Novel two-stage peristaltic micropump optimized for automated drug delivery and integration into polymer microfluidic systems Dissertation zur Erlangung des Doktorgrades der Fakultät für Angewandte Wissenschaften der Albert-Ludwigs-Universität Freiburg im Breisgau vorgelegt von Andreas Geipel Freiburg - 2008 Dekan Prof. Dr. Bernhard Nebel Referenten Prof. Dr. Peter Woias, IMTEK, Universität Freiburg Prof. Dr. Ulrich Massing, Klinik für Tumorbiologie, Freiburg Datum der Promotion: 5. Juni 2008 Abstract This thesis presents a novel concept of an implantable drug delivery system based on microsystem technology that incorporates a high-resolution volumetric dosing unit and a drug reservoir into the space of a conventional subcutaneous port. The controlled release of small drug volumes from a so-called active microport is beneficial, for example, for innovative methods in cancer treatment or pain therapy. The developed release system delivers a flow rate in the range of 0.1 - 50 µl/min and enables a patient-specific release profile. The core of the device is a novel two-stage peristaltic micropump. It features a backpressure independent volumetric dosing capability, such that a stable flow rate is ensured up to a backpressure of 30 kPa. This highly valuable feature is based on a new serial arrangement of two active valves and relies on both an appropriate electrical actuation sequence of the piezo-actuators and an intrinsic limitation of the membrane deflection by the valve seats. A detailed lumped-parameter model is derived in order to reveal the physics behind this pumping principle and to identify the optimum control scheme. For the fabrication of the silicon micropump a comparably simple and robust 2-wafer process based on standard MEMS processes is utilized. A thorough experimental investigation demonstrates the high performance of the micropump. The backpressure independence of the flow rate enables high-resolution volumetric dosing within the aforementioned flow range. A new figure of merit referred to as differential fluidic output resistance is introduced to quantify the degree of backpressure independence within the working range of the micropump. The stroke volume and hence the resolution of the micropump is adjustable between 50 – 200 nl via the upstroke voltage applied to the piezo-actuators. Typical actuation frequencies range from 0.05 to 5 Hz and the flow rate scales in proportion to the frequency within that frequency range. i Abstract A modified actuation scheme referred to as gas pumping mode is proposed for the transport of gas. This actuation mode equips the micropump with a full capability to pump both gas and liquid which enables a reliable self-priming process. For alternate gas and liquid pumping a critical compression ratio is analytically derived which is determined by the capillary pressure drop of a gas-liquid interface trapped in the pump chamber. As a side aspect of this work, an alternative thermal actuation of the two-stage micropump based on the expansion of paraffin wax upon its solid-liquid phase transition is explored. A major technological improvement is achieved by the development of a novel direct heating concept based on the dispersion of conductive carbon black particles into the paraffin matrix. The heat generation inside the paraffin wax yields an increased energy efficiency and enables higher actuation frequencies compared to established thermal approaches. For biomedical applications the encapsulation of devices is a nontrivial task which has to meet ambitious specifications such as biocompatibility, mechanical and chemical stability, fluidic sealing and electrical isolation. In this work, a multilayer soft lithography process for polyurethane is introduced for rapid prototyping in microfluidics. Compared to the widely used soft-polymer polydimethylsiloxane (PDMS) the polyurethane excels by an even better transparency and a slightly higher mechanical stability. Moreover, a large number of glues work with polyurethane whereas adhesives for PDMS are rarely found. Based on the developed process chain, a compact multilayer housing for the developed drug delivery system is presented. The final prototype of the active microport system is housed in an injection molded container made from polypropylene which measures 49 x 36 x 25 mm3 only. A miniaturized highperformance electronic control unit is embedded in the system and enables freely programmable dosing profiles. An implemented pressure sensor is used to permanently monitor the dosing process and to detect a potential catheter occlusion. The electronic control unit is optimized for both energy consumption and weight which are both essential parameters for a portable device. The power consumption of the active microport depends on the actuation frequency and sums up to approximately 50 – 200 mW. The overall weight of the system including a rechargeable battery is 35 g. ii Zusammenfassung Die vorliegende Arbeit stellt ein neuartiges Konzept für einen implantierbaren aktiven Mikroport vor. Der Ansatz verwendet das Potential der Mikrosystemtechnik um eine hochauflösende, steuerbare Dosiereinheit und ein Reservoir in ein System von der Größe eines konventionellen passiven Ports zu integrieren. Ziel dieses Systems ist es, kleinste Mengen eines Wirkstoffs mit Dosierraten im Bereich 0.1 - 50 µl/min kontrolliert und gleichmäßig zu verabreichen. Dies würde neue Therapieformen in der Krebs- und Schmerztherapie ermöglichen, bei denen Patienten-spezifische Dosierprofile zur Anwendung kommen. Die Dosiereinheit des aktiven Mikroports beruht auf einer neuartigen Zwei-MembranMikropumpe, die eine nahezu gegendruckunabhängige Dosierrate aufweist. Die SiliziumMikropumpe wird piezoelektrisch angetrieben und besitzt zwei aktive Ventile. Durch eine geeignete Ansteuersequenz und die definierte Begrenzung der Membranauslenkung durch den Ventilsitz kann die Dosierrate bei niedrigen Frequenzen bis zu einem Gegendruck von 30 kPa stabil gehalten werden. In der Arbeit wird ein umfangreiches Netzwerkmodell zur Beschreibung der Mikropumpe hergeleitet. Durch nachfolgende Netzwerk-Simulationen lassen sich physikalische Zusammenhänge analysieren und die Ansteuersequenzen optimieren. Die Herstellung der Mikropumpe beruht im Wesentlichen auf der Nutzung der bekannten Siliziumtechnologien. Im Vergleich zu anderen Konzepten wird eine relativ einfache Prozessabfolge vorgestellt, wobei die Mikropumpe aus zwei strukturierten Wafern zusammengesetzt wird. Daran schließt sich eine umfangreiche Charakterisierung der Mikropumpe an. Die erreichten Leistungsmerkmale werden gestützt durch die experimentellen Ergebnisse und die zugehörigen Simulationen. Zur Bemessung der Gegendruckunabhängigkeit wird eine geeignete Kennzahl, der differentielle fluidische Ausgangswiderstand, eingeführt. Durch die angelegten Spannungsniveaus kann das Hubvolumen und damit die Auflösung der Mikropumpe zwischen 50 – 200 nl variiert werden. Innerhalb des typischen Ansteueriii Zusammenfassung frequenzbereichs von 0.05 – 5 Hz wird eine gute Linearität zwischen Frequenz und Dosierrate nachgewiesen. Für den Transport von Gasen wird ein modifiziertes Ansteuerschema entwickelt welches einen stärkeren Fluidvortrieb ermöglicht. In diesem Fördermodus, der sowohl für Flüssigkeiten als auch für Gase geeignet ist, wird eine selbstständige Befüllung der Mikropumpe erreicht. Bei wechselweiser Förderung von Flüssigkeiten und Gasen treten Grenzflächeneffekte auf, die bei der Bestimmung eines kritischen Kompressionsverhältnisses berücksichtigt werden. Als Nebenaspekt dieser Arbeit wird ein alternativer Antrieb der Pumpe mittels eines thermischen Paraffin-Aktors untersucht. Paraffin zeigt eine vergleichsweise hohe Volumenausdehnung beim Phasenübergang von fest nach flüssig. Zur technischen Nutzung dieses Effekts werden eine effiziente Heizstrategie und ein geeigneter Fertigungsprozess benötigt. Ein neuartiges Konzept auf der Basis eines leitfähigen Paraffin-Wachses stellt hierzu einen wesentlichen technologischen Fortschritt dar. Durch die Dispersion von leitfähigen Rußpartikeln im Paraffin kann die Wärme unter Stromfluss direkt im Wachs generiert werden. Dies führt zu einer deutlichen Effizienzsteigerung und ermöglicht kürzere Zykluszeiten im Vergleich zu anderen thermischen Konzepten. Insbesondere im Bereich medizinischer Anwendungen ist die Gehäusetechnik ein anspruchsvoller Bereich. Ein medizinisches System muss die gegebenen Kriterien bezüglich Biokompatibilität, mechanischer und chemischer Stabilität, fluidischer Kapselung und elektrischer Isolierung erfüllen. Im Rahmen dieser Arbeit wird ein MehrlagenFertigungsprozess für Polyurethan entwickelt. Dieser Prozess kann flexibel für den Prototypenbau im Bereich der Mikrofluidik eingesetzt werden. Im Vergleich zum oftmals verwendeten Polydimethylsiloxan (PDMS) besitzt das verwendete Polyurethan-Elastomer eine höhere Transparenz und eine leicht höhere mechanische Stabilität. Außerdem besitzt Polyurethan eine wesentlich bessere Verklebbarkeit als PDMS. In der Entwicklung des aktiven Mikroports wird dieser Prozess zur Herstellung erster Gehäuse-Prototypen verwendet. Der abschließende Prototyp des aktiven Mikroports befindet sich in einem spritzgegossenen Gehäuse aus Polypropylen. Die Größe des Gesamtsystems beträgt 49 x 36 x 25 mm3. In das System ist eine miniaturisierte Elektronik integriert, die frei programmierbare Dosierprofile ermöglicht. Ein Drucksensor überwacht den Dosiervorgang und erkennt einen eventuellen Katheterverschluss. Die Elektronik des Systems wurde optimiert bezüglich Größe, Gewicht und Energieverbrauch. Der Energiebedarf hängt von der Ansteuerfrequenz ab und bewegt sich im Bereich 50 – 200 mW. Das Gewicht des entwickelten Systems, einschließlich eines entsprechenden Akkus, beträgt etwa 35 g. iv Contents Abstract ............................................................................................. i Zusammenfassung .......................................................................... iii Contents ........................................................................................... v 1 Introduction ............................................................................. 1 1.1 1.2 Motivation .................................................................................................................... 1 State of the art ............................................................................................................. 2 1.2.1 Micropumps ....................................................................................................................... 2 1.2.2 Actuation principles ............................................................................................................ 5 1.2.2.1 Piezoelectric actuation .............................................................................................. 5 1.2.2.2 Thermal actuation ..................................................................................................... 5 1.2.2.3 Further actuation mechanisms .................................................................................. 6 1.2.3 Drug delivery ...................................................................................................................... 7 1.2.3.1 Diffusion-based systems ........................................................................................... 8 1.2.3.2 Transdermal injections .............................................................................................. 9 1.2.3.3 Active infusion systems ............................................................................................. 9 1.2.3.4 Adaptive systems .................................................................................................... 10 1.3 2 Objective of this thesis ............................................................................................ 11 Fundamentals ........................................................................ 13 2.1 Microfluidics .............................................................................................................. 13 2.1.1 Fluid mechanics ............................................................................................................... 13 2.1.1.1 Density and viscosity............................................................................................... 13 2.1.1.2 Continuum equation ................................................................................................ 15 2.1.1.3 Mach number .......................................................................................................... 15 2.1.1.4 Laminar regime and Navier-Stokes-equation ......................................................... 16 2.1.1.5 Stokes flow .............................................................................................................. 16 2.1.2 Wettability ........................................................................................................................ 18 2.1.2.1 Surface tension and interfacial energy.................................................................... 18 2.1.2.2 Contact angle and Young’s equation ...................................................................... 19 2.1.2.3 Contact angle hysteresis ......................................................................................... 20 2.1.2.4 Kinetic phenomenon ............................................................................................... 21 2.1.2.5 Wetting of silicon surfaces ...................................................................................... 22 2.1.2.6 Capillary effect and Young-Laplace pressure drop................................................. 22 2.1.2.7 Weber number and capillary number ...................................................................... 23 2.2 Piezoelectric membrane actuators ........................................................................ 24 2.2.1 Piezoelectric effect ........................................................................................................... 24 2.2.1.1 Piezoelectric materials ............................................................................................ 25 2.2.1.2 Actuation modes ..................................................................................................... 26 2.2.1.3 Piezoelectric coefficients ......................................................................................... 27 2.2.2 Membrane mechanics ..................................................................................................... 28 v Contents 2.2.2.1 2.2.2.2 2.2.2.3 2.2.2.4 2.2.2.5 3 Pressure induced deflection of a homogeneous membrane .................................. 29 Flexural rigidity of a piezo-membrane-composite ................................................... 32 Pressure induced deflection of the piezo-membrane-composite ............................ 35 Piezoelectric deflection of the piezo-membrane-composite ................................... 36 Displacement volume .............................................................................................. 40 Two-stage micropump .......................................................... 41 3.1 Design and Working Principle ................................................................................ 41 3.1.1 3.1.2 3.1.3 3.1.4 3.2 Concept............................................................................................................................ 42 Design of the piezo-membrane-actuator ......................................................................... 42 Geometry of the pump chamber ...................................................................................... 43 Actuation scheme ............................................................................................................ 44 Modeling and simulation of the micropump ......................................................... 45 3.2.1 FEM simulation of the bending membrane ...................................................................... 46 3.2.2 Lumped parameter model of the elastic membrane ........................................................ 47 3.2.3 Lumped parameter modeling of the piezoelectric actuation ............................................ 50 3.2.4 Superposition of piezoelectric and pressure induced bending ........................................ 53 3.2.5 Lumped parameter model of an active valve................................................................... 56 3.2.6 Fluidic inertance ............................................................................................................... 57 3.2.7 Lumped parameter model of the micropump................................................................... 59 3.2.8 Implementation and evaluation of the lumped parameter model .................................... 61 3.2.8.1 Pressure in the pump chamber ............................................................................... 62 3.2.8.2 Phase setting .......................................................................................................... 63 3.2.8.3 Backpressure characteristic .................................................................................... 64 3.2.8.4 Voltage-controlled adjustment of the stroke volume ............................................... 65 3.3 Transport of gases and gas bubbles ..................................................................... 65 3.3.1 3.3.2 3.3.3 3.3.4 3.3.5 3.4 4 Gas pumping mode ......................................................................................................... 65 Compressibility of entrapped air bubbles ........................................................................ 66 Gas-liquid interfaces ........................................................................................................ 69 FEM simulation of capillary forces ................................................................................... 72 Critical compression ratio ................................................................................................ 74 Single-membrane micropump ................................................................................ 77 Fabrication of the micropump .............................................. 79 4.1 4.2 Silicon manufacturing .............................................................................................. 79 Back-end processes ................................................................................................. 82 4.2.1 4.2.2 4.3 Quality control ........................................................................................................... 83 4.3.1 4.3.2 4.3.3 4.4 5 Gluing of piezo-disks ....................................................................................................... 82 Wire bonding .................................................................................................................... 83 IR inspection of bond quality............................................................................................ 83 Fluidic test setup .............................................................................................................. 84 Electrical capacitance measurement of the actuators ..................................................... 84 Fabrication costs ...................................................................................................... 84 Experimental characterization .............................................. 87 5.1 5.1.1 5.1.2 5.1.3 5.1.4 5.1.5 5.1.6 5.2 5.2.1 vi Experimental setup ................................................................................................... 87 Micro balance................................................................................................................... 88 Flow sensor...................................................................................................................... 88 Hydrostatic pressure method ........................................................................................... 89 Pressure sensor ............................................................................................................... 89 Pressure controller ........................................................................................................... 90 Electronic control unit for the micropump ........................................................................ 90 Variation of pressure ................................................................................................ 91 Backpressure independence of the flowrate ................................................................... 91 Contents 5.2.2 5.2.3 5.3 Variation of frequency .............................................................................................. 93 5.3.1 5.3.2 5.4 5.5 Flow rate versus frequency .............................................................................................. 94 Stroke volume versus frequency ..................................................................................... 95 Phase setting of the actuation sequence .............................................................. 96 Variation of the control voltage .............................................................................. 97 5.5.1 5.5.2 5.5.3 5.6 5.7 5.8 Opening voltage ............................................................................................................... 97 Closing voltage of the outlet valve ................................................................................... 98 Cut-off pressure ............................................................................................................... 98 Variation of the pump chamber geometry ............................................................ 99 Gas pumping mode ................................................................................................ 102 Gas-liquid interfaces .............................................................................................. 102 5.8.1 5.8.2 5.9 6 Impact of forward pressures ............................................................................................ 92 Impact of common mode pressures ................................................................................ 93 Capillary pressure drop.................................................................................................. 102 Membrane hysteresis .................................................................................................... 103 Single-membrane micropump .............................................................................. 104 Discussion ........................................................................... 107 6.1 Backpressure stability ........................................................................................... 107 6.1.1 6.1.2 6.2 6.3 6.4 6.5 7 Differential fluidic output resistance ............................................................................... 107 Comparison of different micropumps ............................................................................. 109 Pump chamber geometry ...................................................................................... 110 Controllability of the micropump ......................................................................... 112 Transport of gases and capillary effect .............................................................. 112 Reliability issues ..................................................................................................... 113 Feasibility study of a paraffin-actuated two-stage micropump ........................................................................... 115 7.1 7.2 7.3 7.4 Paraffin actuators and micropumps .................................................................... 115 Paraffin waxes ......................................................................................................... 117 Single-membrane paraffin micropump................................................................ 117 Direct heating concept ........................................................................................... 119 7.4.1 7.4.2 7.4.3 7.5 8 Conductive paraffin ........................................................................................................ 120 Process chain ................................................................................................................ 121 Results ........................................................................................................................... 123 Discussion ............................................................................................................... 125 Microfluidic devices in soft polymer technology .............. 127 8.1 8.2 8.2.1 8.2.2 8.3 8.3.1 8.3.2 8.3.3 8.4 8.4.1 8.4.2 8.5 8.5.1 MEMS fabricated by soft lithography .................................................................. 127 Elastomer materials................................................................................................ 129 Polyurethane .................................................................................................................. 129 PDMS............................................................................................................................. 130 Replica molding process ....................................................................................... 130 Fabrication of master molds .......................................................................................... 130 Patterning of elastomer layers ....................................................................................... 130 Parallelized replication process ..................................................................................... 131 Surface modifications of polyurethane ............................................................... 132 Hydrophilization by means of flame treatment .............................................................. 132 Hydrophobic microstructuring of the surface ................................................................. 133 Multilayer assembly for polyurethane ................................................................. 134 Investigation of different bond methods ......................................................................... 134 vii Contents 8.5.1.1 Adhesion ............................................................................................................... 135 8.5.1.2 Mixing ratio variation ............................................................................................. 135 8.5.1.3 Adhesive layer....................................................................................................... 135 8.5.2 Measurement setup ....................................................................................................... 135 8.5.3 Results of bond strength measurements ....................................................................... 136 8.6 9 Discussion ............................................................................................................... 137 Active Microport .................................................................. 139 9.1 9.2 System concept ...................................................................................................... 140 Prototype fabrication ............................................................................................. 140 9.2.1 9.2.2 9.2.3 9.3 9.4 9.5 9.6 9.7 Multilayer housing .......................................................................................................... 140 Stability of doxorubicine in contact with polyurethane ................................................... 141 Injection molded housing ............................................................................................... 142 Electronic control unit ........................................................................................... 143 Release monitoring ................................................................................................ 145 Dosing profiles ........................................................................................................ 146 In-vivo experiments ................................................................................................ 147 Application “Pharmaport” ..................................................................................... 148 10 Summary .............................................................................. 149 11 Outlook ................................................................................. 153 Acknowledgements ...................................................................... 155 Appendix A ................................................................................... 157 Appendix B ................................................................................... 159 Appendix C ................................................................................... 161 Appendix D ................................................................................... 163 Appendix E ................................................................................... 165 Appendix F ................................................................................... 167 List of publications ...................................................................... 171 Bibliography ................................................................................. 175 viii Chapter 1 Introduction 1 Introduction 1.1 Motivation Micropumps are considered to be a key component in microfluidics for the controlled propulsion of fluids. A multitude of principles including reciprocating displacement pumps, electrokinetic micropumps as well as osmotic or viscous mechanisms have been explored to satisfy the requirements of the particular target application. While microactuated dispensers such as ink-jet printheads have reached the high volume market the spread of micropumps is still confined to niche markets on an industrial scale. Since off-the-shelf micropumps are rarely found a strong market demand is essential to economically justify the development efforts for an application-adapted solution. In the medical sector, the demand for innovative therapies and for an increased patient compliance generates a pull market for the development of new instrumentations and devices. By now, MEMS technologies have entered the medical sector in many respects, for example in the field of electrostimulation, endoscopy, in-vitro diagnostics or drug delivery. For automated drug delivery systems the development of appropriate micropumps promises to replace conventional principles such as syringe pumps in the future. Here, the MEMS technology paves the way to miniaturize the device size and to equip the instrumentation with novel functionalities. This thesis constitutes a substantial part of a research project that aims to develop a MEMSbased automated drug delivery system. The intended application for this system is the controlled release of diluted chemotherapeutic agents to support innovative methods in cancer treatment. For this purpose a novel two-stage peristaltic micropump was developed and analyzed from both a technical and an analytical point of view. The main focus of this thesis was to develop a detailed understanding of the microfluidic mechanisms that govern the displacement of fluids for the introduced micropump design. Beyond that the interplay between the characteristics of the microactuator and the achieved pumping performance were also a focal point of this work. Finally, the integration of the micropump with supplementary components constituted a technical challenge that was tackled in this thesis. 1 1 Introduction 1.2 State of the art The follow sections will present a detailed overview of the achievements and research efforts in the field of micropumps and MEMS-based drug delivery systems. Additionally, fundamental aspects of microactuation mechanisms are briefly summarized with a focus on the principles employed in the framework of this thesis, i.e. piezoelectric and thermal actuation. 1.2.1 Micropumps In MEMS history the development of micropumps has been playing an important role for more than two decades. As a pioneer in this field Smits [1] proposed a piezoelectrically driven peristaltic micropump already in 1990. Other early micropump concepts have been published by van Lintel et al. [2] or Esashi et al. [3] at roughly the same time. Since that time numerous approaches have been pursued based on a great diversity of technologies and actuation principles. A detailed overview of the efforts and merits in the field of micropumps is given in comprehensive reviews published recently [4 - 6]. Despite the confusingly large number of concepts two aspects are remarkable. First, reciprocating micropumps based on the deflection of a diaphragm appear to be the predominant principle for micropumps [1 - 3, 7 - 26]. Even though various technologies and actuation principles have generated an uncountable number of different designs, the displacement of the fluid by means of an actuated membrane turned out to be the most robust, all-purpose concept. Second, a noticeable large number of micropumps – perhaps even the majority – are designed for life science applications. Here, small volumes have to be displaced for applications such as drug delivery systems, micro-total-analysis systems (µTAS) or biochips [9, 12 - 19, 24 - 34]. For these applications micropumps are predestinated due to their small size, precise metering capability, low power consumption and ability of system integration. While several dynamic pump mechanisms e.g. electroosmotic or electrohydrodynamic pumps depend on certain properties of the fluid such as ionic strength, reciprocating micropumps are generally suitable for the delivery of all gaseous and low viscosity fluids. The field of reciprocating micropumps is still heterogeneous comprising pumps with various numbers of diaphragms and different types of valves. A typical setup with three or more membranes placed in a serial fashion is commonly referred to as peristaltic actuation [1, 12 - 20]. Figure 1-1(a) illustrates the cross-sectional setup of the original design by Smits [1] which was intended to be used for insulin administration. Recently, a conventional peristaltic design has been proposed by Jang et al. [35] for biomedical applications and was tested for PBS injections into a rat. It is operated at high actuation frequencies of about 100 Hz causing a high power consumption of more than 500 mW but is limited to small backpressures below 3 kPa. 2 1 Introduction (a) (b) Figure 1-1: Conventional peristaltic micropump [1] (a) and electrostatic single diaphragm micropump with passive flow rectifiers [8] (b) (figures taken from [4]). Another popular configuration consists of one actuated membrane and two passive check valves for flow rectification [7 - 11, 36]. Figure 1-1 (b) shows the electrostatic micropump design proposed by Zengerle et al. [8] as an example. In this context, either flap valves or ball valves are the most commonly employed types of flow rectifiers. The first bubble-tolerant silicon micropump based on passive flap valves was reported by Linnemann et al. [37]. Other variations of the single membrane concept use valveless diffuser/nozzle elements [23], [38] or tesla valves. These fixed geometry flow rectifiers naturally suffer from high leakage rates and therefore low backpressure stabilities. A typical characteristic for reciprocating and peristaltic micropumps is the backpressure instability of the flow rate, i.e. the flow rate linearly declines with increasing outlet pressure [4, 10, 19]. The piezoelectric micropump offered by Bartels Mikrotechnik GmbH [36] is a commercial example of a single membrane micropump exhibiting the described backpressure characteristic (Figure 1-2). A similar micropump also fabricated in polymer technology via injection molding is available from thinXXS Microtechnology AG [39]. (a) (b) Figure 1-2: Piezoelectric micropump by Bartels Mikrotechnik GmbH (a) and corresponding backpressure characteristic (b) [36]. In our research group, a piezoelectrically driven peristaltic micropump with active inlet and outlet valves was developed in a thesis by Doll [25, 40]. Figure 1-3 (a) illustrates the design of this micropump which is optimized for bidirectional pumping with high flow rates up to 4.3 ml/min. The flow rate of this micropump design still shows the described detrimental decline when exposed to an increased outlet pressure. The target application for this design is an artificial sphincter prosthesis for patients suffering from incontinency [41]. Figure 1-3 (b) shows the integration of the micropump into a prototype of the so-called German Artificial Sphincter System (GASS). 3 1 Introduction (b) (a) Figure 1-3: Schematic drawing of the piezoelectric silicon micropump (a) and integration into the German Artificial Sphincter System (GASS) (b). Only a very few micropumps with a constant, backpressure independent flow rate have been published yet. It is common to most approaches that they implement a mechanical stopper concept which limits the deflection of the diaphragm to a constant value. Recently Inman et al. [42] published a design of a pneumatic micropump with a constant flow rate up to a backpressure of 25 kPa. The authors implemented a contoured-shape pump chamber to eliminate any dead volume. On the other hand, an externally generated actuation pressure of 40 kPa was applied to achieve this performance. Feng et al. [43] observed a nearly constant flow rate up to a backpressure of 2.5 kPa for their single-membrane micropump when operated with a high actuation frequency of 4 kHz. A new product brought to the market by Debiotech is the so-called Nanopump™ which provides a constant flow rate up to 20 kPa backpressure by means of a double limiter concept [26, 27]. It also utilizes piezoelectric actuation and a reciprocating diaphragm (Figure 1-4 (a)). (a) (b) Figure 1-4: The Nanopump™ by Debiotech [27] uses piezoelectric actuation and provides a stable flow rate up to a backpressure of 20 kPa (a). A pneumatic twostage micropump concept has been proposed by Berg et al. [12] (b). A two-stage peristaltic micropump has been presented by Berg et al. [12]. This micropump depicted in Figure 1-4 (b) is made of polydimethylsiloxane (PDMS) and utilizes pneumatic actuation for the fluid displacement. It exposes a pressure dependent flow rate and a maximum backpressure of less than 5 kPa. 4 1 Introduction 1.2.2 Actuation principles This chapter will provide a comprehensive overview of the actuation principles prevailing in MEMS technology. Due to the choice of piezoelectric actuation for the developed micropump this mechanism will be introduced in some more detail followed by thermal concepts. This category includes the paraffin actuators discussed as alternative actuation concept within this thesis. 1.2.2.1 Piezoelectric actuation The inverse piezoelectric effect is the employed mechanism for piezoelectric actuators. Here, an electrical field is applied across the piezoelectric material which causes a geometrical deformation e.g. a contraction or shear strain. Due to their high mechanical energy density piezoelectric actuators excel by sustaining to high mechanical loads (up to several 10000 N). The short response times enable a wide range of operating frequencies. A detailed review about piezoelectric actuation has been presented by Niezrecki et al. in 2001 [44]. Since the piezoelectric effect is a pure solid-state effect these actuators do not suffer from wear. In consequence, piezo-actuators preserve a constant performance over a long period of time (billions of cycles) [45]. Considering the reliability of piezo-actuators, a study investigating both DC and AC degradation mechanisms has been presented by Pertsch et al. [46]. The main reasons for failure are found to be dielectric breakdown or mechanical failure. The malfunction is often caused by the assembly of the actuator, e.g. failure of the electrical connection or the mechanical mount, rather than by failure of the piezoelectric material itself. For DC-operation under exposure to high humidity a dielectric breakdown can also arise from electrolytic degradation of the electrode and subsequent growth of conductive dendrites. The shape deformation of a piezoelectric actuator is proportional to the applied voltage and enables a controlled displacement with a high resolution. Nevertheless, high actuation voltages in the range of 150 to 350 V are typically required and the achieved displacement is only in the range of 0.1 - 0.2 % for piezo-ceramics. Here, multilayer actuators provide an option to increase the axial displacement along the main axis of the piezo-stack. Moreover, the thin layers of a multilayer actuator require only a reduced actuation voltage to achieve a comparable electrical field strength of 1 - 2 kV/mm. In contrast to electromagnetic or thermal principles, piezoelectric actuators exhibit a low power consumption. Since charging currents are the dominating energy drain, virtually no energy is consumed in DC-operation aside from small leakage currents. 1.2.2.2 Thermal actuation Thermal microactuators commonly consist of bimetallic structures, shape memory alloys or deformable cavities and membranes that rely on the expansion of a liquid or gas. Bimetallic structures utilize the non-uniform thermal expansion of two bonded materials to induce mechanical bending. Shape memory alloys rely on the reversible, temperature-activated phase transformation which coincides with a geometric shape deformation. This transformation effect occurs with metallic alloys such as NiTi or CuZnAl but is also observed with ceramic materials such as ZrO2 or even with polymers, e.g. PTFE [47]. Shape memory 5 1 Introduction alloy actuators provide a high energy density [48] and a comparably short response time in the ms-range unlike other thermal actuation principles [49]. Another class of thermal actuators relies on the expansion of gases or liquids upon the influx of heat. A typical design comprises a cavity containing a volume of fluid, with a thin membrane as one wall. Current running through a heater causes the working fluid in the cavity to expand which then deforms the membrane. Thermopneumatic approaches utilize the large expansion of gases and have been investigated by many researcher e.g. for the realization of micropumps [50, 51] or microvalves [52]. Other principles are based on the evaporation of a working liquid to generate an expansion force, e.g. for bubble-driven micropumps [53], dispensers [54] or bubble-jet printheads. To a lesser extent, concepts have been reported to explore the phase transition between the solid and the liquid state. Thermally actuated devices can develop relatively large forces, particularly if they rely on the solid-liquid phase transition. On the other hand, the heating elements consume large amounts of power and the response times are generally long since the actuator has to cool down to return to its original position. Here, the small dimensions of microactuators turn out to be favorable due to the reduced heat capacity and the faster heat dissipation into the surrounding structure. 1.2.2.3 Further actuation mechanisms Besides piezoelectric actuators the electrostatic attraction is the most commonly used actuation mechanism in MEMS devices [55]. This actuation principle uses the attractive force between oppositely charged electrodes as observed in a parallel plate capacitor. Electrostatic comb-drives consisting of interdigitated electrode fingers are a typical implementation of this actuation type [56-58]. The seamless integration into surfacemicromachined devices makes this design appealing for both sensors and actuators. In particular, the comb-drive design generates a comparably large actuation force since it induces a large change of the capacitance with electrode displacement. Nonetheless, the actuation forces of electrostatic actuators are generally well below those of piezoelectric actuators. It should be noted that electrostatic actuators can generate attractive forces only. The applicable voltage range is limited by the electrostatic pull-in voltage. All in all, the specific advantages of electrostatic actuators such as larger displacement distances or CMOS-compatible integration are confronted with a lack of linearity, precision and force strength compared to piezoelectric actuators. Electromagnetic actuation constitutes a further actuation principle frequently used in MEMS technology [59]. The advantages of this actuation type comprise large, long-range actuation forces, high energy densities and low operation voltages. The main concern with electromagnetic actuation is the fabrication process. Miniaturized conventional coils are still rather bulky and their three-dimensional structure is not compatible with the typical MEMS processes. Thus, the focus of many researchers working on electromagnetic microsystems is on the integration of planar coils into MEMS devices [60, 61]. Another inherent drawback of this actuation mechanism is the comparable large current which is even required during stationary holding states. This results in a higher overall power consumption compared to piezoelectric or electrostatic actuation. 6 1 Introduction Beyond that, numerous other actuation principles have been explored for MEMS devices including pneumatical, optical or chemical actuators [62] as well as electroactive polymer actuators [63]. In microfluidics, capillary forces, electrowetting or swelling of hydrogels are further reported actuation concepts. 1.2.3 Drug delivery Among the various methods of drug administration the release of a pharmaceutical substance by means of a pill as well as the delivery via infusion or transdermal injection are widely used. Further common routes of administration comprise the release from polymeric implants, transmucosal administration as well as inhalation. Figure 1-5 presents a schematic overview of the available routes of administration and indicates the main research directions affected by microsystem technology. Figure 1-5: Overview of the importance of microsystem technology for different routes of drug administration. The efforts associated with microsystem technology focus predominantly on improvements in the field of painless injections, automated infusion systems and implantable solutions [64]. Here, numerous approaches have been reported in recent years yielding a great diversity of applied principles and proposed systems. Summarizing the main intention of these approaches, the realization of individual dosing profiles, the site-specific drug release and the increase of the patient compliance are clearly among the first ranked objectives. Especially for implantable solutions, the biocompatibility of the materials and the safety and reliability of the system are vital aspects that compel researches to provide comprehensive studies and thorough testing results. 7 1 Introduction 1.2.3.1 Diffusion-based systems Implantable release systems commonly rely on a biodegradable or sometimes nondegradable polymer matrix hosting the embedded pharmaceutical agent [65]. This carrier is implanted either underneath the skin or close to the target site in order to achieve a sitespecific drug delivery. The release mechanism is based on diffusion and the dosing rate is determined by the properties of the polymer carrier or an enclosing semi-permeable membrane. In recent years, nanotechnology paves the way towards nanoporous membranes and drug delivery by means of nanoparticles, nanospheres or nanocapsules [66, 67]. These nanovectors offer the capability of biomolecular targeting which enables an advanced sitespecific drug release. As an example, the controlled release of doxorubicin from gelatin nanoparticles has been investigated by Leo et al. [68]. The site-specific release of this anticancer agent promises to reduce the toxicity and undesired side effects. A detailed review on the potential of nanotechnology to the field of oncology is given by Ferrari [69]. As an example, the high uniformity of microfabricated nanopores enables the reproducible fabrication of controlled release interfaces [70]. While passive, diffusion-based systems are generally limited to a preset dosing rate, active devices enable a controlled, time-modulated release which enhances the efficacy of the treatment for many diseases. For example, a release profile that mimics the circadian rhythm of the patient (chronotherapy) is considered advantageous for insulin dosing and enables innovative methods in cancer treatment. A silicon microchip for a site-specific release of pharmaceutical agents has been developed by Santini et al. [71]. In his approach, a microstructured silicon chip contains an array with a large number of small reservoirs sealed individually by a thin gold membrane (Figure 1-6). When immersed in a physiological solution the gold anode membrane is dissolve electrochemically and the chemical substance is released from the reservoir by means of diffusion. Since each reservoir is individually addressable the device enables a pulsatile delivery of single or multiple substances with a resolution in the picoliter range. Figure 1-6: Silicon drug release microchip with individually addressable reservoirs [71]. Meanwhile, this development has been commercialized [72] and has been further enhanced by covering the reservoirs with a resorbable polymer membrane [73]. This concept is promising due to its flexible and controlled release mechanism providing a high resolution but 8 1 Introduction it is naturally limited to applications where a pulsatile release is acceptable. Also, the included total volume has to be sufficient for a prolonged period of treatment since the device is not refillable. 1.2.3.2 Transdermal injections Another research direction of drug delivery supported by microsystem technology is the transdermal injection via microneedles. Several research groups have been working in that field focusing on microneedles made of silicon [74, 75], titanium [76] or polymers [77]. Depending on the target application these microneedles typically feature either electrodes, e.g. for neural stimulation, or fluidic channels for drug release. An active drug delivery probe which combines both features for simultaneous stimulation and in vivo drug delivery has been proposed by Papegeorgiou et al. [78]. Microneedles provide a painless method for drug delivery in applications where the transdermal release is the intended route of administration. However, a microneedle itself constitutes only a component of a drug delivery system and has to be combined with a fluidic driving unit e.g. an external pump or a pressurized reservoir. 1.2.3.3 Active infusion systems The state of the art in long-term treatment are still external pump systems [79, 80] which are connected to the patient by means of a passive port system or a subcutaneous catheter. Generally, these devices contain a miniaturized pump that is programmable to realize patient-specific modulated release patterns. While some of these systems are still rather bulky others have shrunken in size to improve their portability. Many of the commercial infusion pumps are optimized for insulin dosing due to the high potential of the diabetes market [29, 81, 82]. Up to now, implantable infusion pumps are predominantly constant flow systems based on osmotic pressure [65] or gas pressure. As an example, Codman [28] offers an implantable system based on a pressurized reservoir and a passive flow restrictor (Figure 1-7 (a)). A water-powered osmotic micropump based on soft micromachining with PDMS has been presented by Su et al. [28]. A clear advantage of these systems is that they do not consume electrical power. On the other hand, in order to gain a functional advantage over diffusion based systems a time-modulated release of the pharmaceutical substance is of high priority. A modified gas pressure pump with a controllable release profile by implementation of an actuated throttle has been presented by Tricumed [83]. A similar approach utilizing active valves has been published by Götsche et al. [32]. A microvalve-regulated system using a micro-spring pressurized balloon reservoir has been reported by Evans et al. [84]. Medtronic [29] offers a commercial infusion system which is programmable but still rather bulky. It is realized by miniaturization of a conventional peristaltic actuation principle, i.e. periodic squeezing of a flexible tube. Eksigent [31] has been working on an electrokinetically driven micropump to achieve highly accurate dosing at extremely low flow rates. An implantable infusion system is announced to be currently under development by Debiotech [26, 27]. It is based on the Nanopump™ as shown in section 1.2.1 and offers promising features such as a freely programmable release profile and an adjustable, backpressure independent flow 9 1 Introduction rate. The proposed design requires a rather complex fabrication process including sophisticated structures for a mechanical double limiter of the pump membrane. A different approach for active drug release based on a novel dispensing unit has been presented by Hu et al. [54]. This concept incorporates the bubble-jet principle extended by a hydrophobic air chamber and provides the capability of discrete chemical release in conjunction with a short response time and absence of leakage. (a) (b) Figure 1-7: Implantable constant flow infusion pump by Codman [28] (a) and miniaturized peristaltic pump with variable flow rates by Medtronic [29] (b). 1.2.3.4 Adaptive systems For a continuous treatment over a prolonged period of time implantable concepts are appealing in terms of patient compliance and in order to avoid inflammation often occurring at the transcutaneous access point. In the ideal case, the patient would retain the uppermost flexibility regarding habitual daily activities since the system automatically monitors vital parameters and controls the appropriate release of the required drug. However, the autonomous and adaptive drug delivery system is still a vision today that requires substantial improvements of the functionality and reliability of existing systems. Nonetheless, several concepts addressing the issue of self-regulated systems have been published yet [85, 86]. Essentially, the pursued strategies can be subdivided into two categories. The first class comprises systems where a biosensor is coupled with the active release mechanism in a feedback loop. Here, a biosensor continuously monitors the concentration level of the target molecule. The measurement data are evaluated by an integrated control unit and the release rate is adjusted accordingly. As an example, the development of a “Smart Pill” has been announced by ChipRx [87]. In this approach the release mechanism is based on an artificial muscle made of a soft, gel-like polymer which responds to electrical charging. Upon stimulation the muscle contracts and clears the microscopic holes of the capsule which enables the diffusive release of the substance. An adaptive release of drugs from an artificial tooth is currently under development in the framework of a research project called “IntelliDrug” [88]. Its specific merit is expected in the treatment of drug addiction and chronic diseases. The second category directly utilizes the response of certain materials to ambient changes. The involved materials, often referred to as smart materials, exhibit the ability to adaptively 10 1 Introduction change their properties or their shape. For example, a pH-sensitive membrane is able to change its permeability which provides a means to regulate the diffusion of drugs out of a capsule. Hydrogels are generally well suited for adaptive approaches since they are susceptible to various stimuli. Their amphiphilic nature enables extensive swelling when exposed to an aqueous environment. A concept for a self-regulated insulin delivery system based on the glucose-sensitive swelling of hydrogel has been proposed by Ziaie et al. [79]. Here, the insulin is released from a pressurized reservoir and the hydrogel swelling is utilized to control a microvalve which then adjusts the delivery rate (Figure 1-8). The degree of swelling has been proven to be dependent on the glucose concentration in the surrounding medium. Figure 1-8: Concept of a glucose-sensitive microvalve [79]. All in all, the promising potential of adaptive release systems is unquestionable and investigations of different concepts are on the way. Nonetheless, virtually all of these systems have not reached a mature state yet and substantial improvements are still awaited considering both material research and system design. The main concern about most of the responsive materials is the reproducibility of their function and the stability of their performance. Once these problems are eliminated the systems are expected to unfold their full potential and to enable novel therapies with an improved efficacy. It is, however, still a long way to go until this ambitious vision may be achieved. 1.3 Objective of this thesis Throughout the last two decades, microsystem technology has been revealed as a promising research direction in the field of controlled drug release. This aspiring technology offers the potential to realize integrated systems with high functionality and superior performance. It enables the realization of accurate, compact and power-efficient drug delivery systems. The availability of those systems is desirable to increase patient safety and comfort in long-term treatment, especially for therapies that rely on a continuous administration of drugs. Nevertheless, automated drug delivery systems based on micropumps are still in a premature state and significant improvements in this field are awaited for the near future. The work presented in this thesis is dedicated to the development of a new silicon micropump optimized for the precise dosing of aqueous drug solutions and its integration into an automated system. To ensure the competitiveness of the concept the developed delivery system has to meet the following specifications: 11 1 Introduction • Delivery rate: 10 – 1000 µl/h • Freely programmable dosing sequences • Stable flow rate up to a backpressure of 30 kPa • System size: approx. 5 x 5 x 2 cm • Weight: max. 35 g • Low power consumption • Expected life time: min. 0.5 year • Guaranteed patient safety As a starting point, this thesis summarizes a selection of fundamental aspects in chapter 2 covering the field of microfluidics and the mechanics of a piezoelectric bimorph actuator. In chapter 3 the novel design of the proposed two-stage micropump will be introduced first and the working principle will be explained. A detailed lumped parameter model will be developed for the sake of a detailed understanding of the physical background and for an optimization of the actuation control scheme. This chapter will also deal with the problem of an alternate gas and liquid transport and will show a design variation featuring an enlarged single membrane. The following chapter 4 will briefly outline the fabrication process. A comprehensive presentation of the experimental results and a discussion of the different phenomena will be given in chapter 5 and 6. Chapter 7 will report about a feasibility study including preliminary experiments for the actuation of the micropump with a paraffin phasechange actuator. Subsequently, chapter 8 will describe new findings in the field of soft polymer technology based on polyurethane. The “Active Microport” - project will be presented in chapter 9 including demonstrative experiments of the achieved system performance and first in-vivo tests in the context of animal experiments. Finally, the thesis will be concluded by a summary and a brief outlook. 12 Chapter 2 Fundamentals 2 Fundamentals The development of a micropump together with the subsequent well-founded analysis of its performance relies on the awareness of the underlying physical principles. This chapter summarizes the respective issues as a fundament for the scientific engineering process. The first part of this chapter will provide a comprehensive background in microfluidics. Apparently the precise dosing of fluids requires the consideration of relevant effects such as the capillary phenomenon. A basic knowledge is also essential to understand the principles of fluidic simulations and to develop a lumped parameter model of the micropump. The second substantial issue associated with the design of reciprocating micropumps is the displacement of fluid by bending membranes. Here, a strong background in structural mechanics is inevitable to establish an analytical model of the membrane deflection due to pressure load or piezoelectric actuation. The second part of this chapter outlines the governing equations and covers the mathematical derivations of design-specific solutions. 2.1 Microfluidics This chapter on microfluidics commences with a recapitulation of the basic properties of fluids. Subsequently, relevant dimensionless numbers are identified and the basic flow equations applying to the laminar flow regime are summarized. The principle of liquid wetting is examined for different solid materials and the implication of the capillary effect for microfluidic applications is discussed. 2.1.1 Fluid mechanics 2.1.1.1 Density and viscosity The density of a fluid is an intensive property that quantifies the mass confined in a unit volume and is calculated as . (2.1) 13 2 Fundamentals For liquids, the density is susceptible to temperature changes, but virtually independent of external forces - a property referred to as incompressibility. In contrast, the density of gases is variable in response to a change of the external pressure. Since gases fully occupy the available volume, a temperature increase does not necessarily change the density, for example, when the available volume remains constant. For an ideal gas the relation between pressure p, volume V and temperature T is constituted by the perfect gas law · · · (2.2) where R is the universal gas constant (R = 8.314 J K-1 mol-1) and n is the amount of gas given in mol. The viscosity accounts for the inner friction of a fluid which arises from molecular forces and collisions. Between adjacent fluid layers moving with different velocities, a dissipative frictional force occurs which is proportional to the viscosity of the fluid. This way, the shear stress τ between adjacent layers and the velocity gradient are related by · . (2.3) Here, the x-direction is supposed to be the direction of flow with the flow velocity ux which leads to a velocity gradient in y-direction (Figure 2-1). For so-called Newtonian fluids the viscosity is independent of the shear stress and therefore the shear stress is proportional to the velocity gradient in accordance to equation (2.3). In consequence, this equation describes the continuous deformation of a fluid volume in response to shear forces and is the backbone for the determination of the flow profile in microchannels (see section 2.1.1.5 on Stokes flow) [89, 90]. Figure 2-1: Shear stress τyx on a fluid confined between two plates and resulting velocity gradient ux(y). The viscosity of fluids is influenced by temperature. For liquids, the viscosity decreases for rising temperatures whereas gases exhibit an opposite behavior with an increasing viscosity at higher temperatures. The viscosity of water at 20°C is η = 1·10-3 Pa s. Some fluids exhibit a viscosity that depends on the shear stress or even on the duration of the applied stress. This class is termed non-Newtonian fluids and exhibits a more complex rheology. A prominent example of a non-Newtonian fluid is oil. Moreover, a constant viscosity cannot be assigned to a gas at very low pressures where the mean free path length λm is in the same range as a typical geometric dimension L of the available volume, for example, the diameter of a microchannel. Here, the Knudsen number 14 2 Fundamentals (2.4) separates different flow regimes which needs to be modeled appropriately. For Kn < 0.1 the fluid is considered as a continuum with an associated constant viscosity η. In this regime the flow is described by the Navier-Stokes-Equation (section 2.1.1.4) and no-slip boundary conditions are generally applied. Despite the small dimensions of microchannels most microfluidic systems fall into the continuous flow regime. This holds true even for gases at standard pressure where the expected mean free path length is in the range of 60 nm [91]. For larger Knudsen numbers, the displacement of individual molecules has to be taken into account by means of appropriate modeling techniques. The interested reader is referred to related textbooks dealing with molecular dynamics for further information on this topic [92]. Sometimes it is useful to relate the viscosity of a fluid to its density which yields another quantity referred to as kinematic viscosity . (2.5) 2.1.1.2 Continuum equation For a given volume element in the fluidic domain the influx and efflux must obey the general principle of conservation of mass. In consequence, the mass flowing into a so-called control volume during a time period Δt can only diverge from the outflow if the density within the control volume changes. In other words, for an incompressible fluid both influx and efflux have to be balanced at each instant of time. This Eulerian approach yields the continuity equation · 0 (2.6) where u is the velocity field at the location of the infinitesimal control volume dV. For an incompressible fluid, the density is constant in time as well as in space and equation (2.6) simplifies to 0 (2.7) which constitutes, that the divergence of the velocity field is zero for an incompressible fluid. 2.1.1.3 Mach number The Mach number (2.8) is a dimensionless number that relates the flow velocity u to the local velocity of sound c. Since the velocity of sound is a function of the compressibility of the fluid, the Mach number 15 2 Fundamentals is an indicator whether a medium can be treated as an incompressible fluid. As a general rule of thumb, a fluidic system can be considered as incompressible for Ma < 0.3 [89, 92]. 2.1.1.4 Laminar regime and Navier-Stokes-equation In microfluidic systems the laminar regime is predominant. Here, the streamlines of fluid flow are aligned in a parallel fashion and the initiation of vortexes is inhibited. Due to the frictional losses within the fluidic system, induced turbulences die out rapidly. The laminar regime is characterized by small Reynolds numbers · (2.9) where u denotes the velocity of fluid flow, l is a characteristic length of the geometry and ν quantifies the kinematic viscosity of the fluid. For a given fluidic system, the Reynolds number basically relates the impact of inertia to the relevance of friction. At low Reynolds numbers, which corresponds to the laminar regime, the impact of friction prevails. A critical Reynolds number describes the transition point to the turbulent regime. It depends on the geometry as well as on the material and the surface properties. For a sphere moving within a liquid basin of infinite size, the critical Reynolds number is approximately 2300. For microfluidic systems, the expected Reynolds numbers are Re < 100 and often even Re < 1 [92, 93]. The laminar flow of viscous fluids is described by the Navier-Stokes-Equation. It arises from the Eulerian model of a continuous fluid subdividing its volume into an infinite number of control volumes. For each control volume, the conservation laws regarding mass, energy and momentum have to be fulfilled at any time. The simplified form of the Navier-Stokes-Equation for incompressible, Newtonian fluids (ρ = constant, ν = constant) · · ·∆ · (2.10) is a fundamental equation in microfluidics. Here, u denotes the velocity vector, p is the external pressure load and g is the constant of gravity. The left hand side is constituted by two terms, the instationary and the convective term, accounting for the local timedependence of the velocity and the change of the velocity due to convection, respectively. The right hand side sums up the impinging force densities which are in general the pressure gradient, the frictional and the gravitational force density. In case of other relevant volume forces, e.g. centrifugal forces, buoyancy or electroosmosis, the force density side is extended with respective describing terms. 2.1.1.5 Stokes flow For certain boundary conditions and flow characteristics a simplified Navier-Stokes-Equation appropriately describes the flow profile for the specific situation. One common example called Couette flow or shear-driven flow describes the flow profile induced by a moving side 16 2 Fundamentals wall. Due to the viscosity of the fluid, a constant velocity of the moving wall transfers a momentum to the fluid and leads to the development of a steady flow profile. For the fluidic modeling of the micropump later in this work the case of pressure-driven laminar flow is more relevant. In microfluidic systems the negligence of gravitational effects is generally acceptable. Moreover, the inertia related terms on the left-hand side of the NavierStokes-Equation (2.10) become zero in case of a stationary flow profile with parallel streamlines. For this situation the Navier-Stokes-Equation simplifies to ·∆ (2.11) which is referred to as Stokes flow. The solution of this Stokes-Equation yields a parabolic flow profile (Figure 2-2). The velocity ux(r) of a pressure-driven flow in x-direction within a cylindrical tube of radius R and length l reads ∆ · 4 . (2.12) Figure 2-2: Parabolic flow profile of pressure driven laminar flow in a tube. Obviously, the maximum velocity is found at the center of the tube whereas the velocity at the wall is zero in agreement to the no-slip boundary condition. The mean velocity is obtained by integration of equation (2.12) over the flow cross section leading to the volumetric flow rate · 8 ·∆ . (2.13) Thus, the flow rate Iv of a pressure-driven flow within a cylindrical tube scales with R4 which is well-known as law of Hagen-Poiseuille. In analogy to electrical circuits, the corresponding fluidic resistance of the tube is expressed as ∆ 8 . (2.14) The Stokes equation (2.11) solved for the parabolic flow profile within a small gap of height h but large width w leads to the volumetric flow rate · 12 ·∆ . (2.15) 17 2 Fundamentals This equation is easily rearranged to obtain the mean flow velocity 12 (2.16) ·∆ and the corresponding fluidic resistance ∆ 12 . (2.17) In this thesis, the assumption of Stokes flow in a small gap will be applied to determine the fluidic resistance for the specific valve geometry of the two-stage micropump (see chapter 3.2). 2.1.2 Wettability Wettability describes the tendency of a liquid to spread on the surface of a solid. From the thermodynamic point of view, it is an energy-driven phenomenon caused by the demand of a system to minimize its total free energy. In microfluidic systems wetting related effects play an important role in the behavior of fluids. Here, the large surface areas and the reduced impact of volume forces such as the gravitational force explain the prevalence of capillary effects in small microchannels. Since water is the most prominent liquid for microfluidic applications the terms “hydrophilic” and “hydrophobic” are widely used to describe the property of a surface to promote or reject wetting by water, respectively. 2.1.2.1 Surface tension and interfacial energy The physical reason behind the existence of interfacial tensions is the preference of molecules to be surrounded by their own species. For energetic reasons a position along the interface of different phases is less favorable and therefore a net force pointing inwardly acts on the molecules along the boundary layer. These effects are summarized in a quantity termed as interfacial tension which indicates the amount of energy necessary to extend the interfacial area by one unit area. The interfacial tension of a liquid phase against a gaseous phase is commonly referred to as surface tension of the liquid. It is calculated by the change of Gibbs surface free energy G with respect to the surface area A at constant temperature and constant pressure , . (2.18) Due to the comparably small interaction forces between the molecules in the gaseous phase the surface tension of a liquid is nearly independent of the actual composition of the gaseous phase. In order to reach an equilibrium state a system consisting of different phases pursues to change the interfacial areas to minimize its Gibbs interfacial free energy. In general, a solid surface promotes wetting by a liquid agent if its solid-gas interfacial tension is significantly larger than the surface tension of the liquid. In this case it is advantageous for the system to 18 2 Fundamentals reduce the area of the solid-gas interface at the cost of increasing the surface area of the liquid. Precisely speaking the solid-liquid interfacial energy has also to be taken into account but for a rough estimation of the wetting properties a comparison of the solid surface tension to the surface tension of the liquid is sufficient. Figure 2-3: Surface tension of common polymers and water [94]. Figure 2-3 gives an overview of the surface tensions of different polymer materials and illustrates the impact regarding the wettability by water. Water itself features a surface tension of σ = 72.5 mJ/m2 at room temperature T = 293 K which is factor 1.5 to 2 larger than the surface tensions of the listed polymers. Therefore, these polymers do not particularly promote wetting by water, but only for PTFE and PDMS the discrepancy between the surface tensions is large enough to observe strictly hydrophobic behavior. Different treatments exist to modify the wetting behavior of a solid-liquid system. On the one hand, wetting is supported by lowering the surface tension of the liquid. This is achieved by adding wetting agents such as surfactants to the liquid phase. On the other hand, the surface energy of the solid could be modified. Here, chemical approaches such as coating layers as well as physical methods including plasma activation or microstructuring of the surface have been report and will be discussed later in chapter 8. 2.1.2.2 Contact angle and Young’s equation The contact angle between the meniscus of a dispensed liquid droplet and the solid surface is a measure for the wettability of the surface with respect to the applied liquid. The contact angle θ is determined by Young’s equation · (2.19) which describes an equilibrium state where the involved surface and interfacial tensions are balanced. Here, σsg denotes the solid-gas interfacial tension (which equals the solid surface tension by approximation [94]), σsl is the solid-liquid interfacial tension and σlg represents the surface tension of the liquid. A common derivation of the Young equation is obtained from the mechanical equilibrium of the forces per unit length acting on the three-phase contact line 19 2 Fundamentals of a droplet (Figure 2-4). Alternatively, the Young equation can be deduced from the energy balance of the interfacial energies which is a more general approach [95, 96]. A small contact angle corresponds to an excellent wettability (e.g. ethanol on SiO2), whereas a contact angle larger than 90° indicates hydrophobic behavior (e.g. θ = 108° for water on PTFE). Figure 2-4: Illustration of the contact angle established by the surface tensions of a three-phase system. 2.1.2.3 Contact angle hysteresis The equilibrium contact angle determined by the Young’s equation is a consequence of the attempt to minimize the free energy of the system which is a strictly thermodynamic interpretation of the wetting problem. On real surfaces other properties such as the surface roughness or heterogeneity also affect the wettability and may avoid that the system reaches its equilibrium state predicted by the Young’s equation. A rough surface obviously has an increased effective surface compared to its geometric area. From the energetic point of view, this increased effective surface area along both the solid-gas and the solid-liquid interface shifts the equilibrium angle determined by the Young’s equation. Wenzel accounted for this impact by introducing a surface roughness factor [97] r effective surface geometric surface (2.20) which enforces a modified equilibrium contact angle θw (Wenzel angle) given by · · . (2.21) For a surface with a Young angle θ < 90° the roughness renders it even more hydrophilic since r is always greater than unity. On the other hand, a smooth hydrophobic surface becomes even more liquid repelling when its roughness is increased. Surface roughness not only shifts the equilibrium contact angle but also leads to a deviation between advancing and receding contact angle. On a microscopic level a rough surface appears very irregular which likely prevents the system from reaching the equilibrium angle. In terms of energy a rough surface introduces local minima and hence a range of possible contact angles is observable on a real surface. This range is bounded by the advancing angle to the upper end and the receding angle to the lower end. Therefore, a single measurement of a static contact angle on a real surface is not very meaningful since one might observe any of the possible states between the receding and the advancing contact angle. 20 2 Fundamentals Figure 2-5: Theoretical behavior of advancing and receding contact angles as a function of roughness for θ < 90° (a) and θ > 90° (b) [94]. Figure 2-5 illustrates the change of the advancing and receding contact angle with respect to surface roughness. Departing from the ideal smooth surface the advancing angle increases with roughness whereas the receding angle diminishes. Further on, the range of possible contact angles shifts in accordance to the Wenzel prediction. Thus, for a hydrophilic surface both receding and advancing angle approach zero with the receding angle decreasing more rapidly. On a hydrophobic surface the receding angle approaches the advancing angle for high roughness values, i.e. only a small hysteresis is observed on extremely rough substrates. This behavior is caused by the capillary effect that comes into play on highly rough surfaces. The liquid does not creep into small cavities of the rough hydrophobic surface which leads to buried air pockets. This way, the liquid droplet faces a composite interface made of solid regions and air pockets which leads to the well-known effect of superhydrophobicity. Due to the large receding angle a droplet can easily roll off a superhydrophobic surface. Wetting of such an inhomogeneous surface with a droplet suspended on top of the rough surface features has been first described analytically by the theory of Cassie et al. [98]. A more detailed coverage of theories dealing with heterogeneous surface compositions is found in related textbooks [94, 99]. 2.1.2.4 Kinetic phenomenon Polymer surfaces are able to change their properties in response to the surrounding environment. Even in the solid state the polymer chains preserve some mobility. Depending on the polymer characteristics such as the mean chain length, the degree of crosslinking or the interaction of attached side groups, some polymers show a more pronounced dynamic behavior than others. As an extreme example, hydrogels such as PHEMA are able to invert their surface properties from hydrophobic to hydrophilic when surrounded by water. This change is the result of a transport process where the hydrophilic hydroxyl groups are moved towards the surface which is favorable in order to minimize the interfacial free energy of the system. 21 2 Fundamentals The time scale of these processes is comparable to typical experiment times. Hence, for experiments investigating the wetting effect on polymer surfaces with pronounced dynamic reorientation of bulk molecules or functional groups, the elapsed time has to be considered as a relevant parameter. This dynamic behavior of polymer surfaces plays a particular role in the field of surface treatment. As an example, polymer surfaces rendered hydrophilic by means of plasma activation eventually recover their hydrophobic characteristic, an effect that is known as ageing. 2.1.2.5 Wetting of silicon surfaces For silicon surfaces a modification of the wetting behavior is achieved by adsorption of molecules, by oxidation of the pure silicon surface or by controlled deposition of a functional layer e.g. silanization. Here, kinetic changes are less pronounced since the self-initiated processes such as adsorption or oxidation occur on a time scale of hours which is typically beyond the experimental time frame. Moreover, surface roughness aspects might be neglected when dealing with polished silicon wafers. For microsystems based on silicon wafers the contact angle of the pure bulk substrate as well as the contact angle observed on a silicon dioxide surface (SiO2) is of great concern in order to describe the fluidic effects of these systems. A pure silicon surface is achieved by cleaning with hydrofluoric acid (HF) and a contact angle of 70° has been reported for these conditions [100]. For silicon microsystems the property of a SiO2-surface is most relevant since a pure silicon surface is covered by a natural oxide layer within a few hours when exposed to atmospheric conditions. For both a natural oxide as well as a silicon surface oxidized in a wet atmosphere at 1100°C a contact angle of 43° has been reported by Hermansson et al. [100]. In discrepancy to these data and the hydrophilic nature of a SiO2surface, Salay et al. presented experimental contact angle measurements on a respective surface with an advancing angle of 87° and a receding angle of 68° [101]. In addition to this information given in the literature, experimental contact angle measurements have been performed in the framework of this thesis. For the bare silicon surface treated with a HF-clean a mean static contact angle of 68° was measured. A subsequent Caro clean (sulfuric-peroxide mixture, H2SO4+H2O2) reduced the contact angle significantly and the measurement yielded an angle of about 25°. For a silicon substrate covered by an 400 nm oxide layer an aging effect was observed and a hydrophilic contact angle of approximately 45° has been obtained eventually (see Appendix A). 2.1.2.6 Capillary effect and Young-Laplace pressure drop Wetting of a surface leads to the capillary effect when the liquid is confined in a small gap or channel. The liquid-gas interface exhibits a curved shape and the liquid meniscus hits each wall with the contact angle determined in equation (2.19). Due to the surface tension, the pressure at the concave side of the interface is elevated with respect to the pressure at the convex side. This so-called Young-Laplace pressure drop 22 2 Fundamentals ∆ 1 1 (2.22) arises across a curved liquid-gas interface and generates the capillary force which displaces the meniscus within capillary channels. Here, σ is the surface tension of the liquid and r1 and r2 denote the radii of curvature. For small contact angles below 90° an advancing force is generated and capillary filling becomes feasible. For large contact angles above 90° a receding force induces a depression of the liquid surface in the capillary. Both cases are illustrated in Figure 2-6 where a liquid droplet is confined between two cylindrical plates. In the case of a wetting liquid, a concave gas-liquid interface is established and the corresponding Young-Laplace pressure drop leads to a decreased pressure in the liquid plug. For a non-wetting liquid a converse behavior is observed and the convex curvature of the meniscus leads to an increased pressure in the liquid droplet. (a) (b) (c) Figure 2-6: Capillary effect in a small gap between two cylindrical plates (a): curvature of the meniscus of a confined droplet in case of wetting (b) and nonwetting (c). For a narrow gap with a small height h but a large width, the strong curvature across the narrow gap is predominant and equation (2.22) simplifies for a known contact angle θ to ∆ · cos θ (2.23) The negative pressure induced in the liquid droplet in Figure 2-6 (a) is the reason behind the well-known effect that two plates stick to each other if a droplet of water is spread between the two of them. This phenomenon is also relevant to microsystems where capillary pressure drops can induce sticking of deflectable microstructures such as cantilevers or membranes. 2.1.2.7 Weber number and capillary number The relevance of the capillary force for the dynamics of a multi-phase fluidic system is described by the Weber number We · · (2.24) 23 2 Fundamentals which relates the inertial force to the capillary force. Here, ρ is the density of the fluid, u denotes the flow velocity and l represents a characteristic length of the fluidic system, e.g. the hydraulic diameter. The Weber number scales proportional to the size of the system ( ) which yields small Weber numbers for microfluidic systems and results in a prevalence of the capillary forces compared to inertial forces. Combining both the Reynolds number and the Weber number leads to another dimensionless number that relates the frictional force attributed to the viscosity of the fluid to the surface tension. In contrast to the Weber number, the capillary number Ca · (2.25) does not depend on the size of the system. It indicates whether the viscosity η or the surface tension σ is predominant for the flow characteristic of the system at hand. For aqueous liquids running through a microchannel with a typical velocity in the range of 10-5 to 0.1 m/s the capillary numbers are in the range of 10-8 to 10-4 [102]. 2.2 Piezoelectric membrane actuators In MEMS technology the piezoelectric effect is frequently used to realize active devices. The piezoelectric material serves as transducer that converts an electric voltage into a mechanic displacement. For reciprocating micropumps the piezoelectric deformation provokes a deflection of the membrane by means of the bimorph effect. At first, this chapter gives a general introduction to the piezoelectric effect with a focus on the physical reason behind this phenomenon. It also outlines the fundamental equations to analytically describe the piezoelectric effect. Thereafter, a necessary foundation in structural mechanics is provided in order to derive constitutive equations for the membrane deflection due to pressure load and piezoelectric actuation. 2.2.1 Piezoelectric effect The piezoelectric effect was first discovered by Jacques and Pierre Curie in 1880. They explored the ability of some single crystal materials such as quartz, gallium phosphate or tourmaline to generate an electric surface charge in response to shear stresses or squeezing. Besides the single crystal materials several ceramics also exhibit a piezoelectric effect. Among those the lead-zirkonate-titanate ceramic (PZT) is the most prominent example which is used for many commercial and scientific applications. The inverse piezoelectric effect is utilized for the design of piezoelectric actuators. Here, an electric field is applied across the piezoelectric material which causes a geometric deformation such as contraction or shear strain. Compared to single crystal materials piezoceramics exhibit a larger piezoelectric effect which makes this class of materials advantageous for applications where large mechanical deformations are required. 24 2 Fundamentals 2.2.1.1 Piezoelectric materials A non-centrosymmetric crystal structure of the material is a prerequisite for piezoelectric behavior. Among the piezoelectric materials two classes have to be distinguished, the single crystal materials such as quartz or tourmaline and the piezo-ceramics. Figure 2-7 illustrates the atomic background of the piezoelectric effect for quartz. Here, the asymmetrically distributed and oppositely charged ions are displaced with respect to each other in response to an external shape deformation. Consequently, a polarization vector arises across the material and surface charges are detected at the two electrodes. Figure 2-7: The hexagonal crystal structure of quartz generates a surface charge upon an applied mechanical stress. For piezo-ceramic materials the physical mechanism behind the piezoelectric behavior is different. These ferroelectric, polycrystalline ceramics are composed of numerous crystallites, each of them exhibiting a permanent electric dipole moment below the Curie temperature. The piezoelectric nature of the individual crystallites arises from the so-called Perovskite crystal structure (Figure 2-8 (a)). For energetic reasons the positively charged ion at the center of the lattice cell is slightly dislocated from its symmetry position which results in a permanent polarization. Initially, the dipole moments are aligned in a parallel fashion only within small domains (Figure 2-8 (b)). These single domains are randomly oriented and have to be polarized to obtain a macroscopic piezoelectric behavior. The polarization process requires a high field strength together with an elevated temperature close to the Curie temperature. After the polarization process the alignment of the domains is nearly preserved as long as the temperature remains below the Curie temperature (polarization hysteresis). Above the Curie temperature (typically 200 - 300 °C for common piezo-ceramic materials [103]) the crystal structure is transformed to a symmetric cubic lattice and hence the dipole moments of the crystallites vanish. 25 2 Fundamentals (b) (a) unpolarized polarized Figure 2-8: Crystal structure of a lead-zirkonate-titanate ceramic (PZT) with a non-centrosymmetric ion (Ti4+, Zr4+) in the middle of the cell (a). After polarization the dipoles of all domains are aligned in parallel and exhibit a macroscopic polarization (b). In the polarized state a shape deformation of the material changes the magnitude of the polarization vector which induces surfaces charges on the electrodes. Vice versa, an applied actuation voltage changes the polarization vector and leads to a shape deformation. If the applied electric field vector is aligned in parallel to the polarization vector an elongation in the direction of the polarization axis occurs. Once the polarization reaches its saturation value the elongation is terminated and a further increase of the voltage would destroy the piezoactuator as soon as the breakdown field strength is reached. Applying a negative voltage with an electrical field vector that opposes the polarization vector leads to a contraction of the material. This process will continue until the coercitive field strength of the hysteresis loop is reached (Figure 2-9 (a)). Here the polarization of domains in the material are reoriented in the opposite direction and a further increase of the negative voltage will result in an elongation again. The resulting displacement curve is depicted in Figure 2-9 (b) and is referred to as butterfly-curve due to its characteristic appearance. (a) (b) Figure 2-9: The polarization of a piezo-ceramic under the influence of an electric field exhibits a hysteresis loop (a). This hysteresis property leads to a characteristic “butterfly” curve for the strain of the piezoelectric material (b). 2.2.1.2 Actuation modes For piezoelectric actuators commonly three actuation modes are distinguished referring to the orientation of the polarization, the applied electric field and the achieved deformation. 26 2 Fundamentals The situation mentioned in the previous section describes the longitudinal actuation mode. Here, the polarization and the electrical field are aligned in parallel and the extension or contraction along the polarization axis is utilized. This geometric deformation along the polarization axis is always accompanied by a transverse deformation similar to the transverse strain for elastic deformation. Many applications in MEMS technology make use of this transverse effect. Typically the piezo-ceramic is mounted on a passive cantilever or membrane and the transverse piezoelectric mode causes a deflection of this so-called bimorph structure. The actuation of the micropump presented in this thesis relies on the transverse actuation mode and the correlation between applied voltage and membrane deflection is illustrated in Figure 2-10. (a) (b) Figure 2-10: Transverse actuation mode utilized for a piezoelectric membrane actuator. The third actuation mode is referred to as shear actuation mode. Here, the direction of the external electric field is perpendicular to the polarization of the material which leads to a shear deformation of the actuator. 2.2.1.3 Piezoelectric coefficients For the inverse piezoelectric effect the strain of the actuator is related to a mechanical stress σ and the applied electric field E by · · . (2.26) Due to the anisotropic nature of this effect the analytical description involves tensors of fourth order to account for the correlation between stress and strain as well as electric field strength and strain. The indices 1, 2 and 3 denote the spatial directions x, y and z, respectively. By convention, the direction of the polarization axis is usually the z-direction. In equation (2.26) the first part relates the strain to the mechanical stress by means of the stiffness tensor (given in units m2/N or Pa-1) at constant electric field strength E. For the directions of the principal axes this term simplifies to Hooke’s law and the stiffness coefficient is the inverse of the Young’s modulus for the specific direction. In addition to this mechanical impact the second term in equation (2.26) quantifies the superimposed piezoelectric strain induced by the electric field. The charge constant (given in units As/N or m/V) is a decisive criterion for the efficiency of the piezoelectric material and for most applications a high charge constant is desired. In many cases a simplification of equation (2.26) is feasible if the strain of interest is limited to one direction. For the transverse actuation mode utilized in this thesis the strain in the 27 2 Fundamentals transverse direction is of main interest. For a squared PZT actuator aligned in parallel to the principal axes in the xy-plane the shear stresses and shear strains vanish and the relevant strain in both x- and y-direction is obtained by [103] · · (2.27) . determines the strain which arises from the impact of the Note, that the charge constant electric field at a constant mechanical stress σ1. The linear superposition in equation (2.27) implies that both contributions are independent of each other. If the piezoelectric deformation induces mechanical stress due to boundary conditions such as clamping this in turn affects the polarization and has to be considered by an equation describing the direct piezoelectric effect. Moreover, if the boundary conditions provoke mechanical stresses in both principal directions the resulting strain equation has to incorporate the transverse stress as well. For isotropic elastic deformation the Poisson’s ratio ν describes this impact and the equation reads · · · . (2.28) The linear equation (2.28) provides an acceptable approximation only within a small elastic range of the piezoelectric material. In case of large stresses non-linear interactions between the mechanical and electrical parameters become noticeable and have to be considered by appropriate terms. Temperature variations are another potential reason for non-linear effects which are also neglected in equation (2.28). Moreover, frequency or hysteresis related phenomena may also add further distortions to this equation [104]. The coupling factor which quantifies the efficiency of the energy transfer between electrical and mechanical energy is then given by [105] · . (2.29) Typical values for different piezoelectric materials are tabulated in [45, 105, 106]. 2.2.2 Membrane mechanics The peristaltic micropump developed in the framework of this thesis relies on the deflection of two membranes in order to displace a fluid volume. For the design engineering as well as for the performance analysis an analytical description of the membrane deflection is desirable. The deflection is provoked either by a pressure difference across the membrane or by piezoelectric actuation. In the following sections mathematical expressions will be derived for both effects. Later on in chapter 3 it will be shown, that the contributions of both effects are simply superimposed in case of small deflections. In accordance to the theory of mechanics the silicon membrane has to be treated as a plate since it is dimensionally stable and capable of generating restoring forces. Nevertheless, in the MEMS terminology those elements are commonly referred to as membranes or diaphragms. Even though the theory of plates will be applied in the following sections the 28 2 Fundamentals common terms membrane and diaphragm will be used interchangeably for the sake of uniformity with other MEMS literature. 2.2.2.1 Pressure induced deflection of a homogeneous membrane A pressure difference across the membrane causes a deflection in accordance to the elastic properties of the membrane. In a first step, the differential equation for a homogeneous membrane of thickness h with a well-know Young’s modulus E and Poisson’s ratio ν will be derived. For the sake of simplicity pure bending is assumed which implies the absence of shear stresses. Axial stresses in x- and y-direction occur within the membrane and induce an internal bending moment but they cancel out by integration over the thickness of a volume element. Figure 2-11 (a) illustrates the orientation of the volume element and indicates the bending moments as well as the radii of curvature. For small deflections of a thin membrane a set of assumptions named after Kirchhoff facilitate the mathematical derivations. First of all, stresses are restricted to the xy-plane i.e. normal stresses in z-direction are neglected. Concluding from that, there is no strain of the membrane in z-direction and consequently the deflection w(x,y) is independent of the vertical position. The deflection w is measured with respect to the flat membrane position. The relation between the deflection and the radius of curvature yields for small deflections 1 (2.30) 1 In addition, the cross-sectional area of the volume element is assumed to remain flat throughout the deformation and to be always perpendicular to the neutral plane (Bernoulli assumption). For the given geometry of the homogeneous volume element the neutral plane is found in the middle of the membrane. The neutral plane experiences zero stress and – due to Hooke’s law – also zero strain. (a) (b) (c) Figure 2-11: Volume element of a membrane exposed to pure bending (a). The normal stress along the principal axis changes linearly across the thickness of the membrane (b). The obtained radius of curvature is inversely proportional to the curvature of the bending line (c). 29 2 Fundamentals Figure 2-11 (b) points out, that the normal stress σx is a linear function of the z-position with its zero-crossing at the neutral plane. In the upper part the membrane experiences compression whereas the lower part is exposed to extension. For geometrical reasons the axial strain is [90] 1 · (2.31) and therefore the normal stress including the contribution of the transverse strain yields in accordance to Hooke’s law · 1 1 1 · · 1 . (2.32) the overall bending Considering a slice z of the membrane with a specific value moment acting with respect to the neutral axis can be determined via integration. Each of these slices contributes to the overall bending moment and the integration over the crosssectional area dAyz (shaded area in Figure 2-11 (a)) gives · · · 1 · 1 1 1 (2.33) · where Iy is the geometrical moment of inertia for bending about the y-axis. Commonly the integral in equation (2.33) is evaluated for unit width and plate thickness h which leads to the expression · 12 1 · · (2.34) incorporating a geometrical parameter called the flexural rigidity of the plate D. Therewith, a differential equation needs to be derived which describes the bending of the whole membrane. Again, deflections are assumed to be small compared to the plate thickness which allows for the negligence of normal in-plane stresses arising from the reactive forces of the clamped boundary conditions. In other words, the neutral plane is believed to remain in the middle of the plate during bending. Based on the equilibrium conditions considering the infinitesimal change of the bending moment and normal forces in z-direction a second order differential equation 2 30 , (2.35) 2 Fundamentals is derived for a distributed load q(x,y) acting in z-direction [107]. Despite the gradual changes of the bending moment it is assumed that the relation between bending moment and deflection derived for pure bending in equation (2.33) is sufficiently accurate for a small volume element of the membrane. This finally leads to the fourth order partial differential equation for bending of a thin plate 2 , . (2.36) From the mathematical point of view the solution of this partial differential equation is ambitious and only feasible for certain geometries. Selected cases are discussed in detail in the textbook by Timoshenko [107] where mathematical solutions for these problems are derived. In this thesis, only square membranes are applied. Nevertheless, a useful approximation to estimate the deflection of the square plate is given by the solution of equation (2.36) for a uniformly loaded circular plate. Here, the differential equation simplifies to 1 · . 2 (2.37) Consecutive integration with respect to the radial position r yields the general solution · 64 1 4 · · . (2.38) with R representing the radius of the plate. For clamped boundary conditions with w(r = R) = 0 and w’(r = R) = 0 the solution reads · 64 (2.39) 1 and thus the deflection at the center of the plate (r = 0) as a function of the load q and the flexural rigidity D is · 64 (2.40) . Timoshenko [107] presents also a calculation based on a square plate with clamped edges. For the derivation of this solution he first solves equation (2.36) for a simply supported rectangular plate with edge lengths a and b exposed to a uniform pressure load q. Following the method of Lévy the solution to this problem is given by an infinite Fourier series , · . (2.41) 0 along all of the The boundary conditions for a simply supported plate are 0 and simply supported edges. The coefficients Ym of the Fourier series are determined in order to satisfy these boundary conditions as well as equation (2.36) which yields 31 2 Fundamentals 4 , / 1 2 , , ,.. 2 2 2 1 (2.42) 2 · 2 with the substitution am = mπb/2a. This solution is based on a coordinate system where x = 0 and y = 0 denotes the center of the membrane and hence the egdes are found at x = ±a/2 and y = ±b/2. Since this Fourier series converges very rapidly a reasonable approximation is obtained by taking only the first few terms of the series. For clamped edges the derived solution for the simply supported case is superimposed with an additional contribution caused by external bending moments. The bending line induced by an external moment My distributed along the edges y = ±b/2 is given by , / 1 2 2 2 , , ,.. (2.43) 2 · A similar solution w2 is provoked by a bending moment Mx along the edges x = ±a/2. The 0 and 0 superimposed solution w+w1+w2 has to satisfy the boundary conditions along all of the clamped edges. This linear superposition is reasonably accurate for small deflections. In result the maximum deflection at the center of a square membrane with edge length a and Poisson’s ratio ν = 0.3 is given by 0.00126 · · . (2.44) The result indicates that the deflection is proportional to the fourth power of the edge length and inversely proportional to the flexural rigidity of the membrane. If the edge length is taken twice the radius R the deflection of the square membrane exceeds the displacement of the circular membrane. In contrast, if both geometries are set to equal areas the circular membrane exhibits a larger deflection. In both cases the deviation is approximately 25%. Thus, the general solution derived for the circular plate (equation (2.38)) can be used to determine both an upper and a lower estimation of the membrane deflection which encloses the solution for the square membrane. 2.2.2.2 Flexural rigidity of a piezo-membrane-composite So far the objective of the discussion was a review of well-known solutions for a homogeneous plate exposed to a uniformly distributed load such as a pressure load. By contrast the piezoelectric membrane actuator considered in this thesis constitutes a composite plate consisting of the silicon membrane, the glue layer and the piezoelectric PZT32 2 Fundamentals disc. In order to apply the equations derived in the previous section, the mechanical properties of this specific actuator design have to be determined. In this section, an analytical expression of the overall flexural rigidity of the piezo-membrane-composite will be derived. Figure 2-12: Close-up of the composite plate consisting of the silicon membrane, the glue layer and the PZT actuator with the corresponding elastic constants. Figure 2-12 depicts the setup of the composite plate. The isotropic Young’s moduli for the PZT ceramic and the adhesive layer are taken from the data sheets of the manufacturer (see Appendix F). The Poisson’s ratio of the PZT ceramic is in the range of ν = 0.3 which is a typical value for most materials. For the anisotropic silicon the Young’s modulus and the Poisson’s ratio depend on the crystal orientation as shown in Figure 2-13. (a) (b) Figure 2-13: Anisotropic Young’s modulus (a) and Poisson’s ratio (b) for silicon [108]. The membranes used in this work are manufactured from an (100)-wafer by means of anisotropic KOH etching. The orientation of the obtained membrane cavities is shown in Figure 2-14. Therefore, the principal in-plane stresses σx and σy in the membrane are aligned in (110)-direction. Schroth [109] constitutes in his comprehensive work on micromechanics that isotropic modeling of a thin silicon membrane is feasible for the given orientation. In this case, the Young’s modulus E(110) and the Poisson’s ratio ν(110) of these principal directions of the silicon membrane are applicable. The corresponding parameter values are E(110) = 169 GPa and ν(110) = 0.064 [110]. Nevertheless, since the theory presented in the previous section relies on a constant Poisson’s ratio for the entire composite the value ν = 0.3 will be used in 33 2 Fundamentals the following derivations for all layers of the composite plate despite the deviation in the silicon layer. For the implementation of different Poisson’s ratios shear stresses between the layers would have to be included. Figure 2-14: Orientation of a KOH-etched cavity fabricated into a (100)-silicon wafer. Due to the different Young’s moduli of these materials the neutral plane deviates from the middle plane of the composite plate. The position of the neutral plane is determined with respect to a reference plane which remains fixed in the middle of the silicon membrane. The neutral plane is characterized as the position of the bending axis. For equilibrium reasons the position of the neutral plane coincides with the minimum of the overall flexural rigidity which is the sum of the flexural rigidities of the individual layers for bending about the neutral plane. This way, the overall flexural rigidity is calculated as 1 1 (2.45) . 1 Instead of directly solving the integrals Steiner’s parallel axis theorem can be applied simplifying equation (2.45) to 1 · 12 1 1 12 12 1 2 1 2 1 2 1 2 · (2.46) · Note, that zN is negative for the given coordinate system (see Figure 2-12). The flexural rigidity D is plotted as a function of zN in Figure 2-15. The curve passes through a minimum for bending about an axis at zN=-68.27 µm with respect to the reference plane. In other 34 2 Fundamentals words, at this position the composite plate exhibits the lowest resistance against bending and therefore zN indicates the position of the neutral plane. Figure 2-15: Plot of the flexural rigidity vs. the position zN of the neutral plane. 2.2.2.3 Pressure induced deflection of the piezo-membrane-composite Based on the determined mechanical and piezoelectrical parameters a derivation of the bending line is essential for the analytical modeling of the actuator. From the mechanics point of view, the piezoelectric actuator comprises a composite plate in the inner region and a passive silicon membrane in the outer region. For a square actuator, this setup tremendously increases the mathematical complexity of an analytical solution based on the equations (2.41) - (2.43). Instead, as a straightforward approximation, the bending line of a pressure-loaded composite plate is calculated for a circular geometry by means of equation (2.38). This general solution serves as basis for the two solutions wi(r) and wo(r) that describe the deflection in the inner and outer region, respectively. The two different flexural rigidities Di for the inner region (see section 2.2.2.2) and Do for the outer silicon region are incorporated into the equations of the respective region. Then, appropriate boundary conditions have to be defined to link both regions. Figure 2-16 illustrates the composition of the membrane together with the stresses and moments at the transition point of the two regions. Figure 2-16: Cross-sectional view of the piezo-membrane-actuator and description of the inner and outer region. 35 2 Fundamentals At the clamped edge of the outer region both the deflection and the derivative of the bending line have to vanish, i.e. 0 0 . (2.47) Also, at the transition point of the two regions, both the bending line and its derivation have to be continuous. In addition, the bending moments at both sides have to balance each other in order to ensure a static equilibrium. The normal stresses σ are neglected due to the uniformly distributed load and the small deflections [107]. The mathematical derivation for this task is given in Appendix B. Figure 2-17 illustrates the calculated bending line for two different pressure loads. The parameter set utilized for the plotted solution is summarized in the table next to the plot. Ri 3.25 mm Ro 4 mm Di 0.2644 Nm Do 0.0142 Nm ν 0.3 Figure 2-17: Analytical solution for the bending line of the membrane spanning both regions. 2.2.2.4 Piezoelectric deflection of the piezo-membrane-composite Besides the pressure induced deflection of the membrane, which corresponds to the fluidic capacitance, the active deflection due to piezoelectric actuation is relevant for the modeling of the micropump. For the separated piezoelectric disc the application of a voltage and hence an electric field results in a strain given by equation (2.28). For the piezo-membranecomposite the strain of the piezoelectric disc is constrained by the restoring forces of the silicon membrane. In consequence, normal in-plane stresses are generated which constitute a bending moment of this bimorph structure about the neutral plane. Figure 2-18 depicts the linear strain distribution and the corresponding normal stresses in case of piezoelectric contraction. The different slopes of the stress curve correspond to the different Young’s moduli of the three layers. 36 2 Fundamentals Figure 2-18: Distribution of stresses and strains in a piezo-plate-composite upon piezoelectric contraction. As depicted in Figure 2-18 the piezoelectric contraction leads to a compression beyond the neutral plane and an extension below that plane with respect to the non-actuated state. Considering simply supported boundary conditions the piezoelectric actuation causes pure bending of the bimorph structure and hence a linear strain distribution is assumed across the thickness of the piezo-membrane-composite · . (2.48) The contraction of the piezoelectric disc is limited by the restoring force of the silicon membrane which results in tensile stresses across the entire thickness of the composite. Therefore, in contrast to pure bending induced by an external bending moment, the integration of the depicted normal stress distribution over the thickness of the piezomembrane-composite returns a net in-plane force which arises from the constricted piezoelectric strain. In order to determine the bending moment arising from this stress distribution the strainstress relation needs to be revealed for all layers. For the assumption of equal strains in xand y-direction the stress of the passive layers is immediately obtained from Hooke’s law 1 1 · 1 · 1 · · (2.49) · · The stress induced by the actuator is calculated for the undeformed state, i.e. strain of the 0) [111]. In vertical piezoelectric disc is inhibited in both transverse directions ( direction the piezoelectric disc is assumed to be unconstrained (σz = 0). Therewith, the normal stress in the piezoelectric disc is calculated from equation (2.28) and yields · 1 · (2.50) is replaced by the inverse of the isotopic Young’s modulus where the stiffness coefficient for PZT. Note that the coefficient d31 is negative i.e. a positive field strength E3 leads to a transverse contraction of the piezoelectric disc which induces tensile stresses (σ > 0). Since the piezoelectric disc is attached to the elastic silicon membrane, the normal stress induced by the actuator leads to a bending of the bimorph composite with an approximated linear strain distribution as explained above. Therefore, the overall stress in the piezo-disc is obtained by superposition of the strain-related contribution following Hooke’s law and the voltage-induced impact 37 2 Fundamentals · 1 · 1 · · · 1 · (2.51) . The position of the neutral plane zN for the piezo-membrane-composite has been derived in section 2.2.2.2 and is used for the following calculations. For the simply supported boundary condition the bending moments caused by the stress distribution σ(z) need to balance across the thickness of the composite. Therefore, the integral · 1 · · 1 · . (2.52) · 1 · · 0 has to vanish. In other words, the pure bending moment induced piezoelectrically is balanced across the thickness of the piezo-membrane-composite. Equation (2.52) is integrated and solved for the slope κ of the strain curve by means of the symbolic toolbox of MatlabTM. The result reads 6 2 12 12 3 6 12 12 3 4 4 2 12 12 6 24 12 12 12 (2.53) 12 . Therewith, the stress distribution across the thickness of the piezo-membrane-composite is exactly determined by the equations (2.49) and (2.51). The induced bending moment is calculated by integration of σ(z) times the distance (z-zN) to the neutral plane over the thickness of the piezo-membrane-composite. Alternatively, the bending moment is also obtained by solely integrating the piezoelectric stress σpiezo times the distance to the neutral plane over the thickness of the PZT-layer · . (2.54) A calculation of the bending line from known bending moments has been published for circular plates by Li et al. [112]. It takes up the general solution for the bending of circular plates developed by Timoshenko [107]. For the outer annular region the solution 38 2 Fundamentals · · (2.55) for bending of a circular plate with a circular hole at its center is applied. The geometric variables r and Ro refer to the cross-sectional view shown in Figure 2-16. The bending moment Mo acting on the outer region at the transition point r = Ri needs to satisfy the equation 2 2 . 0, Together with the clamped boundary conditions ( (2.56) 0 ) the coefficients of equation (2.55) are easily evaluated and the calculation yields · 2 1 · 1 2 · . (2.57) For the inner region pure bending upon the piezoelectric actuation is assumed and the bending line satisfies the general solution · . (2.58) The bending moment Mi responsible for the pure bending of the inner region deviates from the bending moment Mo at the edge of the outer region due to the piezoelectric contribution. Mi is composed of two parts, the piezoelectric bending moment Mpiezo and the bending moment Mo imposed by the outer region. While the piezoelectric bending moment Mpiezo balances itself across the thickness of the composite, the superimposed contribution Mo is balanced by the elastic coupling to the outer region. In sum, the bending moment Mi amounts to (2.59) where the bending moment Mo acts as restoring moment opposing the bending effort of the piezoelectric bending moment Mpiezo. Similar to equation (2.56) the bending moment of the inner region 2 ·2 (2.60) is calculated based on the general solution (equation (2.58)) and determines the coefficient c4. Therewith, the bending line in the inner region reads 2 · 1 . (2.61) 39 2 Fundamentals At the transition point of the two regions appropriate compatibility conditions have to be met, explicitly . (2.62) Applying the first of the two conditions eliminates the coefficient c5 and gives the bending line of the inner region · 2 1 · 1 2 · 2 1 . (2.63) Since only the bending moment Mpiezo is known by value in accordance to equation (2.54) the bending moments Mi and Mo have to be expressed in terms of Mpiezo. Using the second boundary condition stated in equation (2.62) determines the unknown bending moment 1 · 1 (2.64) 1 1 and also the bending moment Mi via equation (2.59). This completes the analytical solution for the bending line of a clamped circular plate deflected by a circular piezoelectric actuator. A comprehensive analytical modeling for this configuration is also presented in a recent publication by Fox et al. [113]. The investigations reveal that the coupling of the stresses and bending moments between the piezoelectric disc and the silicon membrane occur predominantly at the transition point between inner and outer region. This justifies that the coupling of the piezoelectric bending moments via the compatibility conditions stated in equation (2.62) is a valid approach. 2.2.2.5 Displacement volume The volume displaced by the membrane due to its deflection is immediately obtained from the bending line by integration over the area. For the considered circular membrane with an inner and an outer region the displacement volume amounts to ∆ 40 ·2 ·2 . (2.65) Chapter 3 Two-stage micropump 3 Two-stage micropump A novel two-stage micropump concept is proposed and realized in this thesis which bears the potential of high resolution, power-efficient and pressure-independent dosing of fluids in conjunction with a comparably simple 2-wafer silicon fabrication process. Especially in the field of medical devices, but also for many other microfluidic applications, this micropump provides desired and beneficial characteristics which demonstrate the competitiveness of this concept. This chapter introduces the design of the micropump and explains the working principle. A lumped parameter model of the micropump is established in order to analyze key attributes of this concept. For the development of the model, numerical simulations of the membrane deflection are carried out and the results are compared to the analytical model derived in chapter 2. Subsequently, this chapter considers the capability of the micropump to transport gases and gas bubbles. The theoretical background is provided and simulations are utilized to verify the conclusions. Finally, a critical compression ratio for this type of micropump is derived. The last section of this chapter introduces a modified design of the two-stage concept referred to as single-membrane micropump and points out the expected benefits. 3.1 Design and Working Principle The design of the two-stage micropump strictly pursues the aim of high-resolution volumetric dosing. In addition, the intended medical application demands for a small device size, minimum weight and low energy consumption. The main impact factor which commonly prevents the volumetric dosing of a micropump is the backpressure dependence, i.e. the susceptibility of the flow rate to an increase of the static pressure head applied to the outlet of the micropump. The presented novel concept of a modified peristaltic micropump eliminates this detrimental effect by favorably utilizing the fluid dynamics of the system in conjunction with a mechanically restricted membrane deflection. 41 3 Two-stage micropump 3.1.1 Concept A schematic cross-section of the proposed silicon micropump is depicted in Figure 3-1.The design essentially adopts the principle of a peristaltic micropump but skips the middle membrane commonly used as pump membrane. Instead, two back-to-back connected active valves generate a well-defined fluid flow by an alternate switching of the piezo-actuators following a 3-phase actuation scheme. Piezoelectric actuation has been chosen due to the need of fast actuation which turns out to be essential for a precise control of the fluid dynamics. Figure 3-1: Schematic cross section of the piezoelectric micropump with a closeup of the valve geometry. The cross-sectional drawing shows that two structured and subsequently bonded silicon chips constitute the micropump. Both fluidic ports are surrounded by a valve seat which is placed centrically underneath the respective membrane. The design-inherent distance between the valve lips and the flat membrane is set to 1 µm. In the following this distance is referred to as residual gap height h0. Piezoelectric discs are attached to each membrane in order to actuate the valves and to displace the fluid. 3.1.2 Design of the piezo-membrane-actuator The actuator deployed for this micropump is a bimorph structure consisting of a square silicon membrane with a square piezoelectric PZT disc on top. The piezoelectric disc is fixed to the silicon membrane by means of a conductive glue which forms an approximately 10 µm thick bonding layer (for details see chapter 4 on the fabrication process). The piezoelectric deformation of the PZT disc leads to a bending of the bimorph structure as discussed in the previous chapter 2. a 8 mm b 6.5 mm hpzt 200 µm hadh 10 µm hsi 100 µm Figure 3-2: Design of the piezo-membrane-actuator featuring the tabulated geometric parameters. 42 3 Two-stage micropump Figure 3-2 illustrates the geometry of the piezo-membrane-actuator which is applied for the two-stage micropump. The tabulated values indicate that the lateral dimension of the membrane is almost two orders of magnitude larger than its thickness. The decision for a square membrane is based on the intended fabrication by means of anisotropic KOH etching. The piezo-membrane-actuator comprises two different regions characterized by different flexural rigidities since the piezoelectric disc covers only central part of the membrane. The ratio of the edge length of the piezo-disc to the edge length of the membrane is a crucial design parameter for the optimization of the membrane deflection. As many researchers have already focused on this aspect it is not studied in the framework of this thesis again. In particular, a detailed investigation of this issue is presented in the thesis by A. Doll [114]. For his work he uses the same materials and fabrication technologies and hence the results are regarded as foundation for this design. Summarizing his results, a maximum deflection at the center of the disc is obtained if the edge length of the piezo-disc is set to approximately 80% of the membrane side length. For maximum volume displacement a side length of 91% is recommended as optimum configuration. These results are in full agreement with respective work published by others [113, 115]. In consequence, the edge length of the square piezo-disc for this two-stage micropump is set to 6.5 mm which is approximately 81 % of the edge length of the membrane. 3.1.3 Geometry of the pump chamber For the interior of the pump chamber two different geometries are investigated in this work. In a first design I a rectangular pump chamber is realized which spans the whole area underneath the membrane and encloses a volume of VI = 16.9 x 8 x 0.03 mm3 = 3.95 µl (Figure 3-3). In a second design II a smaller pump chamber is implemented. The height of the pump chamber is kept at 30 µm, but the corners are removed from the chamber and the width is reduced yielding a pump chamber volume VII = 1.02 µl. The intention behind design II is to prevent the entrapment of air bubbles near the corners of the rectangular pump chamber and to increase the compression ratio by means of a smaller dead volume. (a) (b) Figure 3-3: Illustration of the rectangular chamber in design I (a) and the diminished pump chamber in design II (b). 43 3 Two-stage micropump 3.1.4 Actuation scheme In Figure 3-4 the actuation scheme is illustrated which comprises the three phases termed as refill phase, transfer phase and delivery phase. This scheme is referred to as standard mode since it is the recommended actuation sequence for the presented concept of the two-stage micropump. Nevertheless, for the transport of compressible media such as gases an adapted actuation sequence will be presented later on which deviates from the standard actuation scheme. In the refill phase fluid is drawn into the pump chamber via the inlet. During this phase the outlet valve is tightly closed in consequence of an applied closing voltage and the negative pressure induced by the opening of the inlet valve. At the end of this phase the pressure within the pump chamber is relaxed to the value of the inlet pressure p = pin. Figure 3-4: A 3-phase actuation scheme is applied to control the two actuators. The pumping mechanism essentially relies on a simultaneous switching step to initiate the transfer phase and requires two different closing voltages of the inlet and the outlet valve. The constant cut-off pressure pc, which remains in the chamber at the end of each pump cycle, enables a pressure independent stroke volume. The fluid transfer phase is initiated by a simultaneous closing of the inlet valve and opening of the outlet valve and is the first key attribute of this concept. It is evident that the fluid displacement within the pump chamber must occur on a significantly smaller time scale compared to the fluid inflow and outflow through the valves in order to achieve a pressure independent volume transfer. Thus, the fast actuation mechanism together with the appropriate adjustment of the fluidic resistances of the inlet and outlet valve are crucial in order to transfer a well defined volume from the left part to the right part of the pump chamber and to avoid substantial backflow at elevated outlet pressures. For this reason the gap between the valve lip and the membrane is set to 1 µm only (with respect to the undeflected membrane). Moreover, a well-timed electrical control is vital for the synchronization of the actuators. During the delivery phase the propelled volume is released from the micropump. The inlet valve remains closed and fluid is pushed through the outlet until the membrane touches the valve lip and tightly seals the valve. As depicted in Figure 3-4, a lower closing voltage is applied to the outlet valve in order to keep the inlet valve closed. The micropump model derived later on in section 3.2.3 will point out that the pressure increase in the pump chamber 44 3 Two-stage micropump is caused by and proportional to the closing voltage applied to the outlet actuator. It is crucial that this pressure increase must not exceed a critical value where leakage of the inlet valve would set in due to the elasticity of the membrane. Therefore, the closing voltage applied to the outlet actuator is reduced to approximately 2/3 of the inlet actuator voltage. The impact of a variation of this voltage on the flow rate will be investigated in chapter 5.5.2. The cut-off pressure pc denotes the pressure value at which the membrane of the outlet valve touches the valve lips. In other words, the outlet valve is held open due to the high pressure in the pump chamber until it drops below the cut-off value pc. The cut-off pressure is determined by the interplay between the actuation voltage of the piezo-actuator and the compliance of the membrane. Considered that the outlet pressure pout acts on approximately 0.8 % of the membrane area only while the rest of the membrane is exposed to the pressure in the pump chamber, the cut-off pressure is virtually independent of the backpressure pout. This constant cut-off pressure pc, which remains in the pump chamber at the end of each cycle, ensures a backpressure independent stroke volume and makes up the second key attribute of the pumping mechanism. Compared to other designs which require rather complex structures with up to four stacked wafers to implement mechanical limiters for the diaphragm displacement, in this design the valve lips themselves are the obstacles which limit the deflection of the membrane. 3.2 Modeling and simulation of the micropump The availability of a system model is of high relevance in order to reveal the physics behind the system and to enable straight-forward redesign cycles to further improve the characteristics of the micropump. A numerical simulation of the whole micropump based on a finite element method (FEM) or a computational fluid dynamics approach (CFD) would be desirable but is hardly achievable due to the high complexity of such a system. An appropriate model would have to couple a structural mechanics problem including the piezoelectric actuation with the fluid displacement. Moreover, the membrane deflection changes the size and shape of the fluidic domain which would enforce a deformable mesh approach. The high aspect ratios found in the fluidic domain – the lateral size of the membrane is 8 mm, but the gap height between the valve lips and the membrane is only 1 µm – would require an extremely fine mesh in the vicinity of this critical point resulting in an extraordinary large number of elements. In sum, commercial simulation tools such as CFDACE+ or COMSOL MultiphysicsTM struggle to solve such a comprehensive model of the micropump. Instead, a lumped parameter approach is chosen which links the fluid flow to the pressure gradients between different compartments of the micropump. This pressure differences may arise from the piezoelectric actuation as well as from external hydrostatic pressures applied to the inlet or the outlet valve of the pump. The derivation of the lumped parameter model provides a generalized tool for the analysis of diaphragm displacement pumps. It can be easily adapted to various designs and, within the limit of fast actuation, holds also true for actuation mechanism other than piezoelectric actuation. Only a small set of design-specific parameters such as the fluidic membrane capacitance C, the displacement volume of the or the valve lip geometry is required for the model. The design-specific values actuator ∆ of these parameters can be determined either by numerical simulations or by experimental 45 3 Two-stage micropump measurements. In this work, FEM simulations with COMSOL MultiphysicsTM (COMSOL AB, Stockholm, Sweden) are carried out to analyze the structural mechanics problem of the membrane deflection due to both piezoelectric actuation and pressure load. In addition, an analytical calculation based on the fundamentals presented in chapter 2 will be given in the following sections and will be compared to the simulation results. 3.2.1 FEM simulation of the bending membrane A 3-dimensional model is established for the design introduced in section 3.1.2. As described above, the lateral size of the membrane is 8 x 8 mm2 and the square PZT-actuators feature an edge length of 6.5 mm. Due to symmetry conditions only one quarter of the membrane is modeled. For the simulation the elastic properties of the materials as well as the piezoelectric properties of the PZT ceramic have to be known. The applied parameters are summarized in the following Table 3-1. The parameter values for the piezo-ceramic and the adhesive are taken from the datasheets of the manufacturers. For the elastic properties of the silicon membrane, the textbook edited by Korvink and Paul [110] is used as a reference for these data. Table 3-1: Parameter set used for the FEM simulation of the membrane deflection PZT ceramic [116] Parameter Value d31 -350 10-12 C/N EPZT = Adhesive [117] Silicon (<110>) . 1 sE11 62.9 GPa νPZT 0.3 Eadh 2.47 GPa νadh 0.3 Esi 168.9 GPa νsi 0.064 The established model couples a structural mechanics problem with piezoelectric actuation. Here, the MEMS module of the COMSOL MultiphysicsTM software provides predefined modes which enable a straight-forward setting of the mechanical properties of the materials and the boundary conditions. The mechanical boundary conditions along the edge of the silicon membrane are set to zero deflection and zero bending, a boundary condition that is commonly referred to as clamped or fixed. Apart from the edge the membrane is free to move. A potential pressure load is applied to the bottom face of the membrane. In the piezoelectric mode an actuation voltage is defined across the piezoelectric domain. 46 3 Two-stage micropump Several simulation runs have been completed to validate the convergence of the simulation result with respect to mesh refinements. In consequence of this study, an unstructured mesh with 67705 elements was chosen as trade-off between accuracy and simulation time. The model geometry together with the results of this convergence study is given in Appendix C. An example of a simulation run is shown in Figure 3-5 for an actuation voltage of -80 V. The maximum deflection Δw at the center of the membrane is 3.87 µm. Integration of the z-displacement over the membrane area yields a displacement volume of ΔV - = 94.3 nl. Figure 3-5: Simulation result of the bending membrane upon piezoelectric actuation at a voltage of -80 V. Due to the apparent symmetry only a quarter of the membrane is modeled and appropriate symmetry boundary conditions are applied. 3.2.2 Lumped parameter model of the elastic membrane This section deals with the pressure-induced deflection of the membrane and its significance for the lumped parameter model of the micropump. FEM simulations are carried out and the results are compared to the prediction of the analytical model derived in chapter 2. In the lumped parameter model the fluidic capacitance is an important parameter that accounts for the elasticity of the membrane. The bending of the membrane is attributed either to piezoelectric actuation or to a pressure difference across the membrane. Taking both effects into consideration, the deflected membrane displaces a volume ∆ ∆ · (3.1) with respect to the flat membrane. The first term counts for the nominal displacement volume caused by the actuator. Depending on the actuation voltage the nominal displacement volume might be either negative (ΔV -) or positive (ΔV+). The second term assumes a linear pressure-induced deflection where p is the pressure in the pump chamber and p0 is the ambient pressure at the opposite side of the membrane. Both effects are considered to be decoupled which leads to a superposition of their contributions. This approach coincides with the assumption that the deflections are small and all materials are in their linear elastic range [118]. 47 3 Two-stage micropump The FEM simulation results confirm that the displacement volume changes linearly with the applied pressure (Figure 3-6 (a)) and the slope of the curve determines the fluidic capacitance of the membrane C = 1.17·10-15 m3 Pa-1. This assumption of a constant capacitance holds true only for small deflections, a condition that is clearly fulfilled for the considered design. For deflections in the range of or beyond the membrane thickness the fluidic capacitance C would become pressure dependent which would imply a non-linear term in equation (3.1) [119]. (a) (b) 1.17 · 10 4.43 · 10 Figure 3-6: FEM simulations show a linear displacement volume (a) as well as a linear center deflection (b) of the membrane in response to an applied pressure which complies with the assumption of a constant fluidic capacitance. The simulation results fall into the solution sector of the analytical approach. In Figure 3-6 the analytical equations derived in chapter 2 are compared to the FEM simulation outcome in order to validate the results. Here, the flexural rigidities of the inner composite region Di and the outer membrane region Do are set to the values calculated in chapter 2 (see Table 3-2). The Poisson’s ratio is set to ν = 0.3 for all materials. Since the analytical model considers a circular membrane with circular piezo-actuators, the radii of the model have to be related to the edge length of the squared geometry. The first option is to set the diameter of the circular geometry equal to the edge length of the squared geometry (edge length matching). This definition determines the lower boundary line of the gray shaded sector indicated in Figure 3-6. The upper boundary line is obtained if the areas of the circular PZT-disc and the circular membrane are set equal to their squared counterpart (area matching). In this case, the deflection is overestimated which is in full accordance to reported results [120]. The following table summarizes the geometric and elastic parameters utilized for the analytical calculations. 48 3 Two-stage micropump Table 3-2: Parameter set used for the analytical calculation of the membrane deflection Parameter Value d31 -350 10-12 C/N PZT ceramic [116] . 1 EPZT = Adhesive [117] 62.9 GPa sE11 νPZT 0.3 Eadh 2.47 GPa νadh 0.3 Esi 168.9 GPa νsi 0.3 Silicon (<110>) a2 π R0 4.51 mm Area matching 3.67 mm 4 mm 2 Edge length matching 3.25 mm 2 Flexural rigidity Di 0.2644 Nm Do 0.0142 Nm Besides the displacement volume ΔV the maximum deflection Δw in the middle of the membrane is important for the lumped parameter model in order to determine the gap height hv. The gap height is obviously related to the actual membrane position. At zero membrane deflection there is a residual gap height h0 which is set to 1 µm in the proposed design. The plot of Δw versus the applied pressure also yields a linear relation as shown in Figure 3-6 (b). An elasticity-related mechanical coefficient fV is deduced from the slope of the curve which allows an analytical description of the deflection ∆ ∆ · . (3.2) 49 3 Two-stage micropump Thus, the coefficient fV is the equivalent to the fluidic capacitance C. Its value for the given membrane and actuator size is fV = 4.43·10-11 m Pa-1. A special situation occurs for the closed valve when the membrane is bent inwardly towards the pump chamber and touches the valve lips. In that case the additional support by the valve seat significantly reduces the fluidic capacitance of the membrane. Appropriate FEM simulations turned out that the reduced membrane capacitance for a closed valve amounts to CC = 2.57·10-16 m3 Pa-1. 3.2.3 Lumped parameter modeling of the piezoelectric actuation For actuation, the membrane is deflected on purpose by switching the voltage applied across the piezoelectric disc. Here, the bimorph effect causes the membrane to bend as described extensively in chapter 2. The analytical equation derived there predicts a linear relationship and the applied voltage U between the nominal displacement volume ∆ ∆ , , , , , , , , , , · (3.3) where χ is an actuator specific parameter depending on the geometry of the design and the elastic properties of the materials. A parameter value 1.18 / is obtained from the result plot of the FEM simulation shown in Figure 3-7 (a). Again, the simulation result is overlaid by the analytical solution to validate the consistency of both approaches. (a) (b) 1.18 48.36 Figure 3-7: Simulation of the piezoelectric bending of the membrane considering the displacement volume (a) and the center deflection (b). The gray-shaded sector confines the range of the analytical solution for the circular membrane and the solid curve indicates the FEM simulation result. The graph on the right-hand side depicts the center deflection as function of the actuation voltage. The analytical prediction is compared to the simulation result as well as to the measurement result obtained with a laser distance sensor (AWL7/0.5, WELOTEC GmbH, Laer, Germany). By using the coefficient of proportionality α the membrane deflection is given by 50 3 Two-stage micropump ∆ · (3.4) . The piezoelectric actuation of the membrane generates a sudden change of the chamber pressure. For an incompressible fluid and for the assumption of fast actuation – i.e. the expansion of the actuation force and the subsequent leveling of pressure oscillations within the pump chamber occur on a shorter time scale than the fluid flow through the valves – a triggered pressure step ∆ ∆ (3.5) total accounts for the actuator impact. Here, ∆ is the nominal displacement volume of the actuator and Ctotal is the overall fluidic capacitance within the pump chamber. In order to model an actuator with a larger response time a damped step response could be utilized instead of a pressure step. The pressure change corresponding to a certain part of the actuation sequence is deduced from an investigation of the displacement volumes and the relevant fluidic capacitances. The analytical expressions for the pressure changes are summarized in the following Figure 3-8. |∆ ∆ ∆ ∆ | |∆ |∆ ∆ | | , |∆ ∆ |∆ | | ∆ , 2 Figure 3-8: Pressure changes in the pump chamber upon the piezoelectric switching of the valves. Here, |∆ | and|∆ | denote the nominal displacement volume of the actuator i = 1,2 for the applied positive and negative actuation voltage, respectively. The quantity ∆ describes the displaced volume when the inwards bended membrane is supported by the valve seat. Pc,i denotes the cut-off pressure of the respective valve while C is the capacitance of the diaphragm and CC represents the reduced fluidic capacitance of the closed membrane. 51 3 Two-stage micropump Figure 3-9: Calculation of the pressure step within the pump chamber at the beginning of the delivery phase. Figure 3-9 exemplary illustrates the calculation of the pressure step occurring at the beginning of the delivery phase. The actuation voltage of the outlet valve is switched from -80 V to +80 V which would cause a nominal volume change of |∆ | |∆ | if the membrane was not hindered by the valve seats. However, due to the fast actuation and the inertia of the fluid, the total volume of the pump chamber cannot change rapidly for incompressible fluids. Thus, based on the assumption of a constant total chamber volume, the pressure step Δp is calculated to be ∆ |∆ | |∆ | |∆ , 2 | |∆ | . (3.6) The simplification of equation (3.6) holds true for zero outlet pressure and for the assumption that the inlet valve is kept closed during the entire delivery phase. Therefore, the cut-off pressure of the inlet membrane pc,1 beyond which the inlet valve would be opened has to be greater than all chamber pressures occurring during the delivery phase. For this reason the closing voltage of the outlet valve is reduced compared to the inlet voltage. Considering the design of the membranes and the actuation voltages additional prerequisites have to be met. First of all, the cut-off pressure pc,2 of the outlet membrane needs to be higher than the maximum applied outlet pressure (3.7) . , Second, the pressure step Δp needs to be sufficiently large in order to satisfy the equation ∆ (3.8) , because otherwise the outlet valve would be closed immediately upon the piezoelectric actuation and no fluid would be ejected through the outlet. Finally, all of the criteria above are summarized in the equation , ∆ , . (3.9) The cut-off pressure pc,2 is calculated from the displacement volumes following the equation 52 3 Two-stage micropump |∆ | |∆ ∆ | , ∆ (3.10) Here, the nominal displacement volume ∆ as well as the center deflection ∆ are given by equation (3.3) and (3.4), respectively. At p = pc the deflection in the middle of the membrane is well known to be ∆wp=pc = -h0 . Substituting both terms into equation (3.10) gives the final result | · | | | , (3.11) A graphical evaluation of equation (3.9) is given in Figure 3-10. The left graph shows the relation between the closing voltages U1 and U2 for the inlet and outlet valve, respectively, in order to satisfy the condition pc,1 > pout + Δp. The right graph indicates the required closing voltage U2 of the outlet valve which is required to withstand the outlet pressure (pc,2 > pout). The third condition pout + Δp > pc,2 is uncritical for the given design and is fulfilled for any of the applicable outlet voltages (U2 < 335 V). (a) (b) Figure 3-10: Graphical solution for the inequation pc,1(U1) > pout + Δp(U2) (a) and the inequation pc,2(U2) > pout (b). The dotted lines in Figure 3-10 indicate the voltage values applied for the standard actuation scheme. Concluding from Figure 3-10 (a), the inlet closing voltage U1 seems to be insufficient to fulfill the requirements in case of an outlet pressure of 30 kPa. However, the voltage levels for the standard actuation scheme are determined not solely due to this analytical result but mainly due to the experimental results presented later on in chapter 5. 3.2.4 Superposition of piezoelectric and pressure induced bending For the lumped parameter model it is important to consider the combined deflection which is obtained from piezoelectric deformation superimposed by a pressure load. For the individual effects linear relationships have been revealed in the previous sections regarding both the displacement volume ΔV as well as the center deflection Δw. Thus, for either sole piezoelectric bending or pure pressure induced deflection a constant individual ratio ΔV/Δw is 53 3 Two-stage micropump achieved. For piezoelectric bending a ratio of ΔV/Δw = 2.44·10-5 m2 is determined whereas the ratio for pure pressure-induced bending yields ΔV/Δw = 2.64·10-5 m2. The difference arises from a deviation of the exact shape of the bending line between the two effects. When both contributions are superimposed the shape of the bending line depends on the predominant effect. This phenomenon has been investigated by means of the FEM simulation model and the result is illustrated in Figure 3-11. Here, the thick solid line encloses the nominal displacement volume for an actuation voltage of -80 V without any applied pressure. This piezoelectric displacement is considered as an offset with a superimposed variable pressure load. Figure 3-11: Bending line of the piezoelectrically deformed membrane with different superimposed pressure loads. The effect of a superimposed pressure load on both the displacement volume ΔV as well as the center deflection Δw is illustrated in Figure 3-12 (a) and (c), respectively. It becomes noticeable that the zero crossing is slightly shifted between the two graphs. The reason behind this phenomenon is explained by the corresponding bending lines depicted on the right-hand side of Figure 3-12. A corrugated bending line is caused by the two opposing forces i.e. the upwards directed piezoelectric deflection and the downwards directed pressure impact. This effect leads to a deviation between the zero crossing of the displacement volume ΔV (-80.9 kPa) and the zero crossing of the center deflection Δw (87.9 kPa) 54 3 Two-stage micropump (a) (b) (c) (d) Figure 3-12: Both the displaced volume ΔV (a) as well as the center deflection Δw (c) are linearly related to the applied pressure. The pressure shift of the zero crossing between both graphs is explained by the corrugated bending line of the membrane (b), (d). In consequence, the ratio ΔV/Δw in the case of superimposed piezoelectric bending and pressure induced deflection is no longer constant. As shown in Figure 3-13 the expression ∆ ∆ ∆ ∆ · · (3.12) exhibits a singularity at the zero crossing of the center deflection Δw. Figure 3-13: The ratio of the displaced volume and the center deflection shows a singularity in case of superimposed piezoelectric actuation and pressure loading. 55 3 Two-stage micropump For the lumped parameter model it is important to know the center deflection in order to determine the gap height and the corresponding fluidic resistance of the gap (see the following section 3.2.5). Due to this corrugated membrane phenomenon it is not possible to immediately convert the displacement volume into the center deflection. The ratios reported above for pure pressure induced bending or pure piezoelectric deformation can be utilized as approximation but a precise calculation of the gap height requires the implementation of the equation ∆ ∆ · (3.13) where h0 is the design-inherent initial gap underneath the flat membrane, ∆ is the center deflection caused by the actuator and the last term accounts for the deflection due to the pressure difference across the membrane. It should be noted that a negative gap height is physically impossible even though negative values might be obtained from equation (3.13). In this case the valve lip prevents the membrane from bending downwards any further. Thus, a negative value indicates that the gap is closed and the membrane is additionally supported by the valve lip which necessitates the use of the reduced fluidic capacitance CC = 2.57·10-16 m3 Pa-1. 3.2.5 Lumped parameter model of an active valve An analytical model for micro-diaphragm pumps has been proposed by Goldschmidtböing et al. [121] earlier and serves as a basis for this model approach. Following [121] the flow rate Q through the valve is related to the pressure gradient ΔpV across the valve lips via ∆ · (3.14) where RV denotes the fluidic valve resistance. The resistance is predominantly caused by the small gap hv between the valve lips and the membrane. Figure 3-14 exemplary depicts a schematic drawing of the inlet valve with the corresponding design parameters and its electric circuit equivalent. (a) (b) Figure 3-14: Schematic drawing of the inlet valve (a) with its fluidic network equivalent (b). Assuming a laminar parabolic flow profile throughout the gap which holds true for the small gap height of the presented design the fluidic resistance of the valve RV can be derived from 56 3 Two-stage micropump equation (2.17) for pressure-driven Stokes flow through a small gap. Due to its circular geometry the valve is divided into circular elements with a width and a circumference 2 , each exhibiting an infinitesimal fluidic resistance of 12 · 2 . (3.15) Subsequent integration from the inner radius r1 to the outer radius r2 yields the fluidic valve resistance 6 · · (3.16) which is proportional to the viscosity of the fluid η and inversely proportional to the third power of the gap height hv. Therewith, the flow rate Q through the valve can be expressed as ∆ , ∆ · ·∆ 6 · (3.17) where Δpv denotes the pressure drop across the valve. This model of a constant valve resistance accounts only for the viscous pressure losses. It is an admissible approach for the presented design where the width of the valve lip w = 100 µm exceeds the gap height h0 = 1 µm by two orders of magnitude. In general, the convective losses would add a non-linear contribution to the pressure loss. A detailed study of the fluidic characteristics depending on the valve geometry is given by Doll et al. [122] and the presented numerical results could be implemented into the lumped parameter model by means of a look-up table. Nevertheless, for the particular valve geometry of this micropump and the low maximum flow rates in the range of µl/s the convective losses are small and hence are consequently neglected for the sake of simplicity. 3.2.6 Fluidic inertance The mass of the fluid exposes a fluidic inertance which inhibits a rapid change of the flow rate. As proposed in section 3.1 the high fluidic resistance of the valve geometry is crucial in order to prevent undesired backflow during the transfer phase. Since the transfer phase coincides with a rapid displacement of a fluid volume the fluidic inertance has to be considered in order to determine the fluidic response to the piezoelectric actuation. The pressure difference Δpinertia which is required to overcome the inertia of the fluid is ∆ (3.18) where L is the fluidic inertance and dIv/dt is the change of the volume flow rate Q with time. For a rigid tube with a cross-sectional area A and a length l the fluidic inertance is given by 57 3 Two-stage micropump · . (3.19) Especially in microfluidic channels that typically feature small diameters the pressure drop upon rapid acceleration or deceleration could be a significant aspect for the dynamic response of the system. The characteristic time which limits the dynamic response of a fluidic system is determined by the ratio (3.20) of the fluidic inertance L and the fluidic resistance R. Figure 3-15: Fluidic network model to describe the dynamic response of the systems upon the simultaneous switching of both actuators at the beginning of the transfer phase. The fluidic situation emerging in the pump chamber during the transfer phase is described by the network model depicted in Figure 3-15. The outlet valve is omitted since it is closed during the entire transfer phase. Upon the simultaneous actuation of both membranes the fluid volume displaced from the capacitance C1 has two possible means of escape: it may either travel backwards through the inlet valve (path 1) or move into forward direction towards the fluidic capacitance C2 at the front part of the pump chamber (path 2). Along path 1 the fluid has to pass the resistance Rv established by the inlet valve and it also faces a resistance Rinlet and an inertance Linlet belonging to a microchannel or tube connected to the inlet port of the micropump. 58 3 Two-stage micropump Following path 2 the fluid displaced within the pump chamber faces a chamber resistance Rchamber and an inertance Lchamber. Both quantities are determined by the pump chamber geometry. From Hagen-Poiseuille’s law the chamber resistance 8· · (3.21) · is estimated by means of the hydraulic diameter Dhd of the rectangular cross-section of the pump chamber and the distance l = 9 mm between inlet and outlet. Due to the extreme aspect ratio of the rectangular cross-section with a width w = 8 mm (Design I) or w = 4 mm (Design II) and a height of only h = 30 µm the hydraulic diameter is calculated by 4 ·cross‐sectional area wettet perimeter 4· 2 · 2 2 60 μ . (3.22) Therewith, a resistance Rchamber = 2.5·1013 Pa s m-3 is estimated for the pump chamber according to equation (3.21). The fluidic inertance Lchamber = 3.75·107 Pa s2 m-3 (Design I) and Lchamber = 7.5·107 Pa s2 m-3 (Design II) is determined by equation (3.19). This yields characteristic response times τ inertia = 1.47 µs (Design I) and τ inertia = 2.94 µs (Design II) for the fluid displacement within the pump chamber. This inertia-related response delay needs to be compared to the expected time constant for the discharge of the fluidic capacitance · . (3.23) If the whole volume displacement would occur along path 1, the characteristic time τ capacitance = Rv ·C = 355 ms applies. Here, the valve resistance Rv was calculated by equation (3.16) for a gap height hv = 5 µm that corresponds to the gap height of the open valve for an upstroke voltage of -80 V. Even the characteristic time for a displacement within the pump chamber τ capacitance = Rchamber ·C = 30 ms is about three to four orders of magnitude higher than the inertia time constant. In consequence, the impact of the inertia-related effects is reasonably neglected and the volume displacement is governed by the ratio of the fluidic resistances 3 · 10 2.5 · 10 / / 10 1. (3.24) Again, this ratio is calculated for an open gap with hv = 5 µm. It further increases during the closing process when the membrane approaches the valve seat. 3.2.7 Lumped parameter model of the micropump In the previous sections lumped parameter models of the individual components of the micropump have been developed. Based on this knowledge the lumped parameter model of the entire two-stage micropump is composed straightforward. The complete model is depicted in Figure 3-16 (a) and covers all fluidic resistances and inductances. It also includes 59 3 Two-stage micropump the capacitances of both membranes as well as those of potential gas bubbles. Voltage signal generators account for the pressure changes upon the piezoelectric actuation and constant voltage sources represent the pressure offsets at the fluidic ports. (a) (b) Figure 3-16: Fluidic network model of the two-stage micropump: complete model accounting for all fluidic resistances as well as the inertia of the fluid (a) and simplified model including only the predominant elements (b). Concluding from the discussions on the fluidic inertance its impact is neglected in a simplified lumped parameter model shown in Figure 3-16 (b). Moreover, the fluidic resistance Rchamber is also assumed to be negligible due to the predominance of the valve resistance. Therewith, the model requires only one pressure variable p to describe the pressure in the pump chamber. This pressure is assumed to be spatially leveled throughout the chamber at any time. 60 3 Two-stage micropump 3.2.8 Implementation and evaluation of the lumped parameter model The symmetry of the micropump design enables to model both valves identically. Substituting equation (3.13) into equation (3.17) yields the expression for the instantaneous flow rate through one valve ·∆ · 6 · ∆ · . (3.25) Here, p denotes the chamber pressure, p0 is the ambient pressure and ΔpV is the pressure drop across the valve. The subsequent integration of the instantaneous flow rate Q in time gives the change of the chamber pressure ∆ 1 . (3.26) This in turn has an impact on the membrane deflection and consequently changes the instantaneous flow rate in accordance to equation (3.25). The lumped parameter model needs to account for this dependency via the implementation of a feedback loop. Figure 3-17: Block diagram of the lumped parameter simulation model. SIMULINKTM (MathWorks Inc., Natick, MA, USA) is utilized to implement and solve the lumped parameter model. Figure 3-17 illustrates the main blocks embedded in the micropump model. Both the inlet and the outlet valve block solve equation (3.25) in order to 61 3 Two-stage micropump determine the instantaneous inflow and outflow which are subsequently summed up to get the net inflow. In accordance to equation (3.26) the chamber pressure p is then obtained by integration of the net inflow. The two main blocks at the bottom cover the piezoelectric actuation following equation (3.5). The lumped parameter model is employed for an investigation of the main parameters of interest. First, it is solved for the time-dependent pressure in the pump chamber in order to identify the minimum durations required for the individual phases. Related to this task, the critical simultaneous closing of the inlet valve and opening of the outlet valve is considered in detail to optimize the timing of the actuation scheme. Subsequently, the flow characteristics are analyzed by means of the lumped parameter model including the backpressure stability and the voltage-controlled adjustment of the stroke volume. Later on in chapter 3.3.2 the impact of an additional fluidic capacitance constituted by a trapped air bubble will also be studied based on the lumped parameter model. 3.2.8.1 Pressure in the pump chamber The actuation follows the 3-phase scheme referred to as standard actuation mode which was introduced in Figure 3-4. As mentioned above, it involves a simultaneous switching step of the inlet and outlet valve and applies the standard closing voltages of 140 V (inlet) and 80 V (outlet) as well as a common opening voltage of -80 V. The time-dependent pressure in the pump chamber obtained for a actuation frequency of 1 Hz is depicted in Figure 3-18. The applied parameters for this simulation are summarized in the table next to the figure. U1,closed 140 V U2,closed 80 V U1,open -80 V U2,open -80 V trefill 350 ms ttransfer 100 ms tdelivery 550 ms Figure 3-18: The lumped parameter simulation shows the time-dependent pressure p within the pump chamber. By the end of each cycle, the cut-off pressure pc remains in the chamber. The pressure curve points out that two of the three phases are time critical, namely the refill phase and the delivery phase. At the beginning of the refill phase an underpressure is generated by the opening action of the inlet membrane. The minimum duration of the refill phase is given by the relaxation time which is required to reach the equilibrium pressure state defined by the external pressure applied to the inlet port. In the depicted case, the external pressure at the inlet port is set to the atmospheric level. For the introduced design of the two-stage micropump the simulation result indicates a relaxation time of approximately 100 ms in this case. 62 3 Two-stage micropump The transfer phase is characterized by the critical simultaneous switching of both valves. While the closing inlet valve generates an overpressure it is partly compensated by the outwards bending of the opening outlet valve. The displacement volumes assigned to the two switching processes determine if a net forward flow is generated during the transfer phase. In addition, the exact voltage-controlled timing of this step is vital. By the end of this phase, the chamber pressure takes the value of the external pressure applied to the outlet port. As can be seen in Figure 3-18 the leveling of the chamber pressure is rapidly completed which enables a short duration setting for the transfer phase. During the delivery phase the chamber pressure slowly approaches the cut-off pressure pc since the valve gap diminishes continuously while the chamber pressure decreases. If the delivery phase is set too short the pressure in the pump chamber cannot reach the value of the cut-off pressure pc at the end of the pump cycle. Since that is a crucial part of the backpressure independent concept of this two-stage micropump, an extended amount of time should be allocated for the delivery phase. 3.2.8.2 Phase setting Instead of simultaneously closing the inlet valve and opening the outlet valve one could also think of a slight delay between the two actions. This effect has been analyzed with the lumped parameter model and the result is shown in Figure 3-19. A negative overlap describes the situation where the inlet is closed prior to the opening of the outlet valve. For a positive overlap the outlet valve is opened before the inlet valve is closed. The impact of this actuation scheme modification on the pressure curve is indicated in the small figures at the respective side of the diagram. Figure 3-19: Simulation of the flow rate at a frequency of 1 Hz in dependence of a simultaneous switching at the beginning of the transfer phase (overlap = 0) or a delayed switching (overlap <> 0). 63 3 Two-stage micropump When the inlet valve closes first (left side with negative overlap) an overpressure peak is generated at the beginning of the transfer phase followed by a negative pressure peak when the outlet valve opens. The overpressure is strong enough to displace fluid through the outlet valve although it is nominally closed. This effect occurs due to the lower outlet closing voltage of 80 V together with the elasticity of the membrane. The pump performance for a negative overlap is the same as for simultaneous switching. A constant flow rate and an excellent backpressure stability are predicted by the simulation run. Intuitively, it seems to be more reasonable to open the outlet shortly before closing the inlet valve. This scheme has already been applied for a two-stage pump by others [123] and indeed increases the flow rate. Nevertheless, a positive overlap adds a short phase to the actuation scheme where both valves are open and hence the backpressure characteristic suffers from this modification. Beyond a positive overlap of 15 ms a rapid decline of the flow rate sets in for an applied backpressure of 30 kPa. 3.2.8.3 Backpressure characteristic The major goal of the proposed two-stage concept is the achievement of a constant flow rate over a wide backpressure range. The lumped parameter model is suitable to explore the impact of various control parameters on the backpressure curve of the micropump. The result obtained for the design parameters introduced previously in this section is given in Figure 3-20. Figure 3-20: Lumped parameter simulation of the backpressure characteristic: A virtually constant flow rate up to the cut-off pressure pc is confirmed for low frequencies. The simulation confirms that a backpressure independent flow rate is feasible with the developed concept. For a low frequency of 0.25 Hz a virtually constant flow rate is predicted up to the cut-off pressure pc = 64.8 kPa. Beyond that pressure value, a rapid decline of the flow rate is expected. For the increased frequency of 1 Hz an undesired backpressure impact becomes noticeable. This effect arises from an inadequate duration of the delivery phase. Due to the shorter cycle 64 3 Two-stage micropump time at elevated frequencies the duration allocated for the delivery face is not long enough to fully complete the cycle and reach the cut-off pressure. This in turn disturbs the backpressure independent characteristic and is an inevitable problem at higher frequencies. 3.2.8.4 Voltage-controlled adjustment of the stroke volume The stroke volume denotes the fluid volume propelled per pump cycle. For obvious reasons it is dependent on the applied actuation voltages. The following diagram in Figure 3-21 indicates the change of the stroke volume with respect to a variation of the common upstroke voltage. The relationship turns out to be nearly linear which qualifies the upstroke voltage as a suitable parameter to adjust the stroke volume. Figure 3-21: A nearly linear relationship between the stroke volume and the upstroke voltage applied to both actuators is confirmed over a wide actuation voltage range. 3.3 Transport of gases and gas bubbles 3.3.1 Gas pumping mode From the micropump point of view the major difference between liquids and gases is the high compressibility of the latter. Considering the actuation scheme in Figure 3-4, the compressibility of the medium has an impact on its response to the simultaneous switching of the inlet and outlet valve. For a liquid, the largest fraction moves from the rear part to the front part of the pump chamber because the small gap at the valve seat, which corresponds to a high fluidic resistance, avoids significant backflow through the inlet. In contrast, the gas can easily escape through the tiny gap and, hence, the forward movement is less pronounced. Therefore, in the gas pumping mode an intermediate phase is inserted in the actuation sequence where both valves are open for a short time (Figure 3-22). As discussed in section 3.2.8.2 this modification strengthens the propulsion of the fluid. A similar actuation sequence has been reported for a two-stage gas micropump on a centrifugal platform by Haeberle et al. [123]. As a drawback of this additional phase the micropump is open for a 65 3 Two-stage micropump short period of time and, in consequence, the flow rate is not independent of the outlet pressure any more. Figure 3-22: Modified actuation sequence optimized for pumping of gases. Two additional short phases strengthen the displacement of fluid in forward direction. The second modification of the actuation sequence concerns the closure of the outlet valve. For an incompressible liquid, the outlet valve cannot close unless liquid is removed from the pump chamber. In contrast, the gas would be simply compressed within the pump chamber and the outlet valve would be closed without gas being removed from the pump chamber. Therefore, in the gas pumping mode the outlet is being closed in a two step process where the outlet membrane is first released to its flat position and subsequently closed by applying a positive voltage (Figure 3-22). 3.3.2 Compressibility of entrapped air bubbles Compared to pure liquid or gas transport the occurrence of gas bubbles or cavities entrapped in the liquid stream tremendously increases the complexity of the situation since interfacial phenomena such as capillary effects or wetting aspects come into play. Depending on the size and location of the bubble enormous pressure drops across the interfaces can likely cause a failure of the micropump. Therefore, it is important to analyze the impact of a twophase gas-liquid suspension entering the two-stage micropump and to reveal the contribution of several related effects. The situation which is most easily described is that a gas bubble is permanently trapped somewhere in the pump chamber. Gas bubbles are preferably trapped near the corners of the micropump chamber away from the flow pathway. This phenomenon has been investigated in a semi-transparent prototype of the pump (Figure 3-23). Here, the bottom part of the micropump is made from polyurethane which is structured in a replica molding process (see chapter 8). 66 3 Two-stage micropump Figure 3-23: Photograph of a semi-transparent micro-pump with entrapped gas bubbles in the pump chamber. From the modelling point of view such a completely entrapped gas bubble constitutes an additional fluidic capacitance depending on the size of the bubble. The capacitance Cbubble is given by ∆ ∆ (3.27) in which ΔV is the change of volume caused by a pressure variation Δp. Assuming that the interfacial energy remains constant the contraction or expansion of the gas bubble follows the gas law. This assumption holds true if the curvature of the gas-liquid interface and hence the pressure drop is not considerably changed due to the contraction. It applies to a situation as depicted in Figure 3-24 where the gas fills the entire right ending of the pump chamber. If the bubble is exposed to a pressure change the meniscus will shift right or left but the size of the gas liquid interface and the curvature will remain unchanged. The overall interfacial free energy of the system is only slightly changed due to the shifted meniscus position along the wall but this effect is almost negligible. Figure 3-24: Gas bubble trapped in the corner of the pump chamber. The contraction or expansion of the gas bubble can be modelled either adiabatic or isothermal. Due to the comparably low thermal conductivity of gases a rapid change of the volume is modeled best by an adiabatic approach. Here, an increased pressure leads to a contraction of the volume but also to a temperature increase. For the adiabatic, reversible expansion of a perfect gas the equation · κ constant (3.28) relates the pressure to the volume. The adiabatic coefficient κ is determined by the quotient of the molar isobaric heat capacity and the isochoric heat capacity and takes the value κ = 1.4 for air. Given an air bubble with an initial volume V0 at the pressure p0, the volume is changed to 67 3 Two-stage micropump (3.29) · at the pressure p. That is, the change of volume amounts to ∆ · (3.30) 1 . In the electric circuit model (see Figure 3-16) the entrapped gas bubble constitutes a capacitance in parallel to the membrane capacitances. Consequently, the overall volume change due to a variation of the chamber pressure is achieved by linear superposition ∆ ∆ bubble membrane,i ·∆ . (3.31) For a pressure change from p1 to p2 the equations (3.30) and (3.31) yield ∆ · membrane,i · . (3.32) The derived equation points out that the capacitance of the gas bubble is nonlinear and depends on both the previous and the actual pressure. For the simplicity of the model it would be desirable to describe the gas bubble via a constant capacitance Cbubble = -ΔV/Δp which is not achieved with the adiabatic approach. For this reason, the expansion or contraction of a gas bubble is frequently modeled by an isothermal approach. Here, the gas law predicts for a perfect gas · constant . (3.33) Therewith, if a gas bubble with an initial volume V0 at the pressure p0 is exposed to a pressure change Δp, the relationship · ∆ ∆ ∆ ∆ ∆ ∆ (3.34) yields for sufficiently small changes – i.e. the term ∆p·∆V is neglected - the constant capacitance ∆ ∆ . (3.35) This way, the capacitance of the gas bubble is given by its volume V0 at the initial pressure p0, e.g. the original volume of the bubble at atmospheric pressure. Introducing the modulus of volume compressibility ∆ ∆ 68 (3.36) 3 Two-stage micropump the capacitance is expressed as . (3.37) For isothermal expansion or contraction of a gas bubble at atmospheric pressure the modulus of compressibility K is constant and takes the value K ~100 kPa which becomes obvious by comparing equation (3.35) and (3.37). Figure 3-25: Simulation result of the flow rate variation at a frequency of 1 Hz caused by trapped gas bubbles. The isothermal modelling of a gas bubble inside the pump chamber has been implemented into the lumped parameter model presented in chapter 3.2.7. The simulation emphasizes the impact of an additional fluidic capacitance arising from the gas bubble on the flow rate of the micropump (Figure 3-25). The indicated range covers the volume of the typically involved gas bubbles. For design II the pump chamber encloses a volume of 1.02 µl i.e. the upper end of the depicted range corresponds to an entirely gas-filled chamber. Especially if the capacitance of the gas bubble matches or exceeds the capacitance of the diaphragm C = 1.17·10-15 m3 Pa-1 the flow rate falls short of the flow rate expected without the gas bubble. 3.3.3 Gas-liquid interfaces In case of alternate pumping of liquid and gas the capillary effect exhibits a strong impact on the micropump dynamics. This situation occurs if a gas bubble exceeds a critical size where its interface spans over the entire cross-section of the micropump. Then, the gas bubble is placed in the flow pathway and the gas-liquid interface has to be displaced by the induced actuation pressures. The evoked capillary forces play an important role due to the critical valve geometry with narrow contractions and sharp edges. The simplest case of two-phase flow is a single gas-liquid interface that is moved from the inlet of the micropump throughout the entire pump chamber to the outlet. This situation 69 3 Two-stage micropump occurs if the liquid reservoir is removed from the inlet of the liquid-filled micropump and subsequently the pumping is continued with the inlet exposed to atmosphere. The interior material of the micropump is SiO2 or Si with subsequent Caro clean, both rendering a rather hydrophilic surface. The discussion in chapter 2.1.2.5 based on reported values as well as on own measurements points out that the expected contact angles for the interior of the micropump are in the range of 30 - 45°. Especially within the small gap between the valve lips and the membrane this hydrophilic behaviour impedes the displacement of the liquid surface. The height of this gap measures 1 µm for the undeflected membrane and reaches a maximum value of approximately 5 µm for the typically applied opening voltages of -80 V. Following the derivation in chapter 2.1.2.6 the curved surface causes a significant Young-Laplace pressure drop (Table 3-3) based on equation (2.23). Table 3-3: Young-Laplace pressure drop of a curved water surface (σ = 0.0725 N/m) Contact angle Gap height [µm] Young-Laplace pressure drop [kPa] 30° 1 125.6 30° 5 25.1 45° 1 102.5 45° 5 20.5 During the pump cycle two scenarios have to be considered where the interface is potentially pinned to the valve seat which consequently disrupts the pumping progress. First, the interface could be trapped at the inlet valve seat as depicted in Figure 3-26 (a). In this case, the pump chamber is still entirely filled with liquid. For the second scenario, the liquid has already retreated to the outlet where the interface gets stuck in between the valve lip and the membrane of the outlet valve (Figure 3-26 (b)). For both situations an equilibrium gap height and chamber pressure can be calculated based on the analytical modelling approach. For the following derivation the reference pressure applied to both inlet and outlet port of the micropump is assumed to be at atmospheric level. The Young-Laplace equation (2.22) constitutes a decreased chamber pressure for the first scenario and an elevated chamber pressure in the second case. On the other hand, the gap height h has been derived as a function of the chamber pressure in equation (3.13). A graphical solution to this problem is given in Figure 3-26 where the intersection points of both curves denote possible equilibrium positions. In Figure 3-26 (a), the solution corresponding to the lower gap height identifies an instable equilibrium position which means that a small displacement from the equilibrium point triggers a repelling force. Thus, only the second solution constitutes a stable equilibrium position and has to be considered as relevant solution for the indicated problem. 70 3 Two-stage micropump (a) (b) Figure 3-26: Gas-liquid interface trapped between valve lip and membrane at the inlet valve (a) or outlet valve (b). Due to the Young-Laplace pressure drop an equilibrium position for the gap height is determined where the negative pressure (a) or excess pressure (b) in the chamber matches the capillary pressure drop (marked by the dotted lines). An actuation step vulnerable to failure due to capillary forces is the opening of the valve. Here, the interface area has to be increased. Particularly, at the beginning of the opening process the heavily curved interface corresponding to the minute gap height provokes exceptionally large Young-Laplace pressure drops. The diagram in figure 3-26 (a) points out that a restoring force attempts to close the gap again unless the gap height has overcome the instable equilibrium point. This phenomenon is well-know from experiments where two plates, which are bonded together by means of a spreading liquid droplet, are attempted to be separated by forces pointing in normal direction (see chapter 2.1.2.6). The second concern is that the calculated Young-Laplace pressure drop for the stable equilibrium positions is in the range of the chamber pressure as expected during the pump cycle. Particularly in the case depicted in Figure 3-26 (b) where the entire pump chamber is filled with gas the compressibility of the medium is critical for the pumping process. Assuming isothermal compression the stroke of the inlet valve induces an excess pressure of Δp = 11.48 kPa which corresponds to a gap height of h = 5.38 µm at the outlet valve (see Appendix D). For these calculations the typically applied upstroke voltage of -80 V has been considered. Thus, depending on the actual contact angle, capillary pressure drops of up to 25 kPa are expected across the curved surface (see Table 3-3) which exceeds the overpressure in the pump chamber. In consequence, the gas volume would be simply compressed within the pump chamber without displacing the interfacial line from the outlet valve. Moreover, the subsequent closing of the outlet valve would further compress the gas but still the overpressure remains below the cut-off pressure pc2 of the outlet valve, i.e. the outlet valve is immediately closed without a net outflow from the micropump. 71 3 Two-stage micropump 3.3.4 FEM simulation of capillary forces For the validation of the impact of capillary pressure drops a 2D-FEM simulation model of the critical valve lip region has been established. The geometry of the model shows a small channel representing the gap between valve lip and membrane (Figure 3-27). The height of this channel is set to h = 5 µm in correspondence to the previous section. To both ends of the channel a sudden widening of the fluidic domain accounts for the pump chamber on the left side and the fluidic port on the right side. The fluidic meniscus is initially located at half way of the channel. Here, the left part representing the pump chamber is assumed to be gas filled, whereas the right part of the fluidic domain is filled by water. This corresponds to the situation when the meniscus is trapped at the outlet valve. A contact angle of θ = 30° is preset as boundary condition for the channel walls. The left and right edges are defined as fluidic inlets with pressure boundary conditions. An unstructured mesh with 7966 triangular elements and a maximum element edge length of 0.75 µm is chosen for the model. Particularly for increasing pressures applied to either of the two fluidic inlets a refined mesh is crucial for the convergence of the solver since steep pressure gradients occur across the meniscus. A direct solver using variable time steps is deployed for this problem. First, an initial solution is determined and saved as a basis for the subsequent time-dependent simulation. The solver couples the Navier-Stokes-equation for the fluidic displacement with a level-set method to trace the gas-liquid interface. This level-set function φ is determined by the equation [124] Φ · Φ · Φ 1 Φ Φ | Φ| · Φ 0 (3.38) and is recalculated for each time step based on the instationary and the convective term of the Navier-Stokes-equation. The level-set function returns values between zero and one with the interface being represented by a value of 0.5. The simulation sequence shown in Figure 3-27 illustrates that the capillary force displaces the meniscus towards the pump chamber until the sharp bending at the end of the gap channel is reached. Here, the equilibrium contact angle is established and the curvature of the meniscus is almost vanished. At this point a further wetting of the side walls would energetically be cancelled by a necessary increase of the gas-liquid interface area and hence an equilibrium position is found. 72 3 Two-stage micropump Figure 3-27: 2D-FEM simulation of the capillary effect in a small gap with a height of 5 µm (no external pressure applied). For increasing pressures applied to the left boundary, the velocity of the meniscus displacement is reduced. The reason behind is that the external hydrostatic pressure gradient moves the fluid to the right whereas the capillary force draws the fluid towards the left side. In consequence, the difference between the hydrostatic pressure difference and the capillary pressure drop constitutes the driving force on the fluid. From theory the meniscus should not move to either side when the external pressure difference exactly matches the capillary pressure drop. The analytical solution for the net mean flow velocity based on the pressure driven laminar flow through a small gap of height h (equation (2.16)) superimposed by the capillary pressure drop reads 12 · ∆ 2 · cos . (3.39) This mean flow velocity equals the displacement velocity of the meniscus for an incompressible liquid and for the assumption that the equilibrium contact angle of the system is preserved regardless of the applied external pressure Δpext. For a comparison of the FEM simulation with the analytical prediction the displacement of the meniscus is extracted from the simulation results. Here, the position of the meniscus in the middle of the gap channel is plotted as a function of the simulation time and the velocity of the meniscus is obtained from the slope of the curve (Figure 3-28 (a)). The slightly non-linear appearance is caused by the non-constant length l of the liquid slug. Initially the meniscus is located at half way of the channel at the position x = 0 which corresponds to a slug length of 73 3 Two-stage micropump l0 = 50 µm. The slope is determined for the linear segment in Figure 3-28 (a) which corresponds to an approximate slug length of l = 75 µm. (a) (b) Figure 3-28: The liquid gas interface is displaced within the small gap due to the capillary force (a). An additional external pressure gradient affects the meniscus displacement velocity (b). The simulation is carried out for small external hydrostatic pressures only. For larger pressures opposing the capillary pressure drop the simulation does not converge and therefore result markers are missing in that range. Obviously, there is a deviation between the velocities extracted from the simulation results and the analytical approach (Figure 3-28 (b)). Nevertheless, the tendency is in good agreement and the intercept point at the pressure axis is identical for both the analytical curve and the extrapolated simulation curve. The solution indicates a capillary pressure drop of Δpc = 25 kPa. 3.3.5 Critical compression ratio The only promising measure to overcome the obstruction caused by the capillary pressure drop is an increase of the compression ratio ∆ (3.40) which relates the nominal displacement volume ΔV± of the actuator to the dead volume V0 of the pump chamber. In the literature a minimum compression ratio of ε = 0.07 has been reported as a criterion to judge the ability of a micropump to transport gas bubbles [7]. Nevertheless this criterion does not strictly apply for the presented two-stage micropump design since it has been derived for a single membrane micropump with passive check valves. There, the pressure drop across the passive check valve needs to be balanced by the excess pressure induced in the pump chamber in order to maintain the forward propulsion of the fluid. Considering the active valves of the presented two-stage micropump design a modified criterion can be derived which defines a critical compression ratio εcrit by accounting for the 74 3 Two-stage micropump Young-Laplace pressure drop across the valve gap. As derived in chapter 2, this capillary pressure drop amounts to · ∆ , (3.41) , with the surface tension σ, the contact angle θ and the gap height h2 of the outlet valve. The gap height is obviously a function of the design-specific gap height h0 underneath the undeflected membrane, the upstroke voltage U2 of the outlet actuator and is - due to the compliance of the diaphragm - also a function of the chamber pressure p. The increase of the chamber pressure upon the closing stroke of the inlet actuator needs to outbalance this capillary pressure drop, i.e. the chamber pressure has to exceed the pressure calculated for the stable equilibrium position in Figure 3-26. For isothermal compression the basic equation · ∆ · (3.42) gives the solution for the expected pressure increase Δp. At the beginning, the atmospheric pressure p0 is found in the pump chamber enclosing a total volume V1. After closing the inlet valve, the total volume of the pump chamber has decreased to the value V2 which causes a pressure increase by Δp. Solving equation (3.42) for Δp yields ∆ · · · (3.43) ⁄ has been substituted by the compression ration ε. Concluding where the term from this inequation the pressure increase Δp definitely exceeds the capillary pressure drop Δpc in case of · ∆ (3.44) which adds a safety margin to the estimation of the critical compression ratio crit 2· · · , . (3.45) Here, the influence of the unknown chamber pressure on the gap height h2 is also neglected which further increases the safety margin of the derived estimation. 75 3 Two-stage micropump Table 3-4: Compression ratio for different chamber designs and upstroke voltages Design Ι Design ΙΙ εΙ εΙΙ 0.26 0.03 0.12 -110 0.2 0.04 0.14 45° -80 0.21 0.03 0.12 45° -110 0.16 0.04 0.14 Contact angle Upstroke voltage [V] εcrit 30° -80 30° Table 3-4 shows how the dead volume of the presented chamber designs and the upstroke voltage of the piezo-actuators affect the compression ratio. The compression ratios obtained for design I are approximately one order below the critical compression ratios, that is, bubble tolerant pumping cannot be expected for a micropump based on design I. The reduced chamber volume of design II increases the compression ratio approximately four-fold. For this design, the compression ratio of the micropump nearly matches the determined critical compression ratio when large upstroke voltages are applied. A further increase of the compression ratio by implementation of an even smaller pump chamber has been investigated experimentally in the master thesis by S. Nadir [125]. However, a stable and reproducible delivery performance was not achieved with these design variations which is presumably attributed to the small volume displacements of the two-stage concept. Moreover, higher upstroke voltages would further increase the compression ratio due to a larger actuator stroke volume. In this work, the maximum upstroke voltage was limited to -110 V by the employed piezoceramic material since depolarization effects set in beyond this voltage. 76 3 Two-stage micropump 3.4 Single-membrane micropump The single-membrane micropump demonstrates a design variation of the proposed twostage concept which is characterized by an enlarged membrane spanning over both valves. The extended membrane still features two actuated regions thus it also belongs to the family of two-stage micropumps. For piezoelectric actuation the position of the PZT discs is identical to the two-membrane concept, i.e. the piezo-actuators are located right above the fluidic valves. In chapter 7 the design of the single-membrane micropump will also be discussed in conjunction with thermal actuation by means of paraffin. The main benefit promised by this modification is a larger deflection and hence an increased displacement volume due to the reduced stiffness of the membrane. The schematic drawing in Figure 3-29 (a) shows the single-membrane setup combined with the established pump chamber geometry. Beyond that a low-cost design of the single-membrane micropump is considered which reduces the fabrication costs (Figure 3-29 (b)). While the standard fabrication process (see chapter 4) requires four lithography masks this low-cost alternative can be produced with two masks only. Moreover, the low-cost design abandons the valve lips fabricated by an expensive dry etching process. Thus, the bottom part of the micropump (chip 2) needs not necessarily to be made from silicon but could be replaced by a cheaper polymer or glass substrate which contains simple through-holes. (a) (b) Figure 3-29: Design options for the single membrane micropump: Standard twostage design equipped with a single membrane (a) and low-cost design with simple through-holes (b). The deflection of the enlarged single-membrane has been investigated by FEM simulations based on the same model as described above. The lateral dimensions are adjusted to the new membrane size of 8 x 16.9 mm2. The result points out that the deflection is increased by approximately 10% in comparison to the two-membrane micropump (Figure 3-30). For this simulation, the inlet valve is kept open by an upstroke voltage of -80 V whereas the outlet valve is closed with the membrane touching the valve lips. 77 3 Two-stage micropump Figure 3-30: The lower flexural rigidity of the extended single membrane enables a 10% larger deflection compared to the two-membrane concept. The single-membrane design is also supposed to strengthen the peristaltic motion during the transfer phase. A transient FEM simulation of the simultaneous closing of the inlet valve and opening of the outlet valve has been carried out to examine this aspect. Indeed, the obtained simulation sequence confirms a peristaltic motion of the membrane from its original state to the final state (Figure 3-31). 0 µs 50 µs 100 µs 150 µs 300 µs Figure 3-31: FEM simulations of the simultaneous switching process confirm the peristaltic motion of the single membrane featuring two piezoelectrically actuated regions. 78 Chapter 4 Fabrication of the micropump 4 Fabrication of the micropump The fabrication of the micropump is based on silicon micromachining using mainly standard MEMS processes such a vapor deposition, lithography or etching. The process chain has been developed and discussed in a thesis by A. Doll [114]. His established process sequence has been adapted to the novel two-stage design in this work. Mainly, a new set of lithography masks was created and the depth of the various etching steps was modified in order to meet the specific requirement of the new design. 4.1 Silicon manufacturing The proposed micropump consists of two microstructured silicon wafers subsequently bonded together by means of a low temperature silicon wafer bonding process [114]. At the beginning, two 4”-silicon wafers (thickness 525 µm, n-doped, polished on both sides) are carefully selected to exhibit complimentary bows of similar magnitudes. The bow of the wafer has been identified as a crucial parameter to ensure a high bond quality. The two wafers referred to as top wafer and bottom wafer are microstructured using conventional bulk silicon processes. Readers interested in the fundamentals of these technologies are referred to comprehensive textbooks dealing with MEMS silicon processes [126, 127]. The process sequence for both wafers is illustrated in the following Figure 4-1. Over all, four masks are required for the standard lithography steps. The cavities which define the membranes as well as the fluidic ports are made by KOH etching, whereas the interior structures of the micropump, i.e. the pump chamber with the valve lips, are produced by deep reactive ion etching (DRIE). Subsequent to the structuring process an oxide is grown on the upper wafer as an insulating layer on the silicon diaphragms and to act as a bonding layer. 79 4 Fabrication of the micropump Top wafer Bottom wafer • Deposition of a photoresist layer (AZ1518) • UV-lithography using a chromium mask • Wet oxidation @ 950°C: SiO2 (300 nm) • LPCVD @ 770°C: Si3N4 (100 nm) • STS-ICP Advanced Silicon Etching: 1 µm (= gap height between valve lip and membrane) • Deposition of a photoresist layer (AZ1518) • UV-lithography using a chromium mask • Wet oxidation @ 950°C: SiO2 (300 nm) • LPCVD @ 770°C: Si3N4 (100 nm) • Structuring of the SiO2/Si3N4-layer • Deposition of a photoresist layer (AZ1518) • UV-lithography using a chromium mask • Fabrication of inlet and outlet ports • Structuring of the SiO2/Si3N4-layer • Deposition of a photoresist layer (AZ1518) • UV-lithography using a chromium mask • Fabrication of 100 µm silicon membrane • Structuring of the SiO2/Si3N4-layer 80 4 Fabrication of the micropump • • 5% HF solution • Backside protection with blue tape • Wet oxidation @ 950°C: SiO2 (400 nm) • STS-ICP Advanced Silicon Etching: 30 µm • Fabrication of the pump chamber RIE Bond activation: Ar+-plasma • 5% HF solution • Deposition of a Cr / Au layer (30 nm / 150 nm) Figure 4-1: Silicon fabrication process of the two-stage micropump. As preparation for the bonding process a thorough cleaning of both wafers is essential. Here, the bottom wafer is first treated with a Caro clean (sulfuric-peroxide mixture), then dipped into a 5%-HF solution before being treated with a Caro clean again. The top wafer covered 81 4 Fabrication of the micropump with the oxide layer is only treated with a Caro clean. Then, an oxide activation by means of an argon plasma is applied to the top wafer. Subsequently, both wafers are aligned to each other using a bond aligner and are bonded together. Upon overnight annealing at 150° C an irreversible bond is established between the two silicon wafers [128]. At the upper side of the bonded wafer stack, a chromium/gold layer is evaporated onto the membrane cavities to serve as ground electrode for the piezo-actuators. The wafer stack is subsequently diced to release the individual micropump chips. (a) (b) (c) Figure 4-2: Photograph of a two-stage micropump chip (a), REM-picture of the valve lip (b) and microscopic photograph of a cross-sectional cut of the valve (c). Figure 4-2 shows a micropump chip together with details of the interior of the pump chamber. As the most critical part of the design, the valve lips exhibit a height and width of 30 µm and 100 µm, respectively (Figure 4-2 (b)). The inner radius of the valve seat is 400 µm. The small gap between the valve lips and the membrane is visible in the microscopic photograph depicted in Figure 4-2 (c). 4.2 Back-end processes 4.2.1 Gluing of piezo-disks After termination of the cleanroom process the silicon microchip has to be equipped with piezo-actuators. Pressed piezoceramic discs (Stelco GmbH, Neumarkt, Germany) with a thickness of 200 µm were used for the actuator. The low porosity of pressed ceramics yields a large d31 coefficient which is advantageous in terms of deflection. Laser cutting was applied to trench the purchased discs. For this micropump the piezo-discs are cut to a standard size of 6.5 x 6.5 mm2. The piezoceramic discs are glued to the diaphragm by means of a low viscosity epoxy glue (Araldite© 2020, Huntsman Advanced Materials GmbH, Basel, Switzerland). For the electrical contact between the ground electrode evaporated onto the micropump chip and the piezodisc conductive carbon black particles are dispersed into the epoxy glue. For the described actuator size 40 µl of the glue are dispensed onto the middle of the membrane by means of a conventional pipette. The piezo-disc is then aligned on the membrane and the glue is cured at 80 °C for 30 min. 82 4 Fabrication of the micropump 4.2.2 Wire bonding Wedge-wedge aluminum wire bonding (Wedge-Wedge-Bonder 5430, F&K Delvotec Bondtechnik GmbH, Ottobrunn, Germany) is used to electrically connect the printed circuit board with the ground electrode (Au layer) and the upper electrodes of the piezo-actuators (Figure 4-3). Three to five redundant wire bonds are placed at each bond pad to reduce the probability of failure. (a) (b) Figure 4-3: Schematic illustration of the electrical connection (a) and photograph of the wire bonded actuators (b). 4.3 Quality control Fabricated micropumps have to pass an initial inspection protocol in order to discover defect chips. This procedure helps to avoid the assembly of systems with faulty micropump chips. The initial inspection includes a visual inspection of the bond quality, a detection of fluidic blockage and a verification of the piezo-actuator contacting. 4.3.1 IR inspection of bond quality Immediately after the bonding process, an infrared image is taken of the wafer stack. This technique is capable of revealing areas which exhibit a bond defect. The appearance of Newton’s rings indicates improperly bonded areas as shown in Figure 4-4. Bond defects Figure 4-4: Infrared image of the bonded silicon wafers. 83 4 Fabrication of the micropump 4.3.2 Fluidic test setup Occasionally, the small design-inherent gap between the valve lips and the membrane causes a permanent bond at the valve seat which leads to a complete blockage of the micropump. This failure is detected in a fluidic penetrability test. The micropump is placed into the test fixture shown in Figure 4-5 and a pressure driven flow between the two fluidic ports is achieved for penetrable micropumps. Inlet Outlet Figure 4-5: Setup for an initial test of the fluidic penetrability of the micropump. 4.3.3 Electrical capacitance measurement of the actuators The glue between membrane and piezo-actuator has to provide both a mechanical bond and an electrical connection via the conductive particles. For the given size of the piezo-discs, the capacitance between ground electrode and upper electrode of the piezo-actuator is expected to be in the range of 7 – 9 nF. A significant deviation of the capacitance value is an indicator for faulty electrical connections. 4.4 Fabrication costs The fabrication process for a micropump chip in the IMTEK cleanroom comprises the cleanroom processes described above as well as back-end processes such as piezo-disc mounting or wire bonding. The calculation summarized in Table 4-1 is based on the process costs for the fabrication of one silicon wafer stack i.e. one top wafer bonded to one bottom wafer. Since the process chain includes a number of time-intensive batch processes such as oxidation, physical vapor deposition of silicon nitride or KOH etching, an optimum batch size depending on the machine capacities would significantly reduce the costs per micropump chip. This way, the presented calculation gives an upper estimation of the expected fabrication costs. It further assumes that 12 of 17 micropump chips of the wafer stack are free from defects which corresponds to a yield of 70 %. This yield assumption has been made on the background of laboratory observations and clearly gives room for further improvements. For the low-cost design introduced in chapter 3.4 a reduction of the overall fabrication costs by approximately 10 % has been determined. This margin would also increase for larger batch sizes since the single-wafer dry etching process for structuring of the bottom wafer is eliminated in this design. 84 4 Fabrication of the micropump Table 4-1: Calculation of the micropump fabrication costs (IMTEK cleanroom). Deposition processes 23,00 € Lithography 9,30 € Etch processes 15,90 € Wafer bonding 18,00 € Other process steps 6,80 € Material costs (Silicon wafer, piezo-discs) 11,50 € Cost per micropump 84,50 € 85 4 Fabrication of the micropump 86 Chapter 5 Experimental characterization 5 Experimental characterization This chapter presents the results of a comprehensive experimental characterization of the micropump. The most important aspect is the variation of the flow rate with respect to the available control variables. An additional focus of the fluidic investigations is set on the impact of the capillary effect. Beyond that, the outcome of the geometry modification of the pump chamber is analyzed and experimental measurements of the single-membrane micropump are reported. 5.1 Experimental setup As a prerequisite for experimental measurements, a set-up needs to be established which is appropriate for the precise measurement of the desired parameters. For the experimental investigation of the micropump properties a sensitive and accurate method to measure the flow rate is required. In the framework of this thesis, different methods are employed for flow measurement (Figure 5-1). Most conveniently, a gravitational method based on a micro balance or a high-precision flow sensor is utilized. The third method is based on the hydrostatic pressure head in a vertical tube. The flow sensor is capable of resolving the individual flow pulses and therefore is the appropriate method for studying the transient response of the micropump. In contrast, the two other measurement methods return an integrated flow signal which is adequate for measurements of the average flow rate, especially if extended time intervals are considered. The hydrostatic pressure method is particularly suited for investigation of the backpressure performance since the static pressure head is continuously increased. In addition, a pressure controller is employed to generate an external pressure head and a pressure sensor is used for monitoring reasons. Finally, an electronic control unit is required for the actuation of the micropump that converts the intended actuation scheme into an appropriate voltage sequence. 87 5 Experimental characterization Figure 5-1: Block diagram of the measurement setup. 5.1.1 Micro balance The first method employed for flow measurements of liquids is a gravitational method based on a micro balance (ME36S, Sartorius AG, Goettingen, Germany, see Figure 5-2 (a)). Here, the liquid is pumped into or off an open vessel placed on the balance. The variation of the weight is detected over time and mathematically converted into the flow rate. In order to prevent evaporation the vessel is covered by a thin film of oil. Therewith, a stable read-out of the scale is achieved which is essential for long term measurements at low flow rates. The connection between the vessel and the fluidic tube is realized by means of an injection needle (Figure 5-2 (b)). (a) (b) Needle Oil layer H2O Figure 5-2: High precision micro balance (ME36S, Sartorius AG) providing a resolution of 1 µg (a). For the measurement, a needle is dipping into an open vessel filled with deionized water. The surface is covered by an oil layer in order to minimize evaporation errors (b). 5.1.2 Flow sensor An alternative method for flow rate measurement is the use of a commercial flow sensor. The high requirements regarding the resolution and precision of the instrument necessitate the availability of a high-performance sensor. As an ambitious specification, the flow sensor should be able to resolve the individual peaks of the pulsatile flow signal of the two-stage micropump. The device chosen for our setup is a thermal flow sensor (SLG1430, Sensirion 88 5 Experimental characterization AG, Staefa, Switzerland, see Figure 5-3) which features a time resolution of up to 5 ms. It is capable of resolving the flow peaks generated by the micropump and provides a robust and uncomplicated method to determine the flow rate. The specified flow range is between 1 - 40 µl/min. The drawback of this system is that the high peaks of the pulsatile flow signal frequently exceed the upper limit of 40 µl/min. Moreover, the small capillary within the flow sensor causes a pressure drop which constitutes a significant and flow rate dependent pressure load. A quantitative measurement of this effect yielded a sensor-attributed pressure drop of approximately 10 µl⁄Pamin . Figure 5-3: High-performance flow sensor (SLG1430, Sensirion AG) resolving flow rate variations of 7 nl/min within a range of 1 - 40 µl/min. 5.1.3 Hydrostatic pressure method A simple yet precise method to measure the flow rate at different pressure loads is the use of the hydrostatic pressure head in a vertical tube. The measurement setup comprises a rigid tube with an inner diameter of 1.2 mm and a length of 3.5 m which is vertically mounted to a pole. At the bottom point of the tube a pressure sensor (see section 5.1.4) measures the hydrostatic pressure head acting on the pump outlet. From the recorded pressure signal the volumetric flow rate Q is immediately obtained by · · · (5.1) where r is the diameter of the tube, ρ is the density of the fluid, g represents the gravitational constant and dp/dt is the slope of the recorded pressure curve. 5.1.4 Pressure sensor A commercial piezoresistive silicon blood-pressure sensor (MPX2300D, Freescale Semiconductor Inc., Austin, TX, USA) is used for pressure monitoring. It covers a pressure range up to 40 kPa. The fluidic interface of the sensor, i.e. a small cavity above the silicon diaphragm, is gel-filled to avoid signal distortion due to gas cavities. The photographs in Figure 5-4 show the small package of the pressure sensor which is mounted to a flowthrough housing made of PMMA. This medical grade sensor is also integrated into the concept of the active microport which will be presented in chapter 9. 89 5 Experimental characterization (a) (b) Figure 5-4: Medical grade silicon pressure sensor (MPX2300D, Freescale Semiconductor) (a) mounted to a fluidic channel (b). 5.1.5 Pressure controller A pressure controller (DPI 520, GE Sensing, Bad Nauheim, Germany) is part of the experimental setup to provide preset external pressures. The instrument is connected to the PC via the RS-232 interface and the pressure setting is controlled by means of LabView (National Instruments Corp., Austin, TX, USA). The provided gas pressures range up to 400 kPa. 5.1.6 Electronic control unit for the micropump For the electronic control of the micropump actuators a specific electronic driving circuit depicted in Figure 5-5 (a) was developed in a diploma thesis by M. Heinrichs [129]. It contains a dual step-up converter to provide voltages between -150 V and +250 V. A flexible, software-controlled implementation of the actuation scheme is realized by means of an 8-bit microcontroller with an associated non-volatile memory (EEPROM). This electronic device is placed into an electronic control unit for experimental evaluations in the laboratory (Figure 5-5 (b)). Herein, an additional circuit board provides the connection to the computer via a RS-232 interface. It also contains a readout circuit for the pressure sensor. At the computer side, a LabView program is utilized to define the actuation scheme and to record the pressure sensor signal. (a) (b) Figure 5-5: The electronic driving circuit for the piezo-actuators (a) is contained in an electronic control unit providing an RS-232-interface (b). 90 5 Experimental characterization 5.2 Variation of pressure The stability of the flow rate against pressure variations applied to the fluidic ports is a decisive criterion for the applicability of the micropump. The experimental measurements presented in this section reveal the characteristics of the developed two-stage micropump when exposed to external pressures. The detailed studies are focused on the pressure range up to 30 kPa which accounts for the requirements of the intended biomedical application. In the physiological environment, i.e. for in-vivo applications such as intravascular drug administration, pressures exceeding 30 kPa are not expected to act on the micropump during regular operation. In addition, an investigation of the maximum backpressure was carried out. 5.2.1 Backpressure independence of the flow rate A main focus of the investigation was the backpressure characteristic of the micropump. The backpressure describes the static pressure head which is applied to the outlet with respect to atmospheric pressure. For this measurement the inlet pressure was set to atmosphere. The voltages applied to the piezo-actuators took the values indicated in the standard actuation scheme in Figure 3-4, i.e. an upstroke voltage of -80 V was applied to both actuators and the closing voltage was set to +140 V and +80 V for the inlet and the outlet actuator, respectively. The result plot in Figure 5-6 was obtained by means of the hydrostatic pressure method for a micropump with a rectangular pump chamber (design I). The measurement was carried out with deionized water at room temperature. Figure 5-6: Flow performance of the micropump P26: a stable flow rate up to a backpressure of 30 kPa is proven for low frequencies. For low frequencies a virtually constant, backpressure independent flow was confirmed within a range between 0 and 30 kPa. For an actuation frequency of 0.25 Hz the flow rate decreased to 90% of the initial value at p90% = 27 kPa. For increasing frequencies the impact of the backpressure became more pronounced and the 90%-barrier of the flow rate was found at p90% ~ 15 kPa. The declined performance is caused by the shortened durations of 91 5 Experimental characterization the refill and delivery phase which has been revealed as a potential deteriorating factor by the previous simulations (see chapter 3.2.8). This effect will be investigated experimentally in section 5.3.2. A maximum backpressure of 65 kPa has been achieved with the two-stage design. The corresponding backpressure curve is depicted in Figure 5-7 (a). The diagram on the right side (Figure 5-7 (b)) redisplays the simulation results as presented in chapter 3.2.8.3. A good agreement is confirmed in terms of expected zero-pressure flow rates and the characteristics of the curves. However, a discrepancy is revealed concerning the maximum backpressure and the value of the cut-off pressure pc. (b) (a) Figure 5-7: Backpressure curve of micropump P32 for an extended pressure range (a) and comparison to the simulation result (b). 5.2.2 Impact of forward pressures The following experiment considered the impact of a pressure offset applied to the inlet. While the micropump is optimized to withstand outlet pressure heads, the effect of a forward pressure is not eliminated by the design. Here, a significant and nearly linear correlation between the applied pressure and the flow rate became apparent (Figure 5-8). μ 0.226 Figure 5-8: Impact of a static pressure head applied to the inlet (forward pressure) (P26). 92 5 Experimental characterization For the intended application in a drug delivery system the micropump has to transport the liquid from a reservoir to the delivery site. The measured characteristic implies that the pressure in the reservoir should be stabilized at atmospheric pressure in order to avoid dosing errors. The stable inlet pressure can be realized either by means of a vented reservoir or by an elastic balloon with a sufficiently large external gas buffer. These methods are considered appropriate as the flow rate of this micropump is low and thus the inlet pressure changes smoothly in response to the withdrawn fluid. 5.2.3 Impact of common mode pressures To combine the two situations analyzed in the preceding subsections, an equal pressure offset was applied to both fluidic ports. In analogy to the electric circuit theory this case is termed common mode pressure stability. Figure 5-9 clearly points out that the common mode pressure affects the flow rate. In this case the susceptibility of the flow rate to forward pressures is the prevailing implication while the impact of the outlet pressure head is still suppressed. Nevertheless, the outlet pressure head causes a reduced slope of the flow rate curve compared to the situation of sole forward pressure (0.144 (µl min-1)/kPa versus 0.226 (µl min-1)/kPa). The lumped parameter simulation confirms this common mode pressure effect even though the observed flow rates for the inspected micropump P26 are larger than the simulated values. μ 0.144 Figure 5-9: The susceptibility of the flow rate to forward pressures yields a common mode pressure effect, i.e. an increased flow rate if the pressure head is applied to both inlet and outlet (P26). 5.3 Variation of frequency For piezoelectric micropumps, the flow rate is usually adjusted via the actuation frequency. Typically, the flow rate scales linearly with the applied frequency within a certain frequency range. This section explores the correlation between the flow rate and the actuation frequency and illustrates the respective impact on the stroke volume. 93 5 Experimental characterization 5.3.1 Flow rate versus frequency The following Table 5-1 lists the time settings for the three phases of the actuation scheme for the typically applied frequencies. Again, the voltages utilized for this investigation were set to the standard voltage levels +140 V / + 80 V and -80 V as denoted above. Table 5-1: Time settings of the actuation sequence for different frequencies. Frequency [Hz] Refill phase [ms] Transfer phase [ms] Delivery phase [ms] 0,125 2000 100 5900 0,25 1500 100 2400 0,5 800 100 1100 1 350 100 550 2 100 100 300 4 70 30 150 6 47 20 100 8 40 15 70 10 30 15 55 12 22 15 45 The result depicted in Figure 5-10 (a) indicates that the low frequency regime up to 4 Hz is characterized by a nearly linear increase of the flow rate. It also points out that the backpressure stability declines for higher frequencies confirming the results of the previous section. (a) (b) Figure 5-10: The measurement of the flow rate for different actuation frequencies revealed a linear frequency scaling for low frequencies (a) and a subsequent decline towards higher frequencies (b) (P19). 94 5 Experimental characterization A linear correlation between the flow rate and the actuation frequency implies a constant stroke volume delivered at each pump cycle. By approximation, this assumption held true for low frequencies but was violated for higher frequencies. In consequence, the flow rate became non-linear towards higher frequencies (Figure 5-10 (b)). This behaviour is typical for micropumps and corresponds to a decreasing stroke volume. The maximum flow rate achieved with this two-stage micropump design was found in the range of 100 – 150 µl/min and was reached for a frequency of approximately 20 – 30 Hz. 5.3.2 Stroke volume versus frequency The stroke volume delivered at each pump cycle is affected by the actuation frequency. As described above, the shorter cycle time at higher frequencies necessitates a shorter duration of the individual actuation phases. This is, in particular, critical for the delivery phase. In Figure 5-11 the pulsatile flow signal was recorded for two different frequencies using the flow sensor. The sampling rate for this recording was set to 50 Hz and 100 Hz, respectively, which enabled a detection of the individual flow pulses. The result indicates that the flow pulses at a frequency of 1 Hz have not completely died off at the arrival of the subsequent pulse. This effect explains the observed decrease of the stroke volume. (a) (b) Figure 5-11: The flow signal recorded by means of the flow sensor points out that the pulse length is in the range of 1s (P32). The data given above in Figure 5-10 (b) are rearranged to illustrate the development of the stroke volume with respect to the actuation frequency (Figure 5-12). Since the flow rate is calculated from the product of actuation frequency and stroke volume, a double logarithmic scale is chosen in order to illustrate the effect of these two parameters on the flow rate. For a low frequency of 0.125 Hz a stroke volume of approximately 200 nl was obtained. A rapid decline of the stroke volume was observed for frequencies beyond 30 Hz. 95 5 Experimental characterization Figure 5-12: Stroke volume versus actuation frequency displayed in a double logarithmic scale (P19). 5.4 Phase setting of the actuation sequence Among the different timing options of the actuation sequence the phase setting at the beginning of the transfer phase is the most crucial parameter. As standard sequence, a simultaneous closing of the inlet valve and opening of the outlet valve is proposed for the working principle of this two-stage micropump. The impact of an asynchronous switching has already been investigated by means of the lumped parameter simulation in chapter 3.2.8.2 and the simulation outcome is redisplayed in Figure 5-13. A negative overlap describes the situation where the inlet is closed prior to the opening of the outlet valve. For a positive overlap the outlet valve is opened before the inlet valve is closed. Figure 5-13: Investigation of the optimal phase setting at the beginning of the transfer phase (P32). 96 5 Experimental characterization The experiments clearly confirmed the effect predicted by the simulation. A small negative overlap was comparable to a simultaneous switching in terms of flow rate and backpressure stability. Towards positive overlaps an increase of the flow rate became noticeable. Up to a positive overlap of 15 ms the backpressure stability was still preserved. For larger positive overlaps the backpressure stability degraded due to the extended time where both valves are open. Thus, the proposed simultaneous switching is appropriate for the two-stage micropump to ensure a backpressure-stable flow rate. As an alternative, an intermediate phase setting with a short positive overlap up to 15 ms is capable of strengthen the fluid propulsion. This characteristic is utilized for the actuation scheme referred to as gas pumping mode (see chapter 5.7). 5.5 Variation of the control voltage Together with the actuation frequency and the phase settings, the applied voltage levels complete the set of parameters to electrically control the performance of the micropump. The subsequent sections will reveal how the voltage levels affect the flow rate and the backpressure stability of the micropump. 5.5.1 Opening voltage The piezoelectric actuation enables a comprehensive electrical control of the flow characteristic. The stroke volume of the micropump (design I) is adjustable between 50 200 nl by variation of the piezo-actuators upstroke voltage (Figure 5-14). This way, the resolution of the flow rate setting is adaptable to the specific application. It is also beneficial for minimizing the required power since for higher flow rates the stroke volume can be increased more power-economic than the actuation frequency. Moreover, an increase of the upstroke voltage also raises the compression ratio which has proven as an appropriate measure to overcome the blockage of the pumping process due to capillary forces or small gas bubbles. Figure 5-14: The stroke volume of the micropump is adjustable by means of the upstroke voltage applied to both actuators (P26). 97 5 Experimental characterization 5.5.2 Closing voltage of the outlet valve As mentioned in chapter 3.1, the closing voltage applied to the outlet valve is another crucial parameter of the working principle of the two-stage micropump. The applied voltage needs to be high enough in order to withstand the external outlet pressure. On the other hand, a maximum closing voltage has been identified by the analytical derivation in chapter 3.2.3 in order to keep the inlet valve closed when closing the outlet valve. There, the graphical solution to this problem defined a set of requirements for the applicable voltage levels in order to comply with the analytical constraints. For experimental investigations, two cases were considered where the closing voltage of the inlet valve was set to 140 V and 185 V, respectively. For both cases, the outlet closing voltage was varied and the impact on the flow rate was studied at different backpressures (Figure 5-15). (a) (b) U1 = 140 V U1 = 185 V Figure 5-15: Flow measurements at a frequency of 1 Hz indicate the appropriate outlet voltage range in order to ensure a backpressure independent flow rate for an inlet closing voltage of 140 V (a) and 185 V (b) (P32). It is apparent that the flow rate declines for an increased closing voltage of the outlet valve. The shaded sector delineates the appropriate voltage range in order to achieve a backpressure independence of the flow rate. In this range, all flow rates coincide, regardless of the applied backpressure. To comply with this requirement, a minimum outlet closing voltage of 80 V is essential for both considered inlet voltage levels. It also becomes evident, that the outlet closing voltage needs to be smaller than the inlet closing voltage. Concluding from these measurements, the outlet closing voltage is recommended to be approximately 2/3 of the inlet value as a rule of thumb. 5.5.3 Cut-off pressure The cut-off pressure is linked to the closing voltage of the outlet valve and constitutes a crucial parameter of the lumped parameter model. It denotes the pressure in the pump chamber where leakage of the closed outlet valve sets in. According to theory and simulations a rapid decrease of the flow rate should set in for outlet pressures above the cutoff pressure (see chapters 3.2.3 and 3.2.8). For the experimental evaluation, the inlet valve was kept open and different closing voltages were applied to the outlet valve. Then, an external pressure provided by the pressure controller was connected to the inlet port and 98 5 Experimental characterization was continuously increased. A flow sensor behind the outlet valve detected the onset of valve leakage. The corresponding pressure value was determined from the graph depicted in Figure 5-16 (a). The threshold flow rate was set to 0.1 µl/min. (a) (b) Figure 5-16: The experimental evaluation of the cut-off pressure at different closing voltages (a) is compared to the analytical prediction (b). It confirms that high cut-off pressures up to 100 kPa are achieved (P26). Figure 5-16 (b) shows the cut-off pressure of the outlet valve as a function of the closing voltage applied to the outlet piezo-actuator. The measured values are in good agreement with the analytical solution derived in equation (3.10) especially within the voltage range between 60 V and 100 V which is typically used for this micropump. 5.6 Variation of the pump chamber geometry The geometry modification introduced in design II is expected to be favorable for an increase of the compression ratio. On the other hand the smaller pump chamber reduces the prevalence of the valve resistance in comparison to the pump chamber resistance which is an undesired side effect. This section summarizes some key characteristics of the micropump design II. First of all, the flow resolution is increased due to a reduction of the stroke volume. Still, the stroke volume is tunable by means of the upstroke voltage and can be adjusted within a range of 10 - 50 nl (Figure 5-17). 99 5 Experimental characterization Figure 5-17: Stroke volume versus upstroke voltage for design II (P302). A setback of the backpressure characteristic is determined for design II (Figure 5-18) in comparison to the excellent backpressure stability of design I. Even though the micropump is still able to sustain large static pressure heads well beyond 30 kPa, the pressure value p90% indicating a 10 % flow rate decline has dropped below 10 kPa. This effect is attributed to the increased fluidic resistance within the pump chamber that hampers the fluid displacement during the transfer phase. Figure 5-18: Backpressure characteristic of the micropump for the reduced pump chamber size of design II (P312). A major benefit of design II is the improved self-priming capability of the micropump. A vital prerequisite for self-priming is the capability of the micropump to propel gases. Generally, in the gas pumping mode both designs are able to transport gases (see chapter 5.7). For both designs self-priming experiments were conducted and the flow rate achieved after this priming process was recorded. It was then compared to the likewise curves obtained for manually pre-primed micropumps. For design I large deviations were observed between consecutive measurements. These discrepancies were particularly pronounced between pre-primed and self-primed experiments (Figure 5-19 (a)). The simulations presented in chapter 3.3.2 point out that trapped air bubbles can explain such deviations of the flow rate. Since the size and location of the trapped air bubbles are randomly distributed, a reproducible flow cannot be expected if 100 5 Experimental characterization gas bubbles are inside the pump chamber. For the improved design II, comparable flow rates were obtained for both self-priming and pre-priming (Figure 5-19 (b)) which is taken as evidence for the absence of air bubbles. (a) (b) Figure 5-19: Flow rate vs. frequency for the pump chamber design I (a) in comparison to design II (b): Large discrepancies were observed between preprimed and self-primed measurements which have been eliminated by the modified design (P26 / P302). Calculating the mean stroke volume as well as the standard deviation for both designs, significantly smaller deviations were confirmed for repeated measurements (three measurements with self-priming and with pre-priming for each design) in case of design II (Figure 5-20). Thus, the decreased stroke volume of design II together with its reliable selfpriming capability improves the dosing resolution of this micropump. Figure 5-20: Dosing volume per pump stroke (double logarithmic scale) for design I and design II: The smaller geometry in design II yields a reduced stroke volume and the reproducibility has significantly improved (P26 / P302). 101 5 Experimental characterization 5.7 Gas pumping mode The gas pumping mode has been introduced in chapter 3.3.1 as actuation scheme for the propulsion of compressible fluids. Concluding from the analysis of the phase setting effect in section 5.4 the duration of 15 ms is recommended for the intermediate phase where both valves are open. The measurements confirm that both designs of the micropump are able to transport gases in the gas pumping mode (Figure 5-21). For gases, the backpressure stability is rather poor due to the rapid backflow of gas during the intermediate phase. The recorded frequency curves are less regular in comparison to the measurements carried out with water. (a) (b) Figure 5-21: Gas pumping capability of design I (P32) (a) and design II (P312) (b) in the gas pumping mode. 5.8 Gas-liquid interfaces In chapter 3.3 the Young-Laplace pressure drop across a gas-liquid interface was identified as a potential failure mechanism. Now, an experimental validation of the capillary pressure drop and its consequence for the deflection of the membrane is presented. 5.8.1 Capillary pressure drop First, as a reference measurement, the pump chamber was fully pre-primed which was achieved by first priming the pump chamber with ethanol and subsequently replacing the ethanol by water. For the measurement itself, the inlet valve was kept open by an upstroke voltage of -80 V, whereas the outlet membrane was in free float, i.e. without actuation. Then, the external pressure at the inlet port was stepwise incremented by means of the pressure controller. The recorded curve shows a non-linear relation between flow rate and applied pressure which arises from an increase of the gap height caused by the pressure induced bending of the membrane (Figure 5-22). Note that the applied inlet pressure and the chamber pressure are assumed to be equal due to the open inlet valve. 102 5 Experimental characterization Figure 5-22: Measurement of the pressure-induced flow rate through the micropump for a fully primed pump chamber (dashed curve) and blockage of the flow by a gas-liquid interface (solid curve) (P312). In a subsequent measurement, a gas-liquid interface was transported into the micropump until the micropump failed. Once again, the inlet valve was kept open and the outlet voltage was in free float mode. A sharp threshold was detected for increased pressures: the capillary effect prevented fluid displacement below 20 kPa whereas beyond this threshold a sudden rise of the flow rate set in. 5.8.2 Membrane hysteresis The active switching of the valves exhibits a hysteresis behavior. For this measurement the entire pump chamber was filled with water. The inlet valve was kept open during the whole experiment and a constant external pressure difference of 10 kPa was applied between the inlet and the outlet. The voltage of the outlet valve started at a closing voltage of +140 V and was then continuously varied ending at an opening voltage of -110 V. Thereafter, the voltage swept back to its initial value of +140 V. The diagram in Figure 5-23 shows the impact of this voltage sweep on the measured flow rate. Aside from the hysteresis of the piezo-actuator itself, the capillary force is supposed to contribute to this hysteresis effect. The opening of the valve necessitates the compensation of the strong capillary forces which arise in the minute gap at the beginning of this process. In consequence, a negative upstroke voltage is required until a pressure driven fluid flow sets in. When the voltage shifts back again, a hysteresis effect becomes apparent. Now, a flow rate of 80 µl/min is still maintained at zero voltage. This is explainable by the design inherent initial gap of 1 µm remaining underneath the flat membrane. The flow rate dies off beyond a closing voltage of +40 V. 103 5 Experimental characterization Figure 5-23: The pressure driven flow rate (p = 10 kPa) visualizes the hysteresis of the membrane deflection (P312). 5.9 Single-membrane micropump The single-membrane micropump is considered as a design variation of the two-stage micropump. Despite the larger deflection expected from the FEM-simulations the fluidic performance of the single-membrane pump appears comparable to the two-membrane designs. Therefore, the simpler fabrication process is considered as the main advantage of this concept. From this point of view, the design featuring simply through-holes in the bottom chip is most appealing. For this configuration, the flow rates obtained for different actuation frequencies are depicted in Figure 5-24 (a). Similar to former results, the characteristic is mainly linear for low frequencies up to 4 Hz and the flow rates are in a comparable range. (a) (b) Figure 5-24: Frequency response (a) and backpressure characteristic (b) of the low-cost micropump design. 104 5 Experimental characterization Even in this simple design, the micropump has proven to withstand large backpressures up to a maximum of 45 kPa (Figure 5-24 (b)). Since the valve seats are missing and consequently the concept of the constant cut-off pressure does not apply any more, the continuous decline of the flow rate towards increasing backpressures is inevitable. Nonetheless, this simplified design is considered as an appropriate alternative for applications that are less demanding in terms of precision or backpressure independence. 105 5 Experimental characterization 106 Chapter 6 Discussion 6 Discussion In this chapter the effect of different parameters will be discussed and typical characteristics of the two-stage micropump will be highlighted. A new figure of merit to quantify the backpressure stability within a given working range will be presented and introduced as a distinctive feature in comparison to conventional reciprocating micropumps. An overview of the experimental results determined for the inspected micropumps will provide a profound background for the conclusions made in this chapter. Moreover, the important issue of reliability will be discussed towards the end of this chapter. 6.1 Backpressure stability The delivery of a constant flow rate at varying backpressures is the outstanding attribute of the presented novel micropump design. This section summarizes the experimental results, presents a conclusive assessment of the backpressure performance and comments on the benefits and restrictions of the two-stage concept. 6.1.1 Differential fluidic output resistance Typically, the characteristics of a micropump exhibit a linear decline of the flow rate with increasing backpressures. Hence, the zero-pressure flow rate together with the maximum backpressure is an appropriate measure to assess the performance of the pump. The novel design proposed in this thesis shows a completely different backpressure characteristic. Within the working range, the flow rate remains nearly constant i.e. independent of the static pressure head applied to the outlet. Beyond the working range the flow rate decreases rapidly since the actuation force is not strong enough to close the outlet valve any more. For the assessment of this micropump, a new figure of merit is introduced referred to as differential fluidic output resistance. It is adopted from the electric circuit equivalent of the micropump which would be a current source. 107 6 Discussion (b) (a) Figure 6-1: The novel flow rate versus backpressure characteristic of the twostage micropump (P26) (a) exhibits a high differential fluidic output resistance within the working range in analogy to an electrical current source (b). Figure 6-1 depicts a typical measurement of the flow rate for increasing backpressures at a pumping frequency of 0.25 Hz. The obtained result shows a characteristic similar to the current versus voltage diagram of a practically implemented electric current source. For the current source the differential output resistance denotes the variation of the delivered current with respect to the voltage applied to the output. In analogy the differential fluidic output resistance can be mathematically expressed as out, fluidic ∆ ∆ (6.1) with the actual flow rate Q, the flow rate at zero backpressure Qmax and the external pressures pin and pout at the inlet and outlet, respectively. Thereof, the virtual maximum backpressure can be calculated as out, fluidic · (6.2) which is the point where the linear extrapolation of the working range would intersect the pressure axis. Alternatively, the same information can be expressed as 90%-barrier where the flow rate has decreased to 90% of its initial value Qmax. This pressure value is easily calculated by % · out, fluidic · (6.3) and has already been used in the previous chapter to compare the backpressure stability of the micropumps. As an example, a flow rate decrease of less than 10% up to a backpressure of 30 kPa, i.e. p90% = 30 kPa, would require a virtual maximum backpressure of 300 kPa. For a flow rate of Qmax = 5 µl/min the absolute value of the differential fluidic output resistance would have to be greater than rout,fluidic = 60 kPa/µl/min. 108 6 Discussion 6.1.2 Comparison of different micropumps The key figures of three selected micropumps concerning the flow rate and backpressure stability are summarized in Table 6-1. These selected micropumps have been characterized comprehensively and exhibit typical characteristics. Overall, more than 40 micropumps were investigated in the framework of this thesis and an excellent performance could be assigned to approximately 40% of the tested micropumps. The quoted results are based on the standard actuation scheme introduced in chapter 3.1.4 with the corresponding time settings given in chapter 5.3.1. Table 6-1: Micropump characteristics (Design I) [ /min] /(µl/min)] µl P19 P26 P32 Mean Standard deviation 0,25 Hz 3.41 2.8 1.92 2.71 0.75 1 Hz 9.76 10.89 6.98 9.21 2.01 0,25 Hz -101.3 -96.3 -124.0 -107.2 14.7 1 Hz -26.2 -13.4 -28.5 -22.7 8.1 0,25 Hz 34.6 27.0 29.3 30.3 3.9 1 Hz 25.5 14.6 24.7 21.6 6.1 p90% [kPa] [ rout,fluidic kPa Qmax Micropump No. A remarkable backpressure stability has been proven for this novel two-stage micropump concept which exceeds the capability of most micropump designs published yet. In particular, no other two-stage approach has been reported that is capable of providing a nearly constant flow rate up to a backpressure of 30 kPa. Additionally, the sustained maximum backpressure of 65 kPa is also an outstanding attribute for a two-stage micropump. The most illustrative figure to assess the backpressure stability is the p90% value which is spread between 25 – 35 kPa for a frequency of 0.25 Hz and between 15 – 25 kPa for a frequency of 1 Hz. As explained before, the reason for the superior performance at lower frequencies is found in the truncated flow pulses towards higher frequencies. This effect is also the reason for the non-linearity of the flow rate towards higher frequencies which corresponds to a decreased stroke volume. Based on the introduced figures of merit, the performance of this two-stage micropump is contrasted with other piezoelectric micropumps in Figure 6-2. The micropump by Maillefer et al. [26] is the only one providing a nearly constant flow rate up to a backpressure 109 6 Discussion of 20 kPa which corresponds with an exceptionally high differential fluidic output resistance. The devices presented by Shoji et al. [130], Linnemann et al. [37], Kämper et al. [131] and Doll et al. [114] are sustaining large backpressures but all feature a linear backpressure dependency. In this case the p90% value is found at 10 % of the maximum backpressure. Only the micropump by Stehr et al. [21] shows a non-linear backpressure curve with the p90% value at approximately 20 % of the maximum backpressure but the extremely high flow rate at zero backpressure causes a low differential fluidic output resistance. Figure 6-2: Backpressure stability and differential fluidic output resistance for various published piezoelectric micropumps. Among the more recent publications on piezoelectric micropumps no significant improvement concerning the backpressure stability has been reported. Either the concepts are restricted to extremely low backpressures [43] or they still obey to a linear flow rate decline. Considering other actuation principles, a strong micropump based on paraffin actuators has been reported without giving details on the backpressure characteristic [132]. A pneumatic micropump with a constant flow rate up to a backpressure of 25 kPa has been announced recently by Inman et al. [42] but here an external actuation pressure of 40 kPa was applied which makes this approach unsuitable for portable systems. 6.2 Pump chamber geometry The interior of the micropump design II exhibits a 4-fold smaller pump chamber which is limited to the region between the inlet and the outlet. It has been introduced to increase the 110 6 Discussion compression ratio and minimize the occurrence of gas bubbles in the pump chamber which are frequently trapped at the corners of design I. This goal has been achieved and an improved reproducibility of the calibrated flow rate has been proven. In particular, the increased compression ratio and the more appropriate shape of the pump chamber enable micropumps of design II to transport small gas bubbles from the inlet to the outlet. However, the backpressure performance of the micropump has declined due to the pump chamber modification. The following Table 6-2 summarizes the key figures of three inspected micropumps of design II to be compared with the above noted results for design I (see Table 6-1). Table 6-2: Micropump characteristics (Design II) [ /min] /(µl/min)] µl P302 P305 P312 Mean Standard deviation 0,25 Hz 0.73 0.89 1.27 0.96 0.28 1 Hz 2.89 3.47 4.01 3.46 0.56 0,25 Hz -22.7 -69.0 -51.5 -47.7 23.4 1 Hz -7.5 -6.7 -11.6 -9.3 2.1 0,25 Hz 6.5 7.2 6.8 6.8 0.4 1 Hz 6.0 1.7 5.0 4.2 2.3 p90% [kPa] [ rout,fluidic kPa Qmax Micropump No. The lower backpressure stability is attributed to the increased fluidic resistance within the smaller pump chamber. Thus, the valve resistance is not solely governing the fluid dynamics in design II which leads to a significant and backpressure dependent backflow during the transfer phase. The increased chamber resistance is also held responsible for the smaller flow rates obtained for design II. In the simulation, the dead volume does not play a role. The simulation is based on the assumption of an unrestricted fluid transfer within the pump chamber and hence cannot explain the flow rate deviation between design I and II. All in all, a performance trade-off appears and the decision between design I and II needs to be made in the context of the target application. For both designs self-priming of the micropump is enabled by means of the gas pumping mode. For applications where gas bubble tolerance is a decisive issue the proposed design II is favored over design I. On the other hand, for the active microport system the backpressure independence of the flow rate is of high priority and therefore design I is the preferred choice in this study. This, in turn, requires a thorough initial priming process of the micropump to entirely fill the pump chamber. 111 6 Discussion 6.3 Controllability of the micropump Due to the piezoelectric actuation the electrical control of the micropump characteristics has proven feasible. Within the relevant frequency range 0.05 – 4 Hz a straightforward frequency setting of the flow rate is possible based on a linear correlation between flow rate and actuation frequency. The phase setting also affects the flow rate, however it is not recommended as a control parameter due to the high non-linearity of this effect. In addition, the flow rate is shiftable by means of the applied voltages. The upstroke voltage applied to both actuators corresponds nearly linear to the stroke volume and is identified as a suitable control variable. This way, the micropump can be calibrated to deliver a standardized flow rate for a certain frequency or to regulate the flow rate in a feedback control loop, e.g. in conjunction with a flow sensor. Moreover, this option can be used to adapt the resolution of the micropump to the intended application. As an example, a larger stroke volume helps to minimize the energy consumption in case of high flow rates. It also provides the possibility to temporarily increase the compression ratio if the micropump is blocked by a gas bubble. The flow rate is also related to the closing voltage of the outlet actuator. In contrast to the upstroke voltage, the closing voltage is not considered as an appropriate control variable since it is coupled to the backpressure stability of the micropump. Therefore, for the closing voltage of the outlet valve a standard value of approximately 2/3 of the inlet voltage is recommended for a proper pumping operation. 6.4 Transport of gases and capillary effect The intention behind the development of this micropump was the robust transport of incompressible, water-based liquids. In case of gases, the comparably low compression ratio of the presented design together with the high compressibility of the medium is a severe problem. Nevertheless, the slightly modified actuation sequence referred to as gas pumping mode enables the transport of gases and is regularly applied for self-priming of this micropump. The gas pumping mode is also applicable to liquids and leads to an accelerated transport of the fluid. In accordance to the phase setting study in chapter 5.4 the backpressure stability of the flow rate for liquids is still preserved in the gas pumping mode if the duration of the intermediate phase does not exceed 15 ms. The modified chamber design II, which reduces the pump chamber volume to 1 µl and hence increases the compression ratio, constitutes a significant progress towards a reliable and robust gas bubble tolerance. A maximum compression ratio of approximately 0.14 is achieved in case of an upstroke voltage of -110 V. Therewith, small gas bubbles with a volume well below 1 µl have successfully managed to pass the micropump. Nevertheless, medium sized air bubbles entrapped at the valve seats are still a potential reason for failure due to the capillary forces provoked by the small gap between valve lips and membrane. The capillary pressure drop across a curved gas-liquid interface pinned to the valve seat has been estimated analytically and verified experimentally. As derived in chapter 3.3.3, a Young-Laplace pressure drop between 20 – 125 kPa is reasonable depending on the actual contact angle and gap height. Both parameters are subject to large uncertainties which does 112 6 Discussion not allow for a more precise prediction. The experimentally determined pressure drop of 20 kPa is found at the lower end of this expected range. The presented calculations have outlined that this pressure drop may already cause a complete blockage of the micropump. A further increase of the actuator stroke volume, i.e. the application of a higher upstroke voltage, would enhance the gas bubble tolerance of this two-stage micropump. This strategy would imply either a substitution of the piezoceramic or a reduction of the membrane thickness. Both measures would change the mechanic properties of the piezo-membraneactuator and hence affect the described characteristics of the micropump. 6.5 Reliability issues After fabrication several circumstances may cause a malfunction of the micropump. First, no membrane deflection at all is observed if the conductive carbon-black particles embedded in the glue layer do not provide an electrical contact to the electrode of the piezoceramic discs. Second, the micropump inhibits the priming of the pump chamber if the membrane is bonded to the valve lips. Third, an insufficient performance of the micropump regarding the flow rate or the backpressure stability may be caused by leakage of one or both valves. All these faults are detectable in an initial inspection and deficient samples can be rejected due to defects. Figure 6-3: Accelerated lifetime test of a bulk PZT actuator using bipolar actuation cycles (+230V / -130V) at 45°C [133]. Micropumps that passed the inspection successfully show an excellent stability and reproducibility of their performance on the long term. Early prototypes have been in operation for more than two years by now and still maintain their initial characteristics. A systematic long-term test to investigate the durability of this technology has been carried out in a diploma thesis by M. Wischke [134] (Figure 6-3). The most likely reasons for long term failure would be fatigue of the glue layer or wear at the valve lips. As the positive result, the membrane deflection did not show any degradation over an operation period of more than 200 million duty cycles [133]. Moreover, a visible inspection did not give any indication of wear at the silicon valve lips. 113 6 Discussion 114 Chapter 7 Feasibility study of a paraffin-actuated two-stage micropump 7 Feasibility study of a paraffin-actuated two-stage micropump The novel two-stage design presented in this thesis is based on the experience of our research group with piezoelectrically actuated micropumps. For the intended target application in an implantable drug delivery system the comparably low power consumption of piezoelectric actuators is considered to be a decisive advantage over other mechanisms such as electromagnetic or thermal actuation. Nevertheless, as a side aspect of this thesis the feasibility of a two-stage micropump actuated by a thermal phase-change actuator based on paraffin has been explored. The comparably large expansion of paraffin upon the solidliquid phase transition within a small temperature interval and the incompressibility of the material in either state enable the realization of a powerful and robust actuation mechanism. The incompressibility is considered as a particular advantage in case of high mechanical loads such as large external pressures. A membrane deflected by means of a paraffin actuator would not exhibit a large fluidic capacitance as in case of piezoelectric actuation. In principle, these actuator characteristics appear promising for the development of a highpressure micropump. Two technology-related concerns are the development of an appropriate design and the integration of paraffin into the MEMS fabrication process. For a proof of concept the single-membrane design of the two-stage micropump was chosen and two different options for the integration of the paraffin actuator were investigated. First a classical approach with a resistive heater was studied and an appropriate fabrication process was demonstrated. Subsequently, a novel direct heating strategy based on the Joule heating of a paraffin compound with embedded conductive particles was developed and first actuator samples were characterized. Based on this method an increased efficiency has been proven which reduces the power consumption and enables shorter actuation cycles. 7.1 Paraffin actuators and micropumps Despite the promising properties of paraffin actuators they are still considered as an uncommon approach in MEMS technology. While thermopneumatic approaches based on the evaporation of a working liquid have been widely explored [135], only a few thermal 115 7 Feasibility study of a paraffin-actuated two-stage micropump actuators using the solid-liquid phase transition of paraffin have been reported. Most of these approaches realized fluidic valves switched by the deflection of a membrane in consequence of the paraffin expansion. A severe concern about paraffin is the partial incompatibility between its wax-like character and typical MEMS fabrication processes. Carlen et al. [136] demonstrated that thin solid state paraffin films can be patterned using surface micromachining techniques. This way, the authors fabricated a thin paraffin microactuator placed onto a resistive heater and sealed by a flexible parylene layer (Figure 7-1). Figure 7-1: Paraffin actuator fabricated by means of surface micromachining [136]. A high force paraffin actuator made from silicon has been proposed by Klintberg et al. [137]. This concept utilizes the high energy density of paraffin actuators to realize a large deflection of a bossed silicon membrane (Figure 7-2 (a)). The paraffin is contained in a ring-shaped cavity around the edge of the membrane. Two structured silicon wafers are bonded to form the cavity which is then filled with the liquid paraffin. Due to the large volumetric expansion and the incompressibility of the liquid paraffin, a large stroke is obtained even under load. This example shows that the high energy density of the actuator can be used to implement mechanical amplification structures that increase the deflection of the actuator. Based on their experience with paraffin actuators, the same research group presented a high-pressure micropump where the resistive heating structure is located in the middle of the paraffin reservoir [132, 138]. It is a conventional peristaltic micropump design featuring three sequential paraffin actuators (Figure 7-2 (b)). A low flow rate of 74 nl/min is delivered by this device at an actuation frequency of 1/32 Hz. The valves do not exhibit leakage up to an external pressure head of 1 MPa. The proposed design relies on a rather complex fabrication process including multiple structured layers and a critical filling process with liquid paraffin but bears the potential of an exceptionally strong, backpressure stable micropump. This presumption has not yet been proven conclusively by the published results. (a) (b) Figure 7-2: Paraffin actuator with mechanical amplification structure [137] (a) and cross section of the pump design with two valves and one pump chamber [132] (b). Further concepts of paraffin actuators comprise a latchable microfluidic valve [139] as well as a paraffin-PDMS composite actuator [140]. For the valve, a combination of paraffin latching and pneumatic switching is employed to realize a static latch without continuous power consumption. The composite actuator is fabricated by dispersing paraffin droplets in a PDMS 116 7 Feasibility study of a paraffin-actuated two-stage micropump matrix. In this concept the nearly incompressible but elastic PDMS serves as encapsulation and provides the restoring force while the powerful volumetric dilatation of the paraffin droplets leads to an overall expansion of the composite even against heavy loads. 7.2 Paraffin waxes From the chemical point of view, paraffin is a long chain polymer consisting of a hydrocarbon backbone with a structural composition CnH2n+2. A typical chain length consists of 20 to 40 carbon atoms. Upon melting the material exhibits a volumetric expansion of 10 - 15 %. The melting temperature can be tailored between -100°C and 100°C by variation of the hydrocarbon chain length [137]. In general, waxes with longer polymer chains posses a higher melting temperature. Most commonly, the melting range is found between 35 – 80°C. Nearly the full expansion is attained over a narrow temperature interval of 2 – 5°C which is favorable in terms of power consumption. Nevertheless, the melting temperature range can be broadened by a mixing of alkanes with different molecular weights. Most technical waxes are obtained as distillates of a petroleum refining process and contain alkanes of different chain length. Depending on the length and orientation of these molecular chains the paraffin waxes are classified as micro-crystalline and macro-crystalline types [141]. The compressibility of paraffin in the liquid state is small and negligible for most MEMS applications [137, 142]. In consequence, the energy density obtained for paraffin actuators is extremely high and reaches values up to 107 J m-3 [136]. As a drawback, paraffin has a low thermal conductivity and features a large heat capacity in the range of 170 kJ/kg [137]. This enforces large heating energies and allows for slow heating/cooling cycles only. Paraffin is an inexpensive material since it is a leftover from the petroleum industry. A popular paraffin wax reported in the literature is purchased from Sigma-Aldrich with melting temperatures either between 44 - 46°C [132, 139] or at about 65°C [137]. In this work, a refined paraffin (107150 Paraffin, 42-44, Blockform, Merck KGaA, Darmstadt, Germany) with a melting temperature of 42 – 44°C as well as a pure n-Tetracosane (Alfa Aesar GmbH & Co KG, Karlsruhe, Germany) are investigated. The refined wax exhibits a volume expansion of approximately 7.4 % within a temperature interval of 2 K. In contrast, the n-alkane paraffin features a two-step expansion process. A solid-solid phase transition at 48°C is associated with a first fractional expansion followed by a second sharp expansion upon melting at 50.1°C. Overall, the volume expansion sums up to 15 % with a portion of 38 % attributed to the solid-solid phase transition. Compared to the ductile refined wax, the pure n-alkane is rather brittle which complicates the patterning of this paraffin type in its solid state. 7.3 Single-membrane paraffin micropump The paraffin micropump concept reverts to the experience on the two-stage silicon micropump and the established fabrication technology. The single-membrane design with the large membrane spanning both fluidic valves was chosen for the prototype of the paraffin micropump. Instead of mounting piezoelectric actuators to the membrane, a resistive heating structure was added to the top side of the silicon membrane. The heating meander was restricted to one half of the membrane in order to initiate an asymmetric melting process. 117 7 Feasibility study of a paraffin-actuated two-stage micropump Then, the cavity above the membrane was filled with paraffin and subsequently sealed by a stiff cap (Figure 7-3). Figure 7-3: Concept of the single-membrane paraffin micropump. As mentioned above, the integration of paraffin actuators into MEMS devices is non-trivial and requires an appropriate fabrication technology. The placement of the paraffin wax should be one of the last process steps for two main reasons. First, a surface contaminated with paraffin is not compatible with the purity standards of most cleanroom processes. Second, all process steps following the paraffin deposition need to be carried out at room temperature due to the low melting temperature of the wax. In the developed concept the fabrication of the silicon micropump is not affected by the back-end actuator integration at all. This way, the aforementioned problems are eluded and the uppermost flexibility is retained to equip the micropump either with a piezoelectric actuator or with a paraffin actuator. For the paraffin actuator, the gold layer evaporated onto the membrane was laser-structured to create a resistive heating meander (Figure 7-4). The heater was connected with two bond pads formed at the edge of the micropump and exhibited a resistance in the range of 300 – 500 Ω. Then, approximately 50 mg of solid paraffin were placed onto the membrane and the edge of the micropump was brought into direct contact with a soldering iron. The melting paraffin wax spread across the membrane and formed a planar paraffin layer. The micropump chip was then inserted into a fabricated silicone master mold to create the sealing cap. For the cap material, the cold cast epoxy resin Stycast® 2057 (National Starch and Chemical Company, Westerlo, Belgium) was used which provided a sufficient stiffness. It was easily replicated and released from a silicone master. The master mold featured small undercuts and spacers to prevent the Stycast from enclosing the micropump and to provide access to the bond pads for the electrical contacting. The assembly was cured for 20 hours at room temperature and was subsequently released from the silicone master. As final process step, the pads of the resistive heater were electrically connected by means of soldering. Alternatively, a conductive glue could also be used to provide an electrical connection. 118 7 Feasibility study of a paraffin-actuated two-stage micropump (a) (b) (c) (d) Figure 7-4: Fabrication process of the paraffin micropump including laser structuring of the gold layer (a), deposition of molten paraffin wax (b), and backside sealing by means of an epoxy resin (c),(d). Experimental measurements for cycles times of 90 s and 180 s were conducted (Figure 7-5). In each case, a 720 mW square wave power signal with a duty cycle of 50% was applied. The flow rate was recorded by means of the flow sensor. The signal shows a distinct volume displacement upon melting and solidification of the paraffin. An average flow rate of approximately 80 nl/min was obtained for both frequencies. Obviously, the shorter cooling interval in case of the cycle time of 90 s is insufficient to complete the solidification of the paraffin actuator. In consequence, a residual expansion remains at the end of the cycle which decreases the stroke volume of the actuator. (a) (b) Figure 7-5: Flow signal of the paraffin micropump for a rectangular power signal with a cycle time of 90 s (a) and 180 s (b). 7.4 Direct heating concept The comparably high energy consumption of paraffin actuators is still a limitation for the use of those actuators. This issue has been addressed by the development of a novel heating strategy to significantly increase the efficiency of the actuator. A resistive heater placed on 119 7 Feasibility study of a paraffin-actuated two-stage micropump the side wall of the cavity suffers from high thermal losses. In contrast, the heat generation in the interior of the paraffin actuator promises a tremendous increase of the efficiency factor. Bodén et al. [132] tackled this problem by burying the resistive heating structure in the middle of the paraffin cavity. The detrimental aspect of this solution is the complex fabrication process including the critical filling of the cavity with liquid paraffin. The concept proposed in this thesis is the generation of Joule heat inside the paraffin material based on the dispersion of conductive carbon black particles. The development of this concept has been carried out by P. Katus within his diploma thesis [141]. Figure 7-6 illustrates the assembly of the actuator including the galvanic sealing with a copper cap. The copper layer is designed to be an order of magnitude stiffer than the silicon membrane. A highly doped n+-silicon wafer is used to provide the electrical contact to the bottom side of the paraffin cavity. Since the paraffincarbon black-composite constitutes the predominant electrical resistance, an applied voltage provokes the generation of Joule heat inside the paraffin material. The subsequent expansion causes a deflection of the silicon membrane. Figure 7-6: Concept of the paraffin – carbon black – composite actuator. 7.4.1 Conductive paraffin For the fabrication of the “conductive” paraffin, the commercial carbon black Printex XE2 (Evonik Degussa GmbH, Essen, Germany) is used as a filler to render the paraffin conductive. The mean particle size of Printex XE2 is 30 nm. Since paraffin is an electrical insulator, the density of the dispersed graphite particles has to exceed the percolation threshold for the formation of continuous current paths. Resulting from preliminary experimental studies, a volume fraction of 2 % - 4 % carbon black is stirred into the paraffin. Figure 7-7 (a) shows the correlation between the electric resistance of the composite and the volume fraction of carbon black. Here, molten samples are examined by means of a spreading resistance measurement. Note that the composite material is paste-like and dimensionally stable even in the molten state which is in contrast to the low viscosity liquid paraffin. The diagram indicates a difference between the initial resistivity and the steady state value. When exposed to an electric field a dynamic reorientation of the carbon black agglomerates is presumed which leads to an exponential decrease of the resistivity within the first 30 – 45 s of the experiment. This effect has already been studied by others [143, 144] and is assigned to induced surface charges. In the solid state the dynamic reorientation is impeded and consequently a stable resistivity value is observed. 120 7 Feasibility study of a paraffin-actuated two-stage micropump (a) (b) Figure 7-7: Resisitivity of the molten paraffin-carbon black-composite at an applied field strength of 550 – 930 V/m (a) and non-linear voltage-currentdiagram for composites filled with different volume fractions (b). The relationship between the applied voltage and the induced current is highly non-linear (Figure 7-7 (b)). A minimum voltage of 0.5 V was required for the initiation of an electric current. Negligible deviations were observed between the depicted diagram recorded in the molten state and the corresponding measurement in the solid state. The electric resistance of filled composites is also known to be frequency dependent [145]. Displacement currents induced in isolated carbon black conglomerates lead to an increased conductivity when subjected to an alternating electric field. Thus, AC operation of the actuator is a robust alternative especially in case of low volume fractions of carbon black since it does not rely on the formation of a conductive particle network ranging from end to end of the actuator. From the fabrication point of view, the challenge in producing conductive paraffin actuators is the avoidance of gas inclusions, as they would drastically reduce the actuator performance. Preliminary degassing of carbon black at a temperature of 350°C and subsequent stirring of the paraffin - carbon black - composite under vacuum in an exsiccator has been confirmed as appropriate measure. For the stirring process, the paraffin is heated to a temperature well above its melting point to obtain a suitable viscosity. Upon completion, the paraffin is solidified at room temperature and filled intro a syringe. In an evacuated container, the syringe is then heated up again to form a rod of the composite material. 7.4.2 Process chain The detailed fabrication process is shown in Figure 7-8. It is based on a highly doped n+silicon wafer and consists of standard MEMS processes like lithography, metal evaporation and KOH etching to create the actuator diaphragm, an insulation layer and the galvanic starting layer. Two lithography masks are required for this process. The investigated membranes feature a size of 6 x 6 mm2 and 8 x 8 mm2. 121 7 Feasibility study of a paraffin-actuated two-stage micropump (a) (e) (b) (f) (c) (d) (g) Figure 7-8: Fabrication process of the paraffin – carbon black – composite actuator [141]. These standard processes are followed by the deposition of the paraffin - carbon black – composite. Different strategies were explored in the diploma thesis by P. Katus. Pressing of thin paraffin discs onto the membrane and subsequent patterning by means of a razor blade has been revealed as a suitable manual procedure. Here, the chip is mechanically supported from the bottom side and a spherical part of the composite material is flattened on the upper side of the diaphragm (Figure 7-9). Paraffin discs with a thickness in the range of 100 µm have been realized utilizing appropriate spacers, e.g. aluminum foils. A sufficient ductility of the composite material is essential in order to prevent cracking of the actuator disc which would cause gas inclusions. For this reason the refined waxes are more appropriate for this deposition strategy whereas the rather brittle n-alkanes need to be processed close to their melting temperature. Figure 7-9: Pressing of a paraffin disc onto the actuator membrane. The second approved strategy is printing of the molten, paste-like composite by means of a molding tool. For this approach, a patterned adhesive foil with an orifice in the area of the diaphragm was used as molding tool. A subsequent lift-off of the foil left the structured 122 7 Feasibility study of a paraffin-actuated two-stage micropump paraffin disc with a thickness of 50 – 70 µm laminated onto the membrane. Even though the homogeneity of the printed actuator discs was not as good as the result obtained by the pressing process, the printing method was the preferred choice for the deposition of the n-alkane paraffin due to its brittle nature in the solid state. The gas-free encapsulation of the actuator disc was achieved by a galvanic copper deposition process. The carbon black particles acted as nucleation sites for the galvanic process. In order to ensure a homogeneous growth of the copper layer, a preparative clean of the wax surface with a diluted isopropanol solution is recommended. Additionally, the surrounding galvanic starting layer has to be thoroughly cleaned with acetone to enable optimum adhesion of the copper cap. The chip was then immersed into a copper electrolyte (Dr. Roperts GmbH, Munich, Germany) and contacted via the n+-silicon chip. As the conductive paraffin electrically connects the silicon membrane and the galvanic starting layer the copper layer starts growing on both the starting layer and the nucleation spots of the paraffin disc. At a current density of 0.63 A/dm2 a deposition rate of 8 µm/h was observed. After 3 – 4 days a homogeneous and stiff copper cap was grown on the top side of the actuator chip (Figure 7-10). Figure 7-10: Top view of the paraffin actuator sealed with a copper cap. 7.4.3 Results The actuators were characterized by a setup consisting of a temperature sensor for the chip temperature, a laser triangulation sensor for measurement of the center deflection of the actuator diaphragm and a controlled power source. Figure 7-11 (a) shows a static measurement of an actuator with a 6 x 6 mm² diaphragm. It contained the refined paraffin with a thickness of 114 µm, corresponding to a paraffin volume of 5.1 µl. The curve shows a slight linear increase of the deflection with chip temperature up to the melting point of 39°C. Here, a sharp rise of the deflection by 33.1 µm is observed within a temperature interval of 3.5 K. This increase corresponds to a volume expansion of the paraffin of 7.7 %. Beyond the melting interval, only a slow further increase is obtained for continuously incremented temperatures. Since the nominal expansion interval of this refined wax is expected between 42 – 44 °C, the measured temperature is obviously approximately 2 K below the mean paraffin temperature which is attributed to thermal losses. Moreover, the broadened melting interval indicates a temperature gradient inside the paraffin which is also estimated to be in the range of 2 K. 123 7 Feasibility study of a paraffin-actuated two-stage micropump (a) (b) Figure 7-11: Volume expansion at the solid-liquid phase transition of a refined paraffin wax (a) and a two-step expansion process of a tetracosan wax (b). A similar measurement with a tetracosan actuator is shown in Figure 7-11 (b). The curve is nearly flat for temperatures below 46°C. A pronounced two-step expansion process becomes apparent which is typical for pure n-alkanes. The preceding solid-solid phase transition causes a deflection jump of about 5.6 µm and is followed by a second expansion surge of 16.1 µm at 49°C. The total deflection corresponds to a volume expansion of 12 %. In sum, approximately one quarter of the overall expansion is caused by the solid-solid phase transition, while the main deflection is attributed to the melting phase change. Dynamic measurements have been performed to evaluate the reaction time and efficiency of the actuator. First, the time response to a single heating pulse was studied. In this case, the actuator was heated from ambient temperature with a voltage pulse (power: 1.9 W; duration: 25 s). The corresponding temperature and deflection for an actuator containing refined paraffin are shown in Figure 7-12 (a). The full actuator stroke was achieved after a heating period of 15 s, but the major contribution due to the solid-liquid phase transition fell into a time interval of only 4 s. In contrast, the contraction of the actuator required a significantly longer time interval of about 13 s before returning to the initial state. (a) (b) Figure 7-12: Response signal of a refined paraffin actuator to a single heating stroke (a) and oscillatory deflection of a n-alkane actuator about its melting point (b). 124 7 Feasibility study of a paraffin-actuated two-stage micropump For applications relying on periodic deflection the temperature would be set to an oscillatory variation around the melting range of the paraffin wax in order to improve both energy efficiency and cycle time. Here, n-alkane actuators are beneficial due to their narrow melting interval which requires only small temperature oscillations. The n-alkane actuator characterized in Figure 7-12 (b) was driven by a 1.9 W square wave power signal with a duty cycle of 50% at a frequency of 0.2 Hz, i.e. an energy of 4.75 J was consumed per cycle. The temperature oscillation was restricted to the range of the solid-liquid phase transition of tetracosan. Note that the depicted temperature curves indicate the measured chip temperature rather than the wax temperature itself. When the paraffin-carbon black-composite is prepared at ambient pressure a certain amount of gas is adsorbed in the composite. This gas is observed to be rapidly released in case of a strong and short heating pulse with subsequent rapid cooling. After the pulse, the gas is apparently not adsorbed again leading to a residual steady-state deflection. (b) (a) Figure 7-13: Reversible expansion due to a moderate heating pulse (a) and residual deflection caused by a rapid, steep heating pulse followed by a moderate heating pulse for relaxation of the deflection (b). Figure 7-13 contrasts a reversible expansion of a refined paraffin actuator caused by a moderate heating pulse with a power of 0.8 W to a rapid and partly irreversible expansion due to a short heating pulse with a power of 1.2 W. During the short heating pulse the chip temperature remains below the melting point of the wax which provokes a fast solidification of the paraffin after switching off the heating power. The application of a subsequent extended heating pulse has proven as appropriate measure to re-adsorb the gas again and thus to reverse the residual deflection. All in all, this effect could be used to realize a bistable actuator which does not consume power in either hold state. The reproducibility of this effect has been demonstrated by P. Katus [141]. 7.5 Discussion In conclusion of this feasibility study, the general compatibility of the proposed silicon micropump and a thermal actuator based on the expansion of paraffin has been demonstrated. As a proof of principle, periodic displacement of fluid by means of heating 125 7 Feasibility study of a paraffin-actuated two-stage micropump cycles has been shown and a net flow was achieved with a single-membrane micropump. The advantages of a paraffin based concept are clearly seen in the low actuation voltage in the range of 3 – 10 V and the robust expansion process associated with the solid-liquid phase transition which enables large stroke forces. In addition, paraffin is a non-toxic and inexpensive material. However, similar to other thermal phase-change concepts, paraffin actuators suffer from a comparably high power consumption, limited cycle frequencies and complicated fabrication processes. These drawbacks have been addressed in the framework of this thesis yielding a novel approach to facilitate the fabrication and integration into MEMS devices. A new concept of a direct heating paraffin actuator based on a paraffincarbon black-composite actuator has been developed. Several advantages are assigned to this approach. The direct generation of the heat inside the actuator clearly increases the transducer efficiency. This not only reduces the power consumption but also enables shorter cycle times, i.e. higher actuation frequencies. Moreover, the fabrication process is fairly simple and robust. The complexity is comparable to the presented concept of the singlemembrane paraffin micropump and meets the requirements for MEMS integration as stated above, such as the deposition of the paraffin actuator at the back-end of the process chain. The galvanic sealing is an adequate solution to serve as hermetic encapsulation, mechanical support and electrical contact at the same time. The choice of the paraffin material depends on the specific application. From the fabrication point of view, refined waxes are favorable due to their higher ductility. On the other hand, pure n-alkane waxes exhibit the phase transition within a smaller temperature interval which is beneficial for increasing the cycle frequency. Moreover, the composition of the paraffin wax enables an application-specific adjustment of the melting range which is typically found between 35 – 80°C. A future integration of the direct heating concept with the proposed micropump would extend the micropump platform of our research group which is so far limited to piezoelectric actuation. An important aspect of this development track will be a revised design to control the heat flux. In particular, a design-inherent asymmetry to strengthen the forward propulsion of the fluid and a controlled heat dissipation to increase the cycle time are of outstanding relevance. 126 Chapter 8 Microfluidic devices in soft polymer technology 8 Microfluidic devices in soft polymer technology For biomedical applications the encapsulation of devices is a nontrivial task which has to meet ambitious specifications such as biocompatibility, mechanical and chemical stability, fluidic sealing and electrical isolation. Especially for microfluidic systems the packaging is considered as a key process since it has to provide the interface between the microchips and the outer world. In the framework of this thesis, a multilayer soft lithography process for polyurethane has been established to fabricate the housing of a drug delivery system with integrated microfluidic structures. For life science applications polydimethylsiloxane (PDMS) has been widely employed, particularly for the formation of multilayer stacks. As an alternative to PDMS, this thesis explores polyurethane (PU) rubber as a suitable material for the multilayer technique. Compared to PDMS it excels by an even better transparency and a slightly higher mechanical stability. Moreover, a large number of glues work with PU whereas adhesives for PDMS are rarely found. For the formation of multilayer stacks the bond strength is an extremely critical parameter due to the demand for reliable sealing. Several bonding techniques such as adhesion, annealing, mixing ratio variation and adhesive layers were compared considering the reversibility of the bond type and the bond strength. Additionally, appropriate surface modifications were investigated to render a polyurethane surface hydrophilic or hydrophobic. 8.1 MEMS fabricated by soft lithography Microfluidic systems are often involved in life science applications e.g. drug delivery systems or labs-on-a-chip. In the latter case, the application of the biocompatible elastomer polydimethylsiloxane (PDMS) as a structural material has been frequently reported [146 - 149]. The variety of fabrication processes referred to as soft lithography comprise 127 8 Microfluidic devices in soft polymer technology replica molding, hot embossing, microcontact printing as well as microtransfer molding [150]. The use of PDMS provides several advantages such as good chemical stability, biocompatibility and transparency [146]. Typically, a stack of several structured PDMS layers is assembled to form the embedded channels and fluidic elements [34, 151]. Alternatively, bonding of PDMS layers to glass substrates [148, 152] or silicon chips [153] are reported techniques. The elastomer layers are commonly replicated from a master by replica molding. A great diversity of master types has been presented, mainly aluminum masters, microstructured silicon wafers [154] or photostructured resist layers, e.g. SU8 [148, 155]. An anti-adhesion layer may be deposited on the master mold in order to facilitate the release of the cured elastomer. Fluorinated silane layers [148], PTFE-coatings or different metal layers (e.g. Au [154]) are some of the most frequently reported methods. A typical casting process is depicted in Figure 8-1. The ease of the process enables flexible prototyping at low costs. The process starts with the careful mixing of the two component prepolymer, typically followed by a degassing step. Then, the prepolymer is cast onto the master mold. If the master mold is a silicon wafer structured by bulk or surface micromachining, the prepolymer is frequently spin-coated onto the silicon master in order to obtain flat and thin layers. After overnight curing at room temperature or accelerated curing in a furnace (e.g. 1h at 100°C), the elastomer layer is peeled off the master mold. Figure 8-1: Illustration of the soft lithography technique. PDMS is clearly the material which draws the main attention in the field of soft micromachining. Nevertheless, the fabrication technique mentioned above is generally applicable also for structuring of polyurethane (PU) elastomer layers. Compare to PDMS, the advantages of PU are seen in its higher mechanical stability and its superior adhesion properties while the resistance against solvents stays slightly behind the durability of PDMS [156]. Polyurethane exhibits a contact angle in the range of 90° – 100° and thus is less water repellent than PDMS. An additional benefit is the large variety of PU formulations including thermoplastic types or shape memory polyurethanes [157]. Published applications of PUbased microsystems are a flow sensor [156] as well as micropumps [158, 159]. 128 8 Microfluidic devices in soft polymer technology 8.2 Elastomer materials 8.2.1 Polyurethane Polyurethane is a common polymer which is widely used for technical applications and products. It is available in a great diversity of formulations which allows an optimization of the PU properties for the target application [160]. For example, a thermoplastic formulation can be processed either by injection molding to produce window frames, skis, car instrument panels or by extrusion molding to obtain thin films. Foamed PU is applied for items such as mattresses, car seats or insulation materials. For soft micromachining and rapid prototyping cold-cast polyurethane rubbers are particularly suited. These elastomer formulations are available as two-component systems similar to PDMS. Polyurethane is created by an addition polymerization of polyols and polyisocyanates. The basic reaction between an alcohol and a diisocyanate is shown in Figure 8-2. The product is characterized by the urethane group. Urethane group Figure 8-2: Basic chemical polymerization reaction for polyurethane [161]. Numerous variations of the involved educts result in the great diversity of available polyurethanes. Common formulations of polyurethane are based on polyethers or polyesters. Polyether-based products have a high resistance to both dynamic and static mechanical loads, a good low-temperature behavior and an excellent abrasion and hydrolysis resistance [162]. Polyester-based compounds particularly excel by their resistance to light and thermal aging [160]. The standard polyurethane elastomer employed in this thesis is the polyester-based VT 3402 KK-NV (Lackwerke Peters GmbH [163]). This product is originally designed for the electronics industry to serve as protective cover or insulation layer. It provides an excellent transparency and a slightly higher stiffness (Shore A: 70) compared to the most frequently applied PDMS Sylgard® 184 (Shore A: 40). Equal amounts of each component are easily mixed and degassed enabling a simple handling procedure. In the prepolymer state, the rather low viscosity (1100 ± 300 mPas) supports the casting process and the shrinking of the material is negligible. The quoted data are taken from the data sheet of the manufacturer. The polyether-based PU Elastocoat® C 6909/1 (Elastogran GmbH [162]) has been evaluated as an alternative PU rubber. It has a non-transparent appearance and exhibits a significantly increased viscosity in the prepolymer state which requires a mixing at elevated temperatures (~45°C). 129 8 Microfluidic devices in soft polymer technology 8.2.2 PDMS Polydimethylsiloxane (PDMS) is obtained by addition polymerization which forms a covalent bond between vinyl-groups and silicon hydride groups yielding a –OSi(CH3)2– backbone [146]. By far, the most popular PDMS formulations used in the field of MEMS technology are Sylgard® 184 or Sylgard® 186 (Dow Corning [164]). The material has been extensively characterized by others [13, 34, 146, 151]. In this thesis, Sylgard® 184 is utilized as reference material to benchmark the polyurethane multilayer bond strength. 8.3 Replica molding process 8.3.1 Fabrication of master molds The original master mold subsequently called primary master is fabricated from aluminum by means of conventional micromachining i.e. micromilling and drilling. For the target application, the fluidic channels connecting the micropump with the reservoir and the catheter are embedded in the elastomer layers. The precision of this master mold is uncritical since the smallest microfluidic features exhibit a size of 500 µm to 1 mm. Thus, the conventional approach is considered advantageous over silicon micromachining. 8.3.2 Patterning of elastomer layers The replica molding process utilized in this work is based on the comparably weak adhesion between silicones and polyurethane. That way, silicones can be used as masters for the fabrication of a polyurethane layer and vice versa [150]. In a first casting step, a negative copy of the aluminum master is produced to serve as a secondary master for subsequent replication steps. For this purpose, highly elastic formulations of silicone (Elastosil M 4642, Wacker Chemie AG, Munich, Germany) or polyurethane (VU 4452/61 HE, Lackwerke Peters GmbH, Kempen, Germany) are chosen as material for the secondary master mold. The cure and release of this secondary master from the aluminum master is followed up by the second casting step. Now, the transparent PU layers (VT 3402 KK-NV) are replicated from the silicon secondary master while the PDMS layers (Sylgard® 184) are obtained as negative copies from the polyurethane secondary master. In each case the liquid prepolymer is degassed in an exsiccator before being poured onto the master mold. The subsequent curing is accomplished at room temperature. Figure 8-3 illustrates the options given by this process chain. Due to the higher elasticity of the silicone material (Elastosil M 4642) it is the preferred choice to copy detailed features form the aluminum master. The polyurethane (VU 4452/61 HE) turned out to be comparably brittle which necessitates the deposition of an antiadhesion layer on the aluminum master. PTFE spray has proven to be an adequate method to facilitate the polyurethane peel off. 130 8 Microfluidic devices in soft polymer technology Figure 8-3: Two-step replica molding process for PU and PDMS. The replication of transparent PU layers immediately from a microstructured silicon wafer has also been investigated. Here, the adhesion between the pure silicon wafer and the polyurethane turned out to be too strong for a non-destructive release of the elastomer. The deposition of a fluorinated layer on the silicon wafer by means of a C4F8 passivation process has proven as appropriate method to enable a facile peel off. 8.3.3 Parallelized replication process The introduced process chain offers the potential to parallelize the fabrication process since an appropriate number of secondary master molds can be easily provided. This would not be possible if, for example, a silicon wafer is used as master due to the high fabrication costs. The parallelization of this process is illustrated in Figure 8-4. Figure 8-4: Process chain with parallelized replication: the availability of an arbitrary number of silicone masters enables the parallel fabrication of numerous PU layers. 131 8 Microfluidic devices in soft polymer technology 8.4 Surface modifications of polyurethane The contact angle is utilized as characteristic property to analyze the effect of surface modifications. The measurements were carried out with the Dataphysic Contact Angle System OCA (Dataphysics Instrument GmbH, Germany). In each of the following diagrams, the mean values of five-point-measurements are given together with the corresponding standard deviations. The measurements show that the equilibrium contact angle of the smooth polyurethane surface is in the range of 90° – 100° for both polyurethane formulations (Figure 8-5). The advancing and receding angles show the typical characteristic as expected for a hydrophobic surface with minor intrinsic roughness (see chapter 2). Figure 8-5: Contact angle on a smooth polyurethane surface (five-pointmeasurements). 8.4.1 Hydrophilization by means of flame treatment While treatment methods to render a PDMS surface hydrophilic have been investigated thoroughly, similar investigations for PU surfaces are rarely found in the MEMS literature. Surface activation in an oxygen plasma is a reported method to hydrophilize a PU surface [156]. The hydrophilization method evaluated in this work is flame treatment with NanoSil05 (NanoFlame NF02-Set, Polytec PT GmbH, Germany) which deposits a silanization layer onto the PU surface. Contact angle measurements have proven that this chemical surface modification significantly reduces the contact angle. Nevertheless, similar to other surface activation methods, an aging process was observed due to the dynamic recovery of the polymer surface. After an elapsed time of 20 days only a small decline of the hydrophilization was noticeable but after a time period of four months the original contact angle of approximately 100° was reestablished (Figure 8-6). The large standard deviation values indicate that the silanization via flame treatment leads to an inhomogeneous distribution of the surface energy across the surface. 132 8 Microfluidic devices in soft polymer technology Figure 8-6: Hydrophobic recovery of the VT 3402 surface after silanization (fivepoint-measurements). 8.4.2 Hydrophobic microstructuring of the surface An increase of the contact angle on the PU surface is achieved by microstructuring the surface. This structure is added to the aluminum master by means of laser ablation and then transferred to the elastomer layer via the replica molding process described above. Photographs of the structured aluminum master featuring a 130 µm grid and a droplet dispensed onto the replicated PU surface are shown in Figure 8-7. A Nd:YAG-laser writes the CAD-data to the aluminum master (parameters are given in Appendix E) and thus provides a simple rapid prototyping method to accomplish microstructuring with arbitrary patterns. (a) (b) Figure 8-7: Microscopic image (laser scanning microscope) of the aluminum master structured with a 130 µm grid (a) and contact angle of a water droplet on the replicated PU surface (b). The impact of different grid sizes on the contact angle has been studied. For all pattern a significantly increased contact angle of 140 – 145° was observed (Figure 8-8 (a)). Thus, obviously the grid size does not affect the mean static contact angle. This result appears reasonable in the context of published research work on the wetting of ultrahydrophobic surfaces. There, microstructured surfaces featuring posts of different size, distance and geometry are compared with respect to their wetting behavior in the so-called Cassie mode 133 8 Microfluidic devices in soft polymer technology (see chapter 2.1.2.3). In this mode, the advancing angle is found to be virtually independent of the post size while the receding angle is affected by the geometric features [165, 166]. The hydophobization technique turned out to be less efficient for the Elastocoat® C 6909/1. This is presumably due to the higher viscosity of the prepolymer which prevents a precise replication of the microstructures. (a) (b) Figure 8-8: Hydrophic microstructured surface analyzed for different grid sizes (a). The small deviation between advancing and receding angle is an indication of superhydrophobic behavior (b) (five-point-measurements). The more detailed investigation for the 130 µm grid points out, that the deviation between the advancing and the receding contact angle has diminished (Figure 8-8 (b)). Reviewing the fundamentals of chapter 2, this behavior is in good agreement with the characteristics of a superhydrophobic surface. Thus, the introduced technique yields an inhomogeneous surface composed of solid compartments and air pockets which promotes wetting in the Cassie mode. The expected roll-off of water droplets has been observed for the structured surface. 8.5 Multilayer assembly for polyurethane The most critical part of the multilayer soft lithography is the bond interface. Several methods for bonding of polymer layers have been published yet [13, 151]. All of these approaches are focused on the bonding of PDMS layers. Even though irreversible methods have been reported [34, 152, 154] the bond strength turns out to be critical in order to ensure a reliable sealing which is essential for most microfluidic devices. In this chapter, potential bonding strategies for polyurethane are evaluated in order to optimize the bond strength between adjacent layers. 8.5.1 Investigation of different bond methods In principle the well-known bonding techniques can be divided into three categories regarding the type of the bond. For the first effect, pure adhesion between adjacent layers is utilized which yields a reversible bond of moderate strength. Second, a chemical linkage is 134 8 Microfluidic devices in soft polymer technology initiated by disequilibrial mixing ratios of the two prepolymer components or by plasma activation of the surface. The third category comprises all sorts of glue layers applied to one or two of the bond faces. In this work, the investigated methods include simple adhesion, annealing, variable mixing ratios of the base polymer and the curing agent as well as bonding by means of an additional glue layer. The methods were tested for the polyurethane VT 3402. As the measured bond strength values depend on the specific measurement setup it is critical to compare these results to published data. Therefore, equivalent measurements were carried out with PDMS layers fabricated from Sylgard® 184 in order to assess the results obtained for polyurethane. Thus, the reported forces have to be understood as relative assessment of the bond strength depending on the specific measurement setup. 8.5.1.1 Adhesion For the adhesion strength test two elastomer layers were attached to each other immediately after their release from the master mold. As an alternative, the impact of an isopropanol clean of the bond face prior to the adhesion bonding was examined. Moreover, the improvement of the bond strength by means of an additional annealing step at 100°C for 60 min was studied. 8.5.1.2 Mixing ratio variation For an evaluation of the mixing ratio effect two layers of the polyurethane were bonded with mixing ratios of 1.3:1 and 0.7:1, respectively, instead of the standard mass mixing ratio of 1:1 (polyester : isocyanate). For PDMS, mixing ratios of 5:1 and 15:1, respectively, were utilized instead of the standard mixing ratio of 10:1 (base polymer : curing agent) which is in accordance to approaches reported in the literature [151]. The layers were cured at 100°C for 20 min only before being peeled off the master mold, attached to each other and cured again for another 60 min. 8.5.1.3 Adhesive layer This concept comprises the application of an adhesive layer to establish a permanent bond between adjacent polymer layers. As a first option, the prepolymer itself was utilized as adhesive layer. Here, the cured bond faces were coated with a thin layer of the liquid prepolymer, attached to each other and cured again. As alternative solution, the medical grade glue Vitralit 1810 (Panacol-Elosol GmbH, Oberursel, Germany) was applied as adhesive layer. This low-viscosity UV-glue was dispensed onto one surface and was subsequently cured under UV-exposure within 30 s. 8.5.2 Measurement setup For an experimental investigation of the bond strength a setup based on a tensile sensor (KD9363S, ME-Messsysteme GmbH, Germany) was established. Elastomer blocks with a size of 45 x 30 x 10 mm were bonded to each other before being pulled apart (Figure 8-9 (a)). The force was applied to loops which were embedded into the blocks during the curing 135 8 Microfluidic devices in soft polymer technology process. In this setup the blocks were fixed to the sensor on one side and to a slide on the other side (Figure 8-9 (b)). The slide was moved by means of a set-screw and the corresponding force was recorded by the sensor and transferred to a data file via a RS232interface. The maximum recorded force at the moment of breakdown was taken as bond strength value. (a) (b) Sensor Slide Elastomer blocks Figure 8-9: Two bonded elastomer blocks are pulled apart by a force (a) which is recorded by means of a tensile sensor (b). 8.5.3 Results of bond strength measurements Figure 8-10 summarizes the results obtained for the investigated bonding methods. For polyurethane, the delamination force for the reversible pure adhesion bond was in the range of 5 N. An additional isopropanol cleaning of the bond face turned out to be clearly disadvantageous. A significantly stronger, but still reversible bond was achieved by means of an additional annealing step as well as by application of different mixing ratios. In both cases, the surface was still sticky after the initial 20 min curing time in the furnace which enabled the subsequent development of a higher bond strength during the annealing process. The two attempts based on an adhesive layer, i.e. the PU prepolymer as well as the UV glue Vitralit 1810, yielded an irreversible bond sustaining a maximum force of more than 45 N. At this point the bulk material was corrupted but there was no delamination along the bond interface. For PDMS, the delamination force obtained for pure adhesion bonding was in the range of 3 N. While the additional isopropanol clean did not exhibit a significant effect, the bond strength was clearly increased by an additional annealing step. With different mixing ratios a similar maximum delamination force of nearly 20 N was measured. For the concepts based on an adhesive layer only the option utilizing the liquid PDMS prepolymer provided a sufficiently strong, permanent bond which sustained a maximum force of more than 30 N. The UV-glue worked only in conjunction with a primer and the obtained bond durability was limited to a reversible bond with a delamination force in the range of 20 N. 136 8 Microfluidic devices in soft polymer technology (a) (b) Figure 8-10: Comparison of the bond strength obtained for different bonding methods of PU (a) and PDMS (b) (three samples each). The repeatability for subsequent bonding and delamination steps was considered for a reversible adhesion bond. The measurements were repeated with three different samples of both materials. Within the limit of five cycles no significant decline of the bond performance could be observed for both PU and PDMS (Figure 8-11). Only the initial bond of the polyurethane samples turned out to be slightly stronger than the subsequent bonds. The mean delamination forces obtained for the PU bonds were approximately 1.5-fold larger than the corresponding delamination forces of the PDMS samples. (a) (b) Figure 8-11: Delamination force for repeated adhesion and delamination cycles of PU layers (a) and PDMS (b) (three samples each). 8.6 Discussion The presented results indicate that multilayer soft lithography based on polyurethane is a worthwhile alternative to the established PDMS technology. The investigated polyester 137 8 Microfluidic devices in soft polymer technology based PU VT 3402 KK-NV excels with its superb transparency and a good mechanical stability. With a measured Young’s modulus of approximately 6 MPa this polyurethane elastomer is about an order of magnitude stiffer than the PDMS Sylgard® 184 and features a greater hardness. The polyurethane is impermeable to water but fairly permeable to gas (0.07 µl min-1 cm-2 measured at a pressure difference of 100 mbar). A significant swelling of up to 50 % in volume is observed when the material is immersed in ethanol. The proposed process chain for the fabrication of the PU layers has proven as a particularly practical method. First, the extremely elastic silicone formulation used for the secondary master enables the precise replication of small features including moderate undercuts. Second, these features are easily transferred to the PU VT 3402 layer. Compared to PDMS, the lower viscosity of the PU prepolymer and the shorter degassing time are convenient features for the replication process. Similar to PDMS, the observed shrinking upon curing is negligible. The other investigated polyurethane Elastocoat® C 6909/1 is considered to be less suitable for the manual replication process. Its high viscosity at room temperature impedes the mixing process as well as the replication of small features. Additionally, a complete degassing is hardly achieved and the appearance of the elastomer is opaque. Both hydrophilic and hydrophobic modifications of the PU VT 3402 surface have been demonstrated. Particularly the hydrophobization technique based on the microstructuring of the aluminum master by means of laser ablation provides a convenient rapid prototyping method for the integration of hydrophobic structures into microfluidic devices, e.g. hydrophobic barriers or bubble traps. A main focus of this chapter was on the bonding strategy to assemble reliable multilayer stacks. Here, the gluability of polyurethane is considered as a tremendous advantage. The applied medical grade glue Vitralit 1810 leads to an irreversible bond between polyurethane layers. The capillary effect supports the formation of a film with uniform thickness between the layers and the spreading of the glue inherently stops at the edges of the channels. This glue also provides a sufficient bond strength between PU layers and silicon chips which enables a seamless integration of the silicon micropump into the polymer device. All in all, this method provides a higher bond quality than adhesion or mixing ratio variations and is favored over prepolymer adhesion layers in terms of handling. From the biocompatibility point of view, polyurethane is generally accepted as an appropriate material. It is created by an addition polymerization which naturally does not leave any decomposition products. Nonetheless, the mixing process of the polyurethane prepolymer is critical due to the existence of isocyanates. Free isocyanates are classified as toxic and may evoke biological or medical hazards. Thus, the mixing ratio has to be precisely controlled which clearly prohibits the use of unbalanced mixing ratios for medical applications. 138 Chapter 9 Active Microport 9 Active Microport An automated drug delivery system, a so-called “active microport”, is the target application of the research work presented in this thesis. The system developed in the framework of an interdisciplinary project funded by the Landesstiftung Baden-Württemberg is intended to facilitate the exploration of innovative therapies such as the metronomic therapy and the chronotherapy. Recent preclinical studies have shown that frequent administration in vivo of low doses of chemotherapeutic drugs ("metronomic" dosing) can affect tumor endothelium and inhibit tumor angiogenesis. It promises to reduce significant side effects (e.g. myelosuppression) involving other tissues, even after chronic treatment [167]. For these therapies, the demand for freely programmable, time-modulated release profiles necessitates the availability of an active device. Our project partners of the Tumor Biology Center in Freiburg tested workable prototypes of the active microport in the context of their studies on the continuous and controlled release of soluble antiangiogenic drugs. In particular, the administration of doxorubicine to rats suffering from cancer tumors was the main focus of this project on the application side. Doxorubicine is a stable, highly potent drug which has rheological properties similar to water. Thus, the precise dosing by means of the proposed novel two-stage micropump appears promising to facilitate the continuous release of minute amounts of the drug over a prolonged period of time. 139 9 Active Microport 9.1 System concept On the technical side, the goal of the project is a fully integrated drug delivery system to replace a common passive port system operated in conjunction with an external dosing pump. Thus, the design and performance of the developed concept have to meet the requirements for a future implantation of the system. An implantable active microport obviously has to be small and compact since the acceptable device size is limited by physiological constraints. It has to feature at least four components: the refillable reservoir, a flow control device, i.e. the micropump, a pressure sensor as monitoring device for the dosing process and an electronic control unit (Figure 9-1). The electronic control unit comprises the micropump driver, a microprocessor for data processing and control tasks, a non-volatile memory for storage of a freely programmable dosing sequence and an internal power management system as well as inductive coils for data and power telemetry. Figure 9-1: Concept of an active microport. 9.2 Prototype fabrication The container of the system has to host the individual components and needs to provide embedded microfluidic structures for the fluidic connection. During the prototyping stage a flexible technology is favorable in order to enable rapid redesign cycles. Thus, the housings used throughout this project were based on the multilayer soft lithography with polyurethane as presented in chapter 8. To the end of this project, an injection molded container has been fabricated to serve as basis for future research work. 9.2.1 Multilayer housing The multilayer design of the device is illustrated in Figure 9-2. It encloses the micropump chip, a pressure sensor and the electronic circuit board. The system integrates the reservoir as well as fluidic channels, a refill port and a catheter outlet. As described in chapter 8, the medical grade glue Vitralit 1810 (Panacol-Elosol GmbH, Oberursel, Germany) was used to provide irreversible bonds between the polyurethane layers and to ensure fluidic sealing. The same adhesive was also used to attach the silicon micropump chip, the pressure sensor and the Luer adapter at the outlet. The size of the systems measures only 50 x 35 x 30 mm3 at a weight of 55 g including a battery. 140 9 Active Microport (a) (b) Figure 9-2: Cross-sectional drawing of the active microport design visualizing the embedded fluidic elements and the integration of the silicon micropump chip and the pressure sensor (a) and photograph of a prototype of the active microport system (b). 9.2.2 Stability of doxorubicine in contact with polyurethane The compatibility between the polyurethane material and the antiangiogenic substance doxorubicine has been experimentally approved. For the investigation, three doxorubicine samples were incubated in a polyurethane reservoir for 148 days. Sterile conditions at 4°C, room temperature and 37°C were applied. In parallel, pure doxorubicine samples without PU were incubated under the same conditions. HPLC measurements were performed to determine the stability of the doxorubicine after 30 days, 72 days, 112 days and 148 days. The result is depicted in Figure 9-3. No deviation could be indentified for the samples with and without polyurethane. An increase of the degradation of doxorubicine with temperature was observed. In consequence, a service interval of 4 weeks has been defined as appropriate time period for an implanted system which is exposed to the body temperature. Figure 9-3: Stability of doxorubicine incubated with polyurethane samples under different conditions. In addition to the HPLC measurements the surface of the polyurethane reservoir was investigated optically after the experiment. A visual inspection by a conventional microscope did not reveal any wear of the surface. Roughness measurements with a laser scanning microscope have confirmed this result. 141 9 Active Microport 9.2.3 Injection molded housing While housings based on multilayer soft lithography are favorable in terms of design flexibility, the low stiffness of the elastomer material requires comparably thick side walls in the range of 2 – 3 mm which results in a rather bulky container. The fabrication of the housing by means of injection molding using a rigid polymer significantly reduces the weight of the overall system. For the final design of the active microport an injection molded housing made of polypropylene (200-CA40, Biesterfeld Plastics GmbH, Hamburg, Germany) has been realized. This thermoplastic material features a good translucence. In general, polypropylene is considered as biocompatible material and is frequently used for medical instrumentations and devices, e.g. catheters. The fabrication of the housing was carried out in our laboratory with an Arburg Allrounder 270 U 400 - 70 injection molding machine. Figure 9-4 shows the stainless steal molding tool utilized for the fabrication process. Figure 9-4: Photograph of the mold cavity fabricated into a stainless steal molding tool. The cross-sectional view in Figure 9-5 (a) shows the final design of the active microport. In contrast to the design presented above in section 9.2.1 the gas volume surrounding the micropump and the electronic circuit board in the upper compartment of the container is now used as gas buffer. The connection between the upper compartment and the reservoir cavity is provided by small through-holes. The assembly of the system is achieved by means of the UV-glue Vitralit 1810. This adhesive has proven to work also with polypropylene. Compared to polyurethane, the capillary gluing effect is even more pronounced which facilitates the assembly process. In a first assembly step, the micropump, the pressure sensor and a small printed circuit board with gold bond pads are glued onto an insertion plate made of polypropylene. The electrical connection between the printed circuit board and the piezo-actuator is established by wedgewedge bonding of aluminum wires. Soldering is used to connect the pressure sensor. Then, the insertion plate is mounted into the housing and the Luer adapter as well as the separation foil (Walopur 4201 AU, 25 µm, Epurex Films GmbH & Co.KG, Walsrode, Germany) is added to the device. In a subsequent step, the electronic circuit board is connected to the printed circuit board by means of short wires. For completion of the device, a polypropylene cap is attached to the housing on each side. The bottom cap features a through-hole which is closed with a silicone plug (Elastosil M 4642, Wacker Chemie AG, Munich, Germany) to serve as refill port. 142 9 Active Microport (a) (b) Figure 9-5: Cross-sectional drawing of the injection molded design (a) and assembled device showing the micropump, the pressure sensor and the electronic control unit (b). The presented final design of the active microport measures 49 x 36 x 26 mm3. The weight of the system, i.e. with an empty reservoir, is 36 g including a rechargeable battery with a capacity of 570 mAh (weight 12.5 g). The reservoir features a volume of 10 ml. 9.3 Electronic control unit For the control of the micropump a specific electronic driver circuit was developed in our laboratory (Figure 9-6). It contains step-up converters to provide voltages up to 150 V that are required to drive the piezo-actuators of the micropump. A flexible, software-controlled adjustment of the actuation scheme is realized by means of an 8-bit microcontroller with an associated non-volatile memory (EEPROM). An A/D-converter is implemented for a recording of the pressure sensor signal and a real-time data preprocessing. Power is supplied via a 3 V rechargeable battery connected to the electronic circuit. Currently the wireless communication is realized by means of an IR-interface which will be replaced by an RF-communication unit prior to implantation. Figure 9-6: Photograph of the electronic circuit board containing the micropump driver and sensor data processing capabilities. The programming of the dosing sequence as well as the readout of the sensor data is achieved by means of a system specific LabView interface (Figure 9-7). While the IR interface provides the functionality to start and stop the micropump and to read the sensor values temporarily stored in the EEPROM, the electronic board has to be connected to the PC via a cable for the programming process. 143 9 Active Microport Figure 9-7: LabView-interface for micropump programming and sensor readout. The overall power consumption depends on the applied actuation frequency and sums up to approximately 100 mW for low frequencies (Figure 9-8). Two different high-voltage levels were analyzed and higher voltages have proven to be more power consuming. The power consumption is in between an upper and a lower limit. The minimum power consumption of 48 mW applies during the resting state, i.e. during the phase of the actuation sequence where voltage levels are held constant. Upon each switching of the voltage levels a temporarily increased power consumption is recognized. If the switching occurs very frequently, i.e. at frequencies beyond 2 Hz, a maximum constant power consumption of 234 mW is obtained. For further increased frequencies, the provided high-voltages were found to fall behind the preset values. (a) (b) Figure 9-8: Measurement of the power consumption of the complete device as a function of frequency (a) or flow rate (b). 144 9 Active Microport 9.4 Release monitoring A pressure sensor (MPX2300D, Freescale Semiconductors Inc., Austin, TX, USA) is implemented at the outlet port of the micropump in order to monitor the static and transient pressure appearing in the system. This pressure sensor fulfils a twofold task. First, it is meant to detect a system failure due to catheter occlusion. This principle is also used in macroscopic syringe infusion pumps. However, the small dead volume of the microfluidic system generates a rapid pressure increase in case of an obstructed drug release which can be detected with a much shorter time lag. Second, real-time sensing of the released volume is crucial to ensure patient safety. The discrete strokes of the micropump cause a pulsatile pressure signal which is correlated to the flow rate. Due to the small diameter of the catheter high pressure drops arise at the micropump outlet upon the piezoelectric actuation. This dynamic signal is superimposed by an offset value determined by the pressure at the delivery site, e.g. the blood pressure for intravenous administration. Figure 9-9: Experimental pressure signals recorded near the pump outlet. As shown in Figure 9-9 the pressure signal clearly resolves the discrete pulses recorded for an actuation frequency of 0.5 Hz. An occlusion at the catheter tip leads to a significant increase of the pressure signal within less than 2 minutes. For this measurement, a 17 cm long catheter with an inner diameter of 240 µm and an outer diameter of 750 µm was employed as flow restrictor. 145 9 Active Microport As mentioned before, the pulsatile pressure signal may be converted into a corresponding flow rate for a known fluidic resistance. For low frequencies, intervals of zero flow arise in between the discrete pulses. Thus, the remaining offset pressure during these intervals enables an integrated measurement of the external blood pressure pblood without additional efforts. In sum, this sensor arrangement is suitable for release monitoring and the released volume is obtained straight forward by integration of the pressure signal 1 . (9.1) Experiments using the described catheter confirmed the monitoring capability of this method (Figure 9-10). Here, the catheter constitutes the predominant fluidic resistance and the measured mean pressure drop was in good agreement with the pressure drop expected from Hagen-Poiseuille’s law. Figure 9-10: The integrated pulsatile pressure signal confirms the expected pressure drop in accordance with Hagen-Poiseuille’s law. 9.5 Dosing profiles A time-modulated dosing profile is delivered by the active microport system to enable a patient-specific administration of the drug. The corresponding actuation sequence is stored in the EEPROM of the electronic control unit and automatically changes the actuation frequency after preset time intervals. A test sequence running for two hours is depicted in Figure 9-11. Here, the gray curve shows the output signal of a flow sensor (SLG1430, Sensirion AG, Staefa, Switzerland). The average flow rate in the respective time interval is indicated by the solid curve and demonstrates that the microport system precisely tracks the intended flow profile. 146 9 Active Microport Figure 9-11: Active microport system tracking a preset release profile. 9.6 In-vivo experiments For testing of the device a prototype was used for animal studies to deliver a physiological buffer solution. A catheter was implanted into the Vena jugularis of a rat and was connected to the outlet of the active microport (Figure 9-12 (a)). An external prototype placed on a side table was employed for this trial. The pressure signal recorded by the pressure sensor of the system was evaluated for various frequencies. The measurements confirmed the expected linear relationship between the flow rate and the average pressure signal (Figure 9-12 (b)). The indicated flow rates were based on a calibration curve of the micropump determined previously. (a) (b) Figure 9-12: Implantation of a venous catheter (a) and recording of the pressure sensor signal at different flow rates (b). 147 9 Active Microport 9.7 Application “Pharmaport” Besides the target application of this system as an implantable active microport an additional promising application has been identified in the field of pharmacological research. Trials of new pharmaceutical substances involve comprehensive animal studies to achieve a medical approval for the developed drug. Usually, discrete amounts of the drug are manually injected into the animal, e.g. mice or rats, or an osmotic micropump (Alzet®, DURECT Corp., Cupertino, CA, USA) is implanted subcutaneously which delivers a constant flow rate for continuous administration. Here, for the proving of modern approaches in cancer treatment such as chronotherapy, a programmable micropump would be desirable which can be carried by the animal. The compact size of our systems enables the animal to carry the infusion pump in a specially designed backpack (Figure 9-13). This way, the clinicians and pharmaceutical researchers obtain a full controllability of the delivery profile which clearly contributes to the significance of their studies. (a) (b) Figure 9-13: Concept (a) and photograph (b) of a remote-controlled pharmaport for the application in pharmaceutical animal studies where the developed infusion system is carried by a rat in a backpack. 148 Chapter 10 Summary 10 Summary The development and evaluation of a novel two-stage micropump concept and its integration into an automated drug delivery system was the core of this thesis. The main objective of the design development was the precise dosing capability based on an adjustable, backpressure independent flow rate in the range of 0.1 – 50 µl/min. This feature constitutes a novelty in the field of reciprocating micropumps and is of particular interest for all micropump applications with exposure to variable pressure heads. The two-stage micropump design comprises two active valves controlled by two piezoactuators. The fabrication of this silicon micropump involves mainly standard MEMS processes. The main advantage of piezo-actuators is seen in their short response time and their low power consumption. The first aspect is important to achieve a rapid valve switching and to ensure an optimum controllability of the valve state. The energy efficiency is a crucial aspect for all portable devices. The relevance of this issue is evident since the concept of the developed active microport system is to design a device appropriate for future implantation. The overall power consumption of the developed system has been reduced to approximately 50 - 200 mW depending on the actuation frequency. A set of control variables and design variations has been systematically investigated in the framework of this thesis. For a deeper understanding, a lumped parameter model of the micropump has been established. The model is based on both analytical approaches and numerical simulations such as FEM simulations of the membrane deflection. It extends the established lumped parameter modeling techniques for reciprocating micropumps and covers the specific characteristics of the two-stage concept. As a result of this work, a general model for the simulation of a two-stage micropump with active valves has been derived. This model was used to identify suitable control variables and to optimize the applied actuation scheme. 149 10 Summary In regard to the investigated control variables such as voltage settings or frequency variations, the simulation results strengthened the phenomena revealed by the experimental examination. As an example, the upstroke voltage applied to the piezo-actuators is almost linearly related to the stroke volume which gives immediate access to calibrate the flow rate. This linear relationship has been experimentally observed for many micropumps and has been confirmed by the lumped parameter simulations. The high insensitivity of this two-stage micropump design to variations in outlet pressure has been consistently confirmed. This performance is based on the concept of a constant cut-off pressure in the pump chamber at the end of each pump cycle which is inherently provided by this design. The differential fluidic output resistance has been introduced as a new figure of merit to account for the specific backpressure characteristic of this micropump concept. A flow rate decline of only 10 % up to a backpressure of 30 kPa has been proven for low frequencies of 0.25 Hz. This result is exceptional for reciprocating micropumps and enables a precise dosing within a wide backpressure range. The maximum backpressure experimentally achieved with this two-stage micropump was approximately 65 kPa. It was essentially limited by the strength of the piezo-actuators when closing the outlet valve. In the gas pumping mode, a modified actuation sequence provides a more pronounced forward transport which equips the micropump with a full self-priming capability and enables the transport of compressible fluids. Obviously, in this actuation mode the micropump cannot sustain high backpressures since there is a phase included where both valves are opened. Nevertheless, the gas pumping mode is ideally suited to prime the micropump or to temporarily increase the throughput in applications with low outlet pressure heads. A severe concern is the capillary effect across a gas-liquid interface which becomes relevant for alternate gas and liquid pumping. As the well-known criterion for the critical compression ratio of reciprocating micropumps with passive check valves does not apply for this two-stage design with active valves, a new critical compression ratio was deduced based on analytical derivations and FEM simulations. It denotes the critical compression ratio which is necessary to overcome the capillary pressure drop and to achieve robust and bubble-tolerant pumping. Here, the chamber modification (design II) reduces the dead volume of the pump chamber and hence yields a higher compression ratio which is advantageous for the transport of gases. An alternative thermal actuation concept based on the expansion of paraffin wax upon its solid-liquid phase transition has been explored in this work. The general feasibility of this approach has been demonstrated with a single-membrane, two-stage micropump. Here, a resistive heater fabricated on top of the diaphragm was used to initiate the expansion process which leads to a membrane deflection. In sum, a net flow rate in the range of 80 nl/min has been verified. Additionally, a novel direct heating strategy has been presented which induces resistive heating by means of a conductive paraffin wax. The dispersion of carbon black particles in the paraffin wax is the main innovation of this concept as it drastically increases the efficiency and time response of the actuator and also reduces the fabrication complexity. Simple silicon membrane actuators have been used for a proof of concept. A periodic 150 10 Summary deflection with a cycle time of 5 s has been demonstrated. The power consumption was determined to be in the range of 1 W. The multilayer technique with polyurethane has been identified as a suitable technology for prototyping in the field of polymer microfluidics. It enables sophisticated designs with different embedded structures and the co-integration with silicon chips. Appropriate surface modifications have been identified to render the polyurethane surface either hydrophobic or hydrophilic. Depending on the application demands different bond methods are available to produce either reversible or irreversible bonds. The benefit of capillary gluing with a lowviscosity UV-adhesive has been clearly demonstrated for the assembly of multilayer polyurethane stacks. The developed active microport system incorporates the two-stage micropump, the associated electronic control unit, a pressure sensor and a drug reservoir into a compact device with dimensions comparable to a conventional subcutaneous port. The miniaturized high-performance electronic control unit enables arbitrary, patient-specific release profiles. This electronic circuit is optimized for both energy consumption and weight, which are both essential parameters for a portable device. The data of the implemented pressure sensor are used to permanently monitor the dosing process and to detect a potential catheter occlusion. The polyurethane soft lithography process is utilized to provide high design flexibility for the production of housings during the different development stages. For the final system prototype, an injection molded container made from polypropylene has been fabricated which measures only 49 x 36 x 25 mm3. The functionality of the system has been tested comprehensively in the laboratory. The presented prototype meets the specifications regarding the micropump as well as the system functionality as stated in chapter 1. The micropump performance covers the desired flow rates and satisfies the pressure specifications. The available control parameters provide a full electrical control of the micropump and enable an automated dosing process with freely programmable profiles. Preliminary in vivo experiments have been successfully conducted in which the system was used to deliver physiological solutions as well as doxorubicine to rats via a venous catheter. 151 10 Summary 152 Chapter 11 Outlook 11 Outlook The prototype of the active micropump system presented in this work has proven the feasibility of the pursued concept. The functionality of the individual components and the assembly into an integrated system have been demonstrated. The system size and weight are considered appropriate for a future implantation. Even though biocompatibility aspects have been considered for the fabrication of the system, the package has to be revised during subsequent development stages. Moreover, a substantial testing of the device as an external delivery system in animal experiments and subsequently in clinical trials has to be completed successfully to eventually achieve a medical approval for the active microport. This assessment procedure will be both time and cost intensive. On the track towards implantation of the device a number of additional technical issues will have to be addressed. A qualified process line for the silicon micropump fabrication needs to be established. In contrast to the prototype fabrication at the IMTEK cleanroom, the involvement of a certified MEMS foundry would help to minimize the performance deviations between different micropump samples and to increase the yield. Additionally, the gluing of the piezo-discs should be automated as far as possible since the manual assembly induces a significant variation of the micropump performance. From the electronics point of view, the determined power consumption of approximately 100 mW is still critical for an implantable device. For the sake of patient compliance recharge intervals of about one day are not acceptable which necessitates a further reduction of the energy consumption. A potential strategy could be the use of multilayer piezo-actuators. The lower actuation voltages required to drive multilayer actuators would enable a higher efficiency of the step-up converters which generate the actuation voltages. On the other hand the higher charging currents expected for multilayer piezo-actuators have to be taken into account in order to 153 11 Outlook determine the potential energy savings. In terms of patient safety, a lower actuation voltage would also reduce the requirements for electric isolation. The electronic control unit also needs to be equipped with a RF telemetry module for transcutaneous data and energy transfer. This communication unit will eventually replace the IR interface currently used for data transmission. The development of a robust and user friendly interface for programming and controlling of the active microport is regarded as an urgent engineering task for future work. The current version of a LabView interface turned out to be partly instable when used on different computers and the readout of the pressure sensor has not been fully integrated with the micropump control window yet. An improved software version would also need to include routines which are capable of coping with user errors in order to be ready for use by patients. For the system assembly the injection molded housing appears appropriate for the external version of the active microport. Nevertheless, degradation or aging of the glue layers used to attach the caps on both sides of the system is considered to be a potential failure mechanism. Here, an additive polymer sheathing could provide a more reliable hermetic sealing. It is evident that the aim of implantation implies the use of biocompatible and nondegradable materials. Even though medically approved polypropylene formulations are available, a titanium container or a titanium sheathing of the polymer housing is considered as a potential alternative. Sterilization of the device is another critical issue. Different strategies are conceivable including ethanol sterilization, gas sterilization or sterilization by means of e-beam or γ−radiation. So far, a sterilization process for the device has not been established as the antiangiogenic substance doxorubicine is destructive for any microorganism such as bacteria. For ergonomic reasons the shape of the container will need to be revised prior to implantation. The refill port needs to be accessible transcutaneously and the shape has to be optimized for the implantation site such as the abdominal region. Finally, systematic long term tests would have to be conducted to ascertain the durability and expected life time of the active microport system. 154 Acknowledgements The work presented in this thesis has been substantially supported by a large number of people and I would like to thank all of them for their assistance. In order to keep this list comparably short I wish to mention only a few of them by name and wish to express my gratitude to • • • • • • • • • • • • • • • • • Prof. Dr. Peter Woias for the opportunity to work in his laboratory, for providing guidance and mentoring for my thesis and for supporting my concepts and ideas Dr. Frank Goldschmidtböing for his continuous support as my group leader in an always cooperative way, for his backing throughout the entire project and for fruitful discussions about thesis related aspects Prof. Dr. Ulrich Massing for his willingness to review my thesis as an expert Dr. Peter Jantscheff, Dr. Norbert Esser and Prof. Dr. Ulrich Massing for the always cooperative, fruitful and pleasant team work in our project team the “Landesstiftung Baden-Württemberg” for financial support of the project “Active Microport” Stefan, Martin and Michael for the excellent and enjoyable atmosphere in our office Marika for doing all the paper stuff and taking care of our emotional balance Franz for turning handwritten sketchy drawings into mechanical devices with outstanding precision Martin and Michael (and also Elmar) for helping out with each and every computer problem My former colleague Michael for telling me about this interesting project 3.5 years ago Our whole laboratory group for an absolutely enjoyable and unforgettable time (including a couple of leisure activities) My diploma students Shahab Nadir and Philip Katus as well as my student workers Benjamin Müller, Chiheb Farhat and Johannes Hartwiger for their excellent work Christian Peters and the Laboratory for Microelectronics for making the tensile sensor available Christian Dorrer and the Laboratory for Chemistry & Physics of Interfaces for providing access to the contact angle measurement system IMTEK cleanroom service center for their support of the micropump fabrication My family for their everlasting support in all respects Meike for her encouragement, patience and for being always by my side 155 156 Appendix A Contact angle Measurement of the contact angle on a SiO2 wafer: Figure A-1: Contact angle measurement of a smooth silicon surface covered by a 400 nm oxide layer (SiO2). The measurement was carried out with the Dataphysic Contact Angle System OCA (Dataphysics Instrument GmbH, Germany). 157 Appendix A 158 Appendix B Pressure induced deflection: Derivation of the bending line The mathematical derivation for the solution of the bending line is presented below. It is based on the general solution obtained for the bending of a circular plate exposed to a uniformly distributed load q. The overall bending line is composed of two different solutions for the inner region wi(r) and the outer region wo(r). The corresponding flexural rigidities are Di and Do and the radii of the regions are Ri and Ro. Given equation (2.40) the general solutions for the two regions are · 64 1 4 · · 64 1 4 · 0 · For the solution of the inner region the logarithmic term vanishes since the deflection at the center of the plate (r = 0) needs to be a finite value. At the clamped edge of the outer region the applying boundary conditions are 0 0 . Therewith, the following equations are determined: 0 · 64 1 4 · · 64 1 4 · · 16 0 · 0 (I) 0 1 2 · 0 (II) For the transition point of the two regions a continuous transition of the bending line and the derivative of the bending line is required: · 64 1 4 · · 64 1 4 · · 0 (III) 159 Appendix B · 16 1 2 · · 16 1 2 · 0 (IV) The bending moment in radial direction is obtained from the bending line by . For a balanced bending moment at the transition point the following equation has to be fulfilled: , 1 2 3 · 16 1 2 1 , · 16 3 · 16 1 2 1 2 · 16 1 2 1 1 1 2 (V) 0 This linear system of equations (I) – (V) has been solved for the coefficients C1 to C5 by means of the symbolic tool box of Matlab: C1 = -1/8*q*1/Di*(- ν Ri4 Do-Ro2 Do Ri2+Ri4 Di-Ri4 Do+Ri2 Do ν Ro2+Ro2 Ri2 Di+ ν Ri4 Di-2 Di Ro4Ri2 ν Di Ro2)/(-Do Ri2-Do Ri2 ν -Ro2 Do+ ν Ro2 Do+Ri2 Di-Di Ro2+Di Ri2 ν - ν Di Ro2) C2 = 1/64*q/(Di Do) *(-Di Ro6 Do-Ri6 Di2-4*log(Ri/Ro) Di Ro2 ν Ri4 DoDi2 Ro6+4*log(Ri/Ro) Di Ro4 Do Ri2-4*log(Ri/Ro) Di2 Ro4 Ri2 ν-4*log(Ri/Ro) Di2 Ro4 Ri2Ri6 Do2 ν-i4 Do2 Ro2+2*Ri6 Do Di+Ri4 Di2 Ro2-Ri6 Di2 ν +Di2 Ro4 Ri2-Di2 Ro6 ν Ri6 Do2+Ri4 Do2 ν Ro2+2 Ri6 Do Di ν+Ri4 Di2 ν Ro2-Di Ro4 Do Ri2 +Di Ro6 ν Do+Di2 Ro4 Ri2 ν 2 Ri4 Do ν Di Ro2-Di Ro4 Do Ri2 ν+4*log(Ri/Ro) Di2 Ro2 Ri4 +4*log(Ri/Ro) Di Ro4 Do Ri2 ν +4*log(Ri/Ro) Di2 Ro2 ν Ri4-4*log(Ri/Ro) Di Ro2 Ri4 Do)/(-Do Ri2-Do Ri2 ν -Ro2 Do+ ν Ro2 Do +Ri2 Di-Di Ro2+Di Ri2 ν - ν Di Ro2) C3 = -1/8*q*1/Do*(-ν Ri4 Do+ν Ri4 Di-Do Ro4-ν Di Ro4+Do ν Ro4+Ri4 Di-Di Ro4-Ri4 Do)/(-Do Ri2Do Ri2 ν-Ro2 Do+ν Ro2 Do+Ri2 Di-Di Ro2+Di Ri2 ν-ν Di Ro2) C4 = 1/16*q/Do*Ro2 Ri2 (Ro2 Do+ν Ro2 Do-Di Ro2-ν Di Ro^2-Do Ri2 ν+Di Ri2 ν+Ri2 Di-Do Ri2)/(-Do Ri2Do Ri2 ν-Ro2 Do+ν Ro2 Do+Ri2 Di-Di Ro2+Di Ri2 ν-ν Di Ro2) C5 = 1/64*q/Do*Ro2*(Ro2 Do Ri2+Ri2 Do ν Ro2-Do Ro4+Do ν Ro4-Ro2 Ri2 Di-Di Ro4-Ri2 ν Di Ro2ν Di Ro4-2 ν Ri4 Do+2 ν Ri4 Di+2 Ri4 Di-2 Ri4 Do)/(-Do Ri2-Do Ri2 ν-Ro2 Do+ν Ro2 Do+Ri2 DiDi Ro2+Di Ri2 ν-ν Di Ro2) 160 Appendix C FEM Simulation C.1. Model geometry The following FEM-model of the piezoelectric membrane actuator has been created with the COMSOL MultiphysicsTM geometry editor and the meshing was conducted by means of the included mesh generator. (b) (a) Figure C-1: Illustration of the FEM-simulation model consisting of the silicon membrane (thickness 100 µm), the adhesive layer (10 µm) and the piezoactuator (200 µm) (a) and visualization of the unstructured mesh (b). Due to symmetry conditions only one quarter of the real structure has been implemented. Table C-1: Details of the mesh geometry Number of elements 67705 Number of nodes 12586 Degrees of freedom 330854 C.2. Convergence of the simulation The convergence of the FEM simulation model for increased numbers of elements is investigated for the following exemplary parameter configuration: • • Actuation voltage: 80 V Pressure load: -50 kPa The number of mesh elements is consecutively increased from 2804 elements up to 130647 elements. The following plot in Figure B-2 shows the obtained deflection for the center of the membrane together with the required simulation time. 161 Appendix C Figure C-2: Membrane deflection and simulation time in dependence of the number of mesh elements. The diagram illustrates that the result is continuously converging towards a final value for increased element number. At the same time the simulation effort increases rapidly leading to a significantly extended simulation time. As a trade of between accuracy and simulation time, an element number of 67705 is chosen for the standard model. The error scale in figure B-1 indicates that the expected deviation from the convergent value is less than 1 %. 162 Appendix D Compression of gas-filled chamber In this section the pressure increase is derived that arises in the gas filled pump chamber upon closing of the inlet valve. For this calculation isothermal compression of the gas volume is assumed. Note that the following calculation is performed under the artificial condition of zero outflow despite the open outlet valve. The intention of this assumption is to determine whether the expected pressure increase is large enough to outbalance the pressure drop across a gas-liquid interface which blockades the outlet valve. Before closing of the inlet valve: p0 100000 Pa V0 Vchamber ∆V1 atmospheric pressure ∆V2 with design determined chamber volume for flat membranes (for design II) 1.02 · 10 chamber Nominal displacement volumes at 80V (simulated values) 9.43·10‐11 9.43·10‐11 ∆V1 ∆V2 After closing of the inlet valve: p1 p0 V1 Vchamber Δp ∆V1 ∆p·Cc1 ∆V2 ∆p·C2 ∆V1 ‐4.65·10‐11 Cc1 2.57·10‐16 C2 1.17·10‐15 Volume displacement of the closed inlet diaphragm (140V) supported by the valve lip Reduced capacitance of the inlet membrane when supported by the valve lips Capacitance of the outlet diaphragm For isothermal compression, the ideal gas law requires p0 ·V0 p1 ·V1 ∆p 11480 . 163 Appendix D Therewith, the center deflection Δw is obtained by ∆w ∆w2 fv ·∆p 3.87·10‐6 m m 4.43·10‐11 Pa ·11480 Pa which implicates a gap height of h 164 h0 ∆w 1μm 4.38 μm 5.38 μm. 4.38 μm Appendix E Laser parameter Table E-1: Parameter set for laser ablation of hydrophobic structures Laser DPL Magic Marker Wave length 1064 nm Intensity 100 % Pulse repitition rate 4 kHz Pulse length 10 µs Velocity 10 mm/s Number of runs 8 165 Appendix E 166 Appendix F Datasheets 167 Appendix F 168 Appendix F 169 Appendix F 170 List of publications Patents 1. Mikropumpe, DE102005038483, erteilt am 14.8.2005 Mikropumpe, PCT/EP2006/007988, angemeldet am 11.08.2006 2. Überwachungseinheit zur Fluiddosierung und Mikrodosieranordnung, DE102005058080.7, angemeldet am 6.12.2005 3. Konträrmembranantriebe für Mikropumpen, DE102006028986.2, angemeldet am 23.06.2006 Journal publications 1. A. Geipel, A. Doll, P. Jantscheff, N. Esser, U. Massing, P. Woias, F. Goldschmidtböing Novel two-stage backpressure-independent micropump: modeling and characterization J. Micromech. Microeng. 17, 949-959, 2007. 2. A. Geipel, F. Goldschmidtböing, A. Doll, P. Jantscheff, N. Esser, U. Massing, P. Woias An implantable active microport based on a self-priming high-performance two-stage micropump Sens. Actuators A, available online since January 9, 2008 doi:10.1016/j.sna.2007.11.024 3. A. Geipel, F. Goldschmidtböing, P. Jantscheff, N. Esser, U. Massing, P. Woias Design of an implantable active microport system for patient specific drug release Accepted for publication, J. Biomedical Microdevices Conference proceedings 4. A. Geipel, F. Goldschmidtböing, A. Doll, P. Jantscheff , N. Esser, U. Massing, P. Woias Vollintegrierter aktiver Mikroport im PDA-Format: System, Performance und Applikationen Proc. Mikrosystemtechnik Kongress 2007, Dresden, Germany, pp. 719 – 722, 2007. 5. A. Geipel, P. Katus, F. Goldschmidtböing, P. Woias Peristaltische Ein-Membran-Mikropumpe mit zwei aktiven Ventilen Proc. Mikrosystemtechnik Kongress 2007, Dresden, Germany, pp. 935 – 938, 2007. 171 List of publications 6. A. Geipel, F. Goldschmidtböing, A. Doll, S. Nadir, P. Jantscheff, N. Esser, U. Massing, P. Woias An implantable active microport based on a self-priming high-performance two-stage micropump Proc. IEEE Transducers '07, Lyon, France, pp. 1943-1946, 2007. 7. A. Geipel, F. Goldschmidtböing, A. Doll , C. Farhat, P. Jantscheff, N. Esser, U. Massing, P. Woias Biocompatible polymer encapsulation with embedded functional structures for medical devices Proc. 5th IASTED Int. Conf. BioMED07, Innsbruck, Austria, pp. 272 – 276, 2007. 8. A. Geipel, A. Doll, F. Goldschmidtböing, B. Müller, P. Jantscheff, N. Esser, U. Massing, P. Woias Design of an implantable active microport system for patient specific drug release Proc. 4th IASTED Int. Conf. BioMED06, Innsbruck, Austria, pp. 161 – 166, 2006. 9. A. Geipel, A. Doll, F. Goldschmidtböing, P. Jantscheff, N. Esser, U. Massing, P. Woias Pressure-independent micropump with piezoelectric valves for low flow drug delivery systems Proc. IEEE MEMS 2006, Istanbul, Turkey, pp. 786-789, 2006. 10. A. Geipel, A. Doll, F. Goldschmidtböing, P. Jantscheff, N. Esser, U. Massing, P. Woias Design of an implantable active microport system for autonomous time-variant drug release Proc. Mikrosystemtechnik Kongress 2005, Freiburg, Germany, pp. 419-422, 2005. Journal publications (co-author) 11. A. Doll, M. Wischke, A. Geipel, H.-J.Schrag, F. Goldschmidtboeing, P. Woias Characterization of Active Silicon Microvalves with Piezoelectric Membrane Actuators Microelectronic Engineering 84 (5-8), 1202-1206, 2007. 12. A. Doll, M. Wischke, A. Geipel, H.-J.Schrag, U.T.- Hopt, F. Goldschmidtboeing, P. Woias A Novel Artificial Sphincter Prosthesis Driven by a Four Membrane Silicon Micropump Sens. Actuators A, available online since April 4, 2007, doi:10.1016/j.physletb.2003.10.071 Conference proceedings (co-author) 13. A novel self-heating paraffin membrane micro-actuator F. Goldschmidtböing, P. Katus, A. Geipel, P. Woias Proc. IEEE MEMS 2008, Tuscon, USA, 2008. 172 List of publications 14. A. Doll, M. Wischke, A. Geipel, H.-J. Schrag, U.T. Hopt, F.Goldschmidtböing, P.Woias A Novel Artificial Sphincter Prosthesis Driven by a Four Membrane Silicon Micropump Proc. APCOT 2006, Singapore, D-36, pp.1-4, 2006. 15. M. Wischke, A. Doll, A. Geipel, F. Goldschmidtboeing, P. Woias Fabrication of multi-layer piezo-actuators and reliable assembling strategy for high performance micropumps Proc. Actuator 2006, Bremen, pp. 276-280, 2006. 16. F. Goldschmidtböing, A. Doll, A. Geipel, M. Wischke, Peter Woias Design of Micro Diaphragm Pumps with Active Valves Proc. FEDSM2006, 2006 ASME Joint U.S. – European Fluids Engineer Summer Meeting, Miami, USA, FEDSM2006-98506, pp.1-9, 2006. 17. A. Doll, A. Geipel, M. Heinrichs, F. Goldschmidtböing, P. Woias, H.-J. Schrag, U.T. Hopt Eine neue Schließmuskelprothese basierend auf einer Hochleistungs-Silizium Mikropumpe Proc. Mikrosystemtechnik Kongress 2005, Freiburg, Germany, pp. 495-500, 2005 173 List of publications 174 Bibliography [1] J.G. Smits, Piezoelectric micropump with three valves working peristaltically, Sens. Actuators A 21, 203-206, 1990. [2] H.T.G. van Lintel, F.C.M. van de Pol, S. Bouwstra, A piezoelectric micropump based on micromachining of silicon, Sens. Actuators 15, 153-167, 1988. [3] M. Esashi, S. Shoji, A. Nakano, Normally closed microvalve and micropump fabricated on a silicon wafer, Sens. Actuators 20, 163-169, 1989. [4] P. Woias, Micropumps - past, progress and future prospects, Sens. Actuators B 105, 2838, 2005. [5] D.J. Laser and J.G. Santiago, A review of micropumps, J. Micromech. Microeng. 14, R35–R64, 2004. [6] N.-T. Nguyen, X. Huang, T. Chuan, MEMS-micropumps: a review, ASME J. Fluids Eng.124, 384–392, 2002. [7] M. Richter, R. Linnemann, P. Woias, Robust design of gas and liquid micropumps, Sens. Actuators A 68, 480–486, 1998. [8] R. Zengerle, J. Ulrich, S. Kluge, M. Richter, A. Richter, A bidirectional silicon micropump, Sens. Actuators A 50, 81-86, 1995. [9] S. Böhm, W. Olthuis, P. Bergveld, A plastic micropump constructed with conventional techniques and materials, Sens. Actuators A 77, 223–228, 1999. [10] D. Accoto, M.C. Carrozza, P. Dario, Modelling of micropumps using unimorph piezoelectric actuator and ball valves, J. Micromech. Microeng. 10, 277–281, 2000. [11] C. Yamahata, F. Lacharme, J. Matter, S. Schnydrig, Y. Burri, M.A.M. Gijs, Electromagnetically actuated ball valve micropumps, Proc. IEEE Transducers'05, Seoul, Korea, pp. 192–195, 2005. [12] J.M. Berg, R. Anderson, M. Anaya, B. Lahlouh, M. Holtz, T. Dallas, A two-stage discrete peristaltic micropump, Sens. Actuators A 104, 6–10, 2003. [13] O.C. Jeong, T. Yamamoto, S.W.Lee, T. Fujii, S. Konishi, Surface modification, mechanical property, and multi-layer bonding of PDMS and its applications, Proc. 9th Int. Conf. µTAS, Boston, MA, pp. 202-204, 2005. [14] C.-H. Wang and G.-B. Lee, Pneumatically driven peristaltic micropumps utilizing serpentine-shape channels, J. Micromech. Microeng. 16, 341–348, 2006. 175 Bibliography [15] W. Mamanee, A.Tuantranont, N.V. Afzulpurkar, N. Porntheerapat, S. Rahong, A. Wisitsoraat, PDMS based thermopneumatic peristaltic micropump for microfluidic systems, J. Phys.: Conf. Series 34, 564–569, 2006. [16] J. Xie, J. Shih, Q. Lin, B. Yang, Y.-C. Tai, Surface micromachined electrostatically actuated micro peristaltic pump, Lab Chip 4, 495–501, 2004. [17] B. Husband, A.G.R. Evans, T. Melvin, Investigation for the operation of an integrated peristaltic micropump, J. Micromech. Microeng 14, 64–69, 2004. [18] M.M. Teymoori and Abbaspour-Sani, Design and simulation of a novel electrostatic peristaltic micromachined pump for drug delivery applications, Sens. Actuators A 117, 222–229, 2005. [19] C.-W. Huang, S.-B. Huang, G.-B. Lee, Pneumatic micropumps with serially connected actuation chambers, J. Micromech. Microeng. 16, 2265–2272, 2006. [20] C. Grosjean and Y.-C. Tai, A thermopneumatic peristaltic micropump, Proc. IEEE Transducers'99, Sendai, Japan, pp. 3P5.10, 1999. [21] M. Stehr, S. Messner, H. Sandmaier, R. Zengerle, The VAMP—a new device for handling liquids or gases, Sens. Actuators A 57, 153–157 , 1996. [22] A. Wego and L. Pagel, A self-filling micropump based on PCB technology, Sens. Actuators A 88, 220–226, 2001. [23] A. Olsson, P. Enoksson, G. Stemme, E. Stemme, A valve-less planar pump isotropically etched in silicon, J. Micromech. Microeng. 6, 87 – 91, 1996. [24] H.-L. Yin, J. Hsieh, H.-H. Hsu, A novel elastomer membrane micropump with embedded coils for off-chip pumping applications, Proc. IEEE MEMS'06, Istanbul, Turkey, pp. 770773, 2006. [25] F. Goldschmidtböing, A. Doll, A. Geipel, M. Wischke, P. Woias, Design of micro diaphragm pumps with active valves, Proc. FEDSM'06, Miami, USA, pp. 1–9, 2006. [26] D. Maillefer, S. Gamper, B. Frehner, P. Balmer, H. van Lintel, P. Renaud, A highperformance silicon micropump for disposable drug delivery systems, Proc. IEEE MEMS'01, Interlaken, Switzerland, pp. 413–417, 2001. [27] Debiotech S.A., Lausanne, Switzerland. www.debiotech.com. [28] Codman & Shurtleff, Inc.,Raynham, MA, USA. www.codmanjnj.com. [29] Medtronic MiniMed Inc., Northridge, CA, USA. www.minimed.com. [30] Tricumed Medizintechnik GmbH, Kiel, Germany. www.tricumed.de. [31] Eksigent Technologies, LLC, Dublin, CA. www.eksigent.com. [32] T. Goettsche, J. Kohnle, M. Willmann, H. Ernst, S. Messner, R. Steger, M. Storz, W. Lang, R. Zengerle, H. Sandmaier, Novel approaches to microfluidic components in highend medical applications, Proc. Transducers'03, Boston, USA, pp. 623–626, 2003. 176 Bibliography [33] V. Namsivayam, R.G. Larson, D.T. Burke, M.A. Burns, Transpiration-based micropump for delivering continuous ultra-low flow rates, J. Micromech. Microeng. 13, 261–271, 2003. [34] M.A. Unger, H.-P. Chou, T. Thorsen, A. Scherer, S.R.Quake, Monolithic microfabricated valves and pumps by multilayer soft lithography, Science 288, 113-116, 2000. [35] L.-S. Jang and W.-H. Kan, Peristaltic piezoelectric micropump system for biomedical applications, Biomed. Microdevices 9, 619-626, 2007. [36] Bartels Mikrotechnik GmbH, Dortmund, Germany, Die Mikropumpe – Herzschlag der Mikrofluidik, Datasheet [online 2007]. www.bartels-mikrotechnik.de. [37] R. Linnemann, P. Woias, C.-D. Senfft, J.A. Ditterich, A self-priming and bubble-tolerant piezoelectric silicon micropump for liquids and gases, Proc. IEEE MEMS'98, Heidelberg, Germany, pp. 532-537, 1998. [38] V.l Singhal, S.V. Garimella, J.Y. Murthy, Low Reynolds number flow through nozzlediffuser elements in valveless micropumps, Sens. Actuators A 113, 226–235, 2004. [39] thinXXS Microtechnology AG, Zweibrücken, Germany. www.thinxxs.com. [40] A. Doll, F. Goldschmidtböing, M. Heinrichs, P. Woias, H.-J. Schrag, U.T. Hopt, A piezoelectric bidirectional micropump for a medical application, Proc. IMECE'04, 61083, Anaheim, Netherlands, 2004. [41] A. Doll, M. Wischke, A. Geipel, H.-J. Schrag , U.T. Hopt, F. Goldschmidtböing, P. Woias, A novel artificial sphincter prosthesis driven by a four membrane silicon micropump, Proc. APCOT'06, Singapore, vol D-36, pp. 1-4 , 2006. [42] W. Inman, K. Domansky, J. Serdy, B. Owens, D. Trumper, L.G. Griffith, Design, modeling and fabrication of a constant flow pneumatic micropump, J. Micromech. Microeng. 17, 891–899, 2007. [43] G.-H. Feng and E.S. Kim, Micropump based on PZT unimorph and one-way parylene valves, J. Micromech. Microeng. 14, 429–435, 2004. [44] C. Niezrecki, D. Brei, S. Balakrishnan, A. Moskalik, Piezoelectric Actuation: State of the Art,The Shock and Vibration Digest 33 (4), 269-280, 2001. [45] PI Ceramics GmbH, Lederhose, Germany, Piezo Actuator Tutorial [online 2007]. www.piceramic.com. [46] P. Pertsch, S. Richter, D. Kopsch, N. Krämer, J. Pogodzik, E. Hennig, Reliability of piezoelectric multilayer actuators, Proc. Actuator'06, Bremen, Germany, 2006. [47] Y. Liu and M. Kohl, Formgedächtnis – Mikroventile mit hoher Energiedichte,Wiss. Bericht FZKA 6934, Forschungszentrum Karlsruhe GmbH, Karlsruhe, 2001. [48] P. Krulevitch, A.P. Lee, P.B. Ramsey, J.C. Trevino, J. Hamilton, M.A. Northrup, Thin film shape memory alloy microactuators, J. Microelectromech. Systems 5 (4), 270-282, 1996. [49] M. Kohl, I.Hürst, B. Krevet, Time response of shape memory microvalves, Proc. Actuator2000, Bremen, Germany, pp. 212-215, 2000. 177 Bibliography [50] S. Zimmermann, J.A. Frank, D. Liepmann, A.P. Pisano, A planar micropump utilizing thermopneumatic actuation and in-plane flap valves, Proc. IEEE MEMS'04, Maastricht, Netherlands, pp. 462-465, 2004. [51] W.H. Song and J. Lichtenberg, Thermo-pneumatic, single-stroke micropump, J. Micromech. Microeng. 15, 1425-1432, 2005. [52] J.A. Potkay and K.D. Wise, An electrostatically latching thermopneumatic microvalve with closed-loop position sensing, Proc. IEEE MEMS'05, Miami, FL, USA, pp. 415-418, 2005. [53] H. Yuan and A. Prosperetti, The pumping effect of growing and collapsing bubbles in a tube, J. Micromech. Microeng. 9, 402-413, 1999. [54] M. Hu, T. Lindemann, T. Göttsche, J. Kohnle, R. Zengerle, P. Koltay, Discrete Chemical Release From a Microfluidic Chip, J. Microelectromech. Systems 16 (4), 786-796, 2007. [55] G. Krijnen and N. Tas, Micromechanical Actuators, Review, MESA+ Research Institute, Netherlands, [online 2007]. http://tst.ewi.utwente.nl/education/elbach/mandt/extra/background/mems_actuators.pdf. [56] W.C. Tang, T.-C. H. Nguyen, R.T. Rowe, Laterally driven polysilicon resonant microstructures, Proc. IEEE MEMS'89, Salt Lake City, UT, USA, pp. 53-59, 1989. [57] V.P. Jaecklin, C. Linder, N.F. de Rooij, J.M Moret, Micromechanical comb actuators with low driving voltage, J. Micromech. Microeng. 2, 250-255, 1992. [58] R. Legtenberg, A.W. Groeneveld, M. Elwenspoek, Comb-drive actuators for large displacement, J. Micromech. Microeng. 6, 320-329, 1996. [59] M. Nienhaus, Electromagnetic Microdrives – Principles and Applications, Proc. Actuator'04, Bremen, Germany, 2004. [60] J.W. Judy, R.S. Muller, H.H. Zappe, Magnetic microactuation of polysilicon flexure structures, J. Microelectromech. Systems 4 (4), 162-169, 1995. [61] E. Ozlu, G.J. Cheng, X. T. Wu, Feasibility Study on a Novel Electromagnetic MEMS Actuator with Planar Coils, Proc. Actuator'04, Bremen, Germany, 2004. [62] M. Tabib-Azar, Microactuators : electrical, magnetic, thermal, optical, mechanical, chemical & smart structures, Kluwer Academic, 1998. [63] P. Sommer-Larsen, R. Kornbluh, Polymer Actuators, Proc. Actuator'02, Bremen, Germany, 2002. [64] D.A. LaVan, T. McGuire, R. Langer, Small-scale systems for in vivo drug delivery, Nature 21 (10), 1184-1191, 2003. [65] Durect Corporation, Cupertino, CA, USA. www.durect.com. [66] R. Muller and C. Keck, Challenges and solutions for the delivery of biotech drugs - a review of drug nanocrystal technology and lipid nanoparticles, J. Biotechnology 113, 151-170, 2004. 178 Bibliography [67] K.S. Soppimath, T.M. Aminabhavi, A.R. Kulkarni, W.E. Rudzinski, Biodegradable polymeric nanoparticles as drug delivery devices, J. Controlled Release 70 (1-2), 1-20, 2001. [68] E. Leo, R. Cameroni, F. Forni, Int. J. Pharmaceutics 180, 23-30, 1999. [69] M. Ferrari, Cancer Nanotechnology: Opportunities and Challenges, Nature Reviews Cancer 5 (3), 161-170, 2005. [70] T.A. Desai, D.J. Hansford, L. Kulinsky, A.H. Nashat, G. Rasi, J. Tu, Y. Wang, M. Zhang, M. Ferrari, Nanopore technology for biomedical applications, Biomed. Microdevices 2 (1), 11-40, 1999. [71] J.T. Santini, M.J. Cima, R. Langer, Nature 397, 335-338, 1999. [72] MicroCHIPS Inc., Bedford, MA, USA. www.mchips.com. [73] A.C. Richards Grayson, I.S. Choi, B.M. Tyler, P.P. Wang, H. Brem, M.J. Cima, R. Langer, Multi-pulse drug delivery from a resorbable polymeric microchip device, Nature Materials 2 , 767-772, 2003. [74] A. Trautmann, P. Ruther, O. Paul, Microneedle arrays fabricated using suspended etch mask technology combined with fluidic through wafer vias, Proc. IEEE MEMS03, Kyoto, Japan, pp. 682-685, 2003. [75] N. Roxhed, P. Griss, G. Stemme, Reliable in-vivo penetration and transdermal injection using ultra-sharp hollow microneedles, Proc. IEEE Transducers'05, pp. 213- 216, Seoul, Korea, 2005. [76] E.R. Parker, M.P. Rao, K.L. Turner, N.C. MacDonald, Proc. IEEE MEMS’06, Istanbul, Turkey, pp. 498-501, 2006. [77] J.M. Lippmann, A.P. Pisano, In-Plane, Hollow Microneedles Via Polymer Investment Molding, Proc. IEEE MEMS’06, Istanbul, Turkey, pp. 262-265, 2006. [78] D. Papageorgiou, S.C. Bledsoe, M. Gulari, J.F. Hetke, D.J. Anderson, K.D Wise, A shuttered neural probe with on-chip flowmeters for chronic in vivo drug delivery, J. Microelectromech. Systems 15 (4), 1025 - 1033, 2006. [79] B. Ziaie, A. Baldi, M. Lei, Y. Gu, R.A. Siegel, Hard and soft micromachining for BioMEMS: review of techniques and examples of applications in microfluidics and drug delivery, Adv. Drug Delivery Reviews 56 , 145-172, 2004. [80] Pegasus GmbH, Kiel, Germany. www.pegasus-gmbh.com. [81] Disetronic Medical Systems AG, Burgdorf, Switzerland. www.disetronic.com. [82] RoweMed, Parchim, Germany. www.rowemed.de. [83] Y.-C. Su and L. Lin, J. Microelectromech. Systems 13, 75-82, 2004. [84] A.T. Evans, J.M. Park, G.F. Nellis, S.A. Klein, J.R. Feller, L. Salerno, Y.B. Gianchandani, Proc. IEEE Transducers’07, Lyon, France, pp. 359-362, 2007. 179 Bibliography [85] A.C. Grayson, R.S. Schawgo, A.M. Johnson, N.T. Flynn, Y. Li, M.J. Cima, R. Langer, A BioMEMS Review: MEMS Technology for Physiologically Integrated Devices, Proc. IEEE, 92 (1), 6-21, 2004. [86] A. Heller, Integrated medical feedback systems for drug delivery, AIChE Journal, 51 (4), 1054-1066, 2005. [87] ChipRX, Ohio, USA. www.chiprx.com. [88] J. Kohnle and A. Wolff, Intelligent Drug Delivery from the Oral Cavity, Medical Device Technology 18 (2), 44-45, 2007. [89] E. Truckenbrodt, Fluidmechanik, Band 1, 4. Auflage, Springer Verlag, 1996. [90] S.D. Senturia, Microsystem Design, Kluwer Academic Publishers, 2001. [91] W. Demtröder, Experimentalphysik 1, 3rd ed., Springer Verlag, 2004. [92] N.-T. Nguyen, Mikrofluidik, B.G. Teubner Verlag, Wiesbaden, 2004. [93] J. Ducree, Free online Resources in Microfluidics [online 2008]. www.myFluidix.com. [94] F. Garbassi, M. Morra, E. Occhiello, Polymer Surfaces - From Physics to Technology, John Wiley & Sons, 1998. [95] J.E. McNutt, G.M. Andes, Relationship of the Contact Angle to Interfacial Energies, J. Chemical Physics 30 (5), 1300-1303, 1959. [96] P. Roura, Thermodynamic derivations of the mechanical equilibrium conditions for fluid surfaces: Young's and Laplace's equations, Am. J. Phys 73 (12), 1139-1147, 2005. [97] R.N. Wenzel, Resistance of solid surfaces to wetting by water, Ind. Eng. Chemistry 28 (8), 988-949, 1936. [98] A.B.D. Cassie and S. Baxter, Wettability of porous surfaces, Trans. Faraday Soc. 40, 546-551, 1948. [99] A. W. Adamson, A. P. Gast, Physical Chemistry of Surfaces, 6th ed., John Wiley & Sons, New York, 1997. [100] K. Hermansson, U. Lindberg, B. Hök, G. Palmskog, Wetting properties of silicon surfaces, Proc. Transducers'91, San Francisco, CA, USA, pp. 193-196, 1991. [101] L.C. Salay, A.M. Carmona-Ribeiro, Synthetic Bilayer Wetting on SiO2 Surfaces, J. Phys. Chem. B 102, 4011-4015, 1998. [102] H.-C. Chang, Bubble/Drop Transport in Microchannels, in: Gad-el-Hak (ed.), The MEMS Handbook, CRC Press, 2002. [103] J.W. Judy, Microactuators, in: J.G. Korvink and O. Paul (eds.), MEMS: A Practical Guide to Design, Analysis, and Applications, William Andrew Inc., 2006. [104] F.E.H. Tay, Microfluidics and BioMEMS Applications, Kluwer Academic Publishers, Boston, 2002. [105] H.-R. Tränkler, E. Obermeier, Sensortechnik, Springer Verlag, Berlin, 1998. 180 Bibliography [106] Stelco GmbH, Neumarkt, Germany, Materialien, Technische Daten, Anwendungen [online 2007]. www.stelco.de. [107] S.P. Timoshenko, S. Woinowsky-Krieger, Theory of plates and shells, 2nd ed., McGraw-Hill, 1959. [108] A. Heuberger, Mikromechanik, Springer Verlag, Berlin, 1989. [109] A. Schroth, Modelle für Balken und Platten in der Mikromechanik, Dresden University Press, Dresden, 1996. [110] J.G. Korvink and O. Paul (eds.), MEMS: A Practical Guide to Design, Analysis, and Applications, William Andrew Inc., 2006. [111] P. Woias, Konstruktion II - Lecture Manuscript, IMTEK, University of Freiburg, Germany, 2007. [112] S. Li, S. Chen, Analytical analysis of a circular PZT actuator for valveless micropumps, Sens. Actuators A 104 (2), 151-161, 2003. [113] C.H.J. Fox, X. Chen, S. McWilliam, Analysis of the deflection of a circular plate with an annular piezoelectric actuator, Sens. Actuators A 133, 180-194, 2007. [114] A. Doll, Entwicklung einer Hochleistungsmikropumpe für eine Schließmuskelprothese, Dissertation, IMTEK, University of Freiburg, Germany, 2008. [115] C.J. Morris and F.K. Forster, Optimization of a circular piezoelectric bimorph for a micropump driver, J. Micromech. Microeng. 10, 459–465, 2000. [116] Stelco GmbH, Neumarkt, Germany. www.stelco.de. [117] Huntsman Advanced Materials GmbH, Basel, Switzerland, Araldite© 2020 Datasheet [online 2007]. www.huntsman.com. [118] C. Liu, T. Cui, Z. Zhou, K. Lian, J. Goettert, Fluid-structure coupling analysis and simulation of a micromachined piezo microjet, Sens. Actuators A 114 (2-3), 460-465, 2004. [119] R. Zengerle, M. Richter, Simulation of microfluidic systems, J. Micromech. Microeng. 4, 192 - 204, 1994. [120] W.C. Young and R.G. Budynas, Roark's Formulas for Stress and Strain, 7th ed., McGraw-Hill, 2002. [121] F. Goldschmidtböing, A. Doll, M. Heinrichs, P. Woias, H.J. Schrag, U.T. Hopt, A generic analytical model for micro-diaphragm pumps with active valves, J. Micromech. Microeng. 15 (4), 673-683, 2005. [122] A. Doll, M. Wischke, H.-J. Schrag, A. Geipel, F. Goldschmidtboeing, P. Woias, Characterization of active silicon microvalves with piezoelectric membrane actuators, Microelectronic Engineering, 84 (5-8), 1202-1206, 2007. [123] S. Haeberle, N. Schmitt, R. Zengerle, J. Ducrée, Centrifugo-Magnetic Pump for Gasto-Liquid Sampling, Sens. Actuators A 135 (1), 28-33, 2007. 181 Bibliography [124] COMSOL AB, Stockholm, Sweden, Tutorial: Filling of a Capillary Channel [online 2007]. www.comsol.com. [125] S. Nadir, Optimization of a peristaltic micropump with an improved gas pumping performance, Master thesis, IMTEK, University of Freiburg / University of Rostock, 2006. [126] M.J. Madou, Fundamentals of Microfabrication: The Science of Miniaturization, 2nd ed., CRC Press, 2001. [127] W. Menz, J. Mohr, O. Paul, Microsystem Technology, 2nd ed., Wiley-VCH, 2001. [128] M. Rabold, A. Doll, F. Goldschmidtböing, P. Woias, The modified Blister Test Method: Versatile test strategies for a Non-Destructive Strength-Characterization of Full-Wafer Bonds, J. Electrochemical Society 154 (7), H647-H652, 2007. [129] M. Heinrichs, Flexibles Steuer- und Regelsystem für unterschiedliche Sensor-/AktorAnwendungen, Diploma thesis, IMTEK, University of Freiburg, Germany, 2004. [130] S. Shoji, S. Nakagawa, M. Esashi, Micropump and sample-injector for integrated chemical analyzing systems, Sens. Actuator A 21, 189-192, 1990. [131] K.-P. Kämper, J. Dopper, W. Ehrfeld, S. Oberbeck, A self-filling low-cost membrane micropump, Proc. IEEE MEMS'98, Heidelberg, Germany, pp. 432-437, 1998. [132] R. Bodén, M. Lehto, U. Simu, G. Thornell, K. Hjort, J.-A. Schweitz, A polymeric paraffin actuated high-pressure micropump, Sens. Actuators A 127, 88-93, 2006. [133] M. Wischke, A. Doll, A. Geipel, F. Goldschmidtboeing, P. Woias, Fabrication of MultiLayer Piezo-Actuators and Reliable Assembling Strategy for High Performance Micropumps, Proc. Actuator'06, Bremen, Germany, pp. 276-280, 2006. [134] M. Wischke, Entwicklung von mehrlagigen Niederspannungs-Piezoaktoren zur Leistungsteigerung von Silizium-Mikromembranpumpen, Diploma thesis, IMTEK, University of Freiburg, Germany, 2006. [135] W.Y Sim, H.J. Yoon, O.C. Jeong, S.S. Yang, A phase-change type micropump with aluminum flap valves, J. Micromech. Microeng. 13, 296-294, 2003. [136] E.T. Carlen and C.H. Mastrangelo, Electrothermally activated paraffin microactuator, J. Microelectromech. Systems 11 (3), 165-174, 2002. [137] L. Klintberg, M. Karlsson, L. Stenmark, J.-A. Schweitz, G. Thornell, A large stroke, high force paraffin phase transition actuator, Sens. Actuators A 96, 198-195, 2002. [138] M. Lehto, Paraffin actuators in microfluidic systems, Dissertation, Uppsala University, 2007. [139] B. Yang and Q. Lin, Compliance-based latchable microfluidic actuators using a paraffin wax, Proc. IEEE MEMS'06, Istanbul, Turkey, pp. 738-741, 2006. [140] P. Dubois, E. Vela, S. Koster, D. Briand, H.R. Shea, N.-F. de Rooij, Paraffin-PDMS composite thermo microactuator with large vertical displacement capability, Proc. Actuator'06, Bremen, Germany, pp. 215-218, 2006. 182 Bibliography [141] P. Katus, Paraffin-Aktoren auf der Basis rußgefüllter Wachse, Diploma thesis, IMTEK, University of Freiburg, Germany, 2007. [142] W. Dollhopf, H.P. Grossmann, U. Leute, Some thermodynamic quantities of n-alkanes as a function of chain length, Coll. and Polymer Science 259, 267-278, 1980. [143] R. Schueler, J. Petermann, K. Schulte, H. Wentzel, Agglomeration and electrical percolation behavior of carbon black dispersed in epoxy resin, J. Appl. Polymer Science 63, 1741-1746, 1997. [144] T. Prasse, L. Flandin, K. Schulte, W. Bauhofer, In situ observation of electric field induced agglomeration of carbon black in epoxy resin, Appl. Physics Letters 72 (22), 2903-2905, 1998. [145] F. Bueche, Electrical resistivity of conducting particles in an insulating matrix, J. Polymer Science: Polymer Physics Edition 11, 1319-1330, 1973. [146] S.K. Sia and G.M. Whitesides, Review: Microfludic devices fabricated poly(dimethylsiloxane) for biological studies, Electrophoresis, 24, 3563-3576, 2003. in [147] S.D. Gillmor, B.J. Larson, J.M. Braun, C.E. Mason, L.E. Cruz-Barba, F. Denes, M.G. Lagally, Low-contact-angle polydimethyl siloxane (PDMS) membranes for fabricating micro-bioarrays, Proc. Microtech. Medicine & Biology, Madison, WI, USA, pp. 51-56, 2002. [148] S.J. Lee, J. C.-Y. Chan, K.J. Maung, E. Rezler, N. Sundararajan, Characterization of laterally deformable elastomer membranes for microfludics, J. Micromech. Microeng. 17, 843-851, 2007. [149] N.L. Jeon, D.T. Chiu, C.J. Wargo, H.Wu, I.S. Choi, J.R. Anderson, G.M. Whitesides, Design and fabrication of integrated passive valves and pumps for flexible polymer 3dimensional microfluidic systems, Biomed. Microdevices 4 (2), 117-121, 2002. [150] Y. Xia and G.M.Whitesides, Soft Lithography, Annu. Rev. Mater. Science 28, 153-184, 1998. [151] D. Armani, C. Liu, and N. Aluru, Re-configurable fluid circuits by PDMS elastomer micromaching, Proc. IEEE MEMS'99, Orlando, USA , pp. 222–227, 1999. [152] H. Wu, B. Huang, R. N. Zare, Construction of microfluidic chips polydimethylsiloxane for adhesive bonding, Lab Chip 5, 1393-1398, 2005. using [153] U. Kloter, H. Schmid, H. Wolf, B. Michel, D. Juncker, High-resolution patterning and transfer of thin PDMS films: Fabrication of hybrid self-sealing 3D microfluidic systems, Proc. IEEE MEMS'04, Maastricht, Netherlands, pp. 745-748, 2004. [154] L. Cui and H. Morgan, Design and fabrication of travelling wave dielectrophoresis structures, J. Micromech. Microeng. 10, 72-79, 2000. [155] M.L. Adams, M.L. Johnston, A. Scherer, S.R. Quake, Polydimethylsiloxane based microfluidic diode, J. Micromech. Microeng. 15, 1517-1521, 2005. 183 Bibliography [156] J.M. Engel, J. Chen, D. Bullen, C. Liu, Polyurethane rubber as a MEMS material: Characterization and demonstration of an all-polymer two-axis artificial hair cell flow sensor, Proc. IEEE MEMS'05, Miami, FL, USA, pp. 279-282, 2005. [157] Y. Zhu, J. Hu, L.-Y. Yeung, Y. Liu, F. Ji, K.-W. Yeung, Development of shape memory polyurethane fiber with complete shape recoverability, Smart Mater. Struct. 15, 13851394, 2006. [158] T.S. Hansen, K. West, O. Hassager, N.B. Larsen, An all-polymer micropump based on the conductive polymer poly(3,4-ethylenedioxylthiophene) and a polyurethane channel system, J. Micromech. Microeng. 17, 860-866, 2007. [159] S. Ramirez-Garcia and D. Diamond, Biomimetic, low power pumps based on soft actuators, Sens. Actuators A 135 (1), 229-235, 2006. [160] E.S. Wilks (ed.), Industrial Polymers Handbook, Vol.1, Wiley-VCH, 2000. [161] Lackwerke Peters GmbH, Grundlagen der chemischen Komponenten-Systemen [online 2007]. www.peters.de. Vernetzung von 2- [162] Elastogran GmbH, Lemförde, Germany. www.elastogran.de. [163] Lackwerke Peters GmbH, Kempen, Germany. www.peters.de. [164] Dow Corning Corp., Midland, MI, USA. www.dowcorning.com. [165] C. Dorrer and J. Rühe, Advancing and receding motion of droplets on ultrahydrophobic post surfaces, Langmuir 22, 7652-7657, 2006. [166] D. Öner and T.J. McCarthy, Ultrahydrophobic surfaces: Effects of topography length scales on wettability, Langmuir 16, 7777-7782, 2000. [167] G. Bocci, K.C. Nicolaou, R.S. Kerbel, Protracted Low-Dose Effects on Human Endothelial Cell Proliferation and Survival in Vitro Reveal a Selective Antiangiogenic Window for Various Chemotherapeutic Drugs, Cancer Res 62, 6938-3943, 2002. 184
© Copyright 2025 ExpyDoc