Heterogeneous Integration Strategy for Obtaining Physically Flexible 3D Compliant Electronic Systems

Dissertation by

Sohail Faizan Shaikh

In Partial Fulfillment of the Requirements

For the Degree of

Doctor of Philosophy

King Abdullah University of Science and Technology

Thuwal, Kingdom of Saudi Arabia

March, 2020 2

EXAMINATION COMMITTEE PAGE

Committee Chairperson: Professor Muhammad Mustafa Hussain Committee Members: Prof. Carlos M. Duarte, Prof. Ahmed M. Eltawil, Prof. Zhenqiang (Jack) Ma (External)

Sohail Faizan Shaikh

ABSTRACT

Heterogeneous Integration Strategy for Obtaining Physically Flexible 3D Compliant Electronic Systems

Sohail Faizan Shaikh

Electronic devices today are an integral part of human life thanks to state-of-the- art complementary metal oxide semiconductor (CMOS) technology. The progress in this area can be attributed to miniaturization driven by Moore’s Law. Further advancements in electronics are under threat from physical limits in dimensional scaling and hence new roadmaps for alternative materials and technologies are chased. Furthermore, the current era of Internet of things (IoT) and Internet of everything (IoE) has broaden the horizon to a plethora of unprecedented applications. The most prominent emerging fields are flexible and stretchable electronics. There has been significant progress in developments of flexible sensors, transistors, and alternative materials, etc. Nonetheless, there remains the unaddressed challenges of matching performance of the status-quo, packaging, interconnects, and lack of pragmatic integration schemes to readily complement existing state-of-the-art technology.

In this thesis, a pragmatic heterogeneous integration strategy is presented to obtain high-performance 3D electronic systems using existing CMOS based integrated circuit (IC). Critical challenges addressed during the process are: reliable flexible interconnects, maximum area efficiency, soft-polymeric packaging, and heterogeneous integration compatible with current CMOS technology. 5

First, a modular LEGO approach presents a novel method to obtain flexible electronics in a lock-and-key plug and play manner with reliable interconnects. A process of converting standard rigid IC into flexible LEGO without any performance degradation with a high-yield is shown.

For the majority of healthcare and other monitoring applications in IoT, sensory array is used for continuous monitoring and spatiotemporal mapping activities.

Here we present ultra-high-density sensory solution (1 million sensors) as an epitome of density and address each of the associated challenges.

A generic heterogeneous integration scheme has been presented to obtain physically flexible standalone electronic system using 3D-coin architecture. This

3D-coin architecture hosts sensors on one side, readout circuit and data processing units embedded in the polymer, and the other side is reserved for antenna and energy harvester (photovoltaic). This thin platform (~ 300 µm) has achieved bending radius of 1 mm while maintaining reliable electrical interconnection using through-polymer-via (TPV) and soft-polymeric encapsulation. This coin integration scheme is compatible with existing CMOS technology and suitable for large scale manufacturing.

Lastly, a featherlight non-invasive ‘Marine-Skin’ platform to monitor deep-ocean monitoring is presented using the heterogeneous integration scheme. Electrical and mechanical characterization has been done to establish reliability, integrity, robustness, and ruggedness of the processes, sensors, and multisensory flexible system. 6

ACKNOWLEDGMENTS

All praises, glory, and thanks to the Almighty for blessing me with the ability to fulfill

His pre-destined will, and providing the opportunity, resources, and everything to accomplish this doctorate. This thesis is dedicated to my family.

I express my deepest respect and gratitude to my advisor for life, Prof.

Muhammad Mustafa Hussain, for providing me with the opportunity to work with him. Ever since our first interaction, he has been a great advisor and motivator for my academic as well as personal life. He stood as a beacon of light through thick and thin during graduate studies and provided amazing ideas and enough space for independent research. He has provided the best possible professional environment, resources, and scholarly encouragement while at the same time acted as a compassionate, caring, and leading family head. I am proud to have a mentor for life in his stature.

I am grateful to Prof. Ahmed Eltawil, Prof. Carlos M. Duarte, and Prof. Zhenqiang

(Jack) Ma for serving in my dissertation committee, reviewing the thesis, and providing timely constructive feedback.

My heartiest love and countless thanks goes to my parents, siblings, in-laws, all family members, and, very special gratitude to my beautiful, understanding, and lovely wife Asfiya and my son Aabaan, for their consistent motivation, moral support, unconditional love, care and prayers. I couldn’t have imagined to achieve anything had they not shown confidence and strengthened me. 7

I express special thanks to my friend-cum-brother Prof. Mohammad Akram, his family for motivating, supporting, and guiding me personally throughout with his valuable life lessons. I can’t thank enough all my friends, colleagues (special mention to Melvin) and core lab members (Syed Ahad & his team) for their direct and indirect valuable contributions in my time as student.

TABLE OF CONTENTS

EXAMINATION COMMITTEE PAGE ...... 2 COPYRIGHT PAGE ...... 3 ABSTRACT ...... 4 ACKNOWLEDGMENTS ...... 6 TABLE OF CONTENTS ...... 8 LIST OF ABBREVIATIONS ...... 11 LIST OF ILLUSTRATIONS ...... 13 LIST OF TABLES ...... 15 LIST OF PUBLICATIONS ...... 16 1. Chapter 1: Introduction ...... 20 1.1. Motivation ...... 20 1.2. Moore’s Law and End of Scaling? ...... 21 1.3. Need for Physically Compliant Electronics ...... 23 1.4. Why CMOS technology? ...... 24 1.4.1. Case A- Biological Signals Processing ...... 26 1.4.2. Case B- Continuous Health Monitoring ...... 26 1.4.3. Case C- Increasing Form Factor...... 27 1.5. Dissertation Overview ...... 29 2. Chapter 2: Modular LEGO Electronics...... 32 2.1. Introduction ...... 32 2.2. Flexible LEGO Concept and Fabrication ...... 36 2.2.1. Modular Assembly for Logic Gates ...... 37 2.2.2. Fabrication of Flexible LEGO modules ...... 40 2.2.3. Contact Reliability and Assembly Yield ...... 42 2.3. Transforming IC into flexible LEGO ...... 45 2.3.1. Rigid IC to LEGO Conversion Flow ...... 46 2.3.2. The Comparator Configuration ...... 49 2.4. Conclusion ...... 52 3. Chapter 3: High-Density (1 Million) Sensors and System Integration ...... 54 9

3.1. Introduction ...... 56 3.2. Sensor Design and Interface ...... 59 3.3. Fabrication Process Flow ...... 61 3.4. Sensory System Performance ...... 64 3.4.1. Individual Sensor Characterization ...... 65 3.4.2. Spatial Resolution ...... 67 3.5. Mapping of 1 Million Sensors ...... 69 3.6. Cellular Growth and Pattern Mapping ...... 74 3.7. Scalability and Yield ...... 76 3.8. Conclusion ...... 78 4. Chapter 4: Heterogeneous 3D Integration Strategies (3D Coin) ...... 81 4.1. Overview of System Integration ...... 82 4.1.1. 3D Integration ...... 82 4.1.1.1. Advantages of 3D Wafer Stacking ...... 83 4.1.2. Flexible Interconnects and Packaging ...... 84 4.1.3. Heterogeneous Integration ...... 85 4.2. Challenges ...... 86 4.3. 3D Coin Architecture ...... 88 4.3.1. Heterogeneous Integration Fabrication Scheme ...... 90 4.3.2. Modular Lego Assembly ...... 93 4.3.3. TPV, Antenna, and Solar Cell ...... 95 4.3.4. Sensory System Integration...... 97 4.4. Results and Discussion ...... 97 4.4.1. Electrical Characterization ...... 97 4.4.2. Sensory System Characterization ...... 99 4.4.3. Process/Mechanical Induced Effects on Bare Dies ...... 102 4.5. Conclusion ...... 105 5. Chapter 5: An Application- (Non-invasive Featherlight Marine-Skin Tagging System) ...... 106 5.1. Introduction ...... 109 5.2. Challenges in the Marine Monitoring Systems ...... 112 5.3. Biocompatible Soft Packaging ...... 114 5.4. Multisensory Complaint Sensor Design ...... 115 10

5.5. Material Optimization ...... 117 5.5.1. Temperature Sensor ...... 117 5.5.2. Salinity Sensor ...... 120 5.5.3. Pressure (Depth) Sensor ...... 122 5.5.4. Numerical Finite Element Analysis ...... 124 5.6. Marine Skin Fabrication Process Scheme ...... 127 5.7. Overall Sensory System Performance ...... 131 5.7.1. Temperature and Salinity (Conductivity) ...... 131 5.7.2. Temperature and Salinity (Conductivity) ...... 133 5.8. Reliability, Ruggedness, and Mechanical Integrity ...... 135 5.8.1. Repeated Cyclic Bending Tests ...... 136 5.8.2. Prolonged Exposure to Saline Environment ...... 140 5.8.3. Non-invasive Attachment Mechanism ...... 141 5.8.3.1. Attachment on Large Species ...... 142 5.8.3.2. Attachment on Large Species ...... 143 5.9. Conclusion ...... 147 6. Chapter 6: Summary and Future Outlook ...... 149 6.1. Materials ...... 149 6.2. Design Strategies ...... 151 6.3. Integration Strategies ...... 152 6.4. Applications ...... 153 6.5. Challenges ...... 157 6.6. Future Outlook ...... 158 6.7. Conclusion ...... 159 REFERENCES ...... 160

LIST OF ABBREVIATIONS

CMOS Complementary Metal Oxide Semiconductor

DIY Do It Yourself

E-skin Electronic Skin

IoE Internet of Everything

IoT Internet of Things

PDMS Polydimethylsiloxane

PI Polyimide

SoC System-on-Chip

RTD Resistive Temperature Detector

AFM Atomic Force Microscopy

SEM Scanning Electron Microscopy

RIE Reactive Ion Etching

DRIE Deep Reactive Ion Etching

1D One-Dimensional

2D Two-Dimensional

IC Integrated Circuit

MOS Metal Oxide Semiconductor

MOSFET Metal Oxide Semiconductor Field Effect Transistor

ITRS International Technology Roadmap for Semiconductors

ULSI Ultra Large Scale Integration

GaAs Gallium Arsenide

TFET Thin Film Transistor 12

SS Subthreshold Slope

ECG Electrocardiogram

ECoG Electrocorticography

EEG Electroencephalogram

OPAMP Operational Amplifier

ECD Electrochemical Deposition

PR Photoresist

MEA Microelectrode Array

DIBL Drain Induced Barrier Lowering LED Light Emitting Diode OLED Organic Light Emitting Diode SSSDSA Sequential Shape and Solder Directed Self Assembly MIC Microwave Integrated Circuit MMIC Monolithically Microwave Integrated Circuit ITRS International Roadmap for Semicondutors MRI Magnetic Resonance Imaging TSV Through Silicon Via TPV Through Polymer Via SiP System in Package SEB Sequential Etch Back MCU Micro Controller Unit BLE Bluetooth Low Energy ECD Electrochemically Deposited

LIST OF ILLUSTRATIONS

Figure 1.1| Data and processing need for IoT based healthcare devices...... 28 Figure 2.1| The process of modular Lego assembly ...... 38 Figure 2.2| Performance of the assembled module...... 40 Figure 2.3| Fabrication schematics for flexible LEGO...... 42 Figure 2.4| Cross-sectional scanning electron microscopic (SEM) images...... 43 Figure 2.5| High yield robotic assembly process using pick-and-place ...... 45 Figure 2.6| Process flow for transforming rigid IC into flexible LEGO: ...... 48 Figure 2.7| Output characteristic waveforms of OPAMP at different process...... 51 Figure 2.8| Output characteristic waveforms of transformed Lego-...... 51 Figure 3.1| Conventional method and the vision of high-density cellular growth...... 56 Figure 3.2| 3D and cross-sectional 2D fabrication process flow...... 65 Figure 3.3| Physical and electrical characterization of fabricated ...... 66 Figure 3.4| Spatial resolution and sensitivity across channels...... 79 Figure 3.5| 1 Million pressure sensor mapping...... 73 Figure 3.6| Cellular growth activity monitoring...... 76 Figure 3.7| 1 Million pressure sensor mapping...... 78 Figure 4.1| Concept of 3D-coin architecture illustrating multilayered...... 89 Figure 4.2| A. Heterogeneous integration flow for obtaining 3D system...... 91 Figure 4.3| 3D coin electronic system illustration and bending...... 92 Figure 4.4| SEM of TPV to study the profile...... 93 Figure 4.5| SEM validating the PDMS as softmask and sidewall spacer...... 95 Figure 4.6| SEM images at different stages of integration scheme...... 96 Figure 4.7| Reliability of interconnects (impedance measurements)...... 98 Figure 4.8| Sensory system and heterogeneous integration on 3D-coin...... 100 Figure 4.9| System architecture for stand-alone electronic systems...... 101 Figure 4.10| Reliability of the integration process demonstrated by electrical,...... 103 Figure 4.11| System demonstration using temperature sensor, antenna,...... 104 Figure 5.1| Robust compliant Marine Skin illustration...... 112 Figure 5.2| Comparison of the performance of 2 different temperature...... 119 Figure 5.3| Comparison of the performance of two variations of salinity ...... 121 Figure 5.4| Dielectric material optimization for increasing sensitivity...... 123 14

Figure 5.5| FEM simulation studies to understand the stress,...... 127 Figure 5.6| 3D schematic process integration for the waterproof...... 130 Figure 5.7| Performance improvement of the wearable multisensory...... 132 Figure 5.8| Harsh marine environment real-time pressure recording ...... 135 Figure 5.9| Robustness and reliability of the devices for prolonged...... 138 Figure 5.10| Robustness of the pressure sensor characterized...... 139 Figure 5.11| Non-invasive attachment mechanism for tagging...... 145 Figure 5.12| Wearable bracelet designs and molds,...... 146

LIST OF TABLES

Table 3.1| 3D printer INL letter mold for pressure monitoring by using………………...….68 Table 3.2| Comparison of high-density electrode and related work with.……...…...... 72 Table 5.1| Comparison of Marine Skin and commercial products..…………….…...... 124

LIST OF PUBLICATIONS

1. Sohail F. Shaikh, Sherjeel M Khan, Muhammad M. Hussain, "3D-coin Heterogeneous Integration Strategy for Obtaining Truly Physically Flexible Standalone Electronic Systems.", Patent USPTO 0338-440- WO/2019-109-02 2. S. F. Shaikh, et al., “Non-invasive featherlight wearable compliant “Marine Skin” – standalone multi-sensory system for deep-sea environmental monitoring”, Small, 15, 1804385, 2019 [Cover Article] 3. Sohail F Shaikh, MT Ghoneim, RR Bahabry, SM Khan, MM Hussain, “Modular-Lego Electronics”. Adv. Mater. Tech. 2017. 4. Sohail F Shaikh, Mohamed T Ghoneim, Galo A Torres Sevilla, Joanna M Nassar, Aftab M Hussain, Muhammad M Hussain, “Freeform Compliant CMOS Electronic Systems for Internet of Everything Applications”. IEEE Trans. on Elect. Dev. 2017. 5. Sohail F. Shaikh, Muhammad M. Hussain, " Multisensory Graphene-Skin for Harsh-Environment Applications", Applied Physics Letters (Revision) 6. Sohail F. Shaikh, Nazek El-Atab, Muhammad M. Hussain, " A Flexible, High- Density 1 Million Pressure Sensory Array for in situ Pressure Mapping", ACS Nano (Under Review) 7. S. M. Khan, S. F. Shaikh, N. Qaiser, M. M. Hussain, “Flexible Lightweight CMOS Enabled Multisensory Platform for Plant Microclimate Monitoring”, IEEE Trans. Elect. Dev. 65(11), 5038 (2018) 8. N. Alcheikh*, S. F. Shaikh*, M. M. Hussain, “Ultra-Stretchable Archimedean Interconnects For Stretchable Electronics”, Extr. Mech. Lett. 24, 6, October 2018 9. N. Alcheikh, S. F. Shaikh, M. M. Hussain, “In-plane deformation mechanics of high stretchable archimedean interconnects”, AIP Adv. 2019 10. El‐Atab, N. , Shaikh, S. F., Khan, S. M. and Hussain, M. M. (2019), “Bi‐Facial Substrates Enabled Heterogeneous Multi‐Dimensional Integrated Circuits (MD‐IC) for Internet of Things (IoT) Applications”, Adv. Eng. Mater., 21: 1900043. 11. Nazek El-Atab, Sohail F. Shaikh, et. al. " A Flexible Nano-porous Template for the Design and Development of Reusable Anti-COVID-19 Hydrophobic Face Mask", ACS Nano 2020 12. R. B. Mishra, S. F. Shaikh, A. M. Hussain, M. M. Hussain, “Metal coated polymer and paper-based cantilever design and analysis for acoustic pressure sensing”, AIP Adv. 2020. 17

13. S. M. Khan, Sohail F. Shaikh, et. al. “Design and Analysis of Paper-Based Wearable Wireless Sensory System for Chronic Monitoring of Respiratory Function", IEEE Trans. on Elect. Dev. 2019 14. Almuslem, A. S., Shaikh, S. F., Hussain, M. M., “Flexible and Stretchable Electronics for Harsh‐Environmental Applications.” Adv. Mater. Technol. 2019 15. N. El-Atab, S. F. Shaikh, et. al., “Heterogeneous Cubic Multi-Dimensional Integrated Circuit (MD-IC) for Water and Food Security in Fish Farming Ponds”, Small 2020 16. N. Qaiser, A. N. Damdam, S. M. Khan, S. F. Shaikh, M. M. Hussain, “Design, Mechanics, and Operation of Spiral-Interconnect based Networked Sensor Array for Stretchable Electronics”, Appl. Phys. Lett. 2020 17. Nazek El-Atab, Sohail Faizan, M. M. Hussain, " Nano-scale transistors for interfacing with brain: design criteria, progress and prospect", Nanotechnology, 2019 (Review) 18. M. M. Hussain, Z. Ma, S. F. Shaikh, “Flexible and Stretchable Electronics – Progress, Challenges, and Prospects”, The Electrochem. Soc. Interface, Winter 2018, volume 27, issue 4, (Invited Review) [Cover Article] 19. J. M. Nassar, S. F. Shaikh, et al., "Compliant Lightweight Non-Invasive Standalone “Marine Skin” Tagging System", npj Flexible Electronics (2018). 20. S. M. Khan, S. F. Shaikh, et al., “Do-It-Yourself (DIY) Integration of Paper Sensor in a Smart Lid for Medication Adherence”, IoP Flexible and Printed Electronics, 2019 21. M. A. Karimi, Sohail F. Shaikh, et. al., " Flexible tag design for semi- continuous wireless data acquisition from marine animals”, Flex. Print. Electron. 2019 22. W. Park, S. Sohail, J. Min, S. Lee, B. H. Lee, M. M. Hussain*, "Contact Resistance Reduction of ZnO Thin Film Transistors (TFTs) with Saw- Shaped Electrode", nanotechnology, (2018) 23. W. Park, J. Min, S. Shaikh, M. M. Hussain, “Stable MoS2 FETs using TiO2 Interfacial Layer at Metal/MoS2 Contact”, physica solidi status A [Cover Article] 2018 24. G. A. Torres Sevilla, S. F. Shaikh, et al., “Fully Spherical Stretchable Silicon Photodiodes Array For Simultaneous 360 Imaging”, Appl. Phys. Lett. (2018) 25. R. R. Bahabry, S. F. Shaikh, et al., “Corrugation Architecture Enabled Ultra- Flexible Wafer-Scale High-Efficiency Mono-crystalline Silicon Solar Cell”, Adv. Energy Mater. 2017. 18

26. A. Gumus, A. Alam, Sohail F. Shaikh, et. al. “Expendable Polymer Enabled Wirelessly Destructible High Performance Solid State Electronics”, Adv. Mater. Tech. 1600264 (2017) 27. A. M. Hussain, S. F. Shaikh, M. M. Hussain, “Design criteria for XeF2 enabled deterministic transformation of bulk silicon (100) into flexible silicon layer”, AIP Adv. (2016) 28. N. Qaiser, S. F. Shaikh et. al. , “Mirror-Symmetry Controlled Mechanical Response of Interconnects for Stretchable Electronics”, Extreme Mech. Lett. 29. S. F. Shaikh, and M. M. Hussain, Design Criteria and Integration Strategy for Fully Compliant Standalone Large Area Electronics Systems in Elements in Flexible and Large-Area Electronics, Ed. R. Dahiya and L. Occhipinti (Cambridge University Press, 2019) (In Press). 30. Sohail F Shaikh, Implantable Electronics, in Handbook of Flexible and Stretchable Electronics, M. M. Hussain and N. El-Atab (CRC Press, 2019). 31. Sohail. F. Shaikh, Nazek Elatab, M. M. Hussain, " 3D Coin Heterogeneous Integration Strategy for Physically Compliant CMOS Electronic Systems”, DAC 2020(Submitted) 32. Sohail. F. Shaikh, Nadeem Qaiser, M. M. Hussain, " 3D Coin Integration for Realizing Next-Generation Flexible Electronic Systems”, IITC 2020. 33. N. El-Atab, R. Suwaidan, Y. Alghamdi, A. Alhazzany, R. Almansour, S. F. Shaikh, S. M. Khan, M. M. Hussain, “Multi-Dimensional Integrated Circuit For Water Security in Fish Farming Ponds”, 2019 IEEE Symposium on Technologies for Homeland Security 2019, November 2019, Waltham, MA, USA 34. N. El-Atab, S. F. Shaikh, S. Khan, J Yun, M. M. Hussain, “Heterogeneous Multi-Dimensional Integrated Circuit for Internet-of-Things Application”, IEEE S3S 2019, October 2019, San Jose, CA, USA 35. Sohail F. Shaikh, M. M. Hussain, “Marine IoT: Non-Invasive Wearable Multisensory Platform for Oceanic Environment Monitoring”, WF-IoT 2019, Ireland. 36. M. M. Hussain, N. Qaiser, A. Damdam, N. Alcheikh, S. F. Shaikh, “Design of stretchable electronics: gap mitigation, mirroring, and reconfiguration”, SPIE Defence + Commercial Sensing 2019, Baltimore. 37. Muhammad M Hussain, Sohail F Shaikh, Galo A Torres Sevilla, Joanna M Nassar, Aftab M Hussain, Rabab R Bahabry, Sherjeel M Khan, Arwa T Kutbee, Jhonathan P Rojas, Mohamed T Ghoneim, Melvin Cruz; “Manufacturable Heterogeneous Integration for Flexible CMOS Electronics”, 2018 76th Device Research Conference (DRC), 2018. 19

38. Muhammad Mustafa Hussain, Nazek Elatab, Amir N Hanna, Aftab M Hussain, Hossain M Fahad, Sohail Faizan Shaikh; “Homogeneous and Heterogeneous Material Based Nanotube Tunnel Field Effect Transistor with Core-Shell Gate Stacks”, Meeting Abstracts, The Electrochemical Society, 2018. 39. R. R. Bahabry, A. C. Sepulveda, A. T. Kutbee, S. F. Shaikh, M. M. Hussain, “Corrugation Architecture Enabled Ultra-Flexible Mono-Crystalline Silicon Solar Cells via Plasma Etching and Laser Ablation”, 7th World Conference on Photovoltaic Energy Conversion, Waikoloa, Hawaii, USA 10-15 June 2018 40. Hussain, M. M.; Nassar, J. M.; Khan, S. M.; Shaikh, S. F.; Sevilla, Galo T.; Kutbee, A. T.; Bahabry, R. R.; Babatain, W.; Muslem, A. S.; Nour, M. A.; Wicaksono, I.; Mishra, K.; Democratized electronics to enable smart living for all, IEEE Sensors 2017, Galsgow, UK. 29 Oct. – 1 Nov. 2017 20

1. Chapter 1: Introduction

1.1. Motivation

Electronics are an integral part of our daily digital life. We enjoy comfort, convenience and safety through a variety of electronic applications in the general area of computation, communication, infotainment, medical instrumentation, healthcare, automobiles, and smart cities. In 2005, the International

Telecommunication Union (ITU) suggested connecting all objects in a sensory and intelligent manner, coined as Internet of Things (IoT) and described four dimensions in IoT: tagging things (item identification), thinking things (embedded systems), feeling things (sensors and wireless sensor network), and shrinking things (nanotechnology) [1]–[3]. Internet of Everything (IoE) is the realization of interconnecting living beings (people, plants, animals, etc.) and non-living things

(data, process, devices, vehicles, etc.). The notion of IoT and IoE provides connectivity of anything-anytime-anyplace for more intelligent health, smarter energy-efficient cities, transportation, smart grids, etc.

Complementary metal oxide semiconductor (CMOS) technology has been instrumental in this rise of the digital age, with micro-and-nanofabrication processes being indisputably the most reliable and advanced existing technologies, which continue to be perfected. CMOS-enabled electronics are omnipresent in the technology domain on account of ultra-large-scale-integration

(ULSI) density, reliable device manufacturing, energy efficiency, and enhanced

21 performance per cost. In this chapter, a section dedicated to scaling (Moore’s law) that drove CMOS technology, physical limits, and the roadmap for the next generation of electronics to keep progressing at a similar rate is presented. It is followed by shortly addressing the field of flexible and stretchable electronics. A case is made for why CMOS technology is needed for coming generations of electronics. Next, different flexing strategies are provided. Lastly, the outline is provided for this dissertation, which is directed towards the challenges in this exciting field

1.2. Moore’s Law and End of Scaling?

In 1965, Gordon Moore predicted the exponential improvement in microprocessor performance. His prediction (Moore’s Law) of doubling the number of transistors (physical scaling) on any processing platform every 18-24 months has been the driving force behind the semiconductor industry. The idea is to scale down devices, allowing more device population in equal chip area, while keeping the same/improved performance without increasing the power consumption [4]–

[6]. The basic governing equations of the MOSFET for drain current, intrinsic delay

(delay of data propagation) and power consumption inherently signify the effect of scaling the devices for improved performances.

푊 ( )2 퐼푑 = ( 퐿 )퐶표푥µ 푉푑푑 − 푉푡 (1.1)

퐶 푉 푇 = 푔 푑푑 (1.2) 퐼푑

퐶 푉 2 푃 = 푔 푑푑 (1.3) 푑푒푛푠푖푡푦 푊퐿푇

Where Id is drain current of the transistor, Cox is the gate capacitance per unit area, μ is the mobility of channel material, Vdd is the supply voltage, Vt is the threshold voltage, W and L are the transistor gate width and length, respectively.

The gate capacitance, Cg, is the product of effective oxide capacitance and the area of the gate (W*L) and T is the intrinsic delay introduced by the CMOS gate.

Pdensity gives power dissipated per unit area of the chip.

Functionality and performance of a chip with the same given area can be improved by increasing the number of transistors by scaling the transistor physically. Stable MOSFET characteristics can be obtained if proportional scaling is applied and hence the scaling factor of s leads to the following modifications in the characteristics with modification in physical dimensions.

L → L/s , W → W/s , 푡표푥 → 푡표푥/s , 퐼푑 → 퐼푑. s

2 3 2 퐶푔 → 퐶푔/s , T → T/푠 , 푃푑푒푛푠푖푡푦 → 푃푑푒푛푠푖푡푦 푠 , DD → DD 푠

Where DD is the device density, representing a total number of devices per unit area of the chip. From the above scaling law, it can be seen that power dissipation density is increased by the cube of the scaling factor and that can lead to temperature rise in the chip. However, this issue is resolved by reducing the operating voltages in proportion, thus:

푉푑푑 → 푉푑푑/s , 푉푡 → 푉푡/s , 퐶푔 → 퐶푔/s , 퐼푑 → 퐼푑/s

2 2 T → 푇/s , P → P/s , 푃푑푒푛푠푖푡푦 → 푃푑푒푛푠푖푡푦 , 퐷퐷 → DD s

Current state-of-the-art CMOS technology has reached sub-10 nm. Further scaling will result in higher leakage from source to drain, velocity saturation, drain induced barrier lowering (DIBL) and hot carrier degeneration. These leakage currents imply higher off currents for a transistor thereby leading to significant increase in static power consumption. In addition, significant infrastructural changes in the existing CMOS process technology are needed for sub-7 nm. Thus, we can consider that the physical limit of transistor scaling has reached its end. It may be the end of the scaling era, however, the principal of Moore’s law remains to be quenched and the quest continues into novel materials, architectural changes, multidimensional stacking and process integration strategies to achieve the new performance heights for next generation applications guided by international technology roadmap for semiconductors (ITRS).

1.3. Need for Physically Compliant Electronics

Natural living beings are marvelous creations at the epitome of the most complex engineering and architecture, illustrating intricate asymmetric contours and soft skin and tissue surfaces. This design makes integration with symmetric and uniformly shaped electronics difficult. Consequently, it is critical to introduce a new dimension in electronics to transform into freeform physically flexible and stretchable electronics. Although IoE is still germinating in its early stages, we strongly believe that electronics will capitalize on flexibility and stretchability, incorporating interactive system integration of data acquisition (sensors) [7]–[9], data processing and decision making (logic circuits and elements), data storage

(memory devices) [10]–[15], communication (radio-frequency and Li-Fi elements)

[16]–[20], decision accomplishment (actuators) [21]–[24], and efficient power management (energy scavenging and energy storage) [25]–[28]. In real-time healthcare monitoring applications, electronic devices might be deployed round the clock on curvilinear body locations (joints, neck, shoulders, wrists, etc.) requiring constant twisting, bending, contracting, and stretching; offering opportunities to exploit flexibility and stretchability of user-friendly electronics [1],

[29]–[31]. Besides these external body parts, internal organs (heart, brain, kidney, etc.) introduce a different level of challenges (biocompatibility and biodegradability). Environmental aspects like marine monitoring or plant monitoring for agricultural productivity are applications beyond healthcare which necessitate capitalization of the physical flexibility, stretchability, and reconfigurability of electronic devices.

1.4. Why CMOS technology?

Electronics have become an indispensable part of our daily life, with constant CMOS scaling that leads to increased performances at lower costs and in constrained areas. Inarguably, silicon (Si) is the most affordable and abundant flagship material of state-of-the-art CMOS technology, leading the consumer electronics market for decades. Integration of microelectronics and nanoelectronics will be critical for achieving data-driven applications in healthcare monitoring (wearable and implantable electronics) and environmental applications.

The core components of any standalone electronic system incorporates: sensor,

25 data processing (logic), data storage (memory), data communication, and power management. There has been remarkable progress in the development of wearable and implantable flexible sensors for health monitoring devices. However, the demonstrations are mostly limited to flexible sensors, discrete devices, and at most a sensory read-out system.

Furthermore, several efforts have been made to improve human life using multi-functional gadgets for monitoring health conditions, for example electrocardiogram (ECG) [32]–[34], blood pressure [35], [36], and pulse oximeter

[37], [38], etc. Applications related to human central processing unit - the brain, include neural prosthetics (cochlea, vision, and brain-computer interface); diagnostics (electrocorticography-ECoG, electroencephalogram-EEG); treatment of neuro-degenerative diseases by neuro-stimulation and deep-brain stimulation.

Soft electrodes mimicking mechanical properties like soft tissues play a pivotal role in implantable monitoring applications. However, bulky and rigid base stations, wireless connectivity, reliability, sensitivity, and battery life are major downsides of available devices. Despite remarkable progress in material and fabrication development, we have not encountered (to the best of our knowledge) any fully flexible health monitoring system that has integrated circuitry for processing, data storage capability for handling big data, communications system, and power supply. A few case studies that emphasize immediate addressable challenges and needs in data storage, integration for higher performance, and energy management capacities are presented. These cases signify why CMOS

26 electronics is superior and must be at the core of any electronic system development rather than new organic or other new material based counter parts.

1.4.1. Case A- Biological Signals Processing

Biological signals are mostly low amplitude signals (µV to mV) in DC with very low frequencies (0-150 Hz for ECG and EEG, 10-200 Hz for EMG, and 0-200 Hz for sphygmomanometer), except speech signals which occur at higher frequencies

(0.1-4 kHz). The processing of these signals for extracting different parameters and conditions requires a minimum sampling at the Nyquist rate, though a higher rate provides better reliability and accuracy. Typical ECG signals are sampled at a minimum frequency of 500 Hz which corresponds to data generation of 1.8 MB per hour, accumulating into 1.2 GB per month, for a single subject. For increased accuracy and reliability, processing at higher sampling rates is desired and the corresponding humongous data that would be generated is depicted in Fig. 1 A.

1.4.2. Case B- Continuous Health Monitoring

For brain and neural mapping applications where even the simplest of brains have 10–1000s of neurons located in multiple locations (the cerebral cortex of a rat has 15-20 million neurons, a cat has approximately 300 million, a human has more than 50 billion neurons), required implantable electrodes will generate immense data. An electrical signal in the neuron is a result of electrical potential fluctuations, response to the fluctuations, and then generation of a required action; all these activities are repeated hundreds of time per second. For instance, recording the electrical activity of 200 electrodes each from 10 subjects which

27 equals ~1 million neurons, 1000 times per second would generate 1 GBps and 4

TB per hour, data that sharply reaches 100 TB per day. This big data generated, even if compressed by a factor of 10, would result in approximately 3 PB annually which is as much data as generated by the 17 mile long large Hadron Collider, most advanced astronomical observations, and other complex science projects [3],

[29].

1.4.3. Case C- Increasing Form Factor

Integration of more functionalities due to continuous transistor scaling have enabled ULSI densities for energy efficient high-performance computational processors and ICs. However, a higher demand for processing speed also means a larger cache memory, which implies increased form factor. For example, Intel

Atom Processor 330 operating at 1.6 GHz frequency (1 MB cache) occupies 484 mm2 area which increases to ~2362 mm2 (388% increase) for Core i7-5930K (15

MB cache) operating at 3.5 GHz (250% rise in frequency), the trend is plotted in

Fig. 1 B. Also, a similar pattern is depicted in form factor, frequency, and cache memory increment in processor variants charted in Fig. 1 C.

From perspectives of energy, wireless communications networks like

Bluetooth, Wi-Fi, and ZigBee, which are ubiquitous in portable electronic systems, need 10-100s of mW at peak power while an average power in 10s of µW to low mW at low duty cycle [39]. High energy demands for flexible electronics makes a solar harvester unsuitable due to area constraints, e.g. with a 100% charge efficiency, 100 mWh of storage charging for one hour would need a 1 m2 solar

28 panel. Directed efforts toward making robust, flexible, and low-cost energy harvesting and storage devices for intuitive truly flexible systems are critical.

Figure 1.1| Data and processing need for IoT based healthcare devices (a) Typical data generated from a real-time ECG monitoring system at different sample rate (500 Hz, 1 kHz, 2 kHz, 4 kHz and 44 kHz). (b) The increase in form-factor (area) and operating frequency as increase in cache memory shown in percentage change for different processors through years from Intel, and (c) consistent increasing trend in operating frequency, cache memory and form- factor for different generation of Intel processors as high computational demands rise.

CMOS-technology based FETs are cutting-edge devices and have established performance matrices over decades. The general trend of low mobility for thin film transistors (TFT) and organic transistors (<10 cm2V-1s-1) is observed in organic transistors while high flexibilities up to 100 µm bending radius and 104 endurance cycles are reported [40], [41]. Large operation voltages (on orders of 10 volts) and high SS restrict their practical implementation as energy efficient devices. Hybrid

FET devices where inorganic FETs are flexed and transferred onto flexible polymers, show better performances than organics, exhibiting reduced SS and

Vth, higher ION/IOFF, and greater mobility, compromising with lower flexibility [42],

[43]. Organic transistors have illustrious reports highlighting best individual performances for SS, minimum bending radius and high ION/IOFF ratios.

However, improving one metric adversely affects another: e.g. increment in gate dielectric thickness reduces threshold voltage and increases effective mobility at the cost of increased SS and lowered ION/IOFF ratio [44]. Furthermore, the big data generation from these devices imposes significant challenges to be addressed for handling, processing, and communicating data. Simple low-cost- garage-fabrication based electronics may be effective in free-form electronics, however, their inferior performances related to data management capabilities outweighs the noted advantages over CMOS based state-of-the-art technologies.

1.5. Dissertation Overview

This doctoral thesis addresses the critical challenges in obtaining a truly flexible standalone electronic system that can match the performance of existing bulky and

30 rigid counterparts. These challenges have been roadblocks, hindering the adaptability of these processes to comply with existing CMOS technology. The dissertation is divided into 6 chapters. The first chapter serves as motivation while the last chapter provides conclusions and future outlook. Each of chapters 2-5 presents a problem, details how that problem was tackled using novel approaches and includes strong support for why the approach is the best pragmatic way to move forward.

 Chapter 2 addresses one of the most neglected issues, that of obtaining

reliable, flexible interconnects. A concept of modular LEGO electronics is

presented for reliable interconnection. We demonstrate how module LEGO

works as lock-and-key mechanism for flexible electronics, eliminating the

need of any rigid soldering and fragile wire-bonding process. In addition, a

CMOS compatible fabrication scheme illustrates the conversion of existing

ICs into flexible LEGO modules for unique identification by naked eye and

reliable interconnects with soft-polymeric encapsulation.

 Chapter 3 focuses on the sensory component of an overall IoT system.

High-density electrodes and sensors are needed for mapping of human

activities for different applications. We present a record 1 million sensor

matrix that can measure the tiny movements of expansion and contraction

of an internal organ with high accuracy. Detailed fabrication,

characterization, and interfacing (mapping) activity is presented.

 In Chapter 4, a generic heterogeneous integration scheme has been

presented to obtain fully flexible 3D standalone electronic systems. This 3D

coin architecture is detailed having sensors (fabricated) on one side of the

coin platform. Sandwiched between the layers are CMOS based

microcontroller, solid-state microbattery, and the transceiver IC, all in

flexible form (< 20 µm thick). The other side of the coin hosts the antenna

and energy harvester for complementing the energy storage. 3D

interconnections across the layers have been established using TPV,

maximizing the area efficiency. Extensive mechanical and electrical

characterizations are detailed to understand the effect of processes,

mechanical deformation and environmental condition on the performance

of obtained flexed system from the integration scheme.

 In Chapter 5, the integration scheme would be useless without application;

accordingly, an electronic system application is demonstrated. Here we

have developed a multi-sensory light weight non-invasive Marine-Skin

tagging platform to monitor deep sea environment. Material optimization,

robustness and ruggedness of the sensory system due to extreme

mechanical, chemical, and environmental conditions have been

investigated, establishing the readily adaptability of our integration scheme

for manufacturability. This chapter also presents multiple non-invasive

attachment mechanisms for the marine species.

2. Chapter 2: Modular LEGO Electronics

2.1. Introduction

Architecture and appearance are often used as means of identity. Living beings such as Japanese male puffer fish - otherwise invisible in the sea environment, naturally create magnificent architecture in the sand to attract the female counterpart [45]. Living cells demonstrate countless examples of functional self-assembly in addition to being able to self-assemble themselves. Such exciting phenomena in nature have inspired research community to study, understand, and design self-assembled objects for applications in assembly, manufacturing and robotics of the future [46], [47].

The construction of modern man-made artefacts like computers, cell- phones, automobiles, etc., rely on modular assembly of components, assisted by robotic assembly lines to place, package, and interconnect a variety of modules at different scales; extensively in millimeter and higher scales. Complementary metal oxide semiconductor (CMOS) technology, which has been the driving force for today’s digital world, depends on scaling of the devices. Miniaturized devices with complex fan-outs are increasingly challenging for automated manufacturing. Large numbers of tiny pins on each side of the integrated circuits (ICs) requiring greater accuracy and precision are a key reason for such complexity. Thus, the assembly and interconnection of such scaled components is one of the bottle necks in modern CMOS technology lowering the throughput of the entire manufacturing process [48].

Different approaches to achieve parallel assembly as an alternative to robotic assembly are presently under investigation, including parallel transfer of chips to a donor wafer which typically contains three steps: fabrication of devices on donor wafer with a sacrificial release layer, pick up using an intermediate elastomeric transfer stamp followed by transfer onto the final substrate. However, this approach is not feasible when the conventional substrates are replaced by curved topologies and flexible platforms which are emerging as new frontiers in the electronics industry. Directed and engineered self-assembly is an alternate route in which maintaining orientation, alignment, and functional density over large areas is possible. Among different directed assembly techniques, surface tension driven assembly has attracted major attention. Smith et al. [49] pioneered this field by demonstrating the quasi-monolithic integration of Gallium Arsenide (GaAs) light emitting diodes (LEDs) on Silicon (Si) substrate by combining geometry and fluidic direction for self-alignment. Several other reported methods for self-assembly use the surface tension in combination with either solder bumps [50]–[53], magnetic

[54], [55] or electrophoretic [56], [57] forces, or capillary transport [58]. Smith et al. and Jacob et al. [59] have reported higher yields for the assembly process in their works.

Other interesting works presented in the area at a scale of hundreds of micrometer are: cylindrical display fabrication using capillary interaction, [59] sequential-shape-and-solder-directed- self-assembly (SSSDSA), [60] mesoscale building blocks assembly using capillary interactions, [61] single angular orientation and contact pad registration based alignment and integration of

34 semiconductor dies in SSSDSA. Sub 100 µm scale devices are reported to be assembled by liquid-liquid-solid surface tension driven approach [62], [63]. All of the above approaches of assembly have a combination of either shape, surface tension driven or gravity driven approach in addition to using liquid or molten solder paste on the connecting pads.

The emerging era of flexible, stretchable and reconfigurable electronics for applications in healthcare monitoring, displays, electronic papers, and environmental monitoring demands the compliance and high throughput assembly process which can be applied to compliant systems. Parviz et al. [47], [50] has demonstrated assembly of single crystal field effect transistors and diffusion resistors on a plastic substrate, which is mediated by fluidic agitation at moderate temperatures of 60-90ºC, to have good interconnection using molten solder pads.

The yield of above mentioned microscale level assembly approaches have been reported to be as high as 80% which enables the assembly of individual devices like field effect transistor (FET), PIN diode, etc. Most of the applications in flexible electronics and internet of things domain, which include health monitoring, wearable gadgets, flexible displays, and lab-on-chip devices, rely on integration of sensors network, data processing units, data storage, communication module, system-on-chip (SoC), and power management. In all these applications, a yield of 80% assembly at an individual device level will lead to failure in achieving the desired performance from a system since the devices are part of the major

35 components and even a single device interconnect failure means failure of the block, which implies entire system failure.

To tackle this problem, here we show a holistic approach of assembling at the level of system integration using modular LEGO-electronics. LEGO based modular approach fundamentally uses the geometry of two blocks/pieces that fit exactly into each other transforming the blocks into a system. These individual blocks can easily be detached and reconfigured into another completely different and independent system when rearranged in a different manner. A typical electronic system will include the microprocessor/microcontroller, memory, energy storage, and communication modules. We present a method of converting these individual integrated circuits (ICs) into a specific geometry that provides a unique identification to each IC –physically identifiable without microscopic details. A typical CMOS based IC consists of the isolated connecting pads (~ 55 µm - 120

µm square pads) on top of the die for establishing interconnects with the outer interface circuitry or fan-out. We convert the IC into one part of the LEGO-module

(part A or male-LEGO) by transforming it into a specific shape without compromising the performance of the chip. This module can be assembled onto its inverse replica (part B or female-LEGO) on the flexible substrate, bridging the designed system schematic. We first demonstrate a complete flexible system that implements an AND and OR logic gate like functionalities by fabricating both male-

LEGO and female-LEGO modules on flexible polyimide substrate. Later, we have translated this approach to transform commercially available ICs to LEGO- electronic modules that promises new opportunities and approaches of integration

36 and assembly techniques for flexible electronic systems. We also demonstrate for the first time how this geometrical modular LEGO-electronic approach empowers the blind assembly of the modules in position with great accuracy, which opens tremendous opportunities for the visually impaired.

2.2. Flexible LEGO Concept and Fabrication

This concept realization has two-layer geometrical shape recognition for modular assembly. The need for two-layer shape recognition arises from the fact that the device geometry itself has to be made unique enabling quick identification, however, the presence of two devices with the same geometry will need additional features (layer) for differentiation. Thus, a unique combination of two layered architecture of module and very well defined binding sites will ensure the accuracy and ability of the host to accept only designated LEGO modules at their site. First, we fabricate the inverted binding site replica (host) on Kapton sheet following the schematic presented in Figure 2.3A and millimeter scale male-LEGO modules using fabrication scheme in Figure 2.3B – both detailed in later section 2.4.

Fabricated host and Lego modules are depicted in Figure 2.3C and 2.3D. Three different component geometries: heart, circle, and square - all with same teeth shape (defined by square at second layer) and with different teeth design (heart, circle, and square) were fabricated. The dimensions of the devices were chosen to be 3 mm × 3 mm × 0.25 mm which mostly resembles typical IC dimensions at die level. Surface mount red LEDs were then mounted on the binding sites at

37 dedicated pads which act as output signal for implemented logic gate-like AND and OR design.

2.2.1. Modular Assembly for Logic Gates

The binding site dimensions for inverse replica of LEGO were carefully chosen to be slightly bigger (~ 20 µm) than the male LEGO-module in order to provide better placement without any added external pressure application when the modules reach the site. Figure 2.1A, 2.1B, and 2.1C show the binding sites before LED attachment, individual LEGO-modules and site after LED attachment, respectively, before assembly. Manually assembled complete LEGO-modules can be observed for accuracy in Figure 2.1D while Figure 2.1E indicates the performance of the system when powered by the turning ON of LEDs as a result of the output of the gates. Top left circuit trace in the Figure 2.1A represents an

AND gate logic which requires combination of heart shaped LEGO with one of the other modules to turn ON the LEDs while the right circuit trace makes OR like gate logic.

For assembly, the LEGO module was gently pressed into the binding site, however this resulted in only momentary illumination of the LED. Though the

LEGO and pads at the binding site seem to exactly fit into each other, it is possible a small gap or voids exist at the interface. To overcome this we applied a very thin

<10 um micro drop of silver epoxy resin onto the electrodes at binding site to ensure connection. This thin epoxy enabled improved adhesion, resulting in instantaneous connection and illumination of the LED. Additionally, the use of

38 epoxy removed the risk of damaging to the LEGO module by removing the need of external pressure.

Figure 2.1| The process of modular Lego assembly represented by A) empty binding sites having pads open for LED attachment and inverse Lego pattern, B) multiple Lego modules, C) binding substrate after LEDs are connected, and D) accurate assembly of Lego-modules on dedicated sites represented by turning on of LEDs when powered by a DC power source. Scale bar is 3 mm.

The accuracy of LEGO module identification by combination of two layer geometric identification is demonstrated when two same sized heart shaped LEGO having different shaped teeth (square and heart) are sequentially attempted to be placed at the same site. We can conclude from the experiment that when teeth

(second layer recognition) do not match with host, the module does not fit the site.

And when both first and second layer recognition matches, the acceptance is evident, connection is easily established and the LED switches ON. Additional

39 geometric features like angles and heights at the second layer will further strengthen the unique identification during assembly in complex and large component sets.

The flexibility of the fabricated modules is observed by bending the substrate after assembly. Similar to our previous comparison analysis of with and without epoxy resin assisted assembly, we show both these methods to evaluate bending performances. Figure 2.2A and 2.2B show performance of an AND logic gate remains intact after it is subjected to a bending radius of 1.25 cm while the OR gate logic configuration for the module does not turn ON. The reason for this is that we have used epoxy at the binding sites of AND gate logic while no epoxy was applied at OR logic sites. Since, both male-LEGO and female-LEGO parts are flexible, they pop out slightly when bent due to inability of the micro-epoxy to hold at higher pressure/stress induced from bending. This is solved with smart packaging of the modules by spinning a thin 100 µm PDMS on top to hold the modules intact. Extreme bending can be achieved in these flexible LEGO-modules as shown in Figure 2.2C and 2.2D, which demonstrate bending of individual LEGO modules up to a radius of 500 µm. Even lower bending radii may be achieved.

Figure 2.2| Performance of the assembled module A. and B. the LEDs are ON for the Lego that have intact connection due to epoxy application LEDs turned off represents the popping out of the Lego which were assembled without epoxy application. C and D shows bending of individual Lego module at two different bending radii of 0.75 mm and 0.5 mm respectively.

2.2.2. Fabrication of Flexible LEGO modules

A complete flexible system concept of LEGO-modular approach on polyimide (PI) based flexible substrate is built. Fabrication schematic of the female-

LEGO module (host substrate) is shown in Figure 2.3A, which is essentially an inverse replica (with AND and OR logic gate circuit traces) of its male LEGO- module counterpart, fabricated using the processing scheme illustrated in Figure

2.3B. Both the fabrication schemes are explicitly designed to be low cost involving no lithography step.

Fabrication of female-LEGO module host site starts with the glass carrier substrate (7 cm × 5 cm × 0.1 cm) on which a 125 µm Kapton (DuPont polyimide

(PI)) sheet is carefully attached to avoid air bubbles. Bubbles created under the

Kapton could create difficulty if they were to expand during subsequent processing steps. A shadow mask is applied to this substrate followed by DC sputtering of 180 nm Au layer with 10 nm of Ti as adhesive layer under Au. After sputtering Ti/Au layer, shadow mask is removed and second Kapton sheet of thickness 100 µm is adhered on this patterned interconnect structure to replicate the second level geometry. The inverse replica of the geometry of the male LEGO-module is imprinted selectively on this top layer using CO2 laser engraving with optimized combination of 30% power, 40% speed, and 75 kHz pulse frequency. Another engraving for the additional second level unique geometric identification is repeated on the bottom Kapton layer. Finally, the substrate module is removed easily from the glass carrier and is shown in Figure 2.3C.

Male LEGO-modules are fabricated in a similar fashion, starting with two

Kapton sheets of 100 µm and 125 µm thickness stacked over one another on a glass career substrate. Using CO2 laser engraving technique, different shapes are outlined on this top layer of the stack, which will act as teeth for each piece. A second laser engraving is applied on the bottom layer to complement the shapes from the female-LEGO module fabrication. This is followed by Ti/Au DC sputtering in exactly the same manner as mentioned previously. Different module shapes

(heart, square, and circle) each having different shaped teeth can be observed in

Figure 2.3D.

Figure 2.3| Fabrication schematics for flexible LEGO A) flexible binding site (inverse replica) of the Lego module on PI, B) flexible Lego-module fabrication, C) digital image of binding sites after fabrication on glass substrate before release, and D) digital image of Lego modules in three different shapes (heart, square, and circle) with same and different teeth. Bottom right shows the contour of the step height for Lego teeth. Scale bar is 5 mm.

2.2.3. Contact Reliability and Assembly Yield

Cross-sectional scanning electron microscope (SEM) images were acquired to investigate contact establishment in order to optimize the process and to determine if conductive silver epoxy resin is imperative. Figure 2.4A represents the cross- section of a LEGO-module assembled fit into host binding site without any silver epoxy resin. We can observe from the SEM image that there is a void between two flexible layers meaning no connection between pads of LEGO module and pads of binding site. A very fine line of interconnect is possible between the LEGO

43 module and its binding site, which intermittingly completes the circuit resulting in momentary turning ON of the LEDs. Cross sectional SEM image of the LEGO- module assembled with very thin epoxy application between the pads is presented in Figure 2.4 B and C, seamless interconnection at the binding site is observed.

There is a possibility of minor voids in the epoxy, however, this is negligible as there is enough contact established to counter it. This is further confirmed by plotting voltage versus current measurements across the pads of the two assembly methods. The resistance is provided by the slope of the line and gives the value of

<10 Ω (7.6 Ω) for LEGO with epoxy whereas ~475 Ω without epoxy, as plotted in

Figure 2.4D. This high resistance restricts the current flow in the circuit and results in a maximum voltage drop across the LEGO instead of the LED.

Figure 2.4| Cross-sectional scanning electron microscopic (SEM) images of the Lego-modules acquired when they are attached at binding sites. A) Lego module and binding site fit accurately;

44 however, there is a void at interconnecting pads that make gentle external pressure application mandatory for connection establishment. B) Strong seamless interconnection of flexible Lego on PI at the site assisted by a minimal amount of conductive epoxy. C) Represents reliable connecting path between Lego and the host site even if there are voids in the epoxy application. D) Resistance measurement for the connected Lego when epoxy is applied and without epoxy application (with a gentle tap applied to make some connection).

The high yield of geometrical identification and assembly is studied using a robotic pick and place tool (Datacon 2200) - custom designed for handling the flexible devices and substrate shown in Figure 2.5A – 2.5C. The tool has been programmed to accurately identify these LEGO-modules, differentiating between same component geometries with different shaped teeth. Tool assembly achieved

>97% yield of exact placement of LEGO as seen in Figure 2.5B and 2.5C, without and with epoxy (LEDs ON), respectively. This robotic assembly due to geometrical unique identification allowed us to set up a novel experiment where a subject is blind folded to mimic the visually impaired person. A colleague was blind folded and tasked to assemble the devices on the binding site by manually identifying the physical geometries. Promising results of successful assembly of the flexible

LEGO-modular electronics were achieved, which can open unprecedented opportunities in the CMOS technology, especially for visually impaired people.

Figure 2.5| High yield robotic assembly process using pick-and-place tool is demonstrated. A) Typical stage area shown at the machine site and few important components of the tool are labelled, inset shows different Lego modules. 100% yield of assembly achieved by programming the tool; B) without application of epoxy which results in off LEDs due to no external pressure to improve connection, and (C) microliter of epoxy application before assembly results in enhanced interconnection and hence LEDs are turned ON for both AND and OR logic gates. Scale bar is 3 mm in (B) and (C).

2.3. Transforming IC into flexible LEGO

In this study, we demonstrate transformation of a commercial IC into LEGO- module by implementing a comparator circuit on a flexible substrate. High precision operational amplifier (OPAMP) constitutes an integral part of many system-on-chip (SoC), microcontroller or microprocessor, and many other analog and digital systems. Here, we converted a state-of-the-artcommercially available

OPAMP bare-die IC into a LEGO-module. The fabrication/conversion procedure is presented in Figure 2.6A and detailed in the fabrication section. The integrity and reliability of the process conversion is validated by device characterizations of unprocessed bare-die module and the converted LEGO-module. A simple comparator circuit on flexible PI is fabricated using schematic in Figure 2.6B. The optical microscope images for the unprocessed bare-die IC and converted LEGO- module are presented in Figure 2.6C and 2.6D, respectively, showing 100 µm contact pads. Digital images of the LEGO-module after assembly can be observed in Figure 2.6E from both front and rear side of the binding site on glass. Conversion of state-of-the-art bare die ICs into a LEGO-module could also potentially be achieved by patterning specific geometry on the backside of the IC using techniques such as laser etching or soft-back-etch.

2.3.1. Rigid IC to LEGO Conversion Flow

This is an extension of the LEGO module fabrication detailed in section

2.2.2, here we convert commercial integrated circuit (IC) bare-die into male-LEGO module and the female-LEGO module is fabricated on flexible platform on glass substrate. The male LEGO-module conversion of bare-die IC is illustrated by schematic in Figure 2.6A. A commercially available bare-die IC (high precision operational amplifier OPA2340 from Texas Instruments-TI) is used here. The contact pads (100 µm) on the devices must be converted into teeth by growing metallic columns of specific shape on these pads. First, we embed these bare-die

(280 µm thick) into uncured PDMS, spin coated on a Si/SiO2 wafer to get a

47 thickness of ~270 µm. This PDMS is cured at 65 ˚C for an hour followed by DC sputtered deposition of 10 nm/150 nm Ti/Au seed layer for thick copper electroplating (or electrochemical deposition (ECD)). ECD Cu growth is carried out in a plating station with a DC current of 147 mA for 25 minutes, growth is controlled to the vertical direction by patterning a positive photoresist (PR) ECI 3027 using direct laser writer (µPG501 Heidelberg). The PR removal follows seed layer removal using reactive ion etching (RIE) to ensure excess metal removal to avoid any shorts. RIE of Ti/Au is performed in two steps; first to remove Au by RIE is done under Ar plasma at 10 ˚C and 5 mTorr pressure for 3 minutes, followed by Ti etching with a mixture of SF6, C4F8, and O2 gases at 10 ºC and 5 mTorr pressure for 30 s. Cu teeth grown on bare-die completes the male LEGO transformation, which then can be easily peeled from PDMS and cleaned from backside for removal of residues using acetone, isopropanol (IPA), and deionised (DI) water.

State-of-the-art bare-die OPAMP IC after conversion to LEGO module can be seen in Figure 2.6C.

Female LEGO module (or host site) fabrication starts with a glass carrier wafer followed by spinning 10 µm polyimide (HD microsystems PI-2611) at 2000 rpm for

30 s. Spin coated PI is initially soft cured at 90 ˚C for 90 s followed by intermediate cure at 150 ˚C for 90 s and final hard bake at 350 ˚C for 30 minutes by gradually ramping the temperature from 150 ˚C to 350 ˚C at a rate of 4 ˚C /min. This step is followed by Ti/Au metal interconnect deposition using the same parameters mentioned above, while patterning is carried out using lift-off process. A positive

PR ECI 3027 is used to pattern interconnects on PI after surface activation under

O2 plasma in RIE at 30 ˚C for 2 minutes. DC sputtering of 10/180 nm of Ti/Au is followed by PR lift-off in acetone sonication bath followed by rinsing in IPA and DI water. This patterned circuit is followed by second PI layer spinning for 20 µm using the same multiple step spinning recipe. This layer will host the teeth formed from the male-LEGO piece counterpart. The vias/ inverse replica of teeth are etched atop PI using RIE. A hard mask of 200 nm Al is deposited on PI to selectively etch

PI by RIE flowing a mixture of 10:1 O2:CF4 gases at 60 ˚C and at 80 mTorr pressure. Cyclic PI etching is performed with breaks in steps to cool the substrate.

Finally, Al hard mask is removed using highly selective gravure Al wet etchant followed by final cleaning step of acetone, IPA and DI water.

Figure 2.6| Process flow for transforming rigid IC into flexible LEGO: A) transforming any bare die or CMOS base IC at wafer level into Lego-modular electronic component and B) inverse replica formation to act as host that incorporates the complete system outline. C) Optical microscopic images of as-received bare die for OPAMP in our study having standard 100 µm contact pad size and D) transformed Lego OPAMP. Scale bar is 300 µm. E) Digital image transformed Lego-OPAMP assembled in comparator layout (top) and female-Lego binding site under Lego-OPAMP (bottom). Scale bar is 1.5 mm.

2.3.2. The Comparator Configuration

The comparator configuration of the OPAMP is implemented by providing a reference voltage (Vref) (50 mV - 550 mV DC) at the inverting terminal of the device with a sinusoidal input signal (Vin = 1.2 Vpp at 1 kHz frequency) applied at non- inverting terminal. Unprocessed OPAMP die characterization for comparator configuration using semiconductor probe-station is recorded on digital oscilloscope in Figure 2.7A with Vref = 500 mV and Vin = 1.2 Vpp at 1 kHz. It is obvious from the waveforms that the output signal jumps to +VCC (upper saturation of supply) whenever the input signal goes higher than Vref and it drops to -VCC (lower saturation) for the signal lower than Vref. The peak-to-peak voltage for the obtained square waveform is 5 V, which resembles the difference of +VCC and –VCC. Pulse width of the output signal is measured to be approximately 18.5 % which closely matches the width of input signal Vin > 0.5 V (= 18.3 %). The similar performance is reported for various reference voltages varying from 100 mV to 500 mV and is represented in Figure 2.8.

The converted LEGO-module OPAMP is characterized using same method as described above and the results are compared to the previous unprocessed

OPAMP. The exact performance is observed in Figure 2.7B for the converted

LEGO-module which when compared to the unprocessed device validates that the conversion process does not alter device properties and hence, performance is retained. Finally, the device characterization after LEGO-module assembly onto the host circuit is carried out. We observed no significant performance degradation or variation in the comparator configuration of the device, prominent from Figure

2.7C. Careful attention must be given to the conductive epoxy application that helps improve the interconnection of the two modules. Previous studies used molten solder at the binding sites, however, the required heating may cause damage and affect performance. The thickness of the conductive epoxy has to be controlled to avoid overfilling the trenches and blocking the matching pattern on the binding site, which is inverse design of the teeth on the male LEGO- module.

Thus, the implementation of the modular LEGO electronic concept has been realized at the IC level with a high yield and a good rationale for not pursuing alternative self-assembly methods of microscale components, mainly FET, PIN diodes, etc. The integration and assembly of other major electronic system components at the macroscale can be further investigated.

Figure 2.7 | Output characteristic waveforms of OPAMP at different processing steps. A sinusoidal input signal of 1.2 Vpp at 1 kHz and a reference voltage of 500 mV is applied for OPAMP in comparator configuration. A) As-received bare die, B) transformed Lego die after final Cu growth using electroplating for teeth formation, and C) output waveform after the assembly at the host site on flexible PI substrate.

Figure 2.8 | Output characteristic waveforms of transformed Lego-module after assembly in flexed state. Bending radius of substrate flexed is 1cm, A sinusoidal input signal of 1.2 Vpp at 1 kHz, at different reference voltages A. Vref = 100 mV, B. Vref = 200 mV, C. Vref = 300 mV, and D. Vref = 400 mV in comparator configuration.

2.4. Conclusion

Ordered assemblies of different geometries found in nature was a motivation to demonstrate the concept of LEGO based modular electronic systems that are compliant and feasible for applications in wearable gadgets, health monitoring, flexible displays, etc.

We provide a unique two level identification to each module that enables recognition and binding of the modules at precise sites. This has been demonstrated by fabricating modules with equivalent geometry (heart, square, and circle) and size while having differentiated teeth (shape, width, length) used for connecting the module to the binding site. This design allows the modules to accurately assemble only at the binding site specific to a unique combination of the two identifiers (module size/shape and teeth size/shape), providing a lock-and- key model approach. Blind folded assembly for electronic LEGO- modules is demonstrated for the first time, which can be foreseen to open a multitude of opportunities for visually impaired persons. The flexibility of these LEGO-modules will certainly enable opportunities for implementation within flexible sensor networks, wearable devices, health and environmental monitoring, and especially for internet of things and internet of everything applications. Further, we believe that fluid directed self-assembly can be studied in the future to complement modular LEGO electronics.

3. Chapter 3: High-Density (1 Million) Sensors and System Integration

In this chapter, we present an extremely high-density pressure sensors array with extreme high-spatial resolution and noise isolation from adjacent sensors. To gain detailed insights of complex biological processes, devices that can manipulate and track the cell parameters are essential. Thousands of recording/stimulating electrodes (or sensors) are required across the large area with high-density and high spatiotemporal resolution, in addition to physical flexibility and seamless integration with existing technology which are often unaddressed. Here, we show a physically flexible pressure sensory patch with the highest density (15,625 sensors/mm2) and highly selective 1 million pressure sensory matrix to understand in situ pressure distributions. Our multi-level cumulative addressing approach provides unprecedented spatial resolution, high sensitivity (detection of ~10 µl droplet pressure) and ease of accessing millions of sensory sites in a small area

(25 mm2). We further demonstrate cellular activity and growth pattern monitoring

(MDCK II cells) using this sensory patch. The flexible, biocompatible, sensitive, high-density, and high-spatial resolution million sensors patch have potential to address the current challenges of next generation implantable and neural interfacing technologies. Unlike high-density MEAs, which have precisely a single electrode pitched at micron scale distance connected to the active matrix based transistor and active devices for sensing, we have designed a pressure sensor passive matrix that responds to extreme low pressures with highest spatial resolution (sub-cellular < 10 µm) without being effected by adjacent sensors and the need of active elements. We show how we have achieved a record 1 million

55 sensor fabrication in smallest area (never shown before even in MEA where state- of-the-art maximum number of electrodes demonstrated are ~59,760 [64]. This concept and application of cellular growth activity monitoring and mapping in situ is illustrated in Figure 3.1.

Figure 3.1| Conventional method and the vision of high-density cellular growth pattern mapping. (a) Digital image of a flexed high density integrated array, (b) a broad schematic of a flexible electronics system with multiple flexible silicon chips, (c) flexed system wrapped around human finger, and (d) flexible sensory patch that can be integrated with silicon chips.

3.1. Introduction

Rapid development of high-performance materials, technologies, and processes has landed human being in the age of smart things known as IoT and

IoE. The notion of things being smart or intelligent is attributed to the ability of making decisions based on feedback after sensing (detecting) a specific signal or circumstance, similar to how a human brain reacts. In the quest for improving human life, there has been a surge in electronic devices that monitor different parameters over a short period of time or constant monitoring depending on the need. Widespread examples in medical diagnostics include ECG, EEG, ECoG, ultrasound, X-ray scan, magnetic resonance imaging (MRI), etc. The central element in all these diagnostic and imaging techniques is the sensor/transducer which converts a specific signal into electrical signal. Tremendous amount of research activities have been directed toward developing sensors for a variety of applications like physiological monitoring [9], [37], [65]–[68], diagnostics [23], [31],

[69]–[71], environmental monitoring [72]–[76], and neural activity and cellular potential mapping [77]–[82], etc.

Cellular and extracellular recordings of electrical activities of cells located in brain, heart, internal organs, and retina can provide huge insights and information about the physiological and pathological conditions of the organs, especially in the area of degenerative diseases like Parkinson’s disease or Alzheimer’s. In addition

57 to electrical signals, electrochemical cues can also be detected due to ion exchange between the cells during specific functions [77], [83], [84]. The chemicals that are transmitted during the synaptic nerve cycle are termed neurotransmitters and play a significant role in the above mentioned diseases. These kinds of signals are trivially detected using electrochemical sensing methods involving voltammetry, and similar techniques. Furthermore, for diagnostics and treatment purposes in diseases like depression, chronic pain, epilepsy, disorders of peripheral nervous systems (PNS), and other degenerative disorders, neurophysiological monitoring is widely practiced. Advanced capabilities play a critical role in mapping and monitoring brain functions during surgical preparations and intricate neurosurgeries, and can assist in areas such as location of tumor, accurate placement of a neurodevice, guiding surgical procedures on congested peripheral nerve mesh like in lumbar, brachial and sacral plexus. These clinical needs provide enough motivation to put efforts in developing technologies for neurophysiological monitoring along with other human physiological monitoring platforms that can provide high-fidelity (as available in ICUs), high-resolution (as offered in the operating room), and reduction in costs by pre-detecting onset of diseases. [85], [86].

Array of sensors interfacing the neural tissues and cellular matrix for hundreds of sites on passive substrate has become a popular tool in the neurotechnology, drug screening and discovery, and other pharmaceutical applications [87], [88]. The advancements in Multi Electrode Arrays (MEAs) technology over the year have further contributed to well-established practices for

58 in vitro cellular and extracellular investigations using multiple parallel electrical recording and stimulating electrodes. Advances of CMOS technology has enabled fabrication and usage of thousands of MEAs in a smaller area down to pitch similar to the cell diameter (tens of micrometer) that is crucial to providing the detailed insights about the physiochemical phenomena at cellular and sub-cellular level

[89]–[95].

Wearable healthcare monitoring devices are used heavily to track fitness/wellness, physical rehabilitation, awareness and cognitive state assessment, and more or less toward human-machine interfaces. An electronic device that enables measurement of body temperature, pressure, pulse oximetry, or electrophysiology is generally deployed on one or more regions of the body.

Mapping activities of temperature and pressure in specific locations of human body can provide vital insights about the health status which can lead to predictive information before onset of the disease. For example, temperature variations during sleep can be used to gauge the sleep phase that can help in treatment of sleep related disorders. Similarly, measuring pressure at the skin interface may provide important information relating to pressure ulcers, and preventive measures alerts can be provided to avoid skin sores, irritation, decubitus ulcers [96].

Furthermore, the evolution of technology has led to successful demonstrations of multisensory platforms in flexible format for physiological monitoring from sweat, blood pressure, hydration, body temperature and other vital information, often regarded as e-skin (electronic skin) platform [65], [67], [97]–[104].

3.2. Sensor Design and Interface

The electronic devices that can sense pressure, shear stress, strain, and deformations (axial or angular twisted) have gathered significant attention from the researcher community. An accurate quantification of these parameters depends on the effective signal transduction converting external stimuli into analog electrical signal using various transduction methods (piezoelectric, piezoresistive, resistive, or capacitive, triboelectric, optical, or electromagnetic mechanisms). Rapid developments are ongoing to meet the new and emerging challenges and opportunities that overload the application of e-skin in prosthetics, robotics, and human-machine interfaces in augmenting artificial intelligence and future generation devices. Recent reviews on the pressure sensing modalities have revealed advantages and disadvantages of each mechanism for a variety of applications [98], [105]–[111].

A generic parallel plate capacitive structure can be used as pressure sensor where the capacitance varies linearly with the variation in the thickness of dielectric between the two metallic plates/electrodes. Traditionally, diaphragm based capacitive pressure sensors have been used in microelectromechanical systems

(MEMS) industry due to their higher sensitivity and mature process, however, complex fabrication processes are involved. Furthermore, MEMS based diaphragm and devices are not suitable for flexible electronics applications due to their fragile nature and change of properties upon bending. In our sensor design, we have used a soft-dielectric elastomeric material, PDMS, which due to its compressive and elastic nature responds to the applied pressure by thickness

60 variation. Hence, we have used this material with a dielectric constant εr= 2.9 as an insulating and pressure responsive layer sandwiched between two metallic electrodes. The response to applied pressure on the parallel plate capacitive structure is expressed by variation in capacitance C due to change in dielectric thickness d proportional to the applied pressure, as expressed in Equation 3.1

ε0εrA⁄ C = d (3.1)

Where εr= 2.9 (relative permittivity of PDMS dependent on the mixing ratio

−1 of elastomer and curing agent), ε0 = 8.854 x 10-12 F⋅m absolute permittivity of free space, and A is the surface area of metallic plate overlapping the dielectric thickness, d.

PDMS is known as a biocompatible material and is also compatible with cleanroom nanofabrication processes; we have used microfabrication processes that are compatible with state-of-the-art CMOS technologies to fabricate the sensors. This allowed us to fabricate the sensors with individual electrode area of

100 µm2 with 7 µm electrode and 5 µm pitch. This allowed us to fabricate 1 Million pressure sensors in a small area of 2.25 cm2 with an astonishing high-density of

10,000 sensors per mm2. With the help of advanced lithography techniques and advanced nanofabrication processes, the sensor dimensions can be further scaled to achieve 1 million sensors in 25 mm2 if the sensory electrode pitch is reduced to

5 µm with electrode size at 3 µm and spacing of 2 µm. This can further be scaled using electron beam lithography technique giving electrodes as small as 100 nm with similar pitch to obtain a density of more than 5 x 105 sensors per mm2.

Although, such a high density will pose extreme challenges with interfacing electronics development.

The rationale behind choosing the specified dimensions come from the interfacing electronics. The design of sensors is essentially a switch matrix or cross bar architecture having electrodes connected electrically at different polymeric layers in rows and columns. These rows/columns can be used as select lines and individual and/or sets of electrodes can be accessed using multiplexing techniques. Further, many applications in cellular or physiological monitoring in vitro or in vivo do not need to have individual cellular level reading (individual sensor access), rather, extracellular or field signals can be combined to analyze the conditions. Thus, our multiplexing technique and the design not only allows individual access to the sensors but also a higher-level access to the set of sensors divided into 10 channels or rows and columns providing an accumulative signal from a local field. Further sensor characterizations and performances are detailed in section 3.4 after the fabrication method detailed in next section.

3.3. Fabrication Process Flow

For the design of parallel plate capacitance based pressure sensor in flexible form, thin metal layers on polymeric flexible substrate is essential with a soft and compressible dielectric medium. Air as dielectric material was not an options as we had high-density sensor fabrication allowing use of standard CMOS based processes and similar compatible materials in micro/nanofabrication process.

We start the fabrication with a bulk Si (100) carrier wafer having thickness of ~500 µm, suitable for handling and subsequent fabrication steps, as illustrated in 3D schematic flow in Figure 3.2. Next, we chose one of the most widely preferred flexible polymer- polyimide (PI 2611 from HD Microsystems) as the substrate due to its mechanical and chemical properties. We spin ~ 8 µm of PI on Si wafer using spin coating at the speeds of 2000 rpm for 30s preceded by a 5s step of spreading the polymer on wafer at 500 rpm. This spinning step is followed by multistep curing procedure for PI; first step is moisture/solvent removal from the liquid spun PI by baking it at 90 °C for 90s followed by initial soft curing at 150 °C for 90s. Next final curing is done at 300 °C for 30 minutes before which temperature of the hot plate is gradually ramped from 150 °C to 300 °C at 240 °C/hour ramping rate providing the best stable mechanical properties.

Next step in fabrication involves deposition and patterning of metallic sensor layer 1 forming electrodes on one side of the capacitive structure. Since PI surface has comparatively poor adhesion to the sputtered/deposited thin film metals, we perform surface treatment of PI using plasma RIE in presence of 10 SCCM of O2 flow at 5 mTorr chamber pressure, 10 °C table temperature and 150 W RF power for 2 minutes. This helps in surface preparation followed by deposition of thin Ti/Au layer (10/180 nm) using DC metal sputtering under 25 SCCM Ar flow, 400 W DC target power. The deposited metal is patterned using nanoimprint lithography technique for which a positive photoresist ECI3027 (4 µm) is spin coated and baked at 100 C for 1 minute followed by exposure to lithography under a constant dosage of 200 mJ/cm2 using a bright-field quartz-chromium mask, followed by post

63 exposure development in AZ 726 MIF developer for 60s. Each chemical step or etching or development is followed by wash and rinse cycle of washing with acetone, isopropyl alcohol (IPA), deionized water (DI water) and dried using N2 gas. This metallic electrode pattern is then etched using RIE in dirty chamber for

2 minutes at 10 °C, physically etching Au layer using Ar plasma (30 SCCM) at 150

W power and 25 mTorr chamber pressure, followed Ti etching in presence of SF6 and Ar gas with same conditions for 30s. The patterned electrodes have PR as mask during etching which is removed in next step using acetone, IPA, DI rinse cycle without need for any ultrasonication.

The dielectric layer for such capacitive structure is compressive PDMS material. PDMS can be prepared by mixing ratios of two components; Sylgard184 elastomer A and curing agent B in various ratios. The standard ratio is 10:1 of elastomer to curing agent which we have modified to 12:1 to get softer and more compressive dielectric layer that will improve the sensitivity of the pressure sensor.

A thin 50 µm layer of PDMS is spin coated after mixing at speeds of 1000 rpm for

1 minute followed by low temperature curing at 60 °C for 75 minutes. This low temperature curing is essential to maintain the compressive nature of PDMS that can alter the properties if cured at relatively higher temperatures like 90 °C or more.

First layer of electrodes and dielectric layer are formed, now second layer of electrodes remain. Exact same process for fabricating layer 1 is repeated with a second mask and aligned using nanoimprint contact aligner for accurately aligning

5 µm pads with a tolerance of < 1 µm. We encapsulate the top layer of the capacitive pressure sensor array with a thin 100 µm PDMS layer to protect devices

64 from wear and tear in handling, followed by peeling the device from the Si surface.

The Si carrier wafer used gives an advantage of poor surface adhesion to PI as compared to SiO2 wafer and hence it assists in easy peel-off the entire flexible 1 million sensors. The peeled sensory array can be seen flexible and conforming to the curved surface when held, flexed, and twisted on the finger, as seen in the

Figure 3.2.

Figure 3.2| 3D and cross-sectional 2D fabrication process flow of 1 million pressure sensory matrix

3.4. Sensory System Performance

The fabricated sensory matrix (array) of 1 million sensors can be optically seen in the digital image Figure 3.2A with the matrix occupying 1.5 cm x 1.5 cm area of the substrate. Individual sensory nodes are accessible using the channels formed

65 using rows and columns, as shown, which forms a multiplexing network for digital acquisition systems in stand-alone mode. These channels have a fan-out at 3 different levels providing access to individual sensor (L0) or a combination of 20

(L1) or 100 channels (L2). In many cases the local field recordings are desired instead of individual sensor, hence this multilevel approach will be beneficial in all cases where there is no extreme necessity for accessing an individual element when a combination of sensors provides vital information. Scanning electron microscopy (SEM) image of the individual sensor array can be seen in the Figure

3.3 that illustrates the accurate overlapping of the two parallel plate elements with channel lines connecting at their respective layers.

3.4.1. Individual Sensor Characterization

Individual pressure sensor characterization is carried out by applying pressure on each node and recording the corresponding change in capacitance due to dielectric compression. Since each sensory element is ~10 µm square pad, we used the metallic probe as a source of external pressure. Repeated measurements were performed by applying high pressure on a single sensor and releasing the pressure by instantaneous lifting of the probe. Instantaneous and rapid response a fast response time of ~100 milliseconds of is recorded (Figure 3.3e) (limited by the measurement tool) with the increase in capacitance corresponding to decreasing thickness of PDMS dielectric layer on application of pressure.

Furthermore, the sensitivity and spatial resolution are important characteristics, and we have observed the sensitivity to extreme low pressures as small as ~ 10

Pa which is equivalent to the pressure of ~ 10 µl water droplet placed on the

66 sensory patch (Figure 3.3f). Detection of such low pressures and fast response times are comparable with most of the advanced works in the domain reported to the best of our knowledge.

Figure 3.3| Physical and electrical characterization of fabricated sensory matrix Physical characterization of high-density sensory matrix. (a) digital image of 1 million sensors and corresponding multiplexing channels (addressing nodes L0, L1, and L2), (b) Scanning Electron Microscopy (SEM) images for nanoscale electrodes (top right), 1 µm pitched electrodes with equal spacing (bottom right), and standard 10 µm sensor (left). Optical microscopic image of device with (c) PHCCBEADS as control experiment, and (d) MDCK II cells after 48 hour of incubation period, arrows in the image points to the clusters of grown cells. (e) Response of the applied pressure on single pixel using tip of the probe and (f) high sensitivity response to microliter droplets of water (10-40 Pa pressure).

3.4.2. Spatial Resolution

One of the major challenges in high-density electrodes is the spatial resolution which means how effectively the individual sensor or the channel responds to the signal without being affected/induced by changes in adjacent electrodes. This essentially points to the effective noise cancellation and isolation from the adjacent sensors, hence, providing the best possible response to the desired signal. In this regard, we needed to characterize the individual channels for their isolation and see if there is any effect from pressure applied on adjacent channels. For such characterization, we fabricated a 3D printed mold of letters INL having a cavity in the middle that can hold liquid (water) as a measure of applying pressure on the sensory channel. These molds can be seen in Figure 3.3B placed on the sensory array and filled with DI water for measurement.

To study the effect of adjacent cells, we apply a fixed pressure on one of the channels (channel 2 column) (inset shows letter mold I filled with DI water) and then measure the observed capacitance change in adjacent and all channels.

From the plot of recorded capacitance change, we did not observe any significant deviation in baseline values of capacitance for the sensors not subjected to

68 pressure. In addition, we only saw the response in capacitance change at channel

2 where the pressure was applied, which conclusively indicates that the sensors are not effected by adjacent sensor signals, and therefore have a high spatial resolution.

Similarly, we repeated the measurements for bunch of channels by placing the letters INL filled with water and also unfilled on the sensors using different orientations to observe sensitivity and response of a channel. We can observe that when letter I is place in vertical orientation having maximum pressure on columns, the recorded value of capacitance is higher at the channels in the vertical column than the channels in horizontal row that do not have the applied pressure. The same row shows higher capacitance when the I letter mold is placed in horizontal orientation corresponding to maximum pressure. Similar trends were observed using N and L letter molds showing the high sensitivity of even combined channels to the small changes in the applied pressure. The varied pressure applied on the sensors due to varying weights of the mold in different conditions and different letter molds are listed in Table 3.1.

Table 3.1 | 3D printer INL letter mold for pressure monitoring by using their varying weight

Letter Mold Stamp-as printed Stamp with cavity Stamp filled with water I 166.56 mg 122.10 mg 177.64 mg N 277.06 mg 218.70 mg 289.3 mg L 153.44 mg 124.65 mg 158.18 mg

Figure 3.4| Spatial resolution and sensitivity across channels (A) Response from all the channels recorded when pressure is applied on single channel line 2, adjacent channels show no significant variation. (B) Recorded response of the channels when the characterizing mold of different letter were placed in different orientations on the channels.

3.5. Mapping of 1 Million Sensors

Mapping of pressures in a given area is critical to a huge variety of applications from physiological monitoring, wellness/fitness tracking, internal organ monitoring, and also in vehicle industries for car aerodynamics mapping. The pressure measured at multiple points provides an aerodynamic map of the vehicle thereby showing the effects of drag forces exerted on the car body that implies variation in acceleration, balance, and speeds of the vehicle, which are the most critical factors in racing cars.

We created a 3D printed mold of letters INL, which represents our lab, each having a cavity that can be filled with DI water. The 3D printed mold initially had wax in the cavity which is removed before filling with water. The molds are

70 designed to cover the majority of the channels of the sensory matrix and then the respective capacitance changes are measured. Here it must be noted that we have

10 channels at L2 for simplified recording from combined channels instead of an individual cell. However, it is worth remembering that each individual sensor is contributing to the local field or the accumulated cellular response from 10,000 sensors in each of the 100 channels from a matrix of 10 x 10 channel arrangement for mapping. One question that arises here is, what is the advantage of having an array of small sensors in an area when their combined response is recorded which could be done by one large sensor covering the similar area? When we use capacitive pressure sensing technique relying on dielectric thickness variation, the boundary condition for the accurate measurement is uniform application of the force/pressure on sensing electrode across the entire area of capacitor. If a point source force is applied, there is deflection in the capacitance top plate (thickness reduction) only at the point of force however at other points it is raised, countering the change in thickness due to in-compressive nature of dielectric material retaining its volume. Therefore, if a single large sensor is used there would be low response to non-uniform pressures. At the same time, our high-density sensors are extremely small which means the response to the point loads is recorded without depletion and error thanks to the array design.

The mapping response for different letter molds filled with water are plotted in

Figure 3.5 A-C , on the left as vertical bars respective to applied pressure, while

2D plots of the same are represented on the right. In 3D bar mapping plot, we can clearly differentiate the areas that were under various pressure which are also

71 labelled. The lowest capacitance, termed as base capacitance, corresponds to atmospheric room pressure exerted on channels when there was no mold placed on those channels (channels in column 1, 2, 8, and 9 in Figure 3.5A (left)). The design of the mold allows for applying different pressure at different points (vertical edges of mold and cavity). The edges are solid material exerting pressure slightly higher than the cavity filled with water. The mapping measurements accurately represent the response to the corresponding pressures by having tall towers at the edges of the mold with slightly shorter towers in the cavity and also low points for atmospheric pressure outside the mold. Similar pixels mapping is plotted as a 2D matrix seen in Figure 3.5A (right) representing all the 100 higher order pixels for combined 1 million sensors with color coded pressure measurements. The maximum recorded capacitance at the edge is 4080 fF while at the cavity it is 4075 fF, which shows the nature of small pressure change detection. Also, the base value recorded for I mold was 4025 fF.

Similarly, we have repeated the mapping experiment using N and L molds and the respective 3D bar mapping and 2D pixel mapping is illustrated in Figure 3.5B and 3.3C for N and L letters, respectively. The maximum value of the capacitance to the applied load for N-mold was measured to be 4205 fF at the edges (higher than I due to larger mold area) whereas the baseline for the sensors that were partially covered was 4180 fF and respective minimum baseline was recorded as

4140 fF. Similar trend were observed for L-mold showing maximum value of 4105 fF at the edges of the mold while outside the edges the values were 4075 fF. The variation in the baseline values throughout the experiments can be attributed to

72 the changing conditions and calibrations. The overall absolute changes in capacitance were observed to be highest for largest mold N (65 fF) while smallest for L shape (30 fF) due to design. Thus, mapping of such a huge number of sensory nodes can enhance the state-of-the-art in health monitoring, neurological signal recording and stimulating, as well as niche applications in racing vehicle aerodynamics. A tabular comparison of our work and other relevant notable works can be found in Table 3.2.

Table 3.2 | Comparison of high-density electrode and related work with this work

Work Number of Spatial Sensing modality Electrode/sensor sensors/electrodes resolution density (#/mm2 ) Dragas [77] 59760 13.5 µm AP (mV) 5,487

Rogers [112] Few 100s NA Pressure NA

Eversmann [87] 32768 8.75 µm AP (mV) 12,987

Chi [113] 144 89.4/357.8 µm AP (mV) 125/7.5

Ballini [82] 26400 17.5 µm AP (mV) 3,265

Chiang et al. [113] 1008 290 µm - 12.12

This work 1 million 10 µm (scalable) Pressure 10,000 (resolution 30 mPa)

Figure 3.5| 1 Million pressure sensor mapping. 3D Mapped response of channels recorded on application of pressure using 3D printed mold having cavity filled with DI water (A) for mold letter I, (B) for letter N, and (C) for letter L, respectively. 2D mapping for the same experiments are plotted on the right plots for respective mapping.

3.6. Cellular Growth and Pattern Mapping

This high-spatial resolution and sensitivity has allowed us to explore applications of high-density mapping in biological activities related to cellular growth, patterning, and cell counting. The concept is to design a matrix of sensors and record the response of the sensors to the growing number of cells within the matrix. We hypothesized that higher density of cells at any pixel will exert the higher pressure on it reflecting a higher capacitance. This concept can be used for pattern mapping and growth activity monitoring as illustrated in Figure 3.1. For this, we started with the control experiment of measuring the response to the droplet of

DI water containing the PHCCBEADS of similar size as the Madin-Darby Canine

Kidney (MDCK II) cells (~15 µm diameter). A consistent trend is observed with the increased capacitance relative to increased number of beads by increasing the droplet volume. We have observed the repeatability across multiple devices for the same control experiment and the trend repeats. The control experiment with droplet of DI water having PHCCBEADS can be seen in Figure 3.3 c.

For cellular growth experiment, first we established that the materials used are biocompatible by observing the growth of cells after 48 hours of incubation. Cell growth was observed through microscope and grown clusters of the cells are visible in Figure 3.3 d around the transparent areas of the device. The cells on top of the electrodes were not visible as the image is acquired using transmission microscopy.

The MDCK II cells were cultured on the prepared device having 10 x 10 matrix of sensors as a proof of concept with each pixel (sensor) covering 300 µm square electrode shape equally spaced at 200 µm. The capacitance was measured after

48 hour of incubation and represented with a bar graph in Figure 3.6a for corresponding density of cells on a specific channel. Since the growth of cells in the overall available space is not uniform, the sensory channels experience different density (cluster) which are reflected from the figure by showing higher capacitance for densely populated cell cluster. Similarly, this behavior is established through repeated measurements against number of channels and scanning the rows in each channel represented in Figure 3.6b. A good fit and repeatability is observed among the nodes in each row with similar density with acceptable tolerances. A persistent trend with increasing capacitance corresponding to the increasing density are observed from Figure 3.6 c & d, retaining similar values across channels scanning the single row. This represents a gradient cellular distribution in the culture space which can be attributed to the unique setup of culture where we had tilted the coverslip while seeding the cells to start proliferation from one side (edge) of the well to other gradually.

Figure 3.6| Cellular growth activity monitoring. (a) Response of a test device to the increasing density of MDCK II cells within the culture well. (b) The repeatability of the measurements established by measuring multiple rows within the matrix. For mapping the growth pattern, (c) a gradient of proliferation was observed through the channels (column with increasing cellular density), and the trend is reflected in all the rows corresponding to the increased density. And (d) shows how the growth of the cells is gradient across the channels.

3.7. Scalability and Yield

We have demonstrated the mapping using a 8 µm pitched 1 million sensory array with a density of 15,625 sensors per mm2. Using advanced lithography and nanofabrication processes, 1 million sensors can be fabricated in 25 mm2 if the

77 sensory electrode pitch is reduced to 3 µm with electrode size at 2 µm. The total coverage area reduction and the increasing density of the sensors is plotted against the decreasing size of the electrodes for scaling purposes. We can observe that if we keep the spacing nearly same as the electrode dimensions, the coverage area reduces to 1/4th for a scaling factor of 1/2 which is a squared relation with scaling as seen. Similar quadratic relationship is observed in the density vs electrode size represented in Figure 3.7b. A nanoscale version using electron beam lithography technique having electrodes as small as 100 nm with 100 nm spacing can give a whopping density of more than 25 million sensors per mm2

(SEM image of the nanoscale electrodes can be seen in Figure 3.2b top right).

Although, such a high density will pose extreme challenges in interfacing electronics development and multiplexing techniques. Multiple devices were tested with the control PHCCBEAD solution added on matrix in steps and removed (dried off) manually. This gives us a very high yield and reliability across the devices with the standard deviation of measurement observed for each measurement point within a device is under ± 1.6% of baseline reading.

Figure 3.7| Scalability and Yield of 1M sensory patch. (a) Comparison of this work with previous status-quo. (b) Scalability of sensors shown by total area coverage per million sensors and the respective density of sensors plotted as a function of sensory electrode size (keeping the pitch equivalent to electrode size). Ideally, design should lie above red line in the blue area close to left axis. (c) The yield of the device performance established using a control experiment of measuring response to (0, 15, 20, and 40 µl) droplet of DI water having PHCCBEADS (size similar to MDCK II cells), acceptable deviation is observed across multiple devices that can be attributed to manual experimental artefacts. Droplet was added using a micropipette and removed manually.

3.8. Conclusion

In this chapter, we have explained the need for flexible pressure sensors. We reported how MEA and multisensory platforms are used for advanced neurological disorders diagnostics, treatment, prosthetics, skin replication using e-skin platforms, and understanding the aerodynamics of racing vehicles. We reported an easy CMOS technology compatible process flow to fabricate high-density pressure sensors on a completely flexible platform (bending radius up to < 500

µm). As the advancements in micro/nanofabrication are progressing, there has been increased interest in understanding and interfacing the human organs for continuous monitoring and possible early diagnosis of chronic diseases. Arrays of neural electrodes/probes are used in advanced neurological disorder diagnostics, treatment, prosthetics, and skin replication using e-skin platforms. We reported a comprehensive CMOS technology compatible process to fabricate a flexible

(bending radius up to < 500 µm) high-density pressure sensory patch for mapping activities. A soft compressive pressure sensor design and the multiplexing/addressing scheme allow highest density integration (15,625 sensors per mm2) with an extremely high-spatial resolutions similar to sub-cellular scale (<

10 µm), at the same time retaining high sensitivity (response to the cellular beads and growth density). We have also demonstrated the scalability of such design at nanoscale using the advanced nanofabrication and lithography techniques to achieve possible density of > 1M sensors/mm2 with the sensory pitch fixed at 1 µm allowing the electrode size variations (50 nm to 500 nm). We devised a simple multi-level addressing technique to access sensors (at different level individually) or local fields (cumulative response) based on the application. The advantage of looking at cumulative sensors/channels response (having hundreds of sensors in a given area) over a single large sensor is established by the ability to detect and respond to small signals like point loads that could often go undetected otherwise.

The mapping of array of million sensors has opened a new opportunities in biomedical applications. We have reported how this million sensors patch can be used to detect the pattern of cellular growth within a media. The response of the

80 individual pixels with high selectivity resulted in excellent response to increasing cellular density and the gradient of proliferation. The usage of conventional materials has established the biocompatibility, ruggedness, flexibility, reliability, and CMOS compatibility. From the density and growth patterns, we believe that number of cells can also be estimated. We believe, nanoscale version of the matrix array will have challenges on the fronts of integrating digital and analog multiplexers on site, processing unit and CMOS based amplifiers for realizing such high-density sensory system for real time applications. We also believe that advanced 3D heterogeneous integration strategies that combine materials, processes, sensors, and different technologies can bring the paradigm shift in flexible electronics on one platform utilizing the maximum area efficiently for advanced applications.

4. Chapter 4: Heterogeneous 3D Integration Strategies (3D Coin)

Technological advances to augment the quality of life need Internet of Things

(IoT) and Internet of Everything (IoE) seamlessly connecting device-data-people- process. An efficient IoT/IoE system needs a pragmatic approach to integrate trillions of sensors on complex, asymmetrical and challenging biological and non- living surfaces. Therefore, robust integration strategy is required for a reliable physically compliant electronic system. There are plenty of efforts going on in the emerging field of flexible electronics, exploring 0D/1D/2D materials, organic materials, polymers, and low-cost processing techniques. Nonetheless, for data management conventional rigid ICs are used [9], [114], [115]. Therefore, different approaches have been explored to use self-assembled layers, thinned devices, micro-bumps, specific bonding layer and high-temperature process-based pressure application for stack bonding. A variety of methods are used to reduce the CMOS devices into a thin and flexible form [77], [116]–[121]. Stacking ICs without substantial thickness reduction for through-silicon-via (TSV) restricts efficient area utilization and yields thicker stack thereby losing flexibility. In-plane single layer assembly of active and passive electronics causes severe skin/cell inflammation by heat dissipation from active devices. This chapter discusses the developed heterogeneous integration strategy using a novel 3D coin based architecture and its reliability.

4.1. Overview of System Integration

The quest for ever-lasting exponential performance improvement in semiconductor technology has led to the investigation of alternative architectures for silicon MOSFET, or integrated circuits. This approach has been called the

More-than-Moore approach by the ITRS, since these advances don’t necessarily follow Moore’s law and are not limited by short channel effects. Several interesting

MOSFET architectures have come into fray during the last few years. These include the FinFET [15], the silicon nanotube FET [16], and wavy channel thin film transistors [17, 18]. To complement these efforts, newer chip architectures are being looked into so that the performance improvement from these transistors can be fully utilized.

4.1.1. 3D Integration

One such research area is the fabrication of 3D integrated circuits. There are mainly two approaches currently being followed for the fabrication of 3D circuits.

In the first approach, the channel material for device fabrication and an insulation layer is deposited repeatedly to obtain devices in a three dimensional structure.

This approach entails the deposition of the channel material on an insulator. Since there is no seed layer for the growth of channel layer, the channel layer has to be polycrystalline at best. Hence, this approach is not applicable to high quality single crystalline channel materials like silicon. In the second approach, the entire silicon substrate is thinned down and separate chips are stacked and bonded to obtain the 3D chip [6], [122]–[124]. In this process, separate circuit blocks are made on conventional silicon (100) wafers. The dies are then thinned down to about 30-40

µm thickness using chemical baths or chemical-mechanical polishing. These thinned dies are bonded together. Lastly, several holes are etched though the thickness of the thinned silicon substrate to form through silicon vias (TSVs).

These TSVs are used as interconnects with the next layer of thinned chip to be placed on top of the previous die [125]–[128]. This process is repeated to stack multiple thinned down chips and form a 3D integrated circuits.

4.1.1.1. Advantages of 3D Wafer Stacking

The stacking of several silicon chips together has several interesting advantages. The number of transistors per unit area of the chip can be increased n times by stacking n chips. This enables circuit designers to incorporate more functionality into a single chip and enables a new generation of powerful but tiny devices. The chips stacked together can be made using different channel materials or specific process modules. Thus, a 3D chip can be designed for a particular application by integrating heterogeneous chips elements together. Since several modules can be stacked together, the on-board wiring length is greatly reduced.

This can make the circuits faster because of smaller interconnects. Smaller interconnect wires also reduce the power dissipated in the wires from Joule heating. Also, shorter wires have smaller parasitic capacitance, which further reduces the power consumption. 3D stacking of separate circuit components reduces the possibility of reverse engineering chip architectures by compartmentalizing the circuit modules. This makes the chip secure against design theft. 3D integration of various circuit models provides an opportunity to introduce higher bandwidth, vertical interconnect buses between the functional

84 blocks. This can greatly increase circuit performance and offers an opportunity for innovative communication protocols.

4.1.2. Flexible Interconnects and Packaging

An electronic product may be folded and twisted to fit into a very limited confined space or through its normal use. Flexibility is also important for many medical applications. As cost continues to decrease, consumers will want electronic products integrated into their garments requiring compatibility with washing cycles in hot water with detergent, hot-air dryers, and the heat and pressure of ironing. ICs and passive devices may be surface-mounted on a flexible interconnecting substrate or integrated as thin film structures into flexible wiring layers. In some applications, these wiring layers must be stretchable. Die thinned to below 20 microns and embedded inside the wiring layers may use compliant bumping techniques for interconnect and thin molding for protection from the environment. Metal-coated polymer fibers could be deployed as conductive wires using a garment as the supporting substrate. Flexibility of interconnect components such as microsprings and compliant bump structures will also be necessary to meet the requirements of future multi-chip packages. SiP with die and other components of varying thickness integrated into a dense package will require compliant structures to compensate for thickness variation, lack of coplanarity, and rigid microbumps.

4.1.3. Heterogeneous Integration

The optimal solution for emerging flexible applications can be at the interface of different materials, processes, technologies. Figure 8 depicts the different technologies that can be incorporated into a complex SiP through heterogeneous integration processes. Efforts to continue the pace of progress of the semiconductor industry have been in two technical directions. In one direction, functional diversification, referred to as “more than Moore,” is the driving force behind SiP based architecture. The second is the use of the third dimension. This allows increase in functional density for a given node and also improves performance and reduces power requirements. Potential benefits include higher performance, reduced power requirement, reduced latency, smaller size and eventually lower cost compared to 2.5D and conventional 2D packaging [129]–

[132].

In this direction, today’s state-of-the-art replaces wire bond in staggered stock with stacked die using the TSV. Specifications of devices using TSV based package have higher speed, decreased package volume, decreased latency, and reduced operating power, 50% (Toshiba data) [6], [122], [123], [133]–[137]. In addition, the heat generation in ICs is highly non-uniform with few locations having very high local heat fluxes. Future trends of thermal design power show an increase in both average power density and local power density (known as hot spot). It limits the thermal solution for the package. Even if the total power of the components remains within the designed constraints, the hot spot power density increase could limit the device performance and reliability.

4.2. Challenges

The advantages of 3D IC have gathered major attentions in semiconductor research in the past few years. Major challenges are power requirement and integrity, thermal management and the limitations in physical density bandwidth.

Total power can be reduced by reducing operating voltages, reducing interconnect distance, reducing capacitance and reducing operating frequency. Stacking the circuit in layers reduces the interconnect length by the square root of the number of layers. However, thermal density is increased and surface area to exhaust the heat is reduced. The heterogeneous integration of CMOS based dies and polymers and passive electronics with the communication and energy management modules has several major advantages over the conventional monolithic integration approaches. Still, several challenges are often unaddressed that must be given due consideration for solving system integration.

Already the semiconductor industry packs in over a billion transistors in a single chip. With stacked integrated circuits, a defect in any of the chips or interconnects can damage the functionality of the 3D IC. Thus, yield will need to be further enhanced. A major problem with stacked integrated circuits is the buildup of heat inside the chip. Specific areas of the chip with high horizontal and vertical density of devices may become hotspots and may cause chip failure.

Careful design of circuits to distribute the heat load as evenly as possible will be required. The use of TSVs for vertical interconnects occupies a large footprint. This reduces the density of devices on all the chips. Further, TSVs introduce complexity in placement and routing designs. Additionally, 3D stacking of separately made

87 parts involves the thinning down of silicon substrate for TSV formationthat results in a significant loss of substrate material.

Majority of work related to flexible electronics involves usage of polymers and one of the key processing steps is transfer of the polymers. There is no technology that can guarantee accurate transfer process for flexible substrates without incorporating wrinkles, alignment errors, and mechanical damage due to handling of flexed platform. One of the important aspects in 3D stacking for heterogeneous materials or thin chips is bonding material selection. Often, microbumps, BCB polymer, and other specialized temporary adhesives are used, which poses new dimensions and constraints on overall integration process. For demonstrating any application, multiple ICs integration is a must. Bare dies of different ICs, like microcontroller, transceiver, and solid-state batteries, all have different thickness. In this case, after assembly, all of the ICs have to be reduced to thickness of < 15 µm using RIE and, hence, achieving uniform thickness of multiple ICs having variable initial thicknesses is a challenge. The major challenge of any stand-alone system is the power source. A major chunk of power is consumed by communication thus, for any module to be stand-alone a paradigm shift in energy harvester is a must. Here, we propose use of multiple solid state batteries in addition to an RF energy harvester for communication and energy management. The solid state batteries can provide small current while the processor can be optimized for such parameters. The final challenge but certainly not least important is that the integration processes scheme must be compatible and scalable to manufactural scale for ready adoption.

In this work, we intend to demonstrate an integration sequence for convolution of heterogeneous materials, processes, and technologies in 3D platform using TSV and through-polymer-via (TPV). In addition, we reflect on the packaging materials and the alternative that we propose for application specific flexible electronics.

4.3. 3D Coin Architecture

To overcome these challenges, here we show a reliable heterogeneous integration scheme to obtain a physically flexible standalone CMOS electronic system using a coin architecture (Figure 4.1). In a 3D-coin architecture, heterogeneous materials and technologies like CMOS, polymers, and graphene are integrated together on a single platform. A sequential-etch-back (SEB) technique is used to transform state-of-the-art CMOS ICs into flexible ICs, converted into flexible LEGO for secure identification and accurate assembly using modular LEGO electronics approach like a plug-and-play (or lock-and-key) mechanism [120]. The 3D coin architecture hosts different materials (copper and graphene) based sensors on one side of the coin interacting with electronics for processing, sandwiched in the polymeric platform using through-polymer-via

(TPV). The other side of the coin architecture is reserved for antenna, for data communication, and energy harvester, to compliment the solid-state micro- batteries that are powering the electronic system embedded in the polymeric package. The polymeric encapsulation for the electronic interface ensures the biocompatibility as well as ruggedness [138], [139]. In addition, this coin- architecture provides physical isolation of the electronics and communication

89 element from the sensory layer, thereby, not interacting with the host body during various monitoring applications.

Figure 4.1 | Concept of 3D-coin architecture illustrating multilayered architecture. Sensors are physically isolated from the electronics embedded in the polymer. RF and solar energy harvesting modules are integrated on the top face of the coin away from the sensing face.

4.3.1. Heterogeneous Integration Fabrication Scheme

Complete fabrication schematic for 3D-coin heterogeneous integration is illustrated in Fig. 4.2 A and B. We start with spinning a thin (< 10 µm) polyimide

(PI) on a bulk silicon (100) carrier wafer and cure gradually in multi-steps (90–300

°C) [139]. This layer acts as a host substrate for LEGO-based lock-and-key assembly and provides mechanical strength to the entire flexible system. Next is trench formation in the PI layer for vias connection to the bottom face of the flexible platform, reserved for sensor integration. TPV though PI is etched using a metallic hard mask of Ti 50 nm sputtered using DC under 25 SCCM of Ar at 400 W power and 5 mTorr chamber pressure. Deposited hard mask needs to be patterned using photolithography, here we use ECI 3027 positive PR (4 µm spin coated and baked at 100 °C for 60s). Following the PR baking, nano-imprint contact aligner is used to expose light at a constant dosage of 200 mJ/cm2, using the mask for the vias fabricated on quartz material. Exposure is followed by development of PR in AZ726

MIF developer for 60s and then hard mask is etched using physical plasma based

RIE at 10 °C for 1 minute in presence of 30 SCCM Ar gas carrier and 5 SCCM of

SF6 at 10 mTorr pressure. PI is then dry etched using RIE with a gaseous mixture of CF4 and O2 flowed at 50 SCCM and 5 SCCM, respectively, at 80 mTorr pressure and 60 °C chamber temperature. Helium gas cooling is enabled at the bottom of table to stabilize the rising temperature during etching process. After trenches in

PI are formed, hard mask can be removed using wet chemical etching or dry etching. If Cu metal is used as hard mask for PI etching, then this can also act as seed layer for next step.

Figure 4.2| Heterogeneous integration flow for obtaining 3D compliant standalone electronic system

Figure 4.3| Demonstration of 3D coin standalone electronic system (a) 3D-coin with (b) MSP430TM, (C) 1 million nanosensors array, (d) flexed solid-state microbattery, and (e) graphene antenna. Flexibility demonstration and comparison with the rigid bare die version of ICs and entire system (f-h)

Next, we sputter thin Cu (100 nm) layer for circuit layout and interconnects on PI to form female LEGO site for multiple die assembly. Since we need to cover the trenches and also have reliable interconnects, we use ECD copper to fill the vias midway providing good reliability to the vertical bottom interconnects. Physical

93 characterization of such half-filled TPV and empty TPV profile are shown in Figure

4.3 A. Parameters of ECD Cu are exactly similar to those used earlier where 100 mA current is passed through copper sulphate solution at 24 °C for 10 minutes in alternate mode to grow ~3-5 µm of Cu, filling the TPV halfway. This metallic interconnect trace is followed by second PI layer for insulation and provides second layer geometrical binding sites for LEGO modules (Figure 4.5 D and E).

Same process is repeated to make different LEGO patterns in the PI to form binding sites.

Figure 4.4 | SEM of TPV to study the profile

4.3.2. Modular Lego Assembly

We presented modular LEGO electronics as solution to interconnect problems and novel integration method in chapter 2. A microcontroller unit (MCU)

(Texas Instruments MSP430™, 8kB Flash and 256B RAM, thickness = 280 µm),

Bluetooth based RF communication IC (µElectronicsTM EM9304, thickness = 260

µm), and solid-state micro-batteries (Cymbet™ CBC005, thickness = 200 µm) are assembled to demonstrate a stand-alone system. We convert these bare die ICs into LEGO modules providing unique identity and locking features using the process described in chapter 2, section 2.3.1. SEM images of converted IC into

LEGO modules are shown in Figure 4.5 F. These LEGO modules in rigid die form are then assembled on the binding site using a soft-vacuum suction-based pick- and-place tool for high yield (7000 UPH) with a placement accuracy of ±2 µm [8].

These rigid LEGO modules do not impose any challenges in handling associated with ultrathin chip and flexed ICs. We have used 30 µm thin layer of polydimethylsiloxane (PDMS) as bonding layer. To protect the lateral IC shrinkage occurring due to DRIE or other flexing methods, hydrophobic PDMS is spin-coated as sidewall spacer, SEM of the assembled dies and PDMS validates the sidewall spacer hypothesis, Figure 4.4. Thus, PDMS layer acts as hard mask for DRIE of bare die ICs to achieve ultrathin ICs (< 20 µm). Next, we transform the rigid bare- die ICs into flexible form by performing sequential back etch process. The DRIE process is a modification of BOSCH process where we perform multiple repeated steps of C4F8 polymer deposition followed by etching using SF6 and Ar to achieve the vertical etch profile. These flexed versions of ICs are now embedded using

PDMS as encapsulating layer (100 µm, cured at 70 °C) to keep the maximum thermal budget less than 90 °C.

Figure 4.5 | SEM validating the PDMS as soft-mask and sidewall spacer to protect lateral etching and shrinkage of bare die IC during DRIE based back etching process.

4.3.3. TPV, Antenna, and Solar Cell

The encapsulated sandwiched electronic system then has via formation through PDMS layer using laser ablation. A CO2 laser (wavelength = 10 µm) is used to create TPV in 100 µm thick PDMS reaching the interconnecting pads on the sandwiched metallic layer to connect BLE IC and batteries with the top surface.

We transfer chemical vapor deposited (CVD) grown graphene on PDMS using

PMMA for fabricating a 24 GHz transparent antenna, patterned using a CO2 laser

(wavelength = 10.6 µm). The connection between the graphene antenna and sandwiched processing unit is established using a laser-ablated through polymer via (TPV) in PDMS layer filled with electrochemically deposited (ECD) copper.

While the process to fill TPV is already described in the previous sections, the point to be noted here is that the antenna is also patterned in the same step during ECD, by reason of same metallic elements are used and ECD Cu provides larger

96 thicknesses, thereby improving the antenna performance. This also creates seamless integration as there is no need of separate bonding/soldering to integrate the components. A solar cell is placed and connected to the system, binding strongly due to sticky nature of PDMS under the contact pads and surroundings.

Thin 50 µm PDMS encapsulation is completed to protect the antenna as well as creating a corrugation pattern atop the solar cell to convert it into flexible form using corrugation technique [140]. Final processing step on front (or top face) of the 3D- coin architecture is thin 100 µm PDMS encapsulation layer acting as soft packaging material to provide mechanical and chemical protection (Figure 4.5 B and C).

Figure 4.6| SEM images at different stages of integration scheme. (A) Thinned bare-dire MCU after DRIE and using PDMS as hardmask and sidewall spacer, (B) and (C) IC die embedded in PDMS, (D) interconnects in the sandwich layer (E) Binding sites and interconnects for LEGO modules and half-filled TPV, and (F) Converted rigid ICs to LEGO modules with locking pads.

4.3.4. Sensory System Integration

The flexed ICs and antenna system encapsulated with 100 µm PDMS can be flip transferred onto a carrier wafer having cured PDMS to perform sensory integration. The encapsulated platform can be easily peeled from the Silicon host carrier wafer due to its relatively low adhesion properties with PI. From the bottom surface, the TPV filled interconnects are readily available to integrate the sensors directly with the platform. The integration can be performed using soldering, conductive epoxy, wire bonding, or by fabricating sensors directly on the platform.

As we have demonstrated high-density 1 million sensors in chapter 3, we integrate similar sensors directly fabricated on the available bottom surface (now flipped).

Variety of sensors like temperature, pressure, salinity, and humidity can be fabricated using the metallic interconnect layers [72], [75], [76], [138], [141]. Finally, the system is encapsulated with another 100 µm PDMS to neutralize possible mechanical stress mismatch from the top and bottom side of the 3D-coin sandwich structure. As a result, a fully flexible 3D electronic system is obtained which can be seen flexed around 5 mm bending radius in Fig. 4.3 inset.

4.4. Results and Discussion

4.4.1. Electrical Characterization

To validate the integration sequence, we have performed electrical characterization of interconnects, TPV, MCU, and sensors in varying conditions to understand the process-induced or mechanical bending induced variations and studied the effect of volumetric thickness reduction. As we know the electrical

98 connections through-out the system are established using TPV, connecting the circuit trace lines and multiple vertical layers, we have characterized the established connection by measuring the impedance of the trace line when a

LEGO module is assembled, TPV at different sites, and also the resistance of the entire circuit trace when all the modules are assembled in the sandwiched platform. Figure 4.6 A and C presents IV curves for the two measurements, while

B and D represents the insignificant change observed in the measured impedance, which can be attributed to the different positions of TPV, and varying length of the trace.

Figure 4.7| Reliability of interconnects (impedance measurements) (A) IV curve to measure resistance of TPV in 3D integration, (B) change in resistance plot. (C) IV curve of the circuit trace when the bare die are assembled using Lego approach sequentially, and (D) change in resistance measured along the trace when all the modules are connected sequentially as LEGO, Vdd, BLE, and Processor respectively.

4.4.2. Sensory System Characterization

Implementation of IoT and IoE rely on the integration of different standalone sensory systems. Typical block diagram of a stand-alone electronics system and the architecture proposed for 3D coin (application shown in Figure 4.8) is illustrated in Figure 4.9. Here, sensors from bottom layer, antenna and solar cell from top layer encapsulated, while the middle sandwich layer consist of associated electronics (processor, RF IC, memory storage, analog to digital converter (ADC), flash, and solid-state batteries).Hence, we have fabricated sensing modalities that are often required for healthcare applications like temperature and humidity sensors. We have used graphene as a sensing material on interdigitated electrodes patterned for humidity sensing with 2 µm pitch and 3 µm electrode in an area of 25 mm2. The design and fabrication of temperature sensor are adapted from [72], [75], [76], [141]. Advanced lithographic techniques can scale it further to integrate same 1M sensors in 0.25 mm2 area, if the sensing electrodes have size of 100 nm spaced at 400 nm and using nanoscale devices for multiplexing these sensing electrodes. SEM images of a million sensory electrodes for humidity sensing and Raman spectra of graphene layer can be seen in Figure 4.8 B and C.

The performance of fabricated temperature sensors using sputtered copper (200 nm) on front (top) face of coin is compared with a commercial packaged reference

100 sensor from Sensiron™. We observe from Figure 4.8 C that the response of both the test and the reference sensors follow same path when the temperature is increased from room temperature to 90 °C, manually using a hot air gun. Similarly, we report the measured variations in the humidity sensor fabricated on the front face (top) of the 3D-coin interface. Figure 4.8 E shows the changes in capacitance of the sensor on application of external stimulus in the form of human breath at different intervals with varying humidity level. Sensors response matches with the reference sensor from Sensiron™, as expected. An efficient multiplexing technique can be used to acquire the data from 1M sensory electrodes.

Figure 4.8| Sensory system and heterogeneous integration on 3D-coin. (A) components of 3D coin (digital image), (B) SEM image of 1 million nanosensor array of graphene electrodes, and (C) Raman spectra of CVD grown graphene. (D) Temperature and (E) Humidity sensor performance compared with reference sensor following the trend as expected.

101

Figure 4.9| System architecture for stand-alone electronic systems. (top) a typical block diagram, (bottom) detailed architecture of 3D coin stand-alone system.

102

4.4.3. Process/Mechanical Induced Effects on Bare Dies

The heterogeneous integration scheme incorporates the bare die form of the commercial silicon based ICs for MCU, BLE, and solid-state batteries.

Conventionally, bare dies are very sensitive to processes, high temperature and environmental parameters, thereby prone to possible performance. Accordingly, it is critical to evaluate if there is any performance degradation due to the different processes or due to mechanical deformations when the dies are in ultrathin flexible forms; known as process or mechanical induced variations. To study these effects, we have selected a single die (MCU MSP430TM unit) as a reference die and measured the current consumption at unprocessed stage and at different steps of the integration scheme. Variation in current consumption of the IC is recorded during code writing, in the idle state after writing, and during the code run sequence

(active state). The same measurements are recorded for the MCU at unprocessed stage, after LEGO transformation of thin IC of, and after making TPVs for vertical interconnects in the flexible bent state with a bending radius of 3 mm (Figure

4.10a). Figure 4.10b illustrates the plots obtained for the average current consumption from the MCU when it is programmed to constantly blink an external

LED. For an unprocessed die, the consumed current of the MCU and integrated

LED is recorded at 3.3mA when LED is ON (, ~0.400mA when LED is OFF in the active state and ~0.108mA in the idle state (after writing). For both the rigid die after LEGO transformation and the thin flexed die (<20 µm) after DRIE processing, the LED OFF active state consumed current did not change significantly, recorded at ~0.402 mA and ~0.405 mA, respectively. The current in the idle state was

103 measured to be 0.111mA for both processing stages and is an insignificant change as it is < ± 3%.

Similarly, the same average current consumptions of the MCU and integrated LED are presented in Figure 4.10 d to understand mechanically induced variations due to bending (Figure 4.10 c bending set-up). It can be clearly observed that the current consumption of thin MCU after DRIE changes negligibly from

~0.402 mA to ~0.405 mA in the bent state (bending radius 3 mm) in active mode.

The same values drop from ~0.111 mA to 0.109 mA in the idle state after writing the code sequence. Thus, we infer that our integration sequence is the most pragmatic route for flexible electronic based IoT and IoE applications.

104

Figure 4.10| Reliability of the integration process demonstrated by electrical characterization of MCU at different processing stages. (a) rigid bare die and flexed version comparison showing the necessity of flexing ICs, (b) average current consumption of MSP430 during writing a programming code, in the idle state, and in active state for blinking LED at different processing steps. (c) Characterization of bending induced performance at different bending radii (d).

Similarly, the performance of the entire system with an integrated BLE and antenna is evaluated by programming the MCU to detect a threshold temperature.

In this demonstration, threshold temperature of 25 °C is set in the MCU which turns on the LED by sending a signal, represented by the current consumption of the

MCU and BLE unit in both active mode and when the signal is transmitted to turn on the LED using BLE IC. The plot in Figure 4.11 a shows 80 µA current in active mode (LED OFF) since we programmed it to run in low power mode, and ~3.4 mA when LED is turned ON after detection of passing a temperature threshold. The designed antenna S11 characteristics and its radiation pattern are also illustrated in Figure 4.11 b and c.

Figure 4.11| System demonstration using temperature sensor, BLE, and antenna radiation pattern. (A) threshold detection using fabricated sensor for > 25 C and responding by turning on LED, (B) S11 reflection coefficient and (C) radiation pattern for the Serpienski fractal design antenna with radiation frequency of 24 GHz and 4 GHz bandwidth.

105

4.5. Conclusion

Heterogeneous integration of materials, processes and technologies is the way forward for efficient implementation of IoT and IoE applications. Therefore, we have presented a low temperature, CMOS compatible, heterogeneous integration scheme, which can give a fully compliant electronic system free from any rigid component and in 3D coin architecture for maximized area efficiency and reduced thermal impact on interfacing skin/cells. We have reported the reliable process of making IC flexible using SEB (final thickness < 10 µm, rms roughness of 5.46 nm, and bending radii of up to 700 µm), without any performance degradation.

Integration of graphene based 2.4 GHz antenna, sensors array, and processing unit is reported. This is the most pragmatic and feasible approach for implementation of advanced applications. We report a CMOS compatible manufacturable heterogeneous integration scheme for achieving physically compliant (< 1 mm bending radius) coin architecture 3D-CMOS electronics with integrated graphene sensors for maximized area efficiency in IoT devices. Sensors are placed on one side of the soft-polymeric platform, like a coin, whereas commercial CMOS based 16 MHz microprocessor and micro-battery (5 µAh @ 3.9

V) are embedded in the middle layer (flexed form) and having a 24 GHz graphene fractal antenna on the other side of the coin. This thin coin-system (< 1 mm) can be used as an add-on electronic tag making things smart.

106

5. Chapter 5: An Application- (Non-invasive Featherlight Marine-Skin

Tagging System)

The marine environment - a continuous provider of valuable goods and services for decades, has been directly or indirectly altered by extensive human impacts. Direct harvesting, rigorous over exploitation by fisheries, run-off of nutrients and pollutants, and pollution contribute to the anthropogenic changes occurring in the oceanic ecosystem [142]. The impact of human activities on marine ecosystem varies with the intensity of resources extraction, pollutant addition, and the changes in the species composition. It is estimated that 41% of the oceans are impacted by multiple activities [143]. To facilitate policy implementations, trade-offs and devise mitigation strategies at global scales, it is of utmost importance to quantify and map environmental variables and human activities throughout the marine environment [143]–[146]. Recent developments in electronic tagging devices and animal-borne bio-loggers have facilitated quantifying ecosystems and the effect of humans. Animal-borne tagging devices are used to understand animal migration (plasticity of dates and routes, and mechanisms), foraging behaviour, physiological performances, habitat selection, and social interaction with other species, as well as abiotic environmental variables

[138], [147]. To maintain the physiology, normal behaviour, and the survival of tagged animals, the weight of the electronic tag should not exceed 2% of the body weight [148]–[150]. Most of the available CTD (conductivity, temperature, and depth) bio-logging devices fail to adhere to the 2% weight regime, in addition, they are rigid, bulky, expensive, and not suitable for smaller species, young specimens,

107 and invertebrates [151], [152]. Invasive methods of attachment, as well as bulky and expensive devices have restricted the focus of studies to larger species such as dolphins, sirenians, sharks, and cetaceans. Commonly adopted practices for the attachment of devices use a crossbow or a shotgun to insert the device in the tissue of the species. Incisions and surgical tools and/or mechanical bolts are frequently used for fixation of the tags to the fins of the animals [153]–[155].

However, these attachment methodologies fail to sustain extended periods underwater due to the drag forces and behavior of the marine species (both tagged and surrounding). Studies have revealed the serious repercussion of the weight and design of the tag, and the invasive attachment methods which influence diving patterns, swimming, foraging, mating and nesting behaviours [156]–[159].

Despite the technological advancements in marine tagging electronic devices (e.g. CTD), the majority of limitations related to animal comfort, non- invasive attachment, and adaptability to the myriad of aquatic species (from a tiny gold fish to large mammals) remain unaddressed. In addition, prolonged lifetime of the tag, increased sensing capabilities, and conformal (flexible) design of the devices that do not influence the natural behaviour of animals, present another level of challenges. Addressing the reliability of performance under any mechanical deformation, extreme pressures and complex behaviour after extended submersion in seawater without any biofouling is also important. Therefore, there is an urgent need to focus efforts towards making marine electronic tagging systems that are lightweight, conformal, and do not exceed the cost of currently available sensors. Moreover, the packaging should be biocompatible, robust, leak-

108 proof, and reliable for sustained underwater exposure. The tag must be non- invasively capable of monitoring the marine environment even at great depths, without hindering the current status-quo in terms of resolution and high performance while having a simple attachment mechanism.

Convolution of state-of-the-art complementary metal oxide semiconductor

(CMOS) technologies, materials, and advances in the field of flexible electronics is driving the technologies of the future, where data, processes, sensors, and living and non-living things interact in synergy for Internet of Everything (IoE) applications [120], [160]–[162]. Recently, we demonstrated “Marine-Skin”- a lightweight non-invasive physically flexible skin-like multifunctional electronic tagging system with waterproof packaging (Figure 5.1) [138]. The Marine-Skin tagging system was lighter (~6 g in air) than ever reported previously, conformal with biocompatible packaging, and a non-invasive tagging method never reported before. The resolution and the accuracies of the sensory system were not compromised after deployment or from the physical deformations when attached to the hard-shell crustaceans using glue. That Marine-Skin system was tested in a controlled environment in the lab for different measurements where maximum depth of water was < 80 cm. In addition, the deployment on crustaceans demonstrated the real-time acquisition of data at shallow water depths (<60 cm).

The majority of the small fishes, many sharks, dolphins and other species primarily inhabit the ocean’s surface (<200 m depth), but also dive deeper than 600 m in search of prey and routine behaviour. Consequently, it is necessity to have sensors that withstand the pressure of water at great depths (at least within 600 m).

109

Furthermore, the attachment method used previously (super-glue or waterproof adhesive) only works on animals with exoskeletons or shells, such as crustaceans or sea turtle.

5.1. Introduction

The marine environment - a continuous provider of valuable goods and services for decades, has been directly or indirectly altered by extensive human impacts. Direct harvesting, rigorous over exploitation by fisheries, run-off of nutrients and pollutants, and pollution contribute to the anthropogenic changes occurring in the oceanic ecosystem. [142] The impact of the human activities on marine ecosystem varies with the intensity of resources extraction, pollutant addition, and the changes in the species composition. It is estimated that 41% of the oceans are impacted by multiple activities [143]. To facilitate the policy implementations, trade-offs and devise mitigation strategies at global scales, it is of utmost importance to quantify and map environmental variables and human activities throughout the marine environment. [143]–[146] Recent developments in the electronic tagging devices and animal-borne bio-loggers have facilitated quantifying ecosystems and the effect of humans. Animal-borne tagging devices are used to understand the animal migration (plasticity of dates and routes, and mechanisms), foraging behaviour, physiological performances, habitat selection, and social interaction with other species, as well as abiotic environmental variables

[138], [147]. To maintain the physiology, normal behaviour, and the survival of tagged animals, the weight of the electronic tag should not exceed 2% of the body weight. [148]–[150] Most of the available CTD (conductivity, temperature, and

110 depth) bio-logging devices fail to adhere to the 2% weight regime, in addition, they are rigid, bulky, and expensive, not suitable for the smaller species, young specimens, and invertebrates. [151], [152] Invasive methods of attachment, as well as bulky and expensive devices have restricted the focus of studies to larger species such as dolphins, sirenians, sharks, and cetaceans. Commonly adopted practices for the attachment of devices use a crossbow or a shotgun to insert the device in the tissue of the species. Incisions and surgical tools and/or mechanical bolts are frequently used for fixation of the tags to the fins of the animals. [153]–

[155] However, these attachment methodologies fail to sustain extended periods underwater due to the drag forces and behavior of the marine species (both tagged and surrounding). Studies have revealed the serious repercussion of the weight and design of the tag, and the invasive attachment methods which influence diving patterns, swimming, foraging, mating and nesting behaviours. [156]–[159]

Despite the technological advancements in the marine tagging electronic devices (e.g. CTD), the majority of limitations related to animal comfort, non- invasive attachment, and adoptability to the myriad of aquatic species (from a tiny gold fish to large mammals) remain unaddressed. In addition, prolonged lifetime of the tag, increased sensing capabilities, and conformal (flexible) design of the devices that do not influence the natural behaviour of animals, present another level of challenges. Addressing the reliability of performance under any mechanical deformation, extreme pressures and complex behaviour after extended submersion in seawater without any biofouling is also important. Therefore, there is an urgent need to focus efforts towards making marine electronic tagging

111 systems that are lightweight, conformal, and that do not exceed the cost of currently available sensors. Moreover, the packaging should be biocompatible, robust, leak-proof, and reliable for sustained underwater exposure. The tag must be non-invasively capable of monitoring the marine environment even at great depths, without hindering the current status-quo in terms of resolution and high performance while having a simple attachment mechanism.

Convolution of state-of-the-art complementary metal oxide semiconductor

(CMOS) technologies, materials and advances in the field of flexible electronics is driving the technologies of the future, where data, processes, sensors, and living and non-living things interact in synergy for Internet of Everything (IoE) applications. [120], [160]–[162] Recently, we demonstrated “Marine-Skin”- a lightweight non-invasive physically flexible skin-like multifunctional electronic tagging system with waterproof packaging (Figure 5.1) [138]. The Marine-Skin tagging system was lighter (~6 g in air) than ever reported previously, conformal with a biocompatible packaging and a non-invasive tagging method never reported before. The resolution and the accuracies of the sensory system were not compromised after deployment or from the physical deformations when attached to the hard-shell crustaceans using glue. That Marine-Skin system was tested in a controlled environment in the lab for different measurements where maximum depth of water was < 80 cm. In addition, the deployment on crustaceans demonstrated the real-time acquisition of data at shallow water depths (<60 cm).

The majority of the small fishes, many sharks, dolphins and other species primarily inhabit the ocean’s surface (<200 m depth), but also dive deeper than 600 m in

112 search of a prey and routine behaviour. Consequently, it is necessity to have sensors that withstand the pressure of water at great depths (at least within 600 m). Furthermore, the attachment method used previously (super-glue or waterproof adhesive) only works on animals with exoskeletons or shells such as crustaceans or sea turtle.

Figure 5.1| Robust compliant Marine Skin illustration. Different marine species swimming at varying depths under the dark see and respective temperature ranges experienced. The digital photograph of the multisensory wearable Marine Skin tagging gadget for marine environmental monitoring.

5.2. Challenges in the Marine Monitoring Systems

Conventionally, CTD based tagging systems have been used to monitor three fundamental parameters (conductivity, temperature, and depth) of the marine environment to understand the composition, water masses and the niche

113 used by marine life [163]. There still exists major challenges for developing marine sensors, including underwater data acquisition. There are huge efforts to develop reliable underwater communication including acoustic and optical communication.

Optical communication under water presents a lot of promise to solve this persistent challenge, however these technologies are in their infancy [164]–[167].

Thus, an alternative approach is to store the data in memory embedded on the platform and transfer the stored data when the animal comes to the surface of the water during its natural behavior. Hueter et al. recently demonstrated the usage of satellite linked MiniPATTM system for tracking the movement of Carcharhinus

Falciformis (silky shark) to understand the ecology of this large, oceanic species

[168]. The attachment method consisted of inserting a plastic or stainless steel anchor into the skin, which is not only invasive but also discomforting to the animal.

Moreover, the device is not attached to the animal’s body, but is hanging with a support, which can be harmful during movement and could be detached (break).

In our Marine-Skin system, we integrated the Bluetooth system on chip

(SoC) and memory module with a small battery, which were encapsulated in a waterproof and biocompatible packaging. Great efforts were taken to make the system flexible, stretchable, and feather light, which can be attached non- invasively using a wearable soft jacket architecture. The SoC and memory module stores the acquired data underwater and can be retrieved from memory after the tag is recovered by connecting it to a Bluetooth enabled device. The flexible, small, and light weight platform combined with the wearable jacket design resulted in a

114 sensor with minimal discomfort to any tagged species irrespective of their size or epidermal type.

5.3. Biocompatible Soft Packaging

Any sensory platform designed for the marine environment must retain its performance under high salinity, high pressure and this corrosive aqueous environment. The salinity of sea water varies in different geographical locations; nevertheless, it can range from 35-40 practical salinity unit (PSU) [169], [170].

Therefore, the first challenge in our Marine-Skin tagging development was to have a compliant packaging material that is soft to adhere to the animals’ skin and elastic to maintain stretchability for comfortable breathing and comply with physical deformation of the device. Different material choices available for such soft encapsulation are Ecoflex©, or polymethyl methyl acrylate (PMMA), or polydimethylsiloxane (PDMS). We chose polydimethylsiloxane (PDMS) (Sylgard

184TM) as the soft encapsulation of the sensory platform due to its hydrophobic, non-toxic, non-decomposing, and non-irritating properties. In comparison,

Ecoflex© undergo biodegradation and PDMS does not decompose or undergo major polymeric deformations when exposed to certain kinds of microorganisms found in seawater [171]–[173]. Further, PDMS is one of the most used materials in microfluidics and biosensors because of its biocompatibility. In addition, PDMS has a hydrophobic surface giving lower surface energythat provides a weaker adhesion of a possible biofouling layer on the PDMS surface. A textured PDMS surface has also been reported to provide a weaker adhesion of biofouling

115 elements thus making it easier to shrug off after a simple rinsing process [174]–

[176].

5.4. Multisensory Complaint Sensor Design

“Marine Skin” is a stretchable and flexible multisensory platform that monitors temperature, pressure (depth), and the salinity of the marine environment. The pressure sensor is based on array of multiple pixels of capacitors all connected in parallel. If the end electrodes are not connected, the pressure at individual pixels can be mapped. We use PDMS as a soft and compressible dielectric material for the pressure sensors. The sinusoidal wavy architecture of metallic interconnects allow stretchability in lateral directions while also maintaining twisting and bending capability due to inherent soft elastic properties of PDMS encapsulating material [34,35]. A resistive temperature detector (RTD) is fabricated for temperature measurements while salinity sensor was incorporated by using an interdigitated electrode architecture for enhancing the sensitivity. The miniaturized version 2 measures 20 mm × 20 mm × 0.3 mm. All metallic conductors have a temperature-dependent resistance governed by the Equation (4).

푅 = 푅푟푒푓 (1 + 휍(푇 − 푇푟푒푓 )) (5.1)

Where ‘R’ is the resistance at a given temperature ‘T’ and Tref is the temperature at a known temperature. ’ζ’ is the temperature coefficient of resistance.

116

Currently available electronic products on the market for marine animal tagging are rigid, with packaging in a stiff acrylic or plastic package. The core of the tag is a printed circuit board (PCB) that hosts different components and the sensor integrated circuit (ICs). Having all the rigid ICs with passive electronic components (resistors and capacitors) soldered on the rigid PCB, no matter how small, will not make a system flexible. There are many groups that focus on polymeric materials for using as a flexible substrate in flexible electronics. It must be noted that merely changing the substrate from PCB and mounting the same rigid ICs on the flexible PCB does not make it flexible for the desired application.

Moreover, many presented flexible systems have flexible sensory parts but the core of the readout circuit is a microcontroller or data acquisition system that in itself is not flexible, also, the sensory platforms are not stand-alone [114], [117],

[177]. Most often, these flexible sensors are connected to the data acquisition system by either wire-bonding or soldering wired connectors that again makes the system vulnerable for failure. Hence, a practical approach for obtaining a fully compliant system needs absolutely no rigid components. This can be achieved if we can integrate the unpacked (bare die) version of the ICs required in the system and make them flexible.

Here we present a complete integration strategy for the multisensory

“Marine Skin” platform and the detailed 3D schematic flow is presented in Figure

5.6. This integration flow was designed taking into account the low-cost CMOS fabrication approach, scalability, high-yield, batch fabrication, and precision in

117 alignment at bare-die level. Precise details of the entire integration flow and the materials are provided in section 5.6.

5.5. Material Optimization

Marine species experience varying oceanic environments during their natural movements, migration, foraging for food and evading predation. An electronic tagging system must be able to withstand these varying extreme conditions and individual experiences. Hence the material choices and the design of each sensor is central to the overall system reliability and sustenance.

5.5.1. Temperature Sensor

The temperature of seawater drops from 27 °C at the surface to 2 °C as the depth reaches approximately 4000 m, highlighting the need for highly sensitive temperature sensors for detection of small variations [145], [178]. Oceanic temperatures are relatively stable in response to changes in climate, for example the change in oceanic temperature is effectively 0.1 °C for 0.6 °C change in global average temperature over the last century. However, the temperature of the seawater varies with increasing depth (200-1500 m) known as thermocline region, which also has significant effect on the marine ecosystems at different depths

[145], [178], [179]. Traditionally, gold has been a primary material choice for biological applications due to desirable properties like biocompatibility, corrosion resistance, and non-toxicity. Consequently, in our Marine Skin version 1, we used

Au as the metal for all the sensors and interconnects. However, platinum is widely

118 known for its usage in RTD sensors therefore, we compared our version 1 and version 2 of Marine Skin having Au versus Pt as temperature sensor.

We compare the performance of the temperature sensor with the performance of the commercial temperature sensor integrated circuit (IC) from

SensironTM. This commercial temperature sensor is attached next to the fabricated temperature sensor on the wafer for accurate temperature detection. Both the reference and test sensors are subjected to heating from room temperature of 21

°C up to 80 °C using a hot air gun. Initially both the sensors record the values for room temperature which starts increasing as the hot air gun is gradually brought closer and closer to the sensor. From Figure 5.2, we can clearly observe that change in resistance of version 2 sensor follows exactly with the reference standard. Instantaneous variations in the resistance change not only suggest that the resolution is high but also the response and recovery times match or are at least on par with the status-quo, if not better. Figure 5.3 represents drastic improvement (10 folds) in the temperature sensitivity from 43.18 mΩ/°C for Au in version 1 to 487.5 mΩ/°C for Pt in version 2.

119

Figure 5.2| Comparison of the performance of 2 different designs for temperature sensors (A) version 1 and version 2 temperature sensor for material choice and improved sensitivity, and (B)Version 2 comparison and calibration with the commercial temperature sensor IC from SensirionTM. Fabricated sensors changes the resistance corresponding to the temperature with the response exactly following the reference sensor.

120

5.5.2. Salinity Sensor

Similarly, the salinity of the seawater is fairly constant at the surface and at great depths (> 800 m) but it varies rapidly within the halocline region (~ 300 –

1000 m) similar to thermocline region for temperature. Variations in the oceanic salinity have been reported to effect water cycles and oceanic circulation. Factors that gradually increase the salinity of oceanic surface are naturally counterbalanced by inflow of fresh water from rivers, global ice melting, and precipitation of rain water[146], [180]. Ocean surface salinity is also one of the key parameters in understanding effects of fresh water intake on ocean dynamics due to occurrences of 86% evaporation and 78% global precipitation over oceans.

Thus, quantifying temperature and salinity can provide us basic understanding of the adaptability, habitat, food habits and growth profiles of marine species. In addition, the density of seawater varies with the variation in temperature, depth, and salinity. Variations in the density of water are significantly observed in pycnocline region (~ 200 – 500 m) (a subset of the halocline and thermocline regions) whereas, the density variation saturates at depths beyond 1000 m [146],

[178]. Conventionally, a salinity sensor is a simple 2 electrode design separated by some distance (2 mm in our first version). To increase the stability and the sensitivity, we have modified the design to interdigitated electrode pattern and hence observed more than ~625% increase in the sensitivity with stable and robust performance (Figure 5.3). We also observed ~135% increment in the sensitivity due to change of material from Au to Cu (Figure 5.4B), however, the effect has

121 been neglected due to corrosive nature of Cu which can degrade performance of salinity sensor upon exposure underwater.

Figure 5.3| Comparison of the performance of two variations of salinity (conductivity) sensor (A) version 1 and version 2 salinity sensor for improved sensitivity by modified design, and (B)Version 2 comparison for two different material choices of Au and Cu. An increase of ~625% sensitivity is observed due to design modification, while increase in sensitivity has been neglected.

122

5.5.3. Pressure (Depth) Sensor

The underwater pressure of the oceanic environment is directly related to the height of the water ( 푷 = 흆 ∙ 풉 ∙ 품 ) where P is the hydrostatic pressure, ρ is density of water, g- acceleration due to gravity, and h represents the height of water. The total pressure exerted (Ptotal) on any object underwater is the combination of partial hydrostatic pressure P and the atmospheric pressure (P0) at sea level (14 psi),

(푷풕풐풕풂풍 = 푷 + 푷ퟎ). Conventional pressure sensors have high sensitivity with caveat of low operating range, which restricts the usage in marine sensors.

Therefore, we have designed our depth sensor based on a parallel plate capacitance principle where the capacitance varies linearly with increasing depth with extremely good sensitivity and instantaneously. We chose PDMS as dielectric material for capacitive pressure monitoring underwater due to its inherent compressive nature. Upon pressure application it can change thickness and hence capacitance change is detected in response to the pressure. Normally

PDMS is prepared by mixing the curing agent to elastomer in ratio of (1:10) and is cured at 90 °C for 60 minutes. However, altering the mixing ratio and curing temperature modifies the compressibility and elasticity of PDMS. To improve the compressibility of the dielectric layer, we modified the PDMS curing agent to elastomer mixture during preparation. Increase in elastomer ratio to (1:12) from

(1:10) and curing at relatively low temperature (60 °C) resulted in increased compressibility- implying increased sensitivity. Also, the thickness of the dielectric layer plays an important role in increasing the detection range of the pressure (or depth). We have observed higher sensitivity for 1:12 composition of PDMS (Figure

123

5.4A) whereas similar increment was observed for thickness of 50 µm (Figure

5.4B). Our material choices and the optimisation has increased the operating range up to 2 km, which was only restricted due to non-availability of the tool that can simulate higher pressure > 2 km.

Figure 5.4| Dielectric material optimization for increasing sensitivity of pressure sensor (A) Effect of the variation in the mixing ratio of PDMS elastomer to curing agent, with (12:1) ratio showing higher sensitivity due to increased compressibility. (B) Effect of dielectric thickness variation on sensitivity and absolute values, optimum thickness is 50 µm.

124

Table 5.1: Comparison of the performance matrices for two versions of Marine Skin.

Parameter Version 1 Version 2 % change

Average temperature 22.66 mΩ/°C 358.8 mΩ/°C ~1500 % increase; sensitivity (14.83 folds increase) Average salinity 3.298 kΩ/PSU 19.655 kΩ/PSU ~500 % increase sensitivity Maximum depth of < 10 m ~ 2000 m 200 fold increase sustenance Weight ~ 6 g < 3 g entire gadget; 50% decrease even after < 0.5 g (sensor with dies, addition of wearable without bracelet ) bracelet

Bending cycles N/A > 1 × 104 N/A

Saline water prolonged 20 days 28 days Both versions can exposure withstand *Note: Comparison with other commercial products have already been presented in the previous work,[6] which stands true for this version as well. 5.5.4. Numerical Finite Element Analysis

Although the optimisation of dielectric thickness and composition resulted in improved sensitivity, we observed saturation in pressure detection at depths greater than 10 m. We hypothesised that this saturation is due to mechanical stress distribution on the thin metallic layer. The metal layers (Au in version 1 or

Cu in version 2) are sandwiched between very compliant materials such as PDMS and PI (Figure 5.5A), so we can postulate that metal/polymer interface might experience the maximum stresses.

A finite element methods (FEM) based model was created using COMSOL

Multiphysics to understand the mechanical stress distribution due to high pressure at the interface of sandwiched metal in polymer. We used numerical analysis technique to compare the performances of Au and Cu in the full stack of marine sensor, under same pressure conditions, i.e. forces experienced by the structure

125 under water at depth of 1000 m. The geometrical parameters taken were identical to the experiments. We used a linear elastic Solid Mechanics module of a commercial FEM program COMSOL™ to map the structural response, i.e. mechanical stress distribution. Since no permanent bending or deformation was observed in our device, it is reasonable to use the linear elastic mechanics in the modeling to simplify the calculations. Furthermore, it is well established that for linear elastic mechanics, the stresses are linearly proportional to the strains which are expected to exceed the yield strength of the materials under large loading when plasticity is not considered in the model. This holds true for our model where we have assumed it to be elastic model based on experimental observation of no plastic deformations in the devices. The bottom surface of the PDMS was kept as a fixed boundary while symmetric conditions on the sides of the stack were applied.

To exert the equivalent forces of water at the corresponding depth, sweep function of boundary load (equivalent pressure) on top surface was implemented. Elastic moduli of Au, Cu, PDMS and Pi were taken as 70 GPa, 120 GPa, 2.9 MPa, and

8.5 GPa, respectively, whereas the Poisson’s ratio as 0.44, 0.34, 0.49, and 0.40, respectively. To make sure that the solution is converged, fine mesh was considered.

Figure 5.5B illustrates the linear increase in the stresses at the metal interface for both thin Au/Ti (150 nm/10 nm) and thick Cu (5 µm) metal. It is evident that stresses are higher for Au whereas corresponding values for Cu are lower. As expected, the thick layer of Cu has lower stresses (σmax = 1.5 GPa) compared to

Au (σmax = 5.8 GPa). The stress contours for the small region with Au is shown in

126

Figure 5.5 C. The thickness of Au/Ti in version 1 deposited on the PI substrate could not sustain the stresses exerted at the metal/PI interface, and hence creates discontinuity in the film to give a saturated response with external pressure. As the elastic modulus of Au is slightly lower than that of Cu, Au is expected to induce lower stresses when using the same thickness for both materials (i.e. Cu and Au).

However, it is clear from the figure that the difference of induced stresses for both materials is insignificant, which does provide a sound reason to adopt the inexpensive Cu material for the device. It is evident that evolved stresses depend on the elastic modulus and feature size, thus the thicker metal would result in lower stresses. It is further proved by simulations that a relatively thick metal (~5 µm compared to 150 nm) significantly reduces the stresses. Figure 5.5 D shows that the stress has reduced significantly, which reveals the benefit of using relatively thicker layer of Cu. And when a thick metal (sub 10 µm) is to be selected, it only makes sense to elect a material that is significantly cheaper and can be readily grown/deposited using a simple technique. Thus, we chose Cu over expensive Au, which can be easily deposited using ECD technique. Conducted experimental results of sustained performance at higher pressure validate the FEM results.

127

Figure 5.5| FEM simulation studies to understand the stress distribution at the metal/polymer interface. (A) Cross section profile of the sensory platform showing different layers (not to scale). Yellow layer represents metal. (B) von Mises stress experienced by the metallic interface at different depths under water, Cu experiencing significantly lower experiences than Au/Ti. (C) and (D) stress distribution contours at the sandwiched interface of polymers and metals.

5.6. Marine Skin Fabrication Process Scheme

We start with two Si (100) wafers and spin 10 µm polyimide (PI 2611 from

HD Microsystems) at the speed of 2000 rpm for 30s. PI needs multiple curing steps of soft-baking, intermediate and final curing at different temperatures. Soft-baking is performed for 90s at 90 °C, intermediate baking at 150 °C for 90s and final curing at 300 °C for 30 minutes with the temperature gradually ramped from 150 °C to

300 °C at the rate of 240 °C /hr. The step of PI spinning and later releasing the PI

128 is depicted in Figure 5.6D. In parallel, we use another wafer as the carrier for the encapsulation from bottom side using PDMS. We take a 300 nm silicon oxide deposited on Si (100) wafer followed by sputtering a thin layer of Ti/Au (10 nm/100 nm). This gold film is used due to its low bonding energy with PDMS and hence it eases the process of peeling the entire sensory platform when final encapsulation is completed. 100 µm of PDMS is spun in at 500 rpm speed for 60s followed by curing at relatively low temperature of 70 °C for 45 minutes. PDMS for encapsulation layers (bottom and top) is prepared by mixing curing agent (Sylgard

184TM) to elastomer in the ratio of 1:10.

We perform O2 plasma treatment of PI on wafer 1 followed by deposition of

100 nm Cu using sputtering. O2 treatment of PI improves adhesion of the deposited metal. Cu is used as a seed layer for deposition of a thick (10 µm) Cu by using

ECD technique. The growth of ECD Cu is restricted to only the desired pressure sensor pattern by using lithography. The growth of the first metal layer ECD Cu is followed by PR removal and removal of Cu seed layer using plasma etching.

Temperature sensor was fabricated in the next step by a lift-off process and sputtering 100 nm of Pt. The plasma etching of PI follows Cu seed removal to pattern the PI for introducing the stretchability by design geometry. Compressive soft dielectric material PDMS (50 µm at 700 rpm for 60s) is then spun on the patterned bottom metal layer. After dielectric PDMS is cured, we transfer second layer of PI on this layer and repeat the steps for top layer ECD Cu.

129

Once the seed layer is removed and PI is etched for top metallic layer, the entire stack is transferred onto wafer 2, which has 100 µm PDMS cured on it.

Transfer is followed by integration of SoC and battery using a modular LEGO electronics approach with the help of pick-and-place robotic tool customised to handle the flexible substrates and ICs [120]. The placement accuracy is as high as

±3 µm with ~7000 UPM (units per module) production rate, which shows the manufacturability of the process. Finally, the entire integrated systems is encapsulated for waterproofing using 100 µm PDMS (1:10 ratio of curing agent: elastomer). Salinity sensor pattern is on the top metal seed layer and it must be directly exposed to the aqueous solution, which was achieved by laser cutting the

PDMS around the active area of the salinity sensor, thereby not compromising any other packaging.

130

Figure 5.6| 3D schematic process integration for the waterproof and featherlight Marine Skin platform, (A-C) represents version 1 (modified with bare die ICs), version 2 and the extreme bending of the tag respectively. (D) Steps involved in process, Wafer 1 for the bottom PDMS encapsulation spun on Ti/Au for ease or removal shown in stage 1. Stage 2 shows polyimide (PI 2611) spun and peeled for transfer at later stages. Steps 3-8 represents fabrication of sensors layers and patterning using different processes. Step 9 present the die integration of microcontroller and battery dies for stand-alone system using pick-and-place tool. Step 10-13 illustrates transfer for bottom and top waterproof encapsulation of entire system with final release from the wafer. Bottom inset shows the final released lightweight, extremely compliant Marine Skin platform.

131

5.7. Overall Sensory System Performance

This Marine Skin has advanced from the previous version 1 in terms of performance in harsh marine environmental conditions. Here we report the performance in harsh conditions with the custom set-up for replicating the real life harsh environment.

5.7.1. Temperature and Salinity (Conductivity)

Since the temperature of seawater drops from 24 °C at the surface to 2 °C for ~4000 m depth, it is important to capture the small variations throughout, which means a sensor with high sensitivity. From Figure 5.7 B, reliable linearity of the temperature sensor is evident by the correlation of R2 = 0.99872. Moreover, the calculated sensitivity of version 2 (Sv2 = 358.8 mΩ/°C) is ~15 times higher than the version 1 (Sv1 = 22.66 mΩ/°C). In addition, version 1 had large variation in performance when immersed in water compared to in air performance, more pronounced in the temperature range of interest (0-21 °C, the range of interest for sea environment). In version 2, we not only observe sustained linearity (constant sensitivity) over the entire regime from ~2 °C to 60 °C, but also the performance matches with the existing commercial solution for the temperature sensor (Figure

5.2 B). Thus, making our version 2 more reliable and feasible alternative for direct integration with the SoC device.

Electrical conductivity of the aqueous solution is the measure of salinity.

Version 1 of the salinity sensor design was a simple two electrodes design (2 mm apart) which conducts an electrical current when an ionic solution bridges the

132 electrodes. Version 1 salinity sensor demonstrated reasonable performance with average sensitivity of 3.298 kΩ/PSU. To increase the reliability and sensitivity further, we adopted an interdigitated electrode pattern. This modified design resulted in tremendous increased average sensitivity to a value of 19.655 kΩ/PSU

(~500% increase). We have observed a change of ± 3% in the baseline values when submerged in seawater for extended periods (28 days), however, the sensitivity of the device has not changed significantly, as shown by the hysteresis plot in Figure 5.7 C.

133

Figure 5.7| Performance improvement of the wearable multisensory Marine Skin gadget with reliable interlocking attachment mechanism shown in (A) Improved sensing capabilities from version 1 for sustaining higher pressure and improved (B) temperature sensitivity, with great linearity and (C) salinity sensor with increased sensitivity and reliable hysteresis for prolonged saline water exposure. (D) Depth measurements for higher depths (high pressure) 1.5 km, increments in the steps of 30 m each, representing the linear fit throughout the entire range. 5.7.2. Temperature and Salinity (Conductivity)

For depth measurements at high-pressure environment, we used a hydraulic pressure simulation tool available in the lab. The tool has the capabilities to control the applied pressure in the chamber filled with water up to a maximum pressure of 3000 psi, which is equivalent to a depth of 2000 m. However, the recommended maximum value for applied pressure was 2300 psi for the equipment, restricting our measurements to a maximum pressure equivalent to depth of 1500 m (Figure 5.8). The experimental setup is shown in Figure 5.8 B in which the sensor is immersed in a closed metallic vessel (~70 cm tall with an internal diameter of 15 cm) with thermal insulation from outside, connected to a digital and manual pressure control system. We started with digitally controlling the pressure applied through a small hydraulic hand-pump, however, observed a lot of noise in the signal recording. This noise was figured to be originating from the electrical interference of the connections of digital control system to the steel vessel. We switched to manual control mode, and applied pressure in an incremental way up to 1500 m with a step size of 30 m (~43 psi). Real-time variation in capacitance of pressure/depth sensor with respect to the applied changing pressure has been plotted in Figure 5.8 A. It can be seen that sensor exhibits a liner relationship to the change in depth of the water with extremely fast response time. Incremental increase in the capacitance with the corresponding

134 step increase in applied pressure appears to be constant throughout the entire range. From this response, it can be conferred that sensor performance neither degrades nor reaches a saturation in the value, and thereby can withstand the pressure of water higher than depths of 2 km. One can observer oscillations in the capacitance change from Figure 5.8, with incremental pressure. These oscillations are arising from the decrease in the pressure during manual application of pressure, and hence the hydraulic pressure drops initially then ramps up to the desired values. These oscillations to the variations in the pressure experienced during the step increment confirms high sensitivity and the resolution of the depth sensor.

In version 1, the pressure sensor has shown very stable response to changes in the water pressure due to increasing depths (shown by 5 cm steps), however, the response saturated when the height of water goes > 10 m. Although, many marine species swim in depths < 600 m, there are also many deep sea species that inhabit depths around 3000 m (Vampire squid) and 4500 m (Pacific

Viper fish). Thus, we have optimised the material for sustaining the performance up to 2000 m. Figure 5.7 D illustrates the linear response to linear increase in depth in steps of 30 m each. We have not observed any saturation in the pressure sensor response even up to 2000 m, which is the ultimate limit of the compression tank.

Thus, we infer that version 2 with optimised design and material shows sustained performance in harsh environments, with all the enhanced performance metrics compared in Table 5.1. Such high-pressure sustainability and capability has never been reported previously.

135

Figure 5.8| Harsh marine environment real-time pressure recording using Marine Skin platform, (A) demonstrating linear increment with the increased pressure, high sensitivity, resolution and fast response time by observed oscillations due to manual handling. (B) Experimental setup for the high-pressure simulation and testing in the central labs. 5.8. Reliability, Ruggedness, and Mechanical Integrity

For the reliability of the Marine Skin, it is important to study the effect of the harsh environmental parameters that it may experience. The two important parameters are high salinity exposure for extended periods (multiple weeks) and no degradation in the performance due to physical deformations.

136

5.8.1. Repeated Cyclic Bending Tests

The reliability of encapsulation material can be established if the Marine

Skin platform shows no degradation in performance under rugged and harsh environmental testing conditions. It must not be effected by any external factors.

Since we have no control over the orientation of the animal movement or their rotation, it is important to have a device that has no dependency on the orientation of the sensor (or the orientation/rotation of the animal). We carried out depth measurements in two different orientations (horizontal and vertical) to correlate if there is any dependency. The rationale for choosing pressure sensor lies in its working principle of capacitive sensing, as the capacitive sensor can experience different pressures in different orientations at the same height inside water. Figures

5.9 A and B plot the hysteresis curves for the depth measurements in horizontal and vertical orientation of the sensor. There was no significant change observed due to orientation of the sensor, in addition, the hysteresis curves represents the consistency of the results.

The benchmark for testing the reliability and robustness of any flexible device is the sustained performance over multiple physical deformations (bending cycles). Here, we establish a record benchmark for the robustness and ruggedness of the soft-polymeric packaging and sensor design by subjecting sensors under extreme number of bending cycles. We subject the Marine Skin version 2 to cyclic bending testing for 10 thousand cycles, with each cycle bending the sensor with radius of 1 mm and stretching (Figure 5.9 F). The device is characterised at the intervals of 0, 100, 500, 1000, 2500, and 10000 cycles. From

137

Figure 5.9 C, we observe that median temperature sensitivity stands at 330.45 mΩ/°C with minute variations in absolute sensitivities at specific cycles. However, the change in sensitivity recorded is < 4% from the value at zero bending cycles.

Similarly, the robustness of the pressure sensor can be observed from Figure 5.9

D where the sensitivity has not changed significantly due to constant cyclic physical deformations. The observed variations are mainly due to manual controlling of the immersion in water, whereas the maximum deviation observed in absolute value is < 2.6% from the initial baseline at zero cycles.

First, the ruggedness of the depth sensors and integrity of packaging were tested by subjecting the fabricated Marine Skin to a large number of bending cycles with the bending radius of 1 mm. Depth measurements in the lab environment are carried out after 100, 500, 1000, 2500, and 104 bending cycles. Real time variations in the capacitance while submerged in water in steps of 10 cm are plotted in Figure 5.10 where the change in each step is consistent with only variation due to the manual handling error in maintaining constant step height increment. The sharp increase in the capacitance occur as soon as the sensor is immersed in water from air. To confirm the reliability for the ruggedness over different cycles, hysteresis tests are performed where measurements are recorded at specific depths during increasing depths (submerging) and decreasing depths

(rising up) of water. This also illustrates excellent hysteresis characteristics of the sensors to variation in depth, nevertheless, the variations in the hysteresis can also be attributed to variations occurring due to manual handling.

138

Figure 5.9| Robustness and reliability of the devices for prolonged exposure to highly saline seawater and the cyclic bending testing. Reliable depth sensing performance and hysteresis (submerging in water and raising from water) independent of the sensor orientation (A) horizontal or (B) vertical orientation with respect to ground plane. Harsh bending cycle testing for (C) temperature sensors and (D) depth sensor, by subjecting to bending radius of 1 mm for a maximum 104 cycles. (E) Effect of prolonged exposure to highly saline Red Seawater on depth sensing for 1, 3, 7, 15, and 28 days. (F) Bending cycles test setup for cyclic testing, Version 2 Skin in stretched and then bent for a bending radius of 1 mm.

139

Figure 5.10| Robustness of the pressure sensor characterized for the pressure sensor when subjected to extreme bending cycle test of (1 mm bending radii) (A) - (F) with 0, 100, 500, 1000, 2500, and 10 thousand cycles respectively. Discrete measurements are plotted at step heights of ~10 cm for observing the hysteresis during submerging and rising from the water in the acrylic tank.

140

5.8.2. Prolonged Exposure to Saline Environment

Similarly, the effect of prolonged exposure to the Red Sea saline water (41

PSU) is studied for establishing robustness and reliability of the packaging. The packaged sensors were immersed in the Red Sea water for extended periods of time while their performance is evaluated after 1, 3, 7, 15, and 28 days. Once again, there is no significant noticeable variation in performance observed for pressure sensors, as illustrated in Figure 5.9 E. Moreover, the maximum deviation in absolute value of recorded pressure is < 2.3% from the average value at maximum depth. It must be noted that the cyclic testing was carried out after exposure testing, the samples were already exposed to Red Sea water for ~ 1 month, and hence it establishes more robust packaging by showing reliability. The real-time change in the capacitance with increasing depth of the water is acquired for these different scenarios, which show no degradation in terms of saturation or sensitivity. Thus, we can conclude that the integrity of packaging, and the reliability of the sensors makes the Marine Skin platform an extremely robust, flexible, and highly lightweight solution for marine environment monitoring.

For deployment on animals and monitoring the required parameters, the utmost important aspect is the sustained performance at high depths, which no other research has reported. In Figure 5.7D, we can clearly observe the linear increment in the capacitance as the pressure increases and retains similar step change over the entire range from 500 m to 1500 m. The real-time oscillations due to fluctuations in the pressure are plotted in Figure 5.8, suggesting high resolution and excellent sensitivity. This linear variation in the depth measurement along with

141 no saturation, even at the depth of 2000 m (limited due the experimental emulator equipment’s range), suggest robust and sustained performance at extreme environments. We report the first instance of sustained performance of devices at such high depth, which has never been reported by any other groups previously.

To summarise, the modified design and material optimisation of Marine Skin has resulted in tremendous increase in sensitivity, reliability, and robustness of the packaging. In addition, version 2 demonstrates extremely robust and rugged performance reported for the first time, in an extremely harsh environment of 104 bending cycles (bending radius 1 mm), high pressure (3000 psi, ~2 km), and prolonged immersion in Red Sea water (1 month at 41 PSU).

5.8.3. Non-invasive Attachment Mechanism

Marine species have huge variations in their skin types. Many fish species have skeletal elements (scales) that cover their skin. Scales are classified based on the composition and structure as Ganoid scales, Placoid scales (or denticles), Cycloid scales, and Ctenoid scales [181]. Careful consideration must be given to the skin type of the species under study for devising the best attachment method. Sharks are one of the most widely studied and tagged marine species for several reasons.

For instance, Silky shark (Carcharhinus falciformis) is one of the most abundant cosmopolitan shark species of the pelagic zone. It feeds on schooling fishes, including tuna, and as a result is a major bycatch item of the tuna fishery. Typically, shark skin is very rough with huge denticles scattered over the skin each with the pulp cavity surrounded at the edge with odontoblasts. In addition, the animals often

142 used for tagging of bio-loggers include hard shell bodied crustaceans and sea turtles [168], [181], [182]. Most of the work on tagging constitutes mainly invasive attachment methods using incisions and anchors. Our initial tagging experiment with version 1 relied on a host body (like a cylindrical CAN platform) for non-hard shell animals, which are attached to the dorsal fin of the sharks using clamps (3D printed and/or metallic) as shown in Figure 5.11 A. These clamps not only exert pressure on the fins of sharks but also attract other individuals to attack this foreign element. Moreover, this methodology does not comply with the smaller fishes and the species that do not have rigid fins.

5.8.3.1. Attachment on Large Species

The tagging of the marine species has always been an invasive method involving incisions through skin, tissues, or usage of metallic and plastic anchors inserted in the skin. Invasive method of attachment always leads to the injury of the species, which not only introduces great discomfort to the species tagged but also can change the natural movements and behavior. Thus, the requirement of

2% bodyweight, flexibility and non-invasive nature of the devices are the goal. Our focus was on making a flexible standalone system that can adhere to all these norms in addition to having a completely non-invasive tagging mechanism. In the past, Marine Skin version 1 was tested for its flexibility and non-invasive method attachment on Tiger Shark, Wobbegong Shark, and Stingray (Figure 5.11). We tried mounting the sensors on a cylindrical CAN host attached to a steel clamp that is used to attach to the dorsal fin of the shark. This method can only be applicable to large species with the dorsal fin acting as an anchor for the sensing system. For

143 other species with comparatively smoother and mucous skin or smaller sized animals, we have to apply other techniques. Superglue can be used on species with hard shell (turtles, crab, etc.) whereas as dental/surgical glue was tested on

Wobbegong and Stingray. The hydrodynamic forces due to water streams detaches the sensors from the body of animal. Also this glue is dissolvable in water within 48 hours, impossible for long term deployment.

We have tagged a wobbegong shark and stingray with a very different skin type using dental glue (Figures 5.11C and 5.11D). The tag stayed perfectly conformal to the body of the animal, however, surgical glue dissolves in water within 48 hours and prolonged deployment was not possible. We observed that, any tagging platform, which is not conformal to the skin will detach in short time due to the hydrodynamic forces acting on the tags opposite to the swimming direction. It has been noticed from veterinary experts’ that permanent glue and/or super adhesives present the danger of skin irritation and injuries to the soft skin of the species. For small and bony fishes, the skin is very soft and has mucous membranes. Furthermore, there are species such as dolphins that are known to replace their skin cells over short time periods (< 1 day). Hence, alternative strategies for attachment needs exploration.

5.8.3.2. Attachment on Large Species

On the line with wearable fitness trackers for human being, increasing in the consumer market due to its conventional smart watch like design, we present a unique, pragmatic, and universal approach for sensory platform tagging. We

144 propose a soft wearable bracelet in which the sensory tags can be embedded

(Figure 5.11 F right) within the soft polymer (PDMS) and can be wrapped around species like a wearable gadget. Flexibility, direct compatibility with the schematic flow, biocompatibility, softness, elasticity, and other properties of PDMS played an important role in material choice. We made 3D printed molds for replicating the wearable modules, followed by pouring PDMS to cure at 60 °C for an hour. This cured wearable jacket design embedded with sensory platform can be easily peeled from the mold (Figure 5.12). In initial design, we used the locking mechanism of soft-pins and the holes made from the same PDMS material (Figure

5.12 C). However, the soft pins design was not strong enough to withstand water stream pressure, in addition, the adhesive strength was not high enough to hold the jacket on the skin. The interlocking mechanism of the jacket consists of 3D printed mushroom pins that locks into holes on the other side of the bracelet

(Figure 5.11 G right). The presence of multiple pins allows adaptability of the same gadget for different sized animals. The strength of the bracelet was tested in the lab by wrapping it around closed fingers and stretching to see if the interlocking parts break (Figure 5.12 E) and the devise did not break even after exerting significant force. We illustrated the attachment mechanism by tagging the

Barramundi (Lates calcarifer) and Seabream (Figure 5.11 F and G). We have observed that this wearable Marine Skin is easy to attach even on small fishes. It presented no hindrances to their natural movements due to softness, conformability, stretchability, and minimal weight (bracelet with the embedded system weighs < 3g). Hence, the feather-light and breathable wearable bracelet

145 pattern is a pragmatic mechanism for the Marine Skin tagging system that is conformal, comfortable for the individual species tagged, and also not noticeable to other species due to its inherent transparency.

Figure 5.11 Non-invasive attachment mechanism for tagging Marine Skin. (A) Shark tagged using a steel clamps attached to a light cylindrical can (B) hosting the Marine Skin platform. Direct tagging version 1 using surgical glue on (C) wobbegong shark, (D) Stingray popping out due to water stream and the rigid component of battery. (E) Scaled version 2 without rigid components adhere strongly on gold fish attached using surgical glue. (F) Direct tagging of Marine Skin version 2 on the barramundi fish (left) using wearable soft bracelet with sensors embedded (right). (H) Direct comfortable tagging on the Seabream (left) and ability to adopt to different sizes of species due to the interlocking mechanism of the soft bracelet (right).

146

Figure 5.12 | Wearable bracelet designs and molds (A) 3D printed mold designs for different wearable bracelet. (B) Easy peeling-off of the flexible and stretchable bracelet from 3D mold, (C) two designs having Marine Skin embedded within the bracelet with soft-mushroom pins to hold. Serpentine structures (bottom) can provide more breathability for animal and stretchability to the bracelet. (D) Modified bracelet design with 3D printed mushroom pins (spherical and trapezoidal shape) to improve inter-locking mechanism. (E) Increased inter-locking strength due to incorporation of 3D printed mushroom pins has little effect on the flexibility.

147

5.9. Conclusion

The ocean ecosystem is affected by multiple human impacts like pollution, over-exploitation, warming and acidification. To facilitate policy implementations, trade-offs and devise mitigation strategies at global scales, it is very important to quantify the distribution of human impacts on the marine environment and hence marine research should include ocean wide monitoring of species and the environment. Monitoring should also include the deep ocean which covers a majority of the earth’s surface, but remains poorly studied. The evolution of the sensors and tracking tools in conjunction with the advances in the technology is making pivotal contributions in understanding the marine ecosystem and animal behaviour and it will continue to grow further. The complex animal geometries, different skin types, size, and varying environmental conditions necessitate a flexible wearable and stretchable gadget that can monitor different environmental parameters without any discomfort to the animal. Recently, we have demonstrated a flexible, and stretchable waterproof Marine Skin platform for monitoring the environment, however, for the harsh environment and the varying conditions, a rugged and robust device is required. Therefore, we demonstrate an extremely rugged and robust version of the Marine Skin tagging platform that withstood 104 severe bending cycles (1 mm bending radius), prolonged exposure to the highly saline (41 PSU) Red Sea water, and the extremely high pressure of ocean depths

(~ 2 km deep). To facilitate ocean monitoring, a unique attachment strategy by means of a wearable stretchable jacket (gadget) architecture is implemented for a

148 non-invasive and easy attachment method without any harm and discomfort to the species irrespective of their skin type. This Marine Skin gadget outperforms any other entities in the domain in terms of flexibility, stretchability, non-invasiveness, comfort, featherlight (weight <0.5g for systems, and <3 g with the entire wearable gadget), with proven ruggedness and sustained performance under harsh environments.

149

6. Chapter 6: Summary and Future Outlook

For the last sixty years, miniaturization of electronics fabricated on prominent active electronic materials like silicon, germanium, III-V materials and gallium nitride has enabled modernization of today’s world especially bringing convenience, safety and efficiency to our daily life. However, we are in continuous pursuit to find alternative materials and process technologies to lower the cost of manufacturing and to increase functionalities of electronics. A radical physical change of rigid electronic components and systems to a mechanically compliant flexible and stretchable version will jettison the interfacial mismatch with nearly all natural lives, enhance the functionalities of existing applications and will usher in new applications, which are not possible today. In this section, we will briefly focus on key areas of this emerging area of electronics [160]–[162].

6.1. Materials

In 1962, Henry Letheby reported the first organic conductive material. Due to its natural compliance, researchers focused on using available soft polymeric materials as vehicle for flexible organic electronics. In the year 2000, the Nobel

Prize was awarded for discovery of a large set of conductive polymers followed by subsequent promising progress in the area of large-area low-cost thin film transistors and organic light emitting diodes. Thousands of scientific papers have been published and hundreds of start-ups have been established since then. As of today, therefore, organic and molecular electronics are considered as the primary hope and home for flexible electronics except a few unresolved significant caveats: low charge transport ability and thermal instability [105], [183], [184].

150

In the late nineties, discovery of fullerene followed by subsequent progress in carbon nanotube and broad class of nanowires (including semiconducting) excited the research community toward one dimensional materials based flexible electronics [185]–[187]. Especially their superior material properties (compared to those of organic materials) and their natural compliance (due to their low dimensionality) lead to a variety of flexible device demonstrations including artificial skin type multi-sensory platform [188]–[190]. Relevant process technology including roll-to-roll printing served as catalyst for this progress. Remaining large- area assembly/integration issues inhibited its further progress.

In the middle of the 2000s, scientific community focused on graphene and other two-dimensional (dichalcogenide) materials as they showed the promise of large area coverage (like thin films) and their pristine atomic crystal structure with superior charge transport ability [186], [191]. Specially for energy storage, transparent conductive thin film and sensor technology – there have been thousands of scientific publications published. However, with non-uniformity in their growth, suitable interfacing for higher conduction and often dielectric and conductive interface formations remain as unsettled challenges. Therefore, even with reduced momentum, new two-dimensional atomically thin materials are being explored every day to overcome these remaining challenges.

From the very beginning, classical crystalline materials have been ignored for flexible electronics due to their already dominant presence, inherent rigidity and brittleness. However, advances in amorphous, polycrystalline thin films (especially

151 oxide based) opened up an alternate door to use nearly identical material properties (those of traditional materials) for flexible electronics. Distinctively, thin film transistors lead the way. Yet, from the beginning of the 2000s, there has been a surge of flexible silicon, gallium nitride and III-V electronics. Their advantage to be fabricated using existing complementary metal oxide semiconductor (CMOS) technology makes them attractive. They are fast, scalable and reliable too.

Another commonly used material, paper has been explored as a potential host substrate for ultra-low cost flexible electronics. Functionalizing them with chemicals and other low-dimensional materials and processes, a variety of applications have been demonstrated [192]–[194]. Nonetheless, their reliability has been questioned often. Recent demonstrations of recyclable papers as active electronic materials and their biocompatibility have now brought them back as alternative flexible electronic materials [65], [195].

6.2. Design Strategies

While innovation in material to achieve flexibility and semiconducting properties requires innovative design of composite materials, stretchable electronics nearly depend on this singular requirement of appropriate design strategy. Stretchable electronics are subjected to maximum physical deformation

(linear and non-linear both). Again, the major design strategy has been to use macroscopic stretchable polymeric material as host substrates followed by deposition/transfer of organic, 1D and 2D material on them. While 0D and 1D materials being nanoscopic easily conform and comply with stretching phenomena

152 in the host substrates[160], [162], 2D materials can face rupture based on their mechanical properties. Some popular stretchable organic and 1D materials are silver nanowire (AgNW). Often graphene has been dubbed as a potential active stretchable electronic material.

From the middle of the 2000s, several innovative ways of transforming the conventional rigid electronics into stretchable electronics have been successfully developed and demonstrated: pre-straining and adoption of fractal design [196],

[197]. In the first case, pre-strained polymeric material acts as host substrate for conventional crystalline and amorphous thin films (including silicon and gallium nitride) and then the pre-strained material is released to relieve the stress. This transforms the continuous thin film into a seemingly deformed (in reality, regular wavy shaped) but one form of stretchable electronics with limited stretchability (up to 10%). In the later approach, islands of active materials are interconnected/bridged through adoption of various fractal designs (serpentine, spiral, etc.) [198]–[201]. Such creative design adoption has resulted into stretchability up to 1020%. Recently, some studies have been exploring out-of- plane, staged/periodic and reversible stretchable platform for stretchable electronics [198]–[201]. In all cases, design strategies must conform the chosen material to suitable properties and deformation and endurance mechanics.

6.3. Integration Strategies

From the very beginning nearly all efforts have been focused to demonstrate discrete devices. Chipfilm has made substantial efforts for manufacturable flexible

153

CMOS system. The persistent challenges include but are not limited to expense and reliability [202]. Several companies including Xerox PARC have been pursuing low-cost printing technology development focusing on organic materials. However, none of those has been proven as efficient as CMOS technology except the cost of the later is obviously higher due to its precision. From system level integration perspective, Rogers et al. have shown multi-sensory platform for various applications including brain-machine-interfacing [117], [203]. Someya et al. use active matrix display type architecture for their organic material focused sensory platform [105]. Bao et al., Javey et al. and Arias et al. have also demonstrated a variety of multi-sensory platforms [189], [204], [205]. Hussain et al. has led a

CMOS based manufacturing strategy which can produce a fully flexible packaged electronic system. They emphasize on non-planar coin like 3D architecture where sensors, actuators, energy harvesters and antenna remain in the outer sides of both planes and other accessorial electronics remain in the middle (like a sandwich) and they are also physically flexible [128], [206].

6.4. Applications

Displays: Practical organic light emitting diode (OLED) device was first demonstrated by Eastman Kodak in 1987. Since then both academia and industry have pushed this technology and billions of dollars have been invested. It is expected that Samsung and LG Electronics will launch the first flexible display in the 2020s.

154

Photovoltaic: Since flexibility is achieved by volumetric reduction, in the case of inorganic material based solar cells, efficiency is compromised. Recently,

Hussain et al. introduced a corrugation structure based flexible crystalline solar cell with record efficiency of 19% at a bending radius of 140 µm [119]. Obviously, using

III-V solar cells will be a rational choice for higher efficiency, except they will also increase costs. Although organic materials based solar cells could have been naturally flexible, their fundamental low efficiency and unreliable operation (until today) impedes their wide scale adaption. Sun power suggests providing commercially available flexible solar cells.

Wearable: Major electronics giants like Apple and Samsung have already acquired a substantial market with their Apple Watch and Samsung Gear. Fitbit became a major player by being the first to introduce a health tracker. Since then hundreds of start-ups have launched nearly the same kind of products with a little variation in the functionalities and major design differences. Their approach is to use miniaturized ICs but at the end they are not completely flexible. Also, they are expensive. Thus, overall market for wearable is not as promising as it was before.

Implantable: Introduction of implantable electronics would be a game changer like what has happened with pacemakers. However, in most cases, rigorous requirement for clinical trial data to prove their long-term reliability and safety looms as a critical concern. This serves as a major basis for public perception and doubt about implantable electronics. Current academic research

155 efforts are restricted to brain machine interface and nanomedicine based targeted drug delivery (which by the way is not an implantable device) [31], [81].

Add-on: This new kind of electronics is introduce by Hussain et al. where they envision using Do-It-Yourself integration strategy to assemble low-cost add- on electronics using recyclable materials [75], [76], [206], [207]. Such electronics are expected to be attached to existing objects to transform them into “smart”

(data and sensing oriented) objects.

Soft robotics: Whitesides et al. have introduced the concept of soft-robotics and we can see major innovations and their practical usage through robotic arms and other organs [208]. An effective integration of both the interactive material based soft robotics and flexible and stretchable electronics can add breakthrough functionalities in soft robotics.

Textile: While we have been using textiles for thousands of years, its basics have remained the same for centuries. Therefore, we have observed limited activities to “smartize” them and the course nature of fabric introduces an interfacial mismatch for traditional electronics to be integrated. Therefore, substantial research is needed to develop reliable high volume manufacturing strategy for low-cost smart textiles.

Communication: Microwave (and millimeter wave) flexible and stretchable electronics are an important sub-field of flexible and stretchable electronics that has demonstrated its impact over the last decade [190], [209]–[212].

Conventional microwave electronics have been widely implemented in mobile

156 devices, wireless communications, radar sensors, radio monitoring and surveillance, etc. However, the present form of microwave electronics is chip- based, which are rigid, brittle, system-bulky, and costly, particularly if implemented in a large area, thereby limiting their applications to be further expanded. For example, a high density array of a millimeter-wave phased-array antenna based on rigid-chip wiring is heavy, costly, and reliability-risky. The sub- field of microwave flexible and stretchable electronics was created to specifically address the need of high frequency electronics that can satisfy the requirements of non-conventional form factors (e.g., non-rigid and large area) and overcome the various shortcomings of the present microwave electronics.

In comparison to the conventional circuit board-based microwave integrated circuits (MIC) and extensively implemented monolithically microwave integrated circuits (MMIC) over the last 2-3 decades, the mechanical flexibility and stretchability features of the microwave flexible electronics, which were demonstrated in the recent decade,[213]–[216] opens numerous new microwave application opportunities that cannot be fulfilled by MIC/MMIC. The unique mechanical features allow the flexible and stretchable microwave electronics to be implemented on uneven or rugged surfaces, wood substrates and skin, dramatically expanding the application boundaries of the traditional MIC/MMIC

[217], [218]. While the microwave performances have been well-maintained and the functionalities have been greatly enhanced in microwave flexible electronics, the non-traditional fabrication methods associated with the new form factors have

157 also led to a great cost reduction in comparison to the fabrication of traditional

MIC/MMIC.

As a sub-field of the broadly defined flexible electronics, the unique feature of the microwave flexible/stretchable electronics that distinguishes it from the rest of the flexible electronics fields, is the high frequency (>1 GHz) characteristics. At such high frequencies, new materials, new design methodology, new fabrication techniques and new characterization tools are required [210], [211], [219]. Within microwave flexible electronics are active devices, passive components and substrates [210], [211], [220]. To satisfy the requirements of high frequency operation with mechanical flexibility and stretchability, the substrates suitable for microwave flexible and stretchable electronics need to exhibit low microwave energy loss (tan δ). As high-frequency operation inevitably causes excessive heat generation, the substrates or any fixtures that are used to carry the active devices on the substrates will also need to have good thermal conductivity. Satisfying the need for operation at different frequency ranges, both lumped and distributed flexible and stretchable passive components have been demonstrated and implemented in flexible and stretchable microwave and millimeter wave circuits.

6.5. Challenges

Several critical challenges remain for wide scale adoption of flexible and stretchable electronics. The first one is need for a go-to application. Obviously major display companies are suggesting flexible display technology. While the display itself is flexible, associated electronics are still non-flexible. Therefore, it

158 falls under hybrid architecture. Also, rollability will allow displays to be easily portable, but its pervasive use is questionable. Second area of potential usage is photovoltaic and battery technology. Both can be benefited by the volumetric reduction based weight loss, flexibility and low-cost fabrication. However, a critical challenge remains for both in context of appropriate material selection. Other concerns are long-term endurance and safety. Finally, lack of coherent manufacturable technology serves as a severe challenge specifically when the overall activity is predominantly lead by the academic community. Major integrated device manufacturers are focused on already established businesses and strengthening that further. Hence, a major technological gap exists.

6.6. Future Outlook

Empowering mankind with electronics can enable a better future for us. That requires electronics which are low-cost, easy to implement and use. From these perspectives, flexible and stretchable electronics can be promising venues for expansion of electronic applications. While many such applications have been demonstrated, rarely any of them has been commercialized for widespread usage.

Therefore, as mentioned before, identification of critical go-to applications is the key. That will happen with the evolution of a valid manufacturable technology.

Some impediments exist in context of implantable electronics including social prejudice and safety related regulations. Innovation in bio-safe material and appropriate communication can be effective to overcome both.

159

6.7. Conclusion

Regardless of the countless progress in flexible and stretchable electronic components, substrates that have been made, and applications which have been demonstrated, they are just a tip of the iceberg. A combination of materials, process technologies, and manufacturable integration strategies focused on effective and impactful applications at an affordable price will expand the horizon of future electronics to empower humanity and to make this world a better place every day. In this thesis, we have presented such pragmatic heterogeneous 3D integration strategy which can readily be adapted for manufactural scale unlike other integrations schemes that involve multiple transfers of flexible polymeric substrates.

160

REFERENCES

[1] A. M. Hussain and M. M. Hussain, “CMOS-Technology-Enabled Flexible and Stretchable Electronics for Internet of Everything Applications,” Adv. Mater., 2015. [2] L. Coetzee and J. Eksteen, “The Internet of Things - promise for the future? An introduction,” IST-Africa Conf. Proceedings, 2011, pp. 1–9, 2011. [3] “The BRAIN Initiative: Surviving the Data Deluge.” [Online]. Available: http://www.kavlifoundation.org/science-spotlights/brain-initiative-surviving- data-deluge#.V_evJY997IW. [4] G. Yeric, “Moore ’ s Law at 50 : Are we planning for retirement ?,” pp. 1–8, 2015. [5] ITRS, “International Technology Roadmap for Semiconductors 2.0: Executive Report,” Int. Technol. roadmap Semicond., p. 79, 2015. [6] J. N. Burghartz, “Thin Chips on the ITRS Roadmap,” in Ultra-thin Chip Technology and Applications, New York, NY: Springer New York, 2011, pp. 13–18. [7] C. Dagdeviren et al., “Conformable amplified lead zirconate titanate sensors with enhanced piezoelectric response for cutaneous pressure monitoring.,” Nat. Commun., vol. 5, 2014. [8] J. Kim et al., “Miniaturized Battery-Free Wireless Systems for Wearable Pulse Oximetry,” Adv. Funct. Mater., vol. 27, no. 1, pp. 1–8, 2017. [9] Y. Khan, A. E. Ostfeld, C. M. Lochner, A. Pierre, and A. C. Arias, “Monitoring of Vital Signs with Flexible and Wearable Medical Devices,” Advanced Materials, vol. 28, no. 22. 2016. [10] M. Ghoneim and M. Hussain, “Review on Physically Flexible Nonvolatile Memory for Internet of Everything Electronics,” Electronics, vol. 4, no. 3, pp. 424–479, 2015. [11] S. H. Jo, T. Chang, I. Ebong, B. B. Bhadviya, P. Mazumder, and W. Lu, “Nanoscale memristor device as synapse in neuromorphic systems,” Nano Lett., vol. 10, no. 4, pp. 1297–1301, Apr. 2010. [12] D. S. Jeong et al., “Emerging memories: resistive switching mechanisms and current status,” Reports Prog. Phys., vol. 75, no. 7, p. 076502, Jul. 2012. [13] S. Kim, H. Y. Jeong, S. K. Kim, S. Y. Choi, and K. J. Lee, “Flexible memristive memory array on plastic substrates,” Nano Lett., vol. 11, no. 12, pp. 5438–5442, Dec. 2011.

161

[14] S. Porro, E. Accornero, C. F. Pirri, and C. Ricciardi, “Memristive devices based on graphene oxide,” Carbon N. Y., vol. 85, pp. 383–396, 2015. [15] S.-T. Han, Y. Zhou, and V. A. L. Roy, “Towards the development of flexible non-volatile memories.,” Adv. Mater., vol. 25, no. 38, pp. 5425–49, Oct. 2013. [16] M. Kubo et al., “Stretchable microfluidic radiofrequency antennas,” Adv. Mater., vol. 22, no. 25, pp. 2749–2752, Jul. 2010. [17] L. Song, A. C. Myers, J. J. Adams, and Y. Zhu, “Stretchable and Reversibly Deformable Radio Frequency Antennas Based on Silver Nanowires,” ACS Appl. Mater. Interfaces, vol. 6, no. 6, pp. 4248–4253, Mar. 2014. [18] A. M. Hussain, F. A. Ghaffar, S. I. Park, J. A. Rogers, A. Shamim, and M. M. Hussain, “Metal/Polymer Based Stretchable Antenna for Constant Frequency Far-Field Communication in Wearable Electronics,” Adv. Funct. Mater., vol. 25, no. 42, pp. 6565–6575, Nov. 2015. [19] D. Kim et al., “On-demand frequency tunability of fluidic antenna implemented with gallium-based liquid metal alloy,” Eur. Phys. J. Appl. Phys., vol. 78, no. 1, p. 11101, Apr. 2017. [20] H. R. Khaleel, “Design and fabrication of compact inkjet printed antennas for integration within flexible and wearable electronics,” IEEE Trans. Components, Packag. Manuf. Technol., vol. 4, no. 10, pp. 1722–1728, 2014. [21] C. Dagdeviren et al., “Recent progress in flexible and stretchable piezoelectric devices for mechanical energy harvesting, sensing and actuation,” Extrem. Mech. Lett., vol. 9, pp. 269–281, 2016. [22] M. M. Hussain, Z. (Jack) Ma, and S. F. Shaikh, “Flexible and Stretchable Electronics – Progress, Challenges, and Prospects,” Electrochem. Soc. Interface, vol. 27, no. 4, pp. 65–69, Dec. 2018. [23] D.-H. Kim et al., “Materials for multifunctional balloon catheters with capabilities in cardiac electrophysiological mapping and ablation therapy,” Nat. Mater., vol. 10, no. 4, pp. 316–323, Apr. 2011. [24] A. Gumus et al., “Expandable Polymer Enabled Wirelessly Destructible High-Performance Solid State Electronics,” Adv. Mater. Technol., vol. 2, no. 5, 2017. [25] G. Kettlgruber et al., “Intrinsically stretchable and rechargeable batteries for self-powered stretchable electronics,” J. Mater. Chem. A, vol. 1, no. 18, p. 5505, Apr. 2013. [26] R. R. Bahabry et al., “Corrugation Architecture Enabled Ultraflexible Wafer- Scale High-Efficiency Monocrystalline Silicon Solar Cell,” Adv. Energy Mater., vol. 8, no. 12, p. 1702221, Apr. 2018.

162

[27] K. Huang et al., “An approach to cost-effective, robust, large-area electronics using monolithic silicon,” in Technical Digest - International Electron Devices Meeting, IEDM, 2007, pp. 217–220. [28] J. Lee, J. Jang, and Y. K. Song, “A review on wireless powering schemes for implantable microsystems in neural engineering applications,” Biomed. Eng. Lett., vol. 6, no. 4, pp. 205–215, 2016. [29] S. F. Shaikh, M. T. Ghoneim, G. A. Torres Sevilla, J. M. Nassar, A. M. Hussain, and M. M. Hussain, “Freeform Compliant CMOS Electronic Systems for Internet of Everything Applications,” IEEE Trans. Electron Devices, vol. 64, no. 5, pp. 1894–1905, May 2017. [30] J. M. Nassar, J. P. Rojas, A. M. Hussain, and M. M. Hussain, “From stretchable to reconfigurable inorganic electronics,” Extrem. Mech. Lett., vol. 9, pp. 245–268, Dec. 2016. [31] S. M. Khan, A. Gumus, J. M. Nassar, and M. M. Hussain, “CMOS Enabled Microfluidic Systems for Healthcare Based Applications,” Adv. Mater., vol. 30, no. 16, pp. 1–26, 2018. [32] Y. Khan et al., “Flexible Hybrid Electronics: Direct Interfacing of Soft and Hard Electronics for Wearable Health Monitoring,” Adv. Funct. Mater., vol. 26, no. 47, pp. 8764–8775, 2016. [33] S. Kwon, J. Kim, S. Kang, Y. Lee, H. Baek, and K. Park, “CardioGuard: A Brassiere-Based Reliable ECG Monitoring Sensor System for Supporting Daily Smartphone Healthcare Applications,” Telemed. e-Health, vol. 20, no. 12, pp. 1093–1102, Dec. 2014. [34] M. M. K. Delano, “A Long Term Wearable Electrocardiogram ( ECG ) Measurement System,” no. 2010, 2012. [35] T. Nakamura et al., “Basic characteristics of implantable flexible pressure sensor for wireless readout using MRI,” Trans. Japanese Soc. Med. Biol. Eng., vol. 52, p. O-409-O-410, 2014. [36] S. L. Swisher et al., “Impedance sensing device enables early detection of pressure ulcers in vivo,” Nat. Commun., vol. 6, no. 1, p. 6575, Dec. 2015. [37] C. M. Lochner, Y. Khan, A. Pierre, and A. C. Arias, “All-organic optoelectronic sensor for pulse oximetry,” Nat. Commun., vol. 5, no. 1, p. 5745, Dec. 2014. [38] M. Rothmaier, B. Selm, S. Spichtig, D. Haensse, and M. Wolf, “Photonic textiles for pulse oximetry,” Opt. Express, vol. 16, no. 17, p. 12973, Aug. 2008. [39] J. D. MacKenzie and C. Ho, “Perspectives on energy storage for flexible electronic systems,” Proc. IEEE, vol. 103, no. 4, pp. 535–553, 2015.

163

[40] T. Sekitani, U. Zschieschang, H. Klauk, and T. Someya, “Flexible organic transistors and circuits with extreme bending stability,” Nat. Mater., vol. 9, no. 12, pp. 1015–1022, Dec. 2010. [41] M. Yi et al., “The mechanical bending effect and mechanism of high performance and low-voltage flexible organic thin-film transistors with a cross-linked PVP dielectric layer,” J. Mater. Chem. C, vol. 2, no. 16, pp. 2998–3004, 2014. [42] G. A. T. Sevilla et al., “Flexible and transparent silicon-on-polymer based sub-20 nm non-planar 3D FinFET for brain-architecture inspired computation,” Adv. Mater., vol. 26, no. 18, pp. 2794–2799, May 2014. [43] H. Xu et al., “Fabrication of Flexible Amorphous Indium-Gallium-Zinc-Oxide Thin-Film Transistors by a Chemical Vapor Deposition-Free Process on Polyethylene Napthalate,” ECS J. Solid State Sci. Technol., vol. 3, no. 9, pp. Q3035–Q3039, Jul. 2014. [44] Z. R. Wang et al., “Low power flexible organic thin film transistors with amorphous Ba0.7Sr0.3TiO3 gate dielectric grown by pulsed laser deposition at low temperature,” Org. Electron., vol. 13, no. 7, pp. 1223– 1228, 2012. [45] BBC, “BBC One - Life Story, Courtship, Pufferfish ‘crop circles.’” [Online]. Available: http://www.bbc.co.uk/programmes/p029nb9g. [Accessed: 05- Jun-2017]. [46] T. D. Clark, J. Tien, D. C. Duffy, K. E. Paul, and G. M. Whitesides, “Self- assembly of 10-??m-sized objects into ordered three-dimensional arrays,” J. Am. Chem. Soc., vol. 123, no. 31, pp. 7677–7682, 2001. [47] C. J. Morris, S. A. Stauth, and B. A. Parviz, “Self-assembly for microscale and nanoscale packaging: Steps toward self-packaging,” IEEE Trans. Adv. Packag., vol. 28, no. 4, pp. 600–611, 2005. [48] S. Biswas, M. Mozafari, T. Stauden, and H. O. Jacobs, “Surface tension directed fluidic self-assembly of semiconductor chips across length scales and material boundaries,” Micromachines, vol. 7, no. 4. Multidisciplinary Digital Publishing Institute, p. 54, 28-Mar-2016. [49] H. J. J. Yeh and J. S. Smith, “Fluidic self-assembly of microstructures and its application to the integration of GaAs on Si,” Proc. IEEE Micro Electro Mech. Syst., vol. 6, no. 6, pp. 279–284, 1994. [50] S. A. Stauth and B. A. Parviz, “Self-assembled single-crystal silicon circuits on plastic.,” Proc. Natl. Acad. Sci. U. S. A., vol. 103, no. 38, pp. 13922–7, Sep. 2006. [51] S. A. Stauth and B. A. Parviz, “Integration of silicon circuit components onto plastic substrates using fluidic self-assembly,” in Proceedings - 2005

164

International Conference on MEMS, NANO and Smart Systems, ICMENS 2005, 2005, pp. 147–152. [52] E. Saeedi, S. Kim, and B. A. Parviz, “Self-assembled crystalline semiconductor optoelectronics on glass and plastic,” J. Micromechanics Microengineering, vol. 18, no. 7, p. 75019, 2008. [53] D. H. Gracias, J. Tien, T. L. Breen, C. Hsu, and G. M. Whitesides, “Forming Electrical Networks in Three Dimensions by Self-Assembly,” Science (80-. )., vol. 289, no. 5482, pp. 1170–1172, 2000. [54] E. E. Kuran and M. Tichem, “Magnetic self-assembly of ultra-thin chips to polymer foils,” IEEE Trans. Autom. Sci. Eng., vol. 10, no. 3, pp. 536–544, Jul. 2013. [55] E. Iwase and I. Shimoyama, “Multistep sequential batch assembly of three- dimensional ferromagnetic microstructures with elastic hinges,” J. Microelectromechanical Syst., vol. 14, no. 6, pp. 1265–1271, Dec. 2005. [56] C. F. Edman et al., “Electric field directed assembly of an InGaAs LED onto silicon circuitry,” IEEE Photonics Technol. Lett., vol. 12, no. 9, pp. 1198– 1200, Sep. 2000. [57] R. H. M. Chan, C. K. M. Fung, and W. J. Li, “Rapid assembly of carbon nanotubes for nanosensing by dielectrophoretic force,” Nanotechnology, vol. 15, no. 10, pp. S672–S677, Oct. 2004. [58] B. Chang, A. Shah, Q. Zhou, R. H. A. Ras, and K. Hjort, “Self-transport and self-alignment of microchips using microscopic rain,” Sci. Rep., vol. 5, p. 14966, Oct. 2015. [59] H. O. Jacobs et al., “Fabrication of a cylindrical display by patterned assembly.,” Science, vol. 296, no. 5566, pp. 323–5, 2002. [60] W. Zheng, P. Buhlmann, and H. O. Jacobs, “Sequential shape-and-solder- directed self-assembly of functional microsystems.,” Proc. Natl. Acad. Sci. U. S. A., vol. 101, no. 35, pp. 12814–7, 2004. [61] K. Reynolds, A. O’Riordan, and G. Redmond, “Self-assembly of a functional electronic circuit directed by capillary interactions,” Appl. Phys. A Mater. Sci. Process., vol. 98, no. 1, pp. 203–209, Jan. 2010. [62] R. J. Knuesel and H. O. Jacobs, “Self-Tiling Monocrystalline Silicon; a Process to Produce Electrically Connected Domains of Si and Microconcentrator Solar Cell Modules on Plastic Supports,” Adv. Mater., vol. 23, no. 24, pp. 2727–2733, Jun. 2011. [63] R. J. Knuesel and H. O. Jacobs, “Self-assembly of microscopic chiplets at a liquid-liquid-solid interface forming a flexible segmented monocrystalline solar cell.,” Proc. Natl. Acad. Sci. U. S. A., vol. 107, no. 3, pp. 993–8, Jan. 2010.

165

[64] J. Dragas et al., “In Vitro Multi-Functional Microelectrode Array Featuring 59 760 Electrodes, 2048 Electrophysiology Channels, Stimulation, Impedance Measurement, and Neurotransmitter Detection Channels,” IEEE J. Solid-State Circuits, vol. 52, no. 6, pp. 1576–1590, 2017. [65] J. M. Nassar et al., “Paper Skin Multisensory Platform for Simultaneous Environmental Monitoring,” Adv. Mater. Technol., vol. 1, no. 1, p. 1600004, Apr. 2016. [66] B. P. L. Lo, H. Ip, and G.-Z. Yang, “Transforming Health Care: Body Sensor Networks, Wearables, and the Internet of Things,” IEEE Pulse, vol. 7, no. 1, pp. 4–8, 2016. [67] J. R. Windmiller and J. Wang, “Wearable Electrochemical Sensors and Biosensors: A Review,” Electroanalysis, vol. 25, no. 1, pp. 29–46, 2013. [68] K. Van Laerhoven et al., “Medical Healthcare Monitoring with Wearable and Implantable Sensors 3 The UbiMon Body Sensor Network Architecture,” Sensors Peterbrgh. NH, no. January, p. 6, 2004. [69] P. Ertl, D. Sticker, V. Charwat, C. Kasper, and G. Lepperdinger, “Lab-on-a- chip technologies for stem cell analysis,” Trends Biotechnol., vol. 32, no. 5, pp. 245–253, 2014. [70] W. Gao et al., “Fully integrated wearable sensor arrays for multiplexed in situ perspiration analysis,” Nature, vol. 529, no. 7587, 2016. [71] G. Schwartz et al., “Flexible polymer transistors with high pressure sensitivity for application in electronic skin and health monitoring,” Nat. Commun., vol. 4, no. 1, p. 1859, Dec. 2013. [72] S. F. Shaikh et al., “Noninvasive Featherlight Wearable Compliant ‘Marine Skin’: Standalone Multisensory System for Deep-Sea Environmental Monitoring,” Small, vol. 15, no. 10, p. 1804385, Mar. 2019. [73] J. M. Nassar et al., “Compliant lightweight non-invasive standalone ‘Marine Skin’ tagging system,” npj Flex. Electron., vol. 2, no. 1, Apr. 2018. [74] A. S. Almuslem, S. F. Shaikh, and M. M. Hussain, “Flexible and Stretchable Electronics for Harsh‐Environmental Applications,” Adv. Mater. Technol., vol. 1900145, p. 1900145, 2019. [75] J. M. Nassar, S. M. Khan, D. R. Villalva, M. M. Nour, A. S. Almuslem, and M. M. Hussain, “Compliant plant wearables for localized microclimate and plant growth monitoring,” npj Flex. Electron., vol. 2, no. 1, p. 24, Dec. 2018. [76] S. M. Khan, S. F. Shaikh, N. Qaiser, and M. M. Hussain, “Flexible Lightweight CMOS-Enabled Multisensory Platform for Plant Microclimate Monitoring,” IEEE Trans. Electron Devices, pp. 1–7, 2018. [77] J. Dragas et al., “In Vitro Multi-Functional Microelectrode Array Featuring

166

59 760 Electrodes, 2048 Electrophysiology Channels, Stimulation, Impedance Measurement, and Neurotransmitter Detection Channels,” IEEE J. Solid-State Circuits, vol. 52, no. 6, pp. 1576–1590, 2017. [78] Y. Kim and M. I. Romero-Ortega, “Material considerations for peripheral nerve interfacing,” MRS Bull., vol. 37, no. 06, pp. 573–580, Jun. 2012. [79] J. Kim et al., “Stretchable silicon nanoribbon electronics for skin prosthesis.,” Nat. Commun., vol. 5, p. 5747, Dec. 2014. [80] I. R. Minev and P. Lacour, “Stretchable Bioelectronics for Medical Devices and Systems,” pp. 257–273, 2016. [81] S. P. Lacour, G. Courtine, and J. Guck, “Materials and technologies for soft implantable neuroprostheses,” Nat. Rev. Mater., vol. 1, no. 10, 2016. [82] M. Ballini et al., “A 1024-channel CMOS microelectrode array with 26,400 electrodes for recording and stimulation of electrogenic cells in vitro,” IEEE J. Solid-State Circuits, vol. 49, no. 11, pp. 2705–2719, Nov. 2014. [83] L. Luan et al., “Nanoelectronics enabled chronic multimodal neural platform in a mouse ischemic model,” J. Neurosci. Methods, Dec. 2017. [84] N. Lei, B. Watson, MacLean J, R. Yuste, and K. Shepard, “A 256-by-256 CMOS microelectrode array for extracellular stimulation of acute brain slices.,” in Proceedings to the International Solid-State Circuits Conference, 2008. [85] S. K. Kang et al., “Bioresorbable silicon electronic sensors for the brain,” Nature, vol. 530, no. 7588, pp. 71–76, 2016. [86] K. J. Yu et al., “Bioresorbable silicon electronics for transient spatiotemporal mapping of electrical activity from the cerebral cortex,” Nat. Mater., vol. 15, no. 7, pp. 782–791, 2016. [87] B. Eversmann, A. Lambacher, T. Gerling, A. Kunze, P. Fromherz, and R. Thewes, “A neural tissue interfacing chip for in-vitro applications with 32k recording / stimulation channels on an active area of 2.6 mm 2,” Eur. Solid-State Circuits Conf., pp. 211–214, 2011. [88] D. T. Simon et al., “Organic electronics for precise delivery of neurotransmitters to modulate mammalian sensory function.,” Nat. Mater., vol. 8, no. 9, pp. 742–746, Sep. 2009. [89] G. Bertotti et al., “A CMOS-based sensor array for in-vitro neural tissue interfacing with 4225 recording sites and 1024 stimulation sites,” in IEEE 2014 Biomedical Circuits and Systems Conference, BioCAS 2014 - Proceedings, 2014, pp. 304–307. [90] Y. S. Lin, S. C. Lee, Y. J. Chen, C. M. Huang, and H. Chiueh, “Live demonstration: A wireless multi-channel physiological signal acquisition

167

system-on-chip for wearable devices,” in Proceedings - 2016 IEEE Biomedical Circuits and Systems Conference, BioCAS 2016, 2016, p. 128. [91] L. Berdondini et al., “Active pixel sensor array for high spatio-temporal resolution electrophysiological recordings from single cell to large scale neuronal networks,” Lab Chip, vol. 9, no. 18, pp. 2644–2651, 2009. [92] U. Frey et al., “Switch-matrix-based high-density microelectrode array in CMOS technology,” IEEE J. Solid-State Circuits, vol. 45, no. 2, pp. 467– 482, Feb. 2010. [93] G. W. Gross, B. K. Rhoades, H. M. E. Azzazy, and Ming-Chi Wu, “The use of neuronal networks on multielectrode arrays as biosensors,” Biosens. Bioelectron., vol. 10, no. 6–7, pp. 553–567, 1995. [94] S. I. Morefield, E. W. Keefer, K. D. Chapman, and G. W. Gross, “Drug evaluations using neuronal networks cultured on microelectrode arrays.,” Biosens. Bioelectron., vol. 15, no. 7–8, pp. 383–96, Oct. 2000. [95] P. Fromherz, “Electrical Interfacing of Nerve Cells and Semiconductor Chips,” ChemPhysChem, vol. 3, no. 3, p. 276, Mar. 2002. [96] S. Han et al., “Battery-free, wireless sensors for full-body pressure and temperature mapping,” Sci. Transl. Med., vol. 10, no. 435, 2018. [97] Y. Wan, Y. Wang, and C. F. Guo, “Recent progresses on flexible tactile sensors,” Mater. Today Phys., vol. 1, pp. 61–73, 2017. [98] X. Wang, L. Dong, H. Zhang, R. Yu, C. Pan, and Z. L. Wang, “Recent Progress in Electronic Skin,” Adv. Sci., vol. 2, no. 10, pp. 1–21, 2015. [99] C.-L. Choong et al., “Highly Stretchable Resistive Pressure Sensors Using a Conductive Elastomeric Composite on a Micropyramid Array,” Adv. Mater., vol. 26, no. 21, pp. 3451–3458, Jun. 2014. [100] S. Wang et al., “Mechanics of Epidermal Electronics,” J. Appl. Mech., vol. 79, no. 3, p. 031022, 2012. [101] X. Huang et al., “Materials and designs for wireless epidermal sensors of hydration and strain,” Adv. Funct. Mater., vol. 24, no. 25, 2014. [102] A. J. Bandodkar, W. Jia, C. Yardımcı, X. Wang, J. Ramirez, and J. Wang, “Tattoo-Based Noninvasive Glucose Monitoring: A Proof-of-Concept Study,” Anal. Chem., vol. 87, no. 1, pp. 394–398, Jan. 2015. [103] D.-H. Kim et al., “Epidermal Electronics,” Science (80-. )., vol. 333, no. 6044, pp. 838–843, Aug. 2011. [104] S. Yang et al., “‘cut-and-Paste’ Manufacture of Multiparametric Epidermal Sensor Systems,” Adv. Mater., vol. 27, no. 41, pp. 6423–6430, 2015. [105] T. Sekitani and T. Someya, “Stretchable, large-area organic electronics,”

168

Adv. Mater., vol. 22, no. 20, pp. 2228–2246, 2010. [106] M. L. Hammock, A. Chortos, B. C. K. Tee, J. B. H. Tok, and Z. Bao, “The Evolution of Electronic Skin (E-Skin): A Brief History, Design Considerations, and Recent Progress,” Adv. Mater., vol. 25, no. 42, pp. 5997–6038, Nov. 2013. [107] J. Kim et al., “Stretchable silicon nanoribbon electronics for skin prosthesis,” Nat. Commun., vol. 5, pp. 1–11, 2014. [108] S. C. B. Mannsfeld et al., “Highly sensitive flexible pressure sensors with microstructured rubber dielectric layers,” Nat. Mater., vol. 9, no. 10, pp. 859–864, 2010. [109] M. Kaltenbrunner et al., “An ultra-lightweight design for imperceptible plastic electronics.,” Nature, vol. 499, no. 7459, 2013. [110] D. H. Kim et al., “Materials for multifunctional balloon catheters with capabilities in cardiac electrophysiological mapping and ablation therapy,” Nat. Mater., vol. 10, no. 4, pp. 316–323, 2011. [111] T. Löher et al., “Stretchable electronic systems: Realization and applications,” in Proceedings of the Electronic Packaging Technology Conference, EPTC, 2009, pp. 893–898. [112] S. Han et al., “Battery-free, wireless sensors for full-body pressure and temperature mapping,” Sci. Transl. Med., vol. 10, no. 435, 2018. [113] T. Chi et al., “A Multi-Modality CMOS Sensor Array for Cell-Based Assay and Drug Screening,” IEEE Trans. Biomed. Circuits Syst., vol. 9, no. 6, pp. 801–814, 2015. [114] Z. Huang et al., “Three-dimensional integrated stretchable electronics,” Nat. Electron., vol. 1, no. 8, pp. 473–480, 2018. [115] N. Lu and D.-H. Kim, “Flexible and Stretchable Electronics Paving the Way for Soft Robotics,” Soft Robot., vol. 1, no. 1, 2014. [116] T. Fukushima, Y. Yamada, H. Kikuchi, and M. Koyanagi, “New three- dimensional integration technology using chip-to-wafer bonding to achieve ultimate super-chip integration,” Japanese J. Appl. Physics, Part 1 Regul. Pap. Short Notes Rev. Pap., vol. 45, no. 4 B, pp. 3030–3035, 2006. [117] J. Viventi et al., “Flexible, foldable, actively multiplexed, high-density electrode array for mapping brain activity in vivo,” Nat. Neurosci., vol. 14, no. 12, pp. 1599–1605, 2011. [118] K. W. Lee, A. Noriki, K. Kiyoyama, T. Fukushima, T. Tanaka, and M. Koyanagi, “Three-dimensional hybrid integration technology of CMOS, MEMS, and photonics circuits for optoelectronic heterogeneous integrated systems,” IEEE Trans. Electron Devices, vol. 58, no. 3, pp. 748–757, 2011.

169

[119] R. R. Bahabry et al., “Corrugation Architecture Enabled Ultraflexible Wafer- Scale High-Efficiency Monocrystalline Silicon Solar Cell,” Adv. Energy Mater., vol. 8, no. 12, 2018. [120] S. F. Shaikh, M. T. Ghoneim, R. R. Bahabry, S. M. Khan, and M. M. Hussain, “Modular Lego-Electronics,” Adv. Mater. Technol., vol. 3, no. 2, p. 1700147, Feb. 2018. [121] S. F. Shaikh, M. T. Ghoneim, G. A. Torres Sevilla, J. M. Nassar, A. M. Hussain, and M. M. Hussain, “Freeform Compliant CMOS Electronic Systems for Internet of Everything Applications,” IEEE Trans. Electron Devices, vol. 64, no. 5, 2017. [122] E. Beyne, “Through-Silicon via Technology for 3D IC,” in Ultra-thin Chip Technology and Applications, New York, NY: Springer New York, 2011, pp. 93–108. [123] M. Koyanagi, “3D-IC Technology Using Ultra-Thin Chips,” in Ultra-thin Chip Technology and Applications, New York, NY: Springer New York, 2011, pp. 109–123. [124] A. Joseph et al., “Through-Silicon via Technology for RF and Millimetre- Wave Applications,” in Ultra-thin Chip Technology and Applications, New York, NY: Springer New York, 2011, pp. 445–453. [125] M. F. Lai, S. W. Li, J. Y. Shih, and K. N. Chen, “Wafer-level three- dimensional integrated circuits (3D IC): Schemes and key technologies,” Microelectron. Eng., vol. 88, no. 11, pp. 3282–3286, 2011. [126] C.-H. Lin, W.-T. Hsieh, H.-C. Hsieh, C.-N. Liu, and J.-C. Yeh, “System-level design exploration for 3-D stacked memory architectures,” Proc. seventh IEEE/ACM/IFIP Int. Conf. Hardware/software codesign Syst. Synth. - CODES+ISSS ’11, p. 389, 2011. [127] X. Sun, X. Peng, Y. Zheng, X. Li, and H. Zhang, “A 3-D Stacked High- $Q$ PI-Based MEMS Inductor for Wireless Power Transmission System in Bio-Implanted Applications,” J. Microelectromechanical Syst., vol. 23, no. 4, pp. 888–898, 2014. [128] A. M. Hussain and M. M. Hussain, “Deterministic Integration of Out-of- Plane Sensor Arrays for Flexible Electronic Applications,” Small, no. 37, pp. 5141–5145, 2016. [129] W. Park, J.-W. Min, S. F. Shaikh, and M. M. Hussain, “Stable MoS2Field-Effect Transistors Using TiO2Interfacial Layer at Metal/MoS2Contact,” Phys. Status Solidi Appl. Mater. Sci., vol. 214, no. 12, 2017. [130] S. Hao, B. Yang, and Y. Gao, “Fracture-induced nanoscrolls from CVD-

170

grown monolayer molybdenum disulfide,” Phys. Status Solidi - Rapid Res. Lett., vol. 10, no. 7, pp. 549–553, 2016. [131] M. S. Cao, X. X. Wang, W. Q. Cao, and J. Yuan, “Ultrathin graphene: electrical properties and highly efficient electromagnetic interference shielding,” J. Mater. Chem. C, vol. 3, no. 26, pp. 6589–6599, 2015. [132] M. Osada and T. Sasaki, “Two-dimensional dielectric nanosheets: Novel nanoelectronics from nanocrystal building blocks,” Adv. Mater., vol. 24, no. 2, pp. 210–228, Jan. 2012. [133] J. N. Burghartz, “Silicon-on-Insulator (SOI) Wafer-Based Thin-Chip Fabrication,” in Ultra-thin Chip Technology and Applications, New York, NY: Springer New York, 2011, pp. 61–67. [134] T. Suwald, “Thin Chips for Document Security,” in Ultra-thin Chip Technology and Applications, New York, NY: Springer New York, 2011, pp. 399–411. [135] M.-U. Hassan, H. Rempp, H. Richter, and N. Wacker, “Electrical Device Characterisation on Ultra-thin Chips,” in Ultra-thin Chip Technology and Applications, New York, NY: Springer New York, 2011, pp. 245–257. [136] S. Mijalković, “MOS Compact Modelling for Flexible Electronics,” in Ultra- thin Chip Technology and Applications, New York, NY: Springer New York, 2011, pp. 259–270. [137] F. Laermer and A. Urban, “Through-Silicon Vias Using Bosch DRIE Process Technology,” in Ultra-thin Chip Technology and Applications, New York, NY: Springer New York, 2011, pp. 81–91. [138] J. M. Nassar et al., “Compliant lightweight non-invasive standalone ‘Marine Skin’ tagging system,” npj Flex. Electron., vol. 2, no. 1, p. 13, 2018. [139] S. F. Shaikh et al., “Noninvasive Featherlight Wearable Compliant ‘Marine Skin’: Standalone Multisensory System for Deep-Sea Environmental Monitoring,” Small, vol. 15, no. 10, p. 1804385, Mar. 2019. [140] R. R. Bahabry, A. C. Sepulveda, A. T. Kutbee, S. F. Shaikh, and M. M. Hussain, “Corrugation Architecture Enabled Ultra-Flexible Mono-Crystalline Silicon Solar Cells via Plasma Etching and Laser Ablation,” in 2018 IEEE 7th World Conference on Photovoltaic Energy Conversion (WCPEC) (A Joint Conference of 45th IEEE PVSC, 28th PVSEC & 34th EU PVSEC), 2018, pp. 0289–0292. [141] N. El-Atab, S. F. Shaikh, S. M. Khan, and M. M. Hussain, “Bi-Facial Substrates Enabled Heterogeneous Multi-Dimensional Integrated Circuits (MD-IC) for Internet of Things (IoT) Applications,” Adv. Eng. Mater., vol. 1900043, pp. 1–6, 2019. [142] J. B. C. Jackson et al., “Historical overfishing and the recent collapse of

171

coastal ecosystems,” Science, vol. 293, no. 5530. American Association for the Advancement of Science, pp. 629–637, 27-Jul-2001. [143] B. S. Halpern et al., “A Global Map of Human Impact on Marine Ecosystems,” Science (80-. )., vol. 319, no. 5865, 2008. [144] C. M. Duarte, “Global change and the future ocean: a grand challenge for marine sciences,” Front. Mar. Sci., vol. 1, p. 63, Dec. 2014. [145] USEPA, “Climate Change Indicators: Stream Temperature,” 2016. [Online]. Available: https://www.epa.gov/climate-indicators/climate-change- indicators-sea-surface-temperature. [Accessed: 01-Oct-2017]. [146] “Chapter 7 The Ocean General Circulation and Climate,” in International Geophysics, vol. 56, no. C, 1994, pp. 171–203. [147] H. A. Broadbent, T. P. Ketterl, C. S. Reid, and J. Dlutowski, “A low-cost, miniature CTD for animal-borne ocean measurements,” in OCEANS 2010 MTS/IEEE SEATTLE, 2010, pp. 1–7. [148] C. J. Bridger and R. K. Booth, “The Effects of Biotelemetry Transmitter Presence and Attachment Procedures on Fish Physiology and Behavior,” Rev. Fish. Sci., vol. 11, no. 1, pp. 13–34, Jan. 2003. [149] N. Jepsen, C. Schreck, S. Clements, and E. B. Thorstad, “Aquatic Telemetry: Advances and Applications : Proceedings of the Fifth ... - Food and Agriculture Organization of the United Nations - Google Books,” Aquatic Telemetry: Advances and Applications : Proceedings of the Fifth Conference on Fish Telemetry, pp. 255–260, 2003. [150] C. C. Wilmers, B. Nickel, C. M. Bryce, J. A. Smith, R. E. Wheat, and V. Yovovich, “The golden age of bio-logging: how animal-borne sensors are advancing the frontiers of ecology,” Ecology, vol. 96, no. 7, pp. 1741–1753, Jul. 2015. [151] R. S. Brown, S. J. Cooke, W. G. Anderson, and A. R. S. Mckinley, “Evidence to Challenge the ‘“2% Rule”’ for Biotelemetry,” North Am. J. Fish. Manag., vol. 19, pp. 867–871, 1999. [152] S. J. Cooke et al., “Biotelemetry: a mechanistic approach to ecology,” Trends Ecol. Evol., vol. 19, no. 6, pp. 334–343, Jun. 2004. [153] K. A. Walker, A. W. Trites, M. Haulena, and D. M. Weary, “A review of the effects of different marking and tagging techniques on marine mammals,” Wildl. Res., vol. 39, no. 1, p. 15, Dec. 2012. [154] S. Andrzejaczek et al., “Temperature and the vertical movements of oceanic whitetip sharks, Carcharhinus longimanus,” Sci. Rep., vol. 8, no. 1, p. 8351, Dec. 2018. [155] L. A. Howey-Jordan et al., “Complex Movements, Philopatry and Expanded

172

Depth Range of a Severely Threatened Pelagic Shark, the Oceanic Whitetip (Carcharhinus longimanus) in the Western North Atlantic,” PLoS One, vol. 8, no. 2, p. e56588, Feb. 2013. [156] A. C. Broderick and B. J. Godley, “Effect of tagging marine turtles on nesting behaviour and reproductive success,” Anim. Behav., vol. 58, no. 3, pp. 587–591, Sep. 1999. [157] K. Alex Shorter, M. M. Murray, M. Johnson, M. Moore, and L. E. Howle, “Drag of suction cup tags on swimming animals: Modeling and measurement,” Mar. Mammal Sci., vol. 30, no. 2, pp. 726–746, Apr. 2014. [158] N. Hammerschlag, S. J. Cooke, A. J. Gallagher, and B. J. Godley, “Considering the fate of electronic tags: interactions with stakeholders and user responsibility when encountering tagged aquatic animals,” Methods Ecol. Evol., vol. 5, no. 11, pp. 1147–1153, Nov. 2014. [159] T. Todd Jones et al., “Calculating the ecological impacts of animal-borne instruments on aquatic organisms,” Methods Ecol. Evol., vol. 4, no. 12, pp. 1178–1186, Dec. 2013. [160] A. M. Hussain and M. M. Hussain, “CMOS-Technology-Enabled Flexible and Stretchable Electronics for Internet of Everything Applications,” Adv. Mater., vol. 28, no. 22, pp. 4219–4249, Jun. 2016. [161] S. F. Shaikh, M. T. Ghoneim, G. A. Torres Sevilla, J. M. Nassar, A. M. Hussain, and M. M. Hussain, “Freeform Compliant CMOS Electronic Systems for Internet of Everything Applications,” IEEE Trans. Electron Devices, vol. 64, no. 5, pp. 1894–1905, May 2017. [162] J. M. Nassar, J. P. Rojas, A. M. Hussain, and M. M. Hussain, “From stretchable to reconfigurable inorganic electronics,” Extreme Mechanics Letters, vol. 9. pp. 245–268, Dec-2016. [163] N. E. Hussey et al., “Aquatic animal telemetry: A panoramic window into the underwater world,” Science (80-. )., vol. 348, no. 6240, 2015. [164] F. Tonolini and F. Adib, “Networking across Boundaries: Enabling Wireless Communication through the Water-Air Interface.” [165] C. Shen et al., “20-meter underwater wireless optical communication link with 15 Gbps data rate,” Opt. Express, vol. 24, no. 22, p. 25502, Oct. 2016. [166] A. Al-Halafi, H. M. Oubei, B. S. Ooi, and B. Shihada, “Real-Time Video Transmission Over Different Underwater Wireless Optical Channels Using a Directly Modulated 520 nm Laser Diode,” J. Opt. Commun. Netw., vol. 9, no. 10, p. 826, Oct. 2017. [167] H. M. Oubei et al., “Light based underwater wireless communications,” Jpn. J. Appl. Phys., vol. 57, no. 8S2, p. 08PA06, Aug. 2018.

173

[168] R. E. Hueter et al., “Movements of three female silky sharks (Carcharhinus falciformis) as tracked by satellite-linked tags off the Caribbean coast of Cuba,” Bull. Mar. Sci., vol. 94, no. 2, pp. 345–358, 2018. [169] R. For and T. H. E. Red, “STATE OF THE MARINE ENVIRONMENT REPORT FOR THE RED SEA AND GULF OF ADEN : 2006 Regional Organization for the Conservation of the,” Environment, 2006. [170] L. S. Balistrieri and J. W. Murray, “The influence of the major ions of seawater on the adsorption of simple organic acids by goethite,” Geochim. Cosmochim. Acta, vol. 51, no. 5, pp. 1151–1160, May 1987. [171] H. Zhang and M. Chiao, “Anti-fouling Coatings of Poly(dimethylsiloxane) Devices for Biological and Biomedical Applications.,” J. Med. Biol. Eng., vol. 35, no. 2, pp. 143–155, 2015. [172] S. Krishnan, C. J. Weinman, and C. K. Ober, “Advances in polymers for anti-biofouling surfaces,” J. Mater. Chem., vol. 18, no. 29, p. 3405, Jul. 2008. [173] E. M. James, “Polymer Data Handbook, 2nd ed.,” J. Am. Chem. Soc., vol. 131, no. 44, pp. 16330–16330, Nov. 2009. [174] M. J. Vucko et al., “Using textured PDMS to prevent settlement and enhance release of marine fouling organisms,” vol. 7014, 2014. [175] S. Krishnan, J. Weinman, and C. K. Ober, “Advances in polymers for anti- biofouling surfaces,” pp. 3405–3413, 2008. [176] C. Liu, C. Ma, Q. Xie, and G. Zhang, “Self-repairing silicone coatings for marine anti-biofouling,” pp. 15855–15861, 2017. [177] A. Furniturewalla, M. Chan, J. Sui, K. Ahuja, and M. Javanmard, “Fully integrated wearable impedance cytometry platform on flexible circuit board with online smartphone readout,” Microsystems Nanoeng., vol. 4, no. 1, p. 20, 2018. [178] S. C. Doney et al., “Climate Change Impacts on Marine Ecosystems,” Ann. Rev. Mar. Sci., vol. 4, no. 1, pp. 11–37, Jan. 2012. [179] J. H. Koo, D. C. Kim, H. J. Shim, T. H. Kim, and D. H. Kim, “Flexible and Stretchable Smart Display: Materials, Fabrication, Device Design, and System Integration,” Advanced Functional Materials, 2018. [180] “Salinity | Science Mission Directorate.” [Online]. Available: https://science.nasa.gov/earth-science/oceanography/physical- ocean/salinity. [Accessed: 06-Jul-2018]. [181] “Integument - Fishes | Britannica.com.” [Online]. Available: https://www.britannica.com/science/integument/Fishes. [Accessed: 29- Aug-2018].

174

[182] R. Bonfil, “The Biology and Ecology of the Silky Shark, Carcharhinus Falciformis,” Sharks Open Ocean Biol. Fish. Conserv., no. 1982, pp. 114– 127, 2009. [183] T. Sekitani et al., “Stretchable active-matrix organic light-emitting diode display using printable elastic conductors,” Nat. Mater., vol. 8, no. 6, pp. 494–499, Jun. 2009. [184] S. Astrophysique, C. E. A. Saclay, G. Yvette, I.- Trieste, S. Padova, and I.- Padova, “10,11,7,” vol. 326, no. December, pp. 1516–1520, 2009. [185] M.-Y. Wu, J. Zhao, N. J. Curley, T.-H. H. Chang, Z. Ma, and M. S. Arnold, “Biaxially stretchable carbon nanotube transistors,” J. Appl. Phys., vol. 122, no. 12, p. 124901, Sep. 2017. [186] N. O. Weiss et al., “Graphene: An emerging electronic material,” Adv. Mater., vol. 24, no. 43, pp. 5782–5825, 2012. [187] Y. Zhu and F. Xu, “Buckling of aligned carbon nanotubes as stretchable conductors: A new manufacturing strategy,” Adv. Mater., vol. 24, no. 8, pp. 1073–1077, 2012. [188] C. Wang et al., “User-interactive electronic skin for instantaneous pressure visualization,” Nat Mater, vol. 12, no. 10, pp. 899–904, Jul. 2013. [189] K. Takei et al., “Nanowire active-matrix circuitry for low-voltage macroscale artificial skin.,” Nat. Mater., vol. 9, no. 10, pp. 821–6, Oct. 2010. [190] Y. H. Jung et al., “High-performance green flexible electronics based on biodegradable cellulose nanofibril paper.,” Nat. Commun., vol. 6, p. 7170, 2015. [191] Y. Yuan et al., “Ultra-high mobility transparent organic thin film transistors grown by an off-centre spin-coating method.,” Nat. Commun., vol. 5, p. 3005, 2014. [192] A. C. Siegel, S. T. Phillips, B. J. Wiley, and G. M. Whitesides, “Thin, lightweight, foldable thermochromic displays on paper,” Lab Chip, vol. 9, no. 19, p. 2775, Oct. 2009. [193] L. Wang et al., “Simple, Rapid, Sensitive, and Versatile SWNT−Paper Sensor for Environmental Toxin Detection Competitive with ELISA,” Nano Lett., vol. 9, no. 12, pp. 4147–4152, Dec. 2009. [194] J. Lessing, A. C. Glavan, S. B. Walker, C. Keplinger, J. A. Lewis, and G. M. Whitesides, “Inkjet Printing of Conductive Inks with High Lateral Resolution on Omniphobic ‘R F Paper’ for Paper-Based Electronics and MEMS,” Adv. Mater., vol. 26, no. 27, pp. 4677–4682, Jul. 2014. [195] J. M. Nassar, K. Mishra, K. Lau, A. A. Aguirre-Pablo, and M. M. Hussain, “Recyclable Nonfunctionalized Paper-Based Ultralow-Cost Wearable

175

Health Monitoring System,” Adv. Mater. Technol., vol. 2, no. 4, pp. 1–11, 2017. [196] J. A. Fan et al., “Fractal design concepts for stretchable electronics,” Nat. Commun., vol. 5, pp. 1–8, 2014. [197] D.-H. Kim, R. Ghaffari, N. Lu, and J. A. Rogers, “Flexible and Stretchable Electronics for Biointegrated Devices,” Annu. Rev. Biomed. Eng., vol. 14, no. 1, pp. 113–128, 2012. [198] J. P. Rojas, A. Arevalo, I. G. Foulds, and M. M. Hussain, “Design and characterization of ultra-stretchable monolithic silicon fabric,” Appl. Phys. Lett., vol. 105, no. 15, 2014. [199] N. Alcheikh, S. F. Shaikh, and M. M. Hussain, “Ultra-stretchable Archimedean interconnects for stretchable electronics,” Extrem. Mech. Lett., vol. 24, pp. 6–13, Oct. 2018. [200] N. Qaiser, S. M. Khan, M. Nour, M. U. Rehman, J. P. Rojas, and M. M. Hussain, “Mechanical response of spiral interconnect arrays for highly stretchable electronics,” Appl. Phys. Lett., vol. 111, no. 21, p. 214102, Nov. 2017. [201] A. C. Cavazos Sepulveda, M. S. Diaz Cordero, A. A. A. Carreño, J. M. Nassar, and M. M. Hussain, “Stretchable and foldable silicon-based electronics,” Appl. Phys. Lett., vol. 110, no. 13, 2017. [202] J. N. Burghartz et al., “Ultra-thin Si chips for flexible electronics Process technology, characterization, assembly and applications,” in 28th Symposium on Microelectronics Technology and Devices (SBMicro 2013), 2013, pp. 1–3. [203] Y. M. Song et al., “Digital cameras with designs inspired by the arthropod eye,” Nature, vol. 497, no. 7447, pp. 95–99, May 2013. [204] C. M. Lochner, Y. Khan, A. Pierre, and A. C. Arias, “All-organic optoelectronic sensor for pulse oximetry,” Nat. Commun., vol. 5, pp. 1–7, 2014. [205] G. Schwartz et al., “Flexible polymer transistors with high pressure sensitivity for application in electronic skin and health monitoring,” Nat. Commun., vol. 4, no. May, pp. 1858–1859, 2013. [206] G. A. T. Sevilla et al., “Decal Electronics: Printable Packaged with 3D Printing High-Performance Flexible CMOS Electronic Systems,” Adv. Mater. Technol., vol. 2, no. 1, p. 1600175, Jan. 2017. [207] S. Khan, N. Qaiser, and M. Mustafa Hussain, Do-It-Yourself (DIY) Flexible Paper Sensor Based Electronic System for Pill Health Monitoring, 1570429385th ed. Ottawa, Ontario Canada: IEEE International Flexible Electronics Technology Conference, 2018.

176

[208] M. T. Tolley et al., “A Resilient, Untethered Soft Robot,” Soft Robot., vol. 1, no. 3, pp. 213–223, Sep. 2014. [209] Y. H. Jung et al., “Stretchable Twisted-Pair Transmission Lines for Microwave Frequency Wearable Electronics,” Adv. Funct. Mater., vol. 26, no. 26, pp. 4635–4642, 2016. [210] J. H. Ahn et al., “High-speed mechanically flexible single-crystal silicon thin-film transistors on plastic substrates,” IEEE Electron Device Lett., vol. 27, no. 6, pp. 460–462, 2006. [211] C. Wang et al., “Self-Aligned, Extremely High Frequency III–V Metal-Oxide- Semiconductor Field-Effect Transistors on Rigid and Flexible Substrates,” Nano Lett., vol. 12, no. 8, pp. 4140–4145, Aug. 2012. [212] T. Chang et al., “A General Strategy for Stretchable Microwave Antenna Systems using Serpentine Mesh Layouts,” Adv. Funct. Mater., vol. 27, no. 46, p. 1703059, Dec. 2017. [213] S. J. Cho, Y. H. Jung, and Z. Ma, “X-band compatible flexible microwave inductors and capacitors on plastic substrate,” IEEE J. Electron Devices Soc., vol. 3, no. 5, pp. 435–439, Sep. 2015. [214] Y. H. Jung et al., “A Compact Parylene-Coated WLAN Flexible Antenna for Implantable Electronics,” IEEE Antennas Wirel. Propag. Lett., vol. 15, pp. 1382–1385, 2016. [215] Tzu-Hsuan Chang et al., “High power fast flexible electronics: Transparent RF AlGaN/GaN HEMTs on plastic substrates,” in 2015 IEEE MTT-S International Microwave Symposium, 2015, pp. 1–4. [216] H.-C. Yuan and Z. Ma, “Microwave thin-film transistors using Si nanomembranes on flexible polymer substrate,” Appl. Phys. Lett., vol. 89, no. 21, p. 212105, Nov. 2006. [217] Y. H. Jung, H. Zhang, S. J. Cho, and Z. Ma, “Flexible and Stretchable Microwave Microelectronic Devices and Circuits,” IEEE Trans. Electron Devices, vol. 64, no. 5, pp. 1881–1893, 2017. [218] X. Huang et al., “Epidermal radio frequency electronics for wireless power transfer,” Microsystems Nanoeng., vol. 2, no. April, p. 16052, 2016. [219] S. Mhedhbi et al., “First Power Performance Demonstration of Flexible AlGaN/GaN High Electron Mobility Transistor,” IEEE Electron Device Lett., vol. 37, no. 5, pp. 553–555, May 2016. [220] H. Mi et al., “Characterizations of biodegradable epoxy-coated cellulose nanofibrils (CNF) thin film for flexible microwave applications,” Cellulose, vol. 23, no. 3, pp. 1989–1995, Jun. 2016.