STICTION REDUCTION AGENTS STUDIES USING QCM

Transcription

1 Abstract HUSSAIN, YAZAN AHED. Stiction Reduction Agents Studies Using QCM. (Under the direction of Christine S. Grant.) The problem of stiction in microelectromechanical systems (MEMS) is highly limiting their fabrication and functionality. The problem occurs during the fabrication, release stiction, as well as during the use of the devices, in-use stiction. Anti-stiction agents are currently an active area of research in the MEMS field to address this issue. These agents are primarily deposited in the form of self-assembled monolayers (SAM s) on the substrate to change its surface properties; to reduce what is known as the stiction problem. One commonly used SAM is the octadecyltrichlorosilane (OTS). On silicon substrates, the most used material in MEMS fabrication, OTS has shown high effectiveness in reducing devices stiction. The use of a quartz crystal microbalance (QCM) as an analytical technique for studying the OTS anti-stiction agent is presented in this work. Using the QCM as the primary analytical tool, we were able to extract comprehensive information about the formed SAM. The dependence of SAM deposition on the bulk phase concentration of the deposit solution is shown. A rough estimation of the adsorption kinetics rate constants were calculated, and the equilibrium constant was determined from their values. The equilibrium constant shows the high favorability of OTS deposition on silicon substrate compared to the reverse desorption process. The complex nature of the OTS SAM and its formation mechanism were also shown. These conclusions were made based upon comparison between the more robust SAM system of thiols on gold and the OTS on silicon. Finally, the interaction between the OTS and a vapor-phase lubricant (tertiary-butyl phenyl phosphate, TBPP), for friction reduction, was studied. Preliminary QCM results show a change in the adsorption of lubricant on bare silicon compared to OTS coated silicon. In addition, the lubricant film is believed to have higher slippage when OTS was present as an underlayer.

2 STICTION REDUCTION AGENTS STUDIES USING QCM by Yazan A. Hussain a thesis submitted to the graduate faculty of north carolina state university in partial fulfillment of the requirements for the degree of master of science department of chemical engineering raleigh November 2003 approved by: Dr. Christine Grant chair of advisory committee Dr. Jacqueline Krim Dr. Saad Khan

3 Biography Yazan A. Hussain was born in Zarqa, Jordan on the 20th of January, He is the fifth of four brothers and one sister. He studied the first five years of school in Saudia Arabia, after then he came back home, where he finished his school. In 1995, Yazan joined Jordan University of Science & Technology, where he obtained his B.S. with honors in Chemical Engineering in In 2001, he came to the USA, where he joined the Chemical Engineering Department at NCSU as M.S. student. He joined Dr. Grant s group, where he worked in the MEMS field. ii

4 Acknowledgments First, I thank my parents, my brothers, and my sister for being my family. I love you! I also thank my advisor, Dr. Christine Grant, for her support during this work, her understanding, and her patience. I also thank the former researchers, Manoj and Jeremy, who started this project, and my current group-mates, it is a pleasant group to be with. I thank Dr. Krim and her students, Manju and B, for the discussions we had and for supplying the data for the lubricant uptake. I also thank Chang, from Dr. Parsons group, for teaching me how to use the AFM. And Micheal, from Dr. Genzer s group, for teaching me how to use the ellipsometer. Finally, I acknowledge the NSF support for this work. iii

5 Table of Contents List of Tables List of Figures vii viii 1 INTRODUCTION Historical Background Microfabrication Bulk Micromachining Surface Micromachining LIGA & Micromolding Materials in Microfabrication Single Crystal Silicon PolySilicon Silicon Dioxide Silicon Nitride Silicon Carbide Diamond Germanium III-V Materials Metals Limitations and Challenges in MEMS Project Overview PRELIMINARIES Stiction Mechanisms Capillary force Solid bridging Van der Waal s force Electrostatic forces Hydrogen bonding Asperity deformation force Adhesion Under Capillary Forces Current Stiction Solutions Process modification Surface modification techniques Surface energy reduction by organic monolayers iv

6 2.4 Self-Assembled Monolayers (SAM) SAM s characterization techniques Merits of SAM s Common SAM systems Organosilicon SAM s EXPERIMENTAL & MATERIALS Quartz Crystal Microbalance Impedance analysis Factors affecting QCM behavior Applications of QCM Experimental setup Contact Angle Atomic Force Microscopy Experimental Procedure Cleaning QCM experiments Contact angle experiments Atomic force microscopy Materials RESULTS & DISCUSSION Validation of Experimental Technique Selection of verification system Results & Discussion Conclusion Studies on the frequency effect on adsorption Theory Approach Results & discussion Conclusion Signal Stability & Analysis Noise reduction Temperature control f determination Thiol Adsorption on Gold Why thiols? Compounds used Previous QCM studies on thiols Results & discussion Silane Adsorption on Si Literature survey on silane studies Results & discussion Kinetics studies Conclusion v

7 5 CONCLUSIONS & FUTURE WORK Conclusions Future Work List of References 106 A Derivation of Electrical Components For QCM 116 B QCM runs summary table 121 vi

8 List of Tables 1.1 Bulk vs. surface micromachining Comparison between different release techniques Kinematic viscosity measured by viscometer and QCM B.1 QCM runs for C 12 E 8 on Au crystals vii

9 List of Figures 1.1 Examples of positive and negative tone photoresists Schematic diagram of bulk micromachining Schematic diagram of the LIGA process Examples of fabricated MEMS devices Capillary forces caused by trapped liquid Solid bridging Equivalent contact model of two rough surfaces Comparison between stiction forces as a function of separation Contact angle relation with wettability Spring-plate-liquid system used to model adhesion in MEMS Constant volume curve showing the total energy change versus vertical spacing Branching diagrams for the spring-plate-liquid system Clamped beam adhering to its substrate Self-assembled-monolayer representation Convention used to represent 2D SAM structure Parameters used to represent the 3D SAM structure OTS adsorption mechanism Typical piezoelecric devices Assignment axes of qruartz crystal Butterworth-Van Dyke equivalent circuit for unperturbed quarts microbalance QCM crystal used in this work QCM crystal s holder used in this work Setup used in QCM experiments Ramé-Hart goniometer model used for contact angle measurements Adsorption isotherm of C 12 E 8 on gold Langmuir plot for C 12 E 8 on gold in aqueous solution at 25 C Adsorption isotherm for C 12 E 8 on 5 MHz gold crystal to 9 MHz Scaled adsorption isotherm for C 12 E 8 on 5 MHz gold crystals to 9 MHz crystals plus the oscillation effect Comparison between obtained and reference data Contact angles for Au slides coated with different concentrations of C 12 E 8 at 25 C Shear waves generated by an oscillating surface Scaled experimental f to account for oscillation effect The difference in f between oscillating and non-oscillating crystals viii

10 4.10 The effect of switching the oscillator ON/OFF on frequency for gold QCM crystals AFM images for 5 and 6 MHz crystals before and after C 12 E 8 deposition Comparison between in-situ and ex-situ runs QCM response to surfactant addition and its relation with signal stability Stirring effect on QCM signal noise The change in solution temperature due to surfactant injection Temperature effect on QCM signal stability Some frequently used thiols for SAMs QCM signal response to different thiols QCM response to thiols with different chain lengths Langmuir isotherm fit to a thiol adsorption run Si QCM crystals response to OTS addition at different concentrations f for 5 MHz Si crystals resulting from the addition of different OTS concentrations The initial rate of change of f for 5 MHz Si crystals as a function of OTS concentration Chang in crystal frequency upon immersion in different solvents Relation between change in QCM frequency and solvent properites Comaprison between f for OTS, thiols, and C 12 E Change in overall V for the three systems used in this work Langmuir isotherm fit to OTS adsorption Determination of OTS adsorption rate constants f for the uptake QCM measurements of vapor phase lubricant TBPP Change in amplitude for the uptake QCM measurements of vapor phase lubricant TBPP ix

11 Chapter 1 INTRODUCTION The problems that arise when a certain device is scaled, either up-to-down or bottom-up, are due to two facts. First, the change in dimensions will be combined with a respective change in surface area-to-volume ratio. For example, if a cubic box of 1 m length is scaled down to 0.5 m, the surface area-to-volume ratio will change from 6 m 1 to 12 m 1. Therefore, the corresponding external forces that act on the body will change in its total magnitude. The buoyancy force, for instance, becomes larger or smaller than the weight. Second, the nature of interaction between surfaces will also change, and, for very small objects, atomic forces start to play an important role. The nature of such forces is not completely understood. Currently, the technology to produce very small devices is available, and some sophisticated structures have actually been made, although their use is still very limited because of the two facts mentioned above. In some instances, these devices, with one or more dimensions on the micron scale, are electrically actuated to generate mechanical movement, and is known as MEMS, which stands for microelectromechanical system. Indeed, the surface area-to-volume ratio is very high for MEMS, and surface forces have proved to play a crucial rule in their operation. Beside the traditional friction forces, the problem of stiction also arises. Stiction is defined as the failure of the restoring forces of the body to overcome the surface forces which will damage the device. The surface problems confronting the MEMS technology are severe. Several research fields are becoming involved in an attempt to find a practical solution. The field of microtribology, which studies surface forces at the very small scale, is one new field concentrating on such studies. In addition, chemical, mechanical and electrical engineering, and material science are also involved. Such multidisciplinary involvement indicates the volume of the problem, and the importance of finding a solution for the friction and stiction problems. To summarize, the following statement was taken from the reports of Proceedings of the NSF/AFOSR/ASME Workshop held in Ohio in 1998 [59]: 1

12 Chapter 1. INTRODUCTION 2 At this point in time, being able to avoid stiction in a surface micromachined device might determine if a company will make it or not. This project is directed toward the study of a potential solution technique for stiction and friction problems in MEMS. This introduction continues with a short historical background about the industry and its development. Then, the major processing techniques for MEMS are described. A brief review of common materials used in the fabrication process is then introduced. It is important to understand the reasons behind the choice of such materials, although materials with better mechanical properties exist. The last two sections will review current opinions about the status of MEMS technology and its challenges, and will describe the general scheme of the project, its goals, and its strategy. 1.1 Historical Background By the 1960s, the integrated circuit (IC) industry had made considerable progress since its first appearance at Bell Labs in the forties. Difficulties in the production of pin point contacts greatly limited the miniaturization process and eventually led to the development of junction transistors. Since the forties and through the next decade, many important advancements were made; a significant one was the invention of transistors which was recognized by the Nobel Prize. Another important advancement, upon which the current IC industry was built, was the oxide masking process developed at Bell Labs in 1954 [45]. Not long after these inventions, the idea to extend this technology to more sophisticated geometries arose. In fact, miniaturized devices were already invented, but their dimensions were too large to be considered micro-devices. Inspired by the performance of biological systems and their ability to perform functions and store information within their microscopic volumes, R. Feynman discussed the possibilities of making miniaturized mechanical devices. Although he was building on the techniques available during his time, Feynman made spectacular speculations about the development of the miniaturization industry in terms of both manufacturing and potential processing and operating problems [22, 23]. Since Feynman s time in the late sixties, the industry has witnessed several technological advancements. Currently, the name microelectromechanical systems (or MEMS) refers to such miniaturized instruments. However, current technology is still unable to produce 3-D miniaturized devices that can perform the function of equivalent larger devices. This inability is due mainly to mechanical limitations such as fast wear rates and structure adhesion. These will be discussed in greater detail later.

13 Chapter 1. INTRODUCTION Microfabrication Microfabrication is currently accomplished by three major techniques, namely, bulk micromachining, surface micromachining, and micromolding [60]. A brief description of each process follows Bulk Micromachining In bulk micromachining, features are sculpted in the bulk materials, such as silicon, quartz, SiC, and GaAs by orientation dependent (isotropic) and/or orientation independent (anisotropic) etchants. This technique was the conventional one utilized in the IC industry; therefore, it was also adapted and utilized, in microfabrication of MEMS, first among other techniques from the IC industry [60]. The initial steps for bulk micromachining start by coating the the substrate, usually Si, with a layer of photosensitive material. The main component of photoresists is a polymer; this polymer changes structure upon exposure to radiation, usually ultraviolet light. A secondary component is a solvent which enables spin coating application and the formation of a thin layer on the substrate. In the photoresist, the sensitizer controls the photochemical polymerization reaction. Two types of resist exist: positive tone and negative tone. Positive tone means the material exposed to radiation becomes soluble in the solvent of choice, while the opposite is true for negative tone resist. Examples of both tones are shown in Figure 1.1. CH 3 O OCH 3 C O C O [ ] CH 2 C CH 2 C CH 2 C O C O n CH 3 O OCH 3 a. Poly(methylmethacrylate), PMMA N 3 X N 3 b. Bis(aryl) azide-sensitized rubber resist Figure 1.1: Positive mode, a, and negative mode, b, photoresist.

14 Chapter 1. INTRODUCTION 4 A pattern is transfered to the resist layer by exposing it through a mask to radiation with a specific wavelength. The mask is made of nearly optical flat glass or quartz plate, with a metal absorber pattern. The soluble portion of the photoresist is stripped or removed from the surface. The stripping can be accomplished by wet stripping with a liquid such as a strong acid (e.g., H 2 SO 4 ) or acetone. Photoresist stripping can also be conducted via a dry process. This occurs through exposure of the resist to plasma (e.g., oxygen), ultraviolet radiation, or a combination of both [60]. The pattern of the stripped photoresist will be a negative image of the resultant structure of the substrate. This structure is created by exposing the system to additional processes, consisting of either etching of the substrate material or the deposition of different materials on the substrate. The etching and deposition processes can be achieved by several techniques. For example, plasma etching, ionetching, deep reactive ion-etching (DRIE), non plasma dry etching, and wet etching (by KOH as instance) are some common etching techniques for silicon based systems. In contrast, silicon doping and physical or chemical vapor deposition (PVD and CVD) are examples of the techniques used for the addition of materials to the film. A schematic of the wet bulk process is shown in Figure 1.2. Examples of devices that are manufactured utilizing the bulk micromachining technique include electrochemical sensors, electrochemical valves,and self-aligned vertical mirrors Surface Micromachining At first glance, surface micromachining may look similar to bulk micromachining. However, some sharp differences exist between the two techniques. In bulk micromachining, the three dimensional structure is built by etching the substrate (e.g., polysilicon or single crystal silicon) whereas in surface micromachining, the structure is built through layer by layer deposition. Also, in surface micromachining, shapes in the x,y-plane are not restricted by crystallography as is the case for bulk processes. Table 1.1 compares bulk and surface micromachining. Some of the limitations associated with surface micromachining, as outlined in Table 1.1, have been overcome by process modification and/or alternative designs [60]. In addition, new processes, such as silicon-on-insulator (SOI), hinged polysi, and SUMMiT had enriched the surface micromachining techniques. Examples of devices that are made by this technique are the analog devices, accelerometers, and TI mirrors LIGA & Micromolding LIGA combines sacrificial molding with x-ray lithography and electrodeposition. The lithography here is based upon synchrotron orbital radiation (SOR) which, in the early 1990s, was expected to

15 Chapter 1. INTRODUCTION 5 Coated resist Substrate Imaging radiation Mask Latent image Distributed catalytic photoproduct, C C C C C C C C C Exposure Postexposure bake Positive image Negative image Figure 1.2: Schematic representation of a generalized chemically amplified resist process [60].

16 Chapter 1. INTRODUCTION 6 Bulk Micromachining z-dimension restricted by wafer thickness. No annealing needed. Devices can be built from single crystal Si. Can use crystallography for dimensional control. Stiction is not a concern. Surface Micromachining z-dimension restricted by deposition technique used. Annealing at high temperatures needed. Only Polysilicon can be used. Crystallography dimensional control not possible. Surface stiction possibility. Table 1.1: Comparison between bulk and surface micromachining techniques [60]. be commercially used by the early 21 st century. However SOR does not present a large business opportunity due to its high cost [60]. In this process, a low atomic number, thin membrane is used as a mask. Mask production for x-ray is one of the most difficult aspects. The pattern is made by means of a high atomic number compound, usually Au. The mask is first aligned to the substrate. However, the alignment process is difficult because an x-ray mask does not allow visible light to pass through. For conventional lithography, the alignment is done by visually matching a specifically made triggering points on the mask with the substrate. As for the substrate, it is required to be electrically conductive or initially plated with a conductive layer. This requirement is for the electrodeposition process. Examples of substrates used include aluminum and silicon wafers with a thin titanium top layer. After this process the substrate is covered with an x-ray resist material. Poly(methylmethacrylate) (PMMA), Figure 1.1, shows excellent contrast and good stability during deep-etch synchrotron radiation lithography. However, low lithographic sensitivity and susceptibility to stress cracking made it necessary to modify the PMMA resist or replace it with other chemicals. Resist application can be done by different techniques. Multiple spin coating and PMMA sheets are examples of commercially available techniques. The resist is developed by exposing to x-ray. This process is illustrated in Figure 1.3. In the final step, the resist is etched using a solution that has a high resist/developer dissolution ratio in the exposed and unexposed areas. An example of a typical solution made by KFK, Germany, which consists, on a volume basis, of 20% tetrahydro 1,4-oxazine, 5% 2-aminoethanol-1, 60% 2(2-butoxyethoxy) ethanol, and 15% water. To stop the development, less concentrated developer solutions are successively applied to prevent precipitation of dissolved resist. The resulting structure is then used as a final product, or it can be used in subsequent metal deposition. An example where

17 Chapter 1. INTRODUCTION 7 X-ray Mask Absorber Exposure PMMA Substrate Developing Metal Galvaning Molding Plastic Figure 1.3: Schematic diagram of the LIGA process.

18 Chapter 1. INTRODUCTION 8 LIGA is utilized is in the fabrication of electromagnetic micromotors. 1.3 Materials in Microfabrication This section will discuss the materials that are used in MEMS fabrication. The term materials in microfabrication does not include the chemicals used during the process, such as the photoresist or etchants for example. The discussion here includes those materials used in the MEMS fabrication which are considered as an integral part of the design, performing a mechanical, or other, function, or materials used in the process of construction of MEMS, for example, etchant stop Single Crystal Silicon Single crystal silicon has a diamond crystal structure, an electron gap of 1.1 ev, and can be doped to alter its conductivity. Silicon is a brittle material, with a Young s modulus comparable to that of steel. Also, silicon is widely abundant on earth which makes it inexpensive [26]. Single crystal silicon has several advantages. In terms of manufacturing, low-cost production processes with very low defects densities exist. The crystallographic orientation of silicon also gives rise to the capability of anisotropic etching [60, 26]. Uses of single crystal silicon include its use for substrate materials as a mechanical platforms on which device structures are fabricated, or as an etch stop PolySilicon Polysilicon, PolySi, is the most widely used material as the primary structure for MEMS devices in surface micromachining. The mechanical properties of PolySi are comparable to that of single crystal silicon. Similar to single crystal silicon, polysi can also be doped with different dopants (e.g., Phosphorous or Boron). Doping polysi not only changes its electrical conductance, but also changes other properties, such as: (i) amorphous-crystalline transition temperature, and (ii) the nature of stress levels found in polysi film (i.e., compressive to expansive or the reverse). One advantage of polysi is the ease of doping compared with single crystal silicon. Another advantage is the ability to conduct in-situ doping, which results in lower resistivity levels. Difficulties with polysi come from the hard reproducibility of a thin film due to the many parameters that affect its properties. For example, the electrical resistance and thermal conductance of polysi depend on the grain structures. This, in turn, depends on different factors including

19 Chapter 1. INTRODUCTION 9 deposition temperature, doping, and deposition method. PolySi is found in several sensors which utilizes its piezoresistive nature in the sensing element Silicon Dioxide One advantage of silicon is its ability to grow a stable oxide film, that is silicon dioxide (SiO 2 ), which acts as a packaging for the device. Silicon dioxide is a dielectric material which can be doped, as is the case for silicon and polysi. However, doping here is not intended to increase the conductivity, but to perform certain functions, such as, lowering the melting point to facilitate smooth topographies. Silicon dioxide is also used as a sacrificial material, insulating intermediate material, protecting devices from impurities and scratch via passivation. Doped silicon oxide serves also as a solid diffusion source to dope silicon with phosphorous Silicon Nitride Silicon nitride has similar properties to silicon dioxide, i.e., ability to be doped, stability, and dielectricity. It can passivate the silicon against moisture and mobile ions. While silicon nitride is a dielectric material, the mechanical properties of silicon nitride are better than those of silicon dioxide; it has a Young s modulus comparable to that of silicon. Etching resistance and residual stresses in silicon nitride depend highly on the deposition technique. Silicon nitride is used as a passiviation barrier to the diffusion of moisture and mobile ions, such as Na +. It is also used as optical waveguides, insulators, etch masks, and as a mechanical part due to its high Young s modulus Silicon Carbide Silicon carbide, SiC, is a polymorphic material, in which some properties may differ from one structure to another. In general, silicon carbide has an electronic band gap of about 3 ev, twice that of silicon. It also has a high thermal conductivity and high breakdown field, higher stiffness than silicon, and higher resistance to most etchants used in micromachining. The properties of silicon carbide make it a potential candidate to replace silicon for harsh environments, e.g., high temperature or wear rates. Therefore, silicon carbide is applied for hightemperature and high-power applications. Its properties also make it attractive to micromachining resonators and filters. However, the use of silicon carbide is limited due to the difficulties to bulk micromachine.

20 Chapter 1. INTRODUCTION Diamond Diamond has exceptional physical properties, such as, hardness, high-wear resistance, and low coefficient of friction. Diamond also has a very large electron gap, making it suitable for high-temperature operation. Beside, it can be doped to make it semiconducting. Young s modulus of diamond is very high. Finally, diamond can be made as an amorphous or polycrystalline material. Diamond is used in the atomic force microscopy cantilever probe. The fact that it is hard to micromachine diamond limits its use in microfabrication. The inertness and strength of diamond make it a potential material for micromold fabrication [88] Germanium Germanium has participated in the semi-conducting materials industry since the early development of transistors. Deposition of germanium can be carried out by low-pressure CVD (LPCVD) at lower temperatures than polysi (325 C). Residual stresses in deposited germanium can be nearly eliminated by annealing for 30 seconds at 600 C. Besides, germanium is resistant to major Si etchants, and it can be doped to yield a electrical resistive material. On the other hand, germanium oxide is water-soluble, which limits its development as a microelectronic devices material. Also, it does not nucleate on SiO 2, which prohibits its use as a substrate. Applications of germanium include its use as a sacrificial layer and as a substrate layer in polysi surface micromachining. Thermistor and Si 3 N 4 membranes are examples of products that utilize these two techniques. An eutectic Al/Ge alloy was also used to create high quality patterned bonds between two silicon dices [100]. Recently, applications of SiGe alloy has received attention. The advantage here is to lower the deposition temperature while keeping silicon comparable material properties. This last fact makes the fabrication processes for SiGe to follow the polysilicon processing technique. Structural elements, such as gimbal/microactuator were made from such an alloy. Finally, GeAs is another germanium based material of interest in microfabrication. This material is better for thermal isolation and higher temperature applications. GeAs possesses a piezo- and optoelectrical properties, which favor the integration of micromachined devices with electrical circuits for fast signal processing. The use of GeAs is limited due to its high cost III-V Materials GeAs has already been discussed in the last section. This is one example of materials that can also be listed under III-V materials category. Other materials like InP is also attractive for use in sensors

21 Chapter 1. INTRODUCTION 11 and optoelectronic devices. The III-V compounds have favorable piezoelectric and optoelectronic properties, high piezoresistive constants, and wide electron band gaps (relative to Si). Fabrication of these materials is straightforward comparable to that of silicon. Some examples of devices built based on InP are the multi-air gap filters and torsional membranes Metals The most widely used metal in microfabrication is aluminum [60]. Aluminum is used as a structural layer in most cases; however, it is also used as a sacrificial layer. Aluminum can be deposited at low temperatures, which makes it possible to be used with some polyimide, where such combination can be used for subsequent fabrication of the IC on wafer. Work on the application of aluminum in microfabrication includes the utilization of anodic aluminum oxide to fabricate high aspect ratio structures [71], fabrication of thin film metallic glasses, e.g., Zr 76 Cu 19 Al 6 [52], as a reflective mirror in micromachined spectrometer [108], and tweaking the properties of aluminum via ion implantation of oxygen in the aluminum layer. Tungsten is another example of metals in microfabrication. It has some unique mechanical properties. It also has the property of being able to deposit selectively on silicon or metal surfaces but not oxides and nitrides. At Sandia National Laboratories, tungsten is being studied as a coating material for silicon to reduce stiction and wear [61, 24]. Other metals used in microfabrication are gold and nickel. Gold is used in contact points, as a high reflectivity, as a high mass density, and as adhesion layer material [60]. Studies on nickel, on the other hand, include its use in electroless deposition (as an alloy with copper) onto MEMS devices [75], in the fabrication of micromachined switches [109], implantation with titanium to enhance its properties, i.e., reduce wear and friction [70], and, finally, in deposition with iron in the form of alloy called Invar to produce low thermal expansion metal [32]. Few other metals are also used in MEMS, for example, titanium and silver. However, studies in this area (use of metal in MEMS) are still limited; as a result, their benefits are yet to be determined [60]. 1.4 Limitations and Challenges in MEMS The previous section briefly summarized the materials involved in MEMS fabrication. Silicon compounds clearly occupy a central role in the process. Heavy reliance on silicon stems, in part, from the fact that these materials are prevalent in the fabrication of ICs. As a result, film deposition and

22 Chapter 1. INTRODUCTION 12 etching technologies are readily available [26]. However, the inherent differences involved between the operation of ICs and MEMS impose few, but important, limitations on the new industry. MEMS devices are characterized by their large surface area to volume ratio, as have been mentioned. Hence, surface forces in MEMS can play a large role relative to gravity and inertia forces, typical dominant forces at the macro scale [19]. Surface, or interfacial, forces, such as capillary, chemical, van der Waals, and electrostatic attractions, make every type of micromechanical devices susceptible to stiction [57]. Stiction occurs when the aforementioned forces exceed the restoration forces in the microstructure, which causes a permanent adhesion between two surfaces of the microstructure. Surfaces in microstructures can come into contact either during fabrication or during normal operation. As mentioned earlier, MEMS fabrication relies on the micromachining technologies from the IC industry. Today, commercial silicon microfabrication techniques consist primarily of wet processes. In these processes, the devices are formed by pattern transfer via successive steps which are performed in liquid media, e.g., etching [60]. At the end of the wet process, the devices are removed from the liquid medium, during which the capillary forces from the draining liquid causes deflection of the surfaces, which may bring them into contact. This process is known as release stiction. The second situation where stiction may occur is during normal operation of the device; this is called in-use related stiction. This type of stiction may result from the unintentional contact between surfaces during their operation. This contact is due to (i) acceleration or electrostatic forces [57], (ii) physical shock during storage, (iii) electrostatic or capillary action caused by the presence of water vapor in the surrounding atmosphere [19]. The contact between the microstructure surfaces may, as well, be a part of its normal operation. This is the case for rotating parts or sliding beams for example. When surface forces exceed restoration forces, the device will adhere to the substrate. However, even when the surfaces fail to adhere, another problem results from the contact of moving surfaces, that is friction. There are two primary negative effects of the resulting friction. First, friction results in the requirement of high break-away forces during startup, for example, in micromotors [27]. The second and most serious is wear. Currently, wear presents the main obstacle for the successful commercialization of MEMS [93]. Because of the small clearance between the microdevice surfaces (e.g., hub and rotor in micromotors) the resulting wear debris may stop the device and cause a permanent performance malfunction [57]. In this instance, similar to friction, wear may increase the chances for surfaces to stick to each other [47]. In summary, micromachines are currently greatly limited by tribological problems, which limits

23 Chapter 1. INTRODUCTION 13 their commercialization. Current micromachines available in the market try to avoid contact during normal use [93]. There is a strong sentiment in the MEMS research community regarding the importance of the development of basic tribological knowledge. This is crucial if the MEMS industry is to expand to applications of devices more complicated than sensors [47]. (a)six-gear Chain (b) Silicon Mirror Assembly. (c) Triple-Piston Microsteam Engine. (d) TI Digital Micromirror Device. Figure 1.4: Examples of MEMS devices produces by Sandia National Laboratories (a,b,c) [49] and Texas Instruments (d) [34]. 1.5 Project Overview Microtribology is an emerging field in surface studies. It is concerned with the application and/or modification of traditional tribological science, the study of friction and wear at macroscale levels, to a microscale level. The tribological challenges in microfabrication consist of two parts. The first part is the release-related stiction which arises during the fabrication process. The second part occurs during device operation or, more generally, after release. This important fact suggests that for a solution to be practical, it must address both problems. Release- and in-use-related stiction exhibit different mechanisms (i.e., different acting forces).

24 Chapter 1. INTRODUCTION 14 This difference complicates solving the problem and requires separate studies on each of them. Another important difference results because, first, release is usually a wet process, while the resulting devices are normally used in dry environment. Also, while release is often done at low temperatures; (e.g., room temperature) the devices may be required to operate at the harsh conditions associated with higher temperatures. As a result, there is a need for a comprehensive technique that addresses both problems. Unfortunately, the majority of current techniques and those being researched are either focused on the release-related stiction or, in the best case, perform poorly during device operation. This project studies a potential, two-step approach to solving both release- and in-use-related stiction. The goal is to enable reliable microfabrication processes, with an emphasis on operation in harsh or extreme environments. For this purpose, the project is divided into two sections, release- and in-use-related stiction studies. Release-related stiction has been studied by the Grant group, through what seems to be the most potential solution of surface modification via the application of certain organic molecules as a monolayer on the top of the surface. In-use tribological problems, on the other hand, are to be studied by our collaborator, Dr. Krim, from the physics department. The Krim group is working on the application of a vapor phase lubricant (tertiary-butyl phenyl phosphate, TBPP) which has demonstrated a high quality performance at elevated temperatures. The strategy for this work combines two techniques, one for release-related and one for in-userelated stiction, so that we can minimize the drawbacks and limitations of one by the other. The drawback is the low durability and reproducibility of release-related stiction solution (i.e. selfassembled-monolayers) when the devices are set to use. The lubricant to be applied here is more powerful for this purpose, and also known to sustain high temperatures, but the limitation is due to its method of application, which must be carried out in gas phase. Therefore, the interactions between the two chemicals and the quality of the final resultant devices are the ultimate objective of this work.

25 Chapter 2 PRELIMINARIES 2.1 Stiction Mechanisms Stiction can be defined as a phenomenon which causes two proximity surfaces to yield unusually high adhesive forces between the normal and tangential directions due to such effects as the presence of a thin liquid layer at the contact interface or a potential differences across the interface [47]. Number of effects can cause surfaces to adhere, some of which occurs before putting the device in use, i.e., during fabrication and storage, this is called release-stiction. The second type occurs during the use of microdevices and is termed in-use stiction. The first type, release-stiction, is the concern of this part of the research done by our group. Common stiction mechanisms are discussed next. These mechanisms are either attractive forces that pull the surfaces closer to each other, or forces that enhance the adhesion once the surfaces are in contact Capillary force Liquids that make a contact angle of less than 90 with the substrate underneath will produce a negative Laplace pressure, and if it is trapped between two surfaces, it will pull them into contact [47]. In micromachining, the final steps usually involve the application of liquid pools. This starts by using HF as the etching solution, which is replaced with de-ionized water, to remove the HF and etching product. During this stage, also upon exposure to room air, a layer of oxide, typically 5-30 Å, is formed on the silicon surface. The presence of hydroxyl groups residue on the surface produces a high surface energy, which generates a strong capillary force [58]. Laplace pressure, for the case of small vertical spacing compared to the lateral dimension, is 15

26 Chapter 2. PRELIMINARIES 16 given by [58]: P L = 2γ L cosθ (2.1) d where γ L is the liquid surface tension, θ is the contact angle, and d is the vertical separation. This is shown in Figure 2.1. θ Figure 2.1: Liquids tend to form a meniscus when a solid-liquid-vapor are in contact, θ, the contact angle, represents the tendency of liquids to spread on the solid surface [47]. When θ < 90, the pressure will be negative pointing toward the center point, which may pull the surfaces closer. Therefore, a liquid with contact angle higher than 90 will be required to prevent such force. Also, the smaller the separating distance, d, the larger the pressure, either positive or negative. A detailed analysis was introduced by Komvopulus [47] where he included the surface roughness effect. Finally, it is important to note that capillary forces can also arise after fabrication. For example during storage when the devices are in contact with room air, where humidity may fill the vertical spacing with liquid [47]. It is interesting to note here that humidity have a positive effect in reducing wear, although it may enhance stiction [94] Solid bridging Non-volatile debris left after the evaporation of rinsing liquid, during the final release step for micromachines, can act as an adhesive enhancer when surfaces are in close proximity, as shown in Figure 2.2 [47]. Debris encountered during release may be introduced with the liquid, or through dissolution of particles and/or substrate by the etching liquid [58]. This problem can be avoided by forming chemical oxide on the surface to protect it from dissolution, followed by the deposition of hydrophobic (OH-terminated) monolayer, to prevent strong capillary forces [47, 58].

27 Chapter 2. PRELIMINARIES 17 Debris Figure 2.2: Debris remaining between microstructure surfaces may enhance the adhesion Van der Waal s force This force exist due to the interaction between instantaneous dipole moments. However, such interaction is retarded for large separation, approximately 20 nm [47]. Therefore, van der Waal s force can be thought, since vertical gaps in MEMS are much larger than 20 nm, to have similar effect as that of solid debris. Van der Waal s force can be estimated using the following relation [58]: Electrostatic forces L vdw = HA 6πh3. (2.2) Charges are known to accumulate from the ambient and migrate across insulating surfaces on silicon chips [58]. Also, electrostatic force arises on such charged surfaces due to externally applied voltage across the interface and differences in the material work function of the surfaces [47]. In IC industry, such surface charges were the source for device instability. This was solved by thickening the oxide and nitride insulating layers. Polysilicon, as well, can be passivated with self-assembled monolayer films anchored on the surface oxide [58]. The electrostatic forces can be calculated using the relation for parallel-plate capacitor: Hydrogen bonding L ei = ɛu2 A 2h 2. (2.3) For a fully hydrated silica surface, the density of hydrogen bonding sites was found to be 5 per nm 2. This will yield, approximately, 200 mj/m 2 adhesion energy [58]. The same authors reported that -OH can form strong hydrogen bonds, as the separation between surfaces becomes small.

28 Chapter 2. PRELIMINARIES Asperity deformation force This is a repulsive force that arises due to non-uniform surface topography, i.e., due to surface roughness. If contact between two surfaces occurs, the highest peaks of the surface will react with a repulsive force, due to its elasticity. The magnitude of this force depends on the surface height distribution function, the mechanical properties of the contacting solid, and the asperity interference that depends on the normal load [47]. See Figure 2.3. Figure 2.3: Equivalent contact model of two rough surfaces [47]. Figure 2.4 shows a plot of the three attractive forces, neglecting hydrogen bonding, as a function of the vertical separation. From the figure it can be seen that capillary forces dominate at larger separation ranges, about two order of magnitude larger than electrostatic force. Also, it should be noted that the magnitude of electrostatic and van der Waals forces become significant when the two surfaces are near contact. To prevent such forces from causing bending and stiction in the structure, it must be treated to reduce the magnitude these forces below the spring-restoring forces of the device material, normally silicon. Techniques for reducing these forces include the removal of hydrophilic groups, making the surface more conductive, and decreasing the actual contact area between the two surfaces by increasing roughness. The minimum (or critical) micromachine stiffness to prevent stiction was given by Komvopoulus [47] as: k = L max d h (2.4) where L max is the magnitude of maximum attractive force, d is the stand free separation between the suspended microstructure and the substrate, and h is the surface separation distance corresponding to the maximum attractive force. 2.2 Adhesion Under Capillary Forces Common use of micromachined structures is in pressure and acceleration sensors. In such devices it is desirable to fabricate suspended structures that have a minimum gap distance but large area

29 Chapter 2. PRELIMINARIES 19 Figure 2.4: Comparison of attractive forces per 1 mm 2 area existing at the dimensions of microstructures as a function of the separation between two perfectly smooth surfaces of silicon [58]. in order to maximize the transducer capacitance [65]. This structure, under certain conditions, can collapse and adhere to its underlying substrates. When a liquid is in contact with a solid, it will exert certain pressure, given by Young s equation (2.1). The resulting angle at the interface can be found from: γ LS = γ SV γ LV cosθ (2.5) at equilibrium. Here, γ represents the surface tension for the subscripted system, i.e., solid (S), liquid (L), and vapor (V), and θ is the contact angle. This is shown in Figure 2.5. When the system satisfy the criterion γ LA γ SL < γ SA it is energetically favorable for the liquid to spread on the surface of the solid. For the case of two parallel plates with a drop of liquid forming a bridge trapped between them, the surface energy stored, neglecting liquid-vapor interaction (since its area is much smaller than that for liquid-solid or solid-vapor) is given by: U s = 2[A SA γ SA + A SL γ SL ]. (2.6)

30 Chapter 2. PRELIMINARIES 20 Figure 2.5: Contact angle relation with wettability And, using Young s equation (2.5): U s = 2π[ro 2 γ SA + γ LA rl 2 cosθ]. (2.7) The force required for moving the plate small distance, dz, will be: ( ) du s F = dz (2.8) which, realizing that the volume of trapped liquid is constant, can be combined with (2.7) to give: The capillary pressure, q, will be: F = 2πγ l cosθ r2 l z. (2.9) q = F A = 2πγ cosθ l z. (2.10) A spring can be added to the system to represent the elasticity stored in the structure, Figure 2.6. Figure 2.6: Schematic of spring-rigid plate-liquid system. The structure is pulled toward the substrate as the liquid vanishes [65]. In the case where the volume of the drop is taken to be changing, the maximum volume of the

31 Chapter 2. PRELIMINARIES 21 fluid before it overflows, V, will be πr 2 h. For this case, the surface energy stored will be: { ) Us + 2πγ l cosθ( r 2 U s = V l πz z z ( ) U s + 2πγ l (1 cosθ) r 2 (2.11) V l πz z z where U s is a constant, and z = V l. πr 2 The total energy of the system, neglecting the plate weight, is: U T = U S + U E = U s + 2πγ l cosθ(r 2 V l πz ) k(h z)2 (2.12) where U E is the elastic energy from spring. Figure 2.2 shows the total energy for different liquid volumes. The curve has one or two minima. One minimum occurs at the break point of U s, implying that an equilibrium liquid radius is r. The other minimum results when dut dz equation (2.12): = 0 or, from These two minima can be written as: z 2 (z h) + 2γ l cosθv l k β : ξ = λ where ξ = V l V, λ = z h, and N kh c = 2 2πγ l. cos θr 2 The two minima, β and β ɛ, will intersect if the equation: = 0. (2.13) β ɛ : ξ = N c (1 λ)λ 2 (2.14) ξ = 1 2 ± ( N ) 2 (2.15) has real roots. Therefore, if N c < 4 the two curves will not intersect. This value, N c = 4, is denoted as N T. A new quantity is then defined as N EC = N N T, and is called elastocapillary number. The new number will have a threshold value at 1; if N EC < 1 the two curves will not intersect. The above analysis is useful in determining weather the suspended structure will bend under capillary force to an extent it touchs the underlying substrate or not. To clarify this, Figure 2.2 shows the total surface energy versus vertical spacing for different liquid volumes. If we start at point A, where z = h (spring relaxed), and then start the evaporation, the trajectory followed will be that of minimum possible energy. However, since the curves have more than one local minimum, the trajectory will be also dependent on the parameter N c (i.e., k,h, γ, and θ). Figure 2.2 shows the paths that will be followed by the system for two different N c values during the evaporation process. Once the structure is bent and in contact with the substrate, its adhesion will be determined by three factors; its elasticity, the solid-solid adhesion forces, and the solid-liquid surface forces. This

32 Chapter 2. PRELIMINARIES 22 Figure 2.7: Change in total energy (U T ) as a function of vertical separation (ɛ) for entrapped liquid with constant volume.

33 Chapter 2. PRELIMINARIES 23 Figure 2.8: Branching diagrams for the spring-plate-liquid system showing the extrema of total energy. The arrows show equilibrium trajectories [47].

34 Chapter 2. PRELIMINARIES 24 dependence can be derived by adding the three energy components (i.e., elastic, S-L interface, and S-S adhesion energies). The relations are [28, 66]: U E = EI 2 l/2 l/2 ( 2 u ) x 2 dx (2.16) U SL = 4γ l w cosθ(x l x s ) (2.17) U SS = 2γ s wx s (2.18) where U E is the elastic energy, E is the Young s modulus, I is the moment of inertia of the beam, U SL is the solid-liquid surface energy, U SS is the solid-solid adhesion energy, and x, x l, x s,w and u are as shown in Figure 2.9. Figure 2.9: Cross section of a doubly clamped beam adhering to its substrate [66]. Taking the sum off these three quantities will give the total energy of the system. Minimizing the total energy with respect to the lateral dimension, x, gives the following parameter [28, 66]: N p = k EH3 t 3 γ s l 4 (2.19) where k is a constant that depends on the geometry of the system, e.g., k = 128/5 for circular disk and 3/8 for suspended beam. N p is defined as the peel number. For N p > 1 the structure will peel, otherwise, it will adhere. As seen in equation (2.19), new system parameters are now playing role in determining its final state, those are: E (Young s modulus), t (structure thickness), and γ s (interfacial adhesion energy). Recall, from section (2.1) in this chapter, that number of forces affect the adhesion process. Such effects have to be taken into consideration. It is possible to do so either by introducing unique relations for each force, e.g., solid bridging effect on adhesion or van der Waal s interaction forces, or, possibly, their effect can be lumped with other system parameters presented above, e.g., solid bridging effect can be lumped with γ s, asperity deformation with E and so on. Such analysis was not found, and it is an interesting study to determine their effect on the accuracy of the model.

35 Chapter 2. PRELIMINARIES Current Stiction Solutions Many attempts were made during the last few years to deal with the problem of stiction. The problem was studied from different approaches. Those approaches can be divided into three major categories. The first category is where a modification is introduced to the fabrication process itself in order to eliminate the surface forces that cause stiction. The second category includes the methods in which surface topography or structure was altered. The final category is that treated the surface to reduce its energy by depositing a self-assembled monolayer. These three categories are discussed next Process modification The major cause of stiction in microstructure, as discussed, is the capillary forces that arise during the release of final structure. The liquid-vapor transition during release is the cause for such phenomenon. Therefore, attempts were made to avoid such transition, i.e., liquid-vapor transition, to eliminate the origin of the problem. Before discussing the techniques under this category, it is useful to mention some of the slight modification that was made to the release process. Originally, after the devices were etched with HF, de-ionized water was introduced to replace the etching solution and to remove etching debris. The water was then evaporated to get the final product [60]. Since silicon and other silicon compounds are hydrophilic materials, this process involved high capillary forces. Methanol was then used to wash the de-ionized water and replace it before evaporation. Methanol has a lower surface tension compared to water, versus mnm 1 [40]. Although better results were obtained by this method compared to water, the problem was not solved. Freeze (Sublimation) drying The liquid-vapor phase change is avoided here by first freezing the liquid, then subliming it to get the free structure. This method first appeared in 1989 [64]. In this method, a mixture of water and methanol, t-butylalcohol, or p-dichlorobenzen were used for sublimation [40, 64]. Although this process is relatively simple, the fact that the rinse solution can undergo a significant volume change upon freezing, enough to destroy the sample [64], and the need for a vacuum chamber and refrigeration are major drawbacks which limited its applicability. Today, freeze drying is rarely used, and has been replaced by the more successful supercritical drying technique [64].

36 Chapter 2. PRELIMINARIES 26 Supercritical drying In this technique, the rinsing water is gradually replaced by liquid CO 2 at elevated pressures inside a high-pressure chamber. The sample is then taken to the supercritical point of CO 2 where the liquid-gas interphase is avoided. This is a very successful technique for eliminating release stiction, with nearly 100% yields [63]. Also, commercial large scale supercritical dryers are now available [64], which is another advantage of the technique. However, two disadvantages present here. First, the process involves very high pressure (72 atm [64]) which requires additional safety precautions. Second, is the need for a complicated setup, in the price range, for small scale dryers, of $10, ,000 [4]. Vapor phase Etching Etching using vapor phase mixture of HF/H 2 O was introduced in 1966 [50]. In this technique, vapor mixture of HF/H 2 O is introduced in a chamber, where the structure is at relatively high temperatures (80 C). This is to prevent water condensation from the reaction mixture. The pressure of HF can be controlled which will affect the etching rate. Lee et al. [50] replaced the water in the etching mixture with methanol. This was done to avoid condensation of water during the reaction which may produce capillary forces. Yet, another modification was made by Cole et al. [16] to overcome the low etching rates in gas phase etching, compared to liquid HF. Cole and his coworkers combined both techniques, dry and wet, so that the bulk of the sacrificial structure will be quickly removed by wet etching, followed by dry process to complete the etching. Xenon difluoride (XeF 2 ) is another gas phase etchant. Whereas HF is used for silicon dioxide etching, XeF 2 is used for silicon etching [14]. Dry etching has the advantages of simplifying the etching process; no additional solutions are needed, better small-geometry penetration, and high selectivity. On the other hand, etching rates are slower than the wet processes, and it requires a batch process to control the pressure. Table presents a summary of different etching techniques [40] Surface modification techniques Since surface forces are the cause of stiction, it is, intuitively, expected that smaller contact areas will give lower forces. To reduce the contact area of a surface, it can be textured to generate rough interface. Indeed this method increased the detachment length by a factor of two. Another way to reduce the contact is to construct a periodic array of small supporting posts called dimples [64]. Yee et al. [106] proposed a new surface roughening technique for the purpose of reducing sticking

37 Release methods Evaporation drying Sublimation drying Supercritical Drying HF VPE SAM Coatings Procedure HF DI HF DI HF DI HF DI HF DI HF vapor HF DI methanol methanol methanol p-dichlorobenzene methanol CO 2 H 2 O 2 IPA CCl 4 OTS IPA DI Melting T ( C) N/A Boiling T ( C) N/A Vapor Pressure 20 C 27 C 20 C 25 C N/A C N/A (torr) 80 C 54.8 C Surface Tension 25 C N/A 1.16 N/A N/A (mn m 2 ) Advantages Simple Lower surface Fast subli- Only hot plate Clean No liquid in- No com- tension mation needed volved plicated than DI setup water Excellent results Excellent results Excellent results Disadvantages Medicore results Medicore results Requires refrigeration Toxic Complicated setup Complicated setup Many chemical and rins- and ing steps vacuum Absorbs water Needs vacuum Under development Contracts upon solidifying Table 2.1: Comparison between different release techniques [40]. Chapter 2. PRELIMINARIES 27

38 Chapter 2. PRELIMINARIES 28 of microstructures. They utilized the different oxidation rates between grain boundaries and grain for heavily doped n-type silicon. The process consists of two steps. First, the silicon is doped with phosphorous before being thermally oxidized. This oxide layer is etched, and only the thick layer over grain boundaries remains. Second, the polysilicon is etched to make the grain holes; the remaining oxide acts as a mask for polysilicon. Another process was proposed by Fujitshka and Sakata [25] where a silicon substrate with low resistivity, underneath a SOI layer, is etched to increase its roughness. Three features of this work were mentioned, those are roughening of substrate, formation of asperities, and increment of gap between SOI and substrate. Surface texturing has two major drawbacks. First, it cannot be applied to all systems, because some structures require a flat surface for operation (e.g., transducers). Second, dimples adhere stronger than expected by force relation to surface area and surface tension, which is believed to be due to the formation of solid bridge of SiO 2 as the water evaporates [64] Surface energy reduction by organic monolayers Applying ultra-thin organic films to solid substrate allows tweaking its characteristics [78], e.g., insulation or hydrophobicity. Such layers have very small thickness of an order of tens of angstroms. This is very advantageous in microelectronics, because the structures of the devices will have a thickness and spacing between different parts in the order of few microns. Therefore, the structure may be affected if thick films were deposited. In addition to their small thickness, two other features give self-assembled monolayers (SAM s) the expected potential for being a possible solution for stiction. First, the deposition of SAM s is relatively simple, compared to other films, e.g., polycrystalline diamond [27]. Simply, a monolayer can be deposited from a suitable solution in few minutes at normal temperatures and pressures. The second feature, and of most importance, is the fact that SAM can handle both release and in-use stiction. This last characteristic is a unique feature for SAM s and no other technique succeeded in doing so without, or at least with slight, effect on the structure. Silanes and flourosilanes show superior properties in terms of reducing the surface energy of silicon substrates [90]. Contact angles for water on silicon were largely increased when such films were deposited, and surface energy drops by 2 to 7 times that of bare silicon [90]. Alkene based monolayers show superior results in reducing surface energy. Compared to silanes, it has the advantage of being more robustly applied, but the deposition process involves heating at high temperatures (180 C) in an organic solution for an extended intervals of time [2].

39 Chapter 2. PRELIMINARIES Self-Assembled Monolayers (SAM) Self-assembled monolayers (SAM s) are molecular assemblies that are formed spontaneously by the immersion of an appropriate substrate into a solution of an active surfactant in an organic solvent [97]. A self-assembling molecule can be divided into three parts, form energetic point of view [97], head, backbone, and tail, Figure 2.4. The head group-substrate pair is used to define the individual SAM system [85]. Figure 2.10: Schematic of SAM. Shaded circle indicates chemisorbing headgroup and open circle endgroup, which can be chosen from variety of chemical functionalities [85]. To define a SAM structure, two features need to be determined [85]: 1. The 2D-structure of the molecules, i.e., their layout, with respect to the substrate. This is done using a convention of SAM film molecule orientation with respect to the substrate, as shown in Figure The 3D-structure of the molecular backbones which includes a possible tilt angle with respect to surface normal, the tilt direction, the twist angle, etc. Figure 2.4. It is important to note, however, that not all SAM s show crystallinity, therefore, the 2D structure may be amorphous SAM s characterization techniques Wide rang of analytical tools are used to study SAM s. In general, the analysis aims to determine the structure of molecules, i.e., chemical, electrical potential, etc., or to determine geometrical orientation, 2- and 3-D structures. A brief description of the most common techniques follows. Ellipsometry This is used to evaluate film thickness, which is usually the first criterion in SAM study [97]. Ellipsometer uses the fact that linearly polarized light will have different speed in its two perpendicular

40 Chapter 2. PRELIMINARIES 30 Figure 2.11: SAM 2D representation: relationships between surface and bulk meshes (above), and unit meshes of five possible surface nets (below) [79].

41 Chapter 2. PRELIMINARIES 31 Figure 2.12: Schematic of angular degrees of freedom of alkanethiol (bound to substrate via thiol group, with alkyl chain fully stretched). Angle θ t refers to tilt of molecular axis with respect to substrate surface normal. χ t defines tilt direction, i.e., it is derived from projection of molecule in substrate plane. χ t is undefined for θ t = 0. Twist angle, ψ, describes rotation about axis of molecule [85].

42 Chapter 2. PRELIMINARIES 32 wave components. The phase shift between the two components, using certain value for the refractive index, is then translated into thickness spectroscopy [95]. Microscopy techniques This category includes, e.g., surface tunneling microscopy (STM), atomic force microscopy (AFM), and scanning electron microscopy (SEM). The strength of such techniques is to provide a direct image of the structure [85]. Different information can be obtained using this technique. Besides probing the surface topography, AFM for example, can be used to probe the electrical conductivity or chemistry of the surface. Diffraction based techniques Diffraction techniques are used to study the 2D structure of SAM s. It includes low-energy electron diffraction (LEED), grazing incidence x-ray diffraction (GIXD), and low-energy atom diffraction (LEAD). Spectroscopy based techniques This includes infrared techniques, e.g., SFG and NXAFS, in which the waves interact with transient dipole which allow drawing conclusion on the structure. X-ray photoelectron spectroscopy operates at high energy levels that can penetrate deeper in the SAM, allowing the study of binding energy of the head groups. Spatially averaging techniques Techniques in this category works on average of a relatively large scale, enough to insure actual representation of the sample. However, such techniques do not give a detailed view of the structure. Contact angle measurements are one simple technique to measure surface energies and wetting behavior. Quartz Crystal Microbalance (QCM) is another example on this category Merits of SAM s Langmuir-Blodgett (LB) films are a mono- or multi-molecular film spread on a liquid surface and transfered to solid by immersion of the solid into that liquid [78]. LB layers suffer from some inherent limitations affecting the stability, homogeneity, and final structure of the resulting built-up film [73]. The interest in SAM s system started in the beginning of the eighties as a substitute of LB films to overcome some of its limitations. Sagiv et al. prepared a homogeneous monolayer that has the following properties [73]:

43 Chapter 2. PRELIMINARIES Unusual chemical and electrical stabilities. 2. Degree of perfection in terms of molecular density, packing, and orientation. 3. Monolayers can be prepared from chemisorbed, physisorbed, or combination of both molecules. 4. Molecules are attached to the surface (e.g., via covalent bonds) and it is intralayer polymerized. Therefore, SAM s can, in principle, be formed on top of a solid surface which can then be separated from the monolayer. This will yield a true 2D surface of one molecule width. In addition, the formed layer will epitaxial transfer information from a molecularly organized solid surface [73, 74]. As a result, from research point of view, SAM s offer unique opportunities to increase fundamental understanding of self-organization, structure-properties relationships, and interfacial phenomena [99]. From a practical point of view, SAM s have the following unique characteristics [85]: 1. Ease of preparation. 2. Tunability of surface properties via modifications of molecular structure and functional groups. 3. The use of SAM s as a building blocks to more complex structures, e.g., multilayer building. 4. The possibility of lateral structuring in the nanometer regime. 5. The applications made possible by these features Common SAM systems Several systems have been used to form SAM s. Among those, SAM systems of thiol (sulfur containing organic compounds) on Au(111) [85] surface are probably the most popular. Organosilicon SAM s on silicon substrate, or silicon dioxide, is another example of well studied systems. These two systems will be discussed next. Thiol/Au systems This is probably the most studied and understood among SAM systems. This is due to the ease of preparation and well-defined order, plus the inertness of the substrate, gold [85]. While thiols SAM can be formed on several metals, e.g., silver, copper, mercury, GeAs, and InP, gold is preferable due to its resistance to oxidation in air, which have made gold substrates convenient and reproducible for laboratory experiments [98]. Several thiols (and organosulfur compounds, in general) have been used to prepare SAM s on gold. Examples are alkanethiol, dialkyldisulfides, dialkylsulfide, alkyl xanthate, and dialkythiocarbonate [99]. In addition, the terminating group of the thiol compound can also be varied to produce

44 Chapter 2. PRELIMINARIES 34 different surface properties. For example, a high surface energy may be produced with a CO 2 H group, lower energies can be obtained using OCH 3, and, yet lower energies are obtained with CH 3 terminating group [98]. The most important factor in thiol deposition is the gold substrate condition. The contaminates on the gold surface are known to cause large variations in its wettability [89]. These contaminations are generally carbonaceous materials, and techniques like UV/O 3 or oxygen plasma are a common methods to remove these contaminations. However, such techniques cause the formation of oxidized surface which cause a systematic variation of the formed layer [83]. The preparation of thiol SAM s is usually done from liquid solutions of, typically, 1 mm concentration. Ethanol is most commonly used for this purpose. Other solvents include hexane, cyclohexane [98], and water [83]. Contact angle measurements, however, showed a dependence on the used solvent [83]. The mechanism of thiol adsorption starts either by ionic dissociation of thiols or by the formation of H, according to the following reactions [98]: RSH + Au RSAu + H 2 (2.20) RSH + Au RSAu + H (2.21) While desorption takes place according to one of the following schemes: 2RSH+2Au 2RSAu + H 2 (2.22) k d RSH + Au RSSR+2Au. (2.23) k d The hydrogen in reaction (2.22) is generated by the association of two hydrogen radicals generated in reaction (2.21). a k a k The increase in the system order, due to molecular organization to form the SAM, will be surpassed by a negative change in enthalpy, making the total Gibbs energy ( G) negative, indicating a spontaneous process. The situation is reversed for the desorption reactions (2.20) and (2.21), with higher energy barrier for disulfide formation than that for thiols. Therefore, it is more probable for the desorped molecules to be converted back into thiol molecules than disulfide. The over all SAM formation reaction will be [98]: RSAu (s) + 1 k d 2 H 2(solv) (2.24) RSH (solv) + Au (s) k a The Adsorption kinetics for thiols on gold in n-hexane solution was studied using quartz crystal microbalance (QCM) for in-situ measurements [98].

45 Chapter 2. PRELIMINARIES 35 With respect to adsorption kinetics, it is believed that the adsorption takes place in two distinct steps. The first step is a very fast step, which lasts a few minutes, by the end of which the layer will have a contact angles close to their limiting values and a thickness of about 80-90% of its maximum. The second step is a slow step which lasts several hours, at the end of which the thinkness and contact angles reach thier final values [99]. The first step is well described by the Langmuir adsorbtion isotherm: dθ dt = k a(1 θ)c k d θ (2.25) where θ is the fraction of occupied sites, and k a and k d are the adsorption and desorption rate constants, respectively. The structure of the formed SAM on Au(111) surface 1 was studied using electron diffraction technique. The structure is believed to be a hexagonal with a S S spacing of 4.97 Å per molecule [99]. However, other crystallinties were recently observed, namely, the dense ( 3 3) packing. The packing of molecules strongly depends on the surface morphology of the substrate and the end group of the thiol molecules [85] Organosilicon SAM s Different silicon containing organic compounds, silanes, have been deposited on many hydroxylated surfaces. Alkylchlorosilanes, alkylalkoxysilanes, and alkylaminosilanes are examples of deposited compounds. Substrates on which the layers were formed include silicon oxide, aluminum oxide, quartz, glass, mica, zinc selenide, germanium oxide, and gold [99]. Alkylchlorosilane compounds have been receiving increasing attention because of thier superior performance in reducing stiction problem and surface energy for silicon based materials [57, 39]. This project emphasizes the potential anti-stiction coatings for MEMS devices, therefore, alkylchlorosilane SAM s on Si substrates will be discussed in more details. Alkylchlorosilanes Octadecyltrichlorosilane (CH 3 (CH 2 ) 17 SiCl 3, OTS) is perhaps the most common anti-stiction agent that is currently being studied. When properly applied, OTS can achieve the following characteristics [57]: 1. Eliminate release stiction by effectively reversing the shape of the water meniscus. 2. Reduce in-use stiction by three to four order of magnitudes 1 This morphology is commonly used for thiol deposition experiments because it is the most stable among different Au morphologies.

46 Chapter 2. PRELIMINARIES Eliminate the need for large input signals in the start-up phase. 4. Reduce friction in microengines. 5. Reduce wear significantly. However, since OTS has three chlorine atoms in the head group, it is highly sensitive to water; the main draw back for OTS [99]. Another trichlorosilane used is 1H,1H,2H,2H-perfluorodecyltrichlorosilane (CF 3 (CF 2 ) 7 (CH 2 ) 2 - SiCl 3, FDTS). This is the most effective reagent to prodcue hydrophopic coating on oxide-coated surfaces [90]. The surface tension of a FDTS coated substrate is about 4 times lower than surfaces coated with OTS, and 2.5 times lower than Teflon. However, FDTS is extremely sensitive to moisture and must be handled and stored under inert atmosphere, e.g., N 2 [90]. Cholorsilanes with two chlorine atoms have been recently studied as a potential anti-stiction coating for silicon surfaces. With two chlorine atoms, there is an improved control over polymerization. Example for such material is dichlorodimethylsilane (DDMS) [39]. OTS SAM formation process Since OTS is sensitive to water, the process to apply such material is quite challenging. While the presence of water in the solution may cause the OTS molecules to polymerize, forming siloxane groups, very small amount of water are required for the reaction to take place. The necessary amount of water is not exactly known, however, a value of 0.15 mg water per 100 ml solution was suggested [99]. Different deposition techniques are used to form the OTS SAM. Commonly, adsorption of OTS molecules from a liquid solution onto silicon substrates is used. In addition, gas phase deposition has also been done, as well as deposition from liquid CO 2 [21]. For MEMS release, it is only feasible to apply the film from a liquid solution, therefore, this technique will be discussed here. The deposition process is usually carried out in a mixture of hexadecane and chloroform solution (4:1 V/V). Chloroform acts as a cosolvent to increase the solubility of chlorine head [9], carbon tetrachlide my also be used for this purpose in place of chloroform. Other solvents, beside hexadecane, are used, those include toluene [5] and bicyclohexyl [87]. The immersion time for OTS monolayer preparation varies among different research groups. Intervals of 1 minute and up to 1 hour had been employed, but usually 10 to 15 minutes [87, 90]. The substrate must be prepared carefully. Water must be removed from the surface, usually by drying at high temperatures. After drying, successive cleaning steps are performed. During the cleaning process, the substrate surface must be hydroxylated to create the necessary deposition sites

47 Chapter 2. PRELIMINARIES 37 on the surface. Hydroxylation is usually done via immersion in piranha solution (H 2 O 2 :H 2 SO 4 ) [90], or exposure to UV/ozone [21]. Reproducibility and Robustness of OTS SAM Silanization reaction has continued to be plagued by a high level of irreproducibility in the macroscopic properties, i.e., wettability and contact angle, of the deposited monolayer [9]. In fact, the high sensitivity of trichlorosilanes, which causes it to polymerize, has been one important reason for searching for an alternative agent such as alkene [2] and dichlorosilanes [39]. The two main important factors that affect the silanization process are water content and temperature [9]. As mentioned, water is essential for this reaction, but excess amounts cause the OTS molecules to polymerize which is not only undesirable but, for MEMS fabrication, harmful. This is because the relatively large size of polymerized OTS particulates on the device surface will enhance stiction through solid bridging. Temperature, on the other hand, can be used to improve the layer quality. There exist a critical temperature below which superior film properties are obtained. This transition temperature has been reported to be approximately 28 C for OTS. For shorter chain lengths, this value decreases [90]. A study on the robustness of different chlorosilane films had been conducted by Iimura and Kato [33]. They concluded that the OTS monolayer is very stable against heating, organic solvents and acids. This stability decreases for di- and mono-chlorosilanes. It can be concluded that any modification on the OTS SAM formation technique will be highly desirable, and will certainly be a great advantage to the MEMS industry. Properties of OTS SAM s It was first proposed, in the beginning of the eighties, that self-assembly can produce monolayers of unique advantages unattainable by the LB approach [73, 74]. This proposal was based on a study done on silane monolayers with different number of functional groups. The presence of three functional groups at the OTS molecule head, i.e., three chlorine atoms, gives OTS the ability, besides bonding to the substrate, to crossbond with other OTS molecules as a result forming a crosslinked network. This is the reason for high stability of OTS layers. Other properties of OTS monolayers are: 1. High thermal stability, up to 400 C in N 2 ambient, and 150 C in air [90]. 2. High chemical resistance to most acids and solvents. However, basic solutions severely attack the monolayer.

48 Chapter 2. PRELIMINARIES The resulting surface will have a very low surface energy, 20 mn/m [9], and a large contact angle, > 110 for water. 4. The ability to bridge over irregularities and form uniform monolayers even on molecularly rough surfaces [73, 74]. 5. The resulting surface will have a low coefficient of static friction, 0.13 versus 2.3 for SiO 2 [91]. 6. Multilayers of OTS can be easily built by introducing an unsaturated bond at the alkane tail, which can be replaced, after deposition, with a hydroxyl group forming the base for the new layer [73, 74]. 7. Among the disadvantages, however, OTS had shown to enhance in-use stiction as compared to non-coated devices [57]. Mechanism of OTS SAM formation The scheme for the adsorption process is shown in Figure The adsorption starts by the diffusion of OTS molecules to the substrate surface. To become reactive, the OTS must encounter an -OH group. The molecule that is formed by the reaction between OTS and -OH group, referred to as active molecule, is unusual since, after its hydrolysis, it can react with both hydrolyzed (active) and unhydrolyzed (inactive) molecules. Therefore, a number of possibilities arise: 1. OTS molecule may encounter a H 2 O molecule in the solution or the H 2 O layer on the substrate surface hydroxyl groups have been shown to exist on the silicon surface, these groups are the reason for the presence of water layer at the surface [31] which will cause the activation of the molecule. 2. Active molecule can reach the surface at an active site, i.e., a hydroxylated silicon atom in the substrate, this will lead to OTS-substrate bonding and SAM formation. 3. Active molecule can reach the surface where it encounters an adsorbed molecule. The -OH group in the active molecule can then react with a Cl atom in the adsorbed molecule or -OH group. Of course, the reverse, i.e., an inactive molecule can reach the surface at a point where a molecule had been adsorbed and has a free -OH group, is possible. 4. Two active, or one active and one inactive molecules, can react in the solution to form a dimer. This process may be repeated and results in the undesirable formation of the polymer. It was though earlier that the grafted OTS molecules on the silicon surface will, at later stages of the process, crosslink and form a stable network. However, it was argued later that, due to

49 Chapter 2. PRELIMINARIES 39 Figure 2.13: OTS adsorption scheme [9]. steric effect between the neighboring molecules, such crosslinking is not possible [92]. Nevertheless, crosslinking seems to take part in the high film stability, but it my be occurring between grafted and ungrafted molecules. This fact also agrees with the island growth mechanism described below. Several studies had been conducted to examine the mechanism in which OTS layer grows. SAM s are known to grow either by uniform growth or island growth mechanisms. Uniform growth mechanism indicates that the layer will grow with even distribution on the surface, where the adsorbed molecules will have an initial high tilt angle which then decreases as the coverage increases. In contrast, island growth indicates the formation of small packed islands in which the molecules have a tilt angle approximately equal to its final value, and these islands will grow with time until maximum coverage is achieved. Atomic force microscopy (AFM) studies strongly suggest that the layer grows by island growth mechanism [5, 99]. Therefore, it was suggested that the molecules within the islands do not have enough freedom to rearrange themselves, which explains the high disorder in silanes monolayer compared to that of thiols, for example [99]. However, Carraro et al. [11] had shown, using AFM to image the surface, that the mechanism is a function of temperature. Three regions were identified: at low temperatures (< 16 C) island growth mechanism regime observed, at high temperatures ( 40 C) uniform growth mechanism

50 Chapter 2. PRELIMINARIES 40 regime, and a mixed regime at intermediate temperatures. Adsorption kinetics In contrast to thiols/au systems, no structure with long range order had been reported for OTS SAM s on oxidizied silicon, the possible crosslinking between adsorbed molecules can have a strong impact on the properties of the formed film, and the poor reproducibility of the prepared films make the study of growth difficult [85]. However, some qualitative analysis had been done. For example, it was found that the islands number density, at short time, is proportional to t 1/3 and that the island area to be proportional to t 2/3 [85]. The adsorption process can be written as [21]: s P (2.26) k ar C b k mc + n k af where C b is the bulk concentration of the reactant, C s is its surface concentration, n is the concentration of active sites, P is the adsorbed species prior to reaction, and k af and k ar are the rate constants for the reversible adsorption reaction. The rate for the system described in reaction (2.26) can be written as: R A = k m (C b C s ) (2.27) R B = k af C s n k ar P = k af C s N(1 φ) k ar Nφ (2.28) [ = k af N C s (1 φ) φ ] (2.29) K eq where φ is the fractional surface coverage, N is the total number of active sites available on surface, and K eq = k af /k ar. Beside the adsorption reaction for OTS, other competitive reactions also take place, those are: 1. Adsorption of OTS onto the substrate. 2. Reaction of OTS with H 2 O molecules in the solution. 3. Reaction with hydrolyzed OTS molecules. Therefore, an OTS molecule in solution can face different fates: 1. Diffuse to an active site on the surface, which lead to the chemisportion of the molecule, thus SAM formation. This point can also include the situation where the OTS molecule reaches at the surface and encounter an adsorbed molecule with a free -OH group. 2. Encounter a H 2 O molecules which causes OTS to hydrolyze and generates a HCl molecule, i.e., RSiCl 3 + H 2 O RSiCl 2 OH + HCl.

51 Chapter 2. PRELIMINARIES Encounter another active OTS molecule in the solution forming a dimer. 4. Encountering an inactive OTS molecule. However, it will be assumed here that the solvent is effectively dissolving the OTS molecules and preventing them from forming micelles. As long as the molecule has an active reaction group, -OH group, it will continue to be susceptible to steps 1 3. The following assumptions can be made for the OTS system in solution: 1. As the OTS molecules oligemrate, according to step 2, the diffusivity of the resulting oligomer decreases, according to Stokes-Einstien equation (D AB = κt/6πrµ B ). Therefore, it will be assumed that species with more than one molecule, i.e., all species other than single OTS molecules, will have poor diffusivity and will not reach the surface. 2. The diffusion rate for OTS is higher than the reaction rate [21]. Further discussion will be given in Chapter 4. Overcoming OTS sensitivity There have been several attempts to overcome the difficulties associated with OTS deposition. Even when the reaction is done in a highly controlled environment, the moisture content will still be difficult to control [2]. Therefore, the solution for such problem must directly address inhibiting the agent reactivity until the last moment of deposition, or reducing the reactivity. Three approaches were studied: 1. Alkylation of Si surfaces using a two-step halogenation/grinard route [6]. 2. Reducing the reactivity of the chlorosilane molecule by using a dichloro- instead of trichlorosilanes [39]. 3. Inhibition of silane polymerization in solution. Recently, a specific class of sulfonal functional silane was reported as a stabilizing agent for silanol, RSiOH, molecules [1].

52 Chapter 3 EXPERIMENTAL & MATERIALS As mentioned in Chapter 1, this project has two primary sections. This document deals with research on the first section, release-stiction, which was planned to progress according to the following approach: 1. Room temperature surfactant uptake studies using QCM on metal and polysilicon substrates. 2. Determine the relationship between surface/interfacial properties and: (a) Surfactant molecular structure. (b) Concentration of surface active agent in solution. (c) Substrate material. (d) Exposure time. (e) Method of submersion. (f) Degree of agitation. The work done here deals primarily with the first point; surfactant uptake of different surface active materials on silicon and gold substrates. The plan of research was as follows: 1. Validation of QCM setup and experimental procedure, to insure the absence of systematic error. This was done by conducting a set of experiments on a well-defined aqueous system of non-ionic surfactant uptake and comparing the results with a study in the literature. In addition, other analytical techniques were used, e.g., contact angle measurements, to validate the performance of the results obtained by our QCM s system. 2. Studies on chemical interactions between surfactant and substrate. This will provide information about the performance of QCM under chemisorption, i.e., chemisorbed molecules, on ideal 42

53 Chapter 3. EXPERIMENTAL & MATERIALS 43 SAM systems (thiols on gold). Thiols on gold is a robust monolayer system which was among the first monolayers that were studied [97]. Thiols have minimal sensitivity to surrounding conditions, such as humidity, which allows the separation of any variables that may affect the QCM response, e.g., solution viscosity or particle formation in solution. In addition, part of this project was originally dedicated to the study of stiction for metal substrates. Non-silicon materials are gaining more and more attention in the industry for their potential applications in MEMS devices. 3. Perform in-situ measurements to study the adsorption process kinetics. At this point, the experimental techniques used was verified and demonstrated the ability to follow a chemical adsorption (chemisorption) process for the robust system of thiols on gold. The next step was to start the analysis for an active surface agent that is being widely used, in research, to increase the contact angle of silicon and reduce its surface energy. This material is in the silane hydrocarbon family. The study will also include the effect of different parameters, e.g., temperature and concentration, on the deposition rate and on the quality, measured by defects density and coverage area, of the formed layer. 4. Determine the optimum deposition conditions for the selected silane. Optimum conditions are to be determined with respect to, first, robustness of the formed layer, second, the applicability and reproducibility of the deposition technique, and, finally, its compatibility with the in-use vapor phase lubricant to be applied. Three analytical techniques are used in this study. Quartz crystal microbalance (QCM), the primary technique, was used as an in-situ measuring technique to study the kinetics of film deposition under different conditions. Contact angle measurements were used to determine the wetability and surface energy of the formed layer. Finally, atomic forces microscopy (AFM) was used to study topography, coverage, and uniformity of the deposited film. The next few sections are dedicated to a theoretical discussion followed by experimental setup description of those techniques, with an emphasis on QCM, the primary technique in this work. 3.1 Quartz Crystal Microbalance Quartz is one of the ionic crystalline solids that crystallizes in structures lacking a center of inversion [53]. Therefore, a single crystal will possess a polar axis associated with the orientation of atoms in the crystalline lattice [10]. As a consequence, if the crystal is set under mechanical pressure, an electric signal will be generated. This is known as the piezoelectric nature of quartz. On the other hand, if an electrical potential is applied across the crystal, the polar dipoles will be displaced in

54 Chapter 3. EXPERIMENTAL & MATERIALS 44 a specific direction, which depends on the relative orientation of the electrodes to the crystal axis, causing elastic deformation in the crystal. This phenomenon is known as reverse piezoelectricity. If a certain potential is applied, the crystal will respond in a certain direction. If the applied potential polarity was inversed, while keeping the same magnitude, the crystal will respond with the same displacement but in the opposite direction. Hence, an applied AC voltage will cause a displacement in the crystal in a vibrational mode, exactly as a swing or pendulum would move under constant forces. The displacement of the crystal will depend on the angle at which the crystal was cut. This gives rise to different vibrational modes. Examples of these modes are thick shear modes, surface acoustic wave, flexural (or Lamb wave), and shear horizontal acoustic plate mode, as shown in Figure 3.1 [102]. Figure 3.1: Representations of typical formats for different piezoelectric devices and their corresponding particle motions. For the SAW and shear horizontal mode, the direction of particle motion is indicated by the arrows [102] The QCM operates under thickness shear mode. For this particular case, the crystal is cut at about 35 from the crystal axis as shown in Figure 3.2. Because of symmetry considerations, only the fundamental shear wave and its odd harmonics can be excited [36]. Therefore, only single frequency, referred to as natural or resonance frequency, will be able of exciting an ideal crystal,

55 Chapter 3. EXPERIMENTAL & MATERIALS 45 i.e., one that has no mechanical losses. However, mechanical losses do exist in quartz, although in a small magnitude. This fact causes the crystal to vibrate at a spectrum of frequencies instead of one discrete value. The quality factor, Q, is used to describe crystal ideality. It can be defined as the ratio of the resonant frequency to the full width of frequency at half the maximum amplitude of conductance [37]. Figure 3.2: The assignment of axes to a quartz crystal, and different common axes of cut. It can be concluded, from the previous paragraph, that the resonance frequency of a crystal occurs when the shear wave has a length equal to half the crystal thickness or its odd multipliers. The wave for this case will have a zero amplitude at the crystal s center, known as the wave s node, and a maximum at the crystal boundaries, know as anti-nodes. This can be written as: f = u q 2t q (3.1) where f is the resonance frequency, u q is the acoustic wave velocity in quartz, and t q is the quartz thickness. By simple manipulations, equation (3.1) can be used to derive a relation between the

56 Chapter 3. EXPERIMENTAL & MATERIALS 46 change in frequency to the change in thickness (or, in other words mass, if the density is known). This relation is called the Sauerbrey equation; written as [53]: f = 2f2 o ρq µ q m. (3.2) Here, f is the change in frequency due to a change of mass equal to m, ρ q is the quartz density, and µ q is the elastic shear modulus for quartz. This equation was behind the use of quartz resonators for sensing mass change. The main assumption, in order to use equation (3.2) to relate the frequency and mass change, is that the applied mass has negligible thickness with respect to the crystal [37]. This assumption and others are discussed in several references [53, 54, 102]. In QCM studies, a useful analogy is made between the vibrating quartz crystal (electromechanical) and an equivalent electrical circuit. If an ideal crystal (i.e., crystal with no internal friction) was excited by applying a DC voltage for a short time and then the voltage was removed, the crystal will continue to oscillate by virtue of its inertial and elastic energies [10]. This is similar to a swinging pendulum in frictionless surroundings, where the oscillation continues by cyclic transformation of potential and kinetic energies. In electrical circuits, the so called tank circuit shows an identical swinging behavior but in terms of current and voltage. Tank circuit is composed of an ideal capacitor and an ideal inductor in series [55]. For such circuit, if a voltage is applied for certain time and then released, the energy will keep oscillating between the inductor, which stores the energy in the form of magnetic field, and the capacitor, which stores the energy in the from of electrical potential. For a non-ideal crystal, friction imposes a resistance that will cause dissipation of the stored energy. Similarly, in the tank circuit, a resistance can be added in series to the inductor and capacitor, which will cause energy dissipation. The resulting aforementioned circuit does not represent the QCM system completely, one more component needs to be added. This is due to the piezoelectric nature of quartz, which couples the mechanical and electrical energies through the piezoelectric constant. A brief description of commonly used commercial QCM crystals follows to help explaining this point. Commercially available QCM crystals are usually made of a thin quartz desk with metal electrodes deposited on both sides. The quartz disk is usually cut in a range of angles with respect to the crystal axis. For quartz, this range is know as the Y-cut family, which includes the AT-cut (at ) and BT-cut (at 49 ) [77]. These two cuts are shown in Figure 3.2. Also, the disk is usually half an inch in diameter with portion of the central region fully plated on both sides [29]. Going back to the equivalent circuit representation, the two metal electrodes deposited on both sides of the crystal with the quartz sandwiched between them will act as a capacitor. Taking into consideration the fact that this is an electromechanical process, it becomes necessary to include this

57 Chapter 3. EXPERIMENTAL & MATERIALS 47 last component, the capacitance due to the metal electrodes and the quartz between them, although it does not directly interfere with the motional crystal. Because this capacitance does not depend on the quartz deformation, it should be added to the circuit in parallel. For this reason, the first group of components (the inductor, capacitor, and resistor in series) is called the motional arm, and the second component (the capacitance) is called the static arm. Figure 3.3 is a representation of this circuit, which is also know as Butterworth-Van Dyke equivalent circuit for unperturbed quarts microbalance [62]. Advantages for such analogy are [62]: 1. With only few lumped elements, this model simulates the electrical characteristics of the QCM over a range of frequencies near resonance. Ideally, the model should explicitly relate the circuit elements to physical properties of the QCM as well as surface mass layer and contacting liquid. This also allows easier prediction of the QCM behavior. 2. Standard circuit analysis software can be used to extract information from electrical measurements. C Rq C q L q Figure 3.3: Butterworth-Van Dyke equivalent circuit for unperturbed quarts microbalance For example, if the mechanical model for an unloaded QCM was solved and compared to the corresponding circuit model (see appendix A.1 for details), the following relations are obtained: C o = ɛ 22A h (3.3) C 1 = 8K2 o C o (Nπ) 2 (3.4) 1 L 1 = ωsc 2 (3.5) 1 where ɛ 22 is the quartz permittivity, A is the active area of quartz (the area of the smallest electrode), h is the crystal thickness, K o is a quartz constant related to the electromechanical coupling constant, N is the harmonic number (1,3,5,...), and ω s is the resonance frequency [62].

58 Chapter 3. EXPERIMENTAL & MATERIALS 48 The usefulness of this equivalent circuit, as stated, is when the crystal is under an external load, e.g., film deposition or liquid medium. To illustrate this, consider a QCM crystal on which a metal film had been deposited. The effect of the film will be as follows. For the motional arm, the inertial component will be affected due to the fact that the film is rigid and, after deposition, acts as a part of the crystal. The mechanical losses due to friction, however, may be affected, but are usually ignored since the film thickness is very small compared to that of quartz. Finally, the compliance component will not be affected because the film is moving freely from one side. Using the same analogy used for the unperturbed crystal, we see that only an inductor (assuming negligible resistance) will be needed. For a viscous medium, one can see that an inductor and resistor must be added, and so on. Details of this analysis can be found in several references, for example, the review paper by Buttry and Ward [10] Impedance analysis Another advantage of expressing the mechanical properties of a quartz resonator in electrical equivalents is the ability to determine the values of the circuit components using impedance, or network, analysis [10]. Impedance analysis involves the measurement of current at known applied voltage over a specified range of frequencies. In the QCM, this can be done because of the piezoelectric nature. Such analysis can elucidate the properties of the quartz as well as the interaction of the crystal with the contacting medium [10]. The way impedance analysis is performed on QCM is analogous to that for electrical circuits. A certain voltage, having known frequency, is applied to the crystal, the piezoelectric effect will cause the crystal to oscillate and generate an electric current, which can be measured. Resonance can be found when the current reaches its maximum, or impedance reaches a minimum. Beside resonance frequency, the magnitude and phase shift of current are also obtained. from which the impedance (or admittance) of the quartz crystal can be calculated. Therefore, instead of only measuring the frequency change, which is the only value affected by mass change, the used of impedance makes it possible to obtain changes in resistance (mechanical losses) and capacitance (compliance) Factors affecting QCM behavior The high sensitivity of QCM, although makes it a unique analytical technique, imposes a substantial challenges due to the large number of factors that can perturb its operation. Therefore, it is important to comprehensively study the different factors that can affect the QCM operation to be able to separate their effects and make correct conclusions based on the results obtained.

59 Chapter 3. EXPERIMENTAL & MATERIALS 49 Nature of deposited film The basic assumption for deriving the Sauerbrey equation for the relation between f and m is that the film thickness is negligible compared to that of the crystal. This assumption implies, implicitly, that the propagation of shear wave is negligible in the film, i.e., the shear wave is confined by the crystal boundaries, and the anti-nodes are still occurring at the crystal surface. Therefore, it has been stated [54] that equation (3.2) is valid up to 2% change in frequency (f) or mass (m). Correction to the assumption of negligible film thickness, and its consequences, had been made. For example [54]: tan( πν c ν q ) = ρ fυ f ρ q υ q tan( πν c ν f ) (3.6) where υ f and υ q are the shear-wave velocities in the film and quartz, respectively, ν f = υ f 2l f, where l f is the film thickness. Another assumption concerning the deposited film is that it is rigid. This eliminates the mechanical losses in the film. If, on the other hand, the film was viscoelastic, the shear wave will continue to propagate through the layer instead of completely reflecting at the quartz-film interface, introducing a new source of error [81, 105]. Finally, it is also common to use the no-slip condition at the quartz-film interface. This leads to continuous displacement and shear boundary conditions at the interface [10]. Mechanical properties of the surrounding medium The nature of the surrounding medium greatly affects the behavior of QCM. In fact, it was initially believed that the mechanical losses of oscillating quartz if immersed in a viscous medium will be very high, making it impossible to continue oscillation [36]. The use of QCM is considerably harder in liquids than under vacuum or gas conditions because [3]: 1. The high damping of the liquid in contact with the quartz surface results in a large loss of the quality factor and a decrease in the phase gradient. 2. The observed frequency shift is not only due to mass loading, but also to the viscous coupling with the surrounding medium. In representing the additional effect of viscous media on the quartz crystal using the analogous electrical components, several analysis were made, for example, Kanazawa in [36]. Commonly, the viscous medium is assumed to increase the mechanical losses and to introduce an additional mass loading on the crystal surface. This is equivalent to adding a resistor and inductor in series with the motional arm on in the Butterworth-Van Dyke equivalent circuit [10], see page 47. Another effect arises from the parasitic capacitance from the test fixture, which, i.e., the capacitance, arises

60 Chapter 3. EXPERIMENTAL & MATERIALS 50 from the external fields surrounding the QCM. This was accounted for by an additional capacitor in parallel with the motional arm [62]. Temperature The advantage of AT-cut quartz crystals is the fact that it has a very small temperature coefficient near room temperature [53]. However, any change in temperature may have an effect on the surrounding medium properties, especially liquids, which will affect the QCM response. At high temperatures range, the piezoelectric nature of quartz will be severely affected, until it reaches the piezoelectric response limit at 573 o C where the quartz losses its piezoelectric nature [80]. Therefore, it is necessary to take into account any temperature change, since this will lead to an error in interpreting the frequency data. Rahtu and Ritala [80] presents a method to compensate for such effect using a reference crystal and a modeled baseline. In our study, temperature was controlled around room temperature. Large variations in temperature did not present. Howevere, even slight variations showed an effect on the frequency, as will be discussed in Chapter 4. Pressure Pressure has a compression effect on the QCM, which causes the frequency to increase linearly with increasing pressure. This is believed to be true for gases up to 1 atm and for liquids up to 104 atm [96]. For the case of liquids, where one side of the crystal is exposed to the liquid medium and the other to air, the difference in hydrostatic pressure causes surface stresses, and the relation between the change in frequency and pressure difference is believed to be parabolic [10]. This effect, however, seems not to be very important, since the pressure generally will be constant during the experiment. Surface Morphology Molecular rough surfaces can entrap molecules from the surrounding medium, causing an increase in the measured frequency change [35]. On the other hand, if the surface roughness is large, also referred to as strong roughness in [17], the liquid will couple viscously with the surface [7]. In general, when the surface of the resonator is rough, the liquid motion, generated by oscillating surface, becomes much more complicated than for the case of a smooth surface [18]. Current studies in our group on the effect of roughness on QCM are being conducted in an attempt to relate surface roughness parameters, obtained by SPM techniques, to the observed frequency shift.

61 Chapter 3. EXPERIMENTAL & MATERIALS 51 Other factors In addition to the aforementioned major factors, other factors have shown an affect on the QCM operation, although, fewer studies can be found on these topics. For example, the effect of liquid medium conductivity on the QCM quality factor was studied by Rodahl et al. [82]. The effect is believed to be due to the enlarged electrode area due to the liquid electrical properties, which extends the applied potential region to an uncoated areas. Another study showed that, for a one side exposed crystal, there is an effect of immersion angle of the QCM on the final frequency shift. A (1 + sinθ) term, where θ is the immersion angle, was suggested to account for this effect [107]. Finally, the QCM experiments setup and procedure requires special care to avoid introducing unwanted errors. One example on this point is the presence of static stress. This stress appears due to improper mounting of the crystal in the holder, which causes uneven distribution of loads causing a serious shift of the crystal orientation with respect to the electrodes [53] Applications of QCM The QCM is a very powerful sensor. It responds to extremely small changes in mass, as low as pg/cm 2. It is sensitive to almost any change in the surrounding system. It can be coated, theoretically, with any material to make it sensitive to specific reagent of interest. It has a very high response time, in order of microseconds. It has a small volume, it can be used for in-situ measurements, and it is inexpensive. For this, the QCM has been used in measuring and probing wide range of variables. However, nowadays, two main challenges largely limit the use of QCM. First, quantitative results obtained from QCM measurements are still very limited and the reliance on such results for numeric data is minimal. Second, reproducibility of QCM is poor due to its high sensitivity and to the large number of variables that can interfere with its operation. Despite these limitations, the QCM had been employed for wide spectrum of measurements. Commercially, the QCM is known to be thickness monitors for vacuum deposition systems. This kind of applications (i.e., measuring or monitoring film deposition and removal) comprises the major section of QCM s applications. Examples on this kind of applications are the monitoring of detergency process [103], removal of solid organic soils from hard surfaces [104], surfactant adsorption [13], mass of submonolayer deposits [72], biological species adsorption [12, 30, 69], and monitoring of electrochemical processes [37, 76]. Chemical analysis is also possible using QCM. For example, Wang et al. used a QCM crystal coated with a polymer film that exhibits large changes in frequency upon changing the ph, enabling the measure of solution s ph [101].

62 Chapter 3. EXPERIMENTAL & MATERIALS 52 Contact angle and surface tension have also been measured using QCM crystal. This was done by placing a droplet with a known volume of the fluid under study on a QCM crystal (either in air or in other fluids) and measuring the frequency change as a result. The contact angle is then obtained by relating the frequency change to the droplet radius, which, in terms, is related to the contact angle when the volume is known [51]. Polymer shear modulus was also determined using a quartz resonator. This was done by measuring the change in the wave amplitude and phase in quartz crystal indicating a change in the physical properties of the film on top of the crystal [56]. QCM was also used as a scanning electrode to determine the surface topography of metal coatings. In this work, a quartz crystal coated with a metal film was actuated using a thin gold wire that scans over crystal surface. The change in frequency from one point to another is then used to determine the local thickness at the scanning point [86]. In our group, QCM is being used to study the thermal degradation of textile lubricants and to study the dissolution of photoresist using super critical CO Experimental setup The QCM crystals were AT-cut crystals obtained from Maxtek Inc., CA, having a cut angle between and with a turn around point between 40 and 80 C. The worst case temperature coefficient over the range from 20 to 100 C is approximately 2 Å of Al per C. A maximum temperature of around 300 C is recommended to avoid irreversible damage of the crystal. The crystals electrodes were gold, nickel, or silicon. The gold (Au) film has a thickness of 3400 Å deposited in a vacuum chamber on top of Ti film (for adhesion). Nickel (Ni) crystals have similar specifications as Au but the adhesion layer here is Cr. Silicon (Si) crystals were % pure vacuum deposition grade, having a thickness of 1000 Å for the Si layer deposited on top of a 2900 Å Au layer which, in turn, was deposited on a 400 Å Cr film. The crystals, as received, have 1 (2.54 cm) diameter for the quartz disk. Metal electrodes were deposited on the center of the crystal with 0.5 (1.27 cm) diameter for the exposed side of the crystal and half this value for the unexposed side. The difference in the electrodes diameter is to minimize the effect of radial change in sensitivity which is a known characteristic of the QCM. A schematic of the crystal is shown in Figure 3.4 Crystals holders were also obtained from Maxtek, Inc. For aqueous solution, a poly(vinylchloride) (PVC) holder was used. For organic solutions, a Teflon holder was used. The PVC holder is more durable to handling, unlike the Teflon holder. This is due to the nature of Teflon which makes it harder to process. The holders mount the crystal on two gold pins which also provide the necessary electrical connection. After mounting on top of the pins, the crystal is fixed with a hollow

63 Chapter 3. EXPERIMENTAL & MATERIALS 53 Figure 3.4: QCM crystal used in this work cover under which a gasket (made of Kalrez for the PVC holder and Viton for the Teflon one) to prevent liquid leaks. Finally, the holder connects to the oscillator through a BNC connector for the PVC holder, and SMB connector for Teflon holder. This is shown in Figure 3.5 Figure 3.5: QCM crystal s holder used in this work The oscillation was provided via a phase lock oscillator, which was obtained from the same company, Maxtek, Inc. (part no. PLO-10). The PLO-10 utilizes an oscillator referred to as voltage controlled oscillator (VCO). The range of operation for PLO-10 is 3.8 to 6.06 MHz frequency. It can operate in a temperature range of 0 to 50 C, and it drives the crystal with an rms voltage value of 12.5 mv. Two outputs can be obtained from the PLO-10 oscillator. First the frequency of the oscillating crystal, and second its resistance (or conductance). These two measured values help monitor changes in the mass load and in the solution properties, e.g., viscosity. The output from the PLO-10 is fed to a frequency counter (Agilent, model no A) which

64 Chapter 3. EXPERIMENTAL & MATERIALS 54 is connected to a PC that records the frequency in real-time. The experimental setup is shown in Figure 3.6 Crystal Holder QCM Crystal VC Oscillator Multimeter Temperature control Using Chiller Freq Counter Magnatic Stirrer To PC For Data Acquisition Figure 3.6: Setup used in QCM experiments 3.2 Contact Angle Wettability of stable monolayers and multilayer films can be estimated from wetting measurements. This is due to the fact that the shape of a liquid drop on a plane, homogenous surface (which is the result of the free energy of this drop) is affected by the free energy of the surface [97]. The contact angle rises from the balance between the interfacial tension of the contacting phases, usually, solid, liquid, and gas. Equilibrium position is attained by maximizing the surface area with minimum surface energy. Therefore, different behaviors may be obtained, and contact angles can vary widely from less that 10 to values up to 110 or more. Although information obtained by contact angle measurement gives a representative picture of the nature of the interaction between the three phases involved, using its value to determine the precise surface energy is questionable. Usually, Young s equation is used to determine the surface energy, but it has been argued that this result may give incorrect interpretation of the surface nature [48]. Different contact angle modes are usually measured. Young s contact angle is typically measured using the sessile drop technique. Here, a drop of fixed volume is formed on the end of a needle,

65 Chapter 3. EXPERIMENTAL & MATERIALS 55 which is then lowered slowly until the drop touches the substrate. If the needle is detached from the drop, a free standing drop is formed. For liquids with larger densities, it is recommended to keep the needle in contact with the drop to minimize vibrations which may lower the measured angle. This is called captive drop technique [48]. Advancing and receding contact angle are another mode of measurements aimed at the study of film uniformity. It is important not to confuse advancing and receding contact angle with the dynamic contact angle, which is a different contact angle mode that will be discussed shortly. To obtain the advancing contact angle, a drop is formed in the same manner as described in the sessile drop technique, but the needle now must be kept in contact with the drop. A fixed volume of the liquid is then added to the drop, and the angle is measured to obtain the advancing contact angle. For the receding contact angle, the same volume that was added in the advancing mode is withdrawn to get this angle. Other techniques for advancing and receding contact angles measurement exist. For example, a plate can be immersed in the fluid, and then tilted until the clinging meniscus disappears; the tilt angle at which this occurs is the contact angle. The plate is then dipped further into the liquid, and the angle formed between the liquid surface and the plate is taken as the advancing contact angle. Withdrawing the plate to a point lower than the original point gives the receding contact angle [67]. The difference between advancing and receding angles is called contact angle hysteresis, which is caused by [67]: 1. Impurities on the surface. 2. Effect of adsorption of the measurement material used. 3. Surface roughness. Dynamic contact angles, on the other hand, are the contact angle associated with a moving solid-liquid-vapor three-phase contact line. Static and dynamic contact angles tend to be identical on smooth solid surfaces. Wetting kinetics, specifically, dynamic contact angles, are discussed in the literature in terms of hydrodynamic or, alternatively, a molecular kinetic description [97, 98]. Sessile drop is one technique for measuring contact angles. Axisymmetric Drop Shape Analysis- Profile (ADSA-P) is a technique in which the complete drop s profile must be fitted to a Young- Laplace curve, and the angle is determined at the contact line. Therefore, to measure the advancing or receding contact angle, the needle must not be in contact with the drop, since the whole circumference is needed to fit the profile to the Young-Laplace equation. To do this, the drop is formed from beneath the substrate by injecting the liquid slowly through a hole [98]. Capillary raise technique can also be used to determine the contact angle through the relation between the contact angle and the capillary raise. This technique has the advantage of very high accuracy in height measurements, which gives a contact angle accuracy of ±0.6 [67, 97].

66 Chapter 3. EXPERIMENTAL & MATERIALS 56 Experimental setup The goniometer used for contact angle measurements was a Ramé-Hart (model no ) goniometer (Figure 3.7). This setup is composed of a specimen stage over which the sample substrate is placed. The specimen stage is a flat stage with two leveling knobs for horizontal level adjustment before measuring the contact angle, which is fixed on a 3-axis stage that can be moved horizontally, for image focusing, vertically, and laterally. This whole assembly is fixed on an optical bench which allows the horizontal movement of the 3-axis stage, and holds the CCD camera and the illuminator lamp. A syringe is attached on an assembly fixed to the 3-axis stage. The assembly allows vertical movement of the syringe for loading/unloading the sample, and for moving the syringe toward the substrate. Figure 3.7: Ramé-Hart goniometer model used for contact angle measurements The syringe was made by Gilmont Instrument (model no. GS-1200); it can hold 2.0 ml of liquid and has a smallest division of ml, with ± 0.5% of ±1 division accuracy. On one side of the optical bench is the microscope assembly. It is equipped with an 2.3 objective lens, 10 render-type eyepiece, and has a CCD black-and-white video camera (Sony, model no. XC-75/75CE) for image capturing. The other side of the bench holds the illuminator lamb. The lamb is attached to a Scott-Fostco power supply (model no. LLC 150W), which allows control of the illumination level for a sharply defined drop. Finally, the CDD camera is connected to a PC where images of the drop are captured using the vendor s software. The software allows adjustment of the image quality, leveling of the sample, capturing of the drop image, and automated measurements of contact angle by fitting the drop

67 Chapter 3. EXPERIMENTAL & MATERIALS 57 profile to Young-Laplace equation. 3.3 Atomic Force Microscopy Atomic Force Microscopy (AFM) uses the interaction between the sample surface and a very thin tip (with diameter less than 100 Å) mounted on a piezoelectric scanner (usually made of lead zirconium titante). The atoms in the tip interact with those on the surface, and the resulting attractive or repulsive forces cause the scanner to deflect. The deflection is followed by a laser detector. AFM has an advantage over scanning tunneling microscopy (STM); the ability to scan conductive as well as non-conductive substrates. In addition, by proper choice of the tip material, different information about the surface can be obtained. For example, surface concentration of certain compound can be probed using a scanner tip coated with the proper reagent. Similarly, other properties can be probed, e.g., charge distribution, using AFM. Topography results from AFM are usually given in terms of rms (root-mean-square). The rms value represent the degree of variation of surface height from one point to another relative to the overall surface. In this work, the AFM was used to scan the surface topography. Information from surface topography that can be obtained includes the surface roughness, uniformity and cleanness, and coverage or tilt angle. Both coated and uncoated samples were scanned [97]. Experimental setup AFM for surface topography analysis was conducted using a Digital Instruments Dimension 3000, with a Nanoscope IIIa controller and vibration shielded hood. Imaging was performed in tapping mode, using c-si cantilever probes with a nominal tip radius of 5210 nm. Image analysis was performed using built-in software functions to calculate the root-mean-square roughness (rms) and the power spectral density (PSD) spectrum. 3.4 Experimental Procedure Cleaning The following protocol was followed in the preparation of all crystals for QCM studies. QCM crystals: 1. The crystals were stored in the container provided by supplier until used.

68 Chapter 3. EXPERIMENTAL & MATERIALS A first wash consisted of tap water and detergent, tap water rinse, followed by DI water for gross cleaning. 3. For the crystals that were previously used in OTS experiments, a 0.05 M NaOH solution was used to clean the substrate by soaking for approximately 30 minutes. (OTS have a strong resistance to acids and organic solvents, but not basic solutions). Other crystals were cleaned with a 0.01 mm H 2 SO 4 solution. 4. For another 30 minutes, the crystal was then soaked in an acidic solution (0.01mM H 2 SO 4 ). 5. DI water rinse followed by purified N 2 drying. 6. UV/ozone was then applied for 30 minutes. This is a crucial step in cleaning, since UV/ozone is a very powerful technique to clean small amounts of organic contaminants from the surface [68]. 7. Ethanol was then used as final solvent for any organics that remained left on the surface (for 30 minutes). 8. A second 30 minutes UV/Ozone step was used to eliminate the final traces of organics and solvent. 9. Before use, the crystals were blown with N 2. This procedure was adapted, with minor modifications, from Brzoska et al. [9]. Silicon wafers: The same surface preparation procedure was followed here as that for QCM crystals, except for the addition of an ultrasonic cleaning step in chloroform before step 4. Ultrasonic cleaning was not possible with the crystals since the metal film on the top will peel under the effect of blasting bubbles. 1 Crystal holder: Before using the QCM holder, it was thoroughly washed in tap-water-detergent solution, rinsed under running tap-water, rinsed with DI water, and lightly rinsed with ethanol (for the PVC holder) or acetone (for Teflon holder). This was followed by a second DI water rinse, and finally, left in a closed container until dry. 1 A batch of the Si crystals bought from the company had a poor adhesion between the Si and Au films, where even DI water rinse caused delamination. In addition, harsh chemical treatments, using H 2 O 2 :H 2 SO 4 solution, caused delamination as well.

69 Chapter 3. EXPERIMENTAL & MATERIALS 59 Glassware: All the glassware used for solution preparation or substrate handling was cleaned as follows: first, it was washed in a solution of hot tap-water and detergent, and rinsed under running tap-water. Then, it was rinsed with DI water, ethanol, DI water, and covered and left to dry at room conditions. For OTS experiments, water was avoided as possible, and in the last three steps DI water was replaced with ethanol or acetone. This is due to the high OTS sensitivity to water traces QCM experiments 1. The crystal to be used was taken from the original container and cleaned as described in section above. 2. During the solution cleaning intervals of the crystal, the holder was cleaned as described in section and left to dry. 3. The solution to be used was measured from a stock solution and prepared for use. Whenever possible, weight measurement was used instead of volume for better accuracy. For a given run, each weighing was repeated 4 times and an average value was taken. 4. After the crystal was cleaned, it was mounted in the holder, and the capacitance of the PLO-10 circuit was adjusted as discussed below. 5. The oscillator was turned on, and the frequency was recorded for the crystal in air. This enabled the check of the initial stability and a measurement of the natural frequency, which was recorded each time the crystal was used in order to compare cleanliness of the substrate. 6. The holder (with the crystal in it) was immersed in the solution, and the capacitance of the circuit was checked again and canceled if necessary, as instructed by the circuit manufacturer. 7. The solution was placed in a 250 ml jacketed beaker. The temperature inside the beaker was kept constant ( 20 C) within 0.5 C. A small ( 1 cm) magnetic stirrer gently stirred the solution at low speed (to reduce mechanical noise). For OTS experiments, a Teflon beaker was used, since glass is reactive to OTS, and it was placed in an organic coolant (ethylene glycol) to avoid water contamination. 8. The acquisition of the frequency data was started. 9. When the frequency reached a stable level (within 1-2 Hz), the surface modifying agent was injected using a micropipette. A fast injection disturbed the QCM frequency with noise, but the solution injection should not be slow, which clearly affect the deposition rate.

70 Chapter 3. EXPERIMENTAL & MATERIALS Frequency acquisition was continued, and a stable signal (within 1-2 Hz) was again achieved before reading the final frequency. 11. The setup was disassembled, cleaned, and stored in ambient atmosphere for next run Contact angle experiments Because of the geometrical design of the QCM crystal, where the metal layer was deposited at the middle of the crystal preventing the ability to take a clearly defined image of the drop and the surface, the contact angle was measured either on metal coated microscopic slides or on silicon wafers. 1. The substrate was cleaned as described in section Required amount of solution was measured and placed in glass (or Teflon in the case of OTS) dish. 3. The substrate was then immersed in the solution, which was stirred with a magnetic stirrer. 4. The Required amount of surfactant solution was measured and added. The substrate was left in this solution for the desired time. The time intervals were determined based on the QCM experiments. For example, for one of the surfactants, C 12 E 8, the QCM response was approximately immediate, therefore, short time of immersion was used ( 15 minutes, to allow for good mixing). 5. The substrate was removed from the solution and dried. 6. The substrate was then placed on the goniometer stage and the image was focused to give a clear image for better readings. 7. A fixed volume drop (about 6 µl) was used for each measurement. 8. The drop was formed at the tip of the needle, which was about 0.5 cm above the substrate, and then slowly lowered until it touched the substrate. At this point, measurements started Atomic force microscopy Both QCM crystal and metal coated slides were used. AFM studies were conducted on a scan size of 1, 5, and 10 µm. The scan rate was set to 1 Hz and the resolution to 256 scan lines. Four scans at for different spots were taken per sample. The final rms value is the average of those values.

71 Chapter 3. EXPERIMENTAL & MATERIALS Materials DI water was obtained from MilliQ-Pore unit(> 12 MΩ). Ethanol and acetone were obtained from Fisher Scientific. n-hexadecane and chloroform were anhydrous grade obtained from Aldrich-Sigma. Octadecyltrichlorosilane (90%+) and Methyloctadecyldichlorosilane (MDCS) were obtained from Sigma-Aldrich, and stored under vacuum until use. Octaethylene glycol monodecyl ether (C 12 E 8 ) was obtained from Sigma-Aldrich, dissolved in DI water and kept in room environment. All chemicals were used as received.

72 Chapter 4 RESULTS & DISCUSSION The general outline for the work in this project had been presented in Chapter 3. The scheme consisted of four phases. This chapter will discuss those phases in detail, state the approach taken, represent and discuss the experimental results. The project started by verifying the creditability of the experimental setup used and the different experimental techniques commonly used in QCM studies. This will be discussed in the first section of this chapter. During this phase, certain problems and questions arose concerning the use of QCM and the analysis and interpretation of the obtained data. These issues will be discussed in the second and third sections. When an acceptable level of experience and confidence in acquiring experimental data was achieved, a new phase was started. This included the study of more relevant systems as will be discussed in section four. Finally, the silanes-on-si (silane/si) system, which is the system of interest for this work will be the subject of the last section. 4.1 Validation of Experimental Technique Verification of the reliability of our QCM data was a major concern in this work. Taking into consideration the large number of variables that can affect the QCM operation and the absence of detailed studies on thin organic films using QCM, required special carefulness in choosing an appropriate system for study Selection of verification system Anti-stiction agents commonly studied for MEMS tribological problems are compounds that can form self-assembled monolayers (SAM) by chemically attaching to the substrate (see section 2.4). Such systems, for example silanes/si, have been extensively studied using different analytical techniques, including contact angles, ellipsometry, surface probing microscopy (SPM) techniques. However, the 62

73 Chapter 4. RESULTS & DISCUSSION 63 full picture for this system is not yet clear. Different interpretations of the experimental data and theoretical analysis, for the silane/si systems, lead to controversial conclusion about the mechanism of deposition and the nature of the formed layer. Therefore, it was not desirable to start with such system, as, by doing so, serious erroneous conclusions can be made. Therefore, the criteria to select an acceptable system was as follows: 1. The system must form an organic thin layer on the substrate through adsorption from solution. This is to be similar to that of silane/si system. 2. Adsorption of active molecules from solution should be the only change known to occur in the system. Other changes, such as viscosity, can also affect QCM results, which are not desirable at this stage. 3. An aqueous media was preferred, since the operation of QCM is known to vary with solution properties. Since water is readily available, the effect of the variation of solution properties can be minimized. Also water is inexpensive. 4. Since extensive studies using QCM on any one particular system is practically absent in the literature, the choice was made such that system should be well characterized. The QCM data available are supported by other studies. For example, the critical micelle concentration (CMC) can be extracted from the QCM curves and agree, within experimental error, with the reported CMC value. Caruso et al. [13, 103] performed a series of studies on C 12 E 8, octaethylene glycol monododecyl ether, using gold coated QCM crystals. The ethoxylated alcohol surfactants family, to which belongs the C 12 E 8 surfactant, are a well known surfactants that have an ethoxylated alcohol with a longcarbon tail and a varying number of ethoxy groups (e.g., 8 groups in the case of C 12 E 8 ). These are non-ionic surfactants, which is an important criterion in the QCM experiment, as discussed in section In addition, aqueous solution is used for deposition. During deposition, the ethoxylated alcohols form hydrogen bonds with water and the hydrophobic carbon tail deposits onto the substrate surface to minimize contact with water [67]. As a result, the ethoxylated alcohol, specifically C 12 E 8, was chosen for the purpose of experimental verification of the QCM technique Results & Discussion A set of experiments was designed so that it would involve a concentration range as that in the reference work [13]. The range extends from well below the CMC point for C 12 E 8 (CMC M) to about 3 times that value.

74 Chapter 4. RESULTS & DISCUSSION 64 The system used in this experiments set was: Gold coated QCM crystal mounted in the PVC crystal holder (section 3.1.4). Chemicals: DI water and C 12 E 8 (section 3.4.5). The QCM runs were performed as described in section Results Figure 4.1 shows the change in frequency upon addition of varying C 12 E 8 concentrations. Each data point represents an average of about 3 runs. The figure also shows the Langmuir adsorption isotherm, solid line, which was obtained by fitting the data to the equation [13]: Figure 4.1: Adsorption isotherm of C 12 E 8 on gold in aqueous solution at 25 C. c f = c f max + 1 K ads f max (4.1) The fitting for c/ f vs. c is shown in Figure 4.2. From the straight line, f max and K ads was found to be 13.5 Hz and M 1, respectively. Discussion To compare the obtained data points with those from Caruso et al., two main differences between the two systems needed to be addressed before making the comparison. Those difference are: 1) the

75 Chapter 4. RESULTS & DISCUSSION 65 Figure 4.2: Langmuir plot for C 12 E 8 on gold in aqueous solution at 25 C. area of the exposed electrode for the two systems was different, and 2) Caruso et al. used crystals with a fundamental frequency of 9 MHz, whereas the crystals for this work have a 5 MHz natural frequency. Difference in electrodes configuration Crystals used in this work had an f of 5 MHz, whereas those used by Caruso et al. had an f of 9 MHz. The crystal used by Caruso et al. had a 5 mm diameter for both electrodes, therefore, the active area of the electrode can be taken as 5 mm, and this will give an exposed area of 6.25π mm 2. In our work, the surface area of one electrode was 0.5 ( 12.7 mm) while the other was 0.25, therefore, the active surface area of the exposed electrode must be taken as 0.25, which gives a total area of 10.1π mm 2. The difference that will be generated from these different electrode configurations is not due only to the difference in deposition area. The QCM is known to show a radial change in the frequency sensitivity to deposition. The configuration of the crystals used in this work is made such that this effect will be minimized by making the non-contacting electrode larger than the contacting one. On the other hand, crystals used by Caruso et al. have both electrodes of the same area. The aerial sensitivity, however, is a function of several parameters and must be taken into account by calibration [53]. Therefore, it is not feasible here to account for such difference, since detailed information on the Caruso et al.

76 Chapter 4. RESULTS & DISCUSSION 66 system is not available, and it must be taken into consideration that such difference will introduce an certain amount of error if a comparison is made. In this regard, the difference between electrodes configuration will be taken only to affect the total active surface area. However, as can be seen from Sauerbrey equation (3.2), the surface area do not affect the resultant f since it is a function of aerial mass density rather than total mass. Difference in natural frequency, f From Sauerbrey equation (3.2), it can be seen that f will depend on the natural frequency f in a hyperbolic relationship. This means that the sensitivity of the crystal will increase in a hyperbolic fashion with the f. Therefore, to correct for this sensitivity difference, f must be multiplied by (9/5) 2 or Figure 4.3 shows the same data as in Figure 4.1 adjusted by multiplying each point by 3.24, in order to directly compare our data with Caruso et al. The results from Caruso et al., together from Figure 4.3: Adsorption isotherm of C 12 E 8 on gold in aqueous solution scaled to a 9 MHz crystal response at 25 C. the obtained data, are shown in Figure 4.4. To compare the two sets of data, it was necessary to use a statistical measure for that purpose. Therefore, the Langmuir isotherm which relates c/ f to c in a linear relationship was used. The two data sets from Caruso s and our work were lumped into one set and linear regression was performed. The result is shown in Figure 4.5. As the figure shows,

77 Chapter 4. RESULTS & DISCUSSION 67 the correlation coefficient, R, is This value is acceptable, since quantitative data are not of interest yet, and taking into consideration the differences between the two systems, as discussed above. Figure 4.4: Experimental data (scaled) compared to Caruso et al. data. The solid lines are Langmuir isotherms fit, equation (4.1) Conclusion From the above discussion, it can be concluded that the system and techniques used in this work are giving acceptable results which, to a good extent, agree with a previous work with another system. The differences in the two systems were not simply a technical difference, but a differences that certainly interfere with the obtained signal and, thus, the final interpretation of the data. Although the use of similar setup to that used in Caruso et al. would be preferable for comparison with previous results; there are some definite advantages to the setup used in this work that made it more desirable. One advantage, that was mentioned above, is the configuration of the electrical electrodes which minimizes the differential aerial sensitivity. Another advantage lies in the design of the holder. The holder used in this work completely prevents the contact of the solution with the unexposed side of the crystal in addition to the electrical connections to the electrodes.

78 Chapter 4. RESULTS & DISCUSSION 68 Figure 4.5: Experimental data lumped with data points from Caruso et al. to perform linear regression. Finally, for comparison purpose the behavior of contact angle as a function of surfactant concentration is shown in Figure 4.6. The data was obtained using Au coated microscopic slides at the same conditions as that for the QCM runs. 4.2 Studies on the frequency effect on adsorption The QCM crystals used in this work had a natural frequency of greater than 5 MHz. During the first validation experiments, the question was asked as for whether such high oscillating frequencies would have an effect on the deposition or not. As Figure 4.4 indicates, the data points obtained in this work, at 5 MHz, were still, after scaling, less than those obtained by Caruso et al. who used a 9 MHz crystals. Different sources of error were discussed in the previous section, however, the results obtained for this work were lower by, approximately, a factor of 1.4 from those in Caruso et al. We conducted a set of experiments to verify if the natural frequency of the crystal affects the rate and extent of deposition.

79 Chapter 4. RESULTS & DISCUSSION 69 Figure 4.6: Contact angles for Au slides coated with different concentrations of C 12 E 8 at 25 C Theory An oscillating surface in viscous media will generate shear waves that propagate through the media to a specific distance, i.e., the boundary layer. These waves can be described in terms of the surface velocity (v) [8]: υ(y, t) = υ e y f/2ν cos(ft y f/2ν) (4.2) where υ is the velocity at the surface, f is the oscillating frequency of the surface, ν is the kinematic viscosity of the liquid, and y is the direction normal to the surface as shown in Figure 4.7. For simplicity, the diffusion occurring from the bulk solution to the surface is assumed to occur via the boundary layer mechanism. The boundary layer thickness is defined as the damping distance for the shear waves, i.e., the distance at which the shear wave amplitude becomes 0 (referred to as δ). This value, δ, can be calculated by setting equation (4.2) equals to zero, which is only possible when the exponential term is equal to zero. Therefore, y at which υ (y, t) equals zero is given by: y υ =0 = δ = 2ν f (4.3) The mass diffusion equation, assuming boundary layer diffusion, is given by: N = C δ (4.4)

80 Chapter 4. RESULTS & DISCUSSION 70 Figure 4.7: A schematic showing the shear waves generated by an oscillating surface in viscous medium. where N is the mass flux of the diffusing material (the surfactant in our case), C is the concentration gradient between bulk and surface, and δ is the boundary layer thickness which is taken here, as discussed above, to be equal to the penetration depth of the shear wave. Combining equation (4.4) and equation (4.3) yields: f N = C (4.5) 2ν It is clear, from equation (4.5) that the flux will be proportional to the oscillating frequency of the surface under the square root. Increased deposition will results in two related changes in the resulted film: 1. Higher film densities, resulting in higher frequency changes for QCM runs. 2. Higher film packing, which will affect the film properties, for example, film roughness and tilt angle. The problem here is that even if the flux changes, eventually, both systems will reach the same concentration after enough time, which means that the frequency change will be identical in both cases. However, the calculated f was taken within a small range around the time at which surfactant was injected. Also, the signal before and after injection was stable to within ±1 Hz in the included area. Finally, and most importantly, is the fact that the packing of the molecules on the surface may affect its final equilibrium state. To account for this proposed effect, the data points obtained in this work must be multiplied by a factor of f 2 /f 1, where f 2 is the natural frequency for the crystals used in Caruso et al., i.e.,

81 Chapter 4. RESULTS & DISCUSSION 71 9 MHz, and f 1 is that for this work, i.e., 5 MHz. Therefore, an additional scaling factor is used here which equals to The scaled results in Figure 4.4 multiplied by the new scaling factor is shown in Figure 4.8. Figure 4.8: Experimental points scaled to account for both f and the proposed effect of oscillation on deposition versus points from Caruso et al. The figure shows a better agreement between both data sets than in Figure 4.5. Nevertheless, this was a very simplified approach, and a closer investigation was needed to confirm the hypothesized effect Approach Since the QCM was used as an analytical technique in this work, and was not the subject of the research, only a straight forward approach to study the effect of oscillation on deposition was feasible. This approach consisted of performing f measurements at known concentrations with an in-situ (frequency is to be recorded during adsorption) and ex-situ measurements (frequency before and after adsorption is to be measured and the crystal was not oscillating during deposition). The procedures for the above two experiments was described in section The crystals that were utilized for this work has a natural frequency of either 5, 6, or 9 MHz. The oscillating circuit used, however, can only operate in the range of MHz. Therefore, it

82 Chapter 4. RESULTS & DISCUSSION 72 was only possible to perform runs with the 5 and 6 MHz crystals. For both of these frequencies, in-situ and ex-situ runs were performed. First, an ex-situ run was made using a 5 MHz crystal. The results from this run was taken as a basis which was scaled to higher frequencies similar to the approach taken in section Also, an ex-situ run using 6 MHz was performed to compare the value with the scaled value from the 5 MHz ex-situ run. Then in-situ runs using both 5 and 6 MHz crystals were performed. To expand the comparison range, a value of f results at the same concentration was taken from Caruso et al. to include the 9 MHz frequency Results & discussion Figure 4.9 shows the obtained frequency response results for the ex- and in-situ runs for the change in frequency for a 5 and 6 MHz upon injection of 3000 µl of C 12 E 8 solution (This resulted in a total concentration of 0.15 mm). The first run was an ex-situ run done using a 5 MHz crystal. The result from this run was scaled to the corresponding 6 and 9 MHz crystals response, as done in section (page 66). Two in-situ runs using 5 and 6 MHz crystals were performed to give, in combination with the result from Caruso et al. a set of three in-situ frequency change values. From the figure, one can see that the difference between the in-situ and ex-situ results is small and is expected to have insignificant effect on the deposition which, probably, lies within the error margin, taking into account the high sensitivity of the QCM which affects the reproducibility of its data. To further test the proposed effect, two experiments were made. First an ex-situ run using a 6 MHz crystal at the same concentration of surfactant was performed, i.e., 0.15 mm. The raw data for this run is shown in Figure 4.10(1). This figure shed the light on one important source of error when performing an ex-situ QCM experiment. As seen, the action of switching the crystal OFF then ON again results in a change in the frequency. The prediction of this change is difficult and seems to depend on the crystal used, the conditions of the surrounding medium, as well as the time interval between switching ON and OFF. For example, Figure 4.10(2) shows the results obtained for the same conditions as that in the 6 MHz run, but for 5 MHz. As a result of the change in f introduced by the process of switching the oscillator ON and OFF, which was comparable to the change introduced by injecting the surfactant, it was necessary to use a different experimental technique in the analysis. If higher frequencies will result in higher mass deposition which will lead to a more dense packing. Therefore, the use of AFM seemed appropriate at this point. The objective was to determine the surface roughness of the QCM crystals used for deposition in both in- and ex-situ runs. The results are shown in Figure The rms (see section 3.3) values for the 5 MHz crystals decreased by nm for the in-situ run, where as for the ex-situ run it increased by nm, and the rms difference between the in- and ex-situ cases was nm. For the 6 MHz crystal, the rms was nm higher for the in-situ than the ex-situ

83 Chapter 4. RESULTS & DISCUSSION 73 Figure 4.9: The difference in f between oscillating and non-oscillating crystals. case. The expected increase of roughness of the resulted surfactant film with oscillation was indeed present. This result, coupled with that from QCM runs, indicate the enhancing effect of oscillation on the deposited film. To this point, the results were consistent with the proposed theory but the confidence level was not big enough to draw a certain conclusion. A final inspection of the effect of natural frequency on the deposition process was needed to see whether or not such effect is significant in this study. A set of in- and ex-situ experiments were performed to measure the change in frequency, f, upon injection of the same concentration of surfactant C 12 E 8. The results are shown in Figure Conclusion Figure 4.12 suggests that the difference in f-which is also translated to change in mass-is lower than can be detectable by the current approach. On the other hand, looking at the overall results, Figures 4.8, 4.9, and 4.11, indicate the presence for the effect of oscillation on the film formation process and on the resulted film properties. Therefore, it was concluded that the study of this phenomena needs to be done with other techniques that can give a clearer, more detailed picture of the properties of the resulted film. For example, scanning tunneling microscopy (STM) may be

84 Chapter 4. RESULTS & DISCUSSION 74 (1) (2) Figure 4.10: An ex-situ QCM run for (1) 6 and (2) 5 MHz Au crystal upon injection of 0.15 mm C 12 E 8 at 25 C (injection time indicated by arrow). The oscillator was switched OFF at and ON at without (a) and with (b) surfactant injection.

85 Chapter 4. RESULTS & DISCUSSION 75 Before injection (1) (2) In-situ after injection (3) (4) Ex-situ after injection (5) (6) 5 MHz crystal 5 MHz crystal 6 MHz crystal Figure 4.11: AFM images for 5 and 6 MHz crystals before and after injection of 2000 µl C 12 E 8 (which gives a final solution concentration of 0.1 mm) for in-situ and ex-situ case. The scanning area is 1 1 µm 2. The rms values are: (1) (2) (3) (4) (5) (6) nm.