Gas viscosity as a sensing element: microbubble chromatography

Air quality monitoring has become increasingly important because of the growing need for public health assessment under inevitable exposure to both indoor and outdoor air pollution. Monitoring air quality requires the detection of a wide range of gases, especially, at a part-per-billion (ppb) level. Gas chromatography (GC), despite its extensive volumes and power consumption, is the most widely utilized technique to separate and detect multiple compounds from a mixture. Among all miniaturized detectors for a gas chromatography system, only a few are capable of detecting gases down in ppb levels. To address such problems, a gas detector must avoid the use of any chemistry films, extensive peripheral equipment, and ion destructive mechanisms. This dissertation first presents the phenomenon of viscosity-induced pressure transients as a potential sensing element and thus as a solution to address the aforementioned challenges. Then, this dissertation secondly presents the use of gas bubbles, when gas is streamed into liquid flow through a micro nozzle, to enable quantification, which was named as microbubble chromatography. The proof-of-concept viscosity-based sensor was realized by constructing coupled microchannels with different diameters and placing a pressure detector in the second channel. This proof-of-concept viscosity-based sensor demonstrated (i) the detection of 12 volatile organic compounds, (ii) a minimum detection limit of 13.1 (15.3 ppb), 13.6 (13.7 ppb), and 17.5 picograms (22.6 ppb) for hexane, heptane, and benzene, respectively, (iii) an average sensitivity of 15×10-6 sccm/pg, and (iv) temperature stability of 94.3% in a temperature range of 40-110 0C for hexane compounds in a microgas chromatography system. Furthermore, this dissertation demonstrated successful quantification of the detected gases by utilizing the viscosity-to-pressure conversion mechanism in a microfluidic device. The device produced different sizes of microbubbles for different types of gases and enabled quantification via the counting of the number of bubbles. The fabricated bubble chromatography system (i) produced unique bubble diameters of 52.5, 93.0, 98.5, and 103.5μm from four types of gases (He, H2, N2, and CH4), (ii) demonstrated bubble chromatogram by injecting 0.1 μL pentane, and (iii) achieved a detection limit of 4.5 nanogram with sensitivity of 0.34 μm/ng and 6.94 bubbles/ng.

GAS VISCOSITY AS A SENSING ELEMENT: MICROBUBBLE CHROMATOGRAPHY by A N M Ashrafuzzaman Bulbul A dissertation submitted to the faculty of The University of Utah in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Electrical and Computer Engineering The University of Utah May 2020 Copyright © A N M Ashrafuzzaman Bulbul 2020 All Rights Reserved The University of Utah Graduate School STATEMENT OF DISSERTATION APPROVAL The dissertation of A N M Ashrafuzzaman Bulbul has been approved by the following supervisory committee members: , Chair 07-22-2019 Carlos Mastrangelo , Member 07-22-2019 Florian Solzbacher , Member 07-22-2019 Massood Tabib-Azar , Member 07-22-2019 Bruce Gale , Member 07-22-2019 Hanseup Kim and by Florian Solzbacher the Department/College/School of Date Approved Date Approved Date Approved Date Approved Date Approved , Chair/Dean of Electrical and Computer Engineering and by David B. Kieda, Dean of The Graduate School. ABSTRACT Air quality monitoring has become increasingly important because of the growing need for public health assessment under inevitable exposure to both indoor and outdoor air pollution. Monitoring air quality requires the detection of a wide range of gases, especially, at a part-per-billion (ppb) level. Gas chromatography (GC), despite its extensive volumes and power consumption, is the most widely utilized technique to separate and detect multiple compounds from a mixture. Among all miniaturized detectors for a gas chromatography system, only a few are capable of detecting gases down in ppb levels. To address such problems, a gas detector must avoid the use of any chemistry films, extensive peripheral equipment, and ion destructive mechanisms. This dissertation first presents the phenomenon of viscosity-induced pressure transients as a potential sensing element and thus as a solution to address the aforementioned challenges. Then, this dissertation secondly presents the use of gas bubbles, when gas is streamed into liquid flow through a micro nozzle, to enable quantification, which was named as microbubble chromatography. The proof-of-concept viscosity-based sensor was realized by constructing coupled microchannels with different diameters and placing a pressure detector in the second channel. This proof-of-concept viscosity-based sensor demonstrated (i) the detection of 12 volatile organic compounds, (ii) a minimum detection limit of 13.1 (15.3 ppb), 13.6 (13.7 ppb), and 17.5 picograms (22.6 ppb) for hexane, heptane, and benzene, respectively, (iii) an average sensitivity of 15×10-6 sccm/pg, and (iv) temperature stability of 94.3% in a temperature range of 40-110 0C for hexane compounds in a microgas chromatography system. Furthermore, this dissertation demonstrated successful quantification of the detected gases by utilizing the viscosity-to-pressure conversion mechanism in a microfluidic device. The device produced different sizes of microbubbles for different types of gases and enabled quantification via the counting of the number of bubbles. The fabricated bubble chromatography system (i) produced unique bubble diameters of 52.5, 93.0, 98.5, and 103.5μm from four types of gases (He, H2, N2, and CH4), (ii) demonstrated bubble chromatogram by injecting 0.1 μL pentane, and (iii) achieved a detection limit of 4.5 nanogram with sensitivity of 0.34 μm/ng and 6.94 bubbles/ng. iv To My family, Mou and Anayah TABLE OF CONTENTS ABSTRACT....................................................................................................................... iii LIST OF FIGURES ......................................................................................................... viii ACKNOWLEDGMENTS ................................................................................................. xi Chapters 1. INTRODUCTION .......................................................................................................... 1 1.1 Importance of Gas Sensing .................................................................................. 1 1.2 Importance of Wide Ranges of Gas Detection and Quantification ..................... 2 1.3 State-of-the-Art for Wide Ranges of Gas Detection in PPB ............................... 6 1.4 Limitations of PPB Level Gas Sensors .............................................................. 10 1.5 Solution: Viscosity as a New Sensing Element ................................................. 11 1.6 Improved Quantification: Microbubble Chromatography ................................. 14 1.7 Organization of Dissertation .............................................................................. 14 1.8 Innovation and Significance .............................................................................. 15 2. LITERATURE REVIEW OF MINIATURIZED GAS SENSORS.............................. 17 2.1 Working Principle of a Gas Sensor.................................................................... 17 2.2 Characteristics of a Microgas Sensor................................................................. 18 2.3 Different Types of Gas Sensors ......................................................................... 19 2.3.1 Chemistry Dependent/Sorption Based ................................................... 20 2.3.2 Chemistry Independent .......................................................................... 28 2.4 Summary of Microgas Sensors .......................................................................... 35 3. VISCOSITY DETECTION: THEORY ........................................................................ 37 3.1 Gas Viscosity: Unique Physical Property .......................................................... 37 3.2 Viscosity-to-Pressure Conversion...................................................................... 38 3.3 Pressure-to-Flow Generation ............................................................................. 41 3.4 Flow Amplification ............................................................................................ 43 3.5 Sensor Design .................................................................................................... 46 3.5.1 Improvement of Flow Rate .................................................................... 46 3.5.2 Signal Amplification .............................................................................. 47 4. GAS DETECTION BY A VISCOSITY-BASED SENSOR ........................................ 49 4.1 Permanent Gas Detection................................................................................... 49 4.1.1 Testing Methodology ............................................................................. 49 4.1.2 Gas Chromatograms............................................................................... 50 4.2 VOC Detection .................................................................................................. 52 4.2.1 Testing Methodology ............................................................................. 52 4.2.2 VOC Chromatograms ............................................................................ 53 4.3 Limit of Detection (LOD) .................................................................................. 55 4.4 Sensitivity .......................................................................................................... 56 4.5 Effect of Temperature ........................................................................................ 58 4.6 Effect of Carrier Gas .......................................................................................... 59 5. MICROBUBBLE CHROMATOGRAPHY ................................................................. 61 5.1 Microbubble Chromatography........................................................................... 61 5.2 Mechanism of Microbubble Formation ............................................................. 63 5.3 Microfluidic Device for Microbubble Formation .............................................. 64 5.3.1 Structure ................................................................................................. 64 5.3.2 Fabrication ............................................................................................. 65 5.4 Testing Methodology ......................................................................................... 67 5.4.1 Overview of Testing Setup .................................................................... 67 5.4.2 Bubble Size Measurement ..................................................................... 68 5.5 Viscosity Effect During Bubble Formation ....................................................... 69 5.5.1 Viscosity Effect at Gas PDMS Interface ............................................... 69 5.5.2 Viscosity Effect at Gas-Liquid Interface ............................................... 70 5.6 Bubble Volume .................................................................................................. 74 5.6.1 Modelling of Bubble Volume ................................................................ 74 5.6.2 Modelling Versus Experimental ............................................................ 75 5.7 Gas Detection..................................................................................................... 77 5.7.1 Bubble Chromatogram ........................................................................... 77 5.7.2 Modelling of Bubble Chromatogram ..................................................... 79 5.8 Gas Quantification ............................................................................................. 80 5.9 Limit of Detection and Sensitivity..................................................................... 83 5.10 Limitations of Microbubble Chromatography ................................................. 84 6. CONCLUSION ............................................................................................................. 87 6.1 Summary ............................................................................................................ 87 6.2 Future Work ....................................................................................................... 89 APPENDIX : PRESSURE SENSING STENT................................................................. 93 REFERENCES ............................................................................................................... 117 vii LIST OF FIGURES Figures 1.1. Percentage of VOC exposure from ambient to personal level, Camden, New Jersey, adapted from reference [8]. ................................................................................................. 3 1.2. Wide ranges of VOCs and importance of quantification: (a) Word cloud of 187 VOCs and five major non-VOCs, adapted from reference [9]; (b) WHO safely level of some pollutants, adapted from reference [11]..................................................................... 4 1.3. Gas chromatography system and popular sensors of a GC system. (a) Principle of gas chromatography: the ability to separate multiple compounds; (b) state-of-the-art gas sensors for a gas chromatography system, adapted after evaluating detection limit (ppb), power (mW), and volume (cm3) from the references [24-74], and each point indicates the best performance among that specific sensor group. .......................................................... 7 1.4. Possible application of a portable gas sensor in home solution, distributed area network, smart phone, and smart watch.............................................................................. 9 1.5. Viscosity-based gas sensor as solution: some features of a viscosity-based gas sensor. ............................................................................................................................... 13 2.1. Principle of a gas sensor ............................................................................................ 18 2.2. Different types of gas sensors .................................................................................... 20 2.3. Principle of a Surface Acoustic Wave (SAW) sensor. .............................................. 21 2.4. Principle of a chemiresistor sensor. ........................................................................... 23 2.5. Working principle of a Metal Oxide Semiconductor (MOS) sensor. ........................ 25 2.6. Principle of a Thermal Conductivity Detector (TCD). .............................................. 29 2.7. Working principle of a Flame Ionization Detector (FID). ......................................... 31 2.8. Principle of Photoionization detector (PID). ............................................................. 33 3.1. Experimental results of viscosity-to-pressure conversion. ........................................ 40 3.2. Conceptual explanation of pressure-to flow generation by contraction. ................... 42 3.3. Flow amplification with nonlinear pressure gradient: (a) Nonlinear pressure distribution across narrow microchannel: pressure gradient increased nonlinearly across 50 μm diameter narrow microchannel; (b) Flow amplification through a narrow microchannel. .................................................................................................................... 45 3.4. Pressure-to-flow generation in T-junction-based expansion. .................................... 46 3.5. Signal amplification in T-junction-based expansion. ................................................ 47 4.1. Flow chromatogram of five different permanent gases (He, N2, CO2, CH4, H2); He carrier: downward signal (flow rate change) for less viscous gas than helium; N2 carrier: upward signal for higher viscous He but downward signal for less viscous CO2, CH4, H2; CO2 carrier: upward signal for higher viscous He, N2 but downward signal for less viscous CH4, H2; CH4 carrier: upward signal for higher viscous He, N2, CO2 but downward signal for less viscous H2; H2 carrier: upward signal for higher viscous gas He, N2, CO2, CH4. Chromatogram of five different gases with different carrier gases. ........ 51 4.2. Chromatogram of 12 different VOCs with (left): viscosity-based sensor, (right) Flame ionization detector; the compounds are: (1) methanol, (2) n-pentane (3) n-hexane, (4) benzene, (5) n-heptane, (6) toluene, (7) n-octane, (8) tetrachloroethylene, (9) chlorobenzene, (10) ethylbenzene, (11) m-xylene, and (12) n-nonane............................. 53 4.3. Limit of detection of viscosity-based gas sensor for hexane gas. .............................. 56 4.4. The sensitivity of viscosity-based detector: (a) sensitivity of hexane, heptane, and benzene; (b) sensitivity of heptane with respect to corresponding FID output. ............... 57 4.5. Signal variation with the variation of temperature from 40 to 1200C........................ 59 4.6. Comparison between helium and nitrogen in terms of senor performance. .............. 60 5.1. Concept of microbubble chromatography. ................................................................ 62 5.2. Bubble formation mechanism. ................................................................................... 63 5.3. Structure and fabrication. ........................................................................................... 65 5.4. Overview of testing setup. ......................................................................................... 67 5.5. Image processing algorithm. ...................................................................................... 68 5.6. Modified device to test gas loss in a liquid phase. ..................................................... 70 5.7. Testing results of gas loss in a liquid phase. .............................................................. 72 ix 5.8. Bubble volume (nL): theoretical versus experimental. .............................................. 76 5.9. Bubble chromatogram of permanent gases and VOCs: (a) target gas: hydrogen, carrier gas: helium, (b) target gas: helium, carrier gas: hydrogen, (c) target gas: pentane, carrier gas: helium............................................................................................................. 78 5.10. Gas loss mechanism to generate chromatogram. ..................................................... 81 5.11. Bubble chromatogram: theoretical versus experimental.......................................... 82 5.12. Limit of detection and sensitivity of bubble chromatography. ................................ 83 5.13. Limitation of bubble size and frequency measurement. .......................................... 85 A.1. Main concept of the proposed pressure sensing stent: One component system: capacitive pressure sensor embedded on stretchable stent (stainless steel wire); Signal amplification: diaphragm displacement by small amount (d) creates larger displacement (100×d) inside liquid microchannel; Signal digitization: capacitive signal increases by a step (digitized) as the liquid moves one electrode to the other; Pressure profiling: pressure sensors at multiple locations will indicate the pressure profiling due to plaque. ............. 97 A.2. Equivalent circuit: (a) model circuit of the pressure sensor forms RLC transient circuit, (b) SPICE simulation model based on the values mentioned in Table A.1, and (c) liquid flow rate in the outlet channel showed transient behavior. .................................... 98 A.3. Surface roughness after two steps of polishing. ...................................................... 108 A.4. Fabrication of on-wire pressure sensor: (a) fabrication steps and (b) fabricated pressure sensing wire (stent). .......................................................................................... 110 A.5. Characterization of glycerol evaporation process. .................................................. 111 A.6. Optical measurement of pressure sensing stent: (a) testing setup and (b) characterization of optically measured liquid displacement. .......................................... 113 A.7. Capacitance measurement of pressure sensing stent: (a) capacitance change from one electrode to another and (b) COMSOL simulation of capacitance change. ............. 114 A.8. Wireless testing setup and change of resonance frequency with pressure. ............. 115 x ACKNOWLEDGMENTS First I would like to thank my advisor Professor Hanseup Kim for guiding me over the entire period of my PhD. life. Without his continuous support and inspiration, this project and dissertation would never have been reached the current level. I also would like to thank my dissertation committee members, Professor Carlos Mastrangelo, Professor Massood Tabib-Azar, Professor Florian Solzbacher, and Professor Bruce gale, for providing valuable comments and directions during my both qualifying exam and proposal defense. This work could not have been continued without financial support from the research award NSF-ECCS-1509912. I thank to my ECE department for providing me tuition support in every semester. I was fortunate to have lots of good friends and colleagues over the course of my PhD. I would like to thank my friends and colleagues from my research group, HaoChieh Hsieh, Mahbubur Rahman, Mahatasin Azad, Seungbeom Noh, Ross Booth, Shakirul Khan, and Yunhao Peng. I also would like to mention some of my friends outside of my research group, Kevin Petersen, Pradeep Pai, and Qingboo Guo. Last but not the least, I am grateful to my wife Mou- whose love, affection, and support encouraged me to find a hope in my everyday work. I am very much thankful to my parents and sisters for their unconditional love and support even though I am 7746 miles away from home. CHAPTER 1 INTRODUCTION 1.1 Importance of Gas Sensing Gas sensing is a critical issue as air pollution has rapidly emerged as a health threatening issue all over the world [1]. The air pollution has been reported to be responsible for the growing number of fatalities [1]. Such air pollutants include various types of gaseous and chemical substances, including sulfur dioxide (SO2), nitrogen oxides (NOx), carbon monoxide (CO), ozone (O3), and >100 types of volatile organic compounds (VOCs). The amounts of air pollutants, being inhaled by humans daily, are estimated to be significant because they are included in the daily inhalation of between 10,800 to 18,000 liters of air [2] while the air is now increasingly polluted from rapid industrialization, emission from fuel combustions, and industrial by-products. According to the World Health Organization (WHO) report, the fatalities from air pollution was totaling over 70,000 lives in 2002 only in the US [3]. This number is twice the number of automobile fatalities and is equal to deaths from breast cancer and prostate cancer combined [3]. Over the last 26 years, from 1990 to 2016, the number of deaths caused by ambient air pollution increased by 23.5% from 3.4 million to 4.2 million per year worldwide. The diseases, responsible for these deaths, include cardiovascular diseases such as stroke and ischemic heart disease, chronic obstructive pulmonary disease, lung 2 cancer, and lower respiratory infections [1]. Indoor air quality imposes higher risk since humans spent more than 80% of their time inside an indoor environment such as offices, schools, workplaces, homes, and public or private buildings [4]. Therefore, individuals are more exposed to particular pollutants over the entire period of a day, even at the low-level existence of that pollutant. According to IHME (Institution for Health Metrics and Evaluation), 2.6 million people died in 2016 because of illness attributed to indoor air pollution [5]. Particular sources of indoor air pollution include tobacco smoke, gas stoves, household cleaning products, paints, adhesive, new carpets, and various chemicals used for interior decoration. For example, paint and glue are significant sources of benzene, toluene, ethylbenzene, and xylene, and incense burning is a significant source of carbon monoxide and nitrogen dioxide [6] [7]. According to a study [8] carried out in a village (Water Front South) at Camden, New Jersey, the mean ambient concentration of carcinogenic VOCs was 27±16 ppm, and benzene was the major contributor. Some of the significant VOCs were MTBE (Methyl tert-butyl ether), benzene, hexane, toluene, ethylbenzene, m/p-xylene, o-xylene, chloroform, and carbon tetrachloride. Percentages of personal level exposure from an ambient of these gases were found as 49%, 22%, 33%, 26%, 22%, 28%, 23%, 17%, and 6%, respectively, as shown in Figure. 1.1. Therefore, monitoring air quality is highly demanded for public health. 1.2 Importance of Wide Ranges of Gas Detection and Quantification Air quality monitoring requires detection and quantification of a wide range of VOCs and non-VOCs. Note that a VOC is a carbon-related compound and quickly 3 6 Toluene (26%) Indoor concentration (μg/m3) 5 Reference 100% Hexane (33%) 4 Ethylbenzene (22%) 3 2 1 0 MTBE (49%) m/p-Xylene (28%) Benzene (22%) o- Xylene (23%) C-tetrachloride(6%) Chloroform(17%) 0 0.5 1 1.5 2 2.5 3 Ambient concentration (μg/m3) Figure 1.1. Percentage of VOC exposure from ambient to personal level, Camden, New Jersey, adapted from reference [8]. evaporates under atmospheric temperature and pressure, where a non-VOC represents a noncarbon gaseous air pollutant. According to EPA (Environmental Protection Agency), there are 187 hazardous VOCs and five major non-VOCs [9], as shown in Figure 1.2. Here highlighted compounds indicated the degree of presence of each compound in our environment. The major non-VOCs were SO2, NO2, O3, CO, and CO2. On another side, AQT (air quality technology) has reported 636 VOCs [10]. Gas quantification is a technique to detect the concentration of a toxic gas present in the testing environment. Gas quantification is another necessary step of air quality 4 (a) Wide ranges VOCs and non-VOCs 10 5 Concentration (PPB) 104 103 (b) Quantification: WHO safety level WHO-safety level CO N2O Benzene (6000 ppb) (21.2 ppb) (5.2 ppb) Toluene Formal(1.5 ppb) dehyde (81.4 ppb) 102 101 0 5 Gas stove Kerosene stove Incense Ice arena 4 Workplace School Home Furniture shop 3 Gas stove BBQ Open fire Incense 2 Grinding Smoking Carpentering Painting 1 Burning Kerosene stove Smoking Painting 10 Figure 1.2. Wide ranges of VOCs and importance of quantification: (a) Word cloud of 187 VOCs and five major non-VOCs, adapted from reference [9]; (b) WHO safely level of some pollutants, adapted from reference [11]. 5 monitoring because there is a health risk when a gas exposure level exceeds a certain safety threshold. Therefore, continuous monitoring and threshold checking are essential at homes, offices, schools, and different occupations. According to WHO guidelines, safety concentration levels of some toxic gases such as benzene, toluene, carbon monoxide, formaldehyde, and nitrogen dioxide are 5.2 ppb, 1.5 ppb, 6 ppm, 81.4 ppb, and 21.2 ppb, respectively, [11] [12], as shown in Figure 1.2(b). These gases often exceed the safety level in our everyday life. For example, benzene concentration varies in 13.651.4 ppb (× 2.6-9.9 above the safety level) in a kerosene stove kitchen, toluene ranges in 30-234 ppb (× 20-156) in some of the major US metropolitan centers, carbon monoxide varies from 12.2 to 45.8 ppm (× 2.0-7.6) inside underground tunnels or tunnel-booths, formaldehyde varies in the range of 17.9-139.2 ppb (× 0.2-1.7) inside a bedroom with different furniture, wooden products, candle or incense burning, textiles, etc., and nitrogen dioxide was measured as 4000 ppb (× 188.7) inside an ice arena with inadequate ventilation from the petrol-fueled ice resurfacing machine [13] [14] [15]. Figure 1.2(b) shows some significant sources among these toxic gases. For example, the emission of benzene from burning, cooking with a kerosene stove, smoking and painting were 36, 51, 554, and 4004 ppb, respectively. Different levels of toxic gas concentrations cause both noncarcinogenic and carcinogenic effects. For example, short term exposure of benzene such as in 50-100 ppm for 30 minutes causes fatigue, and in 250-500 ppm causes dizziness, headache [11]. On the other hand, long term exposure causes chronic issues on health. According to a study, 17 years of exposure to 15-30 ppm of benzene causes shoe factory workers to suffer from leucopenia, thrombocytopenia, and pancytopenia [11]. Therefore, air quality monitoring requires both detection and quantification of a wide 6 range of gases in ppb levels. 1.3 State-of-the-Art for Wide Ranges of Gas Detection in PPB Wide ranges of gases can be detected either by gas chromatography-based sensors or standalone sensors; however, the gas chromatography-based sensors provide a wider range of gas detection as compared to standalone sensors. A gas chromatography column helps to separate wide ranges of gases in time and space and then transfer each gas to detectors one by one. Where a standalone sensor (e.g., SAW, CR, FET, and MOS) detects multiple gases at the same time with the help of arrays of chemistry and previously trained neural network models. These standalone gas sensors cannot detect wider ranges of gases because the degree of uncertainty to identify a gas increases with the increasing number of unknown gases. For example, Hsieh et al. [16] conducted an experiment with an array of six different polymer-coated SAW sensors and reported that the ability to recognize a gas component was reduced to 3% in a ternary mixture from 89% in binary mixtures. A commercial gas chromatography column can separate as high as 112 different compounds [17], where a microscale gas chromatography column can separate 38-45 compounds. Note that there are more than 600 compounds and gases in science and engineering [10]. Reidy et al. [18] detected 30 compounds with a 0.25 m long microseparation column and a resistance temperature detector (RTD). Kim et al. [19] reported the detection of 38 compounds with a 3 m microcolumn and a chemiresistor array. Lambertus et al. [20] reported screening of 45 different compounds with a 3 m long microcolumn and a differential mobility spectrometer (DMS) detector. The principle of a gas chromatography system is shown in Figure 1.3(a). It showed that at first multiple 7 (a) Principle of gas chromatography Mixed Vaporization compounds Separated compounds Separation C7 C6 Detector Chromatogram C5 C6 C7 C5 Time Helium flow Detection limit (ppb) (b) Gas Sensors for a GC system 10 4 10 3 10 2 10 1 10 0 Chemistry Non-chemistry MOS CR QCM Optical 10 TCD FID FET Viscosity PID SAW MS -1 10 -5 10 0 10 5 10 10 Portability (mW.cm3) Figure 1.3. Gas chromatography system and popular sensors of a GC system. (a) Principle of gas chromatography: the ability to separate multiple compounds; (b) state-ofthe-art gas sensors for a gas chromatography system, adapted after evaluating detection limit (ppb), power (mW), and volume (cm3) from the references [24-74], and each point indicates the best performance among that specific sensor group. 8 gas mixtures are vaporized at an elevated temperature of 2200C in a chamber named injector. After that, the vaporized gas mixture enters into a long separation column where the gas mixture becomes separated from each other with the help of a mobile phase (helium carrier gas) and a stationary phase (coated inside the separation column). Then the separated gas is detected by a detector. Often there are two other components connected to a gas chromatography system: one is preconcentrator, and the other is the data processing unit. A preconcentrator amplifies the total mass of a compound to be sent to the detector by absorbing and accumulating compounds over an extended period. In such a case, the accumulated compounds enter into the separation column by thermal desorption process in an elevated temperature. If a preconcentrator is not used, then a liquid/gas sample of a compound is directly injected into the injector. To detect gas in ppb level, a highly sensitive detector is required. Some of these state-of-the-art detectors have been investigated in this dissertation, and their performances were plotted in Figure 1.3(b). In Figure 1.3(b), portability (mW.cm3) in xaxis was calculated from the multiplication of power (mW) and volume (cm3) of each gas sensor [24-74]. Portability is a critical aspect to miniaturize a sensor for application such as in home solution, distributed area network, smart phone, or smart watch-based gas sensor (Figure 1.4). Figure 1.3 (b) represented two different categories of sensors: chemistry dependent (red) and chemistry independent (black) ones. In chemistry dependent gas sensors, a chemically active film is present at the sensing interface, and such sensors chemically interact with a target gas by absorption/adsorption process. These types of gas sensors are metal oxide semiconductor (MOS), surface acoustic wave (SAW), chemiresistor (CR), field-effect transistor (FET), quartz crystal microbalance 9 Back Benzene 3.2 Air Quality Benzene Toluene CO PPB 3.0 PPB PPB 6 4 2 0 Nov 27 Indoor air quality Benzene Distributed area network 6 7 8 9 10 11 AM Smart phone Time Smart watch Figure 1.4. Possible application of a portable gas sensor in home solution, distributed area network, smart phone, and smart watch. (QCM), graphene, and carbon nanotube (CNT). On the other hand, chemistry independent gas sensors utilize gas’s physical properties such as thermal conductivity, ionic emission, or ionization energy. They are photoionization detector (PID), thermal conductivity detector (TCD), flame ionization detector (FID), mass spectrometer (MS), and others. Although chemistry independent gas sensors show robustness against environmental change, they are often equipped with peripheral equipment to operate the detector. Such peripheral equipment often increases power consumption and system complexity. For example, an FID requires one power source to ionize the gas and then another source to measure the ion current. Furthermore, an FID needs additional gas supplies of air and hydrogen. A TCD requires a heater element to elevate the temperature. A detailed description of these sensors and their operation and performance are provided in Chapter 2. Figure 1.3(b) also showed that among chemistry-independent detectors, 10 only PID and MS are capable of detecting at the ppb level, while they require excessive peripheral equipment ultimately being limited in practical miniaturization. 1.4 Limitations of PPB Level Gas Sensors There are three significant limitations of current state-of-the-art ppb level gas sensors, and they are chemistry dependency, high power consumption, and limitation in detectable gas ranges. The first limitation is chemistry dependency. This chemistry dependency causes sensors to suffer from the degradation of signal stability over time. These sensors are, thus, limited in lifetime because of the inevitable interaction of the film with humidity, temperature, or other nonspecific contaminant gases. The purpose of such a thin film is to bind the specific gas molecules to the sensor platform. To increase the ranges of target gases, these type of sensors utilized an array with different binding groups. The second limitation of ppb gas sensors is high power consumption and difficulty in miniaturization. For example, an MS requires a vacuum chamber to ionize the gas. It also requires a diaphragm pump and a turbo-molecular pump to reach a proper vacuum level. These components often consume around 100 watts even at a miniscale (Mini-12, Perdue, 25 kg, 117.1×103 cm3) to support control circuit components associated with high voltage (5kV) spraying, a high voltage RF amplifier, and an ion detector [21]. The third limitation of ppb level gas sensors is a limited detection range. For example, a PID is inherently limited in detection of some gases whose ionization energies 11 are less than a UV lamp power inside a lamp chamber. The ionization power typically available by a lamp is xenon-9.6, helium-10.2, krypton-10.6, and argon-11.7 eV. Although most of the VOCs’ photoionization energies are less than 9-10 eV, there are still quite wide ranges of gases whose photoionization energies are above 10 eV, and thus not detectable. According to the photoionization energy list of different compounds (EPA, [22]), among 326 gases and compounds, xenon, hydrogen, krypton and argon lamps cannot ionize 205, 129, 91, and 43 compounds, respectively. Some of the common compounds those cannot be detected by any PIDs include O2, N2, H2, CO2, SO2, CO, CH4, HF, HCl, F2, and O3, because their photoionization energies are 12.08, 15.58, 15.43, 13.79, 12.34, 14.01, 12.98, 15.77, 12.74, 15.7, and 12.80 eV, respectively. These values are higher than any available UV lamp for a PID. According to the photoionization energy list of different compounds (EPA, [22]), among 326 gases and compounds xenon, hydrogen, krypton, and argon lamp cannot ionize 205, 129, 91, and 43 compounds, respectively. Some of the common compounds that cannot be detected by any PID are: O2, N2, H2, CO2, SO2, CO, CH4, HF, HCl, F2, and O3, because their photoionization energies are 12.08, 15.58, 15.43, 13.79, 12.34, 14.01, 12.98, 15.77, 12.74, 15.7, and 12.80 eV, respectively. These values are higher than any available UV lamp for a PID. 1.5 Solution: Viscosity as a New Sensing Element This dissertation presents a proof-of-concept viscosity-based gas sensor to address the aforementioned problems: chemistry dependency, high power consumption, limited gas ranges, and detection of gases in ppb level. Viscosity distinctively characterizes the 12 collision resistance of a gas while it is in motion in a tube. Such collision resistance is a physical property of a gas and is responsible for a pressure drop when a heterogeneous gas being injected into a background gas. Similarly, a viscosity difference between a carrier/reference gas and a target gas also generates a minute local pressure drop around the group of a target gas. Such pressure drop can be utilized to detect a target gas in a flow of a carrier gas. Note that the viscosities of the majority of VOCs’ are less than 10 μPa.s, while those of inert gases are close to or more than 20 μPa.s. For example, the viscosities of helium, argon, krypton, and xenon are 19.9, 22.7, 25.5, and 23.2 μPa.s, respectively, at 300K. This indicates maximum viscosity difference can be generated when an inert gas will be utilized as a carrier gas, and this differential viscosity can be utilized as an input of a viscosity-based sensor. The developed viscosity-based sensor consumes no power during the signal amplification process. Here a change of viscosity was first converted to another fluidic signal-flow rate. And the fluidic signal was amplified by a fluidic amplifier such as slip mode flow amplifier in a narrow microchannel. These two steps, conversion of analytical property and signal amplification, are fluidic and do not consume any electrical power. Note that other gas sensors consume power in both conversion and amplification procedure. In a TCD, heating is required. In an ionization-based gas sensor, a high voltage source is needed. The developed viscosity-based sensor can detect a wide range of gases because viscosity is a unique property of individual gases. The developed viscosity-based sensor can detect as low as 10s of picogram level because of the inherent ability to generate a localized flow pulse around (Figure 1.5). In principle, a viscosity-based sensor utilizes a 13 Viscous pressure drop (relative) Helium Target No electronics in amplification Designed Fluidic Amplifier: Gas compression and expansion • No chemistry • Applicable to all gases • No power • No high voltage • No ignitor to block by contaminants No peripheral equipment • Easy to miniaturize • Simple Any flow meter Designed amplifier Capillary tubes TConnector Figure 1.5. Viscosity-based gas sensor as solution: some features of a viscosity-based gas sensor. viscosity transient and induces it to a pressure transient and the pressure transient causes a flow pulse generation at a junction of two different diameters microchannel. The magnitude of the pressure transient is directly proportional to differential viscosity and applied flow rate. The magnitude of flow rate depends on the corresponding resistance of the microchannel according to Poiseuille’s equation. Note that the viscosity change can initially be generated by introducing a heterogeneous gas molecule in an existing flow of reference gas. The reason of pressure transient generation is that the heterogeneous gas molecules either create a higher local pressure than reference gas (more viscous than a reference gas) or create a lower local pressure than reference gas (less viscous than a reference gas). Viscosity-change induced flow can be amplified by utilizing the slip mode flow amplification in a narrow microchannel. Gas flows in a narrow microchannel, where 14 a radius is in the order of 10s of micrometer or less, lies in the slip flow regime. In this regime, Knudsen number (Kn=λ/r, λ= mean free path and r= radius of flow tube) falls in the range of 10-3 < Kn < 10-1. 1.6 Improved Quantification: Microbubble Chromatography A microbubble chromatograph is a new type of a postseparation detector in a gas chromatography system where a separated gas can be further distributed in a multiple (10-1000) of bubbles. The distributed bubbles allow an estimation of a specific gas by just measuring its sizes and counts. The working principle of a bubble chromatography is that when a viscosity related pressure pulse reached near the gas-liquid interface, a reverse flow was generated from liquid pressure and influenced the bubble diameter. Here, the gas-liquid interface acted as a T-junction sensor where a certain portion of the inlet gas flows throughout the liquid phase as a mass transfer action, and then the nozzle/orifice of the microfluidic device acted as a narrow microchannel after a Tjunction, as will be mentioned in Chapter 3. 1.7 Organization of Dissertation Chapter 1 gives the introduction of the necessity of gas sensing and setting the goal to improve the research as compared to other existing detectors. Chapter 2 provides a literature review on previous gas sensors by describing their operation principle and corresponding merits and demerits. Chapter 3 introduces the development of the theory of a viscosity-based gas sensing mechanism and corresponding experimental validation. Chapter 4 discusses the experimental results after utilizing the viscosity-based gas sensor 15 in both permanent gas sensing and VOCs sensing. Chapter 5 gives a detail theoretical and experimental investigation on bubble chromatography. Chapter 6 concludes this dissertation with a summary and possible future direction of this research. 1.8 Innovation and Significance This dissertation has made some significant innovations in the field of viscosity effect in gas sensing. The first innovation is viscosity-to-flow generation at the junction of two microchannels with different diameters, and the second innovation is the effect of a narrow microchannel to improve the viscosity-to-flow generation. While developing the concept of the first concept, the effect of change of viscosity at a fluidic junction was explored and observed as a measurement of pressure change, and the effect of gas tube length, diameter, and flow rate on pressure change was also observed. In developing the second concept, at first, effect of slip flow on steady-state flow rate was observed by monitoring the flow amplification ratio from inlet to outlet. Then the effect of a transient flow on output flow rate was explored and experimentally validated. Later in Chapter 6, it was shown for the first time that different gases generate different sized bubbles over time, and this phenomenon was termed as bubble chromatography. This dissertation will have an impact in the field of high sensitivity and low power gas sensors. The developed viscosity-based gas sensor can be integrated with a gas chromatography system in a practical manner to detect a wide range of volatile organic compounds and gases. Also, it can be utilized as a gas contaminant detector in a gas flow pipe. This dissertation will also have an impact in the area microbubble use in gas 16 sensing. The developed principle can be utilized to detect and estimate a specific amount of toxic gas inside a bubble while the bubble is traveling throughout a channel. CHAPTER 2 LITERATURE REVIEW OF MINIATURIZED GAS SENSORS 2.1 Working Principle of a Gas Sensor A gas sensor measures the variation of an analytical property of a target gas based on the interaction between a sensing element and a target gas molecule. Structurally, a gas sensor consists of three essential components: (i) gas inlet, (ii) gas sensing element, and (iii) gas outlet, as shown in Figure 2.1. Usually, a background gas carries a target gas to the location of a sensing element. Then the sensing element can sense the relative change of an analytical property from carrier to target gases before the carrier gas takes away the target gas though outlet. However, for some of the gas sensors such as metal oxide semiconductor sensors (MOS), the sensing element is directly exposed to the environment, and thus eliminates the need for dedicated gas inlet and outlet. Some of the significant analytical properties of a gas are thermal conductivity, ionization, combustion reaction, absorption/adsorption, etc. The effect of these properties can be measured by transforming them into electrical/optical signals. For example, the effect of thermal conductivity can be measured by measuring the resistance of a coil; the effect of ionization can be measured by measuring the corresponding current; the effect of combustion reaction can be measured by measuring the resultant ion current from the combustion; the effect of absorption/adsorption can be measured by monitoring the 18 Sensing element Gas out Resistance/ Current Gas in Figure 2.1. Principle of a gas sensor resistance changes of a coil based on the degree of absorption or adsorption. 2.2 Characteristics of a Microgas Sensor A microgas sensor offers several advantages because of its small size, low cost, and low power consumption. Some of the essential characteristics of a microgas sensor are described below: • Sensitivity: The sensitivity of a gas sensor is defined by a change in the measured signal for a unit change in target gas concentration. • Limit of Detection: Limit of detection or LOD is defined by the least amount of a target gas that can be detected by a gas sensor. Here the amount can be expressed either by weight or its ratio such as parts per million/billion/trillion (ppm/ppb/ppt). Usually, the baseline of a detector signal varies randomly and forms a noise floor due to inconsistent experimental 19 conditions such as temperature, humidity, pressure in a testing room, as well as fluctuations in gas flow rates. Under those random noises, LOD is often empirically decided by 2-3 times of the standard deviation of the baseline noises. • Lifetime: The lifetime of a microgas sensor is an important property related to the degradation of the sensing element. Limited lifetime is a problem of a gas sensor that occurs when the sensing element is exposed to a specific gas molecule with an irreversible binding or interaction. The source of such exposed gas may be other nonspecific gases, oxygen, or vapor molecule under various conditions of humidity and temperature. The sensor loses its signal strengths, drifts signal amounts, and loses its sensitivity over time. If the analytical property depends on chemistry such as sorption, then poisoning of the sensing element is unavoidable, unless there is a controlled environment around the sensing element. • Portability: The portability of a microgas sensor can be defined by a Figure-of-Merit (FOM) value that combines volume, weight, and power consumption. Portability is an essential aspect of a microgas sensor because a field-portable gas sensor can reduce the associated cost of sample carrying from a field location to laboratories. 2.3 Different Types of Gas Sensors Different types of gas sensors can be broadly categorized into two groups (Figure 2.2): (1) sorption-based or chemistry dependent and (2) nonsorption-based or chemistry independent gas sensors. Their performances can be categorized in the following 20 Gas Sensor Chemistry dependent /sorption based Chemistry independent Surface Acoustic Wave (SAW) Thermal Conductivity Detector (TCD) Chemiresistor (CR) Flame Ionization Detector (FID) Metal Oxide Semiconductor (MOS) Photo-ionization Detector (PID) Optical sensor Mass Spectrometer (MS) Figure 2.2. Different types of gas sensors parameters: response time, effect of oxidation and humidity, range of gas type, power consumption, peripheral equipment or bulkiness, detection limit, and gas type identification capability. In terms of these performance parameters, gas sensors can be categorized into the following types. 2.3.1 Chemistry Dependent/Sorption Based Chemistry dependent gas sensors utilize a target gas responsive thin film on top of sensing substrate and then measure a corresponding parameter change such as resonant frequency, resistance, current, or optical wavelength. They are described below. 21 2.3.1.1 Surface Acoustic Wave (SAW) Sensors Surface Acoustic Wave (SAW) sensor work based on a mass sorption property to a polymer thin film. The sorption amount is detected by measuring a frequency shift or output acoustic wave frequency. Saw sensors consist of two sets of interdigitated electrodes deposited and patterned on top of a thin piezoelectric substrate, such as a quartz-crystal wafer [23]. One of the electrodes is a transmitter, and the other is a receiver. Note that interdigitated electrodes or interdigital transducers (IDT) are fingershaped electrodes where fingers are equally spaced and alternately connected to the positive and negative side of the voltage source, as shown in Figure 2.3. When a sinusoidal voltage is applied on the transmitter IDT, the resultant sinusoidal electric field excites the piezoelectric material and crates a wave-like mechanical deformation (acoustic wave). The wave then propagates along the surface of the piezoelectric substrate. When the surface wave interacts with the receiver IDT, then it is converted to an electrical signal and generates a current through the electrodes. The resultant currents can be measured by an external circuit. Since the amplitude of the surface wave Polymer Piezoelectric substrate I2 V1 Transmitter Receiver Figure 2.3. Principle of a Surface Acoustic Wave (SAW) sensor. 22 attenuated gradually through the substrate of the piezoelectric material, such attenuation can be affected by the interaction (e.g., mass density) between the propagating wave and a sorption film that was previously deposited on top of surface layer. The deposited surface film can be properly tuned to detect a specific target providing gas selectivity; thus, it can be utilized as an electronic nose for gas type identification. However, the sensitivity of a SAW sensor is significantly low because it is subject to the variations of temperature, humidity, as well as film nonuniformity. SAW sensors suffer from cross responsivity for a nonspecific target and a relatively slow response/recovery time of a few seconds [24]. However, SAW sensors offer excellent detection limits as low as 1 ppb. Chevailler et al. [25] developed a sensitive coating for SAW sensor array, which demonstrated a detection of 2, 4-Dinitrotoluene at 1 ppb. Nimal et al. [26] demonstrated a SAW sensor for chemical warfare agent detection, which achieved 3 ppb level of sarin gas detection at a power consumption of 300 mW with a dimension of 92 cm3. Another SAW-based detector for hazardous chemical agents (HAZMATCAD, Tradeways Ltd. [27]) detected as low as 30 ppb of nerve gas agents at 1354 mW of power with a device size of 712 cm3. To make a SAW sensor compatible with a gas chromatography system, the sensor response time needs to be reduced as much as possible. Faster response time can be achieved by adding a pulsed heating/cooling element [28]. Staples et al. introduced an additional thermoelectric heating and cooling element to control the adsorption/desorption process. This process was then commercially developed as a portable GC-SAW device [29]. Their portable GC-SAW (EST Model 4650) is 21390 cm3 in size and can detect 10-100 ppb VOC sample at 112 W of power. 23 2.3.1.2 Chemiresistor Sensors Chemiresistor is a small sensing element whose resistance changes based on the interaction of gas vapor. It consists of a polymer thin film deposited on top of two sets of electrodes, as shown in Figure 2.4. The physical property of the polymer, such as mass or electrical conductivity, changes because of the gas vapor exposure. When a gas vapor comes in contact with a polymer film, the following changes might happen: (i) gas vapor might affect the charge transfer between polymer and electrode; (ii) vapor might initiate an oxidation and reduction process by which net charge carrier of the polymer might increase or decrease; (iii) vapor might interact with the mobile charge carrier to lowering down the mobility of the carrier; (iv) vapor could modulate the mobility of the counteranions within the polymer; (v) or vapor could change the interchain hopping inside the polymer and thus change the conductivity of the polymer [30]. The vapor concentration to resistance change is usually linear at low concentration; however, resistance change reaches saturation at higher concentration. There exist different types of chemiresistor sensors based on the material used with a polymer. The materials include carbon black composite [31], organic material coated metal nanoparticle [32], and Polymer Gold electrode Figure 2.4. Principle of a chemiresistor sensor. 24 polypyrrole or polyaniline (PAni) and carbon nanotube. The significant advantages of chemiresistor are the ability to perform at room temperature, detection capability of a wide range of VOCs, low cost of microfabrication, low power consumption, and relatively high selectivity. The sorption process is reversible, and a heating process can do vapor desorption. Multiple gas identification is possible when an array of chemiresistors is utilized. Because, in that case, each polymer interacts with vapor mixture differently and provides an n-number of known resistance values for n-number of unknown gases [33]. Based on the a prior characteristic response of each array, vapor identification from a mixture of gases becomes possible through pattern-recognition algorithms. Mu et al. [34] developed a 4×2 chemiresistor sensor array (2.0×1.2 cm2) with onboard noise cancellation circuitry. Such a sensor array on a PCB board was 56 cm3 in size and could detect as low as 4.2 ng of TCE (trichloroethylene) with a power consumption of 66μW [19]. Their group enhanced the detection limit to ppt level by utilizing a micropreconcentrator and then a microfocuser. Davis et al. [35] showed that without a preconcentrator, the detection limit of a chemiresistor was 13.5 ppm for mxylene, and LOD improved from ppm to 61.8 ppb when a micropreconcentrator was added. Harun et al. [36] developed a large-scale chemiresistor array (300×4) with a dimension of 1.2×0.85 cm2, and such an array achieved an LOD of 150 ppm for ethanol detection. Due to the cost-effectiveness and simplicity of the microfabrication process, chemiresistor arrays have been commercially utilized to make portable VOC detectors by Sandia national lab [37] and Adsistor Technology [38]. 25 2.3.1.3 Metal Oxide Semiconductor (MOS) Sensors Metal Oxide Semiconductor (MOS) sensors consist of a metal oxide layer on top of a heating coil. The operation principle of a MOS sensor is a simple mechanism of chemisorption of oxygen near the surface of the metal oxide. Environmental oxygen is adsorbed near the surface of a semiconductor by extracting an electron from the conduction band of the semiconductor (Figure 2.5). The basic reactions are given below [30]: 1 − }, 𝑚𝑚𝑂𝑂2 + {𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠} + 𝑒𝑒 − → {𝑂𝑂𝑚𝑚 2 −} 𝑋𝑋 + {𝑂𝑂𝑚𝑚 → {𝑋𝑋𝑂𝑂𝑚𝑚 } + 𝑒𝑒 − , {𝑋𝑋𝑂𝑂𝑚𝑚 } → 𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓ℎ𝑒𝑒𝑒𝑒 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟; where {} indicates that they are bound to the surface sites, and X is the unknown vapor. Note that under equilibrium condition, oxygen always presents near the surface of a metal oxide sensor. The first two equations indicate that with an increase of vapor X there will Electrode O2 Metal oxide e- e- e- eHeater Silicon Supply voltage Current measurement Figure 2.5. Working principle of a Metal Oxide Semiconductor (MOS) sensor. 26 be a decrease in Om- ion, because Om- will be converted to XOm, and also there will be an increase of the electron charge carrier. The delocalization of the electron from the bulk trapping of oxygen ion near the surface introduces the formation of holes or carrier depletion region in bulk and negative charges near the surface. Such accumulation of negative charges induces band bending of the semiconductor. Depending on the type of semiconductor, the number of charge carriers may be increased or decreased. Usually, MOS sensors need to be heated between 100 to 5000C. The nature of the oxygen ion, Omwhether it is molecular (O2-) or atomic (O-) ion depends on the temperature. Usually O2exists below 1500C and O- exists above 1500C [39]. However, at temperatures below 1000C water vapor reacts with the oxide and prevents the reaction with oxygen, where below 2000C, the rate of reaction is prolonged. The ideal condition is 300-5500C. Such high-temperature requirement increases the power consumption as high as 800 mW [30]. Also in order to identify gas, MOS sensors need to be arranged in an array coupled with a pattern recognition system. For example, FOX 4000 electronic nose (Alpha MOS, France) used 18 high power MOS sensors and eventually exceeded more than 1-2 W [40]. However, power consumption can be limited to 10 mW by introducing SOI (silicon on insulator) CMOS microfabrication technology [41]. Such technology also enables lower heat loss by allowing a suspended planar structure. The advantages of MOS sensors are low power consumption and low cost of mass production. However, the type of gases it can detect is very much limited, and the sensitivity greatly depends on temperature. To couple a MOS sensor with a gas chromatography system, both response and recovery time need to be low. For atom/ion coupling, both response and recovery should be lower than the VOC elution rate of a chromatographic column. Sanchez et al. 27 [42] successfully integrated a commercial MOS sensor [43] with a gas chromatography system. This commercial MOS sensor had an area of 4 mm2, a response time of 30 seconds, and their integrated system detected 10 ppm for C2H5OH with a power consumption of 40 mW. For better operation, MOS sensors need to be refreshed (desorption) periodically up to 5500C. Alphasense [44] commercially developed a MOS sensor, and the sensor was as small as 1.2 cm2 and consumed not more than 530 mW of power. 2.3.1.4 Optical Sensors Optical sensors consist of a light-sensitive chemically active film, a light source, and a photodetector. When a vapor interacts with the film, the transmitted or reflected wavelength of the incident light changes. Such a vapor dependent wavelength shift can be utilized to detect and identify a target gas. Reddy et al. [45] developed a Fabry-Perot (FP) sensor based on the principle wavelength dependence on vapor sensitive polymer. The sensor was 8×8 mm2 in size and can detect 1.7 ppm. However, such an FP sensor required one 532 diode laser or two spectrometers (Ocean optics 2000), consumed 2390 cm3, 1400 cm3 of spaces, and 20W, 2.2 W of powers, respectively [46]. Scholten et al. [47] developed WGM (whispering gallery mode) based optical ring resonator. The resonator had a size of 0.4 cm3 and a detection limit of 500 ppb of m-xylene. Notably, they were able to reduce the power consumption by utilizing a low power commercial laser (Phillips 1550nm laser) and a photodetector (Newport 2300). Such lasers and photodetectors consumed 1.27 cm3, 204 cm3 space, and 30 mW, 9 mW, respectively. 28 2.3.2 Chemistry Independent 2.3.2.1 Thermal Conductivity Detectors (TCDs) Thermal conductivity detectors (TCDs) are the most versatile gas chromatography-based detector because of its ability to detect a wide range of gases in a nondestructive manner [48]. In a TCD a resistor coil is heated by a current and cooled by a gas flow passing by (Figure 2.6). The temperature rise in the coil changes the resistance of the coil according to the thermal coefficient of resistance (TCR). Here heat dissipation of the resistor will depend on the thermal conductivity of the passing-by gas and its composition with a carrier gas. The composition changes from a carrier gas to targetcarrier mixture gas when a target gas passes through a detector. Heat transfer in a TCD depends on the thermal conductivity of the gas according to the following equation: 𝑄𝑄 = 𝛼𝛼𝛼𝛼(𝑇𝑇𝑐𝑐 − 𝑇𝑇𝑠𝑠 ) where Q, α, K, Tc, and Ts are heat transfer amount, proportional constant, thermal conductivity, temperature of the coil, and surrounding temperature, respectively. Total heat transfer depends on some other factors such as forced convection of the gas flow, thermal conductivity of electrical connectors to the coil, and thermal radiation [49]. Forced convection related to the mass transport by the gas and contributed to around 25% of the overall heat transfer. Such a heat transfer depends on the volumetric velocity, v, and heat capacity, Cp, of the gas as follows: 2 3 𝑄𝑄2 ≈ √𝑣𝑣 𝐶𝐶𝑝𝑝 . Therefore, TCD output signals greatly depend on the gas flow rate. To minimize the convective heat transfer, flow rate needs to be low, so that the Peclet number (Pe) 29 R2 R1 Wheatstone bridge circuit DC voltage R3 Heater Gas flow out RTCD Gas flow in Output voltage Figure 2.6. Principle of a Thermal Conductivity Detector (TCD). becomes lower than 0.1 [50], where the Peclet number is defined by: 𝑃𝑃𝑃𝑃 = 𝜌𝜌𝐶𝐶𝑝𝑝 𝑣𝑣𝑣𝑣 𝐾𝐾 where ρ, L are gas density and characteristic length of the flow path, respectively. Thermal radiation depends on the surface area of the heating coil, and it is not more than 4% of the total heat transfer [49]. Since the sensitivity of a TCD depends on temperature and flow rate, the limit of detection is very much limited up to ppm or sub-ppm. Narayan et al. [51] [52] developed a TCD sensor with an area of 5 cm2, a detection limit of 200 ppm, and power consumption of 13 mW. Where Kaanta et al. [53] improved the detection limit to 260 ppb [54]. In an ideal case, the TCD coil is heated up to a specific temperature by flowing a 30 constant current through it; however, there is a chance that target gas might raise the temperature of the coil and burn it up. To avoid this phenomenon, TCD coil is maintained to a constant average temperature then any target gas-related temperature fluctuation is translated to a current change [55]. Such a constant temperature of a detector also provides temperature stability and eliminates environmental temperature fluctuation [55], [56], [57]. This additional heating requires lots of power (~60 W) and space (171.45 cm2) [Model TCD3, Valco instrument co. inc.]. 2.3.2.2 Flame Ionization Detectors (FIDs) Flame Ionization Detectors are popular detector for benchtop gas chromatography system because of lower detection limit (10-100 femtogram). In FIDs, a target gas undergoes a combustion process, and resultant ionization current gives the detection signal. Here during the combustion process, a compound is burned by oxygen and hydrogen flames, i.e., an oxyhydrogen flame. A FID is limited to carbon-based gases or organic compounds. This is because, during combustion, a carbon-based compound releases high oxidation energy while producing carbon dioxide or carbon monoxide (Figure 2.7). Such oxidation energy is related to an ionization coefficient of a compound and thus defines the sensitivity of FID. Experimental results showed that a pure oxyhydrogen flame contains very few ions, as low as 107 ions/cm3. However, a combustion reaction increases the ionization current by increasing the ion levels up to 1011 ions/cm3 [49]. A typical oxyhydrogen flame generates H, OH, and O radicals in the following way [58]: 𝐻𝐻2 + 2𝑂𝑂2 => 2𝑂𝑂 + 2𝑂𝑂𝑂𝑂 31 Oxyhydrogen flame Electrode Current + + + Target gas Oxyhydrogen inlet Figure 2.7. Working principle of a Flame Ionization Detector (FID). 𝐻𝐻2 + 𝑂𝑂 => 𝐻𝐻 + 𝑂𝑂𝐻𝐻 𝐻𝐻2 + 𝑂𝑂𝑂𝑂 => 𝐻𝐻2 𝑂𝑂 + 𝐻𝐻 Now when a hydrocarbon comes close to an oxyhydrogen flame, it forms hydrocarbon radicals by pyrolysis, and then the hydrocarbon radicals react with the oxygen radicals to form detectable ions such as e- or OH- in the following way: 𝐶𝐶𝐶𝐶 + 𝑂𝑂 => 𝐶𝐶𝐶𝐶𝐶𝐶+ + 𝑒𝑒 − (𝑂𝑂𝑂𝑂 − ) The detectable ions are first accelerated by an anode-cathode electrode pair located top and bottom of the flame, respectively, and then measured by an external circuit. Although a benchtop FID can detect a compound in the femtogram level, it requires more than 10 32 cm3 of space, more than 1 W of power, and additional H2 and Air-gas cylinders for an oxyhydrogen flame. Power consumption reduces when the pyrolysis chamber gets smaller because pyrolysis will then require a micro burner and minimum oxyhydrogen flame. In this way, Zimmermann et al. [58] developed a μFID, and the sensing chamber consumed only a volume of of 25 cm3 at 6W of power consumption. Their sensor successfully detected 104 ppb of gas. 2.3.2.3 Photoionization Detectors (PIDs) In photoionization detectors high energetic photons ionize a VOC inside a closed chamber and corresponding photo ionization current is measured. Usually, the photon energy is in the range of 10eV, and it is good enough to ionize most of the VOCs except some permanent gases such as O2, N2, CO2, SO2, CO, CH4, HF, HCl, F2, SF6, and O3. This is because the ionization energies of these permanent gases are higher than 10eV. A PID consists of a lamp chamber or discharge department and an ionization chamber or detection department, as shown in Figure 2.8. An optically transparent window separates the two chambers, and since standard glass material cannot transmit low wavelength photon. As a glass material, different alkali or alkaline crystals such as salt crystals are used as a window material. Lamp chamber contains certain gases like He, H2, Ar, Kr, or Xe at a lower pressure of 0.1 or 1 torr [49], and high energy photons are generated inside the lamp when a voltage discharges the gas. Note that either dc or ac voltage can be utilized to create a glow discharge; however, ac voltage requires much lower power consumption (fraction of a watt) as compared to that of dc voltage (600-1500 volts and several watts) [59]. A VOC enters the detection chamber after passing through the 33 Membrane filter VOC Current Detection chamber Discharge chamber Cathode Anode Photon Discharge electrode Lamp gas (Ar/Kr/Xe/) Figure 2.8. Principle of Photoionization detector (PID). membrane filter. Then the VOC ionizes with the help of a high energy photon from the lamp chamber. Such an ionization process generates some electrons, and the electron flow generates a current. This current flows through a cathode-anode system coming out from the detection chamber. RAE Systems by Honeywell developed a commercial palmheld PID (Mini RAE3000). This PID consumes 1240 cm3 space and less than a few watt power with a detection limit of 100 ppb. Where Alphaphse [44] developed a commercial miniature (25 cm3) PID, such a PID could detect as low as 100 ppb with only 300 mW power consumption. Lewis et al. [60] used a PID (Model PID-AH) from Alphaphase and detected 100 ppb of benzene. Two crucial advantages of a PID are that a PID can be miniaturized and also can be coupled to a gas chromatography system. Akbar et al. [61] developed an integrated system including μPID, μSC (separation column), and other accessories, where the μPID was approximately 1 cm3 and detected 500 ppb at a power of 34 1.4 mW. They applied 550 V (@2.5μA) across a pair of microelectrodes separated by 20 μm and used a constant helium flow (1mL/min) to discharge helium plasma. 2.3.2.4 Mass Spectrometers (MSs) In a mass spectrometer, a high energy beam of an electron collides into a target molecule and knocks off an electron from the target molecule and turns the molecule into a positive ion. If the molecular ion is unstable, then the collision generates all possible fragments of lighter ions. Mass spectrometer consists of six different components: electron source, ionization chamber, ion optics, mass separator, energy filter, and detector. At first, high energy electron plasma (50-100eV) is generated in a chamber by creating a spark discharge in a noble gas medium. Then the electron travels to an ionization chamber where it ionizes a target gas into fragments of ions. After a target gas undergoes electron impact ionization, the ions then move to ion extraction optics. Such optics consists of electrostatic lenses and is usually developed by three pair of electrodes to extract the ions at first, then to focus, and then to parallelize the ion beam. Then the ions pass through a mass separator. Here the separator is an ion channel consisting of different electrodes on one side and comb structure on the other side of the channel. The frequency of the applied signal in these electrodes generates a time-varying electric field. This frequency can be correlated to an ion with specific mass and thus controls the mass separation according to the following equation [62]: 𝑓𝑓 = 2𝐸𝐸𝑘𝑘 𝑑𝑑2 𝑚𝑚 𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 where f, Ek, d, and mmolar are applied frequency, kinetic energy, the distance between 35 electrodes, and molar mass, respectively. Then ions whose kinetic energies are outside the desired range pass through the energy filter and eventually neutralizes at the detector. Since all of these components are bulky and associates a large amount of power, MS is very difficult to miniaturize. For miniature and portable GC, it requires an integration of all components into a small package so that it can run with battery power. Here the common components include a miniature vacuum, a mass analyzer, atmospheric pressure interface, ion source, and ASIC circuitry. Li et al. [21] from Perdue developed a benchtop miniature MS named as Mini12. Their mini MS is 117.1×103 cm3 (19.6×22.1×16.5 inch3) in size and consumes less than 100 W power with a detection limit of 10 ng/mL. Yang et al. [63] developed even smaller sized palm portable MS (PPMS) with a net dimension of 1546 cm3 (8.2×7.7×24.5 cm3) and a net weight of 1.48 kg. Such MS detected as low as 5 ppm of toluene at a power consumption of only 5 W. 2.4 Summary of Microgas Sensors Some major parameters of different types of gas sensors are summarized in Table 2.1. The table was categorized based on characteristic properties such as portability in terms of power and space consumption, detection limit, sensor lifetime, detectable target gas range, signal drift stability, and required peripheral equipment. In summary, chemistry dependent gas sensors require much less power and space; however, the range of detected gas is very much limited to a chemical reaction. Because of the instability inevitably coming from chemical reaction, their lifetime is limited. On the other hand, chemistry independent gas sensors are robust against signal drift and cross-contamination and immune to the variations in humidity and temperature. They show excellent 36 Table 2.1 Summary of different types of gas sensors Portabilit y Power (mW) Volume (cm3) Detectio n limit Lifetime Target gas range Drift stability Peripher al equipme nt Chemistry independent PID TCD FID High Mediu Mediu m m >85 >2400 >6000 MS Low Chemistry dependent MOS SAW High High CR High Optical High >5000 >530 >300 >0.066 >39 >5.2 >0.1 >205 >1547 >25 >92 >56 >205 100 260 104 1 1000 1 1000 500 Limited Ionizatio n energy depende nt Yes Long Univers al Long Carbon based Long Univers al Limited Polymer dependn et Limited Polymer dependn et Limited Polymer dependn et Limited Polymer dependn et Yes Yes Yes No No No No Gas, Power Heater, power Gas, Power Vacuum , electro n source, power Heater, power Heater, power None Laser, photodetector, power detection limit in the range of femtogram or ppb to ppt. However, they are often limited in terms of miniaturization because of the associated bulky power sources or peripheral equipment. CHAPTER 3 VISCOSITY DETECTION: THEORY 3.1 Gas Viscosity: Unique Physical Property Gas viscosity can be an excellent option to develop a gas sensor because gas viscosity is uniquely correlated to gas types, and its measurement can be easily performed without utilizing sorption-based chemistry or other bulky systems. According to Sutherland et al. [64], the viscosity of the gas, depending on its mass (m) and radius (a), can be expressed in the following way 𝜂𝜂 = 0.064√𝑐𝑐𝑐𝑐𝑐𝑐 , which turns out to be 𝑚𝑚2 𝑓𝑓(1⁄2𝑎𝑎) (2𝑎𝑎)2 1+ 𝑐𝑐𝑐𝑐 sufficiently differentiating among gas types. For example, it was shown that different gases of H2, N2, O2, CO, and CO2 exhibit molecular masses of 2, 28, 32, 28, 44 g/mole, and relative radius parameters, (2𝑎𝑎)2 of 127, 228, 206, 239, and 239 compared to hydrogen, which in combination led to respectively unique viscosities of 8.76, 9.82, 14.80, 17.20, 19.00, 135, and 92×10-6 μPa.s, respectively. When a gas travels through a tube or channel, it maintains a net direction of motion but with different magnitude at different locations. The centerline gas stream has the maximum amplitude while the velocity magnitude of gas layers reduced gradually from center towards the wall of a tube and along perpendicular direction of motion. Thus, when a layer of gas molecules travels with a velocity of vr along a channel length, it applies a shear force on a neighboring 38 layer that has different vr. This shear force is proportional to the velocity gradient among different layers of gas, and the proportionality constant is called as a coefficient of viscosity η. This relationship can be written in the following form: 𝑃𝑃𝑠𝑠ℎ𝑒𝑒𝑒𝑒𝑒𝑒 = 𝜂𝜂 𝑑𝑑𝑑𝑑𝑟𝑟 𝑑𝑑𝑑𝑑 A gas flow always maintains a balance between the degree of such shear force and applied pressure. The relationship between the applied pressure the shear force is often treated by the Poiseuille’s equation (3.1). When a gas experiences a sudden change of flow rate because of change in capillary diameter, it becomes possible to detect a single gas that uniquely depends on viscosity by manipulating capillary tube designs. The Poiseuille’s equation is expressed as below: 𝑄𝑄 = 𝜋𝜋𝑟𝑟 4 ∆𝑃𝑃 8𝜂𝜂𝜂𝜂 ………………………….…………..(3.1) where Q, η, r, l, and ΔP are gas flow rate (m3/s), gas viscosity (Pa.s), a radius of gas flowing tube (m), length of the tube (m), and applied pressure across the tube (Pa), respectively. The principle lies on the fact that when a target gas is present in a reference gas flow, it either increases or decreases the flow rate at the junction of two or more different diameter capillary tubes, depending on the comparative viscosity (either higher (ηtarget > ηref) or lower (ηtarget < ηref)). 3.2 Viscosity-to-Pressure Conversion Injection of gas molecules with different viscosities into a reference gas stream would induce instant pressure drop around because it would provide different atomic collision rates and the corresponding pressure, as follows: 39 ∆𝑃𝑃 = 𝜂𝜂𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 − 𝜂𝜂𝑟𝑟𝑟𝑟𝑟𝑟 𝑑𝑑𝑑𝑑𝑟𝑟 𝑑𝑑𝑑𝑑𝑟𝑟 = ∆𝜂𝜂 𝑑𝑑𝑑𝑑 𝑑𝑑𝑑𝑑 The injected gas molecules stay together and travel as a group along the channel carried by the laminar flow of a reference gas [65]. In a laminar flow, both velocity and concentration of gas remain constant over the channel length; therefore, the group of new gas molecules maintains its characteristic shear force and maintains a constant differential pressure from the reference gas to themselves. Thus, the induced transient pressure drop, traveling with gas molecules along a microchannel, could be measured by a pressure sensor when it passes by. However, it is important to note that the transient pressure, which causes localized pull-in of the gas flow in all directions, would not result in a measurable flow due to its short residence time across a flow meter and symmetry within a microchannel without diameter variations. In order to validate this principle, a target gas of pentane was injected into a reference flow of helium through a long glass capillary microchannel that was connected to both a pressure transducer (Validyne DP-45, range: 0.5-40 torr with 0.2 torr resolution) and a mass flow meter (Omega FMA 1601A), as shown in Figure 3.1. Note that the utilized pressure transducer operated in a capacitive mode, only being sensitive to localized pressure changes instead of other parameters (e.g., heat loss), and thus clearly precluding any interference from a thermal operation such as in a thermal conductivity detector (TCD). The lengths and diameters of the first and the second capillaries were 500 cm/ 250 μm and 10 cm/ 50 μm. The pressure transducer was connected at four different locations (Figure 3.1): (1) at 500 cm away from the inlet of the wide microchannel, (2) at the junction of wide and narrow microchannels, (3) at 5 cm away 40 90 cm × 250 μm 500 cm × 250 μm 134.51 kPa Pressure (Torr) 1 1 δqu 5 cm × 50 μm 5 cm × 50 μm 3 4 85.86 kPa 2 δqd 70 s Time - 609.18 Pa 2 85 s - 555.86 Pa 3 4 Time 90 s Time - 706.49 Pa 95 s Time - 546.53 Pa Figure 3.1. Experimental results of viscosity-to-pressure conversion. from the junction, and (4) at the end of the second microchannel. The reference gas (helium) was flown at a flow rate of 2.5 ml/min under pressure of 134.51 kPa, and the injected gas (pentane) amount was 0.1 μL. Experimental results showed (1) that the change in viscosities from helium reference gas (19.9 μPa.s) to pentane target gas (7.2 μPa.s) generated an average of 604.51 Pa pressure at a given microchannel, (2) that the differential pressure traveled through microchannels, (3) that the magnitude of the pressure pulse was maintained, and (4) that the flow rate was measured as zero or below 41 the detection limit, confirming the principle, as shown in Figure 3.1. The reason of pressure reading in pressure transducer is that when viscosity related pressure vacuum reached near the connecting point, some portion of the trapped helium traveled toward the mainstream flow and caused reading in pressure transducer. 3.3 Pressure-to-Flow Generation The localized pull-in flow, induced by pressure transient, becomes asymmetric at a junction where two microchannels in different diameters meet, and resultantly produces flow rate pulse, as shown in Figure 3.2. When a pressure transient (δP) reaches a junction of two microchannels with different diameters, it produces asymmetric flow due to the resistance differences between upstream (Ru) and downstream (Rd) channels. The magnitude of flow generation will mostly depend on the dominant resistance of a narrow microchannel. Figure 3.2 showed a junction where the flow area suddenly got reduced, and this type of junction is called a contraction junction [66] [67]. In this type of junction, gas rarefied immediately when it enters into the narrow microchannel [68]. And when a less viscous gas undergoes instant rarefaction entering into a narrower microchannel, it increases the velocity and flow rate more than the reference gas at the junction. The net flow generation can be written by inserting flow resistance in Poiseuille’s equation (𝑄𝑄 = 𝜋𝜋𝑟𝑟 4 ∆𝑃𝑃 8𝜂𝜂𝜂𝜂 ), as below. 𝛿𝛿𝛿𝛿𝑖𝑖 ≈ 𝑄𝑄𝑖𝑖,𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝑤𝑤ℎ𝑒𝑒𝑒𝑒𝑒𝑒 𝑘𝑘 = 𝑄𝑄 𝑖𝑖,𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏 = 𝜋𝜋𝑟𝑟𝑛𝑛4𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛 × 𝑘𝑘𝑘𝑘𝑘𝑘 8𝜂𝜂𝑙𝑙𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛 𝜋𝜋𝑟𝑟4 𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛 ∆𝑃𝑃𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛 8𝜂𝜂𝑙𝑙𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛 𝜋𝜋𝑟𝑟4 ∆𝑃𝑃 𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤 𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤 8𝜂𝜂𝑙𝑙𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤 𝑟𝑟𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛 4 ≈ 𝑟𝑟𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤 × 𝑙𝑙 𝑙𝑙𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤 𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛 42 (a) Theory (b) Results 90 cm × 250 μm 500 cm × 250 μm 1 134.51 kPa Pressure δqd No flow δqu Location-3 δqd No flow δqi Location-2 Flow rate (SCCM) δqu 85.86 kPa 1 Location Location-1 10 cm × 50 μm 2 3 4 0.047 sccm 2 85s 0.0455 sccm Time Time 3 85s 0.0795 sccm Time 4 85s Time flow Figure 3.2. Conceptual explanation of pressure-to flow generation by contraction. Note that the flow magnitude mainly depends on the geometric ratio between the first (wide) and the second (narrow) microchannels. This is particularly beneficial for the wide-to-narrow junction (contraction junction) to amplify signals, which becomes our focus. It is also notable that the induced gas flow pulse in the wide-to-narrow configuration can be immediately sensed at the end of a microchannel. This is because the gases in the narrow microchannel, as a lump, act as an incompressible fluid in a narrow microchannel where the corresponding Reynolds number (1.1) and Knudsen number (0.004) are sufficiently low, respectively, thus satisfying the condition for incompressible gas behavior [69]. This principle was experimentally validated by measuring a flow transient with a mass flow meter (OMEGA, FMA 1601A) in a wide-to-narrow microchannel. The mass flow meter was connected at the end of the microchannel (Figure 3.2: location-4). The connected microchannels hold the identical dimensions in the previous section, and the 43 flow connections remain the same. The measurement showed that the flow pulse was measured at 85 seconds which was the same as soon as the pressure transient arrives at the junction at 85 seconds (Figure. 3.1). Results also showed that irrespective of pressure pulse travels through the narrow microchannel flow was measured at 85 seconds, as shown in Figure 3.2, confirming the principle above. 3.4 Flow Amplification The induced gas flow pulse becomes gradually amplified along the second narrow microchannel, where nonlinear pressure distribution in the slip-condition causes increasing pressure differentials. The nonlinear pressure distribution is attributed to two opposing phenomena, as explained in [68] [70] [71] [72]. The first phenomenon is the gas expansion and flow acceleration toward the end of a microchannel due to pressure gradient along a microchannel. The second phenomenon is the opposing tendency against the pressure gradient due to the flow acceleration. The resultant nonlinearity leads to the highest pressure differential at the end of the microchannel, which amplifies the baseline flow rate most within the narrow channel, as validated by previous literature [73] [74] [75]: 𝑄𝑄𝑜𝑜 𝑃𝑃𝑟𝑟2 − 1 + 8𝐾𝐾𝐾𝐾𝑜𝑜 (𝑃𝑃𝑟𝑟 − 1) 8𝐾𝐾𝐾𝐾𝑜𝑜 = =1+ 2 𝑄𝑄𝑖𝑖 𝑃𝑃𝑟𝑟 − 1 𝑃𝑃𝑟𝑟 + 1 where Qo is the outlet steady-state flow rate, Qi is the inlet steady-state flow rate, Pr=Pi/Po is the ratio of the inlet pressure (Pi), and the outlet pressure (Po), and Kno is the Knudsen number at the outlet. The second term represents the addition of flow in the slip flow regime, indicating flow amplification depending on the nonlinear pressure variations 44 across the channel as well as the Knudsen number. The amplification in the flow rates through the second narrow channel was experimentally verified. To monitor the flow rate magnitude of the travelling flow pulse, a measurement was performed by placing a flow meter at nine locations along a narrow microchannel, respectively, at 0.1, 0.5, 1, 4, 5, 6, 7, 8, and 10-cm away from the junction of the two microchannels. To monitor nonlinearity in wide microchannel, nominal pressures were also observed at every 20 cm in the end 100 cm of the 500 cm long wide (250 μm) microchannel. The microchannel dimensions and flow conditions remained identical to the previous sections. Experimental results also showed that the nominal pressure gradient increased nonlinearly from 0.57 to 4.65 kPa/cm from the beginning (at 0.12 cm) to end (a 10 cm) of narrow microchannel, as shown in Figure. 3.3(a). Figure 3.3 also showed that the pressure gradient remained constant at 3.3 Pa/cm in a wide microchannel section, and thus indicated no such nonlinearity in a wide microchannel. Experimental results showed that the flow rates of the flow pulse nonlinearly increased from 0.0047 to 0.0810 as the pulse travels through the second channel (Figure 3.3). This result confirmed that a flow pulse was being gradually amplified along the microchannel, indicating the end of a channel as the best flow sensing location. Notably, the measured amplification ratio reached approximately 23.01, indicating the efficiency of the proposed sensing mechanism. It is hypothesized that such high amplification was caused by the combined effects of a highly nonlinear pressure ratio, a changing Knudsen number, which needs further investigation. 45 (a) Non-linear pressure distribution Pressure gradient, dP/dx (Pa/cm) 5000 4000 3000 Nonlinear pressure 2000 50 μm dia 1000 0 -1000 -10 -6 0 0 0.2 0.4 5 0.6 0.8 -5 Normalized location, x/L (unitless) 110 (b) Flow amplification 0.5 Q o (x) Q o (x)+ q o (x) 0.45 δqo(x) Qo (sccm) 0.4 Qo(x) 0.35 0.3 δqi Qi 0.25 0 0.2 0.4 0.6 0.8 1 Normalized location, x/L (unitless) Figure 3.3. Flow amplification with nonlinear pressure gradient: (a) Nonlinear pressure distribution across narrow microchannel: pressure gradient increased nonlinearly across 50 μm diameter narrow microchannel; (b) Flow amplification through a narrow microchannel. 46 3.5 Sensor Design The final sensor was developed based on the principles described above, however, with an improved nominal flow rate. Here junction pressure at narrow channel side is much lower than that of a wide channel side. 3.5.1 Improvement of Flow Rate To improve the nominal flow rate an additional low resistance path was connected in parallel with the amplifier microchannel. Such additional microchannel reduces the equivalent resistance of the flow amplifier because most of the flow will then bypasses though the added microchannel (Figure 3.4). To observe this phenomenon, a 0.5 cm long 75 μm diameter capillary tube was connected with the mainstream capillary by a T-connector. Experimental results showed that the addition of such a second microchannel improved the nominal flow rate from 0.08 sccm to 0.6 sccm. Expansion Qi Flow meter Resistance Ri Resistance R1 Added channel Resistance R2 Figure 3.4. Pressure-to-flow generation in T-junction-based expansion. 47 3.5.2 Signal Amplification The output signal can be further amplified when a baseline flow rate is increased by forming a T-branch channel to reduce the equivalent resistance value, as shown in Figure 3.5. In the T-configuration, the flow direction is reversed in the second narrow channel leading to the flow pulse amplification in the reverse direction toward the wide channel, indicating the best sensing location at the junction. Notably, the increased flow rate enables a more rapid detection time. To observe improved amplification by a T-branch configuration, a third channel (75μm diameter) was added to the previous structure with a T-connector (IDEX, p-888, 500 cm × 250 μm (b) Pressure (Torr) 134.51 kPa δqu (a) Flow (sccm) (c) 4 cm × 250 μm 1 9s 5 cm × 50 μm 4cm×250μm 2 4 3 85.86 kPa 0.5cm×75 μm - 4.57 kPa 5 cm × 50 μm 5 9s - 4.76 kPa Time 9s 9s 0.2945 sccm 9s 0.0075 sccm Figure 3.5. Signal amplification in T-junction-based expansion. 9s 0.0026 sccm 48 360μm O.D.). The length and diameter of the third channel were 0.5 cm and 75 μm. To confirm the reverse amplification of a flow pulse in the second channel, a flow meter (OMEGA FMA 1601A) was placed at four different sites (location-1, 2, 3 & 4), as shown in Figure 3.5. Experimental results showed that (1) the change in viscosities from helium reference gas (19.9 μPa.s) to pentane target gas (7.2 μPa.s) generated an average of 4.67 kPa pressure at a given microchannel (Figure 3.5(b)); (2) that the differential pressure produced a flow pulse at a junction (Figure 3.5(a)); and that the produced flow pulse traveled while being nonlinearly amplified by 115.49 times through the narrow microchannels (Figure 3.5(c)); (3) that the flow pulse entering the narrow channel was instantly (at 9 seconds) measured at all locations in the narrow microchannel (10 cm long) (Figure 3.5(c)); and (4) the output signal was enhanced by 7.72 times in pressure (Figure 3.5(b)) and 3.64 times in flow rates (Figure 3.5(c)) in comparison to the twochannel configuration. The detection time was also reduced from 85 seconds (Figure 3.2(b)) to 9 seconds (by 9.44 times) through the T-configuration (Figure 3.5(a)). CHAPTER 4 GAS DETECTION BY A VISCOSITY-BASED SENSOR 4.1 Permanent Gas Detection Permanent gas detection is critical because some of the permanent gases are present in various forms in our environment, such as in natural gas, groundwater, and in thermal runway events during overheating or overcharging of Li-ion batteries [76]. For example, natural gas contains 1.0% nitrogen gas, 0.5% carbon dioxide, 94.7% methane, and 2.0% ethane gas [76]. Groundwater contains different permanent gases such as nitrogen, oxygen, carbon dioxide, methane, and ethane. Where overheated Li-ion batteries release both permanent gas and volatile organic compounds. The released permanent gases are carbon monoxide, carbon dioxide, oxygen, methane, ethylene, and ethane. Because of such importance, five different permanent gases were selected for this study, and they are helium, hydrogen, nitrogen, carbon dioxide, and methane. 4.1.1 Testing Methodology In order to observe the viscosity change from one gas to another and to detect permanent gases, five different gases were used: helium (He), nitrogen (N2), carbon dioxide (CO2), methane (CH4), and hydrogen (H2). The viscosities of these gases are 19.9, 17.9, 15.0, 11.1, and 8.9 μPa.s at 300K. To test the viscosity dependence, one of the 50 gases was set as a carrier (reference) gas, and 0.4 mL of rest of the four gases were injected one by one. The carrier gas was supplied from a cylinder, and computer-operated gas chromatography (Thermo Focus GC) system was utilized to control the gas flow rate. The carrier gas at first passed through an injection system (temperature 2200C), then through a 5 m long 250 μm inner diameter (i.d.) capillary tube (Molex, polymicro), and then through a 3-way T connector (micro tee PEEK). One outlet of the T-connector was connected to a 1 cm long 50 μm i.d. capillary tube, and the other outlet was connected to a series connection of 20 cm long 250 μm i.d. and 3 cm long 24 μm i.d. capillary tube. To monitor the volumetric gas flow rate, a flow meter (Omega, FMA 5400A) was connected at the junction of the series connection of 250 and 25μm i.d. capillary. The following conditions were used while a sample was injected: injector temperature 2200C, carrier gas flow rate 1 ml/min, and GC oven temperature 500C. The chromatogram (change-of-flowrate) was then monitored with the flow meter for a 1 minute. 4.1.2 Gas Chromatograms Experimental results showed that the injected gas generated an increase in flow rate if its viscosity was higher than carrier gas, and it generated a downward chromatogram if its viscosity was lower than carrier gas, as shown in Figure 4.1. Figure 4.1 also showed that for a helium carrier gas, all the rest of the four gases, such as nitrogen, carbon dioxide, methane, and hydrogen, generated downward signals. For a nitrogen carrier, only helium generated an upward signals, and the rest of the three gases generated downward signals. For a carbon dioxide carrier, helium and nitrogen generated upward signals, and the rest of the two gases generated downward signals. For a methane 51 He 19.9 Injected gas N2 17.9 CO2 15.0 H2 8.9 CH4 11.1 0.05 0.09 Carrier: He (19.9) 0.08 0.07 0.06 0.05 Carrier: N2 (17.9) Carrier: CO2 (15.0) Flow rate (sccm) 0.04 0.052 0.13 0.03 0 10 20 30 40 50 60 70 80 0 10 20 30 40 50 60 70 80 0 10 20 30 40 50 60 70 80 0.12 0.11 0.1 0.09 0.08 0.06 0.19 0.07 0.18 0.17 0.16 0.15 0.14 0.13 0.07 0.18 0.12 0.17 Carrier: CH4 (11.1) 0.16 0.15 0.14 0.13 0.12 0.35 0.23 0.11 Carrier: H2 (8.9) 0 10 20 30 40 50 60 70 80 0.3 0.25 0.2 0.15 0 0.1 0 10 20 30 40 50 60 70 80 Figure 4.1. Flow chromatogram of five different permanent gases (He, N2, CO2, CH4, H2); He carrier: downward signal (flow rate change) for less viscous gas than helium; N2 carrier: upward signal for higher viscous He but downward signal for less viscous CO2, CH4, H2; CO2 carrier: upward signal for higher viscous He, N2 but downward signal for less viscous CH4, H2; CH4 carrier: upward signal for higher viscous He, N2, CO2 but downward signal for less viscous H2; H2 carrier: upward signal for higher viscous gas He, N2, CO2, CH4. Chromatogram of five different gases with different carrier gases. 52 carrier, only hydrogen generated a downward signal the rest of the three gases generated upward signals. Finally for a hydrogen carrier, all of the gases generated upward signals. This signal matrix is validating the principle of viscosity dependence on a viscosity-based gas sensor. This matrix indicated that any unknown gas flowing with another gas could be detected by utilizing this sensor. 4.2 VOC Detection Detection of volatile organic compounds (VOCs) is critical as there are wide ranges of toxic VOCs present in both outdoor and indoor environments. A detail list of these VOCs, their impact on our health and sources of them have been described in Chapter 1. Among 187 toxic VOCs, the following compounds were selected for detection because of their higher degree of presence in our everyday life. 4.2.1 Testing Methodology To observe the sensor capability to detect volatile organic compounds (VOCs), 12 different VOCs (volatile organic compounds) were utilized as the testing targets. These compounds were (1) methanol, (2) n-pentane (3) n-hexane, (4) benzene, (5) n-heptane, (6) toluene, (7) n-octane, (8) tetrachloroethylene, (9) chlorobenzene, (10) ethylbenzene, (11) m-xylene, and (12) n-nonane. At first, 50 μL of each of the liquid compounds were collected and mixed into a 2 mL vial. Then a 0.1 μL of mixture volume was injected into the GC injector, and the gas flow rate was monitored for 5 minutes. In order to separate the individual compound from each other with respect to time and space, a 25 cm long separation column (coated stationary phase: OV1) was used as a channel that the gas 53 mixture was flown into. Each signal output was then assigned to each gas type, based on the previously known retention time of each sample. Note that retention time is a time required to reach a sample from an injector to a sensor and is a unique property for each gas. 4.2.2 VOC Chromatograms Experimental results of 12 chromatograms (Figure 4.2) showed that the viscositybased gas sensor could be used as a post-separation detector in a gas chromatography system. Figure 4.2 showed negative peaks for 11 VOCs except for tetrachloroethylene, where tetrachloroethylene generated a positive peak. These results indicated that at an oven temperature of 1000C (1) tetrachloroethylene’s viscosity became higher than that of Viscosity-based sensor 0.03 2 450 0.025 4 400 0.02 350 0.015 1 0.01 3 4 0.005 6 5 7 9 10 11 12 FID (mv) Sensor signal (sccm) Flame ionization detector 500 8 9 300 250 3 200 (1) methanol, (2) n-pentane, (3) n-hexane, (4)150 benzene, (5) n-heptane, (6) toluene, (7) n-octane, (8) 100 tetrachloroethylene, (9) chlorobenzene, (10) ethylbenzene, 50 (11) m-xylene, and (12) n-nonane 0 -0.005 2 -0.01 0 50 5 6 7 10 8 11 12 1 0 100 Time (s) 150 200 0 50 100 150 200 Time (s) Figure 4.2. Chromatogram of 12 different VOCs with (left): viscosity-based sensor, (right) Flame ionization detector; the compounds are: (1) methanol, (2) n-pentane (3) nhexane, (4) benzene, (5) n-heptane, (6) toluene, (7) n-octane, (8) tetrachloroethylene, (9) chlorobenzene, (10) ethylbenzene, (11) m-xylene, and (12) n-nonane. 54 helium, and (2) the rest of the 11 VOC’s viscosities became lower than that of helium. Such opposite polarity signal has an inherent advantage of detecting two closely spaced compounds. The measured retention time of each compound in both viscosity sensor and FID was shown in Table 4.1. Table 4.1 showed that the viscosity sensor’s retention time reasonably matched with that of FID for the first five compounds and gradually increased with the rest of the compounds. This was because the viscosity sensor had a slightly reduced flow rate (0.1 sccm), and this caused a delay in sample traveling for compounds with higher molecular weights. This result of VOC chromatograms showed the feasibility of utilizing a viscositybased gas sensor as a postchromatography detector. The upward and downward signals indicated that the relative viscosity difference between a carrier and VOC could be utilized to detect a VOC. The sensor also showed similar retention time as compared to in Table 4.1 Chromatogram Performance of selected VOCs Compounds 1. Methanol Retention time (s) Viscosity FID [77] 57.0 55.2 Peak height Viscosity FID (mV) (sccm) -0.296 0.064 Peak area Viscosity FID (mV.s) (sccm.s) -0.247 0.0067 2. n-Pentane 63.0 60.8 -1.000 1.000 -1.000 1.000 3. n-Hexane 72.0 69.3 -0.372 0.338 -0.374 0.419 4. Benzene 81.8 78.0 -0.393 0.812 -0.439 1.108 5. n-Heptane 90.6 85.5 -0.302 0.317 -0.330 0.383 6. Toluene 111.2 103.6 -0.235 0.352 -0.278 0.570 7. n-Octane 127.0 116.8 -0.293 0.228 -0.480 0.377 8. Tetrachloroethylene 133.0 122.3 +0.165 0.172 +0.228 0.322 9. Chlorobenzene 154.0 137.5 -0.213 0.509 -0.407 1.130 10. Ethylbenzene 164.4 146.8 -0.235 0.293 -0.370 0.583 11. m-Xylene 171.2 151.8 -0.229 0.277 -0.307 0.562 12. n-Nonane 199.0 176.7 -0.302 0.169 -0.584 0.357 55 FID, indicating that the viscosity sensor does not affect the carrier gas flow rate. 4.3 Limit of Detection (LOD) To observe the limit of detection (LOD) of viscosity-based sensor at first hexane was diluted to different concentrations, and then 0.02 μL diluted hexane from different concentration levels were injected. Low concentration of hexane was prepared by adding hexane and decane (solvent, C10H22) at a different ratio, such as 1:10, 1:102, 1:103, and 1:104, 1:105, and 1:106. To do so at first, 180 μl of decane was poured in six different vials (2mL, Thermo Fisher), and then 20 μL of pure hexane was poured and mixed in the first vial. Then the first vial of hexane and decane was appropriately mixed, and a 20 μL of this mixture (0.1%) was poured and mixed with decane in the second vial. In this way 20 μL from one mixture was mixed consecutively to the next vial. Then 0.02 μL of diluted hexane from each mixture was injected into the GC injector, and the corresponding flow rate was monitored. Experimental results of LOD determination showed that the sensor was able to detect as low as 13.1 pg of hexane, as shown in Figure 4.3. This limit of detection in mass can be converted to ppm/ppb estimation. Such conversion can be made by (i) at first turning the minimum detectable mass into molar volume at standard temperature and pressure (STP), (ii) then by calculating the estimated carrier gas volume from measured carrier gas flow rate and measured time width of chromatogram from minimum detectable mass, (iii) and finally dividing the molar volume of target gas by the estimated carrier volume in that time period. For example, for 13.1 picograms of hexane corresponding ppb would become: 56 1.3μg 130ng 0.085 1.3ng 130pg 13pg Flow rate (sccm) 0.08 0.075 0.0043 sccm 0.07 0.0008 0.0002 0.0014 sccm sccm sccm 0.065 0.06 0.0194 sccm 20 30 40 50 60 70 80 Time (s) Figure 4.3. Limit of detection of viscosity-based gas sensor for hexane gas. 𝑝𝑝 86.18𝑔𝑔/𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 ÷ 3.7𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠×1𝑠𝑠𝑠𝑠𝑠𝑠 60𝑠𝑠𝑠𝑠𝑠𝑠 = 15.3𝑝𝑝𝑝𝑝𝑝𝑝. 4.4 Sensitivity To measure the sensitivity of the viscosity-based sensor, three different compounds were tested at a different concentration level. Then 0.02 μL diluted version of each compound was injected to monitor the sensor signal. Experimental results showed that the limits of detection of viscosity sensor were 13.1, 13.6, and 17.5 picograms for hexane, heptane, and benzene, respectively (Figure 4.4(a)), indicating pg detection 57 Signal amplitude (sccm) (a) 10 0 10 -1 10 -2 10 -3 10 -4 10 Viscosity sensor peak area (sccm.s) (b) 10 Hexane Heptane Benzene 0 10 2 10 4 10 6 10 8 10 5 Injected mass (pg) 0 Heptane Fitted line 10 -1 10 -2 10 -3 10 -4 10 0 10 1 10 2 10 3 10 4 FID peak area (mV.s) Figure 4.4. The sensitivity of viscosity-based detector: (a) sensitivity of hexane, heptane, and benzene; (b) sensitivity of heptane with respect to corresponding FID output. 58 capability of the viscosity sensor. These numbers gave a theoretical detection limit of 15.3, 13.7, and 22.6 ppb of hexane, heptane, and benzene, respectively. Note that in this calculation at first these minimum masses were converted to a molar volume at STP and then divided them by helium carrier gas volume at 3.7 sccm (measured) flow rate and for 1 second. Figure 4.4(a) also showed that detection sensitivity were 0.41, 0.47, and 0.4 sccm/picogram for hexane, heptane, and benzene, respectively. Figure 4.4(b) showed a linear detection limit of 10-6 (pg-μg) for the heptane compound. Here it was considered that the FID signal varied linearly from pictogram to microgram. Note that viscosity sensor data were plotted against the corresponding FID peak area to remove any sample preparation error during series of dilution of heptane. 4.5 Effect of Temperature To observe temperature stability, the viscosity sensor was tested at different temperatures ranging from 40 to 1200C. To avoid any manual injection error from one temperature to another, FID was also connected in parallel with the viscosity sensor. In this case, a flow divider (PEEK TEE connector) was utilized to equally divide the flow between FID and viscosity sensor so that retention time and sample concentration profile does not change. Finally, the sensor’s signal was normalized with the corresponding FID signal. Experimental results showed that the sensor’s signal varied only 5.7% (Figure 4.5) for temperature variation from 40 to 1000C. Figure 4.5 also showed that the sensor’s signal varies in between 0.0248 to 0.0303 sccm with a mean and standard deviation of 0.0279 and 0.0016 sccm. Note that the sensor’s signal was normalized with respect to the FID signal. 59 Sensor signal (sccm) 0.04 0.03 0.02 SD = 5.7% Mean 0.01 0 40 60 80 100 120 Temperature (0C) Figure 4.5. Signal variation with the variation of temperature from 40 to 1200C. 4.6 Effect of Carrier Gas To test the effect of carrier gases, helium and nitrogen gases were utilized. During each time of carrier gas, 0.02 μL of hexane, heptane, benzene, and toluene were injected, and corresponding flow changes were monitored. Experimental results showed that helium provided better sensitivity as a carrier gas since helium has a higher viscosity than nitrogen (Figure 4.6). Results also showed that helium carrier gas generated 1.19, 1.63, 1.27, and 1.28 times peak areas as compared to that of in nitrogen as a carrier gas. This was because, like other inert gases, helium’s viscosity (19.9 μPa.s) induced higher differential viscosity with volatile organic compounds. Note that most of the VOCs’ viscosities are less than 10 μPa.s when they are in complete gaseous form. 60 0.06 He carrier Sensor signal (sccm) N 0.05 2 carrier 0.04 0.03 0.02 0.01 0 1 2 3 4 Pentane Hexane Benzene Toluene Figure 4.6. Comparison between helium and nitrogen in terms of senor performance. CHAPTER 5 MICROBUBBLE CHROMATOGRAPHY 5.1 Microbubble Chromatography Microbubble chromatography is a chromatography technique where a series of gas microbubbles represents a specific target gas and quantifies the amount of gas. Here the variation in bubble diameters with respect to time constitutes the chromatogram signal. Figure 5.1 describes the main concept of the bubble chromatography. First, gas is streamed into a liquid to produce bubbles while the sizes of the bubbles are monitored to generate a chromatogram for gas identification and quantification. It is hypothesized that the proposed bubble-based gas sensor is attached at the end of the chromatographic separation column where a gas mixture is already separated in time and space into discrete groups of individual gases (C5: red, C6: black and C7: blue) before entering the bubble-based gas sensor. Each of the separated gas groups is then streamed into the liquid channel where a carrier gas, e.g., helium (He), is continuously flowing. When only the carrier gas flows into the liquid-flowing microchannel, it forms a train of bubbles in a specific and uniform size. Then, the injection of the previously separated gas targets (C5, C6, and C7) into the carrier gas flow changes the sizes of the corresponding bubbles (Figure 5.1). Such size changes depend on the gas types as well as mixture ratios between the two gases. Thus, by monitoring, measuring, and plotting the variation of bubble sizes 62 C10 Gas concentration change He 100% Separation column Gas concentration Liquid C7 He C10 Each column represents each He/C5: bubble He 90% C5 He C5 80% 70% C12 He:Carrier gas (25/75) 60% He He He 50% 40% 30% C7 20% C6 He 10% 0% C5 10 0 20 C6 30 40 50 60 70 80 90 100 Time Gas Gas mixture Separated in time and space Big He Liquid Small C5 Big He Small C10 Big He Small C12 Big He Size Type • Gas bubble generation Microfluidics • Gas type identification Size change Figure 5.1. Concept of microbubble chromatography. in reference to the background carrier bubble sizes, a gas chromatogram can be established. Here each bubble, determining the sensor resolution, is represented by each column in the figure. Then, by counting and adding all the bubble volumes within the corresponding peak in the chromatogram, one can quantify the total volume of each gas. For example, the injected C5 peak produces the corresponding three gas bubbles (third to fifth bubbles in the plot) that showed the reduction in their sizes compared to 100% helium bubbles. Such a size change in the bubble stream indicates the introduction of a particular gas molecule group other than nitrogen, forming a chromatographic peak. These three bubbles have individually-different sizes because they contain different concentrations of the target gas. If the relationship between the bubble size and the concentration is known, one can infer that the three bubbles contain, e.g., 13, 65, and 25% volumes of C5, respectively (Figure 5.1). 63 5.2 Mechanism of Microbubble Formation Microbubble can be generated in different ways such as by ultrasound or through a microfluidic device. In this dissertation, a microfluidic device was utilized to generate microbubbles. In a microfluidic flow-focusing device (Figure 5.2), a gas stream is flown through the center channel, and liquid stream is flown through two channels where all three channels meet at a point and exit though a much narrower orifice. Under a certain flow condition, a gas stream is periodically cut by the liquid stream and generates microbubbles at a periodic rate. The size of microbubble depends on the gas flow rate, liquid flow rate, and device structure, such as each channel’s length, height, and width. The effect of these parameters on bubble size has been well documented; however, the effect of gas types has not previously been observed. It was reported that the sizes of the produced bubbles were related to the nozzle dimensions, flow rates of both gas and liquid, and liquid viscosity. Garstecki et al. related the bubble sizes of a specific gas (N2) with gas pressure, liquid flow-rate, and device geometry within flow focusing and Tjunction devices [78] [79]. Fuerstman et al. investigated the effects of surfactant t=0 Liquid concentrations on bubble sizes via pressure drop calculation [80], and Cubaud et al. t = 1.5 ms Gas Liquid Figure 5.2. Bubble formation mechanism. t = 1.7 ms 64 studied the effect of solubility on the CO2 bubble sizes [81] [82]. 5.3 Microfluidic Device for Microbubble Formation To generate microbubbles, a cofocusing microfluidic device was adopted [83], and its structure and fabrication procedure are described below. 5.3.1 Structure The structure of the bubble-based gas sensor consists mainly of a gas flow channel, two liquid flow channels, a nozzle, and an outlet channel for bubble flow, as shown in Figure 5.3. The gas flow channel introduces the gas stream between the two liquid flows that are provided by the liquid flow channels. At the intersection of both gas and liquid flows, the nozzle is located where the gas stream is cut into a train of discrete bubbles by continuous liquid flows. These channels consist of hydrophilic walls to maintain a stable flow of liquid phase by wetting the channel walls, and their surface roughness is less than 1nm to avoid any disturbance in laminar flow through the channel. The channels are in 100’s micro-meter ranges resulting in low Reynolds number (3×10-6 for liquid; 9.2 for gas phases), 𝑅𝑅𝑅𝑅 = 𝜌𝜌𝜌𝜌𝜌𝜌 𝜇𝜇 , for both liquid and gas phases to ensure the laminar flow through the channels. The width of the nozzle is designed to be narrower than those of the gas and liquid flow channels in order to reduce the size and increase the frequency of bubbles for higher precision measurement. The outlet channel is the path that the produced bubbles flow for optical measurement. The outlet channel extends the observation range through a meander shape within a compact footprint. 65 (a) Fabrication steps (b) Fabricated microfluidic device SU-8 Silicon wafer Liquid inlet SU-8 spin coating at1500 rpm for 60 s Gas inlet Nozzle SU-8 Silicon wafer Outlet channel Patterning to make a SU-8 mold PDMS pouring SU-8 Silicon wafer PDMS on a SU-8 mold PDMS top layer PDMS channel layer Liquid inlet Gas inlet Bubble PDMS top layer PDMS channel layer Fabricated device O2 plasma Bonding between PDMS top layer and channel layer Outlet Bonding between PDMS top layer And channel layer Figure 5.3. Structure and fabrication. 5.3.2 Fabrication Fabrication was performed by molding individual PDMS layers containing a microchannel and stacking them with oxygen plasma bonding techniques through only one mask step (Figure 5.3(a)). First, SU-8 2050 (MicroChem) polymer layer was utilized to construct a raised structure, a sacrificial channel mold for the fluid channels as well as the nozzle. The SU-8 pre-polymer was spin-coated at 2500rpm for 60s on top of a silicon wafer and subsequently cured at 650C for 3 minutes and 950C for 9 minutes. The resultant thickness of 38μm defined the heights. After being patterned with UV lithography at 350W and 30s, the SU-8 mold was post-baked at 650C for 2 minutes and 950C for 7 66 minutes to accelerate the cross-linking mechanism in SU-8 and developed for 3-5 minutes in a SU-8 developer (MicroChem). On top of the fabricated mold, PDMS (Sylgard 186 Silicone Elastomer Kit), diluted in a curing agent at a 10:1 ratio, was poured and then cured at 650C for 6 hours. The same process was repeated to fabricate the top PDMS layer. The resultant widths of the fabricated gas channel, liquid channels, nozzle, and outlet channel were 200, 150, 40, and 600μm, respectively, and the height of all structures was 38μm. The length of the outlet channel was designed to be relatively shorter (15mm) to reduce the viscous resistance of the outlet channel, according to the Poiseuille equation as 𝑅𝑅 ≅ 𝜇𝜇𝑣𝑣/ℎ4 where μ, L, and h are liquid viscosity, outlet channel length, and outlet channel height, respectively, resulting in the reduction of the required pressure to produce gas bubbles. The reduced pressure, in turn, decreased the bubble frequency within the limit of the given camera speed, the relationship of which was confirmed in previous studies by other research groups [84]. The designed volume fraction of the bubbles (the ratio of bubble to outlet channel volumes) ranged between 0.33~0.37 in this particular study. On the PDMS layers, three I/O holes were drilled by a custom-made bio-puncher with a 20-gauge size (ID=0.6mm, OD=0.91mm), resulting in a final diameter of 600μm. The two PDMS layers were then bonded by exposing each layer under oxygen plasma using Dyne-A-Mite 3D treater and Enercon (120V and 4A for 30s) and stacking them on top of each other under pressure at room temperature. The bonded channel was immediately filled by a DI water to maintain the hydrophilic properties of oxidized PDMS walls. The footprint of the fabricated device was 20×5mm2 (Figure 5.3(a)). 67 5.4 Testing Methodology To monitor bubble diameter, the whole experimental setup was divided into two sections: one was to generate gas microbubbles with a gas chromatography system and a microfluidic device, and second was to perform post processing to measure individual bubble diameter. 5.4.1 Overview of Testing Setup To generate bubbles, the target gases were supplied through a commercial GC instrument (Thermo Focus GC) at a pressure of 6~20kPa, and liquid was provided in control by utilizing a syringe pump (KD Scientific, KDS-210) (Figure 5.4). The gas and liquid phase were selected as helium and glycerol (52% w/w) mixed surfactant (2% w/w) water. The produced bubbles were then video-recorded by an optical camera for detailed analysis. Figure 5.4. Overview of testing setup. 68 5.4.2 Bubble Size Measurement Bubble size was measured by first capturing the image of bubbles then by processing each image through a custom python program. The algorithm was developed by (i) narrowing down the scope of the camera only near nozzle area of 128×168 pixels resulting in only 50 kB/frame, (ii) enabling a reduced computation period of 2 ms/frame by selecting and analyzing only a single bubble at the burst location, as shown in Figure 5.5 per frame. The selection of a single bubble rather than measuring and averaging the diameters of all the bubbles significantly enhanced the performance in terms of memory usages and processing time. More importantly, such a process allowed the algorithm to measure the summation of all black pixels (∑95,100,105 𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏) at only three dedicated vertical locations, as shown in Figure 5.5, resulting in the improvements above mentioned. Through such an algorithm, the decision of a bubble diameter was finalized when the standard deviation of measurement across those locations was measured less than 1.3. Algorithm 128×168 binary image 20 40 Std Line_1,2,3 <1.3 60 80 No 100 120 20 40 60 80 100 120 140 160 Line 1,2,3 Figure 5.5. Image processing algorithm. Yes Size 69 The analysis only at three sites reduced the data matrix size from 168×128 to 3×128. Subsequently, it enhanced the speed of the computation by utilizing the Python-based computation (an .exe file) instead of the previous MATLAB-based program. 5.5 Viscosity Effect During Bubble Formation During bubble generation majority of the gases participate in bubble generation; however, some of the gases diffuse through PDMS polymer and some of the gases through the liquid phase. To experimentally measure gas diffusion amounts through PDMS, two microfluidic devices were manufactured with a glass (low permeability: 65×10-11) and a PDMS (high permeability: 8×10-9 cc/s/cm2/mm/cmHg) substrate cover layers. By comparing the resultant volumes of input gas and output bubbles, the loss amount was measured. In this experiment, gas pressure was kept at a fixed value of 20.5 kPa while the liquid flow rate was varied from 2.93-4.53 μLs-1 at every 0.27 μLs-1 interval. 5.5.1 Viscosity Effect at Gas PDMS Interface Experimental results showed that the gas loss difference through PDMS walls in comparison to glass walls was measured by only 1.07%, indicating negligible diffusion through PDMS. Here the gas loss in PDMS-PDMS device was measured as 41.45% average while PDMS-glass device showed 40.38% average gas loss. In both cases, 42 different sets of data (at different liquid flow rates) were utilized to calculate the average. 70 5.5.2 Viscosity Effect at Gas-Liquid Interface To measure gas diffusion through liquid, another microfluidic device with a narrow sampling liquid outlet was fabricated by soft lithography, as shown in Figure 5.6. The narrow sampling channel was 40 μm wide and located right after the nozzle such that some amounts of the gases that were dissolved into liquid could be collected. Note that the rest of the device dimensions remained as same as in the previous section: width of the gas channel, the liquid channel, the nozzle, and the outlet channel were 200, 150, 40, and 600 μm, respectively, while the height was 110 μm. During bubble generation, an amount of a liquid sample was collected in 0.11-2.24mL for the quantitative estimation of diffused gases in liquid. This estimation was performed by the following two methods: titration method and GC-FID method. • Titration Method: The diffused mass of CO2 in liquid was measured by at first collecting some CO2 diffused liquid during bubble generation and then by analyzing the liquid with a quantitative chemical analysis method [85]. In this chemical analysis method, the Liquid in Mass transfer Gas in Liquid in liquid sampling Liquid out Figure 5.6. Modified device to test gas loss in a liquid phase. 71 unknown mass of slightly acidic CO2 was determined in three steps: (i) neutralizing or capturing all of the CO2 with an excess amount of NaOH base, (ii) measuring the required volume (V1) of HCl acid to neutralize excess NaOH, and (iii) measuring the required volume (V2, V2>V1) of HCl acid to release all of the CO2 and make the solution acidic again. The experimental method is further described below. During bubble generation, CO2 diffused though the liquid phase at the location of gas-liquid interaction. Some portions of such liquid were collected from the sampling channel. Then CO2 was extracted from the liquid by 30 minutes physical agitation of the liquid utilizing a sonicator in de-gas mode. While the liquid was under agitation, the evolved CO2 gas continued to transfer from the liquid container to another NaOH (1M 500mL) container though an airtight glass tube. All of the evolved CO2 became Na2CO3 after reacting with NaOH. The excess amount of NaOH was then neutralized by the controlled dropping of each 0.1M HCl with a 0-20μL pipette. Here, Na2CO3 became NaHCO3 after reacting with HCl. To determine the endpoint of HCl addition or the neutral point of such acid-base reaction, a pH indicator, phenolphthalein, was utilized. Note that phenolphthalein changed its color from pink to colorless as the base solution (pH > 7) turned to neutral (pH = 0). To measure the second endpoint, a second pH indicator methyl orange was used by adding two drops of methyl orange into the solution. With the addition of more HCl, methyl orange became red when all of NaHCO3 converts to CO2, and make the solution acidic. The mass of CO2 was calculated by the following equation: the volume difference of HCl from first to second endpoints (V2-V1) × molarity of HCl × molar mass of CO2. By utilizing this equation, the dissolved CO2 masses into different liquid sample volumes (0.6 to 2.3 mL) were determined. The total percentage of CO2 masses with 72 respect to various input were plotted against different volumes of liquid samples, as shown in Figure 5.7(a). Then the liquid phase mass transfer coefficient was determined by utilizing the following equation: 𝑘𝑘𝑙𝑙 𝑎𝑎 = 𝐶𝐶 ∗ −𝐶𝐶 𝑗𝑗𝑙𝑙 𝑙𝑙𝑙𝑙 𝐶𝐶 ∗−𝐶𝐶0 [86] [87], where, jl, L, C*, C0, and 𝐿𝐿 1 C1 were liquid phase velocity, gas-liquid interface length, gas solubility, and initial and final gas amounts in liquid, respectively. Experimental results of the titration method showed that the portion of dissolved CO2 in four different liquid samples in 0.56, 1.12, 1.68 and 2.24 mL were 47.7, 39.7, 42.4, and 39.7 %, respectively, indicating an average loss of 42.4% of CO2 due to diffusion into the liquid phase (Figure 5.7(a)). Note that, in each sample liquid the measured CO2 masses were 66, 110, 176, and 220 μg, respectively, and thus resulted in densities of 0.59, 0.49, 0.52, and 0.49 mg/mL in each sample liquid. The total influx of CO2 gas was calculated based on gas flow meter data (0.07 sccm) and the molar mass of CO2 (44/22.4 g/L), resulting in 0.13 mg/min. By dividing this gas influx rate with the (a) CO2 diffusion test by Titration 100 80 60 % C5H12 in liquid % CO2 in liquid 80 41.45% 40 20 0 (b) C5 diffusion test by GC-FID 100 60 36.28% 40 20 0.5 1 1.5 Liquid volume (mL) 2 2.5 0 0.2 0.4 0.6 0.8 Liquid injection volume (μL) Figure 5.7. Testing results of gas loss in a liquid phase. 1 73 incoming liquid flow rate (ml/min), the incoming gas mass that underwent liquid interaction was calculated as 1.23 mg/mL. The resultant average mass transfer rate of CO2 gas into liquid was kla = 167.48 s-1, indicating 7.9 times higher mass transfer rate as compared to previous CO2 mass transfer study in the microfluidic domain [87]. • GC-FID Method: The diffused amount of another gas, pentane, was determined by first collecting pentane diffused liquid and then analyzing the liquid by a quantitative GC-FID method. Note that, before applying the quantitative method, a calibration curve of pentane mass versus. FID signal areas was generated first by injecting 12.52-62.6 ng of pentane into GC-FID and by measuring the corresponding FID signal area. To achieve the nanogram mass of pentane for calibration, the pentane was diluted to 0.01% by adding C10H22. Once the calibration curve was obtained, the GC system was disconnected from the FID and then connected with a bubble sensor. Then, pure (100%) pentane of 1 μL was injected through a GC injector with helium as a carrier gas to generate microbubbles. During the injection of pentane, a sample liquid in 650 μL was collected from the sample channel. Then out of the 650 μL sample liquid, a small amount of liquid in 0.2-1.0 μL was injected into GC-FID by disconnecting the GC system from bubble sensor and connecting it again to FID. In this latter case, the unknown mass of pentane in each liquid injection was estimated by measuring the corresponding FID signal area and comparing it to the calibration curve. Diffusion results of the GC-FID-based method showed that 0.2, 0.4, 0.6, 0.8, and 1.0 μL of sample liquids contained estimated amounts of 25.73, 35.75, 42.27, 42.24, and 40.74 ng of pentane, respectively, resulting in the corresponding ratios of 53.43, 37.13, 29.26, 21.93, and 16.92% of pentane diffusions, as shown in Figure 74 5.7(b). The calibration equation was𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑙𝑙𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 = 0.214 × 𝑒𝑒𝑒𝑒𝑒𝑒7.41 × 104 × 𝑚𝑚𝑝𝑝 , here mp was mass of pentane. Pentane mass transfer rate in liquid was calculated as 880.78s-1, which was 5.26 times higher than that of CO2 gas. 5.6 Bubble Volume To evaluate bubble volume, at first bubble volume was modelled based on gas loss mechanism in liquid phase, and then the model was validated with experimental measurement of bubble diameter. 5.6.1 Modelling of Bubble Volume A theoretical model of bubble volume was established by subtracting the losses from the input gas volume, which can be expressed as: 𝑉𝑉𝑏𝑏 = 𝑄𝑄𝑔𝑔 𝑡𝑡𝑏𝑏 − 𝑘𝑘𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑄𝑄𝑔𝑔 𝑡𝑡𝑏𝑏 − 𝑘𝑘𝑙𝑙 𝑄𝑄𝑔𝑔 𝑡𝑡𝑏𝑏 ……………………5.1 where Qg, tb, kpdms, kl are gas inflow rates (nL/s), bubble generation time (s), the PDMS, and liquid mass transfer fraction (cm3/cm3), respectively. It was found experimentally (described in the next section) that kpdms was negligible. Also, according to Garstecki et al. [79], bubble generation period tb, defined as the time required to generate a single bubble, was found to be inversely proportional to the ratio of gas pressure (Pg) and liquid flow rate (Ql), such as: 𝑡𝑡𝑏𝑏 ∝ 𝑃𝑃 1 𝑔𝑔 𝑄𝑄𝑙𝑙 . Both gas and liquid phase velocities (jg, jl) could be found by dividing the flow rates with corresponding tube cross-sectional areas, such as: Qg, Qliq, such as: 𝑗𝑗𝑔𝑔 = 𝑄𝑄𝑔𝑔 ⁄𝐴𝐴𝑔𝑔 and 𝑗𝑗𝑙𝑙 = 𝑄𝑄𝑙𝑙 ⁄𝐴𝐴𝑙𝑙 , where jg, jl, Ag and Al are gas superficial velocity, liquid superficial velocity, gas, and liquid tube cross-sectional area, 75 respectively. The superficial velocity is defined as the calculated flow velocity over a known cross-sectional area. After substituting these into (5.1), the bubble volume prediction equation can be expressed as: 𝑉𝑉𝑏𝑏 ≈ => 𝑉𝑉𝑏𝑏 ≈ 1 𝑄𝑄 − 𝑘𝑘𝑙𝑙 𝑄𝑄𝑔𝑔 𝑃𝑃𝑔𝑔 𝑄𝑄𝑙𝑙 𝑔𝑔 𝑗𝑗𝑔𝑔 𝐴𝐴𝑔𝑔 1 [1 − 𝑘𝑘𝑙𝑙 ] 𝑃𝑃𝑔𝑔 𝑗𝑗𝑙𝑙 𝐴𝐴𝑙𝑙 1−𝑘𝑘𝑙𝑙 𝑗𝑗𝑔𝑔 => 𝑉𝑉𝑏𝑏 ≈ 𝑘𝑘 1 𝑃𝑃𝑔𝑔 𝑗𝑗 ………………..……………(5.2) 𝑙𝑙 here k1 is a constant related to gas pressure and gas and liquid tube cross-sectional area ratio. The resultant bubble volume equation was then verified by measuring the bubble size of a single gas (Figure 5.8). 5.6.2 Modelling Versus Experimental To validate the theoretical model of the bubble volume (5.2), microbubble volumes were measured and compared with the model by first generating microbubbles with a bubble sensor and then by measuring their sizes with image analysis. In this experiment, helium was utilized to generated bubbles, gas pressure was fixed at 14.5 kPa, and the liquid flow rates were varied from 1.87-2.67 μLs-1 at every 0.27 μLs-1 interval. At each liquid flow rate, the bubble images were captured with a high-speed camera (Olympus ispeed-2 at 4000 fps). In image analysis, a bubble volume was calculated based on the following equation: 4⁄3 𝜋𝜋𝑟𝑟 3 𝑉𝑉𝑏𝑏 = 4 3 𝜋𝜋ℎ2 (3𝑟𝑟 − ℎ) 𝜋𝜋𝑟𝑟 − 2 × 3 3 ; 2𝑟𝑟 < 𝐻𝐻 ; 2𝑟𝑟 > 𝐻𝐻 76 3 Bubble volume (nL) 2.5 Experimental Estimated 2 1.5 14.5 kPa 16.5 kPa 18.5 kPa 1 20.5 kPa 22.5 kPa 0.5 0.5 1 1.5 2 2.5 3 Gas-liquid velocity ratio (jg/jl) Figure 5.8. Bubble volume (nL): theoretical versus experimental. Here the volumes of two spherical caps were subtracted from a sphere volume when a bubble diameter was higher than a microchannel height, H (110 μm). Note that h and r indicated the height of a spherical cap and the radius of a sphere. The gas pressure was then increased to four different levels to 16.5, 18.5, 20.5, and 22.5 kPa. At each gas pressure, the corresponding bubble volume was measured by repeating the described process for different sets of liquid flow rates in the range of 1.87-4.8 μLs-1. The estimated bubble volume indicated that the calculated bubble volume from the (5.2) matched to the experimental results with errors of 13.93, 6.36, 7.98, 5.97, and 5.39% at five different gas pressure levels of 14.5, 16.5, 18.5, 20.5, and 22.5 kPa, respectively, indicating a reasonably-good agreement of only 7.93% average error 77 (Figure 5.8). Note that percentage error was calculated based on the following equation: 𝑉𝑉𝑡𝑡ℎ𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒 −𝑉𝑉𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒 𝑉𝑉𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒 × 100%. Note that the bubble volume dependence on jg/jl ratio is widely accepted; however, our inclusion of mass transfer fraction makes this bubble volume equation versatile for any type of gases. 5.7 Gas Detection To detect a target gas, at first it was injected in a flow of carrier gas, and then corresponding bubble diameter was monitored. After performing gas detection, bubble chromatograms were further investigated by modelling a chromatogram with respect to an experimental result. 5.7.1 Bubble Chromatogram To observe bubble chromatogram, three different experiments were performed (Figure 5.9). In the first experiment, hydrogen was used as a target gas in helium carrier flow. In the second experiment, helium was injected in hydrogen carrier flow, while in the third experiment, pentane was injected in helium carrier gas. To change the carrier gas, both the GC system and bubble sensor were turned off, and then the corresponding cylinder was reconnected with the GC system. In all of the experiments, the following operating conditions were maintained: carrier gas flow rate: 1 ml/min, gas injection 28.6 µL, pentane injection 0.003 μL (split ratio 1:70), oven temperature 400C, injection temperature 2200C, liquid flow rate 110 μL/min. During bubble generation, bubble size was monitored for 10 seconds by capturing images at 4000 fps with a high-speed camera. 78 Figure 5.9. Bubble chromatogram of permanent gases and VOCs: (a) target gas: hydrogen, carrier gas: helium, (b) target gas: helium, carrier gas: hydrogen, (c) target gas: pentane, carrier gas: helium. 79 Later the images were analyzed by the python-based program, as mentioned above. The first experimental results showed that helium carrier gas generated bubbles with an average diameter of 119.5 μm. When 28.4 μL of hydrogen reached the bubble sensor, bubble sizes were reduced by 28.3 μm. The chromatogram height and width were 28.3 μm and 2.9 seconds. Experimental results also showed that the chromatogram contains 2020 bubbles in a 2.9 seconds time period. The second experimental results showed that hydrogen carrier gas generated bubbles with an average diameter of 126.2 μm. When 28.4 μL of helium reached the bubble sensor, bubble sizes were increased by 31.3 μm. The chromatogram height and width were 31.3 μm and 1.5 seconds. Experimental results also showed that the chromatogram contains 985 bubbles in a 1.5 seconds time period. The third experimental results showed that helium carrier gas generated bubbles with an average diameter of 128.5 μm. When 0.003 μL of pentane reached at bubble sensor, bubble sizes were reduced by 60.3 μm with as low as 68.2μm. The chromatogram height and width were 60.3 μm and 1.1 seconds. Experimental results also showed that the chromatogram contains 745 bubbles in a 1.1 seconds time period. 5.7.2 Modelling of Bubble Chromatogram This bubble generation theory was then extended to bubble chromatography theory by incorporating both carrier and solute phase mass transfer fractions. Considering negligible PDMS diffusion, the (5.1) can be rewritten as: 𝑉𝑉𝑏𝑏 ≈ (1 − 𝑘𝑘𝑙𝑙 )𝑄𝑄𝑔𝑔 𝑡𝑡𝑏𝑏 . Since the bubble generation frequency, Nb is: 𝑁𝑁𝑏𝑏 = 1⁄𝑡𝑡𝑏𝑏 , instantaneous incoming gas volume per single bubble can be estimated by dividing the gas flow rate with Nb. Thus, the single 80 bubble volume equation can be expressed as: 𝑉𝑉𝑏𝑏 (𝑡𝑡) ≈ 𝑘𝑘𝑙𝑙′ 𝑄𝑄𝑔𝑔′ where 𝑘𝑘𝑙𝑙′ = 1 − 𝑘𝑘𝑙𝑙 indicated an input fraction of gas that was involved in each bubble generation. Now to incorporate both carrier and solute phase mass transfer fractions, it was considered that both solute and carrier formed a binary mixture for a specific interval of time. Note that the injected solute sample was distributed in a Gaussian manner as the solute traveled through stationary and mobile phases. Thus, such distribution could be found in Van Deemter et al. [88]: 𝑐𝑐 𝐷𝐷𝑠𝑠 (𝑡𝑡) = 𝑐𝑐 𝐼𝐼 = 0 𝛽𝛽𝑡𝑡0 2𝜋𝜋𝜎𝜎12 +𝜎𝜎22 (𝑧𝑧⁄𝑢𝑢−𝛽𝛽𝛽𝛽)2 𝑒𝑒𝑒𝑒𝑒𝑒 − 2𝜎𝜎2 +𝜎𝜎2 ------(4) where 𝜎𝜎12 = 2𝐷𝐷𝐷𝐷⁄𝑢𝑢3 and β, σ2 1 2 are coefficients related to stationary phase mass transfer. Note that Ds(t) is the instantaneous distribution of solute in a carrier gas. While a mixture of carrier and solute phase generated a single bubble, the total volume loss would be:𝑉𝑉𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 = 𝑘𝑘𝑙𝑙,𝑐𝑐 𝑄𝑄𝑔𝑔′ (1 − 𝐷𝐷𝑠𝑠 ) + 𝑘𝑘𝑙𝑙,𝑠𝑠 𝑄𝑄𝑔𝑔′ 𝐷𝐷𝑠𝑠 . Here, kl,c and kl,s are mass transfer fraction of carrier and solute phase, respectively. Bubble chromatogram can then be found by subtracting volumetric gas loss from incoming gas for a single bubble, 𝑄𝑄𝑔𝑔′ : 𝑉𝑉𝑏𝑏,𝑐𝑐ℎ𝑟𝑟𝑟𝑟𝑟𝑟 = 𝑄𝑄𝑔𝑔′ − 𝑘𝑘𝑙𝑙,𝑐𝑐 𝑄𝑄𝑔𝑔′ (1 − 𝐷𝐷𝑠𝑠 ) + 𝑘𝑘𝑙𝑙,𝑠𝑠 𝑄𝑄𝑔𝑔′ 𝐷𝐷𝑠𝑠 𝑉𝑉𝑏𝑏,𝑐𝑐ℎ𝑟𝑟𝑟𝑟𝑟𝑟 = 𝑄𝑄𝑔𝑔′ 1 − 𝑘𝑘𝑙𝑙,𝑐𝑐 + 𝑘𝑘𝑙𝑙,𝑐𝑐 − 𝑘𝑘𝑙𝑙,𝑠𝑠 𝐷𝐷𝑠𝑠 A plot of such a bubble chromatogram is shown in Figure 5.10. 5.8 Gas Quantification To validate the bubble chromatogram theory, a bubble chromatogram was generated and compared it to the model. Specifically, pentane gas in 0.05 μL was injected 81 Loss Bubble diameter Diffusion into liquid Diffusion into PDMS Time To chromatogram Figure 5.10. Gas loss mechanism to generate chromatogram. into the helium carrier gas to generate microbubbles. The sizes of microbubbles were measured via image analysis at 4000 fps for 4-5 seconds. Both the bubble diameter and frequency were determined by analyzing 16,000-20,000 images with a custom Pythonbased program (computation time 90-120 seconds). For bubble generation, the following conditions were utilized: gas flow rate = 0.7 ml/min, carrier gas = helium, solute gas = pentane (C5H12), liquid flow rate = 2.93 μLs-1, liquid phase = water: glycerol: tween-20 = 25:10:1.75 (mg), and injector temperature = 2200C, split ratio = 1:10. Experimental results showed closely-matched trends traces of the measured bubble chromatograms to theoretical estimation, as shown in Figure 5.11. Here the following values were used for theoretical chromatograms: 𝑄𝑄𝑔𝑔′ = 1.47 nL, kl,c = 0.4 and kl,s =0.5. Theoretical and experimental chromatogram signal amplitudes were 0.74 and 0.67 nL, time widths were 2.3 and 1.8 seconds, and the volume losses during the 82 220 ng pentane injection Bubble chromatogram (nL) 1 0.9 0.8 0.7 Theoretical Experimental 0.6 0.5 0 2 4 6 Time (sec) 8 10 12 Figure 5.11. Bubble chromatogram: theoretical versus experimental. chromatogram periods were 116.58 and 99.63 nL. Experimental results also showed that the bubble diameter reduced from 119.24 μm, an average baseline, to 104.13 μm. Note that 1283 bubbles were generated during the solute interval of 1.8 seconds. Such a theoretical equation explained the phenomenon of a chromatogram that can be generated from the difference of two mass transfer fractions of a binary gas mixture. To quantify an unknown amount of gas at first (i) different two different volumes of pentane were injected, then (ii) corresponding bubble size and count were measured, and (iii) then injected gas volume was estimated by adding all the bubbles’ volume in that bandwidth. Measurement results showed estimated bubble volumes as 0.29 and 0.57 μL for the injected amount of 0.33 and 0.57 μL. This result indicated a successful 83 quantification process with an error of 12.1 and 14.0%, respectively. 5.9 Limit of Detection and Sensitivity To evaluate the performance of bubble chromatography in terms of limit of detection and sensitivity, different volume of a diluted sample was injected in a flow of carrier gas while carrier gas was generating bubbles through a microfluidic device. In these experiments, pentane (C5H12) was utilized as a VOC sample, and decane (C10H22) was used as a solvent to dilute the pentane sample. Experimental results showed that bubble chromatography could detect a minimum of 3.52 nanograms of pentane while the sensitivity was measured as 0.39 μm/ng and 6.94 bubbles/ ng, as shown in Figure 5.12. 6.94 bubbles/ng 600 400 40 30 200 0 0.39 μm/ng 20 10 Bubble diameter (μm) Bubble counts 800 0 0 20 40 60 80 100 Injected mass (nanograms) Figure 5.12. Limit of detection and sensitivity of bubble chromatography. 84 5.10 Limitations of Microbubble Chromatography Microbubble chromatography showed promises in gas detection and quantification. Bubble chromatography enables chip-scale gas quantification because of the inherent ability of a bubble to encapsulate femtoliter to picoliter amount of gas. However, the detection limit of microbubble chromatography suffers from some limitations. These limitations are described below. • Fluidic Fluctuation: In a flow-focusing microfluidic device bubble is generated based on the interaction between two fluids. Such interaction introduces two fluctuations in bubble sizes: low frequency and high frequency fluctuations. A parameter: polydispersity index is often used to indicate the bubble size uniformity. This polydispersity index is defined by a ratio of the standard deviation to the mean diameter of bubbles. It has been reported that the polydispersity index can be as low as 3-5% [84]. However, the index increases at higher gas and liquid flow. • Measurement Limit: Experimental results showed that the current measurement system has two major limitations, as shown in Figure 5.13: (i) high-speed camera cannot record for a longer time at higher frame rate, e.g., increasing the frame rate from 25 to 1000 fps reduces the record time from 178.9 to 4.5 seconds, and (ii) retention time increases exponentially as the bubble generation frequency reduces below 2000. However, to measure the bubble count, recording frame rate should be at least twice as the bubble generation frequency. And the recording time should be at least three times to monitor all of the bubbles covering from the moment of injection to the occurrence of second sample in two sample 85 Record time vs.frame rate Retention time vs. bubble frequency Required record time vs. required frame rate 350 Time (seconds) 300 250 200 Required Zone 150 100 50 0 0 2000 4000 6000 Frame rate or bubble frequency 8000 Figure 5.13. Limitation of bubble size and frequency measurement. mixture. The required zone of bubble recording time and frame rates are shown as shaded parts in Figure 5.13. Note that the required zone is far away from the camera limit. In the current bubble sensor device the gas flow rate and retention time are 77 seconds and 11.29k bubbles/sec, respectively; however, in current measurement system video was recorded for 150 seconds at 50 frame/sec indicating that bubble frequency monitoring was avoided in order to cover the bubble size measurement over the entire 150 seconds period of time. Another problem after video recording process is that image processing speed reduces with a higher number of frame analysis. For example, a 100 seconds video at 2000 fps will require a 100×2000 fame analysis. A computer with a processor frequency of 2.5GHz and 8GB memory takes 0.87 seconds to process a single frame. This indicates that a 100 seconds long video will require 48 hours to finish the analysis. 86 Note that during multiple iterations always the next iteration takes longer time to finish than that of previous because it requires storing additional data from previous iteration. CHAPTER 6 CONCLUSION 6.1 Summary This dissertation proposed and validated mainly two proof-of-concepts: viscositybased gas sensing as a new gas sensing mechanism and a unique bubble-based chromatography technique. The measurement results validated the feasibility of both concepts and demonstrated their practical performance exceeding existing gas sensor options. In particular, the following results were achieved. • Viscosity-to-Pressure Transient: In this dissertation, the effect of gas viscosity on transient pressure development has been proven experimentally. A viscosity transient was generated by introducing a small amount of target gas in a flow of reference gas. It was observed that 0.1 μL of pentane gas in a flow of 0.25 sccm helium generated 453 Pa of transient pressure that traveled with the moving pentane gas molecules. • Pressure-to-Flow generation: The produced pressure transient correspondingly led to the generation of localized gas flows in a micro channel design. Such flow generation was particularly amplified through a contraction or expansion junction. Note that a contraction junction is a junction of a wide to a narrow channels, and an expansion junction is a junction of a narrow to a 88 wide channels. Flow was generated due to the asymmetry in flow between two differently-sized channels. • Flow Amplification: A slip mode narrow microchannel has been utilized as a flow amplifier. It has been well documented that the steady-state flow in a narrow microchannel increased from inlet to outlet when the gas flows in the slip regime. The slip regime can be generated when the Knudsen number falls in between 0.001 and 0.1, i.e., when the microchannel radius is 10 to 1000 times of mean free path length. In this regime, the gas undergoes continuous compression and rarefaction, which causes nonlinearity in pressure distribution, and the pressure drop becomes maximum at the end of microchannel. Such a higher pressure drop creates a higher steady state flow at the end of the channel. In this dissertation, the effect of transient flow in slip mode has been observed and experimentally validated. • Viscosity-Based Gas Sensor: This work showed the operation principle and performance of a viscosity-based gas sensor. The sensor showed following performances: (i) limit of detection 13.7 ppb, (ii) power consumption 30 mW, (iii) sensitivity 15×10-6 sccm/pg, (iv) drift over time 0.83%, (vi) temperature stability 94.3%, and (vii) detection range 130 μg-13 pg. • Microbubble Chromatography: This dissertation showed the working principle of newly proposed bubble chromatography. A bubble chromatography is defined as a type of chromatography where fluidically generated microbubble changes its size and generation rate according to the viscosity of gases. Thus, such chromatography enables to detect and quantify any 89 target gas in a chromatography system. The performance parameters of this microbubble chromatography are: (i) limit of detection 3.52 ng, (ii) bubble size sensitivity 0.34 μm/ng, (iii) bubble count sensitivity 6.94 bubbles/ng, and (iv) detection range 130 μg-4 ng. 6.2 Future Work There are several aspects that have not been addressed in this research. The performance of the sensor can be improved further by exploring the following issues: • Modeling of Transient Slip Flow: The steady-state slip flow model has been well documented; however, there is no transient slip flow model, which can predict the amplification factor. The effect of transient pressure in the slip flow regime has not been modeled in this dissertation. However, the effect has been addressed successfully in terms of enough experimental results. Currently, the steady-state flow amplification has been modeled by a multiplying factor of 4, Knudsen number, TMAC (tangential momentum accommodation coefficient), and pressure ratio. To develop the model the following experiment can be performed. At first, one wide and narrow channel should be coupled in series. The purpose of the narrow channel is to achieve very Knudsen number in the slip flow regime. Then the pressure transient can be applied with the help of a pulse shaped gas plug. Here pulse shaped gas plug will travel through a carrier gas as discussed in chapter 3. Note that this way of pressure pulse application has two advantages: (i) it can be made sure that a pressure transient is applied at a user defined instant other than the gas injection peak and (ii) amplitude and time width of the pressure pulse can be controlled by injecting different amount of target 90 sample. Then the corresponding flow at different locations of the narrow channel will provide the transient flow behavior. A plot can be generated from different pressure versus flow, which will help to support the model equation. • Pressure Sensor-Based Solute Monitoring: In this research, viscosity-to-pressure generation was utilized in flow amplification to ultimately monitor the solute chromatogram. However, in-situ monitoring is also possible by utilizing the concept of viscosity-to-pressure generation. A pressure sensor can be integrated at different locations to observe the time and concentration status of solute while the solute is travelling inside a channel. This pressure sensor-based monitoring will be helpful to develop any solute concentration-based system design. This system could be to turn on or off a certain valve when the solute concentration reduces below a certain level such as alarm-based home solution for carbon monoxide detection. • Gas Identification: This dissertation only discussed the feasibility of the viscosity-based sensor as postchromatography detector. Here the sensor can only detect the presence of a gas and can identify the gas based on previously known retention time. However, this sensor can also be explored to find the feasibility of utilizing it as a standalone sensor. This is because each gas has a distinctive viscosity while generates distinctive pressure pulse. By finding a gas viscosity to flow and then gas mixture to flow relationship a set of known data points can be generated. From this relationship, this viscosity-based sensor can be extended from postchromatography detector to standalone detector. • Internal Standard for Quantification: 91 In this research, a calibration curve or internal standard was generated only for pentane (C5H12) gas. Note that the calibration curve is required to know the percentage volume of a specific sample inside a carrier and sample gas mixed microbubble. However, a detailed characterization of all of the VOCs is required in order to know the bubble size versus percentage sample. For example, to quantify the 0.1ppm C10 sample, we need to know the number of bubbles and the amount of C10 sample in each different sized bubbles. Finding the number of bubbles does not require any calibration curve; however, finding the percentage of sample inside each bubble needs a calibration curve. • Bubble Chromatography: Integrated System Chip: One important research regarding bubble sensor is monolithic integration of both separation microcolumn and bubble sensor. In this research, a PDMS-based microfluidic device was fabricated to generate bubbles, and later a capillary tube was used to connect between the separation column and the microfluidic device. However, microchannel can also be made by the DRIE etching process and bonding it with glass. The advantage of such monolithic integration will be twofold: retention time will be greatly reduced because of eliminating the tubing connections between the separation column and bubble generator, and any gas diffusion through PDMS polymer will be prevented. Another important research is that the whole bubble size measurement system will be miniaturized by (i) replacing the camera with a CMOS line image sensor and (ii) replacing computer processing with microcontroller processing. Hamamatsu 10077 line image sensors will be used, which has a 7×7μm2 pixel size and 10MHz data output. To replace the microscopes system, a 60x mini pocket lens (6×4 mm2) will be used. The video data output from a line image sensor will be updated every 100ns, and the 92 movement of a bubble can be detected. A bubble on the image sensor will generate a lower output voltage as compared to no bubbles on an image sensor. Note that by implementing this system, both bubble size and frequency can be monitored at the same time, which has not been possible in current measurement schemes, mainly because of the higher frame rates and the data amounts. APPENDIX PRESSURE SENSING STENT A.1 Introduction A stent provides an excellent benefit for the patient with coronary heart disease by reopening a blocked artery. A stent procedure reduces the chest pain, increases the chance of recovery from artery blockage, and reduces the cost of multiple balloon angioplasty procedure – a required procedure to reopen an artery blockage. A stent is a hollow, metal, mesh-like expandable tube that is expanded with a balloon to reopen a blocked artery and stays there to support the reopened artery mechanically. After the invention of stent, the rate of coronary intervention to monitor artery condition has been reduced by 25% [89]. However, tissue grows near the stent area and closes the vessel again even after stent placement, and if the vessel diameter narrowing exceeds by 50%, then it is called restenosis. In a medical study carried over 100 patients [89], it was observed that after 6 months of the initial stent placement, the lumen diameter reduces by 33.33% from 2.58±0.44 to 1.72±0.69 mm. In another study run over 212 patients [90], it was observed that because of restenosis, 10.4% of patients had a heart attack and the rest of 89.6% of patients had either stable chest pain or sudden artery blockage (ischemia). A stent demands detection capability of restenosis to make an appropriate clinical decision of recatheterization for further artery reopening. According to the study of aortic 94 and distal pressure measurement, whenever the blood pressure difference (upstreamdownstream) is 10-30 mmHg, then it is considered as stenosis. The current clinical method to monitor pressure is the fractional flow reserve (FFR) method where pressure is monitored by inserting a pressure monitoring guidewires (PressureWire, Radi Medical System, Uppsala, Sweden) during coronary catheterization and both aortic and distal pressure is measured [91]. However, this process is also invasive and also requires catheterization, which costs approximately $13,014 per procedure. A new type of stent, smart stent, has been demonstrated in the research level that provides catheterization-free restenosis occurrence decisions by sensing the pressure remotely. The principle is that a stent can be used as an antenna and inductor (L) at the same time. Therefore a blood pressure-dependent capacitive sensor (C) can be utilized, and in a combination of a stent and a capacitive sensor, the LC resonance can be easily detected by an external inductor. The significant advantage of smart stent is that it does not require further catheterization. This is because after placement of a smart stent, it will continuously provide the restenosis pressure outside of body by inductive telemetry. Several research groups have developed remote pressure sensing stent; however, they have either one or more of the following major disadvantages: (i) the pressure sensors are too large (smallest one = 1.4×1.8 mm2) as compared to stent, (ii) the sensors are separately connected with the stent, and thus increases the chances of blood flow obstruction, and (iii) requires associated circuitry to transmit the pressure signal, which also acts as discrete component. To avoid any blood flow obstruction along the blood flow path, both pressure sensor and stent are required to have the smallest cross-sectional area along the flow path. Therefore, the pressure sensor’s footprint is needed to be as 95 small as stent wire width (0.15 mm), as the same shape of stent or blood vessel wall and monolithically integration of sensor and stent. Besides, blood flow obstruction additional circuit with a stent requires gigahertz range of communication link, and body tissue creates higher absorption loss at gigahertz frequency and, therefore, lossy and harmful at the same time. Wireless readout of such pressure sensing stent has also been developed by different techniques such as associated circuitry, passive LC resonance, magnetic field, or x-ray. The first approach [92], [93] integrated two discrete components with the stent: capacitive pressure sensor (C) to sense and associated circuitry to process and transmit the sensing signal. In this approach, the associated circuitry was used to convert the pressure modulated frequency into digital data and to send such data at high frequency (2.4 GHz). Note that the stent was used as an antenna to transmit the digital data. The second approach [94], [95], [96], [97], [98] [99]eliminated the circuit chip and separately connected the pressure sensor with the stent. Here the stent was utilized both as an inductor (L) and an antenna for passive LC resonance; therefore, the pressure modulated frequency was detected by tuning the resonant frequency with an external inductor. The third approach [100] utilized a magnetoelastic sensor to sense the pressure, and the pressure modulated magnetic oscillation was detected with an external magnetic coil. Note that, in this approach, the magnetoelastic sensor was separately connected with stent. Recently, Gulari et al. [101] developed a radiopaque liquid-based microfluidic pressure sensor and detected the pressure-induced liquid displacement by x-ray. However, all of the approaches raised serious concerns because: (i) multiple components may cause component loss and surgery difficulty, (ii) higher thickness and larger rectangular footprint of the pressure sensor causes higher blood flow obstruction 96 which may lead to higher chances of re-endothelialization, and (iii) x-ray-based detection is potential hazardous to health. According to a study, stenosis incidence is raised by 75.4% when the stent thickness increased by 180% from a 50 μm thick stent. This indicates that any increase in thickness across the blood flow increased the chances of restenosis. Furthermore, any rectangular structure in the stent design introduces nonstreamlined intense blood flow disturbance, and results showed that plaque accumulation was higher right after the rectangular zone. A.2. Main Concept: Microhydraulic Pressure Sensing Stent Operation principle can be explained by two unique properties: (1) hydraulic amplification and (2) fluidic digitization. Hydraulic amplification comes from higher volumetric ratio (α, α>>1) of large reservoir to narrow channel. As shown in Figure A.1, the fluidic chamber of the device has two sections, one is a large reservoir and the other one is a narrow channel. During normal conditions (zero applied pressure), the reservoir contains liquid and the channel is empty. The entire fluidic chamber is covered by parylene which is biocompatible and has large Young’s modulus [102]. Parylene on top of reservoir will act as a diaphragm. Plaque accumulation on the device will apply some pressure and will create a small downward deflection (δ) on diaphragm. This deflection will be amplified by the ratio α, and liquid will travel much larger distance, αδ throughout the channel, as shown in Figure A.1(a). Fluidic digitization comes from the interdigitated electrodes as liquid moves from one electrode to another and capacitance changes by multiples of liquid dielectric. Liquid movement causes the changes in capacitance between the pair of interdigitated electrodes due to the difference in dielectric constant 97 Previous: Multi- components: (Stent + sensor+ ASIC) Takahata et. al. Our approach: One component: Sensor embedded in stent Amplification Pressure Vessel Sensor on stent Small Capacitive pressure sensor (Cs) Chow et. al. Cs Inductive stent (Ls) Channel Chamber 100×d Amplification d Large Digitization C = C0 + ∑ Cn ,ε =80 + ∑ Cn ,ε =1 ≈ C0 + ∑ Cn ,ε =80 Ls C1 C0 P1 C2 C3 C4 P2 P3 P4 Figure A.1. Main concept of the proposed pressure sensing stent: One component system: capacitive pressure sensor embedded on stretchable stent (stainless steel wire); Signal amplification: diaphragm displacement by small amount (d) creates larger displacement (100×d) inside liquid microchannel; Signal digitization: capacitive signal increases by a step (digitized) as the liquid moves one electrode to the other; Pressure profiling: pressure sensors at multiple locations will indicate the pressure profiling due to plaque. between air and liquid. As the liquid passes by each gap of any two electrodes, it flips the corresponding capacitance into a higher value. Since all the capacitors are in a parallel configuration, their total capacitance is the sum of all individual capacitors, resulting in digitized outputs depending on how many electrode gaps the hydraulic liquid crosses. A.2.1. Fluidic Amplification To understand the behavior of the fluidic network of pressure sensor, all of the elements have been modeled with corresponding electrical equivalent resistor, capacitor and inductor. The equivalent circuit that explains the system behavior is shown in Figure A.2 and is explained below. In this fluid system there are two different types of elements: 98 pcm Qi Qc m Qo + pc - Pressure (Pa) or Flow rate (m3/s) (a) Fluidic Model (b) SPICE Simulation 2 2 1 0 0 -2 -1 x 6.8x103Pin -15 x 10 -4 τ=RC= 0.3085 s Qo -2 x 6.8x103Pin x 10-15Qc(Exp) x 10-15Qc(Fitted) -6 τ=RC= 0.3074 s -8 -3 0 (c) Experimental 1 2 3 4 Time (seconds) -10 5 0 5 10 Time (seconds) 15 20 Figure A.2. Equivalent circuit: (a) model circuit of the pressure sensor forms RLC transient circuit, (b) SPICE simulation model based on the values mentioned in Table A.1, and (c) liquid flow rate in the outlet channel showed transient behavior. structural elements and fluidic elements. The structural element such as membrane is denoted as capacitor and the membrane capacitance can be calculated from the following equation [103]: 𝐶𝐶𝑚𝑚 = 6𝑤𝑤 6 (1−𝜗𝜗2 ) 𝜋𝜋4 𝐸𝐸𝑡𝑡 3 ………………………………..(A.1) Another structural capacitance is chamber capacitance, which depends on the compressibility chamber wall flexibility [104]. Since chamber wall’s material is same as diaphragm material, capacitance can be calculated from Equation (A.1) by choosing the width accordingly. Among all of the fluidic elements the first element that comes after diaphragm deflection is chamber liquid inductance, which can be calculated from liquid density and channel height and area by following the equation [104], [105]: 𝐿𝐿𝑙𝑙 = 𝜌𝜌ℎ 𝐴𝐴 Then the liquid flow will be divided into two paths: one through the chamber and the other through the outlet, as shown in Figure A.2. Note that the flow rate Qcm through the 99 chamber would be smaller as compared to the outlet channel (Qo) because of smaller flow path in the chamber. Both of the resistances, one along chamber direction and the other along the outlet channel, can be calculated from the following equation [105] [106]: 𝑅𝑅 = 𝜇𝜇.𝑙𝑙.(𝑤𝑤+ℎ) (𝑤𝑤.ℎ)2.5 12 192 𝜋𝜋 1− 5 𝜀𝜀 tanh (1+𝜀𝜀)√𝜀𝜀 2𝜀𝜀 𝜋𝜋 . From the fluidic circuit model (Figure A.2), it can be observed that all of the components form a RLC transient circuit. Here the trapped air capacitor charges up to the chamber pressure when the pressure is applied and discharges through the outlet resistance when the applied pressure is off. The differential equation of the circuit can be written as: => 𝑝𝑝𝑐𝑐𝑐𝑐 = 𝑅𝑅𝑜𝑜 𝑄𝑄𝑜𝑜 + (𝐿𝐿𝑜𝑜 + 𝐿𝐿𝑙𝑙 ) 𝑑𝑑2 𝑝𝑝𝑐𝑐 𝑑𝑑𝑑𝑑 2 + 𝐿𝐿 𝑅𝑅𝑜𝑜 𝑜𝑜 +𝐿𝐿𝑙𝑙 𝑑𝑑𝑑𝑑𝑐𝑐 𝑑𝑑𝑑𝑑 + (𝐿𝐿 1 𝑜𝑜 +𝐿𝐿𝑙𝑙 )𝐶𝐶𝑏𝑏 𝑑𝑑𝑄𝑄𝑜𝑜 1 + 𝑄𝑄𝑜𝑜 𝑑𝑑𝑑𝑑 𝑑𝑑𝑑𝑑 𝐶𝐶𝑏𝑏 𝑝𝑝𝑐𝑐 = 𝑝𝑝𝑐𝑐𝑐𝑐 ; 𝑄𝑄𝑜𝑜 = 𝐶𝐶𝑏𝑏 𝑑𝑑𝑑𝑑𝑐𝑐 𝑑𝑑𝑑𝑑 . Here Pcm is the chamber pressure, and for simplification, it was assumed that chamber pressure is close to the input applied pressure: 𝑝𝑝𝑐𝑐𝑐𝑐 ≈ 𝑃𝑃𝑖𝑖𝑖𝑖 . The above differential equation is applicable for all of the three conditions of input pressure: gradual pressure increment, constant pressure, and pressure off. In order to solve the above differential equation, at first all of the circuit components’ values were calculated based on the corresponding equation (see Table A.1). By choosing these values from Table A.1, the circuit was simulated in SPICE. For trapped air capacitor, capacitance was assumed in such a way so that the RoCb time constant was matched with that of the experimental values. Outlet channel liquid flow rate Qo both from simulation and experiment are shown in Figure A.2(b-c). Experimental flow rate showed similar behavior as of simulation results such as: (i) during gradual increment of pressure flow rate followed the input pressure, (ii) 100 Table A.1. Values of all elements in fluidic model Symbol Ll Name Values 3.78 × 105 Kg/m4 Chamber liquid intertance Rcm Chamber resistance 5.11 × 1011 N.s/m5 Ccm Chamber capacitance 1.12 × 10-22 m5/N Ro Outlet channel 3.07 × 1017 N.s/m5 resistance Lo 2.10 × 109 Kg/m4 Outlet channel inertance Cb Trapped air capacitance 1.00 × 10-18 m5/N after input pressure reached to a constant value flow rate became zero, and (ii) after the pressure was switched off flow rate quickly discharged to zero according to τ=RoCb time constant. The solution of the differential equation was calculated as: 𝑝𝑝𝑐𝑐 = 𝑃𝑃𝑎𝑎 exp(−3.2578𝑡𝑡) 𝑄𝑄𝑐𝑐 = 𝐶𝐶𝑏𝑏 𝑑𝑑 (𝑝𝑝 ) 𝑑𝑑𝑑𝑑 𝑐𝑐 => 𝑄𝑄𝑐𝑐 = − 𝑡𝑡 > 𝑝𝑝𝑜𝑜𝑜𝑜𝑜𝑜 𝑃𝑃𝑎𝑎 𝐶𝐶𝑏𝑏 × exp(−3.2578𝑡𝑡) 𝑅𝑅𝑜𝑜 𝐶𝐶𝑏𝑏 𝑡𝑡 > 𝑝𝑝𝑜𝑜𝑜𝑜𝑜𝑜 𝑃𝑃 => 𝑄𝑄𝑐𝑐 = − 𝑅𝑅𝑎𝑎 × exp(−𝑡𝑡/𝜏𝜏). 𝑜𝑜 In order to find out the displaced volume under a specific pressure, this discharged flow rate was needed to be integrated over time: 101 𝑉𝑉𝑙𝑙𝑙𝑙𝑙𝑙 = ∞ 𝑡𝑡=𝑝𝑝𝑜𝑜𝑜𝑜𝑜𝑜 => 𝑉𝑉𝑙𝑙𝑙𝑙𝑙𝑙 = − 𝑄𝑄𝑐𝑐 . 𝑡𝑡. 𝑑𝑑𝑑𝑑 𝑃𝑃𝑎𝑎 ∞ 𝑡𝑡. exp(−𝑡𝑡/𝜏𝜏) 𝑑𝑑𝑑𝑑 𝑅𝑅𝑜𝑜 0 => 𝑉𝑉𝑙𝑙𝑙𝑙𝑙𝑙 = 𝑃𝑃𝑎𝑎 𝐶𝐶𝑏𝑏 𝑃𝑃 𝐶𝐶 𝑎𝑎 𝑏𝑏 => 𝑑𝑑𝑐𝑐ℎ = ℎ𝑤𝑤 . 𝑐𝑐ℎ Here dch, wch and h are distance covered by liquid, channel width, and channel height, respectively. By putting values of Pa, Cb, h, wch as 6.89 kPa, 1x10-18 m5/N, 12 μm, and 20 μm, liquid displacement was calculated as 28.7 μm. According to the experiment, liquid displacement was measured as 16 μm at 6.89 kPa. Therefore, the model showed a considerable agreement with the experimental value. A.2.2 Parameter Optimization According to the model circuit, it was obvious that the most of the input energy of diaphragm deflection was used to overcome the hydraulic resistance and trapped air capacitance [107], [108]. In order to find out the optimal channel width and optimal channel length, the following relationship was utilized: 𝐹𝐹 × 𝑑𝑑 = 𝑃𝑃𝑟𝑟 . 𝑑𝑑𝑑𝑑 + 𝑃𝑃𝑎𝑎 . 𝑑𝑑𝑑𝑑 => 𝐹𝐹 × 𝑑𝑑 = 𝑃𝑃𝑟𝑟 × (𝑑𝑑𝑐𝑐ℎ 𝑤𝑤𝑐𝑐ℎ ℎ) + 𝑃𝑃𝑎𝑎 × (𝑑𝑑𝑐𝑐ℎ 𝑤𝑤𝑐𝑐ℎ ℎ)………………..(A.2) where Pa, F, and d are applied pressure, force applied to diaphragm, and downward diaphragm deflection, respectively. Pr is hydraulic pressure drop and can be written in the form of: 128𝜇𝜇𝑙𝑙𝑐𝑐ℎ 𝑄𝑄𝑜𝑜 ⁄𝜋𝜋𝑑𝑑ℎ4 . Here, dch, lch, wch, and h are distance covered by liquid, channel length, channel width, and channel height, respectively. In the hydraulic pressure 102 drop equation, hydraulic diameter dh and flow rate can be written as: 𝑑𝑑ℎ = 𝑄𝑄𝑜𝑜 = 2ℎ𝑤𝑤𝑐𝑐ℎ ℎ + 𝑤𝑤𝑐𝑐ℎ 𝑉𝑉𝑐𝑐ℎ ℎ𝑤𝑤𝑐𝑐ℎ 𝑙𝑙𝑐𝑐 𝑤𝑤𝑐𝑐 ℎ𝑤𝑤𝑐𝑐ℎ 𝑑𝑑𝑙𝑙𝑐𝑐 𝑤𝑤𝑐𝑐 = 𝑑𝑑𝑐𝑐ℎ × = 𝑑𝑑 × × = 𝑡𝑡 𝑡𝑡 ℎ𝑤𝑤𝑐𝑐ℎ 𝑡𝑡 𝑡𝑡 𝑠𝑠𝑠𝑠, 𝑃𝑃𝑟𝑟 = 128𝜇𝜇𝑙𝑙𝑐𝑐ℎ 𝑑𝑑𝑙𝑙𝑐𝑐 𝑤𝑤𝑐𝑐 . 𝜋𝜋(𝑑𝑑ℎ )4 𝑡𝑡 In (A.2), force F can be written as 𝐹𝐹 = 𝑃𝑃𝑎𝑎 × 𝐴𝐴𝑐𝑐 = 𝑃𝑃𝑎𝑎 × 𝑤𝑤𝑐𝑐 𝑙𝑙𝑐𝑐 . Note that wc and lc are chamber or diaphragm width and length. By replacing the force in (A.2), 𝑃𝑃𝑎𝑎 × 𝑤𝑤𝑐𝑐 𝑙𝑙𝑐𝑐 𝑑𝑑 = 𝑃𝑃𝑟𝑟 × (𝑑𝑑𝑐𝑐ℎ 𝑤𝑤𝑐𝑐ℎ ℎ) + 𝑃𝑃𝑎𝑎 × (𝑑𝑑𝑐𝑐ℎ 𝑤𝑤𝑐𝑐ℎ ℎ) => 𝑟𝑟, 𝑑𝑑𝑐𝑐ℎ = => 𝑑𝑑𝑐𝑐ℎ = => 𝑑𝑑𝑐𝑐ℎ = => 𝑑𝑑𝑐𝑐ℎ = 𝑃𝑃𝑎𝑎 × 𝑤𝑤𝑐𝑐 𝑙𝑙𝑐𝑐 × 𝑑𝑑 128𝜇𝜇𝑙𝑙𝑐𝑐ℎ 𝑑𝑑𝑙𝑙𝑐𝑐 𝑤𝑤𝑐𝑐 𝑤𝑤𝑐𝑐ℎ ℎ 𝑃𝑃𝑎𝑎 + 𝜋𝜋(𝑑𝑑ℎ )4 𝑡𝑡 𝜋𝜋𝜋𝜋𝑃𝑃𝑎𝑎 𝑤𝑤𝑐𝑐 𝑙𝑙𝑐𝑐 𝑑𝑑 128𝜇𝜇𝑙𝑙𝑐𝑐ℎ 𝑑𝑑𝑙𝑙𝑐𝑐 𝑤𝑤𝑐𝑐 𝜋𝜋𝜋𝜋𝑃𝑃𝑎𝑎 𝑤𝑤𝑐𝑐ℎ ℎ + 𝑤𝑤𝑐𝑐ℎ ℎ × (𝑑𝑑ℎ )4 𝜋𝜋𝜋𝜋𝑃𝑃𝑎𝑎 𝑤𝑤𝑐𝑐 𝑙𝑙𝑐𝑐 𝑑𝑑 𝑤𝑤𝑐𝑐ℎ + ℎ 4 × 128𝜇𝜇𝑙𝑙𝑐𝑐ℎ 𝑑𝑑𝑙𝑙𝑐𝑐 𝑤𝑤𝑐𝑐 2𝑤𝑤𝑐𝑐ℎ ℎ 𝜋𝜋𝜋𝜋𝑃𝑃𝑎𝑎 𝑤𝑤𝑐𝑐ℎ ℎ + 𝑤𝑤𝑐𝑐ℎ ℎ × => 𝑑𝑑𝑐𝑐ℎ = => 𝑑𝑑𝑐𝑐ℎ = 𝑃𝑃𝑎𝑎 × 𝑤𝑤𝑐𝑐 𝑙𝑙𝑐𝑐 × 𝑑𝑑 𝑤𝑤𝑐𝑐ℎ ℎ(𝑃𝑃𝑟𝑟 + 𝑃𝑃𝑎𝑎 ) 24 𝜋𝜋𝜋𝜋𝑃𝑃𝑎𝑎 𝑤𝑤𝑐𝑐 𝑙𝑙𝑐𝑐 𝑑𝑑 24 𝜋𝜋𝜋𝜋𝑃𝑃𝑎𝑎 (𝑤𝑤𝑐𝑐ℎ ℎ) + 128𝜇𝜇𝑙𝑙𝑐𝑐ℎ 𝑑𝑑𝑙𝑙𝑐𝑐 𝑤𝑤𝑐𝑐 × (𝑤𝑤𝑐𝑐ℎ + ℎ)4 𝑤𝑤𝑐𝑐ℎ ℎ3 16𝜋𝜋𝜋𝜋𝑃𝑃𝑎𝑎 𝑤𝑤𝑐𝑐 𝑙𝑙𝑐𝑐 𝑑𝑑 𝑤𝑤 4 6 4 ℎ 16𝜋𝜋𝜋𝜋𝑃𝑃𝑎𝑎 (𝑤𝑤𝑐𝑐ℎ ℎ) + 128𝜇𝜇𝑙𝑙𝑐𝑐ℎ 𝑑𝑑𝑙𝑙𝑐𝑐 𝑤𝑤𝑐𝑐 × 𝑐𝑐ℎ + 2+ + 2 + 3 3 𝑤𝑤𝑐𝑐ℎ ℎ 𝑤𝑤𝑐𝑐ℎ 𝑤𝑤𝑐𝑐ℎ ℎ ℎ => 𝑑𝑑𝑐𝑐ℎ = 16𝜋𝜋𝜋𝜋𝑃𝑃𝑎𝑎 𝑤𝑤𝑐𝑐 𝑙𝑙𝑐𝑐 𝑑𝑑/ℎ 𝑤𝑤 4 16𝜋𝜋𝜋𝜋𝑃𝑃𝑎𝑎 (𝑤𝑤𝑐𝑐ℎ )+128𝜇𝜇𝑙𝑙𝑐𝑐ℎ 𝑙𝑙𝑐𝑐 𝑤𝑤𝑐𝑐 𝑑𝑑× 𝑐𝑐ℎ 4 + 3+ ℎ ℎ 6 . 4 1 + + 𝑤𝑤𝑐𝑐ℎ ℎ2 𝑤𝑤2 ℎ 𝑤𝑤3 𝑐𝑐ℎ 𝑐𝑐ℎ 103 Considering 𝑘𝑘1 = 16𝜋𝜋𝜋𝜋𝑃𝑃𝑎𝑎 𝑤𝑤𝑐𝑐 𝑙𝑙𝑐𝑐 𝑑𝑑/ℎ, 𝑘𝑘2 = 16𝜋𝜋𝜋𝜋𝑃𝑃𝑎𝑎 ,and 𝑘𝑘3 = 128𝜇𝜇𝑙𝑙𝑐𝑐ℎ 𝑙𝑙𝑐𝑐 𝑤𝑤𝑐𝑐 𝑑𝑑 the above equation becomes 𝑑𝑑𝑐𝑐ℎ = 𝑘𝑘1 , 𝑤𝑤𝑐𝑐ℎ 4 6 4 1 𝑘𝑘2 𝑤𝑤𝑐𝑐ℎ + 𝑘𝑘3 × 4 + 3 + + 2 + 3 ℎ ℎ 𝑤𝑤𝑐𝑐ℎ ℎ2 𝑤𝑤𝑐𝑐ℎ ℎ 𝑤𝑤𝑐𝑐ℎ variation of dch with respect to wch at different applied pressure. For simplification, higher order of 1ℎ terms such as 𝑤𝑤𝑐𝑐ℎ ℎ4 , 4ℎ3 , and 6𝑤𝑤 ℎ2 was neglected, and simulation 𝑐𝑐ℎ results suggested that neglecting these first three terms do not move the location of (𝑑𝑑𝑐𝑐ℎ )𝑚𝑚𝑚𝑚𝑚𝑚 except modifying the value of (𝑑𝑑𝑐𝑐ℎ )𝑚𝑚𝑚𝑚𝑚𝑚 . The last two terms, 4 2 and 𝑤𝑤𝑐𝑐ℎ ℎ 1 , are important; however, neglecting the last term 1 3 3 simplifies the equation 𝑤𝑤𝑐𝑐ℎ 𝑤𝑤𝑐𝑐ℎ further by sacrificing the location of (𝑑𝑑𝑐𝑐ℎ )𝑚𝑚𝑚𝑚𝑚𝑚 by ~2 μm. Then the equation simplified to ′ 𝑑𝑑𝑐𝑐ℎ = 𝑘𝑘 2 𝑘𝑘1 𝑤𝑤𝑐𝑐ℎ ℎ 3 2 𝑤𝑤𝑐𝑐ℎ ℎ+4𝑘𝑘3 . ′ ′ Now solving for optimal channel width 𝑤𝑤𝑐𝑐ℎ where the liquid displacement 𝑑𝑑𝑐𝑐ℎ can be maximum, => 3 (𝑘𝑘2 𝑤𝑤𝑐𝑐ℎ ℎ + 4𝑘𝑘3 ) × 𝑑𝑑 (𝑑𝑑′ ) = 0 𝑑𝑑𝑤𝑤𝑐𝑐ℎ 𝑐𝑐ℎ 𝑑𝑑 𝑑𝑑 2 (𝑘𝑘 𝑤𝑤 2 ℎ) − 𝑘𝑘1 𝑤𝑤𝑐𝑐ℎ (𝑘𝑘 𝑤𝑤 3 ℎ + 4𝑘𝑘3 ) ℎ× 𝑑𝑑𝑤𝑤𝑐𝑐ℎ 1 𝑐𝑐ℎ 𝑑𝑑𝑤𝑤𝑐𝑐ℎ 2 𝑐𝑐ℎ =0 3 (𝑘𝑘2 𝑤𝑤𝑐𝑐ℎ ℎ + 4𝑘𝑘3 )2 3 2 2 ℎ + 4𝑘𝑘3 ) × (2𝑘𝑘1 𝑤𝑤𝑐𝑐ℎ ℎ) − 𝑘𝑘1 𝑤𝑤𝑐𝑐ℎ ℎ × (3𝑘𝑘2 𝑤𝑤𝑐𝑐ℎ ℎ) = 0 => (𝑘𝑘2 𝑤𝑤𝑐𝑐ℎ 4 2 => 𝑘𝑘2 𝑤𝑤𝑐𝑐ℎ ℎ − 8𝑘𝑘3 𝑤𝑤𝑐𝑐ℎ ℎ = 0 3 8𝑘𝑘 => 𝑤𝑤𝑐𝑐ℎ = 𝑘𝑘 ℎ3 . 2 In order to find out the optimal channel length Boyle’s law was used, 104 𝑃𝑃𝑎𝑎𝑎𝑎𝑎𝑎 𝑉𝑉1 = 𝑃𝑃𝑚𝑚𝑚𝑚𝑚𝑚 𝑉𝑉2. Considering a maximum applied pressure as 260 mmHg or 0.34 atm, the equation can be modified to: 1 × 𝑉𝑉1 = 0.34 × 𝑉𝑉1 − 𝑉𝑉𝑐𝑐ℎ,𝑚𝑚𝑚𝑚𝑚𝑚 => 0.34 × 𝑉𝑉1 = 1.34 × 𝑉𝑉𝑐𝑐ℎ,𝑚𝑚𝑚𝑚𝑚𝑚 => 𝑉𝑉1 = 3.94 × 𝑉𝑉𝑐𝑐ℎ,𝑚𝑚𝑚𝑚𝑚𝑚 => 𝑙𝑙𝑐𝑐ℎ 𝑤𝑤𝑐𝑐ℎ ℎ = 3.94 × 𝑑𝑑𝑐𝑐ℎ,𝑚𝑚𝑚𝑚𝑚𝑚 𝑤𝑤𝑐𝑐ℎ ℎ. From the equation of dch, the maximum dch can be calculated as 424 μm, which was used to calculate the optimal channel length as 1.67mm. A.2.3. Capacitive Amplification The capacitance between two coplanar electrodes is defined by the following equation [109]: 𝐶𝐶𝑐𝑐𝑐𝑐𝑐𝑐 2 2𝜀𝜀0 𝜀𝜀𝑟𝑟 𝑙𝑙𝑐𝑐𝑐𝑐𝑐𝑐 2𝑤𝑤𝑐𝑐𝑐𝑐𝑐𝑐 2𝑤𝑤𝑐𝑐𝑐𝑐𝑐𝑐 = ln 1 + + 1 + − 1 𝜋𝜋 𝑔𝑔𝑐𝑐𝑐𝑐𝑐𝑐 𝑔𝑔𝑐𝑐𝑐𝑐𝑐𝑐 where ε0, εr, lcop, wcop, and gcop are permittivity of air, relative permittivity of medium, length, width, and gap distance of the coplanar electrodes. By adopting the measurement data of the coplanar electrode as: wcop = 16 μm and gcop = 6 μm, the logarithmic term gave a value of 2.5, which makes the coplanar capacitance as: 𝐶𝐶𝑐𝑐𝑐𝑐𝑐𝑐 ≈ 5𝜀𝜀0 𝜀𝜀𝑟𝑟 𝑙𝑙𝑐𝑐𝑐𝑐𝑐𝑐 ⁄𝜋𝜋. Since the coplanar capacitance represents the capacitance between two electrodes, as liquid advances, the differential capacitance from electrode to electrode will be same as the above equation, 105 ∆𝐶𝐶𝑐𝑐𝑐𝑐𝑐𝑐 = 𝐶𝐶𝑐𝑐𝑐𝑐𝑐𝑐 = 5𝜀𝜀0 𝜀𝜀𝑟𝑟 𝑙𝑙𝑐𝑐𝑐𝑐𝑐𝑐 𝜋𝜋 . On the other hand, the capacitance of parallel plate capacitor is 𝐶𝐶𝑝𝑝𝑝𝑝𝑝𝑝 = 𝜀𝜀0 𝜀𝜀𝑟𝑟 𝑙𝑙𝑝𝑝𝑝𝑝𝑝𝑝 𝑤𝑤𝑝𝑝𝑝𝑝𝑝𝑝 𝑔𝑔𝑝𝑝𝑝𝑝𝑝𝑝 where lpar, wpar, and gpar are length, width, and gap distance of parallel plate electrodes. Now, if we consider a parallel plate diaphragm capacitor having same length and width and gap distance of our liquid chamber diaphragm, then lpar = 400 μm, wpar = 100 μm, and gpar = 12 μm. In order to compare the capacitance between two, applied pressure needed to be same in both cases. Roughly, that can be calculated by following the fluidic ′ amplification equation: 𝑔𝑔𝑝𝑝𝑝𝑝𝑝𝑝 × 𝑙𝑙𝑝𝑝𝑝𝑝𝑝𝑝 × 𝑤𝑤𝑝𝑝𝑝𝑝𝑝𝑝 = 𝑤𝑤𝑐𝑐𝑐𝑐𝑐𝑐 + 𝑔𝑔𝑐𝑐𝑐𝑐𝑐𝑐 × 𝑤𝑤𝑐𝑐ℎ × ℎ. The new gap ′ distance of the parallel plate capacitor at the new pressure can also be written as: 𝑔𝑔𝑝𝑝𝑝𝑝𝑝𝑝 = (1 − 𝑘𝑘)𝑔𝑔𝑝𝑝𝑝𝑝𝑝𝑝 , where k is a constant and 0<k<1. Then the differential capacitance can be written as ∆𝐶𝐶𝑝𝑝𝑝𝑝𝑝𝑝 = 𝜀𝜀0 𝑙𝑙𝑝𝑝𝑝𝑝𝑝𝑝 𝑤𝑤𝑝𝑝𝑝𝑝𝑝𝑝 𝜀𝜀0 𝑙𝑙𝑝𝑝𝑝𝑝𝑝𝑝 𝑤𝑤𝑝𝑝𝑝𝑝𝑝𝑝 − ′ 𝑔𝑔𝑝𝑝𝑝𝑝𝑝𝑝 𝑔𝑔𝑝𝑝𝑝𝑝𝑝𝑝 => ∆𝐶𝐶𝑝𝑝𝑝𝑝𝑝𝑝 = 𝜀𝜀0 𝑙𝑙𝑝𝑝𝑝𝑝𝑝𝑝 𝑤𝑤𝑝𝑝𝑝𝑝𝑝𝑝 𝜀𝜀0 𝑙𝑙𝑝𝑝𝑝𝑝𝑝𝑝 𝑤𝑤𝑝𝑝𝑝𝑝𝑝𝑝 − (1 − 𝑘𝑘)𝑔𝑔𝑝𝑝𝑝𝑝𝑝𝑝 𝑔𝑔𝑝𝑝𝑝𝑝𝑝𝑝 => ∆𝐶𝐶𝑝𝑝𝑝𝑝𝑝𝑝 = 𝜀𝜀0 𝑙𝑙𝑝𝑝𝑝𝑝𝑝𝑝 𝑤𝑤𝑝𝑝𝑝𝑝𝑝𝑝 𝑔𝑔𝑝𝑝𝑝𝑝𝑝𝑝 𝑘𝑘 1−𝑘𝑘. Therefore, the capacitive amplification can be written as: ∆𝐶𝐶𝑐𝑐𝑐𝑐𝑐𝑐 5𝜀𝜀0 𝜀𝜀𝑟𝑟 𝑙𝑙𝑐𝑐𝑐𝑐𝑐𝑐 𝑔𝑔𝑝𝑝𝑝𝑝𝑝𝑝 1 − 𝑘𝑘 = × × ∆𝐶𝐶𝑝𝑝𝑝𝑝𝑝𝑝 𝜋𝜋 𝜀𝜀0 𝑙𝑙𝑝𝑝𝑝𝑝𝑝𝑝 𝑤𝑤𝑝𝑝𝑝𝑝𝑝𝑝 𝑘𝑘 ∆𝐶𝐶𝑐𝑐𝑐𝑐𝑐𝑐 5𝜀𝜀𝑟𝑟 𝐴𝐴𝑐𝑐ℎ 1 − 𝑘𝑘 = × ∆𝐶𝐶𝑝𝑝𝑝𝑝𝑝𝑝 𝜋𝜋𝐴𝐴𝑑𝑑 𝑘𝑘 where Ach was the cross sectional area of the microchannel because the length of the 106 coplanar electrode (lcop) was same as width of channel (wch), and gap distance in parallel plate capacitor (gpar) was same as height (h) of the channel. The other area Ad represented the area of the diaphragm irrespective of parallel plate capacitor and liquid chamber diaphragm in coplanar case. According to calculation, the above ration can be written as ∆𝐶𝐶𝑐𝑐𝑐𝑐𝑐𝑐 = 0.04𝜀𝜀𝑟𝑟 ≈ 1.6. ∆𝐶𝐶𝑝𝑝𝑝𝑝𝑝𝑝 According to this equation, the capacitive amplification is directly proportional to the relative permittivity of the liquid media. A.2.4. Wireless Resonance The fabricated pressure sensing stent will form a series RLC circuit where Rs, Ls, and Cs are resistance of the stent and pressure sensor, inductance of stent coil, and capacitance of the coplanar capacitive pressure sensor. The impedance and resonance frequency of such circuit can be written as 𝑍𝑍𝑠𝑠 (𝜔𝜔) = 𝑅𝑅𝑠𝑠 + 𝑗𝑗 𝜔𝜔𝐿𝐿𝑠𝑠 − 1 𝑓𝑓𝑟𝑟 = 2𝜋𝜋𝐿𝐿 𝑠𝑠 𝐶𝐶𝑠𝑠 . 1 𝜔𝜔𝐶𝐶𝑠𝑠 Such resonant frequency could be detected by using an external inductor. Through inductive coupling the external inductor energized the sensor inductor and because of such coupling load impedance from the sensor coil was reflected back to the external coil. Such reflected impedance can be found from [110]: (𝜔𝜔𝜔𝜔)2 𝑋𝑋𝑟𝑟𝑟𝑟𝑟𝑟 (𝜔𝜔) = 𝑍𝑍𝑠𝑠 107 where M is mutual inductance: 𝑀𝑀 = 𝐿𝐿𝑒𝑒 𝐿𝐿𝑠𝑠 and Le is the inductance of external coil. Therefore, the total impedance of the external coil would be 𝑍𝑍𝑒𝑒 = 𝑅𝑅𝑒𝑒 + 𝑗𝑗𝑗𝑗𝐿𝐿𝑒𝑒 + 𝑋𝑋𝑟𝑟𝑟𝑟𝑟𝑟 . However, at resonant frequency fr, the sensor impedance Zs would be resistive, and that is why the external coil impedance would be 𝑍𝑍𝑒𝑒 (𝜔𝜔0 ) = 𝑅𝑅𝑒𝑒 + 𝑗𝑗𝜔𝜔0 𝐿𝐿𝑒𝑒 + (𝜔𝜔0 𝑀𝑀)2 𝑅𝑅𝑠𝑠 . This equation indicated that both magnitude and phase of the external coil impedance is being influenced by the sensor coil. A.3. On-wire Pressure Sensor Fabrication The pressure sensor was fabricated on top of a cylindrical copper wire. To perform wafer level fabrication, at first the wire was transferred to a silicon wafer, and then a parylene-based capacitive pressure sensor was fabricated by utilizing standard fabrication technique. A.3.1. Grafting and Polishing First, on a silicon wafer, multiple trenches were formed as wire holders by etching the silicon wafer with DRIE process. Each trench was 300µm wide, 180µm deep, and 7080mm long. In order to fix the wires inside the trenches, at first cyanoacrylate-based adhesive (super glue) liquid was poured inside the trenches. Thirty gauge (260μm) wires were previously cut according to the length of each trench-line. Each wire was then placed into the trenches, and a weight was placed on top of that for several hours. Then 108 two steps of polishing were performed: mechanical polishing (MP) and chemical mechanical polishing (CMP). In the first step, a 600 grit sized (median particle diameter = 15.3μm) silicon carbide paper was used to polish on Allied MultiPrep Polishing System. This mechanical polishing was carried out at 50 rpm and 10-15 psi of hand pressure for 60 minutes, which gave the wire surface roughness of ±1µm with average roughness of 228nm, as shown in Figure A.3. In the next step, CMP was performed in Strausbaurgh 6EC CMP system. The recipe was followed as: table rpm:50, chuck rpm: 25, slurry rate: 100ml/min, applied pressure 6 psi for 10 minutes and additional 8 psi for 10 minutes. This CMP process gave Cu surface roughness of about ±100nm with average of 34nm. Then the wafer was ready to be used to perform the subsequent wafer level fabrication process. Surface roughness Surface roughness (nm) 1000 After MP (mean ±228 nm) After MP After CMP 500 0 -500 After CMP (mean ±34 nm) -1000 0 2000 1500 1000 500 Along Cu wire direction (nm) Figure A.3. Surface roughness after two steps of polishing. 109 A.3.2. Standard Fabrication Process For stent fabrication, at first on the polished wafer 1.2 μm parylene-C was deposited to provide an insulation layer (Figure A.4(a)-(i)). The insulation layer was patterned by photolithography and O2 plasma etching (Oxford 80 RIE) to form two VIAs (vertical interconnect access) at two ends of each wire (Figure A.4(a)-(i)). Before going to the next step of gold deposition, a 30s of O2 plasma etching was done on top of entire parylene to promote the adhesion between gold and parylene. A 300nm gold was then deposited on top of parylene followed by gold patterning to produce coplanar interdigitated capacitive electrode (Figure A.4(a)-(ii)). The interdigitated electrode finger width and gap distance were maintained at 16 and 6μm. To fabricate the hydraulic pressure sensor on top of the wire, an 11μm sacrificial photoresist layer was patterned in rectangular microchannel shape (Figure A.4(a)-(iii)), which consisted of a wider channel (400×100×20μm3) and a narrow channel (500×20×20µm3). Note that the wider channel and narrow channel corresponded to liquid pressure chamber and liquid displacement channel, respectively (Figure A.4(b)). To deposit a diaphragm layer on top of sacrificial layer, 1.2μm parylene-C was deposited (Figure A.4(a)-(iv)). The parylene layer was then patterned to remove parylene from the two ends of the sacrificial photoresist layer (Figure A.4(a)-(v)). The wafer was then kept under acetone for 4-5 hours to completely remove the sacrificial photoresist (Figure A.4(a)-(vi)). The parylene layer formed a freestanding membrane once the sacrificial photoresist was removed. Then the microchannel was ready to be filled with liquid. SEM images of a final fabricated pressure sensing stent is shown in Figure A.4(b). 110 (a) Fabrication stepts (i) Insulation Parylene (400nm) deposition and patterning for via Polished wire Silicon wafer Metal deposition (400nm) and patterning for electrode (ii) (iii) Sacrificial photoresist (12μm) (iv) Diaphragm parylene (1.2 μm) deposition (v) Parylene patterning (vi) Acetone removal of PR (b) Fabricated device Liquid chamber Coplanar capacitor Capacitor leg (20μm width) Channel Cu wire (260μm) as stent Figure A.4. Fabrication of on-wire pressure sensor: (a) fabrication steps and (b) fabricated pressure sensing wire (stent). 111 A.3.3. Liquid Filling and Wire Release The hollow microchannel was filled with working media (glycerol) by placing a drop of glycerol at the micrchannel end (Figure A.5). With the influence of capillary driven flow through micrchannel, glycerol quickly (within 10-15s) filled up the entire microchannel and chamber and then reached to the other end. After that the chamber end was sealed with silicone sealant. Note that the working media was chosen as glycerol because of its biocompatibility, and it does not evaporate at room temperature (boiling temperature 2980C). According to operation principle, liquid needed to be removed from the microchannel portion. Evaporation process was adopted to remove the liquid; however, any thermal process above 1300C would degrade the quality of parylene, and 43.2 120 36.0 100 28.8 21.6 14.4 7.2 0 120 Torr 20 Torr 1.2 Torr Repeatability (102 80 devices) = 4.42% Displacement (μm) Volume (picoliter) glycerol boiling temperature is 2980C. Therefore, to reduce the evaporation temperature, 60 40 20 90 95 100 105 110 115 Evaporation temperature (0C) 120 Figure A.5. Characterization of glycerol evaporation process. 112 evaporation was performed at reduced pressure in a vacuum oven (YES LP II). To observe the effect of vacuum pressure on the evaporation temperature, we performed an experiment by varying the vacuum pressure from 1.2-120 Torr and temperature from 901200C, and for each setup, the amount of liquid evaporation was monitored. Experimental results showed that (Figure. A.5) by reducing the vacuum pressure by 83.3% from 120 Torr, it increased the glycerol evaporation by 2.96, 2.82, 2.10, and 4.14×103 μm3 at 900, 1000, 1100, and 1200C, respectively. Reducing the pressure further by 94.0% from 20 Torr removed even higher volume of as measured by 2.59, 3.13, 6.74, and 10.28×103 μm3 at 900, 1000, 1100, and 1200C, respectively. At each setup of temperature and vacuum pressure, standard deviation of liquid evaporation was calculated to estimate the liquid meniscus controllability over temperature and pressure. Calculated results showed that the higher controllability could be achieved at lower pressure and lower temperature, which was as low as ±4.66μm or 1.12×103 μm3. When the liquid meniscus reached 80100 μm ahead of the chamber, then evaporation was stopped and the microchannel end was sealed with silicone sealant. The pressure sensing wire was then wound around a 2.5mm diameter tube to turn it into an inductor. While winding the wire into coil, the pressure sensor was kept inner side of the coil (i) to prevent any Au electrode cracking because of bending and (ii) to ultimately measure the blood flow change because of stenosis. A.4 Stenosis Pressure Detection In order to measure the stenosis pressure (0-250 mmHg), the fabricated pressure sensing stent was tested in three different ways: optically measurement of pressure versus 113 liquid progression, capacitance measurement of pressure versus coplanar capacitance, and wireless measurement of pressure versus resonant frequency. A.4.1. Optical Method For optical testing, the device was placed inside a custom pressure chamber. The custom pressure chamber was built from acrylic sheet and was sealed both adhesively and mechanically. The mechanical sealing process was done by a screw-nut-tight process by placing an O-ring between the lead and base of the chamber (Figure A.6(a)). Note that the pressure chamber was connected with a gas inlet and a pressure gauge. Experimental results showed that for a pressure range of 0-310mmHg, liquid displacement increased by 1.5 times from 0-21μm to 0-32μm as the diaphragm area increased by 4 times from 100×100 to 100×400μm2 (Figure A.6(b)). However, liquid displacement increased by 9.6 (b) Optical measurement 350 (a) Testing setup To inlet Displacement (μm) To pressure gauge 300 250 lc=400μm lb=1460μm lb lc lc=400μm lb=986μm 200 150 lc=400μm lb=428μm 100 lc=400μm lb=375μm 50 O-ring 0 0 50 100 150 200 250 Applied pressure(mmHg) 300 lc=100μm lb=375μm Figure A.6. Optical measurement of pressure sensing stent: (a) testing setup and (b) characterization of optically measured liquid displacement. 114 times from 0-32 to 0-308μm when the length of the trapped air increased by 3.9 times from 375μm to 1460μm. We chose the diaphragm area as 100×400μm2 and the trapped air length as 600μm because it was also observed that at higher outlet channel length the trapped air formed discrete air-liquid plug inside the channel. A.4.2 Capacitance Measurement Method To measure the capacitance change while liquid moves from one electrode to another, a capacitance to digital converter IC (AD 7746) was used (Figure A.7(a)). Note that this IC can measure capacitance in the range of ±4pF with a resolution of 4aF. At first, the pressure sensing stent coil was placed inside the pressure chamber then the two terminals of the capacitor were connected with the IC via two feed-through connections. (a) Capacitance measurement (b) COMSOL simulation of capacitance 0.43 Volt Liquid block εr=42 Capacitance to digital converter IC Electrode (16×20×0.3 μm3) Δ= 6 fF 0.42 0.415 Δ = 4.5 fF Δ = 3.5 fF 0.41 0.405 0.4 0 Δ = 2.7 fF Mean ΔC=4.5 fF 50 38 60 100 82 Liquid displacement (μm) Capacitance (fF) Capacitance (Picofarad) 0.425 Mean ΔC=4.3 fF 150 104 Figure A.7. Capacitance measurement of pressure sensing stent: (a) capacitance change from one electrode to another and (b) COMSOL simulation of capacitance change. 115 Experimental results showed capacitance increased step by step manner with average step height of 4.5fF. A COMSOL simulation was also performed to mimic the experiment. For COMSOL model, the electrode dimension was chosen as 16×20×0.3μm3, the liquid block dimension was chosen as 16×20×0.3μm3, and the liquid permittivity was selected as 42. Simulation results also showed similar step height of 4.3fF capacitance change when liquid moved from electrode to electrode (Figure A.7(b)). A.4.3. Wireless Ex-vivo Method To detect the pressure wirelessly, the pressure sensing stent coil was placed inside a mock artery tube, and liquid was flown through it to mimic the blood flow (Figure. A.8). A digital pressure transducer (0-765mmHg) was connected as T-connection with the water flow tube. The setup was directly connected with the laboratory DI water supply from where water flow can be controlled. To apply pressure, the tube was constricted by tightening the screw of the worm gear hose clamp. An external planar Pressure transducers -68 0 mmHg 51.7 mmHg 103.4 mmHg 155.1 mmHg 206.9 mmHg 577.3 MHz 579.8 MHz Phase (degree) -68.5 -69 -69.5 Flow in Stent inside tube -70 -70.5 Sensitivity ~1.93 kHz/mmHg -71 5.74 5.76 5.78 5.8 Resonant frequency(Hz) 5.82 8 x 10 Figure A.8. Wireless testing setup and change of resonance frequency with pressure. 116 inductive coil was placed close to the tube, and the coil was connected with an Agilent Network Analyzer 5061B (100k-3GHz). Experimental results showed that the resonant frequency occurred at 579.8MHz with a resolution of 18.1kHz/mmHg (Figure A.8). Previously measured inductance of the stent coil was 250nH. According to such resonant frequency, the base capacitance can be calculated as 0.3pF. According to the capacitive step size measurement, the calculated resolution is 80kHz/mmHg. REFERENCES [1] M. Roser and H. Ritchie. [Online]. Available: https://ourworldindata.org/indoorair-pollution. [Accessed 3 June 2019]. [2] Brake. [Online]. Available: https://studylib.net/doc/8696488/a-guide-to-breathingrates-in-confined-environments. [Accessed 3 June 2019]. [3] B. Fischlowitz-Roberts, "Air pollution fatalities now exceed traffic fatalities by 3 to 1," Eco-Economy Update, 2002. [4] C. Godwin and S. Batterman, "Indoor air quality in Michigan schools," Indoor Air, vol. 17, pp. 109-121, 2007. [5] H. Ritchie and M. Roser. [Online]. Available: https://ourworldindata.org/airpollution. [Accessed 3 June 2019]. [6] J. A. Koziel and J. Pawliszyn, "Air sampling and analysis of volatile organic compounds with solid phase microextraction," J. Air Waste Manage., vol. 51, no. 2, pp. 173-184, 2001. [7] J. A. Koziel and I. Novak, "Sampling and sample-preparation strategies based on solid-phase microextraction for analysis of indoor air," TRAC-Trend Anal. Chem., vol. 21, no. 12, pp. 840-850, 2002. [8] Z. T. Fan, X. Zhu, K. H. Jung, P. Ohman-Strickland, C. P. Weisel, and P. J. Lioy, "Exposures to volatile organic compounds (VOCs) and associated health risks of socio-economically disadvantaged population in a “hot spot” in Camden, New Jersey," Atmos. Environ., vol. 57, pp. 72-79, 2012. [9] EPA, "Hazardous Air Pollutants," [Online]. Available: https://www.epa.gov/haps/initial-list-hazardous-air-pollutants-modifications. [Accessed 16 7 2019]. [10] AQT, "Complete list of VOC's," [Online]. Available: http://aqtvru.com/emissions/complete-list-of-vocs/. [Accessed 17 7 2019]. 118 [11] W. H. Organization, "WHO guidelines for indoor air quality: selected pollutants," World Health Organization, Regional Office for Europe, 2010. [12] W. H. Organization, "Ambient air pollution: A global assessment of exposure and burden of disease," World Health Organization, 2016. [13] A. P. Jones, "Indoor air quality and health," Atmos. Environ., vol. 33, no. 28, pp. 4535-4564, 1999. [14] I. Myers and R. L. Maynard, "Polluted air—outdoors and indoors," Occup. Med., vol. 55, no. 6, pp. 432-438, 2005. [15] L. P. Naeher, B. P. Leaderer, and K. R. Smith, "Particulate matter and carbon monoxide in highland Guatemala: Indoor and outdoor levels from traditional and improved wood stoves and gas stoves," Indoor Air, vol. 10, no. 3, pp. 200-205, 2000. [16] M. D. Hsieh and E. T. Zellers, "Limits of recognition for simple vapor mixtures determined with a microsensor array," Anal. Chem., vol. 76, no. 7, pp. 1885-1895, 2004. [17] J. Cooper, L. Miles, and N. D. Spadafora, "Completely cryogen-free monitoring of PAMS ozone precursors, TO-15 air toxics and OVOCs in ambient air in a single run," Markes International Ltd., Llantrisant, UK, 2018. [18] S. Reidy, D. George, M. Agah, and R. Sacks, "Temperature-programmed GC using silicon microfabricated columns with integrated heaters and temperature sensors," Anal. Chem., vol. 79, no. 7, pp. 2911-2917, 2007. [19] S. K. Kim, H. Chang, and E. T. Zellers, "Microfabricated gas chromatograph for the selective determination of trichloroethylene vapor at sub-parts-per-billion concentrations in complex mixtures," Anal. Chem., vol. 83, no. 18, pp. 7198-7206, 2011. [20] G. R. Lambertus, C. S. Fix, S. M. Reidy, R. A. Miller, D. Wheeler, E. Nazarov, and R. Sacks, "Silicon microfabricated column with microfabricated differential mobility spectrometer for GC analysis of volatile organic compounds," Anal. Chem., vol. 77, no. 23, pp. 7563-7571, 2005. [21] L. Li, T. C. Chen, Y. Ren, P. I. Hendricks, R. G. Cooks, and Z. Ouyang, "Mini 12, Miniature Mass Spectrometer for Clinical and Other Applications - Introduction and Characterization," Anal. Chem., vol. 86, no. 6, pp. 2909-2916, 2014. 119 [22] E. P. Agency, "Photoionization detector (PID) HNU," RI DEM, Providence, USA, 1994. [23] D. S. Ballantine Jr, R. M. White, S. J. Martin, A. J. Ricco, E. T. Zellers, G. C. Frye, and H. Wohltjen, Acoustic Wave Sensors: Theory, Design and PhysicoChemical Applications, California: Academic Press Inc., 1996. [24] A. Afzal, N. Iqbal, A. Mujahid, and R. Schirhagl, "Advanced vapor recognition materials for selective and fast responsive surface acoustic wave sensors: A review," Anal. Chim. Acta, vol. 787, pp. 36-49, 2013. [25] E. Chevallier, E. Scorsone, H. A. Girard, V. Pichot, D. Spitzer, and P. Bergonzo, "Metalloporphyrin-functionalised diamond nano-particles as sensitive layer for nitroaromatic vapours detection at room-temperature," Sens. Actuators B Chem., vol. 151, no. 1, pp. 191-197, 2010. [26] A. T. Nimal, U. Mittal, M. Singh, M. Khaneja, G. K. Kannan, J. C. Kapoor, and D. C. Gupta, "Development of handheld SAW vapor sensors for explosives and CW agents," Sens. Actuators B Chem., vol. 132, no. 2, pp. 399-410, 2009. [27] "TradewaysUSA," [Online]. Available: http://www.tradewaysusa.com/CatalogueNew/Detection_&_Identification/Chemical/Hazmatcad/Hazmatcad.pdf. [Accessed 15 3 2019]. [28] E. J. Staples, T. Matsuda, and S. Viswanathan, "Real time environmental screening of air, water and soil matrices using a novel field portable GC/SAW system," in In Environmental Strategies for the 21st Century, Asia Pacific Conference, Singapore, 1998. [29] "Electronic Sensor Technology," [Online]. Available: https://www.estcal.com/. [Accessed 15 3 2019]. [30] J. W. Gardner and P. N. Bartlett, Electronic Noses. Principles and Applications, New York: Oxford University Press, 1999. [31] M. C. Lonergan, E. J. Severin, B. J. Doleman, S. A. Beaber, R. H. Grubbs, and N. S. Lewis, "Array-based vapor sensing using chemically sensitive, carbon black− polymer resistors," Chem. Mat., vol. 8, no. 9, pp. 2298-2312, 1996. [32] H. Wohltjen and A. W. Snow, "Colloidal metal− insulator− metal ensemble chemiresistor sensor," Anal. Chem., vol. 70, no. 14, pp. 2856-2859, 1998. [33] C. J. Lu, J. Whiting, R. D. Sacks, and E. T. Zellers, "Portable gas chromatograph with tunable retention and sensor array detection for determination of complex 120 vapor mixtures," Anal. Chem., vol. 75, no. 6, pp. 1400-1409, 2003. [34] X. Mu, E. Covington, D. Rairigh, C. Kurdak, E. Zellers, and A. J. Mason, "CMOS monolithic nanoparticle-coated chemiresistor array for micro-scale gas chromatography," IEEE Sens. J., vol. 12, no. 7, pp. 2444-2452, 2012. [35] C. E. Davis, C. K. Ho, R. C. Hughes, and M. L. Thomas, "Enhanced detection of m-xylene using a preconcentrator with a chemiresistor sensor," Sens. Actuators B Chem., vol. 104, no. 2, pp. 207-216, 2005. [36] F. C. Harun, J. E. Taylor, J. A. Covington, and J. W. Gardner, "An electronic nose employing dual-channel odour separation columns with large chemosensor arrays for advanced odour discrimination," Sens. Actuators B Chem., vol. 141, no. 1, pp. 134-140, 2009. [37] "Sandia," [Online]. Available: https://www.sandia.gov/mesa/_assets/documents/Fact_Sheets/sensors/2chemiresis tor.pdf. [Accessed 15 3 2019]. [38] "Adsistor," [Online]. Available: http://www.adsistor.com/home.html. [Accessed 15 3 2019]. [39] V. E. Bochenkov and G. B. Sergeev, "Sensitivity, selectivity, and stability of gassensitive metal-oxide nanostructures," Metal Oxide Nanostruct. App., vol. 3, pp. 31-52, 2010. [40] R. Jaaniso and O. K. Tan, Semiconductor Gas Sensors, Cambridge: Woodhead Publishing Limited, 2013. [41] S. Z. Ali, F. Udrea, W. I. Milne, and J. W. Gardner, "Tungsten-based SOI microhotplates for smart gas sensors," J. Microelectromech. S., vol. 17, no. 6, pp. 1408-1417, 2008. [42] J. B. Sanchez, F. Berger, M. Fromm, and M. H. Nadal, "A selective gas detection micro-device for monitoring the volatile organic compounds pollution," Sens. Actuators B Chem., vol. 119, no. 1, pp. 227-233, 2006. [43] "Microsens," [Online]. Available: http://microsens.ch/products/MSGS.htm . [Accessed 15 3 2019]. [44] "Alphaphase," [Online]. Available: http://www.alphasense.com/index.php/air/downloads/. [Accessed 15 March 2019]. 121 [45] K. Reddy, Y. Guo, J. Liu, W. Lee, M. K. K. Oo, and X. Fan, "On-chip Fabry– Pérot interferometric sensors for micro-gas chromatography detection," Sens. Actuators B Chem., vol. 159, no. 1, pp. 60-65, 2011. [46] J. H. Seo, J. Liu, X. Fan, and K. Kurabayashi, "Fabry-Pérot cavity sensor-based optofluidic gas chromatography using a microfabricated passive preconcentrator/injector," Lab Chip, vol. 13, no. 5, pp. 851-859, 2013. [47] K. Scholten, X. Fan, and E. T. Zellers, "A microfabricated optofluidic ring resonator for sensitive, high-speed detection of volatile organic compounds," Lab Chip, vol. 14, no. 19, pp. 3873-3880, 2014. [48] J. A. Dziuban, J. Mroz, M. Szczygielska, M. Małachowski, A. Gorecka-Drzazga, R. Walczak, and P. Kowalski, "Portable gas chromatograph with integrated components," Sensor Actuat. A-Phys., vol. 115, no. 2-3, pp. 318-330, 2004. [49] J. G. Sevcik, Detectors in Gas Chromatography, vol. 4, New York: Elsevier, 2011. [50] S. Nair, "MicroGC: of Detectors and their Integration," PhD Dissertation, Department of Eelectrical Engineering, Virginia Tech, Blackburg, VA, 2014. [51] S. Narayanan, B. Alfeeli, and M. Agah, " Two-port static coated micro gas chromatography column with an embedded thermal conductivity detector," IEEE Sens. J., vol. 12, no. 6, pp. 1893-1900, 2012. [52] S. Narayanan and M. Agah, "Fabrication and characterization of a suspended TCD integrated with a gas separation column," J. Microelectromech. S., vol. 22, no. 5, pp. 1166-1173, 2013. [53] B. C. Kaanta, H. Chen, and X. Zhang, "Flow rate insensitive thermal conductivity detector," in 16th International Solid-State Sensors, Actuators and Microsystems Conference, Beijing, 2011. [54] B. C. Kaanta, H. Chen, G. Lambertus, W. H. Steinecker, O. Zhdaneev, and X. Zhang, "High sensitivity micro-thermal conductivity detector for gas chromatography," in 2009 IEEE 22nd International Conference on Micro Electro Mechanical Systems, Sorrento, Italy, 2009. [55] R. E. Pecsar, R. B. DeLew, and K. R. Iwao, "Performance of a reduced volume thermal conductivity detector," Anal. Chem., vol. 45, no. 13, pp. 2191-2198, 1973. [56] A. Mahdavifar, M. Navaei, P. J. Hesketh, M. Findlay, J. R. Stetter, and G. W. Hunter, "Transient thermal response of micro-thermal conductivity detector 122 (µTCD) for the identification of gas mixtures: An ultra-fast and low power method," Microsyst. Nanoeng., vol. 1, p. 15025, 2015. [57] M. Ueda, "Thermal conductivity detector," U.S. Patent and Trademark Office, Washington, DC, 2003. [58] S. Zimmermann, S. Wischhusen, and J. Müller, "Micro flame ionization detector and micro flame spectrometer," Sens. Actuators B Chem., vol. 63, no. 3, pp. 159166, 2000. [59] "Raesystem," [Online]. Available: https://www.raesystems.com/sites/default/files/content/resources/pid_handbook_1 002-02.pdf. [Accessed 15 3 2019]. [60] A. C. Lewis, J. F. Hamilton, C. N. Rhodes, J. Halliday, K. D. Bartle, P. Homewood, R. J. P. Grenfell, B. Goody, A. M. Harling, P. Brewer, G. Vargha, and M. J. T. Milton, "Microfabricated planar glass gas chromatography with photoionization detection," J. Chromatogr. A, vol. 1217, no. 5, pp. 768-774, 2010. [61] M. Akbar, M. Restaino, and M. Agah, "Chip-scale gas chromatography: From injection through detection," Microsyst. Nanoeng., vol. 1, p. 15039, 2015. [62] J. P. Hauschild, E. Wapelhorst, and J. Müller, "Mass spectra measured by a fully integrated MEMS mass spectrometer," Int. J. Mass Spectrom., vol. 264, no. 1, pp. 53-60, 2007. [63] M. Yang, T. Y. Kim, H. C. Hwang, S. K. Yi, and D. H. Kim, "Development of a palm portable mass spectrometer," J. Am. Soc. Mass Spectr., vol. 19, no. 10, pp. 1442-1448, 2008. [64] W. Sutherland, "The viscosity of gases and molecular force," London, Edinburgh, and Dublin Philosop. Mag. J. Sci., vol. 36, no. 223, pp. 507-531, 1983. [65] E. L. Cussler, Diffusion: Mass Transfer in Fluid Systems, New York: Cambridge University Press, 2009. [66] F. F. Abdelall, G. Hahn, S. M. Ghiaasiaan, S. I. Abdel-Khalik, S. S. Jeter, M. Yoda, and D. L. Sadowski, "Pressure drop caused by abrupt flow area changes in small channels," Exp. Therm. Fluid Sci., vol. 29, no. 4, pp. 425-434, 2005. [67] W. M. Kays, "Loss Coefficients for Abrupt Changes in Flow Cross Section," Trans. ASME, vol. 72, pp. 1067-1074, 1950. 123 [68] A. Agrawal, L. Djenidi, and R. A. Antonia, "Simulation of gas flow in microchannels with a sudden expansion or contraction," J. Fluid Mech., vol. 530, pp. 135-144, 2005. [69] G. L. Morini, M. Spiga, and P. Tartarini, "The rarefaction effect on the friction factor of gas flow in microchannels," Superlattices Microst., vol. 35, no. 3-6, pp. 587-599, 2004. [70] E. H. Kennard, Kinetic Theory of Gases, with an Introduction to Statistical Mechanics, New York and London: McGraw-Hill Book Company, 1938. [71] S. E. Turner, L. C. Lam, M. Faghri, and O. J. Gregory, "Experimental investigation of gas flow in microchannels," J. Heat Trans., vol. 126, no. 5, pp. 753-763, 2004. [72] G. M. Fryer, "A theory of gas flow through capillary tubes," Proceedings of the Royal Society of London. Series A. Mathematical and Physical Sciences, vol. 293, no. 1434, pp. 329-341, 1966. [73] A. Beskok and G. E. Karniadakis, "Report: A model for flows in channels, pipes, and ducts at micro and nano scales," Nanosc. Microsc. Therm., vol. 3, no. 1, pp. 43-77, 1999. [74] Z. Yang and S. V. Garimella, "Rarefied gas flow in microtubes at different inletoutlet pressure ratios," Phys. Fluids, vol. 21, no. 5, p. 052005, 2009. [75] E. B. Arkilic, M. A. Schmidt, and K. S. Breuer, "Gaseous slip flow in long microchannels," J. Microelectromech. S., vol. 6, no. 2, pp. 167-178, 1997. [76] L. Bai, J. Smuts, P. Walsh, H. Fan, Z. Hildenbrand, D. Wong, W. Wetz, and K. A. Schug, "Permanent gas analysis using gas chromatography with vacuum ultraviolet detection," J. Chromatogr. A, vol. 1388, pp. 244-250, 2015. [77] C. Ho, Properties of Inorganic and Organic Fluids, CINDAS Data Series on Materials Properties, New York: Hemisphere Publishing Corp, 1988. [78] P. Garstecki, M. J. Fuerstman, H. A. Stone, and G. M. Whitesides, "Formation of droplets and bubbles in a microfluidic T-junction—scaling and mechanism of break-up," Lab Chip, vol. 6, no. 3, pp. 437-446, 2006. [79] P. Garstecki, A. M. Ganan-Calvo, and G. M. Whitesides, "Formation of bubbles and droplets in microfluidic systems," Tech. Sci., vol. 53, no. 4, pp. 361-372, 2005. 124 [80] M. J. Fuerstman, A. Lai, M. E. Thurlow, S. S. Shevkoplyas, H. A. Stone, and G. M. Whitesides, "The pressure drop along rectangular microchannels containing bubbles," Lab Chip, vol. 7, no. 11, pp. 1479-1489, 2007. [81] R. Sun and T. Cubaud, "Dissolution of carbon dioxide bubbles and microfluidic multiphase flows," Lab Chip, vol. 11, no. 17, pp. 2924-2928, 2011. [82] T. Cubaud, M. Tatineni, X. Zhong, and C. M. Ho, "Bubble dispenser in microfluidic devices," Phys. Rev. E, vol. 72, no. 3, p. 037302, 2005. [83] A. Bulbul and H. Kim, "A bubble-based microfluidic gas sensor for gas chromatographs," Lab Chip, vol. 15, no. 1, pp. 94-104, 2015. [84] P. Garstecki, I. Gitlin, W. DiLuzio, G. M. Whitesides, E. Kumacheva, and H. A. Stone, "Formation of monodisperse bubbles in a microfluidic flow-focusing device," App. Phys. Lett., vol. 85, pp. 2649-2651, 2004. [85] S. K. Crossno, L. H. Kalbus, and G. E. Kalbus, "Determinations of carbon dioxide by titration," J. Chem. Educ., vol. 73, no. 2, pp. 175-176, 1996. [86] J. Yue, G. Chen, Q. Yuan, L. Luo, and Y. Gonthier, "Hydrodynamics and mass transfer characteristics in gas–liquid flow through a rectangular microchannel," Chem. Eng. Sci., vol. 62, no. 7, pp. 2096-2108, 2007. [87] J. Yue, L. Luo, Y. Gonthier, G. Chen, and Q. Yuan, "An experimental study of air–water Taylor flow and mass transfer inside square microchannels," Chem. Eng. Sci., vol. 64, no. 16, pp. 3697-3708, 2009. [88] J. J. Van Deemter, F. J. Zuiderweg, and A. V. Klinkenberg, "Longitudinal diffusion and resistance to mass transfer as causes of nonideality in chromatography," Chem. Eng. Sci., vol. 5, no. 6, pp. 271-289, 1956. [89] C. Bauters, J. M. Lablanche, E. Van Belle, R. Niculescu, T. Meurice, E. P. Mc Fadden, and M. E. Bertrand, "Effects of coronary stenting on restenosis and occlusion after angioplasty of the culprit vessel in patients with recent myocardial infarction," Circulation, vol. 96, no. 9, pp. 2854-2858, 1997. [90] A. K. Nayak, A. Kawamura, R. W. Nesto, G. Davis, J. Jarbeau, C. T. Pyne, D. E. Gossman, T. C. Piemonte, N. Riskalla, and M. S. Chauhan, "Myocardial infarction as a presentation of clinical in-stent restenosis," Circulation, vol. 70, no. 8, pp. 1026-1029, 2006. [91] N. H. J. Pijls and B. De Bruyne, "Coronary pressure measurement and fractional 125 flow reserve," Heart, vol. 80, no. 6, pp. 539-542, 1998. [92] E. Y. Chow, A. L. Chlebowski, S. Chakraborty, W. J. Chappell, and P. P. Irazoqui, "Fully wireless implantable cardiovascular pressure monitor integrated with a medical stent," IEEE Trans. Biomed. Eng., vol. 57, no. 6, pp. 1487-1496, 2010. [93] E. Y. Chow, Y. Ouyang, B. Beier, W. J. Chappell, and P. P. Irazoqui, "Evaluation of cardiovascular stents as antennas for implantable wireless applications," IEEE Trans. Microw. Theory Tech., vol. 57, no. 10, pp. 2523-2532, 2009. [94] K. Takahata, Y. B. Gianchandani, and K. D. Wise, "Micromachined antenna stents and cuffs for monitoring intraluminal pressure and flow," J. Microelectromech. S., vol. 15, no. 5, pp. 1289-1298, 2006. [95] A. R. Mohammadi, M. S. M. Ali, D. Lappin, C. Schlosser, and K. Takahata, "Inductive antenna stent: Design, fabrication and characterization," J. Micromech. Microeng., vol. 23, no. 2, p. 025015, 2013. [96] X. Chen, D. Brox, B. Assadsangabi, Y. Hsiang, and K. Takahata, "Intelligent telemetric stent for wireless monitoring of intravascular pressure and its in vivo testing," Biomed. Microdevices, vol. 16, no. 5, pp. 745-759, 2014. [97] X. Chen, D. Brox, B. Assadsangabi, M. S. M. Ali, and K. Takahata, "A stainlesssteel-based implantable pressure sensor chip and its integration by microwelding," Sensors Actuat. A-Phys, vol. 257, pp. 134-144, 2017. [98] X. Chen, B. Assadsangabi, D. Brox, Y. Hsiang, and K. Takahata, "A pressuresensing smart stent compatible with angioplasty procedure and its in vivo testing," in 2017 IEEE 30th International Conference on Micro Electro Mechanical Systems, Las Vegas, USA, 2017. [99] A. D. DeHennis and K. D. Wise, "A fully integrated multisite pressure sensor for wireless arterial flow characterization," J. Microelectromech. S., vol. 15, no. 3, pp. 678-685, 2006. [100] S. R. Green and Y. B. Gianchandani, "Wireless magnetoelastic monitoring of biliary stents," J. Microelectromech. S., vol. 18, no. 1, pp. 64-78, 2009. [101] M. N. Gulari, M. Ghannad-Rezaie, P. Novelli, N. Chronis, and T. C. Marentis, "An implantable X-ray-based blood pressure microsensor for coronary in-stent restenosis surveillance and prevention," J. Microelectromech. S., vol. 24, no. 1, pp. 50-61, 2015. 126 [102] P. J. Chen, Rodger, S. S. D. C., M. S. Humayun, and Y. C. Tai, "Microfabricated implantable parylene-based wireless passive intraocular pressure sensors," J. Microelectromech. S., vol. 17, no. 6, pp. 1342-1351, 2008. [103] T. Bourouina and J. P. Grandchamp, "Modeling micropumps with electrical equivalent networks," J. Micromech. Microeng., vol. 6, no. 4, p. 398, 1996. [104] R. L. Bardell, N. R. Sharma, F. K. Forster, M. A. Afromowitz, and R. J. Penney, "Designing high-performance micro-pumps based on no-moving-parts valves," in in 1997 ASME International Mechanical Enginering Congress and Exposition, Dallas, USA, 1997. [105] B. Mosadegh, C. H. Kuo, Y. C. Tung, Y. S. Torisawa, T. Bersano-Begey, H. Tavana, and S. Takayama, "Integrated elastomeric components for autonomous regulation of sequential and oscillatory flow switching in microfluidic devices," Nat. Phys., vol. 6, no. 6, p. 433, 2010. [106] Q. Yang, P. Kobrin, C. Seabury, S. Narayanaswamy, and W. Christian, "Mechanical modeling of fluid-driven polymer lenses," Appl. Opt., vol. 47, no. 20, pp. 3658-3668, 2008. [107] N. Srivastava and M. A. Burns, "Microfluidic pressure sensing using trapped air compression," Lab Chip, vol. 7, no. 5, pp. 633-637, 2007. [108] J. Lee, F. Rahman, T. Laoui, and R. Karnik, "Bubble-induced damping in displacement-driven microfluidic flows," Phys. Rev. E, vol. 86, no. 2, p. 026301, 2012. [109] J. Z. Chen, A. A. Darhuber, S. M. Troian, and S. Wagner, "Capacitive sensing of droplets for microfluidic devices based on thermocapillary actuation," Lab Chip, vol. 4, no. 5, pp. 473-480, 2004. [110] O. Akar, T. Akin, and K. Najafi, "A wireless batch sealed absolute capacitive pressure sensor," Sensors Actuat. A-Phys., vol. 95, no. 1, pp. 29-38, 2001.