Fluid Phase Measurement Using Optical, Microfluidic and Nanofluidic Methods

Fluid Phase Measurement using Optical, Microfluidic and Nanofluidic Methods

Bo Bao

A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy Mechanical & Industrial Engineering University of Toronto

Fluid Phase Measurement using Optical, Microfluidic and Nanofluidic Methods

Bo Bao

Doctor of Philosophy

Mechanical & Industrial Engineering University of Toronto

2016

Abstract

Understanding fluid phase behavior is essential to a wide range of applications, including oil and

gas recovery, chemical reactor engineering, transport and storage of natural gas and carbon

dioxide, and supercritical fluid processing and extraction. In this thesis, novel experimental

methods – optical, microfluidic and nanofluidic - are developed to measure and understand fluid

phase behaviors for carbon dioxide transport/storage and shale gas/oil production. (i) Optical

thin-film interference based bubble and dew point sensor probe: The sensor probe within a small

pressure-volume-temperature (PVT) system offers accurate (< 5% error) and responsive

measurement (1-to-2 orders faster than the conventional method) of bubble and dew point of

both pure fluids and mixtures up to 80 oC and 10 MPa. This approach also allows in situ

measurement of the thickness of condensed liquid film to 1 µm accuracy. (ii) Refractive index based optical fiber sensor: This approach takes advantage of the sharp refractive index difference

between different phases. The optical fiber successfully distinguishes supercritical CO2 and brine at sequestration pressure and temperature conditions. In addition, the CO2-saturated brine is

detectable relative to unsaturated brine – a minute refractive index difference. (iii) Multiplexed

microfluidic-based phase diagram mapping: Demonstrated here is the direct measurement of the ii full Pressure-Temperature phase diagram with 10,000 microwells. The method is tested with a pure fluid and a fluid mixture. Liquid, vapor and supercritical regions are clearly differentiated, and the critical point is measured within 1.2% error on a single chip. This method provides 100- fold improvement in measurement speed over conventional methods. (iv) Nanofluidics-based measurement of bubble nucleation and growth in nanochannels: A nanofluidic platform is developed to investigate vapor bubble nucleation and growth in a pure hydrocarbon confined in sub-100-nm channels. Measured nucleation conditions in the nanochannels are compared with those predicted from the nucleation theory. In addition, different types of bubble growth dynamics are observed and analyzed. Collectively these contributions leverage optics and microfluidics to develop fast and accurate fluid phase measurement methods, and leverage nanofluidics to study the unique effects of nanoconfinement on fluid phase behavior.

iii

Acknowledgments

It is a great joy for me to express my appreciation to all of you who have supported me academically or spiritually during my PhD journey.

Foremost, I would like to express my deep and sincere gratitude to my academic supervisor Professor David Sinton for his incalculable guidance and support to my PhD study and research. It is his intelligence, enthusiasm, patience, encouragement and profound knowledge that helped me achieving the honorable degree. Thank you to my supervisor.

Also, I would like to thank all my thesis committee: Professor Anthony Sinclair, Professor Markus Bussmann and Professor Nikos Varotsis, for their insightful comments and suggestions to my PhD thesis. And, I would like to thank Professor Peter Wild from University of Victoria for his co-supervision on the work in Chapter 4 and Dr. Farshid Mostowfi from Schlumberger Doll-Research Center (moved from Schlumberger DBR Edmonton) for the collaboration on the work in Chapter 6.

Moreover, I would express my thanks to all my colleagues for their help to my PhD work. Specially, thanks to Dr. Hossein Fadaei for his expertise in petroleum and his help to my research in first two years. My sincere thanks also go to Dr. Jason Riordon, Dr. Huawei Li and Dr. Hadi Zandavi who worked closely with me and supported my research work in last two years. In addition, thanks to three summer students who assisted my work, Haiyi Wang, Japinder Nijjer and Yi Xu.

Finally and importantly, I would like to express my deep appreciation to my parents, Pingyuan Bao and Junqing Shi, for their endless and invaluable support throughout my PhD journey and my life. I could not make the achievement without them. Thank you to my parents.

Table of Contents

Contents

Acknowledgments...... iv

Table of Contents ...... v

List of Tables ...... viii

List of Figures ...... ix

List of Appendices ...... xviii

1 Thesis Overview ...... 1

1.1 Research motivation ...... 1

1.1.1 Carbon transport and sequestration ...... 1

1.1.2 Shale gas/oil production ...... 4

1.2 Thesis structure ...... 6

2 Introduction ...... 8

2.1 Fluid phases ...... 8

2.2 Experimental methods to measure fluid phase ...... 12

2.2.1 PVT experiments ...... 14

2.2.2 Optical methods ...... 17

2.2.3 Electrical and acoustic methods ...... 19

2.2.4 Microfluidic methods ...... 20

2.2.5 Nanofluidic methods ...... 23

3 Detection of Bubble and Dew Point using Optical Thin-film interference ...... 29

3.1 Introduction ...... 29

3.2 Experimental ...... 31

3.2.1 Experimental setup...... 31

3.2.2 Sensing mechanism: thin-film interference ...... 32

3.2.3 Experimental procedure ...... 35

3.3 Results and discussion ...... 35

3.3.1 Method validation with pure CO2 ...... 35

3.3.2 Application to industrial CO2 mixtures containing impurities ...... 40

3.3.3 Comparison with existing methods of gas mixture properties ...... 43

3.3.4 Film thickness determination ...... 45

3.3.5 Resolution ...... 45

3.3.6 Repeatability ...... 46

3.4 Conclusion ...... 46

3.5 Supplemental material ...... 47

4 Detecting Supercritical CO2 in Brine at Sequestration Pressure with an Optical Fiber Sensor ...... 50

4.1 Introduction ...... 50

4.2 Experimental ...... 52

4.2.1 Optical fiber sensor ...... 52

4.2.2 High-pressure apparatus...... 54

4.2.3 Characterization of the optical fiber sensor ...... 56

4.2.4 Detection of scCO2 relative to brine ...... 56

4.2.5 Detection of CO2-saturated brine relative to brine ...... 57

4.3 Results and discussion ...... 58

4.3.1 Characterization of the optical fiber sensor ...... 58

4.3.2 Detection of scCO2 relative to brine ...... 59

4.3.3 Detection of CO2-saturated brine relative to brine ...... 62

4.4 Implications...... 65

4.5 Supplemental material ...... 65

5 Direct Measurement of the Fluid Phase Diagram using Multiplexed Microfluidics ...... 67

5.1 Introduction ...... 67

5.2 Measurement of pressure-temperature phase diagram of pure CO2 ...... 69

5.3 Measurement of pressure-temperature phase diagram of 95% CO2 + 5% N2 mixture ...... 73

5.4 Phase-mapping device accuracy and speed ...... 74

5.5 Conclusion ...... 75

5.6 Supporting information ...... 75

6 Bubble Nucleation and Growth in Nanochannels ...... 78

6.1 Introduction ...... 78

6.2 Experimental setup...... 80

6.3 Results and discussion ...... 81

6.3.1 Bubble nucleation ...... 81

6.3.2 Bubble growth ...... 84

6.4 Conclusion ...... 90

6.5 Supplemental material ...... 91

7 Conclusions ...... 98

7.1 Fluid phase measurement using optical methods...... 98

7.2 Fluid phase diagram mapping using multiplexed microfluidics ...... 99

7.3 Fluid phase change in nanochannels ...... 99

7.4 Outlook ...... 100

References ...... 101

Appendix 1 ...... 110

vii

List of Tables

Table 3-1. Dew points and bubble points of pure CO2 ...... 48

Table 3-2. Dew points and bubble points of impure CO2 ...... 49

Table 6-1. Bubble nucleation conditions ...... 91

viii

List of Figures

Figure 1-3. SEM image of Fayetteville shale. A) micro-porosity (scale bar = 5 µm); B) organic matter and nano-porosity (scale bar = 500 nm); C) nano natural factures (scale bar = 1 µm); D) pore size histogram; E) Hydraulic fracturing. Reproduced with permission from Elsevier, [16] copyright 2013...... 5

Figure 2-2. Phase diagrams of pure substance. a) Temperature-Volume (T-v) diagram, b) Pressure-Volume (P-v) diagram, and c) Pressure-Temperature (P-T) diagram. Reproduced with permission from McGraw Hill,[24] copyright 2012...... 9

Figure 2-3. Pressure-Temperature phase diagram of A) water, and B) carbon dioxide; C) Images o of CO2 phases at different Pressures and Temperatures. The critical point of CO2 is 31.0 C and 7.38 MPa. Reproduced with permission from Elsevier,[25] copyright 2014...... 11

Figure 2-4. P-T phase envelope of natural gas. The circle stands for critical point. Composition of natural gas mixture is listed at right. Reproduced with permission from Taylor & Francis Group,[9] copyright 2015...... 12

Figure 2-5. A) PVT Analytical-isothermal methods for phase measurement; B) Synthetic method (visual and non-visual). Reproduced with permission from Annual Review of Chemical and Biomolecular engineering,[10] copyright 2013...... 13

ix differential depletion experiment for oil; E) Schematics of constant-volume depletion experiment for a gas condensate. Reproduced with permission from Taylor & Francis Group,[9] copyright 2015...... 15

Figure 2-7. Optical methods: A) Schematics of automatic chilled mirror method;[28] B) Principle of Tunable Diode Laser Absorption Spectroscopy (TDLAS);[29] C) Schematics of Fabry-Perot hygrometer; [28] D) Schematics of fiber-optic reflectometer;[30] Reproduced with permission from GE Measurement & Control, Ametek Process Instruments, and, AIP Publishing LLC, copyright 2006...... 17

Figure 2-8. Electrical methods: A) Capacitance probe and operating principle; B) Quartz-crystal microbalance (QCM) sensors and operating principle.[29] Reproduced with permission from Ametek Process Instruments, copyright 2011...... 19

Figure 2-9. A microfluidic PVT system. A) Concept of bubble point measurement: bubble forms below bubble point pressure at the inlet restriction. B) Image sequence of bubble evolution at the inlet restriction (5000 fps). C) Image of the microfluidic PVT system in operation, showing bubble and liquid slugs. Reproduced with permission from Royal Society of Chemistry,[11] copyright 2012...... 21

Figure 2-10. A microfluidic system investigating thermodynamics at high pressures and temperatures. A) Schematics of the microfluidic system with the key features of the serpentine channel. B) Analogy of two operation modes: “continuous flow” and “dynamic stopflow mode”. C) Measurement of phase envelop with detected bubble point and dew point. Reproduced with permission from Royal Society of Chemistry,[35] copyright 2014...... 22

Figure 2-11. Two microfluidic systems measuring the phase behavior of multicomponent

samples. A) Detection of water dew point in CO2 stream. B) Measurement of the minimum

Figure 2-13. A) Nanochannel device including sets of nanochannels; B) Cavitation induced by deformed meniscus and bubble entrapment at the nanochannel entrance; C) Image sequence of cavitation in 58-nm nanochannels. Reproduced with permission from Proceedings of the National Academy of Sciences of the United States of America,[23] copyright 2012...... 26

Figure 2-14. Cavitation test of water in nano-porous media. A) Top view and B) Side view schematics of ink-bottle structure filled with water; Porous silicon (poSi) is interconnected to these cavities; C) Optical images show evaporation occurred from edges of sample. Reproduced with permission from American Physical Society, [39] copyright 2014...... 27

Figure 2-15. Study of nanoconfinement effect in hydrocarbon production. A) Structure of nanofluidic chip; B) SEM image of nanochannels; C) Experiment of water displacing nitrogen gas; D) Pentane evaporation in nanochannels; E) Evaporation of a ternary mixture stopped in one of the microchannels. Reproduced with permission from Society of Petroleum Engineers,[20][26] copyright 2014...... 28

Figure 3-1. a) Schematic of experimental apparatus integrating gas cylinder, pump, PVT cell, optical interrogator, and computer; b) Internal features of the PVT cell with optical fiber sensors...... 32

Figure 3-2. a) Schematic of single phase fluid surrounding the fiber tip; b) , i.e. Schematic of a liquid film on fiber tip while surrounding medium is vapor phase, i.e., Phase I = liquid, Phase II = vapor. Thin-film interference is an additive effect of a wave reflected at the Core/Phase I interface (dotted curve) and a wave reflected at the Phase I/Phase II (dashed curve). No phase shift results in a constructive wave returning to the detector, where as a 180° wave phase difference returns a destructive wave. Thus, the presence of thin film – indicating vapor-liquid phase transition - can be very accurately determined by the optical fiber sensor via constructive and destructive wave patterns in the reflection spectrum. It is also possible to extract further measurements from this returning signal, such as film thickness...... 33

o Figure 3-3. Detection of bubble and dew point of pure CO2 at 20 C. a) Reflection spectrum of top sensor under decreasing pressure: The interference pattern observed at 54 bar indicates the xi formation of a thin film corresponding to the bubble point; b) Corresponding average power vs. pressure: A large interference fluctuation band at 54 bar (3.57 dBm) indicates the phase transition; c) Reflection spectrum of the bottom sensor under increasing pressure: The interference pattern observed at 55 bar indicates the dew point; d) Corresponding average power vs. pressure: A large interference fluctuation at 55 bar (3.24 dBm) indicates the phase transition. (Note: error bars are included in plots (b) and (d), however they are smaller than the data-point markers; pressure lines in (a) and (c) appear overlapped in single phase and shift significantly during phase transition.)...... 36

Figure 3-4. Detection of critical point of CO2 – lowest temperature at which no interference pattern is observed a) Reflection spectrum from the top sensor under decreasing pressure; b) Corresponding average power vs. pressure; c) Reflection spectrum from the bottom sensor under increasing pressure; d) Corresponding average power vs. pressure. No interference observed in a) or c) indicates the absence of two phase region, and, smooth variation of average power with constant noise-level standard deviations (0.12 - 0.17 dBm) showed in b) and d) implies a smooth phase change characteristic of supercritical behavior. (Note: error bars are included in plots (b) and (d), however they are smaller than the data-point markers) ...... 39

o Figure 3-5. Detection of bubble and dew point of impure CO2 at 20 C. a) Reflection spectrum of top sensor under decreasing pressure: The interference pattern observed at 87 bar indicates the formation of a thin film corresponding to the bubble point; b) Corresponding average power vs. pressure: A large interference fluctuation band at 87 bar (0.84 dBm) indicates the phase transition; c) Reflection spectrum of the bottom sensor under increasing pressure: The interference pattern observed at 67 bar indicates the dew point; d) Corresponding average power vs. pressure: A large interference fluctuation at 67 bar (0.62 dBm) indicates the phase transition. (Note: Pressure lines in (a) and (c) appear overlapped in single phase and shift significantly during phase transition.)...... 41

Figure 3-6. Detection of maxcondentherm point of impure CO2. a) Reflection spectrum from the top sensor under decreasing pressure; b) Corresponding average power vs. pressure; c) Reflection spectrum from the bottom sensor under increasing pressure; d) Corresponding average power vs. pressure. No interference was observed in a) or c), and a smooth variation of

xii the average power with constant deviation (0.12 – 0.19) implies a continuous phase change. The maxcondentherm point was identified at 26 oC and 82 bar...... 42

Figure 3-7. Comparison of bubble and dew points measured and predicted by two models

(REFPROP and PVTsim) for both the case of pure CO2 and impure CO2. The pure CO2 case provides method validation. The dew point data in the impure case shows a good agreement with the models. Both the models and the experimental data deviate substantially for the bubble point of the case with impurities, highlighting the importance of composition-specific testing enabled by the small scale PVT method...... 43

Figure 3-8. Example of bubble point detection with increased resolution using reduced pressure intervals of 0.1 bar. a) Reflection spectrum from the top sensor under decreasing pressure from 88 to 87 bar: The interference pattern observed at 87.4 bar indicates the formation of a thin film corresponding to the bubble point; b) Corresponding average power vs. pressure: A large interference fluctuation band at 87.4 bar indicates the phase transition, with resolution of 0.1 bar...... 45

Figure 3-9. Repeatability test of sensor for bubble and dew point measurement of impure CO2 at 20 oC...... 46

A contrast between the native brine solution and either scCO2 or CO2-saturated brine is detected through a resonance wavelength shift...... 53

Figure 4-2. Schematic of the high-pressure apparatus used to characterize and test the optical

fiber sensor. (a) The setup employed for detecting scCO2 relative to brine solution, and (b) the

contents of the cylinders used for detecting CO2-saturated brine relative to brine. The pressure of 1400 psi (9.65 MPa) was monitored by pressure gauges (G1 to G3) while the temperature of 40 °C was controlled by the water bath...... 54

RIU) at 40 °C, the sensitivity to refractive index (between 1.00 and 1.33) is determined to be 4.847 nm / RIU. b) The resonance wavelength shift as a function of pressure and temperature. The resonance wavelength shift shows linear correlations with both pressure and temperature, with the correlation coefficients (R2) of 0.9881 and 0.9959, respectively. The sensitivities to pressure and temperature are 0.026 nm / 100 psi (0.69 MPa) (1000 - 1800 psi, or 6.89 to 12.41 MPa) and 0.054 nm / °C (35 – 45 °C), respectively...... 58

Figure 4-4. The resonance wavelength shift corresponds to the alternate cycles of test samples

between brine (red rectangles) and scCO2 (green diamonds). Each cycle includes a 10-min time gap for sample replacement and stabilization plus a 5-min period for data collection. A moving average curve is shown as a solid line, and a dashed trend line between data collections is shown as a guide for the eye. The brine and scCO2 are distinguished repeatedly and significantly in terms of the resonance wavelength shift, 1.149 nm as the average value...... 60

minutes from t = 5 min when scCO2-brine solution replaced the initial brine. A detectable

resonance wavelength shift of 0.192 nm is observed for CO2-saturated brine relative to the original brine solution...... 62

Figure 4-6. a) A high-pressure stainless steel chamber (cross-section view) to enclose and immobilize the optical fiber sensor. The chamber is assembled from 1 threaded pipe nipple, 2 threaded pipe fitting Tees and 4 Yor-lok tube fitting adapters as four ports (namely Port A, B, C and D). The optical fiber sensor is tensioned and suspended through the chamber cavity via Port A and B, and, the long period grating section is adjusted to locate at the center of the pipe nipple. b) A tight sealing between the fitting and the optical fiber. The optical fiber passes through a PEEK sleeve and the PEEK sleeve passes through a compression fitting ferrule. The fitting ferrule compresses on the PEEK sleeve when the nut thread is tightened so that the PEEK sleeve shrinks and tightly clamps on the optical fiber...... 66

Figure 5-1. a) Schematic of the microfluidic fluid phase-mapping device. a) Full device featuring a 2D array of micro-wells subject to a vertical pressure gradient and a horizontal temperature xiv gradient. Only a few channels are displayed for clarity; the actual phase-mapping device contains 100 horizontal channels, each with 100 micro-wells. b) Enlarged view of micro-wells...... 69

Figure 5-2． Measurement of the pressure-temperature phase diagram of pure CO2. a) Phase- mapping device in operation, with liquid, vapor and supercritical regions visualized. b) Microscope image of a region of the phase-mapping device with the critical point. Inset images show enlarged views of three liquid-vapor interfaces. c) Pixel intensity profile across corresponding liquid-vapor interfaces. Inset shows how the height-to-width ratio of the pulses changes near the critical point. d) Pressure-temperature phase transition point measurements and validation with NIST reference points. Inset images show typical fluid behavior within micro- wells at various pressure-temperature conditions...... 72

Figure 5-3. Measurement of the fluid phase diagram of a 95 % CO2 + 5 % N2 mixture and comparison to NIST reference. Inset images show typical fluid behavior within micro-wells at various P-T conditions...... 74

Figure 5-4. a) Schematic of the experimental setup. Water was flowed between baths and chiller and heater blocks above the phase-mapping device to produce a temperature gradient. Two pumps (Teledyne Isco) were attached to the inlet and outlet to pressurize the system, and drive single-phase flow through the resistor channel. A piston (filled with the fluid of interest) was used to isolate the first pump from the device (which was filled with water). A custom stainless- steel manifold was used to provide a strong connection to the device. A microscope monitored the phase-mapping device during operation. Pressure transducers measured pressures at the inlet and outlet throughout the experiments. b) Image of the phase-mapping device operating with pure CO2...... 75

Figure 5-5. Temperature characterization of the phase-mapping device: a) Image of temperature calibration test using propane. b) Temperature distribution of phase-mapping device during test of pure CO2 and mixture of 95% CO2 + 5% N2...... 76

Figure 5-6. The response of liquid-vapor interface position to inlet pressure changes. Insets show microscope images of interface positions at various pressure states...... 77

Figure 6-1. a) Schematic of the experimental setup including micro/nanofluidic chip and top view of the chip; b) Side view, and c) Cross-section view of 85-nm deep nanochannels; d) Cross- section view of single nanochannel by SEM imaging...... 81

Figure 6-2. Measurements of bubble nucleation in 85-nm nanochannels. Plotted for comparison are the saturation vapor pressure, capillary pressure calculated from the Kelvin Equation, Eq(6-

1), the prediction from the classical nucleation theory, Eq(6-2) for tN = 900 s, and the spinodal limit from Eq(6-3)...... 82

Figure 6-3． Mechanisms of bubble column growth at Tl = 347 K and Pl = 1.60 MPa: a) Image sequence of Type A growth where vapor bubble nucleates at channel-end; b) Image sequence of

Type B where vapor bubble nucleates along the channel; c) Bubble length lB versus time of Type

A and B growth; d) Positions of left and right liquid-vapor interfaces, lL and lR, bubble length lB of Type B growth. Three distinct growth regimes can be identified: “Transient start-up”, “Transitional” and “Steady linear growth”...... 85

Figure 6-4. a) Vapor bubble column growth lB(t) of Type A at five different temperatures; inset

shows the calculated evaporation rate dNv/dt in steady linear growth regime; b) The calculated

pressure in the vapor phase Pv versus time of the bubble nucleation experiment at T = 347 K. Saturation pressure and the reservoir pressure are also plotted for comparison...... 89

Figure 6-5. Complete data of bubble nucleation time and associated growth types of the ten nanochannels at five nucleation conditions...... 92

Figure 6-6. Complete data of left interface velocity uL in the steady linear growth regime of the ten nanochannels at five nucleation conditions...... 92

Figure 6-7. Complete data of right interface velocity uR of Type B growth in the nanochannels at five nucleation conditions...... 93

Figure 6-8. The relative position of right liquid-vapor interface in the transient start-up regime of the type B growth in the five nucleation conditions. It is clear that the right interface moves linearly with time in the “transient start-up” regime. The inset shows the right interface velocity,

uR, at five nucleation conditions. The uR increases from 0.3868 to 0.8709 µm/ms as the temperature increases from 343 to 362 K...... 93 xvi

Figure 6-9. Mechanisms of bubble column growth at Tl = 343 K and Pl = 1.10 MPa: a) Bubble

length lB versus time of Type A and B growth; b) Positions of left and right liquid-vapor

interfaces, lL and lR, of Type B growth; c) Predicted bubble pressure Pv (Type A) during bubble column growth...... 94

Figure 6-10. Mechanisms of bubble column growth at Tl = 352 K and Pl = 2.00 MPa: a) Bubble

length lB versus time of Type A and B growth; b) Positions of left and right liquid-vapor

interfaces, lL and lR, of Type B growth; c) Predicted bubble pressure Pv (Type A) during bubble column growth...... 95

Figure 6-11. Mechanisms of bubble column growth at Tl = 357 K and Pl = 2.40 MPa: a) Bubble

length lB versus time of Type A and B growth; b) Positions of left and right liquid-vapor

interfaces, lL and lR, of Type B growth; c) Predicted bubble pressure Pv (Type A) during bubble column growth...... 96

Figure 6-12. Mechanisms of bubble column growth at Tl = 362 K and Pl = 2.90 MPa: a) Bubble

length lB versus time of Type A and B growth; b) Positions of left and right liquid-vapor

interfaces, lL and lR, of Type B growth; c) Predicted bubble pressure Pv (Type A) during bubble column growth...... 97

xvii

List of Appendices

Appendix 1: Silicon-glass micro/nanofluidic chip fabrication procedure

xviii 1

1 Thesis Overview

1.1 Research motivation

Fluid phase measurement is crucial in a broad range of processes, including oil and gas recovery, chemical reactor engineering, transport and storage of natural gas and carbon dioxide, and supercritical fluid processing and extraction. The research topics described in Chapter 3 to 6 are motivated by two broad applications outlined below: Carbon transport and sequestration; and shale gas/oil production.

1.1.1 Carbon transport and sequestration

Global CO2 emission increased 0.5 % in 2014 compared to the previous year, reaching 35.5 Gt total emissions.[1] Current emission rates point to a global temperature increase over 5°C by 2100. The global temperature increase would cause serious consequences, including sea level rise, dislocation of human settlement, extreme weathers, reduced food production and human disease. [2] Despite of fast development of sustainable energy, it is clear that the fossil fuel dominant energy portfolio cannot be significantly changed in a short term, particularly considering the fact that fossil fuels contribute to 86.3 % of global energy in 2014 (23.7% Natural gas, 30% Coal and 32.6% Oil).[1] The International Energy Agency indicates that Carbon Capture and Storage (CCS) is a critical component in the portfolio of low-carbon

solutions. CO2 could be captured from large point sources, including gas processing plants, gas or coal power plants, fertilizer plants, ethanol plants and other large industrial emitters (Figure 1-

1). The captured CO2 could be further used for enhanced oil recovery, or stored in depleted oil/gas field and deep saline aquifers (Figure 1-1). Figure 1-1 B assesses 29 global CCS projects, including 16 in North America, 10 in Europe and 3 in rest of the world. Though deep saline aquifers shows tremendous potential for direct storage (100 – 10,000 Gt estimated [3]), the only currently economic carbon sink option is combined storage and enhanced oil recovery by injecting CO2 into oil fields.

CO2 pipelines connect the sources and sinks. To ensure safe and corrosion-free operation, there

are strict specifications for CO2 in terms of delivery composition, pressure and water content.

First, 95% purity of CO2 is usually required considering common impurities of H2O, N2, O2, H2S and CO in the stream. Second, water content is required to be less than 640 ppmv to avoid corrosion – notably any liquid water separated from the mixture quickly becomes acidic due to

the CO2.[4] It is also desirable to have the entire pipeline run as a single supercritical phase because multiple phases cannot be tolerated by pumps and compressors. The presence of

impurities significantly changes the phase behavior of CO2 stream. Therefore, it is critical to

develop and utilize appropriate technology to characterize the phase behavior of industrially-

relevant CO2 streams. This motivates the research topic of developing novel technique to detect

phase boundary for CO2 pipeline mixtures, as described in Chapter 3.

Environmental and health risks are major concerns surrounding carbon sequestration technology. It is crucial to ensure the process of injection and sequestration are safe without any leakage.

Developing technologies to detect and monitor CO2 plumes underground is necessary to ensure

safety. A wide variety of subsurface CO2 monitoring methods have been investigated, including seismic[5], geoelectric [6] and geochemical methods[7]. Optical methods are relatively new and are becoming more and more popular in subsurface tools, due to their inherent advantages including in-situ measurement, immunity to electromagnetic noise, high sensitivity and capacity for distributed sensing. The most successful deployment of optical fiber sensors underground is distributed temperature sensing (DTS) [8]. This recent work motivates the development of an

optical fiber sensor to distinguish CO2 phase from the formation water phase, as described in Chapter 4.

A major limitation of all available phase measurement technologies is that only a single pressure- temperature condition can be measured at once. The most common configuration is the pressure- volume-temperature (PVT) cell, ubiquitous in petrochemical and polymer processing applications.[9][10] These cells typically vary in size between 100 mL and 1 L, and reach pressures and temperatures up to 60 MPa and 150 °C, respectively.[9] Since thermal and chemical equilibrium within these large systems must be reached between measurements, obtaining a single P-T data point typically takes 8 to 10 h [11]. Thus tens to hundreds of hours are required to generate a full phase diagram at a considerable expense. Microfluidic techniques provide a unique solution to integrate a large number of data points in a relatively small space with accurate control of temperature and pressure. Developed here is a novel PVT system which can generate a P-T diagram in a single run with 10,000 parallel microwells, as described in Chapter 5.

1.1.2 Shale gas/oil production

The single most significant change in the global energy system in the last 10 years has been the emergence of shale gas and tight oil, achieved via hydraulic fracturing in horizontal wells. The result has been a flood of hydrocarbons, a rapid drop in oil and gas prices with associated global economic impacts, and the re-emergence of US production long thought to have peaked (peak oil theory). The implications for Canada are profound. Global technically recoverable resources of shale gas and shale/tight oil reached 7,299 trillion cubic feet and 345 billion barrels in 2013 indicated by a report from U.S. Energy Information Administration. The vast resources of hydrocarbons in these shale/tight formations contribute to 10% increase of crude oil and 48% increase of natural gas due to inclusion of shale oil and shale gas. [12] Figure 1-2 shows the global geological distribution of shale gas and oil formation. In contrast to conventional reservoirs, shale/tight formation pores are much smaller, with typical size of tens to hundreds of nanometers. Due to the nano-porosity, the production of these hydrocarbons out of nanopores requires hydraulic fracturing. During hydraulic fracturing, fracking fluid (water + trace amount of proppants) is injected into a wellbore to create artificial cracks in the formations. Once

the hydraulic pressure is removed from the well, the proppants hold the fractures open. Such fractures expose and connect large areas of hydrocarbon-bearing nanopores and hydrocarbon is released to production wells. This is a complex process involving coupled fluid mechanics and nanoscale fluid phase behavior which is poorly understood, a “blank page with huge impact” as described in a recent review[13]. The immediate need for fundamental and applied research in this area is underscored by both a 2014 special issue of Science [14], and a 2014 report by the Council of Canadian Academies[15].

Figure 1-3 A-C) shows typical structure of shale with nanopores by SEM. The pore size distribution of the imaged shale range from 30 to 60 nm, as presented by the histogram in Figure 1-3 D. Hydraulic fracturing process is illustrated in Figure 1-3 E.

Phase behavior of hydrocarbons in nanopores is a key element in understanding this complex physical process. A better understanding of this physics is crucial to both maximizing the energy efficiency of these operations, and quantifying their environmental impact. Previous work on phase behavior in nanoconfinement is mostly numerical and analytical [17][18][19]. However, there are very few experimental studies to validate these results. Nanofluidic methods – developed primarily for biomolecular separations and biomedical applications – can provide a window into fluid phase behavior at the nanoscale. Some in the oil and gas research community have developed nanofluidic systems to study hydrocarbon transport [20][21]. These recent studies show a rich physics, with deviation from classical theory in terms of capillary pressure and interface shape [22][21]. There are relatively few studies, however, a recent article in PNAS shows highly anomalous behavior on water evaporating from a nanochannel in a water-air system [23]. The lack of experimental investigation in nanofluidic motivates the research topic of hydrocarbon liquid-vapor phase change, as detailed in Chapter 6.

It is worthy to mention that, the developed sensors and systems from Chapter 3 to 6 are not limited to the specific motivations, but can be used in a much wider range of applications. Specifically, (i) the optical bubble and dew point sensor (Chapter 3) can be used to characterize the phase envelop of any fluid mixture (pressure-, temperature- and chemical-tolerant) or to measure the thickness of thin film deposition on the order of 1 µm; (ii) the optical fiber sensor (Chapter 4) can be used to distinguish any two phases with a difference of refractive indices down to ~ 0.04 R.I.; (iii) the phase mapping device (Chapter 5) can be utilized in phase diagram mapping of a variety of pure fluids or the measurement of critical points in fluid mixtures within the broad temperature and pressure range of the device; and (iv) the nanofluidic platform (Chapter 6) could be applied to study phase behavior in nanoconfinement of other fluids, e.g. water, organic solvents, refrigerant and other hydrocarbons such as reservoir relevant tight-oil mixtures or fracture fluids.

1.2 Thesis structure

The thesis describes the author’s contributions in fluid phase measurement by using novel optical, microfluidic and nanofluidic approaches for the applications of carbon transport and sequestration, and, shale oil/gas production. These contributions leverage optics and

microfluidics to develop fast and accurate fluid phase measurement methods, and leverage nanofluidics to study the unique effects of nanoconfinement on fluid phase behavior.

Chapter 1 briefly describes the research motivations from two industrial processes: Carbon Transport and Sequestration; and shale gas/oil production. This chapter discusses the key aspects of these two industrial processes most closely associated with this work and briefly describes the demand for novel phase measurements in that context.

Chapter 2 begins with the key concepts used in this thesis. These concepts include phase, fluid phase diagram, and typical phase behavior of pure substance and mixtures. Then the latter section of this chapter presents a short literature review of existing techniques for fluid phase measurement, including typical pressure-volume-temperature (PVT), electrical, optical, microfluidic and emerging nanofluidic approaches.

Chapter 3 presents the thin-film interference based optical approach to detect bubble and dew point of CO2 and industrial CO2 mixtures. This chapter was published in Sensors and Actuators B: Chemical.

Chapter 4 presents the refractive-index sensing based optical approach to distinguish CO2 phase from brine phase. This chapter was published in Environmental Science and Technology.

Chapter 5 describes the microfluidic approach rapid mapping of pressure-temperature phase diagram of CO2 and a CO2-N2 mixture. This chapter manuscript has been submitted to Journal of the American Chemical Society (Communications).

Chapter 6 presents a nanofluidic platform to study cavitation and bubble growth of pure hydrocarbon (propane). This chapter manuscript has been submitted to Physical Review Letters.

Chapter 7 summarizes the author’s contribution in optical, microfluidic and nanofludic methods measuring fluid phase. This chapter also includes an outlook for the trend of fluid phase measurement technology in the future.

2 Introduction

2.1 Fluid phases

Substances exist in solid, liquid and gas phases. In the solid phase, molecules are at relatively fixed positions with three-dimensional repeatable patterns. The distance between molecules is relatively small which keeps molecules attracted to each other. When the temperature is increased, the velocity of molecules reaches a threshold point where the attractive forces are overcome and the “pattern” breaks away, which is known as the melting process from solid to liquid. The molecules in the liquid phase have weaker inter-molecular forces and the “freedom” of the molecular movement is higher than it is for a solid. Once temperature is increased or pressure is decreased, the molecule structure could not be maintained at a relatively fixed position, instead the molecules start moving randomly which forms gas phase where molecules are spaced in relatively large distance. The molecular arrangement in these three phases is illustrated in Figure 2-1.[24] The phase transition from liquid to gas happens in one of two ways: (1) evaporation: where liquid molecules escape from the liquid-vapor interface and forms a gas phase; (2) boiling: where gas bubbles nucleate and grow inside the entire mass of liquid. Conversely, phase transition from gas to liquid is defined as condensation, where gas molecules form a liquid molecular pattern.

Three basic thermodynamic properties of fluid, i.e. temperature, pressure and volume, are usually studied and expressed by property diagrams. Typical property diagrams are T-v, P-v and P-T diagrams.

Figure 2-2 a) shows a typical T-v diagram of a pure substance. At constant pressure P1, a heated pure substance undergoes compressed liquid, saturated liquid-vapor coexistence and superheated vapor (dash line). Saturation temperature is defined as the temperature at which a substance changes phase at a given pressure. Similarly, saturation pressure is defined as the pressure at which a substance changes phase at a given temperature. The horizontal region, “saturate liquid- vapor”, indicates liquid-vapor coexistence in equilibrium. If pressure is increased, this saturation line becomes shorter. If pressure is kept increasing, this saturation line finally shrinks into a point, defined as “critical point”, where the saturated vapor and saturated liquid phases are identical. In other words, there is no explicit boundary between liquid and vapor phases.[24] Similarly, Figure

2-2 b) is a representative P-v diagram of a pure substance. Instead of P = constant in T-v diagram, temperature T is kept constant and phase change curve goes a downward trend in the direction of increasing v.

Figure 2-2 c) shows the P-T diagram of a pure substance. P-T diagram is usually called phase diagram because all the solid, liquid and vapor phases are distinguished from each other by three lines, i.e., sublimation line, melting line and vaporization line. All the three lines meet at triple point where three phases coexist in equilibrium. Critical point is the end point of the vaporization line when there is no phase boundary between liquid and vapor phases.[24] It is noteworthy that the focus of this thesis is the “fluid” phase transition, or, the “vaporization line”.

Typical phase diagrams of pure fluid are shown in Figure 2-3 A) water and B) carbon dioxide. It is noteworthy that the vaporization line in Figure 2-2 c is generally called the saturation line here in Figure 2-3. The critical point of water is at a temperature of 373.95°C and a pressure of 22.06

MPa. The critical point of CO2 is at a temperature of 31.0°C and a pressure of 7.38 MPa. Since

CO2 is intensively studied in Chapter 3, 4 and 5, visualization images of CO2 phases at different pressures and temperatures are presented in Figure 2-3 C). A clear interface between liquid and vapor phases is observed on the saturation line below the critical point (Figure 2-3 C-a and C-b). Moving along the saturation line towards the critical point, the liquid-vapor interface becomes vaguer and finally vanishes when reaching critical point, as shown from c to g in Figure 2-3 C.

Above the critical point (T > Tcr and P > Pcr), the fluid turns into a “supercritical” phase (h in Figure 2-3 C). The supercritical phase is a uniform phase with gas-like viscosity and liquid-like

density. Generally, liquid phase refers to the region where T < Tcr and P > Psat, while vapor phase refers to the region where T > Tsat and P < Pcr. However, it is also common to distinguish

“compressed fluid” from the liquid phase where T < Tcr and P > Pcr (Figure 2-3 B, and, k in Figure 2-3 C). Similarly, “superheated vapor” can be distinguished from the vapor phase when T >

Tcr and P < Pcr. All phases of vapor, liquid, supercritical, compressed liquid and superheated vapor are labelled in Figure 2-3 B. Typically, there is no sharp boundary between the supercritical phase and the compressed fluid phase (See isobaric transition h-l-k in Figure 2-3 C), and, neither between the supercritical phase and superheated vapor phase (See isothermal transition h-i-j in Figure 2-3 C). [25]

Figure 2-3. Pressure-Temperature phase diagram of A) water, and B) carbon dioxide; C)

The phase diagram of multicomponent systems or fluid mixtures is more complex than that of a pure fluid. Instead of a simple liquid-vapor saturation line in the pure fluid case, a fluid mixture has a two-phase envelope inside which liquid and vapor coexist with different liquid-to-vapor ratios. Figure 2-4 shows the phase envelope of natural gas, a fluid mixture consisting of 90.4%

methane, 5.4% ethane, 2.1% propane, trace amount of CO2 and N2, and other heavier hydrocarbons.[9] The phase envelope boundary is composed of two saturation lines, i.e. a bubble point line and a dew point line. The bubble point is defined as the point where the first bubble is formed out of the liquid phase. Similarly, dew point is defined as the point where the first dew is formed out of vapor phase. The bubble point line and dew point line are joined at the critical point where there is no sharp boundary between the liquid and vapor phases. Other important

points on phase envelope are “cricondenbar” or “maxcondenbar” (P = Pmax), and,

“cricondentherm” or “maxcondentherm” (T = Tmax). It is noteworthy that the maxcondenbar point, maxcondentherm point and critical point are not overlapping for the case of a fluid mixture, as shown in Figure 2-4.

2.2 Experimental methods to measure fluid phase

Understanding the phase behavior of a fluid is particularly important in petroleum reservoir engineering, chemical reactor engineering, the transport and storage of natural gas and carbon dioxide, and, supercritical fluid applications. Fluid phase behavior can be obtained by either

theoretical/simulation approaches or experimental methods. Theoretical/simulation approaches calculate the thermodynamic properties (P,V, T) for mixtures using the cubic Equation of State (EOS), such as Redlich-Kwong (1949), Soave-Redlich-Kwong (1972), and Peng-Robinson (1978) equations.[9] The majority of phase simulation software applies these EOS or modified EOS to predict the phase behavior of fluid mixtures in different scenarios. For example, the Peng-Robinson EOS [26] and modified van der Walls EOS [17] are applied to study the influence of nano-confinement on phase behavior. Molecular simulations of hydrocarbons in nano-confinement have been motivated by the boom of shale oil/gas.[18][27] By using molecular simulations, inter-molecular forces and wall-molecular force can be predicted. The second approach to obtaining phase behavior is through experimental methods. Since the scope of this thesis is within experimental methods, the remainder of this section focuses on existing experimental methods to measure fluid phase.

The existing experimental methods can be categorized into two classes based on whether the compositions of the equilibrium phases are determined in the test (“Analytical”) or the mixture has been prepared with known composition (“Synthetic”).[10] Figure 2-5 A shows analytical- isothermal schematics including charging, mixing & equilibrium, and sampling & analysis. Fluid phase composition is determined by different analytical techniques, e.g. liquid chromatography, gas chromatography and in-situ spectroscopy. The analytical approach is well suited to

investigating strongly composition-dependent phase boundaries and the main error sources come from the precise determination of chemical composition under test pressures. The total use of analytical approaches decreased from 45.6 % to 37.0 % from 2000-2004 to 2005-2008.[10] On the other hand, synthetic methods investigate mixtures with known composition without analysis of phases. Figure 2-5 B shows the detection of phase transitions using the synthetic approach with either visual or non-visual techniques. For visual methods, the fluid phase change is conducted in a pressure-volume-temperature (PVT) cell and visualized through a transparent window. More details of PVT tests are described in Section 2.2.1. For non-visual methods, a representative approach is to characterize the P-V curve where a sharp slope change indicates a phase change. More non-visual methods are grouped into optical (Section 2.2.2) and electrical methods (Section 2.2.3). This chapter also reviews microscale phase measurement techniques (Section 2.2.4) as well as emerging nanoscale methods (Section 2.2.5).

2.2.1 PVT experiments

It is essential to characterize the phase behavior of fluids in many industries. The oil and gas industry is the largest current application area for such measurements, from production to the refinery, to the consumer (upstream, midstream and downstream). Typically, upstream reservoir pressures range from 10 to 200 MPa and temperature ranges from 25 to 200 °C. Routine Pressure-Volume-Temperature (PVT) experiments are designed to emulate the processes in a reservoir produced through natural depletion, or “primary recovery”. One of the most important PVT properties is the saturation pressure at reservoir temperature.[9] For example, when pressure drops below the saturation pressure during production, the fluid will split into gas and liquid phases. However, the gas phase will be produced relatively easier compared to liquid due to lower viscosity and buoyancy, which leaves the heavier (often more valuable) components in the reservoir. Enhanced oil recovery (EOR) PVT experiments are performed with the goal to keep production in single phase with a high concentration of heavier components.

Figure 2-6 A lists the volume, pressure limit and temperature limit of PVT cells used for oil and gas condensate. The typical volume of a PVT cell is from 500 to 650 mL.[9] It is noteworthy that there are PVT cells can reach higher temperature than the ones in this table. For example, a PVT cell from Core Laboratories is rated at 260 oC. In industry, routine PVT experiments include constant-mass expansion, differential liberation and constant-volume depletion.

Figure 2-6. Typical PVT experiments of a reservoir fluid. A) Pressure, volume and temperature conditions of PVT cells used for different reservoir fluids; Schematics of constant-mass expansion experiment for B) an oil mixture and C) a gas condensate; D) Schematics of differential depletion experiment for oil; E) Schematics of constant-volume depletion experiment for a gas condensate. Reproduced with permission from Taylor & Francis Group,[9] copyright 2015.

During the constant-mass expansion of an oil mixture (Figure 2-6 B), the oil mixture is initially stabilized at constant temperature and above reservoir pressure. The volume is increased so that pressure is decreased to the bubble point. The volume is decreased slowly. Pressure and volume are recorded at each time interval. Isothermal compressibility is calculated if the pressure is

above saturation pressure, while the Y-factor (a ratio between the relative change in pressure and the total volume change in the two-phase region) is calculated below the saturation point. For the case of the constant-mass expansion of gas condensate (Figure 2-6 C), liquid is condensed when pressure is slowly decreased, down to a typical level of 5 MPa. The gas compressibility and liquid dropout are two parameters calculated from recorded data for the cases above saturation pressure and below saturation pressure, respectively. The differential liberation (Figure 2-6 D) best emulates the compositional and volumetric changes during oil production. The oil mixture is initially kept at a fixed reservoir temperature and above the reservoir pressure. The valve on top allows the gas to be depleted at constant pressure during pressure reduction. The depleted gas composition is measured at standard conditions. The pressure reduction and gas depletion are repeated typically six stages down to atmospheric pressure. Typically, the following parameters can be calculated: (1) formation volume factor Bo (the ratio of oil volume at certain stage to residual oil volume); (2) gas/oil ratio (GOR); and (3) gas formation volume factor (gas volume at cell condition over that at standard conditions). Another common PVT test is the constant- volume depletion, usually performed on gas condensate mixtures (Figure 2-6 E). In this test, the excess volume of gas on top during expansion is depleted to maintain a constant volume. This procedure is typically repeated six stages down to around 5 MPa. Liquid volume, compressibility factor, gas composition and other parameters can be measured.[9]

In summary, current PVT methods are the standard technique of fluid phase analysis because they offer direct visual quantification of fluid phases, precise control of temperature and pressure, and are the incumbent technology in the oil and gas industry. Commercial PVT systems can be used to measure not only saturation pressure and temperature, but also other properties including viscosity, gas-oil-ratio, density and chemical composition (with chromatography tools). However, a typical PVT system usually takes hours to change temperature for a new data point due to a large heat mass associated with a bulky cell volume (hundreds of mL). On the side of cost, the PVT visualization cell along with complicated accessories, such as temperature-, pressure- and control-system, can cost up to $500,000. The slow measurement times and the high capital costs motivate innovation in this space - novel fluid phase measurement techniques, such as optical, electrical, acoustic or microfluidic methods, targeting specific applications.

2.2.2 Optical methods

Fluid phase change can be measured by the means of optical signals. Several principles of optical methods are briefly discussed here. Chilled mirror method measures the dew point temperature by using a cooled plane surface to induce condensation. The dew formation is detected manually by eye (details can be found in ASTM-1142) or automatically by an optical system. In the

18 automatic approach (See Figure 2-7 A), visible or IR light is emitted from a photodiode, reflected from the cooled mirror surface and received by the photodetector. Once dew is formed, the reflected light is modified due to the absorption and scattering of incident light. The range of a typical chilled mirror is from -80 to 85°C.[28] It is noteworthy that the chilled mirror method is widely accepted as a laboratory reference standard for calibration purposes. Absorption spectroscopy is another approach that uses optical signals to interpret fluid phase change, especially dew point (See Figure 2-7 B). Tunable laser diode offer a broad bandwidth of spectrum to include vibration frequency that is unique to the species of molecule detected. The spectroscopy measurement is based on the non-contact method, which is a distinct advantage. However, the measurement result is very sensitive to the temperature of the laser diode, which should be taken care of in measurements.[29] Other examples using an optical method for dew point detection are based on refractive index sensing, such as the Fabry-Perot hygrometer and Fiber-optic reflectometer. The Fabry-Perot hygrometer (See Figure 2-7 C) consists of multi- layered material with high and low refractive indices. The condensed liquid (usually water) inside the surface material layer changes the refractive index and shifts the signal wavelength. The wavelength shift is proportional to the amount of liquid molecules on the sensor. This technique provides intrinsically safe light signals but has the disadvantage of slow response and coating degradation after long-term use.[28] The fiber-optic reflectometer (See Figure 2-7 D) detects the change in the fluid by responding to the difference in refractive indices between the fiber material and the fluid residing on the end of fiber. This fiber-optic sensor can detect bubble point, dew point and critical point via the reflected signal. The fiber-optic probe installed in an equilibrium cell can work at high temperatures (300 °C) and pressures (30 MPa).[30] Importantly, optical methods have been employed with conventional PVT system, e.g., Solid Detection System (Schlumberger) which measures the saturation/onset pressure by NIR light transmission and Charged-Coupled Device system (Schlumberger) which measures the fluid volume via a movable long-distance microscope.

In summary, optical methods offer several advantages, such as high sensitivity, in situ measurement, automatic detection and potential for distributed sensing over long distances. A wide application of optical methods is the measurement of moisture or dew point of industrial gas especially for CO2 and natural gas.

2.2.3 Electrical and acoustic methods

Figure 2-8. Electrical methods: A) Capacitance probe and operating principle; B) Quartz- crystal microbalance (QCM) sensors and operating principle.[29] Reproduced with permission from Ametek Process Instruments, copyright 2011.

Electrical methods utilize electrical signals to detect dew or moisture content. This subsection reviews a few typical electrical methods including capacitance-based and oscillation-based approaches. A capacitance probe (Figure 2-8 A) generally uses moisture-sensitive dielectric material sandwiched between two electrodes. The condensed liquid (typically water) changes the dielectric constant of the layer material. The change in capacitance is detected by an electrical circuit. This approach has the advantage of a low installation costs but is vulnerable to contamination or fouling from system impurities. A quartz crystal microbalance (See Figure 2-8 B) employs mechanical oscillation to detect the fluid phase transition. The condensed liquid on the coating material increases the mass loading of the oscillator which decreases the resonance frequency. The amount of condensation is related to the oscillation frequency. It is noteworthy that the quartz crystal microbalance is widely used in natural gas moisture measurement in industry because of its high accuracy (down to 10 ppb). Determining fluid phase change by tracking mass change is also the principle of a surface acoustic sensor. A surface acoustic wave sensor was developed to quantify water condensation and dew point.[31] The pioneer work

shows a sensitivity that a phase shift of 30° is equivalent to 3µg/cm2. The sound speed in a fluid is another parameter of fluid phases. In a previous work, the phase behavior of rich gas mixtures (varying methane from 70 to 82 %, up to 22 MPa) were measured by the decompression wave.[32] The change in the speed of sound within a fluid indicates the fluid phase transition. The change in the decompression wave speed between dry gas and the two-phase region agreed with the model predictions within a few percent.

In summary, electrical and acoustic methods take advantages of the changes in electrical or acoustic signals caused by a new phase during phase transition. Capacitance sensors are widely used in the measurement of moisture or dew point of natural gas due to relative low cost.

2.2.4 Microfluidic methods

Microfluidic technologies offer unique perspectives and tools for fluid phase measurement. The inherent advantages provided by microfluidics include small sample volume, precise control of temperature and pressure, time-efficient operation, fast analysis and potential for multiplexing. This subsection reviews microfluidic-based methods for fluid phase measurement.

A microfluidic PVT system was first developed by Mostowfi et. al.[11] The silicon-glass bonded microfluidic device allows direct visualization within a flow-based design. The idea is to establish a pressure-driven steady liquid flow in a long serpentine channel and induce bubble formation at designed inlet restrictions, as shown in Figure 2-9 A and B. The generated bubbles expand as they flow downstream where the gas-liquid volume fraction is measured, as shown in Figure 2-9 C. The experiment is performed at room temperature. Inlet pressure is set slightly above the saturation pressure and the outlet pressure is set at atmospheric pressure. Micro-cavity- based pressure sensors along the microchannel are used to indicate the pressures by measuring the deformation of cavities. Phase change is imaged using a high speed camera operating at 5000 fps. The detected bubble point pressure shows a ± 2.5 % relative accuracy. A linear pressure drop was observed in the liquid-gas slug flow. Notably, this approach requires less than 15 minutes, compared to 8 hours using conventional PVT techniques. This pioneer work introduced a microfluidic approach to PVT studies, specifically bubble point pressure determination. This PVT system was also used to determine the gas-oil-ratio of hydrocarbons in a later study.[33] Mostowfi’s team also developed a microfluidic platform to measure asphaltene of crude-oil

samples, which took less than 30 minutes (compared to days required for the conventional technique) and showed good repeatability of ± 2%.[34]

Later, another microfluidic device with serpentine channel was developed by Aymonier et. al to measure the bubble point, dew point and critical point of a multicomponent system, as shown in Figure 2-10 A and C.[35] The silicon-glass material allows direct visualization, high temperature and high pressure capabilities. The primary advantage here, as compared to Mostowfi’s PVT device, is that it can detect dew point and critical point in addition to bubble point pressure. This device can function at higher temperature (300 - 500 K) and higher pressure (1 - 200 bar). Finally, this device offers a distinct operational mode (“dynamic stopflow” via a bypass channel, see Figure 2-10 B) which can reduce the flowrate significantly and therefore lower the requirement camera imaging speed using 4 fps rather than 300 fps using a continuous mode. The measurement demonstrated a 2 % deviation from modeling results. This microfluidic approach is 5 times faster than conventional optical cell methods.

Recently, more microfluidics-based techniques have been developed for the phase behavior measurement of multicomponent samples. Sinton et. al. developed a microfluidic chip to measure trace amount (~ 10 µL and <0.005 mole fraction) of water droplet in industrial CO2 streams, as shown in Figure 2-11 A.[36] The glass microfluidic chip enables direct visualization of dew droplet (1 – 2 µm) under high pressure, up to 13 MPa. This microfluidic approach provides a 3-fold error reduction compared to existing methods. Sinton group made another microfluidic device to measure minimum miscibility pressure of CO2 in crude oils, as shown in

Figure 2-11 B. The CO2 phase and the oil phase are clearly distinguished by the inherent fluorescence of crude oil. The miscibility pressure can be determined by investigating the

intensity profile across interface between CO2 and crude oil, which is automatic and operator independent. The variation of results by using this technique is 0.5 MPa which is lower than for

conventional methods (e.g. the rising bubble approach). Importantly, this method takes less than 30 minutes in contrast to the days or weeks by using conventional methods.[37]

Figure 2-11. Two microfluidic systems measuring the phase behavior of multicomponent samples. A) Detection of water dew point in CO2 stream. B) Measurement of the minimum

In summary, microfluidic methods exhibit inherent advantages of low sample volume and time- saving operations and analysis over conventional PVT methods. In addition, microfluidic based measurement usually is coupled with precise control of temperature and pressure which provides accurate measurement. Notably, microfluidic (chamber or channel is smaller than 1 mm) or even meso-fluidic (chamber or channel is between 1 mm and 1 cm) present a great potential for multiplexing of multiple parameters into a single run, which could significantly reduce operation time and cost. The challenges of commercialization of the microfluidic-based methods come from fabrication cost and the relative short lifetime due to contamination and mechanical weakness for long-term run.

2.2.5 Nanofluidic methods

In the previous subsection, it was shown that microfluidic methods are mostly motivated by improving the time efficiency of fluid phase measurements. In other words, the microfluidic

systems are targeting for practical applications and competing with existing techniques. In contrast, the phase behavior measurements at the nanoscale are mostly motivated by understanding the fundamental physics of the effect of nano-confinement in phase transition. This subsection summarizes the experimental investigation of fluid phase change in nanoconfinement.

Buongiorno et. al. measured the water nucleation temperature on nanoscale cavities (down to 90 nm in diameter) and posts (down to 60 nm in diameter) fabricated on ultra-smooth silicon wafers, as shown in Figure 2-12 B. A lamp is used to heat the water-inducing bubble nucleation on the surface. The temperature of the surface is measured by using an infrared camera (650 to 1000 fps) with an accuracy of 2°C, as shown in Figure 2-12 A. Bubble nucleation is indicated by a sudden temperature drop. This technique showed that the bubble nucleation happens at a high superheating condition and agrees with the curve predicted by a Young-Laplace model, as shown in Figure 2-12 C. This result provides solid evidence that bubble nucleation in nano-sized features is not happening at the saturation temperature.

Figure 2-12. Study of bubble nucleation by a local heating source. A) Schematics of experimental setup for boiling on surfaces with fabricated cavities and posts; B) Enlarged images of cavities and posts; C) Experimental data of water boiling temperature in experimental conditions and theory predictions. Reproduced with permission from American Institute of Physics, [38] copyright 2012.

The nanoscale confinement is not limited to single geometries of cavity or post. Nanochannels represent another geometry that can be fabricated and applied to study the fluid phase transition. Majumdar et. al. observed evaporation-induced cavitation in nanofluidic channels.[23] Cavitation is defined as the bubble formation in liquids which is under tension, or, understood as the bubble formation at pressures lower than saturation pressure at given temperature. Liquid water is filled in 58-nm nanochannel as well as the connecting microchannels, as shown in Figure 2-13 A. Instead of evaporation (receding liquid-vapor menisci), a vapor bubble is “swallowed” into nanochannel followed by liquid slug pinning at nanochannel entrance, as shown in Figure 2-13 B. Such evaporation-induced cavitation was explained as the presence of

local expansions at the nanochannel entrance. In other words, the expansion at the entrance is responsible for the entrapped bubble that subsequently moved into center, as shown in Figure 2- 13 C. This study provides new insights into fluid phase change at the nanoscale.

Fluid phase change in nanoscale was also investigated in a nanoporous medium. A porous silicon (poSi) layer with ink-bottle geometry was developed by Stroock et. al. and applied to the study of the drying process by cavitation.[39] The layer including nanoscale pores of 1-2 nm in radius is under the glass layer with arrays of micro voids, as shown in Figure 2-14 A and B. The system of 625 voids is filled with water and dried for 60 h. The drying dynamics were visualized and it was observed that evaporation started from all four edges, as shown in Figure 2-14 C. The cavitation pressure was found to be 84 % of saturation pressure.

The research of fluid phase behavior in nanoconfinement is becoming of high interest, and is largely driven by the rise of shale gas/oil production in recent years. The lack of experimental investigations in this field motivates the research of hydrocarbon phase behavior in nanopores. A nanofluidic device including a set of nanochannels was made by Ma and Yin et. al. to study both the transport of water/gas system at the nanoscale as shown in Figure 2-15 A, B and C.[20] By using the same device, Yin and Ozkan studied the liquid-to-vapor phase transition of pure n- pentane, and, a ternary mixture of n-butane, i-butane and n-octane.[26] It was found that pure hydrocarbons evaporated immediately after the complete drying in the service microchannel, as shown in Figure 2-15 D. However, the evaporation of the ternary mixture stopped before entering nanochannels due to the distillation effect which leaves heavier components during continuous evaporation, as shown in Figure 2-15 E. Improvements in experimental design are needed for mixtures to avoid the distillation effect which changes the composition of liquid.

In summary, nanofluidic systems (chamber or channel is smaller than 1 µm) for fluid phase measurement are mostly motivated by understanding the nano-scale confinement effect in phase transition phenomenon such as boiling, cavitation and condensation. Literature shows that the nano-scale confinement brings in some unique phenomena, such as extreme high boiling temperature (super heating), which is different from that in bulk volume. The application of these nanofluidic methods can be extended to investigate more fundamental physics involved in nanoscale related applications, such as Shale/tight oil production. On the other hand, the constraints in nanofluidic methods are mostly associated with the complicated fabrication process, high cost and mechanical weakness due to contamination and clogging. More research effort is needed to overcome these constraints.

3 Detection of Bubble and Dew Point using Optical Thin-film interference

Bubble and dew point data are essential for many practical applications, and particularly the safe

pipeline transport of post-capture CO2 which contain impurities. These mixtures show highly composition-specific phase properties, necessitating much more experimental data and motivating more rapid and inexpensive measurement methods. Here we demonstrate a responsive small-scale pressure-volume-temperature (PVT) cell system enabled by thin-film interference, and its application to an industrially-relevant post-capture CO2 mixture stream. The small (5 mL) volume is one-to-two orders of magnitude faster to equilibrate than conventional PVT cells with viewing windows. Inside the cell, top and bottom optical fiber sensors detect bubble and dew points, respectively. At vapor-liquid transition points, the reflection spectrum from the optical fiber tips report clear interference patterns caused by a thin film on the sensor. In addition to sharply delineating the phase change condition, the sensor also reports the real time thickness of the film (accuracy on the order of 1 µm). The method is validated with the well

characterized pure CO2 test case (average error of 2.8 bar as compared to NIST data), and

applied to an industrially-relevant CO2 stream, characteristic of post-capture oxyfuel combustion

– an important source for downstream CO2 utilization and storage.

Bao, B., Fadaei. H., Sinton. D. Detection of Bubble and Dew Point Using Optical Thin-Film Interference. Sensors & Actuators: B. Chemical, 207 (A) 640–649 (2015). Reproduced with permission from Elsevier.

Publication online link: http://www.sciencedirect.com/science/article/pii/S0925400514012878

3.1 Introduction

Fluid phase behavior at high pressures is central to a wide range of processes including petroleum reservoir engineering, high-pressure chemical reactors, and transport and storage of

natural gas and CO2.[40] Phase behavior of CO2 is of much interest due to concerns over global

CO2 emissions and related industrial applications in enhanced oil recovery and carbon

sequestration. Pipeline transport of CO2 is central to all current and eventual large-scale CO2

applications.[41] Effective transport requires that CO2 remains in a dense phase (either liquid or supercritical), and thus measurement and knowledge of the phase behavior of industrially-

relevant CO2 is essential.

In contrast to laboratory grade CO2, industrially-relevant CO2 streams contain a variety of

impurities, often including N2, O2, Ar, CO, H2O, SOx, NOx, H2S and other components depending on the source process.[42] Typical sources include power plants, refineries,

upgraders, cement plants and steel plants. The presence of impurities in the CO2 stream significantly impacts the phase behavior, as indicated by current models.[43][44][45][46] For instance, phase diagrams of CO2 and N2 binary mixtures studied by Equation of State models

show significant deviation from pure CO2 phase behavior (e.g. a 40% increase in saturation

pressure at 0°C of CO2 with 5 mol % N2).[43] In another approach, a ternary mixture of N2, O2 and CO2 at low temperature was modeled by combining experimental data from the three binary subsystems.[44] In general models indicate that even relatively low concentrations of air-

derived impurities (N2, O2 and Ar) can considerably increase the saturation pressure and decrease

the critical temperature of CO2 streams.[43][45] The implications for transport are particularly significant. Maintaining a dense phase will require either lower temperatures (generally not feasible), or increased pressures leading to increased capital and operating costs.

Current methods of measuring fluid phase transition include analytical (~ 40%) and synthetic (~ 60%) approaches.[40] In analytical methods, fluid is sampled from both phases of an equilibrium mixture and the composition of each phase is measured using chromatography or spectroscopy.[47][48] Challenges include sampling an equilibrium mixture without significantly disrupting the equilibrium, and relatively high capital and labor costs. Synthetic approaches can determine phase behavior of prepared mixtures with known compositions without sampling.[40] Synthetic-visual techniques are the most common methods, involving a pressure-volume- temperature (PVT) cell with a visualization window which allows direct observation of phase transition as a function of cell pressure and temperature.[49][50] Synthetic-nonvisual experimental methods have also been developed, using acoustics,[51][31] quartz sensors[52][53][54] and infrared spectroscopy.[55]

Microfluidic-based PVT devices were developed to measure physical properties such as diffusivity of immiscible fluid mixtures, showing orders of magnitude improvement in speed and

sample volume.[56][57] Another microfluidic PVT approach was developed to analyze phase diagrams of vapor-liquid systems.[11] The depressurization of the sample as it flowed through the microfluidic chip mimicked the fluids transition from reservoir to surface. Recently we developed a microfluidic approach to detecting liquid water condensation at very low

concentrations in supercritical CO2 using direct observation.[58] A localized surface plasmon resonance based sensor was demonstrated to detect dew condensation on gold/ceramic nanocomposite.[59] To enable more accurate and automated optical detection, optical fibers are a preferred approach. We have previously developed ultra-sensitive optical fiber sensors to detect

dissolved CO2 in brine at reservoir temperatures and pressures.[60] Optical fiber sensor has been used to determine relative humidity in air.[61] Reflection-based optical fiber sensors have also been applied to directly detect phase transition.[62][30] In general, optical fiber sensors offer many benefits for PVT related studies, including high accuracy, robustness, fast response, compact size and low cost.

The method presented in this chapter is a new synthetic-nonvisual approach which exploits the principle of thin-film interference on reflection-based optical fiber sensors. During vapor-liquid phase transitions, the reflection spectra of the optical fiber sensors show transient interference patterns caused by accumulation of thin films of vapor or liquid phase on the sensor. This new

PVT system was first applied to determine the vapor-liquid phase transition of pure CO2 - enabling validation of the technique. Then, the phase transition envelope for an industrially- relevant CO2 stream from oxyfuel combustion was measured. The experimental result – a first – was compared with predictions from two common models.

3.2 Experimental

3.2.1 Experimental setup

The schematic of experimental apparatus is shown in Figure 3-1 a. A cross-shaped PVT cell was connected to a high-pressure syringe pump (Teledyne ISCO Model 260D) and a cylinder containing CO2 samples. Optical fibers sensors were inserted both at the top and bottom of the PVT cell. The temperature of the PVT cell was controlled by emersion in a water bath (Polystat Cooling/Heating Circulating Bath, 6.5 L), while the pressure was controlled by the pump. The two optical fiber sensors were connected with an optical interrogator (Micron Optics SM 125).

The optical interrogator delivered a broadband incident light (1510 – 1590 nm) to each of the optical fiber sensors and detected the spectra of reflected light from the sensors in real time.

The internal features of the PVT cell are shown in Figure 3-1 b. The stainless steel PVT cell had an internal volume of about 5 mL and four 1/8” NPT ports. The left and right ports allowed loading and purging the fluid samples, respectively. The top and bottom ports immobilized the top and bottom optical fiber sensors, respectively. The optical fiber sensor was single mode (Core/Cladding/Coating: 9/ 125/ 245 µm). One distal end of the fiber was cleaved to generate a flat fiber tip to serve as the reflection sensor surface. The other distal end was connected to the interrogator. The top and bottom fiber tips were positioned axially to be just inside the fluid cavity of the PVT cell. The optical fiber was sealed with a PEEK sleeve and compression fittings, as detailed elsewhere.[60]

Figure 3-1. a) Schematic of experimental apparatus integrating gas cylinder, pump, PVT cell, optical interrogator, and computer; b) Internal features of the PVT cell with optical fiber sensors.

3.2.2 Sensing mechanism: thin-film interference

Figure 3-2 shows the principle of detection of vapor-liquid phase transition with a fluid sample. If the fluid sample surrounding the fiber tip exists as a single phase (Figure 3-2a), such as single vapor or liquid phase, the incident light travelling inside the fiber core will be partially reflected at the fiber/fluid interface. The reflection light intensity is a function of refractive index of the

fluid sample near the end of the fiber. This single-phase reflection mode is the previous approach to phase detection with reflection-based optical fiber sensors. [62][30]

Figure 3-2b shows the thin-film approach developed in this work. Specifically, in dew-point detection, any liquid forming (from an otherwise gaseous mixture) forms preferentially at the bottom of the PVT cell, and is subsequently detected by the bottom optical fiber sensor. Both types of phase transitions (dew-point and bubble-point) can exhibit temporary thin films on the

sensor surface. In the case of dew formation, a thin layer of liquid film would be condensed on the fiber tip while the surrounding medium is still in vapor phase (Figure 3-2b). In the case of bubble formation, either a thin vapor film or a thin liquid film is possible during transition, depending on the wettability of the sensor (typical silica fibers are highly hydrophilic and thus liquid films are expected during transition). Thin films on the end of the fiber optic generate interference as shown at right in Figure 3-2b. Constructive interference, as shown, is an additive effect of a wave reflected at the first interface (dotted curve) and a wave reflected at the second interface (dashed curve). Specifically, the two reflected waves have the same electromagnetic wave phase (the term ‘EM phase’ is used here to distinguish from vapor/liquid ‘phase’). In contrast, if the reflected waves have an EM phase difference of 180°, the added wave will show a destructive wave pattern. The thin-film interference can be clearly identified by the reflection spectrum as a function of wavelength. This interference will also be accompanied by a change in the average reflected power. For instance, in the case where a liquid film is replaced by gas (bubble point detection), the reflected power will increase.

In addition to measuring vapor-liquid phase transition, the interference pattern also enables the determination of the thickness of the liquid film on the fiber tip. Considering the principle of thin-film interference demonstrated by Figure 3-2b, the wavelength associated with the constructive wave in the thin film is λ1, while the wavelength associated with the neighbor

destructive wave in the thin film is λ2 (λ2>λ1). The thickness of the liquid film on fiber tip, d, is correlated to the wavelength by Eq 3-1, where m is an even integer.

2 1 (3-1)

The thickness of the thin film can thus be determined by Eq 3-2.

(3-2)

The corresponded wavelengths detected by the interrogator, in free space, are λo,1 and λo,2 (λo,1 =

n λ1 and λo,2 = n λ2), where n is the refractive index of the medium in the thin film.

, , (3-3) ,,

3.2.3 Experimental procedure

The optical fiber sensing method was applied to measure dew points and bubble points of pure

CO2 and a representative oxyfuel combustion product composed of 91.7% CO2, 2.5% N2, and

5.8% O2 (mole fraction, from Praxair). The isothermal approach was employed, where pressures o were varied at fixed temperatures (from 10 C to critical point of the CO2). Specifically, the top sensor detected bubble points with decreased pressures from 100 to 30 bar, while bottom sensor detected dew points with increased pressures from 30 to 100 bar. Given the temperature range of

interest, a pressure of 30 bar guarantees that CO2 samples are a vapor phase, and, pressure of 100

bar ensures CO2 samples are a liquid phase.[63] The measurement interval of temperature and pressure were set to 1 oC and 1 bar, respectively. At each point of temperature and pressure, e.g. 20 oC and 80 bar, enough time was given for equilibrium before the measurement, typically 3 to 5 minutes after pressure became stable. This waiting time ensured the equilibrium of fluid sample in the PVT cell. To avoid any potential bending, or insertion, or compressive stress related errors associated with fiber optic sensing, the position of the fibers were maintained constant without any change during all measurements.

3.3 Results and discussion

The reflection-based fiber optic phase sensing method was applied to quantify both the dew point

and bubble point of CO2. First the approach was validated by application to pure CO2, for which the phase transition is sharp and well known. Once validated the approach was applied to an industrially-relevant CO2 stream with impurities.

3.3.1 Method validation with pure CO2

Figure 3-3 shows a representative detection result for bubble and dew point of pure CO2. Figure 3-3a presents the reflection spectrum of top sensor under decreasing pressure from 56 to 52 bar at 20 oC. Initially at 56 and 55 bar, the reflection spectrum shows uniform power signals across the entire wavelength. When the pressure was decreased to 54 bar, the reflection spectrum shows

a transient interference pattern including periodical peaks and valleys (t = t0), as shown in Figure 3-3a. Specifically, the peaks represent the constructive wave while the valleys indicate the destructive wave characteristic of thin-film interference. The thin-film interference marks the onset of phase change at 54 bar (at the top of the cell) from the initial homogeneous liquid phase.

The interference pattern is transient, lasting only ~ 30 s for the pure case, after which the power increased to a steady state value. This higher power level indicates a lower refractive index fluid at the fiber tip and is maintained at remaining lower pressures.

o Figure 3-3. Detection of bubble and dew point of pure CO2 at 20 C. a) Reflection spectrum of top sensor under decreasing pressure: The interference pattern observed at 54 bar indicates the formation of a thin film corresponding to the bubble point; b) Corresponding average power vs. pressure: A large interference fluctuation band at 54 bar (3.57 dBm) indicates the phase transition; c) Reflection spectrum of the bottom sensor under increasing pressure: The interference pattern observed at 55 bar indicates the dew point; d) Corresponding average power vs. pressure: A large interference fluctuation at 55 bar (3.24 dBm) indicates the phase transition. (Note: error bars are included in plots (b) and (d), however they are smaller than the data-point markers; pressure lines in (a) and (c) appear overlapped in single phase and shift significantly during phase transition.)

Figure 3-3b shows the average power as a function of decreasing pressure, with both the initial

(t0) and steady state values (t∞) plotted at P = 54 bar where the interference occurs. The interference fluctuation band is plotted as an orange bar, clearly indicating the phase transition point. The interference deviation (3.57 dBm) was very large, and thus easily distinguished from the noise level of the sensor (~ 0.12 dBm). We note that the actual amplitude of the interference pattern is even larger than the interference deviation (calculated as standard deviation). We calculate it here as deviation, however, in keeping with signal processing norms. The low-to- high average power shift is expected as the lower refractive index of vapor replaces high index liquid at the fiber tip, resulting in a higher reflective coefficient. It is noteworthy that the pressure change due to a local bubble formation is negligible due to i) small ratio of a bubble volume to total liquid volume; and ii) fast response of PID control in pump system.

o Similarly, the dew point measurement for pure CO2 at 20 C with the bottom sensor is reported in Figure 3-3c and d. The procedure for dew point is similar to that of bubble point, but with pressures increasing from vapor to liquid levels. Here, high power levels (indicating vapor) are observed at pressures up to and including P = 54 bar. At P = 55 bar, the dew point is detected as interference is observed and the average power drops effectively instantaneously to the lower power level, indicating liquid formation on the fiber tip.

Comparing the bubble and dew point cases offers some additional insight. In the bubble point case, the initial interference band (t0) is centered at the lower average power level, and then shifts to the higher average power level over time (t∞). This behavior is consistent with a high refractive index liquid film at the tip, drying over time, as would be expected below the bubble point pressure. In contrast to the bubble point case, the dew point case shows no such ‘drying’

time delay. Rather, the average power during initial interference (t0) matches that of the steady

state (t∞) value. Likewise, in both bubble and dew point detection cases, the interference pattern is centered at the lower average power (i.e. high refractive index) level, indicating a liquid film in contact with the fiber tip in both cases. This is in keeping with the highly wetting nature of the silica surface, promoting liquid wetting when both phases are available.

o o Bubble and dew points of CO2 were measured at 1 C temperature intervals from 10 C to the critical point. A full listing of bubble and dew point measurements for the pure CO2 case is provided in Table 3-1 (Supplementary material). Theoretically, bubble and dew points for pure

substance are identical. However, the measured bubble point of CO2 was, in general, 1 bar lower than its dew point. This difference is directly attributable to (i) the pressure intervals (1 bar) used in this test, and (ii) the conservative approach of claiming the dew or bubble point only after a phase change has been observed, which is necessarily beyond the transition point. As with all such measurements, reducing interval size readily enables higher resolution, if required, at the expense of measurement time. Comparing the measured and validated values in Table 3-1, the method applied with 1 bar intervals provided an average error of only 2.8 bar (5.0 %), with the error being less for higher temperature cases, approaching critical.

For all sub-critical cases the dew point and bubble point behavior was similar to that described above, however, determination of the critical point required an adapted procedure. Figure 3-4 shows the case for the first temperature above the critical temperature. Beyond the critical temperature, vapor will transition smoothly to a supercritical state without a sharp change in refractive index. Smooth transition behavior would thus be expected at both the top and bottom sensor. The reflection spectrum of top sensor under decreasing pressure and bottom sensor under increasing pressure are illustrated by Figure 3-4a and c, respectively. No interference pattern is observed in either case. The average powers with the standard deviations of top and bottom sensors are presented by Figure 3-4b and d, respectively. The standard deviations were found between 0.12 and 0.17 dBm, which were on the noise level, and well below the interference deviations observed in all subcritical cases. In addition, the smooth and relatively small transitions of the average power (between -23.76 and -25.58) imply a continuous phase change in the PVT cell, as expected for a vapor to supercritical state transition. Therefore, the o critical temperature of CO2 was indicated as 32 C using this approach (the first temperature interval measured over the validated critical point 31.1 oC [64]). Critical pressure was determined as the point with largest variation rate, i.e. maximum ∆Power / ∆Pressure. In this case, the critical point was found as 32 oC and 74 bar (close to known critical point of 31.1 oC and 73.8 bar [64]).

The time response of the small-scale PVT system is noteworthy. Here, 15 minutes was sufficient for the cell to reach equilibrium once the temperature bath was stable, as verified by the pure

CO2 validation testing. In contrast, changing the temperature of conventional PVT equipment requires up to 8 hours for equilibrium due to a large thermal mass. [11] In comparison with our

PVT cell volume (5 mL), some PVT cells used for CO2 phase behavior have volume in the range

of 200 mL. [65] In addition, the bulky surrounding material of conventional PVT cells requires hours of equilibrium time. As compared to conventional PVT systems with volumes one-to-two orders of magnitude larger, this equilibrium timescale is very short. This faster response is a direct consequence of heat transfer whereby the thermal equilibrium timescale, τ, scales to the power of 2/3 with volume.

3.3.2 Application to industrial CO2 mixtures containing impurities

Oxyfuel combustion is a combustion process whereby air is replaced with near-pure oxygen.[66] Historically used for high temperature flame applications (welding etc.) oxyfuel is also used for fossil fuel based power generation resulting in much reduced flue gas volumes with relatively

high CO2 content.[45] Here, a post-capture mixture characteristic of an oxyfuel process is employed, with 91.7% CO2, 2.5% N2, and 5.8% O2 (mole fraction). Such relatively low levels of impurities can have a significant influence on phase behavior. Here, a preliminary scan at high pressure intervals (~ 10 bar) was applied to determine the rough range of the dew and bubble points, followed by a targeted 1 bar intervals in the region of the transition.

Figure 3-5 shows representative results from bubble and dew point measurements of the impure o CO2 stream. The bubble point of the impure CO2 at 20 C was detected by the top sensor under decreasing pressure from 89 to 85 bar. The reflection spectrum (Figure 3-5a) shows a clear

transient interference pattern at 87 bar. As in the case of the pure CO2, a drying period was also observed whereby the average power of the interference case (t0) drifted upwards to the steady

state value (t¶), as indicated in Figure 3-5b. This transition required on the order of 3 minutes, which was longer than the pure case. A faster transition in the pure case is consistent with a sharp phase transition, as opposed to a two-phase envelope. The dew point measurement by the bottom sensor is demonstrated in Figure 3-5c and d. The reflection spectrum (Figure 3-5c) presents the transient interference pattern at 67 bar with large average power deviation (0.62 dBm). It is noteworthy that with impurities, the bubble and dew point are well separated, indicating a stable two phase envelope as can be expected for mixtures.[67]

o As in the pure CO2 case, bubble and dew points of the impure CO2 were measured at 1 C temperature intervals until the phase boundary disappeared. As described earlier, the phase boundary between liquid and vapor phase of pure CO2 disappeared when it reached the critical temperature. However, for mixtures, this maximum temperature is formally the maxcondentherm o point.[68] The maxcondentherm temperature of impure CO2 was successfully identified as 26 C

by top and bottom sensors, as indicated by Figure 3-6. No interference pattern was captured either under decreasing pressure by top sensor (Figure 3-6a) or under increasing pressure by bottom sensor (Figure 3-6c). The average power with deviations of top sensor (Figure 3-6b) and bottom sensor (Figure 3-6d) show a smooth transition and negligible, consistent deviation (0.12

– 0.19 dBm). Similar to pure CO2 critical pressure measurement, the maxcondentherm pressure was determined as the point with largest variation rate, i.e. maximum ∆Power / ∆Pressure. In this o post-capture oxyfuel CO2 mixture, the maxcondentherm point was found to be 26 C and 82 bar.

3.3.3 Comparison with existing methods of gas mixture properties

Figure 3-7 shows the experimental data of impure CO2, as well as predictions from models provided by Reference Fluid Thermodynamic and Transport Properties Database (REFPROP Version 8.0) from National Institute of Standards and Technology (NIST) and PVTsim software. Peng-Robinson and GERG-2004 are the models used in NIST and PVTsim, respectively. The

experimental data and validation points for the pure CO2 case are also plotted in Figure 3-7 for comparison. Table 3-2 (Supporting Information) provides a full listing of the bubble and dew points for the impure CO2 case, including both models and values measured here.

Figure 3-7. Comparison of bubble and dew points measured and predicted by two models

First, the bubble and dew point pressures of impure CO2 are significantly higher than the phase

transition for pure CO2, as consistent with expected behavior for mixtures with air-derived impurities.[43][45] Second, the difference between the bubble and dew point data show a significant two-phase envelope. Third, both the measurement and PVTsim data show the maxcondentherm point shifting down 5 oC and up 8 bar as a result of the impurities. Four, for dew point the agreement between the models and the measurements is quite good, with an average error of 1.4 bar (PVTsim and measurements). Particularly at higher temperatures, the measurements are within 0.33 bar at 20-25 oC (with error increasing to a maximum of 2.6 bar at 10 oC, in comparison to PVTsim). Five, both models and the experimental results diverge significantly in the case of bubble points. Near the maxcondentherm, the bubble point measurements and PVTsim predictions agree well down to 21 oC. However, below 21 oC all diverge, with the measured bubble points increasing with decreasing temperature, effectively widening the phase envelope. Although not reflected in the models under these conditions, such

behavior is observed in other studies of CO2 mixtures.[43]

One potential source of error in the bubble point measurements is the formation of a small bubble in the PVT cell, prior to migration upward to the sensor. That is, one could imagine missing the initial bubble formation as pressure is decreased. Such an error would result in the measurement underestimating the bubble point. However, the bubble point measurement data in Figure 3-7 are overestimated, as compared to the models, and thus such a ‘hidden bubble’ scenario cannot explain the discrepancy. Rather, we attribute the differences in the bubble point measurements and model predictions to the fundamentally challenging nature of assessing bubble point in mixtures – both for models and for experiments. In experiments, the mixture under test must be exactly that of interest (minimal error in the composition). In models, bubble point is challenging as it requires prediction of the first distinct vapor bubble which at such high pressures would be almost exclusively composed of impurities. The prediction becomes very dependent on the exact composition of impurities and their collective behavior in the presence of

CO2. These challenges, as well as the significant discrepancy between established models, motivate the need for validated small scale PVT based measurements able to provide the highly composition-specific dew and bubble-point data not otherwise achievable.

3.3.4 Film thickness determination

For the case shown in Figure 3-5c, λ, ≅ 1544 nm and λ, ≅ 1562 nm. The refractive index n of liquid layer of CO2 at the specific condition (20 C and 67 bar) determined by Lorentz-Lorenz equation[69] is 1.182. Thus the film thickness at that moment is determined to be d≅28 μm by Eq 3-3. The estimated film thickness is about a quarter of the probe diameter (125 μm), which implies a thin liquid film attached on the fiber tip. Similarly, for example, the film thickness shown in Figure 3-3c ( λ, ≅ 1531 nm and λ, ≅ 1564 nm , n 1.189 ) is determined to be d≅15 μm.

3.3.5 Resolution

The bulk of previous applications and measurements mostly target an accuracy of 1 bar.[65] The resolution of our sensor system can be increased by reducing the pressure intervals. To o demonstrate, the bubble point of impure CO2 at 20 C (from Figure 3-5) was repeated at a pressure interval at 0.1 bar, with the results plotted in Figure 3-8. The sensor clearly reports the bubble point at 87.4 bar, which agrees with the lower resolution results from Figure 3-5 (87 bar). As is common with measurement systems, the increased resolution comes at the cost of measurement time. In practice, increasing the resolution by one order of magnitude requires ~ 10 refined steps once the rough dew point is located. The cost of these additional steps is 30

minutes/step, or 5 hours per order of magnitude. It is noteworthy that although the same equilibrium time was used in all tests, the time for each refined step could have been decreased as the change in conditions is small. Although 1 bar is sufficient resolution for most of the applications of interest, these results indicate that 0.1 bar resolution is readily achievable with this approach, if needed.

3.3.6 Repeatability

o The bubble and dew point measurement tests of impure CO2 at 20 C (same as Figure 3-5) were repeated, with sample results shown by Figure 3-9. As shown, the results in all cases were repeatable, with measurements varying no more than a single pressure increment.

Figure 3-9. Repeatability test of sensor for bubble and dew point measurement of impure CO2 at 20 oC.

3.4 Conclusion

This work has demonstrated a sensing approach for bubble and dew point measurements. The method employs thin film interference on the tip of an optical fiber, enabling both (i) highly accurate detection of two phase formation, and (ii) a small-scale PVT cell with no viewing

window. First with respect to accuracy, the well characterized pure CO2 test case provided method validation and indicated an accuracy of 2.8 bar (or 5.0 %). The method was applied to an industrially-relevant CO2 stream, characteristic of post-capture oxyfuel combustion – an

important source for downstream CO2 utilization and storage. Despite relatively low impurity levels, the phase properties varied markedly from the pure CO2 case. The measured dew point values agreed well with current models while the bubble point values for both models and the experiments deviated significantly, indicating the importance of having composition-specific experimental data as enabled by this approach. Second the small-scale nature of this measurement system – enabled by the optical fiber approach – has important implications for testing time and cost. Specifically, the small-scale cell is much smaller volume than traditional systems, resulting in an one-to-two orders of magnitude improvement in characteristic response time (i.e. much less time is required to meet equilibrium for a given data point). Also related to testing time and cost is the potential for automation offered by the clear signal provided by the interference approach. It is also noteworthy that the power level of the sensing system is around -20 dBm (~ 10 µW). This equipment is designed for lab-based measurement, that is, determining the safe pipeline operating range for a given sample. However, the technique could potentially be extended for use in monitoring over long distances. The optical fiber used in this technique SMF- 28 shows low attenuation coefficient (<0.22 dB/km) at 1550 nm. A 1 km length results in only a 5 % reduction in power. Although not the focus here, the fiber optic based method is amenable to monitoring over relevant pipeline distances.

The implications of faster, cheaper and more accurate sensing are underscored by (i) the composition-specific properties of fluid mixtures that make model prediction (particularly of bubble point) difficult; and (ii) the lack of experimental data available for the many important industrially-relevant CO2 mixtures, such as oxyfuel combustion detailed here. In the particular

context of CO2 transport there are additional implications of this work. The higher bubble point pressures predicted here indicate a higher threshold is required for transporting this mixture (and likely similar mixtures) in a single dense phase, with implications for pipeline operation, efficiency, safety and cost.

3.5 Supplemental material

Table 3-1 shows a full listing of bubble and dew point measurements and validated data from Reference Fluid Thermodynamic and Transport Properties Database (REFPROP) developed by

National Institute of Standards and Technology (NIST) for the pure CO2 case.

Table 3-1. Dew points and bubble points of pure CO2 Validated data from NIST Experimental data REFPROF Temperature Bubble point Dew point Saturation pressure (oC) pressure pressure (bar) (bar) (bar) 10 40 42 44.86 11 42 43 45.98 12 42 44 47.12 13 44 45 48.29 14 45 47 49.48 15 47 48 50.69 16 48 49 51.92 17 49 51 53.18 18 51 52 54.46 19 52 53 55.76 20 54 55 57.09 21 55 56 58.44 22 56 58 59.82 23 58 59 61.23 24 60 61 62.66 25 61 62 64.12 26 63 64 65.61 27 64 66 67.13 28 67 67 68.68 29 68 69 70.27 30 70 71 71.89 31 72 73 73.56 32 74 (Critical point at 32 oC) 73.8 (Critical point at 31.1 oC)

Table 3-2 provides a full listing of bubble and dew point measurements and modeling data from

REFPROP and PVTsim for the impure CO2 case (91.7% CO2, 2.5% N2, and 5.8% O2 mole fraction).

Table 3-2. Dew points and bubble points of impure CO2 Modeling Data Modeling data Experimental data from NIST REFPROF from PVTsim Temperat Bubble Dew Bubble Dew point Bubble point Dew point ure point point point pressure pressure pressure (oC) pressure pressure pressure (bar) (bar) (bar) (bar) (bar) (bar) 10 100 49 85.10 52.04 71.78 51.62 11 99 51 85.74 53.42 72.65 52.99 12 98 52 86.38 54.79 73.55 54.48 13 97 54 87.03 56.32 74.44 55.97 14 95 55 87.68 57.90 75.34 57.46 15 94 57 88.32 59.49 76.23 59.00 16 92 59 88.96 61.07 77.11 60.68 17 91 61 89.60 62.65 77.99 62.36 18 90 62 90.24 64.31 78.87 64.04 19 89 64 90.84 66.08 79.74 65.75 20 87 67 91.43 67.91 80.57 67.61 21 85 69 91.98 69.82 81.36 69.59 22 85 71 92.47 73.60 82.08 71.63 82.82 (Maxcondentherm 23 84 74 82.76 73.73 point at 22.9 oC) 24 83 76 ‐ - 83.16 76.21 25 82 79 ‐ - 83.4 79.23 26 82 (Maxcondentherm - - 82.1 (Maxcondentherm point at 26 oC) point at 25.7 oC)

4 Detecting Supercritical CO2 in Brine at Sequestration Pressure with an Optical Fiber Sensor

Monitoring of sequestered carbon is essential to establishing the environmental safety and the efficacy of geological carbon sequestration. Sequestration in saline aquifers requires the

detection of supercritical CO2 and CO2-saturated brine as distinct from the native reservoir brine.

Here we demonstrate an all-optical approach to detect both supercritical CO2, and saturated brine under sequestration conditions. The method employs a long-period grating written on an optical fiber with a resonance wavelength that is sensitive to local refractive index within a pressure- and temperature-controlled apparatus at 40 °C and 1400 psi (9.65 MPa). The supercritical CO2 and brine are clearly distinguished by a wavelength shift of 1.149 nm (refractive index difference of 0.2371). The CO2-saturated brine is also detectable relative to brine, with a resonance wavelength shift of 0.192 nm (refractive index difference of 0.0396). Importantly, these findings indicate the potential for distributed, all-optical monitoring of CO2 sequestration in saline aquifers.

Bao, B., Melo, L., Davies, B., Fadaei, H., Sinton, D., Wild, P. Detecting Supercritical CO2 in Brine at Sequestration Pressure with an Optical Fiber Sensor. Environmental Science & Technology, 47, 306 – 313 (2013). Reproduced with permission from American Chemical Society.

Publication online link: http://pubs.acs.org/doi/abs/10.1021/es303596a

4.1 Introduction

Geological sequestration of anthropogenic CO2 presents an opportunity to reduce global fossil

fuel CO2 emissions by 20 % - 40 % by 2050.[70] Deep saline aquifers are considered ideal

sequestration sites due to large collective capacity (100 to 10,000 GtCO2), wide distribution with proximity to emission point-sources worldwide, high formation pressure and favorable geochemistry.[3] Establishing the environmental safety and efficacy of this approach will

require widespread monitoring of the process in the deep subsurface. Detection of the CO2

plume and CO2-saturated brine is required both to inform the process and ultimately to detect

any possible CO2 leakage on the long term.[71][72]

Existing subsurface CO2 monitoring techniques include seismic[5][73][74][75], geoelectric[6][76] and geochemical methods[7][77][78]. Seismic monitoring technologies have

detected CO2 saturation in brine by measuring the compressional wave velocity in formations in the Sleipner Project.[5] However, the vertical resolution of seismic data was low, ~17 m and 12 m for brine saturated rocks and CO2-saturated rocks, respectively.[79] Geoelectrical sensing using a vertical electrical resistivity array indicated an increase of electrical resistivity of about

200% due to CO2 injection, however the electrical noise was found to limit the technique.[76]

Geochemical analysis of reservoir fluid samples has also been used to study CO2 transport and reactivity in saline formations. However, this analysis is performed at the surface and is not well- suited to in-situ monitoring.[7] The massive scale of sequestration formations will require monitoring approaches that are capable of widely distributed in-situ measurement of CO2 at sequestration conditions.

Sensors based on optical fibers have several inherent advantages including immunity to electromagnetic interferences, compact size, high sensitivity, robustness, low signal loss and capacity for distributed in-situ sensing over large distances.[80][81] Leveraging these advantages, optical fiber sensors have been successfully used to measure parameters such as temperature, pressure, pH and chemical composition.[81] Applications include aerospace instruments,[82] biomedical devices,[83][84] chemical detection devices,[85] structural health monitoring,[86] and energy and environmental processes.[87][88] In the context of subsurface energy and environmental processes, there is precedent for optical fiber based sensing of pressure, temperature, vibration and flow rate.[87][8]

Optical fiber sensors have also been developed for measurement of CO2.[89][90] A chalcogenide

glass fiber based sensor was developed to detect CO2 gas based on its characteristic optical

absorption spectrum.[89] A related absorption-based approach was developed to detect CO2 dissolved in water under atmospheric conditions.[90] These sensors were applied to gaseous and dissolved CO2 but not supercritical CO2 (scCO2), which is the relevant state under sequestration

conditions (i.e. temperatures and pressures exceeding the CO2 critical point of 1070 psi (7.38 MPa) and 31 . Although not in an environmental monitoring context, optical and

spectroscopic methods have been successfully applied to the study of scCO2.[91] [92] An optical

probe was employed to determine the refractive indices of scCO2 and scCO2-ethanol mixture by measuring the reflected light from the end tip.[91] Infrared spectroscopy has been applied to

study forsterite carbonation in the presence of scCO2 and water.[92] In-situ X-ray diffraction and NMR technique have also been used to study reactions between rock formations and

scCO2.[93][94] Although these related technologies show promise, an optical fiber method

suitable for the detection of CO2 under sequestration conditions has not been demonstrated to date.

In this chapter a long-period grating optical fiber sensor is presented for the detection of scCO2 and CO2-saturated brine relative to brine at sequestration temperature and pressure. The resonance wavelength in the transmission spectrum of the optical fiber sensor is sensitive to the refractive index of the surrounding medium and, thus, to the chemical species and phase in the grating vicinity. A pressure- and temperature-controlled apparatus was developed to test the optical fiber sensor at 40 °C and 1400 psi (9.65 MPa). Both scCO2 and CO2-saturated brine were detected relative to brine through the resonance wavelength shift of the optical fiber sensor. This

approach indicates potential for distributed, all-optical monitoring of CO2 sequestration in saline aquifers using optical fibers.

4.2 Experimental

4.2.1 Optical fiber sensor

The optical fiber sensor was developed based on a long-period grating in which the refractive index of the fiber core is periodically modulated over a small distance along the long axis of the fiber.[95] The physics of optical gratings has been detailed in previous studies.[95][96][97] The concept, as applied in this work, is schematically illustrated in Figure 4-1. Briefly, as light traveling through the fiber encounters the grating, specific resonant wavelengths are scattered outwards into the surrounding medium. These wavelengths depend on the period of the grating, the refractive index of the cladding material surrounding the fiber core, and the refractive index of the medium surrounding the fiber. Thus, the resulting transmission spectrum displays an attenuation band directly influenced by the refractive indices of the chemical constituents surrounding the grating. This sensing capacity has been exploited previously in a range of

applications.[95][96][97][98] Brine solutions, CO2-saturated brine, and scCO2, all have different refractive indices, which provides the mechanism for detection in this work.

Figure 4-1. Schematic of the CO2 sensing method developed here. As light traveling through the fiber encounters the long-period grating, specific resonant wavelengths are scattered outwards into the surrounding medium. The resulting transmission spectrum displays an attenuation band. A contrast between the native brine solution and either scCO2 or CO2-saturated brine is detected through a resonance wavelength shift.

The long-period grating (Technica SA) was fabricated in 9/125 single mode silica optical fiber (Corning SMF-28e). The grating had a period of 400 µm, a total length of 20 mm and attenuation of 30 dB at the resonant wavelength of ~1557.7 nm, in air.

For this work, appropriate selection of optical fiber material is critical to ensure its resistance to high-pressure scCO2 and brine. Polymer fiber, for instance, is not suitable. The diffusion of scCO2 into polymers generally causes swelling and changes the physical and optical properties.[99] Here, the fiber core and the cladding are both made of silica, which is resistant to scCO2. Silica materials have been used extensively in micro-reactors and micro-models for the study of scCO2 and brine.[100][101] In addition, silica-based optical fibers have been

successfully employed in other scCO2 applications.[62][102] In most commercial fibers, the cladding material is surrounded by an additional coating layer for mechanical protection. The silica core and silica clad fiber used here was initially coated with polymer, however, this coating was removed at the sensing section prior to testing with scCO2 and brine.

As discussed earlier, a key advantage of optical fiber based sensing is the low attenuation coefficient and, thus, potential for monitoring over long distances. The maximum specified attenuation for Corning SMF 28e fiber is ~ 0.20 dB/km at λ = 1550 nm. Given these loss characteristics, the optical loss will be 0.4 to 1.2 dB for the 2 to 6 km round-trip fiber length scales of sequestration operations (1 to 3 km deep). This optical loss is much smaller than the 30 dB magnitude of the measured resonance attenuation (i.e. to significantly reduce the 30 dB signal requires over 100 km of transmission). Thus, optical fiber signal transmission is well-suited to the length scales of sequestration operations.

4.2.2 High-pressure apparatus

Figure 4-2. Schematic of the high-pressure apparatus used to characterize and test the optical fiber sensor. (a) The setup employed for detecting scCO2 relative to brine solution, and (b) the contents of the cylinders used for detecting CO2-saturated brine relative to brine. The pressure of 1400 psi (9.65 MPa) was monitored by pressure gauges (G1 to G3) while the temperature of 40 °C was controlled by the water bath.

An apparatus was developed to characterize and test the optical fiber sensor, as shown in Figure 4-2. The apparatus integrated a stainless steel test chamber, a flow path, an optical path and a data collection unit. The test chamber was used to immobilize and submerge the optical fiber sensor in high-pressure samples. Ports C (inlet) and D (outlet) of the test chamber were connected to the flow path (labeled as thickened grey lines), while ports A and B were linked to the optical path (labeled as black lines).

Stainless steel tubing (1/16" diameter) built a linkage among the test chamber, cylinders, pumps and valves as shown. Through the inlet (Port C), the test chamber could be filled with the test sample from either of the two vertically-positioned cylinders (Swagelok, 40 cm3, 1800 psi (12.41

MPa)) using a 3-way valve, V6. The Cylinder-1 was connected to the brine supply and to Pump-1

(Simplex, 10,000 psi (68.95 MPa), 20 cc). The Cylinder-2 was connected to the CO2 tank (800 psi (5.52 MPa)) and to Pump-2 (High Pressure Equipment, 30 cc, 10,000 psi (68.95 MPa)). Through the outlet (Port D), the test chamber could be safely purged. The flow of test samples was controlled by the ball valves, V1 to V5. Details of the test chamber structure and its sealing features can be found in Supporting Information.

The optical path guided the signals between the optical fiber sensor and an optical interrogator (Micron Optics, SM125). This unit supplied a broadband input signal, in the 1510 – 1590 nm range, and detected the transmission spectrum through a single channel. The incident and transmitted optical signals of the optical fiber sensor were coupled into this single channel through an optical fiber circulator (Thorlabs, 6015-3-APC). The transmission spectrum was detected at a frequency of 1 Hz by the interrogator. A computer with supporting software package (Micron Optics, ENLIGHT) recorded the transmission spectrum for subsequent analysis.

The pressure and temperature of test samples were controlled and monitored in all tests. The

pressures in the test chamber and both of the cylinders were monitored by pressure gauges G1,

G2 and G3 (2000 psi (13.80 MPa) range, 5 psi (0.03 MPa) accuracy). The test samples and the optical fiber sensor were maintained at a constant temperature by immersion of the test chamber and the two cylinders in a water bath (Fisher Scientific, Isotemp, 5 L capacity, 0.1°C accuracy), shown as the dashed line in Figure 4-2a. Figure 4-2a also shows the contents of the cylinders for

the tests involving brine and scCO2. Figure 4-2b shows the contents of the cylinders for the tests involving brine and CO2-saturated brine.

4.2.3 Characterization of the optical fiber sensor

The long-period grating based optical fiber sensor approach was chosen based on its sensitivity to the refractive index of the surrounding medium. To characterize the sensor, two samples were used as refractive index standards, deionized (DI) water and air. The refractive index of DI water was measured by an Abbe refractometer (Officine Galileo) at 40 °C controlled by circulating water between the refractometer and a heating circulator bath (Haake Fisons). The refractive index of air at 40 °C was taken as 1.0002.[103] The test chamber was filled with air and DI water separately. The apparatus was controlled at 40 °C and kept at atmospheric pressure. The transmission spectrum was detected and the resonance wavelength was recorded.

The responses of the optical fiber sensor to temperature and pressure were characterized separately. In the temperature characterization, the test chamber was filled with DI water at atmospheric pressure and the data were collected between 35 °C to 45 °C, at increments of 1 °C. The temperature control provided by the water bath demonstrated a stability of 0.1 °C. In the pressure characterization, the optical fiber sensor was submerged in DI water at pressure points from 1000 psi (6.89 MPa) to 1800 psi (12.41 MPa) with increments of 100 psi (0.69 MPa).

Pressure was controlled by Pump-1 and monitored by the pressure gauges G1 and G3, while the temperature was held constant at 40 °C. The resonance wavelength was tracked and recorded for each temperature and pressure point.

4.2.4 Detection of scCO2 relative to brine

The optical fiber sensor was tested for its ability to distinguish between scCO2 and brine. The purpose of this test was to demonstrate that the optical fiber sensor could distinguish a localized

plume of scCO2 from native reservoir brine solution.

Test samples of brine and scCO2 were loaded into Cylinder-1 and Cylinder-2 of the high- pressure apparatus, respectively (Figure 4-2a). The brine was a 3M sodium chloride solution (NaCl) prepared with DI water. Both samples were pressurized to 1400 psi (9.65 MPa) and the

temperature was controlled at 40 °C, under which conditions the CO2 becomes supercritical.

The test procedure was designed to detect alternating test samples of brine and scCO2 within the test chamber. Cycles 1, 3, 5 and 7 corresponded to brine injections while Cycles 2, 4, and 6 corresponded to scCO2 injections. Each replacement cycle included a 10-minute time interval for sample replacement and stabilization as well as a 5-minute period for data collection. Injection of the samples into the test chamber was performed over approximately 10 seconds. This injection ensured that the test chamber was fully occupied by the sample. A pressure drop of about 500 psi (3.45 MPa) was observed as a result of each injection operation. However, the test chamber was immediately re-pressured to 1400 psi (9.65 MPa) using the either Pump-1 or Pump-2, as appropriate. The process of sample replacement, including injection and re- pressurization, took approximately 3-4 minutes.

4.2.5 Detection of CO2-saturated brine relative to brine

The optical fiber sensor was used to detect CO2-saturated brine relative to pure brine. The

purpose of this test was to determine if the sensor could distinguish between CO2–saturated brine and the native brine solution, based on refractive index.

The test samples of brine and CO2-saturated brine were prepared and loaded into the high- pressure apparatus, with cylinder contents as shown in Figure 4-2b. A 3M-brine was loaded into

Cylinder-1and the test chamber. Brine and scCO2 were both loaded into Cylinder-2 and allowed

to equilibrate. The brine and the scCO2-brine solutions were both pressurized to 1400 psi (9.65

MPa) with the temperature controlled at 40 °C. The scCO2-brine solution in Cylinder-2 was

considered saturated when no pressure drop was observed at gauge G2 after valve V3 was temporarily closed.

Initially, the test chamber was filled with brine and data were collected from t = 0 to 5 min. At t

= 5 min, V6 was switched to Cylinder 2, and V5 was temporarily opened. This procedure

injected scCO2-brine solution from Cylinder-2 into the test chamber. The full system was re- pressurized to 1400 psi (9.65 MPa), compensating for the pressure drop caused by the injection. Data were collected from t = 15 to 20 min. Subsequently, data were collected for 5-minute periods at 15-minute intervals, with the pressure maintained. The new saturation equilibrium

was considered to be re-established if no pressure drop was observed at gauge G2 or G3 when valve V3 was temporarily closed. Lastly, the test chamber was injected with the original brine solution (t = 125 min).

4.3 Results and discussion

4.3.1 Characterization of the optical fiber sensor

Figure 4-3. a) The transmission spectrum of the optical fiber sensor corresponding to DI water and air at 40 °C. The resonance wavelengths are found to be 1556.102 nm in DI water and 1557.705 nm in air. Given the refractive index value of air (1.0002 RIU) and DI water (1.3309 RIU) at 40 °C, the sensitivity to refractive index (between 1.00 and 1.33) is determined to be 4.847 nm / RIU. b) The resonance wavelength shift as a function of pressure and temperature. The resonance wavelength shift shows linear correlations with both pressure and temperature, with the correlation coefficients (R2) of 0.9881 and 0.9959, respectively. The sensitivities to pressure and temperature are 0.026 nm / 100 psi (0.69 MPa) (1000 - 1800 psi, or 6.89 to 12.41 MPa) and 0.054 nm / °C (35 – 45 °C), respectively.

Figure 4-3a presents the transmission spectrum of the optical fiber sensor obtained during the calibration tests. The resonance wavelengths of the sensor surrounded by DI water and air at 40 °C are 1556.102 nm and 1557.705 nm, respectively, which corresponds to a resonance wavelength shift of 1.603 nm. In comparison, the refractive index of DI water at 40 °C is 1.3309 RIU and the refractive index of air at 40 °C is 1.0002,[103] corresponding to a refractive index difference of 0.3307. The resonance wavelength shift of long-period gratings has been shown to be approximately linear in the region of refractive indices from approximately 1 to 1.4 RIU.[95] Therefore, the sensitivity of the optical fiber sensor to changes in refractive index is calculated to be 4.847 nm / RIU.

Figure 4-3b shows the resonance wavelength shift of the transmission spectrum as a function of pressure and temperature. A linear relationship is observed between the resonance wavelength and the temperature with a correlation coefficient (R2) of 0.9959. The sensitivity to temperature is 0.054 nm / °C in the range of 35 – 45 °C. A linear correlation is indicated between the resonance wavelength and the pressure with a correlation coefficient (R2) of 0.9881. The sensitivity to pressure is 0.026 nm / 100 psi (0.69 MPa) in the range of 1000 to 1800 psi (6.89 to 12.41 MPa).

4.3.2 Detection of scCO2 relative to brine

Figure 4-4 shows the resonance wavelength shift corresponding to the alternating cycles of brine

and scCO2. Each cycle consisted of a 10-minute period for sample replacement and stabilization, and a 5-minute period for data collection (1 Hz sampling frequency). Each 5-minute data collection period gives 300 data points indicating either brine, labeled as red rectangles, or

scCO2, labeled as green diamonds. A moving average of 20 data points was applied to each 5- minute data collection period. A dashed trend line over the 10-minute sample replacement and stabilization period is provided as a guide for the eye. As shown, the scCO2 and brine can be clearly distinguished based on resonance wavelength shift. The average value of the wavelength shift corresponding to each of the measurement intervals are, in chronological order: 0.000, 1.125, - 0.063, 1.111, 0.018, 1.199 and 0.063 nm. The absolute values of the differences between adjacent periods are 1.125, 1.188, 1.174, 1.093, 1.181 and 1.136 nm, giving an average shift of

1.149 nm which demonstrates a significant contrast between scCO2 and brine.

Figure 4-4. The resonance wavelength shift corresponds to the alternate cycles of test

samples between brine (red rectangles) and scCO2 (green diamonds). Each cycle includes a 10-min time gap for sample replacement and stabilization plus a 5-min period for data collection. A moving average curve is shown as a solid line, and a dashed trend line between data collections is shown as a guide for the eye. The brine and scCO2 are distinguished repeatedly and significantly in terms of the resonance wavelength shift, 1.149 nm as the average value.

The repeatability of the optical fiber sensor was investigated by analyzing the standard deviations of all cycles corresponding to the same test sample. The standard deviations of the resonance wavelength shifts related to brine and scCO2 are 0.045 nm and 0.039 nm, respectively.

Both deviations are below 4% of the resonance wavelength shift observed between scCO2 and brine (1.149 nm). These results demonstrate repeatability and, therefore, feasibility in detecting scCO2 relative to brine using the method developed here.

As further validation, the observed wavelength shifts are compared to those expected from the measured sensitivity and literature values for refractive index. Given the measured refractive index sensitivity of 4.847 nm / RIU, the observed refractive index difference between the brine

and scCO2 corresponds to 0.2371 RIU. The refractive index of the brine (3M) is 1.3559 RIU as measured by the Abbe refractometer at 40 °C. Therefore, by subtracting the refractive index

difference from the refractive index of the brine, the refractive index of scCO2 is determined to be 1.1188 RIU at 40 °C and 1400 psi (9.65 MPa). This value is within 1.4% of the refractive index of scCO2 reported previously [69] at the same temperature and pressure condition (1.1347 RIU).

The refractive index of pure brine at 1400 psi (9.65 MPa) is assumed here to be that measured at atmospheric pressure. The pressure effect on refractive index of water has been investigated previously. [104] Specifically, measurements up to 54 MPa indicate a change in refractive index with respect to pressure of ~ 1.5x10-4 RIU/MPa. [104] Thus the test pressure of 1400 psi (9.65 MPa) is expected to change the refractive index by ~ 0.0014 RIU. Given the sensitivity of the long-period grating (4.847 nm/RIU), this pressure effect is expected to induce a resonance wavelength shift of approximately 0.007 nm, which is small compared to the measured shifts of 1.149 nm and 0.192 nm, percentages of 0.61 % and 3.63 %, respectively. Alternatively, the refractive index of liquid can be related to its density by Lorentz-Lorenz formula.[105] [106] [107] Specifically, the density of NaCl solution changes 1.4% when pressure increases to 20 MPa,[108] which indicates a change of ~ 0.7% in brine density for the pressure employed here which would correspond to a ~ 0.0028 RIU shift in refractive index. While it is important to note the effect of pressure on refractive index, from both estimation approaches discussed above, the expected influence is small compared to the shifts measured here.

The potential influence of changes in temperature and pressure on the measurement can also be quantified. Sensor characterization indicated sensitivities of 4.847 nm / RIU, 0.054 nm / °C and 0.026 nm / 100 psi (0.69 MPa) to refractive index, temperature and pressure, respectively. The temperature control of the water bath has an accuracy of ± 0.1°C which could generate a fluctuation in resonance wavelength of ± 0.005 nm. The pressure is controlled within an error of approximately ± 25 psi (± 0.17 MPa) which would correspond to a fluctuation in resonance wavelength of ± 0.006 nm. Therefore, the maximum possible error caused by the sensitivities to

temperature and pressure would be up to ± 0.011 nm, which is less than 1% of the measured resonance wavelength shift of 1.149 nm.

4.3.3 Detection of CO2-saturated brine relative to brine

Figure 4-5. The resonance wavelength shift corresponding to CO2-saturated brine solution (blue circles) as compared to the original brine solution (red rectangles). Data are collected in 5-min periods at 15-min intervals. A moving average is plotted as a solid line, and a dashed line provides a guide for the eye between data sets. The saturation equilibrium was achieved after 100 minutes from t = 5 min when scCO2-brine solution replaced the initial brine. A detectable resonance wavelength shift of 0.192 nm is observed for CO2-saturated brine relative to the original brine solution.

Figure 4-5 shows the measured resonance wavelength shift corresponding to CO2-saturated brine as compared to the original brine solution. The test chamber was initially filled with the brine

solution and after 5 minutes, this solution was replaced with the CO2-saturated brine solution using the cylinder configuration in Figure 4-2b. A solid line in Figure 4-5 indicates the moving average, and the dashed line provides a guide for the eye between data sets. The resonance wavelength shift corresponding to brine (t = 0 – 5 min) provides a reference. The resonance

wavelength observed while injecting the CO2-saturated brine solution increased from t = 15 min and finally approached a stable level (t = 105 - 125 min). This increase in resonance wavelength

is consistent with a decrease in refractive index accompanying CO2 dissolution into brine. The

resonance wavelength of the CO2-saturated brine is a detectable 0.192 nm higher than the

original brine solution. This is the first reported optical measurement of CO2-saturated brine.

It is important to note that the sensor is responsive to carbon dioxide within the solution only so far as CO2 modifies the refractive index of the solution. At the given conditions of temperature

and brine salinity, saturation with CO2 results in a low pH solution (pH = 3.0), in which the majority of the dissolved inorganic carbon is in the form of aqueous CO2 (aq). Specifically,

according to the data for CO2 solubility [109] and the equilibrium model for dissolved inorganic carbon [110], the expected saturation concentrations in the test are as follows: 0.6916 mol/kg - 2- CO2 (aq), 0.009 mol/kg HCO3 and negligible CO3 . As shown, CO2 (aq) represents 99.87% of the total dissolved inorganic carbon (0.6925 mol/kg), and is, therefore, chiefly responsible for measured resonance wavelength shift (0.192 nm) caused by the refractive index change.

Unlike the tests in Figure 4-4, where one pure solution replaces another within the test chamber,

the replacement fluid in Figure 4-5 is initially a CO2-saturated mixture. The solution

replacement process necessarily disrupts this scCO2-brine equilibrium (producing CO2 bubbles) and longer times are required for the solution in the test chamber to re-equilibrate. According to Henry’s law, the pressure drop (at relatively constant temperature) would cause a decrease of solubility so that some of the dissolved CO2 would be temporarily released from the solution. Although the solution was re-pressurized within a few minutes, it takes time for the new solution in the test chamber to recover the original equilibrium composition. During this period, the temperature was maintained by the water bath and the test chamber pressure was maintained by Pump-2. It is important to note that this delay is a function of the solution replacement process, local equilibrium chemistry, and the geometry of test chamber. The inherent sensor response time is governed only by the speed of light and the optical instrumentation and is effectively instantaneous.

After 125 minutes, the CO2–saturated brine solution was replaced with the original brine solution. The corresponding resonance wavelength measurements are shown at the far right of Figure 4-5. The resonance wavelength shift between the brine (recorded during t = 0 through 5 min) and the re-introduction of the brine was -0.039 nm. This difference is small, compared to the intermediate measurements corresponding to CO2 saturation.

Using these results, the refractive index of CO2–saturated brine can also be estimated. By dividing the measured resonance wavelength shift of 0.192 nm by the established refractive index sensitivity of 4.847 nm / RIU, the refractive index difference between CO2-saturated brine and brine is calculated to be 0.0396 RIU (at 40 °C and 1400 psi (9.65 MPa)). Therefore, given that the refractive index of the brine, under the set test conditions, is 1.3559 RIU, the refractive index of the CO2-saturated brine, under the same conditions, can be determined to be 1.3163

RIU. To our knowledge, this is the first published value of the refractive index of CO2-saturated brine solution – a critical parameter for the design of optical instrumentation related to carbon sequestration in saline aquifers.

The detection of CO2-saturated brine also indicates potential to detect CO2 concentrations in brine at levels below saturation. Given the ability to differentiate resonance wavelength shifts

over 0.001 nm, and the measured shift of 0.192 nm corresponding to saturated CO2 concentrations (~0.6916 M), the ultimate limit of detection is on the order of ~ 0.0036 M. It is important to note, however, that in practice the minimum concentration is likely much larger than this ultimate minimum due to associated errors in, for instance, temperature and pressure controls.

In the context of future deployment within or adjacent to geologic sequestration sites, several aspects of the technology are noteworthy. With respect to potential weaknesses, it will be critical to ensure the physical protection of the fiber in the harsh downhole environment while still exposing the periodic sensors to reservoir fluids. Fortunately there is precedent for optical fiber-based measurements in other downhole applications. [8] It will also be necessary to measure local temperature separately and, potentially, pressure using additional parallel or in- line fiber optic sensors, as demonstrated previously.[111] In application, the local solution refractive index is expected to change from reservoir-specific chemistries as a result of dissolution and precipitation of minerals (in addition to the presence of inorganic dissolve carbon

as detected here). Thus similar to controlling for temperature and pressure of a reservoir, reservoir-specific chemistry will need to be considered in practice. In-lab testing as applied here, but with both the applicable reservoir rock and brine, could be used to assess or calibrate for reservoir-specific geochemistries. With respect to strengths, the optical fiber based detection method offers high sensitivity, immunity to electromagnetic interferences, compact size, robustness, low signal loss and - perhaps most important in the context of large scale carbon sequestration operations - capacity for multiplexed sensing over large distances. The efficacy presented here indicates the potential for distributed, all-optical monitoring of CO2 sequestration in saline aquifers.

4.4 Implications

The most significant implication of this work is the demonstration that both scCO2 and CO2- saturated brine are detectable at sequestration conditions using optical fiber sensors. Importantly, the established benefits of optical fiber based sensing for environmental monitoring can be leveraged for monitoring carbon sequestration operations. These results indicate feasibility for deployment in observation wells in test injection sites and full-scale commercial projects. An example of a candidate test injection site with observation wells is the current pilot project in Ketzin, Germany.[112] Planned full-scale commercial projects with drilled observation wells, include the Bell Creek Project (ten wells) and the Shell Quest Project (seven wells).[113] In such

operations, the sensor could report both increasingly CO2-saturated brine, as would be expected

far from the injector in advance of the CO2 front, as well as a scCO2 plume, as would be expected close to an injection well, or through a leakage pathway/fracture. The detection of saturated brine is particularly significant as it could serve as an early indication of the advancing

CO2 front, and also serve to inform models and policy surrounding carbon sequestration operations.

4.5 Supplemental material

Additional information on the test chamber employed in the test setup is provided here. A high- pressure stainless steel chamber was developed to enclose and immobilize the optical fiber sensor, as shown in Figure 4-6a. The chamber was assembled from one threaded pipe nipple (1/8" Pipe Size by 1.5" Length), two threaded pipe fitting Tees (1/8" Pipe Size) and four Yor-lok tube fitting adapters (1/16" Tube OD and 1/8" NPT Male Adapter). The four fitting adapters

function as four ports of the chamber, namely Port A, B, C and D. The optical fiber sensor was tensioned and suspended through the chamber cavity via Port A and B, and, the long period grating section of the optical fiber was adjusted to locate at the center of the pipe nipple. A tight sealing between the fitting and the optical fiber is illustrated in Figure 4-6b. The optical fiber passed through a PEEK sleeve (1/16" OD and 250 µm) and the PEEK sleeve passed through a compression fitting ferrule (1/16" ID). The fitting ferrule compressed on the PEEK sleeve when screwing the nut so that the PEEK sleeve shrunk d and tightly clamped on the optical fiber.

a) C (Inlet) D (Outlet)

Tube Fitting Adaptor

Pipe Nipple Pipe Fitting Te e

A B

Optical Fiber Sensor Long Period Grating b) PEEK Sleeve Fitting Nut Tube Fitting Adaptor

Optical Fiber Sensor Fitting Ferrule

5 Direct Measurement of the Fluid Phase Diagram using Multiplexed Microfluidics

The thermodynamic phase of a fluid – liquid, vapor or supercritical – is fundamental to all chemical processes, and the critical point is particularly important for supercritical chemical extraction. Conventional phase measurement methods require hours to obtain a single datum on the pressure and temperature diagram. Here we present the direct measurement of the full pressure-temperature phase diagram, with 10,000 micro-wells. Orthogonal, linear, pressure and temperature gradients are obtained with 100 parallel microchannels (spanning the pressure range), each with 100 micro-wells (spanning the temperature range). The phase-mapping approach is demonstrated with both a pure substance (CO2) and a mixture (95 % CO2 + 5 % N2). Liquid, vapor and supercritical regions are clearly differentiated, and the critical pressure is measured at 1.2 % error with respect to the NIST standard. This approach provides over 100-fold improvement in measurement speed over conventional methods.

Bo Bao, Jason Riordon, Yi Xu, Huawei Li, David Sinton. Direct Measurement of the Fluid Phase Diagram. Manuscript has been submitted to Journal of the American Chemical Society (Communications).

5.1 Introduction

A major limitation of traditional phase measurement technologies is that only a single pressure- temperature condition can be measured at once. The most common configuration is the pressure- volume-temperature (PVT) cell, common in petrochemical and polymer processing applications.[9][10] These cells typically vary in size between 100 mL and 1 L, and reach pressures and temperatures of 60 MPa and 150 °C, respectively.[9] Since thermal and chemical equilibrium within these large systems must be reached between measurements, obtaining a full map of fluid phase behavior can take months, at considerable expense. The size of the fluid volume also limits the temperature accuracy attainable to 0.5 °C.[114] For extreme temperatures and pressures, high-pressure optical cells (HPOCs) are used.[115]

New methods have recently been introduced that seek to reduce measurement times, often by detecting physical properties instead of direct observation. These techniques include spectroscopy,[116][117] dynamic light scattering,[118] acoustics,[31][119] microwave resonance,[120] shear-mode quartz sensing,[52] quartz-crystal microbalance,[121] pressure-drop measurement [122] and optical fibers.[62] [123] While PVT cells suffer from long measurement times, these new techniques have yet to achieve their success and widespread commercial use.

Microfluidic technologies have emerged as a powerful tool for rapid, parallel measurements, fully leveraging short microscale diffusion times. For example, Shim et al. demonstrated a PDMS microfluidic device that mapped the phase behavior of water/solute fluid mixtures.[124] Glass-silicon microfluidics can convey many of the control and speed benefits associated with lab-on-a-chip applications, with solvent- pressure- and temperature-tolerance required of energy

applications and process industries.[125] The dew point of water in CO2 and minimum

miscibility pressure of CO2 in oil were both measured using a microfluidic approach.[36][37] Pinho et al. and Mostowfi et al. demonstrated microfluidic techniques that rapidly measures multicomponent pressure-temperature phase properties within a single microchannel, including dew point and bubble point, faster than PVT.[35][11] However, phase mapping within a continuous flow suffers from poor precision due to multiphase flow instabilities, high-speed imaging limitations, subjective operator assessment, and impurity accumulation at phase change interfaces.

We here demonstrate direct measurement of the full fluid phase diagram, where a fluid’s physical state is observed within 10,000 individual micro-wells simultaneously, each at a distinct pressure and temperature. Micro-wells are positioned in a square grid, where orthogonal, linear, pressure and temperature gradients are applied (Figure 5-1a). The phase-mapping device is made of glass-silicon to enable high pressures and temperatures, and the high thermal conductivity of the silicon (k ~ 149 W·m−1·K−1) ensures local control of temperature. We characterize the device

using both pure CO2 and a 95 % CO2 + 5 % N2 mixture, and validate these results using NIST reference values. We obtain highly accurate critical points, with critical pressures at 1.2 % of expected values. As opposed to traditional methods that require several days to complete, our phase-mapping device exploits short length-scales and generates the full phase diagram quickly within a single run.

Figure 5-1. a) Schematic of the microfluidic fluid phase-mapping device. a) Full device featuring a 2D array of micro-wells subject to a vertical pressure gradient and a horizontal temperature gradient. Only a few channels are displayed for clarity; the actual phase- mapping device contains 100 horizontal channels, each with 100 micro-wells. b) Enlarged view of micro-wells.

5.2 Measurement of pressure-temperature phase diagram of pure CO2

The device was first characterized using pure CO2. A linear pressure gradient (Plow = 5.5 MPa to

Phigh = 8.0 MPa) was distributed across the network by maintaining a continuous vertical single- phase liquid flow through a serpentine resistor channel. The resistor channel cross-section area (A = 25 µm × 25 µm) was low enough to (i) provide an elevated hydraulic resistance to render out-of-chip resistances negligible, (ii) ensure a manageable flow rate (Q = 0.08 ml/min), and (iii) a low Reynold’s number (Re = 520). One hundred dead-end channels were positioned at

intervals of 250 µm, and run horizontally across the chip from the resistor channel. Each horizontal channel has one hundred orthogonally connected micro-wells (diameter d = 100 µm) at 200 μm intervals (Figure 5-1b), to allow direct observation of phase properties without magnification. Under normal operation, the pressure distributes linearly along liquid phase in the resistor channel (vertically), effectively assigning each horizontal channel (and all 100

corresponding micro-wells) an identical pressure. A temperature gradient (Tlow = 13.7 °C to Thigh = 37.8 °C) was applied horizontally using external cooler and heater blocks. Details of experimental setup and temperature characterization are presented in Supporting Information (Figure 5-4 and 5-5).

Figure 5-2a is a consumer camera image (image size ~ 1” x 1”) of the chip in operation for a preliminary run, which shows how the liquid, vapor and supercritical states are clearly distinguishable by eye. A liquid-vapor saturation line traverses the chip, and becomes increasingly blurred as it nears the critical point. For precise phase mapping, a microscope was used to identify micro-wells nearest the liquid-vapor interface position. Figure 5-2b shows a microscope image of an area containing the critical point. The distinct liquid-vapor interface vanishes at Row 78. In Figure 5-2c, intensity profiles across the liquid-vapor interfaces are plotted for pressures approaching the critical point. A sharp pulse corresponds to sharp light- dark-light liquid-vapor transitions. Broader, shallower pulses are produced nearer the critical point. Through Gaussian fitting, the height and width of these pulses were extracted. The height- to-width ratio, an expression of peak quality, is plotted in the inset of Figure 5-2c as a function of position along the horizontal channels, and fitted to a sigmoidal curve. This method of evaluating liquid-vapor interface quality provides a quantified method of establishing the critical point value, without relying on subjective operator assessment – typical of traditional methods. Here, critical point was measured at 7.47 ± 0.07 MPa / 31.7 ± 0.5 °C, which compares favorably to the NIST reported value of 7.38 MPa / 31.1 oC – a 1.2 % deviation in critical pressure.

To compile the full phase diagram, the position of all liquid-vapor interfaces (for all horizontal channels) was measured. Figure 5-2d shows the measured saturation line and critical point plotted with the NIST reference data. For all phase transition measurements, the standard deviation between the measured and NIST values was 0.03 MPa. The maximum pressure difference with respect to NIST reference data was -0.07 MPa. Similarly for temperature, the standard deviation was 0.2 °C and the maximum temperature difference was 0.5 °C. The

estimated error band based on the pressure (pump) and temperature (heater/chiller) uncertainties was ±0.07 MPa and ±0.5 °C, and all NIST data fall within this range. With regard to resolution, the discretization of the field into micro-wells corresponds to 0.025 MPa and 0.24 °C per micro- well in this test. It is important to note that reducing both the temperature and pressure range across the chip would improve accuracy and resolution – effectively zooming in to a narrower range. The accuracy achieved for even the relatively large P-T range here (Figure 5-2) is on par with existing technologies, such as PVT cells. [114]

While each micro-well could be considered a single “point” on the P-T phase diagram, there is in fact a small temperature gradient of 0.14 °C within each well, providing additional information on phase kinetics. Figure 5-2d shows three microscope images of micro-wells taken along the saturation line. The surface tension of CO2 changes significantly from Plow to the critical point, decreasing from 1.3 mN/m to zero. This marked difference in surface tension results in distinct bubbling kinetics at the interface at various pressures. At lower pressure and temperature, a high interfacial tension prevents the formation of bubbles – only a single liquid-vapor interface is observed (sharp, well-defined meniscus shown in Figure 5-2d, inset i). At higher pressures and temperatures, interfacial tension decreases and rapid bubbling was observed – biased to the high- temperature right-hand side of the micro-well (Figure 5-2d, insets ii and iii). When the liquid- vapor interface in the horizontal channel was directly below a micro-well, nucleation (boiling) occurred on the right “hot” sidewall, with bubbles growing and flowing out of the well before collapsing at a position directly above the interface. This sidewall boiling mechanism is inherent to micro-wells nearer the critical point, where interfacial tension - a barrier to bubble formation – is low. A video showing the bubbling phase dynamics in near-critical micro-wells are provided in Supporting Information (Video 1).

Figure 5-2． Measurement of the pressure-temperature

phase diagram of pure CO2. a) Phase-mapping device in operation, with liquid, vapor and supercritical regions visualized. b) Microscope image of a region of the phase- mapping device with the critical point. Inset images show enlarged views of three liquid-vapor interfaces. c) Pixel intensity profile across corresponding liquid-vapor interfaces. Inset shows how the height-to-width ratio of the pulses changes near the critical point. d) Pressure- temperature phase transition point measurements and validation with NIST reference points. Inset images show typical fluid behavior within micro-wells at various pressure- temperature conditions.

5.3 Measurement of pressure-temperature phase diagram of 95% CO2 + 5% N2 mixture

To demonstrate the applicability of our phase-mapping device to mixtures, the phase diagram of

a binary 95 % CO2 + 5 % N2 mixture was measured. For this experiment, pressures were set at 8.5 MPa and 6.0 MPa for the inlet and outlet, respectively, and the temperature gradient was the

same as for the pure CO2 experiment. In contrast to pure substances that are characterized by a single saturation line, the phase diagram of a fluid mixture is characterized by a phase envelope, bounded by an upper bubble point line and a lower dew point line. Within this envelope, both liquid and gas phases coexist in varying proportions. Figure 5-3 shows the measured pressure- temperature diagram for the mixture, with the expected phase envelope (solid line - NIST reference). The single liquid-vapor saturation line observed was centered between the dew point line and bubble point line of the NIST reference (Figure 5-3). This effective averaging is due to the interconnected nature of the micro-wells, and equilibration to a global state reminiscent of

the fractional distillation process. Specifically, the lighter component (N2) vaporizes first and preferentially accumulates on the right end of horizontal channels, resulting in a single effective liquid-vapor saturation line leading to the critical point. Near the critical point, intense bubbling behavior is observed as shown inset in Figure 5-3. At the region just below critical point, several unique bubbling phenomena were observed at the interface (See Figure 5-3 insets i, ii, iii, and iv). The dynamic behavior of bubbles in the four insets was recorded in videos (see Supporting Information, video 2). While a traditional phase envelope was not produced, the system nonetheless led to highly accurate critical point measurement: 8.05 ± 0.07 MPa /26.2 ± 0.5 °C and compares favorably to the NIST reference critical point, 8.15 MPa / 27.0 °C predicted for this mixture – a 1.2% difference in critical pressure.

5.4 Phase-mapping device accuracy and speed

Notably, the addition of 5 % N2 to the CO2 sample caused a 10 % increase in the critical point pressure, and the device was shown fully capable of measuring each of these critical pressures at 1.2 % error with respect to NIST reference values. The ability of the phase-mapping device to correctly measure critical point over a wide range underscores the importance and effectiveness of the method at measuring the critical point.

Regarding fundamental measurement time, two metrics are noteworthy, (i) equilibrium time and (ii) degree of multiplexing. The phase-mapping device has 10,000 micro-wells operating in parallel (2 × 10-4 μL per micro-well, or 2 μL for the entire array) that equilibrate in seconds. The phase-mapping device required ~20 s to re-equilibrate after a 0.1 MPa pressure change (See Figure 5-6 and Video 3 in Supporting Information). In practice, the experiment with 10,000 micro-well data points described in Figure 5-2d required 3 h to complete, including 1 h of preparation, 1 h of test, 1 h of temperature calibration and post-processing. In contrast, a traditional PVT system typically takes 8 to 10 h for a single P-T data point.[11] A minimum of

800 hours (100 P-T data points) would be needed using a traditional PVT system to achieve similar resolution – an over 100-fold longer measurement time.

5.5 Conclusion

In conclusion, we demonstrated direct mapping of the P-T diagram – where 10,000 data points were measured simultaneously, fully leveraging rapid micro-scale equilibrium times for 100-fold improvement in P-T mapping speed over the conventional PVT method. The device provided accurate measurement of the critical point, demonstrating critical pressure measurements with

1.2 % error with respect to NIST values for both the pure CO2 and a 95 % CO2 + 5 % N2 mixture.

5.6 Supporting information

S1. Experimental setup

Figure 5-4. a) Schematic of the experimental setup. Water was flowed between baths and chiller and heater blocks above the phase-mapping device to produce a temperature gradient. Two pumps (Teledyne Isco) were attached to the inlet and outlet to pressurize the system, and drive single-phase flow through the resistor channel. A piston (filled with the fluid of interest) was used to isolate the first pump from the device (which was filled with water). A custom stainless-steel manifold was used to provide a strong connection to the device. A microscope monitored the phase-mapping device during operation. Pressure transducers measured pressures at the inlet and outlet throughout the experiments. b)

Image of the phase-mapping device operating with pure CO2.

S2. Temperature characterization

Figure 5-5. Temperature characterization of the phase-mapping device: a) Image of temperature calibration test using propane. b) Temperature distribution of phase-mapping

device during test of pure CO2 and mixture of 95% CO2 + 5% N2.

The temperature distribution across the phase-mapping device was first characterized by a calibration fluid with known thermodynamic properties: propane. Propane was chosen since it’s

critical temperature is higher than for either pure CO2 and the 95% CO2 + 5% N2 mixture, and thus allows temperature calibration over the full temperature operation range. During the phase- mapping experiments with both pure CO2 and the 95% CO2 + 5% N2 mixture, propane was filled in the temperature calibration channel (see TCal in Figure 5-1a), a dead-end channel (with micro- wells) fully isolated from the phase-mapping channel network, but running parallel to the horizontal channels and thus subject to an identical temperature distribution. The temperature distribution along the calibration channel (and thus along the network as well) was measured by incrementing the pressure at the channel entrance from 0.7 to 1.3 MPa, and measuring the coordinates of each equilibrium liquid-vapor interface position (Figure 5-5). NIST values of known saturation pressures for propane were used to associate these coordinates (at set pressure)

with theoretical temperatures – providing a full temperature map for use in the pure CO2 and the

95% CO2 + 5% N2 mixture tests. This calibration was repeated during both the pure CO2 and the

95% CO2 + 5% N2 mixture tests.

S3. Dynamic response of the phase-mapping device to pressure change

Figure 5-6. The response of liquid-vapor interface position to inlet pressure changes. Insets show microscope images of interface positions at various pressure states.

The liquid-vapor interface position response to rapid pressure changes was evaluated. Once the system reached equilibrium, the inlet pressure was decreased from 8.0 to 7.9 MPa, and the interface position was monitored. After a new equilibrium was reached, the pressure was increased from 7.9 to 8.0 MPa. Figure 5-6 shows the corresponding changes in interface positions. Insets show the images of equilibrium interface positions. The interface position response to pressure decrease and increase were fitted to exponentials, with characteristic response times of 4.3 s and 0.7 s, respectively. This much faster response to pressure increase is explained by free moving gas molecules hitting the liquid-vapor interface and condensing – a much physically easier process than liquid molecules escaping from a confined liquid phase through evaporation. This test also showcases the reproducibility of the system: equilibrium high pressure states are near-identical. Additionally, the temporal fluctuations in the signal showcase the inherent noise in our signal – largely due to pressure fluctuations from the pump. The standard deviation on the baseline position was 8 µm, which corresponds to ~4 % of the micro- well spacing. This rapid response to pressure change (a few seconds) is a key feature of our phase-mapping device, in sharp contrast to macroscale devices which require hours to equilibrate.

6 Bubble Nucleation and Growth in Nanochannels

We investigate vapor bubble nucleation and growth in a pure liquid confined in sub-100-nm channels through isothermal cavitation. Measured cavitation pressures for the test fluid, propane, are compared with those predicted from the nucleation theory. We find that cavitation pressure in the nanochannels corresponds closer to the spinodal limit than that predicted from classical nucleation theory. Three of the five cavitation experiment measurements exceed the spinodal limit. Depending on the initial nucleation location – along the channel or at the end - two types of bubble growth dynamics were observed. Bubble growth was measured experimentally at five temperature and pressure conditions. A fluid dynamics model is developed to determine the evaporation rate, instantaneous bulk liquid velocity, and bubble pressure. Collectively these results demonstrate, characterize, and quantify isothermal bubble nucleation and growth of a pure substance in nanochannels.

Bao, B., Zandavi, S. H., Li, H., Mostowfi, F., Sinton. D. Bubble Nucleation and Growth in Nanochannels. Manuscript has been submitted to Physical Review Letters.

Note that the theory in Section 6.3.1 is a collaboration work.

6.1 Introduction

Liquids become metastable by either super-heating beyond the liquid–vapor saturation temperature or reducing pressure below the saturation pressure [126][127]. Metastable liquids have been observed in many systems, such as, water transport in trees [126], the sonocrystallization of ice [128], filtering processes, hydraulic fracturing of hydrocarbon-rich nanoporous shale [26], hydraulic machinery [129], and targeted drug delivery in nanoporous organs [130]. The stability of metastable liquids has been extensively studied for a long time with a variety of theoretical, experimental, and numerical methods. The metastable liquid will eventually transform into the more thermodynamically stable phase, vapor, via a bubble nucleation mechanism [131]. The classical homogeneous nucleation theory is typically used to study the fundamental physics of bubble nucleation in a pure liquid [131][132]. However, there is a huge discrepancy between experimental results and nucleation theory predictions[131][133].

This discrepancy is usually attributed to the inhomogeneities of surfaces, external initiating actions, and the presence of microbubbles in the liquid or solid particles on the surface [131][134].

The rupture of the metastable liquid phase occurs by the nucleation of a nano-sized bubble. Therefore, investigation of the stability of the liquid at nanoscale is crucial. The development of micro- and nanofluidics techniques provides a unique solution to realize vapor bubble nucleation in liquids [133][135][136]. Ando et al. [135] reported homogeneous bubble nucleation achieved by a laser-induced shock on a free surface. Nagashima et al. [137] detected a single bubble nucleation in a single nanopore with the radius of 43.5 nm by extreme pulsed superheating. Vincent et al. [136] observed that microbubbles appear within microseconds in a confined cavity when the liquid is at a tension condition. Vincent et al. [39] used microfabricated structures to observe the nucleation and growth of cavitation bubbles during drying in the ink-bottle nanostructure. They found that the liquid phase pressure at cavitation was less than that expected from the nucleation theory. A recent study [38] found that larger micrometer-sized cavities nucleate vapor at a lower superheated temperature than smaller nanometer-sized cavities. Duan et al. [138] reported evaporation-induced cavitation of water in 20-120 nm deep nanochannels.

Once a bubble nucleates, the rate of bubble growth is dictated by several factors such as surface tension, viscosity, expansion due to pressure difference and heat transfer [139]. The dynamic of the freely growing bubble has been studied for many years starting from the works of Plesset [140] and Forster [141] in the 1950’s. The investigation of bubble growth in a confined geometry was first initiated in inkjet printing [142] studies, and it was accelerated by the development of microfluidics systems [143][144]. The growth and collapse of a vapor bubble inside a microfluidics system has been studied both experimentally [144][145] and theoretically [146]. In the experimental studies, the vapor bubbles were created by the initial localized heat or short duration laser pulse [145]. However, bubble nucleation and growth of a pure substance in nanochannels has not been demonstrated to date.

In this chapter, we study the bubble nucleation and growth in liquid propane in 85-nm channels. The micro/nanofluidic chip allows direct visualization of bubble nucleation and growth. The measured nucleation temperatures and pressures are compared with nucleation theory. We also investigate the growth rate of vapor bubble column after nucleation. Based on the location of

bubble nucleation, two different growth mechanisms are distinguished. The cavitation-induced vapor column growth is used in a fluid dynamics model to predict the evaporation rate, liquid velocity, and bubble pressure.

6.2 Experimental setup

A schematic of the experiment is shown in Figure 6-1. The test region consists of 10 identical dead-ended nanochannels of 85-nm depth (7 μm x 35 mm). The nanochannels are fabricated by deep-reactive-ionic-etching on a silicon wafer and anodic bonding with glass. The assembled nanofluidic chip is installed on a customized manifold enabling appropriate sealing for high pressures. The chip and manifold are connected to a sample source cylinder and pump via tubing and valves. During this process, the temperature and pressure are well maintained and monitored. To ensure nucleation occurs within the bulk liquid phase instead of the liquid-vapor interface [138], we introduce a temperature difference across the chip by pairing a heater and chiller clamped on top of the chip, isolating the end portion of the relatively long nanochannels.

Specifically, the dead-end portion of the nanochannel is set at a higher temperature (Thigh), while

the entrance portion (near the inlet port) is kept at a lower temperature (Tlow). The field of view

falls into the Thigh region. Therefore the temperature of the liquid phase, Tl , is also Thigh. The heater and chiller are copper-made blocks with ethylene-glycol circulated from a temperature- controlled bath. Pressure is provided and controlled (± 0.5% accuracy) by a pump and monitored by a pressure transducer near the chip. An isolation piston cylinder is used between the pump and chip to avoid contamination of the sample.

The tubing, pump, piston chambers and valves are cleaned thoroughly and connected to the nanofluidic chip. The entire system is vacuumed for 15 minutes. Research-grade propane (Praxair 99.99%) is filled into the nanochannels. Isothermal cavitation tests are performed

systematically, with Thigh and Tlow kept constant and pressure set far above the saturation pressure

at Thigh (up to 4 MPa). We waited for 15 minutes at each pressure step to observe bubble nucleation. Cycles are repeated until bubble nucleation occurs. In total, five different

temperatures (Tl = 343, 347, 352, 357 and 362 K) are performed for the investigation of the temperature effect on the metastability of the liquid phase. All individual conditions in all tests are performed four times to ensure repeatability. Bubble nucleation and growth are visualized in the bright-field mode by an inverted microscope mounted with a 10 x objective. The field of

view is a rectangle (~ 2.0 x 1.5 mm) encompassing on the dead-end portion of the group of 10

nanochannels. The measured value of the cavitation pressures ΔPcav = Psat (Tl) - Pl and the corresponding superheat ΔTsup = Tl – Tsat (Pl) are provided in Supplemental Material.

6.3 Results and discussion

6.3.1 Bubble nucleation

The superheating and cavitation values along with the saturation curve are shown in the phase diagram of propane in Figure 6-2. The dependence of the cavitation pressure on liquid phase temperature has been obtained. The values of ΔTsup and ΔPcav decrease as the experimental condition moves toward the critical point of propane. The maximum value of the superheat

measured in the experiments is ΔTsup = 39.1 K at Pl = 1.10 MPa. Similarly, the maximum value

of cavitation pressure in the experiments is ΔPcav = 1.49 MPa at Tl = 343 K. Each ΔPcav value in Figure 6-2 is verified by four independent rounds of tests.

Figure 6-2. Measurements of bubble nucleation in 85-nm nanochannels. Plotted for comparison are the saturation vapor pressure, capillary pressure calculated from the

Kelvin Equation, Eq(6-1), the prediction from the classical nucleation theory, Eq(6-2) for tN = 900 s, and the spinodal limit from Eq(6-3).

The cavitation process initiates by an explosive growth of seeding bubbles in liquids. The necessary conditions for thermodynamic equilibrium lead to an expression for the critical radius,

Rc , which is a thermodynamic property of the system. When the radius of a nuclei bubble is

bigger than the critical radius Rc it can grow unlimitedly [8]. For a confined geometry, a

necessary, but not sufficient, condition for bubble nucleation in a nanochannel is 2 ,,

where h and w are the channel depth and width[129]. This inequality provides the minimum degree of superheating and cavitation required for bubble nucleation. For propane in a nanochannel with a depth of h = 85 nm, the minimum conditions required can be calculated from the Kelvin equation [147]:

, (6-1)

where γ is the liquid-vapor surface tension; kB and are the Boltzmann constant and the density of the liquid at the saturation condition, respectively. Here, the minimum superheating required

83 calculated from Eq. (6-1) is plotted (Figure 6-2), and corresponds closely to the saturation condition of bulk propane. This means that the minimum required degree of superheating in the 85-nm channel is not dictated by the channel dimensions.

The kinetic limit of the metastable liquid phase or the onset of homogeneous nucleation can be obtained from the classical nucleation theory [131]. The liquid phase cannot remain in a metastable state for an arbitrarily long period of time and the nucleation theory predicts the time, tN, for which a liquid can be held at pressures below (or temperatures above) the saturation, before the first bubble nucleation would be expected from the motion of the molecules in the bulk liquid phase. A simple form of nucleation theory is used to obtain the nucleation pressure,

,[147]

, (6-2)

where is the volume of the liquid under observation, the nucleation time and η is simply

exp .

In Eq. (6-2), Z is the rate constant and estimated to be 25 x 1030 (cm3 s)-1 by Volmer [147]. Since

the liquid-vapor surface tension is a decreasing function of temperature, will vary with the liquid temperature. As can be seen in Figure 6-2, the nucleation theory, given a waiting time of tN = 900 s and an observation volume of 85 nm x 7 μm x 1 mm, over predicts the cavitation pressures.

The theoretical extreme limit of the metastable liquid is the spinodal limit; the point at which the slope of the P-V phase diagram at a constant temperature, (dP/dV)T, changes from negative to positive. Here, the Furth formula is used for the spinodal limit: [148]

1.32 3 . (6-3)

This equation has been shown to work well in the region of positive pressures [149]. As shown in Figure 6-2a, closer agreement is observed between the experimental results and those from the spinodal limit rather than the theoretical values calculated from the nucleation theory. Two of the five measurements do not exceed the spinodal limit while three of them do. The maximum

difference between the measured superheat and the predicted one from the Furth formula is 4 K at the closest condition to the critical point. The maximum superheat value reported for propane is 52.9 K at 0.1 MPa [150]. This extreme experimental value of superheat agrees closely with the predicted value from the nucleation theory, 55.2 K [150]. The difference between the measured nucleation pressures and the predictions from the spinodal limit can be related to the critical point shift in the nanosized pores (channels) [151]. The reported value of this shift is below 1 K in a 23 nm pore. Importantly, these results show bubble nucleation onset under uniform temperature and controlled pressure in nanochannels. It is noteworthy that 1) these are the first experimental result of hydrocarbon bubble nucleation temperatures and pressures in sub-100 nm channels; and 2) Critical point shift effect is not experimentally determined here because it is not in the scope of this chapter.

6.3.2 Bubble growth

Following nucleation of a vapor bubble, a vapor column grows in the nanochannels. An image sequence showing bubble growth in a nanochannel versus time is recorded by a video camera with a sampling speed of 50 frames per second (Supplemental video shows bubble nucleation and column growth in the ten nanochannels at all nucleation conditions). Figure 6-3 shows the typical bubble column growth mechanisms found in our experiment. In all the cavitation tests, we observed two types of bubble column growth mechanisms according to the location of bubble nucleation: Type A where the bubble nucleates at channel-end, and Type B where the bubble nucleates along the channel. Figure 6-3a and 6-3b show image sequences of Type A and B

growth, from our cavitation test at Tl = 347 K and Pl = 1.60 MPa, respectively. The vapor and liquid phases are clearly distinguished by color or intensity. The vapor phase appears bright while the liquid phase appears darker. A scale bar on the plot indicates the channel dimension and the vapor column length.

Figure 6-3 ． Mechanisms of bubble column growth at

Tl = 347 K and Pl = 1.60 MPa: a) Image sequence of Type A growth where vapor bubble nucleates at channel-end; b) Image sequence of Type B where vapor bubble nucleates along the channel; c)

Bubble length lB versus time of Type A and B growth; d) Positions of left and right liquid-vapor

interfaces, lL and lR, bubble

length lB of Type B growth. Three distinct growth regimes can be identified: “Transient start-up”, “Transitional” and “Steady linear growth”.

The bubble column length and liquid-vapor interface positions are extracted from image sequences using an image processing algorithm based on pixel counting with a threshold intensity. As shown in Figure 6-3c, Type A growth experiences two regimes, namely “transient start-up” and “steady linear growth”. In the case shown in Figure 6-3c, transient start-up regime

takes about 200 ms. In this figure, lB is defined as vapor bubble column length. The velocity of liquid-vapor interface (vapor bubble column growth rate) is initially high (~ 4.6 µm/ms), and becomes slower afterward. The transient start-up regime smoothly transitions into the steady linear growth regime, where the bubble column grows at a lower and constant velocity of ~ 0.32 µm/ms. In contrast, the Type B growth curve presents three different regimes, namely “transient start-up”, “transitional” and “steady linear growth” regimes, as shown in Figure 6-3c.

In the Type B growth, Figure 6-3b, both the right and left interface movement contribute to the vapor bubble column growth. Therefore, both left and right interface positions are tracked, as

shown in Figure 6-3d. lL and lR are defined as the distance between the channel-end and the left

and right liquid-vapor interfaces, respectively. The vapor bubble length is simply lB = lL - lR (Figure 6-3d). In the transient start-up regime, the right liquid-vapor interface moves at a constant velocity. While the right interface moves linearly in Type B, the left liquid-vapor interface moves non-linearly in this regime, similar to Type A. When the right liquid-vapor

interface hits the channel-end (lR = 0), there is a significant slowing in the bubble growth – an effective pause prior to subsequent steady linear growth (Fig 3d). We term this ~ 0.62 s period the transitional regime, which is unique to Type B growth. It was seen that lB could grow slightly, or remain unchanged, or even shrink in this regime depending on the experimental conditions, temperature and pressure (examples of both Type A and B growth at each of the other four nucleation conditions are presented in Supplemental Material). Finally, in the steady linear growth regime, the bubble column grows at a constant velocity (0.32 µm/ms), which is equal to the velocity of the steady linear growth regime in Type A growth. Surprisingly, for all five conditions tested, the bubble growth curves converge eventually in steady linear growth regime regardless of type or initial nucleation location. That is, while Type A and B growth show distinct dynamics in earlier times, the ultimate interface position and rate of advance are the same for all cases once in steady linear growth regime. As shown in Figure 6-3c at 347 K, the liquid-vapor interface for the two growth types reach the same position (620 ± 10 μm) after 870 ms from the time of nucleation (See Supplemental Material for all cases).

Figure 6-4a shows Type A growth of the vapor bubble column at five different temperatures. In order to show the data clearly, only one channel is selected as representative at each temperature. As shown, the growth of Type A in the nanochannels depends on both temperature and pressure in the liquid phase.

Early in the transient start-up regime, the pressure in the bubble is near its maximum, and the rate of bubble expansion is governed by the balance of pressure force and surrounding fluid inertia[139]. The bulk liquid flow in the nanochannel is approximated as one-dimensional flow. A schematic diagram of a simple one-dimensional model of the system is shown in the inset of Figure 6-4b. The liquid phase pressure at the entrance of the nanochannel is assumed to be equal to the pressure in the reservoir Pres (the pressure drop in microchannels is neglected), and the

pressure in the vapor phase Pv(t) is assumed to be uniform across the bubble. The resulting governing equation of the liquid motion in the nanchannel is [144]

, (6-4)

where uliq(t) is the bulk liquid velocity in the channel, Dl = 35 mm is the distance from the liquid-

vapor interface to the reservoir and can be assumed to be constant since Dl >> lB, and f is the flow friction. For laminar flow in rectangular channels and for h / w << 1 the flow friction is 2 approximately f = 12µDl/h , where µ is the bulk viscosity [30].

The time derivative of the bubble length can be viewed as a summation of uliq(t) and a component related to the evaporation at the liquid-vapor interface

, (6-5)

where Nv is the moles of vapor and its time derivative dNv/dt is the evaporation rate at the liquid- vapor interface, M is the molar mass of propane, and A is the cross sectional area of the nanochannel. The vapor phase is approximated as an ideal gas. Differentiating the ideal gas law with respect to time for an isothermal system gives:

, (6-6)

where R is universal gas constant.

The differential equations (6-4) to (6-6) form a coupled system with three unknowns, uliq(t), Nv(t) and Pv(t). Solving this system of equations requires a boundary condition on uliq(t), Nv(t), or Pv(t). For the steady linear growth regime, the pressure inside the bubble column is assumed to be close to the saturation pressure at the vapor temperature, i.e., Pv = Psat. This assumption is valid only in the steady linear growth regime[132]. By using this assumption, Eq. (6-6) is applied to

predict dNv/dt in the steady linear growth regime. The evaporation rate dNv/dt is assumed to be constant for each experimental condition and matching that of the steady linear growth regime. -13 The inset in Figure 6-4a shows that the evaporation rate, dNv/dt, decreases from 2.02 x 10 to 0.99 x 10-13 mol/s as the temperature increases from 343 to 362 K (superheat decreases from 39 to 13 K).

The calculated value of dNv/dt and the measured growth data lB(t) are applied in Eq.(6-5) to

calculate the liquid velocity uliq(t). We take the case of Type A growth at T = 347 K (same as the one shown in Figure 6-3a). The result is that 89% of this liquid-vapor interface movement

dlB(t)/dt results from the liquid flow uliq(t) and the remaining is due to evaporation dNv/dt. Similarly for the other nucleation conditions, the ratios are 90% (343 K), 88% (352 K), 86% (357 K) and 84% (362 K).

Figure 6-4. a) Vapor bubble column growth lB(t) of Type A at five different temperatures; inset shows the calculated evaporation rate dNv/dt in steady linear growth regime; b) The

calculated pressure in the vapor phase Pv versus time of the bubble nucleation experiment at T = 347 K. Saturation pressure and the reservoir pressure are also plotted for comparison.

In the transient start-up regime, a high vapor pressure is required to counter the fluid inertia. The

momentum equation Eq.(6-4) with calculated uliq(t) is used here to predict Pv as a function of

time. Figure 6-4b shows the resulting value of Pv(t). As shown, Pv is predicted to be as high as

14.7 MPa, based on the first two measurement points, i.e. 20 ms interval. The vapor pressure Pv decreases sharply with time, approaching a plateau level of 2.52 MPa in the steady linear growth regime. The transient regime of Pv agrees with the general trend observed previously for growth and collapse of a vapor bubble column with transient heating in a microtube [22]. Similarly, the calculated vapor pressure for all cases is provided in the Supplemental Material.

Comparison between the saturation pressure Psat and the Pv in the steady linear growth regime

provides a measure of validation for the model. Here, the calculated Pv in the steady linear

growth regime (2.52 MPa) is 10 % lower than Psat (2.80 MPa). Similarly for the other nucleation conditions, the deviations are 14% (343K), 13% (352 K), 17% (357 K), and 14% (362 K). The

somewhat lower prediction of pressure (pv) is an indication of higher viscous friction in the nanochannels as compared to the theoretical value. Previous studies have also shown that the predicted values of the velocity from one-dimensional theoretical models are higher than the actual measured values in nanoconfinements[153][154]. In general, this phenomenon has been attributed to surface roughness, electroviscous effects (in polar fluids), variations in dynamic contact angle, the presence of bubbles, and/or the formation of highly viscous layers near the interface[153][32][33][34]. The maximum deviation (17%) observed here is relatively small as compared to reported values for water (23% in 53-nm channel [34]), ethanol (77% in 45-nm channel [32]) and isopropanol (95% in 45-nm channel [32]) in capillary filling experiments. Lastly, the maximum Reynolds number in the channel is 2.5, which supports the viscous flow assumption.

6.4 Conclusion

In this chapter, we investigate bubble nucleation and growth in liquid propane confined in 85-nm channels. In contrast to the bulk phase, it is found that the cavitation pressure in the nanochannels corresponds closer to the spinodal limit than that predicted by classical nucleation theory. For the growth of the vapor column following nucleation, two different growth mechanisms are distinguished depending on the location of the bubble nucleation. Finally, a fluid dynamics model is developed to determine the evaporation rate, liquid velocity, and bubble

91 pressure during bubble growth. Collectively these results demonstrate, characterize, and quantify isothermal bubble nucleation and growth of a pure substance in nanochannels.

Supplemental Material contains a complete list of bubble nucleation conditions, left and right interface velocities in ten nanochannels, and bubble growth curves of both Type A and B at other four conditions (similar to Figure 6-3c and 6-3d). The supplemental video shows the bubble nucleation and column growth in the ten nanochannels at all conditions.

6.5 Supplemental material

Nomenclature Symbol Meaning Symbol Meaning A cross sectional area of the channel R universal gas constant Dl distance from liquid-vapor interface to reservoir Rc critical radius dNv/dt evaporation rate at the liquid-vapor interface T temperature (dP/dV)T slope of P-V diagram at a constant temperature Tsat saturation temperature f flow friction Thigh temperature of copper heater h channel depth Tlow temperature of copper chiller kB Boltzmann constant Tl temperature of liquid phase lB vapor bubble column length tN waiting time for bubble nucleation lL position of left liquid-vapor interface uliq(t) bulk liquid velocity vs. time lR position of right liquid-vapor interface V volume M molar mass of propane volume of liquid under observation Nv moles of vapor w channel width P pressure Z constant rate Pl pressure of liquid phase γ liquid-vapor surface tension nucleation pressure predicted by nucleation theory µ viscosity of liquid Pres pressure in the reservoir density of liquid phase Psat saturation pressure ΔPcav cavitation pressure Pv(t) pressure in vapor phase vs. time ΔTsup superheat temperature

The five bubble nucleation conditions are listed in Table 6-1.

Table 6-1. Bubble nucleation conditions

Tl (K) Pl (MPa) ΔTsup (K) ΔPcav (MPa) 343 1.10 39.1 1.49 347 1.60 27.0 1.20 352 2.00 21.6 1.08 357 2.40 17.7 0.97 362 2.90 13.0 0.80

Figure 6-5. Complete data of bubble nucleation time and associated growth types of the ten

nanochannels at five nucleation conditions.

Figure 6-6. Complete data of left interface velocity uL in the steady linear growth regime of the ten nanochannels at five nucleation conditions.

Figure 6-7. Complete data of right interface velocity uR of Type B growth in the nanochannels at five nucleation conditions.

Figure 6-9. Mechanisms of bubble column growth at Tl = 343 K and Pl = 1.10 MPa: a)

Bubble length lB versus time of Type A and B growth; b) Positions of left and right liquid-

vapor interfaces, lL and lR, of Type B growth; c) Predicted bubble pressure Pv (Type A) during bubble column growth.

Figure 6-10. Mechanisms of bubble column growth at Tl = 352 K and Pl = 2.00 MPa: a)

Bubble length lB versus time of Type A and B growth; b) Positions of left and right liquid-

vapor interfaces, lL and lR, of Type B growth; c) Predicted bubble pressure Pv (Type A) during bubble column growth.

Figure 6-11. Mechanisms of bubble column growth at Tl = 357 K and Pl = 2.40 MPa: a)

Bubble length lB versus time of Type A and B growth; b) Positions of left and right liquid-

vapor interfaces, lL and lR, of Type B growth; c) Predicted bubble pressure Pv (Type A) during bubble column growth.

Figure 6-12. Mechanisms of bubble column growth at Tl = 362 K and Pl = 2.90 MPa: a)

Bubble length lB versus time of Type A and B growth; b) Positions of left and right liquid-

vapor interfaces, lL and lR, of Type B growth; c) Predicted bubble pressure Pv (Type A) during bubble column growth.

7 Conclusions

This thesis first provides a short literature review of the existing experimental methods for fluid phase measurement (Chapter 2). According to the inherent principles and scales, these methods are categorized into PVT experiments, optical methods, electrical and acoustic methods, microfluidic methods and nanofluidic methods. This literature review summarizes the working principles, advantages and shortcomings of each method in each category.

This thesis then describes my contribution of four novel methods for fluid phase measurement. Specifically, the methods enrich and diversify the portfolio of optical, microfluidic and nanofluidic methods: (i) Optical thin-film interference based bubble and dew point sensor probe (Chapter 3); (ii) Refractive index based optical fiber sensor (Chapter 4); (iii) Multiplexed microfluidic based phase diagram mapping (Chapter 5); and (iv) Nanofluidic based measurement of bubble nucleation and growth (Chapter 6). These fluid phase measurement methods are tested

and verified with reservoir fluids, including CO2, CO2-O2-N2 ternary mixture, brine, CO2- saturated brine, CO2-N2 binary mixture for the industrial process of carbon dioxide transport/storage, and, pure hydrocarbon propane for shale gas/oil production.

7.1 Fluid phase measurement using optical methods

The bubble and dew point sensor probe employs thin film interference on the tip of an optical fiber, enabling both (i) highly accurate detection of two phase formation, and (ii) a small-scale PVT cell with no viewing window. First, with respect to accuracy, the well-characterized pure

CO2 test case provided method validation and indicated an accuracy of 2.8 bar (or 5.0%). Second, the small-scale nature of this measurement system – enabled by the optical fiber approach – has important implications for testing time and cost. The small-scale cell is much smaller volume than traditional systems, resulting in an one-to-two orders of magnitude improvement in characteristic response time (i.e. much less time is required to meet equilibrium for a given data point). Also related to testing time and cost is the potential for automation offered by the clear signal provided by the interference approach. The technique is primarily designed for laboratory based measurement. Due to a lower power loss in the optical fiber, It could potentially be extended for use in monitoring over long distances on pipelines.

The second contribution to optical methods is the development of refractive index based optical fiber sensor. The resonance wavelength in the transmission spectrum of the optical fiber sensor is sensitive to the refractive index of the surrounding medium and, thus, to the chemical species and phase in the grating vicinity. A pressure- and temperature-controlled apparatus was developed to test the optical fiber sensor at 40 °C and 1400 psi (9.65 MPa). Both scCO2 and

CO2-saturated brine were detected relative to brine through the resonance wavelength shift of the optical fiber sensor. This approach indicates potential for distributed, all-optical monitoring of

CO2 sequestration in saline aquifers using optical fibers. Specifically, the results indicate feasibility for deployment in observation wells in test injection sites and full-scale commercial

projects. The detection of scCO2 relative to brine in native reservoir is particularly significant as

it could serve as an early indication of the advancing CO2 front, and also serve to inform models and policy surrounding carbon sequestration operations.

7.2 Fluid phase diagram mapping using multiplexed microfluidics

The phase-mapping system developed here is very unique. It presents the first direct visualization of the full Pressure-Temperature (P-T) phase diagram – a staple of fluid and thermodynamics textbooks. The strategy involves multiplexing a range of pressures and temperatures in a microfluidic device and observing the phases in each of 100 x 100 microwells. Such a system provides a complete mapping of the P-T phase diagram – where 10, 000 data points are measured simultaneously. The phase-mapping device provides accurate measurement of the critical pressure, demonstrating a 1.2 % deviation from NIST values. The experiment with

10,000 data points of CO2 requires 3 h to complete, including 1 h of preparation, 1 h of test, 1 h of temperature calibration and post-processing. In contrast, a traditional PVT system typically takes 8 to 10 h for a single P-T data point. At least 800 hours (100 P-T data points) are needed using a traditional PVT system to achieve the same number of data points and resolution as our phase-mapping device. Therefore, our microfluidic system demonstrates a 100-fold faster mapping of the P-T diagram than the traditional PVT method.

7.3 Fluid phase change in nanochannels

While microconfinement generally does not influence phase behavior, nanoconfinement can have profound effects on phase behavior. The nanofluidic platform developed here is used to investigate bubble nucleation and growth in liquid propane in nanochannels. Isothermal

100

cavitation experiments are performed at five temperatures, 343, 347, 352, 357 and 362 K. In contrast to the bulk phase, it is found that the cavitation pressure in the nanochannels corresponds closer to the spinodal limit than that predicted by classical nucleation theory. For the growth of the vapor column following nucleation, two different growth mechanisms are distinguished depending on the location of the bubble nucleation. Finally, a fluid dynamics model is developed to determine the evaporation rate, liquid velocity, and bubble pressure during bubble growth. Collectively these results demonstrate, characterize, and quantify isothermal bubble nucleation and growth of a pure substance in nanochannels or nanopores, as relevant for shale gas and tight oil reservoirs.

7.4 Outlook

Regarding to future trend of fluid phase measurement techniques, the author believes that PVT experimental methods would remain as the cornerstone technique at least in a short term future, due to its direct visual features as well as a wide utilization in industry. However, these PVT systems suffer high infrastructure expense ($100,000 to 500,000) and low time-efficiency (hours per data point) due to the bulky chamber size (hundreds of mL). The author believes the fluid phase measurement techniques are evolving towards to real time, miniaturized, rapid, automated and more cost effective systems. In this sense, optical or acoustic probes show great potential in terms of in situ ability, size, speed, cost and inherent automatic features. Another emerging approach is microfluidics (chamber or channel is smaller than 1 mm) or meso-fluidic (chamber or channel is between 1 mm and 1 cm) systems, because of the compact size and rapid measurement by multiplexing multiple parameters into a single run. The challenge of these systems would be the relatively high cost associated with micro-fabrication as well as the short lifetime caused by contamination and mechanical weakness under high pressure operations. Finally, nanofluidic systems (chamber or channel is smaller than 1 µm) will be required for specific understanding of the fundamentals of fluid phase phenomena at the nanoscale. That is, nanofluidics will be applied for the specific application of nanoconfined fluids, and not more broad application due to the challenges of operating at this scale and the influence of nanoconfinement on fluid phase.

101

References

[1] BP Statistical Review of World Energy June 2015, (2015). [2] International Energy Agency, Technology Roadmap: Carbon capture and Storage, 2013.

[3] R.G. Bruant, A.J. Guswa, M.A. Celia, C.A. Peters, Safe Storage of CO2 in Deep Saline Aquifers, Environ. Sci. Technol. 36 (2002) 225A–2536. [4] IEAGHG: CO2 Pipeline Infrastructure, (2013). [5] R. Arts, O. Eiken, A. Chadwick, P. Zweigel, L. van der Meer, B. Zinszner, Monitoring of CO2 injected at Sleipner using time-lapse seismic data, Energy. 29 (2004) 1383–1392. [6] S. Wang, M. Lee, M.K. Park, J.-M. Kim, Box experiments on monitoring the CO2 migration in a homogeneous medium using electrical resistivity survey, Geosci. J. (Seoul, Repub. Korea). 14 (2010) 77–85. [7] S. Emberley, I. Hutcheon, M. Shevalier, K. Durocher, W.D. Gunter, E.H. Perkins, Geochemical monitoring of fluid-rock interaction and CO2 storage at the Weyburn CO2- injection enhanced oil recovery site, Saskatchewan, Canada, Energy. 29 (2004) 1393– 1401. [8] N.G. Skinner, J.L. Maida, Downhole fiber optic sensing: the oilfield service provider’s perspective, Proc. SPIE. 5589 (2004) 206–220. [9] K.S. Pedersen, P.L. Christensen, J.A. Shaikh, Phase Behavior of Petroleum Reservoir Fluids, 2015. [10] R. Dohrn, J.M.S. Fonseca, S. Peper, Experimental methods for phase equilibria at high pressures, Annu. Rev. Chem. Biomol. Eng. 3 (2012) 343–67. [11] F. Mostowfi, S. Molla, P. Tabeling, Determining phase diagrams of gas-liquid systems using a microfluidic PVT, Lab Chip. 12 (2012) 4381–4387. [12] EIA: Technically Recoverable Shale Oil and Shale Gas Resources : An Assessment of 137 Shale Formations in 41 Countries Outside the United States, 2013. [13] L. Bocquet, P. Tabeling, Physics and technological aspects of nanofluidics, Lab Chip. 14 (2014) 3143–3158. [14] D. Malakoff, The gas surge, Science (80-. ). 344 (2014) 1464–1467. [15] Environmental Impacts of Shale Gas Extraction, 2014. [16] B. Bai, M. Elgmati, H. Zhang, M. Wei, Rock characterization of Fayetteville shale gas plays, Fuel. 105 (2013). [17] Y. Ma, L. Jin, A. Jamili, SPE 166476 Modifying Van Der Waals Equation of State to Consider Influence of Confinement on Phase Behavior, in: SPE, 2013. [18] B.R. Didar, I.Y. Akkutlu, A. Texas, SPE 164099 Pore-Size Dependence of Fluid Phase Behavior and Properties in Organic- Rich Shale Reservoirs, in: SPE, 2013. [19] A. Sanaei, A. Jamili, J. Callard, A. Mathur, SPE 169493 Production Modeling in the Eagle Ford Gas Condensate Window : Integrating New Relationships between Core Permeability , Pore Size , and Confined PVT Properties, in: SPE, 2014: pp. 1–20.

102

[20] Q. Wu, B. Bai, Y. Ma, J.T. Ok, K.B. Neeves, X. Yin, SPE 164549 Optic Imaging of Two- Phase-Flow Behavior in 1D Nanoscale Channels, in: SPE, 2014. [21] S. Kelly, SPE 167635 Experimental Investigation of the Influence of Molecular Surface Interactions on Imbibition in Shale Nano-Pore Proxies, in: SPE, 2013. [22] P. Kim, H.-Y. Kim, J.K. Kim, G. Reiter, K.Y. Suh, Multi-curvature liquid meniscus in a nanochannel: evidence of interplay between intermolecular and surface forces., Lab Chip. 9 (2009) 3255–3260. [23] C. Duan, R. Karnik, M. Lu, A. Majumdar, Evaporation-induced cavitation in nanofluidic channels, Proc. Natl. Acad. Sci. U. S. A. 109 (2012) 3688–3693. [24] Y.A. Cengel, J.M. Cimbala, R.H. Turner, Fundamentals of Thermal-Fluid Sciences 4th edition, McGraw-Hill, 2012. [25] V. Anikeev, M. Fan, T. Adschiri, Supercritical Fluid Technology for Energy and Environmental Applications, Elsevier B.V., 2014. [26] L. Wang, E. Parsa, Y. Gao, J.T. Ok, K. Neeves, X. Yin, et al., SPE 169581 Experimental Study and Modeling of the Effect of Nanoconfinement on Hydrocarbon Phase Behavior in Unconventional Reservoirs, in: SPE, 2014. [27] M. Barisik, A. Beskok, Scale effects in gas nano flows, Phys. Fluids. 26 (2014). [28] G. McKeogh, Moisture Measurement Technologies for Natural Gas, n.d. [29] D.R. Potter, Analytical Devices for the Measurement of Water Vapor and Hydrocarbon Dew-Point in Natural Gas, 2011. [30] W. Wu, J. Ke, M. Poliakoff, New design of fiber-optic reflectometer for determining the phase boundary of multicomponent fluid mixtures at high pressures and high temperatures, Rev. Sci. Instrum. 77 (2006) 023903. [31] D.W. Galipeau, J.D. Stroschine, K. a. Snow, K. a. Vetelino, K.R. Hines, P.R. Story, A study of condensation and dew point using a SAW sensor, Sensors Actuators B Chem. 25 (1995) 696–700. [32] K.K. Botros, W. Studzinski, J. Geerligs, A. Glover, Determination of Decompression Wave Speed in Rich Gas Mixtures, Can. J. Chem. Eng. 82 (2004) 880–891. [33] R. Fisher, M.K. Shah, D. Eskin, K. Schmidt, A. Singh, S. Molla, et al., Equilibrium gas-oil ratio measurements using a microfluidic technique, Lab Chip. 13 (2013) 2623–2633. [34] M.H. Schneider, V.J. Sieben, A.M. Kharrat, F. Mostowfi, Measurement of Asphaltenes Using Optical Spectroscopy on a Microfluidic Platform, Anal. Chem. 85 (2013) 5153– 5160. [35] B. Pinho, S. Girardon, F. Bazer-Bachi, G. Bergeot, S. Marre, C. Aymonier, A microfluidic approach for investigating multicomponent systems thermodynamics at high pressures and temperatures, Lab Chip. (2014) 3843–3849. [36] W. Song, H. Fadaei, D. Sinton, Determination of Dew Point Conditions for CO2 with Impurities Using Micro fluidics, Environ. Sci. Technol. 48 (2014) 3567–3574. [37] P. Nguyen, D. Mohaddes, J. Riordon, H. Fadaei, P. Lele, D. Sinton, Fast Fluorescence- Based Micro fl uidic Method for Measuring Minimum Miscibility Pressure of CO2 in

103

Crude Oils, Anal. Chem. 87 (2015) 3160–3164. [38] S. Witharana, B. Phillips, S. Strobel, H.D. Kim, T. McKrell, J.B. Chang, et al., Bubble nucleation on nano- to micro-size cavities and posts: An experimental validation of classical theory, J. Appl. Phys. 112 (2012). [39] O. Vincent, D.A. Sessoms, E.J. Huber, J. Guioth, A.D. Stroock, Drying by cavitation and poroelastic relaxations in porous media with macroscopic pores connected by nanoscale throats, Phys. Rev. Lett. 113 (2014) 1–5. [40] R. Dohrn, J.M.S. Fonseca, S. Peper, Experimental methods for phase equilibria at high pressures., Annu. Rev. Chem. Biomol. Eng. 3 (2012) 343–67. [41] R. Svensson, M. Odenberger, F. Johnsson, L. Strömberg, Transportation systems for CO2 – application to carbon capture and storage, Energy Convers. Manag. 45 (2004) 2343– 2353. [42] J. Serpa, J. Morbee, E. Tzimas, Technical and Economic Characteristics of a CO2 Transmission Pipeline Infrastructure, 2011. [43] E. Goos, U. Riedel, L. Zhao, L. Blum, Phase diagrams of CO2 and CO2–N2 gas mixtures and their application in compression processes, Energy Procedia. 4 (2011) 3778–3785. [44] J. Vrabec, G.K. Kedia, U. Buchhauser, R. Meyer-Pittroff, H. Hasse, Thermodynamic models for vapor–liquid equilibria of nitrogen+oxygen+carbon dioxide at low temperatures, Cryogenics (Guildf). 49 (2009) 72–79.

[45] IEA, Effects of impurities on geological storage of CO2, 2011. [46] X. Ji, C. Zhu, Predicting Possible Effects of H2S Impurity on CO2 Transportation and Geological Storage, Environ. Sci. Technol. 47 (2012). [47] S.-P. Kang, H. Lee, Recovery of CO2 from Flue Gas Using Gas Hydrate: Thermodynamic Verification through Phase Equilibrium Measurements, Environ. Sci. Technol. 34 (2000) 4397–4400. [48] A. Oancea, B. Hanoune, C. Focsa, B. Chazallon, Cross determination of the vapor liquid equilibrium of formaldehyde aqueous solutions by quadrupole mass spectrometry and infrared diode laser spectroscopy., Environ. Sci. Technol. 43 (2009) 435–40. [49] T. Ahmed, Equations of state and PVT analysis: applications for improved reservoir modeling, Gulf Publishing Company, 2007. [50] A. Shariati, C.J. Peters, Measurements and modeling of the phase behavior of ternary systems of interest for the GAS process: I. The system carbon dioxide+1- propanol+salicylic acid, J. Supercrit. Fluids. 23 (2002) 195–208. [51] J.C.R. Reis, N. Ribeiro, A. Aguiar-Ricardo, Can the speed of sound be used for detecting critical states of fluid mixtures?, J. Phys. Chem. B. 110 (2006) 478–84. [52] R.M. Oag, P.J. King, C.J. Mellor, M.W. George, J. Ke, M. Poliakoff, Probing the Vapor - Liquid Phase Behaviors of Near-Critical and Supercritical Fluids Using a Shear Mode Piezoelectric Sensor, Anal. Chem. 75 (2003) 479–485. [53] H. Ziegler, K. Rolf, Quartz sensor for automatic dew-point hygrometry, Sensors Actuators B Chem. 11 (1987) 37–44.

104

[54] O.J. Joung, Y.H. Kim, Dew point measurement for organic vapor mixture using a quartz crystal sensor, Sensors Actuators B Chem. 113 (2006) 335–340. [55] A.A. Novitskiy, J. Ke, G. Comak, M. Poliakoff, M.W. George, A modified golden gate attenuated total reflection (ATR) cell for monitoring phase transitions in multicomponent fluids at high temperatures., Appl. Spectrosc. 65 (2011) 885–91. [56] H. Fadaei, B. Scarff, D. Sinton, Rapid Microfluidics-Based Measurement of CO2 Diffusivity in Bitumen, Energy & Fuels. 25 (2011) 4829–4835. [57] H. Fadaei, J.M. Shaw, D. Sinton, Bitumen−Toluene Mutual Diffusion Coefficients Using Microfluidics, Energy & Fuels. 27 (2013) 2042–2048. [58] W. Song, H. Fadaei, D. Sinton, Determination of Dew Point Conditions for CO2 with Impurities Using Microfluidics, Environ. Sci. Technol. (2014). [59] K. Iwami, S. Kaneko, R. Shinta, J. Fujihara, H. Nagasaki, Y. Matsumura, et al., Plasmon- resonance dew condensation sensor made of gold-ceramic nanocomposite and its application in condensation prevention, Sensors Actuators B Chem. 184 (2013) 301–305. [60] B. Bao, L. Melo, B. Davies, H. Fadaei, D. Sinton, P. Wild, Detecting Supercritical CO2 in Brine at Sequestration Pressure with an Optical Fiber Sensor., Environ. Sci. Technol. 47 (2013) 306–313. [61] I.M. Raimundo, R. Narayanaswamy, Simultaneous determination of relative humidity and ammonia in air employing an optical fibre sensor and artificial neural network, Sensors Actuators B Chem. 74 (2001) 60–68. [62] M. V. Avdeev, A.N. Konovalov, V.N. Bagratashvili, V.K. Popov, S.I. Tsypina, M. Sokolova, et al., The fibre optic reflectometer: A new and simple probe for refractive index and phase separation measurements in gases, liquids and supercritical fluids, Phys. Chem. Chem. Phys. 6 (2004) 1258. [63] Properties of carbon dioxide, Union Engineering, n.d. [64] NIST, Carbon Dioxide, NIST Webb. (n.d.). http://webbook.nist.gov/cgi/cbook.cgi?Name=carbon+dioxide&Units=SI/. [65] J.M.S. Fonseca, R. Dohrn, S. Peper, High-pressure fluid-phase equilibria: Experimental methods and systems investigated (2005–2008), Fluid Phase Equilib. 300 (2011) 1–69. [66] I. Miracca, K.I. Aasen, T. Brownscombe, K. Gerdes, M. Simmonds, Chapter 25 Oxyfuel Combustion for CO2 Capture Technology Summary, in: Carbon Dioxide Capture Storage Deep Geol. Form., 2005: pp. 441–449. [67] U.K. Deiters, T. Kraska, Phenomenology of Phase Diagrams, Elsevier, 2012. [68] J. Ke, Y. Sanchez-Vicente, G.R. Akien, a. a. Novitskiy, G. Comak, V.N. Bagratashvili, et al., Detecting phase transitions in supercritical mixtures: an enabling tool for greener chemical reactions, Proc. R. Soc. A Math. Phys. Eng. Sci. 466 (2010) 2799–2818.

[69] Y. Sun, B.Y. Shekunov, P. York, Refractive index of supercritical CO2-ethanol solvents, Chem. Eng. Commun. 190 (2003) 1–14. [70] IPCC Special Report: Carbon dioxide capture and storage, summary for policymakers, 2005.

105

[71] UK DTI, Technology Status Report: Monitoring Technologies for the Geological Storage of CO2, 2005.

[72] US DOE/NETL, Monitoring, verification, and accounting of CO2 stored in deep geologic formations, 2009. [73] R.J. Arts, R.A. Chadwick, O. Eiken, M. Trani, S. Dortland, Synthetic versus real time- lapse seismic data at the Sleipner CO2 injection site, 2007.

[74] D. Lumley, 4D seismic monitoring of CO2 sequestration, Lead. Edge. 29 (2010) 150–155. [75] J.P. Verdon, J.-M. Kendall, D.J. White, D.A. Angus, Linking microseismic event observations with geomechanical models to minimise the risks of storing CO2 in geological formations, Earth Planet. Sci. Lett. 305 (2011) 143–152. [76] C. Schmidt-Hattenberger, P. Bergmann, D. Kießling, K. Krüger, C. Rücker, H. Schütt, Application of a Vertical Electrical Resistivity Array (VERA) for monitoring CO2 migration at the Ketzin site: First performance evaluation, Energy Procedia. 4 (2011) 3363–3370.

[77] B.P. Spalding, D.B. Watson, Measurement of dissolved H2, O2 and CO2 in groundwater using passive samplers for gas chromatographic analyses, Environ. Sci. Technol. 40 (2006) 7861–7. [78] S.C.M. Krevor, T. Ide, S.M. Benson, F.M. Orr, Real-time tracking of CO2 injected into a subsurface coal fire through high-frequency measurements of the 13CO2 signature., Environ. Sci. Technol. 45 (2011) 4179–86. [79] F.C. Boait, N.J. White, M.J. Bickle, R. a. Chadwick, J. a. Neufeld, H.E. Huppert, Spatial and temporal evolution of injected CO2 at the Sleipner Field, North Sea, J. Geophys. Res. 117 (2012) 1–21. [80] B. Lee, Review of the present status of optical fiber sensors, Opt. Fiber Technol. 9 (2003) 57–79. [81] R. Bogue, Fibre optic sensors: a review of today’s applications, Sens. Rev. 31 (2011) 304– 309. [82] V. Latini, V. Striano, G. Coppola, I. Rendina, Fiber optic sensors system for high- temperature monitoring of aerospace structures, Proc. SPIE. 6593 (2007) 65930S– 65930S–9. [83] L.C.L. Chin, W.M. Whelan, I.A. Vitkin, Optical-Thermal Response of Laser-Irradiated Tissue, 2nd ed., Springer Netherlands, Dordrecht, 2011. [84] C.R. Dennison, P.M. Wild, Enhanced sensitivity of an in-fibre Bragg grating pressure sensor achieved through fibre diameter reduction, Meas. Sci. Technol. 19 (2008) 125301. [85] M. Fan, P. Wang, C. Escobedo, D. Sinton, A.G. Brolo, Surface-enhanced Raman scattering (SERS) optrodes for multiplexed on-chip sensing of nile blue A and oxazine 720., Lab Chip. 12 (2012) 1554–60. [86] O. Abdi, M. Kowalsky, T. Hassan, S. Kiesel, K. Peters, Large deformation polymer optical fiber sensors for civil infrastructure systems, Proc. SPIE. 6932 (2008) 693242– 693242–12. [87] J.W. Berthold, Overview of advanced fiber optic sensor equipment for energy production

106

applications, Proc. SPIE. 5589 (2004) 197–205. [88] N. a. David, P.M. Wild, J. Jensen, T. Navessin, N. Djilali, Simultaneous In Situ Measurement of Temperature and Relative Humidity in a PEMFC Using Optical Fiber Sensors, J. Electrochem. Soc. 157 (2010) B1173. [89] F. Charpentier, B. Bureau, J. Troles, C. Boussard-Plédel, K. Michel-Le Pierrès, F. Smektala, et al., Infrared monitoring of underground CO2 storage using chalcogenide glass fibers, Opt. Mater. 31 (2009) 496–500. [90] R. Orghici, U. Willer, M. Gierszewska, S.R. Waldvogel, W. Schade, Fiber optic evanescent field sensor for detection of explosives and CO2 dissolved in water, Appl. Phys. B Lasers Opt. 90 (2008) 355–360. [91] T.-J. Lin, C.-T. Lou, Reflection-based localized surface plasmon resonance fiber-optic probe for chemical and biochemical sensing at high-pressure conditions, J. Supercrit. Fluids. 41 (2007) 317–325. [92] J.S. Loring, C.J. Thompson, Z. Wang, A.G. Joly, D.S. Sklarew, H.T. Schaef, et al., In situ infrared spectroscopic study of forsterite carbonation in wet supercritical CO2, Environ. Sci. Technol. 45 (2011) 6204–10. [93] E.S. Ilton, H.T. Schaef, O. Qafoku, K.M. Rosso, A.R. Felmy, In situ X-ray diffraction study of Na+ saturated montmorillonite exposed to variably wet super critical CO2, Environ. Sci. Technol. 46 (2012) 4241–8. [94] J.A. Surface, P.A. Skemer, S.E. Hayes, M.S. Conradi, In Situ Measurement of Magnesium Carbonate Formation from CO2 Using Static High-Pressure and -Temperature 13C NMR, Environ. Sci. Technol. (2012). [95] H.J. Patrick, A.D. Kersey, F. Bucholtz, Analysis of the response of long period fiber gratings to external index of refraction, J. Light. Technol. 16 (1998) 1606–1612. [96] X. Shu, L. Zhang, I. Bennion, Sensitivity characteristics of long-period fiber gratings, J. Light. Technol. 20 (2002) 255–266. [97] S.W. James, R.P. Tatam, Optical fibre long-period grating sensors : characteristics and application, Meas. Sci. Technol. 14 (2003) R49–R61. [98] A. Kapoor, E.K. Sharma, Long period grating refractive-index sensor: optimal design for single wavelength interrogation., Appl. Opt. 48 (2009) G88–94. [99] P. Alessi, A. Cortesi, I. Kikic, F. Vecchione, Plasticization of Polymers with Supercritical Carbon Dioxide : Experimental Determination of Glass-Transition Temperatures, J. Appl. Polym. Sci. 88 (2003) 2189–2193. [100] S. Marre, Y. Roig, C. Aymonier, Supercritical microfluidics: Opportunities in flow- through chemistry and materials science, J. Supercrit. Fluids. 66 (2012) 251–264. [101] Y. Kim, J. Wan, T.J. Kneafsey, T.K. Tokunaga, Dewetting of silica surfaces upon reactions with supercritical CO2 and brine: pore-scale studies in micromodels., Environ. Sci. Technol. 46 (2012) 4228–4235. [102] M.J. Carrott, C.M. Wai, UV-Visible Spectroscopic Measurement of Solubilities in Supercritical CO2 Using High-Pressure Fiber-Optic Cells., Anal. Chem. 70 (1998) 2421– 2425.

107

[103] S.M. Selloy, Handbook of Chemistry and Physics: Index of Refraction of Air, 48th ed., The Chemical Rubber Co., Cleveland OH, 1967. [104] K. Vedam, G.A. Samara, Refractive index of liquids at high pressures, Crit. Rev. Solid State Mater. Sci. 11 (1983) 1:45. [105] P. Schiebener, J. Straub, J.M.H. Levelt Sengers, J.S. Gallagher, Refractive Index of Water and Steam as Function of Wavelength, Temperature and Density, J. Phys. Chem. Ref. Data. 19 (1990). [106] K. Vedam, P. Limsuwan, Piezo- and elasto-optic properties of liquids under high pressure. II. Refractive index vs density, J. Chem. Phys. 69 (1978) 4772. [107] E. Reisler, H. Eisenberg, A.P. Minton, Temperature and Density Dependence of the Refractive Index of Pure Liquids, J. Chem. Soc., Faraday Trans. 2. 68 (1972) 1001–1015. [108] J.M. Simonson, C.S. Oakes, R.J. Bodnar, Densities of NaCl(aq) to the temperature 523K at pressures to 40 MPa measured with a new vibrating-tube densitometer, J. Chem. Thermodyn. 26 (1994) 345–359.

[109] Z. Duan, R. Sun, An improved model calculating CO2 solubility in pure water and aqueous NaCl solutions from 273 to 533 K and from 0 to 2000 bar, Chem. Geol. 193 (2003) 257–271. [110] F. Millero, F. Huang, T. Graham, D. Pierrot, The dissociation of carbonic acid in NaCl solutions as a function of concentration and temperature, Geochim. Cosmochim. Acta. 71 (2007) 46–55. [111] X. Zhang, Pressure and temperature sensing system based on fiber optic Bragg gratings for the oil and gas downhole, Proc. SPIE. 5279 (2004) 636–640. [112] B. Prevedel, L. Wohlgemuth, J. Henninges, K. Krüger, B. Norden, A. Förster, The CO2SINK Boreholes for Geological Storage Testing, Sci. Drill. (2008) 32–37.

[113] Knowledge Sharing in MVA/MMV in CCS Demonstration Projects and Large scale CO2 Injection Tests: Workshop summary. A report for the European CCS Demonstration Project Network, 2012. [114] H.M. Cota, G. Thodos, Critical Temperatures and Critical Pressures of Hydrocarbon Mixtures, J. Chem. Eng. Data. 7 (1962) 62–65. [115] S. Nakano, M. Moritoki, K. Ohgaki, High-Pressure Phase Equilibrium and Raman Microprobe Spectroscopic Studies on the CO2 Hydrate System, J. Chem. Eng. Data. 43 (1998) 807–810. [116] A.A. Novitskiy, J. Ke, G. Comak, M. Poliakoff, M. George, A modified golden gate attenuated total reflection (ATR) cell for monitoring phase transitions in multicomponent fluids at high temperatures, Appl. Spectrosc. 65 (2011) 885–891. [117] E. Anim-danso, Y. Zhang, A. Dhinojwala, Freezing and Melting of Salt Hydrates Next to Solid Surfaces Probed by Infrared-Visible Sum Frequency Generation Spectroscopy, J. Am. Chem. Soc. 135 (2013) 8496–8499. [118] K.I. Saitow, D. Kajiya, K. Nishikawa, Dynamics of Density Fluctuation of Supercritical Fluid Mapped on Phase Diagram, J. Am. Chem. Soc. 126 (2004) 422–423. [119] J. Ke, H. Buxing, M.W. George, H. Yan, M. Poliakoff, How does the critical point change

108

during a chemical reaction in supercritical fluids? A study of the hydroformylation of propene in supercritical CO2, J. Am. Chem. Soc. 123 (2001) 3661–3670. [120] E.F. May, T.J. Edwards, a. G. Mann, C. Edwards, R.C. Miller, Development of an automated phase behaviour measurement system for lean hydrocarbon fluid mixtures, using re-entrant rf/microwave resonant cavities, Fluid Phase Equilib. 185 (2001) 339–347. [121] Y. Okahata, K. Kimura, K. Ariga, Detection of the phase transition of Langmuir-Blodgett films on a quartz-crystal microbalance in an aqueous phase, J. Am. Chem. Soc. 111 (1989) 9190–9194. [122] D. Kodama, J. Miyazaki, M. Kato, T. Sako, High pressure phase equilibrium for ethylene + 1-propanol system at 283.65 K, Fluid Phase Equilib. 219 (2004) 19–23. [123] B. Bao, H. Fadaei, D. Sinton, Detection of bubble and dew point using optical thin-film interference, Sensors Actuators B Chem. 207 (2015) 640–649. [124] J.-U. Shim, G. Cristobal, D.R. Link, T. Thorsen, Y. Jia, K. Piattelli, et al., Control and measurement of the phase behavior of aqueous solutions using microfluidics, J. Am. Chem. Soc. 129 (2007) 8825–8835. [125] D. Sinton, Energy: the microfluidic frontier., Lab Chip. 14 (2014) 3127–3134. [126] T.D. Wheeler, A.D. Stroock, The transpiration of water at negative pressures in a synthetic tree, Nature. 455 (2008) 208–212. [127] M.A. Gonzalez, G. Menzl, J.L. Aragones, P. Geiger, F. Caupin, J.L.F. Abascal, et al., Detecting vapour bubbles in simulations of metastable water, J. Chem. Phys. 141 (2014). [128] K. Ohsaka, E.H. Trinh, Dynamic nucleation of ice induced by a single stable cavitation bubble, Appl. Phys. Lett. 73 (1998) 129–131. [129] R. Zhang, Y. Ikoma, T. Motooka, Negative capillary-pressure-induced cavitation probability in nanochannels, Nanotechnology. 21 (2010) 105706. [130] A. Kiwilsza, B. Milanowski, K. Drużbicki, L.E. Coy, M. Grzeszkowiak, M. Jarek, et al., Mesoporous drug carrier systems for enhanced delivery rate of poorly water-soluble drug: nimodipine, J. Porous Mater. 22 (2015) 817–829. [131] C.E. Brennen, Cavitation and bubble dynamics, Oxford University Press, Oxford, 1995. [132] A.S. Tucker, C.A. Ward, Critical state of bubbles in liquid-gas solutions, J. Appl. Phys. 46 (1975) 4801–4808. [133] Z.G. Li, S. Xiong, L.K. Chin, K. Ando, J.B. Zhang, A.Q. Liu, Water’s tensile strength measured using an optofluidic chip, Lab Chip. 15 (2015) 2158–2161. [134] D. Fuster, K. Pham, S. Zaleski, Stability of bubbly liquids and its connection to the process of cavitation inception, Phys. Fluids. 26 (2014). [135] K. Ando, A.Q. Liu, C.D. Ohl, Homogeneous nucleation in water in microfluidic channels, Phys. Rev. Lett. 109 (2012) 1–5. [136] O. Vincent, P. Marmottant, P.A. Quinto-Su, C.D. Ohl, Birth and growth of cavitation bubbles within water under tension confined in a simple synthetic tree, Phys. Rev. Lett. 108 (2012) 1–5. [137] G. Nagashima, E. V. Levine, D.P. Hoogerheide, M.M. Burns, J.A. Golovchenko,

109

Superheating and homogeneous single bubble nucleation in a solid-state nanopore, Phys. Rev. Lett. 113 (2014) 1–8. [138] C. Duan, R. Karnik, M.-C. Lu, A. Majumdar, Evaporation-induced cavitation in nanofluidic channels, Proc. Natl. Acad. Sci. 109 (2012) 3688–3693. [139] V.P. Carey, Liquid-vapor phase-change phenomena, Taylor & Francis Group, New York. Abingdon, 2008. [140] M.S. Plesset, S.A. Zwick, The growth of vapor bubbles in superheated liquids, J. Appl. Phys. 25 (1954) 493–500. [141] H.K. Forster, N. Zuber, Growth of a vapor bubble in a superheated liquid, J. Appl. Phys. 25 (1954) 474–478. [142] A. Asai, Application of the nucleation theory to the design of bubble jet printers, Jpn. J. Appl. Phys. 28 (1989) 909–915. [143] A. Mazouchi, G.M. Homsy, Thermocapillary migration of long bubbles in polygonal tubes, Phys. Fluids. 15 (2003) 308–314. [144] C. Sun, E. Can, R. Dijkink, D. Lohse, A. Prosperetti, Growth and collapse of a vapour bubble in a microtube: the role of thermal effects, J. Fluid Mech. 632 (2009) 5–16. [145] E.D. Torniainen, A.N. Govyadinov, D.P. Markel, P.E. Kornilovitch, Bubble-driven inertial micropump, Phys. Fluids. 24 (2012). [146] S. Kumar Saha, G. Piero Celata, S.G. Kandlikar, Thermofluid Dynamics of Boiling in Microchannels, Elsevier Inc., 2011. [147] S.H. Zandavi, C.A. Ward, Contact angles and surface properties of nanoporous materials, J. Colloid Interface Sci. 407 (2013) 255–264. [148] V.P. Skripov, Metastable liquids, John Wiley & Sons, New York. Toronto, 1974. [149] V.E. Vinogradov, P.A. Pavlov, Cavitation Strength of Liquids With Large Molecules, Int. J. Mod. Phys. B. 27 (2013) 1–10. [150] M. Blander, J.L. Katz, Bubble nucleation in liquids, AlChE J. 21 (1975) 833–848. [151] M. Thommes, G.H. Findenegg, Pore Condensation and Critical-Point Shift of a Fluid in Controlled-Pore Glass, Langmuir. 10 (1994) 4270–4277. [152] Frank M. White, Viscous Fluid Flow, McGraw-Hill, New York etc., 1991. [153] S. Kelly, M.T. Balhoff, C. Torres-Verdin, Quantification of bulk solution limits for liquid and interfacial transport in nanoconfinements, Langmuir. 31 (2015) 2167–2179. [154] V.N. Phan, N.T. Nguyen, C. Yang, P. Joseph, L. Djeghlaf, D. Bourrier, et al., Capillary filling in closed end nanochannels, Langmuir. 26 (2010) 13251–13255. [155] M.N. Hamblin, A.R. Hawkins, D. Murray, D. Maynes, M.L. Lee, A.T. Woolley, et al., Capillary flow in sacrificially etched nanochannels, Biomicrofluidics. 5 (2011). [156] N.R. Tas, J. Haneveld, H. V. Jansen, M. Elwenspoek, A. Van Den Berg, Capillary filling speed of water in nanochannels, Appl. Phys. Lett. 85 (2004) 3274–3276.

110

Appendix 1 Silicon-glass micro/nanofluidic chip fabrication procedure

1. Design chip pattern using AutoCAD

2. Mask writing using mask writer

 Convert the AutoCAD file and start mask writing

 Develop the exposed mask for 30 s, rinse the mask with DI water for 1 min, and dry

the mask with N2.

 Immerse the mask in Cr etchant for 30 s, rinse the mask with DI water for 1 min, and

dry the mask with N2.

111

 Immerse the mask in photoresist stripper for 1 min to remove the residual photoresist,

then dry the mask with N2

3. Spin coating of photoresist S1818

 Preheat Si wafer at 100oC for 1-2 min.

 Spin coating of HMDS on the Si wafer at 2000 RPM for 90 s.

 Spin coating of photoresist S1818 on top of HMDS layer at 2000 RPM for 90 s.

 Soft bake of the Si wafer for 100oC for 2-5 min.

4. Exposure

2 2  Set exposure dosage to 100 mJ/cm (nanochannel) or 200 mJ/cm (microchannel).

5. Develop the exposed photoresist in wet bench

112

 Immerse the exposed Si wafer in developer (MF312 : DI water = 1 : 1) for 60 s.

 Rinse the wafer with DI water for 60 s.

 Dry the wafer with N2.

6. Etch the pattern using RIE when the target depth is less than 1 µm, otherwise use DRIE.

 RIE recipe (120 nm depth = 20 s)

Pressure (mT) RIE RF Power Set (W) SF6 Set (sccm) Helium Set (sccm)

30 100 20 130

 DRIE recipe: Bosch process: ~10 µm/min (100 µm depth = 12 min or 252 cycles)

113

7. Measure the pattern’s depth using profilometer.

If there is only one layer of pattern, direct to step 14;

If there are two layers, e.g., both micro and nano features, continue to step 8.

8. Piranha (H2SO4 : H2O2 = 2.5 : 1) cleaning of the Si wafers for at least 3 hours.

 Strip the residual photoresist using acetone or photoresist stripper.

 Place Si wafers vertically into a clean container.

 Carefully pour 2.5 portions of H2SO4, and then 1 portion of H2O2 into container.

114

 3 h immersion

 Dry the wafer with N2

9. Spin coating of photoresist for the second layer of pattern, refer to step 3.

10. Alignment and exposure of the second layer of pattern, refer to step 4.

11. Develop the exposed photoresist, refer to step 5.

12. Etch the pattern using RIE or DRIE, refer to step 6.

13. Measure the pattern’s depth using profilometer, refer to step 7.

14. Drill the inlet and outlet holes on Si wafer.

15. Strip the residual photoresist on the Si wafer using acetone or photoresist stripper.

16. Ultrasonic cleaning of the Si wafer with DI water for 5 min.

17. Piranha (H2SO4 : H2O2 = 2.5 : 1) cleaning of the Si wafer and glass for at least 3 hours.

 Place Si wafers and glass vertically into a clean container.

 Carefully pour 2.5 portions of H2SO4, and then 1 portion of H2O2 into container

115

 3 h immersion

 Dry the wafer and glass with N2

18. Anodic bonding of Si wafer and glass

 Place chip into chamber and close the chamber.

 Vacuum the chamber < 10-5 bar.

 Set force to 100 N.

 Heat temperature 400 oC.

 Set voltage to 600 V and current to 4 mA, and, enable HV supply.

 Disable HV supply when current is < 0.1 mA.

 Set temperature back to 3oC to be cooled down

 Open the chamber, take sample out and close the chamber.

116

19. Dice the glass/silicon chip

 Seal the chip holes to avoid contamination of cutting water

 Dicing speed 0.3 mm/s.

20. Glass/silicon chip fabricated.