Fabrication and Characterization of Sintered Porous Glass Emitters for Electrospray Propulsi on ACM by ACHUSETTS MTMffE OF TECHNOLOGy Julie Xie JUN 162014 B.S., Harvard University (2012) LIBRARIES Submitted to the Department of Aeronautics and Astronautics in partial fulfillment of the requirements for the degree of Master of Science in Aeronautics and Astronautics at the MASSACHUSETTS INSTITUTE OF TECHNOLOGY June 2014 @ Massachusetts Institute of Technology 2014. All rights reserved. Signature redacted Author ................... ........ I Department ofeAeronautics and Astronautics May 22, 2014 Signature redacted C ertified by .......................... Paulo C. Lozano Associate Professor Thesis Supervisor Signature redacted A ccepted by ............................ \Paulo C. Lozano Chairman, Department Committee on Graduate Theses I 2 Fabrication and Characterization of Sintered Porous Glass Emitters for Electrospray Propulsion by Julie Xie Submitted to the Department of Aeronautics and Astronautics on May 22, 2014, in partial fulfillment of the requirements for the degree of Master of Science in Aeronautics and Astronautics Abstract Ionic electrospray thrusters are promising candidates for CubeSat propulsion systems in space, due to their low power requirement and small form factor. Current technology has demonstrated thrust levels of 10 - 40pN, enabling station keeping and attitude adjustment maneuvers. Densification of emitter arrays could increase the thrust density and potentially expand the application space for electrospray propulsion, but current fabrication processes have intrinsic densification limits. A novel, alternative fabrication process to produce microstructure emitter arrays additively by molding is presented in this paper to enable studies into densification as well as wafer-level processing. MEMS techniques are used to process a silicon-on-insulator wafer to produce molds. Soda-lime glass microspheres with a median diameter of about 4pmn are then deposited into these molds and sintered to form porous columns with a diameter of 25pmt, 50prm, or 75jm. These columns become emitters when the device layer is etched with XeF 2 . A porous sintered glass piece is tested as an emitter to characterize the suitability of the mateiral for electrospray propulsion, and an emitted current of 6pA was measured when a voltage of 2.5kV was applied. Currents from 0.1 - 10P.A per emitter tip have been observed from established metallic porous emitters; this demonstrates that the sintered glass emitters are a competitive candidate for electrospray propulsion. Thesis Supervisor: Paulo C. Lozano Title: Associate Professor 3 4 Acknowledgments The author would like to thank: Lockheed Martin and the Asian Office of Aerospace Research and Development (AOARD) for funding this project; the Microsystems Technology Laboratory (MTL) at the Massachusetts Institute of Technology for use of their facilities; Professor Lozano for advising and setting the direction of this research, as well as for teaching me a lot over the past two years; Michael Canonica for being a great research mentor and project partner; Hanqing Li for processing the version 2 molds and various arrays on demand; Kurt Broderick for teaching me a variety of processes as well as showing me how to use many different machines; Todd Billings for answering all my machining questions; Steven Arestie for exchanging and bouncing ideas with me every day: Louis Perna, and Chase Coffman for providing me materials or parts when I needed them; Natalya, Tom, Caleb, and Catherine for being great office-mates in a. room with no window; my parents and my brother for supporting me; Katherine, Kriti, Sabrina, and Susana for always being on my side; and the members of Festivus Miracles for sharing this experience with me. 6 Contents 1 2 1.1 Electrospray principles . . . . . . . . . . . . . . . . . . . . . . . . . . 17 1.2 Electrospray emitter geometries . . . . . . . . . . . . . . . . . . . . . 19 23 Background and motivation 2.1 2.2 3 15 Introduction Areas of electrospray technology improvement . . . . . . . . . . . . . 23 . . . . . . . . . . . . . . . . . . . . . . 23 . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 Fabrication processes . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 2.2.1 Subtractive processes . . . . . . . . . . . . . . . . . . . . . . . 26 2.2.2 Additive processes . . . . . . . . . . . . . . . . . . . . . . . . 27 2.1.1 Wafer-level processing 2.1.2 Densification 31 Methods 3.1 C oncept . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 3.2 Substrate material . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 3.3 Fabrication processes . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 3.3.1 Mold fabrication and design . . . . . . . . . . . . . . . . . . . 36 3.3.2 Deposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 3.3.3 Sintering glass microspheres . . . . . . . . . . . . . . . . . . . 49 3.3.4 De-molding . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57 61 4 Results 4.1 Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 61 5 4.2 Arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66 4.3 Characterization of sintered porous glass for electrospray applications 68 4.3.1 Filling with ionic liquid . . . . . . . . . . . . . . . . . . . . . . 68 4.3.2 Experimental setup . . . . . . . . . . . . . . . . . . . . . . . . 69 4.3.3 Current-voltage curve . . . . . . . . . . . . . . . . . . . . . . . 71 4.3.4 Current stability . . . . . . . . . . . . . . . . . . . . . . . . . 75 77 Discussion and future work 5.1 Material evaluation . . . . . . . . . . . . . . . . 77 5.2 Process evaluation . . . . . . . . . . . . . . . . 78 5.3 Future work . . . . . . . . . . . . . . . . . . . . 79 . . . . . . . . . . . . . . . 80 . . . . . . . . . . 80 5.3.1 Densification 5.3.2 Wafer-level integration 8 List of Figures 1-1 An array of individual emitters. . . . . . . . . . . . . . . . . . . . . . 16 1-2 An assembled thruster with a quarter for scale. . . . . . . . . . . . . 17 1-3 Illustration of a single porous electrospray emitter . . . . . . . . . . . 18 1-4 Models of two emitter geometries created using COMSOL. . . . . . . 20 1-5 Plots showing the change in electric field strength across a line drawn over the surface of the models created with COMSOL. . . . . . . . . 21 1-6 Various types of electrospray emitters. . . . . . . . . . . . . . . . . . 22 2-1 Two different assembly schemes. . . . . . . . . . . . . . . . . . . . . . 24 2-2 A series of electrochemically etched emitter tips to illustrate process variability. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 . . . . . . . . . . . . . 28 2-3 Laser ablated borosilicate glass tip and array. 2-4 Micro-powder compression molded stainless steel emitter array. ..... 29 3-1 Capillaries filled with 5pm silica inicrospheres. . . . . . . . . . . . . . 32 3-2 A concept drawing of the desired emitter arrays. . . . . . . . . . . . . 33 3-3 A scanning electron microscope (SEM) image of glass microspheres on top of a flat silicon surface. The nicrospheres have D50 D1O = 2amn, and D90 = = 3 or 4pm, 6pn. . . . . . . . . . . . . . . . . . . . . . . 35 . . . . . . . . . . . . . . 36 3-4 A schematic of the fabrication process flow. 3-5 An example gold-coated wafer with several different designs etched into . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 3-6 Top-down view of segmented capillary throughholes . . . . . . . . . . 38 3-7 Second iteration mold design features. . . . . . . . . . . . . . . . . . 40 it . 9 3-8 A schematic of the process flow used to create the second iteration molds. 41 3-9 An array of emitters with particle bleed-through scraped off...... 43 3-10 Failure modes of deposition. . . . . . . . . . . . . . . . . . . . . . . . 44 3-11 Emitter structure processed with back pressure. . . . . . . . . . . . . 48 3-12 Stereomicroscope image of second iteration mold array. . . . . . . . . 49 3-13 Progression of sintering between two particles. . . . . . . . . . . . . . 50 . . . . . . . . . . . . . . . . . . . . . . . 52 3-15 Slow sintering trial results. . . . . . . . . . . . . . . . . . . . . . . . . 53 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54 3-17 SEM images demonstrating sintering variability. . . . . . . . . . . . . 55 3-14 Flash sintering trial results. 3-16 Sintering profile. 3-18 Diagram of an emitter structure with a single microsphere forming the tip . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56 . . . . . . . . . . . . . . . 58 3-20 XeF 2 etched glass microspheres. . . . . . . . . . . . . . . . . . . . . . 59 3-19 SEM image of sintered glass microspheres. 4-1 SEM image of 75pn diameter emitter structure with a height of 60im. 62 4-2 SEM image of 25mrn. diameter emitter structure with a height of 60pm. 62 4-3 SEM images of several 50pn diameter emitter structures with a, height of 60p m . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-4 63 SEM cutaway image of an entire emitter structure, with both device . . . . . . . . . . . . . . 65 4-5 An array of emitters with a diameter of 25pm. . . . . . . 67 4-6 Plot of white light interferometry measurements of 25pum array. 68 4-7 Porous sintered glass emitter filled with EMI-BF 4 . . . . . 69 4-8 Sharpened glass tip clamped in an XYZ stage. . . . . . . 70 4-9 Schematic of experimental setup in vacuum chamber. . . 70 4-10 I-V curve of a single sintered glass emitter. . . . . . . . . 71 . . . . . . 72 layer and handle layer columns. 4-11 I-V curve of a single porous tungsten emitter. 4-12 Relative interception current with respect to the current collected by the Faraday cup. . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 74 4-13 Current stability for a single sintered glass emitter fired for 10 hours. 11 75 12 List of Tables 3.1 First iteration mold designs . . . . . . . . . . . . . . . . . . . . . . . 38 3.2 Second iteration mold designs . . . . . . . . . . . . . . . . . . . . . . 39 3.3 Various stopper qualities . . . . . . . . . . . . . . . . . . . . . . . . . 46 3.4 Sintering trials - Set #1 . . . . . . . . . . . . . 51 3.5 Sintering trials - Set #2 "Slow Sintering" . . . . . . . . . . . . . 53 3.6 Sintering profile parameters . . . . . . . . . . . . . 54 3.7 Structural analysis - minimum neck diameters . . . . . . . . 57 "Flash Sintering" . . . . . . . . . . 13 14 Chapter 1 Introduction Electrospray propulsion is a means of electric propulsion that has recently gained increasing attention. Though it was conceptualized very early on, with origins at the beginning of the 20th century, development of electrospray technology has been undertaken mostly in recent times, with the growth of satellites and deep space exploration missions. Like Hall thrusters and ion engines, electrosprays offer high specific impulses of 2000-3000 s [5], especially compared to those of chemical propulsion systems, which are typically 200-400 s. As specific impulse is given by the impulse per unit weight of propellant, f Fdt' g f 1h dt' = it is a. measure of fuel efficiency. Thus, high specific impulses allow for less fuel to be required for a mission. However, these high specific impulses require a tradeoff with the available thrust force, as is shown by the equation for required power: Fc =(1.1) As electric propulsion systems require an external power supply, the power available is essentially fixed. This means that gains in specific impulse must be accompanied with an inversely proportional decrease in thrust force. In practice, Hall thrusters and ion engines can provide at most up to 1ON of thrust, while bipropellait chemical thrusters 15 can deliver hundreds of kN. While this means that electric propulsion cannot supply enough thrust to overcome gravity and launch a spacecraft into space or perform highthrust maneuvers, there are certainly missions for which electric propulsion is a better choice. It may require more time to complete a task than a chemical system would, but it also has the potential for vast fuel savings. However, it must be noted that this comparison also derives from the fact that electric propulsion systems must rely on power supplies such as solar panels, which are unable to provide nearly the same amount of power that a, large chemical reaction can. Thus, this is also a, question of power density. Given the available technology, then, current electric propulsion systems are prime candidates for deep space robotic exploration missions that are less sensitive to mission duration and do not carry heavy human life-support systems. MIT1221 2010AM1 23:14 x100 1 mm Figure 1-1: An array of individual emitters[12]. In particular, electrosprays are an excellent candidate for CubeSat propulsion. CubeSats, which are small satellites constrained in volume to 1000cc and in mass to about 1kg, are increasingly used by universities as well as the private sector for scientific missions but do not yet have a reliable propulsion system. Electrosprays are able to provide thrust without requiring much space or power. Individual electrospray emitters are as small as tens of microns in diameter, and in the ion electrospray 16 propulsion system (iEPS) design, an electrospray thruster device consists of an array of these emitters arranged in an honeycomb pattern, as shown in Figure 1-1; the arrays are themselves small in size, with 12 x 12 x 2mm' dimensions[12], as seen in Figure 1-2. The assembled thrusters have been shown to operate at a power of about 1W, emitting about 400pA, which corresponds to a thrust of about 10 - 40PN[3, 5]. Thus, electrosprays, with their small form factor and low power requirements, could provide for low-thrust maneuvers such as attitude control and slew, which could increase the mission capabilities of CubeSats[14]. Figure 1-2: An assembled thruster with a quarter for scale. 1.1 Electrospray principles Unlike Hall thrusters and ion engines, electrosprays do not need to ionize neutral gases; this is because electrospray propulsion systems make use of ionic liquids, which are room temperature molten salts, as a propellant. Thus, electrosprays do not need to use energy to create ions; energy is only used to extract and accelerate ions. This leads to an improved power efficiency over other electric propulsion devices electrosprays have an efficiency of 80%, while other systems typically have efficiencies from 40% - 70%. In electrospray systems, the potential of the ionic liquid is biased relative to a downstream extractor grid, as depicted in Figure 1-3. This generates an electric field, and, as a result, the ionic liquid becomes electrostatically stressed, with ions 17 Extractor Emitter e, *.e *...**. .~ . E V Figure 1-3: Illustration of a single porous electrospray emitter. of the polarity that is attracted by the extractor grid moving to the surface of the ionic liquid. The electrostatic pressure causes a structure similar to a Taylor cone to form on the surface, and at the tip of this electrified meniscus, the local electric field is strengthened, which causes the cone to sharpen in a positive feedback loop. Eventually, the electric field is strong enough for the field emission of ions from the surface of the meniscus. These ions are accelerated by the potential difference, and, moving through the apertures of the extractor grid, they provide thrust. The description above details a thruster operating in the ion regime, as do the iEPS thrusters; however, an electrospray thruster can also emit charged droplets or a combination of both, corresponding to the droplet and mixed regimes, respectively. The regime of operation depends largely on the flow rate of the propellant; decreasing the flow rate increases the specific charge of a droplet, which corresponds to decreasing droplet size; further decreases in flow rate will push emission towards the ion regime. It is important to note that electrosprays operate at high efficiencies at either end of the spectrum, at either pure droplet or pure ion, with losses occurring during mixed 18 operation. 1.2 Electrospray emitter geometries While in theory ions can be evaporated from a flat ionic liquid surface, in practice, this can be difficult, as the electric field strength required for emission of ions from a. meniscus is given by: 1 - oE2 = 2 2 (1.2) where -y is the surface tension, about 0.05N/m for a. typical ionic liquid, and r, is the radius of curvature of the meniscus, taken to be 3pm. for this approximation. This yields a necessary electric field strength of about 8.5 x 107 , which, when created by two flat plates, would require a. very high potential difference, even at close distances. However, this challenge can be overcome by introducing protruding structures to the surface geometry. If the ionic liquid can travel to the tips of such structures, this allows the ionic liquid to form a strong local electric field without creating a strong field throughout space, as is the case for parallel surfaces. This principle is illustrated by Figures 1-4 and 1-5. The applied voltage is then allowed to be lower and therefore more attainable; avoiding strong fields throughout the space between electrodes can also help to avoid shorting. 19 X10~1 '10-1 X10 (a) Model of flat emitter surface (b) Model of emitter surface with conical structures Figure 1-4: Models of two emitter geometries created using COMSOL. 20 Line Graph; Normalized E-field (V/m) with flat geometry 0.65 0.6 0.55 0.5 0.45 0.4 0.35 0.3 0.25 0.2 0.15 0 E 0.1 0.05 0 -0.05 -E-field -0.1 -0.15 Ey - -0,2 - -0.25 -1200 -1000 -800 -600 -400 200 0 -200 x-coordinate (pm) 400 magnitude- Ez 800 600 1000 120( (a) Electric field across flat emitter surface Line Graph; Normalized E-field (V/m) with conical emitter structures 3 2.5 k 2 1.5 E 0 Z 0.5 0 E-field magnitude Ex -0.5 -- Ey Ez -11 -1200 -1000 -800 -600 -400 200 0 -200 x-coordinate (pm) 400 600 800 1000 1200 (b) Electric field across emitter surface with conical structures Figure 1-5: Plots showing the change in electric field strength across a line drawn over the surface of the models in Figure 1-4. The electric fields are normalized with respect to the expected field strength between two parallel plates separated by the same distance and potential difference as in the models. (a) The field strength barely changes; (b) The field peaks over each individual emitter. In fact, the fields over the flat surfaces in both models are actually equivalent, as can be seen. Then, the peaks over the emitters are 8 to 9 times the flat surface field strength. 21 These protruding structures can be produced in several different ways. The Ecole Polytechnique Fedrale de Lausanne, has used capillaries fabricated through microelectromechanical systems (MEMS) processing; ionic liquid is pumped through these capillaries, and typically, droplets are emitted[17]. Velasquez et al. have produced externally roughened silicon cones, also produced using MEMS processes, that transport ionic liquid to the cone tips passively via capillary forces, although these externally wetted emitters have had challenges emitting stably and reliably[18]. Courtney, Legge, and Coffnan also make use of this phenomenon; emitter tips are etched in different porous material subtrates to create internally porous structures that passively wick up ionic liquids[2, 5, 11]. These different types of structures are shown in Figure 1-6. (b) Externally roughened emitter[7] (a) Capillary emitter[9] MIT1226 2010W9I 2321 xSQQ 100 u (c) Porous emitter[12] Figure 1-6: Various types of electrospray emitters. 22 Chapter 2 Background and motivation Although there are several different types of electrospray emitters as described previously, this paper will discuss the development and fabrication of porous emitters. In considering a novel fabrication method, it is necessary to examine the uniformity of the structures produced; ideally, all the structures would be the same with minimal variations throughout an array. The process should also be repeatable and robust, without significant differences between batches. Finally, it is important to address the areas of future improvement for electrospray technology. 2.1 2.1.1 Areas of electrospray technology improvement Wafer-level processing Currently, electrospray thrusters consist of three major components: the frame, the substrate with emitter tips, and the extractor grid. The frame is a, structure that aligns each emitter to an extractor hole. These components are bonded together with epoxy, forming the thruster package shown in Figure 1-2. The frame and extractors are both produced using MEMS processes, which allows for parallel fabrication. Each silicon wafer can be processed and then diced to produce over 40 frames or extractor grids at once. Established substrate processes dictate that the substrate arrays are produced serially, which means each thruster must be assembled at the die level, as 23 shown in Figure 2-1(a). However, if the substrates could be fabricated in parallel, as the frames and extractors are, then assembly and alignment could occur at the wafer level, with dicing producing full thrusters, as shown in Figure 2-1(b). Then, the thruster device throughput could be greatly increased. I I I'll t - t -~ I I I + 'I + I t t .- - .I. (a) Die-level assembly - I+I II MIII - -I I t mI~ (b) Wafer-level assembly Figure 2-1: Two different assembly schemes. (a) The extractor and frame wafers are diced, and assembly proceeds on the die-level; (b) The three wafers are first aligned and bonded - dicing produces assembled thrusters. 24 2.1.2 Densification Additionally, another point to consider is the thrust an electrospray system can provide. The iEPS system described earlier requires low amounts of power, which, when coupled with a high specific impulse, results in very low levels of thrust, as can be seen from (1.1). While this suits CubeSats very well, as they often have limited power available, larger spacecraft with solar panels or batteries may be able to provide more than that. In that case, being able to translate that power into additional thrust could help to increase the application space of electrospray propulsion. Moreover, thrust cannot be considered alone, as it can be observed that more powerful propulsion systems are usually both bigger and more massive; as examples, consider Hall thrusters, ion engines, or chemical rocket engines. Although they have greater thrust forces, the acceleration is penalized for the correspondingly larger masses. Electrosprays, however, already have an established small footprint (mass and volume). While footprint increases due to power and fuel are largely inevitable, if the propulsion device itself could maintain this small footprint while increasing its thrust to that of a larger system, then higher accelerations could be attained. Additional thrust in a given electrospray device, then, can be created in two ways - by increasing the velocity of emitted ions or by increasing the amount of ions emitted. The former would require increasing the potential difference between emitter and extractor, and as explained previously, those are not optimal operating conditions. As for the latter, it is also impossible to indefinitely increase the current emitted by a. single emitter, as current emitted is intrinsically tied to propellant flow rate, and increasing the flow rate will move operation out of the ion regime. However, it is possible to increase the amount of emitters, thereby increasing the net current emitted. Electrospray thrusters are inherently scalable, seeing as individual emitter tips have been multiplexed to form arrays of tips[3]. An array could simply be expanded to include more tips and thereby provide more thrust. However, this would also increase the footprint of the propulsion device. It is important, here, to note that 25 electrospray propulsion devices can be characterized by a thrust density, or thrust 3 per unit area; the iEPS system described previously can deliver about 10PNIa/cmn or 2 0.1N/m 2 . The maximum theoretical thrust density is estimated at 10 MN/m ; this value is calculated by using the current emitted by a. single Taylor cone and the area of the emitter tip on which that cone forms. Increasing the thrust density can be accomplished by decreasing the distance between emitter tips, or pitch, in the arrays. The current pitch in the iEPS devices is 450pmir; if the emitters were instead spaced 2 350pn. apart, the thrust density would be about 1N/rn , which is already comparable to existing ion engines[5, 16]. If the pitch were to be decreased further to 45jm, one hundred times the current thrust density could be achieved; this is equivalent to the thrust density of high-power Hall thrusters, yet the thruster device would have a significantly smaller footprint. Thus, much greater accelerations could be given by such a device. However, established substrate fabrication methods have a built-in limit to densification; therefore, it is important to develop a fabrication process to enable the study of these possibilities. 2.2 Fabrication processes The following is a short survey of fabrication processes that have previously been explored. 2.2.1 Subtractive processes A subtractive process is one which takes a bulk porous substrate and etches material out of it to produce emitter arrays. The currently established processes for producing arrays are all subtractive processes. Electrochemical etching Electrochemical etching has been shown to effectively form tips in porous metals, such as tungsten or nickel. First, the porous substrate is mounted in the frames for alignment purposes; the substrate is then placed in an acid solution, and a potential 26 difference is applied between the metal substrate and a cathode[5, 11, 4]. This then induces the dissolution of the substrate. By appropriately masking the substrate, Courtney et al. have successfully produced arrays of emitter tips[4]. The process has been shown to be fairly reliable, and Courtney's work has greatly improved the uniformity of the resultant arrays, but there is still some variation within each array, as shown in Figure 2-2. Additionally, the process requires that each substrate die (1cm 2 ) is etched serially, and it may also limit densification due to the resulting geometry of the isotropic etch. Figure 2-2: A series of electrochemically etched emitter tips to illustrate process variability [4]. Laser ablation Porous borosilicate glass substrates have been shown to be able to be formed into tips through laser ablation; a focused laser beam can be used to carve out emitter geometries, with the energy of the laser radiation causing the surface material to be vaporized[2]. A single tip and array produced with this process are shown in Figure 2-3. This laser ablation process has shown a high level of uniformity and repeatability, and although each substrate die is still processed serially, as in electrochemical etching, with the substrate first mounted to a frame for alignment, the processing time can be very quick when using a state-of-the-art laser micromachining system. Furthermore, this process has potential for wafer-level integration with a supply of 4" or 6" borosilicate wafers and additional alignment features. However, there is still an intrinsic limit to densification due to both the laser's finite beam width and the non-uniform and relatively large particle size of the borosilicate glass substrate. 27 (aT4)74B s a m1 tp[2 (a) Borosilicate emitter tip[2] (b) Borosilicate emitter array[2] Figure 2-3: Laser ablated borosilicate glass (a) tip and (b) array. 2.2.2 Additive processes An additive process is one in which material is built up into the desired emitter tip geometry. As an example, molding or targeted deposition are additive processes. Micro-powder compression molding Molding requires that the substrate material take on initially a more fluid, or mobile, state. For example, it could be a liquid or a powder of micro- or nano-particles. This liquid or particulate must be deposited into the mold and then be formed into a more solid, or rigid, body while it is in the mold. This may be an intermediate body, called a green body, that is robust enough to be demolded but is not yet in its final state, or it may be the final . In micro-powder compression molding, a feedstock of particles mixed with a binder is deposited into the mold, pressure is applied to form the green body, and then the green body is demolded, debinded, and sintered, joining them to one another in order to form a, solid macrostructure. Diaz Gomez Maqueo used this strategy to mold stainless steel particles with a diameter of about 1pm, using a pressure of 14000psi, showing that an array of emitter tip structures with a base diameter of 150[pn and a height of 400pmrn could indeed be produced with this method, as shown in Figure 2-4[13]. The results of this process were promising, but the process itself required additional development of the debinding and sintering steps[13], and the 28 material was not characterized for electrospray emission. Furthermore, the process as it stands cannot accommodate a full wafer for wafer-level integration. However, densification here is only limited by the processable thinness of the walls of the molds. These results are important as a proof of concept for molding, and this paper will aim to build upon these efforts. (a) MIT1086 2010/0813 0046 x250 300 um (b) Figure 2-4: Micro-powder compression molded stainless steel emitter array[13]. 29 30 Chapter 3 Methods The fabrication process developed in this work involves the molding of a. particulate powder. The mold is produced through MEMS processing of a silicon wafer so as to be compatible with the two areas of future electrospray improvement, as described previously. If each individual mold is a, part of a wafer, then so long as the powder can be shaped and formed in each mold, or die, in parallel, then the mold will take the place of the substrate layer in Figure 2-1. Then, wafer-level alignment marks can be used for bonding of the layers for a wafer-level assembly. Furthermore, densification is readily achievable to the point where the desired pitch is on the order of the inaccuracy of the machine or process, and in MEMS, this is typically less than 1011m. For this project, however, neither of these improvements will be explored in detail; rather, the goal is to outline a process that will enable future improvements in these areas to occur. 3.1 Concept The design concept was inspired largely by the work of Krpoun and Sheal10], who worked on developing capillary arrays as electrospray emitters, which have in the past struggled with too low of a hydraulic impedance, especially for emission in the pure ion regime. To combat this, the capillaries were filled with silica inicrospheres with a diameter of about 5p-m, as shown in Figure 3-1, which effectively increased the hy31 draulic impedance and decreased the flow rate, leading to successful ion emission[10]. (a) (b) Figure 3-1: Capillaries filled with 5pm silica microspheres[9]. This configuration, then, is similar to the iEPS design, which uses a bulk porous material for emitters. Here, the silica microspheres simulate a porous core. However, there are also advantages to having a. porous external surface; namely, any extra ionic liquid flow can be recaptured into the porous body, rather than remain on a silicon surface, where it can accumulate and potentially lead to an electrical short circuiting of the device. For that reason, this work will reconsider the results of Krpoun and Shea and instead aim to keep the internal porous columns as the desired emitter structures, using silicon wafers to produce a mold for these columns. The microspheres used to fill the capillaries will be sintered to form a rigid body, and then the silicon mold 32 material will be removed. The concept drawing for the targeted devices is shown in Figure 3-2. The porous columns will be embedded in a silicon wafer, open to both ends. One end will be flush with the silicon such that ionic liquid can be deposited and transported through the columns to the other end, which protrudes from the silicon surface, forming the emitter structure. Ion electrospray Extractor plate Taylor cone Porous emitter Silicon Nm - x + (a) TV (b) Figure 3-2: A concept drawing of the desired emitter arrays: (a) details the various components of the device; (b) is a cutaway view, showing the open porous columns. 33 3.2 Substrate material In the past, porous metallic electrospray emitters were desirable, since as a conductor, a metallic emitter could also function as an electrode. However, degradation of the emitter tips, caused by electrochemistry occurring at the electrode-liquid interface, was a significant challenge to the long-life operation of such emitters. Brikner and Lozano found that by using an upstream distal electrode and allowing the emitter to vary in potential with the ionic liquid, electrochemical degradation could be shifted to the distal electrode, thus avoiding any erosion of the emitter tips[1]. This motivates the use of dielectric materials for emitter tips, since the substrate no longer needs to function as an electrode, and dielectrics are less subject to electrochemical decay. The material also needs to be both space-compatible, meaning it will not outgas in space, as well as chemically inert in the radiative environment of space. Given these reasons, a glass or ceramic would be a good choice. Finally, because a sintering process will be used to bond the microparticles together, a. material with a lower nelting temperature will help to facilitate the process. Consequently, commercially available soda-lime glass microspheres supplied by Cospheric were selected as the material of choice. This glass is composed of SiO 2 , CaO, MgO, Na 2 0, K 2 0, and AL 2 0 3 , and it has a relatively low melting temperature of 730 C. The microspheres are also polydisperse, as seen in Figure 3-3, meaning that the individual particle size varies. The batches used in this work have a median diameter of 3 or 4pm; 10% of the beads have a diameter of about 2pm. or less, and 90% of the beads have a diameter of about 6ptm or less. It is notable that microspheres with a diameter or 10nmn or greater are also present in the population. While glass microsphere supplies with a more narrow distribution were pursued and sought after, it was found that they are not readily available, due to the manufacturing method. Nevertheless, a more mnonodisperse microsphere population can be achieved in the future by filtering the supply; however, this avenue has not been pursued in this work. 34 MIT4835 2013/05/22 14:38 xl.Ok 100 um Figure 3-3: A scanning electron microscope (SEM) image of glass microspheres on top of a flat silicon surface. The microspheres have D50 = 3 or 4pmn, D1O = 2pn, and D90 = 6pm. 3.3 Fabrication processes Several major fabrication steps are required to produce these emitter arrays. First, the molds themselves must be designed and manufactured using MEMS processing techniques. Then, the soda-lime glass microspheres are deposited into the mold channels and sintered. Finally, the structures are demolded to produce the emitter tips. The overall process is shown in more detail in Figure 3-4, and specifics relating to each fabrication step are described in the following sections. 35 (a) (b) (c) (e) (f) (g) Silicon EGlass Silicon dioxide * Stopper II (d) (h) Photoresist Figure 3-4: A schematic of the fabrication process flow: (a) processing begins with an SOI wafer; (b) channels are etched into the handle layer of the wafer with DRIE; (c) channels are etched into the device layer of the wafer with DRIE; (d) the BOX is opened with a BOE etch; (e) glass microspheres are deposited into the open mold channels; (f) the glass is sintered; (g) XeF 2 is used to etch away the device layer; (h) a film protecting the handle layer is removed, producing the emitters. 3.3.1 Mold fabrication and design The molds are processed out of a silicon-on-insulator (SOI) wafer. These wafers consist of a silicon handle layer about 500pin thick, a thinner silicon device layer in this work, either 60pn or 100p m thick - and a 2pm thick silicon dioxide layer sandwiched between the two silicon layers. This inner layer is called the buried oxide (BOX). 36 Mold designs Two different iterations of mold designs were developed in this process. The first iteration corresponds with the 60pm device layer SOI wafers, and the second iteration uses wafers with a 100pm device layer. The goal of the first design was to prove the feasibility of the fabrication of these emitter structures, while the second design is intended to be compatible with the current iEPS frames and extractors such that fabricated arrays can be fired and tested. Specific differences between the two will be explained below. Figure 3-5: An example gold-coated wafer with several different designs etched into it. Each wafer is able to produce several different mold designs at once, as indicated by Figure 3-5. This allows for the concurrent evaluation of several different mold designs and thus helps to accelerate the development process. For example, the first mold design featured two major alternatives for the device layer design. One consists of segmented capillary-like structures, as shown in Figure 3-6 and was designed for rapid de-molding; however, these structures proved to be very fragile and easy to break. The other device layer option was much more robust. It is comprised of 37 1 mm MIT4767 (a) (b) Figure 3-6: Top-down view of 6pm tall segmented capillary throughholes: (a) shows an array; (b) shows a single tower. throughholes in the bulk silicon layer - this is the design that is described in Figure 3-4. Once the device layer macro-design was decided, however, there were still four alternate throughhole design choices, outlined below in Table 3.1, where FS refers to the front side, or device layer, of the wafer, while BS refers to the back side, or handle layer. An open hole refers to a hole with a simple circular cross-section, while a tri-section hole has three 20pm thick silicon walls forming the shape of a "Y" that divide the hole cross-section into three equal areas. Table 3.1: First iteration mold designs Identifier FS diameter BS diameter BS geometry 1-AL 75prm 150pum open hole 2-BIL 50pn 150pm tri-section hole 3-CAL 25prn 50pm open hole 4-DAN 50,m 45im open hole Using these molds, emitter structures with a diameter of 25pim, 50pzm, and 75pn were all successfully produced. However, producing full arrays was much more difficult. While the 3-CAL design has the smallest hole diameter, deposition trials showed 38 that the 2-BIL and 4-DAN designs, the two 50prn FS diameter designs, were actually harder to fill uniformly. High yield arrays (> 80%) were produced using both the 1AL and 3-CAL designs but not 2-BIL or 4-DAN. This is likely due to the fact that the 1-AL and 3-CAL designs both have larger BS diameters than FS diameters, creating a, funnel-like throughhole. The 2-BIL design, while also possessing a large overall BS diameter, has that area broken up into 3 smaller irregular areas, creating three less effective "funnels". Finally, the 4-DAN design is unique in its "inverse-funnel" design - it has a BS diameter that is smaller than its FS diameter; in fact, it has the smallest BS diameter of all the designs. This was also less effective, as it likely rejected more glass microspheres. Therefore, in selecting designs for the second iteration of the molds, all of the options were created using the "funnel" configuration, with a larger BS diameter than FS diameter. Furthermore, a diameter of 50im was chosen as the FS diameter for most of the designs, in spite of there having been no 50pm arrays produced with the first iteration of mold designs. This is due to the success of both the 75pm arrays and the 25pm arrays; it is reasonable to hypothesize that a 50pn array will be able to be filled with a revised BS design. Furthermore, the 75prm molds had been demonstrated to be significantly easier to fill than the 25p1m. molds, but a smaller FS diameter is more desirable for a greater emitter aspect ratio; as a result, 5Opmn was deduced to be a more readily attainable compromise for the testing aims of the second mold designs. The full list of second iteration emitter array mold designs is included in Table 3.2. Table 3.2: Second iteration mold designs FS diameter BS diameter Pitch 50pi' 200pin 450jm 75pnm 200pim 450pm 50pman. 200pim 300pm 25in 200pin 250pjim 39 As shown in Table 3.2, pitch size is an important feature in the second iteration of the mold designs. Because these molds are to enable testing with the current iEPS devices, it was necessary that some designs have the same pitch as that of the iEPS design, 450pn. Furthermore, die-level alignment features were added so that the soda-lime glass emitter arrays could be mounted onto the iEPS frames such that emitters would line up with extractor holes. These features are shown in Figure 37(a). Additionally, an inset reservoir was added to the handle layer to aid in the deposition of glass microspheres; this is also shown in Figure 3-7(a). (a) A second iteration mold. (b) An iEPS frame. (c) An empty mold mounted into the frame. Figure 3-7: Second iteration mold design features: (a) shows a, slightly recessed reservoir pool around the mold holes, as well as alignment marks, especially visible on the right side of the mold; (b) shows the iEPS frame that has a slightly protruding lip, especially visible on the top side of the device, that fits into the mold alignment feature; (c) shows the mold flipped and mounted into the frame. Mold fabrication Once the mold configurations are decided upon, they are printed out onto masks. For this application, the mask is a quartz plate with a layer of chrome patterned into the desired design. There is one mask for each height layer of the mold design, which means that there are two masks for the first iteration mold design and three masks for the second. Figure 3-4(a)-(d) shows the overall processing steps used to create the molds. Deep reactive ion etching (DRIE) is a dry, or gaseous, process used to etch away silicon to produce channels on both sides of the wafer. As the BOX layer is composed of 40 (a) Clean SOI wafer. (b) Grow 1.5pm ox- (c) Deposit photore- (d) Use BS mask for features sist on both handle alignment ide. reservoir inset and layer. device and layer and FS holes mask; photoresist. develop This patterns the BS delay mask. (e) Etch in BOE. This (f) Strip the remain- (g) Deposit photore- (h) Use BS holes mask; develop phosist on BS again. will pattern the oxide ing photoresist. toresist. layer and form the delay mask on BS. (i) Etch BS about (j) Strip the remain- (k) Etch an additional (1) Etch 1Mimi. to reach BOX. DRIE. with ing photoresist. 400prn deep This forms two height DRIE. layers on BS. FS with ITi (m) Etch in BOE to remove surface oxide as well as BOX. Figure 3-8: A schematic of the process flow used to create the second iteration molds. A cross-section of the resultant molds is shown in (m). 41 silicon dioxide, it is resistant to ion etching, thus serving as an etchstop and allowing for distinct heights and diameters in the channels of the device and handle layers. The BOX then requires an additional removal step using buffered oxide etch (BOE), which is a buffered solution of hydrofluoric acid, an etchant that is highly selective for glass. For this application, the BOE etch step can be difficult, since the channels can get air bubbles trapped inside, blocking the BOX from the etchant. However, an ultrasound is very effective in releasing any bubbles, and a 30min BOE etch with ultrasound will typically open up the throughholes. The second iteration of the molds has a slightly more complex design, requiring the use of a delay mask to produce the two different height levels on the handle layer. Figure 3-8 explains the processing steps in detail. 3.3.2 Deposition After the molds are produced, the next step is the deposition of the glass microspheres, as shown in Figure 3-4(e). The glass beads are first mixed into a. suspension and then deposited on the BS of the molds and induced to travel through the channels to the FS by a combination of gravitational and capillary forces. The solvent, is then allowed to evaporate, leaving behind a. glass microsphere microstructure that is held in place through static forces. In this work, the deposition process proved to be more complicated than that described by Krpoun[9], likely due to a combination of a few factors. For one, the glass microsphere population, as described earlier, is polydisperse. Typically, the mean diameter of the particles used to fill a mold should be an order of magnitude less than the diameter of the mold channel. While this is true for the 75ptm and 50pim. holes, it is very close for the 25prm holes. Indeed, given that the largest particles have a diameter > 10pjm, two of these could block a. throughhole. Another complicating factor is the bleed-through of the microspheres. Bleed- through here refers to the overflowing of particles onto the FS of the molds. For this application, it is a problem because sintering will join these overflowed particles to the emitters, forming a thin film at the level of the emitter tips, thus disrupting the 42 tip geometry. These particles can be scraped off with a razor blade before sintering, but doing so will damage the particle packing structure at the tip, generating irregularities. Moreover, as it is a manual step, it is imperfect and will likely leave behind residual particles, still allowing small films to form. These effects can be seen in Figure 3-9. Thus, it is necessary to introduce a stopper material placed on the FS of the device to prevent this bleedthrough. However, a stopper can work too well, not allowing anything to escape out the FS, and with deposition beginning on the BS, air bubbles can then get trapped in the channels and prevent filling. MIT8863 2013/10/11 17:13 x100 1 mm Figure 3-9: An array of emitters with particle bleed-through scraped off. The tips have irregular features. These effects have made deposition somewhat of a challenge, as evidenced by the failure modes depicted in Figure 3-10. Several variables in the deposition process, however, can be adjusted to improve the yield. These will be discussed next, and 43 then a deposition process that has resulted in successful array production will be given. MITS142 2013109111 15:13 xl.Ok MIT8299 100 um 2013109/4 11:15 xlOk lO0um 14:16 X1.Oc 100 UM (b) (a) MIT8469 201319117 16:31 xl.Ok MIT8321 lO0um 2013109/17 (d) (c) Figure 3-10: Failure modes of deposition: (a) an empty hole; (b) a partially filled structure; (c) a shell structure; (d) a, structure with compromised height. Concentration The concentration of beads in the suspension is an important parameter for a few reasons. Firstly, the concentration affects the amount of beads deposited, since only a certain volume of the suspension can be deposited without flooding the mold. Then, there must be enough of the microspheres to fill the throughhole completely, given also that a certain amount will remain on the BS surface and not travel into the holes. While the structures are very small and do not require a large mass of glass, initial 44 concentrations of 1% by weight of beads appeared to be insufficient. Furthermore, additional weight can only help deposition, as it provides additional filling force. However, it can be observed that increasing the concentration of glass microspheres also increases the contact angle of the suspension, which is indicative of decreased wettability. This is then a barrier to filling. Current experinental results have shown successful array filling using a 10% by weight suspension of glass microspheres. Solvent The solvent can also affect the behavior of the suspension. Notably, choice of solvent can affect the wettability and viscosity of the suspension, which can affect the fill rate. Additionally, the vapor pressure of the solvent could be a point of interest. Initial efforts used deionized water (DIW), which successfully produced structures but not full arrays. Increasing the wettability, then, could help improve the fill rate, since it would increase adhesion between the suspension and the silicon mold walls. Both ethanol and isopropanol were good candidates, as they both have a lower contact angle with silicon than DIW does. However, there was a concern about their higher vapor pressures - if the solvent evaporated before the microspheres had a chance to settle into the channels, it could be problematic. For that reason, the alcohols were combined with water to form solvents. In fact, the ethanol and isopropanol suspensions performed similarly and were an improvement over DIW alone; these produced high-yield emitter arrays. Viscosity was also considered a. property of interest, as it was possible that the water was not pulling the glass beads into the throughholes, even if it was flowing through itself. To address that hypothesis, a glycerol-water solvent was used. However, the glycerol suspension did not wet the silicon well, and filling was mediocre. A glycerol-isopropanol solvent was attempted, but the results were not conclusive. Future development of a high viscosity solvent with good wettability and low vapor pressure, such as ethylene glycol or formamide, could produce even better fill rates. 45 Stopper As mentioned previously, the stopper is an important ingredient in this process and performs a specific task: to prevent glass microspheres from reaching the FS. Thus, it is important that the stopper makes good contact with the FS silicon. Removing the stopper, however, should not damage the settled glass structure. Furthermore, in light of the air bubbles possibly caused by the stopper, allowing the escape of these bubbles through utilization of a porous stopper would be ideal. In fact, the stopper could even be like a sieve, allowing the solvent and air to pass through but holding the glass microspheres back. The stopper's porosity is the property that most affects the fill rate; the other properties will instead affect the tip geometries. Several stopper materials were evaluated for suitability, and the results are summarized in Table 3.3. Table 3.3: Various stopper qualities Contact with FS Porosity Gentle removal 3M Parafilm Good Poor Good Porous alumina, disk Poor Good Good Porous nickel disk Good Fair Good 3M Micropore tape Good Good Poor Stopper material As seen in Table 3.3, each of the stopper choices are good options with perhaps one weakness. Parafiln is minimally permeable to gases and not to liquids, so it does not allow for any flow through the FS. The porous alumina and porous nickel need to be attached to the molds with tape, and since the porous alumina is very thin and fragile, it can be difficult to attach tightly without breaking; moreover, it occasionally breaks from flexing under the weight of the inicrosphere suspension. The 3M Micropore tape can form a very good seal and is definitely porous, but removal of the tape also removes the very top layer of beads in the molds. The porous nickel, then, is typically the best choice, as its weakness is that its surface is not as porous as some of the other options. However, it does require flat surface for taping, so if 46 the mold design does not have enough of a lip, as some do not, then the Micropore tape is the next best option. Additional processing steps Several additional processing steps can be used to aid and facilitate the deposition process. An ultrasound bath can help to make the particles in the suspension more mobile, pushing them towards local depletion zones, dislodging microspheres that are blocking a channel, or freeing air bubbles. A dessicator pulls a vacuum; this can bubble out any trapped air bubbles and allow for smoother suspension flow. Multiple deposition is also a. valuable tool; earlier it was discussed that additional weight could help filling, but increasing the microsphere concentration of the suspension decreased the wettability. However, additional weight can be applied by simply waiting for some of the solvent to evaporate and then depositing more. While this may also increase the effective bead concentration on the mold, it is a gradual, staged increase and has been demonstrated to improve deposition. Finally, adding a pressurized gas flow to the BS can provide additional forcing of the beads into the channels. However, in experimental trials, this would blow out the deposited suspension, creating a bleedthrough as well as a disordered microstructure, as shown in Figure 3-11. If the applied pressure could be finely tuned and controlled, this could prove to be an effective step for future deposition processes. Current process The currently established deposition process is as follows: 1. Prepare a 10% by weight glass nicrosphere suspension. The solvent is a solution of ethanol/isopropanol and DIW. 2. Clean (ultrasound in acetone, isopropanol, DIW) and dry (using N 2 gas) molds. 3. Clean porous nickel chips, if using as stopper (ultrasound in DIW, isopropanol, acetone) and evaporate to dry. 4. Prepare molds with stoppers: attach porous nickel stoppers (with Kapton tape) or attach Micropore tape. 47 MIT7250 2013108/02 12:02 L . x1.0k MIT8280 100 um 2013/09/13 14:27 x1.Ok 100 um (a) Top-down view of emitter structure before (b) Tilted view of same structure after short XeF 2 etch. XeF 2 etch. Figure 3-11: Emitter structure processed with back pressure: (a) shows the hole appearing to be mostly filled, although bleed-through is evident; (b) shows the structure is actually shorter than it appeared, and the packing of the beads is disjointed and disordered. 5. Prepare rig to suspend mold - this increases the available area for liquid or gas flow in the stopper. 6. Deposit suspension into the BS reservoir. 7. Dessicate the molds. The liquid will bubble and then quiet down. At this point, it can be removed. 8. Deposit more of the suspension to replace any liquid that evaporated. 9. Place the molds in an ultrasound bath for 15min. 10. Again, deposit more of the suspension to replace any liquid that evaporated. 11. Ultrasound the molds for 1hr. 12. Allow the molds to evaporate dry fully. This process has been used to produce arrays with the first iteration of mold designs as well as an array with the second iteration designs with the thicker device layer, shown in Figure 3-12. However, the second iteration arrays were notably not uniform in height; thus, the process likely can be improved. It appeared that with the additional height and larger BS diameter, there were not enough beads. Thus, the concentration of the suspension could be raised, or more deposition steps could be added. 48 Figure 3-12: Stereomicroscope image of second iteration mold array. 3.3.3 Sintering glass microspheres In Figure 3-4(f), after the stopper is removed, the filled molds are taken to a box furnace to sinter the glass microspheres. Sintering is the process of forming a solid body from a particulate by applying heat and/or pressure. In this work, it is the process that transforms the molded powder into rigid emitter structures. Sintering is an ancient craft, having been practiced for thousands of years to produce pottery and art[6]. There are a few different atomic level events that can cause particles to join together, and in glass, the dominant atomic event is viscous flow. Sintering is an irreversible process, as it consumes surface energies of the particles to form bonds between them. This means that small particles that have a greater surface energy will sinter more readily. The rate of sintering can depend on a number of other factors as well, such as temperature, pressure, and environmental conditions. In this work, sintering will be conducted at atmospheric pressure with the environment as controlled as possible. Temperature will be the only variable controlling the rate of sintering. The progression of sintering can be classified into four stages. In the first stage, 49 necks begin to form between adjacent particles. In the second, the neck grows. In the third, pores between particles begin to grow and round themselves out. In the last stage, the pores are no longer interconnected; a fluid or gas cannot move from one face of the solid to the other, although there may be individual round pores still present. Figure 3-13 illustrates the corresponding progression of these stages in two adjacent particles with time scales. It is important to note that the rate of sintering decreases with the progression of sintering. This is due to the fact that the surface energy of the particles is being depleted. In fact, complete coalescence, as shown in Figure 3-13, would require a theoretically infinite amount of time. Initial point D contact Spherical particle D = diameter Neck Grain bouindary Early stage neck growlh (short time) Late stage neck growlh (long time) Terminal cordwation fully coalesced (infinite time) 1 26 D Figure 3-13: Progression of sintering between two particles[6]. For this application, however, this property of sintering is beneficial, as a fully 50 solid body is not desirable. Instead, to create a porous network, sintering should be paused during the second stage; the necks between particles will have formed, but the body will still be highly porous. Furthermore, the pore size and the porosity can be tuned to some extent by varying the parameters of temperature and time, as indicated by Kingery et al. [8] Moreover, it is important to note that the same relative density or porosity can be achieved with different combinations of temperature and time. The following, then, will describe the experimental trials used to identify an effective sintering process. Due to the fact that the vent of this furnace was open to atmosphere, these trials were necessary to adjust existing sintering formulas for the furnace. Sintering trials and final process Sintering processes commonly take place at temperatures below the melting point[15]. Since the soda-lime glass microspheres have a melting temperature of 730 0 C, initially it was decided that 700 0 C would be the sintering temperature. Since it is close to the melting point, it was anticipated that these would be relatively quick sintering runs, which is beneficial in decreasing the overall process time. A few sample trials at this temperature are described in Table 3.4, with the corresponding results shown in Figure 3-14. Table 3.4: Sintering trials - Set #1 "Flash Sintering" # Sinter temp. Rise time (h:m) Hold time (h:m) Fall time (h:m) 1 700 0 C 2:20 0:40 5:00 2 700 C 2:20 0:20 5:00 3 700 C 2:20 0:10 5:00 Trial As can be seen from the results, sintering in Trial #1 and Trial #2 clearly pro- gressed too far; the desired internally porous network was not produced. Trial #3, 51 2012/11/28 15:08 x1.0k 100 um MIT3758 (a) Trial #1 2012/12110 13:32 (b) Trial #2 (c) Trial #3 Figure 3-14: Flash sintering trial results: (a) is completely melted, with no visible microstructure - Stage 4 sintering; (b) is mostly melted, with some large particles and pores visible - Stage 3 sintering; (c) has a very porous network with individual particles still visible - Stage 2 sintering. however, did produce a porous substrate that could be used for electrosprays. However, it is evident that there is a large difference in the microstructure between each trial run, even though the runs only varied in the hold time by a maximum of 30 minutes. In fact, Trial #2 and Trial #3 differed by only 10 minutes, yet only one result was acceptable. Given the fact that the furnace is open to atmospheric conditions, such as humidity, which can affect the sintering rate, it was decided that this sintering profile would not be robust enough for repeated processing. Thus, it was determined that a slower sintering process at a lower temperature, while requiring more time, could provide a more robust profile. A second set of trials 52 instead targeted a sintering hold time of 4hr and sought an appropriate sintering temperature. These runs are summarized in Table 3.5, and the results are shown in Figure 3-15. Table 3.5: Sintering trials - Set #2 "Slow Sintering" Trial # Sinter temp. Rise time (h:m) Hold time (h:m) Fall time (h:m) 4 600 C 2:00 4:00 4:00 5 650 C 2:30 4:00 4:30 6 625 C 2:15 4:00 4:15 2013102127 15:19 L X1.0K 100 um MIT4,0d (a) Trial #5 201304/02 17:5b (b) Trial #6 Figure 3-15: Slow sintering trial results: (a) is completely melted, with no visible microstructure - Stage 4 sintering; (b) is highly porous, with individual particles still visible - Stage 2 sintering. While the rise and fall times in Table 3.5 were not held constant, this was because the ramping slopes of the furnace were instead held as the constants. The important variable, then, is the sintering temperature. At 600 0 C, Trial #4 did not sinter, so a picture was not provided. Trial #5, sintered at 650 C, progressed too far and melted all the particles together. Trial #6 at 625 0 C, however, did successfully produce a porous network with small pore sizes. This sintering profile has also been shown to be more robust; changing the sintering temperature to 635 C produced a similarly 53 porous microstructure, with slightly more extended necking between the particles. Finally, the porosity of this sintered glass substrate was measured at about 43%. The final sintering profile used in these fabrication efforts is given in Table 3.6 and illustrated by Figure 3-16: Table 3.6: Sintering profile parameters Sintering temperature 625'C Rise time (h:m) 2:15 Hold time (h:m) 4:00 Fall time (h:m) 4:15 600 500 400 4D - 0 1 10 11 300 a) F- 200 100 0 1 2 3 4 5 6 7 8 9 Time [h] Figure 3-16: Sintering profile. The total time for the sintering process is 10hr : 30min. Additionally, the process is run with a 2500O N2 flow to control the environment as much as possible. Even so, there is still some minimal variability inherent in the process due to the furnace. For one, there is some lag between the programmed temperature and the actual temperature, as it usually starts at roughly 17 C, not 25 C. takes time to heat up the volume of the furnace, which is more than Ift 3 It also . Another artifact of this is the presence of a temperature gradient in the box furnace; since heat is applied from the walls, it will be cooler towards the center of the box. To 54 account for this, the samples were placed close to the same location every time; still, especially when sintering multiple arrays at once, it is unavoidable that they will each be subject to slightly different conditions. Lastly, as previously explained, the furnace is open to atmosphere, which can also affect sintering. These effects can cause the kind of variability shown in Figure 3-17. However, that level of variability is still acceptable for this application, as both of the microstructures are still openly porous. Furthermore, after running extensive trials with this sintering profile, it can be observed that these variations are very uncommon. Overall, the sintering profile of Table 3.6 is satisfactorily robust for producing porous electrospray emitter arrays. (b) (a) Figure 3-17: SEM images demonstrating sintering variability. Both samples were produced using the same sintering profile. Structural analysis As these sintered glass structures are intended to function as electrospray emitters, they will be subject to the same electrostatic pressure that the ionic liquid propellant is subject to. This pressure, if strong enough, can break apart the sintered bonds. This is clearly not an optimal scenario. In the best case, a few beads falling off could cause temporary efficiency losses and unexpected increases in thrust; in the worst case, the emitters would decay over time, eventually leaving behind no structure. Regardless, any unpredictable behavior should be avoided. Figure 3-18, then, shows the worst case scenario, where a single neck joint between 55 El Figure 3-18: Diagram of an emitter structure with a single microsphere forming the tip. The neck joint identified must be able to bear the stress of the electrostatic pressure defined by the electric field, E. two particles must bear the electrostatic pressure. In equation (1.2), the pressure balance of ionic liquid at the emitter tip gives that the electrostatic pressure, !COE2, . This same electrostatic pressure also is equal to the Laplace or surface pressure, acts on the glass microsphere. Multiplying by the projected area cross-section, then, will give the force on the particle. Thus: F = P x A= 1 oE2 x 27rr=2 =2 rc 2 x 27rr2 = 47rre This is the force that must be able to be sustained by the neck joint. The minimum neck diameter, then, can be determined by the following: F = a x Aneck = 9x 7 d 2k "c 4 Taking -y, the surface tension of the ionic liquid, again as 0.05N and -, the yield stress, conservatively as 19MPa, different values of rc will yield different allowable neck diameters, as summarized in Table 3.7. 56 Table 3.7: Structural analysis - minimum neck diameters re dneck dneck 1pm 0.2pm 0.10 2pm 0.3pm 0.07 3pm 0.4pm 0.06 7pm 0.5pm 0.04 As shown above in Table 3.7, the ratio of the neck diameter to the particle diameter is larger for smaller particles and decreases with particle diameter. Fortunately, sintering also follows a similar profile. As the smaller particles have a higher surface energy, they sinter more readily and are therefore more progressed in neck growth than the larger particles. In fact, for particles with a diameter of 2pm or less, the sintering profile in Table 3.6 usually produces neck sizes that are roughly 50% of the particle diameter or greater. This can be observed in Figure 3-19. The larger particles also satisfy the minimum neck diameter requirements. In fact, these results originally helped to motivate the decision to increase the sintering temperature to 635 C and thereby increase the neck diameter. 3.3.4 De-molding The final processing step to produce the emitter arrays is de-molding. This is illustrated in Figure 3-4 (g) and (h). These two steps serve to remove the remaining bulk silicon of the device layer while providing some protection for the handle layer, as the handle layer provides structural support and holds the arrays together. Initial efforts attempted to use tetramethylammonium hydroxide (TMAH), a wet anisotropic etchant, to remove the silicon. However, it proved to be ineffective. XeF 2 , a dry isotropic etchant with a relatively high etch rate, was then successfully used to remove the device layer with a 60min etch. XeF 2 has the additional property of being highly selective for silicon over glass. Consequently, the oxide layer serves as an ef57 Figure 3-19: SEM image of sintered glass microspheres. Necks between particles are clearly visible fective etch-stop, and the soda-lime microspheres are also minimally affected. Figure 3-20 shows the resultant surface etching. In fact, the roughening of the surface can serve to improve the wettability of the soda-lime emitter tips. Then, in order to protect the handle layer, it is necessary to cover the BS. As shown in Figure 3-4 (g), photoresist can be deposited on the BS for this function and subsequently removed using an oxygen plasma asher. While this does sufficiently protect the handle layer, the liquid photoresist can get inside the pores of the sintered glass on the BS, making it more difficult to remove. This then blocks off the BS pores, which can inhibit filling of the arrays with ionic liquid. A simpler solution is to instead cover the BS with a glass slide. This has also been shown to be effective and does not affect the soda-lime microstructure. 58 Figure 3-20: XeF2 etched glass microspheres. While all of the silicon was removed, the glass only received a surface roughening. 59 60 Chapter 4 Results The fabrication process detailed in Chapter 3 has been used to produce both individual emitter structures and arrays. These will be introduced and evaluated in the following sections. Additionally, the sintered porous glass material itself is characterized for its suitability for electrospray propulsion. This in particular reflects upon the material's wettability with ionic liquid as well as its porosity. A test piece of this material is fired to determine its emission characteristics; these results are also included in the following. 4.1 Structures Single emitter structures with diameters of 75pm, 50pm, and 25pm have all been successfully fabricated. Figure 4-1 shows an SEM image of a 75pm diameter structure, Figure 4-3 shows several SEM images of 50pm diameter structures, and Figure 4-2 shows an SEM image of a 25pum diameter structure. All of these structures were produced with the first iteration molds, so they all have a height of 60pm and respective aspect ratios of 0.8, 1.2, and 2.4. The structures typically have a very tightly-packed microstructure, as can be seen in the figures. However, a few of them are more sparse, as in Figure 4-1 and Figure 4-3 (f). The latter can be explained by the 4-DAN mold design; recall that this design had a larger FS diameter of 50pm with a smaller BS diameter of 45pum. In fact, this 61 MIT8867 2013/10/11 17:18 x1.0k 100 um Figure 4-1: SEM image of 75pum diameter emitter structure with a height of 60pm. MIT7491 2013/08/20 16:31 x3.Ok Figure 4-2: SEM image of 25pm diameter emitter structure with a height of 60pm. 62 (a) 2013/06/27 (b) 16:43 L x1,5k 50 um MrT5776 (d) (C) 2013/06/27 2013/06/27 16:47 L x1.5k 50 um MIT5782 (e) 2013/06/27 (f) Figure 4-3: SEM images of several 50pm diameter emitter structures with a height of 60pm. "inverse funnel" design can be seen in Figure 4-3 (f) upon careful observation. These molds were difficult to fill, as the smaller BS diameter rejected more microspheres. 63 Thus, the loose packing in Figure 4-3 (f) could be the result of an insufficient amount of beads filling the channel. Figure 4-1, however, also had a looser microstructure. This could be caused by the solvent used in deposition. As alluded to in Chapter 3, a solvent that evaporated too quickly may not provide enough time for the microspheres to settle into a lowest-energy configuration. It could also be caused by a solvent with a lower viscosity and/or surface tension. Typically, as the solvent evaporates, surface tension forces pull the remaining solvent into itself. Viscosity would then also pull the glass microspheres along, packing them in. Insufficient surface tension forces or viscosity compounded with a higher vapor pressure could thus lead to a looser packing configuration. In fact, Figure 4-1 used a 5:4 ratio of ethanol to DIW as the solvent, while Figure 4-3 (c), for example, used only DIW as a solvent. Variations in the microstructure could then also be an effect of different choices of solvent. Additionally, several of the structures in Figure 4-3 especially can be seen to taper slightly in diameter, increasing towards the base. A cusp at the tip of the emitter, with the edges raised above the center, can also be noticed. The tapering may be attributed to some uneven shrinkage during sintering. Since the mold is sintered upside down, the top of the emitters is is closest to the furnace walls and therefore exposed to a slightly higher temperature during sintering before the temperature profile of the box evens out. It is possible that due to the tighter initial packing in these structures, sintering was able to progress slightly further, emphasizing these effects. The cusped tip can be attributed to specific parameters of the deposition trial that yielded these structures. Parafilm, a non-porous stopper was used, and in conjunction with the higher surface tension of the DIW solvent, it is possible that trapped air pockets in the channels helped to create the cusp shape, or it could simply be an artifact of the suspension's meniscus. The variations between the images in Figures 4-1, 4-2, and 4-3 discussed above are due to differences in the specific fabrication trial processes that produced the structures. These images thus show the feasibility of fabricating electrospray emitters by using the methodology outlined in Chapter 3. In fact, it is evidenced that characteristics of the emitter structures can be altered by adjusting the fabrication process. This 64 can be applied to create structures that are the best suited for electrospray emission. MIT5871 2013/07/03 11:06 L x300 300 um Figure 4-4: SEM cutaway image of an entire emitter structure, with both device layer and handle layer columns. While the emitter structures have thus been established, it is also necessary to verify that the columns continue into the handle layer. Without this component, the emitters would be left without a structural basis and more readily disattach from the silicon handle layer. Figure 4-4 shows an image of a whole emitter structure, with both FS and BS components. Normally, the BS is protected from etching along the sides of the mold by a buffer layer of silicon, but this picture is taken of a mold that had been broken in half yet still filled. This structure happened to be very close to the break line and therefore was successfully etched by the XeF 2 . The oxide layer, as expected, has been untouched by the XeF 2 and can be seen as the flat plate around the emitter structure. The BS column is clearly present, sintered, and attached to 65 the FS column. It is also connected to a thin film of sintered microspheres that lies on the BS surface of the array. This film connects all of the emitters on the BS and helps to transport ionic liquid to the emitters. Another feature of interest is a ring of sintered beads just under the oxide surface with a diameter larger than the BS column. This layer is caused by DRIE overetching of the BS channel when fabricating the molds; it can be exposed by an extended BOE etch of the oxide layer. This region is actually beneficial, as it serves to increase the area that can catch and reuse ionic liquid spillover, preventing shorts. There is a slight gap in the overetched region that can be improved by better deposition, but the full emitter structure can be verified as intact by Figure 4-4. 4.2 Arrays A few high-yield arrays have been fabricated thus far, as indicated earlier, though none with a diameter of 50Am. One 75pm diameter array is shown in Figure 3-9, and a 25pm diameter array is shown in Figure 4-5. Variability across an array is an important metric for the fabrication process. In Figure 4-3, (c), (d), and (e) were from the same array (although it wasn't high yield); it can be observed that (d) appears to be slightly shorter than the other two structures. This shows then that the filling was not consistent across the array. The arrays in Figures 3-9 and 4-5, however, appear to be very uniform. The 25pm array was characterized by measuring 56 of the 480 emitters using white light interferometry. The results are summarized in a box-and-whisker plot in Figure 4-6. The red line gives the median, the blue box marks the 50% of all measurements closest to the median, the black whiskers show the range of the data, and the red crosses indicate any outliers. The plot shows that the bottom and top diameters are both very precisely controlled, as shown by the small ranges. The bottom diameter has a little more variance; this could be due to any amount of overetching of the FS channels. The top diameter is also definitively smaller than the bottom diameter; the medians are separated by about 3pm. This indicates that even though it wasn't 66 MIT9115 2013/10/28 10:38 x100 1 mm Figure 4-5: An array of emitters with a diameter of 25pm. A broken emitter from mishandling the array can be seen at the top. noticeable, there is indeed some tapering of the structures in the 25pum emitters as well as the 50p-m emitters. As explained previously, this is caused by the sintering process. The height of the structures is more varied than either of the diameters, with a wider range and several outliers. The fact that a full quartile of the measurements lies above the 60pm line is unusual, as 6pm is the thickness of the device layer and therefore the maximum possible height. It may be indicative of some measurement error. Nonetheless, 50% of the measured structures had a height between about 57pum and 60Mm; this shows that the filling was largely uniform. In fact, all but three of the emitters had a height above 54pm. These three outliers, then, represent the variability of the current fabrication process. 67 Top Diameter [pm] 65 + - . 60 - . . . - -.- . 50 - - - 55 - .- 45-.+.... -~40 -- 30- - 2520 . ... - - - 35 .. .... --- Bottom Diameter Bottom Diameter fpmI Top Diameter [pm] Height [pm] Figure 4-6: Box-and-whisker plot of white light interferometry measurements of array shown in Figure 4-5 is on the right. On the left is a representative schematic image of an emitter structure with the median dimensions given in the plot. 4.3 Characterization of sintered porous glass for elect rospray applications The arrays fabricated thus far using the first iteration molds are not compatible with the iEPS frames, so alignment with an extractor grid for firing is not feasible at this point. However, the sintered porous glass material is tested for compatibility with ionic liquid, and a single emitter tip fashioned out of the sintered glass is fired to determine emission characteristics. 4.3.1 Filling with ionic liquid It is necessary to determine whether or not the sintered porous glass is wettable by ionic liquid, as it needs to be to function as an electrospray emitter. A simple wetting test, then, was conducted using 1-ethyl-3-methylimidazolium tetrafluoroborate (EMIB4,an ionic liquid commonly used as an electrospray propellant. A few drops of EMI-BF 4 was deposited on the BS of a sintered glass array and allowed a few seconds to travel to the tips of the emitters. A resultant wetted emitter is shown in Figure 68 4-7. After subsequent cleaning of the array in acetone to remove the ionic liquid, the array remained intact and undamaged. This indicates that the sintered soda-lime glass material is likely wettable and compatible with ionic liquid. MIT6039 2013107106 18:47 L x5O0 200 um Figure 4-7: Porous sintered glass emitter filled with EMI-BF 4 . The ionic liquid is the darker substance surrounding the glass microspheres. 4.3.2 Experimental setup In order to determine the suitability of the sintered glass material for electrosprays, a single emitter tip was fashioned with a razor blade out of a piece of the porous glass. The tip was mounted in an XYZ stage, as shown in Figure 4-8, and filled with a small drop of EMI-BF 4 . The extractor, a 1.6mm diameter hole in a stainless steel plate was affixed over the tip, and the tip was then aligned to the hole manually at a distance of about 1mm away from the extractor. This setup was mounted in a vacuum chamber, as shown in Figure 4-9. A Faraday cup with a secondary electron suppression grid was placed 6.5mm downstream of the extractor to collect the emitted current. Current was also collected at the extractor plate to measure any current intercepted by the extractor. A high voltage power source was connected to the emitter through the copper stage, while the extractor was grounded. The two following characterization 69 tests were conducted at a pressure of 5 x 10- 6torr. Figure 4-8: Sharpened glass tip clamped in an XYZ stage. Faraday cup Suppression grid Extractor r Vw V U E~~-~u--------U U I Glass porous emitter I (f _ I I I I XYZ stage Ceramic rod / Vacuum chamber Ammeter A A Hiah voltaae source HV GND Figure 4-9: Schematic of experimental setup in vacuum chamber. 70 4.3.3 Current-voltage curve First, an I-V analysis was performed to determine the current emitted at various applied voltages. A triangle voltage wave was applied to the emitter, with an amplitude of ±2.5kV and a period of 10s, as shown in Figure 4-10 (a). The currents measured in response to this signal are shown in Figure 4-10 (b). Voltage Applied 0 5 Time [sI 0 10 (a) Voltage signal applied to emitter. Single Sintered Glass Emitter -Current collected by Faraday cup -Interception current 87 6 5:3 0 432 1 -200 0 -500o -2000 -1000 -100 0 0 1000 200 1000 2000 3000 Voltage between emitter and extractor [V (b) I-V curve. Figure 4-10: I-V curve of a single sintered glass emitterl. It can be seen from the I-V curve that emission is fairly symmetric across the two polarities. The startup voltages are 1601V and -1713V, and both reach a maximum current of just over 6pA. These results can be compared to previously measured emission from a single porous tungsten emitter[5], shown in Figure 4-11. It can be seen 71 that the emission from the glass emitter is comparable to that of the tungsten emitter in both polarities; however, the glass emitter has significantly higher startup voltages, which are unfavorable. However, this can be attributed to the fairly significant distance between the emitter tip and the extractor hole as well as the hand-sharpened emitter tip that likely had a relatively large radius of curvature at the tip. Overall, these results indicate that the sintered glass could be a promising candidate for electrospray propulsion. Single Tungsten Emitter 6 2 5 00 -3 -2 -1 0 Voltage (kV) 1 2 3 Figure 4-11: I-V curve of a single porous tungsten emitter[5]. The I-V curve in Figure 4-10 has a few other interesting features. Firstly, there is a non-zero resting current that switches in polarity across the OV line. This is an artifact of the inherent capacitance of the vacuum between the Faraday cup and the emitter. As Q = CV, = I= C dt . Thus, the ramping up or down of the voltage generates a small amount of current. Additionally, it is notable that there is a non-negligible interception current. It could be caused by the manual alignment of extractor and emitter, but careful inspection shows that the interception current appears to begin at higher voltages than the current collected by the Faraday cup does. Taking the ratio of interception cur72 rent to the Faraday cup current, then, yields Figure 4-12. The relative amount of current being emitted is increasing with the magnitude of the voltage. This, then, indicates that the active emission region along the emitter tip is growing; as the voltage increases, additional emission sites, or electrified mensici, are being formed further down the emitter. Since the glass emitter is triangular, these emission sites are firing more or less directly into the extractor plate. While ideally, there would be no interception current, the current shown in Figure 4-12 is mostly a function of the emitter tip geometry, and the emitters in the molded arrays have a very different shape. 73 Single Sintered Glass Emitter: Current Ratio - Negative Polarity :3 0.8 h 0.6 ................... . .. . . .. .. . .. . . ... . . ... ... ... . . .. - 0.4 . . ......... .. ... .-.-.. .. - -.. ...... .. - .........- ...... ~0.2 'V -2900 -2500 -2400 -2300 -2200 -2000 -2100 -1900 Voltage between emitter and extractor [V] (a) Negative polarity. Single Sintered Glass Emitter: Current Ratio - Positive Polarity o0.8 cc K ....................... ... ... ........ ............ ........... ................................ L ~0.2 .... ... ..... : ................ .. ............... ............................. ........... ..... C k- ...... . . . . . . . . . . . . . . . . . . . .. . . . . . . . . 1800 1900 2000 .. ................. . ......... 2100 2200 2300 2400 Voltage between emitter and extractor [V] (b) Positive polarity. Figure 4-12: Relative interception current with respect to the current collected by the Faraday cup. This is also fairly symmetric in both polarities. 74 Current stability 4.3.4 In order to determine if the emitter could output a stable current, it was then fired for 10 hours. A square voltage wave with an amplitude of ±2.5kV and a period of 60s was applied, as illustrated by Figure 4-13 (a). The current response is plotted in Figure 4-13 (b). Voltage Applied 5 5, ca 0 -5 0 20 40 60 Time [s] (a) Voltage signal applied to emitter. 2 1.5 1 0- Collected current - Extractor current - 1 0.5 J. I 0 U I.. I -0.5 -1 -1.5 -20 2 6 4 8 10 Time [h] (b) Current response over time. Figure 4-13: Current stability for a single sintered glass emitter fired for 10 hours. In Figure 4-13, there is an initial startup transient current that is slightly higher than the stable current it settles to. It fires at this stable current, 1pLA, for about 75 4 hours before the current begins a slow decay over the final 6 hours, appearing to run out of fuel, which is likely, since no fuel feed system was connected. This can be corroborated by integrating the current to estimate the mass of propellant that was fired. Doing so yields a value of about 1.5 x 10- 5g, which indeed corresponds to the mass of EMI-BF 4 that would be expected to fill a thin triangular emitter about 2mm wide, 1mm tall, and roughly 50% porous. This test demonstrates that the porous sintered glass material is able to emit a steady, stable current. Moreover, the analysis indicates that the emission beam is largely composed of ions, although no time-of-flight analysis has yet been performed to verify this. 76 Chapter 5 Discussion and future work In this work, a full fabrication process for building electrospray emitter arrays out of soda-lime glass microspheres was developed. The fabricated structure arrays were characterized, and the material was fired to determine suitability for electrospray propulsion. The results indicate that this fabrication method is highly promising. Future work should compare the firing characteristics of a full sintered array with those of the state-of-the-art electrospray emitter arrays. This process can also be utilized to investigate electrospray technology improvements in densification and wafer-level processing. 5.1 Material evaluation The soda-lime glass itself is chemically inert and therefore well-suited for use in space, while the sintered glass material has been demonstrated to be competitive in porosity, at an approximately measured 43%, with the porous nickel that Courtney[5] worked with. It has also been shown to have competitive emission characteristics, as compared to porous tungsten. As such, the material can be deemed appropriate for electrospray applications. However, there are some possible non-idealities in the sintered soda-lime glass. For one, the particle distribution is distinctly polydisperse, with a full range of over 10pm in particle diameters. This then leads to polydisperse pore sizes. While not 77 necessarily a problem, this may, for example, lead to preferred flow pathways within the bulk sintered material and possibly favor certain emission sites over others. This can be ameliorated in part by filtering out the larger particles from the population. In fact, if successful, applying a pore size gradient by filling the tops of the emitters with smaller particles and the bottom with larger particles could serve to better draw ionic liquid toward the tips and also better tailor the emitters for emission in the pure ion regime. Additionally, while a structural analysis confirmed the stability of the structures during electrospray operation, to be used in space, the devices must also be able to survive launch. These structures have been shown to break when the array is dropped, striking a hard surface, but a pure vibrational analysis has not been performed. Such results would be desirable for future development of the material. 5.2 Process evaluation The full fabrication process presented in Chapter 3, including mold fabrication, deposition, sintering, and de-molding, has successfully produced not only single emitter structures but several full arrays, which demonstrates that it is reliable. Furthermore, the second iteration molds with an increased device layer thickness were still able to be processed and developed, resulting in an array with a high fill rate. This shows that the process is then also fairly flexible; individual parts of the process can be swapped out without major changes in the rest of the process. A few areas of improvement, then, are identified below. Figure 4-5 shows that the over-etched region forming a porous base around the emitter is not consistently exposed. While this is not a problem, this region may be highly beneficial to electrosprays in terms of being able to catch spillover ionic liquid. If this is the case, the BOE etch step may need to be better defined or extended to expose it in all the structures. Additionally, alignment of the molded arrays is difficult. A complicating factor is in fact the XeF 2 etch step, as it removes the entire device layer; any visual alignment 78 features, then, cannot be used. The mechanical alignment feature included on the BS of the second iteration molds is a good alternative, but it is subject to the tolerances of the fabrication processes, and the alignment joint is not a tight fit. Testing will be required to determine the appropriateness of this alignment. The sintering process has been proven to be very robust, as indicated by the numerous emitter features produced, but any residual variability in the process can be removed by re-adjusting the process for a furnace that is closed to the environment. Meanwhile, the deposition process, while shown to be reasonably flexible in producing arrays of different heights using the same deposition parameters, will need some amount of alterations to yield fully uniform arrays with the second iteration mold designs. Moreover, it has been shown that the choice of solvent for the deposition glass microsphere suspension may be able to affect both the packing of particles and the macrostructure geometry. The sintering profile can also control the progression of sintering and therefore the progression of pore closure. Thus, by tuning the deposition in combination with the sintering process, it may be possible to produce a range of permeability profiles in the sintered glass and thereby allow for the selection of a profile best suited for ion emission. Finally, it may be beneficial to add a processing step to sharpen the emitter tips. The columns produced in this work have a flat top, which may not be optimal for electrospray, as the sharp corners are directed outwards and not necessarily through the extractor hole. A quick dip in HF may be able to smooth out the corners and form more conical shaped emitters, as in the work of Coffman[2] and Courtney[3, which can better direct emission through an extractor hole. 5.3 Future work The arrays produced in this work are aimed at enabling the improvement of electrospray technology in the two areas identified below. The second iteration molds, in particular, allow for future testing and development steps. 79 5.3.1 Densification The second iteration molds have several mold designs that enable densification studies. There are two array designs with decreased pitch sizes, as identified in Table 3.2. One features structures with a 5 Optm diameter and a 300pm pitch, and the other has 25pm diameter structures with a 250pm pitch. All have the same BS diameter, 200pm. Additionally, molds for an alternate emitter geometry - ridges - have been processed. A ridge may be able to allow for densification to the natural separation distance between individual emitter sites. The ridges in these molds have a FS width of 50m, a BS width of 200pm, and are each separated by 400pm. Arrays using these molds can be fabricated using the process described herein, and the arrays can then be characterized. It may be necessary to develop matching extractor grids for firing. 5.3.2 Wafer-level integration The second iteration molds also have a feature to potentially allow for wafer-level integration; this is the reservoir pool. In the first iteration molds, it became evident that a film of beads would form on the BS surface, often to the very edge of each mold die. A wafer-level deposition, then, would lead to a film of beads across the entire BS surface. Upon dicing of this mold wafer, then, it can be imagined that the BS film of beads could shatter, possibly breaking off the embedded structures. This reservoir pool, then, is intended to constrain the formation of the film to each die individually. Then, a wafer-deposition would have many separated BS films, and dicing through edges of the silicon molds would not cut through any glass. The reservoir, then, in addition to directing deposition, is also a feature to enable future wafer-level integration experimentation. 80 Bibliography [1] Natalya Brikner and Paulo C. Lozano. The role of upstream distal electrodes in mitigating electrochemical degradation of ionic liquid ion sources. Applied Physics Letters, 101(19), 2012. [2] Chase Coffman, Louis Perna, Hanqing Li, and Paulo C. Lozano. On the manufacturing and emission characteristics of a novel borosilicate electrospray source. In 49th AIAA/ASME/SAE/A SEE Joint Propulsion Conference and Exhibit, 2013. [3] Daniel G. Courtney, Hanqing Li, and Paulo C. Lozano. Emission measurements from planar arrays of porous ionic liquid ion sources. Journal of Physics D: Applied Physics, 45, 2012. [4] Daniel G. Courtney, Hanqing Li, and Paulo C. Lozano. Electrochemical micromachining on porous nickel for arrays of electrospray ion emitters. Journal of MicroelectromechanicalSystems, 22(2):471-482, April 2013. [5] Daniel G. Courtney and Paulo C. Lozano. Ionic Liquid Ion Source Emitter Arrays Fabricated on Bulk Porous Substrates for Spacecraft Propulsion. PhD thesis, Massachusetts Institute of Technology, 2011. [6] Zhigang Zak Fang and R. M. German. Sintering of advanced materials: Fundamentals and processes. Woodhead Publishing Limited, 2010. [7] Blaise Gassend, Luis Fernando Velisquez-Garcia, Akintunde Ibitayo Akinwande, and Manuel Martinez-Sanchez. A microfabricated planar electrospray array ionic 81 liquid ion source with integrated extractor. Journal of Microelectromechanical Systems, 18:679-694, 2009. [8] W. D. Kingery, H. K. Bowen, and D. R. Uhlmann. Introduction to Ceramics. New York: Wiley, 2nd edition, 1976. [9] Renato Krpoun. Micromachined Electrospray Thrusters for Spacecraft Propulsion. PhD thesis, Ecole Polytechnique F6derale de Lausanne, 2009. [10] Renato Krpoun and Herb R. Shea. Integrated out-of-plane nanoelectrospray thruster arrays for spacecraft propulsion. Journal of Micromechanics and Microengineering,19, 2009. [11] Robert S. Legge. Fabrication and characterization of porous metal emitters for electrospray applications. Master's thesis, Massachusetts Institute of Technology, 2008. [12] Hanqing Li, Daniel G. Courtney, Pablo Diaz Gomez Maqueo, and Paulo C. Lozano. Fabrication and testing of an ionic electrospray propulsion system with a porous metal tip array. In Solid-State Sensors, Actuators and Microsystems Conference (Transducers), 2011 16th International,2011. [13] Pablo Diaz Gomez Maqueo and Paulo C. Lozano. Electrospray emitters for diffusion vacuum pumps. Master's thesis, Massachusetts Institute of Technology, June 2011. [14] Juergen Mueller, Richard Hofer, and John Ziemer. Survey of propulsion technologies applicable to cubesats. Technical report, Jet Propulsion Laboratory, 2010. [15] E. M. Rabinovich. Review: Preparation of glass by sintering. Journal of Materials Science, 20:4259-4297, 1985. [16] George C. Soulas, Matthew T. Domonkos, and Michael. J. Patterson. Performance evaluation of the next ion engine. In 39th AIAA/ASME/SAE/ASEE Joint Propulsion Conference and Exhibit, 2003. 82 [17] John Stark, Bob Stevens, Matthew Alexander, and Barry Kent. Fabrication and operation of microfabricated emitters as components for a colloid thruster. Journal of Spacecraft and Rockets, 42(4):628-639, July-August 2005. [18] Luis Fernando Veliisquez-Garcia, Akintunde Ibitayo Akinwande, and Manuel Martinez-Sinchez. A planar array of micro-fabricated electrospray emitters for thruster applications. Journalof MicroelectromechanicalSystems, 15:1272-1280, 2006. 83