Steam-Generating Device Based on the Decomposition of
High-Test Hydrogen Peroxide
by
MASSACHUSETTS INS
FERAS EID
OFTECHNOLOGSEP 0 1 2010
Master of Science in Mechanical Engineering
I
Massachusetts Institute of Technology, 2006
LIBRARIE,
Bachelor of Engineering in Mechanical Engineering
American University of Beirut, 2004
ARCHIVES
Submitted to the Department of Mechanical Engineering in Partial Fulfillment of the Requirements for the Degree of
DOCTOR OF PHILOSOPHY IN MECHANICAL ENGINEERING
at the
Massachusetts Institute of Technology June 2010
© 2010 Massachusetts Institute of Technology All rights reserved
A
Sign atu re of A u th or ... ...
Department of Mechanical Engineering May 19, 2010
Certified by ...
Accepted by ...
Carol Livermore Associate Professor of Mechanical Engineering Thesis Supervisor
David Hardt Chairman, Department Committee on Graduate Students
IUTE
by
Feras Eid
Submitted to the Department of Mechanical Engineering on May 19, 2010 in Partial Fulfillment of the Requirements for the Degree of
Doctor of Philosophy in Mechanical Engineering
Abstract
Microscale ejector pumps offer the potential for high flow rate pumping of gases, a functionality that is greatly needed in MEMS technology. These pumps have many additional characteristics, such as their simplicity of design and their lack of moving parts, which favor them over other state-of-the-art MEMS gas pumps. One of the challenges associated with driving ejector pumps, however, is providing a compact source of motive fluid. This fluid is the high-speed gas that drives the pumping action. The current thesis presents a MEMS device capable of generating steam at speeds suitable for driving an ejector pump in a compact fashion. To that end, the device utilizes the homogeneous catalytic decomposition of hydrogen peroxide. Analysis shows that a MEMS ejector pump driven by this device is capable of handling mass flow rates per unit pump volume on the order of 10-2 g/s/cm 3, which are two orders of magnitude higher than those of state-of-the-art MEMS gas pumps. In addition to pumping, the steam generator may also be used for microrocket thrust generation in micropropulsion applications.
In this thesis, the design, fabrication, testing, and successful demonstration of the MEMS steam generator are presented. The device consists of a mixing section for the peroxide and catalyst streams, a reactor section where the peroxide decomposes, and finally a nozzle section where the gaseous products of the decomposition are accelerated to the required velocities. To design the device, multidomain (chemical, thermal, and fluidic) numerically-implemented modeling is used to study the underlying physics and arrive at an optimized, microfabricatable design. The modeling takes into account the key challenges of thermal management, achieving fast mixing, and boundary layer compensation. The device is then fabricated from a stack of four silicon wafers and one Pyrex wafer using deep reactive ion etching and wafer bonding. The modeling also guides the design of a mica-based ceramic package which provides both thermal insulation and piping ports. The system is then experimentally tested using high-test hydrogen peroxide and ferrous chloride tetrahydrate solution as the catalyst. The overall initial peroxide mass fraction is
Successful performance is demonstrated via the full decomposition of the peroxide and the complete vaporization of the water produced. The experimental results are also compared with those from the simulation. Good agreement is observed between experiment and theory, providing comprehensive model verification. The realization and demonstration of this steam generator promise significant enhancements in MEMS technology, particularly in the fields of gas pumping and micropropulsion.
Thesis Supervisor: Carol Livermore
I feel very lucky for having had such an enjoyable and rewarding PhD experience. I owe a lot of that to the support of my family and friends, but also to the good fortune of being part of an amazing research group ever since I joined MIT. My supervisor, Prof. Carol Livermore, has been a great mentor and friend. I cannot thank her enough for her continuous support and guidance, for always being available to discuss any questions or concerns I had, and for her constant cheerfulness and understanding, which can make a world of difference in a student's graduate school experience. My labmates have also been
an integral part of this experience. Thank you Frances, Gunjan, Nader, Lei, Aalap, and Eric for all the fun outings, get-togethers, and surprise birthday parties. I will definitely miss those times! Thanks in particular to my officemates Frances, Gunjan, and Lei for all the conversations and laughs, and for their willingness to listen to my rants when the research was facing some obstacles.
I also thank Dr. Luis Fernando Veldsquez-Garcia for the immense technical help he provided during the different stages of the project. One aspect of this work that I particularly pride myself on is that the very first generation of devices that I fabricated worked as expected. This is largely due to Luis's helpful suggestions and willingness to answer my numerous questions, especially during the fabrication stage. I also thank my committee members, Prof. Jeffrey Lang and Prof. Evelyn Wang, for their helpful ideas during our meetings. My thanks also extend to our administrative assistant Natalie Weaver for her help in ordering parts that I needed for the project, to the MTL staff for fab training and consultations, to Kimberlee Collins for her CFD simulation of the mixer, to Frances Hill for her help in taking the SEM image, to Tyrone Hill for showing me the components of his test setup, and to Daniel Herrick, Andrew Kalil, and Steven Yuan for their help in ensuring safety during experiments. The project was financially supported by Prof. Livermore, the Mechanical Engineering Department at MIT (through teaching assistantships), and partially by DARPA, MDA, and AFRL.
I'd also like to acknowledge some friends outside MIT who have made my life during this period much more enjoyable: Joe, Amal, Lama, and Mounir, who have become my "home away from home," Siddarth for being a great roommate, friend, and movie companion, Levi for memorable gatherings and Thanksgiving dinners, and Andrew for many interesting conversations and for his help when I was changing apartments.
Finally, I'd like to acknowledge my parents, sister, two brothers, and my siblings' families, for their continuous moral support during all this time that I've spent abroad. Our
1 INTRODUCTION 14
1.1 Overview 14
1.2 Literature Review of MEMS Pumps 15
1.3 Ejector Pumps 18
1.3.1 O verview ... 18
1.3.2 O perating principle ... 18
1.3.3 M acroscale characteristics and applications... 19
1.3.4 M icroscal e ejector pum ps ... 20
1.4 Steam Generation from the Decomposition of High-Test Hydrogen Peroxide 23 1.4.1 O verview and advantages ... 23
1.4.2 Other examples and applications of high-test hydrogen peroxide decomposition... 25
1.5 Thesis Roadmap 28 2 MODELING AND DESIGN 31 2.1 Background and Challenges 31 2.2 Conceptual Design 34 2.3 Mixer Design 36 2.3.1 M ixer overview ... ... 36
2.3.2 M icrom ixing literature review ... ... 37
2.3.3 M ixer m odeling and design ... 39
2.3.4 CFD sim ulation of the m ixer ... 42
2.4 Reactor 42 2.4.1 Reactor overview ... 42
2.4.2 Reactor m odeling and design ... ... 43
2.5 Nozzle 53 2.5.1 Nozzle overview ... ... ... 53
2.5.2 N ozzle m odeling and design...53
2.6 Parametric Study 59
2.9 Analysis of an Ejector Pump Based on the Current Work.
3 FABRICATION
3.1 Overview
3.2 Wafer-Level Die Layout 3.3 Alignment Marks
3.4 Layer 1
3.5 Layer 3 3.6 Layers 2 and 4 3.7 Wafer Bonding
4 TEST RIG SETUP
4.1 Overview
4.2 Hydrogen Peroxide and Safety 4.3 Package
4.4 Test Rig Components 4.5 Component Passivation_
5 TESTING
5.1 Overview
5.2 Experiments with Mixer-Testing Devices
72 79 100 100 103 108 112 114 114 114
5.3 Experiments with Nominally-Designed Devices 116
5.3.1 Experimental conditions...116
5.3.2 Visual inspection ... 117
5.3.3 Refractive index analysis ... 120
5.3.4 Device wall temperature measurements ... 123
5.3.5 Effluent temperature measurements ... 125
5.3.6 Summary of the implications of the experimental results...130
6.1 Summary and Important Findings 131
6.2 Challenges and Future Improvements 134
6.3 Design Modifications for Future Applications 138
6.4 Concluding Remarks 140
APPENDIX A: Detailed Process Flow 141
APPENDIX B: Practices Followed For Successful Wafer Bonding 145
Figure 1.1. Schem atic of an ejector pum p ... 19
Figure 2.1. Conceptual design ... 35
Figure 2.2. Schematic of a mixer with dimensions ... 40
Figure 2.3. Differential reactor elem ent ... 48
Figure 2.4. Plot of the flow temperature along the reactor length for the design conditions, with inserts magnifying the vaporization stages... 52
Figure 2.5. Plot of the species' mass fractions along the reactor length for the design conditions, with an insert magnifying the region near the reactor inlet... 52
Figure 2.6. Schematic of the nozzle showing the known quantities at different locations... 54
Figure 2.7. Differential nozzle elem ent... 56
Figure 2.8. Nozzle width profile before and after boundary layer compensation with insert show ing the throat region ... 58
Figure 2.9. Plot of static and stagnation pressures along the nozzle for the design conditions ... 5 9 Figure 2.10. Plot of the Mach number along the nozzle for the design conditions... 59
Figure 2.11. Plot of peroxide mass fraction at the reactor exit versus initial peroxide mass fra ctio n ... 6 0 Figure 2.12. Plot of static and stagnation temperatures at nozzle exit versus initial peroxide m ass fractio n ... 6 1 Figure 2.13. Energy balance on the entire reactor ... 61
Figure 2.14. Plot of reactor pressure versus initial peroxide mass fraction ... 64
Figure 2.15. 3D m odel of entire device... 65
Figure 2.16. Schematic of a cross-section of the device at distance x along the flow direction, illustrating the dominant heat transfer mechanisms for the configuration in which the device is encased inside a package ... 68
Figure 2.18. Package design ... 71
Figure 2.19. O-ring gland design. Dimensions and tolerances are in mm. ... 71
Figure 2.20. Schematic for ejector pump analysis... 73
Figure 3.1. Schematic section-view of a microfabricated device ... 80
Figure 3.2. Process flow summary for transferring the alignment marks ... 83
Figure 3.3. "Alignm ent M arks" m ask ... 84
Figure 3.4. Rotated die-level zoom of the mask "Alignment Marks" ... 84
Figure 3.5. Zoorns of the wafer-bonding and the top-bottom alignment marks... 85
Figure 3.6. Process flow sum mary for Layer 1 ... 86
Figure 3.7. "Holes" mask with insert showing the complementary top-bottom alignment fe a tu re s... 8 7 Figure 3.8. Rotated die-level zoom of the mask "Holes" ... 88
Figure 3.9. Process flow summary for Layer 3 ... 89
Figure 3.10. "Deep Features" m ask... 90
Figure 3.11. Rotated die-level zoom of the mask "Deep Features"... 90
Figure 3.12. Process flow summary for Layers 2 and 4 (continued on next page)... 92
Figure 3.13. "All Features" m ask... 94
Figure 3.14. Rotated die-level zoom of the mask "All Features"... 94
Figure 3.15. Schematic of the fusion bonding process ... 96
Figure 3.16. Schematic of the anodic bonding step ... 97
Figure 3.17. Photograph of a microfabricated device with the nominal design... 98 Figure 3.18. Photograph of a microfabricated mixer-testing device having an extra outlet port in the bottom for fluid discharge, which in the fully-functional devices occurs through th e n o zzle ... 9 8
Figure 4.1. Peroxide storage in lab ... 102
Figure 4.2. Safety gear... 102
Figure 4.3. Machined package made of Rescor 914 ... 105
Figure 4.4. Modified package bottom half for facilitating thermocouple insertion... 106
Figure 4.5. Schematic explaining the difference in the thermocouple insertion methods between the original and modified designs of the bottom half of the package ... 107
Figure 4.6. Modified package bottom half for usage with the mixer-testing devices... 108
Figure 4.7. Schem atic of test rig setup... 110
Figure 4.8. Assem bled package, fittings, and tubing ... 111
Figure 4.9. Photograph of the test rig highlighting the main components... 111
Figure 5.1. Frame-grabs of the device during experiment 1 showing the effluent in the different stages: (a) startup, (b) intermediate transient period, and (c) steady state ... 119
Figure 5.2. Procedure used to perform the refracive index analysis on the effluent ... 122
Figure 5.3. Plot of the peroxide mass fraction at the reactor exit (from the simulation) and at the nozzle exit (from experiments 1 and 2) versus initial peroxide mass fraction... 122
Figure 5.4. Plot of the silicon wall temperature from the simulation and the experiments versus initial peroxide m ass fraction ... 124
Figure 6.1. Schematic of proposed modified setup with compressed air line for purging the device after experim ents ... 136
Figure 6.2. Photograph of the cracked bottom half of the package... 137
Figure 6.3. Modified design for the bottom half of the package, which allows the use of a m aterial w ith higher therm al conductivity... 137
Table 1.1. Comparison of the specifications of some MEMS gas pumps in the literature with
those of an ejector pump based on the current work ... 21
Table 2.1. Phases of H20 and H202 during the five reactor stages ... 44
Table 2.2. Variation of state variables during the different reactor stages... 46
Table 2.3. Ejector pum p param eters ... 78
Table 4.1. Comparison of the required package properties and those of Rescor 914 ... 104
Table 5.1. Conditions during the peroxide experiments on nominally-designed devices.. 117
Table 5.2. Recovery factors, predicted adiabatic wall temperatures, and measured effluent tem peratures for the three experim ents... 129
1 INTRODUCTION
1.1 Overview
The past two decades have witnessed a marked increase in the interest in micro-electro-mechanical systems (MEMS), both at the commercial and research levels. Accelerometers, pressure transducers, inkjet printer cartridges, chemical and flow sensors, and lab-on-chip devices for point-of-care medical testing are only a few examples of systems that have been enabled or strongly improved by MEMS technology [1]. This technology broadly refers to a wide range of fabrication methods that allow the mass-production of small-scale systems and components. Some of those fabrication methods include photolithography, physical and chemical material deposition, wet and dry etching for material removal, doping, chip and wafer bonding, and soft lithography (imprinting). Some of the materials used include silicon and its oxides and nitrides, Pyrex, and various metals and polymers. Apart from reducing the manufacturing costs due to mass production, MEMS technology exploits some scaling benefits that allow physics-dictated improvements in the operation of many systems upon downscaling. This technology can have a significant impact on many fields such as the automotive industry, the electrical appliance market, biotechnology, and healthcare. The potential impact has sparked a strong interest in, and funding of, MEMS research by many agencies to further develop this technology.
A significant portion of the current MEMS research is centered on developing integrated systems having many components that work in tandem. For example, some
components of a MEMS power converter may include a combustor, a generator, and various pumps and valves. In many cases, an individual component with a specific function is often required by different systems that perform different tasks. Because of this overlap, small-scale devices are sought that can deliver certain functionalities currently lacking in the MEMS field. One of those much-needed functionalities is the pumping of gases at high flow rates.
1.2 Literature Review of MEMS Pumps
The majority of MEMS pumps that have been designed and/or demonstrated so far handle liquids, not gases [2]. Liquid pumping finds many applications in the MEMS field, such as fuel and propellant supply in microcombustors, microthrusters, and microengines [3], diagnostic sampling of bodily fluids [1], drug delivery [1], and water management in fuel cells [4]. Many MEMS liquid pumps have been reported in the literature. Some examples include turbopumps [1], elastomeric fluid pumps [5], electro-osmotic fluid pumps [6, 7], and a variety of reciprocating displacement pumps [8-13].
Gas pumping, on the other hand, has not received as much attention in MEMS research. Most MEMS gas pumps that have been demonstrated so far fall into one of two categories: displacement pumps or Knudsen pumps. Both of these kinds of pumps are discussed below. There are also pumps that utilize the entrapment and adsorption of gas particles, such as ion pumps. These pumps, however, consume the gas and are only good for vacuum generation, not for other pumping applications. Such pumps will not be addressed here.
Displacement pumps operate by exerting pressure on the working fluid in a periodic manner by means of a moving boundary [2]. In microscale pumps, this boundary is usually a deformable plate or diaphragm with fixed edges, as opposed to a piston in macroscale displacement pumps. The basic components of a displacement pump are the diaphragm-bounded pump chamber, inlet and outlet check or actively actuated valves, and an actuation mechanism. During operation, the externally-powered actuator displaces the diaphragm, causing the chamber to periodically expand and contract. Fluid is sucked into the chamber during expansion and then pushed out during contraction. The valves are oriented (in the case of check valves) or actuated in a way to favor suction at the chamber inlet and discharge at the outlet. Different diaphragm and valve actuation mechanisms have been reported and/or proposed for MEMS displacement gas pumps, including thermopneumatic [14], piezoelectric [15], electromagnetic [16], and electrostatic [17] actuation.
Knudsen pumps operate based on the principle of thermal transpiration of rarefied gases [18]. When two gas chambers are connected by a tube with a characteristic dimension much smaller than (one tenth or less of) the mean free path of the gas, free molecular flow exists. In this flow regime, the gas pressure and temperature ratios between the two chambers are directly related, and this conclusion can be derived by balancing the equilibrium molecular fluxes. As a result, varying the temperature between the two chambers leads to a pressure differential which can be exploited for pumping. Knudsen pumps are usually cascades of multiple, individually heated stages. Each stage has a capillary section in which the gas is rarefied, and a connector section with larger
dimensions in which the flow approaches the continuum regime. A temperature increase is imposed on the capillary section to produce a rise in the gas pressure by virtue of thermal transpiration. The temperature is then reduced to its initial value in the connector section, in which gas rarefaction is prevented due to the larger dimensions of this section. Due to the absence of thermal transpiration in this section, the pressure drop is smaller in magnitude than the pressure rise in the capillary section. This causes the gas to flow across a net pressure increase in each stage, producing pumping action. Various micropumps based on this operating principle have been reported in the literature [19-22].
Both kinds of gas pumps above suffer from certain disadvantages. The moving parts in a displacement pump diminish its robustness and shorten its lifespan. Some moving parts also require lubricants, which limit the use of these pumps in vacuum applications. Since these pumps require many parts (actuator, valves, and chamber), their fabrication can be challenging and dead volume can be an issue. Knudsen pumps do not contain any moving parts, but they have other problems, such as their inefficient use of energy [18]. In addition, Knudsen pumps' upper and lower pressures are usually limited, because the mean free path of the gas has to be kept well above the dimensions of the capillary sections and well below the dimensions of the connector sections for effective pumping. As the pressure is reduced, the mean free path increases, and vice versa. In many cases, extending each pressure limit separately requires functionalization and treatment of the pump's internal surfaces. Increasing both limits simultaneously is usually challenging if possible at all [18, 21]. For example, McNamara et al. [22] created a MEMS Knudsen pump for evacuating a small cavity in a micromachined structure. The pump discharged into the
atmosphere, but the lowest pressure it was capable of creating in the cavity was 0.46 atm, which is very high even for rough vacuum applications. Moreover, displacement and
Knudsen pumps have two major shortcomings in common: both require external power sources for operation (to drive the actuator of a displacement pump and to heat the capillary section of a Knudsen pump), which increase the overall pump size, and both have low pumping capacities.
1.3 Ejector Pumps
1.3.1 Overview
Microscale ejector pumps offer a promising solution to the high flow rate gas pumping requirement of many MEMS systems. Moreover, these pumps do not have the drawbacks and limitations of the displacement and Knudsen pumps that were discussed earlier.
1.3.2 Operating principle
The operation of an ejector pump is illustrated schematically in Figure 1.1. The purpose of the system is to pump a fluid, referred to as the suction fluid, from a low pressure Pio to a high pressure Phigh. To achieve this action, the pump mixes the suction fluid with a high-speed stream referred to as the motive fluid. The motive fluid is accelerated to a high speed before mixing to produce a low static pressure in the mixing region. The low pressure causes the entrainment of the suction fluid. The mixing results in momentum exchange: the suction fluid is accelerated and the motive fluid is slowed down until both are moving at a common velocity. The combined flow is then slowed down by
Chapter 1. Introduction
passing it through a de Laval nozzle. The nozzle allows pressure recovery by converting part of the kinetic energy of the combined flow into pressure rise. As a result, the flow exits at a pressure Phigh that is higher than Pion.
Suction
fluid at
pressure de Laval nozzle
Plow
Combined
flow at
pressure
Phigh> low
Figure 1.1. Schematic of an ejector pump
1.3.3 Macroscale characteristics and applications
Macroscale ejector pumps have been used historically for many purposes. Sample applications include the mixing and compression of gases to obtain fuels with specific heating values in petrochemical refineries, steam recirculation/recompression in the evaporators of power plants and chemical and food processing systems, condensate removal in heat exchangers, vacuum refrigeration and processing, and effluent pumping in chemical lasers [23-25]. These pumps are attractive because they contain no moving parts and require very little maintenance. They can operate at high efficiencies, and have been used at the macroscale to produce a wide range of pressures, from high vacuum to above-atmospheric [23].
1.3.4 Microscale ejector pumps
Microscale ejectors offer the same benefits of robust simplicity and no moving parts as large-scale ejectors, plus one scale-dependent benefit. For a suction fluid at given stagnation conditions, the mass flow rate that can be pumped through an ejector is, to first order, proportional to the nozzle's throat area. The weight of the device, on the other hand, scales with its volume. If a macroscale ejector is miniaturized while retaining the proportions between its dimensions, the ratio of the pumping capacity to the system weight scales inversely with length:
pumping capacity nozzle area 1
system weight volume length
Downscaling an ejector pump therefore allows for higher pumping capacity per unit system weight. This is true as long as the system is not scaled down to the point at which the thickness of boundary layers becomes comparable to the internal dimensions of the pump. At that point, the boundary layers can cause significant reductions in mass flow rate compared to the design value.
In microscale ejector pumps, the simplicity of design (e.g., no need for valves) and the robustness due to the absence of moving parts and fatigue-inducing mechanical cycling are significant advantages over displacement pumps. The potential for efficient operation and wider pressure ranges favor ejectors over Knudsen pumps. Most importantly, microscale ejector pumps are capable of handling higher mass flow rates per unit pump volume than both displacement and Knudsen pumps, while achieving moderately high pressure ratios. Multiple stages can be used to reach higher vacuum levels if desired. Table 1.1 compares the performance of some MEMS gas pumps in the literature, in terms of
pressure ratio and pumping capacity per device volume, to that of an ejector pump based on the current work. The ejector pump specifications that are listed in the last row of this table are derived in the analysis presented in Section 2.9.
Table 1.1. Comparison of the specifications of some MEMS gas pumps in those of an ejector pump based on the current work
the literature with
Research group Schomburg et aL [14] Kamper et aL [15] Bohm et aL [16] McNamara et aL [22] Our group Pump type Thermopneumatic displacement pump Piezoelectric displacement pump Electromagnetic displacement pump Knudsen pump Ejector pump (predicted) based on current work
Maximum gas flow rate (g/s/cm3 of pump volume) -7 10 -4 10 -7 10 Maximum pressure ratio 1:1.05 1:2.8 1:1.14 1:2 1:10 per stage
MEMS ejector pumps can thus potentially open up a wide range of opportunities for MEMS applications if properly designed and demonstrated. Examples of these applications include fuel and air feeds in microcombustors for power generation, and vacuum maintenance in certain MEMS systems such as gas chromatographs. Another important application is in pumping out the effluent of microscale chemical oxygen iodine laser systems, as discussed in Section 1.4.1.
Very few microscale ejector pumps have been demonstrated so far. One challenge in designing these pumps is providing a compact source for the motive fluid. Doms et al. [26] and Chuech et al. [27] reported bulk-micromachined ejectors that were fabricated by processing and bonding silicon and Pyrex wafers. The pump by Doms was tested with two motive fluids: compressed nitrogen gas, and water vapor that was obtained by heating liquid water outside the device. Chuech's pump was driven using air as the motive fluid. Fan et al [28] reported an electro-discharge machined miniaturized stainless steel pump with microscale features that used compressed air as the motive fluid. All of these pumps were capable of achieving suction flow rates in the range 10-?-10-s kg/s, which are relatively high compared to other MEMS gas pumps [2]. There are disadvantages, however, associated with the motive fluids that were used to drive the reported ejector pumps. Because of their low densities, gases such as air and nitrogen require large storage volumes. This poses challenges for keeping the pump compact, especially in portable systems. With pumps utilizing water vapor, more compact storage reservoirs are possible if liquid water is stored inside the system and later vaporized during operation. This approach, however, suffers from two significant drawbacks. One is the need for an external power source to vaporize the water, which increases the size of the system. The second is the integration of a vaporizing mechanism (such as an electric heater) into the system, which can complicate the design and the fabrication process.
One approach that eliminates the needs for both large gas-storage reservoirs and external power sources for liquid vaporization is the chemical generation of high-speed gases from liquid reactants, usually via exothermic decomposition. For example, liquid
hydrazine (N2H4) decomposes catalytically to produce ammonia, nitrogen, and hydrogen gases at elevated temperatures. This decomposition has been widely employed in propulsion applications to control and adjust the attitude of spacecraft, usually using an iridium-alumina catalyst [29]. The gaseous products of the decomposition can be used, in principle, to drive an ejector pump. Hydrazine's extreme toxicity and instability, however, would pose serious safety and health challenges if this approach were implemented in practice to generate the pump's motive fluid [30]. Storage is also a challenge because hydrazine reacts with atmospheric carbon dioxide to produce corrosive compounds [31]. Other concerns are the toxicity and polluting capabilities of the generated ammonia [32, 33]. These problems translate into significant increases in costs for such systems [34].
1.4 Steam Generation from the Decomposition of High-Test
Hydrogen Peroxide
1.4.1 Overview and advantages
Another chemical reaction that allows the generation of a high-speed gas mixture from a liquid precursor is the catalytic decomposition of high-test hydrogen peroxide, or HTP. HTP refers to water-diluted hydrogen peroxide mixtures in which the peroxide mass fraction is generally greater than about 70%. Hydrogen peroxide decomposes catalytically to produce water, oxygen gas, and heat, according to the reaction:
1
H202 -+ H20 +-02 + heat (1.2)
When the initial peroxide concentration in a peroxide-water mixture is greater than about 67%, the energy released by the above decomposition in an adiabatic container originally at room temperature is enough to vaporize all the liquid water present or produced. As a result, a high-temperature mixture consisting predominantly of steam (and containing some oxygen gas) can be generated. The mixture can then be passed through a nozzle to convert some of its thermal energy to kinetic energy. This method can be used to produce a high-speed flow suitable for driving an ejector pump.
The above approach offers the same benefits as using hydrazine, namely allowing a compact container and not requiring any external power sources for vaporization, with some advantageous differences. Unlike hydrazine, hydrogen peroxide is a nontoxic, or "green", substance [31]. Despite posing some health hazards which are discussed in Section 4.2, hydrogen peroxide's low vapor pressure (1.95 mm Hg at room temperature [35]) prevents HTP mixture fumes from easily entering the human body. Exposure to peroxide vapors emanating from an open HTP container at room temperature in a reasonably ventilated area is not lethal to humans [31]. Also, hydrogen peroxide does not react with the atmosphere, and can be stored for long periods of time in properly-designed containers. The decomposition of peroxide according to (1.2) produces nontoxic, environmentally-friendly products. All these features allow for increased safety and cost reduction in peroxide systems compared to others that use more hazardous substances [36].
Concentrated hydrogen peroxide has been used in many large-scale systems to generate steam for driving ejector pumps. One such system is the Thiokol Hyprox, which
was successfully installed and used in many facilities to generate vacuum for rocket testing
[31]. A very promising potential application of an ejector pump that uses this approach at
the microscale is the aspiration of a micro-chemical oxygen iodine laser (pCOIL) system as proposed by Wilhite et al. [37]. In this flowing-gas system, a 1.315 pIm wavelength laser output is produced from the stimulated emission of excited atomic iodine, which is obtained from the reaction of molecular iodine with singlet oxygen. The singlet oxygen is generated from the multiphase reaction of basic hydrogen peroxide (a mixture of H202 and KOH) and chlorine gas [38]. After laser emission, the low-pressure flow gas must be discharged to the atmosphere to maintain stable operation. Since hydrogen peroxide is already utilized in this application to generate the singlet oxygen, employing the above ejector approach to pump out the flow gas is very desirable since it allows for a limited-input portable p.COIL system.
1.4.2 Other examples and applications of high-test hydrogen peroxide
decomposition
Historically, highly-concentrated hydrogen peroxide has found its most widespread use in propulsion applications [31]. HTP can be used as either a monopropellant or bipropellant for rocket, torpedo, and submarine engines. As a monopropellant, hydrogen peroxide is decomposed catalytically to produce a high-energy mixture of steam and oxygen gas. This approach is very similar to that outlined in Section 1.4.1 for generating a motive fluid for an ejector pump. The main difference is that in propulsion, the high-speed exhaust mixture is used for thrust generation, whereas in an ejector the mixture is utilized for momentum exchange. As a bipropellant, the peroxide is used as an oxidizer, in
conjunction with a fuel such as kerosene. The peroxide decomposes, either under the action of a catalyst, or upon contact with the fuel itself, i.e. hypergolically, to produce oxygen which causes the fuel to combust. The exhaust gases from the combustion are then accelerated to provide thrust.
The first propulsion application of hydrogen peroxide dates back to 1933 Germany, when Hellmuth Walter proposed using HTP mixtures at 80%-82% concentrations as submarine monopropellants [31]. In the following years and during the World War II era, the use of hydrogen peroxide expanded to include the propulsion of torpedoes and rocket-assist devices for military aircraft. After the war, the technology moved from Germany to the UK, USA, and USSR, and hydrogen peroxide received a great deal of interest. Some famous applications included reaction control systems' thrusters (such as the X-1 and X-15 systems) that used peroxide as a monopropellant, and rocket engines (such as Black Knight-Black Arrow, the AR engine series, and LR-40) that used peroxide as a bipropellant oxidizer. This interest peaked during the 1950's and 1960's. Following this period, peroxide was gradually displaced by chemicals with enhanced performance, such as hydrazine in monopropellant applications and liquid oxygen and nitrogen tetroxide in bipropellant applications. Ever since the 1990's, however, peroxide has seen a renewed interest because of its previously-discussed advantages, such as its minimal environmental impact, non-toxicity, and relative ease of handling. Many labs, agencies, and companies are currently conducting heavy research on hydrogen peroxide-related systems. Examples of these organizations include Lawrence Livermore National Laboratories, NASA, Rocketdyne, Beal Aerospace, and Orbital Sciences Corporation [31].
Another important historical use of HTP has been as a gas generator for turbopumps [31]. Peroxide is attractive for this application since the amount of water in HTP mixtures can be controlled to produce decomposition temperatures below the operational limits of the uncooled blades. One of the most famous systems that used a peroxide-driven turbopump was the V-2 rocket which was developed toward the end of World War II in Germany.
One interesting naturally-occurring system which uses peroxide as an oxidizer and gas generator is the bombardier beetle [39]. As a defense mechanism, this beetle emits a jet of noxious, boiling chemical spray that can be fatal to insects and small creatures. The active components of this spray are benzoquinones produced from the oxidation of hydroquinone by hydrogen peroxide. The beetle has two separate glands that enable this mechanism: a large "reservoir" containing an aqueous hydrogen peroxide mixture (25% peroxide by mass) and hydroquinone, and a smaller "reaction chamber" containing enzymes such as catalase and peroxidase. When threatened, the beetle opens a muscle-controlled valve that forces the reservoir contents into the reaction chamber. The enzymes catalyze the peroxide decomposition and the subsequent oxidation of the hydroquinone by the generated oxygen gas. Both the decomposition and oxidation reactions are exothermic. These reactions release enough heat to bring the mixture in the chamber to near its boiling point, vaporizing about one fifth of it. Under the pressure of the vaporized gas and the remaining (unused) oxygen from the peroxide decomposition, the mixture is expelled explosively to the atmosphere through tips in the beetle's abdomen. The pressure causes the entrance valve to close during this process, thereby protecting the insect's internal
organs. The defensive spray is ejected cyclically at about 500 pulses per second, and is accompanied by audible "pop" sounds.
1.5 Thesis Roadmap
This thesis presents a MEMS device that generates steam (and oxygen gas) from the catalytic decomposition of hydrogen peroxide. The device's effluent can be used to drive a MEMS ejector pump. The different stages involved in developing the device are discussed, including the modeling, design, fabrication, and testing.
In this chapter, the need for high flow rate MEMS gas pumping was discussed, and ejector pumps were presented as an attractive option capable of delivering this functionality. After describing the operation of these devices and comparing their performance with other state-of-the-art pumps, the advantages of using hydrogen peroxide to generate steam for driving ejectors were discussed. The chapter concluded with a discussion of other examples in which the decomposition of high-test peroxide is utilized, in industrial and military applications and even in nature.
Chapter 2 describes the design of the steam generator. After a discussion of the challenges involved in designing this device, the conceptual design is presented along with ways to counter those challenges. Detailed numerically-implemented physical models are then developed to simulate the conditions in each of the different sections of the device. The modeling results are used to design those sections. In addition, a thermal model is developed to manage heat losses from the device, and it guides the design both of the device and of a thermally-insulating package. The chapter also presents a parametric study that was developed to examine the device operation under a range of conditions. The
chapter concludes with an analysis of an ejector pump based on the current device. This analysis quantifies the proposed pump's improved performance over the state of the art.
Chapter 3 describes the fabrication of the device. The wafer-level die layout is first discussed, along with the functions of the different die variations. The process flow of each device layer is then presented, including cross sections of the layers after the main processing steps, and the masks that were use for photolithography. The wafer bonding process is then described. The chapter concludes with photographs of two diesawed devices and cross-sectional SEM images.
Chapter 4 describes the task of setting up the test rig for experiments. The safety concerns that accompany the use and storage of hydrogen peroxide are first discussed. The machined package is then presented, and its material properties and design variations are discussed. This discussion is accompanied by photographs of the original package and its variations. The test rig components and overall setup are then described and displayed. The chapter concludes with a discussion of the technique used to passivate those
components in order to make them ready for peroxide use.
Chapter 5 describes the experimental work. It starts with a brief discussion of the experiments that were carried out using the mixer devices. The experiments on the nominal devices are then presented. The operating conditions and the different characterization methods, including visual inspection, refractometry, and temperature measurements are described. The experimental results are then presented and used to demonstrate successful operation. These findings are also compared with the modeling results of Chapter 2 to provide comprehensive model verification.
Chapter 6 summarizes the presented research. Some of the challenges that were encountered during the experiments are then described, and ways of alleviating those challenges for improving future tests are suggested. The chapter concludes with a discussion of some design modifications that will enable future use of the device in MEMS pumping and micropropulsion applications.
2 MODELING AND DESIGN
2.1 Background and Challenges
Despite the many advantages of using hydrogen peroxide for steam generation, a number of challenges exist, including the choice of catalyst, thermal management, boundary layer effects, and safety considerations. For devices that generate steam via this approach, thermal management and boundary layers are more important at the microscale than at the macroscale.
The type of catalyst used can greatly affect the success of the peroxide decomposition process. Typically, heterogeneous catalysts have been used in macroscale decomposition devices [40-42]. These catalysts are usually solid materials such as silver or manganese oxide. They are placed inside macroscale devices as meshed layers or pellet beds, and inside microscale devices as channel wall coatings. Such static catalyst layers, however, have a limited lifespan due to surface ablation upon repeated use [43]. The lifetime of silver catalyst, for example, ranges from a few minutes to a maximum of about 30 minutes when using ultra pure peroxide [44]. Solid catalysts also necessitate the use of rocket grade hydrogen peroxide which is relatively unstabilized. The use of stabilizers allows longer-term storage and easier handling of the peroxide, but causes the poisoning of solid catalysts. Furthermore, in microscale applications where very narrow, arrayed channels are required to increase the surface area of the catalyst, flow clogging can become a problem. Hitt et al. [45], in conjunction with the NASA Goddard Space Flight Center, created a prototype monopropellant MEMS thruster to decompose hydrogen peroxide
using a heterogeneous catalyst. The device failed to achieve complete peroxide decomposition and effluent vaporization. One of the causes to which this failure was attributed was the formation of gas bubbles that may have blocked some of the reactor channels. Using a homogeneous (liquid) catalyst [46] eliminates the limited lifespan issue and allows the use of highly stabilized peroxide without causing poisoning. It also allows wider reactor channels that are not as susceptible to clogging. To be able to use this kind of catalyst, however, one extra challenge has to be met. For fast and complete peroxide decomposition, the peroxide and catalyst streams must be well-mixed. This necessitates the design and addition of a proper mixer section to the device. Mixing is especially tricky at the microscale, where the laminar nature of flows usually makes diffusion the dominant mixing mechanism [47], which significantly slows down the mixing.
Two further challenges arise from scaling considerations. The first challenge is thermal management. The heat released from the peroxide decomposition is necessary to keep the reaction going and to vaporize the water produced. The rate of heat generation scales with the device volume (as in any homogeneous reaction). On the other hand, the rate of heat loss from the device scales with surface area. If a macroscale device that generates steam via the above approach is shrunk down while retaining the proportions between its dimensions, the ratio of heat losses to heat generation thus scales inversely with length:
heat loss surface area 1
ocx (2.1)
heat generation volume length
As seen in (2.1), downscaling makes the heat losses significant compared to the heat generated by the reaction. When the scale gets small enough, the energy remaining inside
the device will become insufficient for sustaining the reaction and vaporizing the water produced. This necessitates paying extra attention to thermal management to ensure successful device operation. Osaki et al. [48] designed a microthruster that decomposes hydrogen peroxide heterogeneously, and they found that the heat losses were so significant that an electric heater was needed for effective operation. It is suspected that the increased role of heat losses at the microscale may have also contributed to the failure of the device by Hitt et al. That device lacked any thermal insulation.
The second scale-dependent challenge is boundary layer formation. The thickness 6 of boundary layers inside the device channels scales with the square root of distance x in the flow direction. If a macroscale internal flow device is downscaled while retaining the proportions between its dimensions, the ratio of the boundary layer thickness to the width
w of the flow channel thus increases:
oc (2.2)
w w length
Downscaling can therefore cause boundary layers to occupy a large fraction of the flow field, leading to a significant reduction in the effective flow area [49, 50]. In devices designed to achieve supersonic flow, this reduction will prevent the flow from reaching the sonic point if the scale gets small enough. This can be seen by examining the dependence of Mach number M on the effective flow area A [51]:
dM2 1 dA
2- - - . (2.3)
M2 1-NM2 A
As the flow approaches sonic conditions (M = 1), the denominator of the right-hand side of (2.3) approaches zero. In this case, even very small area changes dA can cause large
deviations dM from the design Mach number. Therefore, proper compensation for the boundary layer presence has to be incorporated into the device design.
One final challenge arises from practical considerations. Despite being nontoxic, hydrogen peroxide is a strong oxidizer. This calls for careful safety provisions in any experimental setup where high-test peroxide is used, to prevent explosions and detonations.
In the current work, the catalyst, thermal management, and boundary layer challenges were addressed by implementing multi-domain physical modeling. This modeling was used to simulate the flow in different sections of the device and to evaluate the heat losses from the device. The results guided the design of both a MEMS device that decomposes hydrogen peroxide using a homogeneous catalyst and a package with sufficiently high thermal resistance to enable full peroxide decomposition and complete water vaporization. The model included the effects of boundary layers in the flow; these effects were compensated for in the design. Finally, by using a continually-supplied homogeneous liquid catalyst, the poisoning problem of heterogeneous catalysts was eliminated, at the expense of adding a mixer section for the peroxide and catalyst streams. Many safety measures for storing, handling, and experimenting with peroxide were taken during the testing stage, as described in detail in Section 4.2.
2.2 Conceptual Design
In the current work, steam is generated from the decomposition of hydrogen peroxide, and the reaction is facilitated by a homogeneous catalyst solution that is mixed with the peroxide inside the device. The exothermic decomposition first produces oxygen 34
Chapter 2. Modeling and design
gas and liquid water. The water is then vaporized by the heat generated, which leads to
all-gaseous products at the reactor exit. The products are subsequently accelerated to the required speed, or expanded to the desired pressure. To enable these functionalities, the device consists of three sections: a mixer for mixing the peroxide and catalyst streams, a reactor for decomposing the peroxide and vaporizing the water, and a nozzle for accelerating the gaseous products. A schematic of the device is shown in Figure 2.1.
Hydrogen peroxide reservoir
Reactor Nozzle
Mixer
Catalyst reservoir
Figure 2.1. Conceptual design
The choice of catalyst solution was based on optimizing different performance criteria, such as providing a strong and fast catalytic activity without containing large insoluble particles that could clog the device channels [43, 46]. The ratio of catalyst to peroxide flow rates was first guided by the findings of other groups that attempted to optimize the reactor length required for full decomposition [52]. Preliminary experiments were then conducted in which this ratio was slightly varied from the reported findings to determine the value that lead to optimal operation. This optimal value was used in the final device-characterization tests.
In the following sections, the models that were used to design the mixer, reactor, nozzle, and thermally-insulating package are described. Even though these models are presented sequentially, they are in fact interdependent, and iteration was required between them to arrive at the final design.
2.3 Mixer Design
2.3.1 Mixer overview
The device includes a mixer section to enable the use of a homogeneous catalyst. Homogeneous catalysts are supplied in solution form and make the catalysis process volumetric. For uniformity and fast decomposition in the reactor, it is desired to achieve thorough mixing of the peroxide and catalyst streams before the flow enters the reactor. Passive micromixing is challenging since it tends to be dominated by diffusion due to the laminar nature of microflows. This causes mixing to be slow and calls for long channels. Active mixing can be faster, but it necessitates external energy sources to stir and mix the flow.
In the current work, a fast passive mixer design (not more than a few millimeters in length) was sought that would allow a compact device without requiring any external energy sources for mixing. In addition, the mixer was designed to minimize the pressure drop across it in order to avoid high supply pressures. This is necessary because the interface between the device and the package consists of thin 0-ring seals, and a supply pressure higher than 5 atm could lead to the failure of these seals [53]. All pressures listed in this thesis are absolute pressures.
2.3.2 Micromixing literature review
Various groups have developed MEMS devices that can mix two or more flows. Some of these designs achieve fast mixing by having out-of-plane (3D) features, such as the split-and-recombine design by Branebjerg et al. [54] and the serpentine design by Liu et al. [55]. In these designs, fast mixing comes at the expense of a complicated fabrication process and a generally large pressure drop. Other designs have simpler fabrication, such as the T-shaped micromixer by Wong et al. [56], the Herringbone mixer by Stroock et al. [57], and the impinging-jets mixer by Yang et al. [58]. These designs, however, suffer from other problems when adapted to the current work. The Herringbone mixer and the standard T-shaped mixer are slow and require very long channels (with lengths on the order of centimeters and meters, respectively) for sufficient mixing. The impinging jet mixer is fast and compact enough (i.e. with a few millimeters long channel) only when one jet of peroxide and one jet of catalyst are used. This makes the design not very robust.
Despite its slow mixing rate, the basic T-shaped mixer has an easily fabricatable two-dimensional design that makes it very attractive from a MEMS-manufacturing point of view. Methods for enhancing this mixer's performance were thus sought. Mengeaud et al. [59] showed that zigzag channels provide better mixing than straight ones when the flow's Reynolds number exceeds 80, due to laminar recirculations that are generated at the zigzag angles. Wong et al. [60] showed that mixer walls with protrusions allow better mixing than smooth walls, due to the generation of eddies and lateral velocity components as the flow crosses these protrusions.
Furthermore, the Woias group [61, 62] showed that in a T-shaped micromixer, three distinct laminar flow regimes can exist, each having different mixing characteristics. The streamlines of the fluids to be mixed have different shapes at the mixer inlet for each of these regimes. At very low flow speeds, there is the "stratified flow" regime in which the streamlines are mostly straight and mixing is diffusion-dominated. At medium speeds, there is the "vortex flow" regime in which vortices start building up inside the channels. In this regime mixing is still primarily dominated by diffusion, but is slightly enhanced by the swirling motion that drags fluid from the middle to the top and bottom of the mixing channel. At high speeds, there is the "engulfment flow" regime in which the axial symmetry of the flow breaks up and the streamlines interweave and reach to the respective opposite half of the mixing channel. This causes a significant improvement in mass transfer and allows this regime to have much faster mixing than the other two. Quantitatively, the three regimes are distinguished by a dimensionless identification number K equal to the ratio of the channel's hydraulic diameter dh,mixer to the Kolmogorov length scale Ak, which is the scale of the smallest eddies in a turbulent flow:
K = dhmixer (2.4)
/1k
Conceptually, K is a measure of the free space for the growth of vortices. At low values of K, the fluid viscosity damps the starting of eddies, and mixing is slow, whereas at higher K values, the flow conditions allow for the formation and growth of eddies and vortices that
enhance mixing. The Woias group has found in experiments and simulations that when K exceeds a critical value of about 45, engulfment flow is achieved and mixing becomes fastest.
2.3.3 Mixer modeling and design
The mixer was designed to have engulfment flow. By following the derivation in
[61], K can be expressed as:
K = lPdh,mixer Re3 Y. (2.5)
pV2 Ly
In (2.5), Lv is the volume-to-area ratio of a control volume encompassing the mixer inlet region as described in [61, 62], AP is the pressure drop in this control volume, p is the flow density, V is the flow velocity, and Re is the Reynolds number based on dh,mixer. The ratio
AP/pV2 across the mixer inlet region can be estimated to first order by using the
Hagen-Poiseuille equation:
AP 32 L(
- = ---- (2.6)
pV2
Re
dh,mixerwhere Li is the equivalent length of the mixer inlet region, as described in [61, 62]. This allows K to be expressed as:
%
K =32 L Re2 . (2.7)
Ly
The Reynolds number can be determined using:
Re= ,4 (2.8)
ipdh,mixer
where rn is the mass flow rate and p is the flow viscosity.
For the mixer, it is desired to increase K to achieve engulfment flow while minimizing the pressure drop to allow successful 0-ring sealing. Since each of these requirements exhibits opposite dependencies on the mass flow rate as well as on the
channel's hydraulic diameter, an optimization study was carried out to find a combination of flow rate and hydraulic diameter that meets the above criteria. The study was subject to the constraint that the mass flow rate through the mixer has to be less than or equal to the total flow rate across the device, which is fixed based on the pumping requirements discussed in Section 2.9. A lower mixer mass flow rate simply means using multiple mixers in parallel. Care was taken to ensure the device-to-mixer flow rate ratio produces a whole number of mixers. The objective of the optimization was to find this number and the dimensions of the mixers. Based on the results of this study, four identical mixers having the design shown schematically in Figure 2.2 are used in parallel.
Peroxi 100 pm
Catalyst
Peroxide 30 pm x 40 pm protrusions
2.9 mm
Depth of channel and protrusions = 100 im
Figure 2.2. Schematic of a mixer with dimensions
Each mixer has three inlets: a middle catalyst branch sandwiched between two peroxide branches. Compared to a two-inlet T mixer, this configuration cuts the distance that the peroxide particles need to travel to meet the catalyst in half. In each mixer, the Reynolds number is on the order of 500 under design conditions, and Li and Lv are found following [61, 62] to be about 4 and 6 times the mixer width, respectively. This results in a
K value around 48, ensuring engulfment flow. Once the cross section is determined, the
length is chosen so that the residence time in the mixer (around 1 ms) matches that of engulfment flow mixers reported in [62] to achieve high mixing qualities at their exit. This
time is 3 orders of magnitude less than the mixing timescale of a diffusive mixer with the same width as the current design; the latter would have led to a much larger device. For the original HTP concentration considered (see the discussion in Section 2.4.2 preceding Figure 2.4), the residence time in the mixer is less than the timescale for the onset of phase change due to reaction. This was done on purpose to minimize the amount of bubbles forming inside the mixer, which has narrower channels than any other section in the device, and hence is the most susceptible to clogging. Ultimately, however, the device was operated at higher peroxide mass fractions for which phase change begins in the mixer. This was not found to cause any problems in operation, suggesting that some phase change in the mixer can occur without causing adverse effects. To further enhance mixing following [59] and [60], each mixer in the current design consists of 5 connected zigzag segments and has 12 wall protrusions that extend along the mixer depth.
The pressure loss across each mixer is estimated by first using (2.6) and replacing Li by the total mixer length to determine the losses in a straight pipe of the same length and cross section. Then the losses due to the bends are calculated by using the equivalent length method [63], and those due to the protrusions are approximated by following the experimental findings of [60]. This analysis shows that the total pressure drop in the mixers is about 2 - 2.5 atm, which for a reactor pressure around 2 atm (see Section 2.4) keeps the supply pressure below 5 atm as desired.
2.3.4 CFD simulation of the mixer
A numerical simulation of the mixer was carried out by Collins [64] using ADINA software to study the mixing quality. The streams to be mixed were assigned different colors, and non-reacting flow was assumed. The mixing quality was assessed based on the color of the combined flow at the end of the mixer. The CFD analysis confirmed the design's capability of achieving good mixing for the desired flow rates of peroxide and catalyst. The simulation was two-dimensional due to computational limits, and therefore its pressure calculations, which predicted much lower losses than the above analysis, were considered inaccurate.
2.4 Reactor
2.4.1 Reactor overview
The reactor section is where most of the peroxide decomposition takes place. After thorough mixing with the catalyst in the mixer, the peroxide starts decomposing in the reactor according to (1.2). Along with the chemical reaction, other physical phenomena take place inside the reactor. The hydrogen peroxide and water are initially in the liquid phase, but are subsequently vaporized due to the heat released by the reaction. Some heat is also lost through the reactor walls to the environment. The reactor has a rectangular cross-section for ease of microfabrication. The design process consists of determining the dimensions of the reactor chamber that ensure complete peroxide decomposition and full liquid vaporization. Although this may sound like a simple task at first, the intertwining of physical phenomena occurring simultaneously inside the reactor makes the design optimization challenging, since different phenomena point in different design directions.
For example, a larger reactor volume allows more residence time for the reaction to take place, but also produces larger heat losses. A higher reactor pressure speeds up the reaction after vaporization is complete, but requires higher peroxide and water boiling temperatures, which can prevent the flow from ever reaching the vaporization stages if the chamber is too short. The reactor model therefore had to account for all of the relevant underlying physics, which necessitated a multi-domain modeling approach.
2.4.2 Reactor modeling and design
The reactor was designed to achieve complete peroxide decomposition and full water vaporization. A qualitative understanding of what happens inside the reactor is necessary for modeling, and is presented here first. Inside the reactor, hydrogen peroxide decomposes according to (1.2) into H20 and oxygen while releasing heat. The flow in the reactor passes through five stages based on the thermodynamic phases of the species present, as summarized in Table 2.1. In stage 1, liquid peroxide decomposes into liquid water and oxygen gas, and the heat released causes the reactor temperature to rise. Once the boiling temperature of water at the reactor pressure is reached, stage 2 commences. In this stage, liquid peroxide continues decomposing and the water starts vaporizing at constant temperature. Once all of the water has been vaporized, stage 3 begins. In stage 3, liquid peroxide decomposes to produce steam and oxygen gas, and the heat released again causes the temperature to increase. This continues until the boiling point of the peroxide at the reactor pressure is reached and stage 4 begins. In stage 4, the decomposition continues while the peroxide changes phase at constant temperature. Once all of the peroxide has been vaporized, stage 5 starts. In stage 5, the peroxide and H20 are both in the gas phase.
