An Analysis of Interfacial Effects in Electronic Devices

Material Information

An Analysis of Interfacial Effects in Electronic Devices
Scheuermann, Andrew
Publication Date:


Subjects / Keywords:
Annealing ( jstor )
Crystals ( jstor )
Diodes ( jstor )
Electronics ( jstor )
Etching ( jstor )
Molecules ( jstor )
Nanoparticles ( jstor )
Ozone ( jstor )
Solvents ( jstor )
Surface areas ( jstor )
Undergraduate Honors Thesis


A variety of interfaces are investigated in a series of experiments spanning from inorganic synthesis to materials characterization and physics. Each aims to illuminate the underlying chemistry and physics at work in novel electronic devices now reaching the nano- and single molecule-scale. In chapter one, the sputter coater is found to cause point defects leading to device failure. An annealing study indicates that thermal properties of the interface are governed by GaN and not the metal at the temperature range investigated. Chapter two builds on these results studying the GaN/dielectric interface present in more relevant devices, HEMTs and MOSFETs. The electrical performance of e-beam fabricated diodes before and after thermal treatment is studied. CL is performed under the diode region to reveal defect formation. In chapter 3, Cu(dbm)2 is synthesized among other molecules as a component in a macromolecular gate. The single molecule/metal interface is studied and a switching mechanism is discovered. Finally in chapter 4, magnesium oxide nanoparticles for thermal batteries are investigated. Here again, the crucial interfaces are probed in terms of seeking to illuminate the chemistry and physics at work at the nanoscale for these novel electronics applications. ( en )
General Note:
Awarded Bachelor of Science; Graduated May 3, 2011 summa cum laude. Major: Chemistry
General Note:
College/School: College of Liberal Arts and Sciences
General Note:
Chapter 1,2 Authors: A.G. Scheuermann, A. Gerger, T. Stewart, B.P. Gila, C.R. Abernath
General Note:
Chapter 3 Authors: A.G. Scheuermann, J. Bonvoisin, S. Gauthier
General Note:
Chapter 4 Authors: A.G. Scheuermann, D.F. Smith, D. Bencoe, K. Waldrip
General Note:
Advisor: Cammy Abernathy

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright Andrew Scheuermann. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.


This item is only available as the following downloads:

Full Text


S c h e u e r m a n n 0 An Analysis of Interfacial Effects in Electronic Devices Andrew Galen Scheuermann Thesis Advisor: Dr. Cammy Abernathy Chapter 1,2 Authors A. G. Scheuermann 1 A. Gerger 1 T. Stewart 1 B.P. Gila 1 C.R. Abernathy 1 1) Department of Material Science and Engine ering, University of Florida, Gainesville, Florida 32611, USA Chapter 3 Authors A.G. Scheuermann 1,2 J. Bonvoisin 2 S. Gauthier 2 2) CNRS, CEMES (Centre d'Elaboration des Matriaux et d'Etudes Structurales), BP 94347, 29 rue Jeanne Marvig, F 31055 Toulouse, France Chapter 4 Authors A.G. Scheuermann 1,3 D.F. Smith 4 D. Bencoe 3 K. Waldrip 3 3) Sandia National Laboratories, PO Box 5800, Albuquerque, New Mexico 87185 4) EMF Sytems,Inc. 14670 Highway 9 Boulder Creek, CA 95006


S c h e u e r m a n n 1 Table of Contents Subject Page numbe r Abstract and Introduction 2 Chapter 1: GaN/Metal Interfaces 4 Chapter 2: GaN/Dielectric Interfaces 13 Chapter 3: Metal/Organic Interfaces 23 Chapter 4: Nanoparticles 43 Acknowledgements 55 References 56


S c h e u e r m a n n 2 Abstract A variety of interface s are investigated in a series of experiments spanning from inorganic synthesis to materials characterization and physics. Each aims to illuminate the underlying chemistry and physics at work in novel electronic devices now reaching the nano and single m olecule scale. In chapter one, the sputter coater is found to cause point defects leading to device failure. An annealing study indicates that thermal properties of the interface are governed by GaN and not the metal at the temperature range investigated Chapter two builds on these results studying the GaN/dielectric interface present in more relevant devices, HEMTs and MOSFETs. The electrical performance of e beam fabricated diodes before and after thermal treatment is studied. CL is performed under the diode region to reveal defect formation. In chapter 3, Cu(dbm) 2 is synthesized among other molecules as a component in a macromolecular gate. The single molecule/metal interface is studied and a switching mechanism is discovered. Finally, magnesium o xide nanoparticles for thermal batteries are investigated. Here again, the crucial interfaces are probed in terms of seeking to illuminate the chemistry and physics at work at the nanoscale for these novel electronics applications. Thesis Introduction As the ever moving march of technology pushes into the twenty first century, we find the a finite limit. 1 Still mired in recent economic stagnancy, the m provided that intellectuals and innovators alike come together in the name of scientific advancement. A myriad of new electronics technologies are seeking to answer this question and to be the next transistor or the next integrated circuit. As we breech new realms of science, it is more important than ever to investigate the fundamental chemistry and physics underlying these processes. With the advent of nanoscale technologies, the complexity of interfacial reactions has skyrocketed. Bulk and even surface properties, well known in most materials for some time, are no longer sufficient to understand these devices. This thesis is an investigation of interface chemistry and physics as it relates to a number of modern electronics applications. Chapters one and two detail studies of the high impact III V semiconductor gallium nitride and its interface chemistry and physic s with both metals (chapter 1) and dielectrics (chapter 2). The concept of using wide bandgap semiconductors in power RF electronic devices has been discussed in the professional community over the last few years but due to some fundamental issues that ha ve yet to be understood and solved, it has not been fully implemented. One of the chief developments that must take place is in the area of improving the performance and stability of insulators used in conjunction with the wide bandgap devices. This proj ect has worked to further provide essential insight needed to help overcome this hurdle. Successful development of this type of device technology would lead to new applications in hybrid drivetrain automobiles, next generation battleships and submarines, Compound semiconductors, however, are only one thrust of many in this broad effort to revolutionize electronic devices. Quantum computation has bee n getting a lot of attention both positive and negative for its huge potential effect. There is another field, however, with great potential that has


S c h e u e r m a n n 3 thus far received much less hype. Molecular electronics is an effort to reengineer device hardware out o f single molecules thereby achieving the ultimate size minimization. The goal is to integrate logic bits into single molecules accomplishing what once took large solid state structures with a single consolidated unit. However, as with all nanoscale techn ologies, our knowledge for so many years has concerned only the bulk and surface properties of chemicals. Single molecules are subject to a whole different set of restraints and we must therefore begin a new investigation of single molecule chemistry and single molecule physics. In this chapter, we embark on this study by examining the organic/metal interface. Chapter 3 details my summer research synthesizing a molecular component for a macromolecular gate. The molecule is studied on metal and salt bi la yer surfaces and a switching mechanism is discovered. Lastly, in chapter 4, nanoparticle properties are investigated. The nanoparticles of interest are magnesium oxide nanoparticles used for thermal battery applications. While trying to reengineer a spec ific nanopowder it was found that the properties governing many of the nanoparticles properties were multivariate or involved aspects of the material that we do not currently have methods to characterize. This study then set off to illuminate the underlyin g physics and chemistry behind nanoparticle growth as we sought to control MgO particle properties from the precursor phase Mg(OH) 2 The triple phase interface between nanoparticle, electrolyte, and anode or cathode is essential to proper battery function ality. All four chapters seek to address the issue of interface interactions in these novel electronic devices. Whether it is solid state thin films like gallium nitride, organic devices like macromolecular gates, or nanoparticle technology like magnes ium oxide enabled thermal batteries, developing a deeper understanding of the fundamental chemistry and physics at work is essential. Through this type of scientific advancement the next generation, the nano generation of electronic devices can be fully r ealized.


S c h e u e r m a n n 4 Chapter 1: Metal/GaN Interfaces 1 1 Introduction While GaN is a fairly new electronics technology, initial results using either MESFET or HEMT structures have been encouraging. However, several important problems persist, namely poor linearity and premature device failure or degradation of performance over time. While the former is most likely due to high parasitic resistances and knee voltages, which arise from the high contact resistivities and high sheet resistances between the sou rce contact and the gate, the latter remains somewhat irresolvable. Conventional n+/n layer structures for GaAs technology cannot be applied in the nitride based material system since no adequate gate recess technology is presently available. Both of these problems may be overcome, however, by using a MOSFET approach. MOSFET devices have a wide gate modulation range, usually resulting in better linearity than either MESFETs or HEMTs. In addition, current FET processing does not have a gate fabrication techn ology that can tolerate the high temperature implant activation needed to form the low resistance source and drain region in the self aligned MESFET configuration. By contrast, MOSFETs can use implantation to form a highly doped source and drain without co mpromising the gate contact. The temperature sensitivity of MOSFETs will also be superior since they do not suffer from the severe leakage encountered in Schottky based devices. Ultimately, the use of and lower p ower consumption. One important issue involving compound semiconductors is the stability of the device to device electrical isolation. Device operation at elevated temperatures and high fields results in a breakdown of this isolation over time and thus a n increase in device leakage from the drain to an adjacent source of another device. Thin passivation layers protect the surface preventing surface leakage paths, but failures still occur. Controlling the oxide dielectric interface is critical. However, almost no information exists about how prolonged device operation or exposure to elevated temperature affects the chemical and electrical nature of this interface. To truly illuminate the underlying physics and chemistry behind these failure mechanisms, it is important to first characterize the interactions at the metal semiconductor interface before even introducing a dielectric. Figure 1 1: Schematic of a MESFET (MEtal Semiconductor Field Effect Transistor) The MESFET device is similar to the MOSFE T, but differs primarily by the fact that no insulating layer separates the gate from the surface. The interface between the metal and semiconductor forms a Schottky junction when it has rectifying properties. The first specific question investigated he re concerns the metal deposition method. It has been known for some time that energetic bombardment from the sputter coater can cause significant damage to a surface. In 1998, Ushikie et. al. replaced xenon with argon and found that the significantly sma ller size of argon reduced the damage to the surface 1 This became a standard for most sputtering system. In 2003, another group reported significant


S c h e u e r m a n n 5 damage from molybdenum sputtering 2 More recently, there has been a thrust toward using plasma charge t raps (PCT) in sputter systems to capture high energy particles that would otherwise damage the sample surface 3 A 2006 study by Ye et. al. compared sputtering damage to the damage caused by earlier techniques in an effort to begin exploring the fundament al science at work. In addition, they studied the effect of varying the power, pressure, and pulse DC parameters on induced damage 4 Our system has a number of changeable variables as well. However, each represents a crucial trade off. A higher sputter ing voltage induces more damage, but a lower one results in slower growth rate. At lower ambient pressures, particles balistically fly in straight line paths impinging on the surface with sufficient force to cause re sputtering of the surface material. A s the pressure is raised, the mean free path is dropped leading to collisions with gas particles. Here, the ions arrive after a random walk contributing to slower growth and poorer film quality. Other possibilities include changing the ambient gas specie s from argon, controlling the temperature, or setting a substrate bias voltage. The magnetron sputter coater operates using three magnets to create a dual effect (Figure 1 2). A plasma is created whereby Figure 1 2: Magnetron sputter coating schematic positive ions are accelerated toward the cathode. For a particle to be sputtered, the energy transferred to it normal to the surface must be around three times the binding energy or approximately the heat of sublimation. Sputtered particles are ejected toward the anode, the substrate, and impinge upon it with a force subject to the variables discussed before. In the first experiment, Au/Pt Schottky diodes are fabricated on GaN samples by photolithography and magnetron sputter coating. A control grou p of diodes is fabricated via e beam deposition. E beam or electron beam physical vapor deposition (EBPVD), is a much softer technique. Electrons are created from a charged tungsten filament under high vacuum. The electrons bombard the target material b ut not to cause sputtering, rather to push target molecules into the gas phase. This gas, then, precipitates leaving a thin metal film over all surfaces within the line of sight. In this way, e beam can accomplish the equivalent metal film without the sp uttering mechanism allowing for a direct study of sputter induced damage. After the sputtering experiment, a new series of diodes are fabricated and annealed at different temperatures to further explore the thermal characteristics of the metal/GaN interf ace. 1 2 Experimental Procedure 1 2.1 Fabrication of Schottky diodes Standard photolithography was used with a positive photoresist meaning that any area exposed to light becomes soluble in the photo developer. A mask was used with a series of circular areas of varying diameter. In addition to positive photoresist, LOR was used. Bi layer resist processing allows for a clean lift off and a well defined remaining metal pattern. In the case of single layer resist processing, the deposited metal sticks t o the walls of the


S c h e u e r m a n n 6 photoresist. When lifted off, these vertical artifacts are often left in place. Additionally, increased mechanical stress on the metal can lead to cracking and peeling of the metal intended to remain intact. In bi layer processing, th e lower layer is soluble in the photo developer so that when the solvent cuts through the exposed areas above, it cuts back further into the bottom layer (Figure 1 2). Figure 1 2: a) Bi layer photo resist set up b) The top layer is exposed to light c) I n the presence of developer the exposed area etches away. Beneath it, the LOR also is etched back pulling beyond the reaches of the top layer. This over hang is crucial producing the break in the metal thin film (d). After the lift off procedure (e), th e unwanted metal is easily pulled away without inducing undue mechanical strain on the remaining metal. The surface is left cleanly patterned 5 7 A Laurell Spin Coater was used to lay down LOR and then a Headway Spin Coater was used to spread positive photoresist on top. A Karl Suss MA6 mask aligner was used to expose the samples. A Kurt J. Lesker CMS 18 Multi Source Magnetron Sputter Coater was used to deposit 200 nm of platinum and 800 nm of gold on GaN. The ambient gas is argon. The diode array after lift off is shown in Figure 1 4. From biggest (A) to the smallest (J), diameters are 1138 m, 815 m, 574 m, 426 m, 290 m, 215 m, 154 m, 112 m, 82 m, and 58 m 1 2.2 Diode etching A diamond scribe was used to etch a cross pattern indicating the beginning of the diode series. This allowed characterization at the correct locations both before and after etching the diodes. Diodes were etched in diluted aqua regia solution, 1HNO 3 : 3HCl: 3H 2 O. Gold etches away at a rate around 10 m/min in such a solution, but platinum requires more time. Samples were submersed in dilute aqua regia mildly boiling at 40 C maintained with a hot water bath. It took approximately 12 minutes per sample to remove all traces of platinum (verified by optical microscope later by EDS as well). With only a Figure 1 3: Kurt J. Lesker CMS 18 Multi Source Magnetron Sputter Coater used to deposit metal Figure 1 4: Diode array pattern lettered A through J. A is the largest diode with a 1138 m diameter (not shown). T he pattern is repeated across the entire samples later allowing for IV mapping of the substrate as well as providing a large number of devices for testing and statistical data compiling.


S c h e u e r m a n n 7 few exceptions, the diodes were perfectly fabricated across the sam ple with smooth edges as shown in Figure 1 4. 1 2.3 Sample characterization Before any processing, substrates were characterized by a JEOL 6400 SEM (scanning electron microscope) and a SPM/AFM dimension 3100 (atomic force microscope). AFM was operated in the tapping mode whereby topology. A laser beam is focused on the back of the sample and reflected to a photodiode array that works through a feedback mechanism to control the piezodielectric. Constant force is maintained between the tip and the surface by varying the height of the probe through a feedback mechanism. This technique provides incredibly high resolution and also allows for the calculation of a roughness measurement derived from the surfac e morphology. All substrates were scanned prior to any treatments and root mean square and average roughness values were calculated. After annealing and etching treatments, scans were taken where the diodes had been etched away to image the interface r egion. Scans and roughness values were also obtained for annealed and etched control samples including under the diode area of e beam deposited samples. Current voltage measurements were performed on the e beam and sputter sample diodes with an analyti cal I V probe station. After observing extensive Ohmic like behavior (see discussion section), Hall measurement was also performed on sputtered GaN samples. 1 2.4 Annealing treatments In addition to studying the damage induced by sputter deposition, th e thermal characteristics of the GaN/metal interface were investigated between 200 and 500 C. A rapid thermal annealing process (RTA) was used to expose the samples to 200, 300, 400, or 500 C. After a 30 second delay, the temperature was ramped over 60 s econds to the set value. The treatment was maintained for an additional 60 seconds before shutting off the heat lamp. The oven was not opened until the temperature was at least 90 C to prevent undue thermal stress on the samples. 1 3 Results and Discu ssion 1 3.1 Sputter induced damage As mentioned briefly in the introduction section, energized particles bombarding the surface can cause damage in a number of ways. Figure 1 5 depicts a number of such possibilities. In the back, secondary electrons a nd backscattered particles are depicted as well as the re scattered particle talked about in the introduction. Probably the most likely type of Figure 1 5: Schematic of damage possibilities from impingent sputtered ions damage, however, is the point defect shown at the front of the schematic. In particular, sputtering is likely to cause nitrogen vacancies leading to gallium rich areas scattered across the surface. This hypothesis was corroborated by the I V characterization. Almost all of the diode s fabricated through sputter coating behaved like Ohmic devices as opposed to the expected Schottky rectifying characteristic. Hall


S c h e u e r m a n n 8 measurement revealed that the GaN semiconductor surface had become metal like with a carrier concentration of 5.923 10 18 cm 3 At this level of carrier concentration we see significant band bending as shown in Figure 1 6. Figure 1 6: Band bending leading to thermionic field emission ( top) or even field emission (bottom ). Thermionic emission occurs when thermal energy provides the electron sufficient energy to pass a potential barrier. At intermediate levels of band bending, particles are able to tunnel through the potential barrier aided by thermal energy. This process is referred to as thermionic field emission and leads to leaky devices. At more extreme band bending, the barrier is thin enough for particles to tunnel through without a thermal energy requirement (Figure 1 6, bottom). At this point, the semiconductor surface is effectively a metal, passing electrons almost freely, thus, the Ohmic like behavior. Nitrogen vacancies are a shallow donor and could indeed cause sufficient band bending for field emission. Point defects would also be more subject to the etching process. Figure 1 7, shows the AFM height image of the interfacial region beneath the diode. The left image shows the diode area beneath the e beam deposited diode. The right image shows the corresponding area beneath the sputter deposited diode. The root mean square roughness is 34% higher on t he left image, 8.86 nm over 6.60 nm. Additionally, a visual comparison is quite revealing. While both samples have a number of pits and a degree of overall roughening, the sputtered sample shows distinct patches of roughened area (yellow) and trenches et ched out between them (shown in red). This pattern is possibly a result of point defects compromising certain areas of the surface that then etched away in the aqua regia solution. SEM (Figure 1 8) also showed this patch like macro morphology beneath the diode area. EDS was performed on the dark and light areas as well as the field. The results are summarized in Table 1 1 below. Looking at the nitrogen percentages, the dark area had the lowest percentage of nitrogen while the light areas had the highes t percentage. Data collected from the field area of the sputtered samples correspond very closely with that of the e beamed diodes. In other words, the area protected by the photoresist polymer has a similar composition to that under the diode area in th e e beam deposited samples. This shows both that the e beam did not damage the surface and it provides a baseline off which to compare the light and dark area EDS spectra. This data further corroborates the hypothesis that nitrogen vacancies are involved and play a role in increased sputter damage. Table 1 1: EDS data. E% refers to the elemental percentage while A% is the atomic percentage. Element type E% (light ) A% (light) E% (dark) A % (dark) E % (field) A% (field) E % (e beam diode) A% (e beam di ode) N 9.53 34.41 8.41 31.36 9.03 33.07 9.12 33.31 Ga 90.47 65.59 91.59 68.64 90.97 66.93 90.88 66.69


S c h e u e r m a n n 9 Figure 1 7: 10 m AFM height images of the interfacial region from the (left) e beam deposited samples and (right) sputter deposited samples. Roughn ess values are 6.60 and 8.86 nm respectively. Figure 1 8: SEM image at 10kx of sputtered sample diode area post etch. The rough patches and trenches in between are visible again.


S c h e u e r m a n n 10 1 3.2 Annealing studies Moving forward from our study of sputter co ater induced damage, another set of GaN Pt/Au diodes was fabricated for thermal studies. AFM scans were taken at 10 m, 5 m, and 1 m. Roughness values are almost always lower as the scan size decreases due to the increased ability to select a homogeneous square area (without pits or artifacts). However, all three offer important insights into the surface morphology. Table 1 2 presents the roughness values for the samples. The first line is the original GaN substrate before annealing or etching. The hig h 10 m roughness shows that the surface contained a number of pits and artifacts to begin with. In this case, the smaller scans provide a better measure of the true roughness of the continuous surface. The first clear trend is the notable increase of rou ghness after etching which is not altogether surprising. Without any annealing treatment the surface roughness increases from 0.489 to 2.90 nm. Then with a 200 C anneal the roughness rockets to 7.69 nm. From this point on, however, it decreases to less than the non annealed, etched value. This suggests that we are situated in a bi variate regime. It seems clear that annealing the metal contributes to failure, roughening the surface. However, additional annealing does not exacerbate the problem and the n the effect of the GaN takes over. As seen in the non etched samples, the surface roughness seems to decrease after an initial increase with higher annealing temperature as well. To study the rate of change of roughness with respect to annealing tempe rature, these roughness values were normalized, setting the value for the 200 C anneal equal to 1. Figure 1 9 shows that this rate of change, the slopes of the two lines, are approximately equal and are relatively robust statistically speaking as indicate d by the R 2 value. This indicates that while the metal is an important part of the interface reaction, the temperature effect in the 200 to 500 C range is governed more by the GaN. The changing morphology of the GaN surface after different annealing temp eratures and diode etching is shown in Figure 1 10. Table 1 2: Raw AFM root mean square roughness data from the control samples, the set without diodes, and the experimental set with diodes. 10m scan R q (nm) 5m scan R q (nm) 1m scan R q (nm) GaN s ubstrate 6.94 1.92 0.489 GaN no diode 200C Anneal 3.48 1.25 0.629 GaN no diode 300C Anneal 3.30 2.83 0.554 GaN no diode 400C Anneal 4.74 1.42 0.402 GaN no diode 500C Anneal 3.80 2.18 0.327 GaN etched no anneal 6.60 8.23 2.90 GaN etched 200C Anneal 8. 96 8.09 7.69 GaN etched 300C Anneal 5.28 3.39 3.13 GaN etched 400C Anneal 4.96 4.03 0.613 GaN etched 500C Anneal 2.90 3.20 1.49


S c h e u e r m a n n 11 Figure 1 9: Comparison of the trend of decreasing roughness with annealing temperature. Roughness values are normalized so that the 200 C point is equal to one for both series allowing for direct overlap of the two best fit lines. R 2 values are shown next to the corresponding line. Figure 1 10: AFM amplitude images showing the changing morphology of the area under the diode after annealing and etching. After 200 C, the roughness was at its highest 7.69 nm (left); After 300 C, 3.13 nm roughness (middle); After 500 C anneal, 1.49 nm roughness. 0.2 0.4 0.6 0.8 1 R = 0.8974 R = 0.9793 Normalized Roughness Annealing Temperature Etched vs Non Etched Control Normalized Comparison Etched diodes Control


S c h e u e r m a n n 12 1 4 Conclusions The chemistry and physics of the GaN metal interface are essentia l for understanding simple devices like MESFETs as well as the more complicated MOSFETs and HEMTs, which will be discussed in the next chapter. The damage induced by the sputter coater has been noted in the literature for a number of years. Previous stud ies focused mainly on quantifying this damage, making parameter adjustments, or the construction of plasma charge traps. In this work, we have compared the technique of magnetron sputter coating to the control, electron beam physical vapor deposition, to illuminate the mechanism of failure. Current mapping revealed Ohmic like behavior across the substrate, especially in the larger diode areas. Hall measurement revealed a carrier concentration of 10 18 sufficient to allow field emission. This now effecti vely metal like surface would conduct freely as opposed to the expected rectifying character of the Schottky metal diode. By etching the diode and performing AFM on the exposed area, we revealed the surface morphology. Trenches and roughened areas led to the hypothesis of nitrogen deficiencies from sputter induced point defects that were more easily etched away. EDS scans of these various areas corroborated this hypothesis by showing nitrogen deficiency in the dark areas as opposed to normal nitrogen lev els in protected areas and on the e beam prepared control samples. Moving forward from this analysis, e beam prepared diodes were annealed to investigate the thermal characteristics of the interface. Etching significantly roughens the surface. After an initial anneal, the roughness is dramatically increased again. However, as annealing temperature is increased, this roughness decreases again until it is lower than even the etch control samples. This same effect is observed in the control where a bare s ubstrate is annealed and characterized. These observations led to the hypothesis that in the 200 to 500 C range, the temperature effect of roughening is governed more by the GaN than by the metal.


S c h e u e r m a n n 1 3 Chapter 2: Dielectric/GaN Interfaces 2 1. Introductio n The first step in understa nding the mechanisms of failure was going back to characterize the interactions at the metal and semiconductor interface (Chapter 1). With a baseline understanding of interfacial reactions and thermal characteristics, we then sought to investigate the more relevant dielectric/GaN interface prese nt in MOSFET and HEMT structures. Building on previous results, we moved to study the electrical performance of e beam fabricated diodes before and after thermal treatments. To date, the Abernathy group has developed two highly successful crystalline die lectrics for use with GaN: Sc 2 O 3 and MgO. These dielectrics have also been used in power HEMT devices demonstrating improved power performance. Similarly Sm 2 O 3 has been utilized for use with GaAs as a filed dielectric and as a gate dielectric. 1 For both s emiconductor systems, we are attempting to fill the gap where little information is known about the long term stability of the dielectric semiconductor interface 2,3 concentrated areas were addressed here: the thermal stability of AlGaN/GaN HEMT devices, Figure 2 1: Vertical schematic of a AlGaN /GaN High Electron Mobility Transistor (HEMT) and an investigation of the yellow defect through the ozone processing step. HEMT devices or High Electron Mobility Transistor s are field effect transistor like MESFET and MOSFET, but instead of a doped region in the channel, HEMTs have a heterojunction. In this case, it is AlGaN and GaN that form the heterojunction crucial to HEMT performance. Figure 2 1 shows a vertical cross section schematic for a AlGaN/GaN HEMT. In most semiconductors, charge carriers are donated by impurities. A Figure 2 2: Band gap diagram of AlGaN/GaN HEMT coordination indicating the 2D electron gas. fundamental trade off results in the fact that the impurities themselves then slow mobility by increasing the number of electron collisions. The heterojunction is able to cleverly avoid this pitfall. As seen in Figure 2 2, a quantum well is formed where the AlGaN and GaN come together as the Fermi le vel aligns from the two different band gaps (E B = 6.1 eV for AlGaN and E B =3.4 eV for GaN). The AlGaN layer is highly doped, these electrons slide into the quantum well and become trapped there. Additionally, since the GaN layer is undoped, the electrons have an incredibly high mobility. This two dimensional electron gas, 2DEG, is indicated in the diagram above.


S c h e u e r m a n n 14 With applied thermal stress, a number of defects can form, most of which exhibit different optical properties than the semiconductor surrounding s. Cathodoluminescence (CL) takes advantage of this fact whereby electrons from a cathode ray tube impinge on the surface and the differing emitted luminescence is captured. This technique allows for the visualization of defects within the upper layers o f the sample. In the second part of this study, an ozone processing step is studied by AFM and photoluminescence. Another important area still needing to be fully addressed in III V semiconductor technology is the presence of deep levels. Deep levels c an be caused by native defects, interface defects, and deep impurities among other factors. While some important deep levels have been identified in other III 2 level in GaAs and the DX center in Al x Ga 1 x As 4 more work needs to be done to illu minate the critical defect physics at work in GaN. Previous photoluminescence region at 2.2eV 5 10 Other studies have shown how varying the excitation density changes the relative intensities of the band gap emission and 11 Only two weeks ago at the APS meeting, researchers from Santa Barbara gave a talk purporting to have solved the yellow defect origin. They claim that carbon substituting for nitrogen acts as a deep acceptor as o pposed to the believed shallow acceptor. Using first principles, they have calculated the absorption and emission of such a deep acceptor and shown that it aligns with the yellow defect region 12,13 We are interested, in particular, in how ozone treatment s can affect this luminescence spectra. Control and understanding of GaN luminescence properties are crucial for GaN based optical device development as well as for a better understanding of the defect physics at work. 2 2 Experimental Procedures 2 2.1Th ermal experiment design The AlGaN/GaN substrate was characterized by tapping AFM before annealing, after annealing, and after the diode etching process. Electrical characterization was done on diodes before and after annealing with an analytical IV prob e station. Diodes were tested across the sample surface to test for continuity and different sized diodes were tested to measure the area dependence of rectifying behavior. The JEOL 6400 was used for SEM and CL imaging after etching on the area beneath t he diode pad. CL measurements were taken at 15 and at 5kV in the field as well as in the exposed diode area. 2 2.2 Fabrication of Schottky diodes Diodes were fabricated using bi layer photolithography and electron beam physical vapor deposition. The p rocess is analogous to that described in section 1 2.1 yielding 1m stacks with 200 nm of Pt and 800 nm of Au. The diode array pattern is also the same and can be seen in Figure 1 4. 2 2.3 Diode etching The etching process was also analogous to that pr esented in chapter one using diluted aqua regia solution for approximately 12 minutes per sample. 2 2.4 Ozone experiment design GaN samples were subjected to ozone treatments of 1, 5, 10, 20, 40, and 80 minutes and the photoluminescence was measured bo th before and after ozonation. A HeCd laser source at 325 nm is used for excitation. GaN luminescence is directed through a monochromator and collected by a photomultiplier tube. The intensity of each spectra was carefully collected allowing for a


S c h e u e r m a n n 15 direc t comparison of the relative intensity of the emission spectra. Scans are taken of the full spectra, 350 to 750 nm, as well as exclusively around the GaN bandgap 3.4eV ~365 nm. In addition to photoluminescence spectroscopy, the substrates are imaged by t he AFM, and roughness values are calculated both before and after ozone treatments to monitor any changes. 2 3 Results and Discussion 2 3.1 Electrical performance of HEMTs Two samples of AlGaN/GaN Schottky diode arrays were processed and electrically c haracterized as a starting point (see appendix 2A 1 through 2A 4 for full IV data). A current density plot showing moderate rectifying behavior is shown in Figure 2 4. Three diode areas are plotted here, D with diameter 426 m, F with diameter 215 m, an d H with diameter 112 m. As the cross sectional area decreases, the reverse breakdown increases slightly although the effect is moderate. The forward compliance appears to be unaffected. This is possibly a function of surface defects in the semiconduct or that cause device leakage. As the cross sectional area of the diode increases, more of these defects by probability are in the interface contributing to the lower breakdown voltage. Diode performance decreased after both 5 shows the current density measurement for the H diode compared before and after annealing. Rectifying behavior decreases by almost 2V. Similar decreases were observed across th e diode array leading to near Ohmic like behavior for the largest diodes. These dramatic changes in diode performance indicate a clear effect from the thermal treatment. This is not surprising given the high temperature regimes selected, 400 and On the other hand, GaN is known for its stability at high temperatures. The Debye temperature is the highest attainable with a single normal vibration within the crystal. For an industry metric, AlGaN based switc hes are eliminating the need for costly and inefficient cooling systems 14 So, device failure at these temperatures is particularly interesting and relevant as GaN based electronics become more widespread. To help differentiate interface effects from surface effects, AFM measurements were taken at each stage of the experiment (Figure 2 3 and table 2 1 shown below). The original step flow morphology is shown in the far right with root mean square roughness of 0 .686 nm. After with roughness of 0.641 nm. The third image Figure 2 3: From left to right AFM height images of 1) AlGaN substrate, 2) AlGaN substrate after anneal, 3) AlGaN substrate after anneal a nd etch treatment, 4) area under diode after anneal and etch. Roughness values are 0.686, 0.641, 0.408, and 0.973nm respectively.


S c h e u e r m a n n 16 Table 2 1: Root mean square roughness values calculated from the above AFM height measurements Sample Root mean square roughness on 5um scan (nm) Original AlGaN substrate 0.686 AlGaN after 400C anneal 0.555 AlGaN after 700C anneal and etch (no diode) 0.408 AlGaN after 400C anneal and diode etch 0.973 AlGaN after 700C anneal and diode etch 1.71 Figure 2 4 : Current Density plot showing the rectifying behavior of three differently sized Schottky diodes. Figure 2 5 3.00 2.00 1.00 0.00 1.00 2.00 3.00 3.00 2.00 1.00 0.00 1.00 2.00 Current Density (A/m 2 ) Voltage (V) Current Density of Diodes D,F,H D F H 8.00 4.00 0.00 4.00 8.00 4.00 3.00 2.00 1.00 0.00 1.00 Current Density (A/m 2 ) Voltage (V) Diode H after 700 C Anneal Before anneal After anneal


S c h e u e r m a n n 17 from the left shows the surface after being subject to the etching aqua regia solution. The surface is now even smoother with R q = 0.408 nm. Finally we look at the area beneath th e etched off metal diode and see significant roughening and loss of the step flow spotted morphology was observed with roughness values of 0.973 nm and 1.71 nm respectively. As in the GaN/metal experiments of ch apter 1, EDS scans confirmed that the surface morphology was all GaN and not residual metal. 2 3.2 Cathodoluminescence under the diode area Cathodoluminescence was performed on the exposed diode areas after etching as well to further illuminate potenti al sources of failure. The first image in figure 2 6 shows the SEM pictograph of the exposed diode area edge. The diode had already been etched away, but a distinct shadow was left behind. Figure 2 shows the CL of this image, 200x magnification and 5eV energy. There is no distinct difference between the diode area and the field indicating the absence of outstanding defects in one area over the other. Images were then taken at greater magnification, 15,000x. The 15kV electrons, which will penetrate dee per into the sample, did not return any outstanding differences. The 5kV scans also returned nearly identical images for the exposed diode area (3) and the field region (4). This indicates that the thermal treatments did not cause outstanding defects in the underlying GaN layer at any of the depths probed. Equation 2 1 shows that the penetration depth of the electrons varies according to the density of the material and of course the energy of the electrons themselves l and E is given in kilovolts: (2 1) Figure 2 6: 1 SEM image 200x of the exposed diode area (darker region) 2 CL on the same spot reveals no outstanding characteristics. A close up of the surface at 15kx magn ification and 5kv energy reveals no difference between the diode area (3) and the field (4).


S c h e u e r m a n n 18 3 so at E = 15kv the CL indicates that annealing treatments are not causing luminescent defect s at or below 140 nm in the semiconductor. 2 3.3 Photoluminescence of ozone treated samples The other aspect of this chapter involves a photoluminescence study of the ozone processing step often used during GaN device fabrication. Figure 2 7 shows the AFM (height image on the left, amplitude on the right) of the GaN films before and after ozone treatment. Figure 2 7: AFM of n GaN samples pre ozone treatment (A height and B amplitude images) and post ozone treatment (C height, D amplitude). Roughne ss values are 0.374 nm (pre) and 0.417 nm (post). The step flow across the surface indicates a smooth and well grown sample. This flow is conserved after the ozone treatment with the roughness level increasing only slightly to 0.417 nm from 0.374 nm. Ozo ne interacts with the surface first by reacting with and removing contaminating carbon, but also by incorporating O 2 into the surface layer to depths on the order of half a nanometer. The PL spectra on the next page (Figures 2 8 and 2 9) present the befo re and after data from the 80 minute ozone treatment. As seen in Figure 2 8, both the shape and position of the GaN band gap peak were unaffected by the ozone process. In Figure 2 9, a saw toothed peak does decrease notably in relative PL emission with r espect to the band gap and satellite band gap peaks. This saw tooth peak results from the so emitting between 450 nm and 650 nm. No clear trend emerged from the samples ozoned at different times. All showed a slight relative redu ction of the yellow defect peak and preservation of the band gap peaks. 2 3 Conclusions Annealing treatments of Schottky diodes on AlGaN/GaN showed decreased electrical also increased with increasing diode are a, an effect that was even more pronounced after annealing. This indicates that surface defects in the GaN are playing a role in failure; probabilistically more of these are included under the larger diode areas. AFM studies reveal that neither anneali ng nor etching alone have a dramatic effect on surface morphology and roughness (in fact, both actually slightly reduce the roughness while maintaining step flow morphology). The etching of diodes does significantly increase roughness. Further, the rough ness almost opposed to the GaN/metal interactions studied in chapter 1 where the roughness then decreased with higher annealing temperatures. The exact mechanism for this roughening is still unknown at this po int. EDS confirmed that the


S c h e u e r m a n n 19 Figure 2 8: Photoluminescence spectra of the GaN peak associated with the band gap. The shape and position of the band gap peak is almost perfectly identical before and after ozone treatment. Figure 2 9: Full photolumin escence spectra of GaN taken before and after the ozone treatment. The blue line is almost perfectly eclipsed at the 350 360 370 380 390 400 Photoluminescence Intensity Wavelength (nm) GaN Peak Ozone Treatment Pre Ozone Post Ozone 350 450 550 650 750 Photoluminescence Intensity Wavelength (nm) Full Spectra Ozone Treatment Pre Ozone Post Ozone


S c h e u e r m a n n 20 composition was indeed 100% GaN, meaning no residual metal was left on the surface. Ca thodoluminescence on the exposed diode area did not reveal defects between 15kv and 5kv, calculated as 970 up to 140 nm of depth. CL images of the exposed diode area and the field area were practically identical indicating that defects were not formed at or below 140 nm from the annealing process. The resolution of the CL spectra quickly degrades closer to the surface although 2kV scans (~30 nm) are feasible. It is possible that the annealing treatment is still forming defects in the near surface region above 140 nm. Moving on from this work, a number of modern techniques could be used to better probe this region. An electrostatic force microscopy atomic force microscopy tandem method, for example, could probe the electrical activity beneath specific re gions revealing areas with higher defect concentration. Another tool that reveals sub surface defects in a particularly clever way is the scanning ultrasound holography probe. Acoustic waves passed up through the sample will be modified by defects and ma terial irregularities. These waves can then be picked up by the AFM above the surface allowing for the imaging of defects in the 10 to 100 nm depth range 15 Either of these techniques would help illuminate the defects in the very near surface region and thus, further explore the chemistry and physics of device failure at elevated temperature. In the second part of this investigation, GaN samples were treated with ozone leading to an upper oxide layer that had only a minor effect on surface roughness in creasing it from 0.374 nm to 0.417 nm. Photoluminescence spectroscopy showed that the position and shape of the band gap peak was maintained for all ozone treatments. The yellow defect region was decreased in relative intensity by the ozone treatment in all cases. This indicates that a passivation layer is formed rapidly that restricts further oxygen incorporation. One possible explanation for this decrease in yellow defect emission is the ozone removal of carbon impurities that have been considered cau ses for this emission. Another theory is that gallium vacancies in n type GaN contribute to the yellow defect. In this case, dangling gallium bonds could be passivated by the ozone treatment reducing this emission. Recent work by Lyons, Janotti, and Van de Walle has shown, using first principles, that carbon replacing nitrogen acts as a deep acceptor. Their calculations place the resultant emission and absorption patterns in the same area as the observed yellow defect signal 12,13 In this case perhaps t he mechanism is a combination of the two: removing carbon and satisfying dangling gallium bonds to decrease the yellow defect emission. Given the shallow nature of the ozone passivation layer, EFM could be used here as well to show changes in electrically active defect regions before and after ozone treatment. XPS spectra would also offer further insight by investigating the shift in the oxygen, carbon, and GaN peaks. One additional future direction would be measuring the PL spectra at elevated temperatu re to increase the magnitude of the yellow defect region. In this case, more minute changes could be quantified statistically.


S c h e u e r m a n n 21 Chapter 2 Appendix Figure 2A 1: IV plot of diode array 1 before annealing treatment. Diodes are listed accorded t o their assigned letter (Figure 1 4). Diameters are from A to H, 1138 m, 815 m, 574 m, 462 m, 290 m, 215 m, 154 m, and 112 m. Figure 2A behavior of the diodes is reduced resulting in Ohmic like behavior, in particular for diodes D through A. 1.50E 07 1.00E 07 5.00E 08 0.00E+00 5.00E 08 1.00E 07 1.50E 07 4 3 2 1 0 1 2 Current (A) Voltage (V) Pt/Au HEMT Diode Array 1 Intial A B C D E F G H 1.50E 07 1.00E 07 5.00E 08 0.00E+00 5.00E 08 1.00E 07 1.50E 07 2 1.5 1 0.5 0 0.5 1 Current (A) Voltage (V) Pt/Au HEMT Diode Array 1 Post 700 C Anneal A B C D E F G H


S c h e u e r m a n n 22 Figure 2A 3: IV plot of diode array 2 before annealing treatment. Diode A was a complete short both before and after the anneal and is thus not shown. Figure 2A like behavior for almost all diodes. 1.50E 07 1.00E 07 5.00E 08 0.00E+00 5.00E 08 1.00E 07 1.50E 07 3 2.5 2 1.5 1 0.5 0 0.5 1 1.5 Current (A) Voltage (V) Pt/Au HEMT Diode Array 2 Initial B C D E F G H 1.50E 07 1.00E 07 5.00E 08 0.00E+00 5.00E 08 1.00E 07 1.50E 07 2 1.5 1 0.5 0 0.5 1 1.5 Current (A) Voltage (V) Pt/Au HEMT Diode Array 2 Post 400 C Anneal B C D E F G H


S c h e u e r m a n n 23 Chapter 3: Organic/Metal Interfaces 3 1. Introduction 3 1.1 Background Nanotechnology, often heralded as the savior of the twenty first century, is broadly defined as technological advances and science performed at the nanoscale. Within thi s broad category of research, however, there are a number of powerful, if not game changing, ideas being investigated, and these last two chapters will investigate two particular thrusts into nanotechnology energy science As a participant in the UF/France Summer 2010 Research Experience for Undergraduates (REU) program, I had the unique opportunity to spend twelve weeks in Toulouse, France under the tutelage of Dr. Jacques Bonvoisin. The research, spanning inorganic chemistry to materials science to physic s, was performed in the GNS group at the national labs of France (Centre National de la Recherche Scientifique). Active research at CEMES is exploring two cutting edg e areas of nanotechnology: Molecular scale machines and molecular scale electronics. The specific project discussed here involves novel planar organic molecules synthesized and then imaged and manipulated on metal substrates. These molecules are componen ts of larger macromolecular assemblies hypothetically capable of performing a variety of logic functions. The concept of molecular electronics strives to meet the ultimate size minimization currently conceivable by science; through molecular electronics t he first molecular calculator can be realized. In addition to working towards these novel devices, we are clearing new ground investigating the properties of the individual unit. Up to this point, the vast majority of scientific knowledge has concerned s ubstances as a collective of molecules all more or less acting together. Thus, with the advent of our ability to harness and control individual units, a new chapter must be opened on single molecule properties both chemical and physical. Fig. 3 1 Represe ntation of a Ruthenium based molecular SWAP Figure 3 1 depicts a metal organic macromolecule capable of performing a SWAP function. Conjugated organic backbone connects the four metal centers acting as nanowire. Two of the four arms will contain logic, a s shown in fi gure 3 2 and the other two will operate the SWAP. Figure 3 2 shows how incident light changes the spin pairing of the two switching metal centers to accomplish a swap between the molecule. Spin up is encoded as 0, wh ile spin down is 1 1,2 One key component to functionality is optical selectivity between the metal centers. It is hypothesized that through manipulating ligand


S c h e u e r m a n n 24 Figure 3 2 Schematic of a molecular SWAP. Binary logic is encoded in the spin of transition metal valence electrons, in this case the 5s electron of a ruthenium center. Incident light drives an electrochemical reaction pairing ctron spin and thus, accomplishes the binary SWAP. coordination chemistry, the optical reactivity of the metals can be selectively tuned. The chemical properties in relationship to stability in the complex and appropriate geometry will be essential as we ll. In addition to this chemistry, however, the physical realm represents a new challenge as we move to operating single molecules independent of bulk solution. 3 1.2 Our Approach Two approaches are underway to achieve the optical selectivity discussed in the last section. 1) Differentiating ruthenium (III) centers with new constituent groups and 2) synthesizing other compounds with different paramagnetic metal centers. This study investigates planar complexes with Cu(II), Ni(II), and Zn(II). The plana rity of these complexes yields ideal substrates for UHV STM and AFM imaging of both diamagnetic and paramagnetic centers, the latter of which represents a novel thrust in this type of imaging. 3 2 Experimental Procedures 3 2.1 Phase 1: Symmetric Cu(dbm) 2 The first attempted reaction was the synthesis of bis(dibenzoylmethane) copper(II) as carried out previously by Ma and Gao in 1999 3 As shown in Fig. 3 3 the reaction scheme is quite simple involving a Cu(II) salt and the desired ligand (dibenzoylmethane) The reaction was carried out in a number of ways but primarily with the salt KHCO 3 acting as a base to deprotonate H dbm (deprotonated form show n in Fig. 3 3), which then, being made Figure 3 3: Ligand replacement synthesis scheme of Cu(dbm) 2


S c h e u e r m a n n 25 Fig ure 3 4: CuCl 2 solution in MeOH (left), reaction flask after 4 hours of mixing (middle), two batches of identical powder from CH 2 Cl 2 (right) active, attacks the central metal ion displacing the chlorine. In the initial experiment, 1.000g of Cu(Cl) 2 was mixed with 100mL of methanol in a 250mL round bottom flask producing a vibrant green solution. 3.340g of dbmH and 1.492g of KHCO 3 were added to the reaction flask producing a slow change to opaque yellow green (cf. Fig. 3 4 ). This solution was filtered a nd then extracted with CH 2 Cl 2 /H 2 O to remove residual salt. A more pure product was obtained by recrystallization in CH 2 Cl 2 yielding a slightly darker yellow green powder. 3 2.2 Phase 2: Ni and Zn analogs While copper systems constitute the primary focu s for the specific macromolecular gate applications in mind nickel and zinc analogs were also synthesized and characterized to provide a breadth of scientific understanding. These two synthesis were carried out in 90% ethanol as done by Soldatov and Heneg ouwen in 2001. 2 Figure 3 5: Zinc salt in solution (left), Zn(H 2 O) 2 (dbm) 2 (middle), dehydrated Zn(dbm) 2 (right) Figure 3 6: Nickel salt in solution (left), Ni(H 2 O) 2 (dbm) 2 (middle), dehydrated Ni(dbm) 2 (right)


S c h e u e r m a n n 26 3 2 .3 Phase 3: Synthesis of Asymmetric Cu(acac)(dbm) A significant portion of the work done focused on a multitude of attempts to better synthesize and isolate Cu(acac)(dbm). The first documented appearance of the asymmetric copper complex was in September 2 009. Prasad and Kushwaha performed two variations on a one pot synthesis citing 85% yield for Cu(acac)(dbm). 4 The specific equilibrium between the three likely species Cu(acac) 2 Cu(acac)(dbm), and Cu(dbm) 2 was investigated and found dependent on solvent choice, addition rate, and choice of metal source (i.e. which copper salt was used). Dependencies on temperature, pH, and other factors are also expected although not explicitly explored here. My various experiments are detailed below: 1) One pot synthes is starting from Cu(Cl) 2 in methanol 2) One pot synthesis starting from Cu(Cl) 2 in 2 EtOH: 1 H 2 O: 1 (CH 3 ) 2 CO 3) Triflic acid reaction from Cu(dbm) 2 4) Two pot synthesis starting from CuSO 4 in 2 EtOH: 1 H 2 O: 1 (CH 3 ) 2 CO 5) Two pot synthesis starting from Cu( NO 3 ) 2 3H 2 O in 2 EtOH: 1 H 2 O: 1 (CH 3 ) 2 CO 6) One pot synthesis starting from Cu(OAc) 2 H 2 O in 1 ethanol: 2 n pentane: 2 toluene 7) One pot synthesis starting from Cu(OAc) 2 H 2 O in 5 ethanol: 1 THF: 4 toluene 8) One pot synthesis starting from CuSO 4 in 4 THF: 1 H 2 O 3 2.4 Triflic Acid Experiment In 1990 Kasahara and co. introduced a very interesting Ruthenium chemistry not completely understood, but very useful for the types of applications with which the Bonvoisin dik etonato)ruthenium(III) complex Kasahara illustrated how treatment with strong acid in acetonitrile uniformly removed one of the three bidentate ligands replacing it with two acetonitrile groups. A 1:1 quantitative reaction was observed across a range of a cids showing that even in the presence of a large excess of diketonato group was removed. 5 This experiment has been reproduced in the labs here with good success. In fact, this is a principal step through which we synthesize asymmetric rut henium compounds, another component of the macromolecular gate. The same experiment was attempted with the copper, zinc, and nickel compounds previously described. The dibenzoylmethane form of the metal complex was refluxed in acetonitrile and degassed wi th argon. Triflic acid is added, 120% of the stoicheometric quantity. The solution was refluxed for 2 hours, cooled to room temperature, and evaporated. The residue was then solvated in 20mL of ethanol while 280mL of n pentane was added to precipitate o ut the product which was gathered by vacuum filtration. Figure 3 7: Asymmetric ligand replacement, Cu(acac) 2 and Cu(dbm) 2 are also expected in various proportion


S c h e u e r m a n n 27 3 2.5 Ni(acac)(dbm) and Zn(acac)(dbm) Synthesis attempts of the nickel and zinc asymmetri c analogs were both made toward the end of my stay in France. Both syntheses followed similar schemes to (8) outlined for copper above. 3 2.6 Phase 4: Single molecule imaging and manipulation A solution of 10 4 M Cu(dbm) 2 was created from purified sing le crystals dissolved in CH 2 Cl 2 Solvent was obtained via a solvent purification system (SPS) and argon was used throughout solvation to minimize contact to air and other contaminants. A single drop of this solution was spread on a glass slide and examine d with the AFM. The distribution and character of the compound was deemed such that the UHV STM could then successfully be used for single molecule imaging and manipulation. In addition, TGA DTA on the Cu(dbm) 2 powder confirmed that the decomposition tem perature and character were appropriate for the necessary sample preparations (Figure 3 8). Figure 3 8: TGA DTA of Cu(dbm) 2 in He. Unlike in the ruthenium analogs where we a step wise thermal decomposition is observed, this c haracterization predicts that both ligands are removed simultaneously. For UHV STM imaging, molecules were distributed on a NaCl/Cu(111) substrate and measured at 5K. 3 2.7 Materials and Characterizations All chemicals used were of reagent grade or bet ter. H NMR was rec orded on a Bruker AMX 500 in CD 2 Cl 2 IR spectra of samples in KBr pellets were taken on a Perkin Elmer 1725 FT IR spectrophotometer. Mass Spectroscopie University using ES (Perkin Elmer SciexSystem API 365). Both FAB and DCI spectrograms were obtained. Cyclic voltammograms were obtained with an Autolab system (PGSTAT 100) in dry dichloromethane (DCM) (0.1M tetrabutylammonium hexafluorophosphate, TBAH) at 25 C with a th ree electrode system consisting of platinum disk working (1 mm diameter), platinum wire counter, and saturated calomel reference electrodes. UV Vis was performed with a Cary 5000 UV Vis NIR Spectrophotometer in both dichloromethane and tetrahydrofuran (THF ). For the physical characterization of single molecules, a commercial low temperature (5K) microscope (Omicron Nanotechnology Taunusstein,Germany) equipped with a tuning fork of the qPlus sensor type 5 and a control electronic system from SPECS (Zurich, Switzerland) was used for this study. All the data shown in this report was obtained at 5K. The Cu(111) and Ag(111) substrates were cleaned by cycles of Ar+ sputtering (600 eV) followed by annealing at 750 K. NaCl was deposited by thermal evaporation on t hese substrates at 300 K in order to get a partial coverage of NaCl bilayer islands. The molecules were deposited from a heated crucible on the sample maintained at low temperature in the microscope.


S c h e u e r m a n n 28 3 3 Results and Discussion 3 3.1 Phase 1: Results for Symmetric Cu(dbm) 2 Crude yields for Cu(dbm) 2 were obtained in the 80 to 95% range differing by the base and solvent choices used for synthesis. This more than two fold increase over the reported crude yield in 1999 (42%) is believed attributable to clos er attention to product solubility properties. The Cu(dbm) 2 complex is very insoluble in almost all regularly encountered solvent atmospheres. In two of the more effective solvents, CH 2 Cl 2 and THF, the copper complex was found to have solubilities of appr oximately 1mg/1mL and 3mg/1mL respectively. In most cases the purity of ancillary crops of crystals is identical to that of the first as detectable by DCI mass spectroscopy and FT IR (cf. Fig. 3 4, right). Product Characterization H NMR, FAB and DCI mass spectroscopy, cyclic voltammetry, UV vis spectroscopy, FT IR, TGA DTA, and XRD analyses were utilized to characterize the product. Characterization was done before and after each recrystallization, on the different fractions isolated, and across different experiments studying solvent and base effects. H NMR was carried out initially, but, not surprisingly, the paramagnetic nature of the Cu (II) center convoluted the results beyond reliable analysis. Electrochemistry revealed a quasi reversible two step red uction, likely a reflection of Cu (II) to Cu (I) to Cu (0) and the corresponding confirmation change to tetrahedral in the Cu (I) state (See appendix figures 3A 4 and 3A 5 ). Due to the high insolubility of the complex and the very different solubility charact ers of possible contaminants, significant attention was required to yield spectra of reliable integrity. Ultimately, DCI mass, UV vis, and infrared spectroscopy all gave reproducible results indicating the presence and purity of the compound. In powder f orm the compounds were made pure enough so as to have no unidentifiable mass spectra peaks above 1% abundance. Crystallization was then used to increase the purity further to the levels required for single molecule imaging, eliminating all by products. Cr ystal Structure The high insolubility prompted the use of assays spanning a myriad of possible crystallization strategies. Two were identified as most successful: 1) Slow diffusion in long test tubes of narrow diameter (~1cm) filled about of the way wi th a saturated solution of THF (d=0.8892g/mL) under a miscible lighter solvent -various were successful although n pentane (d=0.6262g/mL) was particularly effective. (Figure 3 9, left). 2) Slow evaporation (~5 days) with 100% CH 2 Cl 2 (Figure 3 9, right). XR D performed by Nicolas Ratel Ramond, also of the CEMES laboratory, gave the expected almost perfectly square planar structure of the complex (Figures 3 10 and 3 11). Bond angles found (see Appendix Table 3 A 3) were statistically identical to those reporte d by Ma and Gao. 3


S c h e u e r m a n n 29 Fig ure 3 9: Slow diffusion tubes (left), crystals left by slow evaporation in CH 2 Cl 2 (right) Figure 3 10: Ortep diagram of Cu(dbm) 2 with 50% probability for the thermal ellipsoids Figure 3 11: Crystal packing of Cu(db m) 2 as found by XRD


S c h e u e r m a n n 30 3 3.2 Phase 2: Results of Ni and Zn analogs For zinc (cf. Fig. 3 5) the diamagnetic Zn(dbm) 2 was obtained with 62% yield (as compared to 65% reported in 2001). On a second experiment 77% yield was achieved by altering the synthesis a nd dehydration processes. In both cases the products were confirmed by DCI mass spectroscopy and H NMR. The NMR spectra revealed peaks only for the solvent and the Zn(dbm) 2 protons with accurate p eak area ratios (see Appendix 3A 4 ). The nickel compoun d (cf. Fig. 3 6 ) was prepared in the analogous manner with a crude yield of 97.6% (compared with 95% reported in 2001). A mass loss of 7.3% was observed after theoretical loss). The product was confirmed by DC I mass spectroscopy as well as XRD. Crystal Structure Both Ni(dbm) 2 and Zn(dbm) 2 proved more difficult to crystallize than the Cu(dbm) 2 parallel. The same set of strategies was applied resulting in a set of distinct observations. Both the nickel and zi nc solutions created two different crystal structures with unique degradation properties. As seen in the picture below, bulky crystals were grown that initially formed at the solvent interface in slow diffusion tubes, then dropping to the bottom of the gla ss. Higher up in the tube thin crystal striations were observed. With zinc these striations dried and persisted but when washed with the same solvent easily dissolved. The face of the bulk crystals immediately reverted to powder form when exposed to air The striated crystals were not adequate for XRD study and the bulk crystals did not persist long enough for time consistent data. Figure 3 12 : (Left to right) bulky zinc crystals, striated zinc crystals, bulky nickel crystals, striated nickel crystals Nickel also produced two distinct crystal phases, but with marked differences. The thin upper crystals also turned powder like when exposed to air, unlike with zinc. The bulkier Ni(dbm) 2 crystals were also subject to degradation in air, but did persist just long enough for XRD study. It is worth noting that treating any of the above species with ether resulted in immediate and complete reconversion to the powder form. XRD revealed that the solvent, THF, integrated into the structure to create a n octahedral coordination (Figure 3 13) This octahedral coordination lends light to the typically with the tetrahedral and octahedral


S c h e u e r m a n n 31 Figure 3 13: Ortep diagram of Ni(dbm) 2 (THF) 2 with 50% probability for the thermal ellipsoids coordinations, but not with square planar) In fact, Solodatov and Henegouwen isolate four variations of Ni(dbm) 2 in their study. The crystal structure is given, is indeed found to be diamagnetic. Ni(dbm) 2 (THF) 2 seems to be a new crystal structure, but this type of crystal is by no means novel. Ni(dbm) 2 form octahedral complexes upon crystallization has been well chronicled with other ligands, quinoline 7 isoquinoline 7 methanol 8 and benzene 9 among others. This also explains its instability at room temperature and easy redissolution in the mother liquor. 3 3.3 Phase 3: Results for Asymmetric Cu(acac)(dbm) The work of Prasad and Kushwaha was mentioned in the experime ntal section where by they cite 85% yield for Cu(acac)(dbm). However, their paper, only mentioning Cu(acac)(dbm) as an aside, does not go into detail to characterize the composition of this yield. My work has gone on to show that such a one pot synthesis n ot surprisingly creates a mix of three possible species Cu(acac) 2 Cu(dbm) 2 and the desired Cu(acac)(dbm). Early experiments were quantified with FAB mass spectroscopy as done by Prasad and Kushwaha, but further experiments led us to believe this was n ot the best analytical technique for this product. Prasad and Kushwaha provide a spectrum where the most salient Cu(acac)(dbm) peak is no higher than 3% abundance relative to the matrix. At this kind of threshold the signal is practically lost in the noise making it impossible to extricate the purity of the compound and in particular its ratio to the other major species Cu(acac) 2 and Cu(dbm) 2 The nature of FAB mass spectroscopy and the respective solubility of each species in the solvent chosen can also sk ew this apparent abundance. Mass spectra of my first experiment yielded peaks for all three species around 2% abundance, very comparable to the peaks cited by Prasad and Kushwaha. My second experiment, however, gave incredibly different


S c h e u e r m a n n 32 results with the initial precipitate exhibiting 100% abundance for Cu(acac) 2 over 10% for Cu(acac)(dbm), and hardly 1% for Cu(dbm) 2 There were many other fractions from various experiments that almost exclusively yielded Cu(dbm) 2 The most enriched powder of the mixed ligand complex to date was procured by the 6 th scheme outlined above. DCI m ass spectra of this powder showed 60% relative abundance for Cu(acac)(dbm) it is also worth noting that DCI mass spectra were done in retrospect on earlier powders and that the Cu( acac)(dbm) signal was still small to absent. So, the s e latter studies did show a significant enrichment in Cu(acac)(dbm) apart from the different ionization schemes. An accumulation of observations from the various solvent, salt, and reaction ordering cho ices allowed for the formulation of a selectivity mechanism hypothesis: Both the acac and dbm ligand have a comparable affinity to conjugate to the metal center and with the given geometry no steric effects will swing the equilibrium one way or the other Therefore, I hypothesize that the primary driving force behind species selectivity is solubility. The most insoluble product will dominate as it drops out of solution and then continues to form. Thus, the solubility of the product in coordination with the rate of reactant addition and the ordering of steps will have a profound effect on the species distribution. For example, with the 2 EtOH: 1 H 2 O: 1 (CH 3 ) 2 CO solvent mixture the insolubility is Cu(dbm) 2 > Cu(acac)(dbm) > Cu(acac) 2 From the previous as sertion we would expect the product composition to follow a similar pattern, but with certain rates and reaction sequences we observe alternate effects. The most common result with this solvent mixture (the same used by Prasad and Ku shwaha) is by percent abundance Cu(dbm) 2 2 >> Cu(acac)(dbm). I hypothesize further then, that the rate of Cu(dbm) 2 formation and precipitation is such that the dibenzoylmethane ligand becomes adequately depleted in solution early on. With a low free dbm concentration, we then observe a large amount of Cu(acac) 2 regardless of its relative solubility. behind many of the later manipulations. The 6 th and 7 th schemes saw a return to the one pot synthesis but now with a new mixture of solvents. By combining solvents like ethanol, with a higher affinity for Cu(acac) 2 and toluene, that should interact most with the phenyl rings on Cu(dbm) 2 we hoped to encourage a more even distribution if not possibly favor the asymmetric adduct. As noted earli er, we did in fact observe the highest concentration of Cu(acac)(dbm) to date with the 6 th scheme. 3 3.4 Triflic Acid Experiment Results A s documented in the literature 6 ruthenium undergoes a dramatic color change as the triflic acid removes the attac hed ligands and replaces them with the acetonitrile solvent molecules. In all three cases of the planar compounds (copper, nickel, and zinc), no dramatic color change was not observed. An overall dulling of the solution, rather, was the primary effect. Mass spectroscopy revealed different results for each metal complex. For copper, the Cu(dbm) 2 peak was reduced from 50% to 1% relative abundance (relative to the largest peak, in both cases the ligand fragment). The powder gathered after acid treatment a lso showed a major peak that corresponded to [Cu(dbm) + 4CH 3 CN] and then the associated [Cu(dbm) +4CH 3 CN + NH 3 ] (a peak 17 higher is observed for all peaks using ammonia with desorption chemical ionization); The abundances were around 10% and 22% respectiv ely. For the zinc trial, small peaks were present for a [Zn(dbm) + CH 3 CN] and for the


S c h e u e r m a n n 33 desired [Zn(dbm) + 2CH 3 CN]. Nickel on the other hand yielded a plethora of peaks with abundances in the 5 to 15% range that were impossible to delineate; a multitude of both fragment and dimer peaks further convoluted this spectra. In all three cases, the treatment lacked the robust character observed with the ruthenium complexes. Even in the case of zinc, where the desired compound was observed, the lack of a solubi lity gradient already complicated by an overall insolubility made this treatment impractical. The exact mechanism behind this unique chemistry with ruthenium is still not known, and its extension to other metal complexes is an area that merits scientific study. At this point in time, however, this manipulation was deemed insufficient for our purposes. 3 3.5 Ni(acac)(dbm) and Zn(acac)(dbm) Ni(acac)(dbm) and Zn(acac)(dbm) were synthesized as in scheme (8) detailed in 3 2.3. This THF/H 2 O solvent system le d to complicated outcomes for all three metal systems. The desired complex was e xpected to remain soluble enabling it to then be collected by solvent evaporation. Insoluble product, however, was collected for copper and zinc that remained undissolved in p ractically all solvent mediums. It is possible that the specific solvent atmosphere utilized led to the formation of metal clusters or even metal organic polymers. By example, the starting material, zinc acetate, has been known to form clusters under cert ain conditions for over 50 years. 10 Typical analyses, while attempted on these compounds, were made difficult if not impossible by their complete insolubility. The solutions gathered from these two experiments as opposed to the insoluble precipitates, w ere evaporated and the new powders collected. For these products we were able to carry out mass spectroscopy and determine the powder compositions. For zinc, the only peak corresponded to the dbm ligand indicating that either this powder too had low so lubility and was missed by the detector or that this really was just residual starting material and that all the zinc had been effectively consumed in the production of the sticky cluster like precipitation described earlier. Nickel, on the other hand, we nt more or less according to theory. There was no sticky precipitate in the initial reaction and the gathered crop from the evaporated solution had a nice texture and bright green color. Mass spectra of this powder revealed Ni(acac)(dbm) and Ni(dbm) 2 pro portioned approximately 1:4. This, again, is technically a new compound -Prasad and Kushwaha cite two attempted syntheses of Ni(acac)(dbm), both of which only formed Ni(dbm) 2 4 however, at this early unpurified stage it is of little historical importance. 3 3.6 Single Molecule Imaging and Manipulation The STM image of a single Cu(dbm) 2 molecule is shown in figure 3 13. Four bright lobes correspond roughly to the four phenyl rings situated around the copper center. Upon closer inspection, one of the gro oves between these lobes is more pronounced. This suggests that the ligands, with their conjugated bidentate structures, are positioned along the connected lobes (as shown with superimposed molecular structure). The apparent molecular height is 100 pm. Beyond imaging these molecules, we needed to discover switching mechanisms that would allow for logic encoding functionality. Figure 3 14 shows how a bias of +2V on the tip caused an oxidation state change of the copper center. The geometry changes im mediately from square planar to 3D with an increased


S c h e u e r m a n n 34 height over 200 pm (Fig. 3 Applying 2V restored the original configuration reversibility of the process; an essential aspect for switching behavior. To better monitor this switching behavior, the tunneling current was recorded showing sharp transitions at +2V and 2V (Figure 3 15). The waiting time between switching events was a few hundred milliseconds. The Cu(dbm) 2 was also monitored amb iently for several hours and no spontaneous switching behavior was observed demonstrating stability on the NaCl bilayer. Another experiment was done whereby Cu(dbm) 2 molecules were directly adsorbed on Cu(111). In this latter case, it was not possible to induce the switching mechanism. The NaCl bilayer is playing an essential role either through its specific adsorption properties or by its electronic decoupling effect of the adsorbed molecule relative to the metallic surface. The geometric transition di scussed above is well documented in the literature of copper complexes corresponding to the transition from Cu(II), preferring the square planar geometry, to Cu(I), which adopts a tetrahedral formation. Cyclic voltammetry of these compounds revealed a par tially irreversible transition, which has also been documented in the literature 11 ,1 2 Cu(II) is a d 9 metal, thus the structure is stabilized by the Jahn Teller effect. The negatively charged Cu(I) complex is d 10 meaning there is no crystal field stabili zation energy preference. In lieu of this effect, steric and electrostatic effects favor a tetrahedral geometry 1 3 To better probe the electronic structure of the adsorbed molecule tunneling spectroscopy experiments were performed giving IV and dI/dV sp ectra (Figure 3 15). The dI/dV spectrum begins increasing sharply around 1.8V and is truncated by the transition to the tetrahedral state at 2V. Rapid switching made it difficult to image the molecule at this stage. Nevertheless, an image was obtained a t 1.85V that shows notably different characteristics. Figure 3 13: Constant current STM images of Cu(dbm) 2 on an NaCl bilayer on Cu(111). (a) Tunne ling current I t = 1pA, bias voltage V t = 1V. (b) I t = 700 fA, V t = 1.2 V.


S c h e u e r m a n n 35 Figure 3 14: Reversible switching of two Cu(dbm) 2 molecules. (a) (e) Sequence of STM images showing the reversible switching of two molecules. The position of the tip above the molecule during the application of the bias voltage is marked by a black circle. (f) 3D representatio n of image b. Image size: 6.9 nm x 5.5 nm. Imaging conditions: It = 400 fA, Vt = 100 mV. Figure 3 15: STM and IV data of the two redox states. I(V) and dI/dV spectroscopy curves obtained on the square planar (black lines) and the tetrahedral (red line) molecular species. The numbers that appear on the I V curves refer to the numbering of the STM images shown below the graph (size: 3.5 nm x 3.5 nm). They are located at the corresponding imaging bias voltage.


S c h e u e r m a n n 36 The structure has 6 lobes, larger lateral dim ensions, and has a height more than 300 pm. A similar phenomenon occurs around 1.7V although this image reaches ~500 pm in height Recent work 14 including research at CEMES 15 has shown similar behavior resulti ng from molecular ion resonances, which dominate the spectrum due to the electronic decoupling of the molecule from the metal surface by the NaCl bilayer. For the open shell Cu(II) d 9 complex, the resonance likely originates in the singly occupied molecul ar orbital (SOMO). In the Cu(I) complex the resonance is attributed to the first occupied level (HOMO). 3 4 Conclusions In phase 1, the synthesis of Cu(dbm) 2 was further explored and the yield was improved over the first documented synthesis in 1999 th rough a deeper investigation of solubility properties. A number of crystallization schemes were identified as successful, in particular, slow diffusion of n pentane into THF and evaporation of dichloromethane. Crystals were grown to sufficient purity to obtain an XRD crystal structure that corroborates the 1999 literature result. A study of other transition metals, nickel and zinc, broadened our chemical understanding of the system. In addition, a new compound was synthesized, crystallized, and charac terized Ni(dbm) 2 (THF) 2 The mixed ligand synthesis of Cu(acac)(dbm) was also improved reaching higher levels of purity. The compound has not been crystallized successfully yet, but as better synthesis routes and purification schemes increase the ratio of Cu(acac)(dbm) to Cu(dbm) 2 and Cu(acac) 2 in solution, obtaining a pure crystal will become inevitable. The compound Cu(dbm) 2 was successfully imaged on an NaCl bilayer on Cu(111) with both AFM and STM. The mechanism for a redox based molecular switch has been discovered and characterized chemically, electrically, and physically 16 This is a big finding that opens new perspectives for the future of molecular engineering. The molecular switch is not only highly reliable but is also robust, not requiring d elicate preparation procedures on the metal substrate to be operative. The stark conformational change unambiguously reveals the charge state of the molecule without requiring repeated delicate AFM measurements. In addition to the lucrative opportunities in molecular electronics this molecule could be used as an electrochemical transducer whereby the rearrangement caused by the oxido reduction cycle could drive a synthetic molecular motor 17. Future and ongoing work includes more trials to crystallize Cu (acac)(dbm) so that it too can be investigated by AFM and STM. Imaging and tunneling experiments are being conducted with other metal substrates as well such as Ag(111) to further investigate the metal dependence on this organic/metal interface. Eventual ly this type of technology could lead to the realization of the molecular calculator, a truly momentous step for not just nanoscience but science as a whole.


S c h e u e r m a n n 37 Chapter 3 Appendix Figure 3A 1 : FT IR spectrum of Cu(acac) 2 Figure 4: FT IR spectrum of a Cu(acac) 2 Cu(acac)(dbm), and Cu(dbm) 2 mixture Figure 3A 2 : FT IR spectrum of a Cu(acac) 2 Cu(acac)(dbm), and Cu(dbm) 2 mixture


S c h e u e r m a n n 38 Figure 3A 3: H NMR of Zn(dbm) 2 with integrations and proton associations Figure 3A 4: Electroche mistry on Cu(dbm) 2 run forwards (0 to 1.5). A quasi reversible process is shown. The first curve (green) shows that oxidation does not occur until after the corresponding reduction has already taken place.


S c h e u e r m a n n 39 Figure 3A 5 : Electrochemistry on Cu(dbm) 2 run b ackwards (0 to 1.5). As before the behavior in reduction is not exactly mirrored in reduction. The reaction can also seen to change with multiple cycles (3 cycles shown here). Figure 3A 6: UV vis spectra of Cu(dbm) 2


S c h e u e r m a n n 40 SUPPLEMENTARY TABLES for Cu(dbm) 2 Crystal Structure Table 3 A 1 Crystallographic data Formula C 30 H 22 CuO 4 Crystal System monoclinic FW (g/mol) 510.02 Space Group C 2/c a() 26.066 b() 6.017 c() 16.568 115.08 Z 4 1 ) 0.963 calc (g/cm 3 ) 1.439 Table 3 A 2 Complete listing of bond distances (Angstroms) C(1) C(3) 1.392(5) C(1) C(4) 1.510(4) C(1) O(1) 1.277(4) C(2) C(3) 1.399 (4) C(2) C(10) 1.506(4) C(2) O(2) 1.271(4) C(3) H(3) 0.90(4) C(4) C(5) 1.384(5) C(4) C(9) 1.383(5) C(5) C(6) 1.391(5) C(5) H(5) 0.89(4) C (6) C(7) 1.366(5) C(6) H(6) 0.90(4) C(7) C(8) 1.366(5) C(7) H(7) 0.89(4) C(8) C(9) 1.388(5) C(8) H(8) 0.92(4) C(9) H(9) 0.86(3) C(10) C(11) 1.396(5) C(10) C(15) 1.384(4) C(11) C(12) 1.378(5) C(11) H(11) 0.89(3) C(12) C(13) 1.370(5) C(12) H(12) 0.91(4) C(13) C(14) 1.375(5) C(13) H(13) 0.90(4) C(14) H(14) 0.930(4) C(14) C(15) 1.391(5) C(14) H(19) 0.50(30) C(15) H(15) 0.90(4) CU(1) O(1) 1.918(2) CU(1) O(1) 1.918(2) CU(1) O(2) 1.907(2) CU(1) O(2) 1.907(2) T able 3 A 3 Complete listing of bond angles (degrees) C(3) C(1) C(4) 120.5(3) C(3) C(1) O(1) 124.5(3) C(4) C(1) O(1) 115.0(3) C(3) C(2) C(10) 120.5(3) C(3) C(2) O(2) 124.6 (3) C(10) C(2) O(2) 114.9(3) C(1) C(3) C(2) 124.5(3) C(1) C(3) H(3) 119.5(20) C(2) C(3) H(3) 115.9(20) C(1) C(4) C(5) 123.2(3) C(1) C(4) C(9) 118.8(3) C(5) C(4) C(9) 118.0(3) C(4) C(5) C(6) 120.7(4) C(4) C(5) H(5) 123.2(22) C(6) C(5) H(5) 116.1(22) C(5) C(6) C(7) 120.2(4) C(5) C(6) H(6) 118.0(21) C(7) C(6) H(6) 121.8(21) C(6) C(7) C(8) 120.0(3) C(6) C(7) H(7) 119.0(23) C(8) C(7) H(7) 120.9(23) C(7) C(8) C(9) 120.0(4) C(7) C(8) H(8) 121.2(23) C(9) C(8) H(8) 118.7(23)


S c h e u e r m a n n 41 C(4) C(9) C(8) 121.1(3) C(4) C(9) H(9) 119.6(19) C(8) C(9) H(9) 119.1(19) C(2) C(10) C(11) 118.7(3) C(2) C(10) C(15) 123.2(3) C(11) C(10) C(15) 118.1(3) C(10) C(11) C(12) 120.6(3) C(10) C(11) H(11) 117.2(19) C(12) C(11) H(11) 122.1(19) C(11) C(12) C(13) 120.6(4) C(11) C(12) H(12) 120.0(24) C(13) C(12) H(12) 119.3(24 ) C(12) C(13) C(14) 120.0(4) C(12) C(13) H(13) 122.3(21) C(14) C(13) H(13) 117.6(21) C(13) C(14) H(14) 120.1(4) C(13) C(14) C(15) 119.7(4) C(13) C(14) H(19) 146.9(326) H(14) C(14) C(15 ) 120.1(4) H(14) C(14) H(19) 82.2(309) C(15) C(14) H(19) 45.4(326) C(10) C(15) C(14) 121.0(3) C(10) C(15) H(15) 121.1(21) C(14) C(15) H(15) 118.0(21) O(1) CU(1) O(1) 180.0 O(1) CU(1) O(2) 87.0(1) O(1) CU(1) O(2) 93.0(1) O(1) CU(1) O(2) 93.0(1) O(1) CU(1) O(2) 87.0(1) O(2) CU(1) O(2) 180.0 C(1) O(1) CU(1) 126.4(2) C(1) O(1) CU( 1) 126.4(2) CU(1) O(1) CU(1) 0.0 C(2) O(2) CU(1) 126.8(2) C(2) O(2) CU(1) 126.8(2) SUPPLEMENTARY TABLES for Ni(dbm) 2 (THF) 2 Crystal Structure Table 3 A 4 Crystallographic data Formula C 30 H 22 NiO 4 ( C 4 H 8 O) 2 Crystal System monoclinic FW (g/mol) 649.4 Space Group P 2 a() 10.837 b() 7.774 c() 19.365 92.43 Z 4 1 ) 0.641 calc (g/cm 3 ) 1.323 Table 3 A 5 Complete listing of bond distances (Angstroms) C(1) H(1) 0.930(4) C(1) C(2) 1.385(5) C(1) C(10) 1.374(6) C(2) H(2) 0.930(4) C(2) C(8) 1.387(5) C(4) C(6) 1.398(5) C(4) C(7) 1.507(5) C(4) O(2) 1.264(5) C(5) C(6) 1.395(5) C(5) C(8) 1.512(5) C(5) O(3) 1.264(5) C(6) H(6) 0.930(4) C(7) C(11) 1.382(5) C(7) C(14) 1.382(5) C(8) C(12) 1.3 89(5) C(9) H(9) 0.930(5) C(9) C(13) 1.370(7) C(9) C(15) 1.364(6) C(10) H(10) 0.930(4) C(10) C(16) 1.373(6) C(11) H(11) 0.930(4) C(11) C(15) 1.385(6 ) C(12) H(12) 0.930(4) C(12) C(16) 1.383(5)


S c h e u e r m a n n 42 C(13) H(13) 0.930(4) C(13) C(14) 1.378(6) C(14) H(14) 0.930(4) C(15) H(15) 0.930(5) C(16) H(16) 0.930(5) C(20 ) H(20A) 0.970(7) C(20) H(20B) 0.970(6) C(20) C(21) 1.474(9) C(20) O(1) 1.425(6) C(21) H(21A) 0.970(8) C(21) H(21B) 0.970(8) C(21) C(22) 1.513(11) C(22) H(22A) 0 .970(7) C(22) H(22B) 0.970(8) C(22) C(23) 1.522(9) C(23) H(23A) 0.970(6) C(23) H(23B) 0.970(7) C(23) O(1) 1.428(7) Table 3 A 6 Complete listing of bond angles (degrees) H(1) C(1) C(2) 120.3(4) H(1) C(1) C(10) 120.3(4) C(2) C(1) C(10) 119.4(4) C(1) C(2) H(2) 119.4(4) C(1) C(2) C(8) 121.3(4) H(2) C(2) C(8) 119.4(4) C(6) C(4) C(7) 119.7(3) C(6) C(4) O(2) 124.7(4) C(7) C(4) O(2) 115.6(3) C(6) C(5) C(8) 119.8(3) C(6) C(5) O(3) 124.9(3) C(8) C(5) O(3) 115.2 (3) C(4) C(6) C(5) 126.4(4) C(4) C(6) H(6) 116.8(4) C(5) C(6) H(6) 116.8(4) C(4) C(7) C(11) 124.0(3) C(4) C(7) C(14) 118.3(3) C(11) C(7) C(14) 117.8(4) C(2) C(8) C(5) 118.5(3) C(2) C(8) C(12) 118.2(3) C(5) C(8) C(12) 123.3(4) H(9) C(9) C(13) 120.1(5) H(9) C(9) C(15) 120.1(5) C(13) C(9) C(15) 119.8(4) C(1) C(10) H(10) 119.8(4) C(1) C(10) C(16) 120.4(4) H(10) C(10) C(16) 119.8(5) C(7) C(11) H(11) 119.4(4) C(7) C(11) C(15) 121.1(4) H(11) C(11) C(15) 119.4(4) C(8) C(12) H(12) 119.7(4) C(8) C(12) C(16) 120.6(4) H(12) C(12) C(16) 119.7(4) C(9) C(13) H(13) 119.9(5) C(9) C(13) C(14) 120.2(4) H(13) C(13) C(14) 119.9(5) C(7) C(14) C(13) 121.1(4) C(7) C(14) H(14) 119.5(4) C(13) C(14) H(14) 119.5(4) C(9) C(15) C(11) 120.0(4) C(9) C(15) H(15) 120.0(5) C(11) C(15) H(15) 120.0(4) C(10) C(16) C(12) 120.1(4) C(10) C(16) H(16) 120.0(4) C(12) C(16) H(16) 119.9(4) H(20A) C(20) H(20B) 108.7(5) H(20A) C(20) C(21) 110.6(6) H(20A) C(20) O(1) 110.6(5) H(20B) C(20) C(21) 110.6(6) H(20B) C(20) O(1) 110.6(6) C(21) C(20) O(1) 105.7(5) C(20) C(21) H(21A) 110.8(7) C(20) C(21) H(21B) 110.8(7) C(20) C(21) C(22) 104.9(6) H(21A) C(21) H(21B) 108.8(7) H(21A) C(21) C(22) 110.8(7) H(21B) C(21) C(22) 110.8(7) C(21) C(22) H(22A) 110.9(7) C(21) C(22) H(22B) 110.9(7) C(21) C(22) C(23) 104.2(5) H(22A) C(22) H(22B) 108.9(7) H(22A) C(22) C(23) 110.9(7) H(22B) C(22) C(23) 110.9(7) C(22) C(23) H(23A) 110.9(6) C(22) C(23) H(23B) 110.9(6) C(22) C(23) O(1) 104.3(5) H(23A) C(23) H(23B) 108.9(6) H(23A) C(23) O(1) 110.9(6) H(23B) C(23) O(1) 110.9(5) C(20) O(1) C(23) 105.1(4)


S c h e u e r m a n n 43 Chapter 4: Nanoparticles 4 1. Introduction 4 1.1 Background As mentioned in the introduction to chapter 3, nanotechnology is an umbrella term encompassing a myriad of scientific thrusts all now taking place on the nanoscale. Another course, in actuality, we have been dealing with nanoparticles as lon g as man has known about fire it is now known that burning wood creates carbon nanotubes in the ashes. Today, however, we are able to control and characterize nanoparticles opening the path for new innovations. During the summer of 2009, I worked as an i ntern/contractor at Sandia National Labs in Albuquerque, New Mexico operated by Lockheed Martin for the Department of Energy and the Department of Defense under contract DE AC04 94AL85000. An old brand of magnesium oxide called Maglite S had been used s electrolyte binder in a thermal battery. While Figure 4 1: Thermal battery schematic thermal batteries contain many different layers, the underlying principle is still relatively simple. The anode and cathode are sep arated by a frozen electrolyte so that the battery is non operative until the electrolyte is melted by a thermal component. This allows the latent power to be maintained over decades without decay. The thermal element can be accomplished in a number of w ays. In the diagram shown, a Zr/BaCrO 4 fuse strip is stretched across all the layers. Ignited from the top, the fuse strip sends heat all the way to the core and bottom of the battery melting the electrolyte which is a mix of various salts of potassium a nd lithium with chloride, fluoride, and bromide; often a eutectic mixture is preferred. The thermal battery is under significant internal pressure during manufacturing and storage, 150 to 250 psi, to maintain good interfacial contact. Upon electrolyte me lting, these pressures relax to 10 to 40 psi. In order to maintain control of the electrolyte solution and ensure 100% reliability of these batteries, an electrolyte binder is needed, an additive that will modify the properties of the electrolyte lending i t strength and preserving its electrical properties. Figure 4 2: The dependence of viscosity on particle size and particle packing density. A ceramic binder, in this case magnesium oxide, is the additive used that can accomplish this task; lending str ength and simultaneously preserving conductivity. The electrolyte and binder are blended into a pellet, the form used in the thermal battery. Originally


S c h e u e r m a n n 44 Freon TF was used in this process, but since being banned liquid nitrogen has taken its place as the blending agent 1 The pellets are placed to measure the deformation behavior; in most cases deformation occurs within 30s ending with a steady state thickness. Competing mechanisms have been shown to cause a bove optimal deformation when using both large and small magnesium oxide particles. With larger particles, the deformation increases with a higher porosity. The pore structure collapses under the applied pressure and the pellet sinks down into itself. W ith smaller particles, the rheology of the pellet becomes important. As the porosity increases there is actually less deformation. This is due to the fact that at low porosity i.e. close packing, the pellet starts to rotate, spreading material laterally and decreasing the height of the particle again by a large amount. 2 Apropos to these sensitive property regimes, controlling the properties of the ceramic binder is essential. The same applies with papers and tapes, as in tape casting of MgO for thin cer amic substrates 3,4 Maglite S, the binder used for so many years, was trade secreted so when the company died out so did the synthetic procedure. Initial efforts at Sandia focused on reverse engineering the powder. MgO starting material was purchase d and processed through the necessary steps comparing wetting angle, surface morphology, pore size, and deformation data along the way. In 2008, the Advanced Power Sources R&D group created a replacement that worked. After a time, it was necessary to pur chase more starting material. When this second lot of the same powder went through the treatments something was drastically different (Figure 4 2). The second lot of powder, while supposedly identical to the first, acted incredibly differently throughout the processing steps. As seen in Figure 4 2 above, the agglomeration Figure 4 2: SEM micrographs of the two lots showing an incredible difference at 5000x in material is not reproducible despite having iden tical characterization. properties are incredible different. This is evidence that the performance/properties of the MgO powders reside in a multivariate space with complicated dependencies. It hints, further, that there are likely important characteris tics of the material that we do not know how to ascertain at this point. To continue working with such a multivariate system, Sandia would have to perform this analysis for each powder. Clearly this is an incredibly inefficient if not impossible task. A new undertaking began in 2009 for Sandia to create its own MgO nanoparticles


S c h e u e r m a n n 45 from the Mg(OH) 2 precursor phase. Specific control over the dehydration, or calcining step, would finally take Mg(OH) 2 to MgO yielding nanoparticles of the desired characteri stics. By illuminating these simple steps we embarked on a project to publish this data for the wide scientific community. By so doing, we hope to end this cycle of the DOD and DOE continually funding the reinvention of the wheel and at last producing a stable, well known, and reproducible processing path to our nano composite electrolyte thermal battery separator. 4 2. Experimental 4 2.1 Materials and reactor calibration A typical precipitation procedure was employed by adding dilute precipitant (2.41M KOH) drop wise to large volumes of MgCl 2 6H 2 O (1000ml of 0.43M or 1200ml of 0.51M) in a two liter jacketed glass reactor (Figure 4 3) with a lower drain valve and upper ground glass joints. Six experiments were initially carried out to effectively debug t he reactor, minimizing process variables. A thermal control system (heated water) is used which maintains the reactant temperature within + 1 o C. A Masterflex peristaltic pump with a 751810 head is used to meter the precipitant (OH ) at a consistent, repea table rate. The reactor vessel is swept with nitrogen to minimize the effects of CO 2 4 2.2 Mg(OH) 2 nanoparticle (NP) synthesis Using a two level four factorial design of experiments matrix, binomial variables were tracked over sixteen separate batches of Mg(OH) 2 ; factor 1 being temperature (75 or (0.43 or 0.51M), factor 3 the rate of precipitant injection (2.0 or 3.5 on the peristaltic pump), and factor 4 the agitation sp eed (1.5 or 2.0 on the motor); see Fi gure 4 3 (right). Run # Factor 1 (Temp.) Factor 2 (Conc.) Factor 3 (Rate) Factor 4 (Agitation) 1 high high high low 2 low low high low 3 low high high high 4 low high low high 5 low high low low 6 high high low high 7 low low high high 8 hig h low low low 9 high low high low 10 high low high high 11 high high low low 12 low low low high 13 low low low low 14 low high high low 15 high high high high 16 high low low low Figure 4 3: DOEx matrix. Samples done in random blind order to el iminate human bias and extraneous effects; almost the engineering version of the double blind experiment. 4 2.3 Characterization of Mg(OH) 2 NPs All sixteen batches were ground for 15s in a CRC micromill and oven Surface morphology (partic le size and geometry) and agglomeration characteristics were measured using a Zeiss Supra 55VP scanning electron microscope (SEM). Specific surface area and particle size were calculated from the BET method using a Micromeritics Acclerated Surface Area an d Porosimetry (ASAP) 2020 instrument; eight isotherm (P/P 0 ) points were adsorptions rejecting four of the higher and lower values. Bulk and tap de nsity were also measured.


S c h e u e r m a n n 46 Figure 4 3: Reactor set up (left), reactio n scheme(middle), reactor and variable schematic (right). The four variables are shown on the schematic (right). All other conditions are held constant. 4 2.4 Calcination Two sets of the Mg(OH) 2 nanoparticles were calcined in different ovens at different locations to prove repeatability. The first set was calcined at the AML lab, at the University of New Mexico under a fume hood using a F48000 Thermolyne furnace with an Omega digicator monitoring the thermocouple. The second set was calcined in a dry roo m at Sandia National Laboratories. In both cases the samples thermal ramping in either the warming or cooling phases. 4 2.5 Characterization of the MgO NPs All characterizations performed on the Mg(OH) 2 NPs were repeated with the derived M gO for direct comparison. In addition, pore volume and area distribution calculations were carried out through the BJH method using the ASAP 2020. 4 3. Results and Discussion 4 3.1 Mg(OH) 2 NP synthesis and characterization After an initial six sample r un, the reactor was calibrated. The 16 sample blind DOEX set was created and the NPs were characterized in a number of ways. Figure 4 4 (a), (b), and (c) show three SEM pictographs of the different morphologies observed in the as grown Mg(OH) 2 As seen i n (a) broader and thinner plates resulted in stacking both horizontally and vertically. This is contrasted starkly with (c) w here equilateral growth of the a and c axes resulted in an observed rounding and agglomeration via clustering of the primary part icles rather than stacking. Initially, since the data was blind, a sorting of the data was carried out without reference to the growth conditions. Crystal and agglomeration properties were assigned subjectively based on a large number of SEM measurements at a range of magnifications. This was only one of many analyses made, the rest of which conformed to accepted scientific standards. It is important to synthesize data in new ways, however, with reference to the nature of the problem being investigated. As mentioned in the introduction, subsequent lots of MgO powder wound up our known scientific characterizations. As we move into the new realm of nanotechnology, new property regimes are emerging that me rit new approaches. After this analysis, the synthesis parameters were revealed and the 16 samples were reorganized with respect to these


S c h e u e r m a n n 47 (a) (b) (c) Fig 4 4 0.43M MgCl 2 low rate and 2


S c h e u e r m a n n 48 Table 4 1: The 16 blind DOEX samples reorganized by their synthesis parameters. The last two columns show the crystal category and agglomerate size score that were subjectively assigned before the synthesis conditions were revealed. Run # Factor 1 (Temperature) Factor 2 (Concentration) Factor 3 (Rate) Factor 4 (Agitation) Crystal Category Agglomerate Siz e 1 75 0.51 3.5 1.5 1 2 15 75 0.51 3.5 2 2 1 11 75 0.51 2 1.5 1 1 6 75 0.51 2 2 1 2 9 75 0.43 3.5 1.5 1 2 3 10 75 0.43 3.5 2 1 2 2 16 75 0.43 2 1.5 1 1 8 75 0.43 2 2 1 2 1 14 5 0.51 3.5 1.5 2 1 3 5 0.51 3.5 2 2 3 5 5 0.51 2 1.5 2 2 4 5 0.51 2 2 2 3 2 5 0.43 3.5 1.5 3 2 7 5 0.43 3.5 2 3 2 13 5 0.43 2 1.5 2 3 2 12 5 0.43 2 2 2 3 1 Table 4 2: Particle size, surface area, bulk, and tap density data for the blind DOEX set Run # Particle Size BET Surface Area Bulk Density (g/cm3) Ta p Density (g/cm3) 1 59.6623 nm 42.7941 0.1978 m/g 0.609 0.859 15 51.6204 nm 49.4609 0.2415 m/g 0.463 0.638 11 50.0642 nm 50.9983 0.2296 m/g 0.649 0.539 6 67.9388 nm 37.5808 0.1639 m/g 0.468 0.646 9 48.9880 nm 52.1188 0.2359 m/g 0.458 0.639 10 46.3026 nm 55.1415 0.2461 m/g 0.478 0.649 16 40.1101 nm 63.6545 0.3196 m/g 0.336 0.481 8 51.1455 nm 49.9202 0.2988 m/g 0.396 0.559 14 41.5863 nm 61.3950 0.2651 m/g 0.543 0.761 3 45.6433 nm 55 .9379 0.2065 m/g 0.533 0.755 5 44.2008 nm 57.7635 0.2904 m/g 0.519 0.724 4 41.7812 nm 61.1086 0.3073 m/g 0.522 0.704 2 33.4256 nm 76.3842 0.2485 m/g 0.499 0.685 7 31.6590 nm 80.6466 0.2988 m/g 0.508 0.698 13 35.0136 nm 72 .9200 0.3132 m/g 0.509 0.701 12 36.9314 nm 69.1333 0.3167 m/g 0.500 0.698


S c h e u e r m a n n 49 parameters (as opposed to the initial random numbering). Agglomerate size ratings did not correspond strongly with any of the four variables implying that no one or combination of the four variables studied controls agglomeration with high accuracy. Crystal characteristics, however, corresponded strongly with temperature and concentrations and corresponded lightly with agitation and injection rate (Table 4 1). Table 4 2 presents the particle size, surface area, bulk, and tap density data for these powders in their reorganized order. On average, the higher temperature setting and higher MgCl 2 concentration led to larger nanoparticles. The larger nanoparticles, not s urprisingly had lower surface area, but were also found to have not only higher bulk density, but also higher tap density. 4 3.2 Calcination The DOEX set was calcined next to study the dehydration process. Additionally, the 2 nd 6 th 7 th and 11 th pow der were also calcined in a second location, in a different oven, as outlined in the experimental section, to reveal the reproducibility and sensitivity of the process. At least 8 images were taken of each sample from each batch at different locations, a few of which are shown below in Figure 4 5. As can be seen here and in table 4 3, samples 2 and 7had almost identical particle size after calcinations 33 and 32 nm particles before and 30 and 31 nm particles afterwards. Samples 6 and 11 went from 68 and 50 nm particles to 80 and 77 nm particles after calcination. The images from the calcined samples in room 894 show similar particle size and surface morphology. Most of the discrepancy in the images was due to greater charging difficulties on the 894 sam ples resulting in stronger image contrast. 4 3.3 Reproducibility Trials An essential component of any to be industrial process is a reproducibility demonstration. From the data so far we identified too major regimes of nanoparticle growth relying most heavily on temperature and also to an extent on MgCl 2 concentration. The rounding morphology observed in the 7 th sample is considered an outlier and, furthermore, calcining this sample produced approximately the same end result as the other low temperatur e morphologies. Samples 14 and 16 were interesting in particular. Despite completely opposite synthesis parameters, besides the agitation setting, the nanoparticle size and surface area were only 3.5% different. This region was hypothesized as being mor e robust towards the synthesis parameters and was thus selected for reproducibility trials. Samples R3, R5, and R9 were created and characterized. Before calcinations, all three samples gave similar morphologies and particle sizes 41.7 1.2 nm, presenti ng at the 95% confidence interval. After calcinations two of the samples showed a dramatic increase in particle size while one showed a dramatic decrease (Table 4 4). Run # Particle Size BET Surface Area 2 29.8050 nm 56.2313 0.1713 m/g 6 79.5118 nm 21.0784 0.0613 m/g 7 30.9147 nm 54.2130 0.1548 m/g 11 77.1610 nm 21.7205 0.0540 m/g Table 4 3: Particle size and surface area data for the AML calcined samples corresponding to the original set DOEX 2, 6, 7, and 11. As seen in the table and in Figure 4 5, powders 2 and 7 have similar characteristics distinct from 6 and 11.


S c h e u e r m a n n 50 Figure 4 5: SEM pictographs of samples DOEX 2, 6, 7, and 11. The Mg(OH) 2 precursor phase is compared to MgO calcined in the two different laboratories, the AM L (Sandia offsite lab with University of New Mexico) and room 894. Run # Mg(OH) 2 Particle Size Mg(OH) 2 BET Surface Area MgO Particle Size MgO BET Surface Area R3 40.6314 nm 62.8379 0.2990 m/g 88.1581 nm 19.0110 0.0614 m/g R5 41.7461 nm 61.1601 0.2566 m/g 95.5348 nm 17.5431 0.0535 m/g R9 42.8129 nm 59.6361 0.2398 m/g 27.8245 nm 60.2340 0.1743 m/g Table 4 4: Pre and post calcinations data for the reproducibility trials. As seen, the pre calcining data appears very reproducible, bu t the post calcination data shows two distinct outcomes, neither of which closely follows the trend observed with previous samples.


S c h e u e r m a n n 51 Figure 4 6 (a) through (c): Cumulative pore volume charts for the R3, R5, and R9 nanopowders. R3 and R5 are almost pe rfectly identical, but R9 is radically different. The total pore volume is an order of magnitude larger and there are almost no pores in the 60 to 100 nm diameter range.


S c h e u e r m a n n 52 Figure 4 6 (d) through (f): Derivative plots. Again R3 and R5 are very similar R9 lacks pores of larger diameter but contains many pores in the intermediate range while the former two powders contain many small pores and a few large pores.


S c h e u e r m a n n 53 This surprising difference was further probed by investigating the porosity characteristics of the powders. Distributions can be calculated through either gas desorption or adsorption, here the desorption plots are shown. In Figure 4 6, plots a through c show the cumulative pore volume distributions of the three powders, while plots d through f show the derivate plots. R3 and R5 (images a, b, d, and e) are almost perfectly identical in terms of porosity. They contain large and intermediate pores in moderate levels, with a comparatively high number of pores under the 10 nm range. The R9 powde r almost completely lacks pores in the 60 to 100 nm range, but contains a large number of intermediate sized pores. Previous studies have shown that the pore structure plays an essential role in governing the mechanical properties of the electrolyte binde r pellets 2 essential for correct thermal battery operation. In reference to this fact, this unexplained variance represents a major concern before any synthesis scheme can be certified. On the other hand, despite having diverged from strong correlation i n the precursor phase, both of these MgO profiles still fall within the range of powders that have worked in thermal battery pellets historically. While this variance certain merits more research it does not undermine the powders usefulness for their appl ication. 4 4 Conclusion By starting from the precursor phase, we are attempting to gain control over the precise morphology and nanoparticle characteristics of MgO powder. In particular, these powders are to be used as a binder in thermal battery elect rolyte pellets, and thus, need to meet external characterization requirements. Four distinct binomial processing variables were investigated and overall a high degree of robustness was found. Nanoparticles between 32 and 68 nm are reported in a binomial distribution oriented most strongly around the two temperature regimes and secondly around the two concentration settings. These nanopowders were calcined to MgO leading to a divergence of results. Despite the reproducibility and apparent control over th e precursor phase, the latter phase still displayed a significant degree of variance. Regimes of temperature and concentration have emerged where Mg(OH) 2 nanoparticles are indeed robust to other factors. Plate like particles are grown resulting in sta cking agglormeration. One case of rounding, equilateral growth along the two axis, was reported. This resulted in clustering as opposed to stacking in the secondary structure. This potential outlier, however, yielded a similar MgO powder to the other tr eatments after calcinations. During reproducibility trials, an unexpected divergence showed that there is another factor controlling MgO nanoparticle properties that we have not yet characterized or may not be able to characterize at this point. Future work continuing now is illuminating this issue while simultaneously testing a number of the powders developed for their intended application in thermal batteries. It is expected that the large majority of the powders synthesized will work in thermal batt eries. Even then, given the sensitive nature of this work, higher degrees of reproducibility are desired. When dealing with nanoparticles, both the chemistry and the physics of the interfacial reactions are important. This issue comes to head here as our nano enabled electrolyte binder mix must stand up to the physical restriction of deforming pellets and simultaneously meet the requirement of high and consistent electrical conductivity when melted between the battery anode and cathode. As with molecu lar electronics, nanotechnology brings new challenges as we move towards dealing with smaller units. The science of interface


S c h e u e r m a n n 54 interactions is more crucial today than ever in developing, understanding, and controlling electronic devices.


S c h e u e r m a n n 55 Acknowledgements For research performed at the University of Florida, acknowledgements are graciously extended to the REM scholars program, the Arnold and Mabel Beckman Foundation, the Barry M. Goldwater Foundation, and th e Department of Materials Science and Engineering for funding. I would like to acknowledge Dr. Ivan Krachenko for assistance with sputter coater parameterization and the graduate students Andy Gerger, Ryan Davies, Tony Stewart, and Andrew Herrero for trai ning and assisting me throughout this learning process. The Major Analytical Instrumentation Center (MAIC) at the University of Florida is acknowledged as well for making available the SEM and AFM instruments used. For the research performed in France, s pecial thanks are extended to the University of Florida program coordinators Dr. Michael Scott and Dr. Adrian Roitberg, to the coordinator in Toulouse Dr. Gwnal Rapenne, to my mentor Jacques Bonvoisin, and to my coworkers for their assistance and guidan ce, Sabrina Munery and HP Jacquot among others. Acknowledgement for the funding that made this work possible is given to the National Science Foundation (NSF) through the REU program, the Beckman Program for their continued support of my scientific endeav ors, and to the Universit Paul Sabatier for housing and accommodation. In addition, I would like to acknowledge Nicolas Ratel Ramond for XRD work and le Service Commun de Spectromtrie de masse at Universit Paul Sabatier for mass spectroscopy. At Sand ia National Labs, special thanks are given to Dr. Karen Waldrip for her mentorship and continuing involvement in my research career. Bonnie McKenzie is acknowledged for providing SEM work, Eric Coker for EDS, and EMF Systems Inc. for their collaboration i n synthesizing the nanoparticles. Funding was provided by the Beckman Program and by the Material Science department at the University of Florida, as well as a travel grant provided by the HHMI Science for Life program at UF. For overseeing and guiding my training, personal and professional development throughout my undergraduate research career, greatest thanks are extended to my mentors Dr. Cammy Abernathy and Dr. Brent Gila. Without their mentorship and friendship none of this would have been possib le.


S c h e u e r m a n n 56 References Chapter 1 References 1) T. Ushiki, K. Kawai, M. C. Yu, T. Shinohara, K. Ino, M. Morita, T. Ohmi, IEEE Trans. Electron Devices 45, 2349 ( 1998 ) 2) Jeong Soo Lee, Daewon Ha, Yang Kyu Choi, Tsu Jae King, Jeffrey Bokor, IEEE Electr onic Device Letters 24, 741 43 (2003). 3) Hideki Takeuchi Min She, Member, Kozo Watanabe, and Tsu Jae King IEEE Transactions on Semiconductor Manufacturing 18, 894 99 (2005). 4) M. Ye Z. Liu, P. Ding, S. Hung, K. Ahmed, J. Vac. Sci. Technol. B24 5 ( 2006). 5) < > 6) M. Hatzakis, B.J. Canavello J.M. Shaw. IBM Journal of Research and Development. 24, 452 60 (1980). 7) P. Carlberg M Graczyk, E. L. Sarwe, I. Maximov, M. Beck, L. Montelius, Microelectronic Engineering 67 203 207 (2003). Chapter 2 References 1) A. Scheuermann A. Gerger, B. Gila, C. Abernathy. Based HEMTs a nd MESFETs S.P. Harn Museum of Art and Florida Museum of Natural History, Jan 11, 2009. 2) A.D. Stewart, A G. Scheuermann A.P. Gerger, B.P. Gila, C.R Abernathy, S.J. Pearton zation MRS Conference Proceedings Fall 2008. 3) A. M. Herrero, B. P. Gila, A. Gerger, A. Scheuermann R. Davies, C.R. Abernathy, S.J. Pearton. Ba Journal of Applied Physics 106 074105 (2009). 4 ) S. T. Pantalides, Deep Centers in Semiconductors ~Gordon and Breach, Philadelphia, PA, 1992. 5 ) R Singh, R. J. Molnar, M. S. U nlu, T. D. Moustakas, Appl. Phys. Lett 64 336 (1 994). 6) X. Zhang, P. Kung, M. Razeghi, Mat. Res. Soc. Symp. Proc. 395 625 (1995). 7) T. Suski, P. Perlin, H. Tesseyre, M. Leszczynski, I. Grzegory, J. Jun, M. Bockowski, S. Porowski, T. Moustakas, Appl. Phys. Lett. 67 2188 (1995). 8) D. M. Hofmann, D. Kovalev, G. Steude, B. K. Meyer, A. Hoffmann, L. Eckey, R. Heintz, T. Detchprom, H. Amano, I. Akasaki, Phys. Rev. B 52 16702 (1995).


S c h e u e r m a n n 57 9) S. Chichibu, T. Azuhata, T. Sota, S. Nakamura, J. Appl. Phys 79 2784 (1996). 10) G. D. Chen M. Smith, J. Y. Lin, H. X. Jiang, A. Salvador, B. N. Sverdlov, A. Botchkarv, H. Morkoc, J. Appl. Phys. 79 2675 (1996). 11) W. Grieshaber, E.F. Schubert, I.D. Goepfert, R.F. Karlicek, M.J. Schurman, C. Tran. J. Appl. Phys. 80 4615 4620 (1996). 13 ) J. L. Lyons, A. Janotti, and C. G. Van de Walle, Appl. Phys. Lett 97 152108 (2010). 14) S.J. Pearton et. al. Mat. Sci. & Eng. B 82 227 231 (2001). 15) G.S. Shekhawat, V.P. Dravid. Science 310, 89 92 (2005). Chapter 3 References 1) C. Joachim, J.K. Gimzewski, A. Aviram, Nature 408 J41 (2000). 2) S. Ami, M. Hliwa, C. Joachim, Nanotechnology 14 283 (2003). 3) B. Ma, S. Gao, Z. Wang, C. Liao, C. Yan, G. Xu. J of Chem Cryst. 29, 793 (1993). 4) R.L. Prasad, A. Kushwaha, B.P. Gautam, Coord. Chem. Rev., 62, 2983 (2009). 5) F.J. Giessibl Appl. Phys. Lett 76 1470 1472 (2000) 6) Y. Kasahara, Y. Hoshino, K. Shimizu, G. P. Sato, Chem. Letters 1990 381 (1990). 7) D.V. Soldatov G.D. Enright, J.A. Ripme ester Supramol. Chem. 11 35 47 (1999) 8) D.V. Soldatov, A.T. Henegouwen, G.D. Enright, C.I. Ratcliffe, J.A. Ripmeester. Inorganic Chemistry 40 1626 (2001). 9) H. Suzuki, S. Matsumura, Y. Satoh, K. Sogoh, H. Yasuda React. Funct. Polym. 59 253 266 (200 4). 10) H. Koyama, Y. Saito Chem. Soc. Japan 27 112 114 (1954). 11) V. Amendola, et al. Chem. Eur. J 5 3679 3688 (1999). 12) J.R. Bradbury, J.L. Hampton, D.P. Martone, A.W. Maverick Inorg. Chem 28 2392 2399 (1989). 13) M.S. Owen A.T. Brooker A Guide to Modern Inorganic Chemistry (Longman Scientific & Technical, Harlow, 4th edition, 1995). 14) J. Repp G. Meyer, S.M. Stojkovic A. Gourdon, C. Joachim Phys. Rev. Lett 94 026803 (2005).


S c h e u e r m a n n 58 15) C. Villagomez, et al. Chem. Phys. Lett 450 107 11 1 (2007). 16) T. Leoni O. Guillermet, H. Walch, V. Langlais, A. Scheuermann, J. Bonvoisin, S. Gautheir. Controlling the charge state of a single redox molecular switch. Pending publication (2011). 17) A. Carella, C. Coudret, G. Guirado, G. Rapenne, G. V ives, J.P. Launay. Dalton Trans 177 186 (2006). Chapter 4 References 1) R.A. Guidotti, F.W. Reinhardt, A.H. Andazola. Blending study of MgO based separator materials for ther mal Batteries. Technical Report, SAND2002 1458. 2) R.A. Guidotti, F.W. Rei nhardt, E.V. Thomas Deformation study of separator pellets Technical Report SAND 90 2318. 3) E.L. Corral A. Ayala, R.E. Loehman, R Shah, M. Relterer, D. Bencoe Ta pe casting of magnesium oxide. Technical Report SAND2008 0528. 4) R.A. Guidotti, F.W. R einhardt Evaluation of ceramic papers and tapes for use as separators. Technical Report, SAND2006 2244.