wafer bonding concepts as well as different PZT ceramic to silicon bonding concepts were investigated regarding their feasibility to fulfill the aimed requirements. The wafer bonding concept has been split into the 'wafer bonding first, followed by the PZT ceramic bonding' and the 'PZT ceramic bonding first, followed by the wafer bonding' approaches.
The PZT ceramic bonding concept concerning several bonding techniques such as bonding with reactive layers, bonding with the negative photoresist SU8, bonding with epoxies and bonding with different types of adhesive tapes have been investigated. Finally, corresponding the aimed requirements such as a low bonding temperature and a low bonding pressure as well as a low stress input into the whole system especially during the PZT ceramic bonding step one approach could be proven with excellent results. This approach is characterized by the first wafer bonding concept that is a direct bonding step with a special pre-treatment of the bonding areas and a followed temperature step at 200 °C. Afterwards the bonded μSJA system is diced into chips and the PZT ceramics are bonded on the thin silicon membrane of the μSJAs. This is realized with an electrically conductive adhesive foil with a bonding temperature of 130 °C and a bonding pressure of 0,5 bar. The connection to the overall system is realized by the integration of the μSJAs into the system and by wire bonding.
After bonding of the transducers and the wafer, the actuators are tested to evaluate the functionality and to verify the single actuator characteristics. Beside the development and manufacturing of the silicon based actuators the design and the development of an electronics system for the controlling of the actuator are done. The system is divided into signal generator and transducer electronics. The signal generator provides the actuation signal depending on the simulation results and the geometrical data of the actuator. The transducer electronics amplifies this signal to meet the requirements of the actuators power consumption depending on the used actuation principle. To test the whole system including a number of actuators and the driving electronics, the single components are integrated into a mock-up. This mock-up will be system designed in that way that it can be integrated into a wind tunnel model. After finalizing the manufacturing and assembly of all the components of the mock up, the system performance is verified by measuring the output velocities of the actuator. Besides that, test and characterisation of the test board are performed. All the results and manufacturing details are summarized in this final report.
Project Context and Objectives:
1. Introduction
A Synthetic Jet Actuator (SJA) is a low power, highly compact micro fluidic device which has potential application in boundary layer flow control. It is a generator that requires zero mass input and produces non-zero momentum output [1]. It consists of a membrane located on one wall of a small cavity, which has an orifice in another face, typically opposite the membrane. The membrane is forced to oscillate, with fluid being expelled through the orifice as the membrane moves upwards. The flow separates at the edge of the orifice, inducing a vortex ring that moves outwards under its own momentum generating a jet in the ambient flow. When the membrane moves downwards, fluid is entrained into the cavity. If the vortex ring is sufficiently distant from the orifice, it is not influenced by the entrainment of fluid into the cavity [2]. To maximize the exit velocity of the actuator, the SJA is normally driven in resonance frequency of the membrane. There are many types of transducers that can be used in actuators for active flow control, such as piezoelectric [1], [3]-[6], electromagnetic [7]-[9] and shape memory alloys [10].
According to the different types of actuation, several investigations have been done to obtain best geometry dimensions of a SJA for these points of actuation. The quality of a SJA depends on several parameters, such as the diameter of the membrane and the orifice, the height of the cavity, the resonance frequency and the deflection of the membrane. The piezoelectric actuation is used because of its low power consumption, fast response, high reliability and low costs. In [1],[3],[5] SJAs are presented, that are produced with membrane diameters from 15 mm to 40 mm, orifice diameters from 0.5 mm to 1.2 mm and a resonance frequency from 1.2 kHz up to 2.5 kHz. With these parameters they reached a jet stream velocity of up to 130 m/s. Mane et al. [4] used a special piezoelectric configuration - Bimorph and Thunder - with a large membrane diameter of 63.5 mm and orifice diameters of 2 mm (Thunder) and 3.67 mm (Bimorph). In resonance at 90 Hz they achieved a jet stream velocity of 28 m/s. A Lumped Element Model (LEM) for a SJA was developed by Gallas [6], in which he simulated a SJA with piezoelectric actuation in different dimensions at a resonance frequency of 800 Hz and a membrane diameter of 20 mm. With this LEM jet streams velocities up to 60 m/s were evaluated.
Another actuation concept is the electromagnetic transducer. Presented in [7] a conventional loudspeaker is used as transducer for the SJA. A constant maximum velocity of 301 m/s is realized due to a nozzle of 4 mm and a resonance frequency of 50 kHz. Coe et al. [9] developed a special kind of SJA, an electromagnetic actuated SJA with additional pneumatic actuation. A membrane with 89 mm is forced to oscillate by a loudspeaker and generates the synthetic jet. The pneumatic actuator deflects valve structures under the orifices, which are arranged in an array. Thereby the synthetic jet can be directly addressed to a special point up to 20 m/s at a resonance frequency of 20 Hz. Liang et al. [10] designed a SJA based on ferromagnetic shape memory alloy (FSMA) composite with different membrane diameters (64 mm and 72 mm) and a slot like orifice with a dimension of 5 mm x 1 mm. With this FSMA SJA Liang gained jet velocities up to 190 m/s at a resonance frequency of 220 Hz. A very special kind of actuation was invented by Miller [11]. He used a DC piston driven SJA and obtained jet velocities of 124 m/s at a resonance frequency of 100 Hz, at an orifice diameter of 0.4 mm. Until now there is just one publication on μSJA based on silicon manufacturing techniques available. Malinson et al. [2] presented a μSJA which is micro machined in silicon. The top wafer is etched to produce a thin diaphragm with a thickness of 50 μm to 100 μm on which an orifice with a dimension of 4 μm x 4 μm is formed in different types. The bottom wafer is also etched to form the 10 μm to 20 μm thin membrane with a dimension of 200 μm x 200 μm. Two 300 μm silicon wafers were used, which were bonded together to form a cavity with an orifice on one side and an opposing thin diaphragm. For actuation he used a PZT mounted on the thin membrane. Malinson simulated the μSJA with a maximum stream velocity of 4 m/s and than based on the simulation Malinson produced the μSJA.
2 Overall strategy of the work plan
The objective of the project μSAM is to transform the SJA concept in a μSJA that is micro machined in silicon with optimized design based on LEM modelling to increase exit velocity and decreasing weight at the same time. To reach the required output velocities of the μSJA, the chamber and the exit channel or nozzle are optimized using different analytic and numeric methods. Based on the results of the optimization, a design will be investigated and transferred to silicon based structures. The production of the μSJA is performed in the clean rooms of the ZfM and is realized on wafer-level. The two different parts of the μSJA, the cavity and the diaphragm are micro-machined and wafer bonded in order to form the actuator.
To integrate the electro mechanical transducer, in this case the PZT element, into the system, investigations of bonding technologies for the use of PZT silicon bonding are done. Thus, the strength of the low temperature bond of the ceramics, concerning low inducted mechanical stress, is optimized. The objective of the bonding process is the integration of the PZT element in the micromachined silicon substrate. Integration of a mechanically stable bond means, allowing the propagation of the generated deflection to the thin silicon membrane with almost no losses.
After bonding of the transducers and the wafer, the actuators are tested to evaluate the functionality and to verify the single actuator characteristics. Beside the development and manufacturing of the silicon based actuators the design and the development of an electronics system for the controlling of the actuator are done. The system is divided into signal generator and transducer electronics. The signal generator provides the actuation signal depending on the simulation results and the geometrical data of the actuator. The transducer electronics amplifies this signal to meet the requirements of the actuators power consumption depending on the used actuation principle. To test the whole system including a number of actuators and the driving electronics, the single components are integrated into a mock-up. This mock-up will be system designed in that way that it can be integrated into a wind tunnel model. After finalizing the manufacturing and assembly of all the components of the mock up, the system performance is verified by measuring the output velocities of the actuator. Besides that, test and characterization of the test board are performed.
3 Structure of the project
The project is managed by the Project Coordinator (PC) Prof. Dr. Karla Hiller, who is also the leader of the Work Package (WP) 1: Management. The Project Coordinator is the main interface between the Chemnitz University of Technology and the European Commision (EC) as well as the JTI SFWA consortium. She has the overall responsibility for the achievement of the technical and business objectives of the project. Every single work package has been led by a Work Package Leader (WPL). His task is the organisation of the work in the work package as well as the communication with related work packages. The WPLs worked closely together with the PC and the SFWA consortium.
The project is divided in 6 work packages (WP1: Management, WP2: Design, WP3: Wafer preparation, WP4: Wafer bonding, WP5: Electronic design and WP6: Integration and Test), each has been coordinated by one WPL. To report the status of work and the outcome of the different WPs, 6 main deliverables were scheduled. The monitoring of the technical status and progress has been managed by Milestones and quarterly Meetings.
WP1 coordinates the WP interactions in order to ensure the fulfillment of the tasks and milestones as well as the work plan. Therefore, regular meetings with all team members are organised. Outputs in form of external deliverables have been created and submitted to the cleansky consortium. Different fabrication and bond steps have been compared as well as alternative electronic concepts.
WP2 provides the basement of the project. It is fundamental to have a design concept for the fabrication process to ensure the functionality further μSJAs. For this the summarising and investigation of requirements for a μSJA, the supply of an optimised design of a μSJA, an efficient modelling for optimisation purpose and investigations of important constraints and their influence on the system have been done.
In WP3 the enticing masks are designed and fabricated. The wafer processing steps, including the etching (KOH) of Si wafers for the manufacturing of silicon membranes, cavities and nozzles are performed. Furthermore the pretreatment of silicon wafers for wafer bonding and for PZT silicon bonding are realized. Finally the (In situ) measurement of the membrane thickness and the cavity geometry characterizes the achieved results within this WP.
The development and test of novel bonding technologies for PZT ceramics are the main objectives of WP4. Therefore low temperature bonding technologies, the optimization and the characterization of the bonding process as well as the process adaption are the tasks to be performed.
WP5 includes the design and the development of an electronics system for the controlling of the actuator. The system is divided into signal generator and transducer electronics. The signal generator provides the actuation signal depending on the simulation results and the geometrical data of the actuator. The transducer electronics amplifies this signal to meet the requirements of the actuators power consumption depending on the used actuation principle.
Objective of the WP6 is the integration of the developed and manufactured MEMS based SJAs in a mock-up. To test the whole system including a number of actuators and the driving electronics, the single components are integrated into a mock-up and characterized. This mock-up is system constructed that way that it can be integrated into a wind tunnel model.
Project Results:
1. Design of the μSJA
In the design and optimization work package a Combined Simulation μSJA-model was created. Therefore, the functionality was virtually split into two sub-systems. The Finite Element Method (FEM) was used to determine relevant lumped parameters of the piezoelectric transducer (electromechanical sub-system), which can be used with the lumped element method (LEM). It could be shown, that it is feasible to use the Finite Element Method to simulate the behaviour of the electromechanical subsystem, because for the aimed geometry no analytical expressions are available and a derivation would be extremely time consuming. Furthermore, lumped element models for the acoustic sub-system already exist. These were adapted for the μSAM geometry and a complete electro-acoustic simulation was done by bringing together and solving the LE models. This is a time-efficient concept because of its high degree of abstraction. Unfortunately fluid mechanical expression can only be used for simple axial symmetrically geometries.
2. Design and optimization
The preparation of a model for the Combined Simulation of μSJAs was done by splitting the μSJA into sub-systems. In 1a) the Finite Element Method (FEM) has been used to determine relevant lumped parameters of a piezoelectric transducer, which can be used with the lumped element method (LEM) 1b). Lumped element models can be treated with commercial circuit simulation software like LTSpice. Lumped element models for acoustic components already exist, thus these were adapted for the μSAM geometry (2). A complete simulation can be done by bringing the LE models together (3).
Furthermore all relevant constraints for the simulation had been determined. This includes the availability for piezoelectric materials and materials for the bonding processes. Additionally for the design, relevant parameters can be defined. These are the resonant frequency of the piezoelectric transducer as well as the representative parameters acoustic compliance, acoustic mass and the electrical capacitance as well as the transducer coefficient. All these parameters can be influenced by the material choice and the geometrical dimensioning. Some of them are preset, such as the thickness of the piezoelectric material and the bonding layer by the manufacturers.
2.1 Combined Simulation of the electromechanical sub-system
It could be shown, that it is feasible to use the Finite Element Method to simulate the behavior of the electromechanical subsystem (piezoelectric membrane transducer), because for the aimed geometry no analytical expressions are available and a derivation would be extremely time consuming. Therefore FE model of the μSAM actuator was prepared. The mechanical influence of the metal electrodes on electromechanical behaviour can be neglected because of the small electrode thickness. Hence coupled electrical degrees of freedom were used on top and on the bottom of the piezoelectric material. Standard partly anisotropic material parameters of the manufacturers (piezoelectric material, bonding intermediate layer, silicon wafers) were used.
The preliminary tests have been done to proof the different concepts and their possibilities in the manufacturing of the μSJA parts (membrane thickness of 70 µm and 140 μm, etching of the silicon nozzle with the typical pitch of approximately 54 ° ) and the bonding techniques of the PZT cermic to silicon bonding step. FE-model results can be compared with Laser-Doppler-Vibrometer measurement results. Concerning 610-007-4 B4 a maximum deflection in resonance (Q=80) of 1.78 μm could be determined via FEM, the measured value is 1.64 μm. A difference of the resonance frequency can be seen between measurement 10.3 kHz and simulation 12.9 kHz. A quasi-static deflection of 30 nm was measured compared to 42.8 nm as a predicted value.
For understanding the motion of the transducer, FE-simulations are helpful. Different constraints, at first the mounting condition of the device were discussed.
It can be seen, that the clamping of the full frame is a realistic boundary condition for the assembly in use. The damping under standard conditions can be estimated and should be comparable with standard SJAs. Therefore a Q-factor between 1 and 100 can be achieved. We assume 10 and 80 corresponding to different measurement setup with different clamping conditions. Corresponding to the preliminary tests, a prediction was made for the optimized designs 610-028-2 and 610-028-6.
As a conclusion of the preliminary tests the following aspects can be named to describe differences between FE-simulations and measurements:
1)Induced mechanical stress while bonding procedure (different coefficients of thermal expansion (CTE) of the piezoelectric material, the silicon and the intermediate bond layer lead to mechanical stress as a result of temperature and pressure treatment).
2)Variations in the process (e. g. notching was determined) leads to the point, that geometries might be slightly different.
3)Variations of the used piezoelectric material (PIC 255) properties can vary max. +/- 20%.
4)Behaviour of the intermediate bond layer is assumed as linear elastic.
5)Failures caused by the measurement setup (clamping of the chip, wire bond and soldering contact).
2.2 Lumped Element Simulation of the acoustic sub-system
For modelling the acoustic sub-system, lumped elements were used. This is a time efficient concept because of its high degree of abstraction. However, for complex distributed geometries it is hard to derive exact models, but for the aimed μSAM geometries analytical acoustic expressions partially exist [12]. Hence the model [13] was adapted with aimed geometries. Unfortunately in the TANG model fluid mechanical expressions were considered, which can only be used for simple axial symmetrically geometries. Within the framework of μSAM it is impossible to derive fluid mechanical models of the μSAM geometry because of the complex mathematical descriptions. In conclusion an acoustic sub-model was set up using the classical theory of linear acoustics and has been extended for the geometries like the truncated cone and truncated pyramid. Due to the different orifice geometries, which should be investigated, two cases have been taken into account - the rectangular and the circular cross-section with edge length a. In case of the μSAM geometry extensions for wider frequency range were made because of the small dimensions, for details see [12]. Further extensions have to be made to extend the model considering the truncated shape of the orifice due to KOH-processes generating a 54.7° etch profile and laser-ablation used to generate the cylindrical and truncated cone shapes. The dimensions rcyl = rtc = b = a/2 = 50 μm / 100μm / 150 μm / 250 μm. This is a discretisation and summation of calculated lumped values and gives more accurate results than averaging over the inlet and outlet area.
2.3 Combined Simulation of the μSJA
At this point results for the different orifice shapes and the two electromechanical systems are shown.
As an interpretation the following aspects are relevant:
-A smaller orifice leads to a higher exit velocity in electromechanical resonance
-A larger orifice shows a higher exit velocity in the acoustic resonance
-610-028-6 is stiffer, has a higher resonance frequency but also a lower exit velocity in electromechanical resonance than 610-028-2
-Truncated cone and truncated pyramid show similar behaviour, both are more efficient than the simple cylinder
-Electromechanical and acoustic subsystem are coupled (position of the resonance frequencies influence each other)
-610-028-2 with 100 μm outlet (a/2 = 50 μm) truncated pyramid can be determined as an optimal design of this study
Further acoustic aspects are important, i. e. the eigenfrequencies of the cavity, which are in the range of the working frequency. To calculate these, in a first step the truncated pyramid shape of the cavity can be approximated as a cuboid.
In consequence, the acoustic modes of the cavity will influence the frequency response (shift of the resonance frequency of the electromechanical transducer) and causes differences in the amplitudes. Especially the mode at 27.3 kHz could be critical. Finally the simulation results are compared to measurement results of the test chip B4 of the 610-028-2 sample. It can be seen, that the damping has to be improved, but this could only be done by using fluid mechanics or CFD simulations. Regarding the short timeframe of the project μSAM, fluid mechanics or CFD simulations cant be realized. Finally, the Combined Simulation models are ready to be studied in interaction with an electronic power circuit. It could be shown, that the created models can be verified by measurement results.
3. Manufacturing of the µSJAs
In this chapter the manufacturing of the components is presented. The decision for the design concept has been explained in chapter '2 Design of the μSJA'. Based on that design the silicon manufacturing concept of the whole actuator is developed and presented in '3.1 Dimensions and geometry of the μSAM-μSJA'. The results of these manufacturing steps are presented in '3.2 Manufacturing of the silicon membranes' and in '3.3 Manufacturing of the silicon and laser micro-machined nozzle configurations'. In chapter '3.4 Characterization of the PZT ceramics' the properties of the piezoelectric ceramics are presented. Finally the membrane part, the nozzle part and the piezoelectric ceramic are ready for integration.
3.1 Dimensions and geometry of the μSAM-μSJA
The presented μSJA design consists of two KOH-etched silicon wafers in order to form a cavity. This cavity and an etched orifice build up the acoustic subsystem. A bonded bulk piezoelectric transducer at the opposite wafer is used to drive the electromechanical subsystem. For this purpose a new bonding technology has been obtained to assemble the transducer membrane of the μSJAs with standard bulk PZT ceramics.
Different investigations have been done using the Combined Simulation to predict the deflection of the piezoelectric transducer to find an advantageous geometric arrangement in order to increase output parameters like the exit velocity.
3.2 Manufacturing of the silicon membranes
The manufacturing of the membrane-part is defined by a KOH-etching step into the silicon substrate according the required membrane thickness.
The diaphragm-wafer's backside is wet etched in order to produce a membrane in different thicknesses. In preliminary tests the thickness of 40 μm and 70 μm was chosen to evaluate the overall manufacturing and bonding concept that showed good results. Therefore, in the finally manufacturing based on the results of the work package 2 'Design' the three membrane thicknesses 100 μm, 200 μm and 300 μm are aimed to test different types of the electromechanical subsystem.
Time-dependent wet etching without any defined etch stop causes notching due to the inhomogeneous concentration of the etchant, which appears especially in huge lateral and transversal etching dimensions.
This leads to a warping of the silicon membrane, depending from the etching depth. The 100 μm membranes results in a warping of 21 μm (largest etch depth), the 200 μm membranes in a warping of 14 μm and the 300 μm membranes in a warping of 10 μm. The measurements of the membrane thickness have been done by an optical membrane thickness measurement sensor (BMT). Measurement was carried out in the middle of the membrane of each chip and then the values were averaged. The characterization of the warping of the membranes has been performed by a withe light interferometer (Zygo NewView 6200).
This non-planar profile must be taken into account when investigating different bonding methods for mounting the PZT ceramics on the backside of the membrane.
The etch process has been split in two steps for the 100 μm membranes in order to reduce the strain caused by the masking layerstack. The strain of this layerstack would lead to a deflection of the 100 μm silicon membrane.
The Si surface is directly used for direct bonding with the nozzle wafer. In order to maintain the required surface quality, the masking oxide layer on the bond side is removed by wet etching just prior to bonding.
3.3 Manufacturing of the silicon and laser micro-machined nozzle configurations
The manufacturing of the nozzle-part is characterized by two different micro-machining technologies which are used to produce different nozzle configurations.
One technology is a sequential etching process with the same (KOH) etchant, which is used for the manufacturing of the silicon membranes.
For this the orifice will be preliminary etched with a depth of 300 μm. Afterwards, part of the masking layer protecting the cavity is removed. The second step is the etching of the orifice and the production of the cavity geometry. The etch process is continued until perforation. The result of this procedure is the typical pyramid structure with a pitch of approximately 54 °, due to the anisotropic etch behavior related to the crystal orientation. In contrast to the truncated pyramid structure of the KOH-etching nozzles two other nozzle structures that are manufactured by laser micro-machining are investigated. The first laser micromachined nozzle structure corresponds to a cylinder with a pitch of approximately 90 °, the second laser micro-machined nozzle structure corresponds a truncated cone with a pitch of approximately 54 °.These three nozzle structures are manufactured with the equal geometrical dimensions of the cavity and the four different nozzle exit geometries: 100μm, 200μm, 300μm and 500μm concerning the nozzle outlet.
The achievements of these processes are 3D nozzle structures of the orifice in a required aspect ratio and a very small cavity, resulting in a higher pressure gradient which leads to a much higher exit jet velocity compared to a dry anisotropic etched orifice combined with a huge cavity. This creates the main advantage compared to known layouts. Now the two parts (membrane and cavity) are ready for bonding to form a small cavity with an orifice on the one side and a diaphragm on the opposite side. Finally, the piezoelectric ceramic disc is mounted on the back side of the diaphragm into the wet etched cavity.
3.4. Characterization of the PZT ceramics
Measurements of the piezoelectric ceramics have been done concerning the warping of the ceramics and their roughness. It can be shown that the piezoelectric ceramic itself has a not negligible warping and roughness. This warping (in the range of -17 μm up to +10 μm) and roughness (in the range of 0,47 μm up to 0,60 μm) of the piezoelectric ceramic itself and the warping of the KOH-etched silicon membrane can limit the bonding technology of the piezoelectric ceramic bonding.
4. Bonding
In this chapter the bonding concepts as well as the different bonding techniques are presented. Furthermore, investigations of the different bonding techniques and further characterizations of the proposed bonding for the use in the μSJA are described.
4.1 Bonding concept
There are two options for the integration concept: the PZT ceramic-to-wafer bonding first- and the wafer-to-wafer bonding first concept. The challenges of the whole fabrication process are the bonding steps which are needed to convert the SJA concept into a μSJA concept using standard bulk PZT ceramics.
4.1.1 PZT ceramic-to-wafer bonding first
For this bonding approach, dies of the PTZ ceramic need to be bonded to the rough and warped backside of the membrane wafer. Furthermore, the PZT actuators themselves show averagely a certain roughness and a warping. In addition, due to the CURIE temperature of the PZT material, which must not be exceeded, the temperature budget for the bonding process and all subsequent processes is limited. Thus, a low temperature adhesive bonding method with a relatively thick adhesion layer in order to overcome the warp and roughness problems can be chosen. As the membrane can be mechanically supported, there are no pressure limitations. This leads to the opportunity of using higher bonding pressures and results in a higher bonding strength. For the subsequent wafer-to-wafer bonding step only low temperature direct bonding- or adhesive bonding techniques are suitable.
4.1.2 Wafer-to-wafer bonding first
The alternative is the wafer-to-wafer bonding followed by the PZT ceramic bonding step. Due to the fact that the wafer-to-wafer bonding concept followed by the PZT ceramic bonding step shows more benefits compared to the other concept it was finally chosen. Therefore, the wafer-to-wafer bonding step is not limited to temperature regimes or process restrictions. This means that high temperature direct bonding techniques can be used. However, now the PZT ceramic-to-membrane bonding step faces some limitations of the bonding process. The cracking of the membrane due to a large deflection must be avoided. This is tested with cross section- and shear tests with bonded test chips, with different bonding pressures.
4.2 PZT ceramic bonding techniques
Several bonding techniques were tested due to their ability for the PZT ceramic-to-membrane bonding step. Therefore test wafers were used for these preliminary tests. The main focus is to realize the bonding of the rough and warped PZT ceramic to the silicon membrane. Furthermore, the bonding techniques should be able to realize the electrical connection to the lower electrode of the PZT ceramic, because this will reduce technology steps, like deposition, structuring and connection steps. The used bonding techniques are detailed described in the followed subchapters.
4.2.1 Bonding with the negative photoresist SU8
In these tests, the negative photoresist SU8 is structured in 10 mm x 10 mm squares in different thicknesses (5 μm, 10 μm and 20 μm) to overcome the roughness and the warping of the PZT ceramics. But during the bonding step of twelve PZT ceramics to the silicon by using the SU8 in the three different thicknesses, it was not able to compensate the roughness and warping. Furthermore, the test with the thickness of 20μm showed that the surface tension of the SU8 is too high. This surface tension results in removed SU8 structures from the test wafers.
4.2.2 Bonding with adhesive tape Nitto Denko 5601
The adhesive tape 5601 from Nitto Denko is a double side adhesive tape which can be structured by the use of laser micro-machining. However, the bonding tests with this adhesive tape showed no bonding between the PZT ceramics and the silicon membranes. Reasons are the roughness and the warping of the PZT ceramic - and that the adhesive tape with a thickness of 10 μm is not able to compensate this.
4.2.3 Bonding with Loctite electrically conductive adhesive 3880
To test the opportunity of electrically conductive adhesives the Loctite 3880 was chosen. This is one component electrically conductive epoxy which showed good bonding results. Nevertheless disadvantages such as reproducibility especially in the layer thickness and the overall filling, limit the applicability of this bonding techniques for bulk PZT ceramic bonding.
4.2.4 Bonding with reactive multilayer systems
The stacked multilayer, each layer with a thickness in the range of a few nanometers, provides a high energy storage which will be released during the bonding process. A raise in temperature, the application of a slight electronic current or even a laser pulse are able to ignite the self-propagating exothermic reaction. Although the temperature during the reaction is high, the surrounding components are not affected despite of a few lattice planes due to the ultra fast propagation. This implies a bonding process at nearly room temperature. The two different PZT ceramics were bonded against a complete with gold sputtered 4 inch silicon wafer. The preliminary investigation was done by using a razor blade to test the bonding strength between the PZT ceramic, the reactive multilayer system and the golden silicon wafer. It could be shown that there is a really huge bonding strength between the bonding partners. It was possible to destroy the PZT ceramic instead of the bond of the PZT ceramic electrode and the golden silicon wafer.
The used reactive multilayer system is 40 μm thick and consists of alternating layers of aluminium and nickel. In bond test this bonding technique shows only very good results if the bonding partners were pressed together during the bonding step with a high bonding pressure. Otherwise the bonding is characterized by enormous unbounded areas and voids. Referred to the usage in μSJAs with their thin silicon membrane this bonding technique seems not be feasible to fulfill the requirement of smaller bonding pressures. The cracking of the membrane has to be avoided.
4.2.5 Bonding with electrically conductive adhesive foils
The used conductive layer is an electrically conductive adhesive foil which has a thickness of 35 μm and showed excellent bonding results in first preliminary bonding tests. With this bonding foil it is possible to use lower bonding temperatures and bonding pressures. This foil can be structured e.g. by the use of a laser and is able to compensate the roughness and the warping of the PZT ceramic. Therefore, this adhesive foil is used to bond the PZT ceramics to the thin silicon membranes after the wafer-to-wafer bonding step.
4.3 Characterization and Optimization of the PZT-ceramic to silicon wafer bonding technique
Different characterization steps are performed to verify the quality of the bonding. Especially for bonding of PZT ceramics on thin membranes, a minimum of bonding pressure is allowed in order to ensure a crack free assembly. To evaluate the required bonding force which results in a high quality bonding, bulk PZT ceramic-to-membrane bonding tests with different bonding forces have been done. The characterization of the bonding strength is realized for a first estimation by yield tests and for a quality evaluation by shear tests of bonded bulk PZT ceramic-interlayer-membrane test chips with different bonding pressures. Three batches with bonding pressures of 2 bar, 1 bar, 0,5 bar, 6E- 3 bar and 1,5E-4 bar (the weight of the bulk PZT ceramic) have been chosen. These five types are equally bonded together and have been investigated according to their bonding strength. The 10x10 mm² test chips of the first batch have been cut in 1x1 mm² pieces to analyse the bonding yield after dicing for a first strength estimation between the interlayer and the bulk PZT ceramic. The test chips of the second batch were cut into cross-section sheets for the characterization of the wetting to the different materials and the interlayer. The test chips of the third batch have been cut in 4x4 mm² pieces to analyse the shear strength.
4.3.1 The first batch: 1x1mm² - bonding yield after bonding
The 1x1 mm² diced test chips affirmed the assumption, that a low bonding pressure (1.5E-4 bar = bulk PZT ceramic itself) results in a poor strength, compared to a high bonding pressure (1 bar and 2 bar). The bonding yield with regard to the different presuures and strengths is between 0% (1,5E-4 bar) and 100% (2 bar). Furthermore, a very good yield of 80% can be guaranteed already with a bonding pressure of 0,5 bar. This fact is especially for the μSJAs very positive because this lower bonding pressure leads to a lower membrane stress and to a lower deflection of the membrane during the bonding process. These results validate the feasibility of the new bonding technology especially for the usage on thin membranes.
4.3.2 The second batch: cross-section
The cross-section test chips have been evaluated regarding cracks, blowholes, voids and of course the homogeneity of the interlayer with respect to the whole cross-section. The reason for the poor strength of the '1,5E-4 bar' and the '6E-3 bar' test chips results are due to the enormous number of unbonded areas and voids . An example of the cross-section of a test chip which is bonded with a higher bonding force (2 bar).
4.3.3 The third batch: shear strength test
The shear tests have been done with the shear strength test tool (TIRAtest 2805). Due to the fact, that the 1,5E-4 bar and 6E-3 bar test chips do not show such good results in the yield- and cross-section tests compared to the other test chips, the 1,5E-4 bar and 6E- 3 bar test chips have not been characterized with the shear strength test. The results of the shear strengths test confirmed the 1 mm² yield tests. There is no huge variation between the average shear strengths of the 0,5 bar (30,.2 MPa) to the 1 bar (30,9 MPa) and 2 bar (29,9 MPa) bonded test chips. The shearing took place in the conductive polymeric interlayer system.
The alignment of the whole bulk PZT ceramic-interlayer-membrane stack, as well as the interlayer (with respect to the streamed interlayer material during the joining process and the different bonding pressures) has been analyzed. It could be shown, that only a little amount of the conductive polymeric has been streamed out. A high positioning accuracy could be reached by using the flip chip bonding tool for the assembly of the joining partners.
4.4 Bonding of the processed wafers to µSJAs
The final approach is characterized by the wafer bonding first concept that is a direct bonding step. Afterwards the bonded μSJA system is diced into chips and the PZT ceramics are bonded on the thin silicon membranes of the μSJAs. This is realized with the electrically conductive adhesive foil with a bonding temperature of 130°C and a bonding pressure of 0,5 bar. The link to the overall system is realized by the integration of the μSJAs into the system and by wire bonding.
4.4.1 Wafer bonding
Due to the used PZT ceramic bonding technology the wafer-to-wafer bonding (silicon fusion bonding with a special plasma pre-treatment of the bonding areas) is the first bonding step, followed by the PZT ceramic-to-membrane bonding. This silicon fusion bonding is done at room temperature followed by the temperature step of 200°C for 2 hours.
After the silicon fusion bonding of the membrane- to the nozzle-wafer, the bonded wafer pair is sputtered with chrome and gold. These layers will realize the electrode for the connection of the lower PZT ceramic electrode. The electrode and an isolation pad for the electrically connection are manufactured into the chrome-gold electrode by laser micro-machining. The dicing of the wafer to single chips is the last manufacturing step in the manufacturing of the μSJA chips.
4.4.2 PZT bonding
The following steps concerning the integration of the PZT ceramics into the μSJA chips, as well as the electrical connection of the PZT ceramics. Therefore the bonding of the PZT ceramic-to-the membrane with the bonding interlayer is the first integration step. This bonding interlayer is structured by the use of the Laser in 10 mm x 10 mm pieces, which where aligned in the middle of every μSJA chip by the use of a flip chip bonding tool (Tresky T-3002-FC3). On this bonding interlayer the PZT ceramic itself is aligned with this flip chip bonding tool as well. The PZT ceramic-to-membrane bonding step is realized in a pressure and temperature depending process, where the PZT ceramics are bonded on the silicon membranes at 130°C and 0,5 bar (according to the best result of the pretests, described in 4.3 Characterization and Optimization of the PZT-ceramic to silicon wafer bonding technique). Finally the μSJA is ready for integration.
4.4.3 Wire bonding
For the electrical connection of the upper electrode and the electrical connection to the actuator electronics bond wires are used and the μSJA chips are integrated in devices for the characterization and measurement of parameters like the exit velocity of each single μSJA chip.
5 Integration and characterization
The development of actuators for active flow control based on micro technologies also includes the integration of the actuators in a panel for characterization including the necessary electronics to drive and test the actuators. The activities performed for integration and characterization of the actuators are summarized in the following chapter.
5.1 Actuator characterization
For the characterization the electric behaviour of the μSJAs, the electric behaviour was evaluated with the impedance and phase. The AC voltage and the AC current were measured over a determined frequency range. The characterization of the performance of the actuators was done by measuring the exit velocity of the actuators depending on the actuation frequency.
5.1.1 Measurement Equipment
The measurements were performed in a special soundproofed characterization chamber. An automated measurement and storing of the measurement data was realized in LabVIEW™.
It was designed to control the actuators and automatically measure the parameters using and controlling specific equipment:
- Agilent E4980A Precision LCR Meter
- LabVIEW™ Run-Time Engine 2010
- Hameg HMF2525 Arbitrary Function Generator
- TEGAM High-Voltage Amplifier 2350
- PicoScope 4227
- DantecDynamics Hot wire Probe
- MiniCTA for Hot wire probes
- Tektronix TDS3054 oscilloscope
With the automated measurement tool in LabVIEW™ impedance, phase, AC voltage and - current of the actuators in a determined frequency range was measured. With preliminary tests the frequency range was detected.
On the upper side the device is selected via VISA Component. A delay time could be added to delay each frequency step for error detection on the LCR. The signal type and level are adjustable. A frequency sweep is configured via start and stop frequency and the increment.
In the middle of the GUI the output parameters are presented during measurement: Impedance (Ohm), phase (degree), Vac (V) and Iac (A). Measurement data is stored in a file. The location of the file is displayed on the right side. The graphs below the output parameters show the output parameters over the frequency range.
5.1.2 Impedance and phase of the actuators, Voltage and current
Basic measurement parameters are
- Sinusoidal actuation
- Frequency range μSJAs: 50Hz-20kHz
- Frequency range μSJAs: 50Hz-20kHz
5.1.3 Capacitance
The capacitance of the μSJA is measured with charging and discharging behaviour. The measurement is described in the next part. A constant DC source charges the DUT over R. After the charging process is finished, the switch S connects the resistance to GND and the capacitor uncharged. VDC and S had been realized by the function generator Hameg HMF2525. To drive the DUT the high voltage amplifier TEGAM 2350 was used. The amplifier has an amplification factor of 50.The output signal type of the function generator was set to rectangular waveform. The offset was adjusted until the low pulse of the output signal on the amplifier reaches GND level. In this way the signal on the DUT is only in the positive signal space and complete discharging of the device could be reached. The high state of the function generator output conducts VDC with R. This state represents charging process of the DUT. The low state of the function generator represents discharging of the DUT due to connection of R with Ground. The output frequency was set to 100 Hz for time intervals corresponding to complete charging and discharging process. The resistor R was measured with the FLUKE 187 (R=50,91 kΩ). The Voltage over the DUT was captured with the PicoScope 4227.
For the measurement of the capacitance of the SJA the charging and discharging behaviour was measured. The time constant t is measured by using the tangent line in the starting point of charging and discharging. With this time constant t and the resistor R the capacitance can be calculated.
Measurements with the μSJA were performed using the following input parameters:
- Input amplitude (HMF2525): 200 mV
- Amplitude on μSJA: 10 V
- Offset: 112 mV (GND)
- Pulse width: 5 ms
- Frequency: 100 Hz
5.1.4 Behaviour of the electromechanical subsystem
Two μSJA-systems with different membrane thicknesses have been measured. The system #1 '610-28-6+610-29-5' with the 200μm membrane and the system #2 '610-28-2+610-29-1' with the 300μm membrane have been compared. Both μSJA-systems are realized with four different nozzle configurations (100μm x 100μm, 200μm x 200μm, 300μm x 300μm and 500μm x 500μm). The manufacturing and the bonding of the single devices have been done in the same way. The membrane deflection per volt of every single μSJA has been analyzed with a Laser Doppler Interferometer.
5.1.4.1 The 200μm membrane μSJA-system
The membrane deflection per volt and the resonance frequency of the electromechanical subsystem of the system #1 '610-28-6+610-29-5' varies slightly between 9 to 12kHz and 4E-7m/V to 9.5E-7m/V.
5.1.4.2 The 300μm membrane μSJA-system
The membrane deflection per volt and the resonance frequency of the electromechanical subsystem of the system #2 '610-28-2+610-29-1' can be divided into two parts. The chips with the 100μm and 200μm nozzle configuration result in lower resonance frequencies in the range of 9.5 to 11.5kHz and in membrane deflections of 2E-7m/V. The chips with the 300μm x 300μm and the 500μm x 500μm nozzle configuration result in higher resonance frequencies between 14 to 17kHz and in membrane deflections in the range of 1.8E-7 to 4.5E-7m/V. This variation can be explained due to the back coupling of the different nozzle configurations to the 300μm electromechanical subsystem.
5.1.5 Velocity measurement of the mounted μSJAs
The performance of the whole system #1 '610-28-6+610-29-5' with the 200μm membrane and the system #2 '610-28-2+610-29-1' with the 300μm membrane have been characterized by measuring the exit jet velocity. This has been done with a special clamping of the μSJAs and a hot wire velocity sensor.
The measurement results show that exit velocities in the range of up to 80 m/s are possible for the μ-technology based SJAs. This result could be shown using a μSJA with 100 μm membrane thickness, a 500 μm x 500 μm KOH-nozzle configuration and an actuation voltage of 100V.
5.2 System Integration
In previous tests, the single actuators have been mounted onto single mounting parts. That means one actuator for one mounting support. To use the μSJAs in wind tunnel tests, it is necessary to mount the actuators in a single row and to create an actuator surface that match the shaped surface of the wing configuration. Therefore a single mounting support, called panel was designed. It is possible to adapt such a panel for different wing configurations and to use the panel for a variable number of single actuators.
In this design the panel includes five single actuators which are located in the centre of the panel that has a total length of 492 mm. The exit holes are located at the top surface and they lead to an outlet slot. A frame on the inner side of the panel is used to fix the orientation of the actuators. The required width for a single actuator frame is 20.2 mm, so it would be possible to design a panel with much more actuators. For this first demonstrator, five actuators are arranged in a straight line.
Two screws are necessary to mount this part. Each actuator is fixed by one holder which allows the separate mounting of the actuators.
To mount the whole panel onto a wing model, the panel is designed with two holes at the endings of the panel. For this, just two screws are necessary to fix the whole panel. In addition, the screws are covered by a curved plate to ensure a surface without any flow influencing structures.
5.3 Control Electronics
In WP5 the goal was to design control electronics for the μSJAs. With reference to the measured resonance frequency in deliverable 4 a signal generator was designed.
The requirements of the control electronic for the μSJAs are
- frequency tuneable around resonance frequency
- sinusoidal actuation
- output amplitude: 2 Vpp
The output signal of this frequency generator has to be transformed to high voltage signals with a high voltage amplifier. Due to the lack of integrable high voltage amplifiers the TEGAM 2350, which was also used for characterization, had to be used for electronic testing. The high voltage output of the amplifier is distributed to the μSJAs with a Distribution box.
5.3.1 Actuation frequency generation
In this part the frequency generator is described. An integrated circuit, the XR2206 is used on the PCB layout for frequency generation. There are 3 frequency ranges which can be adjusted with a Jumper on the PCB. The frequency can be adjusted according to the resonance frequency of the μSJAs with knobs for coarse (left) and fine (middle) adjusting. The amplitude can be regulated with the right knob. On the PCB with two trimmer resistors it is possible to change the symmetry output of the sinusoidal signal output.
Features:
- Frequency range: 1..100000 Hz in 3 ranges
- Voltage Input: 12..20 V
- Amplitude regulation
- Enable switch
As input a voltage of 12..20 V has to be used. The integrated circuit 7812 regulates the input voltage to 12 V. Trimmer resistors are used for symmetric output adjustment. The capacitors 10nF, 220nF and 10μF set the frequency range. TL074P connected as a voltage follower moves the output Output No: 007 - Version: 001 Date: 29.11.2012 Page 55 of 59 level with capacitor C8 to swing symmetrically around GND. The output can be enabled and disabled with a switch. Output connector is a BNC-Connector.
5.3.2 Signal distribution
The signal provided by a high voltage amplifier has to be distributed to 10 actuators. With the distribution box the 2 poles of the input signal is wired to 10 actuators each pole. On the output of the box a SUB-D connector connects the wires for the actuators. The SUB-D output pins are isolated for high voltage safety requirements. The cabling is built in a Fibox housing with two TS35 rails.
6 Conclusion
The aimed objective of this project, the design, manufacturing and testing of microtechnology based synthetic jet actuators, could be reached. These µSJAs have been manufactured by microtechnology at the clean rooms of the ZfM. Furthermore a novel bonding technology for piezoelectric ceramics has been investigated. Finally, tests and characterizations of the manufactured μSJAs show their ability for generation of increased exit velocity and hence they can be applied in active flow control.
A novel bond technology regarding very low temperature bonding of bulk PZT ceramics to thin silicon membranes has been investigated successfully. The conductive polymeric interlayer system has been characterized and evaluated for a first estimation by yield tests and cross-section tests. For a quality evaluation bonded bulk PZT ceramic-interlayer-membrane test chips have been characterized by shear tests. A good compromise between a high bonding yield and a low bonding pressure has been achieved at a bonding pressure of 0.5 bar. Furthermore, this bonding pressure results in a minimal stress distribution.
First measurement results of samples and the characterization of the electromechanical transducer are presented and the results show a good fit to the simulation. Therefore, the described joining concept represents a promising concept especially for transducers based on thin silicon membranes but also for other applications like cantilever actuators. Due to the usage of standard bulk PZT ceramics, a bonding process for MEMS has been established without the need of special pre-treatments. Finally a complex and integrated manufacturing process including different materials for smart systems integration has been developed.
After the development of the technology for the manufacturing of the micro synthetic jet actuator a system to integrate the actuators was designed and manufactured.
The first tests of the actuators showed that the μSJA performance of the system with the 200μm membrane reaches maximum exit jet velocities of about 28m/s at resonance frequencies of about 18.3kHz with the 500μm x 500μm nozzle configuration and an actuation voltage of 100V. The μSJA performance of the system with the 300μm membrane reaches maximum exit jet velocities of about 38m/s at resonance frequencies of about 21.2kHz with the 500μm x 500μm nozzle configuration and an actuation voltage of 100V. Therefore, the 300μm membrane system with the 500μm x 500μm nozzle configuration is preferred because of its higher resonance frequency and the higher exit jet velocities compared to the other results. Nevertheless the variation of the results depends on different points such as the manufacturing challenges (notching of the membranes, roughness of the membranes and the PZT ceramics, ) as well as geometrical restrictions (bonding interlay, variation of the PZT ceramics).
In a final test of the second manufacturing batch exit jet velocities in the range of up to 80 m/s could be reached for the microtechnology based SJAs. This result could be shown using a μSJA with 100 μm membrane thickness, a 500 μm x 500 μm KHO-nozzle configuration and an actuation voltage of 100V.
The successful manufacturing process and the satisfying measurement results show that it is possible to design and manufacture synthetic jet actuators based on micro-technology. All the manufacturing steps could be realized and presented and the measured exit velocity are in the range of expectation.
7 References
[1] C. Lee et al., Sensors and Actuators A 108 (2003) pp 168-174
[2] S.G. Mallinson et al,., Sensors and Actuators A 105 (2003) pp 229-236
[3] L. Gomes et al.: Towards a Pratical SJA for Industrial Scale Flow Control Applications pp.1-8
[4] P. Mane et al., Journal of Intelligent Material Systems and Structures 18, 2007, pp 1175-1190
[5] M. Watson et al., The Aeronatical Journal (2007), Paper No. 3160
[6] Q. Gallas, Thesis, 2002
[7] V. Tesar, J. Kordk., Sensors and Actuators A : Physical (2008), 1-11
[8] Z. Travnicek et al., Sensors and Actuators A 120 (2005) pp 232-240
[9] D. J. Coe et al, Sensors and Actuators A 132 (2006) pp 689-700
[10]Y. Liang et al., Sensors and Actuators A 125 (2006) pp 512-518
[11]A. C. Miller, Thesis (2004)
[12]Lenk, A.; et al.: Electromechanical Systems in Microtechnology and Mechatronics; Springer-Verlag Berlin Heidelberg (2011), ISBN 978-3-642-10805-1
[13]Tang, H.; Zhong, S.: Lumped element modelling of synthetic jet actuators. In: Aerospace Science and Technology 13 (2009), pp. 331-339
[14]Starke, E.: Kombinierte Simulation - eine weitere Methode zur Optimierung elektromechanischer Systeme. Dresden, Technische Universität Dresden, Dissertation, 2009.
Potential Impact:
μSJA's concepts are new and show a large potential for active flow control applications. High exit velocities are possible with a lower energy consumption compared with standard SJA's. A large impact can be observed in the field of simulation. For μSJA's no models are available. This situation can be improved with μSAM. Another fact is, that with the help of micro technologies, well defined SJA's are producible.
Silicon based µSJAs have been manufactured by microtechnology at the clean rooms of the ZfM by a novel bonding technology for piezoelectric ceramics. This novel bonding technology will cause a large impact on the implementation of piezoelectric ceramics in MEMS. Advantages of these new technologies are strong bonds with only a small amount of temperature inducted mechanical stress. Thus, the Curie temperature is not exceeded and there is no depolarisation of the piezoelectric ceramics.
On the scientific field, the integration of pre-poled bulk PZT is expected to lead into new and innovative applications. Particularly in the areas of acoustical and optical microsystems as well as motion sensors, MEMS have great market potentials. The project partner's competencies are able to cover the required supply chain from design of application-relevant microsystems with sensor-actuator coupling to fabrication of prototypes to generate high potential for success. For the demonstration of the feasibility of the 'hybrid technology', realization of test structures has a crucial role and is considered as an essential step to the conversion into economically interesting products in cooperation with companies. The new composite technology leads to specific application aspects for the user through having:
- a cost advantage due to parallel processing and reduced production costs
- a time advantage due to faster bonding processes
- higher quality, reliability and longer lifetime due to lower joining temperature
- an improved functionality
The ZfM is part of the Chemnitz University of Technology and, therefore, a non-profit organisation. The ZfM is interested in utilizing the technology knowledge and expanding this know-how for further research and development projects. The dissemination was performed and will be further performed trough publications in technical and scientific papers in international journals and presentations of the results at leading European and international conferences. We also aim to exploit the knowledge gained in other regional, national and EU projects. This process is ongoing but not yet finished. After the end of the project, the results could be exploited for other related industry, such as automotive, electronics (e.g. by active cooling), printing technologies, micro acoustics and ultrasonic applications by further research and development projects.
Due to the fact, that the manufacturing process was successful and the measurement results showed that it is possible to design and manufacture synthetic jet actuators based on microtechnology, industrial partner already showed their interest in several aspect of the manufacturing process as well as in the use of the μSJAs for other applications.
List of Websites:
http://www.zfm.tu-chemnitz.de
In this report the design, manufacturing and characterization of the micro-technology based synthetic jet actuator is described. In the design and optimization work package a Combined Simulation μSJA-model was created. Therefore, the functionality was virtually split into two sub-systems. The Finite Element Method (FEM) was used to determine relevant lumped parameters of a piezoelectric transducer - electromechanical sub-system, which can be used with the lumped element method (LEM). Furthermore, lumped element models for the acoustic sub-system already exist. These were adapted for the μSAM geometry and a complete electro-acoustic simulation was done by bringing together and solving the LE models. Relevant constraints and parameters for the simulation had been defined and determined. Further acoustic investigations identified possible failure sources due to additional eigenmodes. Other failures sources were discussed. The simulation results were compared with optical deflection measurement results of the piezoelectric transducer and with flow measurements of the μSJA. Out of the preliminary simulation results an optimal design had been chosen.
The wafer preparation and wafer bonding work packages manufactured the whole μSJA chip. This includes the silicon manufacturing of the nozzle- and the membrane wafers, the bonding of the membrane- and the nozzle wafers to the µSJA system, the dicing of these bonded μSJA wafers to μSJA chips as well as the bonding of the piezoelectric transducer, (PZT ceramics) to the diced μSJA chips. The most important challenge of the μSJA manufacturing is the integration of the PZT ceramic into the μSJA system because this faces some limitations concerning the whole manufacturing- as well as the integration concept. Therefore two wafer bonding concepts as well as different PZT ceramic to silicon bonding concepts were investigated regarding their feasibility to fulfill the aimed requirements. The wafer bonding concept has been split into the 'wafer bonding first, followed by the PZT ceramic bonding' and the 'PZT ceramic bonding first, followed by the wafer bonding' approaches.
The PZT ceramic bonding concept concerning several bonding techniques such as bonding with reactive layers, bonding with the negative photoresist SU8, bonding with epoxies and bonding with different types of adhesive tapes have been investigated. Finally, corresponding the aimed requirements such as a low bonding temperature and a low bonding pressure as well as a low stress input into the whole system especially during the PZT ceramic bonding step one approach could be proven with excellent results. This approach is characterized by the first wafer bonding concept that is a direct bonding step with a special pre-treatment of the bonding areas and a followed temperature step at 200 °C. Afterwards the bonded μSJA system is diced into chips and the PZT ceramics are bonded on the thin silicon membrane of the μSJAs. This is realized with an electrically conductive adhesive foil with a bonding temperature of 130 °C and a bonding pressure of 0,5 bar. The connection to the overall system is realized by the integration of the μSJAs into the system and by wire bonding.
After bonding of the transducers and the wafer, the actuators are tested to evaluate the functionality and to verify the single actuator characteristics. Beside the development and manufacturing of the silicon based actuators the design and the development of an electronics system for the controlling of the actuator are done. The system is divided into signal generator and transducer electronics. The signal generator provides the actuation signal depending on the simulation results and the geometrical data of the actuator. The transducer electronics amplifies this signal to meet the requirements of the actuators power consumption depending on the used actuation principle. To test the whole system including a number of actuators and the driving electronics, the single components are integrated into a mock-up. This mock-up will be system designed in that way that it can be integrated into a wind tunnel model. After finalizing the manufacturing and assembly of all the components of the mock up, the system performance is verified by measuring the output velocities of the actuator. Besides that, test and characterisation of the test board are performed. All the results and manufacturing details are summarized in this final report.
Project Context and Objectives:
1. Introduction
A Synthetic Jet Actuator (SJA) is a low power, highly compact micro fluidic device which has potential application in boundary layer flow control. It is a generator that requires zero mass input and produces non-zero momentum output [1]. It consists of a membrane located on one wall of a small cavity, which has an orifice in another face, typically opposite the membrane. The membrane is forced to oscillate, with fluid being expelled through the orifice as the membrane moves upwards. The flow separates at the edge of the orifice, inducing a vortex ring that moves outwards under its own momentum generating a jet in the ambient flow. When the membrane moves downwards, fluid is entrained into the cavity. If the vortex ring is sufficiently distant from the orifice, it is not influenced by the entrainment of fluid into the cavity [2]. To maximize the exit velocity of the actuator, the SJA is normally driven in resonance frequency of the membrane. There are many types of transducers that can be used in actuators for active flow control, such as piezoelectric [1], [3]-[6], electromagnetic [7]-[9] and shape memory alloys [10].
According to the different types of actuation, several investigations have been done to obtain best geometry dimensions of a SJA for these points of actuation. The quality of a SJA depends on several parameters, such as the diameter of the membrane and the orifice, the height of the cavity, the resonance frequency and the deflection of the membrane. The piezoelectric actuation is used because of its low power consumption, fast response, high reliability and low costs. In [1],[3],[5] SJAs are presented, that are produced with membrane diameters from 15 mm to 40 mm, orifice diameters from 0.5 mm to 1.2 mm and a resonance frequency from 1.2 kHz up to 2.5 kHz. With these parameters they reached a jet stream velocity of up to 130 m/s. Mane et al. [4] used a special piezoelectric configuration - Bimorph and Thunder - with a large membrane diameter of 63.5 mm and orifice diameters of 2 mm (Thunder) and 3.67 mm (Bimorph). In resonance at 90 Hz they achieved a jet stream velocity of 28 m/s. A Lumped Element Model (LEM) for a SJA was developed by Gallas [6], in which he simulated a SJA with piezoelectric actuation in different dimensions at a resonance frequency of 800 Hz and a membrane diameter of 20 mm. With this LEM jet streams velocities up to 60 m/s were evaluated.
Another actuation concept is the electromagnetic transducer. Presented in [7] a conventional loudspeaker is used as transducer for the SJA. A constant maximum velocity of 301 m/s is realized due to a nozzle of 4 mm and a resonance frequency of 50 kHz. Coe et al. [9] developed a special kind of SJA, an electromagnetic actuated SJA with additional pneumatic actuation. A membrane with 89 mm is forced to oscillate by a loudspeaker and generates the synthetic jet. The pneumatic actuator deflects valve structures under the orifices, which are arranged in an array. Thereby the synthetic jet can be directly addressed to a special point up to 20 m/s at a resonance frequency of 20 Hz. Liang et al. [10] designed a SJA based on ferromagnetic shape memory alloy (FSMA) composite with different membrane diameters (64 mm and 72 mm) and a slot like orifice with a dimension of 5 mm x 1 mm. With this FSMA SJA Liang gained jet velocities up to 190 m/s at a resonance frequency of 220 Hz. A very special kind of actuation was invented by Miller [11]. He used a DC piston driven SJA and obtained jet velocities of 124 m/s at a resonance frequency of 100 Hz, at an orifice diameter of 0.4 mm. Until now there is just one publication on μSJA based on silicon manufacturing techniques available. Malinson et al. [2] presented a μSJA which is micro machined in silicon. The top wafer is etched to produce a thin diaphragm with a thickness of 50 μm to 100 μm on which an orifice with a dimension of 4 μm x 4 μm is formed in different types. The bottom wafer is also etched to form the 10 μm to 20 μm thin membrane with a dimension of 200 μm x 200 μm. Two 300 μm silicon wafers were used, which were bonded together to form a cavity with an orifice on one side and an opposing thin diaphragm. For actuation he used a PZT mounted on the thin membrane. Malinson simulated the μSJA with a maximum stream velocity of 4 m/s and than based on the simulation Malinson produced the μSJA.
2 Overall strategy of the work plan
The objective of the project μSAM is to transform the SJA concept in a μSJA that is micro machined in silicon with optimized design based on LEM modelling to increase exit velocity and decreasing weight at the same time. To reach the required output velocities of the μSJA, the chamber and the exit channel or nozzle are optimized using different analytic and numeric methods. Based on the results of the optimization, a design will be investigated and transferred to silicon based structures. The production of the μSJA is performed in the clean rooms of the ZfM and is realized on wafer-level. The two different parts of the μSJA, the cavity and the diaphragm are micro-machined and wafer bonded in order to form the actuator.
To integrate the electro mechanical transducer, in this case the PZT element, into the system, investigations of bonding technologies for the use of PZT silicon bonding are done. Thus, the strength of the low temperature bond of the ceramics, concerning low inducted mechanical stress, is optimized. The objective of the bonding process is the integration of the PZT element in the micromachined silicon substrate. Integration of a mechanically stable bond means, allowing the propagation of the generated deflection to the thin silicon membrane with almost no losses.
After bonding of the transducers and the wafer, the actuators are tested to evaluate the functionality and to verify the single actuator characteristics. Beside the development and manufacturing of the silicon based actuators the design and the development of an electronics system for the controlling of the actuator are done. The system is divided into signal generator and transducer electronics. The signal generator provides the actuation signal depending on the simulation results and the geometrical data of the actuator. The transducer electronics amplifies this signal to meet the requirements of the actuators power consumption depending on the used actuation principle. To test the whole system including a number of actuators and the driving electronics, the single components are integrated into a mock-up. This mock-up will be system designed in that way that it can be integrated into a wind tunnel model. After finalizing the manufacturing and assembly of all the components of the mock up, the system performance is verified by measuring the output velocities of the actuator. Besides that, test and characterization of the test board are performed.
3 Structure of the project
The project is managed by the Project Coordinator (PC) Prof. Dr. Karla Hiller, who is also the leader of the Work Package (WP) 1: Management. The Project Coordinator is the main interface between the Chemnitz University of Technology and the European Commision (EC) as well as the JTI SFWA consortium. She has the overall responsibility for the achievement of the technical and business objectives of the project. Every single work package has been led by a Work Package Leader (WPL). His task is the organisation of the work in the work package as well as the communication with related work packages. The WPLs worked closely together with the PC and the SFWA consortium.
The project is divided in 6 work packages (WP1: Management, WP2: Design, WP3: Wafer preparation, WP4: Wafer bonding, WP5: Electronic design and WP6: Integration and Test), each has been coordinated by one WPL. To report the status of work and the outcome of the different WPs, 6 main deliverables were scheduled. The monitoring of the technical status and progress has been managed by Milestones and quarterly Meetings.
WP1 coordinates the WP interactions in order to ensure the fulfillment of the tasks and milestones as well as the work plan. Therefore, regular meetings with all team members are organised. Outputs in form of external deliverables have been created and submitted to the cleansky consortium. Different fabrication and bond steps have been compared as well as alternative electronic concepts.
WP2 provides the basement of the project. It is fundamental to have a design concept for the fabrication process to ensure the functionality further μSJAs. For this the summarising and investigation of requirements for a μSJA, the supply of an optimised design of a μSJA, an efficient modelling for optimisation purpose and investigations of important constraints and their influence on the system have been done.
In WP3 the enticing masks are designed and fabricated. The wafer processing steps, including the etching (KOH) of Si wafers for the manufacturing of silicon membranes, cavities and nozzles are performed. Furthermore the pretreatment of silicon wafers for wafer bonding and for PZT silicon bonding are realized. Finally the (In situ) measurement of the membrane thickness and the cavity geometry characterizes the achieved results within this WP.
The development and test of novel bonding technologies for PZT ceramics are the main objectives of WP4. Therefore low temperature bonding technologies, the optimization and the characterization of the bonding process as well as the process adaption are the tasks to be performed.
WP5 includes the design and the development of an electronics system for the controlling of the actuator. The system is divided into signal generator and transducer electronics. The signal generator provides the actuation signal depending on the simulation results and the geometrical data of the actuator. The transducer electronics amplifies this signal to meet the requirements of the actuators power consumption depending on the used actuation principle.
Objective of the WP6 is the integration of the developed and manufactured MEMS based SJAs in a mock-up. To test the whole system including a number of actuators and the driving electronics, the single components are integrated into a mock-up and characterized. This mock-up is system constructed that way that it can be integrated into a wind tunnel model.
Project Results:
1. Design of the μSJA
In the design and optimization work package a Combined Simulation μSJA-model was created. Therefore, the functionality was virtually split into two sub-systems. The Finite Element Method (FEM) was used to determine relevant lumped parameters of the piezoelectric transducer (electromechanical sub-system), which can be used with the lumped element method (LEM). It could be shown, that it is feasible to use the Finite Element Method to simulate the behaviour of the electromechanical subsystem, because for the aimed geometry no analytical expressions are available and a derivation would be extremely time consuming. Furthermore, lumped element models for the acoustic sub-system already exist. These were adapted for the μSAM geometry and a complete electro-acoustic simulation was done by bringing together and solving the LE models. This is a time-efficient concept because of its high degree of abstraction. Unfortunately fluid mechanical expression can only be used for simple axial symmetrically geometries.
2. Design and optimization
The preparation of a model for the Combined Simulation of μSJAs was done by splitting the μSJA into sub-systems. In 1a) the Finite Element Method (FEM) has been used to determine relevant lumped parameters of a piezoelectric transducer, which can be used with the lumped element method (LEM) 1b). Lumped element models can be treated with commercial circuit simulation software like LTSpice. Lumped element models for acoustic components already exist, thus these were adapted for the μSAM geometry (2). A complete simulation can be done by bringing the LE models together (3).
Furthermore all relevant constraints for the simulation had been determined. This includes the availability for piezoelectric materials and materials for the bonding processes. Additionally for the design, relevant parameters can be defined. These are the resonant frequency of the piezoelectric transducer as well as the representative parameters acoustic compliance, acoustic mass and the electrical capacitance as well as the transducer coefficient. All these parameters can be influenced by the material choice and the geometrical dimensioning. Some of them are preset, such as the thickness of the piezoelectric material and the bonding layer by the manufacturers.
2.1 Combined Simulation of the electromechanical sub-system
It could be shown, that it is feasible to use the Finite Element Method to simulate the behavior of the electromechanical subsystem (piezoelectric membrane transducer), because for the aimed geometry no analytical expressions are available and a derivation would be extremely time consuming. Therefore FE model of the μSAM actuator was prepared. The mechanical influence of the metal electrodes on electromechanical behaviour can be neglected because of the small electrode thickness. Hence coupled electrical degrees of freedom were used on top and on the bottom of the piezoelectric material. Standard partly anisotropic material parameters of the manufacturers (piezoelectric material, bonding intermediate layer, silicon wafers) were used.
The preliminary tests have been done to proof the different concepts and their possibilities in the manufacturing of the μSJA parts (membrane thickness of 70 µm and 140 μm, etching of the silicon nozzle with the typical pitch of approximately 54 ° ) and the bonding techniques of the PZT cermic to silicon bonding step. FE-model results can be compared with Laser-Doppler-Vibrometer measurement results. Concerning 610-007-4 B4 a maximum deflection in resonance (Q=80) of 1.78 μm could be determined via FEM, the measured value is 1.64 μm. A difference of the resonance frequency can be seen between measurement 10.3 kHz and simulation 12.9 kHz. A quasi-static deflection of 30 nm was measured compared to 42.8 nm as a predicted value.
For understanding the motion of the transducer, FE-simulations are helpful. Different constraints, at first the mounting condition of the device were discussed.
It can be seen, that the clamping of the full frame is a realistic boundary condition for the assembly in use. The damping under standard conditions can be estimated and should be comparable with standard SJAs. Therefore a Q-factor between 1 and 100 can be achieved. We assume 10 and 80 corresponding to different measurement setup with different clamping conditions. Corresponding to the preliminary tests, a prediction was made for the optimized designs 610-028-2 and 610-028-6.
As a conclusion of the preliminary tests the following aspects can be named to describe differences between FE-simulations and measurements:
1)Induced mechanical stress while bonding procedure (different coefficients of thermal expansion (CTE) of the piezoelectric material, the silicon and the intermediate bond layer lead to mechanical stress as a result of temperature and pressure treatment).
2)Variations in the process (e. g. notching was determined) leads to the point, that geometries might be slightly different.
3)Variations of the used piezoelectric material (PIC 255) properties can vary max. +/- 20%.
4)Behaviour of the intermediate bond layer is assumed as linear elastic.
5)Failures caused by the measurement setup (clamping of the chip, wire bond and soldering contact).
2.2 Lumped Element Simulation of the acoustic sub-system
For modelling the acoustic sub-system, lumped elements were used. This is a time efficient concept because of its high degree of abstraction. However, for complex distributed geometries it is hard to derive exact models, but for the aimed μSAM geometries analytical acoustic expressions partially exist [12]. Hence the model [13] was adapted with aimed geometries. Unfortunately in the TANG model fluid mechanical expressions were considered, which can only be used for simple axial symmetrically geometries. Within the framework of μSAM it is impossible to derive fluid mechanical models of the μSAM geometry because of the complex mathematical descriptions. In conclusion an acoustic sub-model was set up using the classical theory of linear acoustics and has been extended for the geometries like the truncated cone and truncated pyramid. Due to the different orifice geometries, which should be investigated, two cases have been taken into account - the rectangular and the circular cross-section with edge length a. In case of the μSAM geometry extensions for wider frequency range were made because of the small dimensions, for details see [12]. Further extensions have to be made to extend the model considering the truncated shape of the orifice due to KOH-processes generating a 54.7° etch profile and laser-ablation used to generate the cylindrical and truncated cone shapes. The dimensions rcyl = rtc = b = a/2 = 50 μm / 100μm / 150 μm / 250 μm. This is a discretisation and summation of calculated lumped values and gives more accurate results than averaging over the inlet and outlet area.
2.3 Combined Simulation of the μSJA
At this point results for the different orifice shapes and the two electromechanical systems are shown.
As an interpretation the following aspects are relevant:
-A smaller orifice leads to a higher exit velocity in electromechanical resonance
-A larger orifice shows a higher exit velocity in the acoustic resonance
-610-028-6 is stiffer, has a higher resonance frequency but also a lower exit velocity in electromechanical resonance than 610-028-2
-Truncated cone and truncated pyramid show similar behaviour, both are more efficient than the simple cylinder
-Electromechanical and acoustic subsystem are coupled (position of the resonance frequencies influence each other)
-610-028-2 with 100 μm outlet (a/2 = 50 μm) truncated pyramid can be determined as an optimal design of this study
Further acoustic aspects are important, i. e. the eigenfrequencies of the cavity, which are in the range of the working frequency. To calculate these, in a first step the truncated pyramid shape of the cavity can be approximated as a cuboid.
In consequence, the acoustic modes of the cavity will influence the frequency response (shift of the resonance frequency of the electromechanical transducer) and causes differences in the amplitudes. Especially the mode at 27.3 kHz could be critical. Finally the simulation results are compared to measurement results of the test chip B4 of the 610-028-2 sample. It can be seen, that the damping has to be improved, but this could only be done by using fluid mechanics or CFD simulations. Regarding the short timeframe of the project μSAM, fluid mechanics or CFD simulations cant be realized. Finally, the Combined Simulation models are ready to be studied in interaction with an electronic power circuit. It could be shown, that the created models can be verified by measurement results.
3. Manufacturing of the µSJAs
In this chapter the manufacturing of the components is presented. The decision for the design concept has been explained in chapter '2 Design of the μSJA'. Based on that design the silicon manufacturing concept of the whole actuator is developed and presented in '3.1 Dimensions and geometry of the μSAM-μSJA'. The results of these manufacturing steps are presented in '3.2 Manufacturing of the silicon membranes' and in '3.3 Manufacturing of the silicon and laser micro-machined nozzle configurations'. In chapter '3.4 Characterization of the PZT ceramics' the properties of the piezoelectric ceramics are presented. Finally the membrane part, the nozzle part and the piezoelectric ceramic are ready for integration.
3.1 Dimensions and geometry of the μSAM-μSJA
The presented μSJA design consists of two KOH-etched silicon wafers in order to form a cavity. This cavity and an etched orifice build up the acoustic subsystem. A bonded bulk piezoelectric transducer at the opposite wafer is used to drive the electromechanical subsystem. For this purpose a new bonding technology has been obtained to assemble the transducer membrane of the μSJAs with standard bulk PZT ceramics.
Different investigations have been done using the Combined Simulation to predict the deflection of the piezoelectric transducer to find an advantageous geometric arrangement in order to increase output parameters like the exit velocity.
3.2 Manufacturing of the silicon membranes
The manufacturing of the membrane-part is defined by a KOH-etching step into the silicon substrate according the required membrane thickness.
The diaphragm-wafer's backside is wet etched in order to produce a membrane in different thicknesses. In preliminary tests the thickness of 40 μm and 70 μm was chosen to evaluate the overall manufacturing and bonding concept that showed good results. Therefore, in the finally manufacturing based on the results of the work package 2 'Design' the three membrane thicknesses 100 μm, 200 μm and 300 μm are aimed to test different types of the electromechanical subsystem.
Time-dependent wet etching without any defined etch stop causes notching due to the inhomogeneous concentration of the etchant, which appears especially in huge lateral and transversal etching dimensions.
This leads to a warping of the silicon membrane, depending from the etching depth. The 100 μm membranes results in a warping of 21 μm (largest etch depth), the 200 μm membranes in a warping of 14 μm and the 300 μm membranes in a warping of 10 μm. The measurements of the membrane thickness have been done by an optical membrane thickness measurement sensor (BMT). Measurement was carried out in the middle of the membrane of each chip and then the values were averaged. The characterization of the warping of the membranes has been performed by a withe light interferometer (Zygo NewView 6200).
This non-planar profile must be taken into account when investigating different bonding methods for mounting the PZT ceramics on the backside of the membrane.
The etch process has been split in two steps for the 100 μm membranes in order to reduce the strain caused by the masking layerstack. The strain of this layerstack would lead to a deflection of the 100 μm silicon membrane.
The Si surface is directly used for direct bonding with the nozzle wafer. In order to maintain the required surface quality, the masking oxide layer on the bond side is removed by wet etching just prior to bonding.
3.3 Manufacturing of the silicon and laser micro-machined nozzle configurations
The manufacturing of the nozzle-part is characterized by two different micro-machining technologies which are used to produce different nozzle configurations.
One technology is a sequential etching process with the same (KOH) etchant, which is used for the manufacturing of the silicon membranes.
For this the orifice will be preliminary etched with a depth of 300 μm. Afterwards, part of the masking layer protecting the cavity is removed. The second step is the etching of the orifice and the production of the cavity geometry. The etch process is continued until perforation. The result of this procedure is the typical pyramid structure with a pitch of approximately 54 °, due to the anisotropic etch behavior related to the crystal orientation. In contrast to the truncated pyramid structure of the KOH-etching nozzles two other nozzle structures that are manufactured by laser micro-machining are investigated. The first laser micromachined nozzle structure corresponds to a cylinder with a pitch of approximately 90 °, the second laser micro-machined nozzle structure corresponds a truncated cone with a pitch of approximately 54 °.These three nozzle structures are manufactured with the equal geometrical dimensions of the cavity and the four different nozzle exit geometries: 100μm, 200μm, 300μm and 500μm concerning the nozzle outlet.
The achievements of these processes are 3D nozzle structures of the orifice in a required aspect ratio and a very small cavity, resulting in a higher pressure gradient which leads to a much higher exit jet velocity compared to a dry anisotropic etched orifice combined with a huge cavity. This creates the main advantage compared to known layouts. Now the two parts (membrane and cavity) are ready for bonding to form a small cavity with an orifice on the one side and a diaphragm on the opposite side. Finally, the piezoelectric ceramic disc is mounted on the back side of the diaphragm into the wet etched cavity.
3.4. Characterization of the PZT ceramics
Measurements of the piezoelectric ceramics have been done concerning the warping of the ceramics and their roughness. It can be shown that the piezoelectric ceramic itself has a not negligible warping and roughness. This warping (in the range of -17 μm up to +10 μm) and roughness (in the range of 0,47 μm up to 0,60 μm) of the piezoelectric ceramic itself and the warping of the KOH-etched silicon membrane can limit the bonding technology of the piezoelectric ceramic bonding.
4. Bonding
In this chapter the bonding concepts as well as the different bonding techniques are presented. Furthermore, investigations of the different bonding techniques and further characterizations of the proposed bonding for the use in the μSJA are described.
4.1 Bonding concept
There are two options for the integration concept: the PZT ceramic-to-wafer bonding first- and the wafer-to-wafer bonding first concept. The challenges of the whole fabrication process are the bonding steps which are needed to convert the SJA concept into a μSJA concept using standard bulk PZT ceramics.
4.1.1 PZT ceramic-to-wafer bonding first
For this bonding approach, dies of the PTZ ceramic need to be bonded to the rough and warped backside of the membrane wafer. Furthermore, the PZT actuators themselves show averagely a certain roughness and a warping. In addition, due to the CURIE temperature of the PZT material, which must not be exceeded, the temperature budget for the bonding process and all subsequent processes is limited. Thus, a low temperature adhesive bonding method with a relatively thick adhesion layer in order to overcome the warp and roughness problems can be chosen. As the membrane can be mechanically supported, there are no pressure limitations. This leads to the opportunity of using higher bonding pressures and results in a higher bonding strength. For the subsequent wafer-to-wafer bonding step only low temperature direct bonding- or adhesive bonding techniques are suitable.
4.1.2 Wafer-to-wafer bonding first
The alternative is the wafer-to-wafer bonding followed by the PZT ceramic bonding step. Due to the fact that the wafer-to-wafer bonding concept followed by the PZT ceramic bonding step shows more benefits compared to the other concept it was finally chosen. Therefore, the wafer-to-wafer bonding step is not limited to temperature regimes or process restrictions. This means that high temperature direct bonding techniques can be used. However, now the PZT ceramic-to-membrane bonding step faces some limitations of the bonding process. The cracking of the membrane due to a large deflection must be avoided. This is tested with cross section- and shear tests with bonded test chips, with different bonding pressures.
4.2 PZT ceramic bonding techniques
Several bonding techniques were tested due to their ability for the PZT ceramic-to-membrane bonding step. Therefore test wafers were used for these preliminary tests. The main focus is to realize the bonding of the rough and warped PZT ceramic to the silicon membrane. Furthermore, the bonding techniques should be able to realize the electrical connection to the lower electrode of the PZT ceramic, because this will reduce technology steps, like deposition, structuring and connection steps. The used bonding techniques are detailed described in the followed subchapters.
4.2.1 Bonding with the negative photoresist SU8
In these tests, the negative photoresist SU8 is structured in 10 mm x 10 mm squares in different thicknesses (5 μm, 10 μm and 20 μm) to overcome the roughness and the warping of the PZT ceramics. But during the bonding step of twelve PZT ceramics to the silicon by using the SU8 in the three different thicknesses, it was not able to compensate the roughness and warping. Furthermore, the test with the thickness of 20μm showed that the surface tension of the SU8 is too high. This surface tension results in removed SU8 structures from the test wafers.
4.2.2 Bonding with adhesive tape Nitto Denko 5601
The adhesive tape 5601 from Nitto Denko is a double side adhesive tape which can be structured by the use of laser micro-machining. However, the bonding tests with this adhesive tape showed no bonding between the PZT ceramics and the silicon membranes. Reasons are the roughness and the warping of the PZT ceramic - and that the adhesive tape with a thickness of 10 μm is not able to compensate this.
4.2.3 Bonding with Loctite electrically conductive adhesive 3880
To test the opportunity of electrically conductive adhesives the Loctite 3880 was chosen. This is one component electrically conductive epoxy which showed good bonding results. Nevertheless disadvantages such as reproducibility especially in the layer thickness and the overall filling, limit the applicability of this bonding techniques for bulk PZT ceramic bonding.
4.2.4 Bonding with reactive multilayer systems
The stacked multilayer, each layer with a thickness in the range of a few nanometers, provides a high energy storage which will be released during the bonding process. A raise in temperature, the application of a slight electronic current or even a laser pulse are able to ignite the self-propagating exothermic reaction. Although the temperature during the reaction is high, the surrounding components are not affected despite of a few lattice planes due to the ultra fast propagation. This implies a bonding process at nearly room temperature. The two different PZT ceramics were bonded against a complete with gold sputtered 4 inch silicon wafer. The preliminary investigation was done by using a razor blade to test the bonding strength between the PZT ceramic, the reactive multilayer system and the golden silicon wafer. It could be shown that there is a really huge bonding strength between the bonding partners. It was possible to destroy the PZT ceramic instead of the bond of the PZT ceramic electrode and the golden silicon wafer.
The used reactive multilayer system is 40 μm thick and consists of alternating layers of aluminium and nickel. In bond test this bonding technique shows only very good results if the bonding partners were pressed together during the bonding step with a high bonding pressure. Otherwise the bonding is characterized by enormous unbounded areas and voids. Referred to the usage in μSJAs with their thin silicon membrane this bonding technique seems not be feasible to fulfill the requirement of smaller bonding pressures. The cracking of the membrane has to be avoided.
4.2.5 Bonding with electrically conductive adhesive foils
The used conductive layer is an electrically conductive adhesive foil which has a thickness of 35 μm and showed excellent bonding results in first preliminary bonding tests. With this bonding foil it is possible to use lower bonding temperatures and bonding pressures. This foil can be structured e.g. by the use of a laser and is able to compensate the roughness and the warping of the PZT ceramic. Therefore, this adhesive foil is used to bond the PZT ceramics to the thin silicon membranes after the wafer-to-wafer bonding step.
4.3 Characterization and Optimization of the PZT-ceramic to silicon wafer bonding technique
Different characterization steps are performed to verify the quality of the bonding. Especially for bonding of PZT ceramics on thin membranes, a minimum of bonding pressure is allowed in order to ensure a crack free assembly. To evaluate the required bonding force which results in a high quality bonding, bulk PZT ceramic-to-membrane bonding tests with different bonding forces have been done. The characterization of the bonding strength is realized for a first estimation by yield tests and for a quality evaluation by shear tests of bonded bulk PZT ceramic-interlayer-membrane test chips with different bonding pressures. Three batches with bonding pressures of 2 bar, 1 bar, 0,5 bar, 6E- 3 bar and 1,5E-4 bar (the weight of the bulk PZT ceramic) have been chosen. These five types are equally bonded together and have been investigated according to their bonding strength. The 10x10 mm² test chips of the first batch have been cut in 1x1 mm² pieces to analyse the bonding yield after dicing for a first strength estimation between the interlayer and the bulk PZT ceramic. The test chips of the second batch were cut into cross-section sheets for the characterization of the wetting to the different materials and the interlayer. The test chips of the third batch have been cut in 4x4 mm² pieces to analyse the shear strength.
4.3.1 The first batch: 1x1mm² - bonding yield after bonding
The 1x1 mm² diced test chips affirmed the assumption, that a low bonding pressure (1.5E-4 bar = bulk PZT ceramic itself) results in a poor strength, compared to a high bonding pressure (1 bar and 2 bar). The bonding yield with regard to the different presuures and strengths is between 0% (1,5E-4 bar) and 100% (2 bar). Furthermore, a very good yield of 80% can be guaranteed already with a bonding pressure of 0,5 bar. This fact is especially for the μSJAs very positive because this lower bonding pressure leads to a lower membrane stress and to a lower deflection of the membrane during the bonding process. These results validate the feasibility of the new bonding technology especially for the usage on thin membranes.
4.3.2 The second batch: cross-section
The cross-section test chips have been evaluated regarding cracks, blowholes, voids and of course the homogeneity of the interlayer with respect to the whole cross-section. The reason for the poor strength of the '1,5E-4 bar' and the '6E-3 bar' test chips results are due to the enormous number of unbonded areas and voids . An example of the cross-section of a test chip which is bonded with a higher bonding force (2 bar).
4.3.3 The third batch: shear strength test
The shear tests have been done with the shear strength test tool (TIRAtest 2805). Due to the fact, that the 1,5E-4 bar and 6E-3 bar test chips do not show such good results in the yield- and cross-section tests compared to the other test chips, the 1,5E-4 bar and 6E- 3 bar test chips have not been characterized with the shear strength test. The results of the shear strengths test confirmed the 1 mm² yield tests. There is no huge variation between the average shear strengths of the 0,5 bar (30,.2 MPa) to the 1 bar (30,9 MPa) and 2 bar (29,9 MPa) bonded test chips. The shearing took place in the conductive polymeric interlayer system.
The alignment of the whole bulk PZT ceramic-interlayer-membrane stack, as well as the interlayer (with respect to the streamed interlayer material during the joining process and the different bonding pressures) has been analyzed. It could be shown, that only a little amount of the conductive polymeric has been streamed out. A high positioning accuracy could be reached by using the flip chip bonding tool for the assembly of the joining partners.
4.4 Bonding of the processed wafers to µSJAs
The final approach is characterized by the wafer bonding first concept that is a direct bonding step. Afterwards the bonded μSJA system is diced into chips and the PZT ceramics are bonded on the thin silicon membranes of the μSJAs. This is realized with the electrically conductive adhesive foil with a bonding temperature of 130°C and a bonding pressure of 0,5 bar. The link to the overall system is realized by the integration of the μSJAs into the system and by wire bonding.
4.4.1 Wafer bonding
Due to the used PZT ceramic bonding technology the wafer-to-wafer bonding (silicon fusion bonding with a special plasma pre-treatment of the bonding areas) is the first bonding step, followed by the PZT ceramic-to-membrane bonding. This silicon fusion bonding is done at room temperature followed by the temperature step of 200°C for 2 hours.
After the silicon fusion bonding of the membrane- to the nozzle-wafer, the bonded wafer pair is sputtered with chrome and gold. These layers will realize the electrode for the connection of the lower PZT ceramic electrode. The electrode and an isolation pad for the electrically connection are manufactured into the chrome-gold electrode by laser micro-machining. The dicing of the wafer to single chips is the last manufacturing step in the manufacturing of the μSJA chips.
4.4.2 PZT bonding
The following steps concerning the integration of the PZT ceramics into the μSJA chips, as well as the electrical connection of the PZT ceramics. Therefore the bonding of the PZT ceramic-to-the membrane with the bonding interlayer is the first integration step. This bonding interlayer is structured by the use of the Laser in 10 mm x 10 mm pieces, which where aligned in the middle of every μSJA chip by the use of a flip chip bonding tool (Tresky T-3002-FC3). On this bonding interlayer the PZT ceramic itself is aligned with this flip chip bonding tool as well. The PZT ceramic-to-membrane bonding step is realized in a pressure and temperature depending process, where the PZT ceramics are bonded on the silicon membranes at 130°C and 0,5 bar (according to the best result of the pretests, described in 4.3 Characterization and Optimization of the PZT-ceramic to silicon wafer bonding technique). Finally the μSJA is ready for integration.
4.4.3 Wire bonding
For the electrical connection of the upper electrode and the electrical connection to the actuator electronics bond wires are used and the μSJA chips are integrated in devices for the characterization and measurement of parameters like the exit velocity of each single μSJA chip.
5 Integration and characterization
The development of actuators for active flow control based on micro technologies also includes the integration of the actuators in a panel for characterization including the necessary electronics to drive and test the actuators. The activities performed for integration and characterization of the actuators are summarized in the following chapter.
5.1 Actuator characterization
For the characterization the electric behaviour of the μSJAs, the electric behaviour was evaluated with the impedance and phase. The AC voltage and the AC current were measured over a determined frequency range. The characterization of the performance of the actuators was done by measuring the exit velocity of the actuators depending on the actuation frequency.
5.1.1 Measurement Equipment
The measurements were performed in a special soundproofed characterization chamber. An automated measurement and storing of the measurement data was realized in LabVIEW™.
It was designed to control the actuators and automatically measure the parameters using and controlling specific equipment:
- Agilent E4980A Precision LCR Meter
- LabVIEW™ Run-Time Engine 2010
- Hameg HMF2525 Arbitrary Function Generator
- TEGAM High-Voltage Amplifier 2350
- PicoScope 4227
- DantecDynamics Hot wire Probe
- MiniCTA for Hot wire probes
- Tektronix TDS3054 oscilloscope
With the automated measurement tool in LabVIEW™ impedance, phase, AC voltage and - current of the actuators in a determined frequency range was measured. With preliminary tests the frequency range was detected.
On the upper side the device is selected via VISA Component. A delay time could be added to delay each frequency step for error detection on the LCR. The signal type and level are adjustable. A frequency sweep is configured via start and stop frequency and the increment.
In the middle of the GUI the output parameters are presented during measurement: Impedance (Ohm), phase (degree), Vac (V) and Iac (A). Measurement data is stored in a file. The location of the file is displayed on the right side. The graphs below the output parameters show the output parameters over the frequency range.
5.1.2 Impedance and phase of the actuators, Voltage and current
Basic measurement parameters are
- Sinusoidal actuation
- Frequency range μSJAs: 50Hz-20kHz
- Frequency range μSJAs: 50Hz-20kHz
5.1.3 Capacitance
The capacitance of the μSJA is measured with charging and discharging behaviour. The measurement is described in the next part. A constant DC source charges the DUT over R. After the charging process is finished, the switch S connects the resistance to GND and the capacitor uncharged. VDC and S had been realized by the function generator Hameg HMF2525. To drive the DUT the high voltage amplifier TEGAM 2350 was used. The amplifier has an amplification factor of 50.The output signal type of the function generator was set to rectangular waveform. The offset was adjusted until the low pulse of the output signal on the amplifier reaches GND level. In this way the signal on the DUT is only in the positive signal space and complete discharging of the device could be reached. The high state of the function generator output conducts VDC with R. This state represents charging process of the DUT. The low state of the function generator represents discharging of the DUT due to connection of R with Ground. The output frequency was set to 100 Hz for time intervals corresponding to complete charging and discharging process. The resistor R was measured with the FLUKE 187 (R=50,91 kΩ). The Voltage over the DUT was captured with the PicoScope 4227.
For the measurement of the capacitance of the SJA the charging and discharging behaviour was measured. The time constant t is measured by using the tangent line in the starting point of charging and discharging. With this time constant t and the resistor R the capacitance can be calculated.
Measurements with the μSJA were performed using the following input parameters:
- Input amplitude (HMF2525): 200 mV
- Amplitude on μSJA: 10 V
- Offset: 112 mV (GND)
- Pulse width: 5 ms
- Frequency: 100 Hz
5.1.4 Behaviour of the electromechanical subsystem
Two μSJA-systems with different membrane thicknesses have been measured. The system #1 '610-28-6+610-29-5' with the 200μm membrane and the system #2 '610-28-2+610-29-1' with the 300μm membrane have been compared. Both μSJA-systems are realized with four different nozzle configurations (100μm x 100μm, 200μm x 200μm, 300μm x 300μm and 500μm x 500μm). The manufacturing and the bonding of the single devices have been done in the same way. The membrane deflection per volt of every single μSJA has been analyzed with a Laser Doppler Interferometer.
5.1.4.1 The 200μm membrane μSJA-system
The membrane deflection per volt and the resonance frequency of the electromechanical subsystem of the system #1 '610-28-6+610-29-5' varies slightly between 9 to 12kHz and 4E-7m/V to 9.5E-7m/V.
5.1.4.2 The 300μm membrane μSJA-system
The membrane deflection per volt and the resonance frequency of the electromechanical subsystem of the system #2 '610-28-2+610-29-1' can be divided into two parts. The chips with the 100μm and 200μm nozzle configuration result in lower resonance frequencies in the range of 9.5 to 11.5kHz and in membrane deflections of 2E-7m/V. The chips with the 300μm x 300μm and the 500μm x 500μm nozzle configuration result in higher resonance frequencies between 14 to 17kHz and in membrane deflections in the range of 1.8E-7 to 4.5E-7m/V. This variation can be explained due to the back coupling of the different nozzle configurations to the 300μm electromechanical subsystem.
5.1.5 Velocity measurement of the mounted μSJAs
The performance of the whole system #1 '610-28-6+610-29-5' with the 200μm membrane and the system #2 '610-28-2+610-29-1' with the 300μm membrane have been characterized by measuring the exit jet velocity. This has been done with a special clamping of the μSJAs and a hot wire velocity sensor.
The measurement results show that exit velocities in the range of up to 80 m/s are possible for the μ-technology based SJAs. This result could be shown using a μSJA with 100 μm membrane thickness, a 500 μm x 500 μm KOH-nozzle configuration and an actuation voltage of 100V.
5.2 System Integration
In previous tests, the single actuators have been mounted onto single mounting parts. That means one actuator for one mounting support. To use the μSJAs in wind tunnel tests, it is necessary to mount the actuators in a single row and to create an actuator surface that match the shaped surface of the wing configuration. Therefore a single mounting support, called panel was designed. It is possible to adapt such a panel for different wing configurations and to use the panel for a variable number of single actuators.
In this design the panel includes five single actuators which are located in the centre of the panel that has a total length of 492 mm. The exit holes are located at the top surface and they lead to an outlet slot. A frame on the inner side of the panel is used to fix the orientation of the actuators. The required width for a single actuator frame is 20.2 mm, so it would be possible to design a panel with much more actuators. For this first demonstrator, five actuators are arranged in a straight line.
Two screws are necessary to mount this part. Each actuator is fixed by one holder which allows the separate mounting of the actuators.
To mount the whole panel onto a wing model, the panel is designed with two holes at the endings of the panel. For this, just two screws are necessary to fix the whole panel. In addition, the screws are covered by a curved plate to ensure a surface without any flow influencing structures.
5.3 Control Electronics
In WP5 the goal was to design control electronics for the μSJAs. With reference to the measured resonance frequency in deliverable 4 a signal generator was designed.
The requirements of the control electronic for the μSJAs are
- frequency tuneable around resonance frequency
- sinusoidal actuation
- output amplitude: 2 Vpp
The output signal of this frequency generator has to be transformed to high voltage signals with a high voltage amplifier. Due to the lack of integrable high voltage amplifiers the TEGAM 2350, which was also used for characterization, had to be used for electronic testing. The high voltage output of the amplifier is distributed to the μSJAs with a Distribution box.
5.3.1 Actuation frequency generation
In this part the frequency generator is described. An integrated circuit, the XR2206 is used on the PCB layout for frequency generation. There are 3 frequency ranges which can be adjusted with a Jumper on the PCB. The frequency can be adjusted according to the resonance frequency of the μSJAs with knobs for coarse (left) and fine (middle) adjusting. The amplitude can be regulated with the right knob. On the PCB with two trimmer resistors it is possible to change the symmetry output of the sinusoidal signal output.
Features:
- Frequency range: 1..100000 Hz in 3 ranges
- Voltage Input: 12..20 V
- Amplitude regulation
- Enable switch
As input a voltage of 12..20 V has to be used. The integrated circuit 7812 regulates the input voltage to 12 V. Trimmer resistors are used for symmetric output adjustment. The capacitors 10nF, 220nF and 10μF set the frequency range. TL074P connected as a voltage follower moves the output Output No: 007 - Version: 001 Date: 29.11.2012 Page 55 of 59 level with capacitor C8 to swing symmetrically around GND. The output can be enabled and disabled with a switch. Output connector is a BNC-Connector.
5.3.2 Signal distribution
The signal provided by a high voltage amplifier has to be distributed to 10 actuators. With the distribution box the 2 poles of the input signal is wired to 10 actuators each pole. On the output of the box a SUB-D connector connects the wires for the actuators. The SUB-D output pins are isolated for high voltage safety requirements. The cabling is built in a Fibox housing with two TS35 rails.
6 Conclusion
The aimed objective of this project, the design, manufacturing and testing of microtechnology based synthetic jet actuators, could be reached. These µSJAs have been manufactured by microtechnology at the clean rooms of the ZfM. Furthermore a novel bonding technology for piezoelectric ceramics has been investigated. Finally, tests and characterizations of the manufactured μSJAs show their ability for generation of increased exit velocity and hence they can be applied in active flow control.
A novel bond technology regarding very low temperature bonding of bulk PZT ceramics to thin silicon membranes has been investigated successfully. The conductive polymeric interlayer system has been characterized and evaluated for a first estimation by yield tests and cross-section tests. For a quality evaluation bonded bulk PZT ceramic-interlayer-membrane test chips have been characterized by shear tests. A good compromise between a high bonding yield and a low bonding pressure has been achieved at a bonding pressure of 0.5 bar. Furthermore, this bonding pressure results in a minimal stress distribution.
First measurement results of samples and the characterization of the electromechanical transducer are presented and the results show a good fit to the simulation. Therefore, the described joining concept represents a promising concept especially for transducers based on thin silicon membranes but also for other applications like cantilever actuators. Due to the usage of standard bulk PZT ceramics, a bonding process for MEMS has been established without the need of special pre-treatments. Finally a complex and integrated manufacturing process including different materials for smart systems integration has been developed.
After the development of the technology for the manufacturing of the micro synthetic jet actuator a system to integrate the actuators was designed and manufactured.
The first tests of the actuators showed that the μSJA performance of the system with the 200μm membrane reaches maximum exit jet velocities of about 28m/s at resonance frequencies of about 18.3kHz with the 500μm x 500μm nozzle configuration and an actuation voltage of 100V. The μSJA performance of the system with the 300μm membrane reaches maximum exit jet velocities of about 38m/s at resonance frequencies of about 21.2kHz with the 500μm x 500μm nozzle configuration and an actuation voltage of 100V. Therefore, the 300μm membrane system with the 500μm x 500μm nozzle configuration is preferred because of its higher resonance frequency and the higher exit jet velocities compared to the other results. Nevertheless the variation of the results depends on different points such as the manufacturing challenges (notching of the membranes, roughness of the membranes and the PZT ceramics, ) as well as geometrical restrictions (bonding interlay, variation of the PZT ceramics).
In a final test of the second manufacturing batch exit jet velocities in the range of up to 80 m/s could be reached for the microtechnology based SJAs. This result could be shown using a μSJA with 100 μm membrane thickness, a 500 μm x 500 μm KHO-nozzle configuration and an actuation voltage of 100V.
The successful manufacturing process and the satisfying measurement results show that it is possible to design and manufacture synthetic jet actuators based on micro-technology. All the manufacturing steps could be realized and presented and the measured exit velocity are in the range of expectation.
7 References
[1] C. Lee et al., Sensors and Actuators A 108 (2003) pp 168-174
[2] S.G. Mallinson et al,., Sensors and Actuators A 105 (2003) pp 229-236
[3] L. Gomes et al.: Towards a Pratical SJA for Industrial Scale Flow Control Applications pp.1-8
[4] P. Mane et al., Journal of Intelligent Material Systems and Structures 18, 2007, pp 1175-1190
[5] M. Watson et al., The Aeronatical Journal (2007), Paper No. 3160
[6] Q. Gallas, Thesis, 2002
[7] V. Tesar, J. Kordk., Sensors and Actuators A : Physical (2008), 1-11
[8] Z. Travnicek et al., Sensors and Actuators A 120 (2005) pp 232-240
[9] D. J. Coe et al, Sensors and Actuators A 132 (2006) pp 689-700
[10]Y. Liang et al., Sensors and Actuators A 125 (2006) pp 512-518
[11]A. C. Miller, Thesis (2004)
[12]Lenk, A.; et al.: Electromechanical Systems in Microtechnology and Mechatronics; Springer-Verlag Berlin Heidelberg (2011), ISBN 978-3-642-10805-1
[13]Tang, H.; Zhong, S.: Lumped element modelling of synthetic jet actuators. In: Aerospace Science and Technology 13 (2009), pp. 331-339
[14]Starke, E.: Kombinierte Simulation - eine weitere Methode zur Optimierung elektromechanischer Systeme. Dresden, Technische Universität Dresden, Dissertation, 2009.
Potential Impact:
μSJA's concepts are new and show a large potential for active flow control applications. High exit velocities are possible with a lower energy consumption compared with standard SJA's. A large impact can be observed in the field of simulation. For μSJA's no models are available. This situation can be improved with μSAM. Another fact is, that with the help of micro technologies, well defined SJA's are producible.
Silicon based µSJAs have been manufactured by microtechnology at the clean rooms of the ZfM by a novel bonding technology for piezoelectric ceramics. This novel bonding technology will cause a large impact on the implementation of piezoelectric ceramics in MEMS. Advantages of these new technologies are strong bonds with only a small amount of temperature inducted mechanical stress. Thus, the Curie temperature is not exceeded and there is no depolarisation of the piezoelectric ceramics.
On the scientific field, the integration of pre-poled bulk PZT is expected to lead into new and innovative applications. Particularly in the areas of acoustical and optical microsystems as well as motion sensors, MEMS have great market potentials. The project partner's competencies are able to cover the required supply chain from design of application-relevant microsystems with sensor-actuator coupling to fabrication of prototypes to generate high potential for success. For the demonstration of the feasibility of the 'hybrid technology', realization of test structures has a crucial role and is considered as an essential step to the conversion into economically interesting products in cooperation with companies. The new composite technology leads to specific application aspects for the user through having:
- a cost advantage due to parallel processing and reduced production costs
- a time advantage due to faster bonding processes
- higher quality, reliability and longer lifetime due to lower joining temperature
- an improved functionality
The ZfM is part of the Chemnitz University of Technology and, therefore, a non-profit organisation. The ZfM is interested in utilizing the technology knowledge and expanding this know-how for further research and development projects. The dissemination was performed and will be further performed trough publications in technical and scientific papers in international journals and presentations of the results at leading European and international conferences. We also aim to exploit the knowledge gained in other regional, national and EU projects. This process is ongoing but not yet finished. After the end of the project, the results could be exploited for other related industry, such as automotive, electronics (e.g. by active cooling), printing technologies, micro acoustics and ultrasonic applications by further research and development projects.
Due to the fact, that the manufacturing process was successful and the measurement results showed that it is possible to design and manufacture synthetic jet actuators based on microtechnology, industrial partner already showed their interest in several aspect of the manufacturing process as well as in the use of the μSJAs for other applications.
List of Websites:
http://www.zfm.tu-chemnitz.de