A Complete Interfacial System Solution for Liquid Metal Electronics

A Complete Interfacial System Solution for Liquid

Metal Electronics

A dissertation submitted to the Graduate School of the University of Cincinnati in partial fulfillment of the requirements for the degree of

Doctor of Philosophy

in the Department of Materials Science & Engineering of the College of Engineering & Applied Science by

Sarah E. Holcomb

B.S., Rensselaer Polytechnic Institute, 2013

Committee Chair: Jason C. Heikenfeld, Ph.D.

Abstract

Liquid metal electronic devices have numerous advantages over traditional solid devices such as the ability to be flexed and stretched or reconfigured. Examples of such devices are wires, switches, polarizers, and antennas. Previously, mercury has been used as the room temperature liquid metal of choice but has been recently replaced by gallium liquid metal alloys

(GaLMAs) which are non-toxic, have extremely low vapor pressures, and can remain liquid at temperatures as low as -19°C. A key difference in the performance of GaLMAs vs. mercury is the mechanically stabilizing, passivating oxide which forms instantly on the surface of GaLMAs in as little as 1 ppm oxygen environments. This oxide presents a significant challenge for reconfigurable device applications because it “sticks” to most surfaces, preventing reversible shape change, which alters the desired electrical performance. Proposed here are two novel methods of overcoming this challenge. These methods enable new capabilities for reconfigurable electronic devices.

The first approach involves removing the oxide in situ as it is continuously formed in all practically achievable device environments. Oxide removal is commonly done through the use of hydrochloric acid (aqueous or vapor), which reacts with the gallium oxide to produce gallium chloride, which is not mechanically stabilizing, and water. The water produced from this reaction can have detrimental effects for interface-sensitive methods of actuation, one example is electrowetting. Furthermore, surrounding the GaLMA in a conductive environment diminishes the performance capabilities of the device. Combining HCl with an insulating and hygroscopic environment, such as silicone oil, allows for the GaLMA to retain its fluidic properties of shape change without sticking while not interfering with the electrical performance of the device.

Importantly, the water produced at the interface diffuses into the bulk of the oil solution where it has a negligible effect on the system.

The second approach utilizes the oxide to modify and tune the surface properties of the liquid metal. Phosphonic acids (PAs) bind strongly to metal oxide to form monolayers and have

been used on transparent metal oxides to tune surface properties for applications such as organic solar cells. More recently, this approach has also been applied to prevent sticking or alloying of GaLMAs to other metals. Because there are not competing reactions, controlling the

GaLMA interface by modifying the native oxide yields a more stable system than removing the oxide.

Acknowledgements

First and foremost, I would like to express my sincerest gratitude to my advisors, Dr. Jason

Heikenfeld and Dr. Christopher Tabor, for their invaluable advice and guidance on both academic and personal levels. Without their encouragement and knowledge to support me throughout this thesis, my

Ph.D. would have never been as productive and efficient. I would also like to thank my doctoral dissertation committee, Dr. Ashley Paz y Puente, Dr. Dale Schaefer and Dr. Je-Hyeong Bahk, for their time and their inspiring suggestions.

I owe my deepest appreciation to Dr. David Mast and Dr. Michael Dickey (NCSU) for technical discussions and sharing with me their expertise. I would also like to give special thanks to Dr. Michael

Brothers (AFRL) for his precious help with chemical characterization and Dr. Alex Cook (AFRL) for help with electrical measurements and automating test setups. I am heartily grateful to Aaron Diebold for his collaborations in electrowetting experiments and device design/testing. I would like to extend my gratitude to the staffs of the Department of Electronics and Computing Systems and of the Department of

Mechanical and Materials Engineering, especially Tony Seta, for all their help.

I wish to thank my colleagues of the Novel Devices Laboratory and Air Force Research

Laboratory for the memorable discussions and all the help. This journey would not have been as fun and enjoyable without all of you.

I would like to express a special thanks to my husband, Sean, for his continued and unfailing love, support and understanding during the pursuit of my Ph.D. You were always around at times I thought that it is impossible to continue and you helped me to keep things in perspective. Finally, I would like to dedicate this work to my parents who have inspired me in life as well as in pursuing higher education. I greatly value their contribution and deeply appreciate their belief in me.

Table of Contents

Abstract ...... ii Copyright Page ...... iv Acknowledgements ...... v Table of Contents ...... vi List of Figures ...... viii Chapter 1: Introduction ...... 1 1.1 Introduction ...... 1 1.2 Research Aims and Outline ...... 2 Chapter 2: Literature Review ...... 5 2.1 Introduction ...... 5 2.2 Key Challenges ...... 6 2.3 Gallium Liquid Metal Alloys ...... 7 2.4 Oxide Characteristics ...... 8 2.4.1 Mechanical Properties ...... 8 2.4.2 Electrical Properties ...... 11 2.4.3 Structure ...... 11 2.5 Oxide Removal Methods ...... 14 2.5.1 Chemical ...... 14 2.5.2 Electrochemical ...... 16 2.6 Electrical Contacts ...... 18 2.6.1 Importance of Reliable Electrical Contacts ...... 18 2.6.2 Chemical Interactions ...... 18 2.6.3 Physical Interactions ...... 20 2.6.4 Electrical Interactions ...... 22 2.7 Conclusions ...... 22 2.8 Objective of Research ...... 23 2.9 References ...... 24 Chapter 3: Oxide-Free Actuation of Gallium Liquid Metal Alloys Enabled by Novel Acidified Siloxane Oils ...... 27 3.1 Introduction ...... 27 3.2 Experimental Materials and Methods ...... 30 3.2.1 Synthesis of Acidic Siloxane ...... 30 3.2.2 Chemical Characterization of Siloxane...... 30 3.2.3 Dielectric Characterization of Siloxane ...... 31 3.2.4 Electrowetting of Liquid Metal ...... 32 3.3 Rationale for Choice of Oil Preparation and Characterization Methods ...... 32 3.4 Acidic Siloxane Characterization by 1H-NMR ...... 34 3.5 Identification of Chemical Species Generated Using Electrospray Ionization Mass Spectrometry (ESI-MS) ...... 36 3.6 Dielectric Characterization of the Acidic Siloxane ...... 38 3.7 Electrowetting Demonstrations of GaLMA ...... 40

3.8 Conclusions ...... 44 3.9 Supplemental ...... 44 3.10 References ...... 48 Chapter 4: Electrowetting-Actuated Liquid Metal for RF Applications ...... 51 4.1 Introduction ...... 51 4.2 Materials and Fabrication of Test Devices ...... 53 4.3.1 Electrowetting Characterization ...... 53 4.3.2 Polarizer Device Demonstration ...... 54 4.3 Theory / Liquid Metal Electrowetting Device Design ...... 57 4.4 Experimental Results ...... 59 4.5 Discussion ...... 67 4.6 Conclusions ...... 69 4.7 Supplemental ...... 69 4.8 References ...... 71 Chapter 5: Reversible Electronic Contacts between EGaIn Liquid Metal and Solid Copper Electrodes ...... 75 5.1 Introduction ...... 75 5.2 Experimental Materials and Methods ...... 77 5.2.1 Materials and Preparation ...... 77 5.2.2 Surface Characterization ...... 78 5.2.3 Contact Cycle Testing...... 78 5.2.4 Electrical Measurements ...... 79 5.3 Results and Discussion ...... 79 5.4 Conclusions ...... 88 5.5 Supplemental ...... 89 5.6 References ...... 92 Chapter 6: Conclusions and Future Outlook ...... 96 6.1 Reconfigurable GaLMA Contacts with Dielectric Materials ...... 96 6.1.1 Results Achieved ...... 96 6.1.2 Future Outlook ...... 97 6.2 Reconfigurable GaLMA Contacts with Solid Metals ...... 98 6.2.1 Results Achieved ...... 98 6.2.2 Future Outlook ...... 99 6.3 Overall Outlook for the Future ...... 100 6.4 References ...... 100

List of Figures

Figure 2.1: Self-healing stretchable wires. Figure 2.2: Switching of resistive networks using vacuum pressure to actuate GaLMA. Figure 2.3: Direct writing of liquid metal 3D structures. Figure 2.4: Visco-elastic rheological plots of EGaIn. Figure 2.5: Contact angles of EGaIn on various surfaces. Figure 2.6: Ga-O phase diagram. Figure 2.7: Proposed atomic arrangement in the oxide layer and corresponding electron density profile. Figure 2.8: Reaction of GaLMA surface oxide with HCl vapor. Figure 2.9: Controlling GaLMA movement using applied potential in a confined channel. Figure 2.10: Graphene barrier layer for GaLMA electrical contacts. Figure 2.11: Schematic illustration of oxide-surface interfaces.

Figure 3.1: Acidic oil removing GaLMA oxide. Figure 3.2: 1H-NMR results. Figure 3.3: ESI-MS results. Figure 3.4: Dielectric characterization and breakdown voltage of various concentrations of acidic oil. Figure 3.5: Electrowetting contact angle of GaLMA in varying concentrations of acidic oil. Figure 3.6: Demonstration of a switchable wire grid polarizer device. Figure 3.7: 1H-NMR supplemental results. Figure 3.8: 1H-NMR supplemental results. Figure 3.9: FTIR supplemental results. Figure 3.10: ESI-MS supplemental results.

Figure 4.1: Device diagrams and relevant dimensions for switchable polarizer. Figure 4.2: Wetting and dewetting speeds in polarizer. Figure 4.3: Signal transmission for various electrode materials including fluorine doped tin oxide and indium tin oxide. Figure 4.4: Transmission results for device on and off, with the liquid metal wires perpendicular and parallel to the electric field polarization. Figure 4.5: Contact angle change of EGaIn vs. AC voltage using a 1 kHz square wave in acidic oil. Figure 4.6: Photos of operating device. Figure 4.7: Water formation on EGaIn droplet.

Figure 5.1: Mode of contact. Figure 5.2: Effect of surrounding solvents on first contact. Figure 5.3: Mitigation strategies for reducing damage of multiple contacts. Figure 5.4: Effect of different contact modes on copper/EGaIn interface. Figure 5.5: Test setup and effect of voltage. Figure 5.6: BSE vs SEI images showing footprint of contact. Figure 5.7: Full survey spectra XPS of both touch and expand contacts in DPA ethanol. Figure 5.8: EDS of damaged area after 1000 cycles with continuously applied voltage.

Chapter 1: Introduction

1.1 Introduction

Room-temperature liquid metals have seen an explosion of recent reports on self- healing, reconfigurable, and stretchable electromagnetic devices such as wires, switches, antennas, polarizers, light valves, diffraction gratings, and filters. Fundamentally, liquid metals can provide a combination of electrical conductivity and analog reconfigurability far exceeding what can be achieved with semiconductor, micro-electro-mechanical systems (MEMS), or other approaches. The most promising alternatives to mercury are gallium liquid metal alloys

(GaLMAs), which can remain a liquid down to -19° C. However, GaLMAs present a significant challenge because a viscoelastic oxide skin rapidly forms in environments with ppm levels of oxygen. This oxide then prevents reversible re-shaping; however it also has certain advantages for enabling the liquid to take on non-equilibrium shapes and as a barrier to alloying.

To enable device demonstrations, researchers have removed this oxide by use of acidic vapor or electrochemical reduction in electrolyte solution. In these cases, the actuation of the

GaLMA shape has been limited to pre-determined geometries imparted by microfluidic confinement (channels). However, from an electronic or electromagnetic perspective, neither acid nor electrochemical approaches are ideal as the required acid or electrolyte solutions are inherently ionically conductive (electrically lossy). Furthermore, surrounding a GaLMA droplet with a conductive solution then makes it impossible to use most reconfigurable electrostatic methods of control such as electrowetting. There is hence an unmet need to find a materials system that can both eliminate the GaLMA surface oxide while simultaneously providing a geometrically open and electrically insulating environment. With such an advance, GaLMAs could achieve previously unseen levels of reconfigurability, enabling new and more sophisticated electromagnetic effects such as multi-way switches, fast-switching optical effects such as diffraction or polarization, or even writeable/erasable meta-materials.

While removal of the oxide surface is necessary for many applications, there are other instances where the oxide is beneficial. In cases where the GaLMA needs to be electrically connected to other devices, for example, the oxide acts as a barrier to alloying. The gallium metal aggressively alloys with most other metals, which leads to failure from either damage of the solid electrode metal or phase change of the gallium alloy. Often this problem is simply ignored for short term testing or else an inert barrier is placed between the metals to prevent alloying. These barrier layers can add a significant amount of contact resistance to an otherwise metallic contact and therefore counteract some of the advantages of using the GaLMA. Finding another method to protect unwanted alloying or damage while minimizing the contact resistance would enable reconfigurable and robust electrical contacts with greater performance and reliability than has been shown previously.

1.2 Research Aims and Outline

The central hypothesis of this work is that GaLMAs and an adjacent solid-phase material can provide robust and reversible electronic contacts that enable a wide-array of reconfigurable electronic devices. Such devices rely on four distinct types of contacts: (1) permanent contacts with dielectrics, (2) reversible contact with dielectrics, (3) permanent contact with conductors, and (4) reversible contact with conductors. The following research aims focus on investigating novel strategies that deal with reversible contacts, which are more difficult than permanent contacts due to the rigid oxide skin. Completion of these aims addresses the difficult problems of device reconfiguration without compromise of electronic performance as well as interfaces that do not degrade over time (e.g. sticking, continuous alloy formation). Since contacting

GaLMAs with dielectrics vs. conductors present different challenges topics are examined separately in the two aims of this thesis.

The first research aim of this dissertation is to create a novel acidic oil for reversible contacts with dielectrics and non-alloying, corrosion resistant conductors and demonstrate a reversibly switchable device using electrowetting actuation. This aim addresses reversible contact with dielectric materials through removal of the oxide and byproducts from the interface in an electrically insulating environment. While work has been done in removing the oxide using acid vapor or fluids to enable reversible shape change and prevent sticking of the oxide to dielectric surfaces these properties could only be achieved in an ionically conductive environment. This approach is detrimental in applications such as electrowetting where any conductive phase at the GaLMA interface would prevent the GaLMA from electrowetting. A conductive environment is also detrimental to other electrical devices (e.g. switches, antennas).

Chapters 3 and 4 support this 1st aim. This work is based on published manuscripts that discuss the characterization of the novel acidic siloxane created, its effects on the GaLMA surfaces, and demonstrate its use in an electrowetting device application. The importance of this aim is that it allows oxide removal in electrically insulating environments as well as absorption of detrimental byproducts such as water from the surface.

The second research aim of this dissertation is to explore phosphonic acid monolayers to enable reversible electrical contacts with electrical conductors and demonstrate reversible and stable electrical contact between GaLMAs and electrodes. This aim addresses reversible contacts with conductive materials through use of barrier layers that have minimal impact on the contact resistance of the GaLMA to another metal electrode. The rationale behind this aim is that GaLMA to solid metal contacts offer minimal contact resistance; however GaLMAs react with most other metals to and thus stable contacts require some sort of barrier layer that can significantly increase the contact resistance. Chapter 5 supports this 2nd aim and is based on a manuscript in preparation for publication that discusses the effect of several variables (i.e. contact mode, environmental fluid, and phosphonic acid surface modification) on the failure of reversible contacts between GaLMAs and copper over 1-1000 cycles. This work exploits a new

method to closely evaluate reversible GaLMA- electrode contacts and identifies ways to overcome failure and enable high cycles of reliable contacts.

Chapter 2: Literature Review

2.1 Introduction

Flexible, stretchable, and reconfigurable electronic devices are of great interest.

Advantages over traditional rigid and static devices are the ability to conform to changing shapes and display tunable functionalities. Some examples of electronic devices that can benefit from innovations in stretchable or reconfigurable materials are wires, switches, antennas, polarizers, light valves, diffraction gratings, and filters1–4. The key materials enabling recent investigations of such devices are gallium liquid metal alloys (GaLMAs). Similar to mercury, GaLMAs can have melting points below room temperature and conductivities similar to other metals. GaLMAs are, however, non-toxic and have extremely low vapor pressures which make them safer alternatives to mercury.

Figure 2.1 Self-healing stretchable wires showing how a straight section can be cut and put back

together in different configurations yet still connect a circuit to light an LED5.

GaLMAs can be used as interconnects (e.g. wires) or as the functional component (e.g. antennas) of electronic devices. An example of a GaLMA wire that is encased in a self-healing polymer is shown in Figure 2.15. The advantages of using the liquid metal over a traditional

solid wire material are that it can stretch or bend as much as the material it is encased in. An example of a functional component is shown in Figure 2.2 where a resistive network can be switched between an on and an off state simply by applying vacuum pressure6. Numerous additional reports have been published showing potential applications of GaLMA electronic devices, but often their functionality and lifetime are severely limited by critical materials issues involving the interfacial interactions.

Figure 2.2 Switching of (a),(b) resistive networks using vacuum pressure to actuate the GaLMA6.

2.2 Key Challenges

GaLMAs spontaneously and rapidly form a surface oxide in environments with as little as

1 ppm oxygen. Due to the critical role of the oxide on the interactions at GaLMA interfaces it is important to understand its structure and properties. Herein lies the first challenge: the oxidized surface is very fragile and hard to characterize. This issue has led to conflicting reports on properties of the oxide itself as well as on other surface-dependent properties of GaLMAs. The existence of the oxide has largely been deemed a hindrance to applications and oftentimes a chemical reaction is used to remove it in order to allow device demonstrations. However, oxide-

elimination reactions have not been sustainable in a closed system. The oxide enables many unique functions of the liquid metal. If the oxide can be controlled rather than eliminated it can be used to tune the interfacial properties. Understanding and controlling the interface will lead to advances in overcoming a major obstacle thus far in GaLMA electronic devices: making robust connections between the liquid metal and the outside electrical world.

This review will first introduce and compare the different GaLMAs studied in literature and discuss the characteristics of the oxide. These characteristics include the mechanical and electrical properties as well as the structure. Next, the methods and limitations of removing the oxide will be addressed. Chemical, physical, and electrical influences on connections between

GaLMAs and other materials will then be discussed. Finally, the future work needed to advance the area of GaLMA electronics devices will be proposed.

2.3 Gallium Liquid Metal Alloys

Gallium is an elemental metal with a melting point of 29.77°C, which is just above room temperature. When alloyed with certain other elements the melting point can be depressed much lower7. Two such alloys are used extensively in studies of liquid metal devices. First is the eutectic composition of the gallium indium system which is termed EGaIn (75% Ga, 25% In by weight). Second is a commercially sold ternary alloy of gallium, indium, and tin which is called Galinstan (68.5% Ga, 21.5% In, and 10% Sn by weight). Both alloys have similar bulk properties including very low vapor pressures, electrical resistivities comparable to platinum, and viscosities similar to water7,8. The largest discrepancy between these alloys is in reported melting temperatures of approximately 15°C for EGaIn and -19°C for Galinstan. However, gallium has a strong tendency to undercool so both alloys can exist as a liquid well below the freezing point. Furthermore, the surface chemistry of both alloys is dominated by gallium, the most abundant component9. In light of these strong similarities, no further distinction will be

made between the two alloys or between alloys that behave similarly. The following review will refer to both simply as gallium liquid metal alloys (GaLMAs).

2.4 Oxide Characteristics

2.4.1 Mechanical Properties

In the presence of greater than 1 ppm oxygen, GaLMAs instantaneously form a thin, solid “skin” composed of gallium oxides10. The oxidized surface exhibits drastically different properties than the bulk liquid. A unique effect of the oxide skin is that it allows the formation of mechanically stable non-equilibrium shapes of the liquid (Figure 2.311). Parallel plate rheology experiments determined that the oxide is viscoelastic with an elastic modulus of ~10 N/m and yield stress of ~0.5 N/m in ambient environments (Figure 2.48).

Figure 2.1 Direct writing of liquid metal 3D structures. Photographs of the diverse free standing, liquid metal microstructures that can be direct printed at room temperature. (a) Liquid metal ejected rapidly from

a glass capillary forms a thin wire. (b) These fibers are strong enough to suspend over a gap despite being composed of liquid. (c) A free standing liquid metal arch. (d) A tower of liquid metal droplets. (e) A

3D cubic array of stacked droplets. (f) A metal wire and an arch composed of liquid metal droplets. (g)

An array of in-plane lines of free standing liquid metal fabricated by filling a microchannel with the metal

and dissolving away the mold. Scale bars represent 500 µm11.

These values did not change with longer exposure time to air, giving evidence that the oxide is passivating. The result is consistent with both pendant drop surface tension measurements (0.63 N/m) and the stress needed to force the GaLMA through a microfluidic channel (~0.6 N/m)8. The parallel plate method used to determine the mechanical properties of the oxide skin provided reasonable evidence (by control experiments having different sample volumes) that the properties of the bulk liquid were negligible. Therefore the observed results were due only to the nature of the surface.

Figure 2.2 Visco-elastic rheological plots of EGaIn. a) A plot of the surface shear stress (σs, N/m), the surface elastic modulus (G’s, N/m), and the surface viscous modulus (G”s, N/m) vs. strain amplitude (γ0) for EGaIn. At low values of strain amplitude (<0.01), EGaIn has elastic like behavior (G’s>>G”s). Notably,

the stress did not increase significantly beyond a strain of 0.1; this result suggests that the EGaIn flows readily once it yields. b) A plot of G’s and G”s vs. stress for EGaIn. The data shows that the EGaIn flows

readily beyond a critical surface stress (~0.5 N/m)8.

Figure 2.3 Contact angles ( θc) of EGaIn (eutectic GaIn) on various surfaces including a thermally- evaporated flat Au film, stretchable Au nanosheet (NS) film, stretchable Ag nanowire (NW) film, a PDMS

substrate, and a PEDOT:PSS film13. Inset is the blown up image of GaLMA drop on Ag nanowire. The

drop appears to have a rough oxidized surface in the reportedly “inert” environment.

Although the mechanical stability of non-equilibrium shapes of oxidized GaLMAs is well accepted, contact angles are still being reported measured in oxidizing environments. Liu reports the most systematic study of these properties in controlled environments. Liu found that the oxide stabilized irreversible distortion of the GaLMA drop in environments of >1 ppm oxygen, which means an arbitrary contact angle could be observed not related to the surface energy10. Regardless of the clear evidence to the contrary, many papers report and assign significance to contact angle measurements done in oxidizing environments. In some cases, the implications of the oxide are simply ignored and contact angles in ambient environments are discussed12. In other cases, attempts are made to prevent oxidation but clearly fall short. An example is the reporting of contact angles in 10 ppm oxygen by Jang which was said to be free of oxide13. Not only does this work contradict the more detailed study by Liu, but oxidation of

Jang’s samples can be observed by the rough appearance of the liquid metal shown in Figure

2.5. By contrast, non-oxidized GaLMAs appear smooth and shiny. The majority of reported

contact angles are essentially meaningless because the existence and viscoelastic effect of the oxide are often not considered.

2.4.2 Electrical Properties

The bulk of GaLMAs have metallic conductivity, however the gallium oxide at the surface is more resistive. The resistivity of gallium oxide (Ga2O3), a common semiconductor material, can range between 1 and 1013 Ω∙cm and strongly depends on the conditions of formation14, therefore the oxide formed on the GaLMA surface can vary greatly from other forms. Variable resistivity is caused by the disorder in the crystal lattice15. To determine resistivity of the surface oxide it is best to try to measure the native oxide skin on the liquid gallium directly. Nijhuis used two copper wire electrodes to compare the charge transport in three different configurations: 1) both wires in contact with bulk GaLMA, 2) one wire in contact with the bulk and the other resting on the oxide surface, and 3) both wires resting on the oxide surface. The measured resistances for each case were 0.0006, 0.04, and 0.7 Ω∙cm-2, respectively16. This test was only able to estimate the relative resistances instead of directly measure the resistivity because adventitious contamination and contact resistances cannot be eliminated from the experiment. The relatively low resistance of the native oxide results from poor crystallization17.

2.4.3 Structure

In the gallium-oxygen system three oxides have been reported to exist: Ga2O3, GaO, and Ga2O. Ga2O3 is the only form which is thermodynamically stable in the solid state, as shown in the Ga-O phase diagram in Figure 2.6. Ga2O could only be produced using the reaction

Ga2O3 + 4Ga = 3Ga2O, yielding an amorphous brown-colored substance. Ga2O3 has five polymorphs; however all but one are metastable and so transform to the β form at sufficiently

high temperatures. The structure of β-Ga2O3 is a slightly deformed cubic close-packed oxygen sublattice with gallium ions uniformly distributed over octahedral and tetrahedral sites15.

Figure 2.4 Ga-O phase diagram15.

The structure and composition for the thermodynamically stable β-Ga2O3 has been characterized; however the film thickness, interfacial roughness, and atomic structure of the native passivating oxide layer are difficult to ascertain due to the difficulty in applying x-ray and neutron scattering to a liquid-metal surface. Regan reports x-ray reflectivity as well as grazing incidence x-ray scattering (GIXS) measurements of gallium oxide as grown on the liquid metal17.

Reflectivity vs. wave vector qz was compared for bare gallium and gallium exposed to 206 L (for oxygen, 1 Langmuir (L) = 1x10-7 Torr for 24 seconds), showing oscillations appear in the latter.

These so called Kiessig fringes are indicative of a surface layer of uniform thickness, in this case oxide of approximately 5 Å. This is in agreement with the SIMS measurement estimating a

few monolayers thick oxide18. The reflectivity data also showed that the thickness and uniformity of the oxide layer remained unchanged over the mm2 area in which data was collected, as evidenced by the minima and maxima of qz having fixed positions across several samples and the two maxima having consistent amplitude within each sample, respectively. Figure 2.7 shows the proposed atomic arrangement of the oxide and its agreement with the average electron density along the surface normal, which is related to the measured reflectivity. The thin solid line is the oxidized gallium and the thin dashed line is the bare gallium. It is proposed that the top layer of liquid gallium surface contains Ga ions that bond with the first layer of O2- ions in

3+ 2- the Ga2O3 structure, and the oxide layer consists of alternating layers of Ga and O ions, terminated by an O2- layer. The Ga ions at the liquid-oxide interface form tetrahedral with the first plane of close-packed O2- ions, the next layer consists of Ga3+ ions in octahedral sites, and additional layers of O2- and Ga3+ continue the film. Fits to density models indicate the average oxide layer density is lower than that for pure gallium, which is inconsistent with the electron density models of known crystal forms of Ga2O3. This result showed that only ~76% of the liquid surface was covered by the oxide; therefore bare and oxidized regions are simultaneously present on length scales comparable to the x-ray coherence length of 1000-3000 Å. GIXS measurements were taken on the oxide but no sharp in-plane peaks were observed, showing that the structure of the oxide is amorphous or poorly crystallized. Disorder is attributed to the film growing from a disordered liquid substrate.

Figure 2.5 Proposed atomic arrangement in the oxide layer and corresponding electron density profile

(thin solid line) with its separate components of underlying liquid Ga (large dash line), and Gaussians

representing the Ga3+ (dash-dotted line) and O2- (short-dashed line) layers17.

2.5 Oxide Removal Methods

2.5.1 Chemical

The unique mechanical stability imparted on GaLMAs has largely been treated as a nuisance since it prevents true liquid behavior and is not very well understood. A common method in literature to circumvent this issue has been the use of hydrochloric acid to chemically react and remove the surface oxide19,20. The gallium oxides are replaced by gallium chlorides and/or indium chlorides and water (Figure 2.821), which do not mechanically stabilize the surface as the oxide does9. Furthermore the gallium and indium chlorides are deliquescent, absorbing additional water from the environment and then dissolving in the water. The key reactions are given by equations (1) and (2)9.

−1 (1) 퐺푎2푂3(푠) + 6퐻퐶푙(푔) → 2퐺푎퐶푙3(푠) + 3퐻2푂(푙) Δ퐺0 = −50 푘퐽 푚표푙

−1 (2) 4퐺푎(푙) + 3푂2 → 2퐺푎2푂3(푠) Δ퐺0 = −1002 푘퐽 푚표푙

Figure 2.6 (a) Schematic diagram of chemical reaction with HCl vapor. The phase of oxides

(Ga2O3/Ga2O), chlorides (GaCl3/InCl3), and water is solid, aqueous, and liquid, respectively. (b) Optical

image of the surface-modified Galinstan droplet on a Teflon-coated glass with an illumination21.

The change in Gibbs free energy is negative for both reactions, showing that they occur spontaneously. The surface composition will favor GaCl3 as long as the rate of consumption of

9 Ga2O3 in (1) is greater than the rate of formation of Ga2O3 in (2) . The rate of consumption of the gallium oxide depends on the concentration of HCl. Using concentrated acid in a device nullifies one of the primary advantages, greater safety, gained from using GaLMAs instead of mercury, clearly identifying a major problem of this solution. Some tricks have been found to reduce the hazard while still utilizing the oxide removing abilities. One such trick is confining the HCl to an ion-exchange membrane or sealed in a closed system9,19. It is obvious from the much more negative value of Gibbs free energy for oxide formation that as long as oxygen is present and/or can enter the system appreciably, formation of oxide on the GaLMA surface is inevitable.

Although HCl oxide removal has successfully been used to enable demonstrations of GaLMA technologies which rely on well-established principles of microfluidics, it is clearly not a viable

option for long term applications, especially since the byproduct, water, is electrically conductive.

The chemical reactions involving HCl to remove the oxide are extensively covered in

GaLMA literature, but there is much less discussion regarding other acids or bases.

Electrolytes with either high or low pH seem to work similarly to remove the oxide and likely form gallium hydroxide instead of gallium chloride or halide22,23. Perhaps the specific oxide removal mechanisms of these materials are not reported in GaLMA literature because they are used exclusively within an aqueous environment and no noteworthy products are formed on the surface. Any chemical reaction will inherently have the same limitation discussed above that the reactant for removing the oxide will be depleted and the thermodynamically favored oxide will inevitably return.

2.5.2 Electrochemical

Although affected by the same limitation as for chemical oxide removal, electrochemical reactions have far greater potential advantages because of their ability to tune the surface properties by controlling the oxide. Electrical control of the surface oxidation can be achieved by applying low voltages (approximately -1 V relative to a saturated Ag–AgCl reference electrode) in certain electrolyte solution (such as 1 mM – 1 M NaF), causing the oxide to either be formed or removed by choosing oxidative or reductive potentials23. The mechanism is not fully understood, but is believed to be caused by reduction of gallium oxide to gallium and water.

Competing reactions of chemical dissolution (or formation) of oxide in the electrolyte and electrochemical oxidation (or reduction) by the applied voltage enable reversible oxidation.

Through studying electrochemical removal of oxides, a variety of intermediate species were produced that provided more insight into how tunable the gallium oxide can become. For example, the oxide layer acts as a surfactant by replacing the high-energy metal-electrolyte

interface with two new interfaces of metal-metal oxide and metal oxide-electrolyte. Metal oxides tend to form hydroxyl groups on their exterior surfaces (which makes them hydrophilic) leading to low interfacial energy with aqueous solutions, and the interior surface of the Ga2O3 most likely has a low interfacial energy with the bulk liquid metal14,17. While traditional surfactants change surface tensions by ~20-50 mJ/m2, the oxide can lower the GaLMA’s surface tension from ~500 mJ/m2 to near zero. The surface tension controls the liquids ability to spread and controlling this property over such a wide range of values can be applied to manipulating the flow of GaLMAs in closed and open systems. Figure 2.9a,b shows examples where oxidative potentials are applied in conjunction with confined channels to induce oxide, thereby dramatically decreasing the surface tension, which causes spreading of the GaLMA through the channel in the direction of the applied potential. Similarly, Figure 2.9c,d shows that, by switching conditions for or against spreading, either drops or wires can be dispensed from the same source23.

Figure 2.7 (a) Inducing liquid metal into an upward-tilted capillary channel (∼0.9 mm i.d.) by application of

a voltage in the presence of 1 M NaOH. (b) Controlling the shape and direction of a metal drop into an

open T-shaped Plexiglas channel submerged in 1 M NaOH solution using only voltage. Switching the

position of the counterelectrode at different points (B, ii-iv), guides the direction of the metal droplet. (c)

Side view of a small droplet of EGaIn pumped out of a 0.5-mm-i.d. polymer tube at 20-mL/h in 1 M NaOH.

The metal forms droplets in the absence of potential. (d) Formation of an oxide-coated liquid metal fiber

coming out of the tube at 5 V23.

2.6 Electrical Contacts

2.6.1 Importance of Reliable Electrical Contacts

The ability to make reliable electrical contacts is critical for most electronic devices.

“Reliable” implies that the physical and electrical properties should not change over the lifetime of the device. Contacts between GaLMAs and solid electrode materials are of interest for many applications, for example to connect multiple rigid components to one another on a stretchable substrate or to make extremely fast switches with unlimited actuation cycles. The interfacial interactions that determine the stability of electrical contacts can be divided into three classifications: chemical, physical, and electrical.

2.6.2 Chemical Interactions

GaLMAs are very reactive with most other metals and readily alloy. Grain boundary penetration and liquid metal embrittlement are well studied phenomena24. Many literature reports on GaLMA devices mention the concern of alloying but offer little insight to a solution.

For example Cumby coated a solid metal connector with conductive carbon ink to prevent alloying during RF device measurements but did not elaborate any further as to how or what it changed19. Both Yoon25 and Jang13 reported making electrical connections with GaLMAs using silver nanowires even though GaLMAs react aggressively with Ag26. A possible explanation of not observing alloying may be that native gallium and/or silver oxides may have acted as a barrier layer for the time scales of their experiments. The native oxide layer could have an especially profound effect for the Ag nanowires because of the high surface area to volume ratio at the nanoscale.

Barrier layers may be an effective way to prevent alloying of GaLMAs with metal electrode materials, since alloying cannot occur without diffusion across the interface.

Graphene was able to prevent GaLMA attack on aluminum for at least seven months on the condition that defects were not present through the thickness of the graphene27. Defects were present by two ways: 1) defects in deposition, including cracking, pinholes, etc. that are intrinsic to any coating process. Such defects can be overcome through deposition of multiple layers of graphene to reduce the possibility defect alignment and 2) damaging of the carbon layer during the physical process of spray coat depositing the GaLMA. Even though defects allowed diffusion of gallium into the aluminum substrate, the graphene still prevented surface Al oxides from penetrating the surface carbon layer. Figure 2.10a shows that blocking penetration of the

Al oxide had a significant effect on keeping the resistance low through the bulk of the sample compared to not blocking them, and b and c illustrate how the defective graphene blocks the oxide but not the GaLMA27. An important follow-up to this study would be to test more conformal and less fragile barrier layers.

Figure 2.8 (a) Two-terminal resistance as a function of time after the spray-coating of a galinstan line

without and with a 4-layer single layer graphene (SLG) stack interlayer. After more than 3 months, the resistance became too high to measure for the sample without graphene protection, while it was 3.5 Ω for the sample with the 4-layer SLG stack. (b) Schematic illustration of the galinstan deposited on the Al thin

film without graphene interlayer by means of spray coating (left) and the subsequent dissolution of Al in the galinstan where a discontinuous and loose Al-oxide layer forms on the surface of galinstan as well as at the interface between Al and galinstan represented by the dash line (right). (c) Schematic illustration of

the galinstan deposited on the Al thin film with the interlayer of 4-layer SLG stack represented by a solid

back bold line by means of spray coating where the part of graphene under the galinstan is partially damaged, which is represented by the dashed bold line (left) and the subsequent dissolution of Al in the

galinstan where no Al-oxide layer forms at the interface between Al and galinstan (right)27.

2.6.3 Physical Interactions

In order for electrical contact to occur two conductive materials must be touching. Since electrical resistance is inversely proportional to area, better conformal contact will yield less resistance at the interface. Multi-scale roughness of solid surfaces can cause very little conformal contact area with a GaLMA drop in air as it exhibits non-wetting Cassie behavior where the liquid is suspended mostly above the surface with pockets of air underneath it12,28.

The resistance to penetration of the surface of the GaLMA drop is probably enhanced by the mechanical stability of the oxide skin. The presence of the oxide also slowed the onset of alloying with the metallic textured surface when compared to similar smooth surfaces. It is important to keep in mind the fragility of the oxide skin when considering physical interactions with solids.

The implications of the macroscopically fragile oxide surface can be understood by considering two different modes of forming an interface in the presence of oxygen (Figure

2.1129). In the first mode a fixed volume of GaLMA is dispensed from a reservoir in air the static drop is brought into contact with a solid surface. In the second mode the drop is contacted with the solid surface before being expanded to its final volume. The interface of the former can be described to have a continuous layer of oxide between the GaLMA and the solid, which is probably rough due to vibrations and movement while bringing the surfaces into contact. The latter can be described by a discontinuous layer of rough oxide, smooth oxide, and bare

GaLMA. When the drop expands in air the oxide layer repeatedly forms and fractures which, in the latter mode, allows the solid surface to be directly contacted by the bare liquid metal. An

important implication of this analysis is that the interfacial structure and chemistry of the surface depend on how a contact is made between an oxidized liquid metal drop and a solid substrate.

Figure 2.9 (a) Schematic illustration of the rough oxide−surface interface responsible for low adhesion

measured using the height variation method… (d) Schematic illustration of the smooth oxide−surface

interface formation around perimeter of drops formed using the volume addition method (a composite

interfacial area consisting of fractured pieces of old oxide, new oxide, and bare GaInSn is likely present

underneath the drop)29.

2.6.4 Electrical Interactions

The elements of a multilayer junction can be treated as a combination of resistors in series and in parallel, regardless of the mechanisms of charge transport. Cademartiri applied this approach to junctions that consist of self-assembled monolayers (SAMs) of various types of thiols (written as SR, where R is a functional group that may range in structure from simple n- alkyl groups to more complex functionalities such as aromatics or ferrocenes) supported by a

template-stripped silver electrode and contacted by a “top” EGaIn liquid electrode at room temperature and covered with a thin metal oxide film and adventitious contaminants14. The thinnest areas of these multi-layer junctions (Ag-SR//Ga2O3/EGaIn) dominate the charge transport through the junction but the most resistive element dominates the resistance of each area. Measurements of junctions without the SAM showed that the surface layer (oxide and adventitious contaminants) on the GaLMA electrode contributed the most resistance to the circuit (Ag/Ga2O3/EGaIn). By comparing that “surface” resistance to the resistance of the entire

SAM junction the influence of the electrode surface on the charge transport of the junction was inferred. It was found that the oxide surface is still significantly less resistive than other common junction materials such as SAMs or other semiconductors, and therefore electrical properties of junctions of these types are not dominated by the oxide14,16,30.

2.7 Conclusions

The interfaces of GaLMAs are unique from either other liquids or other metals. Such interfaces are very difficult to characterize, which leads to challenges in utilizing GaLMAs. The mechanical properties of the oxide enable GaLMAs to assume non-equilibrium shapes, which are not possible for other fluids. GaLMAs can be manipulated in channels by switching the non- equilibrium behavior on and off. Altering the surface of the GaLMA can have a tremendous impact on the shape and behavior of the fluid. The electrical conductivity of the bulk GaLMA is on the same order of magnitude as other metals used for electronic devices. However the oxide layer contributes the most resistance in a metal-GaLMA contact. Since the native oxide is thin and poorly crystallized it has lower resistance than most reports of gallium oxides. The native oxide does not contribute more contact resistance than other common junction materials; therefore having a layer of the oxide at an electrical contact is not necessarily undesirable.

Devices having GaLMA-solid electrode contacts have been reported on, but usually with the

focus on the short term device performance even when obvious issues such as alloying are present. To improve liquid metal electronic devices it is necessary to understand the properties and structure of GaLMA interfaces, which are dominated by the native oxide.

2.8 Objective of Research

The spontaneous oxide that forms on gallium liquid metal alloys (GaLMAs) presents a substantial challenge to studying applications and devices. This challenge is further complicated in cases when the GaLMA is contacted by another material, creating a multilayer interface.

Some literature reports that in environments of less than 1 ppm oxygen the oxide will not form, however it was found by researchers at AFRL that the oxide still forms rapidly (under one minute) in a mere 0.5 ppm oxygen environment. To further the field of liquid metal applications and devices, it is therefore necessary to develop ways to manage the oxide and control the

GaLMA interface.

Since preventing oxide growth is not feasible, there are two strategies that can be taken to manage the oxide skin that is formed: remove it or modify it. Once the surface of the GaLMA can be controlledinterfaces can be tuned and studied for specific applications (e.g. electronic interfaces involving organic materials). The most straightforward strategy is to remove the oxide, which has been done for device demonstrations using aqueous or vaporous hydrochloric acid.

This strategy creates a new problem, however, as the reaction of oxide and HCl produces water on the surface of the GaLMA. This water on the surface, in either case of being in an HCl vapor or in an aqueous environment, limits the usefulness of the GaLMA in reconfigurable electronic devices. There is a clear need to remove the oxide while maintaining an electrically insulating environment. Most notably, this allows for voltage controlled actuation of the liquid metal by electrowetting.

While removing the oxide is a straightforward approach to realizing the full advantages of reconfigurable electromagnetic devices, thermodynamics eventually reestablishes the oxide.

Functionalizing by binding molecules with highly controllable structures to the oxide itself thus offers a complementary approach to gain control over GaLMA interfaces. Phosphonic acid molecules provide a barrier layer between the GaLMA oxide and another material, e.g. an electrode, which can prevent alloying (in case with a metal) as well as control wetting behavior and work function. Research on oxide management is critical to the development of robust, controllable devices.

2.9 References

1. Fassler, A. & Majidi, C. 3D structures of liquid-phase GaIn alloy embedded in PDMS with freeze casting. Lab Chip 13, 4442–50 (2013). 2. Mohammed, M. G. & Dickey, M. D. Strain-controlled diffraction of light from stretchable liquid metal micro-components. Sensors Actuators A Phys. 193, 246–250 (2013). 3. Aïssa, B.et al. Fluidic patch antenna based on liquid metal alloy/single-wall carbon- nanotubes operating at the S-band frequency. Appl. Phys. Lett. 103, 063101 (2013). 4. Zhu, S. et al. Ultrastretchable Fibers with Metallic Conductivity Using a Liquid Metal Alloy Core. Adv. Funct. Mater. 23, 2308–2314 (2013). 5. Palleau, E., Reece, S., Desai, S. C., Smith, M. E. & Dickey, M. D. Self-healing stretchable wires for reconfigurable circuit wiring and 3D microfluidics. Adv. Mater. 25, 1589–92 (2013). 6. Cumby, B. L. et al. Reconfigurable liquid metal circuits by Laplace pressure shaping. Appl. Phys. Lett. 101, 174102 (2012). 7. Sen, P., Kim, C. C. J. & Microelectromechanical, A. Microscale Liquid-Metal Switches — A Review. IEEE Trans. Ind. Electron. 56, 1314–1330 (2009). 8. Dickey, M. D. et al. Eutectic Gallium-Indium (EGaIn): A Liquid Metal Alloy for the Formation of Stable Structures in Microchannels at Room Temperature. Adv. Funct. Mater. 18, 1097–1104 (2008). 9. Ilyas, N., Butcher, D. P., Durstock, M. F. & Tabor, C. E. Ion Exchange Membranes as an Interfacial Medium to Facilitate Gallium Liquid Metal Alloy Mobility. Adv. Mater. Interfaces (2016). 10. Liu, T., Sen, P., Kim, C. C. J. & Measurements, A. C. A. Characterization of Nontoxic Liquid-Metal Alloy Galinstan for Applications in Microdevices. J. Microelectromechanical Syst. 21, 443–450 (2012).

11. Ladd, C., So, J.-H., Muth, J. & Dickey, M. D. 3D printing of free standing liquid metal microstructures. Adv. Mater. 25, 5081–5 (2013). 12. Kramer, R. K., Boley, J. W., Stone, H. a, Weaver, J. C. & Wood, R. J. Effect of microtextured surface topography on the wetting behavior of eutectic gallium-indium alloys. Langmuir 30, 533–9 (2014). 13. Jang, J. et al. Interfacing Liquid Metals with Stretchable Metal Conductors. ACS Appl. Mater. Interfaces (2015). doi:10.1021/am508899z 14. Cademartiri, L. et al. Electrical Resistance of AgTS-S(CH2)n-1CH3//Ga2O3/EGaIn Tunneling Junctions. J. Phys. Chem. C 10848–10860 (2012). doi:10.1021/jp212501s 15. Zinkevich, M. & Aldinger, F. Thermodynamic Assessment of the Gallium-Oxygen System. J. Am. Ceram. Soc. 87, 683–691 (2001). 16. Nijhuis, C. a, Reus, W. F. & Whitesides, G. M. Molecular Rectification in Metal - SAM - Metal Oxide - Metal Junctions. 02138, 17814–17827 (2009). 17. Regan, M. J. et al. X-ray study of the oxidation of liquid-gallium surfaces. Phys. Rev. B 55, 10786–10790 (1997). 18. Chabala, J. M. Oxide-growth kinetics and fractal-like patterning across liquid gallium surfaces. Phys. Rev. B 46, (1992). 19. Tabor, C., Cumby, B., Heikenfeld, J., Mast, D. & Dickey, M. Robust Pressure-Actuated Liquid Metal Devices Showing Reconfigurable Electromagnetic Effects at GHz Frequencies (POSTPRINT). IEEE Trans. Microw. Theory Tech. 63, 3122–3130 (2015). 20. Kim, D. et al. Hydrochloric acid-impregnated paper for gallium-based liquid metal microfluidics. Sensors Actuators, B Chem. 207, 199–205 (2015). 21. Kim, D. et al. Recovery of Nonwetting Characteristics by Surface Modification of Gallium- Based Liquid Metal Droplets Using Hydrochloric Acid Vapor. ACS Appl. Mater. Interfaces 179–185 (2013). doi:10.1021/am302357t 22. Khan, M. R., Trlica, C. & Dickey, M. D. Recapillarity: Electrochemically Controlled Capillary Withdrawal of a Liquid Metal Alloy from Microchannels. Adv. Funct. Mater. 25, 671–678 (2015). 23. Khan, M. R., Eaker, C. B., Bowden, E. F. & Dickey, M. D. Giant and switchable surface activity of liquid metal via surface oxidation. Proc. Natl. Acad. Sci. 111, 14047–14051 (2014). 24. Kobayashi, M. et al. Preferential penetration path of gallium into grain boundary in practical aluminium alloy. Philos. Mag. 86, 4351–4366 (2006). 25. Yoon, J. et al. Design and fabrication of novel stretchable device arrays on a deformable polymer substrate with embedded liquid-metal interconnections. Adv. Mater. 26, 6580– 6586 (2014). 26. Mohammed, M., Sundaresan, R. & Dickey, M. D. Self-Running Liquid Metal Drops that Delaminate Metal Films at Record Velocities. ACS Appl. Mater. Interfaces 7, 23163– 23171 (2015). 27. Ahlberg, P. et al. Graphene as a Diffusion Barrier in Galinstan-Solid Metal Contacts. IEEE Trans. Electron Devices 1–5 (2014).

28. Kim, D., Lee, D. W., Choi, W. & Lee, J. B. A super-lyophobic 3-D PDMS channel as a novel microfluidic platform to manipulate oxidized galinstan. J. Microelectromechanical Syst. 22, 1267–1275 (2013). 29. Doudrick, K. et al. Different shades of oxide: from nanoscale wetting mechanisms to contact printing of gallium-based liquid metals. Langmuir 30, 6867–77 (2014). 30. Sivan, V. et al. Liquid metal marbles. Adv. Funct. Mater. 23, 144–152 (2013).

Chapter 3: Oxide-free Actuation of Gallium Liquid Metal Alloys Enabled by Novel

Acidified Siloxane Oils

This chapter includes adapted text and figures from Holcomb, S. et al. “Oxide-free actuation of gallium liquid metal alloys enabled by novel acidified siloxane oils”. Langmuir 32, 48, 12656-

3.1 Introduction

Room-temperature liquid metals have seen an explosion of recent reports on self- healing1, reconfigurable2, and stretchable electromagnetic devices such as wires, switches, antennas, polarizers, light valves, diffraction gratings, and filters3–6. Fundamentally, liquid metals can provide a combination of electrical conductivity and analog reconfigurability far exceeding what can be achieved with semiconductor, MEMs, or other approaches. Mercury has been the material of choice historically, but because of its toxicity to neuronal cells7, is now avoided in recent liquid-metal device investigations.

The most promising alternatives to mercury are clearly gallium liquid metal alloys

(GaLMAs), which can remain a liquid even down to -19° C8–10. However, GaLMAs present a significant challenge because in only a matter of seconds a viscoelastic oxide skin rapidly forms even in environments with mere ppm levels of oxygen. This oxide then prevents reversible re- shaping. To enable device demonstrations, researchers have removed this oxide by use of acidic vapor 11–13or electrochemical reduction in electrolyte solution14,15. In these cases, the actuation of the GaLMA shape has been limited to pre-determined geometries imparted by microfluidic confinement (channels). Also undesirable from an electronic or electromagnetic perspective, neither acid nor electrochemical approaches are ideal as the required acid or electrolyte solutions are inherently ionically conductive (electrically lossy). For example, consider a simple application of an all-electronic switch, where liquid metals can provide ON

state conductivity superior to semiconductor switches, but where the OFF state current would be severely limited by the ionically conductive surrounding fluid. Furthermore, surrounding a

GaLMA droplet with a conductive solution then makes it impossible to use some of the most reconfigurable electrostatic methods of control such as electrowetting18. There is hence a clear unmet need to find a materials system that can both eliminate the GaLMA surface oxide while simultaneously providing a geometrically open and electrically insulating environment. With such an advance, GaLMAs could achieve previously unseen levels of reconfigurability, enabling new and more sophisticated electromagnetic effects such as multi-way switches, fast-switching optical effects such as diffraction or polarization, or even writeable/erasable meta-materials.

Other groups have shown removal of oxides from GaLMAs using “strong” acids,

13 primarily hydrogen halides (X) . In general, 6 HX + Ga2O3  3 H2O + 2 GaX3. In many cases, the water generated from this reaction causes problems for use in electronic devices since water itself is a conductive liquid16. In water and most other protic solvents the HX will primarily or completely dissociate, which substantially increases the ionic strength of the solution and thereby increases the effective conductivity17. As noted previously, surrounding a GaLMA with a conductive solution diminishes its possible electromagnetic device performance. Furthermore, even a miniscule amount of conductive solution can prevent actuation techniques such as electrowetting18 because the conductive solution readily forms an annulus at the contact line which itself electrowets instead of the GaLMA.

Siloxanes are known to be good insulating fluids for electronics and electrowetting applications because they are electrochemically inert, chemically stable in water-free environments, and are highly electrically insulating19. However, with respect to GaLMA systems, siloxanes cannot prevent oxygen diffusion to the liquid metal surface nor can they remove an already formed oxide. Interestingly, siloxanes also naturally absorb water through hydrolysis20 yet remain electrically insulating16. However, the challenge of removing the surface oxide from the GaLMA still remains.

Presented here is the characterization and demonstration of a novel acidified siloxane that contains hydrochloric acid, which remains insulating due to the unique associative incorporation of the acids in the siloxane (i.e. HCl acting as a weak acid in the siloxane solvent system). Furthermore, we demonstrate that other acids, such as HBr produce similar results in our system. We present a straightforward and effective protocol to generate these acidic and anhydrous siloxanes, a protocol which can easily be reproduced in any laboratory setting without the need for specialized equipment. Demonstrations show that this novel siloxane is appropriate for even the most electrically challenging applications such as electrowetting

(requiring high electric fields and therefore near perfect electrical insulation). The native oxide layer, which is nearly impossible to avoid when dealing with GaLMAs, is removed in a matter of seconds by the acidified siloxane. This accomplishment is important because previous efforts required use of GaLMAs in extremely low oxygen or vacuum environments (e.g. a glove box of

<1 ppm oxygen and moisture, ultra-high vacuum of 10-9 Torr)10,21 yet often still fail due to oxide formation in matter of minutes if any trace oxygen is present. With our acidified siloxane, device assembly and rudimentary testing can be performed in a regular ambient environment, which greatly simplifies construction and operation of GaLMA devices.

An additional and critical advantage of the acidic siloxane over an acidic vapor or aqueous solution is that the water produced by the reaction of the acid and oxide is hydrolyzed

(consumed) by the siloxane, which produces silanol species20. Since there is only a small amount of oxide present to react, the siloxane degradation is minimal at the time of processing and is then rate-controlled by the diffusion of new oxygen into the system. Therefore the better the system can be sealed after assembly, the better the stability and insulating properties of the siloxane will be maintained over time. Furthermore, our experiments suggest that the acidified siloxane interacts with the liquid metal to yield gallium and indium compounds or adducts.

Synthesis and isolation of unusual compounds involving Group 13 metals has been an area of strong research interest because of their implications in organometallic synthesis, new materials

development, and in biological, medical, and environmental systems22. To our knowledge this is the first electrically insulating and chemically stable fluid capable of removing metal oxide and reaction byproducts even in the presence of atmospheric oxygen concentrations. We speculate this work is also more broadly relevant to anyone interested in acidic siloxanes for applications such as electrowetting, MEMs, coordination chemistry, electrochemistry, chemical vapor deposition (CVD) techniques, and charge transfer complexes.

3.2 Experimental Materials and Methods

3.2.1 Synthesis of acidic siloxane

In order to make acidified siloxane, 12N HCl or 12 N HBr was incubated with the siloxane (OS-20, Dow Corning) 1:1 (v/v) for 12 hours, forming an immiscible solution, with the top layer composed primarily of the siloxane and the bottom layer being the aqueous acid. This solution is allowed to equilibrate for 12-16 hours. After 12-18 hours, the water is initially removed by pipetting, and then further dried by addition of anhydrous magnesium sulfate until no clumping is observed. The subsequent solution is filtered by gravity through filter paper and stored in a sealed vial until it is used (caution, imperfect sealing can lead to 15% of HCl evaporating per week). The as-prepared siloxane contained 1.5 M HCl. More dilute samples were created as needed by adding additional OS-20 siloxane.

3.2.2 Chemical characterization of siloxane

1H NMR spectra were acquired on all OS-20 siloxanes containing both HCl and HBr at various concentrations, as well as virgin OS-20, HCl OS-20 doped with 5% DI water, OS-20 doped with 5% Silanol (Xiameter PMX-0156), and both OS-20, acidified and virgin, after exposure to the liquid metal. NMR spectra were acquired on a Bruker DMX-500 NMR spectrometer operating at proton frequency of 500.13MHz and using a 1H/13C/15N triple- resonance probe. All spectra were acquired on neat solutions, and were acquired in the same

run in order to ensure the same referencing. Fresh OS-20 was acquired during all runs to act as a consistent reference point between runs. Spectra were processed and analyzed using

Advanced Chemistry Development NMR Suite. 5 Hz line broadening was applied for all spectra reported.

Electrospray ionization mass spectrometry (ESI-MS) was performed on all samples as follows: from each one, 2 μL sample was further diluted in 100 μL MeOH. Each sample solution was then analyzed using a Q-Tof 2TM mass spectrometer (Micromass/Waters) through direct infusion at 4 μL/min flow rate. The instrument was tuned and calibrated in positive ion mode with the source de-solvation temperature set at 150°C and the cone voltage at 10 V. Presented data are a summation of one minute acquisition for each sample. Data acquisition and processing were done using MassLynx 4.0 software which is part of the Waters instruments.

3.2.3 Dielectric characterization of siloxane

Conventional cavity perturbation theory (as derived by ASTM D2520) was employed to determine the dielectric constant and loss of the pure and acidified siloxane samples at discrete frequencies ranging from approximately 2-20GHz. The combined used of three standard rectangular waveguides, WR-284, WR-90, and WR-62 (S, X, and Ku band, respectively), allowed for such a discrete spectrum to be formed. A vector network analyzer (PNA Model

N5222A, Keysight [formerly Agilent]) was then utilized to detect the shifting of resonant frequencies within the cavity upon insertion of a given sample. Finally, low frequency behavior was measured by a breakdown voltage test. A few drops of each oil sample were sandwiched between two slides of ITO coated glass separated by two thin strips of Kapton tape (3 mils thick). The slides were offset so that alligator clips could connect to them, and the ITO coated sides were placed facing each other. An AC voltage (100Hz) was applied across the oil sample and the current was measured and recorded at 50 V increments. The test was concluded when

the current spiked, indicating breakdown of the oil sample. The voltage recorded was converted into V/μm by dividing by the thickness of the oil sample, 76.2 μm.

3.2.4 Electrowetting of liquid metal

A borosilicate glass slide coated with a 2.1 thick layer of indium tin oxide was used as a substrate, on which 3 μm of Parylene C (Specialty Coating Systems, 휺풓 = ퟑ. ퟏ) was deposited, and finally a 1-10 nm layer of fluoropolymer (Flouropel, PFC 1601V, Cytonix, 1 wt. %) was deposited by dip coating. This substrate was submerged in the acidified oil and a droplet of liquid metal (eutectic gallium indium, Sigma-Aldrich) was deposited on the surface. The electrode which was probed into the liquid metal was a tungsten wire. A function generator

(Tektronix AFG320), phase inverter (FLC Electronics Model INV10) and amplifier (Trek Model

603) were used to generate a 1 kHz square wave with a voltage differential of up to 330 Vpp.

Images and analysis of the contact angle were taken and performed on a VCA Optima contact angle system.

3.3 Rationale for Choice of Oil Preparation and Characterization Methods

Acidified solutions of anhydrous solvents can be produced by bubbling gaseous acid

(including HCl) through the solvent. For example, 2.0 M hydrogen chloride in diethylether is produced and sold by Sigma Aldrich. Previous work has demonstrated that the HCl behaves as a weak acid within a diethyl ether solvent23,24. In this work we preferred an alternative protocol because vaporous acids are corrosive, highly toxic, and require special apparatus. Thus, we relied on equilibrium of the two liquid phases to generate the acidic siloxane from commercially available siloxane and aqueous acids. To prepare the acidified siloxane we combined octamethyltrisiloxane (Dow Corning, OS-20) with 12 M HCl (aq) overnight, approximately 16-18 hours, and then dried the siloxane (see method section for details).

Siloxanes (Figure 3.1) have a dipole moment similar to but weaker than that of diethyl ether17. This suggests that HCl will hydrogen bond to the oxygen in the siloxane as it has been shown to for diethyl ether. However, the decreased polarity of siloxane means that we are less likely to observe dissociation of HCl into H+ and Cl-. While some of the hydrogen chloride still dissociates into the separate H+ and Cl- salts, the majority of it should remain covalently bound as a H-Cl molecular complex. We can therefore determine the concentration of the HCl complex versus the H+ ion based on the proton chemical shift using 1H-NMR.

Figure 3.1 A GaLMA droplet mechanically constrained by a surface oxide in deformed shape (left) and the resulting energetically favorable spherical shape of the GaLMA fluid when the oxide is removed (right)

by HCl. HCl can be incorporated into siloxane, a polar, aprotic solvent. The now-acidified siloxane

contains a combination of associated HCl that forms a hydrogen bonding complex with the oxygen, as well as a dissociative complex, whereupon the oxygen is protonated and the chloride acts as a counter-

ion.

3.4 Acidic Siloxane Characterization by 1H-NMR

We acquired 1H-NMR to quantify the amount of acid taken up by the siloxane, identify impurities, and to identify any chemical changes that occurred upon acidification and/or mixing with the GaLMA.

Figure 3.2 (a) 1H-NMR of the siloxane, the siloxane after combination with HCl overnight and after

several dilutions with virgin siloxane. No changes are observed for peaks corresponding to methyl

protons (0.6 and 0.8 ppm). However, a peak appears after incubation with HCl at 1.02 ppm. This peak

reduces in intensity upon dilution with virgin siloxane. Integration of the acid peak area enabled us to

determine the concentration of HCl. (b) Comparison of the oil and acidic oils after exposure to the

GaLMA, demonstrates the generation of a third, unique peak that we attribute to GaCl3/InCl3 siloxane

adducts only after exposure to the GaLMA.

In Figure 3.2a, we observe that there is minimal change in the linewidth, the position, and the intensity of the peaks upon subsequent dilutions of the acidified siloxane with virgin OS-

20. The only difference between the spectra is the presence and intensity of the peak at 1.02 ppm, which decreases proportionally with dilution. No other peaks were observed above 1.02 ppm. In contrast, an acid peak was observed in diethylether + 2 M HCl (Sigma Aldrich) at 2 ppm as well as at 5 ppm (Figure 3.7). We quantified the concentration of the acid by integration of the peak areas of both the siloxane and the HCl, determining that overnight combination of the siloxane and the acid results in 1.5 M HCl.

After verifying the chemical composition and integrity of the acidified siloxanes and the absence of potential impurities and degradation products (Figures 3.8 and 3.9), we looked for any changes in the siloxane after exposure to GaLMAs. Figure 3.2b shows the 1H-NMR spectra of the regular siloxane, HCl siloxane, and HBr siloxane after exposure to the GaLMA for 1 week

(~50 µL of GaLMA immersed in ~10 mL of the oil). It should be noted that the oxide is visually observed (as in Figure 3.1) to be removed in the HCl and HBr siloxane samples, but not in the regular siloxane. Interestingly, even in an environment with minimal oxygen (glove box with 0.5 ppm oxygen), the HCl peak disappears and we observe a change in the number of distinct peaks corresponding to methyl protons from two to three. The central peak that appears after exposure of the GaLMA with the acidified siloxane is believed to be either solubilized gallium and indium chloride adducts interfacing with the siloxane or the byproduct (i.e. silanol) of the gallium acting as an oxide scavenger. In the latter case, water is generated when the oxide is etched away.

The NMR clearly shows that the HCl siloxane reacts with the GaLMA and reveals a limitation of this acidic siloxane approach: it works initially to remove and prevent the oxide, but the acid will continue to be consumed afterwards as it etches the metal and metal oxide ((In + 3

HCl  InCl3 + 1.5 H2 (g)). This is not unexpected, as most etching agents have limited selectivity for the oxide over the metal, especially hydrogen halides. While etching limits the

effective potlife of the mixture to days for building and sealing a device (i.e. upon addition of the

GaLMA, the acid will be depleted both by oxide removal as well as metal etching), there is still sufficient time to fabricate the device and to consume residual oxygen remaining within the device. Furthermore, we do not believe that continuous acid depletion is a major limitation.

There are at least two options which could improve long-term stability in a sealed device. (1)

Oxygen barriers and/or scavengers should be effective in minimizing oxide formation upon depletion of the acid. The application advantage of this acidified siloxane is that a device can be fabricated in the presence of oxygen and assembled in a 1 step process (e.g. not chemically treated in one solution then transferred to another). (2) A device could also be sealed with a small amount of HCl or HBr that is placed in a sealing sub-assembly inside the outer sealing.

The only requirement would be that the release of HCl or HBr into the oil over time would be at a greater rate than that of oxygen entering the sealed system. Since hermetic sealing can reduce oxygen entry to extremely low levels, the required acid release rate could also be extremely low as well. Any resulting continual loss of GaLMA mass would be negligible for any application that we currently envision.

3.5 Identification of Chemical Species Generated Using Electrospray Ionization Mass

Spectrometry (ESI-MS)

1H-NMR clearly identified the methyl protons in the siloxanes as well as the proton in the

HCl complex. 1H-NMR also enabled us to observe significant changes in chemical composition as a function of how the acidic siloxane was prepared. However, we were unable to observe coordinate covalent complexes and/or polymers we believed were being generated through our process due to sensitivity limitations. We used ESI-MS to answer the following questions: 1) were gallium/indium chloride species being produced; 2) were these species associating with the siloxane in some manner; 3) was chemical rearrangement occurring upon adding acid to the

siloxane, causing both longer chain and shorter siloxanes to be produced; and 4) could we identify any additional chemical species that may be produced through the process.

Figure 3.3 (a) ESI-MS of the OS-20 oil shows it is of high purity and of a singular species (n=1), where n is

the number of dimethyl siloxane groups. (b) Upon addition of HCl, we now observe a peak at n=0 and c)

polymerization of the siloxane where n=2, n=3… . (d) Addition and incubation of the GaLMA in the acidified siloxane oil maintained the same patterns of polymerization, but now (e) we also observe gallium and indium complexes and/or adducts, demonstrating the solubility of GaCl3 and InCl3 into the siloxane oil.

The ESI-MS in Figure 3.3a shows the spectrum of virgin OS-20, where only the sodium adduct of OS-20 is observed. The peak at 107 mass units was observed in a background spectra

as a contaminant. Upon acidification of the siloxane by HCl (Figure 3.3b, 3.3c) we observed rearrangement and polymerization. Instead of solely observing n=1, where n is the number of dimethyl siloxane groups, there is now n=2, n=3, n=4…where the intensity of the observed peak decreases as chain length increases. We clearly observe the protonated siloxane (M+1), as well as the peak (M-15) corresponding to loss of a methyl group. This is unsurprising, considering how acid is known to catalyze hydrolysis and/or rearrangement of siloxanes20.

ESI-MS of the acidic siloxane after exposure to the liquid metal showed additional species

(Fig 3d, 3e). We were able to identify these unknown peaks as gallium chloride/siloxane adducts due to the unique 40% M, 60% M+2 isotope pattern observed (Figure 3.10). Further analysis identified the corresponding indium chloride/siloxane adducts. Interestingly, there is an absence of literature identifying or characterizing these adducts. It is likely that GaCl3 and InCl3 exist in a coordinate covalent complex analogous to diethyl ether adducts previously identified and observed25,26. It is apparent that the gallium and indium chloride produced on the shell of the liquid metal are being solubilized by the siloxane solution and exist stably in solution. It is interesting to note that adducts are observed only when n=3 or greater. The reason for this is unknown, but we believe it is worth calling attention to.

3.6 Dielectric Characterization of the Acidic Siloxane

As electrical insulation is one of the stated advantages of our acidified siloxane, the dielectric properties of the 1.5 M HCl acidic siloxane were characterized and compared to the behavior of the neat siloxane. Because the regular siloxane is proven to be effective as the insulating fluid for electronics applications, including electrowetting27, the acidic siloxane should ideally show similar electrically insulating properties to be useful in this manner. It is also important to quantify these properties such that various electronic devices can at least be conceptually designed and theoretically modeled.

Figure 3.4 (a) Waveguide perturbation analysis showed minimal difference in the dielectric properties of the oil versus the acidic oils, all were found to be highly insulating. The one sigma of the measurements at each frequency was typically equal to or less than the size of the data points shown in the figure. (b) Breakdown voltage test showed that the I-V response of the oil versus acidic oil is similar until the breakdown voltage is reached.

The high frequency complex permittivity was calculated using a resonant cavity perturbation method28,29 (details in methods section) at frequencies ranging from about 2-20

GHz, shown in Figure 3.4a. The real and imaginary parts of the permittivity are calculated by

measuring the slight shift in the resonant frequency and peak Q before and after the sample is placed in the cavity, using the equations below.

풇 푸 = ퟎ (ퟑ. ퟏ) ∆풇(−ퟑ풅푩)

푽풄(풇풄−풇풔) 휺′풓 = + ퟏ (ퟑ. ퟐ) ퟐ푽풔풇풔

푽풄 ퟏ ퟏ 휺"풓 = ( − ) (ퟑ. ퟑ) ퟒ푽풔 푸풔 푸풄

Where f0 is the frequency at the maximum resonant signal, Δf(-3dB) is the frequency width of the peak at 3 decibels below the peak maximum, V is volume, and the subscripts c and s refer to the values of no sample and with a sample, respectively. The dielectric behavior of both HCl and HBr acidic siloxanes is shown to be essentially the same as that of the regular siloxane.

This is quite encouraging and exciting considering that the tested HCl acidic siloxane is at 1.5 M

HCl, and in sealed devices far less HCl content would be needed.

The low frequency current-voltage response of the HCl and regular siloxane oils were measured by an electrical breakdown test (see methods for details). Figure 3.4b shows that the

HCl siloxane behaves identically to the regular siloxane at lower voltages up until the breakdown voltage is reached. The experiment was repeated 3 times with similar results for each oil type. We observed breakdown at approximately 10 V/μm for the 1.5M HCl siloxane compared to 16 V/μm for the regular siloxane. We hypothesize that this is due to the Wein effect (also known as the Dissociation Field Effect). In this effect the electric field induces greater alignment of the siloxane and the acid molecules along the electric field. By orienting the molecules according to their dipole moments, greater proton conduction occurs. This results in a shift in the pKa which frees up more ions, and thus increases current flow30.

3.7 Electrowetting Demonstrations of GaLMA

The ability to control and manipulate the GaLMA through electrowetting is a potential breakthrough for reconfigurable electronic device design and applications. Historically, liquid

metal electrowetting has relied on mercury, or as noted in the introduction, alternative approaches to electrowetting are used which bring additional challenges if used in electronics applications. Furthermore, electrowetting is an excellent validation of the acidic siloxane, as electrowetting requires nearly perfect electrical insulation around the electrowetting liquid, and is very sensitive to changes in interfaces between any of the three phases in a siloxane/liquid metal/solid system. For example, our own experiments have shown that simply trying to use

HCl vapor does not work for electrowetting because, upon reacting with the gallium oxide, water forms on the surface of the GaLMA along with dissociated metal chloride ionic byproducts.

When water is present it will preferentially electrowet and the GaLMA contact angle does not change.

Electrowetting18 requires having a conducting fluid, surrounded by an insulating fluid, which is separated from an electrode by a solid dielectric layer. A low interfacial surface tension between the insulating fluid and the solid dielectric layer surface is used to achieve a high zero- voltage contact angle (up to 180 degrees). When a voltage is applied between the electrode and the conducting fluid, charges build up at the interface of the liquid metal and the electrode beneath the dielectric, creating an electromechanical force that alters the force-balance of interfacial surface tensions. The result is a decrease in the observable contact angle. However, the actual microscopic contact angle, governed by the surface energies of the three phases involved, is unchanged31.

Figure 3.5 A schematic showing the process of electrowetting: a conducting fluid (liquid metal) surrounded by an insulating fluid (oil) on an electrode but separated by a dielectric layer. The equilibrium state of the liquid metal droplet is a high contact angle (~180°) due to the interfacial energy balance. When a voltage is applied between the liquid metal and electrode, the droplet is forced out of equilibrium by the applied electric field, lowering its macroscopic contact angle.

The plot shows the experimental GaLMA relationship of the applied voltage to the contact angle

(dots) as well as the theoretical relationship based on the interfacial energies (solid line).

Figure 3.5 shows the contact angle vs. applied voltage for the GaLMA in the HCl siloxane of approximately 0.135 M HCl (see methods for fabrication details). The contact angle was measured at increasing voltages over a time period of about 5 minutes in an open container; therefore the exact concentration of acid in the siloxane at each measurement cannot be guaranteed as some HCl leaves the system as vapor. The zero voltage contact angle

(Young’s angle) of the LM is expected to be 180° but due to gravity and the limit of using images for measurement, it appears as 160°. Using electrowetting, the contact angle can be reduced to approximately 120° with 300 V. In electrowetting, there is a saturation limit where the decrease

in contact angle plateaus as voltage is increased32. We do not see such behavior because our system’s maximum peak to peak voltage was 300 V, which might not be enough to reach the saturation limit. The experimental data also shows good agreement with the theoretical curve, which was calculated using the electrowetting equation:

훾푖푑−훾푐푑 휀표휀푟 2 cos 휃푉 = + 푉푝푝 (ퟑ. ퟒ) 훾푐푖 2푧훾푐푖

Where θV is the apparent contact angle when voltage is applied, γ is the interfacial energy between any two of the conducting liquid (c), insulating liquid (i), and dielectric (d) phases, ε is the dielectric constant, z is the thickness of the dielectric, and Vpp is the applied voltage. Even this relatively small change in contact angle can be levied in useful device applications33. We also present a simple demonstration of GaLMA reconfiguration (shape-change) by electrowetting. Figure 3.6 shows zoom-in photo demonstration of a switchable wire-grid polarizer device using electrowetting to actuate the GaLMA. When no voltage is applied the liquid metal splits into droplets but when a voltage is applied the liquid metal is pulled to follow the electrode lines. Details on the device design and testing can be found as published elsewhere33.

Figure 3.6 Photos of a portion of wire-grid polarizer device in both on and off states where the dark areas are the GaLMA and the light areas are the substrate. In the off state, the liquid metal is split into droplets by shallow channels patterned on the top plate33. In the on state, the liquid metal takes the line shape of the electrodes patterned on the bottom plate.

3.8 Conclusions

We have demonstrated a new material enabling a breakthrough in non-toxic (GaLMA) liquid metal reconfigurable devices. The acidified siloxane has numerous advantages over other methods being explored to remove/prevent oxide without interfering with device functionality.

Laboratories and fabrication lines can now use GaLMAs to fabricate devices without resorting to use of cumbersome ultra-low oxygen environments. Furthermore this work demonstrates the first path to make long-lasting functional devices, because even sealed devices will allow some minute oxygen entry over time which will cause GaLMA oxidation. As a result, GaLMA is now more practical for a variety of applications. We believe our novel acidified siloxane can also be used in other applications requiring the presence of acid in an insulating aprotic environment.

The siloxane itself has been shown to eliminate oxides as well as solubilize and incorporate the metal ions into the liquid with minimal impact on the insulating properties. In our case, we have used this novel solvent in order to enable electrowetting of GaLMAs.

3.9 Supplemental

NMR of the HCl in ether from Sigma Aldrich shows two peaks, other than those known for ether, corresponding to the acid (Figure 3.7a). One peak at ~ 3 ppm is from associative combination of the HCl and ether and the other at ~7.5 is dissociative. In comparison to this, the HCl in OS-20 has only one peak downfield showing it is mostly associative.

Figure 3.7 1H-NMR showing a) both an associative and dissociative peak of HCl-ether. b) Water and silanol doped oil compared to only silanol showing how potential contaminants would appear in the acidic oil.

We verified what the spectra of some expected contaminants would look like (Figure

3.7b), including water and silanol (Xiameter PMX-0156). In the silanol spectra (red), we see both a shift in the methyl protons as well as a small peak at ~3.5 ppm corresponding to the hydroxyl proton. In the siloxane doped with silanol (purple), these features are significantly reduced. However, if the siloxane is doped instead with water (light blue), we observe a peak corresponding to the acidified water at ~5.5 ppm. The chemical shift indicates the acidity, as the chemical shift for water is typically between 4-5 ppm25). Comparing these potential byproducts to the spectra of our acidified siloxane (black), it is clear that not only has water been eliminated, but that the bulk of the oil remains unhydrolyzed.

Figure 3.8 1H-NMR showing a) acid concentration for various incubation times of the HCl and oil and b) differences in oil combined with HCl and HBr.

The effects of varying aqueous acid exposure time (Figure 3.8a) as well as acid composition (Figure 3.8b) were explored, both with regards to byproducts as well as stability.

Figure 3.8a shows samples that were exposed to the 12 M HCl for 1 hour (light grey), overnight

(grey), or 4 days (black). It is clear that minimal acid is retained in the siloxane after only 1 hour of exposure. Exposure for up to 4 days increases the volume (and therefore concentration) of the acid peak by 25% compared to the overnight exposure, however a third peak appears that we attribute to hydrolysis of the siloxane. We concluded that 16-18 hours of exposure provides sufficient time to absorb HCl, but insufficient time for the siloxane to hydrolyze appreciably.

We then tested whether we could acidify the siloxane using another acid (Figure 3.8b).

We chose HBr which has a higher boiling point and a slightly greater dissociation constant than

HCl. Bromide ions also tend to be less reactive than chloride ions. We wanted to determine

whether replacing HCl with HBr affected the electronic properties of the siloxane or its ability to remove metal oxides. The HBr siloxane appears identical to the virgin OS-20 and has no peak relating to the acid like the HCl siloxane does. We did observe a small peak in HBr siloxane after extended exposure (4 days) at ~2.5 ppm, indicating that the acid is incorporated in an associative manner. However, the peak volume indicates that even after 4 days of exposure the siloxane contains only approximately 50 μM HBr, compared to 1.5 M HCl after 16-18 hours of exposure. We attribute this to the weaker dissociation constant of HCl, which would preserve a higher concentration of the associated HCl in the 12N aqueous solution.

Figure 3.9 FTIR of the oil, acidic oil, and acidic oil doped with water.

FTIR was used to identify functional groups and to confirm that no substantial changes were occurring to the oil. A peak at ~2350 nm was attributed to HCl, but was very weak (Figure

3.9b). No other difference was observed between the regular oil and the HCl oil (Figure 3.9a).

Comparison of the regular oil, HCl oil purposefully contaminated with water, and HCl oil exposed to GaLMA showed that the water produced by the oxide-acid reaction is taken up into the HCl oil and perturbs the functional groups minimally (Figure 3.9b).

Figure 3.10 shows unique isotope pattern helped to identify gallium chloride adducts

(60% M, 40% M+2). Upon closer inspection, indium chloride adducts were also identified

(standard isotopic pattern).

Figure 3.10 ESI-MS showing the unique isotope patterns of the gallium chloride adducts in the sample of EGaIn in dry oil with HCl.

3.10 References

1. Palleau, E. et al. Self-healing stretchable wires for reconfigurable circuit wiring and 3D microfluidics. Adv. Mater. 25, 1589–92 (2013). 2. Koo, C. et al. Manipulating Liquid Metal Droplets in Microfluidic Channels With Minimized Skin Residues Toward Tunable RF Applications. J. Microelectromechanical Syst. 24, 1–8 (2014). 3. Fassler, A., & Majidi, C. 3D structures of liquid-phase GaIn alloy embedded in PDMS with freeze casting. Lab Chip 13, 4442–50 (2013). 4. Mohammed, M. G. & Dickey, M. D. Strain-controlled diffraction of light from stretchable liquid metal micro-components. Sensors Actuators A Phys. 193, 246–250 (2013).

5. Aïssa, B. et al. Fluidic patch antenna based on liquid metal alloy/single-wall carbon- nanotubes operating at the S-band frequency. Appl. Phys. Lett. 103, 063101 (2013). 6. Zhu, S. et al. Ultrastretchable Fibers with Metallic Conductivity Using a Liquid Metal Alloy Core. Adv. Funct. Mater. 23, 2308–2314 (2013). 7. Ralston, N. V. C., Azenkeng, A. & Raymond, L. J. Methylmercury and Neurotoxicity. (Springer US, 2012). 8. Sen, P. & Kim, C. C. J. Microscale Liquid-Metal Switches — A Review. IEEE Trans. Ind. Electron. 56, 1314–1330 (2009). 9. Dickey, M. D. et al. Eutectic Gallium-Indium (EGaIn): A Liquid Metal Alloy for the Formation of Stable Structures in Microchannels at Room Temperature. Adv. Funct. Mater. 18, 1097–1104 (2008). 10. Chabala, J. M. Oxide-growth kinetics and fractal-like patterning across liquid gallium surfaces. Phys. Rev. B 46, 11346-11357 (1992). 11. Ilyas, N., Butcher, D. P., Durstock, M. F. & Tabor, C. E. Ion Exchange Membranes as an Interfacial Medium to Facilitate Gallium Liquid Metal Alloy Mobility. Adv. Mater. Interfaces 3, 1500665 (2016). 12. Cumby, B. L. et al. Reconfigurable liquid metal circuits by Laplace pressure shaping. Appl. Phys. Lett. 101, 174102 (2012). 13. Kim, D. et al. Recovery of Nonwetting Characteristics by Surface Modification of Gallium-Based Liquid Metal Droplets Using Hydrochloric Acid Vapor. ACS Appl. Mater. Interfaces 5, 179–185 (2013). 14. Khan, M. R., Eaker, C. B., Bowden, E. F. & Dickey, M. D. Giant and switchable surface activity of liquid metal via surface oxidation. Proc. Natl. Acad. Sci. 111, 14047–14051 (2014). 15. Wang, M., Trlica, C., Khan, M. R., Dickey, M. D. & Adams, J. J. A reconfigurable liquid metal antenna driven by electrochemically controlled capillarity. J. Appl. Phys. 117, 194901 (2015). 16. Vincent, G. & Dow C. C. The Effects of Water and Silanol on the Electrical Properties of Silicone Fluids. Annual Report; National Acadamies 559-566 (1975). 17. Izutsu, K. Electrochemistry in Nonaqueous Solutions. (John Wiley & Sons, Inc., 2003). 18. Mugele, F. & Baret, J.-C. Electrowetting: from basics to applications. J. Phys. Condens. Matter 17, R705–R774 (2005). 19. Bartnikas, R. Electrical Insulating Liquids. (ASTM International, 1994). 20. Cypryk, M. & Apeloig, Y. Mechanism of the Acid-Catalyzed Si-O Bond Cleavage in Siloxanes and Siloxanols. A Theoretical Study. Organometallics 21, 2165–2175 (2002). 21. Liu, T., Sen, P., Kim, C. C. J. & Measurements, A. C. A. Characterization of Nontoxic Liquid-Metal Alloy Galinstan for Applications in Microdevices. J. Microelectromechanical Syst. 21, 443–450 (2012). 22. Aldridge, S. & Downs, A. J. The Group 13 Metals Aluminium, Gallium, Indium and Thallium: Chemical Patterns and Peculiarities. (John Wiley & Sons, 2011).

23. Giebels, I. A. M. E., van den Broek, M. A. F. H., Kropman, M. F. & Bakker, H. J. Vibrational dynamics of hydrogen-bonded HCl-diethyl ether complexes. J. Chem. Phys. 112, 5127 (2000). 24. Kapoor, K. P., Luckcock, R. G. & Sandbach, J. A. Solubility of hydrogen chloride in ethers. Measurement of vapour pressures and infra-red studies of ether-hydrogen chloride solutions. J. Appl. Chem. Biotechnol. 21, 97–100 (2007). 25. Greenwood, N. N. & Srivastava, T. S. A Nuclear Magnetic Resonance Investigation of Some Gallium Trihalide Complexes. J. Chem. Soc. A Inorganic, Phys. Theor. 703–706 (1966). 26. Fairbrother, F., Flitcroft, N. & Prophet, H. Complex compounds of indium trihalides part I. Ether complexes. J. Less Common Met. 2, 49–63 (1960). 27. Sawane, Y. B., Datar, S., Ogale, S. B. & Banpurkar, A. G. Hysteretic DC electrowetting by field-induced nano-structurations on polystyrene films. Soft Matter 11, 2655 (2015). 28. Baker-Jarvis, J. et al. Dielectric and Conductor-Loss Characterization and Measurements on Electronic Packaging Materials. Natl. Inst. Stand. Technol. 1520 (2001). 29. Baker, J. et al. Measuring the Permittivity and Permeability of Lossy Materials: Solids, Liquids, Metals, Building Materials, and Negative-Index Materials. Natl. Inst. Stand. Technol. 1536 (2004). 30. Cassone, G., Giaquinta, P. V, Saija, F. & Saitta, M. Proton conduction in water ices under an electric field. J. Phys. Chem. B 118, 4419–4424 (2014). 31. Mugele, F. & Buehrle, J. Equilibrium drop surface profiles in electric fields. J. Phys. Condens. Matter 19, 375112 (2007). 32. Chevalliot, S., Kuiper, S. & Heikenfeld, J. Experimental Validation of the Invariance of Electrowetting Contact Angle Saturation. J. Adhes. Sci. Technol. 26, 1909–1930 (2012). 33. Diebold, A. et al. Electrowetting actuation of liquid metal wires for RF applications. J. Micromech. Microeng. 27, 025010 (2017).

Chapter 4: Electrowetting-Actuated Liquid Metal for RF Applications

This chapter includes adapted text and figures from Diebold, A. et al. “Electrowetting-actuated liquid metal for RF applications”. Journal of Micromechanics and Microengineering 27, 2 (2017) which is Copyright © 2017 by IOP Publishing Ltd.

4.1 Introduction

The need to reconfigure RF electronic components is becoming increasingly important in areas ranging from free space applications such as antennas to breadboard components like transmission lines. Often these reconfigurable components are rigid and introduce significant signal loss to the system, while providing only limited tunability. For example, digitally configurable electrode array antennas require lossy semiconducting diodes or power-limited

MEMS switches to achieve discrete effective length changes in radiating elements1,2,3. Others require complicated placement of parasitic elements relative to a fixed antenna structure, offering modest variation of the frequency response or radiation pattern4,5. In contrast, liquid metals, such as Hg and eutectic gallium indium (EGaIn)6, offer the potential for true continuous analog physical reconfigurability7, physical flexibility8,9, and high power thresholds10 due to their high electrical conductivities (Hg conductivity = 1.0 × 106 푆/푚11, EGaIn conductivity = 3.4 ×

106 푆/푚12). Recent work has demonstrated effective pneumatic actuation methods for liquid metal based RF components utilizing Laplace pressure shaping13,14 and pressure-driven flow15.

These techniques are limited in switching speed by additional peripherals such as pumps or syringes which are not readily integrated into devices architectures. Electrochemical methods to mobilize liquid metals have also been reported such as continuous electrowetting (CEW) and electrocapillarity16,17,18,19,20,21. These devices eliminate the need for peripheral pneumatic controls and can operate using only several volts. However, electrochemical techniques to date

are significantly limited in actuation speed and require immersion in electrically lossy electrolytic solutions, which have limited their use in RF applications.

Electrowetting has emerged as an attractive alternative for liquid metal actuation22,23,24,25 which, in theory, can resolve many of the challenges that exist for the above approaches.

Electrowetting in other applications has allowed for automated transport and mixing of biofluids26, tuning of microprism arrays27, adjusting variable-focus lenses28, and altering reflective displays29. The phenomenon relies on variation of droplet contact angle by electromechanical force, resulting in a variety of fundamental operations30.

There are two prominent obstacles in utilizing electrowetting to reconfigure liquid metals for RF applications. Firstly, there is a prohibitively large contact angle change (>90°) required to cause reversible capillary wetting/dewetting of a droplet into and out of an open-ended channel or capillary. To date, electrowetting of liquid metal has been reported to only evoke modest changes in shape and contact angle of a small drop of metal in a few limited switch based applications. Feinerman et al. have reported two types of micromirrors which function by varying the contact angle of a sessile drop of Hg23,24. A frequency reconfigurable antenna has been presented previously31 where frequency tuning results from the change in capacitive loading as the wetting state of a droplet of Hg is varied electrostatically. A similar method drives the frequency sweep in an RF MEMS resonator32. Finally, Kim et al. have made significant contributions to the development and optimization of liquid metal switches based on electrowetting of Hg25,33. The contact angle change required for more significant fluid mobilization and RF tuning is prohibited by the high surface tensions of liquid metals (>400 mN/m34,35) which results in excessive voltage requirements (>1 kV) to achieve large angle changes that are often beyond or near the breakdown limit of practical dielectric films.

Further complicating matters, when attempting to use non-toxic alternatives to mercury such as gallium liquid metal alloys (EGaIn and Galinstan), rapid gallium oxide formation occurs on the metal surface renders electrowetting inoperable due to the highly adhesive and

viscoelastic properties of the oxide6,7,37. Simply using an inert gas is impractical as even < 1 ppm oxygen levels over time will cause some degree of oxide formation38,39. Kim et al.40 have performed extensive characterization of Galinstan in a nitrogen glove box held below 0.5 ppm oxygen, concluding that true liquid behavior is attained at levels below 1 ppm oxygen.

Reversible electrowetting was successfully demonstrated in that report. However, these levels of O2 concentration are incredibly difficult to maintain in a device. An acidic vapor background for chemical removal7,41 of the oxide skin is also unusable with electrowetting, because an aqueous annulus forms near the electrowetting contact line42, which in our experience will electrowet instead of the liquid metal (see supplemental information).

We report here fully integrated devices that utilize several key novel approaches to resolve these challenges for liquid metal electrowetting, and furthermore, we demonstrate a functional RF device in the form of an electromagnetic polarizer that can be actuated on demand. The enabling innovations include a geometric microchannel design that greatly reduces the required contact angle range needed for actuation, the use of an RF transparent conducting polymer as the contacting electrode, and the use of a novel acidic and electrically insulating oil that eliminates the oxide-induced limitations associated with utilizing non-toxic

Gallium liquid metal alloys (GaLMAs) instead of mercury in these devices43. Material optimization then enables the demonstration of a switchable wire-grid polarizer, which provides an average signal attenuation of 12.91 dB in the on state and 1.46 dB in the off state, over the range of 8-9.2 GHz, with a switching speed of 12 ms.

4.2 Materials and Fabrication of Test Devices

4.2.1 Electrowetting Characterization

Mercury and eutectic gallium indium (EGaIn) were used as purchased (Aldrich, Hg,

>99.99% trace metals basis; Aldrich, Gallium-Indium eutectic, >99.99% trace metals basis).

Contact angle characterization with voltage was performed on glass substrates coated with

In2O3:SnO2 (Kaivo, < 10 Ω/square), 3.8 μm of chemical vapor deposited Parylene C (Specialty

Coating Systems, 휺풓 = ퟑ. ퟏ), and a top monolayer of hydrophobic FluoroPel PFC 1601V

(Cytonix). The monolayer was created using a surface grafting technique, described elsewhere[44]. Electrowetting of mercury was performed in Dow Corning® OS-20 silicone oil purchased from Krayden. Electrowetting of EGaIn was performed in an open bath of specially formulated acidic silicone oil. Synthesis and characterization of this oil has been performed in detail by our group. More details on the oil are provided in the experimental results section and are reported elsewhere [43]. Contact angle change was driven by a 1 kHz square wave voltage.

Contact angle measurements were performed using a VCA Optima contact angle analysis system.

4.2.2 Polarizer Device Demonstration

The polarizer fabrication consists of a bottom plate supporting channels aligned with patterned bottom electrodes coated with a thin dielectric. The top plate is coated with an electrode material in addition to thin ridges patterned perpendicular to the bottom channels

(Figure 4.1).

The bottom plate of the polarizer was fabricated by first spin-coating positive photoresist

(Microposit S1818) onto a glass slide, then patterning with 500 mJ/cm2 of i-line energy in the negative of the electrode pattern. The electrode pattern consists of a series of parallel stripes,

475 μm in width, pitch ퟏ. ퟔퟖ mm center to center, connected by a single contact pad at the bottom. A thin layer (< 100 nm) of gold was then sputtered onto the glass/photoresist structure and patterned using the lift-off technique.

Figure 4.1 Device diagrams and relevant dimensions for switchable polarizer.

Next, channels were formed on the bottom plate by spin-coating 72.3% solids SU-8

3000 series (Microchem) negative photoresist at 1800 rpm, yielding a thickness of 40 μm, followed by a soft bake at 100° C for 15 minutes. Channels with width 958 μm were aligned and exposed as above, and then a post exposure bake was performed for 5 minutes at 100° C.

Finally, the photoresist was developed in PGMEA, and the entire structure hard baked at 180° C for 30 min. Next, a 3.5 μm thick coating of Parylene C and monolayer of FluoroPel PFC 1601V were deposited as previously described for the electrowetting test plates.

Poly(3,4-ethylenedioxythiophene) polystyrene sulfonate (PEDOT:PSS) (Aldrich, 1.3 wt% dispersion in H2O) was chosen for the electrode material on the top plate for reasons described below and was spun on at 2500 rpm, and then baked at 120° C for 15 min. This was repeated twice in order to achieve the desired conductance. The sample was then heated under vacuum for 30 minutes at 110° C, to remove any water absorbed by the PEDOT:PSS.

Ridges of thickness 10 μm, width 1.46 mm, and spacing 2.46 mm were patterned by spinning 50% solids SU-8 3000 series negative resist at 1000 rpm onto the cured PEDOT:PSS, then baking and patterning as described above. This structure was coated with a monolayer of

FluoroPel PFC 1601V to achieve hydrophobicity while still maintaining electrical contact45.

Polarization measurements consisted of a signal from a Hewlett-Packard 8684B Signal

Generator being fed through an X-band waveguide and emitted by a Budd Stanley E-plane sectoral horn antenna (X4100-611). Transmitting and receiving antennas were placed a distance of 1.45 cm apart, and the polarizer device placed directly between them (≈ ퟎ. ퟕ cm from each antenna). The antennas measure 27.6 mm in the direction of the E-field and 22.8 mm in the H-plane (629.28 mm2), whereas the active area of the polarizer measures 1,760 mm2.

Signal amplitude values were obtained using an IFR AN930 spectrum analyzer. High speed photos were taken using a Fastec Troubleshooter camera.

4.3 Theory / Liquid Metal Electrowetting Device Design

Electrowetting involves an apparent change in contact angle of an electrically conductive fluid under an applied voltage46. The Young-Lippmann equation relates interfacial surface tensions to the equilibrium contact angle of a sessile drop:

ퟐ 휺ퟎ휺풓푽 풄풐풔휽푽 = 풄풐풔휽풀 + (ퟒ. ퟏ) ퟐ휸풄풊풕 where 휺ퟎ is the permittivity of free space, 휺풓 is the relative permittivity of the dielectric used, 휸풄풊 is the interfacial surface tension between the conducting fluid and the insulating ambient, 풕 is the thickness of the dielectric coating, 푽 is the applied voltage between the bottom and top electrode, and 휽풀 is the resting Young’s contact angle under zero applied potential.

As the intent of this work is to change the shape of a liquid metal droplet, we choose to model our problem in terms of droplet pressure, which has explicit dependence on droplet shape. Laplace pressure ∆풑, the pressure difference across the boundary of a liquid, is related to the shape of a droplet by the Young-Laplace equation46:

ퟏ ퟏ ∆풑 = 휸풄풊 ( + ) (ퟒ. ퟐ) 푹ퟏ 푹ퟐ where 푹ퟏ and 푹ퟐ are the principal radii of curvature of the droplet. We see that the high surface tension of the liquid metal, 휸풄풊, results in high values for the droplet Laplace pressure.

In electrowetting, the electrostatic driving pressure on a conducting droplet in a channel of height (풉 + 풄) may be expressed as47,48:

휺 휺 푽ퟐ ∆풑 = ퟎ 풓 (ퟒ. ퟑ) 풆 ퟐ(풉 + 풄)풕

Transitions between droplet shape configurations may thus be achieved by supplying adequate electrostatic pressure. That is, the supplied electrostatic pressure must exceed the difference between some initial state, ∆풑ퟏ, and the desired final state ∆풑ퟐ:

∆풑ퟐ − ∆풑ퟏ < ∆풑풆 (ퟒ. ퟒ)

Our goal is to minimize this required electrostatic pressure through proper device design.

In the case of the liquid metal polarizer, the desired final shape of each droplet (Fig. 1) is a strip of width 풅 (radius of curvature 푹푯ퟐ = 풅/ퟐ, dictated by electrode width) and height 풉

(radius of curvature 푹푽ퟐ ≈ 풉/ퟐ, dictated by channel height). Representation of 푹푽ퟐ as half of the channel height is a reasonable assumption for EGaIn given its very high surface tension which allows use of the approximation of a 180° Young’s angle on the fluoropolymer surface.

In order to minimize the electrostatic pressure required to achieve this state, we place our droplet in an initial state ∆풑ퟏ as close to its final state ∆풑ퟐ as to still allow proper device functionality. This is achieved by confining the droplet in a channel of width 풘 (radius of curvature 푹푯ퟏ = 풘/ퟐ) which is larger than the electrode width 풅 only by the amount required by volume conservation. The droplet is confined vertically to a height larger than that achieved in its final state by a small amount 풄 (radius of curvature 푹푽ퟏ ≈ (풉 + 풄)/ퟐ). The result of this confinement is twofold. First, it reduces the pressure difference between the two states which must be supplied electrostatically, as discussed above. Second, it enhances the electrostatic pressure contribution according to Eq. 4.3 by substantially reducing the height of the droplet.

Additional channel height in the amount 풄 is enabled by patterning ridges of thickness 풄 perpendicular to the channels of height 풉. The resulting variation in channel height ensures that droplets preferentially dewet into areas of larger channel height, i.e. lower Laplace pressure, upon removal of voltage. Contributions to required voltage due to this variation in channel height were minimized by making dewetting ridges wide and thin, since dewetting reliability increases both with ridge width and thickness, but electrostatic pressure requirements increase only with ridge thickness.

Thus, each droplet wets as desired (Figure 4.1) once the supplied electrostatic pressure exceeds the pressure difference between the initial array configuration and the final, merged state. According to Eq. 4.4, this condition is expressed as:

ퟐ휸 ퟐ휸 ퟐ휸 ퟐ휸 휺 휺 푽ퟐ 풄풊 + 풄풊 − 풄풊 − 풄풊 < ퟎ 풓 (ퟒ. ퟓ) 풉 풅 풉 + 풄 풘 ퟐ(풉 + 풄)풕

We see that because the liquid metal is confined it is presented with a starting Laplace pressure that aids our goal of electromechanically changing its shape into features such as wires. Ohta et al.49 interpret this mechanism with a surface energy approach, where confinement effectively biases the droplet such that minor changes in electrostatic energy effect large in-plane deformations. As noted in the introduction, electrowetting a sessile liquid metal droplet into a capillary or open-face rectangular channel would require a much larger change in droplet pressure, which in our experience is often prohibitive in terms of voltage and dielectric reliability (Figure 4.5).

The dimensions employed in the polarizer design indicate that the onset of diffraction occurs at a frequency of 89 GHz for the wire-grid polarizer50, well above the frequencies tested here.

4.4 Experimental Results

A device was fabricated according to the above dimensions with mercury as the conducting fluid in a silicone oil bath and was verified to function as predicted from the above equations. Operation of the liquid metal polarizer device was first characterized by measuring droplet edge velocity in an Hg-loaded device using a standard ITO electrode, captured by a high speed camera. Figure 4.2 plots the length change of Hg as a function of time. For these plots a voltage of 330 V was utilized, corresponding to an expected minimum contact angle of 90°

based on the Young-Lippmann equation. A maximum length change of about 78% is observed over 12 ms, corresponding to the switch on time. A similar response is observed in the case of dewetting, or switching off. This implies an average wetting speed of 10.3 cm/s.

Figure 4.2 Wetting and dewetting speeds in polarizer.

Before the device could be tested as a wire-grid polarizer, a replacement for the

In2O3:SnO2 (<10 Ω/square) was needed because of excessive transmission loss in the GHz frequencies. Figure 4.3 illustrates the transmission profiles for various possible electrode materials. A thin layer of sputtered gold (< 100 nm) was employed for the bottom electrodes due to its low attenuation and ease of patterning. PEDOT:PSS conductive polymer was used as the top electrode due to its similarly low signal attenuation as well as its availability in dispersion form and potential application in flexible electronics. Most importantly, compatibility with EGaIn and Hg requires that the bare top electrode be of a non-alloying material. Conductive polymers,

such as PEDOT:PSS, are a natural candidate in this regard, although their RF properties have not been thoroughly studied to date.

Figure 4.3 Signal transmission for various electrode materials including fluorine doped tin oxide

(FTO) and indium tin oxide (ITO).

After substituting the above materials into the device, polarizer attenuation for the Hg device was measured in the X-band using horn antennas. The device was actuated with 330 V

AC, using a frequency of 100 Hz, in order to allow for proper capacitive discharge while still preventing dielectric charging that results from the electrowetting process51. RF transmission measurements are depicted in Figure 4.4 for the polarizer in the off (4A) and on (4B) states, with liquid metal “wires” oriented both perpendicular and parallel to the polarization of the electric field. When aligned perpendicular to the electric field polarization, there was no observed signal attenuation, which indicates that losses due to reflection are negligible. In this orientation, transmission greater than that of air may be due to near field effects, or to measurement error.

When aligned parallel to the electric field polarization, the device exhibited an average attenuation of 12.91 dB in the on state and 1.464 dB in the off state over the range 8-9.2 GHz.

As comparison, a patterned copper polarizer of similar dimensions offered an average signal attenuation of 14.51 dB over the same range.

Figure 4.4 Transmission results for device on and off, with the liquid metal wires

(a)perpendicular and (b)parallel to the electric field polarization.

An important consideration for RF liquid electronics, is the choice of conducting fluid. As mentioned previously, the use of Hg is highly contentious due to toxicity concerns and potential loss of material over time due to the high vapor pressure. The substitution of Gallium liquid metal alloys (GaLMAs) for mercury as a non-toxic, near-0 vapor pressure alternative has seen recent attention in the literature, but presents the new concern of surface oxide formation and its effects on electrowetting actuation. Here, we demonstrate for the first time the electrowetting actuation of a common GaLMA fluid, namely eutectic gallium indium (EGaIn), in a device architecture nearly identical to the wire-grid polarizer presented above and enabled by the use

of a novel acidic silicone oil. Oxidized EGaIn droplets are viscoelastic, so that arbitrary shapes attained by any means of actuation are maintained once the stimulus (e.g. voltage) is removed7

(Figure 4.5). Further complications arise from the highly adhesive quality of the gallium oxide skin, as well as the consequent oxide residue which irreparably fouls devices37,52. Achieving reversible and reliable electrowetting of EGaIn thus requires oxide-free droplets. In our work, this is achieved through the use of an electrically insulating acidic silicone oil43.

Figure 4.5 Contact angle change of EGaIn vs. AC voltage using a 1 kHz square wave in acidic oil with (a) low concentrations of HCl and (b) high concentrations of HCl. The percentage values refer to the wt% HCl (see text).

The electrowetting response data shown in Figure 4.5 displays the dependence of the contact angle on voltage as a function of varying HCl concentrations in the silicone oil43. The black curve corresponds to the theoretical response predicted by the electrowetting equation, using a surface tension of 445 mN/m34.

To fabricate the acidic oil, HCl was added to Ar-purged silicone oil until the oil is saturated, i.e. until further HCl is no longer integrated into the structure, yielding a maximum molarity of 1.5M (6.28 wt% HCl)43. We define percent concentrations as vol% saturated oil in a saturated/pure oil system (Figure 4.5).

Our tests were performed in a 1 cm-thick acidic oil bath open to air, so that after the initial oxygen-free oil is dispensed, the diffusion rate of oxygen through the oil determines the concentration of oxygen in the oil.

The effects of silicone oil with low HCl concentrations on enabling EGaIn electrowetting in ambient environments are presented in Figure 4.5a. The 0% curve illustrates the contact angle change of an EGaIn droplet in silicone oil with no HCl content. Though dissolved oxygen content was minimized by purging the oil with Ar, the droplet was observed to be oxidized immediately after being dispensed. This was evidenced by obvious deformation of the droplet from a spherical shape. Oxidation was confirmed by the extreme contact angle hysteresis observed, which derives from the viscoelastic and adhesive characteristics of oxidized EGaIn.

This result highlights the irreversibility of electrowetting with oxidized EGaIn. Fluctuations in this curve are due to gradual shape change as oxidation progresses. The curves corresponding to

0.2% and 0.4% HCl content show good agreement with theory, as well as no observed hysteresis, indicating that oxide is minimized by continual reaction of HCl such that the droplet regains elasticity.

At higher HCl concentrations (Figure 4.5b), the contact angle response of EGaIn becomes less consistent, and at higher voltages (>200 V) the droplet vibrates/oscillates visibly, so that accurate contact angle values become difficult to retrieve. Measurements for the 1.98%

and 3.87% cases terminate at 250 V due to motion of the droplet becoming so significant that the droplet leaves the probe. Similar behavior has been reported in22 and observed in our lab for the case of Hg. At significantly higher concentrations, i.e. 5.09% and 6.28%, total contact angle change from 0-300 V is drastically reduced. This behavior may possibly be attributed to screening effects due to charging of the oil and dielectric, as well as water generation which is a product of the reaction of HCl with gallium oxide species41. The exact mechanism responsible has not been definitively investigated in this work. These high concentrations represent an upper bound for the range of practical concentrations when electrical insulation is required at such voltages. It should be noted that we fully expect that in a sealed system, much lower concentrations of HCl in the oil would be permitted.

Device actuation was then tested with EGaIn in a low-oxygen glovebox held below 1 ppm oxygen, which has been reported to increase the longevity of oxide-free gallium liquid metal alloys40. In our observation, in a glove box at these oxygen levels, the droplets oxidized enough to prevent device operation within minutes when no HCl oil was utilized. Indeed, others report more stringent environmental requirements for oxide prevention, such as ultrahigh

(10−9 Torr) vacuum systems, with oxidation rates steadily increasing at higher pressures38,39.

We thus concluded that optimal device performance would be achieved by assembling the device in the glove box in concert with the use of acidic oil for active oxide removal. Slight modifications to the wire grid polarizer were made to enable this assembly, as detailed in the supplemental information.

Figure 4.6 Photos of operating device, including (a) initial self-loading, (b) with voltage off, and (c) with voltage on.

Photos of the fully functioning EGaIn wire-grid polarizer device are provided in Figure

4.6, with prominent device features indicated by dotted white lines. One of the challenges with arrayed electrowetting devices, such as a wire-grid polarizer, is dosing discrete volumes of fluids. In our previous work, we have developed self-assembly dosing techniques for electrowetting53 and electrofluidic displays29,54. Here, we again demonstrate a self-assembly approach to initially dose the EGaIn into the device. As shown in Figure 4.6a, self-assembled dosing of the device is enabled by the effect of the electrowetting electrodes in combination with a pressurized reservoir at the bottom of the device which spans all of the stripe electrodes.

Simply, the liquid metal is pulled into the channels by electrowetting (Figure 4.6a), and when the voltage is removed, is deterministically split into discrete droplets by the greater Laplace pressure imparted by the top SU-8 ridge (Figure 4.6b). The transition between the electrowetted wire-grid polarizer state of Figure 4.6c and the discrete droplet array of Figure 4.6b was reversible and fast (~12 ms, Figure 4.3)—the same speed demonstrated for Hg in Figure 4.2.

4.5 Discussion

If the mechanism demonstrated in this paper is to be widely employed, several remaining challenges should be mentioned. Some improvements can be expected if one were to shrink device dimensions. As switching speed is determined partially by ridge width, a uniform scaling of device dimensions would result in an increased switching speed.

Variations in droplet size and line width may be seen in the device photos and supplemental video. Such effects are a result of imprecise device loading techniques (syringe injection, by hand) and of device modifications (discussed in the supplemental information), namely the replacement of channel walls with discrete posts in order to maximize the volume available to the acidic oil. Because of this, droplets are not as well-confined and repeatability suffers. Optimization of confinement pressures and alternate forms of oil delivery were not

investigated in this work, but such factors would require consideration in sophisticated applications.

Our demonstration of device functionality has not ruled out partial oxidation of the droplets, and suboptimal performance may also result from this. In fact, due to the dynamic nature of oxide formation, partial oxidation may occur depending on the acid concentrations employed. An upper limit for these concentrations is indicated by contact angle behavior illustrated in Fig. 5, though the mechanism responsible for such fluctuations has not been precisely determined. A possible cause is electrical screening by water generated in the reaction of gallium oxide species with HCl41 (see supplemental information). If this is the case, it may be beneficial to include an active desiccant into the device design. Nevertheless, we have shown that device functionality may still be achieved, so that reversible actuation exhibits some tolerance for the gallium oxide.

Though environmental temperature may affect polarizer performance, such effects were not considered in this work besides the obvious criterion that the operating temperature was above the melting point of the metals of interest. Therefore, all tests were performed at room temperature.

Regarding device design, sputtered gold was used as the bottom electrodes due to its ease of patterning. The 1.464 dB signal loss in the parallel, off state is most likely caused by this gold layer. Substitution of patterned PEDOT:PSS would adequately reduce these off-state losses. Methods of patterning PEDOT:PSS have been demonstrated55,56.

To promote device longevity with gallium liquid metal alloy materials, devices should be assembled with proper hermetic sealing in order to prevent acid evaporation, oxygen diffusion and the resulting oxidation of EGaIn. We have found that certain acids react with EGaIn in undesirable ways43, such that lower concentrations are preferred, or else alternative formulations must be used to achieve longer device lifetimes. Also, lower acid concentrations

will decrease the amount of acid which partitions into the dielectric materials, which is preferred from an electrical reliability perspective.

4.6 Conclusions

Reliable electrowetting of liquid metal for RF applications has been demonstrated in this work following the investigation of an improved device architecture and several enabling materials. A newly-developed acidic oil was used to enable EGaIn mobility due to its ability to remain electrically insulating while mitigating detrimental surface oxide effects and the use of a low loss conducting polymer, PEDOT:PSS, was used as the electrode that electrically contacts the liquid metals without chemically alloying with them. A switchable polarizer utilizing electrowetting has been demonstrated which offers an average signal attenuation of 12.91 dB over the frequency range 8-9.2 GHz, with switching speeds of about 12 ms. This method lays the groundwork for future devices allowing arbitrary physical reconfiguration of liquid metal components for RF applications, bridging the gap between two areas of vast potential in the field of microfluidics. The approach has significant promise in areas where increased efficiency requires active components to be segregated from the actuation components, such as frequency-shifting antennas, electronic switches, variable impedance transmission lines, and reconfigurable metamaterials.

4.7 Supplemental

EGaIn Integration into Electrowetting RF Device: To demonstrate the ability to replace mercury with non-toxic gallium liquid metal alloy alternatives (EGaIn in this case), the device architecture was slightly modified. These slight modifications are needed to overcome effects from the thin oxide skin that forms. Acidic silicone oil was fabricated according to previously reported techniques and used to mitigate the oxide formation. Actuation tests were also performed inside a low oxygen environment (2-10ppm). To ensure more consistent and reliable

dosing, materials were fabricated on polyethylene terephthalate glycol-modified (PETG) substrates, which are transparent and chemically resistant. This allowed inlet and outlet holes to be machined using a drill press so that the EGaIn could be directly injected into the device.

Holes were machined around the active area so that the plates could be secured using nylon nuts and bolts. Use of a torque screwdriver ensured that these were fastened evenly, yielding a uniform channel height throughout the device. Top ridges were thickened (12 μm) to promote better reservoir confinement and dewetting reliability. Finally, bottom SU-8 channels were replaced with SU-8 Laplace barriers in order to increase the volume available to the acidic oil which should preferably surround each EGaIn droplet and increases the rate at which HCl is available to etch the any oxide formation on the EGaIn.

Water Formation: When attempting to perform electrowetting tests by employing high acid concentrations for oxide removal (i.e. when using aqueous HCl vapor or acidic oil concentrations approaching saturation), water is observed to gather at the droplet contact line, due to the reaction of oxidized EGaIn with HCl (Figure 4.7). In this reaction, gallium oxide species (Ga2O and Ga2O3) are replaced by gallium chloride (GaCl3) and water (H2O) components. The electromechanical force that enables electrowetting occurs only for a conducting droplet surrounded by an insulating fluid, so that contact angle change is drastically reduced as the chemically-formed water screens the electromechanical force at the triple contact line.

Figure 4.7 Water formation on EGaIn droplet.

4.8 References

1. Roscoe, D. J. et al. Tunable Dipole Antennas. Proc. IEEE 672–5 (1993). 2. Piazza, D. et al. Design and Evaluation of a Reconfigurable Antenna Array for MIMO Systems. IEEE Trans. Antennas Propag. 56, 869–81 (2011). 3. Fries, M. K., Grani, M. & Vahldieck, R. A Reconfigurable Slot Antenna with Switchable Polarization. IEEE Microw. Wirel. Compon. Lett. 13, 490–2 (2003). 4. Pringle, L. N. et al. A Reconfigurable Aperture Antenna Based on Switched Links Between Electrically Small Metallic Patches. IEEE Trans. Antennas Propag. 52, 1434–45 (2004). 5. Zhang, S., Huff, G. H., Feng, J. & Bernhard, J. T. A Pattern Reconfigurable Microstrip Parasitic Array. IEEE Trans. Antennas Propag. 52, 2773–6 (2004). 6. Harrington, R. Reactively controlled directive arrays. IEEE Trans. Antennas Propag. 26, 390–5 (1978). 7. Dickey, M. D. Emerging Applications of Liquid Metals Featuring Surface Oxides Appl. Mater. Interfaces 6, 18369–79 (2014). 8. Dickey, M. D. et al. Eutectic gallium-indium (EGaIn): A liquid metal alloy for the formation of stable structures in microchannels at room temperature. Adv. Funct. Mater. 18, 1097– 1104 (2008). 9. Hayes, G. J., So, J.H., Qusba, A., Dickey, M. D. & Lazzi, G. Flexible Liquid Metal Alloy (EGaIn) Microstrip Patch Antenna. IEEE Trans. Antennas Propag. 60, 2151–6 (2012). 10. Zhu, S., So, J. H., Mays, R., Desai, S., Barnes, W. R., Pourdeyhimi, B. & Dickey, M. D. Ultrastretchable fibers with metallic conductivity using a liquid metal alloy core. Adv. Funct. Mater. 23, 2308–14 (2013).

11. So, J. H., Thelen, J., Qusba, A., Hayes, G. J., Lazzi, G. & Dickey, M. D. Reversibly Deformable and Mechanically Tunable Fluidic Antennas. Adv. Funct. Mater. 19, 3632–7 (2009). 12. Surmann, P. & Zeyat, H. Voltammetric analysis using a self-renewable non-mercury electrode. Anal. Bioanal. Chem. 383, 1009–13 (2005). 13. Zrnic, D. & Swatik, D. S. On the Resistivity and Surface Tension of the Eutectic Alloy of Gallium and Indium. J. Less-Common Met. 18, 67–8 (1969). 14. Cumby, B. et al. Reconfigurable liquid metal circuits by Laplace pressure shaping. Appl. Phys. Lett. 101, 174102 (2012). 15. Cumby, B., Heikenfeld, J., Mast, D., Tabor, C. & Dickey, M. Robust pressure-actuated liquid metal devices showing reconfigurable electromagnetic effects at GHz frequencies. IEEE Trans. Microw. Theory Tech. 63, 3122–30 (2014). 16. Wang, J., Liu, S., Vardeny, Z. V. & Nahata, A. Liquid metal-based plasmonics. Opt. Express 20, 2346 (2012). 17. Eaker, C. B., Dickey, M. D., Eaker, C. B. & Dickey, M. D. Liquid metal actuation by electrical control of interfacial tension. Appl. Phys. Rev. 3, 31103 (2016), 18. Wang, M., Trlica, C., Khan, M. R., Dickey, M. D. & Adams, J. J. A reconfigurable liquid metal antenna driven by electrochemically controlled capillarity. J. Appl. Phys. 117, 194901 (2015). 19. Khan, M. R., Eaker, C. B., Bowden, E. F. & Dickey, M. D. Giant and switchable surface activity of liquid metal via surface oxidation. Proc. Natl. Acad. Sci. 111, 14047–51 (2014). 20. Lee, J. & Kim, C. J. Surface-tension-driven microactuation based on continuous electrowetting. J. Microelectromechanical Syst. 9, 171–80 (2000). 21. Gough, R. C., Morishita, A. M., Dang, J. H., Hu, W., Shiroma, W. A. & Ohta, A. T. Continuous Electrowetting of Non-toxic Liquid Metal for RF Applications. IEEE Access 2, 874–82 (2014). 22. Khan, M. R., Trlica, C., So, J., Valeri, M. & Dickey, M. D. 2014 Influence of water on the interfacial behavior of gallium liquid metal alloys. ACS Appl. Mater. Interfaces 6, 22467–73 (2014). 23. Wan, Z., Zeng, H. & Feinerman, A. Reversible Electrowetting of Liquid-Metal Droplet. J. Fluids Eng. 129, 388 (2007). 24. Zeng, H., Feinerman, A. D., Wan, Z., & Patel, P. R. Piston-motion micromirror based on electrowetting of liquid metals. J. Microelectromech. Syst. 14, 285–94 (2005). 25. Wan, Z., Zeng, H. & Feinerman, A. Area-tunable micromirror based on electrowetting actuation of liquid-metal droplets. Appl. Phys. Lett. 89, 201107 (2006). 26. Sen, P. & Kim, C. A Fast Liquid-Metal Droplet Microswitch Using EWOD-Driven Contact- Line Sliding. J. Microelectromech. Syst. 18, 174–85 (2009). 27. Hua, Z. et al. Multiplexed real-time polymerase chain reaction on a digital microfluidic platform. Anal. Chem. 82, 2310–6 (2010).

28. Smith, N. R., Abeysinghe, D. C., Haus, J. W. & Heikenfeld, J. Agile wide-angle beam steering with electrowetting microprisms. Opt. Express 14, 6557–63 (2006). 29. Kuiper, S. & Hendriks, B. H. W Variable-focus liquid lens for miniature cameras. Appl. Phys. Lett. 85, 1128–30 (2004). 30. Heikenfeld, J. et al. Electrofluidic displays using Young–Laplace transposition of brilliant pigment dispersions. Nat. Photonics 3, 292–6 (2009). 31. Cho, S. K., Moon, H. & Kim, C. J. Creating, transporting, cutting, and merging liquid droplets by electrowetting-based actuation for digital microfluidic circuits. J. Microelectromech. Syst. 12, 70–80 (2003). 32. Damgaci, Y. & Cetiner, B. A frequency reconfigurable antenna based on digital microfluidics. Lab Chip 13, 2883–7 (2013). 33. Irshad, W. & Peroulis, D. A 12-18 GHz Electrostatically Tunable Liquid Metal RF MEMS Resonator with Quality Factor of 1400-1840. IEEE MTT-S Int. Microwave Symp. Digest (MTT) 2, 3–6 (2011). 34. Xu, Q., Oudalov, N., Guo, Q., Jaeger, H. M. & Brown, E. Effect of oxidation on the mechanical properties of liquid gallium and eutectic gallium-indium. Phys. Fluids 24, 63101 (2012). 35. Aqra, F. & Ayyad, A. Theoretical calculations of the surface tension of liquid transition metals. Metall. Mater. Process. Sci. 42, 5–8 (2011). 36. Kim, J., Shen, W., Latorre, L. & Kim, C. J. A micromechanical switch with electrostatically driven liquid-metal droplet. Sensors Actuators A 97–98, 672–9 (2002). 37. Doudrick, K., Liu, S., Mutunga, E. M., Klein, K. L., Damle, V., Varanasi, K. K. & Rykaczewski, K. Different Shades of Oxide: From Nanoscale Wetting Mechanisms to Contact Printing of Gallium-Based Liquid Metals. Langmuir 30, 6867–77 (2014). 38. Chabala, J. M. Oxide-growth kinetics and fractal-like patterning across liquid gallium surfaces. Phys. Rev. B 46, 11346 (1992). 39. Su, C. Y., Skeath, P. R., Lindau, I. & Spicer, W. E. 1982 Photoemission Studies of Room- Temperature Oxidized Ga Surfaces. Surf. Sci. 118, 248–56 (1982). 40. Liu, T., Sen, P. & Kim, C. J. Characterization of nontoxic liquid-metal alloy galinstan for applications in microdevices. J. Microelectromech. Syst. 21, 443–50 (2012). 41. Nekrasov, S. Y., Migdisov, A. A., Williams-Jones, A. E. & Bychkov, A. Y. An experimental study of the solubility of Gallium(III) oxide in HCl-bearing water vapour. Geochim. Cosmochim. Acta 119, 137–48 (2013). 42. Kim, D. et al. Recovery of Nonwetting Characteristics by Surface Modi fi cation of Gallium- Based Liquid Metal Droplets Using Hydrochloric Acid Vapor. Appl. Mater. Interfaces 5, 179–85 (2013). 43. Holcomb, S., Brothers, M., Diebold, A., Mast, D., Tabor, C. & Heikenfeld, J. Acidified siloxane oil for oxide-free actuation of gallium liquid metal alloys. Langmuir 32, 12656–63 (2016).

44. Berry, S., Fedynyshyn, T., Parameswaran, L. & Cabral, A. Reversible Electrowetting on Dual-Scale-Patterned Corrugated Microstructured Surfaces. J. Microelectromech. Syst. 21, 1261–71 (2012). 45. Kreit, E., Mognetti, B. M., Yeomans, J. M. and Heikenfeld, J. Partial-post laplace barriers for virtual confinement, stable displacement, and >5 cm s−1 electrowetting transport. Lab Chip 11, 4221 (2011). 46. Mugele, F. & Baret, J. C. Electrowetting: from basics to applications. J. Phys. Condens. Matter 17, R705–74 (2005). 47. Lee, J., Moon, H., Fowler, J., Schoellhammer, T. & Kim, C. J. Electrowetting and electrowetting-on-dielectric for microscale liquid handling. Sensors Actuators A 95, 259–68 (2002). 48. Fair, R. B. Digital microfluidics: Is a true lab-on-a-chip possible? Microfluid. Nanofluid. 3, 161–8 (2007). 49. Gough, R. C., Morishita, A. M., Dang, J. H., Moorefield, M. R., Shiroma, W. A. & Ohta, A. T. Rapid electrocapillary deformation of liquid metal with reversible shape retention. Micro Nano Syst. Lett. 3, 1–9 (2015). 50. Bass, M. Handbook of Optics Vol. II (McGraw-Hill, Inc., 1995) 51. Thomas, D., Audry, M. C., Thibaut, R. M., Kleimann, P., Chassagneux, F., Maillard, M. & Brioude, A. Charge injection in dielectric films during electrowetting actuation under direct current voltage. Thin Solid Films 590, 224–9 (2015). 52. Koo, C., LeBlanc, B. E., Kelley, M., Fitzgerald, H. E., Huff, G. H. & Han, A. Manipulating Liquid Metal Droplets in Microfluidic Channels With Minimized Skin Residues Toward Tunable RF Applications. J. Microelectromech. Syst. 24, 1069–76 (2014), 53. Sun, B. & Heikenfeld, J. Observation and optical implications of oil dewetting patterns in electrowetting displays. J. Micromechanics Microengineering 18, 25027 (2008). 54. Yang, S., Zhou, K., Kreit, E. and Heikenfeld, J. High reflectivity electrofluidic pixels with zero-power grayscale operation. Appl. Phys. Lett. 97, 143501 (2010). 55. Li, D. & Guo, L. J. Micron-scale organic thin film transistors with conducting polymer electrodes patterned by polymer inking and stamping. Appl. Phys. Lett. 88, 2–5 (2006). 56. Xiao, S. Picosecond Laser Direct Patterning of Poly(3,4-ethylene dioxythiophene)- Poly(styrene sulfonate) (PEDOT:PSS) Thin Films. J. Laser Micro/Nanoengineering 6, 249–54 (2011).

Chapter 5: Mitigating the Effects of Sticking and Failure in EGaIn Electrode Contacts

This chapter includes adapted text and figures from Holcomb, S. et al. “Mitigating the Effects of

Sticking and Failure in EGaIn Electrode Contacts”. Pre-Press, (2019).

5.1 Introduction

Physically reconfigurable electronics such as foldable electronics and MEMS technologies present opportunities to rewire hardware in a way that has been increasingly explored over the last couple decades. A newer technology that takes advantage of physical reconfiguration of electronic materials is the implementation of liquid electronics within microchannels to connect, disconnect, and modify electrical circuits based on the location of the confined fluids. One of the most useful materials in this space has been liquid metal alloys, specifically those primarily composed of gallium such as eutectic gallium indium (EGaIn) and galinstan1–7. Gallium liquid metal alloys are widely studied for various flexible, stretchable, and reconfigurable electronics applications because of their room temperature liquidus properties, high conductivity, and the fact they are non-toxic8,9. In addition to these attributes, gallium alloys have the remarkable ability to spontaneously form a self-containing passivating oxide when they come into contact with oxygen. This oxide can be used to contain the liquid if damage to the channel occurs10, physically constrain the liquid in non-equilibrium states as a result of the oxide’s viscoelastic properties11, and can be functionalized readily to modify the surface electronic properties12.

One of the primary challenges with physically reconfigurable electronics have been interfaces, such as stiction and contamination in MEMS devices13 or delamination of flexible electronics14. Liquid metal electronics possess similar challenges at interfaces between liquid and solid conductors which have caused significant problems in the past15–17. There are two primary failure modes that have been observed at solid/liquid metal contacts. The first is atomistic diffusion of material at the interfaces as gallium leaches out of the alloy and diffuses

into grain boundaries or alloys with the solid metal conductors when physical contact is made18,19. This can result in delamination of the solid electrodes20 and/or loss of liquid conducting material through unwanted phase changes in the resulting alloy21. The second failure mode results from undesired adhesion of the gallium oxide skin with the solid metal electrode which also usually has a thin surface oxide. These two challenges are in a bit of tension, since the absence of the oxide skin results in increased alloying due to increased metal/metal contact while the presence of the oxide skin leads to unwanted physical adhesion and potential loss of mobility.

Many of the reported demonstrations in the literature utilizing solid / liquid metal contacts such as reconfigurable switches, free space polarizers, and antennas22–25 have avoided cycled contact between liquid metal and solid metal through fabricating free space devices, using static liquid/solid contacts21,26 27–29, employing capacitive coupling between the solid and liquid metal electrodes, or through the application of a thick organic conducting barrier such as carbon black that adds resistive loss to the final device30. Examples of devices which do utilize cyclic contacts of EGaIn and solid electrodes have not explored the previously highlighted detrimental effects, only showing operation over a limited number of cycles31,32.

There have been two complimentary approaches to the challenge of interfacing liquid and solid metals which have been demonstrated in the literature. The first is the use of a favorable solvent layer at the interface along with the liquid gallium alloy33. The second is the use of covalently bound organic ligand such as phosphonic acids12 to chemically bind to the oxide surfaces and prevent adhesion between the gallium alloy and the solid conductor. Often times the solid conductors also have thin oxides as well which can be simultaneously coated, which leads to further interfacial barriers. These approaches are advantageous, as the oxide and organic layers are often nanometers in thickness and can easily be tunneled through with little increase in contact resistance34,35. However, while the ability to significantly reduce or eliminate sticking

of the oxide with the solid metal electrodes has been demonstrated, the long term effects of these chemical approaches to protecting the electrical interface have yet to be reported.

In this work, we present an analysis of the failure mechanisms that plague liquid metal / solid metal connections in reconfigurable electronics with the prototypical EGaIn / copper connection and provide an explanation of the various mitigation strategies presented to overcome them. Specifically, the mode of contact is shown to have a dramatic effect on the interface behavior, and the effect of utilizing solvent as well as solvent with an organic ligand that binds to the oxide interface have been analyzed herein. Additionally, the application of a voltage during switching is seen to affect the longevity of these interfaces, with a final optimized approach shown to be effective up to 1000 cycles without failure or degradation.

5.2 Experimental Materials and Methods

5.2.1 Materials and Preparation

Decyl phosphonic acid (DPA) was purchased from Lancaster Synthesis, Inc. and used without further purification. Prior to use, 1 × 10−3 M ethanolic solutions of DPA were sonicated for at least 10 min. EGaIn was prepared by mixing of gallium (GalliumSource, 99.99%) and indium (GalliumSource, 99.99%) at the eutectic composition. Copper substrates were prepared by sputtering 150 nm of Cu on solvent cleaned glass slides with a 20 nm Cr seed layer using a

Denton sputtering system. Copper substrates were treated with UV-Ozone for 30 minutes just prior to testing or just before DPA pretreatments. DPA pretreatments consisted of submerging the substrate in a 1 mM DPA-ethanol solution for approximately 18 hours and then rinsing with ethanol and drying with nitrogen upon removal. For each experiment, all variable sets were tested from the same batch and on the same days to control for time or other variables affecting the surfaces. Preparation, testing, and characterization were done at consistent time intervals between batches within 10 day cycles. Each batch of tests was performed at least three times.

5.2.2 Surface Characterization

An FEI Quanta scanning electron microscope (SEM) was used to look at the copper surfaces before and after contacts. Secondary electron imaging, backscatter imaging, and energy-dispersive X-ray spectroscopy (EDS) were all used to examine the surfaces at 10 keV. A

Kratos Axis Ultra X-ray photoelectron spectroscopy (XPS) was used to further characterize surface differences. For all samples, the measurements and regions were as follows: survey spectrum, high-resolution spectrum 30 eV wide centered at 19 eV BE (Ga 3d), high-resolution spectrum 30 eV wide centered at 130 eV (P 2p), high-resolution spectrum 30 eV wide centered at 285 (C 1s), high-resolution spectrum 30 eV wide centered at 445 eV BE (In 3d), high- resolution spectrum 20 eV wide centered at 532 eV (O 1s), high-resolution spectrum 50 eV wide centered at 950 eV (Cu 2p), and a high-resolution spectrum 20 eV wide centered at 1117 eV

(Ga 2p). Comparisons were done between the area of GaLMA contact and undisturbed portions of the substrate.

5.2.4 Contact Cycle Testing

Two custom built test rigs were used to control the cycling for contact testing, a manual setup for low cycles of less than 10 and an automated setup for up to 1000 cycles. The manual rig consisted of a syringe held in a syringe pump and suspended over the substrate which was in a square sided container to hold the solution. The syringe pump controlled the expanding of the EGaIn droplet for the expand contact tests, however due to the nature of EGaIn and its oxide skin it could not retract the droplet and so a new droplet was used for each contact. A lab jack was used to move the substrate up and down in contact with the EGaIn droplet for the touch contact tests. Since only the touch contact tests were performed for higher than 10 cycles the automated setup used an Aerotech printing stage for control and programing of the higher cycle contact tests.

5.2.5 Electrical Measurements

A Kiethley 2400 voltage supply was used to take the electrical measurements using a current limiting 4 point probe setup. The probes were attached to the stainless steel needle of a

1 mL syringe containing the EGaIn and a push pin which held the substrate in place. The current was set to 0.001 Amps and voltage was limited to 4 V DC.

5.3 Results and Discussion

In order to best characterize the effects of interface contact of a liquid metal alloy such as EGaIn and a solid copper electrode, two modes of contact were identified. These modes are detailed in Figure 5.1. The first contact mode is labeled Touch and Retract, whereby the surfaces of both liquid and solid metals coming into contact with a near constant interfacial area.

In this mode, the liquid metal alloy does not change in size once contact is made and more importantly the interfacial contact area does not increase. The second contact mode is labeled

Expand and Retract, whereby an “expanding” process occurs of the liquid metal alloy while it is in contact with the solid metal electrode. The interfacial contact area increases, thereby potentially opening up the liquid metallic core contact sites between the electrode surface and the liquid metal alloy without the barrier of the gallium oxide skin. The resulting disconnected liquid metal remaining on the copper electrodes after retraction using both modes of contact are evident in Figure 5.1 over just one cycle. Copper was chosen as the solid electrode because of its common use in electronics and was intentionally oxidized through UV-Ozone treatment to ensure a consistent surface (see experimental methods for more details). However, this effect is observed on nearly every electrode metal available.

Figure 5.1. Modes of contact of EGaIn droplets and Cu substrates. (top) Touch and retract contact where the gallium oxide layer is less disturbed. (bottom) Expand and retract contact where the gallium

oxide layer is ruptured, exposing EGaIn directly to substrate.

A common mitigation strategy employed in the literature has been the use of a solvent to serve as a barrier layer between the liquid metal and the solid electrode33. Figures 5.2a and

5.2b demonstrate this strategy by performing both modes of contact between EGaIn / copper in a solvent bath of absolute ethanol. Ethanol was chosen here because of the favorable polar interaction with the oxide surfaces. In both modes of contact, the EGaIn drop disconnected from the surface, however with visually different impacts on the liquid metal. In Touch and

Retract mode, the viscoelastic oxide is “dented” and permeant deformation in the oxide results in a somewhat flattened liquid metal drop. During the Expand and Retract mode, there is adhesion to the surface, evident by the point that permanently forms at the bottom of the liquid metal drop.

Figure 5.2: Demonstration of two strategies, contact in a solvent (a and b) and in the presence of a

surface ligating species (c and d) to eliminate irreversible adhesion of EGaIn to a copper electrode.

A complimentary approach to mitigating adhesion at this interface is to introduce a surfactant that binds to the oxide interface such as a phosphonic acid12. Figure 5.2c and 5.2d demonstrate this strategy in conjunction with ethanol by introducing 1mM decylphosphonic acid

(DPA) to the ethanol solution prior to initiating contact between EGaIn and copper. It has previously been shown that DPA strongly binds to the surface of gallium oxide36 and copper oxide37 and would be expected to provide an additional physical barrier to inhibit adhesion between the metal surfaces. This effect is observable in the Expand and Retract contact where the adhesion between the liquid drop and the electrode is significantly reduced in the presence of DPA.

While the beneficial effects of a solvent and a ligating species to enable successful retraction of liquid metal from a copper surface are obvious from the above results, devices utilizing liquid metals to reconfigure electrical functionality will utilize many thousands if not millions of cycles of connection / disconnection. Therefore, we next evaluate the accumulative effect of multiple contact cycles on the interface between EGaIn and copper electrodes. Figure

5.3a and 5.3b shows SEM taken in secondary electron mode for up to ten contacts (or taken to failure - permanent adhesion of liquid metal to the electrode, whichever occurred first) in both

Touch and Retract mode Expand and Retract mode, respectively.

The Touch and Retract mode contact in absolute ethanol after 10 cycles resulted in an obvious contact damage mark and the appearance of a “footprint”, indicated by the contrast change on the copper surface (5.3.a.i). Interestingly the presence of excess DPA in solution resulted in no observable change to the surface of the copper electrode after 10 cycles (5.3.a.ii).

The effect of DPA was also explored as a thin DPA film deposited on the copper/copper oxide prior to contact testing in an absolute ethanol solution; however this approach also resulted in noticeable contact damage and a “footprint” forming on the copper electrode following 10 contact cycles in Touch and Retract mode (5.3.a.iii). Again, this contact damage was not noticeable for the DPA coated copper/copper oxide electrode if excess DPA was present in the ethanol solution (5.3.a.iv). The presence of excess DPA in the solvent was helpful in replenishing any defects on the molecular barrier layers on the EGaIn and copper surfaces.

When the Expand and Retract mode was used (Figure 5.3.b) to contact EGaIn and a copper

electrode, all of the experiments resulted in observable contact damage and most contacts left large residues of EGaIn on the copper electrode due to significant adhesion and stopped operating after only 2-3 cycles (5.3.b.i-iii). The exception to this was when there was both an initial coating of DPA on the copper/copper oxide electrode and excess DPA in solution

(5.3.b.iv) where ten cycles were completed with no large residues of EGaIn sticking to the surface. However, further inspection showed microscopic traces of EGaIn on the contact area.

Figure 5.3. a) Secondary electron SEM micrographs of a copper electrode (i and ii) and a copper

electrode treated with DPA (iii and iv) after Touch and Retract contact with EGaIn after ten cycles in

ethanol (i and iii) and 1mM DPA/EtOH (ii and iv). The presence of DPA in the solvent during contact

protects the surface from attack as indicated by a “footprint” forming on the copper without excess DPA present. b) Secondary electron SEM micrographs of a copper electrode (i and ii) and a copper electrode

treated with DPA (iii and iv) after Expand and Retract contact with EGaIn after ten cycles in ethanol (i and

iii) and 1mM DPA/EtOH (ii and iv). EGaIn residue occurred quickly in all instances except where the

surface was pre-coated with DPA and there was excess DPA in the solvent.

The contact damage or “footprint” left on the copper surface from EGaIn is evident in several of the SEM images in Figure 5.3. This “footprint” is likely caused by decreased surface charging in the secondary electron imaging mode, since there was little to no observable changes to the copper surfaces in backscattering electron SEM mode (Figure 5.X). To better understand the effect of EGaIn on the electrode surface during contact, the damage areas were chemically analyzed using X-ray Photoelectron Spectroscopy (XPS) of the copper electrode.

Figure 5.4a and 5.4b provide spectra from the copper surface after a single contact in a 1mM

DPA / ethanol solution was made using Touch and Retract and Expand and Retract modes, respectively. In both contact modes, the contact area was compared to a pristine area on the same electrode. There was no gallium signal detected on either electrode at the interface region in the XPS analysis, indicating both methods restricted gallium diffusion at this interface over the first cycle of contact (Figure 5.XX). The most noticeable change in the contact regions was a loss of copper oxide signal that correlated with the presence of the “footprint” observed in the

SEM images (Figure 5.4c and 5.4d). After the Expand and Retract contact, the ratio of the oxidized copper 2p3/2 signal (933.6 eV) compared with the ground state metalized copper 2p3/2 signal (932.6 eV)38 significantly decreased. Additionally, the satellite peak associated with the oxidized Cu signal also decreased in intensity after contact with EGaIn. Alternatively, there was no noticeable reduction in the copper oxide signals following contact to EGaIn using the Touch and Retract mode, which correlated with no observable damage area on the electrode in the

SEM image.

Figure 5.4. Effect of different contact modes on Copper/EGaIn interface. a) Touch and retract contact in

DPA-ethanol XPS spectra (top) and SEM image (bottom) of contact area (gold) and control area (red). b)

Expand and retract contact in DPA-ethanol XPS spectra (top) and SEM image (bottom) of contact area

(gold) and control area (red).

It appears that over one cycle, the presence of DPA at the interface between the copper electrode and EGaIn has a barrier effect that helps to retard the adhesion of the two metals to one another in both contact modes. The reduced copper oxide at the contact area between

EGaIn and copper during the Expand and Retract is likely caused by the gallium oxide ripping and opening up at the buried interface which would result in exposing metallic gallium directly to the CuO39. The significantly lower reduction potential of gallium (-0.529 V) compared with copper (0.340 V)40 would result in a transfer of the oxygen species to the gallium surface through reduction of the copper and oxidation of the gallium, as shown in Equation 5.1.

ퟑ퐂퐮(퐈퐈)퐎 + ퟐ퐆퐚(ퟎ) → ퟑ퐂퐮(ퟎ) + 퐆퐚(퐈퐈퐈)ퟐ퐎ퟑ (ퟓ. ퟏ)

The reduction of the copper surface by gallium occurs much more readily during the more abrasive contact mode (Expand and Retract), where the organic solvent and DPA layer are likely worn away during the contact, thereby allowing more intimate contact between the solid and liquid metals. It should be noted that the less abrasive contact mode Touch and

Retract also resulted in an observable “footprint” over multiple cycles, as shown in Figure 5.3, however this could be a result of removing organic residue from the surface (DPA or advantageous carbon). It is also possible that minor perturbations and lateral movement of the

EGaIn in the Touch and Retract contact mode could result in strain on the liquid metal at the contact area and exposure of the core metal alloy to the copper surface. However, if this occurs, it would be at a much lower frequency the less abrasive contact likely results in a more robust organic barrier to create a more resilient contact, as evidenced above in Figure 5.3.

Figure 5.5. Effect of applied voltage. a) Test setup and schematic of contact resistance test. b)

Resistance vs. cycles with voltage continuously applied and image of substrate after 1000 cycles (right).

c) Resistance vs. time with voltage continuously applied and image of substrate after 2500 seconds

(right). d) Resistance vs. cycles with voltage controlled to only be applied during contact and image of

substrate after 1000 cycles (right).

The chemical reduction of the copper oxide by EGaIn leads to the possibility that there could be additional electrical effects of the contact on the interface behavior. For this reason, contact resistance tests were performed for 1000 cycles on the DPA pretreated substrates in 1 mM DPA-ethanol using the Touch and Retract contact mode. The test setup is shown in Figure

5.5a. Figure 5.5b shows the contact resistance when the voltage was continuously applied

during contact cycling. The resistance limit of the test was reached within the first few contacts.

The resulting damage to the substrate after 1000 cycles is shown in the image in Figure 5.5b.

EDS of the damaged area showed a large increase in carbon, oxygen, silicon, and phosphorous compared to the undamaged areas of the substrate (Figure 5.XXX). The contact resistance was then measured for a single contact over an extended period of time with no contact cycling, shown in Figure 5.5c. The resistance remained constant at 1 ± 0.5 Ohm for over the course of

40 minutes, with no adhesion to the surface after retracting the EGaIn and no visible damage to the substrate. From this it appears that it is not the presence of an applied voltage that leads to degraded contact resistance observed in Figure 5.5b but likely the effect of hot-switching in an organic medium (ethanol/DPA) that leads to arcing when the metals are in very close proximity but not touching. Hot-switching is often a source of contact degradation in MEMS devices41, and as is indicative here can be a source of degradation for liquid metal / solid metal contacts as well. To prove this out, the voltage was only applied during EGaIn-Cu contact for 1000 cycles, as shown in Figure 5.5d. The resistance shows a slight upward trend starting at 0.3 Ohms during the first 400-500 cycles which then levels off at approximately 0.5 Ohm. There was no observable damage or residue after the 1000 cycles.

5.4 Conclusions

In conclusion, the interface between a liquid metal eutectic gallium indium (EGaIn) and a solid copper electrode was characterized and failure modes were identified. Two modes of contact were identified, namely (1) Touch and Retract, where the oxide of the liquid metal alloy is not largely disturbed and (2) Expand and Retract, where the volume of the liquid contact and interfacial contact area is increased. In most instances, Expand and Retract contact mode resulted in a significant damage to any organic barrier layer between the electrodes, which allowed reduction of the passivating copper oxide on the electrode surface, as measured through spectroscopy. The mechanism is postulated to be due to the strongly negative redox

potential of gallium which reduces the copper oxide by stripping the oxygen from the copper.

The effectiveness of common mitigation strategies to poor solid/liquid metal contact were explored, specifically contact within a polar organic solvent (ethanol) and the inclusion of an oxide binding ligand, decyl phosphonic acid (DPA). It was also shown that the application of a voltage while approaching contact (as opposed to waiting for physical contact to be established) can be detrimental to stripping and deteriorating the organic barrier layer between the solid electrode and the liquid metal. Ultimately, it was found that Touch and Retract mode imparts the least amount of abrasion at the liquid/solid metal interface and leads to prolonged operation, while avoiding hot-switching is preferred to increase the efficacy of the organic barriers up to

1000 cycles. These results may be helpful for other liquid / solid metal interfaces as well, which should improve the lifetime of reconfigurable liquid metal alloy devices in the future.

5.5 Supplemental

Both secondary electron images (SEI) and backscatter electron images (BSE) were taken of the copper surfaces as shown in Figure 5.6. Contrast can be seen in the SEI image

(Figure 5.6.a) as well as some gallium residue along the edge of that contact area. No contrast can be seen in the BSE image for the contact area, except for the gallium residue left at the edge.

Figure 5.6. SEM of the copper surface after one expand and retract contact in ethanol. a) SEI images show a dark footprint where the contact was made whereas BSE images do not show any contrast where

the contact was made.

Full survey spectra XPS of both touch and expand in DPA ethanol are shown in Figure

5.7. No gallium or gallium oxide peaks were observed (1116-1118 eV). It is important to note that the areas scanned were near the center of contact so although gallium residue was visible in the SEM images mostly around the edge of the contacts there was no continuous transfer of gallium or gallium oxide to the surface.

Figure 5.7. XPS survey spectra of the copper surface after an a) touch and retract contact in DPA- ethanol and b) an expand and retract contact in DPA-ethanol. No gallium (~1116 eV) or gallium oxide

(~1118 eV) peaks were observed in either case.

SEI images and energy dispersive x-ray spectroscopy (EDS) of damaged areas after

1000 cycles with continuous applied voltage are shown in Figure 5.8. A large area of visible damage was seen on the surface on and around the contact area. Elemental analysis showed that the damaged area mostly consisted of carbon and also had a higher amount of phosphorous than we typically could detect from our prepared samples.

Figure 5.8. a) Secondary electron images of a touch and retract contact in DPA-ethanol after 1000 cycles

with continuous applied voltage. The red circle shows the approximate area of EDS analysis. b) EDS of

the damaged area around the contact area showed high levels of carbon and also copper, oxygen, and

phosphorous.

5.6 References

1. Dickey, M. D. Emerging Applications of Liquid Metals Featuring Surface Oxides. ACS Appl. Mater. Interfaces 6, 18369–18379 (2014).

2. Dickey, M. D.; Chiechi, R. C.; Larsen, R. J.; Weiss, E. a.; Weitz, D. a.; Whitesides, G. M. Eutectic Gallium-Indium (EGaIn): A Liquid Metal Alloy for the Formation of Stable Structures in Microchannels at Room Temperature. Adv. Funct. Mater. 18, 1097–1104 (2008). 3. Boley, J. W.; White, E. L.; Chiu, G. T.-C.; Kramer, R. K. Direct Writing of Gallium-Indium Alloy for Stretchable Electronics. Adv. Funct. Mater. 24, 3501–3507 (2014). 4. Cheng, S.; Wu, Z. Microfluidic Electronics. Lab Chip 12 (16), 2782 (2012). 5. Chortos, A.; Lim, J.; To, J. W. F.; Vosgueritchian, M.; Dusseault, T. J.; Kim, T.-H.; Hwang, S.; Bao, Z. Highly Stretchable Transistors Using a Microcracked Organic Semiconductor. Adv. Mater. 26 (25), 4253–4259 (2014). 6. Dickey, M. D. Stretchable and Soft Electronics Using Liquid Metals. Adv. Mater. 1606425 (2017). 7. Kubo, M.; Li, X.; Kim, C.; Hashimoto, M.; Wiley, B. J.; Ham, D.; Whitesides, G. M. Stretchable Microfluidic Radiofrequency Antennas. Adv. Mater. 22 (25), 2749–2752 (2010). 8. Lu, Y.; Hu, Q.; Lin, Y.; Pacardo, D. B.; Wang, C.; Sun, W.; Ligler, F. S.; Dickey, M. D.; Gu, Z. Transformable Liquid-Metal Nanomedicine. Nat. Commun. 6 (1), 10066 (2015). 9. Joshipura, I. D.; Ayers, H. R.; Majidi, C.; Dickey, M. D. Methods to Pattern Liquid Metals. J. Mater. Chem. C 3 (16), 3834–3841 (2015). 10. Palleau, E.; Reece, S.; Desai, S. C.; Smith, M. E.; Dickey, M. D. Self-Healing Stretchable Wires for Reconfigurable Circuit Wiring and 3D Microfluidics. Adv. Mater. 25 (11), 1589– 1592 (2013). 11. Ladd, C.; So, J.-H.; Muth, J.; Dickey, M. D. 3D Printing of Free Standing Liquid Metal Microstructures. Adv. Mater. 25 (36), 5081–5085 (2013). 12. Ilyas, N.; Cook, A.; Tabor, C. E. Designing Liquid Metal Interfaces to Enable Next Generation Flexible and Reconfigurable Electronics. Adv. Mater. Interfaces 1700141 (2017). 13. Huang, Y.; Sai Sarathi Vasan, A.; Doraiswami, R.; Osterman, M.; Pecht, M. MEMS Reliability Review. IEEE Trans. Device Mater. Reliab. 12 (2), 482–493 (2012). 14. Chen, H.; Lu, B.-W.; Lin, Y.; Feng, X. Interfacial Failure in Flexible Electronic Devices. IEEE Electron Device Lett. 35 (2014). 15. Ilyas, N.; Butcher, D. P.; Durstock, M. F.; Tabor, C. E. Ion Exchange Membranes as an Interfacial Medium to Facilitate Gallium Liquid Metal Alloy Mobility. Adv. Mater. Interfaces 1500665 (2016). 16. Li, G.; Parmar, M.; Kim, D.; Lee, J.-B. J. B.; Lee, D.-W. PDMS Based Coplanar Microfluidic Channels for the Surface Reduction of Oxidized Galinstan. Lab Chip 14 (1), 200–209 (2014). 17. Koo, C.; Leblanc, B. E.; Kelley, M.; Member, S.; Fitzgerald, H. E.; Huff, G. H.; Member, S. Manipulating Liquid Metal Droplets in Microfluidic Channels With Minimized Skin Residues Toward Tunable RF Applications. J. Microelectromechanical Syst. 1–8 (2014).

18. Sen, P.; Kim, C. C. J.; Microelectromechanical, A. Microscale Liquid-Metal Switches — A Review. IEEE Trans. Ind. Electron. 56 (4), 1314–1330 (2009). 19. Koo, C.; LeBlanc, B. E.; Kelley, M.; Fitzgerald, H. E.; Huff, G. H.; Han, A. Manipulating Liquid Metal Droplets in Microfluidic Channels With Minimized Skin Residues Toward Tunable RF Applications. J. Microelectromechanical Syst. 24 (4), 1069–1076 (2015). 20. Mohammed, M.; Sundaresan, R.; Dickey, M. D. Self-Running Liquid Metal Drops That Delaminate Metal Films at Record Velocities. ACS Appl. Mater. Interfaces 7, 23163– 23171 (2015). 21. Secor, E. B.; Cook, A. B.; Tabor, C. E.; Hersam, M. C. Wiring up Liquid Metal: Stable and Robust Electrical Contacts Enabled by Printable Graphene Inks. Adv. Electron. Mater. 1700483 (2018). 22. Sen, P.; Kim, C. A Fast Liquid-Metal Droplet Microswitch Using EWOD-Driven Contact- Line Sliding. J. Microelectromechanical Syst. 18 (1), 174–185 (2009). 23. Kim, J.; Shen, W.; Latorre, L.; Kim, C.-J. A Micromechanical Switch with Electrostatically Driven Liquid-Metal Droplet. Sensors Actuators A Phys. 97–98, 672–679 (2002). 24. Bilodeau, R. A.; Zemlyanov, D. Y.; Kramer, R. K. Liquid Metal Switches for Environmentally Responsive Electronics. Adv. Mater. Interfaces 4 (2017). 25. Wissman, J.; Dickey, M. D.; Majidi, C. Field-Controlled Electrical Switch with Liquid Metal. Adv. Sci. 4, 1700169 (2017). 26. Ozutemiz, K. B.; Wissman, J.; Ozdoganlar, O. B.; Majidi, C. EGaIn Metal Interfacing for Liquid Metal Circuitry and Microelectronics Integration.Pdf. Adv. Mater. Interfaces 5 (2018). 27. Cumby, B. L.; Hayes, G. J.; Dickey, M. D.; Justice, R. S.; Tabor, C. E.; Heikenfeld, J. C. Reconfigurable Liquid Metal Circuits by Laplace Pressure Shaping. Appl. Phys. Lett. 101 (17), 174102 (2012). 28. Tabor, C.; Cumby, B.; Heikenfeld, J.; Mast, D.; Dickey, M. Robust Pressure-Actuated Liquid Metal Devices Showing Reconfigurable Electromagnetic Effects at GHz Frequencies. IEEE Trans. Microw. Theory Tech. 63 (10), 3122–3130 (2015). 29. Diebold, A. V; Watson, A. M.; Holcomb, S.; Tabor, C.; Mast, D.; Dickey, M. D.; Heikenfeld, J. Electrowetting-Actuated Liquid Metal for RF Applications. J. Micromechanics Microengineering 27 (2), 025010 (2017). 30. Tabor, C.; Champion, M.; Cumby, B. Materials for Liquid RF Electronics: Long Term Operation of Gallium Liquid Metal Alloys in Reconfigurable RF Applciations. In 2017 IEEE International Symposium on Antennas and Propagation & USNC/URSI National Radio Science Meeting; IEEE, 409–410 (2017). 31. Harada, S.; Arie, T.; Akita, S.; Takei, K. Highly Stable Liquid-Solid Metal Contact Toward Multilayered Detachable Flexible Devices. Adv. Electron. Mater. 1 (7), 1500080 (2015). 32. Ahlberg, P.; Jeong, S. H.; Jiao, M.; Wu, Z.; Jansson, U.; Zhang, S.; Zhang, Z. Graphene as a Diffusion Barrier in Galinstan-Solid Metal Contacts. IEEE Trans. Electron Devices 1–5 (2014).

33. Khan, M. R.; Trlica, C.; So, J.; Valeri, M.; Dickey, M. D. Influence of Water on the Interfacial Behavior of Gallium Liquid Metal Alloys. ACS Appl. Mater. Interfaces 6, 22467–22473 (2014). 34. Reus, W. F.; Thuo, M. M.; Shapiro, N. D.; Nijhuis, C. A.; Whitesides, G. M. The SAM, Not the Electrodes, Dominates Charge Transport in Metal-Monolayer//Ga 2O 3/Gallium- Indium Eutectic Junctions. ACS Nano. 6 (6), 4806–4822 (2012). 35. Cademartiri, L.; Thuo, M. M.; Nijhuis, C. A.; Reus, W. F.; Tricard, S.; Barber, J. R.; Sodhi, R. N. S.; Brodersen, P.; Kim, C.; Chiechi, R. C.; et al. Electrical Resistance of AgTS-S(CH2)n-1CH3//Ga2O3/EGaIn Tunneling Junctions. J. Phys. Chem. C 116, 10848–10860 (2012). 36. Farrell, Z. J.; Tabor, C. Control of Gallium Oxide Growth on Liquid Metal Eutectic Gallium/Indium Nanoparticles via Thiolation. Langmuir 34 (1), 234–240 (2018). 37. Fonder, G.; Minet, I.; Volcke, C.; Devillers, S.; Delhalle, J.; Mekhalif, Z. Anchoring of Alkylphosphonic Derivatives Molecules on Copper Oxide Surfaces. Appl. Surf. Sci. 257 (14), 6300–6307 (2011). 38. Biesinger, M. C. Advanced Analysis of Copper X-Ray Photoelectron Spectra. Surf. Interface Anal. 2 49, 1325–1334 (2017). 39. Doudrick, K.; Liu, S.; Mutunga, E. M.; Klein, K. L.; Damle, V.; Varanasi, K. K.; Rykaczewski, K. Different Shades of Oxide: From Nanoscale Wetting Mechanisms to Contact Printing of Gallium-Based Liquid Metals. Langmuir 30 (23), 6867–6877 (2014). 40. Bard, A. J.; Parsons, R.; Jordan, J.; International Union of Pure and Applied Chemistry. Standard Potentials in Aqueous Solution; (M. Dekker, 1985). 41. Dickrell, D. J.; Dugger, M. T. Silicone Oil Contamination and Electrical Contact Resistance Degradation of Low-Force Gold Contacts. J. Microelectromechanical Syst. 16 (1), 24–28 (2007).

Chapter 6: Conclusions and Future Outlook

This chapter concludes the dissertation by summarizing the results achieved and discussing the future outlook of liquid metal electronics. The two phases of the research plan - namely the investigation of interface interactions in reconfigurable GaLMA contacts with either dielectric materials or solid metal electrodes - are reviewed.

6.1 Reconfigurable GaLMA Contacts with Dielectric Materials

The oxide skin, which forms instantaneously at ambient conditions, is a major challenge in working with reconfigurable GaLMA devices and applications1-3. Removal of the oxide is common in order to allow shape change and prevent sticking to surfaces; however the removal processes involve surrounding the GaLMA in a conductive fluid4-7 or producing water and other byproducts at the interface8,9. Oxide removal while in a completely insulating fluid enables electrowetting, an extremely useful actuation method in microfluidics10,11.

6.1.1 Results Achieved

This dissertation presented the characterization and demonstration of a novel acidified siloxane that contains hydrochloric acid, which remains insulating due to the unique associative incorporation of the acids in the siloxane.

A procedure for producing the acidic oil, which is easily scalable and repeatable in any laboratory environment, has been detailed. It was found that the siloxane oil is highly hygroscopic and will absorb water into the bulk from the processing steps as well as the oxide removal reaction. The concentration of acid incorporating into the oil was quantified as approximately 1.5 M per our process of 18 hours incubation, which can then be diluted to the desired lower concentration. Increasing incubation times by roughly 400% resulted in a 25% increase in concentration but also hydrolysis of the siloxane oil. This process was also

compatible with another acid (HBr) but required different processing times for which optimizations were not further explored.

High frequency (2-20 GHz) dielectric characterizations of the acidic oil showed that it was essentially identical to the original oil (both highly insulating). The low frequency current- voltage response of the HCl and regular siloxane oils behaved identically to the regular siloxane at lower voltages up until the breakdown voltage was reached. We hypothesize that this is due to the Wein effect: the electric field induces greater alignment of the siloxane and the acid molecules along the electric field.

Electrowetting GaLMA in the acidic oil was further proof of the insulating behavior of the oil itself, as well as its ability to remove conductive byproducts from the interface. It was found that electrowetting failed if either the oxide was present (preventing contact angle rebounding) or water byproducts were able to accumulate on the surface (conductive phase electrowets instead of GaLMA). The electrowetting curves showed that acid concentrations below 1.5 M resulted in reversible contact angle changes from 165° to 110°. The change in contact angle was utilized along with Laplace pressure barriers in the design of a switchable polarizer device.

The switchable GaLMA polarizer was shown to perform similar to a copper device when on and similar to air (no device) when in the off state.

These results show that this novel siloxane is appropriate for even the most electrically challenging applications such as electrowetting (requiring high electric fields and therefore near perfect electrical insulation). The first aim of the research work, which consisted of creating an insulating fluid that removed the oxide and demonstrating a switchable electrowetting device, has thus been achieved.

6.1.2 Future Outlook

A breakthrough in non-toxic (GaLMA) liquid metal reconfigurable devices was achieved through development of the acidified siloxane, which has numerous advantages over other

methods being explored to remove/prevent oxide without interfering with device functionality.

Laboratories and fabrication lines can now use GaLMAs to fabricate devices without resorting to use of cumbersome ultra-low oxygen environments. Furthermore this work represents a first method to make long-lasting functional devices, because even sealed devices will allow some minute oxygen entry over time which will cause GaLMA oxidation. As a result, GaLMA is now more practical for a variety of applications. We believe our novel acidified siloxane can also be used in a many broader applications requiring the presence of acid in an insulating aprotic environment. The siloxane itself has been shown to eliminate oxides as well as solubilize and incorporate the metal ions into the liquid with minimal impact on the insulating properties.

6.2 Reconfigurable GaLMA Contacts with Solid Metals

Reconfigurable contacts between GaLMA and other metal electrodes are one of the main challenges that have to be overcome to utilize the unique advantages of liquid metal devices12,13. Extensive research efforts have been focused on permanent contacts with metals, taking advantage of sticking or alloying14,15. The failure and reliability of reversible contacts is often overlooked or fixed with short-term measures such as lossy barrier layers16,17. Phosphonic acid surface modification show promise as barrier layers which prevent sticking and also have low contact resistance18-20.

6.2.1 Results Achieved

This dissertation presented the surface effects of reversible electrical contacts between

GaLMA and solid copper electrodes. The external environment, PA surface modification, and mode of contact have been compared. Furthermore, surface damage to the copper electrode and failure mechanisms for multiple cycle contacts are discussed as well as the effect of a single contact cycle.

It was found that direct Ga metal contact to the electrodes, as when the GaLMA is expanded on the surface, is damaging and therefore not suitable for repeated cycles. Instead, the barrier of native Ga oxide, as when the GaLMA is gently touched to the surface, is critical for protecting the electrode surface. A combination of ethanol and phosphonic acids was found to further protect the interface during repeated cycles.

Failure of the contact cycling tests was observed as either 1) sticking of the GaLMA to the copper surface or 2) an increase in resistance by 2 orders of magnitude or more. Tests in air all resulted in sticking failure in 1-30 contact cycles. Tests in ethanol or 1 mM DPA-ethanol solution failed electrically and caused visible damage to the substrate at the area of contact over longer testing of 1000 cycles. It was found that the substrate damage was worsened by PAs and was not seen in contact tests with no applied voltage. Furthermore, voltage versus time tests (no contact cycles) showed electrical failure with no PAs but showed consistent resistance of approximately 1 Ohm for the pre-treated sample in PA-ethanol solution.

Better performance of up to 1000 cycles was achieved for the PA pre-treated sample in

PA-ethanol solution by controlling the voltage to only be applied for measurement during

GaLMA-Cu contact versus being constantly applied during cycling. The resistance was shown to be quite consistent for 1000 cycles at approximately 0.5 Ohm.

These results advance the understanding of factors affecting the reliability of reversible

GaLMA-metal contacts, which meets the requirement of the second research aim.

6.2.2 Future Outlook

GaLMA’s affinity for sticking to other metallic surfaces makes reaching tens or thousands of cycles already quite challenging, however most practical applications require operation in the millions of cycles. Future work is still needed to extend the testing cycles well beyond 1000 as well as exploring different types of liquid environments and surface bound molecules to optimize performance. The systematic approach detailed in this dissertation lays

the groundwork for future studies to expand upon the variables and methods in need of exploration. Once all of these variables’ effects on the interface are better understood they will then need to be integrated together into a device which can control the movement and the surface of the GaLMA.

6.3 Overall Outlook for the Future

This thesis established a first complete picture of the implications of materials choices on performance of gallium liquid metal electronics. Interfaces of GaLMAs have been shown to have a huge influence on behavior as well as reliability of liquid metal devices and therefore considerations of the full materials systems must be made for each application. This is especially true for reconfigurable GaLMA applications which were the main focus of this thesis.

Future work should use the results presented herein, to launch longer term reliability studies and further optimize materials choices for given reconfigurable applications. GaLMAs are unique materials with a wide range of capabilities that are often controlled by their challenging interfaces. Overall, this dissertation has clearly advanced both understanding of, and the capability for, reliable and robust liquid metal electronic systems.

6.4 References

1. Palleau, E., Reece, S., Desai, S. C., Smith, M. E. & Dickey, M. D. Self-healing stretchable wires for reconfigurable circuit wiring and 3D microfluidics. Adv. Mater. 25, 1589–92 (2013). 2. Koo, C. et al. Manipulating Liquid Metal Droplets in Microfluidic Channels With Minimized Skin Residues Toward Tunable RF Applications. J. Microelectromech. Syst. 1–8 (2014). 3. Aïssa, B. et al. Fluidic patch antenna based on liquid metal alloy/single-wall carbon- nanotubes operating at the S-band frequency. Appl. Phys. Lett. 103, 063101 (2013). 4. Cumby, B. L. et al. Reconfigurable liquid metal circuits by Laplace pressure shaping. Appl. Phys. Lett. 101, 174102 (2012).

5. Kim, D. et al. Recovery of Nonwetting Characteristics by Surface Modification of Gallium-Based Liquid Metal Droplets Using Hydrochloric Acid Vapor. ACS Appl. Mater. Interfaces 179–185 (2013). 6. Khan, M. R., Eaker, C. B., Bowden, E. F. & Dickey, M. D. Giant and switchable surface activity of liquid metal via surface oxidation. Proc. Natl. Acad. Sci. 111, 14047–14051 (2014). 7. Wang, M., Trlica, C., Khan, M. R., Dickey, M. D. & Adams, J. J. A reconfigurable liquid metal antenna driven by electrochemically controlled capillarity. J. Appl. Phys. 117, 194901 (2015). 8. Ilyas, N., Butcher, D. P., Durstock, M. F. & Tabor, C. E. Ion Exchange Membranes as an Interfacial Medium to Facilitate Gallium Liquid Metal Alloy Mobility. Adv. Mater. Interfaces (2016). 9. Doudrick, K.; Liu, S.; Mutunga, E. M.; Klein, K. L.; Damle, V.; Varanasi, K. K.; Rykaczewski, K. Different Shades of Oxide: From Nanoscale Wetting Mechanisms to Contact Printing of Gallium-Based Liquid Metals. Langmuir. 30 (23), 6867–6877 (2014). 10. Mugele, F. & Baret, J.-C. Electrowetting: from basics to applications. J. Phys. Condens. Matter 17, R705–R774 (2005). 11. Diebold, A.; Watson, A.; Holcomb, S.; Tabor, C; Mast, D.; Heikenfeld, J. Electrowetting- Actuated Liquid Metal for RF Applications. J. Micromech. Microeng. (27), 025010 (2017). 12. Dickey, M. D. Emerging Applications of Liquid Metals Featuring Surface Oxides. ACS Appl. Mater. Interfaces (2014). 13. Sen, P.; Kim, C. A Fast Liquid-Metal Droplet Microswitch Using EWOD-Driven Contact- Line Sliding. J. Microelectromechanical Syst. 18 (1), 174–185 (2009). 14. Sen, P.; Kim, C. C. J.; Microelectromechanical, A. Microscale Liquid-Metal Switches — A Review. IEEE Trans. Ind. Electron. 56 (4), 1314–1330 (2009). 15. Ahlberg, P.; Jeong, S. H.; Jiao, M.; Wu, Z.; Jansson, U.; Zhang, S.; Zhang, Z. Graphene as a Diffusion Barrier in Galinstan-Solid Metal Contacts. IEEE Trans. Electron Devices. 1–5 (2014). 16. Secor, E. B.; Cook, A. B.; Tabor, C. E.; Hersam, M. C. Wiring up Liquid Metal: Stable and Robust Electrical Contacts Enabled by Printable Graphene Inks. Adv. Electron. Mater. (2017). 17. Cumby, B.; Mast, D.; Tabor, C; Dickey, M.; Heikenfeld, J. Robust Pressure-Actuated Liquid Metal Devices Showing Reconfigurable Electromagnetic Efects at GHz Frequencies. IEEE Trans. Microw. Theory Tech. 63 (10), 3122-3130 (2015). 18. Reus, W. F.; Thuo, M. M.; Shapiro, N. D.; Nijhuis, C. A.; Whitesides, G. M. The SAM, Not the Electrodes, Dominates Charge Transport in Metal-Monolayer//Ga 2O 3/Gallium- Indium Eutectic Junctions. ACS Nano. 6 (6), 4806-4822 (2012). 19. Ilyas, N.; Cook, A.; Tabor, C. E. Designing Liquid Metal Interfaces to Enable Next Generation Flexible and Reconfigurable Electronics. Adv. Mater. Interfaces. 1700141, 1700141 (2017).

20. Paniagua, S. A.; Giordano, A. J.; Smith, O. L.; Barlow, S.; Li, H.; Armstrong, N. R.; Pemberton, J. E.; Bredas, J. L.; Ginger, D.; Marder, S. R. Phosphonic Acids for Interfacial Engineering of Transparent Conductive Oxides. Chem. Rev. 116, 7117-7158 (2016).