Investigation of Thermally-Actuated Pumping During Pool Boiling of a Dielectric Liquid on an Asymmetric Microstructured Silicon Heat Sink by Naveenan Thiagarajan A dissertation submitted to the Graduate Faculty of Auburn University in partial ful llment of the requirements for the Degree of Doctor of Philosophy Auburn, Alabama December 14, 2013 Keywords: reentrant cavities, pool boiling, FC72, microgravity, anemometry Copyright 2013 by Naveenan Thiagarajan Approved by Sushil H. Bhavnani, Chair, Professor of Mechanical Engineering Roy W. Knight, Assistant Professor of Mechanical Engineering Jeyhoon Khodadadi, Professor of Mechanical Engineering Daniel Mackowski, Associate Professor of Mechanical Engineering Narendra Govil, Professor of Mathematics and Statistics Abstract Developments in the eld of electronics fabrication have led to signi cant miniaturization of devices. Such a reduction in foot-print of the electronic devices coupled with increased capabilities, such as computing power, have o ered numerous advantages to human comfort in the form of connectivity and portability. But these developments have also resulted in a bottle-neck which is the demand to dissipate the resulting high heat densities in the electronic packages. With heat dissipation demands exceeding more than 1000 W/cm2, traditional air cooling techniques and even evaporative liquid cooling techniques like heat pipes have been pushed to their limits of operation. Future high powered, micro-electronics, thermal management could potentially migrate to liquid cooling involving phase change which is considered as one of the potential solutions for high heat dissipation demands. While a technique such as ow in microchannels with phase-change has proven to meet the demands, it comes with the cost of power required to pump the uid through narrow passages. With modern electronics getting leaner in power consumption, an ideal cooling technique would be one that while dissipating a large volume of heat, also self-propels the uid to enhance the heat transfer characteristics. Such a cooling system with a pump-less ow loop will be power-free, compact and self-regulating. The system proposed will also be applicable to thermal management of space electronics, where power is a precious commodity. The study conducted by the author in collaboration with a heat transfer research group from Oregon State University, investigates the liquid self-propulsion e ects during pool boiling of a dielectric liquid on asymmetric surfaces. The study describes a novel silicon heat sink with an asymmetric saw-tooth cross- sectioned surface structure, which has the potential to be translated into a liquid propulsion system while dissipating heat e ciently. The heat sink was fabricated using a combination of ii gray-scale lithography, deep reactive ion etching and wet etching techniques. The novelty of the heat sink lies in the ability to e ect lateral motion of bubbles due to nucleation from re- entrant cavities fabricated on the shallow slope of the saw-toothed surface. The asymmetric nucleation, growth and departure of bubbles leads to an angular momentum imparted to the liquid, thereby resulting in a net lateral ow. The study investigates the ability of surface structure to propel the liquid in its immediate vicinity under a variety of test conditions. The tested conditions include heat ux in the range of 0-4 W/cm2, liquid subcooling ranging from 0-20 C, and gravity ranging from 0-1.8g. Due to the unique pro le of the surface, the bubble characteristics are very di erent from those reported in the literature for common surfaces and uids. One of the primary objectives of the study is the characterization of bubble dynamics from such a surface. In the experiments conducted, bubble growth and departure from re-entrant cavities on the asymmetric structures were studied using high speed photography and image processing techniques. The asymmetry in shape of the ratchet and location of re-entrant cavities re- sulted in nucleation only from the shallow slope of the ratchets. Interestingly, the bubbles were \light-bulb" shaped which otherwise would be more circular for a highly wetting uid such as FC-72. It was observed that the bubble growth and departure were normal to the shallow slope of the ratchet surface structure. Bubble dynamics such as growth rate, bubble departure diameter and frequency were studied as a function of heat ux and subcooling. Asymptotic growth relationships were expressed as D = At and D = tm for the inertia and heat transfer controlled regimes respectively. The values of A, varying between 48-181, increased with increasing heat ux and decreasing subcooling. Similarly, in the heat transfer controlled growth regime, the value of , termed as the growth constant, was observed to increase between 0.25-0.3 with increasing heat ux and decreasing subcooling. Subcooling or heat ux did not a ect the value of m signi cantly which varied between 0.20 - 0.25, compared to the value of 0.5 that has been widely reported in the literature. The bubble de- parture frequency was estimated to be increasing between 0 - 60 with increasing heat ux iii and decreasing subcooling. However, under the tested conditions bubble departure diameter was not found to be a ected signi cantly with heat ux or diameter. A similar saw-toothed surface was tested at microgravity to understand the e ects of gravity on bubble dynamics and self-propelled bubble motion across the surface. Pool boiling experiments were conducted aboard NASA?s reduced gravity ight. In FC-72, vapor bubbles six times larger in diameter compared to 1g were observed, due to lack of buoyancy. Interestingly, bubbles were observed to be sliding across the asymmetric surface at velocities as high as 27.4 mm/s. This motion was observed at all tested conditions. Pool boiling on a plain surface would result in stagnant, surface residing vapor bubbles that would a ect the heat transfer characteristics adversely causing burn-out of the chip being cooled. The ability to move the bubble along the surface at high velocities prevented any heat transfer deterioration, and actually leads to an enhancement. The sliding motion was attributed to pressure di erences in the thin liquid lm existing between the saw-toothed surface and the vapor bubble. A model has been proposed based for the sliding velocity of bubbles, which proves that the force due to pressure di erences in the liquid lm is a potential driving force for the bubbles among other forces. iv Acknowledgments The author would like to extend his sincere thanks to his advisory committee, chaired by Prof. Sushil H.Bhavnani, and comprising Prof. Jeyhoon Khodadadi, Prof. Roy W. Knight, Prof. Daniel Mackowski and Prof. Narendra Govil for their contributions and time. Special thanks are extended to Prof. Roy Hart eld for serving as the external reader. The author gratefully acknowledges the support of sponsors, NSF and NASA, for the opportunity to work in this eld of research. Words cannot express the joy and pride that this research has brought along, especially, in the form of microgravity testing. The author would like to sincerely thank his adviser and committee chair Prof. Sushil H. Bhavnani for being a constant source of inspiration and motivation through a lot of di cult times, and for all the opportunities provided. Sincere thanks and appreciation are extended to Prof. Vinod Narayanan and his graduate students, Florian Kapsenberg and Logan Strid, for the guidance and ideas provided. The author is also thankful to AMSTC, especially Mr. Charles Ellis, for providing the facilities and knowledge for the fabrication of test devices. The author would like to acknowledge the guidance and assistance provided by Dr. Shakib Morshed, and Dr. Tamara Isaacs-Smith in fabricating the test sections. Thanks are extended to Mr. Travis Wheeler for his precious assistance during microgravity testing. Much gratitude is extended to my colleagues at the Heat Transfer Research laboratory for their help, support, and company. Above all, the author would like to thank his parents and sister for their endless love, support, encouragement and sacri ces. v Table of Contents Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xix List of Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xx 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.1 Thermally actuated pumping systems: Literature review . . . . . . . . . . . 7 1.1.1 Self-propulsion in single phase liquids . . . . . . . . . . . . . . . . . . 7 1.1.2 Self-propelled ow in phase change systems . . . . . . . . . . . . . . . 9 1.1.3 Fabrication of re-entrant cavities and sloped surfaces: Background . . 14 1.1.4 Bubble dynamics in pool boiling on plain and enhanced surfaces . . . 19 1.1.5 Bubble dynamics under reduced gravity . . . . . . . . . . . . . . . . 23 2 Fabrication of Silicon Test Devices . . . . . . . . . . . . . . . . . . . . . . . . . 32 2.1 Fabrication of silicon ratchet array . . . . . . . . . . . . . . . . . . . . . . . 35 2.2 Fabrication of re-entrant cavities . . . . . . . . . . . . . . . . . . . . . . . . . 39 2.3 Fabrication of heater . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 2.4 Bonding of saw-tooth ratchet wafer to the heater wafer . . . . . . . . . . . . 46 3 Terrestrial Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 3.1 Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 3.2 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 3.2.1 Bubble Dynamics using High Speed Photography . . . . . . . . . . . 55 3.2.2 Bubble growth - Experiments . . . . . . . . . . . . . . . . . . . . . . 56 3.2.3 Bubble departure frequency and diameter . . . . . . . . . . . . . . . 67 vi 3.2.4 Bubble growth - Comparison of experimental data with models . . . 70 3.2.5 Heat Dissipation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74 3.2.6 Lateral Liquid Velocity Measurements Using Hot Wire Anemometer: Parametric E ects . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76 3.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82 4 Microgravity Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84 4.1 Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84 4.2 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87 4.2.1 Bubble Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88 4.2.2 Bubble Transit Hypothesis . . . . . . . . . . . . . . . . . . . . . . . . 96 4.2.3 Heat Dissipation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104 5 Conclusions and Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . 106 5.1 Bubble dynamics: Terrestrial experiments . . . . . . . . . . . . . . . . . . . 106 5.2 Net lateral liquid velocity under terrestrial gravity . . . . . . . . . . . . . . . 108 5.3 Bubble dynamics: Microgravity . . . . . . . . . . . . . . . . . . . . . . . . . 108 5.4 Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109 Appendices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118 A Thermophysical Properties of FC-72 and Water . . . . . . . . . . . . . . . . . . 119 B Preliminary Closed Loop Experiments Under Terrestrial Gravity Using FC-72 . 120 C Calibration of Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124 C.1 Calibration of thermistor and thermocouple . . . . . . . . . . . . . . . . . . 124 C.2 Calibration of test section heater . . . . . . . . . . . . . . . . . . . . . . . . 125 C.3 Calibration of hot wire anemometer . . . . . . . . . . . . . . . . . . . . . . . 126 D Emperical Model for Lateral Liquid Velocity in Nucleate Boiling . . . . . . . . . 128 E Particle Image Velocimetry (PIV) . . . . . . . . . . . . . . . . . . . . . . . . . . 131 F Microgravity Experiments: Test Equipment Data Package (TEDP) . . . . . . . 137 G Microgravity Experiments: Sample Calculations for the Bubble Velocity Model . 186 vii H Design Draft Files for Boiling Chamber Parts and Other Experimental Fixtures 188 viii List of Figures 1.1 Evolution of supercomputers since its introduction illustrating the increase in number of Floating point operations per second (Flops) and the reduction in cost per operation. The corresponding evolution of cooling systems used in these supercomputers from a simple liquid cooling system such as in car radiators to advanced phase change cooling systems is illustrated. A popular current laptop is used for comparison which is much faster than the supercomputers in 1985. . 2 1.2 Chart illustrates the adaptation of advanced cooling techniques for high heat density applications compared to the simple, lower heat transfer air cooling tech- niques for low power, low heat density applications . . . . . . . . . . . . . . . . 3 1.3 Schematic diagram of the concept under 1g showing the cross-section of the saw- toothed structure. Bubbles grow and depart from the re-entrant cavities at an angle that is normal to the surface. This is hypothesized to result in a net lateral velocity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 1.4 Schematic diagram of the concept under 0g showing the larger size of the bubble due to lack of buoyancy. Bubble slides from left to right due to pressure di erence in the liquid lm underneath, between the crest and trough, that arises due to di erences in radii of curvature (r) . . . . . . . . . . . . . . . . . . . . . . . . . 6 1.5 (a) Image illustrates the local convective ow at the liquid surface and the result- ing convective cells (b) The surface tension gradients due to alternating di erence in temperatures of warmer and cold liquids cause de ection of the free surface. Image was acquired from Carey [6]. . . . . . . . . . . . . . . . . . . . . . . . . . 8 ix 1.6 (a) Oval ow loop with saw-toothed heated ratchet sections used by Jo [1] (b) Asymmetric temperature gradients induced by the shape of the ratchet pro le, which leads to surface tension gradients leading to liquid movement. The direc- tion of liquid movement is marked in (a) . . . . . . . . . . . . . . . . . . . . . . 10 1.7 Flow pattern in oil observed by Stroock et al. [3] over asymmetric mini-ratchets at velcities up to 2mm/min. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 1.8 Surface tension induced ow towards the trough of a symmetrical triangular groove which is marked by the movement of glass particles. Alexeev et al. [2] reported that this results in liquid ow towards the wall at the trough and away from the wall at the crest, thus setting up a vortex ow. . . . . . . . . . . . . . 10 1.9 (a) Leidenfrost droplet motion over a saw-toothed surface demonstrated by Linke et al. [8] (b) Droplet propelled by pressure gradients in the thin liquid layer induced by the di erences in radii of curvature of liquid-vapor interface (c) The direction of liquid movement due to the pressure gradient is shown. R-134a droplet velocities of up to 5 cm/s were reported . . . . . . . . . . . . . . . . . . 13 1.10 Bubble based micropumps that rely on asymmetry in surface structure, surface wetting characteristics and heater locations. (i) Meng and Kim [10] demonstrated preferential gas bubble motion due to surface asymmetry in the form of a neck in the microchannel structure as seen in (a). The gas bubble translation was promoted due to a hydrophilic-hydrophobic junction causing bubble translation to the right as seen in (b). Liquid ow rates of up to 65 nL/s were reported. (ii) Jun and Kim [11] demonstrated bubble motion due to surface tension gradients induced by sequential powering of the heaters. Liquid (isopropanol) velocities of up to 160 m/s were observed. . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 x 1.11 (a) Pyramidal re-entrant cavities used in immersion cooled surfaces studied by Nimkar et al. [14]. Such re-entrant cavities are formed by anisotropic potassium hydroxide etch (b) Bulb shaped re-entrant cavity used in immersion cooled heat sinks by Baldwin et al. [13] and Bhavnani et al. [17] . . . . . . . . . . . . . . . 16 1.12 (a) Arti cial cavities for nucleation formed by punch marks that were used in microchannel heat sinks by Kandlikar et al. [15] (b) Inter-connected re-entrant cavities to form slots that were used in silicon microchannel heat sinks by Ko sar et al. [16] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 1.13 Steps involved in deep reactive ion etching (DRIE) (a) Formation of a photomask using photolithography (b) Etching of silicon using a plasma containing SF6 (c) Passivation cycle involving the deposition of a polymer material which protects the side walls from being etched in the subsequent etching cycles (d) Etching cycle is repeated. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 1.14 Steps involved in gray-scale lithography as illustrated in Waits et al. [23] (a) Design of gray-scale mask (b) Exposure and development of photoresist to obtain a gradient height in photoresist (c) DRIE to obtain a gradient height (slope) in silicon. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 1.15 Types of bubble growth models (a) Uniform superheat theory, involving spherical vapor growth in an uniformly superheated liquid (b) Bubble growth in non- uniform temperature eld; temperature highest at the wall[25]. . . . . . . . . . 21 1.16 Regimes of bubble growth illustrated as shown in Carey [6] (a) Ambient cooler liquid is brought in to contact with the heater surface (b) Growth of boundary layer (c) Bubble nucleation and inertia controlled regime (d) Relaxation micro- layer thickness decreases (d) Bubble departure. . . . . . . . . . . . . . . . . . . 22 xi 1.17 Bubble growth process from a cylindrical cavity on a silicon surface using FC-72 - Hutter et al. [34] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 1.18 Bubble growth rate from a cylindrical cavity on a silicon surface using FC-72 - Hutter et al. [34]. The asymptotic growth relationship is marked as the equation of t in the gure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 1.19 (a) Parabolic trajectory of a C-9B ight and the achieved reduced gravity periods (b) G-jitters in a parabolic ight due to disturbances such as air turbulence - Straub [37]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 1.20 Large bubble of diameter 2.5 mm in pool boiling of FC-72 under reduced gravity at a wall superheat of 30 C - Kim and Benton[38]. . . . . . . . . . . . . 28 1.21 (a) Lower heat transfer observed under microgravity in pool boiling of FC-72 at higher wall superheat. Critical heat ux (CHF) also reduced compared to earth gravity - Kim and Benton[38] (b) Heat transfer increases with increasing heater sizes and liquid subcooling - Henry and Kim [41] . . . . . . . . . . . . . . . . . 29 1.22 Large hovering bubble in R-113; smaller adjacent bubbles near the larger bubble were noticed to migrate towards the larger bubble at high subcooling - Lee and Merte Jr. [43]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 2.1 Illustration of steps involved in fabrication of the silicon test devices to be used in the large and small array experiments . . . . . . . . . . . . . . . . . . . . . . 34 2.2 Steps involved in the fabrication of silicon saw-tooth ratchets . . . . . . . . . . 36 2.3 Illustration of DRIE parameters discussed in Table 2.1 . . . . . . . . . . . . . . 37 2.4 SEM image of the ratchets with an angle 19 . The inset shows an SEM image of silicon nano-structures, known as \silicon grass", formed during the DRIE process with Recipe A. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 xii 2.5 Cross-section of a ratchet array with a nominal saw-tooth angle of 24 . . . . . 38 2.6 Illustration of fabrication steps involved in etching re-entrant cavities from the back side of saw-teeth to form trapezoidal cavity mouths only on the shallow slope 41 2.7 Isometric and back side view illustrations of small and large array cavities on the shallow slope of saw-toothed surface . . . . . . . . . . . . . . . . . . . . . . . . 42 2.8 SEM image showing the top view of saw teeth and cavity mouth measurements in the test device with a saw-tooth angle of 24 . . . . . . . . . . . . . . . . . . 42 2.9 Sketch of aluminum serpentine heaters used in large and small array experiments that were fabricated on silicon. The width and thickness of heater traces are 500 m and 2 m respectively. The electrical contact pad con guration is labeled in the gure. The arrangement of electrical contact pads for the small array heater is based on the arrangement of electrical contacts in the wafer xture. . . . . . 44 2.10 Fabrication of aluminum serpentine heaters used in large and small array test devices. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 2.11 Fabrication of isolation trenches on the back side of heater footprint and redution of wafer diameter to 3 inches. Silicon nitride was coated to eliminate re ection of radiation during infrared measurements made during the small array experiments at OSU. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46 2.12 Images of fabricated aluminum serpentine heaters used in large and small array experiments. The diameter of heater wafer for the large and small array test experiments are 4 inches and 3 inches respectively. . . . . . . . . . . . . . . . . 47 2.13 Illustration of fusion bonding process used to bond the large array of silicon saw-teeth to the heater wafer. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 xiii 2.14 Illustration of gold eutectic bonding process used for bonding the small array saw-tooth ratchets to the 3 inch heater wafer . . . . . . . . . . . . . . . . . . . 50 2.15 X-ray image showing the bond quality of gold eutectic bonding process . . . . . 50 3.1 Silicon test device . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 3.2 Boiling chamber assembly . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 3.3 Image showing bubble departure from re-entrant cavities at an angle normal to the sloped surface of saw-teeth during saturated pool boiling at a heat ux of q" = 2.20 W=cm2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56 3.4 Bubble growth and departure from re-entrant cavities at q" = 2:0W=cm2 and saturated pool conditions. Bubbles nucleating from neighbouring cavities are marked with a di erent color. Relative time stamps are marked at the bottom right of each image. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57 3.5 Ebullition cycle of a bubble at q" = 1.6 W=cm2 and 21 C subcooling . . . . . . 59 3.6 Image processing steps of a single bubble frame at 178.44 ms in Fig. 3.5 . . . . 60 3.7 Estimated bubble growth for images shown in Fig. 3.5 at q"=1.6 W=cm2 and 21 C subcooling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61 3.8 E ect of heat ux on bubble growth at a liquid subcooling of 21 C . . . . . . . 64 3.9 E ect of subcooling on bubble growth at a nominal heat ux of 1.0 W=cm2 . . . 64 3.10 Growth constant from experiments at a liquid subcooling of 0-20 C. The gure shows a linear curve t used to approximately express the growth constant. The equations 3.3 and 3.5 used for comparison from the studies Fritz and Ende [52] and Cole and Shulman [29] respectively, were calculated for saturated pool conditions. 66 xiv 3.11 E ect of heat ux on bubble growth at a liquid subcooling of 20 C . . . . . . . 68 3.12 E ect of subcooling on bubble growth at a nominal heat ux of 1.0 W=cm2 . . 68 3.13 E ect of subcooling and heat ux on bubble departure frequency . . . . . . . . 69 3.14 E ect of subcooling and heat ux on bubble departure diameter . . . . . . . . . 69 3.15 Comparison with existing models for bubble growth . . . . . . . . . . . . . . . . 73 3.16 E ect of subcooling on pool boiling curve for the test device with a saw-tooth angle of 24 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76 3.17 (a) and (b) TSI constant temperature anemometer and power supply used for liquid velocity measurements. Electrical circuit and connections are shown in (c) and (d) [55] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 3.18 Illustration of the probe set up over the test device with a saw-teeth angle of 24 79 3.19 Raw and temperature corrected voltage measurements using hot wire anemome- ter over the test device with a saw-tooth angle of 24 . . . . . . . . . . . . . . . 81 3.20 Temperature corrected liquid velocity measurements using hot wire anemometer over the test device with a saw-tooth angle of 24 . . . . . . . . . . . . . . . . 82 4.1 Zero gravity ight used for parabolic maneuvers - (a) A NASA Zer0-g ight on it upward ascent. Boeing 727 was used for the ight experiments (b)-(c) Gravity pro le and recorded accelerometer readings of the achieved parabolic maneuvers 85 xv 4.2 Experimental setup for reduced gravity experiments - (a) 31 test device with two re-entrant cavities per saw-tooth (b) Assembled test device with the polycar- bonate ow channel. Co-ordinate system shows the orientation of the test device with respect to the aircraft. (c) Boiling chamber used for the reduced gravity experiment. Bellows assembly with double containment is used to maintain a constant chamber pressure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88 4.3 Assembled view of the experimental structure used for microgravity experiments. Details of the individual components in the structure and the related structural analysis are provided in Appendix F. . . . . . . . . . . . . . . . . . . . . . . . . 89 4.4 Bubble dynamics in water and FC-72 under g (a) Vapor bubble 10 times larger than at 1g was observed at microgravity. Image inset shows vapor nucleation from re-entrant cavities, under the foot print of the larger bubble, illustrating the presence of a liquid lm. (b) Sliding motion of vapor bubbles at a velocity of 10 mm/s, across the saw-teeth (left to right in images) at q00 = 0.5 W/cm2. Sliding motion was observed at all tested conditions. (c) Sliding motion of vapor bubbles (left to right in the images) at q00 = 1.4 W/cm2. . . . . . . . . . . . . . 91 4.5 Bubble dynamics in FC-72 under 1:8g. Bubble departure diameters are very small compared to those at 1g ( 0.25 D1g. Bubbles were observed to grow and depart at an angle normal to the shallow slope of the saw-teeth, a phenomenon that was previously demonstrated under 1g. . . . . . . . . . . . . . . . . . . . 92 4.6 Comparison of experimental data with departure diameter estimated using Equa- tion 4.1 for water and FC-72 under g . . . . . . . . . . . . . . . . . . . . . . . 93 4.7 Sliding velocity of bubbles in FC-72 under g. Uncertainty in bubble velocity is 0:5%. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95 xvi 4.8 Schematic diagram of bubble motion over a saw-tooth in FC-72 under g (not to scale). The arrows marked between the saw tooth and the vapor bubble indicate the direction of forces acting on the bubble due to pressure di erences. . . . . . 97 4.9 Liquid-vapor interface of a sliding vapor bubble in FC-72 under g . . . . . . . 98 4.10 Estimated bubble sliding velocities for liquid lm thickness varying from 0-25 m. The experimental bubble velocity of 27.4 mm/s is predicted closely for a H value of 17 mm. From bubble images, a value of H. . . . . . . . . . . . . . . . . 104 4.11 Boiling curves for large array experiments with FC-72 under g and 1g. Tsub = Tsat Tpool. Both experiments were performed with a highly subcooled pool and pool temperature was not controlled. Uncertainties in q00 and (Tw-Tsat) are 1% and 0:3% respectively. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105 B.1 Bottom plate used in the closed loop experiments with FC-72. The \race-track" is the ow channel in which ow is intended to occur in the clockwise direction. 120 B.2 Assembly of the closed loop experimental set up showing the polycarbonate ow loop and the assembly of the test device. . . . . . . . . . . . . . . . . . . . . . . 121 B.3 Schematic representation of the closed ow loop used in the study. . . . . . . . 122 B.4 Schematic representation of the net ow e ected mainly by the condenser location.123 C.1 Calibration of Omega ON44007 thermistor and a K-type thermocouple using a NIST certi ed thermistor as a standard . . . . . . . . . . . . . . . . . . . . . . 124 C.2 Calibration of test section heater using a NIST certi ed thermistor as a standard 125 C.3 Flow loop for the calibration of hot wire probe . . . . . . . . . . . . . . . . . . . 126 C.4 Flow loop for the calibration of hot wire probe . . . . . . . . . . . . . . . . . . . 127 xvii D.1 Illustration of the angular momentum imparted by a growing bubble normal to the shallow slope, leading to a net lateral liquid velocity . . . . . . . . . . . . . 129 D.2 Control volume de ned for estimation of resulting net lateral liquid velocity due to bubble growth at an angle normal to the shallow slope of the ratchet . . . . . 129 E.1 Illustration of a PIV system showing the laser sheet that is used to illuminate the particles. The image of the particles is recorded at two time instances using a camera to measure velocity [59] . . . . . . . . . . . . . . . . . . . . . . . . . . 131 E.2 Assembled view of the PIV chamber - basic parts include borosilicate glass cham- ber, and a top polycarbonate lid with ports for sensors . . . . . . . . . . . . . . 133 E.3 Exploded view of the PIV chamber showing the assembly of the copper heating block, borosilicate glass chamber, and a top polycarbonate lid with ports for sensors133 E.4 Velocity vectors obtained from a PIV experiment that are overlaid on top of the captured image. In this image, the shallow long slopes of the saw teeth face towards the left side of the image, and any expected ow due to asymmetric bubble growth will be from right to left in the image. Vectors colored in green and yellow represent measured and interpolated velocities, respectively. . . . . . 134 E.5 Magnitude of velocity vectors obtained from the PIV experiment. Color chart indicates liquid velocities of up to 17 mm/s, although scattered in di erent di- rections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135 xviii List of Tables 2.1 DRIE etch parameters and resulting saw-tooth angle . . . . . . . . . . . . . . . 37 2.2 Characteristics of di erent types of test sections fabricated . . . . . . . . . . . . 51 3.1 Summary of curve tting parameters from Fig. 3.8 and Fig. 3.9 . . . . . . . . . 65 3.2 Summary of available models and experimental data for bubble growth during inertia controlled (IC) and heat transfer controlled (HTC) regimes. . . . . . . . 70 A.1 Thermophysical properties of FC-72 and saturated water at 1 atm . . . . . . . . 119 xix List of Abbreviations Acronyms CVD Chemical Vapor Deposition CHF Critical Heat Flux DAQ Data Acquisition System DRIE Deep Reactive Ion Etching HWA Hot Wire Anemometry LUT Look Up Table MABE Microheater Array Boiling Experiment PECVD Plasma Enhanced Chemical Vapor Deposition PID Proportional Integral Di erential Controller PAE Phosphoric Acid and Acetic Acid Etchant PIV Particle Image Velocimetry TEDP Test Equipment Data Package Symbols A constant in Eq. 3.7, mm=s A area, m2 Cp speci c heat, J=kgK xx CD drag co-e cient D diameter, m F Force, N g gravity, m=s2 H height, m hlv latent heat of vaporization, J=kg I supply current, A Ja Jacob number, m k thermal conductivity, W=mK L length, m m mass, kg m exponent in power law curve t (Table 3.1) P pressure, Pa P momentum, kg:m=s Pr Prandtl number q power supplied to the heater, W q00 heat ux, W=cm2 qloss heat lost to the ambient, W R radius, m r radius of curvature, m xxi Re Reynolds number, m Rh heater resistance, t time, s T temperature, C v velocity, m=s V voltage, V Greek symbols thermal di usivity, m2=s growth constant 4P pressure drop, kPa 4Tsub inlet subcooling, C dynamic viscosity, Nsm 2 kinematic viscosity, m2=s density, kg=m3 surface tension, N=m saw tooth angle, degree shear stress, N=m2 Subscripts 1 ambient c cavity xxii h heater l liquid le equilibrium liquid superheat p particle sat saturation w saw tooth surface v vapor xxiii Chapter 1 Introduction Advances in electronics processing and fabrication techniques have led to signi cant miniaturization, allowing manufacturers to pack a billion transistors, carrying out a billion calculations into an area smaller than a thumbnail. Consequently, current supercomputers, which are indispensable for storm tracking, DNA mapping, social networking, and numerous other scienti c research applications, are more than 20 billion times faster than the rst ones built in 1960s. During the same period, the cost per million operations in a supercomputer has reduced by more than a billion fold. Central to these developments has been the evolu- tion in cooling technologies for the dissipation of resultant high heat densities. Liquid-vapor phase change (boiling) of dielectric uids has been at the forefront of cooling techniques starting with the immersion cooling technique employed in Cray 2 supercomputers to the ow boiling technique used in the current TITAN supercomputers. These advances in com- puting technologies aided largely by miniaturization are not limited to supercomputers, but have translated to laptops and hand-held devices. The microprocessors used in the latest computers are much faster than the supercomputers in 1960s and often uses a combination of air cooling and liquid cooling techniques such as forced convection heat sinks and heat pipes, respectively, to dissipate the resulting high heat density. A summary of this evolution in advances in computing power and cooling techniques is shown in Fig. 1.1 and 1.2. To further reduce the footprint of the billions of transistors, signi cant advances are required in the eld of electronics cooling to push the limits of super-computing and bring even more powerful computing to desks and palms. Such advances in electronics cooling, especially liquid cooling, will also be essential and critical to modern space electronics, like Martian rovers, and critical electronics including life support systems in space. 1 Figure 1.1: Evolution of supercomputers since its introduction illustrating the increase in number of Floating point operations per second (Flops) and the reduction in cost per oper- ation. The corresponding evolution of cooling systems used in these supercomputers from a simple liquid cooling system such as in car radiators to advanced phase change cooling systems is illustrated. A popular current laptop is used for comparison which is much faster than the supercomputers in 1985. 2 Figure 1.2: Chart illustrates the adaptation of advanced cooling techniques for high heat density applications compared to the simple, lower heat transfer air cooling techniques for low power, low heat density applications 3 Boiling or liquid-vapor phase change, as a mode of high heat ux dissipation technique can be realized in two di erent ways, namely, pool boiling or ow boiling. Pool boiling involves immersing the electronics in a dielectric uid with a low boiling point, like in the early supercomputers such as Cray 2. The heat from the microprocessor converts the liquid to vapor providing very high heat removal rates. These rates are increased even further if the liquid is owing such as in ow boiling. Flow boiling involves heat rejection to a liquid owing in a closed loop causing phase change to vapor as shown in Fig. 1.2. Such a cooling technique o ers a number of advantages over single phase ow of liquid such as higher heat transfer, lower ow rates required for the same heat transferred, and chip temperature uniformity. The heat transfer rate involved in phase change liquid cooling is also dependent on the regime of boiling. The highest heat transfer rates in boiling are associated with the lower surface temperature nucleate boiling regime which involves the cyclic nucleation, growth and departure of individual vapor bubbles. In contrast, a lm boiling regime, which involves the formation of a thin vapor layer between the surface and liquid, results in lower heat transfer compared to nucleate boiling but higher heat transfer than natural convection, at temperatures high enough to cause failure of any electronic components. Such advanced cooling techniques, however, consume power to facilitate pumping of uid through narrow passages. Reduction of required power to cool an electronic component is a critical factor to reduce the operational cost of the system. Further, it is of paramount importance in space systems, as power is a precious commodity. To further illustrate this, the current martian rover, Curiosity, uses a \pumped uid loop" to e ciently cool the on- board electronic components. If the pumping of uid in a an electronics cooling system can be achieved by devising a technique to achieve self-propelled ow of liquid, the operational cost of the cooling system could potentially be reduced signi cantly. Such a self-propelled liquid cooling system also o ers other advantages such as reduction in system weight by the removal of prime movers such as pumps. 4 The study reported in this dissertation, conducted in collaboration with Oregon State University, describes an innovative concept wherein a heat sink with microscopic asymmetric surface features in silicon is developed to create a self-propelled ow of a dielectric liquid, dissipating heat e ectively while operating in the high heat transfer nucleate boiling regime. Nucleate boiling is sustained by the presence of re-entrant cavities, which are very e ective vapor trapping sites. Under terrestrial conditions (1g), in such a heat sink, heat from the microprocessor (heater) causes nucleation of vapor bubbles (boiling) from the fabricated microscopic cavities located only on the shallow slopes of the silicon saw-tooth structures as shown in Fig. 1.3. The bubbles grow and depart from the sloped asymmetric surface of the ratchets at an angle normal to the shallow slope of the ratchets with signi cantly high velocities. This asymmetric bubble growth imparts an angular momentum to the uid thus e ecting liquid motion in the adjacent liquid. Under microgravity, due to lack of buoyancy, bubbles generated tend to reside on the surface and grow to several times larger than under terrestrial conditions. In a plain at surface, this would cause temperature to rise under the bubble leading to the failure of electronic device being cooled. However, using the micro-structured surface, the residing bubble which conforms to the shape of the structured surface is propelled due to pressure di erences in the trapped thin liquid lm underneath. This prevents any temperature rise and also leads to a self-propelled ow. This phenomenon is illustrated in Fig. 1.4 Primary objectives of the current study include: MEMS fabrication of the novel heat sink with an asymmetric saw-toothed surface with nucleation sites (re-entrant cavities) on only the long slope of the ratchets analysis of bubble growth and departure from such asymmetric silicon structures using high speed imaging environments analysis of e ects of asymmetric bubble growth on net lateral liquid movement using experimental techniques such as hot wire anemometry and particle image velocimetry 5 Net Liquid F low Vapor bubble Liquid Re-entrant Cavity Saw-t ooth Ratchet Heater Figure 1.3: Schematic diagram of the concept under 1g showing the cross-section of the saw-toothed structure. Bubbles grow and depart from the re-entrant cavities at an angle that is normal to the surface. This is hypothesized to result in a net lateral velocity. Figure 1.4: Schematic diagram of the concept under 0g showing the larger size of the bubble due to lack of buoyancy. Bubble slides from left to right due to pressure di erence in the liquid lm underneath, between the crest and trough, that arises due to di erences in radii of curvature (r) 6 experimentation aboard reduced gravity ight to analyze e ect of microgravity on bubble dynamics and thereby its e ects on net liquid movement 1.1 Thermally actuated pumping systems: Literature review Thermally actuated pumping of a working uid has been demonstrated in the past, in both single phase liquid and liquid-to-vapor phase change systems aided by di erent convective mechanisms. The current section reviews the applicability of both these types of systems for cooling of electronics. 1.1.1 Self-propulsion in single phase liquids In the case of heating a liquid, a convective ow can be obtained by Marangoni con- vection and Rayleigh-Benard convection. Marangoni convection [1{5] is a surface tension driven ow while Rayleigh-Benard convection is driven by buoyancy. As the liquid is heated, warmer liquid rises to the top and can move either to the left or right along the surface of the liquid, as illustrated in Fig. 1.5a. In this process, the liquid loses heat to the colder gas. This di erence in temperature along the surface of the liquid causes a gradient in surface tension, since surface tension is inversely proportional to temperature in pure liquids. The colder liquid moves down due to di erence in density. This causes the colder descending liquid to be surrounded by warmer liquid. The alternating pattern between colder liquid with high surface tension and warmer liquid with low surface tension, causes the de ection of liquid surface as shown in the Fig. 1.5b. This local convective ow due to surface tension gradients is known as the Marangoni e ect [6]. A large number of studies have been conducted to analyze the aforementioned concept of Marangoni convection and to understand the e ects of surface geometry [1{4], and the thickness of liquid layer [1, 3, 4] on the obtained convective transport. The surface geometry, especially, plays a very signi cant role since having a asymmetric heated surface geometry aids in setting up temperature gradients in the uid which in turn aids Marangoni convection. 7 Figure 1.5: (a) Image illustrates the local convective ow at the liquid surface and the resulting convective cells (b) The surface tension gradients due to alternating di erence in temperatures of warmer and cold liquids cause de ection of the free surface. Image was acquired from Carey [6]. This was demonstrated by Jo [1] where an asymmetric periodic saw tooth pattern was used for the heated surface and it was observed that the ratchets created asymmetric surface tension forces which induced a pumping e ect in a thin layer of silicone oil over the ratchets. The e ect of ratchet surface on the temperature gradients as observed by Jo [1] is shown in Fig. 1.6. Velocity measurements in oil using die-tracking demonstrated a maximum velocity of 0.86 mm/min at an optimum lm thickness of 1 mm. Stroock et al. [3] analyzed the e ect of liquid thickness and temperature gradient across the liquid on the Marangoni-Benard convective ow over similar saw-tooth pro led ratchets. It was reported that the direction of ow was dependent on the lm thickness and the temperature gradient. The maximum ow rate of oil observed was 2 mm/min at a temperature gradient of 130 C across the oil layer. The observed ow pattern over the asymmetric structures is shown in Fig. 1.7. In a thin lm of oil over symmetrical triangular grooves, it was observed by Alexeev et al. [2] that surface tension gradients induce a ow 8 at the free surface of the lm towards the trough of the groove and the ow is directed downwards at the trough as seen in Fig. 1.8, thus setting up symmetrical vortices over the period of a groove. In addition to Marangoni convection, surface geometry also a ects the heat transfer from the surface. In experiments conducted with 0.5mm of silicone oil lm heated on parallel symmetrical grooves, it was observed that the heat transfer rate was 30% more in the case of sinusoidal surface shape compared to a trapezoidal surface. From the aforementioned studies, it can be inferred that Marangoni convection leads to a very low ow rate and operates in the natural convection regime which may not be applicable for high heat ux applications. The next section will continue the review of phase-change propulsion systems. 1.1.2 Self-propelled ow in phase change systems A number of studies have been conducted in the recent past to develop self-propelled ow systems utilizing liquid-vapor phase change and asymmetry in heated surface structure or surface wetting characteristics. A self-propelled ow obtained during the nucleate boiling regime holds signi cant potential to be used as a high heat transfer, pump-less, phase change cooling system which is one of the objectives of the current study. In phase change systems, bubble dynamics play a signi cant role in the promotion of the pumping e ect. During phase change of liquid on a plain surface, vapor bubble ebullition cycle consisting of nucleation, growth and departure of the vapor bubble has been shown to induce a pumping e ect by driving the superheated liquid away from the surface and bringing colder liquid to the surface [6]. The kinetic energy of the liquid due to spherical bubble growth on a plane heated wall has been studied in the past by Witze et al. [7], and 9 Figure 1.6: (a) Oval ow loop with saw-toothed heated ratchet sections used by Jo [1] (b) Asymmetric temperature gradients induced by the shape of the ratchet pro le, which leads to surface tension gradients leading to liquid movement. The direction of liquid movement is marked in (a) Figure 1.7: Flow pattern in oil observed by Stroock et al. [3] over asymmetric mini-ratchets at velcities up to 2mm/min. Figure 1.8: Surface tension induced ow towards the trough of a symmetrical tri- angular groove which is marked by the movement of glass particles. Alexeev et al. [2] reported that this results in liq- uid ow towards the wall at the trough and away from the wall at the crest, thus setting up a vortex ow. 10 the liquid kinetic energy was expressed as, K:E: = 9:35 l dR dt 2 R3 where, K:E: = kinetic energy of liquid dR dt = bubble growth rate (1.1) From this equation it is also evident that if the bubble dynamics could be altered to cause bubble growth at an angle inclined to the vertical, then the velocity of the resulting liquid motion has a horizontal component which contributes towards the net lateral horizontal velocity. Linke et al. [8] demonstrated that liquid R-134a droplets over a surface with periodic saw-tooth structures, similar to the structure used in the current study, were propelled at velocities as high as 5 cm/s in the lm boiling regime at a signi cantly high temperature. The propulsion of the droplet was attributed to pressure di erences under the droplet in the thin layer of vapor as observed in Fig. 1.9. Pressure di erences arise due to di erence in radii of curvature of the liquid-vapor interface at the trough and crest of the saw-teeth. This is also known as the Leidenfrost phenomenon and the droplets are called Leidenfrost droplets. Theoretical models developed by Yuan and Prosperetti [9] have shown that phase change in uids could be utilized for producing a pumping e ect. It was shown that repetitive growth and collapse of a vapor bubble nucleating at heaters located in a small liquid- lled tube, that connects two reservoirs, can lead to a net lateral liquid ow as long as the bubble is not generated from the mid-point of the tube. The net ow was attributed to unequal liquid columns on both sides of the bubble. Net pumping velocities of 500 cm/s were reported when the bubble is at the ends of the tubes and the velocity approached zero as the bubble move towards the center of the tube. Such systems in the past have led to the development of 11 bubble based micropumps based on pumping due to surface asymmetry which are attractive for micro uidic biological applications. Bubble-based micropumps have also been developed by using asymmetry in structure and surface wetting characteristics. In experiments conducted by Meng and Kim [10], gas bubbles were generated in an asymmetric microchannel structure as shown in Fig. 1.10(i)a. The directional growth of gas bubbles due to asymmetry and bubble displacement using a hydrophobic-hydrophilic junction resulted in a net liquid ow. Liquid ow rates of up to 65 nL/s was reported. A similar bubble based micro-pump involving phase change was demonstrated by Jun and Kim [11]. A net pressure di erential was obtained as a result of surface tension imbalance arising from asymmetric heating in a microchannel. A bubble produced by a heater as shown in Fig. 1.10(ii)a can be translated to the right by switching on the heater to its right, while switching the previous (upstream) heater o . This sequential powering up of the heaters, causes a pressure gradient due to gradients in vapor pressure of the bubble and surface tension. This causes the bubble occupying the entire cross-section to move towards the location of the heated area, thereby sweeping the liquid along. Liquid velocities of up to 160 m/s were reported using isopropanol as the uid in microchannels of hydraulic diameter 3.4 m and a length 726 m. All the aforementioned bubble based micropumps involving single phase ow of liquid and multi-phase ow of gas-liquid can nd applications in electronics cooling applications as a way to pump uid over the hot surface. As described above, activation of liquid motion at the micropump may require an additional mechanism. Also, the discussed mechanisms were not designed for e ective heat transfer from the hot surface. The current study aims at developing a self-propelled ow mechanism, in which the heat transfer from a hot surface, such as a microprocessor, when immersed in a dielectric liquid leads to nucleate boiling which in turn propels the liquid. The current study aims at constructing a heat sink, as illustrated in Fig. 1.3, with periodic, asymmetric, saw-tooth structures on a silicon surface with a pro le that is similar to the surface used by Linke et al. [8]. Nucleate boiling is promoted 12 Figure 1.9: (a) Leidenfrost droplet motion over a saw-toothed surface demonstrated by Linke et al. [8] (b) Droplet propelled by pressure gradients in the thin liquid layer induced by the di erences in radii of curvature of liquid-vapor interface (c) The direction of liquid movement due to the pressure gradient is shown. R-134a droplet velocities of up to 5 cm/s were reported Figure 1.10: Bubble based micropumps that rely on asymmetry in surface structure, surface wetting characteristics and heater locations. (i) Meng and Kim [10] demonstrated preferen- tial gas bubble motion due to surface asymmetry in the form of a neck in the microchannel structure as seen in (a). The gas bubble translation was promoted due to a hydrophilic- hydrophobic junction causing bubble translation to the right as seen in (b). Liquid ow rates of up to 65 nL/s were reported. (ii) Jun and Kim [11] demonstrated bubble mo- tion due to surface tension gradients induced by sequential powering of the heaters. Liquid (isopropanol) velocities of up to 160 m/s were observed. 13 by fabricating re-entrant cavities only on the shallow slope of the saw-tooth structure to promote asymmetric bubble growth which leads to a net lateral ow of liquid. 1.1.3 Fabrication of re-entrant cavities and sloped surfaces: Background The surface structure of the designed silicon heat sink to be used in the current study is unique and hence literature corresponding to its fabrication methods are not very widely available. Hence the fabrication methods that were devised for obtaining the sloped surfaces and re-entrant cavities that open up on the sloped surfaces, required extensive calibration. In the past, fabrication of re-entrant cavities in silicon that are very similar to the pro les developed in the current study, and other bulb shaped re-entrant cavities, have been demonstrated in studies conducted by Goyal et al. [12] Baldwin et al. [13] and Nimkar et al. [14]. The aforementioned studies investigated bubble dynamics involved in nucleation from re-entrant cavities on plain surfaces, associated heat transfer enhancement and optimum spacing between the cavities for phase-change immersion cooling of electronics. An image of the re-entrant cavity used in [14] and a bulb shaped cavity used in [13] are shown in Fig. 1.11. The pyramidal re-entrant cavities used in [12, 14] were fabricated by transferring the patterns of the cavity base that are printed on a glass mask to a silicon wafer coated with photoresist using standard photolithographic procedure. The developed wafer is then dipped in a solution of potassium hydroxide that etches the exposed silicon along the grain angle of silicon, which is 54.7 , to result in a pyramidal structure that opens a mouth on the other side of the silicon wafer. In immersion cooling applications with dielectric liquids such as FC72, due to the low contact angle and highly wetting nature of the uid, the naturally existing cavities or im- perfections on a plain surface are ooded and hence vapor nucleation is delayed, thereby causing signi cant overshoot in wall temperature. One of the ways to mitigate overshoot as shown by Nimkar et al. [14] is to use re-entrant cavities of the shape shown in Fig. 1.11a, 14 that trap vapor more e ciently due to the presence of four corners. Re-entrant cavities also lead to signi cant reduction in onset of nucleate boiling as demonstrated by Baldwin et al. [13]. The mouth diameter of the cavities also signi cantly a ect the bubble dynamics such as bubble growth rate, bubble departure diameter and departure frequency. Numerous other shapes and forms of cavities have been investigated by researchers in an e ort to enhance boiling characteristics and thereby the performance of the phase-change cooled heat sink. Kandlikar et al. [15] studied the e ect of cavities, that are 100 m punch marks as shown in Fig. 1.12a on ow instabilities during ow boiling in microchannels. Ko sar et al. [16] investigated the performance of circular inter-connected re-entrant cavities as shown in Fig. 1.12. It was concluded that such re-entrant cavities decrease the heat ux required to initiate boiling and mitigate ow instabilities in microchannels. Fabrication of sloped surfaces in silicon, similar to the saw-tooth structures in the current study, has been performed in the past primarily for di ractive optical elements such as deep phase Fresnel lens [18]. Such sloped surfaces in silicon are developed by gray-scale lithography and deep reactive ion etching (DRIE) which was rst demonstrated by Whitley et al. [19]. Gray scale lithography is mainly used to develop structures in silicon that have a gradient in height. To achieve such structures in silicon, gray scale lithography is followed by DRIE which is a very common process to etch deep anisotropic structures such as channels or trenches in silicon. To better understand gray-scale lithography, the DRIE process and the factors a ecting the etching process have to be understood. The DRIE process has four main steps as illustrated in Fig. 1.13[20, 21]. The pattern to be etched is transferred on to a photoresist layer on silicon. The photoresist is then developed to form a photomask and expose the silicon layer to be etched Fig. 1.13a. The DRIE process in carried out in an STS ICP etcher, where silicon is etched in two steps - etching and passivation. In the etching step, silicon is etched using SF6 gas, as shown in Fig. 1.13b, and in the passivation step, a Te on like lm is deposited along the side walls and oor of the channel (Fig. 1.13c). In the subsequent etching cycle, the polymer lm only on the oor of the channel is removed by 15 Figure 1.11: (a) Pyramidal re-entrant cavities used in immersion cooled surfaces studied by Nimkar et al. [14]. Such re-entrant cavities are formed by anisotropic potassium hydroxide etch (b) Bulb shaped re-entrant cavity used in immersion cooled heat sinks by Baldwin et al. [13] and Bhavnani et al. [17] Figure 1.12: (a) Arti cial cavities for nucleation formed by punch marks that were used in microchannel heat sinks by Kandlikar et al. [15] (b) Inter-connected re-entrant cavities to form slots that were used in silicon microchannel heat sinks by Ko sar et al. [16] 16 SF6 ion impingement, thereby protecting the walls of the channel from being etched in the subsequent cycles. This process is repeated until the required depth is achieved. It has to be noted that in the etching cycle, the plasma inside the etching chamber consists of a mixture of gases, usually SF6 and O2. The increased concentration of SF6 in the plasma makes the etch more isotropic, thereby lending a slight concave shape to the surface. Increased concentration of O2 in the plasma leads to an increase in anisotropy but also increases the rate at which the photoresist is etched [21]. The ratio of silicon to photoresist rate is called etch selectivity. A typical photolithography process involves transferring a pattern in a mask to the photoresist deposited on a silicon layer, as discussed above. The patterns form either the transparent or the opaque region of the mask. Ultra-violet light passes through the transpar- ent regions of the mask to expose the photoresist layer on the mask, which is later developed to remove the exposed regions of the photoresist, to expose the silicon. In comparison, a gray-scale mask consists of a pattern that ranges from being opaque to transparent as shown in Fig. 1.14a. When ultra-violet light is passed through such a mask, the photoresist de- posited on silicon is exposed to varying depths (Fig. 1.14b). This leads to a gradient in height in photoresist. The next step is DRIE (Fig. 1.14c) which transfers the slope in pho- toresist to a slope in silicon, although the angle of slope achieved in silicon is a function of etch selectivity. Hence, the DRIE process parameters such as the ow rate of gases, and du- ration of etching and passivation cycles have to modi ed signi cantly to achieve the required angle. This step requires extensive calibration, as it involves optimization of a number of parameters. This step of tailoring the etch selectivity by manipulating the DRIE parameters has been investigated by Waits et al. [22, 23]. Although the steps of gray-scale lithography for optical elements, DRIE for forming channels, and wet etching for the formation of re-entrant cavities have been performed in the past separately for individual applications, the combination of three to form a large array of periodic delicate saw-tooth pattern to form a heat sink has never been attempted. 17 Figure 1.13: Steps involved in deep reactive ion etching (DRIE) (a) Formation of a photomask using photolithography (b) Etching of silicon using a plasma containing SF6 (c) Passivation cycle involving the deposition of a polymer material which protects the side walls from being etched in the subsequent etching cycles (d) Etching cycle is repeated. Figure 1.14: Steps involved in gray-scale lithography as illustrated in Waits et al. [23] (a) Design of gray-scale mask (b) Exposure and development of photoresist to obtain a gradient height in photoresist (c) DRIE to obtain a gradient height (slope) in silicon. 18 1.1.4 Bubble dynamics in pool boiling on plain and enhanced surfaces Studies of bubble dynamics involved in liquid-vapor phase change and their interactions with the heated surface and the liquid is critical for understanding the underlying mech- anism responsible for the enhanced heat transfer associated with boiling or phase change. Knowledge of bubble dynamics provides signi cant advantages in the design of phase-change heat exchangers and engineered surfaces. Bubble dynamics, in general, over the past decades were investigated for common uids such as water and/or refrigerants such as R-113, R-134a and surfaces made of common metals. However, with the rising demands of high heat dissi- pation in modern high power electronics, phase-change cooling is well placed as an attractive cooling technique involving a range of surfaces such as silicon with micro-scale features and carbon nanotubes, and dielectric uids such as FC-72 which promote direct contact cooling at low surface temperatures. These new combinations of surface structures, surface scales, and liquids alter the bubble dynamics signi cantly necessitating further study. Investigations on this front have led to signi cant developments in phase change heat transfer for microelectronics such as ow boiling in microchannels, direct contact liquid immersion cooling and enhancements to engineered micro-surfaces. In the recent past, bubble dynamics have also been manipulated for the development of bubble based micro-pumps which are attractive for applications in electronics cooling since the latent heat required for the bubble production is acquired from the microelectronics leading to pump-less, power-free pumping with high heat dissipation. It has been discussed earlier that the structure of heat sink used in the current study alters the bubble dynamics signi cantly to aid in asymmetric bubble growth and departure. This was also illustrated in the schematic diagram shown in Fig. 1.3. To analyze the e ects of structural asymmetry and asymmetry in bubble growth on the net lateral liquid motion, it is important to characterize the bubble dynamics for the surface considered in this study and the experimental conditions tested. Analysis of bubble dynamics, which is one of the primary objectives of the study, includes the study of bubble growth rate, bubble departure 19 frequency and bubble departure diameter as a function of heat ux, subcooling, and type of uid. Such information would also aid in the development of a model for the liquid velocity due to asymmetric bubble growth. Apart from contributing to the model developed, the analysis contributes to the scarce data available for bubble growth from re-entrant cavities and highly wetting uids such as FC-72. Analytical bubble growth models developed in the past can broadly be classi ed into two types - bubble growth in uniform and non-uniform temperature elds. In the former, a vapor bubble grows in a uniformly superheated pool of liquid where the bubble growth is symmetrical, as illustrated in g. 1.15. In models using the uniform superheat theory, bubble growth is mainly attributed to the pressure di erence between vapor inside the bubble (Pv) and the surrounding liquid (P1), which in-turn is related to Tv T1 using the Clausius Clapeyron equation. On the other hand, a non-uniform temperature eld results from the presence of a heated surface at a temperature Tw which is greater than Tsat, leading to bubble nucleation and growth as the bubble stays attached to the surface. Compared to the uniform superheat theory, bubble growth on a heated surface is often not spherical and growth is not symmetrical. Bubble growth from a naturally occurring or a structured cavity on a heated surface, as illustrated in Fig. 1.16, can be classi ed into two types - inertia controlled and heat transfer controlled bubble growth [6]. After the departure of a bubble, colder ambient liquid is drawn on to the surface, during when heat transfer from the surface to the liquid is by transient conduction and this period is known as waiting period which lasts until the boundary layer thickness grows enough to sustain nucleation. During the initial stages of bubble growth, the growth rate is quite high as the radius of curvature increases suddenly, due to the presence of a highly superheated liquid layer adjacent to the vapor interface. This rapid bubble growth is only limited by the inertia of the liquid and is hence known as inertia controlled growth. As the bubble grows radially, a very small thin lm of liquid underneath the bubble, known as the evaporation microlayer exists which evaporates at the vapor interface. With further growth, the superheated liquid lm surrounding the bubble, 20 Figure 1.15: Types of bubble growth models (a) Uniform superheat theory, involving spher- ical vapor growth in an uniformly superheated liquid (b) Bubble growth in non-uniform temperature eld; temperature highest at the wall[25]. known as the relaxation microlayer, depletes during which the bubble growth rate reduces. This regime of bubble growth is known as heat transfer controlled bubble growth. This process is illustrated in Fig. 1.16. Models developed for the case of uniform temperature eld can be divided into inertia controlled growth (Rayleigh [24]) and heat transfer controlled growth (Mikic et al. [25], Plesset and Zwick [26], Scriven [27]). These models were later extended for the case of bubble growth in a non-uniform temperature eld like those of Zuber [28], Cole and Shulman [29], Van Stralen [30] and Mikic and Rohsenow [31]. However, the results in this study will show the inadequacy of these models to predict bubble growth for structured surfaces and highly wetting uids such as FC-72. Models for experimental conditions that are similar to those in the current study are mostly empirical in nature and often present asymptotic growth relationships as a function of time (Lee at al. [32], and Ramaswamy et al. [33]). Such models are, however, simplistic in nature and do not adequately explain the e ects of liquid, and surface such as roughness or re-entrant cavities. The study presented here is an e ort to contribute to the fundamental understanding of the bubble dynamics such as growth rate, departure frequency and diameter experimentally using high speed imaging for the unique surface structure with re-entrant 21 Figure 1.16: Regimes of bubble growth illustrated as shown in Carey [6] (a) Ambient cooler liquid is brought in to contact with the heater surface (b) Growth of boundary layer (c) Bubble nucleation and inertia controlled regime (d) Relaxation microlayer thickness decreases (d) Bubble departure. 22 cavities that was tested with FC-72. Some of the studies that are similar to the conditions of the current study, in the aspect of pool boiling on silicon surfaces using FC-72 include (Hutter et al. [34], Nimkar et al. [14], Ramaswamy et al [33], Moghaddam and Kiger [35], and Demiray and Kim [36]). Of the aforementioned studies, [14, 17, 33, 34] conducted pool boiling studies on silicon surfaces with structured cavities. No studies have been conducted for the nucleation of FC-72 on sloped surfaces with re-entrant cavities. Growth rate of FC-72 was analyzed by Hutter et al. [34] on silicon surfaces with cylin- drical cavities. The observed bubble growth and bubble shape is shown in Fig. 1.17. It was observed that the growth rates vary with time on the order of t 1=2 as noticed in Fig. 1.18, which is similar to the trends reported by Ramaswamy et al. [33] who conducted pool boiling experiments using FC-72 on structured surfaces with square cavities. Nimkar et al. [14] and Bhavnani et al. [17] used pyramidal re-entrant cavities and bulb shaped re-entrant cavities, respectively, during pool boiling with FC-72. E ects of heat ux, cavity spacing, bubble departure frequency were studied. It was observed that the pyramidal re-entrant cavities completely avoided nucleation hysteresis compared to the bulb-shaped cavities. Also, bubble departure diameter and frequency were found to increase with wall superheat. The data thus collected on bubble dynamics at 1g will be corroborated with data from microgravity experiments conducted using the same set up. 1.1.5 Bubble dynamics under reduced gravity It was earlier stated that one of the objectives of the study would be to develop and test a pump less ow loop utilizing surface asymmetry and re-entrant cavities for space electronics applications. This requires a better understanding of vapor bubble dynamics under reduced gravity and hence its e ect on pumping. Due to the complexity of such experiments, literature on micro gravity experiments and the related bubble dynamics is very limited. 23 Figure 1.17: Bubble growth process from a cylindrical cavity on a silicon surface using FC-72 - Hutter et al. [34] Figure 1.18: Bubble growth rate from a cylindrical cavity on a silicon surface using FC-72 - Hutter et al. [34]. The asymptotic growth relationship is marked as the equation of t in the gure. 24 Microgravity testing methods Microgravity, or reduced gravity, conditions can be produced in a number of ways by compensating for the earth?s gravity. Some of the common methods to produce \weightless- ness" or microgravity include a free fall in di erent trajectories, a freely drifting spacecraft or space station, and ballistic rockets. A free fall could be conducted in drop towers and drop shafts where the free falling object on guide rails experiences microgravity due to the absence of forces that oppose gravity [37]. The relationship between free fall height and time can be expressed as, t = s 2H g (1.2) where, t = time of free fall H = free fall height Key drop tower and drop shaft facilities in United States include NASA Glenn Research Center in Cleveland, Ohio, which has a drop tower of 24 m and a draft shaft of 132 m providing a microgravity duration of 2.2 and 5.2 seconds respectively. Marshall Space Flight Center in Huntsville, AL, can produce 4.6 s of microgravity with a drop tower of 106 m. The drawback of testing in a drop tower or drop shaft is the low duration of microgravity and the quality. The low duration signi cantly a ects analysis of processes with longer time scales such as bubble growth in microgravity, which is considerably slower compared to that under earth gravity. Also, in a drop tower or a drop shaft the highest quality of reduced gravity is achieved during the initial period of fall, extending less than 2 seconds, after which the free falling object experiences a residual gravity due to the drag forces of surrounding gas. However, residual gravity can be reduced using evacuated conditions. The short duration of microgravity is also a drawback since any natural convection that started before the drop will not be eliminated completely within the short period of microgravity. 25 Figure 1.19: (a) Parabolic trajectory of a C-9B ight and the achieved reduced gravity periods (b) G-jitters in a parabolic ight due to disturbances such as air turbulence - Straub [37]. A similar method to produce microgravity involves free fall along parabolic trajectories that are own by special aircraft such as KC-135, DC-9, C-9B, and recently Boeing 727- 200F, by NASA from the Johnson Space Center in Houston (Fig. 1.19a). Advantages of such a method include longer durations of microgravity, up to 20 seconds, and the volume of experiments that can be tested. Also, di erent gravity levels can be achieved by altering the trajectory of the ight to simulate lunar and martian gravities. However, the quality of microgravity could be a ected by air turbulence and ight maneuvers, producing gravity uctuations on the order of 10 2g as shown in Fig. 1.19b, known as g-jitters. However, such g-jitter could be dampened by free oating the experiment inside the ight. Parabolic trajectories to produce reduced gravity are also achieved in ballistic missiles known as \sounding rockets" which are categorized as sub-orbital ights. The sounding rockets are powered by a solid propellant rocket motor to achieve the trajectory and the rockets reach altitudes of up to 165 km where the air density is very low, thereby reducing the air drag to produce a high quality ( 10 6g), long duration ( 200 seconds) microgravity regime [38]. The payloads are retrieved using a parachute for re-use and data retrieval. Some 26 of the commonly used facilities include the sounding rockets under the German TEXUS program and NASA?s Terrier-Orion sounding rockets. The best method for microgravity testing is aboard the International Space Station (ISS) which provides unlimited duration of high quality microgravity. The ISS facility used to conduct boiling experiments is known as Microheater Array Boiling Experiment (MABE) located in a module known as the Microgravity Science Glovebox (MSG) [39]. Bubble dynamics and heat transfer under microgravity In boiling at terrestrial gravity, bubble departure from the heated surface is primarily a balance between surface tension which causes the bubbles to remain attached to the surface, and, drag and buoyancy forces which cause the bubbles to detach from the surface. But under reduced gravity, the forces due to buoyancy are negligible and hence the bubble tends to remain at the surface for longer periods resulting in larger bubble departure diameters as has been widely reported in previous studies [40{43]. An image of the large bubble diameters observed by during pool boiling of FC-72 under microgravity is shown in Fig. 1.20. The dependence of departure diameter on gravity can be expressed as, Dd; g Dd;1g = a g 1/2 (1.3) where, Dd = bubble departure diameter g = earth gravity a = tested gravity Because of the bubble size being as large as the heater as seen in Fig. 1.20, the heat transfer is signi cantly reduced and many studies have reported reduction in heat transfer compared to terrestrial gravity using a variety of uids such as FC-72 [38, 41] and R-113 [40]. Heat transfer from a 2.7 mm x 2.7 mm array of heaters to FC-72, studied in [38] shows 27 Figure 1.20: Large bubble of diameter 2.5 mm in pool boiling of FC-72 under reduced gravity at a wall superheat of 30 C - Kim and Benton[38]. that at high wall temperatures heat transfer at microgravity is signi cantly a ected (Fig. 1.21a). It was also demonstrated by Henry and Kim [41] that heat transfer in microgravity is a function of heater size and liquid subcooling. In a large heater, the bubbles are distributed and the presence of a large number of smaller satellite bubbles lead to a higher heat transfer compared to that with smaller heater sizes (Fig. 1.21b). Also, high liquid subcooling resulted in a higher heat transfer compared to lower subcooling. Bubbles of larger sizes, observed under microgravity, also cause dry-out of the surface under the bubble which leads to critical heat ux (CHF). CHF is the maximum heat ux condition beyond which heat transfer decreases and wall temperature increases signi cantly (Fig. 1.21a). Pool boiling experiments on at heater surface using R-113 was performed in space shuttle ights by Lee and Merte Jr. [43]. E ects of heat ux and liquid subcooling were studied. While observations of larger bubble diameters similar to a number of other studies were reported at all tested conditions, an interesting observation of bubble migration was reported. It was observed that a number of small bubbles coalesced to form a very large bubble which departed from the surface and hovered close to the surface. Bubble departure was mainly attributed to the momentum and coalescence of smaller bubbles. Also, the migration of smaller bubbles towards the larger bubble at velocities of 2.5 cm/s noticed at 28 Figure 1.21: (a) Lower heat transfer observed under microgravity in pool boiling of FC-72 at higher wall superheat. Critical heat ux (CHF) also reduced compared to earth gravity - Kim and Benton[38] (b) Heat transfer increases with increasing heater sizes and liquid subcooling - Henry and Kim [41] 29 high subcooling led to a 40% increase in heat transfer compared to similar conditions at earth gravity. An image of the large hovering bubble and adjacent smaller bubbles are shown in Fig. 1.22 Figure 1.22: Large hovering bubble in R-113; smaller adjacent bubbles near the larger bubble were noticed to migrate towards the larger bubble at high subcooling - Lee and Merte Jr. [43]. Similar migration of vapor bubbles on plain heated surfaces has also been reported previously under reduced gravity attributed to g-jitters [42]. Sliding velocities of up to 2.25 cm/s were reported in pool boiling of distilled water. Marangoni and thermocapillary convection [38, 44{46] have also been shown to cause bubble migration due to boiling in a subcooled pool of liquid. From the aforementioned studies it could be observed that the bubble dynamics in pool boiling are considerably altered under microgravity. Also, literature for such microgravity studies is extremely limited owing to the experimental constraints involved in a microgravity experiment, such as duration of microgravity, number of experiments, and the complicated experimental setup required for safety. Data is often limited to pool boiling on at sur- faces and wires in uids such as water, R-113 and, FC-72. However, no studies have been conducted for surface pro les like the one used in current study, and hence no information is available on the e ects of surface asymmetry and re-entrant cavities on bubble dynam- ics. Hence, the current study to analyze the e ect of surface asymmetry at reduced gravity 30 on bubble dynamics and liquid pumping will be a novel attempt and will provide valuable information about bubble dynamics. 31 Chapter 2 Fabrication of Silicon Test Devices The silicon heat sinks used in this study consist of saw-tooth cross-sectioned, asymmetric surfaces and re-entrant cavities that are located only along the long slope of the surface. The heat sinks fabricated in this study are of two types and were designed to be be used in two di erent sets of experiments. One set of experiments, termed as large array experiments, were designed for global measurements of liquid velocity, bubble dynamics and heat transfer. The test devices used in large array experiments consisted of 80 ratchets in silicon. The other set of experiments were designed to study local bubble dynamics and resulting e ects on the adjacent liquid using a small array of silicon ratchets, termed as small array experiments. The small array test device consisted of only eight ratchets of smaller footprint compared to the large array experiments. The small array test devices were fabricated for use at Oregon State University, a partner institution on this research project. While the overall objective of the study, which is to obtain net lateral liquid ow using structural asymmetry, is the same in both sets of experiments, there are di erences in the scale of the experimental setup, methods of experimentation, and techniques used. Because of these di erences, the test devices used were designed and fabricated to cater to speci c requirements of the techniques used and experiments conducted. Some of the important characteristics of the test device required for the di erent experiments are outlined below: Fabricate saw-tooth cross-sectioned ratchets on silicon to obtain a nominal pro le of 30 - 60 - 90 . The surface of the resulting ratchets required a smooth nish as the presence of any surface imperfections will act as nucleation sites. 32 Etch re-entrant cavities from the back side of the ratchets to open up trapezoidal mouth on the sloped face of the saw-teeth. The presence of these features only on the shallow slope is to cause an asymmetry in vapor bubble nucleation thereby leading to an asymmetry in momentum imparted to the liquid. Design and fabricate aluminum serpentine heaters on separate silicon wafers. The traces were designed to prevent electromigration while providing adequate power to induce vapor nucleation. Electromigration is the severance of current carrying metal paths due to excessive current density in a metal of given cross-section [47, 48]. The heaters for the small array experiments were also designed for infrared imaging of cavities from the back side of test devices, requiring optimum spacing between the aluminum heater traces and anti-re ective coating on the heater to prevent re ection from the aluminum traces. Dice the silicon wafer consisting of saw-tooth ratchets into large and small array of ratchets for the two di erent sets of experiments. Bond the silicon wafer consisting of ratchet array to the heater wafer with minimal thermal interface resistance. Due to the di erence in the footprint area of the large and small array test devices di erent bonding schemes were used. Because of the unique cross-section of the test structures utilized, literature correspond- ing to the fabrication processes was limited and fabrication steps involved elaborate calibra- tion and trials to precisely obtain the required structure. Broadly, the fabrication of the heat sink consists of four steps. Gray-scale photolithog- raphy to obtain the saw-tooth ratchets in silicon, etching re-entrant cavities on the shallow slope of the saw-tooth ratchets, fabrication of serpentine aluminum heater on silicon, and nally dicing and bonding the two silicon layers using techniques such as fusion and gold eutectic bonding. These basic fabrication steps are represented in Fig. 2.1. 33 Figure 2.1: Illustration of steps involved in fabrication of the silicon test devices to be used in the large and small array experiments 34 2.1 Fabrication of silicon ratchet array The saw-tooth structures were fabricated on a 100 mm diameter and 0.5 mm thick double side polished <100>silicon wafer. Gray-scale photolithography process involves deposition of photoresist with a gradient in height. The height gradient in photoresist is obtained by allowing UV light to pass through a gray-scale optical mask consisting of patterns ranging from being completely transparent to opaque. This causes light penetrating to increasing depths in photoresist, which when developed results in a gradient in height [18, 22, 23]. The photoresist used in the current study is SPR 220-7. This asymmetric photoresist structure is transformed into a silicon structure of required dimensions using DRIE, as illustrated in Fig. 2.2. The angle of surface etched in silicon is dependent upon the etch selectivity (ratio of silicon to photoresist etch rate) of the photoresist used, the duration of the etching and passivation cycles, and the ratio of gas ow rates in the plasma during DRIE. Extensive calibration was conducted to analyze the e ects of ow rate of the constituent gases in the plasma during the DRIE etching cycle and cycle duration, on the etched angle of the silicon structure and the surface quality. Prior to DRIE, the wafer with the photoresist structure was subjected to an oxygen plasma, known as the plasma de-scum process, to remove a thin layer of photoresist so as to expose a few micrometers of silicon at the trough. Without this step, etching would begin unevenly across the wafer leading to a poor saw-tooth pro le. All the DRIE processes were performed using an STS ICP etcher. As discussed earlier, the DRIE process consists of an etching cycle, during which silicon is etched using SF6 ow, and a passivation cycle which involves deposition of a Te on layer (C4F8) along the side walls. The rst wafer was etched using a standard Bosch process with the ow rate of SF6 set at 130 SCCM and O2 at 13 SCCM in the etching cycle, and 85 SCCM of C4F8 in the passivation cycle. The time duration used for etching and passivation cycles were 13s and 7s respectively. This resulted in a very high etch selectivity of 100 which in turn resulted in a very high angle of the saw-tooth and a rough surface. Hence to reduce the selectivity i.e., increase the photoresist 35 Silicon Photoresist Silicon Silicon 1. Double side polished 100mm silicon wafer 2. Gray scale deposition of photoresist at a desired angle 3. Deep Reactive Ion Etching (DRIE) to obtain the saw tooth angle of 30 deg. 30 Figure 2.2: Steps involved in the fabrication of silicon saw-tooth ratchets removal rate, the oxygen ow rate was increased sequentially in the subsequent cycles as oxygen has been shown to etch the photoresist faster [20, 23]. Reduction of the duration of passivation cycles also has the e ect of increasing the photoresist etch rate and hence in all subsequent recipes the duration of passivation cycle was reduced to 3s. The recipes followed for DRIE in the subsequent trials have been tabulated in Table 2.1. In recipe A, the ow rate of O2 was increased drastically which decreased the selectivity resulting in an angle of 19 . However, short passivation cycles resulted in poor surface quality due to the formation of black oxide or silicon grass as shown in the SEM image in Fig. 2.4. These nanostructures in silicon or grass are generally formed when the polymer material is not completely etched due to short etching cycles or low concentration of etching species (F radicals) in the plasma [20]. This scenario of low concentration of etching species arises when O2 ow rate is increased in the plasma. It has also been reported that high concentration of oxygen forms micromasks which results in grass like structures due to the directional etching [21]. Hence, to prevent the silicon grass formation, the concentration of 36 the etching species can be increased by increasing SF6 ow rate. But increasing SF6 also leads to deterioration of anisotropy [21]. Hence, in all the subsequent recipes B, C and D short periodic etching lasting not more than 10 seconds using SF6 in the absence of O2 and C4F8 was introduced in between sets of etching/passivation cycles. Figure 2.3: Illustration of DRIE parameters discussed in Table 2.1 Table 2.1: DRIE etch parameters and resulting saw-tooth angle Recipe Flow Rate in the Etching Cycle O2=SF6 (SCCM) Duration of Etching/ Passivation Cycles (s) Angle ( ) Depth d ( m) A 80/100 13/3 19 235 B 65/100 13/3 24 305 C 50/100 13/3 32.15 440 D 53/100 13/3 30.5 415 Since the objective is to achieve an angle of 30 for the saw-tooth, it was required to slow down the photoresist etching rate and hence in recipe B, O2 ow rate was decreased to 65 SCCM keeping the other parameters constant along with periodic 10 second SF6 etching. This procedure resulted in a saw tooth angle of 24 without the silicon grass. In recipe C, by further decreasing the O2 ow rate to 50 SCCM an angle of 32.5 was achieved and nally with 53 SCCM of O2, noted as recipe D, an angle of 30.5 was achieved without any grass formation. An image of a cross-section of the ratchet array with a nominal saw-tooth angle of 24 and without any silicon grass is shown in Fig. 2.5. 37 Figure 2.4: SEM image of the ratchets with an angle 19 . The inset shows an SEM image of silicon nano-structures, known as \silicon grass", formed during the DRIE process with Recipe A. Figure 2.5: Cross-section of a ratchet array with a nominal saw-tooth angle of 24 38 2.2 Fabrication of re-entrant cavities The next step in the process was to etch the re-entrant cavities from the back side of the saw-tooth structure using potassium hydroxide (KOH), known as anisotropic wet etching process. The fabrication steps involved in etching the re-entrant cavities are illustrated in Fig. 2.6. Before the etching process, a 2000 A silicon nitride layer was deposited using low pressure chemical vapor deposition (LPCVD), which acts as a barrier layer preventing the KOH from etching silicon. To prevent any damage to the delicate ratchets during the process of etching cavities and subsequent handling of the wafer, a 500 m thick, 100 mm diameter Pyrex wafer was stuck on top of the ratchets. The Pyrex wafer was coated with photoresist AZ 5214 after treatment with hexamethyldisilazane (HMDS) vapor for 5 minutes which improves adhesion of photoresist to the surface. The Pyrex wafer with the photoresist is stuck to the ratchet side of silicon wafer and baked on a hot plate at 100 C. The back side of the saw teeth was then coated with photoresist AZ 5214 after treatment with HMDS. This is followed by patterning of the photoresist using an optical mask to de ne an array of squares of side 0.655 mm and 0.246 mm for each pair of cavities - large and small respectively as shown in Fig. 2.7. Each saw tooth had eight such pairs spaced equidistantly along the transverse length of the saw tooth. The pattern of cavities on the optical mask held on the back side of the saw teeth was aligned with the saw teeth on the other side. This ensured that the cavities were etched only on the shallow slope. The alignment process was performed using a mask aligner consisting of two microscopes - one on the side of ratchets and other on the side of cavities, and the position of mask was adjusted until the cavity pattern was aligned to the ratchets. The photoresist was then developed to expose the nitride layer as shown in Fig. 2.6(iii). This exposed silicon nitride is etched to transfer the pattern to silicon. The wafers were then dipped into a solution consisting of 30% KOH in DI water with 20% isopropyl alcohol. The temperature of the solution was maintained at 80 C during the etching process. The beaker containing the mixture is covered by a re ux condenser that was mounted on top to prevent evaporation of the solution thereby maintaining the 39 integrity of mixture. The etching process occurs along the grain angle of silicon, which is 54.6 . The anisotropic etching process was periodically monitored by measuring the depth of etch under the microscope until the cavities breached the other side of the wafer on the long slope of the saw teeth to form a trapezoidal mouth Fig. 2.6(vi). Once the cavities were formed, the photoresist was stripped by cleaning with acetone, isopropyl alcohol and DI water. The remaining silicon nitride was removed using a phosphoric acid etch at 154 C. After the completion of all the steps, the Pyrex wafer was removed by dipping the wafer in acetone until the Pyrex wafer parted from the silicon wafer. It has to be noted that test devices with a saw-tooth angle of 24 has only one cavity along the shallow slope of the ratchet compared to the two cavities along the long slope in other test devices. An SEM image with cavity mouth measurements for a 24 test device is shown in Fig. 2.8. This can be attributed to misalignment of the mask used to pattern the cavities on the back side of saw teeth. During DRIE, due to non-uniformity in etching, the saw tooth angle across the wafer varies between the outer periphery and the center of the wafer. This also leads to a variation in cavity mouth sizes. So even a slight misalignment could cause huge shifts in the cavity mouth diameter. 2.3 Fabrication of heater An aluminum serpentine heater was used in the study to supply heat for the pool boiling experiments. Two di erent heater designs were used for the study - one for the large array test device and another for the small array test device. The two di erent heater designs are shown in Fig. 2.9. The steps involved in the fabrication of heater are shown in Fig. 2.10. All the heaters were fabricated on a 100 mm diameter and 0.5 mm thick double side polished < 100 > silicon wafer. The wafer was oxidized in a furnace to deposit 5000 A of silicon oxide. This is followed by evaporating a 2 m thick layer of aluminum on to the SiO2 surface using e-beam evaporation process. Photoresist AZ 5214 was deposited on the aluminum layer for photolithography. The heater design was then transferred to the photoresist from an optical 40 Silicon Silicon Nitride Silicon Silicon Nitride Silicon (ii) Chemical vapor deposition of nitride Silicon Nitride Silicon Nitride Silicon Photoresist Photoresist Silicon Nitride Silicon Nitride Silicon Silicon (iv) Nitride etch to expose silicon (vii) Nitride etch using phosphoric acid Photoresist 56 56 Silicon Nitride Silicon Nitride Silicon (vi) Strip photoresist using acetone (i) Saw-tooth ratchet after DRIE (iii) Patterning of cavities using photolithography 56 56 Silicon Nitride Silicon Nitride Silicon (v) Anisotropic wet etching of silicon using KOH Figure 2.6: Illustration of fabrication steps involved in etching re-entrant cavities from the back side of saw-teeth to form trapezoidal cavity mouths only on the shallow slope 41 Figure 2.7: Isometric and back side view illustrations of small and large array cavities on the shallow slope of saw-toothed surface Figure 2.8: SEM image showing the top view of saw teeth and cavity mouth measurements in the test device with a saw-tooth angle of 24 42 mask using a mask aligner. This is followed by the development of photoresist to expose the aluminum layer as per the heater design. The heater was formed by etching the exposed aluminum in a solution of phosphoric acid and acetic acid etchant (PAE) for 60 minutes. The heater was then diced to the size of the footprint of large array test section. The fabrication of heater for small array test section involved additional steps. A thermal isolation trench was etched using DRIE along the perimeter of the heater to promote one dimensional heat transfer across the heater wafer to the silicon ratchets. This process is conducted along with the process to reduce the diameter of the heater wafer to three inches to t in the test setup for the small array experiments. The wafer diameter was reduced by etching a ring of diameter equal to three inches through the silicon using DRIE. Prior to the etching of isolation trenches and the three inch ring, photoresist was coated on the side of heater and patterned using one mask for both the aforementioned steps. Since the trench was only designed to be 350 m deep, after etching to this depth the trench was covered with photoresist to prevent any further etching. However, etching along the three inch ring continued until the silicon wafer was completely etched through to reduce the wafer to a diameter of three inches. The photoresist was stripped from the wafer using acetone. Further, to facilitate IR temperature measurements the entire surface of the heater was coated with a 1.1 m thick layer of silicon nitride using plasma enhanced chemical vapor deposition (PECVD). This layer of silicon nitride serves to prevent the re ection of infrared radiation. Photoresist was spun on top of the silicon nitride layer and patterned using a mask in a mask aligner to expose the nitride layer over the electrical contacts. This was followed by etching the nitride layer from the electrical contact pads of the heater to enable power supply as silicon nitride acts as an insulator. The additional steps required in the heater for small array experiments are illustrated in Fig. 2.11. Images of fabricated aluminum serpentine heaters on silicon for the large and small array test devices are shown in Fig. 2.12. 43 A D CB B A D C Figure 2.9: Sketch of aluminum serpentine heaters used in large and small array experiments that were fabricated on silicon. The width and thickness of heater traces are 500 m and 2 m respectively. The electrical contact pad con guration is labeled in the gure. The arrangement of electrical contact pads for the small array heater is based on the arrangement of electrical contacts in the wafer xture. 44 SiO2 Silicon SiO2 (iii) E-beam evaporation of aluminum of thickness 2 m Aluminum SiO2 Silicon SiO2 (iv) Deposition of photoresist AZ 5214 Aluminum Photoresist AZ 5214 SiO2 Silicon SiO2 (v) Exposure of development of photoresist using the heater mask Aluminum SiO2 Silicon SiO2 (vi) Aluminum etch by immersing in PAE for 60 min SiO2 Silicon SiO2 (vii) Photoresist stripped by immersing in acetone Aluminum heater traces Silicon (i) Double side polished 100mm diameter, 500 m thick silicon wafer SiO2 Silicon SiO2 (ii) Wet oxidation of wafer at 1000 C to obtain 5000A of oxide Figure 2.10: Fabrication of aluminum serpentine heaters used in large and small array test devices. 45 Silicon SiliconNitride SiliconSilicon Silicon SiO2Al heater Silicon SiliconNitride SiliconSilicon Silicon Al heater Figure 2.11: Fabrication of isolation trenches on the back side of heater footprint and redution of wafer diameter to 3 inches. Silicon nitride was coated to eliminate re ection of radiation during infrared measurements made during the small array experiments at OSU. 2.4 Bonding of saw-tooth ratchet wafer to the heater wafer Bonding of the silicon wafer containing the saw-tooth ratchet array to the heater wafer was performed di erently for the large array and small array test structures. Large array test structures were bonded to the heater wafer using fusion bonding where the bonding between two ultra- at extremely clean surfaces is due to bonding bridge re- placements and bonds between OH groups [49]. Surfaces to be fusion bonded are required to be extremely clean, particle free, at, smooth and hydrophilic. The surfaces to be bonded were rendered hydrophilic using an RCA-1 clean which is used to remove organic residues from the silicon wafer. The recipe for RCA-1 clean is 5 parts of DI water, 1 part hydrogen peroxide (H2O2), and 1 part ammonium hydroxide (NH4OH). The solution was heated up to 70 C when it starts boiling. The surfaces were immersed in to the solution for not more than 5 minutes. This was followed by rinsing the surfaces with DI water and thoroughly drying it. The surfaces to be bonded were immediately brought into contact with each other 46 (a) Fabricated aluminum serpentine heater on a silicon wafer for large array test devices (b) Diced aluminum serpentine heater (c) Fabricated aluminum serpentine heater on a 3" silicon wafer for small array test devices (d) Back side of the small array heater showing the isolation trenches (350 m deep) along the perimeter of the heater footprint Isolation trenches Figure 2.12: Images of fabricated aluminum serpentine heaters used in large and small array experiments. The diameter of heater wafer for the large and small array test experiments are 4 inches and 3 inches respectively. 47 mechanically at room temperature. Under the right conditions, the two surfaces would bond together instantly, although weakly. This was followed by placing the weakly bonded surfaces in an oxidization furnace under pressure at 1100 C for 2 hours. The resulting bonds have a very high fracture strength of up to 20 MPa [49]. Because of the high temperature, any aluminum on the surface would soften or melt away. Hence, for the large array test devices, the aforementioned heater fabrication process was carried out after the bonding process. The fusion bonding process required multiple trials at di erent oven temperatures and durations before achieving one successfully bonded test device. By conducting initial calibration tests for fusion bonding at two di erent temperatures of 450 C and 1100 C, it was learned that the voids were signi cantly reduced at higher temperature. The fusion bonding process of large array silicon saw-tooth ratchet array with the heater is schematically represented in Fig. 2.13 Silicon ratchet Silicon heater wafer Flat, ultra clean, hydrophilic surfaces mechanically brought into contact at room temperature. Bond strength increased by heating at 1100 C in a furnace for 2 hours Fusion bonding Figure 2.13: Illustration of fusion bonding process used to bond the large array of silicon saw-teeth to the heater wafer. The bonding of the small array test devices by fusion bonding process could not be carried out due to the very small area available for bonding. For the small array test devices, bonding of the saw-teeth to the heater was achieved using gold eutectic bonding. Gold eutectic bonding is a low temperature process which involves the use of a solder made of gold-tin alloy which when held between the surfaces to be bonded and heated to the eutectic temperature of the solder, the materials fuse to provide a high strength bond. In this process, both the bonding surfaces were prepared by evaporating 1000 A titanium followed by 2700 A 48 of nickel and nally 1500 A of gold. Titanium improves the adhesion of gold to silicon [49] and nickel serves as a di usion barrier between gold and silicon [50]. The deposition of layers of Ti-Ni-Au on the back of heater wafer was performed after coating photoresist and patterning the back side to only expose the silicon spanning the footprint of the heater. After the deposition of Ti-Ni-Au a lift-o process is performed by immersing the heater wafer in acetone. This causes the thin layer of photoresist under the layers of Ti-Ni-Au to dissolve, thereby causing the lift-o of Ti-Ni-Au from the surface except for the area where the layers are directly in contact with silicon, thus leaving layers of Ti-Ni-Au to span only the footprint of the heater. The bonding process was carried out using an FC150 automated die bonding machine where an upper arm holds the die and the substrate (heater) rests on a stage. A 1 mil thick alloy of 80% Au 20% Sn was used as a solder in the bonding process. The solder with a eutectic temperature of 280 C was advantageous since this temperature is well below the melting point of aluminum used in the heater. The bonding process was carried out at a temperature of 300 C by applying a weight of 200g for a period of 100 s. This process is illustrated in Fig. 2.14. The bonded test structures were inspected for voids using x-ray imaging. Presence of voids at the interface would appear dark in an x-ray image. From the x-ray images shown in Fig. 2.15 it can be noticed that the bonding was void free. The dark square regions in the x-ray image are the re-entrant cavities. It can also be noticed that some of the cavities appear lighter as the solder melted into the cavities. In summary, two types of large array test sections were fabricated - one with a saw tooth angle of 24 and the other with an angle of 31 . The 24 test section has one cavity along the long slope between the crest and trough of each saw tooth, and 31 test section has two cavities. The small array test sections that were fabricated, similar to the large array test section, di ered in saw tooth angle and number of cavities on the long slope of the saw tooth. Additionally, small array test sections also di ered in the number of rows of cavities along the transverse length of saw-tooth. In only one type of small array test sections, the 49 Figure 2.14: Illustration of gold eutectic bonding process used for bonding the small array saw-tooth ratchets to the 3 inch heater wafer (a) X-ray image of bonded 24 small array silicon saw-tooth test section with 3" heater wafer. Voids appear dark in the image. The dark square spots in the image are the re-entrant cavities. No voids in the bond were noticed. Some cavities appear lighter since the solder melts and flows into the cavities partly. (b) X-ray image of bonded 31.5 small array silicon saw-tooth test section with 3" heater wafer. No voids in the bond were noticed. A few smaller cavities appear to be filled partly by the solder Figure 2.15: X-ray image showing the bond quality of gold eutectic bonding process 50 number of rows of cavities were reduced from four to one to reduce the nucleation activity. Characteristics of all the test devices built are summarized in Table 2.2 Table 2.2: Characteristics of di erent types of test sections fabricated Test section parameters Large array test section Small array test sectionA B A B C Number of saw teeth 80 80 7 7 7 Saw tooth angle 24 31 24 31 31 Number of cavities along the long slope of the saw tooth 1 2 1 2 2 Number of rows of cavities along the transverse length of saw tooth 8 8 4 4 1 Foot-print area (mm x mm) 80 x 20.32 80 x 20.32 8 x 11.75 8 x 11.75 8 x 11.75 51 Chapter 3 Terrestrial Experiments 3.1 Experimental Setup The silicon test device used for the large array, terrestrial, pool boiling experiments shown in Fig. 3.1a, is made of two layers - the asymmetric saw-toothed heat sink, and a serpentine aluminum heater. The heat sink has an asymmetric saw tooth cross-section with a 24 90 66 pro le and 1 mm pitch. The heat sink constitutes 80 such saw-teeth spanning a foot print of 80 mm x 20.5 mm. The asymmetric cross-section of the saw teeth was obtained using a combination of novel gray-scale photolithography and Deep Reactive Ion Etching (DRIE). The long, shallow slope of each saw tooth is structured with re-entrant cavities, or vapor trapping sites, that aid in triggering bubble nucleation. The pyramidal re-entrant cavities, fabricated using a wet etching procedure, have a trapezoidal mouth of size ranging between 200 250 m. An SEM image of the cavities is shown in Fig. 3.1b. Each saw tooth has 8 cavities spaced equally along the ratchet on the long slope and in total the heat sink consists of 640 cavities acting as nucleation sites. Heat was provided using an aluminum serpentine heater shown in Fig. 3.1c. The voltage leads on the heater also aid in surface temperature measurements. The fabrication details of the asymmetric structure with re-entrant cavities and heater were discussed in the previous chapter. The bonded test section is soldered on to a printed circuit board for electrical connections and under- lled for structural integrity. The board also serves as the base of a transparent polycarbonate open channel which is 130.8 mm long, 25.4 mm wide and 7 mm deep, and mounted right on top of the test device as illustrated in Fig. 3.1d. The channel serves to con ne the volume of uid for realization of net liquid ow. The test board was suspended from the lid into a pool of FC-72 (C6F14), a highly dielectric highly wetting uid, contained in 52 Figure 3.1: Silicon test device Figure 3.2: Boiling chamber assembly 53 an aluminum boiling chamber (Fig. 3.2) of dimensions 33 cm x 23 cm x 22 cm (L x W x H). Properties of FC-72 are summarized in Appendix A. A thermistor and a pressure transducer were used to measure liquid temperature and absolute pressure respectively. A re ux coil condenser open to the atmosphere helped maintain atmospheric pressure inside the boiling chamber and condense the vapor. Prior to experiments, FC-72 was thoroughly degassed by boiling using the submerged cartridge heaters in the boiling chamber. The cartridge heaters were controlled using a PID controller to obtain the required pool temperature along with chilled water circulation through copper cooling coils immersed in the pool. Note that an earlier embodiment of the uid chamber, in the form of a closed loop was built, but had to be replaced in light of issues associated with vapor management in the small volume. Details of that ow loop are in Appendix B. The boiling chamber was also equipped with quartz windows for high speed imaging. A Phantom V310 high speed camera was tted with K2SC long working distance microscope lens for capturing bubble images. Also, to obtain di erent levels of magni cation close up objectives CF-1, CF-2 and CF-3 were tted to the lens. Images were captures at 3200 fps with a resolution of 1280 x 800. A 250 W halogen lamp was used as a light source. The test section heater was powered by an AMREL 220V programmable DC power supply. Data including the supplied current, voltage measurements from the test section heater, thermistor and pressure transducer was recorded using NI data acquisition system and LabView. The four slot NI data acquisition system chassis consisted of three data ac- quisition modules - NI 9201 for voltage measurements from the pressure transducer, NI 9219 (universal input module) for anemometry, heater voltage, and thermocouple measurements, and NI 9227 for measurements of current supplied to the test section heater by the DC power supply. 54 3.2 Results and Discussion Pool boiling experiments were conducted using FC-72 at atmospheric pressure. Exper- imental parameters tested include a heat ux range of 0 - 4.5 W=cm2 and pool subcooling ranging from 0 C to 20 C. After degassing the uid, the pool temperature was increased us- ing the cartridge heaters until the required subcooling was achieved. The applied heat ux was increased and once steady state was achieved data were acquired and high speed images of bubbles were recorded. This procedure was repeated for a cycle comprising increasing heat ux followed by decreasing heat ux. It is to be noted that the low heat ux range is mainly due to constraints in heater design and not limited by critical heat ux. The results from the large array experiments discussed in this section include bubble dynamics such as bubble growth, bubble departure frequency and diameter, heat dissipation characteristics of the surface, and liquid velocity measurements. 3.2.1 Bubble Dynamics using High Speed Photography Bubble dynamics were studied using the high speed images of bubble growth recorded during the experiments. The captured high speed images were processed and analyzed using NI Vision Assistant software to estimate bubble characteristics. Fig. 3.3 shows bubble departure from the re-entrant cavities on the shallow slope of the saw-teeth at a heat ux of 2.2 W=cm2. It was observed that the bubbles grew and departed at an angle normal to the slope of the ratchets. The angle of departure could be noted by observing the angled line of departed bubbles in the saturated pool at the left end of saw-tooth array. To further illustrate the bubble dynamics at the saw-toothed surface, Fig. 3.4 shows a close-up image of bubble growth and departure from re-entrant cavities on the shallow slope of the saw-teeth. As discussed earlier, it can be observed that the bubble growth is normal to the slope of the saw-teeth and the bubbles depart in a direction normal to the sloped surface at high velocities. This asymmetric growth with respect to the vertical axis, was consistent across the entire surface and at all tested conditions. The growth of 55 Angle of bubble departure Saw-toothed surface Figure 3.3: Image showing bubble departure from re-entrant cavities at an angle normal to the sloped surface of saw-teeth during saturated pool boiling at a heat ux of q" = 2.20 W=cm2. the bubble continues until the buoyancy force overcomes the surface tension force to cause bubble departure. The departed bubble, as noticed in Fig. 3.4 transits in the pool at an angle that is approximately normal to the sloped surface. In the sections to follow, it will be demonstrated that such asymmetric growth of bubbles has the potential to impart a net lateral motion of liquid in the immediate vicinity using a semi-empirical momentum transfer model. 3.2.2 Bubble growth - Experiments Fig. 3.5 shows one complete cycle of bubble ebullition cycle, consisting of bubble nu- cleation, growth and departure from a re-entrant cavity at 2.0 W=cm2 and saturated pool conditions. The vapor bubble nucleating from the cavity is initially hemispherical in shape. The hemispherical bubble grows rapidly, entirely in a direction normal to the shallow slope of the saw-toothed surface and not laterally. In a naturally occurring nucleation site on a planar surface, bubble growth would be concentrated in both lateral and vertical directions. This 56 1 mm 22.188 ms 17.188 ms 10.625 ms 4.062 ms 0 ms 1 mm 1 mm 1 mm 1 mm Figure 3.4: Bubble growth and departure from re-entrant cavities at q" = 2:0W=cm2 and saturated pool conditions. Bubbles nucleating from neighbouring cavities are marked with a di erent color. Relative time stamps are marked at the bottom right of each image. 57 initial rapid bubble growth is inertia controlled where the driving potential for bubble growth is the temperature di erence between the thin superheated liquid layer (relaxation micro- layer) surrounding the bubble, and the vapor. During this initial rapid growth of the bubble normally into the pool, the bubble remains pinned to the cavity as the bubble foot-print remains constant while the dome becomes more spherical. With increasing time, the dome grows larger in size, although at a considerably slower rate as the microlayer depletes. The growth continues until the buoyancy force is large enough to overcome the surface tension force that holds the bubble. As the bubble begins to depart, the neck of the bubble narrows as it breaks and leads to bubble departure normal to the surface. It was also observed that the waiting time, the period between bubble departure and nucleation of the next vapor bubble, is less than 312.5 s which is the time interval between two successive image frames. An interesting observation from these bubble images was the shape of the bubble as it is attached to the surface. Vapor bubbles in highly wetting uids such as FC-72 are generally observed to be more spherical and possess a very low contact angle due to the low surface tension of the liquid. However, in all the bubble images captured in this study the bubbles are consistently \light-bulb" shaped and the contact angles appear to be very high between 60 90 . Similar bubble shapes (Fig. 1.17) were reported by Hutter et al. [34] during pool boiling of FC-72 on a plain silicon surface with cylindrical cavities. This high contact angle also suggests that the evaporation microlayer, which often exists as a thin liquid layer underneath a bubble, as shown in Fig. 1.16, is less signi cant or nearly absent . It is the evaporation of this microlayer which has been shown in the past to contribute signi cantly to the rapid bubble growth in the inertia controlled regime and high heat transfer rates involved in nucleate boiling [30, 51]. However, Demiray and Kim [36] studied the nucleate boiling heat transfer under a single FC-72 bubble and observed that the microlayer contribution is not signi cant compared to transient conduction and microconvection contributions, which also supports the earlier observation about the evaporation microlayer in this study. Similar conclusions were also drawn in [35] in pool boiling of FC-72. 58 Figure 3.5: Ebullition cycle of a bubble at q" = 1.6 W=cm2 and 21 C subcooling In the current study, bubble diameter was estimated from the captured images utilizing standard image processing techniques using NI Vision Assistant software. The main steps involved in processing a single captured frame is shown in Fig. 3.6. The raw image is rst enhanced using a Look Up Table (LUT) with a power (1/x) function to increase the contrast. This is followed by a ltering technique using a convolution theorem to highlight the details such as bubble edges. Next, image thresholding is performed to binarize the image using a threshold value of image intensity. In this process, the pixels with an intensity above the threshold are identi ed (these appear white in the image). This aids in eliminating the noise emanating from light re ection such as at the center of the bubble. An image mask is applied to remove the saw-tooth from the images. A clear image of the saw-tooth without nucleation is selected to be used as a mask. This step is followed by other ltering techniques such as removing smaller particles and regions near the border. The hole in the bubble due to image thresholding is lled and the nal image in the process is used to calculate projected bubble area and perimeter among other parameters. Using the projected area of the bubble from the nal step of image processing shown in the gure, bubble diameter of the non-spherical bubble is calculated as the equivalent diameter of a projected circular disk with the same area as that of the bubble. This estimate of diameter for non-spherical entities is also known as the Waddel disk diameter which can be expressed as D = 2 r A (3.1) 59 This process is repeated for hundreds of frames involved in a single bubble growth cycle and multiple bubbles at di erent experimental conditions. Figure 3.6: Image processing steps of a single bubble frame at 178.44 ms in Fig. 3.5 The bubble diameter estimated using this process during a single growth cycle is shown in Fig. 3.7 as a function of time. As described earlier, it could be noted that the growth of bubble involves a rapid initial inertia controlled growth phase right after nucleation followed by a slower heat transfer controlled growth phase leading to departure. This observation agrees with the results reported in studies by Ramaswamy et al. [33] and Hutter et al. [34] for FC-72, and Lee et al. [32] for R-11 and R-113. Using a linear curve t for the inertia controlled growth and a power law t for heat transfer controlled growth for the single bubble shown in Figures 3.5, the asymptotic bubble growth rate relationship between bubble diameter and time can be obtained. For the inertia controlled growth, the bubble diameter was observed to scale linearly with time expressed as [24, 25], D = At (3.2) where A = 2 7 hlv v T lTsat 1=2 This linear relationship is similar to the experimental growth rate reported in a number of other studies in the literature for the inertia controlled regime . In the heat transfer controlled regime, the asymptotic relationship is of the form, D tm where, and m 60 D 1.43t 0.24D 150.8t Figure 3.7: Estimated bubble growth for images shown in Fig. 3.5 at q"=1.6 W=cm2 and 21 C subcooling are curve tting parameters. The parameter is also termed as a growth constant [27, 29] for the heat transfer controlled regime which is given as a function of wall superheat and uid properties. The bubble growth expression and growth constant as reported in Cole and Shulman [29] is, D = 4 p t (3.3) where, = cp Ja c = experimentally determined constant Ja = Jacob number = (Tw Tsat)Cp;l l vh lv In the above equation, the value of c was experimentally determined to be 1, =2, p3 by Fritz and Ende [52], Forster and Zuber [53], and Plesset and Zwick [26] for bubble growth in a uniformly superheated liquid. 61 The value of m for the experimental conditions shown in Fig. 3.7 is 0.24 which is much lower than the t1=2 relationship reported by a large number of studies in the literature for water [25, 29, 54] and for FC-72 [33]. In a recent study conducted with FC-72 and structured cavities in silicon, Hutter et al. [34] reported values of 0.37 and 0.53 for m, and in studies with saturated R-11 and R-113 Lee et al. [32] observed m to be between 1/5 and 1/3. The lower value of m reported in the current study could be due to the lower liquid equilibrium superheat, 0.1 K for the conditions shown in Fig. 3.7 , experienced due to larger cavities. In comparison, an equilibrium superheat of 0.92 K was reported by Hutter et al. [34]. The equilibrium liquid superheat is the temperature rise of the liquid above the saturation temperature (Tle Tsat) which is a requirement for vapor nucleation to be sustained. It is de ned as, Tle Tsat = 2 Tsat vhlvrc (3.4) where, rc =cavity mouth radius Fig. 3.8 shows the e ect of heat ux at a liquid subcooling of 21 C. As the heat ux increases, the wall superheat increases which vaporizes the liquid near the bubble interface at a faster rate and hence the bubble growth rate increases with heat ux. Also as the wall superheat increases, surface tension of pure uids which holds the bubble on the heated surface decreases. Additionally, since the bubble grows at a faster rate, at a given time the bubble size is larger at a higher heat ux thereby leading to increased buoyancy force which is responsible for bubble detachment. This increasing magnitude of buoyancy forces and the decreasing magnitude of surface tension forces, lead to faster bubble detachment, or smaller overall growth periods as heat ux increases which is evident in Fig. 3.8. These observations corresponding to the in uence of heat ux, and hence wall temperature, on growth rate accord well with studies conducted especially with FC-72 such as Hutter et al. 62 [34] and Ramaswamy et al. [33]. The increase in rate of bubble growth with increasing heat ux was also noticed in the inertia controlled regime which is supported by the increasing value of A in the linear relationship, D = At. The experimental values of A for di erent heat ux values are summarized in Table 3.1. From Fig. 3.8, it could also be noticed that, at a constant subcooling of 21 C, the transition from inertia to heat controlled regime occurs within a narrow range of bubble diameter, 0.45 to 0.55, and observed to be una ected by heat ux. Similarly, when the liquid subcooling is decreased (increasing pool temperature), the wall superheat increases leading to faster bubble growth rates as seen in Fig.3.9 for a nominal heat ux of 1.0 W=cm2. But unlike Fig. 3.8, with decreasing subcooling the inertia controlled regime was sustained for higher bubble diameters which could be attributed to the increasing thermal boundary layer thickness with decreasing subcooling. The e ect of independent parameters - heat ux and subcooling, on the curve tting parameters and m in the asymptotic relationship for heat transfer controlled regime can be studied from Figures 3.8 and 3.9, which are summarized in Table 3.1. In Table 3.1, the units for ?A? and ?B? are mm=s and mm=sm, if the diameter is expressed in mm and time in s. The value of m ranges between 0.13 and 0.27 without being a ected much by heat ux or subcooling, and it is also similar to the range reported by Lee et al. [32]. The value of was found to increase with increasing heat ux and decreasing subcooling and hence , increases with increasing wall superheat which agrees well with the observation by Scriven [27] for bubble growth in a uniformly superheated liquid (Equation 3.3). Similarly, in the experiments conducted by Cole and Shulman [29] with R-113, it was observed that the growth constant increased with increasing wall superheat. The values of c used in Equation 3.3 using the uniform superheat theory signi cantly over estimated the growth constant. Cole and Shulman [29] reported better predictions of growth constant by using a modi ed Jacob number with the wall superheat approximated as 0:5(Tw Tsat), which is valid since 63 D a1 1.72t 0.24 D a11.43t 0.24 D a11.09t 0.19 D a1 1.11t 0.27 D=At A 2.19 W/cm2 = 175.36 A 1.60 W/cm2 = 150.79 A 1.33 W/cm2 = 94.46 A 1.07 W/cm2 = 48.03 Figure 3.8: E ect of heat ux on bubble growth at a liquid subcooling of 21 C D a1 1.11t 0.27 D a11.22t 0.14 D a11.95t 0.21 D=At A 0 oC = 181.32 A 10.92 oC = 165.96 A 21.04 oC = 48.03 Figure 3.9: E ect of subcooling on bubble growth at a nominal heat ux of 1.0 W=cm2 64 the liquid superheat away from the heated wall approaches zero. Cole and Shulman [29] reported a modi ed equation for growth constant ( ) as, = 54Ja34 (3.5) In comparison, the experimental values of growth constant from the current study are plotted in Fig. 3.10 as a function of wall superheat. It can be observed that the modi ed equation (Eq. 3.5) for bubble growth near heated surfaces reported by Cole and Shulman [29] still over-estimates the growth constant. The equation for growth constant obtained using curve tting the experimental data from the current study can be expressed as, = 0:034Ja (3.6) Table 3.1: Summary of curve tting parameters from Fig. 3.8 and Fig. 3.9 Experimental conditions Bubble growth regime Fig. #Inertia controlled Heat transfer controlledD = At (mm) D = tm (mm) Tsub q00 A m ( C) (W=cm2) (mm=s) (mm=sm) 21.04 2.19 175.36 1.72 0.24 Fig. 3.8 21.04 1.6 150.79 1.43 0.24 Fig. 3.8 21.04 1.33 94.46 1.09 0.19 Fig. 3.8 21.04 1.06 48.03 1.11 0.27 Fig. 3.8, 3.9 10.92 0.94 165.96 1.22 0.14 Fig. 3.9 0 1 181.32 1.95 0.21 Fig. 3.9 Figures 3.11 and 3.12 also represent the e ect of heat ux and subcooling on growth rate but with overall asymptotic relationships for growth rate obtained by a power law curve t for both the inertia and heat transfer controlled regimes combined. The growth relationships are only slightly di erent from Figures 3.8 and 3.9 in the heat transfer controlled regime. In comparison with Hutter et al. [34], who conducted a similar analysis for bubble growth rate 65 30 40 50 60 70 80 900 5 10 15 20 25 30 35 40 45 50 Wall Superheat, Tw ? Tsat (oC) Growth Constant, ? Experimental ? Subcooling 0?20oC ?=0.034*Ja Eq. 3.3 with ?c=1 Eq. 3.5 Figure 3.10: Growth constant from experiments at a liquid subcooling of 0-20 C. The gure shows a linear curve t used to approximately express the growth constant. The equations 3.3 and 3.5 used for comparison from the studies Fritz and Ende [52] and Cole and Shulman [29] respectively, were calculated for saturated pool conditions. 66 in FC-72, observed the values of m to be in the range of 0.37 - 0.53 which is slightly higher than those reported in this study. 3.2.3 Bubble departure frequency and diameter As discussed earlier, bubble departure from a nucleation site is a balance between the surface tension and buoyancy forces, in the absence of forces due to liquid ow. As surface temperature increases, surface tension decreases and hence buoyancy forces tend to cause the bubbles to depart from the nucleation site earlier than at lower surface temperatures or wall heat ux, thereby leading to increased bubble departure frequencies or truncated bub- ble growth period. The average bubble departure frequency was estimated by calculating the average time period between a bubble?s nucleation and departure. As shown in Fig. 3.13, bubble departure frequency is a function of both subcooling and heat ux. As heat ux increases the bubble departure frequency also increases, although with a few anoma- lies. Similarly, the bubble departure frequency increases with decreasing subcooling and the highest frequency of 60 Hz was observed under saturated pool conditions and high heat ux. The frequencies observed under saturated conditions agree well with other pool boiling experiments conducted with FC-72 [14, 34]. However, in those studies, the bubble departure diameter was not observed to be a ected by subcooling or heat ux as seen in Figures 3.11, 3.12 which was unexpected. In other pool boiling studies conducted with FC-72 [14, 17, 33, 34], the bubble departure diameter was observed to increase with heat ux, or wall superheat. In [17], the bubble departure diameter increased between 0.26 mm to 0.38 mm over a heat ux range of 0 - 40 W/cm2. However, in the current study the range of heat ux tested was very low, and this could be a reason for the absence of any trend in the variation of bubble departure diameter. 67 D a11.07t 0.24 D a11.39t 0.22 D a11.63t 0.23 D a11.08t 0.18 Figure 3.11: E ect of heat ux on bubble growth at a liquid subcooling of 20 C D a11.23t 0.14 D a11.07t 0.24 D a12.03t 0.22 Figure 3.12: E ect of subcooling on bubble growth at a nominal heat ux of 1.0 W=cm2 68 0 0.5 1 1.5 2 2.5 30 10 20 30 40 50 60 70 80 Wall Heat Flux, q" (W/cm2) Bubble Departure Frequency, f (1/s) Saw teeth angle = 24oCavity mouth radius = 96.7 ?m?Tsub=21.04oC ?Tsub=16.04oC ?Tsub=10.92oC ?Tsub=5.31oC ?Tsub=0oC Figure 3.13: E ect of subcooling and heat ux on bubble departure frequency 0 0.5 1 1.5 2 2.50.75 0.8 0.85 0.9 0.95 1 Wall Heat Flux, q" (W/cm2) Bubble Departure Diameter, D d (mm) Saw teeth angle = 24oCavity mouth radius = 96.7 ?m?Tsub=21.04oC ?Tsub=16.04oC ?Tsub=10.92oC ?Tsub=5.31oC ?Tsub=0oC Figure 3.14: E ect of subcooling and heat ux on bubble departure diameter 69 3.2.4 Bubble growth - Comparison of experimental data with models The experimental results presented in this study show that the bubble growth rates are dependent on the regime of growth. Models developed in the past to predict bubble growth rates, in general, are also regime speci c and universal analytical solutions are only available for very speci c experimental conditions. As discussed earlier models for bubble growth can be classi ed into two types - one developed for the growth of a spherical symmetric bubble in a uniformly superheated pool and the other for bubble growth near heated walls where the temperature eld is non-uniform. Models for inertia controlled regime are very limited [24, 30] and most of the models developed apply mainly for the heat transfer controlled bubble growth. Models developed for the heat transfer control regime are often empirical in nature [27{29] and depend on the power law relationship of bubble growth with time (D t1=2). However, no analytical models are available for experimental conditions similar to those in the current study such as bubble growth in highly wetting uids on surfaces with structured cavities. Bubble growth for such conditions is often expressed as an asymptotic relationship of bubble diameter as a function of time. Analytical models and empirical relationships discussed in the literature that are used for comparison with the current study are summarized in Table 3.2 Table 3.2: Summary of available models and experimen- tal data for bubble growth during inertia controlled (IC) and heat transfer controlled (HTC) regimes. Model Conditions Bubble growth correlation Rayleigh [24] Uniform super- heat theory D = At where A = 2 7 hlv v T lTsat 1=2 (3.7) 70 Model Conditions Bubble growth correlation Mikic and Rohsenow [31] Non-uniform temperature eld D = 4Ja p3 ltp ( 1 Tw T1T w Tsat " 1 + twt 1=2 t w t 1=2#) where tw = 14 l 8 < : rc erfc 1 h Tsat T1 Tw T1 + 2 Tsat(vv vl) (Tw T1)hlvrc i 9 = ; 2 rc = cavity mouth radius (3.8) Van Stralen [30] Bubble growth due to evap- oration and relaxation mi- crolayer near heated surfaces. Applicable for both IC and HTC. D = D1 (t)D2 (t)D 1 (t) +D2 (t) D1 (t) = 1:633 vu ut vhlv (Tw Tsat) exph (t=td)1=2i lTsat t D2(t) = 3:9088 ( b exp " t td 1=2# + T1 TsatT w Tsat ) Jap lt+ 0:746 1=6 Pr l( exp " t td 1=2# + T1 TsatT w Tsat ) Jap lt b = 0:6964 D2 (t)Jap lt 0:1908 1=6 Pr l (3.9) Cole and Shulman [29] Non-uniform temperature eld D = 5Ja3=4p t (3.10) 71 Model Conditions Bubble growth correlation Zuber [28] Non-uniform temperature eld D = c 4 Jap t 1 qw p t 2k(Tw Tsat) (3.11) Lee et al. [32] Non-uniform temperature eld - R11 and R113 R+(t+) = 2 t+1=5 tanh(t+1=5) +R+0 where R+ = R=Rc, t+ = t=tc Rc = p27 2 Ja r lRd tc = 94Ja lRd , = tting parameters R +0 = dimesionless critical radius (3.12) Experiments with R11 and R113 D/t (IC) D/t1=2 (HTC) Ramaswamy et al. [33] Experiments with FC-72 on structured surfaces D/t (IC) D/t1=2 (HTC) Hutter et al. [34] Experiments with FC-72 on silicon with cylindrical cavities D/t1=2 72 0 0.005 0.01 0.015 0.02 0.025 0.03 0.0350 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 Time, t (s) Bubble Diameter, D (mm) Experimental ?Tsub=0oC; q = 1.9 W/cm2 Mikic and Roshenow ? Eq. 3.8Van Stralen ? Eq. 3.9 Lee et al. ? Eq. 3.12Cole and Shulman ? Eq. 3.10 Zuber ? Eq. 3.11Rayleigh ? Eq. 3.7 Figure 3.15: Comparison with existing models for bubble growth From Fig. 3.15, it can be observed that all the tested models deviate signi cantly from the experimental results shown for saturated conditions. Rayleigh?s equation (Equation 3.7) captures the high growth rate involved in the inertia controlled growth regime adequately. All the other equations for the heat transfer controlled growth, although they represent the shape of the bubble growth rate, deviate signi cantly from the experimental data. The model developed for refrigerant R-113 by Lee et al. [32] showed signi cant deviations although the asymptotic growth relationships represented in that study were similar to those reported in the current study. These results suggest that all the available models perform poorly for the experimental conditions tested in the current study. This re-emphasizes the need for further studies and models to analytically explain the e ects of structured surfaces and low surface tension uids. 73 3.2.5 Heat Dissipation Heat transfer characteristics of the surface can be analyzed by plotting wall superheat (Tw Tsat) against wall heat ux (qw), which is known as the boiling curve. Wall temper- ature was calculated from the test section heater resistance measurements using a linear relationship that was previously obtained by calibration. Calibration of heater resistance was performed in a convection oven using an NIST calibrated thermistor as a standard, by varying the oven temperature from 25 C to 85 C. Heater calibration data are presented in Appendix C. Saturation temperature was calculated at the measured pressure inside the tank using a pressure transducer. Heat ux was calculated as, q" = q qlossA st (W=cm2) (3.13) where, q = V I (W) Ast = Surface area of saw-toothed surface, cm2 V = Applied voltage, V I = Applied current, A Heat loss (qloss) was quanti ed by attaching a thermocouple to the back of the test device. Heat loss, which is heat transfer from the back side of the test device, was estimated by assuming 1D steady conduction across the device. It has to be noted that such an estimate of heat loss is only a conservative measure as it does not account for any spreading associated within the PCB. Fig. 3.16 shows the boiling curve at di erent liquid subcooling values. At a constant subcooling, as heat ux increases, heat transfer from the surface is by natural convection de ned by Newton?s law of cooling. As wall superheat increases high enough to trigger vapor nucleation from the cavities, the heat transfer increases sharply, marked by the change in 74 slope of the curve, without causing a signi cant increase in wall temperature. This can be attributed to the e ects of microconvection due to mixing induced by bubble growth and departure, transient conduction due to the constant disruption of the boundary layer and also due to the microlayer evaporation which is less signi cant in the current study owing mainly due to the shape of the bubble [6, 36]. As heat ux increases further, more nucleation sites become active increasing the heat transfer further. It has to be noted that the highest heat ux applied to the device was only 4.5 W=cm2, which was not limited by critical heat ux but only due to limitations of the designed heater, as discussed earlier. Also, at the highest tested heat ux the wall temperature was 92 C and observations of bubble dynamics showed that the boiling regime was still not fully developed suggesting that the CHF was much higher than the highest heat ux tested. This could also be attributed to the re- entrant cavities which prevents bubble coalescence by controlling the location of nucleation sites. Another important characteristic of boiling curve for a highly wetting uid such as FC-72 is the nucleation overshoot which is due to the ooding of cavities caused by the low contact angle of the the liquid. However, in the current study nucleation overshoot was not observed mainly because of the wide variation in cavity mouth radius. According to Hsu?s analysis [6], cavities with larger mouth radius are activated rst and as wall temperature increases cavities with smaller radius nucleate. This causes nucleation to happen in stages with increasing wall superheat and hence overshoot was not observed. As subcooling increases, pool temperature decreases causing increase in heat transfer in the natural convection regime. Also, in the partially developed boiling regime heat transfer is higher as subcooling increases. These observations accord well with the results of number of other studies in the literature. As heat ux increases, it could also be observed that the boiling curves at di erent subcooling tend towards merging with each other, as fully developed nucleate boiling is independent of both liquid subcooling and mass ux and mainly controlled by bubble ebullition cycle. 75 Increasing Subcooling Figure 3.16: E ect of subcooling on pool boiling curve for the test device with a saw-tooth angle of 24 3.2.6 Lateral Liquid Velocity Measurements Using Hot Wire Anemometer: Parametric E ects In the previous sections it has been demonstrated that asymmetry in the location of re-entrant cavities and asymmetry in surface structure can be used to cause bubble growth and departure at an angle normal to the slope of the surface. Using a semi-empirical model it was shown that such asymmetric bubble growth has the potential to impart an angular momentum to the surrounding liquid, thereby causing a net lateral ow. It was shown that in water, the asymmetric growth of bubbles can lead to liquid velocities up to 20 mm/s and in FC-72 the velocities are much lower at 1-2 mm/s due to higher liquid inertia associated with FC-72. In this section, experimental results of liquid velocity measurements over the saw- toothed surface using hot wire anemometry are reported. Hot wire anemometry involves a micro-wire or lm made of platinum or tungsten that is electrically heated to a constant 76 temperature. The heat convected from the probe due to liquid ow causes a change in cur- rent which in turn causes a change in voltage at the anemometer output. The measured voltage output is directly proportional to the liquid velocity (V). In the current study, a TSI Model 1750 constant temperature anemometer was used for the liquid velocity measurements. The anemometer setup includes a power supply, a probe, probe support and an angle adapter. An image of the anemometer along with the power supply is shown in Fig. 3.17a and b. Current supply to the probe and voltage measurements are made using a Wheatstone bridge circuit shown in Fig. 3.17c which is built in to the anemometer. The probe used in the experiments is a TSI Model 1210-20 straight probe with a platinum lm. The probe was held inside the polycarbonate ow channel as shown in Fig. 3.18 using a Model 1152 angle adapter which aids in positioning the probe inside the ow channel horizontally. The angle adapter is connected to a probe support which is connected to the anemometer power supply connection pins. One of the pins in the anemometer is connected to a control resistor which is used to set the operating temperature of the probe. The resistance of the control resistor is calculated based on the prescribed operating resistance for the probe and the cable resistance. The operating temperature of the probe 1210-20 is 67 C and the operating resistance is 48 Ohms. Based on these speci cations the control resistor used has a resistance of 34 Ohms. The selection of control resistor is explained in the manual [55]. The bridge output from the anemometer ranging between 0-5V is connected to the data acquisition system. The assembly of the probe setup is shown in Fig. 3.18. The probe is positioned 4mm over the saw-toothed surface and midway along the length of the test device facing the direction of liquid ow. The hot lm of the probe is enclosed in an aluminum sleeve with a slit at the top. The aluminum sleeve serves to prevent the contact of the vertical convection currents from the surface with the probe, so that the probe measures only the horizontal component of velocity. The heat transfer from the probe (hot wire/ lm) maintained at a constant tempera- ture, apart from liquid velocity, is also strongly a function of liquid temperature and liquid 77 (a) TSI Constant Temperature Anemometer (Model 1750) (a) Power supply used with the anemometer (Model 1751) (c) Wheatstone bridge circuit used for voltage measurement (d) Block diagram of the anemometer showing the pin connections Voltage Output Figure 3.17: (a) and (b) TSI constant temperature anemometer and power supply used for liquid velocity measurements. Electrical circuit and connections are shown in (c) and (d) [55] 78 (b) Exploded view of the probe assembly. Probe assembly includes the Model 1210-20 hot wire probe, Model 1152 angle adapter, and a sleeve 4 mm (c) An image showing the position of the probe over the surface in the experimental setup (a) Illustration of the anemometry setup inside the flow channel. Image inset shows a close-up illustration of the hot wire probe enclosed in an aluminum sleeve Aluminum sleeve slitted at the top Probe support with electrical connector Saw-teeth surface Hot wire probe Aluminum sleeve Hot wire Figure 3.18: Illustration of the probe set up over the test device with a saw-teeth angle of 24 79 properties. Hence for a constant liquid and operating temperature of the probe, the voltage output of the anemometer is primarily a function of liquid velocity and liquid temperature. To estimate the velocity of the liquid from the bridge output of the anemometer, the probe was calibrated in FC-72. A metered magnetic drive gear pump was used to pump the liquid upwards through a vertical tube in which the probe was placed. The calibrated velocity ranges from 1 - 50 mm/s at room temperature. By plotting output voltage against velocity, a fth order polynomial expressing velocity as a function of output voltage was obtained by curve tting. Calibration data for the hot wire probe are presented in Appendix C. The data from the anemometer was also corrected for actual temperature of liquid in the experiment by using the expression, E2cor = E2meas CF CF = Ts Te;calT s Te (3.14) where, Ecor = corrected output of the anemometer, V Ecor =measured output of the anemometer, V CF =correction factor Ts = lm operating temperature, C Te =liquid temperature during the experiment, C Te;cal =liquid temperature during calibration, C Experiments were conducted at a pool subcooling range of 0-20 C and a heat ux range of 0 - 4.5 W=cm2. For a constant liquid subcooling, power to the test device was increased periodically and the bridge output from the anemometer was measured after steady state conditions were achieved. This step was repeated while decreasing the applied heat ux and for experiments at other liquid subcooling values. Figure 3.19 shows the raw and temperature corrected bridge output of the anemometer. The output of the anemometer showed a lot of uctuations due to two reasons. One, was 80 due to operation at the lower limit of hot wire anemometer?s output. The lower limit of velocity measurement using a hot wire anemometer is 10 mm/s for liquid ows. Secondly, the uctuations are also caused by contact between departing vapor bubbles and the probe. The voltage measurements shown in Fig. 3.19 represent the peak amplitude of the recorded data.Using the temperature corrected voltage, liquid velocity can be measured using the calibration equation which is represented in Fig. 3.20. It can be noticed that at high subcooling, the measured velocity increases with heat ux thus indicating the pumping potential although at low velocities. However, at low subcooling the velocity was observed to be independent of heat ux and the measured velocities were very low. This net horizontal velocity component increased slightly at higher subcooling but increased consistently with heat ux. 0 0.5 1 1.5 2 2.51 1.1 1.2 1.3 1.4 1.5 1.6 Wall heat flux q??, W/cm2 Anemometer Voltage, V Saw teeth angle = 24oSubcooling = 21.04oC Raw HWA OutputTemperature Corrected HWA Output Figure 3.19: Raw and temperature corrected voltage measurements using hot wire anemome- ter over the test device with a saw-tooth angle of 24 81 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 50 1 2 3 4 5 6 7 8 9 Heat Flux, W/cm2 Velocity, mm/s ?T sub=0oC?T sub=5.31oC?T sub=10.92oC?T sub=16.04oC?T sub=21.04oC Figure 3.20: Temperature corrected liquid velocity measurements using hot wire anemometer over the test device with a saw-tooth angle of 24 3.3 Summary Pool boiling experiments were conducted on asymmetric silicon heat sinks using FC- 72. The independent parameters controlled in the experiment include heat ux and liquid subcooling. In the experiments that were discussed in this section, the parameters measured include bubble diameter, bubble departure frequency, wall surface temperature and liquid velocity. High speed imaging was used to capture bubble images and image processing techniques were used to estimate the characteristics of vapor bubble such as shape, diameter, and departure frequency as a function of time, heat ux and inlet subcooling. Based on the observed bubble characteristics, bubble growth was characterized as inertia and heat transfer controlled and asymptotic relationships for bubble growth were presented as a function of time, for both inertia and heat transfer controlled regimes. Based on the wall temperature measurements and heat ux, pool boiling curves were presented which illustrated the heat dissipation characteristics of the surface. Net lateral liquid velocity due to asymmetric bubble 82 growth, measured using hot wire anemometry, showed potential pumping velocities of up to 6 mm/s. To further corroborate the results from hot wire anemometry, PIV experiments were conducted with FC-72 which are described in Appendix E. Due to light re ection from bubbles and non-availability of uorescent particles with densities similar to that of FC-72, results from PIV experiments were inconclusive about lateral velocity of liquid. 83 Chapter 4 Microgravity Experiments In pool boiling experiments conducted at 1g, it was observed that bubble growth and departure were normal to the sloped surface of the saw teeth. The buoyancy force acting on the asymmetrically growing and departing bubble, thereby, has a horizontal and a vertical component. The vertical component of buoyancy at 1g acting on a an asymmetrically growing bubble is less than the buoyancy force acting on a vertically growing bubble on a plain surface. In the previous sections, it was observed that this led to signi cant change in bubble dynamics. To further understand the role of buoyancy, it is important to study the bubble dynamics and the resulting lateral motion in the absence of buoyancy which is only possible under microgravity. Hence, microgravity experiments were conducted to study bubble dynamics during phase change on an asymmetric surface in the absence of buoyancy and its e ect on lateral velocity of liquid. The experiments were conducted aboard a Boeing 727 aircraft (Zero-g Inc.) (Fig. 4.1) carrying out parabolic maneuvers to achieve reduced gravity. The maneuvers were carried out on two separate days. Each day consisted of 40 consecutive parabolas with each parabola consisting of a 1.8g pull up, 0g and a 1.8g pull out with periods of 60 seconds for hyper-gravity ( 1:8g) and up to 17 seconds for microgravity ( g) as shown in Fig. 4.1b. Fig. 4.1c shows the recorded accelerometer readings during a single parabolic maneuver consisting of 1.8g, g, and 1.8g. The ight operations initiated and terminated at Ellington eld, Johnson Space Center in Houston, Texas. 4.1 Experimental Setup Fig. 4.2 shows the experimental setup used in the study. The silicon test device used for the pool boiling experiments, shown in Fig. 4.2a, is made of two layers - the asymmetric 84 (b) Parabolic maneuver of the aircraft and period of gravity regimes (c) Accelerometer data of the gravity profile (a) A NASA Zero-g flight on a 1.8g pull up maneuver 1.8g Figure 4.1: Zero gravity ight used for parabolic maneuvers - (a) A NASA Zer0-g ight on it upward ascent. Boeing 727 was used for the ight experiments (b)-(c) Gravity pro le and recorded accelerometer readings of the achieved parabolic maneuvers 85 saw-toothed heat sink, and a serpentine heater layer. The heat sink has an asymmetric saw tooth cross-section with a 31 90 59 pro le and 1 mm pitch (hereafter referred to as the 31 test section). The heat sink used for large array experiments with FC-72, shown in Fig.4.2a consists of 80 such saw-teeth spanning a foot print of 80 mm x 20.3 mm. Each saw tooth has a long slope that is structured with re-entrant cavities. The cavities have a trapezoidal mouth of size ranging between 50 100 m. Each saw tooth has 8 pairs of cavities spaced equally along the transverse length of ratchet on the long slope, with each pair consisting of a large and small cavity (marked as (a) and (b) in Fig.4.2a) spaced between the crest and the trough. In total, the heat sink consists of 1280 cavities acting as nucleation sites. Heat was provided using an aluminum serpentine heater fabricated on silicon. The voltage leads on the heater also aid in surface temperature measurements. The test device is mounted on a printed circuit board for electrical connections. The board also serves as the base of a transparent polycarbonate open channel which is 130.8 mm long, 25.4 mm wide and 7 mm deep, and mounted right on top of the test device as shown in Fig. 4.2b. The channel serves to con ne the volume of uid for realization of net liquid ow. The test board was suspended from the lid into a pool of FC-72 contained in an aluminum boiling chamber (Fig. 4.2c) of dimensions 33 cm x 23 cm x 22 cm (L x W x H). A thermistor and a pressure transducer were used to measure liquid temperature and pressure respectively. Prior to ight experiments, the uid was charged into the system and degassed thoroughly at the ground station. Once the uid was charged, the bellows on top of the boiling chamber were compressed and the valve on top of it was closed with pressure inside the chamber remaining at atmospheric conditions. During the ight experiments, the boiling chamber was completely sealed and any increase in pressure due to the boiling process was compensated by the expansion of rubber bellows, thereby maintaining ground atmospheric pressure within the boiling chamber. A similar experimental apparatus was used by Oregon State University research group for small array experiments with deionized water as working 86 uid. The test section used for small array experiments with water is similar in cross-section consisting of only 8 saw-teeth ratchets with a footprint of 8 mm x 11.3 mm. During the experiment, data from sensors were recorded using NI Compact DAQ, and a Phantom high speed camera tted with an In nity K2SC microscope lens and a CF2 close-up objective was used to record bubble images. The electrical equipment in the setup were grouped in to 110 V and 220 V equipment and power was drawn from separate sources on the ight. Emergency cut-o was setup for disconnecting both the 110 V and 220 V equipments from the power source. The other signi cant di erence between terrestrial and microgravity gravity experimental setup was that the entire boiling chamber compartment was covered by polycarbonate sheets to form a double containment to prevent the liquid from entering the aircraft cabin in the event of a leak. The assembled view of the experimental setup is shown in Fig. 4.3. Experimentation aboard the ight also required adherence to strict mechanical design guidelines which required a factor of safety of 4 while withstanding forces up to 9g at every joint and all the load bearing members of the experimental setup. Details of the experimental design and structural analysis were furnished in a report, Test Equipment Data Package (TEDP), that was submitted to NASA. The submitted version of TEDP is presented in Appendix F. 4.2 Results and Discussion The data reported in this section were collected from experiments on large and small arrays of saw-teeth ratchets with highly subcooled FC-72 and 10 C subcooled deionized water respectively (subcooling, Tsub, is the di erence between saturation and pool temperature, Tsat Tpool) at g ( 10 2 g). The following sections carry discussion of bubble dynamics such as bubble diameter, bubble sliding motion, heat dissipation and a model for the estimation of bubble sliding velocity. 87 Figure 4.2: Experimental setup for reduced gravity experiments - (a) 31 test device with two re-entrant cavities per saw-tooth (b) Assembled test device with the polycarbonate ow channel. Co-ordinate system shows the orientation of the test device with respect to the aircraft. (c) Boiling chamber used for the reduced gravity experiment. Bellows assembly with double containment is used to maintain a constant chamber pressure. 4.2.1 Bubble Dynamics At reduced gravity, due to lack of buoyancy it was observed that the bubbles reside on the heater surface and grow to several diameters larger than those in 1g, as previously reported by a number of studies discussed in the literature. In experiments with water, because of the small size of the heat sink the entire surface of the test device was covered by a single large vapor bubble of diameter as large as 8.5 mm. It was observed that the single large bubble was pinned to the edges of the test section and the bubble interface appeared undulated at all tested heat ux values. Smaller vapor bubbles nucleating from the cavities under the footprint of the large bubble, at an angle normal to the shallow slope of the surface, were observed as shown in the image inset shown in Fig. 4.4a. This phenomenon is only 88 Figure 4.3: Assembled view of the experimental structure used for microgravity experiments. Details of the individual components in the structure and the related structural analysis are provided in Appendix F. 89 possible by the presence of a thin liquid lm between the larger bubble and the surface; the presence of this thin liquid lm is supported by the low measured surface temperature of 113 C, which is well below the Leidenfrost temperature. Straub [37] discusses a similar phenomenon with circulating vapor inside a bubble which was attributed to the shear forces induced by thermocapillary convection at the interface. In the case of large array experiments with FC-72, the bubbles were = 6 times larger compared to 1g. Interestingly, in the large array experiments, the bubbles departed the nucleation site laterally and continued to slide at high velocities across the saw teeth and along the length of test device as shown in Fig. 4.4b and 4.4c. At a low heat ux of 0.5 W/cm2, the interface of sliding bubbles appeared smooth and hemispherical as observed in the time sequence of images shown in Fig. 4.4b, and no departure from the surface was observed. As the heat ux was increased the bubble interface exhibited an unstable behavior as the surface appeared wrinkled, corrugated and more spherical. The corrugated sliding bubbles lifted-o and hovered over the surface which could be attributed to the momentum associated with bubble coalescence. This was also reported in studies with R-113 by Lee and Merte Jr. [43]. Fig. 4.4c at time 325 ms shows a corrugated bubble hovering over the surface at a heat ux of 1.4 W/cm2. Lee and Merte Jr.[43] observed similar corrugated rough bubble surface which was to Rayleigh-Taylor instabilities. The smaller bubbles that appear in Fig. 4.4c resulted from bubble departure during the preceding 1.8g regime that remain suspended in the pool during the next microgravity regime. At 1.8g (Fig. 4.5) with FC-72, due to increased buoyancy forces, the bubble departure diameters (Dd;1:8g) were very small compared to microgravity or 1g, Dd;1:8g