Sensors and Actuators A 96 (2002) 189±195
A large stroke, high force paraf®n phase transition actuator Ê ke Schweitz, Greger Thornell Lena Klintberg*, Mikael Karlsson, Lars Stenmark, Jan-A Department of Materials Science, The AÊngstroÈm Laboratory, Uppsala University, Box 534, SE-751 21 Uppsala, Sweden Received 25 August 2001; received in revised form 12 November 2001; accepted 16 November 2001
Abstract An actuator that uses the volume expansion related to the solid-to-liquid phase transition of paraf®n wax has been fabricated and evaluated. The actuator consists of a ring-shaped paraf®n cavity con®ned by two joint silicon diaphragms with rigid centers. When the paraf®n is melted, the resulting hydrostatic pressure de¯ects the joined rigid centers in one direction only. The magnitude of the de¯ection is primarily a function of the geometrical relation between the two diaphragms, giving the opportunity to tailor the behavior of the actuator in a large range. Conventional IC-processing techniques have been used to fabricate a prototype with a width of 68 mm and a thickness of 825 mm. The prototype attained a maximum de¯ection of ca. 90 mm. Loaded with 3 N it still exhibits a de¯ection of ca. 75 mm. The device can be used as a thermal switch. # 2002 Elsevier Science B.V. All rights reserved. Keywords: Thermal actuator; Paraf®n; Phase transition; Silicon
1. Introduction De¯ecting diaphragms are common in applications such as micropumps and valves [1]. Electrostatic attraction can be used to de¯ect the diaphragm, but more common is to pressurize it. A simple way to do this is to heat gas in a cavity sealed by the diaphragm. A large expansion over a narrow temperature interval can be achieved by ®lling the cavity with liquid and use the expansion associated with the liquid-to-gas transition [2]. At atmospheric pressure the volume expansion of the phase transition is in the order of a factor one thousand. When the device is miniaturized, it is however, not easy to ®ll or seal the cavity. Also, the load capacity of a gas ®lled diaphragm actuator is restricted by the large compressibility of the gas. The compressibility of liquids is, on the other hand, only a fraction of the compressibility of gases, and by using the hydrostatic pressure resulting from the solid-to-liquid phase change, it is possible to have both a large expansion and a large force. With its moderate melting temperature paraf®n is a promising material. It has been used as an active material in a number of ®ne-mechanical actuators to create linear motion. To transmit the expansion of a paraf®n ®lled chamber to a linear motion they all include a shaft. When a squeeze boot seal is not used [3], the paraf®n is either hermetically sealed in a bellow [4] or in a silastic tube [5].
* Corresponding author. Tel.: 46-18-471-3077; fax: 46-18-471-3572. E-mail address:
[email protected] (L. Klintberg).
The potential of this material in microstructure technology has been shown in paraf®n actuated surface micromachined valves fabricated by encapsulating ®lms of evaporated paraf®n by polymeric Parylene diaphragms [6]. Here we present a structure where the volume expansion is utilized for a large stroke de¯ection in essentially one direction. Except for the ®nal ®lling procedure, only conventional IC-processing techniques have been used. 2. Thermal properties of paraffin wax Paraf®n wax consists primarily of straight hydrocabon chains with the composition CnH2n2 and a typical molecular weight of 250±700 g/mol. It typically expands 10±15% when melted [7]. The melting temperature and hence the actuation temperature, is possible to tailor between 100 and 100 8C by mixing paraf®n molecules of different sizes. In addition the temperature change that is required from activation to full stroke can be varied from 2 to 100 8C with different compositions. In this work, a paraf®n wax from Aldrich (#41,166-3) with a melting point of 65 8C has been used. Paraf®n has a heat of fusion of ca. 170 kJ/kg [7], an unusually high value, suggesting a minimized amount of paraf®n to reduce the energy consumption and actuation time of a device. Needless to say, this is in favor of miniaturized components. It is the compressibility of the melted paraf®n that limits the force attainable. Literature data of density as a function of pressure and temperature for C36H74 is found in the work
0924-4247/02/$ ± see front matter # 2002 Elsevier Science B.V. All rights reserved. PII: S 0 9 2 4 - 4 2 4 7 ( 0 1 ) 0 0 7 8 5 - 3
190
L. Klintberg et al. / Sensors and Actuators A 96 (2002) 189±195
counterbalanced already at 16 kPa above atmospheric pressure. With paraf®n as an active material it is possible to make actuators combining large forces and large displacements. 3. The device structure
Fig. 1. The density as a function of temperature (8C) and pressure (MPa) for C36H74. For comparison, nitrogen is also included. Note that the density for the gas has been scaled with a factor of 3000.
of Zolle [8] and is presented in Fig. 1. This paraf®n has a molecular weight close to the one used in this work. For comparison, the density of nitrogen gas, calculated from the law for perfect gases is also included. The melting temperature of C36H74 is about 75 8C at atmospheric pressure and the volume increases by 18% as it is heated from 30 to 80 8C, Fig. 2. Half of this expansion is still retained when a pressure of 60 MPa is applied. If the paraf®n is heated above 140 8C the volume is still more than 10% larger than at room temperature at a pressure of 200 MPa. This should be compared with a perfect gas heated to 80 8C, where the expansion from heating is
Fig. 2. The density of C36H74 as a function of pressure for different temperatures. The paraffin is liquid for all pressures at temperatures of 110.2 8C and above.
We chose to construct the actuator from silicon to make future integration of electronics possible. Microelectronics research and fabrication facilitate high precision machining of this material in this size range, and allow for further miniaturization. The actuator consists of two bonded circular diaphragms sandwiching a ring-shaped paraf®n ®lled cavity, Fig. 3. The diaphragms contain ¯exible membranes of different radii but equal widths, causing the rigid center to de¯ect only in one direction as the cavity is pressurized, Fig. 3 (bottom). The magnitude of the de¯ection is a function of the volume of the cavity and the difference in radii between the two diaphragms. The maximum de¯ection increases with decreasing difference in radius. By just changing the relation between the radii of the diaphragms it is possible to set the maximum de¯ection for a given cavity volume. This
Fig. 3. The actuator consists of two bonded circular diaphragms with rigid centers. The rigid center deflects (bottom), when the ring-shaped cavity is pressurized because of the different radii of the diaphragms.
L. Klintberg et al. / Sensors and Actuators A 96 (2002) 189±195
191
where a is the outer membrane radius Fig. 3, h the membrane thickness, v Poisson's ratio and Ap and Bp are given by 3
1 v2 b4 b2 a 1 Ap 4 2 ln (2) 16 b a4 a and Bp
v2 =
1vb2 =a2
7 v=3
1
b2 =a2
b4 =a4
3
1
v
1
b4 =a4
1
b2 =a2 2 (3)
where b is the inner membrane radius, Fig. 3. The maximum radial bending stress, sr, occurs at the inner and outer boundaries and can be calculated from sr
Fig. 4. The actuator working as a part of a thermal control system for nanosatellites. In normal mode (top), the component is passive, working as a shield for incoming radiation. If the temperature is raised and the paraffin is melted, the actuator makes contact with the IR-radiator (bottom). Heat is now effectively transported to the radiator and emitted into deep space.
mechanism is advantageous when miniaturizing the actuator and maintaining large strokes. Furnished with an IR-radiator, this actuator could be part of the thermal control system in for instance nanosatellites, Fig. 4. In normal mode the component is passive working as a radiation shield. Heat is kept out of the satellite because of the insulating gap between the actuator and the IR-emitter, Fig. 4 (top). But, if the internal temperature of the satellite for some reason is raised over a critical value that can cause overheating, the paraf®n melts and the actuator makes contact with the IR-emitter, Fig. 4 (bottom). This heat is effectively transported by conduction to the radiator which emits it into deep space. Our prototype has been fabricated with this application in mind. A large area for effective heat transfer in contact mode is desirable, hence for the micromechanical technology rather large lateral dimensions. If heat is fed to the actuator by some stiff thermal bridge, strength and high load capacity might be necessary. If heat is generated directly at the backside of the actuator, for instance by some power electronics, the strength is of minor interest.
3p 2
a 4h2
b2 :
The tangential stress is in this case less than the radial stress and does not need to be considered. The dimensions of the structure have been chosen to maximize the contact area between the actuator and an IR-emitter. To ensure that the maximum de¯ection does not cause stresses exceeding the yield stress of 7 GPa [1] a safety factor of about 10 was used.
Fig. 5. Actuator dimensions.
4. Design Following Giovanni [9], the connection between the pressure, p, and the de¯ection, y, for a pressurized, ¯at diaphragm with a rigid center is given by p
Eh3 Eh y B p 4 y3 ; a A p a4
(1)
(4)
Fig. 6. Fabrication process in cross section.
192
L. Klintberg et al. / Sensors and Actuators A 96 (2002) 189±195
Fig. 7. The finished actuator cut out by a dicing saw. The actuator is 68 mm 68 mm and 825 mm thick.
5. Fabrication Two standard (1 0 0) silicon wafers, 300 and 525 mm thick respectively, were patterned using standard photolithography methods, leaving a 1.9 mm thick silicon dioxide as a mask. In the ®rst step the backside of wafer 1 and both sides of wafer 2 were patterned, Fig. 5. The wafers were etched using deep reactive ion etching (DRIE) according to the standard Bosch process (Plasma-Therm SLR Series), which makes etching with high aspect ratio independent of the crystallographic orientation of the wafer possible. Wafer 1 was stripped and reoxidized before the upper side was patterned for inlet/outlet holes. Since the DRIE process does not readily permit etching of vias, holes were wet etched in potassium hydroxide (40 g KOH/100 ml water, 80 8C). The wafers were ®nally joined by silicon fusion bonding (annealing for 3 h at 1050 8C). The fabrication scheme is shown in Fig. 6. Prior to ®lling the device with liquid paraf®n using a syringe equipped with a silicone socket, its rigid center was
manually de¯ected about 100 mm. When the internal air had been replaced by paraf®n as veri®ed by some over®lling, the outlet was sealed temporarily with a silicone plug. This step took place at an ambient temperature of 80 8C. After ®lling, the actuator was cooled and as the paraf®n started to solidify, the forced de¯ection was released. The syringe and the plug remained pressed to the inlet during cooling to prevent air to leak into the actuator and was not removed until the structure had reached room temperature. Finally, the inlet/outlet holes were sealed by gluing cover glasses over them. A total number of six actuators were fabricated in this way. One of them is shown in Fig. 7. 6. Evaluation An optical pro®lometer (WYKO NT-2000) with a ®eld of view of 5 mm 4 mm was used to measure the de¯ection of the actuator by relating the rim of the central part to the rigid frame outside the membrane at three locations separated by
Fig. 8. IR-light revealed small air bubbles gathered preferentially along the inner rim of the cavity.
L. Klintberg et al. / Sensors and Actuators A 96 (2002) 189±195
1208. The overall de¯ection pro®le of the actuator was investigated by a surface pro®lometer (Dektak 200-Si). The temperature of the structure was measured with a Pt-100 element located above the outer boundary (r 30 mm) of the cavity. In these experiments the actuator was placed in a ®xture and heated from below. A layer of heat conductive paste was used to ensure good contact between the actuator and a heated copper cylinder with a diameter corresponding to the rigid part of the smaller diaphragm. The copper cylinder was heated by a resistive heater and the ¯exible heat paste ensured that the thermal expansion of the copper cylinder did not in¯uence the measurements of the actuator de¯ection signi®cantly. The load carrying capacity was investigated by applying weights to the rigid center-part of the actuator and monitoring the resulting decrease in de¯ection in the optical pro®lometer. To avoid the possibility that the loaded actuator would be supported by the copper cylinder, it was replaced here by a weak, resistively heated Kanthal wire that in itself had insigni®cant load bearing capacity. All structures were also investigated by an IR-camera to reveal trapped air in the cavities. 7. Results The IR-camera revealed some trapped air, gathered preferentially along the inner rim of the cavity as shown in Fig. 8. When the structure was heated, the small air bubbles gathered and formed one or two larger bubbles. The de¯ection as a function of temperature for all six actuators measured by the optical pro®lometer is shown in
Fig. 9. The deflection measured at the rim of the rigid center part of the actuator as a function of temperature for six different actuators.
193
Fig. 10. The profile of the actuator at room temperature 55, 60 and 74 8C.
Fig. 9, where the values presented are the de¯ections at the three measurement locations averaged. The de¯ection above the rim was approximately 90 mm when the temperature registered by the Pt(1 0 0) element was 80 8C. The typical shape of an actuator during activation is shown in Fig. 10. For a registered temperature of about 75 8C the mid point of the actuator de¯ected about 55 mm, whereas the rim of the central part de¯ected about 90 mm. It should be mentioned that the de¯ection at room temperature, directly after ®lling, differed from that achieved after the ®rst melting±solidifying cycle, which is the room
Fig. 11. The deflection vs. applied load.
194
L. Klintberg et al. / Sensors and Actuators A 96 (2002) 189±195
temperature curve presented in the ®gure. After the ®rst heating cycle, the actuator always returned to the same de¯ection at room temperature. The de¯ection as a function of load was determined for one structure at a temperature of 70 8C, Fig. 11. An extended test showed that the de¯ected actuator carried a load of up to 13 N before a crack originating from one of the inlet/outlet holes arose. 8. Discussion Since the paraf®n cavity has a width of 1.5 cm and the actuator was heated at the center in order to mimic the situation where it acts as a thermally activated switch, the paraf®n gradually melts as the heat spreads outwards. This explains the discrepancy between the measured actuation temperature as seen in Fig. 9 and the actual melting temperature of 65 8C. For general actuation purposes, resistive heaters at the walls of the cavity would be preferable. The actuators exhibited a small de¯ection at room temperature as a result of ®lling with too large an amount of paraf®n because of the dif®culty to manually de¯ect the actuator to the right extent. In order to have zero de¯ection at room temperature, the amount of paraf®n in the structure must be controlled more precisely. The inclination of the diaphragm at room temperature as visible in Fig. 10 is probably a result of non-uniform cooling conditions over the large diaphragm. No inclination of the rigid center was observed when the actuator was active. Non-uniform cooling is probably also responsible for an initial de¯ection at room temperature differing from the de¯ection after a single heating cycle. In the ®lling procedure the actuator was cooled very rapidly not giving the actuator enough time to attain its equilibrium shape. The center of the diaphragm was supposed to be rigid, but as Fig. 10 reveals, the rim, protruding over the cavity, is too weak and consequently prone to de¯ection. Apart from the fact that this decreases the de¯ection of the membrane by providing extra room for the expanding paraf®n, this is an unwanted behavior in the application as a switch in the thermal control system described previously, since the contact area to the IR-emitter would become smaller. Consequently, some changes in the design must be made for optimal performance regardless of application. The air bubbles in the paraf®n cavity are unwanted since they reduce the load carrying capacity of the actuator and being dif®cult to control, give a performance spread. However, calculations based the law for perfect gases and Eq. (1) indicate that the air has negligible in¯uence on the de¯ection at the investigated temperatures when the actuator is not heavily loaded. As mentioned previously, the response time of a thermal actuator decreases with size. The micromachined paraf®n actuators in [10] have a response time of 3±5 ms, which should be compared with a typical response time of hundreds
of seconds for the linear shaft actuators also described. In our work the response time depended entirely on the heating method. With the Kanthal wire, full de¯ection of the actuator could be achieved in about 15 s, whereas in the case with the copper cylinder, it took almost 30 min to fully activate the actuator from room temperature. If the actuator is to be further miniaturized, the relationships between the size of the diaphragm and the maximum de¯ection, and the stress and the pressure as given by Eqs. (1) and (4), respectively have to be considered. If the actuator is scaled isometrically, the maximum de¯ection decreases approximately linearly with decreasing diaphragm radius, Fig. 12 (a). Decreasing just the lateral dimensions, i.e. keeping the membrane thickness constant, is no problem as long as the radius is larger than 4 mm, as shown in Fig. 12 (b). Whereas a uniformly scaled diaphragm with a radius of 1 mm and a 2 mm thick membrane would have a tolerable de¯ection of about 7 mm, a diaphragm of the same lateral size with a 50 mm thick membrane would allow for a de¯ection of only about 2 mm.
Fig. 12. The deflection corresponding to a radial bending stress equal to half the yield stress vs. radius of a rigid center diaphragm when the structure is isometrically miniaturized, as well as just laterally diminished. In (b) the part of (a), where a fixed membrane thickness becomes a disadvantage is magnified.
L. Klintberg et al. / Sensors and Actuators A 96 (2002) 189±195
We believe that the problem with initial de¯ection and trapped air will be solved by developing a system where the structure is evacuated and the paraf®n degassed before ®lling the structure with a precise amount of paraf®n. This has to be con®rmed in the future. Acknowledgements The Swedish Foundation for Strategic Research, SSF, is greatly acknowledged for ®nancing this project, so is Ê ke Gustafsson for valuable help with the heating Mr. Jan-A equipment and Dr. Ram Gupta for helpful discussions on solid state mechanics. References [1] G. Kovacs, Micromachined Transducers Sourcebook, McGraw-Hill, New York, 1998. [2] M. Zdeblick, R. Anderson, J. Jankowski, B. Kline-Schoder, L. Christel, R. Miles, W. Weber, Thermopneumatically actuated microvalves and integrated electro-fluidic circuits, Technical digest, IEEE Solid-State Sensor and Actuator Workshop, 1994, pp. 251± 255. [3] D. Maus, S. Tibbitts, Heat-activated drug delivery system and thermal actuators therefor, US Patent 5-738-658 (1998). [4] D. Dowen, Design and implementation of a paraffin based micropositioning actuator, SPIE 3132 (1997) 127±134. [5] N. Kabei, M. Kosuda, H. Kagamibuchi, R. Tashiro, H. Mizuno, Y. Ueda, K. Tsuchiya, A thermal-expansion-type microactutor with paraffin as the expansive material, JSME Int. J. 40 (4) 1997. [6] E.T. Carlen, C.H. Mastrangelo, Paraffin actuated surface micromachined valves, in: Proceedings of IEEE Thirteenth Annual International Conference on Micro Electro Mechanical Systems, IEEE, Piscataway, NJ, USA, 2000, pp. 381±385. [7] M. Freund, R. CsikoÂs, S. Keszthelyi, G.Y. MoÂzes, Paraffin Products Properties, Technologies, Applications, Elsevier, Amsterdam, 1982. [8] P. Zolle, Standard Pressure±Volume±Temperature Data for Polymers, Technomic, Lancaster, 1995. [9] M. Giovanni, Flat And Corrugated Diaphragm Design Handbook, Marcel Dekker, New York, 1982.
195
[10] E.T. Carlen, C.H. Mastrangelo, Simple, high actuation power, thermally activated paraffin microactuator, in: Proceedings of Transducers'99, Sendai, Japan, 7±10 June 1999.
Biographies Lena Klintberg was born in 1972 and graduated in 1997 with a MSc in engineering physics at Uppsala University, Sweden. She belongs to the Microstructure Technology Group at the Department of Materials Science, Uppsala University, as a PhD student. Her research interests include microprocessing and actuators. Mikael Karlsson was born in 1974 in Kristinehamn, Sweden and earned his MSc degree in materials science at Uppsala University in 2001. He Ê ngstroÈm Space Technology Center, at the same recently joined The A university, as a research engineer. He is presently working with paraffin as the actuator material in micromechanical thermal switches and micro propulsion cold gas thruster systems. Lars Stenmark was born in 1944 and received his MSc in 1969 from Stockholm University, Sweden. He worked as a development engineer and later as a project manager at the Swedish Space Corporation until 1979 when he started a private company Micro Nova KB to develop new electrooptical instruments. In 1982, he started ACR Electronic AB, a hardware developer and supplier entirely dedicated to space applications. Lars Stenmark has designed, tested and flown numerous experiments and subsystems. In 1998, he left all conventional space technology to promote the introduction of microsystems technology in space. He is the founder Ê ngstroÈm Space Technology Center at The A Ê ngstroÈm and director of The A Laboratory, Uppsala University. È rnskoÈldsvik, Sweden, 1945. BSc at Jan-AÊke Schweitz was born in O Gothenburg University in 1967, MSc in 1967 and DrSc in 1975 at Uppsala University. Presently Professor of Materials Science, Head of the Microstructure Technology (MST) Group and Head of the Strategic Center for Advanced Microengineering (AME) at Uppsala University. Author or coauthor of some 100 international publications; about half of them in the MST field. Greger Thornell was born in 1969 in LidingoÈ, Sweden. He joined the Microstructure Technology Group at Uppsala University in 1994, where he conducted his PhD studies fairly focused on non-conventional processing. Ê ngstroÈm Since 1999, he has worked as an Assistant Professor at The A Laboratory where he gradually became more and more interested in actuators and resonators.
