The Development of High-Throughput and Miniaturized ...
Transcript of The Development of High-Throughput and Miniaturized ...
The Development of High-Throughput and Miniaturized Differential Scanning Calorimeter for
Thermodynamic Study of Bio-Molecules
Shifeng Yu
Dissertation submitted to the faculty of the Virginia Polytechnic Institute and State University in
partial fulfillment of the requirements for the degree of
Doctor of Philosophy
In
Mechanical Engineering
Lei Zuo, Chair
Bahareh Behkam
Ming Lu
Jiangtao Cheng
Masoud Agah
Jan 22nd, 2019
Blacksburg, VA
Keywords: MEMS, Differential Scanning Calorimeter, Thermodynamic, Protein Stability,
Transition Temperature
Copyright © 2019, Shifeng Yu
The Development of High-Throughput and Miniaturized Differential Scanning Calorimeter for
Thermodynamic Study of Bio-Molecules
Shifeng Yu
ABSTRACT
Biomolecular interactions are fundamentally important for a wide variety of biological processes.
Understanding the temperature dependence of biomolecular interactions is hence critical for
applications in fundamental sciences and drug discovery. Micro-Electro-Mechanical Systems
(MEMS) technology holds great potential in facilitating temperature-dependent characterization
of biomolecular interactions by providing on-chip microfluidic handling with drastically reduced
sample consumption, and well controlled micro- or nanoscale environments in which biomolecules
are effectively and efficiently manipulated and analyzed. This dissertation is focused on a high-
through and miniaturized differential scanning calorimeter for thermodynamic study of bio-
molecules using MEMS techniques.
The dissertation firstly introduces the overall design and operation principles. This miniaturized
DSC was fabricated based on a polyimide (PI) thin film. Highly temperature sensitive vanadium
oxide was used as the thermistor material. A PDMS (Polydimethylsiloxane) microfluidic chamber
was separately fabricated and then bonded firmly with the PI substrate by a stamp-and-stick
method. Meanwhile, the micro heater design was optimized to reach better uniformity. A heating
stage was constructed for fast and reliable scanning. In this study, we used syringes to deliver the
0.63 μL liquid sample into both the sample and reference chambers. All the testing processes were
functionalized using the LabVIEW programs.
The sensing material was also characterized. To seek a higher temperature coefficient of resistance
(TCR) and less resistive behavior, explorations about various PVD (physical vapor deposition)
parameters and annealing conditions were conducted for optimization. In this research, we found
vanadium oxide deposited under certain conditions leads to the highest TCR value (a maximum
of 2.51%/oC). To better understand the material’s property, we also did the XRD (X-ray
Diffraction), SEM (Scanning electron microscope).
The micro calorimeter was calibrated using a step thermal response. The time constant was around
3s, the thermal conductance was 0.6mW/K, and the sensitivity was 6.1V/W. The static power
resolution of the device at equilibrium is 100 nW, corresponding to 250 nJ/K. These performances
confirmed the design and material to be appropriate for both good thermal isolation and power
sensitivity.
We demonstrated the miniaturized DSC’s performance on several different kinds of protein
samples: lysozyme, and mAb (monoclonal antibody) and a DVD IgG (double variable domain
immunoglobulin G). The results were found to be reasonable by comparing it with the commercial
DSC’s tests.
Finally, this instrument may be ideal for incorporation into high throughput screening workflows
for the relative comparison of thermal properties between large numbers of proteins when only
small quantities are available. The micro-DSC has the potential to characterize the thermal stability
of the protein sample with significantly higher throughput and less sample consumption, which
could potentially reduce the time and cost for the drug formulation in the pharmaceutical industry.
The Development of High-Throughput and Miniaturized Differential Scanning Calorimeter for
Thermodynamic Study of Bio-Molecules
Shifeng Yu
GENERAL AUDIENCE ABSTRACT
Virtually all biological phenomena depend on molecular interactions, which is either intra-
molecular as protein folding/unfolding or intermolecular as in ligand binding. A basic biology
problem is to understand the folding and denaturation processes of a protein: the kinetics,
thermodynamics and how a protein unfolds and folds back into its native state. Both
folding/unfolding and denaturation processes are associated with enthalpy changes. The
thermodynamics of binding compounds helps a great deal to understand the nature and potency of
such molecules and is essential in drug discovery. As a label-free and immobilization-free method,
calorimetry can evaluate the Gibbs free energy, enthalpy, entropy, specific heat, and stoichiometry,
and thus provides a fundamental understanding of the molecular interactions. Calorimetric systems
including isothermal titration calorimeters (ITC) and differential scanning calorimeters (DSC) are
the gold standard for characterizing molecular interactions.
In this research, a micro DSC is developed for direct thermodynamic study of bio-molecules.
Compared with the current commercial DSC, it is on a much smaller scale. It consumes much less
sample and time in each DSC measurement. It can enable comprehensive high-content
thermodynamics study in the early stage of drug discovery and formulation. It also enables direct,
precise, and rapid evaluation of the folding and unfolding of the large biomolecules like proteins,
DNAs, and enzymes without labeling or immobilization. It can also be used as a powerful tool to
study the membrane proteins, which is often impractical or impossible before.
v
Acknowledgments
This study would not be possible without support and encouragement of my colleagues, friends
and family. I would like to extend my sincerest thanks to all of these people for their friendship,
support, and love throughout all these years.
First, I would sincerely thank my advisor, Professor Lei Zuo. He is so important that he had
influenced my life and my understanding of the world. During my graduate study, I have benefited
from him greatly in many aspects of academic research, communication skills, and personality.
His profound knowledge has been a constant source of information for my research.
Second, I feel gratitude that Professor Ming Lu, Bahareh Behkam, Jiangtao Cheng and
Masoud Agah could be my thesis committee. Thanks for the useful discussions and support in my
Ph.D. study.
Third, I want to thank my colleagues: Dr. Shuyu Wang, Jiancheng Yin, Dr. Gaosheng Fu, Dr.
Xiaoming Chen, Yongjia Wu, Xiaofan Li and many others for their valuable help and discussions
over time. Many helpful discussions with other faculty and students are gratefully appreciated as
well. I would like to thank the CFN in BNL for granting me access to the fabrication and testing
facilities for my research. I would also like to thank the NSF and AbbVie Inc for their funding
support. Especially, I feel grateful for the collaboration with AbbVie (Dr. Michael Siedler, Dr.
Peter Ihnat and Dr. Dana Filoti).
Finally, I would like to thank my parents, my brother, my brother-in-law, whose love, support, and
consistent encouragement kept me motivated and concentrated on my research throughout these
years.
vi
Contents Chapter 1. Introduction ................................................................................................................................. 1
1.1 Calorimetry ......................................................................................................................................... 1
1.2 MEMS Fabrication Techniques .......................................................................................................... 8
1.2.1 Deposition of Electronic Materials .............................................................................................. 8
1.2.2 Photolithography ...................................................................................................................... 11
1.2.3 Plasma Etching and Wet Etching ............................................................................................... 12
1.3 MEMS-Based Calorimetry ............................................................................................................... 13
1.3.1 High Thermal Insulation Design for MEMS-Based DSC .......................................................... 16
1.3.2 Temperature Sensing Methods in MEMS-Based DSC .............................................................. 18
1.3.3 MEMS-based Calorimeter for Protein Stability Study .............................................................. 20
1.4 Motivation and Significance of This Dissertation ............................................................................ 21
1.5 Dissertation Outline .......................................................................................................................... 24
Chapter 2. Design and Fabrication of the MEMS DSC .............................................................................. 26
2.1 Introduction ....................................................................................................................................... 26
2.1.1 Thermodynamics in DSC Test ................................................................................................... 26
2.1.2 Temperature and Power Sensing Principle ................................................................................ 31
2.2 Micro Heater Design ......................................................................................................................... 33
2.2.1 Introduction of the Micro Heater Design ................................................................................... 33
2.2.2 Micro Heater Design Principle .................................................................................................. 35
2.2.3 Characterization of the Micro Heater Design ............................................................................ 40
2.3 Thermistor Design ............................................................................................................................ 46
2.4 Microfluidic Chamber Design .......................................................................................................... 48
2.4.1 Introduction of Microfluidic Device .......................................................................................... 48
2.4.2 Design and Fabrication of the Microfluidic Chamber ............................................................... 50
2.5 Heating Stage Design ........................................................................................................................ 55
2.6 Fabrication of the MEMS Based DSC .............................................................................................. 57
2.6.1 Fabrication of the Polyimide Thin Film ..................................................................................... 57
2.6.2 Micro Heater and Thermistor Fabrication .................................................................................. 59
2.6.3 Microfluidic Device Fabrication ................................................................................................ 60
2.6.4 The Heating Stage Fabrication ................................................................................................... 62
2. 7 Summary .......................................................................................................................................... 63
Chapter 3 Thermistor Material Study ......................................................................................................... 65
3. 1 Introduction ...................................................................................................................................... 65
vii
3.2 Electrical Characterization Method .................................................................................................. 67
3.3 Characterization of the VOx Thin Film as a Thermistor ................................................................... 68
3.4 Metal-Insulator Transition Study of Vanadium Oxide ..................................................................... 73
3.4.1 Background ................................................................................................................................ 73
3.4.2 The Fabrication of the Vanadium Oxide Samples ..................................................................... 75
3.4.3 Characterization and Discussion of the Vanadium Oxide Material ........................................... 77
3.5 Summary ........................................................................................................................................... 85
Chapter 4. MEMS-Based DSC Characterization ........................................................................................ 86
4.1 Abstract ............................................................................................................................................. 86
4.2 Temperature Distribution Analysis of the MEMS-Based DSC ........................................................ 86
4.2.1 Heat Transfer in Thin Films ....................................................................................................... 86
4.2.2 “Cross-Talk” Effect between the Sample and Reference........................................................... 89
4.3 Characterization of the MEMS-based DSC ...................................................................................... 93
4.3.1 Characterization of the Heating Stage........................................................................................ 93
4.3.2 Response Study of the MEMS-based DSC ................................................................................ 97
4. 4 Analysis of the Defects in the Fabricated Micro Heater ................................................................ 100
4.4.1 Background .............................................................................................................................. 100
4.4.2 Experiments of Using the Micro Heater for Temperature Scanning. ...................................... 102
4.5 Summary ......................................................................................................................................... 105
Chapter 5. Application of the MEMS DSC in Protein Tests .................................................................... 106
5.1 Introduction ..................................................................................................................................... 106
5.2 Thermodynamic Parameters ........................................................................................................... 108
5.3 Characterization of the MEMS-Based DSC in Protein Sample Tests ........................................... 110
5.3.1 Baseline Characterization ........................................................................................................ 110
5.3.2 Lysozyme Sample Tests........................................................................................................... 113
5.3.3 The DSC Measurements of Two Antibodies. .......................................................................... 118
5.4 Influence of the Response Time to the Kinetic Parameters of Protein ........................................... 121
5. 5 Influence of the Scanning Rate to the Kinetic Parameters of Protein ............................................ 124
5.6 Thermal Stability Study of a Fab, a mAb and the Corresponding DVD Ig Using the MEMS-Based
DSC ....................................................................................................................................................... 128
5.6.1 Background .............................................................................................................................. 128
5.6.2 Denaturation Measurements of the Fab, mAb and the DVD Ig ............................................... 131
5.7 Summary ......................................................................................................................................... 136
Chapter 6. Concluding Remarks ............................................................................................................... 137
6.1 Thesis Summary .............................................................................................................................. 137
viii
6.2 Future Work .................................................................................................................................... 140
6.2.1 MEMS based Isothermal Titration Calorimeter (ITC) Development ...................................... 140
6.2.2 MEMS-Based DSC Sensor Array Development. .................................................................... 146
Reference .............................................................................................................................................. 149
ix
List of Figures
Figure 1-1. (a). The schematic diagram of DSC. (b). The DSC curve of the reference material. (c). The DSC curve of
the sample material. (d). Heat capacity change of the sample material along the temperature change. ............... 6
Figure 1-2. The schematic diagram of thermal/ebeam deposition system ................................................................... 10
Figure 1-3. Illustration of the CVD process. ............................................................................................................... 11
Figure 1-4. Illustration of the photolithography process[64] ....................................................................................... 11
Figure 1-5. Schematic diagram of plasma etching system[65] .................................................................................... 13
Figure 1-6. Challenges to develop MEMS DSC compared to the conventional DSC. Figure 1-6.a is the schematic set
up for commercial macro DSC[74]. 6b shows a microfluidic chamber device[75]. 6c and 6d show the
suspended membrane for the MEMS DSC and the temperature sensing and heating unit[28] .......................... 15
Figure 1-7. Two typical microfluidic chamber fabrication method. (a) Glass-based microfluidic chamber fabrication
process. (b) PDMS microfluidic chamber fabrication process ........................................................................... 18
Figure 1-8. Two temperature sensing examples in calorimetry application. (a) Four SiC thermistors form a
Wheatstone bridge for the temperature measurement[82]. (b) Thermopile junctions consist of Sb and Bi (40μm
wide, 2 mm long, 0.5 and 1μm thick, respectively)[89]. .................................................................................... 20
Figure 1-9. (a) Sensitivity calibration of a MEMS DSC using Joule heat[92]. (b) DSC scans showing the analysis of
spray-dried lactose at a variety of high heating rates using fast-scan DSC [93]. ............................................... 21
Figure 1-10. (a) The schematic diagram of the MEMS based DSC device. (b) The image of the integrated micro
heater and thermistor under microscope. (c) The schematic drawing of the MEMS based DSC with the
microfluidic device on it. (d) The MEMS based DSC on the heating stage....................................................... 23
Figure 2-1. The DSC model. (a) The schematic diagram for DSC which consists of sample and reference cells. The
cells are located in a well-designed test chamber. (b) The thermal model for the DSC. .................................... 27
Figure 2-2. DSC thermogram for the hen lysozyme denaturation process. ................................................................. 30
Figure 2-3. The schematic diagram of the Wheatstone bridge consists of two thermistors (R1 and R2) and two
external decade resistor boxes (R2 and R4) and the lock-in amplifier for differential detection. ....................... 32
Figure 2-4. (a) The numerical model of the micro heater in the FEA analysis. (b) The design for the micro heater. (c)
The micro structure of the heater, the width of each circle is W1, the space between them is W2. ..................... 38
Figure 2-5. The temperature distribution in the polyimide membrane by FEA simulation. ........................................ 40
Figure 2-6. Temperature distribution along the diameter of the heater for different heater design. ............................ 40
Figure 2-7. The peeled off micro heater array. ............................................................................................................ 42
Figure 2-8. High temperature probe station for the characterization. .......................................................................... 42
Figure 2-9. R-T characterization of the micro heater................................................................................................... 43
Figure 2-10. (a) The gold heater image under the microscope. (b) The IR image of the gold heater under 6.28 mW
heating power. (c) The temperature distribution in the heating area measured by the IR camera...................... 44
Figure 2-11. The measured heater temperature under different heating power consumption in the atmosphere. ....... 45
Figure 2-12. The step response of 8mW heating power for the micro heater when there is 2μL water on the top. ..... 45
Figure 2-13. Thermistor design. (a) normal resistor design. (b) side to side resistor design. (c) parallel thermistor
design ................................................................................................................................................................. 48
x
Figure 2-14. The design of the microfluidic device. (a) the top layer consists of inlets, outlets and a chamber to form
the top airgap. (b) the assembled microfluidic device consists of a top layer and bottom layer. ....................... 52
Figure 2-15. The three types of microchamber design and the corresponding final temperature distribution for the
FEA models. ...................................................................................................................................................... 53
Figure 2-16. The thermal response of the gravity point for all the three types of chamber design. (Design 1 is the
chamber with no air gaps, design 2 is the chamber with side air gaps and design 3 is the chamber with both top
and side air gaps.) ............................................................................................................................................... 54
Figure 2-17. The 3D model of the heating stage ......................................................................................................... 55
Figure 2-18. The top view of the heating stage. .......................................................................................................... 56
Figure 2-19. The cross-section view of the heating stage. ........................................................................................... 56
Figure 2-20. The polyimide sample. (a) Polyimide on silicon wafer carrier after curing. (b) Flexible polyimide thin
film after peeling off from the silicon wafer. ..................................................................................................... 59
Figure 2-21. The whole fabrication process of the micro heater and thermistor. ........................................................ 60
Figure 2-22. (a) The fabricated micro heater. (b) The thermistor on a polyimide substrate. (c) The MEMS DSC
device with the integrated microfluidic chamber, the micro heater, and thermistor. (d), (e) and (f) show the
micro structure of the micro heater, thermistor and the structure in the overlap area. ....................................... 62
Figure 2-23. (a) The prototype of the heating stage. (b) The heating stage without the thermal cover, the marked
position shows the placement of the temperature sensor (TE Tech MP-3022). ................................................. 63
Figure 3-1. Van der Pauw method for thin film resistivity measurement[151]. .......................................................... 68
Figure 3-2. Oxygen flow rate dependence of the resistivity of VOx under room temperature. ................................... 70
Figure 3-3. Resistivity variation of VOx films with different annealing temperature. ................................................. 70
Figure 3-4. Resistance–temperature characteristic of VOx thin film .......................................................................... 71
Figure 3-5. lnρ-1/T linear fitting. ............................................................................................................................... 71
Figure 3-6. The SEM image of the VOx thin film ....................................................................................................... 72
Figure 3-7. The XRD pattern of the VOx thin film ...................................................................................................... 72
Figure 3-8. The relationship between the oxygen flow rate and the pressure inside the tube of the furnace. .............. 76
Figure 3-9. The resistivity VS temperature curves for VO2 thin film (sample AII in Table 6) with different scanning
rates. The inset shows the measured amplitude of the TCR (-1ρ(dρdT)). Tm is the transition temperature at
which the vanadium oxide undergoes abrupt change in resistivity. It is determined by the largest TCR
amplitude. ........................................................................................................................................................... 77
Figure 3-10. The resistivity records of the series A vanadium oxide thin films for both the heating and cooling
cycles.................................................................................................................................................................. 78
Figure 3-11. The amplitude of the TCR of the series A vanadium oxide thin films in the temperature range from
30oC to 110oC. TmH is the transition temperature in the heating cycle, while the TmC is the transition
temperature in the cooling cycle. ∆T(TmH-TmC) is the difference of the transition temperature between the
heating and cooling cycle, which indicates the thermal hysteresis of the VO2 thin film. .................................. 78
Figure 3-12. The resistivity records of the series B vanadium oxide thin films for both the heating and cooling
cycles.................................................................................................................................................................. 80
Figure 3-13. The thermal hysteresis of the vanadium samples oxidized under different temperatures. ...................... 80
xi
Figure 3-14. The XRD patterns of series A vanadium oxide thin films oxidized at 425oC with different .................. 81
Figure 3-15. The XRD patterns of series B vanadium oxide thin films oxidized at different temperatures with the
same oxygen flow rate of 2sccm. ....................................................................................................................... 81
Figure 3-16. SEM graphs for the vanadium oxide thin films with different oxidation conditions (a) 400oC, 2sccm
(BII), (b) 425oC, 2sccm (BIV), (c) 410oC, 2sccm (BIII), (d) 425oC, 1sccm (AI), (e) 425oC, 5sccm (AIII), (f)
390oC, 2sccm (BI). ............................................................................................................................................. 82
Figure 4-1. The thin film model and the coordinates ................................................................................................... 88
Figure 4-2. The design of the MEMS-based DSC ....................................................................................................... 90
Figure 4-3. (a) The MEMS based DSC model. (b) The grid system for the DSC model. ........................................... 91
Figure 4-4. Temperature distribution in the polyimide membrane (the distance between the two heaters was 4mm.).
........................................................................................................................................................................... 91
Figure 4-5. The temperature distribution in the path AA shown in Fig. 4-4. (a) The distance between the two heaters
is 4mm (∆TAmax = 0.0024). (b) The distance is 3mm (∆TBmax = 0.069). ........................................... 92
Figure 4-6. (a) The heating stage with thermal cover. (b) The heating stage without thermal cover. ......................... 94
Figure 4-7. The linearity of the temperature scanning process .................................................................................... 95
Fig. 4-8. (a) The temperature stability study of the heating stage with the two shielding layers. (b) The temperature
stability study of the heating stage without the two shielding layers. ................................................................ 96
Figure 4-9. (a) The IR image of the sample stage. (b) The 3D diagram of the temperature distribution on the surface
of the sample stage. ............................................................................................................................................ 97
Figure 4-10. The schematic diagram of the Wheatstone bridge consists of two thermistors (R1 and R2) and two
external decade resistor boxes (R2 and R4) and the lock-in amplifier for differential detection. ....................... 99
Figure 4-11. The schematic diagram of the whole MEMS based DSC measurement system. .................................... 99
Figure 4-13. Relationship between the sensitivity and the temperature of the heating stage. ................................... 100
Figure 4-14. Micro heater sample with cracks before temperature scanning ............................................................ 101
Figure 4-15. The micro heater sample shows cracks after the temperature scanning. The left figure is the image of
the heater before the temperature scanning. The right figure shows the image of the heater after temperature
scanning. .......................................................................................................................................................... 102
Figure 4-16. The real-time recording of the power added to the heater 1 (heater located in the sample area) and
heater 2 (heater located in the reference area). ................................................................................................. 103
Figure 4-17. The differential power between the two micro heaters. The inset shows the recording between 0s to
100s as the power value is too small. ............................................................................................................... 104
Figure 4-18. The output voltage recording for the temperature scanning process. .................................................... 104
Figure 5-1. DSC data for interpolation. (a) The blue line is the raw data of protein unfolding obtained by a
commercial DSC; the red line indicates the buffer heat capacity (b) after baseline subtraction, the final result
of the protein denaturation heat capacity. ........................................................................................................ 108
Figure 5-2. (a)The MEMS based DSC system in protein sample tests. (b) The TA discovery DSC for protein sample
tests .................................................................................................................................................................. 111
Figure 5-3. (a) The baseline tests of the TA discovery DSC. (b) The baselines after removing the linear part. ....... 112
Figure 5-4. (a) the baseline tests of the MEMS based DSC. (b) the baselines after removing the linear parts. ....... 112
xii
Figure 5-5. The molar heat capacity change and enthalpy of the lysozyme over the temperature in the scanning
process using both the MEMS based DSC. (a) and (b) show the heat capacity change of the lysozyme while
(c) and (d) show the enthalpy change. .............................................................................................................. 114
Figure 5-6. The molar heat capacity change and enthalpy change of the lysozyme over the temperature in the
scanning process using the TA discovery DSC. (a) and (b) show the heat capacity change of the lysozyme
samples while (c) and (d) show the enthalpy change over the scanning temperature. ..................................... 115
Figure 5-7. Service life tests of the MEMS based DSC. (a) and (b) show two MEMS DSC devices were reused in
lysozyme tests. ................................................................................................................................................. 116
Figure 5-8. The lysozyme tests using the MEMS based DSC with different pH values. .......................................... 117
Figure 5-9. The lysozyme sample tests using the MEMS based DSC with different concentrations. ....................... 117
Figure 5-10. The structure of antibody ABT 981 and mAb 696 ................................................................................ 118
Figure 5-11. Differential scanning calorimetry thermogram of ABT 981. The unfolding transitions for the individual
domains of ABT-981 are separated. ................................................................................................................. 119
Figure 5-12. Differential scanning calorimetry thermogram of mAb-696. The unfolding transitions for the individual
domains of mAb-696 are separated. ................................................................................................................. 119
Figure 5-13. The enthalpy change over the scanning temperature for both AbT-981 and mAb-696. ....................... 120
Figure 5-14. The temperature distortion induced by the response time. .................................................................... 123
Figure 5-15. The relationship between the response time and the Tm shift. .............................................................. 123
Figure 5-16. The DSC measurements of lysozyme samples with different scanning rates. ...................................... 126
Figure 5-17. Linear fitting between the lnvTm2 and 1000/Tm. ................................................................................. 127
Figure 5-18. The molecular structure of the Fab, mAb and DVD IgG. ..................................................................... 131
Figure 5-19. The DSC curves of the three proteins obtained by the MEMS based DSC. (b) The DSC curves of the
three proteins obtained by the MicroCal VP-Capillary DSC for comparison. ................................................. 134
Figure 5-20. The deconvolution of the normalized DSC curves for the three protein samples (Fig. 5-20a,c,e are the
DSC curves measured by the MEMS DSC, Fig. 5-20b,d,f are the curves measured by the VP DSC). ........... 135
Figure 6-1. (a) The microfluidic device with a mixer. (b) The detailed structure of the micro mixer. (c) The
temperature sensor on the polyimide substrate. ............................................................................................... 144
Figure 6-2. A simplified model of a typical receptor/ligand-binding interaction. The ligand in this representation is
shown to geometrically match the binding site on the receptor to indicate a specific binding interaction [206].
......................................................................................................................................................................... 145
Figure 6-3. (a) The recording of the heat release for multiple titrations with the different molar ratio. (b) The
integration of all the peaks in Fig. 6-3a. ........................................................................................................... 145
Figure 6-4. (a) The schematic diagram for the MEMS based calorimeter array. (b) The single unit of the calorimeter
array. ................................................................................................................................................................ 148
xiii
List of Tables
Table 1. The comparison of the proposed MEMS DSC and the current DSC ............................................................. 23
Table 2. The material properties of the heater in the FEA analysis. ............................................................................ 38
Table 3. The properties of the materials in the FEA simulation. ................................................................................. 52
Table 4. Specifications of the Peltier module .............................................................................................................. 57
Table 5. A summary of different deposition techniques for the growth of vanadium oxide with the process
temperature and the reported TCR values. ......................................................................................................... 67
Table 6. The vanadium oxide sample series and the correlated oxidation conditions. ................................................ 77
Table 7. Specifications of the FLIR A655sc infrared camera. ..................................................................................... 96
Table 8. The statistical analysis of the temperature distribution in the box 2 area shown in Fig. 56a. ........................ 97
Table 9. Lysozyme test results using both the commercial TA discovery DSC and the MEMS DSC. ..................... 113
Table 10. MicroCal VP-capillary DSC test results VS the MEMS based DSC test results. ...................................... 136
1
Chapter 1. Introduction
1.1 Calorimetry
Differential scanning calorimetry (DSC) as a thermo-analytical technique was developed by E.S
Watson and M.J O’Nell in 1964[1]. Since then, a lot of research has focused on developing high
throughput, high sensitivity DSC and the application of DSCs into different research areas such as
polymer study, biomolecular study, drug design, etc[2-4]. A DSC consists of twin cells (sample
and reference) in which temperature sensor and heating module are integrated and operate in
differential mode. During a differential temperature scanning process, the difference in
temperature (power consumption) between the sample and reference material is measured as a
function of the linear temperature cycle. The difference directly reveals the heat release or
absorption of the sample over temperature which further indicates the heat capacity change. There
are mainly two types of DSCs based on the working principle: heat flux DSC[5] and power
compensated DSC[6]. For the power compensated DSC, the temperature of the sample and
reference material are always kept the same by varying the heat flow to the sample and reference
during the linear temperature scanning process. While in a heat flux DSC, the temperature
difference is directly recorded during the same procedure. Together with the thermal resistance,
the temperature difference can be converted to the heat flow difference[7]. Recently, another type
of DSC was developed which is called temperature modulated differential scanning calorimetry
(TMDSC) or alternative current differential scanning calorimetry (AC DSC)[8, 9]. The basic idea
of the TMDSC is to add a controlled temperature modulation to the conventional linear heating.
The heating process of TMDSC can be divided into two parts. The first part is to heat the sample
at a certain temperature scanning rate just like the conventional DSC. In the second part, the heat
2
capacity component of the heat flow is obtained by applying a controlled oscillating temperature
modulation with a zero net temperature-change. In a TMDSC thermal analysis, the average
scanning rate, the period of modulation and the temperature amplitude of modulation are three
important variables that are tuned to optimize the experiment.
Virtually all biological phenomena depend on molecular interaction. A basic biology problem is
to understand the folding and denaturation processes of a protein, i.e., the kinetics and
thermodynamics how a protein unfolds and folds back into its native state[10, 11]. Both
folding/unfolding and denaturation processes are associated with enthalpy changes. The
intermolecular interactions such as protein-ligand association, protein/DNA interaction, antigen-
antibody binding processes are either enthalpically or entropically driven, depending upon the
binding modes of the small molecules [12-14]. The determination of the affinity thermodynamics
of binding compounds helps greatly to understand the nature of such molecules [15, 16]. The
specificity of binding reactions has fascinated biologists from the very beginning to quantitatively
describe the driving forces that govern the formation of biomolecules [17-19]. This has great
significance to the development of vaccines, new drugs and other molecular compounds [20, 21].
There is a rapid growing number of high-resolution crystal structures of biomolecules. However,
the theoretical concepts which are developed in the tradition of physical-organic chemistry are not
enough to understand such driving force of the large complex biomolecular structure
straightforwardly. The proteins behave cooperatively and undergo structural rearrangements
during the binding reactions. Such binding process has a complicated energy profile involving
different energetic expenditures in going from free components to the final complex which can be
detected directly by DSC [22, 23].
3
There are several commercial DSCs for large biomolecular study. One is the MicroCal VP-DSC
developed by Malvern. It can directly measure the intramolecular stability of structured
macromolecules as well as the intermolecular stability of complexes such as proteins, nucleic acid
duplexes, and lipid and detergent micellar systems. It consumes 500µL sample for each test and
has a high enthalpy resolution as 0.01J/oC[24]. The sample concentration typically ranges from
0.1mg/ml to 2mg/ml. As a capillary DSC, it is important to clean the reservoir each time after the
test, and this limits the concentration of the bio-sample since high concentration protein sample
would stick on the wall of the capillary after denaturation and it is hard to be washed away. The
maximum scanning rate of the MicroCal VP-DSC is only 1.5oC/min which leads to relatively low
throughput. Another typical commercial DSC is the TA DSC[25]. As the most cost-effective DSC
in industry, it has a different sample preparation strategy compared to the MicroCal VP-DSC.
Unlike the VP-DSC which use a syringe to inject the liquid sample to the reservoir through a
capillary system, the TA DSC utilizes aluminum pan for the sample preparation and sealing. The
sample pan and reference pan are then put on the pedestals in the test chamber for temperature
scanning. It can measure protein sample with high concentration as 100mg/ml. It can also be used
to conduct thermoanalysis of polymers, solid crystals, etc. One limit of the TA DSC is its relatively
low-temperature accuracy (0.025oC for DSC 2500).
Although the commercial DSC can provide high repeatability and reliability for biomolecular
study, one major drawback of such macroscale DSC is the large thermal mass of the calorimetric
cell itself and its associated hardware. Consequently, the required energy input may be large
compared with the energy associated with the thermal transitions to be measured. Hence, they
usually require a large amount of sample. This can pose problems in cases where only minute
sample quantities are available for testing. There has been a strong demand to develop high
4
throughput and high sensitivity calorimeter for rapid detection of biomolecular interaction for
decades [26]. The microfabrication and nanofabrication technologies allow the miniaturization of
the conventional bench-top scale instrument for the building of micro/nano DSC with microfluidic
embedded systems, micro temperature sensors and heaters [27-29]. Such miniaturized DSC has
many advantages compared to the current commercial DSC such as parallel operation, less sample
consumption, shorter measurement time and higher sensitivity [30, 31]. The current commercial
DSC consumes tens to hundreds of minutes to test one sample. It would take months to do the
thermoanalysis for 1000 compound samples with the current commercial DSC in drug
development industry. The MEMS based DSC has the potential to shorten the measurement time
from months to days by offering higher scanning rate and parallel test strategy. The MEMS based
calorimeter usually consumes micro littler or even nanoliter of sample per test which is more than
ten times less than the commercial DSC. This can help to save the cost for biomolecular sample
consumption. Protein samples such as MAN-9 Glycan would cost $43B/mol, the miniaturized
MEMS DSC would help to save a lot of budget in the study of such expensive compound[32].
On the other hand, since the sample consumption is much less, the total thermal mass of the MEMS
DSC is significantly reduced which permits higher scanning rates [33, 34]. The miniaturized DSC
unit will also allow us to integrate multiple DSC on one silicon wafer or polymer substrate. There
were already some attempts to develop such calorimeter array [35-38].
In this section, we will firstly discuss about the basic thermodynamic relationship in the binding
process of biomolecular sample during the temperature scanning test, then typical DSC
thermodiagram will be analyzed. We will then review the most up to date progress in the
development of MEMS based DSC and its challenges at last.
5
Calorimetry is a principal science in testing and measuring the thermal properties of chemical
reaction, physical changes and phase transitions [39, 40]. It allows researcher to establish a
connection between temperature and specific properties of samples. Meanwhile, it’s the only
approach for direct measurement of the heat transfer. Different kinds of calorimeters are used
frequently in various areas, such as physics, chemistry, biology, biotechnology, pharmacology,
which are to characterize thermal properties of materials.
In the field of science, it’s essential to determine the thermal properties of the experimental
samples. There are several environmental factors will contribute to different properties, such as
temperature, pressure, and humidity. Among these environmental factors, temperature plays a vital
role in the thermal properties.
In order to characterize the temperature dependence of material properties, hundreds calorimetric
and thermal analysis approaches have been developed since 18th century, Scottish physician and
scientist Joseph Black, who discovered the difference between heat and temperature, he came up
with the new concept of measurement of the temperature and heat transfer during the different
stage of thermal transition. And recently, in the field of nanoscience, nanocalorimeters have been
already widely employed to measure thermodynamic properties of the biomolecules and materials
in nano-scale, such as heat transfer in protein folding and ligand binding[41, 42].
In the thermoanalytical technique, DSC is the most practical modern tools in the determination of
heat transfer with detection of phase transitions. In 1962, E. S Watson and M. J. O’Neill invented
this important technique, and they brought this technique into public commercially in the 1963
Pittsburgh Conference on Analytical Chemistry and Applied Spectroscopy[1].
6
Differential scanning calorimeter can describe the how physical and chemical properties of sample
changes as a function of temperature and time. In other words, this technique is an experimental
approach for measuring the energy necessary to maintain the same temperature of the test sample
and the reference, while the sample is undergoing some physical or chemical reaction.
In a standard DSC experiment, as shown in Fig.1-1, the sample cell and reference cell are heated
by furnace at same time, temperatures of sample cell and reference cell will be raised along the
time, the amount of excess heat absorbed or released by the molecule in the sample will be exactly
the same as the difference in the input energy required to match the temperature of the sample to
that of the reference.
Figure 1-1. (a). The schematic diagram of DSC. (b). The DSC curve of the reference material. (c).
The DSC curve of the sample material. (d). Heat capacity change of the sample material along the
temperature change.
In the past decades, techniques based on DSC have been developed into functional instrument to
advance the molecular measurements of biomolecules, such as conventional/basic DSC[43],
MEMS-DSC[44, 45], infrared (IR) heated DSC[46, 47], modulated-temperature DSC
(c)
(d)
(a) (b)
7
(MTDSC)[48, 49], gas flow-modulated DSC (GFMDSC)[50, 51], parallel-nano DSC
(PNDSC)[52], pressure perturbation calorimetry (PPC)[53-55], self-reference DSC (SRDSC)[56],
and high-performance (HPer) DSC[57]. These DSC-based techniques have been widely used in
the determination of structural-phase transition, melting point, the heat of fusion, percent of
crystallinity, crystallization kinetics and phase transition, oxidative stability, thermodynamic
analysis of biomolecules, and curing kinetics of non-biological materials.
Depending on the operation principle, differential scanning calorimeters can be divided into two
types, heat-flux DSCs and power-compensated DSCs[58]. In a heat-flux DSC, the sample material
and empty reference are enclosed in separate pan and are both placed on a thermoelectric disk,
meanwhile, the whole disk is heated in a furnace which is heated at a linear heating rate, and the
energy will be transferred to the sample and reference material through the thermoelectric disk.
Due to the different heat capacity of sample and reference, there will be a temperature difference
(∆𝑇 ) between the sample and reference, the ∆𝑇 will be measured by thermocouples and
thermistors. Reversely, in a power-compensated DSC, the sample and reference will be placed in
separated furnaces powered by different heaters. The temperatures of sample (Ts) and inertia
reference (Tr) are controlled in separate ovens with different heat sources. The temperature of the
sample and inertia reference is maintained the same with feedback sensor to control the power
input to two ovens, in other words, different energy are applied to sample and inertia reference so
as to get the same temperature. Then the energy difference (∆P) in the process is a measure of heat
capacity or enthalpy changes in the sample.
8
1.2 MEMS Fabrication Techniques
Regarding MEMS fabrication process, we can easily distinguish it from traditional machining and
manufacturing, at least at the present stage, it does not include some basic machining methods
such as milling, lathing, polishing, joining and welding, for most MEMS and IC devices are
commonly fabricated on single crystal silicon wafers whose resolution always stays in micron
range[59].
There are a bunch of technologies and processes which are employed in microelectronics and
MEMS fabrication. To understand the process involved in the MEMS and IC fabrication, we need
to get in touch with several subjects such as biology, physics, and chemistry. In other words,
MEMS is an interdisciplinary technology which covers mechanics, electronics, chemistry and
material science[60].
All the processes related to MEMS fabrication can be sorted according to the major categories
such as additive processes, subtractive steps, patterning, and material modification. Here several
representative technologies and machines will be introduced, a metal evaporator and sputter for
deposition of the metal thin film, the contact aligner for the lithograph, and a plasma etcher for
removal of materials.
1.2.1 Deposition of Electronic Materials
For many nano-scale electronics parts, such as thermistors, transistors and conductors can be
fabricated on a silicon wafer with the additive deposition process. Two most popular methods for
the additive deposition processes are physical vapor deposition and chemical vapor deposition.
Among the two methods, physical vapor deposition process (PVD) is a direct transfer of material
from a source to the wafer surface in an atom-by-atom, layer-by-layer form. The second approach
9
for placing thin film materials on a wafer surface is called chemical vapor deposition (CVD), two
or more than two active particles will react under certain conditions such as heating atmosphere or
plasma bombardment and then transferred to the surface of the wafer.
D. M. Mattox once mentioned that physical vapor deposition (PVD) is atomic deposition processed
in which material is vaporized from a solid or liquid source in the form of atoms or molecules,
transported in the form of a vapor through a vacuum or low pressure gaseous (or plasma)
environment to the substrate where it condenses[61].
Physical vapor deposition (PVD) works with the principle which directly transfers material from
source to certain substrate, including metal evaporation and metal sputtering. In most cases, the
deposition process occurs in a relatively low-pressure environment, between 10e-5 and 10e-9 Torr,
to maintain the stable transport of desired atoms and particles to the surface of the wafer without
the interruption of air molecules.
For thermal or e-beam thin film deposition (metal evaporation), the system is arranged in Fig. 1-
2. The substrate wafer and the metal source are both placed in a vacuum chamber, then the source
(metals) can be transferred by heating it till evaporation, the heating methods can be resistive heater
and e-beam heater, what’s more, as the source is relatively far from the substrate, the heating
source will leave ignorable damage on the substrate wafer. In sputtering deposition system. The
chamber is filled with inert gas such as argon (Ar), meanwhile, DC voltage or RF source are
applied to the diode for the ionization of argon, which allows the generation of free electron, the
free electron and high-energy ions in the chamber can be accelerated by magnetic field, then the
metal can be transferred by bombarding it with high-energy ions. The thickness of the thin film is
10
proportional to the power and time. In most cases, the thickness of metal thin films ranges from
1nm to 2 µm.
Figure 1-2. The schematic diagram of thermal/ebeam deposition system
Another method for depositing thin film material on a substrate wafer is chemical vapor deposition
(CVD). In chemical vapor deposition, two or more active species (in the form of vapor) arrives at
the wafer surface simultaneously. Then certain chemical reaction occurs under desired conditions
such as energy provided by heating or plasma, which results into the solid materials produced by
these species[62, 63]. Process parameters (gas flow, reactor pressure, and substrate temperature)
are highly regulated so that the precursors dissociate into the proper reactive components such that
the desired material is formed on the substrate surface and not in the vapor. The solid phase is
absorbed onto the wafer surface nearby. The gaseous byproduct of the chemical reaction may be
removed by the nearby media. As the chemical reaction going on, a layer of desired material will
be built on the wafer surface. The whole process is illustrated in Fig. 1-3.
11
Figure 1-3. Illustration of the CVD process.
1.2.2 Photolithography
Photolithography plays an indispensable role in the MEMS fabrication process, the purpose of
lithography is to structure concise features on the wafer surface. The lithography patterning results
(thickness) varies according to the spinning speed, the viscosity, and types of photoresist and the
temperature and time of bake process. In most cases, the typical thickness of the photoresist is
from 1µm to 10µm.
Figure 1-4. Illustration of the photolithography process[64]
12
The detailed lithography process could be described as following, (a) clean the prepared wafer, (b)
enhance the adhesion on the wafer surface by plasma treatment, (c) spin coat photoresist on wafer,
and soft bake the wafer, (d) process alignment, (e) exposure the wafer to light with mask, (f)
development, (g) hard bake the wafer if needed. Mask is consisting of a transparent substrate with
opaque images which can block partial light from the mask aligner, avoiding the chemical
composition of resist being changed. For the part of resist exposed to lights directly, the chemical
composition has been changed which makes the resist more soluble in certain chemical solvent
(developer), this process can transfer the features on the mask onto the wafer precisely (Fig. 1-4).
1.2.3 Plasma Etching and Wet Etching
The principle of plasma etching is to deliver the chemically reactive species to the etching surface
in ionized gas (plasma). It is an important approach in removing materials from a wafer surface
(Fig. 1-5), plasma etching process does not involve any wet chemicals, so the plasma etching is
often referred to dry etching. In plasma etching, the plasma etcher is the main facility, which is
filled with chemically active gas species. During the process, the pressure in the etch chamber is
relatively low, so the gas species are broken up by the electric field, resulting in active gaseous
radicals that are electrically charged. The electrically charged radicals may have a chemical
reaction with the materials on the surface, and at the same time, the electrical field will accelerate
the charged radicals to high speed, then the charged radicals will collide with the materials on
wafer physically, in the form of bombardment or sputtering. In the plasma etching process,
physical etching and chemical etching are both important in etching materials, for physical etching
is more directional and anisotropic while chemical etching is isotropic and material selective.
13
Figure 1-5. Schematic diagram of plasma etching system[65]
Wet etching means the process which employs etchants in aqueous solution form to remove the
materials out of the wafer. Generally, material removal is accomplished by redox and solvation
processes. For example, in the process of silicon wet etching, the chemical process can be
described as following,
Si+2H+ → Si2+ , Si2++2(OH)- → Si(OH)2
Si(OH)2 + 6HF → H2SiF6+ 2H2O + H2
After the whole process, the H2SiF6 and some other reactants will be removed in etchant liquid.
1.3 MEMS-Based Calorimetry
MEMS-based calorimeter applies the standard micro-electro-mechanical system technique to scale
down the conventional benchtop size DSC to the wafer or even smaller size. A typical MEMS
DSC consists of three parts: the liquid delivers system, the thermal insulation structure, and the
temperature sensing and heating unit[66]. A MEMS DSC employs a liquid deliver system which
is similar to the current commercial DSC (MicroCal VP-capillary DSC). The liquid deliver system
is used to deliver the liquid sample and reference material to the thermal cells. It consists of a high
precision syringe pump, micro capillary, and other connection parts. The thermal insulation
14
structure includes the suspended thin film or polymer substrate with low thermal conductivity and
microfluidic chamber integrated with an air gap. The suspended substrate can be inorganic material
such as Si3N4, SiO2 or polymer such as SU8 or polyimide [45, 67, 68]. There is a trend to use
polymer to replace the traditional silicon complexes since their thermal conductivity is one or two
orders smaller. The microfluidic chamber is also fabricated using materials with low thermal
conductivity such as PDMS[69] or glass[70, 71]. Another requirement for the material in
microfluidic fabrication is the high transparency which permits real-time monitor of the
temperature scanning process. The microfluidic chamber can be fabricated with the MEMS
technique or using 3D printing with high resolution [72, 73]. The temperature sensing and heating
unit is the core of the MEMS DSC. It consists of the temperature sensing and heating part. There
are mainly two techniques to achieve temperature sensing with high sensitivity: thermopile or
thermistor. The function of the heating unit is to provide heat flux to the sample and reference
material in the thermal cells. In such micro-level heating with high thermal insulation, one big
challenge is to minimize the thermal gradient in the heating material. Because these systems
incorporate a very small thermal mass and use reagent quantities in the nanogram/nanoliter range,
rapid and uniform heating and cooling are highly expected while maintaining a high level of
temperature homogeneity. These miniaturized nano DSC can offer enhanced sensing capabilities
in an inexpensive portable format.
15
Figure 1-6. Challenges to develop MEMS DSC compared to the conventional DSC. Figure 1-6.a
is the schematic set up for commercial macro DSC[74]. 6b shows a microfluidic chamber
device[75]. 6c and 6d show the suspended membrane for the MEMS DSC and the temperature
sensing and heating unit[28]
Fig. 1-6 shows the challenge to fabricate the corresponding components in a MEMS DSC
compared to the commercial DSC. The crucible to contain the sample and reference material is
replaced by a microfluidic chamber. The pedestal structure is abandoned and replaced by a
suspended membrane provides a substrate for the integrated heating and temperature sensing unit
and the microfluidic chamber with excellent thermal isolation. Unlike the conventional DSC which
utilize cumbersome heating/cooling block and the macro thermocouple for temperature scanning
and sensing. In a MEMS DSC, the micro heater and temperature sensor are fabricated on the
suspended membrane directly using standard MEMS technique. The performance of a MEMS
DSC heavily depends on these three parts: microfluidic chamber, suspended membrane,
temperature sensing, and heating unit. The improvement of MEMS DSC is achieved by integrating
(a) (b)
(c)
(d)
16
novel microfluidic chamber structure with better thermal insulation, smaller addenda thermal mass
(heat capacity), a robust suspended membrane with lower thermal conductivity, temperature
sensing and heating unit with higher temperature sensitivity and more uniform temperature
scanning.
1.3.1 High Thermal Insulation Design for MEMS-Based DSC
The simplest way to achieve high thermal insulation is using suspended membrane under a high
vacuum environment. Since biomolecules are often resolved in liquid, there would be serious
evaporation problem without careful chamber system during the temperature scanning process
[76]. Hence, the simple open type structure which is utilized by most MEMS isothermal titration
calorimeters cannot be applied to the MEMS DSC [77, 78]. A carefully designed microfluidic
chamber system is designed to prevent the evaporation and provide high thermal insulation
performance. The thermal insulation system for MEMS DSC consists of two parts: the suspended
membrane as the substrate for the temperature sensing and heating unit and the microfluidic device
[79, 80].
Two types of the suspended membrane material are the silicon compound such as SiO2,
polysilicon, Si3N4 [81] and polymers[82]. Among the silicon compounds, Si3N4 is the best for its
chemical inertness and easy fabrication process. Fabrication of membrane using Si3N4 is mastered
in all cleanroom in the world using chemical etching of silicon coated with silicon nitride on both
sides by KOH based etchant. Thermometer and heater can be lithographed on top of the silicon
nitride membrane which permits the measurement of thermal properties. Polymer membranes are
also used to ensure good thermal insulation for MEMS DSC. In the temperature range applied to
MEMS DSC (room temperature to 110 oC for biomolecular study), polymers such as SU8,
17
polyimide has unique advantages such as robustness, low cost, low thermal conductivity, and
inertness.
The microfluidic chamber is the essential part of the thermal insulation system for MEMS DSC.
One method to fabricate the microfluidic chamber is to use Pyres glass for its high transparency,
low thermal conductivity, low cost and standard micro machining process[83]. There are several
steps to fabricate the microfluidic chamber with Pyres glass wafer. (1). Silicon etching. Using
silicon wafer as the master mold, the desired cavities are obtained by wet etching or DRIE etching.
(2) Anodic bonding. After cleaning, the silicon wafer is hermetically anodic bonding to the Pyrex
glass under a high vacuum environment. Typically, 1000V voltage is added between the silicon
wafer and the glass wafer. (3). Heat treatment. In this step, the bonded wafer is heated up to the
strain point of the Pyres glass (850oC) in a furnace under the standard atmospheric pressure. At
this temperature, the Pyres glass will be softened and fill the cavities on the silicon wafer. (4).
Annealing. In this step, the bonded wafer is put into the heat treatment furnace to take an annealing
process under the temperature of the annealing point of Pyrex glass. (5) Microfluidic device
separation. After the annealing process, the Pyres glass already forms the microfluidic chamber on
the silicon wafer by filling the cavities. In this step, the residual silicon is removed by wet etching.
Another popular method is the polydimethylsiloxane (PDMS) based microlithography technique.
There are several steps in such microfluidic chamber fabrication process [84, 85]. (1). SU8
patterning. In this step, SU8 is used to fabricate the master mold for the PDMS based microfluidic
chamber. It is firstly spin-coated on a cleaned silicon wafer and then patterned with standard
photolithography technique to form the predesigned cavities. (2). PDMS casting. The surface of
the SU8 is firstly treated under plasma for later release of the PDMS chamber. The PDMS is then
mixed with the base and agent under proper volume ratio (10:1 in weight) [86]. After the PDMS
18
casted to the silicon wafer, the whole device is put in a vacuum chamber for degassing. (3). Heat
treatment. In this step, the device is put onto a hot plate to slowly increase the temperature to the
curing temperature and hold for two hours (The curing temperature and time may vary depends on
the expected properties of the PDMS chamber such as Young’s modulus.). (4). PDMS microfluidic
device release. After the curing, the PDMS chamber can be released from the silicon wafer by
simply pilling off. The master pattern can then be cleaned and reused for the next fabrication. The
two fabrication methods are shown in Fig. 1-7.
Figure 1-7. Two typical microfluidic chamber fabrication method. (a) Glass-based microfluidic
chamber fabrication process. (b) PDMS microfluidic chamber fabrication process
1.3.2 Temperature Sensing Methods in MEMS-Based DSC
The sensitivity of a calorimeter is determined by the minimum thermal energy or power it can
detect. In most cases, calorimetric measurement is achieved by measuring temperature changes
[29]. Therefore, a high-sensitivity temperature sensor is needed in developing a high-sensitivity
calorimeter. There are mainly two temperature sensing methods in calorimetry: thermistor and
19
thermopile (Fig. 1-8). Thermopiles (thermocouples connected in series) are used as thermometers
in most MEMS DSC[87-89]. The thermocouple converts the temperature difference to electric
voltage via thermoelectricity. As thermopiles can have very high sensitivity by simply increasing
the number of thermocouples and does not consume electric power to measure temperature, they
are ideal for calorimetric applications. However, there are also limits for such method. First, the
space will limit the number of thermocouples integrated into the thermopile. Second, the
thermocouple usually has large resistance, and they are connected in series, the electrical resistance
will lead to large thermal noise. The other method is thermistor. The working principle of the
thermistor is the electrical resistance is temperature dependent. Some semiconductor materials
such as SiC, VOx have an extremely high-temperature coefficient of resistivity (-2% to -5% per
Kelvin) which are tens or even hundreds of times larger than the common metal materials [90, 91].
The problem with the thermistor is they need current for the temperature sensing which will
generate additional Joule heat to the sample cells. Hence the current to the thermistor must be
minimized for the temperature sensing or carefully controlled.
(a) (b)
20
Figure 1-8. Two temperature sensing examples in calorimetry application. (a) Four SiC thermistors
form a Wheatstone bridge for the temperature measurement[82]. (b) Thermopile junctions consist
of Sb and Bi (40μm wide, 2 mm long, 0.5 and 1μm thick, respectively)[89].
1.3.3 MEMS-based Calorimeter for Protein Stability Study
MEMS DSC can be extended to apply for biomolecular study with less sample consumption, lower
sample concentration, and higher temperature scanning rate duo to its minimized size. One of the
most important parameters of a MEMS DSC is its power sensitivity measured by the
corresponding output voltage to a certain heating power difference between the sample and
reference cells (V/W). A typical sensitivity calibration method is to apply a certain amount of
Joule heat to the sample cell (microwatts level) and leave the reference cell unheated or vice versa
while both cells are filled with buffer. This requires the heating module can heat the cell uniformly.
By this method, we can obtain the continuous response to the different power difference by simply
changing the current applied to the micro heater. Other methods such as the mixing of a certain
amount of acid and alkaline in the sample chamber while filling the reference cell with buffer to
produce heat can also be used to calibrate the sensitivity. These methods can lead to more accurate
sensitivity calibration since they use chemical reactions instead of Joule heat to produce the power
difference which is more close to the real test. However, these methods also require precise
reaction rate control to generate constant power for stable output which will make the system too
complicated. Fig. 1-9a shows the sensitivity calibration of a MEMS DSC using Joule heat. The
sensitivity of the MEMS DSC is 6.1V/W which is high enough to detect weak signal during the
scanning process.
MEMS DSC can increase the scanning rate of the liquid sample to 100oC/s or even higher by
greatly reducing the total thermal mass of the DSC system which is impossible for traditional DSC.
21
The super high scanning rate can be extremely useful to detect weak binding and lead to other
discoveries. Fig. 1-9b shows one example of using a fast MEMS DSC for lactose study. Under
high scanning rate (100-500 oC/min) not only the melting temperature is shifted, but also the
thermal domains which cannot be detected under lower scanning rate (20oC/min) is clearly shown
in the diagram.
Figure 1-9. (a) Sensitivity calibration of a MEMS DSC using Joule heat[92]. (b) DSC scans
showing the analysis of spray-dried lactose at a variety of high heating rates using fast-scan DSC
[93].
1.4 Motivation and Significance of This Dissertation
The calorimeter is considered to be the golden standard for characterizing biomolecule interaction
since it is label-free, requires no immobilization and free from user bias. In the drug discovery
industry, there is a growing need for high throughput calorimeter. The usage of calorimeter for this
field mainly lies in the evaluation of the stability of thousands of mutant proteins in the later
experimental stage and to test the binding affinity for detailed characterization of the screened drug
candidate. These thermodynamics characterizations relates to molecular structure and could
expand our understanding of how the biological system works. If this process failed, the previous
effort: on average ten year of scientific investigation and hundreds of millions of expenditure could
(a) (b)
22
be in vain due to toxicity[94]. Conventional calorimeter generally suffers from long-time
measurement (about 30 min to 2h), large consumption of molecular samples(20-300µL, 0.5-
5mg/ml) and hard for parallel measurement [95]. For new drug design and Even if equipped with
robotic hands to manipulate continuously day and night, it will still require months to years to
finish it all. Potentially, MEMS based calorimeter could be the solution, since it could be able to
decrease the time by three orders of magnitude or even more for shorter time constant and parallel
measurement and with µL instead of mL for sample consumption.
This research is supposed to develop highly sensitive DSC potentially could be applied in the drug
discovery industry. The proposed device is supposed to enhance the SNR (signal to noise ratio)
and resolution, so that previous undetectable thermal phenomenon of the biological process could
be investigated. The novel MEMS-based calorimeter could substantially reduce the sample
requirement amount down to µL scale and shorten the measurement time from hours to minutes
(Fig. 1-10). This could greatly boost the efficiency and save cost in the pharmaceutical industry.
The new developed DSC will be used in probing thermodynamics analysis of protein stability and
drug screening. Currently, at Abbott, a fluorescence-based thermal shift instrument is used for
thermodynamic analysis of protein stability at impressive small sample volume (15-50ul).
However, these thermal shift assays can only obtain the shift of Tm due to ligand binding, which
does not always indicate changes in affinity. In this task, the full thermodynamics of some drug
hits will be studied to complement the approach used in Abbott Laboratories. A highly efficient
instrument for generating the entropy and enthalpy of ligand macromolecule interactions would
contribute greatly to prioritizing among the many chemical candidates in the early stages of
pharmaceutical lead discovery. In the later stage of the research, the MEMS based calorimeter will
23
be assessed for screen protein drug candidates and formulations option to identify the most stable
molecules and guide the selection of the drug substance and drug product composition. The
comparison of the proposed MEMS DSC and the current available DSC is shown in Table 1.
Figure 1-10. (a) The schematic diagram of the MEMS based DSC device. (b) The image of the
integrated micro heater and thermistor under microscope. (c) The schematic drawing of the MEMS
based DSC with the microfluidic device on it. (d) The MEMS based DSC on the heating stage.
Table 1. The comparison of the proposed MEMS DSC and the current DSC
My DSC TA Discovery DSC [92]
Baseline repeatability
(dynamic noise)
0.4µW 4µW Not available
Static noise 10nW Not available 14nW
Scanning rate 10oC/min-45oC/min 1oC/min-10oC/min 1oC/min-6oC/min
Measurement time 8-20min 20-90min 30-90min
Sample consumption 0.63µL(10-100mg/ml) 20µL (1-20mg/ml) 1µL (1-20mg/ml)
(a)
(d)
(c)
(b)
24
1.5 Dissertation Outline
In this dissertation, a MEMS based differential scanning calorimeter with VOx based thermistor
will be introduced. This calorimeter can be applied in biomolecular interaction, such as protein
folding, and could eventually be applied in drug discovery for rational molecular design.
Chapter 1 is the introduction and background. First, it starts with the concept of the calorimeter.
Calorimeter applied in biological field is one that operates by detecting heat signal. This method
is superior in that it is label-free and does not require immobilization. Furthermore, it can be
applied to a ubiquitous biological phenomenon. ITC and DSC are two types of calorimeter
classified by the operation mode of the calorimeter. The dissertation then looked into their working
principle and the correlation of drug discovery with ITC and DSC, (especially the DSC). MEMS
technology caters the need for miniaturization and potentially could achieve parallel measurement
and high throughput, and thus it has been incorporated into latest calorimeter technology.
In chapter 2, the working principle, design, and fabrication of the calorimeter system will be
presented. The MEMS based calorimeter utilizes polyimide suspended thin membrane for thermal
isolation and Vanadium Oxide thermistors which further with a variable resistor to form a
Wheatstone bridge. The Flexdym/PDMS double layer microfluidic device is used to deliver the
bio-samples with high thermal insulation property. A heating stage with high scanning rate is
developed for linear temperature scanning.
In chapter 3, the sensing material and its characterization will be introduced. It firstly starts by a
theoretical introduction to thermistor sensing basics, thin film resistivity characterization method
and semiconductor conduction as well as TCR theories. Following this is the VOx related topics.
We explored the optimized condition and parameter for PVD deposition of VOx to satisfy the
25
requirement for this application: high TCR and low resistivity. To better understand the material,
SEM and XRD measurements are also conducted. We also extend the research topic to the
optimization of VOx thin film fabrication and characterization. In this chapter, the metal-insulation
transition of VO2 grown under certain conditions.
In chapter 4, the performance of the MEMS-based DSC will be characterized. In the first section,
the optimization of the distance between the sample and reference heaters is carried out by
conducting the heat transfer analysis. The second part of chapter 4 deals with the characterization
of the MEMS-based DSC. After the thermistor being calibrated, it is used for thermal response
study. This quantitative calibration leads to the device’s major parameters: time constant,
sensitivity and power resolution. These results demonstrated the calorimeter to be suited and
capable of ultralow power bio-sensing. In the last part, the cracks in the micro heater trace are
studied.
In chapter 5, the protein stability tests based on the MEMS-based DSC will be presented. The
chapter introduces the data analysis method used in protein sample DSC data analysis. Firstly, the
performance of the MEMS-based DSC is verified by comparing the lysozyme test results using
both the MEMS-based DSC and the TA discovery DSC. Then the thermal stability of some mAb
samples is studied. The influence of the response time of the DSC and the scanning rate to the
kinetics of the DSC is then studied. In the last section of Chapter 5, a systematic study of the
thermal stability of a Fab, the mAb and the corresponding DVD Ig is conducted to study the effect
of the structure to the thermal stability of proteins.
This dissertation concludes with a summary and a discussion of future work in Chapter 6.
26
Chapter 2. Design and Fabrication of the MEMS DSC
2.1 Introduction
2.1.1 Thermodynamics in DSC Test
DSC measures the difference in thermal power between a sample and a reference material, as a
function of temperature. When the sample and reference materials are subjected to identical
temperature scanning rate, their temperatures are varied at a defined rate within a range of interest.
The thermally induced activity of the sample molecules, which is either exothermic or endothermic,
causes a small temperature difference between the sample and reference materials. This differential
temperature can be detected to reflect the differential power. The MEMS DSC device consists of
two identical PDMS microfluidic chambers, which hold the sample and reference materials for
calorimetric measurements. These chambers, referred to as the sample and reference chambers,
respectively, are each connected to inlet and outlet ports through microfluidic channels and
situated on a freestanding polyimide substrate. The diaphragms, along with air cavities
surrounding the chambers, provide thermal isolation that enables sensitive calorimetric
measurements. The device is composed of multiple material layers, including a layer of thin film
VOx based thermistor embedded located underneath the centers of the calorimetric chambers to
measure the temperature difference between the sample and reference solutions. Each calorimetric
chamber is also integrated with a thin-film resistive micro-temperature heater, in a separate
material layer underneath the chamber center. These heaters provide heating to the calorimetric
chambers to generate a constant differential power for calorimetric calibration. VOx based
thermistors are chosen for differential temperature sensing due to their high TCR and ease of
fabrication. Polymers are used to construct the calorimetric chambers to facilitate thermal isolation.
In particular, polyimide is chosen as the diaphragm material due to its excellent mechanical
27
stiffness (Young’s modulus: 2.5 GPa), thermal stability (glass transition temperature: 285 °C) and
low thermal conductivity (0.12W/mK), while PDMS and Flexdym are used to fabricate the
calorimetric chambers for its ease of fabrication and packaging, low thermal conductivity
(0.15W/mK for PDMS and 0.27 W/mK for Flexdym) as well as biocompatibility.
Differential scanning calorimetry can characterize the complete thermodynamic profile for both
intermolecular binding and intramolecular folding by measuring the enthalpy ΔH, specific heat
ΔCp, and the melting temperature Tm. The thermal model is shown in Fig. 2-1. In Figure 2-1b, T0
is the temperature of the environment. Cs and Cr are the heat capacity of the sample and reference
material respectively (In the real test, it will also include part of the heat capacity of the material
holder). Ps and Pr are the heat flux to the sample and reference respectively to achieve linear
temperature scanning. Gs is the heat conduction between the sample cell and the environment while
Gr is the heat conduction between the reference cell and the environment. G0 is the heat conduction
between the sample cell and reference cell.
Figure 2-1. The DSC model. (a) The schematic diagram for DSC which consists of sample and
reference cells. The cells are located in a well-designed test chamber. (b) The thermal model for
the DSC.
(a) (b)
28
Controlled heat flux is added both to the reference and sample material to achieve linear
temperature scanning. Part of the heat flux is absorbed by the tested material, the other is lost to
the surrounding air by convection, conduction, and radiation. Most of the heat loss is through heat
convection in a typical DSC test. The thermal equation for heating single unit is expressed in Eq.
2-1:
C𝑑𝑇
𝑑𝑡+ 𝐺(𝑇 − 𝑇0) = 𝑃 (2 − 1)
C is the heat capacity of the heating sample, G is the thermal conductance, which can be expressed
by ℎ ∙ 𝑠 where h is the coefficient of thermal convection and s is the effective area of thermal
convection since convection is the main heat transfer mechanism in a DSC test. T is the real time
temperature of the sample and T0 is the temperature of the environment. Generally, T0 can be
regarded asconstant and the system is first order system. For MEMS DSC, the sample
consumption is from nano-litter to micro litter which means the heat capacity would be quite small.
To achieve high sensitivity, it is essential to increase the thermal insulation of the system which
means minimize the thermal conductance G. In general, Go is small enough to be neglected then
we can have Eq. 2-2:
(𝐶𝑠 − 𝐶𝑟)𝑑𝑇
𝑑𝑡+ (𝐺𝑠 − 𝐺𝑟)(𝑇 − 𝑇0) = 𝑃𝑑𝑖𝑓𝑓 (2 − 2)
Pdiff is the differential heat flux between the sample and reference material which is recorded
directly during the scanning test. In Eq. 2-2, (𝐺𝑠 − 𝐺𝑟)(𝑇 − 𝑇0) is a linear part and can be removed
simply by baseline subtraction. dT/dt is the scanning rate which is fixed during the linear
temperature scanning process. A temperature domain differential heat capacity between the
29
biomolecular sample and reference can be evaluate by dividing the differential heat flux Pdiff by
the scanning rate after baseline subtraction shown in eq. 1-3 ( is 𝑑𝑇
𝑑𝑡 represents the temperature
scanning rate). To obtain the heat capacity of the sample, precise knowledge of the reference
material (chemical buffer for biomolecular test) is needed for the correction. The differential power
Pdiff is usually small. In MEMS based DSC, the sample consumption is from nano litter to micro
litter, higher concentration is needed to get detectable Pdiff compared to the current commercial
DSC.
∆𝐶𝑝 =𝑃𝑑𝑖𝑓𝑓
(2 − 3)
The more thermodynamic parameter of the biomolecule during the reaction such as denaturation
and binding in the scanning process can be derived from the heat capacity change. Among all the
thermodynamic parameters, the Gibbs free energy change ∆G is the key parameter since its value
under particular reactant concentrations dictates the direction of biomolecular equilibria, and it is
the balance between enthalpy and entropy. It is temperature depended and can be described by Eq.
2-3.
∆G = ∆H(𝑇𝑐) + ∫∆𝐶 𝑑𝑇 − 𝑇∆𝑆(𝑇𝑐) − 𝑇 ∫∆𝐶𝑑(𝑙𝑛𝑇) (2 − 4)
∆H is the enthalpy change under constant temperature 𝑇𝑐. This can happen during phase change.
∆𝑆 is the entropy change and ∆𝐶 is the heat capacity change(𝐶𝑠 − 𝐶𝑟) . If ∆𝐶 is temperature
independent during the temperature range, Eq. 2-4 can be simplified to Eq. 2-5.
∆G = ∆H(𝑇𝑐) − 𝑇∆𝑆(𝑇𝑐) + ∆𝐶(𝑇 − 𝑇𝑐 − 𝑇𝑙𝑛 (𝑇
𝑇𝑐)) (2 − 5)
30
If ∆G is negative, the reaction or transition will proceed spontaneously to an extent governed by
the magnitude of ∆G. Otherwise, ∆G specifies the energy needed to drive the reaction to happen.
The enthalpy change ΔH which is described in Eq. 2-6 reflects the energy needed to transfer the
material to a new state. T is the absolute temperature in Kelvin. The entropy can be simply
described as an index to show how easily the energy might be distributed among various energy
levels. For protein binding reactions, the enthalpy is usually negative which reflects the tendency
for the system to fall to lower energy levels by bond formation. In a DSC test, the enthalpy change
∆H can be directly obtained from the DSC curve shown in Fig. 2-2, then the Gibbs free energy
change is derived from Eq. 2-5.
∆H = ∆H(𝑇𝑐) + ∆𝐶(𝑇 − 𝑇𝑐) (2 − 6)
Figure 2-2. DSC thermogram for the hen lysozyme denaturation process.
31
2.1.2 Temperature and Power Sensing Principle
The MEMS based DSC consists of two thermal equilibration areas for sample and reference
liquids. The thermistors, compensation heater, and electric traces are fabricated on a free-standing
polymer membrane. This free-standing polymer membrane significantly decreases the heat
transfer from the sample or reference region to other area and reduces the thermal mass of the
whole measurement region. Because of the minimized thermal mass and the improved thermal
isolation, the device can have a sufficiently fast time response to allow real-time measurement.
Two high sensitivity micro-thermistors are formed on the membrane in each thermal equilibration
areas for temperature sensing. The feedback heater (platinum) is located in the sample area for
power compensation. The temperature difference between the sample and reference regions is
detected by four thermistors on the membrane. The thermistors are Vanadium Oxide thin trace
with high-temperature coefficient resistance (TCR). The temperature difference induced by
molecular interactions will be sensed by the resistance change of the thermistor, which will be
picked up by Wheatstone bridge. In the Wheatstone bridge, there are two variable resistors and
two thermistors to balance the circuit as is shown in Fig. 2-3. This design is to prevent unequal
resistance of thermistor during fabrication.
32
Figure 2-3. The schematic diagram of the Wheatstone bridge consists of two thermistors (R1 and
R2) and two external decade resistor boxes (R2 and R4) and the lock-in amplifier for differential
detection.
In an ideal case (R1=R2= R3=R4), the output voltage is
𝑉𝑜𝑢𝑡 =1
4
∆𝑅
𝑅𝑉𝑖𝑛 =
1
4𝑉𝑖𝑛𝛼∆𝑇 (2 − 7)
𝑉𝑜𝑢𝑡 is the output voltage of the Wheatstone bridge, inV is the input voltage of the bridge, T is
the temperature difference between the sample and reference area, is TCR of the thermistor. The
bridge used here is mainly for common mode rejection in the sample and reference regions, such
as room temperature fluctuations. The model for the MEMS DSC is a first-order system, the
temperature difference ∆𝑇 is proportional to the differential power Pdiff.
∆𝑇 = ℎ𝑃𝑑𝑖𝑓𝑓 (2 − 8)
ℎ is the thermal resistivity of the MEMS based DSC system which indicates the thermal insulation
quality of the system. Combine eq. 2-7 and 2-8:
33
𝑉𝑜𝑢𝑡 =1
4𝑉𝑖𝑛𝛼ℎ𝑃𝑑𝑖𝑓𝑓 = 𝑆𝑃𝑑𝑖𝑓𝑓 (2 − 9)
In Eq. 2-9, S is called the power sensitivity of the DSC, it is determined by the input voltage of the
Wheatstone bridge, the TCR of the thermistor and the thermal insulation of the whole system.
2.2 Micro Heater Design
2.2.1 Introduction of the Micro Heater Design
Micro heaters consist of a heating element and substrate, and have been applied widely for
temperature control in MEMS sensors, especially in bio-calorimeters [96], gas sensors [97],
microfluidic actuators [98] and IR applications [99]. There has a lot of work investigating different
materials used as the heating source such as polysilicon [100], SiC [101], TiN [102], Pt [103], and
Au [104]. When a voltage is applied to the terminals of the heater, which is often made of resistor
materials, it will generate a certain amount of Joule heat. A thin film insulator such as SiO2, Si3N4
or polymer membrane is usually used as the substrate for the heater. To reach high thermal
insulation, the thin film substrate is usually suspended using several beams for support. Since
MEMS fabrication is usually carried out on silicon wafers, which require a lot of etching work, it
is time-consuming and costly [105]. Thin films like SiO2 and Si3N4 can only sustain small loads
due to their poor mechanical strength, so the application of the heater based on these substrates is
limited to some gas sensors and IR devices [106-108]. Recently, however, polymer membranes,
especially polyimide, have gained a lot of attention [109-111] due to their robustness, low thermal
conductivity, and simple fabrication process. Also, the thermal expansion coefficient of the
polyimide membrane is quite close to that of a silicon wafer, which makes it quite comparable
34
with the IC fabrication process. Some companies such as Minco and Omega already have similar
products by employing a meandering resistive track on polyimide to form a heating element.
Compared to the traditional hotplate, a MEMS based micro heater needs much lower driven
power/voltage and costs much less. Due to the small size and simple structure, a micro heater is
much faster in various types of thermal response. Recently, many papers reported their work in
developing novel MEMS heaters which can achieve high heating temperatures and short response
times. However, while most of them mention the temperature distribution in the heating area, only
a few works report about how to improve the temperature uniformity [112]. Good heating
uniformity can increase the selectivity and sensitivity of the device which is potentially important
in ultrasensitive NDIR gas sensing and micro calorimetry. For an NDIR gas sensor, the emissivity
is both temperature and wavelength sensitive [113] while calorimetry measurements are direct
temperature related [114]. Previous research results show that uniform heating power density will
lead to non-uniform temperature distribution [115]. By thorough analysis of the heat transfer on
the surface of the heater and carefully distributing the heating power, the heating area can reach
high-temperature uniformity.
In this section a novel micro heater on a polyimide membrane with high-temperature uniformity
and short thermal response time. A novel process is introduced to fabricate the flexible substrate
using liquid polyimide. A gold resistor is utilized to release Joule heat. For the heater design, a
numerical method is developed to analyze the relationship between the heating power distribution
and the temperature uniformity. The optimization of temperature uniformity in the heating area is
completed using an ANSYS thermal MEMS simulation. A prototype is then fabricated based on
the simulation and tested. The measurement and simulation results are consistent. The time
35
constant for the step response is also measured when two µL water droplet is loaded on the
substrate.
2.2.2 Micro Heater Design Principle
The micro heater is designed for the MEMS based DSC calibration. The goal of the design is to
achieve high-temperature uniformity. The micro heater is composed of a resistive Au/Ti trace on
a polyimide membrane. When a certain amount of heating power is applied to the heater, the
temperature is quickly redistributed over the suspended polyimide membrane. The response time
is on the milli-second scale due to the small thermal mass of the suspended polyimide membrane.
The heat will ultimately be transferred through conduction to the periphery and convection with
the air will occur along with some thermal radiation. When the temperature in the heating area is
less than 500K, the heat loss through radiation can be neglected compared to the total heating
power supply [116]. Due to the low thermal conductivity of the polyimide membrane, the
temperature attenuates rapidly in the non-heating area. When stable, most of the heat will be
transferred to the air from the heating area through heat convection directly. Natural heat
convection is caused by buoyancy forces due to the density difference caused by temperature
variations in the air. When the sample area is heated, the fluid near the boundary is warmed and
becomes lighter. The buoyancy causes the fluid to rise and be replaced by a cooler fluid that will,
in turn, be heated continuing the process [117]. When the polyimide-based micro heater is placed
horizontally, the coefficient of heat convection for the upper surface is different from that for the
lower surface. The coefficient of heat convection depends on the gas which serves as the cooling
fluid, the surface temperature of the sample and the geometrical configuration of the heater. The
temperature distribution in the heating area is greatly affected by the contour of the heater and the
36
distribution of the heating trace. The temperature distribution can be expressed by the Poisson
equation [118] as follows:
−∇(𝑘𝑡 ∙ ∇𝑇) = 𝜎 + ℎ(𝑇𝑎𝑖𝑟 − 𝑇) (2 − 10)
ℎ =0.54𝑘𝑅𝑎𝐿
0.25
𝐿+ 0.27𝑘𝑅𝑎𝐿
0.25
𝐿=0.81𝑘𝑅𝑎𝐿
0.25
𝐿 (2 − 11)
𝜎 =𝑑𝑃
𝑑𝑆 (2 − 12)
where: 𝑘 is the thermal conductivity, and 𝑡 is the thickness of the membrane, 𝜎 is the heating
power density in the heating area, ℎ is the total coefficient of heat convection for the upward and
downward surfaces, 𝐿 is the characteristic length of the heater, specified by the area over the
perimeter, 𝑇𝑎𝑖𝑟 is the temperature of the air which is assumed constant, and T is the local
temperature in the membrane. Since the heater consists of a series of concentric gold traces, and
the polyimide membrane is much larger than the heater itself, the temperature distribution in the
plane of the membrane is isotropic, so a polar coordinate system is used to express Eq. 2-10:
𝑘𝑡
𝑟
𝑑
𝑑𝑟(𝑟
𝑑𝑇
𝑑𝑟) + 𝜎 + ℎ(𝑇𝑎𝑖𝑟 − 𝑇) = 0 (2 − 13)
For the horizontally placed micro heater, ℎ is around 7.8W/m2K[119]. Based on Eq. 2-13, a
numeric method is carried out to calculate the temperature distribution for the given normalized
heating power density distribution.
The gold trace of the heater is a double spiral which can be simplified by a series of concentric
circles. In the FEM model and the mask design, the circle is replaced by an octagon since it is
much easier to design and fabricate an octagon mask, but its performance is similar. The
37
commercial software ANSYS 15 is used as the finite element modeling and simulation tool. A
thermal-electric coupled field analysis is performed to observe the heat release and heat
distribution on the micro-heater as current is applied. The properties of the materials are listed in
Table 2.
In the FEA analysis, the thermal conductivity of the gold/titanium is also considered. The width
of the space between the adjacent trace 𝑊2 is fixed at 20𝜇𝑚. The loading of the heating power is
based on Joule’s law expressed by equation (5). The width of the circle 𝑊1 varies from the center
to the edge to change the heating power density determined by eq. 2-14, 15, 16 and 17.
𝑃 = 𝐼2𝑅 (2 − 14)
𝑅 = 𝜌𝑙
𝑡1𝑊1 (2 − 15)
𝑆 ≈ 𝑙(𝑊1 +𝑊2) (2 − 16)
𝜎 = 𝐼2𝜌1
𝑡1𝑊1𝑊∗ (2 − 17)
In Eq. 2-14, P is the heating power, I is the current, and R is the resistance of each circle gold trace.
Eq. 2-15 shows the resistance R is determined by several parameters: the resistivity of the gold
heater, the length of the circle l, the width of the circle W1, and the thickness of the thin film t1.
The heating power density is expressed in Eq. 2-17, it’s the derivative of the power over the heating
area. The heating area includes the gold circle and the space between them is expressed in Eq. 2-
16. 𝑊∗ is roughly the sum of 𝑊1 and 𝑊2. Since 𝑊2 is fixed, the only parameter that determines
the power density in the heating area is 𝑊1. By gradually reducing the width of the gold circle
38
𝑊1from the center of the heater to the edge, the power density is increased accordingly. Fig. 2-4
shows the simplification process of the micro-heater model.
(a) (b) (c)
Figure 2-4. (a) The numerical model of the micro heater in the FEA analysis. (b) The design for
the micro heater. (c) The micro structure of the heater, the width of each circle is W1, the space
between them is W2.
Table 2. The material properties of the heater in the FEA analysis.
electrical
resistivity (Ω ∙ 𝑚)
Thermal conductivity
(𝑊/𝑚𝐾)
Density
(𝑘𝑔/𝑚3)
Specific heat
(𝐽/𝑘𝑔 ∙ 𝐾)
Polyimide − 0.12 1420 1090
Gold 2.4 × 10−8 314 19800 1290
Titanium 4.2 × 10−7 20.8 4506 540
In the steady-state thermal FEA analysis, the side face of the polyimide membrane is set to be the
ambient temperature 𝑇𝑎𝑖𝑟 which is 20. Another equivalent boundary condition is the boundary
39
temperature gradient is zero. This setting can be explained by the sharp temperature attenuation in
the non-heating area of the membrane. The size of the polyimide membrane is 6 × 6𝑚𝑚2 while
the radius for the outer circle is 1.2mm. A larger polyimide substrate is unnecessary since it’s
enough for the temperature to attenuate to ambient at 3mm from the center. Solid 90 is used as the
meshing element, and the free meshing method with local refinement in the heating area is used
for the meshing process. Fig. 2-5 shows the thermal characteristics of the polyimide heater. The
total heating power is 6.28 mW. From the simulation results, the temperature stays uniformly high
in the central heating area and attenuates rapidly to the room temperature in the non-heating area.
For this double spiral shaped heater, how to determine the width of each circle of the heater is the
key factor to reach high temperature uniformity. With the command file of ANSYS 15, it is easy
to change the parameters and compare the results. The result shows how the change of the width
of the circle from the center to the edge of the heater greatly affect the uniformity of the
temperature which is shown in Fig. 2-6. From design 1 to design 3, we try to optimize the heater
to reach better temperature uniformity in the heating area by adjusting the width of each circle. In
design 3, while the average temperature is 68.7 , the largest temperature difference is less than
0.3 in the area where the radius is less than 1mm. Considering the radius of the whole heating
area is 1.2mm, this result is quite satisfying.
40
Figure 2-5. The temperature distribution in the polyimide membrane by FEA simulation.
Figure 2-6. Temperature distribution along the diameter of the heater for different heater design.
2.2.3 Characterization of the Micro Heater Design
The fabrication starts with the preparation of the silicon chip, which is then cleaned by acetone,
IPA, and water in an ultrasonic cleaner. After cleaning, the silicon wafer is baked on a hotplate
41
with a surface temperature of 160 for fully drying. Liquid polyimide PI2611 is used to fabricate
the flexible polyimide substrate with a thickness of 7 µm. After thawing at room temperature for
30 min, the liquid polyimide sample is then put in a vacuum chamber for degassing. After
degassing, the liquid polyimide is spin-coated on the silicon carrier. After the spin-coating, the thin
film sample is soft-baked on the hotplate under 100 for 2 min, and then the baking temperature
is slowly ramped to 350 in about 2 hours. After fully cured, the sample is gradually cooled down
to the room temperature and then removed from the hotplate. After the patterning of the positive
photoresist on the polyimide substrate by photolithography, a 10 nm layer of titanium is deposited
on the polyimide substrate as an adhesive layer and then followed by the deposition of a 100 nm
layer of gold by e-beam deposition. After a 2 hours lift off, the gold heater is patterned on the
polyimide substrate. After being coated with another protective polyimide layer, the sample is then
peeled off from the silicon carrier easily. Fig. 2-7 shows the sample of the peeled off heater array.
Conventionally, people use commercial polyimide as the substrate by attaching the polyimide
membrane onto the silicon wafer with some adhesion layer or using a relatively thick PI substrate
(100µm) without silicon carrier. However, ultrathin polyimide is needed to reach high thermal
isolation. For a polyimide thin film less than ten 𝜇m thick, it would be difficult to mount on the
silicon wafer without leaving any bubbles inside. Besides, the introduction of the adhesion layer
would also cause difficulty for the separation of the polyimide thin film and the silicon carrier in
the later step.
42
Figure 2-7. The peeled off micro heater array.
The characterization of the gold heater is carried out in the chamber of a high temperature probe
station (AO 600 Compact Rapid Thermal Annealing System, MBE). A small constant current
(10µA) is added to the heater to measure its resistance at a different setting temperature in the
chamber. The high-temperature probe station is integrated with a test chamber with excellent
thermal isolation, a PID temperature controller, a quartz beam heater, and an I/O system, as shown
in Fig.18. The resistance of the gold heater and the ambient temperature share a linear relationship
and is expressed in Eq.18.
Figure 2-8. High temperature probe station for the characterization.
R = 𝑅0 + 𝛼𝑅0(𝑇 − 𝑇0) (2 − 18)
43
Figure 2-9. R-T characterization of the micro heater
Fig. 2-9 shows the characterization of the temperature coefficient of the resistance for the gold
heater which is treated several times in an annealing process. In the measurement range, the
resistance changes linearly with the rise of temperature. The temperature coefficient of gold is
0.09% in the temperature range of 40 to 90 which is significantly smaller than that of the bulk
gold which is possibly due to the porosity of the evaporated thin film.
A sophisticated feedback control system was applied to achieve a programmable constant power
supply for the heater. A thermal image camera was used to measure the temperature distribution
of the heater as shown in Fig. 2-10a and b. The temperature of thousands of points was measured
by the camera, and each was used to draw the map of the temperature distribution. Fig. 2-10c
shows the temperature of the heating area has about 0.45 in variance and 53.7 average when
6.28 mW heating power was loaded. The average temperature is significantly smaller than the
simulation result (53.7 compared to 68.7 ) which may be caused by the difference between
the real heat convection coefficient and the input simulation parameter. The value of the
44
convection coefficient is greatly affected by the forced convection and will increase when the
temperature of the heating area increases. The temperature variance is consistent with the
simulation result and is much smaller than the previous result.
Figure 2-10. (a) The gold heater image under the microscope. (b) The IR image of the gold heater
under 6.28 mW heating power. (c) The temperature distribution in the heating area measured by
the IR camera.
45
Figure 2-11. The measured heater temperature under different heating power consumption in the
atmosphere.
Fig. 2-11 shows the temperature of the heating area as a function of the heating power. The
temperature increase rate is 5.6/mW when the temperature of the heating area is below 100 .
This rate indicates the amount of heat loss which is related to the thermal isolation of the heater.
The temperature increase rate could be larger if the chamber is under vacuum.
Figure 2-12. The step response of 8mW heating power for the micro heater when there is 2𝜇𝐿
water on the top.
The micro heater could be used for various types of heating processes. The response time is
determined by the thermal mass of the load C and the equivalent heat conduction coefficient H.
The device could be regarded as a 1st order linear system which is expressed by Eq. 2-1. Fig.2-12
shows when there is two 𝜇𝐿 liquid water loaded on the micro heater, the average temperature could
46
rise up by 20 under 8mW heating power. The time constant for this first order heating system is
4.3s.
In this section, we have designed and fabricated a novel gold heater on a polyimide substrate with
high-temperature uniformity in the heating area and short response time. The double spiral heater
is a series of concentric octagons. By FEA simulation and optimization, the heater achieves high-
temperature uniformity in the heating area which makes it possible for use in the application of
gas and other types of sensors. The temperature variance is only 0.45 when the heating area
reaches 53.7 in average with 6.28mW power supply. By introducing the polyimide as the
substrate, the micro heater has high thermal insulation; the power consumption is relatively low to
drive the heater to a certain temperature with heating coefficient 5.6/mW. The step response
time for two µL water droplet is only 4.3s, shows its potential in the application in calorimetry and
micro actuators.
2.3 Thermistor Design
Temperature sensing is the most important part of the MEMS based DSC. Ideally, the sensors
should have high-temperature sensitivity and low intrinsic noise. Thermal couples have been used
in calorimetric biosensors [120], but the thermal Johnson noise is large because of large resistance.
Resistance temperature detectors (RTDs) are usually made of metal materials and have low noise
level, but the temperature sensitivity is small. Thermistors, such as amorphous germanium or
silicon [121], silicon-germanium-boron alloy [122] and amorphous YCBO [123] have excellent
temperature sensitivity but the noises are also large due to their high resistivity. The sensing noise
includes thermal noise (Johnson noise) and flicker noise (1/f noise), which is related to material
conductivity and measurement frequency bandwidth. Some research results show that the wide
47
band gap materials such as silicon carbide and diamond have high TCR and moderate resistivity
values [124]. Furthermore, the wide bandgap semiconductors are very stable in the mechanical,
thermal, chemical and biochemical properties, which make them suitable for the thermistor
applications [125].
Besides optimization of fabrication conditions, we propose a novel pattern design of thermistor
with thermal noise two orders of magnitude lower than traditional design. The concept is
illustrated in Fig. 2-13. Fig. 2-13a shows the normal heater design, in which the current flows from
one end to the other. Fig. 2-13b shows the concept for the thermistor design where the current
flows from one side to the other of the thermistor material. In this new design we sandwich
thermistor trace with electrical conductor trace (gold trace). Therefore the electrical current flows
in the side direction, and thus the whole resistance decreases greatly. Since the resistivity of the
gold trace (2.2 × 10−6Ω𝑐𝑚)is much smaller than that of the thermistor material (20Ω𝑐𝑚). We
can directly deposit the gold trace on the top of the thermistor layer without additional work on
the thermistor patterning.
48
Figure 2-13. Thermistor design. (a) normal resistor design. (b) side to side resistor design. (c)
parallel thermistor design
In the saw structure thermistor design, each main trace has 15 fins, the current I flows from the
main trace to the fins and then to the fins of the other main trace through the 𝑉𝑂𝑥 thin film. Ignore
the marginal, the current distribution on each fin is identical (2𝑖 =𝐼
15).The voltage (𝑈) different
between the two main trace is:
𝑈 = 𝑖𝜌𝑑
𝑡𝐿=𝐼𝜌𝑑
30𝑡𝐿 (2 − 19)
𝑅 =𝑈
𝐼= 𝜌
𝑑
30𝑡𝐿 (2 − 20)
𝑅 is the resistance of the thermistor, 𝜌 the resistivity. 𝑑 the distance between two adjacent fins
(40µm) , 𝑡 the thickness of the VOx thin film (100nm), 𝐿 the effective length of the fin (1.2mm).
The calculated resistance of the thermistor is 2.22kΩ which is slightly smaller than the measured
value (2.3kΩ). For the normal resistor design shown in Fig. 23a, the resistance is 4 orders in
magnitude larger which can reach to several Megaohms. It can cause huge Johnson noise (𝑣𝑛)
expressed in Eq. 2-21:
𝑣𝑛 = √4𝑘𝐵𝑇𝑅∆𝑓 ( 2 - 2 1 )
𝑘𝐵 is the Boltzmann constant, 𝑇 the temperature, ∆𝑓 the bandwidth.
2.4 Microfluidic Chamber Design
2.4.1 Introduction of Microfluidic Device
Microfluidic structures all need sealing to form hollow embodiment. PDMS
(Polydimethylsiloxane) has been widely used in microfluidic systems due to its optical
49
transparency, biocompatibility [126], and more importantly, it can be easily bonded to glass or
itself after oxygen plasma treatment to formulate the microchannel [127, 128]. In recent years,
growing interests and efforts were made towards flexible plastic-based microfluidic devices.
Various plastic materials were used, such as polyimide (PI), polystyrene (PS) [129], polyethylene
terephthalate (PET)[130], polycarbonate (PC) [131]and polymethylmethacrylate (PMMA)[132].
These polymers have various attractive material properties, e.g., low mass, high chemical
resistance, low thermal conductance, good optical properties, and good biocompatibility, which
make them potential candidate materials for electrochemical lab-on-a-chip, bio-MEMS (Micro-
electromechanical systems) calorimeter, and biosensors applications. However, the heterogeneous
bonding of PDMS with plastic material can be challenging.
Previously reported bonding techniques can be divided by direct bonding and indirect bonding.
Thermal bonding[133], laser bonding [134] were the common seen direct bonding methods. They
might also need oxygen plasma or UV ozone treatment [135] for surface modification. Direct
bonding often requires high temperature or voltage and thus might introduce unwanted troubles.
For example, thermal bonding might not be suitable for heterogeneous plastic bonding since
channel deformation might occur during heating.
Meanwhile, indirect bonding, which required an intermediate layer to bond two incompatible
surfaces, could circumvent the problems. These adhesives are sometimes based on chemical
effects like polymerization (e.g. acrylics), polycondensation (e.g. silicones) or polyaddition (e.g.
epoxies) and can often be cured at elevated temperatures or using UV light. Recently, (3-
Aminopropyl)triethoxysilane (APTES), (3-Mercaptopropyl)trimethoxysilane (MPS) and low-
50
temperature co-fired ceramics (LTCC) have also been demonstrated effective to bond PDMS with
plastic material.
While adhesive bonding enabled high strength bonding at much lower temperature, adhesive
bonding between plastic materials also confronted several challenges. The uncured adhesive might
flow into the microfluidic channel and cause the clogging. Minimizing the contact area at the
interface between the adhesive and the fluid could also be crucial when labile biomolecule or cell
culturing was involved. As a result, stamp-and-stick, a selective bonding method, was used to
address the problem. It provided a simple method to meet the above constraint by transferring an
intermediate adhesive layer to other surfaces using a patterned stamp [136]. One challenge was the
ununiformed topography of the thin flexible plastic substrate [137]. Direct laminating a thin layer
of flexible plastic substrate on a rigid substrate could easily introduce unwanted bubbles, and such
uneven surface might cause troubles for good sealing. However, few papers reported solutions to
this problem. Another challenge during the stamp-and-stick process was the adhesive might not
spread out on the transferring substrate, especially when the substrate is not hydrophilic enough.
This was because when a substrate had low surface energy compared to the adhesive, then the
adhesive would be attracted to itself rather than to the substrate.
2.4.2 Design and Fabrication of the Microfluidic Chamber
The microfluidic chamber with inlets and outlets is used to deliver and hold the bio-samples for
thermodynamic characterization. To ensure better thermal isolation and easy fabrication process,
we proposed a double layer microfluidic chamber design using Flexdym (a soft thermoplastic
elastomer) as the bottom layer and polydimethylsiloxane (PDMS) as the top layer shown in Fig.
24. Both the reference and sample chambers are surrounded by air gaps to reduce the effective
51
heat capacity in the measurement process and increase the thermal insulation property at the same
time. To date, a preferred and well-established method is the soft lithography (SL) of
polydimethylsiloxane (PDMS). The fabrication of microfluidic devices based on PDMS is an easy
and robust approach, and the simplicity of PDMS manipulation has been key to its success. Indeed,
with minimal training and equipment, the microfabrication of PDMS chips can be performed by
non-experts without stringent methodological constraints. However, there are three major
properties of PDMS that have three specific negative impacts: (1) Channel deformation due to its
high mechanical compliance. (2) Evaporation, sorption and gas permeability. (3) The bonding of
PDMS on polyimide needs further adhesive layer assistance. Gas permeability property of the
microfluidic device is critical when the operation temperature is significantly higher than the room
temperature. Compared to PDMS, Flexdym has much lower gas permeability, and the bonding
process is much simpler. Thus, the Flexdym is used as the bottom layer to prevent evaporation
while PDMS as the top layer to provide a top air gap and serves as the frame for liquid sample
delivery.
(a) (b)
52
Figure 2-14. The design of the microfluidic device. (a) the top layer consists of inlets, outlets and
a chamber to form the top airgap. (b) the assembled microfluidic device consists of a top layer
and bottom layer.
The bottom layer is made of Flexdym and the top layer in made of PDMS. The chamber size is
1.4𝑚𝑚 × 1.8𝑚𝑚 × 0.25𝑚𝑚 with a total volume 0.63𝜇𝐿. To evaluate the effectiveness of the
double layer design in increasing the thermal resistance of the DSC system, a set of FEM
simulations are carried out. The property of related materials is listed in Table 3.
Table 3. The properties of the materials in the FEA simulation.
thermal conductivity
𝑊/(𝑚𝐾)
specific heat
𝐽/(𝑔𝐾)
density
𝑔/𝑐𝑚3
Polyimide 0.12 1.09 1.42
Air 0.026 1.01 0.0012
PDMS 0.15 1.46 0.965
water 0.591 4.186 1
We consider three types of the microfluidic chambers: a chamber with no air gaps, a chamber with
side air gap and chamber with both side and top air gaps shown in Fig. 2-15 (a1), (b1) and (c1)
respectively. The distance between the top air gap and the chamber is 0.5𝑚𝑚, the wall between
the chamber and the side air gap is 0.2𝑚𝑚 and, the width for the side air gap is 0.2𝑚𝑚 in favor
of fabrication. In the simulation, the microfluidic device is bonded to the polyimide substrate
(10µm in thickness). The chamber is filled with water. The influence of the inlet, outlet and the
microchannels to the thermal response of the microchamber is neglected, hence the FEA model
can be simplified. A 5mW power is added to the water to study the thermal response. The
53
calculation time is 40s which is divided to 200 steps. The three models and the corresponding
temperature distribution at the final step is shown in Fig. 2-15.
Figure 2-15. The three types of microchamber design and the corresponding final temperature
distribution for the FEA models.
a1-a2 shows the chamber with no air gaps design and the temperature distribution around the
chamber at 𝑡 = 40𝑠 in simulation , b1- b1 shows the chamber with side air gap design and the
temperature distribution around the chamber at 𝑡 = 40𝑠, c1-c2 shows the chamber with both top
and side air gaps design and the temperature distribution around the chamber at 𝑡 = 40𝑠 in
simulation.
In the simulation, we only consider the thermal response for one chamber since the sample and
reference chambers are identical. We pick the gravity point of the water in the chamber and draw
the thermal response plot for all the three types of chambers shown in Fig. 2-16.
54
Figure 2-16. The thermal response of the gravity point for all the three types of chamber design.
(Design 1 is the chamber with no air gaps, design 2 is the chamber with side air gaps and design 3
is the chamber with both top and side air gaps.)
When it reaches equilibrium, the temperature rise for all the three types of chamber design: is
∆𝑇1 = 7.11, ∆𝑇2 = 7.89, ∆𝑇3 = 13.94 . Since chathe mber is 1st order system, the
temperature rise for a certain amount of power load is: ∆𝑇 =∆𝑃
𝐺. 𝐺 is the thermal conductance of
the system, and 1 𝐺⁄ is the thermal resistance. Comparing ∆𝑇1 , ∆𝑇2 and ∆𝑇3 , the side air gap
increases the thermal resistance of the chamber by 11%, while side air gap and the top air gap
together help to increase the thermal resistance of the chamber by 96%. The simulation results
suggest the double layer chamber design to improve the thermal insulation of the microfluidic
system.
55
2.5 Heating Stage Design
The heating stage design is shown in Fig. 2-17, 18 and 19. The heating stage is constructed to raise
the temperature and provide the thermal shielding. It mainly consisted of a heating sink, a copper
plate, two Peltier heaters, a sample stage, a glass cover, the thermal shielding (made of Teflon
tube), and the metal shielding. The two-layer shielding blocked the outside disturbance and
minimized the temperature fluctuation. The wires were connected outside via a hole on the sample
stage and then soldered to an electrical box with BNC adapters. The heaters were controlled by a
temperature controller TC-720 (TE Tech) and powered by a power supplier. A commercial
thermistor sensor was attached to the sample stage to measure the temperature. The temperature
control process was automated by a LabVIEW program.
Figure 2-17. The 3D model of the heating stage
56
Figure 2-18. The top view of the heating stage.
Figure 2-19. The cross-section view of the heating stage.
The heat source consists of two Peltier heaters connected in series. The heater has two sides, and
when a DC electric current flows through the device, it brings heat from one side to the other, so
that one side gets cooler while the other gets hotter. The "hot" side is attached to a heat sink so that
it remains at ambient temperature, while the cool side goes below room temperature. In some
applications, multiple coolers can be cascaded together for lower temperature. The specifications
of the Peltier heater is listed in Table 4.
57
Table 4. Specifications of the Peltier module
TE module
VT-127-1.4-1.15-71
Material specifications (27oC
at the hot side)
Material specifications (50oC
at the hot side
Vmax (V) 16.1 17.9
Imax (A) 8.0 8.0
Qmax (W) 80.0 87.8
∆𝑇max (oC) 71 80
2.6 Fabrication of the MEMS Based DSC
2.6.1 Fabrication of the Polyimide Thin Film
Polymer materials have superior characteristics such as low thermal conductivity, high-
temperature resistance, a low dielectric constant, good mechanical strength, low cost, inertia to
chemicals and dimensional stability. Recently, numerous MEMS devices have been realized on a
flexible substrate, exploiting these properties. These devices include bolometers, capacitive
accelerometers, temperature sensors, and flexible, transparent transistors. The fabrication of
bendable, light-weight electronics on flexible substrates needs to perform equally to electronics on
rigid substrates like silicon to integrate into MEMS fabrication.
One method is to bond the flexible polymer membrane film on a rigid substrate, and after the
fabrication process, the membrane is detached to form a flexible substrate. This bonding has to be
firm and adhesive enough so that no liquid chemicals can go into the gap between the polymer and
the rigid substrate during the fabrication. Meanwhile, the bonding force has to be gentle enough
so that later peel off would not require great force, which might damage the metalized traces. This
would make the fabrication process harder. PI2611 (HD Microsystem) is selected to be the raw
material. This is an NMP based liquid polyimide that requires spin coating and post-treatment on
58
the hotplate for solidification. This material has low thermal conductivity (lowest among PI
polyimide family), good dielectric property, large Young's modulus, and relatively higher glass
transition temperature. The coefficient of thermal expansion is almost the lowest among polymers
and is close to the silicon wafer. In this way, the post heat treatment of the thermistor would not
introduce bubbles or stress at the low heating rate.
The fabrication starts with the preparation of the silicon chip, which is then cleaned by acetone,
IPA, and water in an ultrasonic cleaner. After cleaning, the silicon wafer is baked on a hotplate
with a surface temperature of 160 for fully drying. Liquid polyimide PI2611 is used to fabricate
the flexible polyimide substrate with a thickness of 7 µm. After thawing at room temperature for
30 min, the liquid polyimide sample is then put in a vacuum chamber for degassing. After
degassing, the liquid polyimide is spin-coated on the silicon carrier. After the spin-coating, the thin
film sample is soft-baked on the hotplate under 100 for 2 min, and then the baking temperature
is slowly ramped to 350 in about 2 hours. After fully cured, the sample is gradually cooled down
to the room temperature and then removed from the hotplate. The fabricated polyimide thin film
is shown in Fig. 2-20.
(a) (b)
59
Figure 2-20. The polyimide sample. (a) Polyimide on silicon wafer carrier after curing. (b) Flexible
polyimide thin film after peeling off from the silicon wafer.
2.6.2 Micro Heater and Thermistor Fabrication
The fabrication of the micro heater fabrication starts from patterning a 100nm Au/Ti layer on the
polyimide substrate. First, a positive photoresist S1811 was spin coated and baked on the
polyimide substrate and then exposed under the UV light with the self-designed micro heater
photomask. The photoresist was then developed in the S1811 developer. After cleaned and dried,
a 100nm Au/Ti layer was deposited using ebeam deposition. In the lift-off process, the whole
device was put into the PG1165 remover to remove the extra part of the Au/Ti layer to form the
micro heater.
After the fabrication of the micro heater layer, another polyimide layer was prepared on the top
for electric insulation. The thermistor fabrication starts from the deposition of VOx thin film
performed in a Kurt Lesker PVD 75 magnetron sputtering equipment. DC power was applied on
a 3-inch diameter VOx target in a mixture of argon and oxygen atmosphere. Polyimide thin film
was coated on the top Si substrates first, and then the VOx was deposited on the polyimide layer.
Before the sputtering process, the vacuum chamber was evacuated to the middle of 5×10−6 Torr.
The properties of the thermistor thin films depended on the oxygen flow rate, the Argon pressure,
and the power. Then a positive photoresist S1811 layer was patterned on the top of the VOx thin
film using standard photolithography process. Then a 100nm Au/Ti layer was deposited on the top
using ebeam deposition and then patterned by lift-off to form the electrical trace. Finally, another
polyimide layer was spin-coated and cured on the top to protect the thermistor and micro heater.
The whole fabrication process is shown in Fig. 2-21.
60
Figure 2-21. The whole fabrication process of the micro heater and thermistor.
2.6.3 Microfluidic Device Fabrication
We propose the double layer design for the microfluidic device fabrication to increase the thermal
insulation and bonding quality. First, SU-8 master mold was fabricated on a silicon wafer with a
standard soft lithography process. We chose SU-8-2100 with a final thickness of 250µm after
developing.
Second, a Flexdym thin film (1mm) was prepared from Blackhole Lab. The embossing
experiments were performed with an applied pressure of ∼0.4 bar and an isothermal process
temperature of 120 and 180 °C for 120 s and 30 s, respectively. The typical zero shear viscosity of
the sTPE material is 3–6 orders of magnitude lower than that of hard TP, which explains the
attractive processing conditions such as the short embossing time and low pressure needed for
(a)
(e) (f)
(d) (c)
(b)
61
replication [138]. The Flexdym thin film formed the bottom layer of the microfluidic device after
peeling off from the SU-8 master mold.
Third, the top layer of the microfluidic device was prepared with the PDMS liquid. PDMS
prepolymer (Sylgard 184) and a curing agent were purchased from Dow Corning (Midland, MI,
USA). To protect the master mold from damages during demolding, we placed a bottle containing
ten µL of chlorotrimethylsilane and the master mold together in a vacuum desiccator for 10
minutes, so that a thin layer of chlorotrimethylsilane was coated. After that, the mixed PDMS (10:1
portion) was poured onto the mold and was degassed by the vacuum chamber. After that, the
PDMS was cured at 80 oC for 1 hour. The solidified PDMS was peeled off from the master to form
the air gap.
The final step was to bond the Flexdym based bottom layer and the PDMS based top layer together
by hot pressing. After hole-drilling at the inlet and outlet ports, the microfluidic chamber was
bonded to the micro DSC device by hot pressing at 120oC for 6 min. The microfluidic device is
shown in Fig. 2-22.
62
Figure 2-22. (a) The fabricated micro heater. (b) The thermistor on a polyimide substrate. (c) The
MEMS DSC device with the integrated microfluidic chamber, the micro heater, and thermistor.
(d), (e) and (f) show the micro structure of the micro heater, thermistor and the structure in the
overlap area.
2.6.4 The Heating Stage Fabrication
The fabricated heating stage is shown in Fig. 2-23. The thickness of the sample stage was
optimized to achieve high-temperature scanning rate (45oC/min) and temperature uniformity. The
thermal shielding part consists of the Teflon bush; glass cover was sealed on the sample stage with
O-ring on the V-groove. The temperature sensor (MP-3022) was placed into a drilled hole (shown
in Fig. 22b) to measure the temperature of the sample stage and provide a reference for the linear
temperature scanning control.
(a) (b) (c)
(d) (e) (f)
63
Figure 2-23. (a) The prototype of the heating stage. (b) The heating stage without the thermal
cover, the marked position shows the placement of the temperature sensor (TE Tech MP-3022).
2. 7 Summary
In this chapter, the design and fabrication of the MEMS based DSC for the characterization of
liquid-phase biomolecular samples are presented. First, the working principle of MEMS DSC is
introduced. Based on the governing equations and the working principles, we proposed the novel
design for the micro heater, thermistor, microfluidic device, and the heating stage. Then follows
by the fabrication of the whole MEMS based DSC system. The micro heater on a polyimide
membrane with high-temperature uniformity and short thermal response time is used for the
sensitivity calibration. The branch design of the thermistor enables easy fabrication process and
moderate resistance of the thermistor (~10 kΩ). In the thermistor material fabrication, vanadium
oxide is chosen for its high temperature coefficient of resistance (-2.51% per oC). A novel double
layer microfluidic design is used to increase the thermal insulation of the MEMS DSC system and
reduce the noise. The thermal insulation of the double layer design 1.8 times better than the single
(a) (b)
64
chamber design. The heating stage is powered by a temperature controller and controlled by the
Labview Program. It can achieve high temperature scanning rate with high linearity.
65
Chapter 3 Thermistor Material Study
3. 1 Introduction
Temperature sensors can be classified into contact and noncontact sensors. The contact sensors
include thermocouples and resistance temperature detectors (RTD). The latter detects temperature
changes with varying resistance values. It may be classified as resistance wire RTDs and
thermistors (thermally sensitive resistors). Thermistors can be sensitive, low noise and it provides
absolute temperature. Thermocouples utilized the thermoelectric principle. It does not introduce
external heat during thermal sensing, while thermistors usually would require small current for
sensing.
It is convenient to define the change of resistance when temperature change as the temperature
coefficient of resistance (TCR), which is conventionally denoted by α. At any temperature, TCR
is defined as α(T) = (dR/dT)/R(T). The change in the temperature of the resistor (ΔT) can then be
directly obtained from ΔT = ΔVRTI/ (IsR(T)α), where ΔVRTI is obtained by dividing the voltage
change (ΔVout) at the output of the amplifier by the amplifier gain G. We note that the voltage
changes are measured in a bandwidth Δf. This equation suggests that the equivalent noise
temperature (NET) that represents the temperature resolution (ΔTRes) of a resistance based
thermometer. Important considerations when choosing the thermal sensitive material for the
temperature sensor include a high TCR, low noise and compatibility with IC fabrication. Metals
such as platinum and gold were previously used for temperature sensing due to high reliability and
a linear temperature vs. resistance relationship. This provides convenience in the calibration of the
thermistor yet they typically suffer from low TCR. Semiconductor materials have negative TCRs.
Despite being nonlinear between 46 temperature and resistance, the sensitivity to temperature can
66
be upgraded to an order. A wide variety of materials have been used for thermistor applications.
Vanadium oxide (VOx) is one of the most widely used thermistor material due to its high TCR of
−2 to −3%/C and its low-temperature process. Many other materials were also seen in NTC
thermistor application, such as SiGe[139], SiC[140], YBCuO[141] and Si[142]. They either suffer
from too high processing temperature or high resistance or intrinsic noise.
Vanadium oxide is gaining increasing interests among researchers in recent years due to its ideal
thermal sensing property: its high activation energy leads to a high TCR with relatively lower
resistance. At the same time, the low 1/f noise and lower processing temperature meet the demand
for electrical test and CMOS technique. Thus, the applications in IR cooled bolometer; thermal
sensors are more frequently seen. The vanadium-oxygen system is a complex system. Wide ranges
of structures can form due to the multivalent vanadium ion. There are many phases in vanadium
oxides, such as VO2, V2O5, and V2O3. Any subtle change in the fabrication process may result in
a different phase and property [143-145]. Up to date, many deposition methods have been utilized
to deposit the thin film, such as sol-gel, RF and DC sputter, IBED, PLD, CVD, and ALD. Each
method might result in a different characteristic, and the condition for annealing can be quite
different. Single-crystal VO2 and V2O5 have large TCR of above 4%/ oC. However, the deposition
of VO2 thin film is very difficult and needs a high-cost ion beam method. V2O5 is easily formed
in high O2 partial pressure, but its resistance at room temperature is very high. V2O3 has low
formation energy. So its resistance is very low at room temperature. Since high electrical resistance
of a device results in a high level of noise, the use of V2O3 phase showing low resistance is
important to the fabrication of low-noise sensors. Vanadium oxides have been applied to
microbolometers in mixed phases formed by ion beam sputtering at low temperature with TCR in
the range from 1.5 to 2.0% /oC due to the limits of the low processing temperature and the thin
67
film thickness required in the microbolometer fabrication process. VOx thin films with high TCRs
and low resistance need to form a suitable mixed phase of VO2, V2O5 (for high TCR) and V2O3
(for low resistance).
Table 5. A summary of different deposition techniques for the growth of vanadium oxide with the
process temperature and the reported TCR values.
Technique Material Processing
temperature (K)
TCR (K-1) Reference
DC sputtering+oxidation VOx 673 -2.0% [146]
PLD VOx 300 -2.8% [147]
Ion beam sputtering+oxidation VO2 473 -2.6% [148]
RF sputtering V2O5/V/V2O5 573 -2.6% [149]
RF sputtering V-W-O 573 -2.6 [150]
3.2 Electrical Characterization Method
Surface resistivity could be defined as the material’s inherent surface resistance to current flow
multiplied by that ratio of specimen surface dimensions. Surface resistivity does not depend on the
physical dimensions of the material. The surface resistivity of a test sample with unit thickness (t)
is expected to equal the resistance of the sample in square dimension regardless of it's in-plane
dimensional surface term approximate resistance. The term surface resistivity in Ohms/square is
the indication of this measurement calculation. Van der Pauw method is used for determining the
surface resistivity of a material.
The Van der Pauw method does not need high accuracy of the geometry of the tested sample
compared to the traditional resistivity measurement method. A common geometry for such a
measurement has four electrical contacts at the four corners of a roughly square sample (Fig. 3-
1b).
68
Figure 3-1. Van der Pauw method for thin film resistivity measurement[151].
exp (−𝜋𝑅12,34𝑑
𝜌) + exp (−
𝜋𝑅23,41𝑑
𝜌) = 1 (3 − 1)
𝑅12,34 =𝑉34𝐼12 (3 − 2)
𝑅23,41 =𝑉41𝐼23 (3 − 3)
The governing equation for the Van der Pauw method is shown in Eq. 3-1. 𝜌 is the resistivity of
the thin film. 𝑑 is the thickness. 𝑅12,34 and 𝑅23,41 are expressed in eq. 3-2 and 3-3 respectively.
3.3 Characterization of the VOx Thin Film as a Thermistor
Vanadium oxide is deposited on 1cm×1cm glass substrates for electrical property characterization,
morphology study, and component analysis. The substrates are first cleaned by a 10min ultrasonic
bath in acetone and DI water, followed by 20 mW plasma clean for 2min in March RIE (Reactive-
ion etching). The VOx thin film is prepared by DC sputtering in a mixed O2/Ar atmosphere. The
1 2
4 3
69
resistivity measurement is carried out using the Van der Pauw method by Signatone Probe Station.
Silver paste is used as the metal contact. To obtain the TCR value, the High-Temperature Probe
Station is used to measure the resistivity at elevated temperatures using the Van der Pauw method
as well. The temperature of the chamber is controlled by a program. In 30 minutes, the temperature
slowly ramps up from 20oC to 100oC. The LabVIEW software controls the incessant measurement
of the resistivity and temperature, and plot the resistivity versus temperature relationship diagram
simultaneously.
The properties of the VOx thin film is sensitive to the O2 flow rate during the DC magnetron
process. During the sputter, the vanadium atoms ejected from the target by ion bombast and flew
to the wafer substrate. During this process, the vanadium particle has sufficient contact with the
oxygen atom and therefore could be fully oxidized. Since vanadium has different valences, the
as-sputtered thin film is the compound of VO2 and V2O5 and usually named VOx. The
experiments showed that the deposition conditions such as the sputtering pressure and the power
had some effects on the properties of the VOx thin film, but not so significantly. The main factor
that affects the electrical property of the VOx thin film is the O2 flow rate (Fig. 3-2). To obtain
low intrinsic noise resistor and high TCR. The resistivity of the VOx under different annealing
temperature is also studied. Fig. 3-3 presents the resistivity change as a function of the annealing
temperature at room temperature (no annealing), 300 and 400 while the VOx thin films were
fabricated under the oxygen flow at 8 sccm, 10 sccm, and 12 sccm. With increasing annealing
temperature, the resistivity of the as-VOx thin films reduced rapidly.
70
Figure 3-2. Oxygen flow rate dependence of the resistivity of VOx under room temperature.
Figure 3-3. Resistivity variation of VOx films with different annealing temperature.
For most ceramics material, they exhibit a negative temperature coefficient of resistivity, governed
by the Arrhenius Equation. The VOx semiconductor thin film is also experimentally proved that
follows such rule shown in Eq. 3-4. Where ρ0 is the prefactor, the kb is the Boltzmann’s constant,
and EaIs the activation energy, which is related to the TCR by α=-Ea/kT2.This indicates that the
slope in Ln()-l/T plots is constant for thermistors. Fig. 3-4 and 3-5 show the activation energy of
0
10
20
30
40
50
60
70
80
6 7 8 9 10 11 12 13
VO
x r
esis
tivit
y
(Ωcm
)
oxygen flow (sccm)
0
10
20
30
40
50
60
70
80
0 100 200 300 400 500
Res
itiv
ity (
Ωcm
)
T ()
8sccm
10sccm
12sccm
71
the sample is 0.19eV (equivalently, temperature coefficient of resistance at 23°C=-2.51%).
(sample was deposited with 10sccm oxygen flow and annealed at 275oC.)
ρ = ρ0 exp (EaKbT
) (3 − 4)
Figure 3-4. Resistance–temperature characteristic of VOx thin film
Figure 3-5. ln 𝜌 − 1/𝑇 linear fitting.
5.00
7.00
9.00
11.00
13.00
15.00
17.00
30.00 40.00 50.00 60.00 70.00 80.00 90.00
(o
hm
cm)
T ()
y = 2258.5x - 4.4933R² = 0.9999
2
2.1
2.2
2.3
2.4
2.5
2.6
2.7
2.8
2.9
3
2.80E-03 2.90E-03 3.00E-03 3.10E-03 3.20E-03 3.30E-03 3.40E-03
Ln(r
)
1/T (K-1)
72
The morphology of the VOx thin film is further studied using scanning electron scope (SEM,
Hitachi 4800) and x-ray diffraction (Ultima III). The SEM image is shown in Fig. 3-6. The VOx
thin film is crystallized with the bar-like shape. The XRD pattern is shown in Fig. 3-7. It shows
the thin film consists of pure V2O5 with the orientation of (501)
Figure 3-6. The SEM image of the VOx thin film
Figure 3-7. The XRD pattern of the VOx thin film
73
3.4 Metal-Insulator Transition Study of Vanadium Oxide
3.4.1 Background
Vanadium oxides exhibit strong correlation effects which greatly depend on the oxidation state of
vanadium [152, 153]. A set of vanadium oxides, such as VO, V2O3, VO2, and V2O5, exist due to
the half d shell of the vanadium atom. Among these oxides, V2O5 is widely studied for its high-
temperature coefficient of resistivity (TCR) in a large temperature range. On the other hand, VO2
shows an intriguing reversible structural phase transition near 68oC. This phenomenon has
attracted a lot of attention since the first observation by Morin in 1959[154]. The bulk VO2 crystal
exhibits a femtosecond time range transition from monoclinic insulating phase (VO2(M)) at low
temperature to rutile metallic phase (VO2(R)) at high temperature in 0.1oC[155]. In the phase
transition, the resistivity of VO2 can change up to 5 orders of magnitude, accompanied by an abrupt
near-infrared transmission change. These properties make VO2 not only an ideal example to
understand the metal-insulator transition (MIT) phenomenon in condensed matter physics, but also
an attractive material in applications such as electrical switches[156], smart window coating[157]
and storage device[158]. Recently, VO2 has been applied to develop uncooled micro-bolometer in
the MIT region due to its large TCR in this region[159, 160].
Compared to the bulk VO2, VO2 thin film has attracted more attention due to its robustness and
capability to withstand numerous heating-cooling cycles without any mechanical cracks, although
its transition usually covers a broader temperature range with larger thermal hysteresis and less
significant optical and electrical properties change. The properties of VO2 thin films, such as
transition temperature, thermal hysteresis and resistivity change during the transition, can be
greatly affected by many factors including grain boundary[161, 162], defect density[163], strain
caused by the lattice mismatch between the VO2 and the substrate[164], and the existence of
74
multivalent vanadium oxides[165]. Various methods, such as low-pressure metalorganic chemical
vapor deposition (LP-CVD)[166], pulsed laser deposition (PLD)[167], atomic layer
deposition[168], sol-gel processing[169] and sputtering[170], have been examined to fabricate
VO2 thin films with high quality for various device development. Some of these methods need
further post heat treatment to achieve the correct phase. Due to the multivalent nature of vanadium
and the existence of several non-hydrate metastable polymorphs of VO2, stabilization of phase
pure VO2 with a single oxidation state (VO2 (M/R)) remains challenging.
In this study, the fabrication of vanadium oxide thin films in two steps. First, pure vanadium thin
films were deposited on sapphire substrates by DC magnetron sputtering. Second, the vanadium
thin films were oxidized in a pure oxygen filled tube furnace. The oxidation temperature and
oxygen flow rate (pressure) were varied systematically to study their effects on the micro structure
and phase transition behaviors of the VO2 thin films. The oxidation time was optimized and fixed
to 4 hours. Electrical resistivity measurements were conducted to characterize the MIT properties.
The crystal orientation and microstructure of these films were studied as a function of deposition
temperature and oxygen flow rate. We found the MIT can be observed only in thin films annealed
from 400oC to 450oC with oxygen flow rates from 1sccm to 20sccm. The transition temperature
varied from 68oC to 80oC in the heating cycle. The transition is reversible, with thermal hysteresis
from 5oC to 30oC and a resistivity change from 0.2 to 4 orders of magnitude. The transition
magnitude and sharpness are correlated with the annealing temperature and oxygen flow rate,
which leads to a better understanding on the growth of VO2 thin films and further paves the way
to the development of VO2 based sensors and other smart devices.
75
3.4.2 The Fabrication of the Vanadium Oxide Samples
Vanadium thin films were deposited on sapphire substrates (c-axis orientation) by DC magnetron
sputtering using the Lesker PVD 75 system. The chamber of the PVD system was pumped down
to 10-3 mT, and subsequently, the argon flow rate was adjusted to a constant pressure of 6mT in
the chamber. The deposition rate was 0.4Å/s under a DC power of 200W. The thickness of the
vanadium thin films was around 100nm measured by the crystal monitor inside the chamber of the
PVD system and the Dektak stylus profiler. A rotation speed of 20rpm was applied to the substrate
sample holder to ensure the thickness uniformity of the deposited thin films.
The as-deposited vanadium thin films were placed in the Lindberg M 55000 series small tube CVD
furnace for oxidation. In the oxidation process, the furnace was pumped down to 100mT and then
filled with oxygen at a certain flow rate. The relationship between the oxygen flow rate and the
stabilized tube pressure is shown in Fig. 3-8. The temperature of the tube furnace was ramped up
to the desired oxidation temperature at 10oC/min, held for 4 hours and then cooled down naturally.
A set of experiments were conducted with different oxidation temperatures and oxygen flow rates
to fabricate a series of vanadium oxide samples for structural and electrical characterization. The
oxidation temperature range was from 390oC to 475oC, while the oxygen flow rate was from 1sccm
to 20sccm. The thickness of the vanadium oxide was 200nm, which was two times thicker than
the as-deposited vanadium thin films.
The TCR measurements were carried out using the Van der Pauw method in a high-temperature
probe station (AO 600 Compact Rapid Thermal Annealing System, MBE). It can measure the real-
time resistivity of the vanadium thin film in a preprogrammed heating/cooling process. Previous
research has demonstrated the MIT as a first-order transition that is temperature scanning rate
76
dependent, where the larger heating/cooling rate will lead to the higher/lower transition
temperature. Fig. 3-9 shows the effect of the heating rate on the MIT of VO2. The increase of
heating rate from 1.3oC/min to 5.3oC/min leads to a Tm shift from 79.2oC to 82.6oC. To reduce the
effect of the scanning rate on the MIT, the chamber of the probe station was preprogrammed to
raise the temperature linearly from 30oC to 110oC in 60 minutes and then cooled down using the
same rate (1.3oC/min). The micro structures were characterized using scanning electron
microscope (SEM) (Hitachi 4800) and x-ray diffraction (Ultima III).
Figure 3-8. The relationship between the oxygen flow rate and the pressure inside the tube of the
furnace.
77
Figure 3-9. The resistivity VS temperature curves for VO2 thin film (sample AII in Table 6) with
different scanning rates. The inset shows the measured amplitude of the TCR (−1 𝜌⁄ (𝑑𝜌𝑑𝑇⁄ )).
𝑇𝑚 is the transition temperature at which the vanadium oxide undergoes abrupt change in
resistivity. It is determined by the largest TCR amplitude.
Table 6. The vanadium oxide sample series and the correlated oxidation conditions.
sample
Condition
Series A Series B
I II III IV V I II III IV V VI
Oxygen flow rate
(sccm)
1 2 5 10 20 2 2 2 2 2 2
Temperature
(oC)
425 425 425 425 425 390 400 410 425 450 475
3.4.3 Characterization and Discussion of the Vanadium Oxide Material
Thin films oxidized at a temperature lower than 400oC were amorphous, whereas the thin films
oxidized at a temperature between 400oC and 475oC, with oxygen flow rates between 1sccm and
20sccm, were found to be well crystalized. To systematically study the influence of the oxidation
78
temperature and the oxygen flow rate to the properties of vanadium oxide separately, two sample
series were fabricated and characterized, listed in Table. 6.
Figure 3-10. The resistivity records of the series A vanadium oxide thin films for both the heating
and cooling cycles.
Figure 3-11. The amplitude of the TCR of the series A vanadium oxide thin films in the
temperature range from 30oC to 110oC. 𝑇𝑚𝐻 is the transition temperature in the heating cycle,
while the 𝑇𝑚𝐶 is the transition temperature in the cooling cycle. ∆𝑇(𝑇𝑚𝐻 − 𝑇𝑚𝐶) is the difference
of the transition temperature between the heating and cooling cycle, which indicates the thermal
hysteresis of the VO2 thin film.
79
Fig.3-10 shows the variation of the electrical resistance of the series A VO2 thin films as a function
of oxygen flow rate for both heating and cooling cycles. Fig. 3-11 shows the TCR of the samples
along temperature in both heating and cooling cycles which is used to define the transition Tm and
thermal hysteresis. The characteristic VO2 transition from the low-temperature insulator phase to
the high-temperature metallic phase was observed.
All the series A samples were oxidized under 425oC. For the thin film oxidized under 425oC and
1sccm oxygen flow rate, a resistivity variation as large as 3.3 decades occurred in a temperature
range of 5oC. As the oxygen flow rate increased up to 20sccm, the resistivity change gradually
decreased to only 0.2 decades. Fig. 42 shows all the series A samples had very high 𝑇𝑚𝐻 and large
thermal hysteresis. The transition temperature 𝑇𝑚𝐻 in the heating cycle is around 80oC which was
significantly larger than that of the bulk VO2 crystals. The thermal hysteresis (∆𝑇) was around
30oC while the oxygen flow increases from 1sccm to 20sccm. It shows the oxygen flow rate can
greatly affect the resistivity change during the MIT, but has little effect on the Tm and the thermal
hysteresis value.
The oxygen flow rate for the series B samples during the oxidation was fixed at 2sccm since better
crystallized VO2 thin films were obtained at this flow rate under different oxidation temperatures
from 390oC to 475oC. The phase transition was observed in sample BII, BIII, BIV and BV with an
abrupt resistivity change around the transition temperature. For sample BII and BV, the thermal
hysteresis was only 6.2oC shown in Fig. 3-13. There was almost no abrupt resistivity change during
the temperature scanning in the vanadium oxide samples BI and BVI, which were oxidized under
390oC and 475oC separately shown in Fig. 3-12.
80
Figure 3-12. The resistivity records of the series B vanadium oxide thin films for both the heating
and cooling cycles.
Figure 3-13. The thermal hysteresis of the vanadium samples oxidized under different
temperatures.
81
Figure 3-14. The XRD patterns of series A vanadium oxide thin films oxidized at 425oC with
different oxygen flow rate.
Figure 3-15. The XRD patterns of series B vanadium oxide thin films oxidized at different
temperatures with the same oxygen flow rate of 2sccm.
82
Figure 3-16. SEM graphs for the vanadium oxide thin films with different oxidation conditions (a)
400oC, 2sccm (BII), (b) 425oC, 2sccm (BIV), (c) 410oC, 2sccm (BIII), (d) 425oC, 1sccm (AI), (e)
425oC, 5sccm (AIII), (f) 390oC, 2sccm (BI).
Fig. 3-14 and Fig. 3-15 show the XRD patterns of both series A and series B vanadium oxide
samples. In Fig. 3-14, samples AI and AII (BIV) have strong peaks appearing at 2θ values of 37o
and 14o, respectively, corresponding to the orientation of (200)VO2 and (001)VO2. It indicates the
high orientation of these epitaxial VO2 thin films. Samples AIII, AIV, and AV, have small peaks
at a 2θ value of 46o, corresponding to (005)V6O13. In Figure. 46, the sample oxidized under
temperature 390oC (BI) only has an obvious peak around 41.67o, corresponding to (006)Al2O3,
indicating the thin film is amorphous. This is consistent with the absence of MIT behavior during
the electrical characterization shown in Figure. 3(a). Thin films oxidized under 400oC (BII), 410oC
(BIII) and 425oC (BIV) only have peaks corresponding to VO2, while the sample oxidized under
450oC (BV) shows a small peak at a 2θ value of 26.7o, corresponding to (110)V6O13. Sample BVI
oxidized under 475oC only has one peak at a 2θ value of 44o which corresponds to (501)V2O5
besides the peak around 41.67o. This can explain the absence of the phase transition during
83
temperature scanning from 30oC to 110oC in Fig. 32. Sample BII only has one strong peak at a 2θ
value of 20o which corresponds to the orientation of (110)VO2.
The surface morphology was characterized by the SEM, and representative graphs are shown in
Fig. 3-16. It shows the oxidation conditions such as temperature and oxygen flow rate can affect
the grain size, shape and arrangement. Specifically, samples BII, BIV, AI, and AIII are well
crystallized with highly oriented grains. The insets show the grain size (width) distribution. The
grain size distribution of BII and AIII are more uniform with smaller average grain size and less
porosity compared to BIV and AI. The crystalline quality decreases in sample BIII compared to
the other four VO2 samples. This can be explained by the stress in the thin films induced by the
multiple orientations of the VO2 crystallites. The VO2 grains vary from spheroidal to worm-like in
shape as the grain size increases from 0.3µm to 0.6µm in sample BIII. Fig. 3-16f doesn’t release
any features which again demonstrates sample BI is amorphous.
The coexistence of monoclinic and tetragonal phases of VO2 during the transition has been widely
reported. As a first order transition, there is a thermal hysteresis between the heating and cooling
cycles during the MIT process. The hysteresis is determined by the potential barrier between the
metallic and insulator phases which can be greatly affected by the strain inside the thin film and
other defects such as grain boundary density, porosity. It can be found that the transition behaviors
of the vanadium oxide thin films are closely related to their composition, micro structure, and
surface morphology, which are largely impacted by the oxidation conditions. For Series A and B
samples, BII, which shows the smallest thermal hysteresis, transition temperature (𝑇𝑚𝐻), and larthe
gest resistivity change during the phase transition, is composed of pure VO2. It has the smallest
average grain size, uniform grain size distribution, high orientation and is well crystallized with
84
compact grain pattern. The larger grain size leads to lower grain boundary density, which
contributes to more significant resistivity change and smaller thermal hysteresis in the MIT region.
Pure VO2 samples, such as AI and AII (BIV), which have larger grain size show relatively larger
thermal hysteresis compared to BII. This phenomenon may be related to porthe osity of these thin
films. All the samples which have resistivity change of more than 3 orders of magnitude during
the phase transition are composed of pure VO2 with sina gle orientation and are well crystallized.
The less significant MIT of other samples can be due to multiple orientations, mixture with other
vanadium oxide states and less crystallized grains. The MIT occurs at higher transition
temperatures (Tm) for all the VO2 thin films in the experiments compared to the single crystal VO2.
It has been widely reported that the residual strain along the c-axis in VO2 thin films strongly
influences the Tm for MIT. The higher measured Tm may be caused by in-plane tensile strain
induced by the c-axis orientation sapphire substrate and the scanning rate effect.
In conclusion, VO2 thin films were deposited on sapphire substrates in two steps. Vanadium thin
films were firstly obtained by DC magnetron sputtering and then oxidized in a CVD furnace for 4
hours under different oxygen flow rates and temperatures. VO2 thin films with monoclinic
insulator phase were obtained. The VO2 thin films show an MIT behavior in the temperature range
from 40oC to 85oC. During the transition, the resistivity change of the thin film ranges from 0.2 to
4 orders of magnitude. A thermal hysteresis over 30oC was observed in some VO2 samples and
can be reduced to 6oC by changing the oxidation conditions. The relationship between the
transition behavior and the microstructure of VO2 thin films were further studied using SEM and
XRD. This study contributes and leads to a better understanding of how the oxidation conditions
affect the micro structures of the VO2 thin films and further affect their MIT properties. It shows
highly oriented VO2 thin films have more abrupt resistivity change during the MIT process. It also
85
indicates a more compact and highly oriented VO2 grain arrangement can significantly reduce the
thermal hysteresis which can be potentially applied to develop uncooled bolometer.
3.5 Summary
This chapter presents the vanadium oxide material study. There are many phases in vanadium
oxides, such as VO2, V2O5, and V2O3. Any subtle change in the fabrication process may result in
a different phase and propert. Up to date, many deposition methods have been utilized to deposit
the thin film, such as sol-gel, RF and DC sputter, IBED, PLD, CVD, and ALD. Each method might
result in different properties. Furthermore, the properties of vanadium oxide as an excellent
thermistor material is compared with other materials. The fabrication method is introduced and
followed by the electrical and micro structure characterization. The temperature coefficient of the
vanadium oxide thin film can reach to -2.51% per oC while the resistance of the fabricated
thermistor is around 10kΩ. The grain of the thin film is bar-like and the average size is 2µm. The
XRD study shows the thin film is mainly composed of (501)V2O5. In Section 3.5, we specifically
study the metal-insulator of vanadium oxide under certain fabrication conditions. Vanadium thin
films were deposited on sapphire substrates by DC magnetron sputtering and then oxidized in a
tube furnace filled with oxygen under different temperatures and oxygen flow rates. The
fabricatedVO2 thin films show an MIT behavior in the temperature range from 40oC to 85oC.
During the transition, the resistivity change of the thin film ranges from 0.2 to 4 orders of
magnitude. A thermal hysteresis over 30oC was observed in some VO2 samples and can be reduced
to 6oC by changing the oxidation conditions. The fabricated VO2 thin films showed the potential
to be applied to the development of electrical sensors and other smart devices.
86
Chapter 4. MEMS-Based DSC Characterization
4.1 Abstract
In this chapter, we focus on the thermal analysis and characterization of the MEMS based DSC.
First, we introduce the numerical method for the thin film heater transfer analysis. In this model,
we take the thermal delay in consideration. A FEM model based on the numerical method is further
developed to study the heat transfer in the DSC thin film gauge. The FEA analysis is used to
determine the optimized distance between the sample and reference cells and provide the design
guidance for highly integrated MEMS DSC array. In the characterization of the MEMS based
DSC, the technical specifications such as the response time and power sensitivity are of vital
importance. The last part is the discussion about the failure of the using the micro heater for
temperature scanning.
4.2 Temperature Distribution Analysis of the MEMS-Based DSC
4.2.1 Heat Transfer in Thin Films
Characterizing the thermal behavior of thin films is of vital importance in micro and nano heat
transfer application. Thin films such as metals, Si, SiO2, polymers are important components of
microelectronic devices. The reduction of the device size to microscale enhances the performance
of the device. On the other hand, size reduction increases the rate of heat generation, which leads
to a high thermal load on the micro-device [171, 172]. Heat transfer at the microscale is also
important for the processing of materials with a pulsed-laser fabrication. Examples in metal
processing are laser micro-machining, laser patterning, laser processing of diamond films from
carbon-ion implanted copper substrates, and laser surface hardening. Hence, studying the thermal
behavior of thin films or micro objects is essential for predicting the performance of a
87
microelectronic device or for obtaining the desired microstructure. In the macro level, the
governing equations for heat transfer are:
−𝛻 ∙ 𝑞 + 𝑄 = 𝜌𝐶𝑝𝜕𝑇
𝜕𝑡 (4 − 1)
𝑞(𝑥, 𝑦, 𝑧, 𝑡) = −𝑘𝛻𝑇(𝑥, 𝑦, 𝑧, 𝑡) (4 − 2)
Where 𝑞 is the heat flux, 𝑄 is the heat source, 𝜌 is the mass density, 𝐶𝑝 is the heat capacity, 𝑘 is
the thermal conductivity. In the micro-level, considering the increased heat generation rate and
high thermal load, there are lags for the heat flux and temperature gradient. Eq. 4-2 should be
rewritten as:
𝑞(𝑥, 𝑦, 𝑧, 𝑡 + 𝜏𝑞) = −𝑘𝛻𝑇(𝑥, 𝑦, 𝑧, 𝑡 + 𝜏𝑇) (4 − 3)
Using Taylor series for the first order approximation, Eq. 4-3 can be written as:
𝑞 + 𝜏𝑞𝜕𝑞
𝜕𝑡= −𝑘 (𝛻𝑇 + 𝜏𝑇
𝜕(𝛻𝑇)
𝜕𝑡) (4 − 4)
𝜏𝑞 and 𝜏𝑇 are positive constants for the heat flux and temperature gradient respectively. For the
thin films, the heat transfer in the x and y directions still follow the traditional Fourier’s Law, only
the heat transfer in the z direction needs to consider the thin film effect. The Eq. 4-4 can be
separated into 3 equations in the x, y and z directions as Eq. 4-5, 4-6 and 4-7. The model for the
thin film is shown in Fig. 4-1.
88
Figure 4-1. The thin film model and the coordinates
𝑞1 = −𝑘𝜕𝑇
𝜕𝑥 (4 − 5)
𝑞2 = −𝑘𝜕𝑇
𝜕𝑦 (4 − 6)
𝑞3 + 𝜏𝑞𝜕𝑞3𝜕𝑡= −𝑘(
𝜕𝑇
𝜕𝑧+ 𝜏𝑇
𝜕
𝜕𝑡(𝜕𝑇
𝜕𝑧)) (4 − 7)
Conducting the differential in the z-direction, Eq. 4-8 is derived from Eq. 4-7.
𝜕𝑞3𝜕𝑧+ 𝜏𝑞
𝜕
𝜕𝑡(𝜕𝑞3𝜕𝑧) = −𝑘 (
𝜕2𝑇
𝜕𝑧2+ 𝜏𝑇
𝜕
𝜕𝑡(𝜕2𝑇
𝜕𝑧2)) (4 − 8)
Eq.1 can also be written as:
−𝜕𝑞3𝜕𝑧= 𝜌𝐶𝑝
𝜕𝑇
𝜕𝑡− 𝑄 +
𝜕𝑞1𝜕𝑥+𝜕𝑞2𝜕𝑦 (4 − 9)
Combining Eq. 4-5, 4-6, 4-9:
−𝜕𝑞3𝜕𝑧= 𝜌𝐶𝑝
𝜕𝑇
𝜕𝑡− 𝑄 − 𝑘
𝜕2𝑇
𝜕𝑥2− 𝑘
𝜕2𝑇
𝜕𝑦2 (4 − 10)
89
Combining Eq. 4-8 and 4-10:
1
𝑎
𝜕𝑇
𝜕𝑡+𝜏𝑞
𝑎
𝜕2𝑇
𝜕𝑡2= 𝛻2𝑇 + 𝜏𝑞(
𝜕3𝑇
𝜕𝑡𝜕𝑥2+𝜕3𝑇
𝜕𝑡𝜕𝑦2) + 𝜏𝑇
𝜕3𝑇
𝜕𝑡𝜕𝑧2+ 𝐺 (4 − 11)
Where 𝑎 is 𝑘
𝜌𝐶𝑝 and G is expressed as:
𝐺 =1
𝑘(𝑄 + 𝜏𝑞
𝜕𝑄
𝜕𝑡) (4 − 12)
4.2.2 “Cross-Talk” Effect between the Sample and Reference
The MEMS based DSC device consists of 2 parts shown in Fig. 4-2: The micro heaters and the
thermistors. The polyimide serves as the substrate for the thermistors and heaters. It is also used
as the dielectric layer between the heater and thermistor and the protective layer on the thermistors
due to its high mechanical strength, low gas permeability and ease fabrication process. The micro
heater is designed for the MEMS based DSC calibration. The goal of the design is to achieve high-
temperature uniformity. The micro heater is composed of a resistive Au/Ti trace on a polyimide
membrane. When a certain amount of heating power is applied to the heater, the temperature is
quickly redistributed over the suspended polyimide membrane. The heat will ultimately be
transferred through conduction to the periphery and convection with the air. A novel pattern design
of thermistor with thermal noise two orders of magnitude lower than traditional design. In this
new design we sandwich thermistor trace with electrical conductor trace. Therefore the electrical
current flows in the side direction, and thus the whole resistance decreased greatly since the
resistivity of the gold trace is much smaller than that of the thermistor material. In the MEMS
based DSC, VOx is used as the thermistor material due to its low fabrication temperature
(~400K), and high-temperature coefficient of resistivity.
90
Figure 4-2. The design of the MEMS-based DSC
To determine the distance between the sample and reference cell, COMSOL 5.0 was utilized to do
the FEA study about the “cross-talk” effect between the sample and reference. The FEA model for
the MEMS based DSC consists of two identical micro heaters and a polyimide substrate (Fig. 4-
3a). The gold trace of the heater is double spiral which could be simplified by a series of concentric
circles. The detailed heater design is described in ref. [173]. The size of the polyimide membrane
was 6 × 12𝑚𝑚2 while the radius for the outside heater circle was 1.2mm. A larger polyimide
substrate is unnecessary since it’s enough for the temperature to attenuate to ambient at 3mm from
the center of the heater. The total grid numbers for the DSC model is 491906 (Fig.4-3 b). The
temperature rise was caused by the heat release in bio-reaction which was represented by a constant
power added to the heater in the simulation. The heating power for each heater is 50 µW to simulate
the heat release in the bio-sample. In the steady-state thermal FEA analysis, two typical cases wthe
side face of the polyimide membrane was set to be the ambient temperature 𝑇𝑎𝑖𝑟 which was 25.
Need to mention that the temperature for the reaction is around 70oC. However, since the heat
convection coefficient is independent and constant, the “cross talk” effect will be much the same
polyimide
91
when the ambient temperature is set to be 70oC. In this way, the second boundary condition which
the boundary temperature gradient was zero can be used in this study. This setting can be explained
by the sharp temperature attenuation in the non-heating area of the membrane. From the simulation
results, the temperature attenuates rapidly to the room temperature in the non-heating area.
Figure 4-3. (a) The MEMS based DSC model. (b) The grid system for the DSC model.
Figure 4-4. Temperature distribution in the polyimide membrane (the distance between the two
heaters was 4mm.).
(a) (b)
92
Figure 4-5. The temperature distribution in the path AA shown in Fig. 4-4. (a) The distance
between the two heaters is 4mm (∆𝑇𝐴𝑚𝑎𝑥 = 0.0024). (b) The distance is 3mm (∆𝑇𝐵𝑚𝑎𝑥 =
0.069).
Fig. 4-4 shows the temperature distribution on the polyimide thin film. Path AA is a line from the
left side of the polyimide substrate to the right side passing the centers of the two micro heaters.
The distance between the two micro heaters was 4mm. The maximum temperature rise was 2.71oC.
In the simulation, two heating strategies (single heater heating VS double heaters heating) were
applied to study the thermal interaction between the two heaters. The distance between the two
heaters was also changed gradually to study the influence of the heater distance to the temperature
distribution in the path AA. Fig. 4-5(a) and (b) shows the temperature distribution in path AA
when the distance between the two heaters was 3mm and 4mm respectively. The blue and red lines
showed the temperature distribution in path AA for the single heater heating mode while the black
lines represented the double heater heating mode. The temperature distribution in the path AA for
the two different heating strategies showed how the two heaters interacted with each other
expressed by the peak temperature difference (∆𝑇𝑚𝑎𝑥) between the two heating modes. If the two
heaters are two close, the heat released in one heater will also transfer to the other heater area and
generate a temperature difference. In Fig. 4b, when the distance between the two heaters was 3mm,
(a) (b)
93
there is a significant temperature difference in the heating area when both heaters were heated up
compared to single heater heating (∆𝑇𝐵𝑚𝑎𝑥 = 0.029). If the distance is too large, the extra space
will be wasted which will hinder the integration of the MEMS DSC array. For the DSC we
proposed, the optimized distance between the sample and reference cell is around 4mm.
4.3 Characterization of the MEMS-based DSC
4.3.1 Characterization of the Heating Stage
The heating stage is constructed to raise the temperature and provide the thermal shielding. It
mainly consisted of a heating sink, a copper plate, two Peltier heaters, a sample stage, a glass cover,
the thermal shielding (made of Teflon tube), and the metal shielding. The two-layer shielding
blocked the outside disturbance and minimized the temperature fluctuation. The temperature of
the heating stage is controlled by the temperature controller (TC-720 from TE tech). The properties
of the heating stage such as the temperature scanning rate, linearity, uniformity of the temperature
distribution were studied.
The TC-720 is a bipolar temperature controller capable of automatically reversing power to Peltier
thermoelectric (TE) devices to provide heating or cooling as required to maintain a specific set
point temperature. It incorporates a keypad and a liquid-crystal display housed in a die-cast
aluminum box. The display allows the user to monitor the sensor temperatures, output level, alarm
conditions, and menu settings. The integrated keypad accesses an easy-to-use menu system,
allowing the user to adjust all of the basic controller parameters such as the set temperature, tuning
parameters, and alarm parameters. The controller can also be connected to a computer via a USB
port for advanced programming, data graphing, and data logging. All of the controller parameters,
including the advanced parameters which are not adjustable through the keypad, can be adjusted
94
with the included software and saved to EEPROM. The command set for the controller is also
provided which allows the creation of custom software applications using National Instruments
LabVIEW.
The heating stage is characterized in two different cases: with the thermal cover (Fig. 4-6a) and
without the thermal cover (Fig. 4-6b). The thermocouple is placed in the small hole on the surface
of the heating stage shown in Fig. 4-6b.
Figure 4-6. (a) The heating stage with thermal cover. (b) The heating stage without thermal cover.
(a) (b)
95
Figure 4-7. The linearity of the temperature scanning process
Fig. 4-7 shows the linearity analysis of the temperature scanning process of the heating stage. The
scanning rate is 0.757oC/s (45oC/min). In the linearity study, we put the two shielding layers on
the sample stage. As illustrated in equation (3), a higher scanning rate leads to higher DSC
sensitivity, the temperature scanning rate of the MEMS based DSC is much higher than the current
commercial DSC (1oC/min for the Microcal VP-DSC).
(a) (b)
96
Fig. 4-8. (a) The temperature stability study of the heating stage with the two shielding layers. (b)
The temperature stability study of the heating stage without the two shielding layers.
Fig. 4-8 (a) and (b) show the temperature stability of the sample stage with and without the
shielding layers respectively. For both cases, the temperatures are fixed at 90oC; the temperature
variance is 0.31oC when the heating stage is covered with the two shielding layers, and it is
increased to 0.42oC when the shielding layers are removed.
The temperature distribution study on the sample stage is studied using the FLIR A655sc high-
resolution infrared camera. FLIR A655sc is equipped with an uncooled, maintenance free,
vanadium oxide microbolometer detector that produces thermal images of 640 x 480 Pixels. These
pixels generate crisp and clear detailed images that are easy to interpret with high accuracy. The
FLIR A655sc will make temperature differences as small as 50 mK visible. Infrared was firstly
discovered in 1800 by Sir William Herschel as a form of radiation beyond red light. There are four
basic laws of IR radiation: Kirchhoff's law of thermal radiation, Stefan-Boltzmann law, Planck’s
law, and Wien’s displacement law. More specifications of the FLIR A655sc are listed in Table. 7
Table 7. Specifications of the FLIR A655sc infrared camera.
System overview FLIR A655sc infrared camera
Detector type Uncooled microbolometer
Spectral range 7.5-14µm
Resolution 640 × 480
Detector pitch 17µm
NETD <30mK
Time constant 8ms
Frame rate 50Hz
(b)
97
Figure 4-9. (a) The IR image of the sample stage. (b) The 3D diagram of the temperature
distribution on the surface of the sample stage.
Table 8. The statistical analysis of the temperature distribution in the box 2 area shown in Fig. 56a.
Tmax=90.67oC Tmin=89.18oC Tmean=89.56oC 𝜎=0.34oC
The temperature analysis of the sample stage is shown in Fig. 4-9. We use the FLIR A655sc IR
camera to take the IR image and further export the data for the 3D mapping of the temperature
distribution. In the box 2 area which we placed the MEMS based DSC, the variance of temperature
is only 0.34oC when the average temperature reaches 89.56oC. The temperature analysis was
carried out without the thermal shielding and metal shielding shown in Fig. 4-8. A smaller
temperature variance in the box is expected when the two shielding layers are installed.
4.3.2 Response Study of the MEMS-based DSC
The MEMS based DSC is a first order system. Important parameters which indicate the
performance of the DSC are studied systematically. Thermal response tests were often used to
calibrate the calorimeter and to show how the sensor reacts to a small amount of heat. It was
operated by applying a known power to the micro heater for a certain time and read the voltage
(a) (b)
98
signal. The temperature sensing unit is a Wheatstone bridge consists of two identical thermistors
and two-decade resistor boxes (Fig. 4-10). The whole MEMS based DSC measurement system is
shown in Fig. 4-11. In the measurement system, the DSC device is placed in the heating stage. The
source-meter is used to provide constant source voltage for the Wheatstone bridge and power for
the sensitivity calibration. The temperature controller is controlled by Labview and powered by
the power supply to directly control the heating stage to heat the MEMS based DSC device
linearly. The lock-in amplifier is to measure the output of the Wheatstone bridge. The whole
system is controlled by the Labview program. By dividing the output voltage by the power applied,
the sensitivity could be determined. It also characterized the sensor’s performance on thermal
insulation and time constant. Fig. 4-12 showed the situation when 50 μW power was applied, and
0.63 μL of DI water was loaded in the microchambers. Integrating the voltage over time and
dividing the known heat revealed a 6.1 V/W sensitivity. By fitting the rising curve and downward
curve to the exponential equation, we found the time constant to be 3 s, which is smaller than some
commercial products (time constant of VP-DSC is 7s). We further studied the relationship between
the sensitivity and the temperature of the heating stage by repeating the experiment above at
different heating stage temperature. The relationship is shown in Fig. 4-13.
99
Figure 4-10. The schematic diagram of the Wheatstone bridge consists of two thermistors (R1 and
R2) and two external decade resistor boxes (R2 and R4) and the lock-in amplifier for differential
detection.
Figure 4-11. The schematic diagram of the whole MEMS based DSC measurement system.
100
Figure 4-12. Step response of the MEMS based DSC (50µW input, 30oC).
Figure 4-13. Relationship between the sensitivity and the temperature of the heating stage.
4. 4 Analysis of the Defects in the Fabricated Micro Heater
4.4.1 Background
The MEMS based DSC device consists of two main parts: The micro heaters and the thermistors.
The polyimide serves as the substrate for the thermistors and heaters. It is also used as the dielectric
101
layer between the heater and thermistor and the protective layer on the thermistors due to its high
mechanical strength, low gas permeability and ease fabrication process. The micro heater is
primarily designed for the MEMS based DSC heating as well as the calibration. The goal of the
design is to achieve high-temperature uniformity. The micro heater is composed of a resistive
Au/Ti trace on a polyimide membrane. When a certain amount of heating power is applied to the
heater, the temperature is quickly redistributed over the suspended polyimide membrane and heat
the sample and reference.
However, in the primary design of the MEMS based DSC, the micro heaters tend to have defects.
Some of the samples have cracked before the temperature scanning while others show cracks after
the temperature scanning shown in Fig. 4-14 and Fig. 4-15 separately.
Figure 4-14. Micro heater sample with cracks before temperature scanning
102
Figure 4-15. The micro heater sample shows cracks after the temperature scanning. The left figure
is the image of the heater before the temperature scanning. The right figure shows the image of the
heater after temperature scanning.
4.4.2 Experiments of Using the Micro Heater for Temperature Scanning.
A LabVIEW program is used to achieve identical linear temperature scanning with the heaters
located in the sample and reference areas. In the experiments, both sample and reference chambers
are filled with DI water. The volume is 0.63µL. The power added to the micro heater is linear to
time since to achieve linear temperature heating since the MEMS based DSC can be regarded as a
first order system. Considering the resistivity of the micro heater (made of gold) changes with
temperature. The resistance of the micro heater is measured at the end of each heating cycle and
used to determine the next power input at the beginning in the next heating cycle. The real-time
power added to the two heaters, the differential power between them and the output voltage are
recorded for analysis. Fig. 4-16 shows the real-time recording of the power added to the two heaters
for linear temperature scanning. The targeted temperature is 80oC. The power scanning rate is
8mW/min, and the temperature scanning rate is 11oC/min. Fig. 4-17 shows the differential power
between heater 1 and heater 2. It can seem that the differential power is less than 0.2µW when the
measurement time is less than 200s. The differential power is significantly increased to microwatt
level after 200s’ scanning and can reach to 60µW at some random peaks. At 200s, the heating
103
power for both heater 1 and heater 2 is 23.4mW. Fig. 4-18 shows the output voltage of the
Wheatstone bridge as a response to the differential power. The output voltage keeps in microvolt
level before 200s and suddenly increase to millivolts level after 200s which is consistent with
differential power recording. The sudden change indicates there are some defects in the micro
heater after 200s. The cracks in the image shown in Fig. 41-15 demonstrates such an assumption.
Figure 4-16. The real-time recording of the power added to the heater 1 (heater located in the
sample area) and heater 2 (heater located in the reference area).
104
Figure 4-17. The differential power between the two micro heaters. The inset shows the recording
between 0s to 100s as the power value is too small.
Figure 4-18. The output voltage recording for the temperature scanning process.
The defects in the micro heater significantly affect the performance of the micro heater. This is the
main reason that we later changed to the heating stage method for the temperature scanning instead
of using the micro heaters. However, experiments also show the micro heaters work well when the
heating power is small (less than 10mW) so the micro heater is still can be used for the sensitivity
calibration which only needs 50µW.
The stress between the polyimide substrate and the gold thin film can lead to the defects of the
micro heater. High-temperature polyimide is widely used in the microelectronics industry as
interlayer dielectrics and passivation layers are owing to excellent mechanical properties, high
thermal stability, low dielectric constant, and high chemical resistance. It is commonly known that
for multilayered devices there are often reliability problems, such as displacement, cracks, and
delamination at the interface. Interfacial stress generation is linked to potential reliability problems
in integrated circuits, such as loss of adhesion and dimensional stability due to the mismatch of
105
physical properties between layers. Thus, it is desirable to have low-stress level materials such as
SiO2 that are inert to ambient fluctuations in conditions such as temperature and humidity.
4.5 Summary
In this Chapter, we characterized the key components of the MEMS DSC (the micro heater and
the heating stage) as well as the power sensitivity and static noise level of the whole MEMS DSC
system. At first, a model for the thin-film thermal analysis was proposed. The model is used to
study the temperature distribution on the polyimide thin film and the cross-talk between the sample
and reference areas. The heating stage which is constructed to raise the temperature and provide
the thermal shielding.. The temperature of the heating stage is controlled by the temperature
controller (TC-720 from TE tech). The properties of the heating stage such as the temperature
scanning rate, linearity, uniformity of the temperature distribution were studied. The sensitivity of
the MEMS DSC is 6.1V/W, and the static noise level is 10µK. In the last section of this chapter,
we studied the micro heater as the heat source for temperature scanning. It showed the with high
power input (25mW), there were cracks on the micro heater which leaded to large output
fluctuation. Literature shows this crack may be induced by the stress generated in the flexible
polyimide substrate fabrication process.
106
Chapter 5. Application of the MEMS DSC in Protein Tests
5.1 Introduction
Differential scanning calorimetry (DSC) is a technique able to study thermally induced transitions
and particularly, the conformational transitions of biological macromolecules (for example
between the folded and the unfolded structure of a protein or between single and double-stranded
DNA). DSC measures the excess heat capacity of a solution (𝐶𝑝) of the molecule of interest as a
function of temperature. The transition is recognized as a sharp endothermic peak-centered at 𝑇𝑚
and the maximum in 𝐶𝑝 occurs directly at 𝑇𝑚 . Integration of the 𝐶𝑝 versus T curve yields the
Transition enthalpy (∆𝐻𝑜(𝑇𝑚)) and the shift in the baseline yields the Cp. DSC is the only
method for the direct determination of ∆𝐻𝑜(𝑇𝑚). From ∆𝐻𝑜(𝑇𝑚) and ∆𝐶𝑝, ∆𝐺𝑜(𝑇) and ∆𝑆𝑜(𝑇)
can be determined. As a result, DSC is able to provide all the thermodynamic parameters of a
conformational transition.
One goal of protein engineering and biopharmaceutical formulation (galenic formulation) is the
development, production, and storage of stable proteins with full functionality. It is therefore
essential to understand how proteins fold into their biological states and how these active states
are stabilized. A denatured protein has a higher heat capacity than the native one. As a result, the
∆𝐶𝑝for protein u,nfolding is almost always positive. The variation of heat capacity associated with
protein unfolding is primarily due to changes in hydration of side-chains that were buried in the
native state, which become exposed to the solvent in the denatured state. Many factors are
responsible for the folding and stability of native proteins, including hydrophobic interactions,
hydrogen bonding, conformational entropy. Moreover, 𝑇𝑚 is an indicator of thermal stability and
107
generally, the higher is the Tm, the more thermodynamically stable is the protein. Proteins with
higher Tm are less susceptible to unfolding and denaturation at lower temperatures.
DSC measures the heat change associated with the molecule's thermal denaturation when heated
at a constant rate. Measuring the 𝑇𝑚, provides a quick and easy indication of stability. To obtain
the enthalpy, the excess heat capacity function must be integrated. Before this can be done, the
instrumental baseline which is observed when both the sample and the reference cells contain
solvent only must be removed (Fig. 5-1). For technical reasons, the instrumental baseline does not
give a zero excess heat capacity between the cells. After baseline subtraction, the linear regions on
either side of the transition peak, which represent the heat capacity of the native and denatured
states of the protein, is extrapolated into the transition and then merge them about the progression
through the transition. This is done with software. Finally, the area under the resulting peak is
integrated to give the excess energy that the DSC requires to denature the protein in the sample
cell. Provided that the concentration of the protein solution and the operational volume of the
calorimeter cell are known, this energy can be converted into ΔH in calories or Joules per mol of
protein.
(a)
)
(b)
)
108
Figure 5-1. DSC data for interpolation. (a) The blue line is the raw data of protein unfolding
obtained by a commercial DSC; the red line indicates the buffer heat capacity (b) after baseline
subtraction, the final result of the protein denaturation heat capacity.
The irreversible step may also result from association or aggregation (Ag) of the denatured state.
This will manifest itself in a concentration dependence of the initial scan parameters. At higher
sample concentrations, the concentration of misfolded proteins will be increased so that the
irreversible step is accelerated, leading to a lower Tm for the protein. In cases of significant and
rapid aggregation, it may even be possible to distort and truncate the denaturation transition with
the exothermic aggregation event and observe a typical noisy data above the transition region.
5.2 Thermodynamic Parameters
The stability of biological macromolecules, and biomolecular associations, is quantified by the
standard free energy G° or the difference in Gibbs energy between different states (the native and
denaturated structures in the case of protein and nucleic acid stability). At equilibrium (when
G=0), G° is related to the equilibrium constant (K) between the two states by:
∆𝐺𝑜(𝑇) = −𝑅𝑇𝑙𝑛 𝐾(𝑇) (5 − 1)
Where R is the universal gas constant and T the absolute temperature in Kelvin. G° is the sum of
two contributions:
∆𝐺𝑜(𝑇) = ∆𝐻𝑜(𝑇) − 𝑇∆𝑆𝑜(𝑇) (5 − 2)
Where ∆𝐻𝑜 and ∆𝑆𝑜 are the enthalpy and entropy changes at the temperature at which is
∆𝐺𝑜 being evaluated. When values of K can be determined experimentally as a function of
temperature (this can be done by various spectroscopic methods) the data can be fitted to yield
109
values for ∆𝐻𝑜. If ∆𝐻𝑜 is independent of the temperature (∆𝐶𝑝= 0), a plot of lnK in function of
1/T will give a straight line with an angular coefficient of –∆𝐻𝑜/R. This case is however very
unusual, ∆𝐻𝑜 being generally temperature-dependent. In this case, it is necessary to make a
hypothesis on this dependence in order to be able to determine DH°. It has also to be noted that
the fit assumes a simple two-state equilibrium, without for example stable folding intermediates
in the case of protein folding. This non-calorimetric and indirect approach to obtain
thermodynamic parameters is called the Van’t Hoff analysis.
ln𝐾(𝑇) = (−∆𝐻𝑜(𝑇)
𝑅) (1
𝑇) + (
∆𝑆𝑜(𝑇)
𝑅) (5 − 3)
𝑑 ln (𝐾)
𝑑 (1𝑇)
=−∆𝐻𝑜(𝑇)
𝑅 (5 − 4)
When a macromolecule changes its thermodynamic state (e.g., unfolds), a heat capacity change
(∆𝐶𝑝) is observed. This change is due to the fact that the heat required to raise the temperature of
a solution of unfolded protein is greater than that required for a solution of folded protein. Heat
capacity changes are primarily due to rethe structuring of the solvent molecules around the non-
polar sidechains exposed to the solvent during the unfolding process. Assuming a constant
temperature-independent (∆𝐶𝑝), ∆𝐺𝑜 is described by:
∆𝐺𝑜(𝑇) = ∆𝐻𝑜(𝑇𝑅) − 𝑇∆𝑆𝑜(𝑇𝑅) + ∆𝐶𝑝(𝑇 − 𝑇𝑅) − 𝑇𝑙𝑛 (
𝑇
𝑇𝑅) (5 − 5)
Where 𝑇𝑅 is the reference temperature and T the temperature of interests. The unfolding transition
occurs at a characteristic temperature called transition midpoint, 𝑇𝑚 . Assuming a two-state
110
transition, ∆𝐺𝑜 is equal to zero if 𝑇𝑅 is equal to 𝑇𝑚. As a result, ∆𝑆𝑜(𝑇𝑚) is just ∆𝐻𝑜(𝑇𝑚)/𝑇𝑚.
∆𝐺𝑜(𝑇) thus becomes:
∆𝐺𝑜(𝑇) = ∆𝐻𝑜(𝑇𝑚) (1 −𝑇
𝑇𝑚) + ∆𝐶𝑝 [(𝑇 − 𝑇𝑚) − 𝑇𝑙𝑛 (
𝑇
𝑇𝑚)] (5 − 6)
To characterize the thermodynamics of an unfolding/folding process means to determine∆𝐺𝑜, ∆𝐻𝑜
and ∆𝑆𝑜 at a given temperature and to obtain ∆𝐶𝑝 to predict the change of these three parameters
with temperature. The denaturation enthalpy and entropy at temperature 𝑇 can be calculated
according to the following relations (Kirchhoff’s law):
∆𝐻𝑜(𝑇) = ∆𝐻𝑜(𝑇𝑚) + ∆𝐶𝑝(𝑇 − 𝑇𝑚) (5 − 7)
∆𝑆𝑜(𝑇) = ∆𝑆𝑜(𝑇𝑚) + ∆𝐶𝑝 ln (𝑇
𝑇𝑚) (5 − 8)
Where ∆𝑆𝑜(𝑇𝑚) is ∆𝐻𝑜(𝑇𝑚)/𝑇𝑚, the denaturation entropy at 𝑇𝑚.
5.3 Characterization of the MEMS-Based DSC in Protein Sample Tests
5.3.1 Baseline Characterization
We characterize the device’s performance in protein sample tests by comparing it with the
commercial DSC. All the protein and buffer samples are prepared in AbbVie Inc (Worcester, MA).
The MEMS based DSC us to obtain the thermodynamic properties of the protein sample during its
unfolding processes, such as partial specific heat capacity, the total change of molar enthalpy (ΔH),
and the mid-point of the transition temperature Tm. We compare the performance of the MEMS
based DSC and the TA DSC in protein sample tests. The MEMS DSC system and the TA discovery
DSC are shown in Fig. 5-2.
111
Figure 5-2. (a)The MEMS based DSC system in protein sample tests. (b) The TA discovery DSC
for protein sample tests
There are some key technical specifications to evaluate the performance of the DSC in protein
tests. Heat flow noise: the fluctuation level of the heat flow when the system is not working
(static). Baseline repeatability: the fluctuation level of heat flow when the system is running
buffer-buffer scanning. Repeatability: the agreement between the results of successive
measurements, expressed by the mean and standard deviation. Reproducibility: the agreement
between the results of the same measurements on the different fabricated device. Scanning rate
limit: the lowest and highest scanning rates. Other important specifications are the baseline drift,
sample consumption volume, and sample concentration range.
(a) (b)
(a) (b)
112
Figure 5-3. (a) The baseline tests of the TA discovery DSC. (b) The baselines after removing the
linear part.
Fig. 5-3 shows the baseline tests of the TA discovery DSC. For the baseline tests, the sample pan
is filled with buffer and runs under a scanning rate of 5oC/min from 30oC to 90oC. The buffer is
15mM histidine (pH6), and the volume is 20µL. The baselines are shown in Fig. 68a consists of
two parts: the linear part which is the baseline drift and the fluctuation part. Fig. 68b shows the
baseline fluctuation after the linear part subtraction. The baseline fluctuation range shows the
repeatability of the baseline which is ±4𝜇𝑊 for the TA discovery DSC. The largest baseline drift
is 117𝑛𝑊/𝑠.
Figure 5-4. (a) the baseline tests of the MEMS based DSC. (b) the baselines after removing the
linear parts.
Fig. 5-4 shows the baseline tests of the MEMS based DSC. The sample and reference chambers
are filled with buffer and run under a scanning rate of 5oC/min from 30oC to 90oC. The buffer is
15mM histidine (pH6), and the volume is 0.63𝜇𝐿. The baselines shown in Fig. 69a consists of two
parts: the linear part which is the baseline drift and the fluctuation part. Fig. 69b shows the baseline
fluctuation after the linear part subtraction. The baseline fluctuation range shows the repeatability
(a) (b)
113
of the baseline which is ±2𝜇𝑉 for the MEMS based DSC. The largest baseline drift is 0.1𝜇𝑉/𝑠.
Considering the sensitivity of the MEMS based DSC is 6.1V/W, the baseline repeatability is
0.33µW which is 10 times smaller than that of the TA discovery DSC. The linear drift is 0.02µW/s
which is 6 times smaller than that of the TA DSC.
5.3.2 Lysozyme Sample Tests
We demonstrated the MEMS based DSC’s performance with the lysozyme protein (10 mg/mL).
We raised the temperature of the chamber from 25oC to 95oC at the rate of 10 oC /min. After
baseline subtraction, the measured voltage outputs from the Wheatstone bridge showed the typical
DSC curve of protein denaturation. We could further obtain the partial specific heat capacity of
lysozyme as a function of temperature (Fig. 5-5) based on equation (3). The melting temperature
were at around 75oC.These results were further integrated to obtain the enthalpy change (ΔH) of
the lysozyme samples. The values obtained in this study were around 429kJ/mol. For the lysozyme
sample test using the MEMS DSC, we prepared 5 MEMS DSC devices and each device was used
only once to evaluate the reproducibility. While for the test using the TA discovery DSC, the
experiment was repeated five times. The results are shown in Fig. 5-6. The performance
comparison of the MEMS DSC and the TA discovery DSC is shown in Table 9.
Table 9. Lysozyme test results using both the commercial TA discovery DSC and the MEMS DSC.
Lysozyme MEMS DSC TA discovery DSC
concentration 10gm/ml 10mg/ml
buffer 15mM histidine, pH6 15mM histidine, pH6
Scanning rate 10oC/min 10oC/min
Sample volume 0.65µL 20µL
Transition temperature 75.6 ± 1.2 75.5 ± 0.35
Enthalpy change 429 ± 35𝑘𝐽/𝑚𝑜𝑙 428 ± 25𝑘𝐽/𝑚𝑜𝑙
114
Figure 5-5. The molar heat capacity change and enthalpy of the lysozyme over the temperature in
the scanning process using both the MEMS based DSC. (a) and (b) show the heat capacity change
of the lysozyme while (c) and (d) show the enthalpy change.
(b)
(c) (d)
(a)
115
Figure 5-6. The molar heat capacity change and enthalpy change of the lysozyme over the
temperature in the scanning process using the TA discovery DSC. (a) and (b) show the heat
capacity change of the lysozyme samples while (c) and (d) show the enthalpy change over the
scanning temperature.
The service life of the MEMS based DSC was further studied. For such high scanning rate
(10oC/min), the denaturated sample was easy to stick on the wall of the polymer chamber after
chamber cleaning. The protein residue will influence the accuracy of the test result and hence
shorten the service life of the DSC. Fig. 5-7 shows the lysozyme test results using the same MEMS
DSC device.
∆𝐻 = 428 ± 25𝑘𝐽/𝑚𝑜𝑙
(a)
(d)
(b)
(c)
116
Figure 5-7. Service life tests of the MEMS based DSC. (a) and (b) show two MEMS DSC devices
were reused in lysozyme tests.
The MEMS based DSC was used to further study the denaturation of the lysozyme sample with
different pH and concentration. After baseline subtraction, the linear regions on either side of the
transition peak, which represent the heat capacity of the native and denatured states of the protein,
is extrapolated into the transition and then merge them about the progression through the transition
to show the heat capacity change of the protein. Finally, the area under the resulting peak is
integrated to give the excess energy that the DSC requires to denature the protein in the sample
cell. Provided that the concentration of the protein solution and the operational volume of the
calorimeter cell are known, this energy can be converted into ΔH in calories or Joules per mol of
protein. The tests results are shown in Fig. 5-8 (different pH value tests) and Fig. 5-9 (different
lysozyme concentration).
(a) (b)
117
Figure 5-8. The lysozyme tests using the MEMS based DSC with different pH values.
Figure 5-9. The lysozyme sample tests using the MEMS based DSC with different concentrations.
(a) (b)
(a) (b)
118
5.3.3 The DSC Measurements of Two Antibodies.
The 10mg/mL monoclonal antibody (mAb 696, molar weight 148kD) and ABT 981 (a double
variable domain immunoglobulin (DVD-IgG), molar weight of 140.5kD) were prepared The
structure of the mAb 696, and ABT 981 is shown in Fig. 5-10. It contains one Fc (fragment
crystallizable) region and two Fab (fragment antigen-binding) regions. In the Fc region, there are
constant heavy chain 2 (CH2) and constant heavy chain 3 (CH3). In the Fab region, there are a
light chain (variable light (VL) and constant light (CL) and heavy chain (variable heavy (VH) and
constant heavy chain 1 (CH1)).
Figure 5-10. The structure of antibody ABT 981 and mAb 696
119
Figure 5-11. Differential scanning calorimetry thermogram of ABT 981. The unfolding transitions
for the individual domains of ABT-981 are separated.
Figure 5-12. Differential scanning calorimetry thermogram of mAb-696. The unfolding transitions
for the individual domains of mAb-696 are separated.
120
Figure 5-13. The enthalpy change over the scanning temperature for both AbT-981 and mAb-696.
The DSC thermogram of ABT 981 is shown in Fig. 5-11. The three thermal domains are separated.
The largest peak (peak 2) is corresponding to the Fab domain. The peak 1 and peak three are
corresponding to the CH2 domain and CH3 domain respectively. The CH3 domain is slightly
shifted to the left (lower temperature) may due to the uncertainty of the deconvolution of its
corresponding peak.
The DSC thermogram of mAb-696 is shown in Fig. 5-12. The three thermal domains are separated.
The largest peak (peak 3) is corresponding to the Fab domain. The peak 1 and peak two are
corresponding to the CH2 domain and CH3 domain respectively.
By integrating the heat capacity over temperature in both Fig. 5-11 and Fig. 5-12 we got the
enthalpy change over temperature for both mAb-696 and ABT-981 shown in Fig. 5-13. The DSC
measurements of the protein samples (lysozyme, mAb-696, and ABT-981) demonstrates the
121
proposed MEMS based DSC is effective to measure the thermodynamic parameters of proteins
with a complex structure (antibody).
5.4 Influence of the Response Time to the Kinetic Parameters of Protein
The deconvolution of heat capacity profiles into individual components as well as the application
of statistical thermodynamic methods of analysis requires access to an experimentally defined heat
capacity function whose shape accurately reflects the equilibrium properties of the process under
study. Unfortunately, DSCs introduce distortions in the shape of the heat capacity function as a
result of their slow instrumental time response [174, 175]. The 1st order response also introduces
the shift of the transition temperature. Typically, high-sensitivity DSCs have intrinsic time
responses ranging from 3 to 10 s, thus imposing severe restrictions on the upper scan rate limit that
can be used without introducing significant distortions in the heat capacity function[176]. Since
the calorimeter sensitivity is directly proportional to the scanning rate, it is impossible to arbitrarily
reduce the scanning rate to obtain distortion-free curves. The shape of the measured heat capacity
function is also affected by dynamic components arising from the transition kinetics and the
scanning rate. These effects are more prominent for slow reactions and can be used to obtain
kinetic parameters for the transition under study. Once the other hand, the thermally non-
equilibrium nature of the DSC also caused a difference between the measured transition
temperature and the “true” transition temperature. There were a lot of models developed to express
the relationship between the measured transition temperature and the scanning rate.
The MEMS based DSC is a first order system (eq.1). A step response test was used to calibrate the
calorimeter to measure the sensitivity and response time. It was operated by applying a certain
power to the micro heater for a certain time and measuring the output voltage signal. By dividing
122
the input voltage by the power applied, the sensitivity could be determined. The limited response
time will introduce distortion in the heat capacity function and lead to the shift of the transition
temperature. In the temperature scanning process, the temperature is linear to the time (Eq. 5-10).
Where C is the heat capacity of the sample/reference cell, T is the temperature of the cell, H is the
thermal conduction coefficient, 𝑇0 is the environment temperature, P is the heating power, 𝑣 is the
temperature scanning rate and b is inthe itial temperature of the cell. The ideal DSC curve of a
single domain protein denaturation is a Gaussian distribution in the temperature domain. However,
as the DSC instrument has lia mited response time, the DSC curve will be distorted. Fig.5-14
shows the input ideal DSC curve of lysozyme sample (10mg/ml, 10mM phosphate, pH7) measured
by the MEMS DSC and the outputs with rea sponse time of 3s, 6s and 9s. The scanning rate is
40oC/min. The transition temperature (Tm) shift (the peak temperature difference between the two
curves) increases as the response time increases. The relationship between the Tm shift and the
response time was further studied by manually changing the response time using simulation using
fithe rst-order model shown in Fig. 5-15. It can be seen that when the response time goes up to
10s, the Tm shift reaches to 3.6oC which should not be neglected. Since the temperature is linear to
time in the scanning process, for a given response time, a higher scanning rate can also lead to
higher Tm shift which put a limit to the scanning rate of the DSCs when the transition temperature
is a key parameter.
𝐶𝑑𝑇
𝑑𝑡+ 𝐻(𝑇 − 𝑇0) = 𝑃 (5-9)
𝑇 = 𝑣𝑡 + 𝑏 (5-10)
123
Figure 5-14. The temperature distortion induced by the response time.
Figure 5-15. The relationship between the response time and the Tm shift.
124
5. 5 Influence of the Scanning Rate to the Kinetic Parameters of Protein
Irreversible protein denaturation is thought to consist of two main steps: (a) reversible unfolding
of the native protein (N) to the unfolded protein (U), (b) irreversible alteration of the unfolded
protein (U) to yield a final state (F) that cannot be folded back to the native state (N). For the
processes responsible for the irreversible step (aggregation, chemical alteration of residues, etc.),
the two-step nature of irreversible denaturation is depicted in the following simplified scheme
which is usually known as the Lumry and Eyring model expressed in the following scheme.
𝑁𝑘1,𝑘2⇔ 𝑈
𝑘3⇒ 𝐹
For the irreversible protein denaturation in the DSC measurement, the Lumry and Eyring model
can also be simplified to a two-state case:𝑁𝑘⇒ 𝐹 as a first order transition when the reversible
unfolding is of less importance [177, 178]. 𝑘 is the rate constant which is related with the
temperature expressed by Arrhenius equation (Eq. 5-11). The relationship between the transition
temperature (Tm) and the scanning rate is derived from Eq. 5-11-17.
𝑘 = 𝑒𝐴−𝐸 𝑅𝑇⁄ (5-11)
The heat capacity change is linear to the reaction rate:
∆𝐶𝑝 = 𝐾1𝑑𝑁
𝑑𝑡 (5-12)
The denaturation rate can be expressed:
𝑑𝑁
𝑑𝑡= −𝑘𝑁 (5-13)
The temperature is linear to the scanning rate:
125
𝑇 = 𝑣𝑡 + 𝑏 (5-14)
The transition temperature is corresponding to the largest heat capacity change that:
𝑑𝑑𝑁
𝑑𝑡
𝑑𝑇= −𝑣 (
𝑑𝑘
𝑑𝑇𝑁 + 𝑘
𝑑𝑁
𝑑𝑇) = −𝑣𝑁 (
𝑑𝑘
𝑑𝑇− 𝑘2/𝑣) = 0 (5-15)
Combine eq. 3 and eq. 8, the relationship between the transition temperature(Tm) and the scanning
rate and can be expressed by Eq. 5-17.
ln (𝑣
𝑇𝑚2) = 𝐵 −
𝐶
𝑇𝑚 (5-17)
𝐴, 𝐵, 𝐶, 𝐾1 are constants. E is the activation energy, R is the gas constant. Eq. 5-17 shows that
irreversible DSC transitions are strongly rate dependent. Such relaa tionship is determined by the
temperature dependent nature of the rate constant (𝑘).
A set of experiments were carried out to verify the model. Lysozyme samples (molar weight is
14307 Da) at a concentration of 10 mg/ml were prepared. The buffer used was 15 mM Histidine
pH 6.0. The buffer control was filtered 15 mM Histidine pH 6. The irreversibility of the lysozyme
denaturation in the DSC measured was verified by changing temperature scanning. In each
measurement, the baseline drift was subtracted from the data to yield the net signal (Fig. 5-16).
𝑑𝑘
𝑑𝑇=𝑘2
𝑣 (𝑇 = 𝑇𝑚) (5 − 16)
126
Figure 5-16. The DSC measurements of lysozyme samples with different scanning rates.
127
Figure 5-17. Linear fitting between the ln (𝑣
𝑇𝑚2) and 1000/Tm.
The linear fitting between the ln (𝑣
𝑇𝑚2) and 1000/Tm (Fig. 5-17) suggests the model suits well in the
lysozyme measurement which demonstrates the two state Lumry and Eyring model can be used to
explain the Tm shift in the DSC measurements with different scanning rate for the irreversible
denaturation after data correction. Hence, for a raw DSC data about irreversible denaturation, the
transition temperature is actually shifted from the real one. The Tm shift is determined by two
factors: the response time and scanning rate. Need to mention that the response time induced Tm
shift is also closely related with the scanning rate since the temperature is linear to time in the
scanning process.
In this section, firstly, the response time induced distortion of the heat capacity curve was studied.
It is possible to remove instrumental distortions arising from the finite time response of the
128
calorimeter and obtain correctly shaped heat capacity profiles even at very fast scanning rates.
Then, the Lumry-Eyring models are used to further study the relationship between the scanning
rate and the transition temperature after correction of the DSC curve for the irreversible lysozyme
denaturation. This study can be used for the guidance for MEMS based DSC array design and
correction of the transition temperature in protein denaturation study as well as the shape of the
heat capacity thermogram.
5.6 Thermal Stability Study of a Fab, a mAb and the Corresponding DVD Ig Using the
MEMS-Based DSC
5.6.1 Background
Since the health authority approval of the first monoclonal antibody in 1986 for the treatment of
organ rejection, monoclonal antibodies and antibody-based therapeutics have dominated the
biopharmaceutical market [179]. There is an increasing interest in understanding the factors that
affect the stability of antibodies as the therapeutic antibody market grows rapidly [180].
The mAb consists of two identical heavy chains and two identical light chains. Each chain folds
into different domains. There are two domains in the light chain: constant light (CL) and variable
light (VL) and four domains in the heavy chain: variable heavy (VL) and constant heavy (CH1,
CH2, and CH3) [181]. Papain digestion of mAb can divide the molecule into two fragments: the
Fab fragment and the fragment of the crystallizable region (Fc fragment) [182]. In one particular
format of bispecific antibodies, two pairs of variable regions, on outer and inner domains of the
Fab, tare linked by peptides of variable length to form dual variable domain immunoglobulins
(DVD-Ig). As a result, the DVD-Ig is capable of simultaneously binding two separate antigens that
are related to a particular disease state. Therefore, DVD Ig with bispecific activities has the
129
potential to improve clinical efficacy through simultaneous targeting of dual antigens and provide
a more extensive modification of biochemical pathways related to a particular disorder (Fig. 5-18)
[183].
Thermal stability is a direct assessment of the conformational integrity of proteins. [184]. Poor
conformational stability may lead to tertiary structure misfolding in the variable regions, a decrease
in antigen binding and aggregation in solution. Moreover, the conformational stabilities of mAbs
and DVD-Igs are directly related to their solution environments which include pH and various
cosolutes. Solution conditions impact the tertiary structures by modifying ionization states of
amino acids, direct interactions, and hydration of the surfaces of proteins [185].
When antibodies or DVD-Igs are subjected to increasing temperature, the native tertiary state
transits to a fully denatured or unfolded state. DSC measures temperature required for this
transition (Tm) along with the enthalpy needed to achieve the unfolded state (∆H). Generally, the
higher the Tm, the more stable the protein under specified conditions. As a label-free and
immobilization-free method, DSC simultaneously evaluates the temperature (Tm) and energy
required for the protein to transit from the native to denatured states thus providing a direct
understanding of molecular interactions of the bio-samples [186]. Consequently, DSC is a critical
tool for protein engineers to assess the intrinsic conformational robustness based on amino acid
sequence and for the formulation scientist to evaluate conformational stability in solution [11].
Since the DVD-Ig Fab are comprised of the two variable domains from separate mAbs, their
conformational stabilities and thermal profiles are more complex. During DVD-Ig engineering,
the precursor variable regions are selected from mAbs that demonstrate relatively high
conformational stabilities prescreened by DSC. With the advent of automated and high-throughput
130
cell culture processes, many hundreds of mAbs and DVD-Igs may be produced over a short time
frame. Consequently, it is advantageous to develop thermal analysis tools that may screen
biomolecules quickly and simultaneously while using little material in order to obtain relative
conformational differences between therapeutic candidates.
In this section, we reported the thermal stability study of a model mAb, its Fab fragment and the
corresponding model DVD IgG (the molecular structure is shown in Fig. 1) using the MEMS based
DSC. The MEMS based DSC is being developed as an ultrasensitive, high throughput calorimeter
which requires microgram quantities of sample to provide a thermal denaturation profile
comparable with standard instruments. The DSC results showed the thermal domain of the Fab
fragment matches the main peak in the intact mAb (corresponding to the Fab domain) which means
the temperature induced denaturation of the Fab in the intact mAb is independent. The transition
temperatures of the DVD IgG domains are approximately 10oC lower compared with the mAb
which indicates that the overall conformational stability is lower as a result of additional
interacting domains. These results are comparable with the conformational differences detected
using the standard DSC. Consequently, the MEMS based DSC is capable of generating
comparable results to the conventional instrument while using a fraction of the sample
requirements and at a faster thermal scan rate.
The MEMS based DSC was fabricated on a polyimide flexible membrane. A PDMS/Flexdyne
microfluidic device with 0.63 µL volume for thermal insulation and evaporation suppression was
bonded to the substrate to form the reaction chamber for the sample and reference materials. The
MEMS based DSC used vanadium oxide thermistors to monitor the temperature difference
between the reference and sample side which can be converted to power difference between them
131
after the sensitivity calibration. A temperature controlled heating stage was used to linearly raise
the temperature and provide thermal shielding. The MEMS based DSC has a higher scanning rate
(45oC/min) and lower sample concentration (0.63µL) compared with the commercial DSC. This
was the first time that a MEMS based DSC was used to study the structure factors that affect the
thermal stability of antibodies.
Figure 5-18. The molecular structure of the Fab, mAb and DVD IgG.
5.6.2 Denaturation Measurements of the Fab, mAb and the DVD Ig
The mAb (molar weight: 144.9kD), Fab (molar weight: 47.4kD) and DVD Ig (196.4kD) were
prepared at the AbbVie Bioresearch Center (Worcester, MA). The DVD-Ig was designed by
combining the CDRs of two precursor IgG1s into one dual-targeting protein using naturally
occurring peptide linkers to attach the domains to the constant framework (Figure 1). A template
was created from the variable domain DNA of the two precursor mAbs and cloned into an
appropriate expression vector. The mAb and DVD-Ig proteins were expressed in Chinese hamster
ovary (CHO) cells and grown in bioreactors. After an appropriate incubation period, the CHO
cells were harvested, lysed and the supernatant was purified by protein A chromatography and
132
anion filtration and diafiltered into a stabilizing buffer. The corresponding Fabs were prepared by
digesting the mAb with papain and purification by size-exclusion chromatography (SEC). The
variable domains of the DVD-Ig, mAb, and Fab are antigen specific for tetanus toxoid. These
proteins were dialyzed before the DSC measurements against the same buffer containing 15mM
histidine at pH 6. The concentrations of the intact mAb, the Fab region, and the DVD Ig were
5mg/ml.
The DSC measurements were carried out using the MEMS based calorimeter at 45oC/min scanning
rate from 30oC to 100oC. The heat capacity thermograms were obtained by subtracting the baseline
and interpolating a cubic baseline in the transition region. The final thermograms were normalized
to the molar concentration of each protein. The Tm represented the peaks in the experimental
thermograms.
The temperature-induced denaturation of the three protein samples (mAb, Fab and DVD Ig) was
shown in Fig. 5-19a. To verify the performance of the MEMS based DSC, we also determined the
thermal stability profiles of the three proteins using the commercial MicroCal VP-capillary DSC
(Malvern/MicroCal, Northampton, MA)(Fig. 5-19b). A sample volume of 300µL at 1 mg/mL was
analyzed at scanning rate of 1oC/min between 25⁰C - 95⁰C. Deconvolution of the thermal profiles
from both the VP-capillary DSC and MEMS DSC to obtain the transition temperatures and
calorimetric transition enthalpies was accomplished using Origin Pro 2015 software.
The performance of the MEMS based calorimeter to determine the temperatures and relative areas
of the thermal unfolding transitions of the Fab, mAb, and DVD were compared with the profiles
obtained from the commercial DSC (Fig 5-20). The Tms and relative calorimetric areas (∆Hcal,
the enthalpy) of the deconvoluted transitions were compared between the three proteins and both
133
instruments. Generally, both Tm and ∆Hcal are parameters that are used to evaluate the
conformational stability of proteins in solution. ∆Hcal is proportional to the endothermic energy
required to unfold the corresponding domain of the protein.
The Fab thermal denature profiles from both the MEMS and commercial calorimeters showed one
transition at approximately 80oC (Table. 8). As expected, the denaturation profiles for the mAb
from both the DSC and MEMS unit demonstrated three transitions. The Tms for the first and
second transitions obtained from both instruments were similar. However, the Tm3 for unfolding
profile of the mAb from the MEMS unit was overestimated by more than 10⁰C. Additionally, the
relative area of the second transition in the denaturation profile was overestimated by the MEMS
calorimeter. The more complex DVD-Ig was deconvoluted to reveal four thermal transitions by
both instruments. The Tm values for the transitions obtained from both instruments were similar
(Table 9). However, whereas the DSC showed that transition one and transition two likely
corresponded to the outer and inner domains of the bispecific Fab (requiring the most energy to
unfold), the areas from the MEMS calorimeter showed that transitions two and three had the
greatest areas.
Overall, the transition temperatures determined by both instruments are more similar compared
with the relative areas of domain denaturation (Fig. 5-20a, c, e and Fig. 5-20b, d, f). This is likely
caused by the differences in scan rate and sample amounts used in the thermal denaturation
experiments. The thermal denaturation profile of immunoglobulins and other proteins consists of
transitions associated with the individual subunits. The temperatures and enthalpies of unfolding
correspond to the conformational stabilities of these individual domains. Domains that have more
complexity and internal peptide contacts generally require more energy to unfold and have greater
134
enthalpies. The Fab domain (contains the variable or antigen binding region) of the bispecific or
monoclonal antibody will have a higher enthalpy of unfolding compared with the CH2 or CH3
regions (Fig. 5-18). Additionally, the profile shows the interactions between domains during
unfolding or cooperativity. Differences in thermal denaturation profiles may be caused by
variation in primary amino acid sequence between largely similar proteins or differences in
solution conditions such as pH or excipients for the same protein. Consequently, the comparison
between proteins based on transition temperatures (Tm) appears to be more suitable using the
MEMS calorimeter rather than enthalpies or interactions between domains. As a result, the Fab
region may be misidentified in thermal denaturation profiles using the MEMS calorimeter.
Figure 5-19. The DSC curves of the three proteins obtained by the MEMS based DSC. (b) The
DSC curves of the three proteins obtained by the MicroCal VP-Capillary DSC for comparison.
(a) (b)
135
Figure 5-20. The deconvolution of the normalized DSC curves for the three protein samples (Fig.
5-20a,c,e are the DSC curves measured by the MEMS DSC, Fig. 5-20b,d,f are the curves measured
by the VP DSC).
We demonstrated the Fab domain transition is independent of other domains in the mAb. We also
showed the transition temperature of the AB095 DVD IgG had much poorer stability compared to
(a) (b)
(c) (d)
(f) (e)
136
AB095 mAb which indicates binding additional variable domains greatly affected the thermal
stability of the antibody. The summary of the test results is shown in Table 10.
Table 10. MicroCal VP-capillary DSC test results VS the MEMS based DSC test results.
Sample Method Transition 1 Transition 2 Transition 3 Transition 4
Tm1 area
∆H1 %
Tm2
area
∆H2 %
Tm3
area
∆H3 %
Tm4
area
∆H4 %
Fab VP DSC 78.8 100% - - - - - -
MEMS DSC 79.5 100% - - - - - -
mAb VP DSC 70.7 12% 79.2 45% 81.3 43% - -
MEMS DSC 71.9 3% 81.1 95% 92.5 3% - -
DVD-
Ig
VP DSC 67.5 45% 74.4 31% 76.1 18% 82.2 7%
MEMS 64.3 16% 69.7 29% 74.7 34% 78.0 21%
5.7 Summary
In this chapter, we firstly studied the performance of the MEMS-based DSC and compared it with
the commercial one (TA Discovery DSC). The results show the MEMS DSC has a better baseline
repeatability (dynamic noise level) than the TA DSC. The service life and reproducibility of the
MEMS DSC are characterized using lysozyme for the DSC measurements. The thermodynamic
parameters such as the transition temperature and enthalpy change are verified by the TA DSC.
We demonstrated that even the service life of single MEMS DSC device is short, the
reproducibility of difference DSC chips is excellent. The MEMS based DSC was used to further
study the denaturation of the lysozyme sample with different pH and concentration. The MEMS
DS was also used to study the effects of scanning rate and response time to the DSC curve. The
performance of the MEMS DSC in more complicated protein characterization was demonstrated
by using it to measure the denaturation of monoclonal antibodies and DVD Ig.
137
Chapter 6. Concluding Remarks
6.1 Thesis Summary
MEMS based DSC with high sensitivity, fast response and low sample consumption are developed
for the thermodynamic study of biomolecular samples. They are normally based on thin films for
enhanced thermal insulation and optimized sensing materials for higher sensitivity. Meanwhile,
the pharmaceutical industry demands high-performance calorimeters with higher throughput and
lower cost. There are mainly two challenges to develop MEMS DSC with high performance. One
is the evaporation problem of the liquid sample due to the large surface to volume ratio. Any slight
evaporation will introduce large noise to the output and cause an error for the heat capacity
measurement. The other challenge is the ultra-small net signal in the denaturation process duo to
the small sample consumption. Hence it requires the MEMS DSC has high power resolution. The
high resolution can be achieved by increasing the thermal insulation and the temperature sensitivity
of the MEMS based DSC device. The heat released during the denaturation process was ultralow,
which required the highly sensitive sensor to detect it.
This research developed an ultra-sensitive MEMS based DSC for the thermodynamic study of bio-
molecular samples. First, the overall concept design was proposed which consists of 4 key
components: The microfluidic device for sample loading, the thermistor for power detection, the
micro heater for sensitivity characterization and the heating stage for linear temperature scanning.
Based on the working principles of DSC, 3 ways can be used to increase the detection limit:
increase the scanning rate, increase the sensitivity and increase the resolution. To be specific, the
main contributions of this PhD work are:
138
1. A novel fabrication process was developed to fabricate the polyimide substrate. Polyimide
is used to replace the rigid substrate such as Si3N4 duo to its robustness, inertness in
chemical reactions and ease fabrication process. The polyimide also serves as the dielectric
layer and protective layer for the MEMES DSC device.
2. A double layer design was developed to fabricate the microfluidic device. The double layer
design introduced the top and side air gaps to increase the thermal resistance of the DSC
system and also isolate the system from the impact of environment noise. Further more, in
this double layer design, the Flexdym was used as the bottom layer to fabricate the chamber
and side air gap duo to its extremely low gas permeability. Traditionally, PDMS was
widely used in microfluidic device fabrication. However, these devices tend to suffer from
the high gas permeability at high temperature.
3. VOx based thermistor was developed for sensitive power/temperature detection. The VOx
thin film fabricated using DC sputtering and oxidation has a moderate resistivity (~10Ω𝑐𝑚)
and high TCR (-2.51% per oC). The new comb design was used to simply the fabrication
process and further reduce the resistance of the resistor to 10kΩ.
4. A novel double spiral design was developed for the micro heater fabrication. This design
optimized the power distribution along the sample area so the temperature distribution can
be uniform for the sensitivity characterization.
5. The performance of the MEMS DSC was verified by comparing the lysozyme test results
with the commercial TA discovery DSC. The results fit each other well which indicate the
high accuracy of the MEMS DSC. The new DSC was further used to study the thermal
stability of lysozyme under various conditions (different concentrations, pHs and scanning
139
rates). Protein samples with more complex structures such as mAb, Fab and IgG were also
studied using the MEMS DSC. The results showed the structure of the antibody has impact
on the thermal stability. Chapter 2 presents the working principle, design and fabrication
of the calorimeter system. The MEMS based calorimeter utilizes polyimide suspended thin
membrane for thermal isolation and Vanadium Oxide thermistors which further with a
variable resistor to form a Wheatstone bridge. The flexible substrate material, polyimide,
is utilized for its robustness, low thermal conductance, and easy fabrication process. Three
different microfluidic devices designed are presented and compared to determine the
optimal design for high thermal insulation.
In chapter 3, the sensing material and its characterization will be introduced. It first starts by a
theoretical introduction to thermistor sensing basics, thin film resistivity characterization method
and semiconductor conduction as well as TCR theories. Following this is the VOx related topics.
We explored the optimized condition and parameter for PVD deposition of VOx, to satisfy the
requirement for this application: high TCR and low resistivity. To better understand the material,
SEM and XRD measurement are also conducted. We also extended the research topic to the
optimization of VOx thin film fabrication and characterization. In this chapter, the metal-insulation
transition of VO2 grown under certain conditions.
In chapter 4, the performance of the MEMS based DSC is studied. In the first section, the
optimization of the distance between the sample and reference heaters is carried out by conducting
the heat transfer analysis. The second part of chapter 4 deals with the characterization of the
MEMS based DSC. After the thermistor being calibrated, it is used for thermal response study.
This quantitatively leads to the device’s major parameters: time constant, sensitivity and power
140
resolution. Both step response and impulse response were carried out. These results demonstrated
the calorimeter to be suited and capable of ultralow power bio-sensing. In the last part, the crack
in the micro heater trace is studied.
Chapter 5 presents the MEMS based DSC in various protein sample tests. The chapter firstly
introduces the data analysis method used in protein sample DSC data analysis. The performance
of the MEMS based DSC is verified by comparing the lysozyme test results using both the MEMS
based DSC and the TA discovery DSC. The baseline tests suggest that the MEMS based DSC has
much higher resolution and lower signal drift during the scanning process. The MEMS based DSC
was further used to study the lysozyme samples with different pH and concentration. The influence
of the response time of the DSC and the scanning rate to the kinetics of the DSC is then studied.
In the last section of Chapter 5, a systematic study of the thermal stability of a Fab, the mAb and
the corresponding DVD Ig is conducted to study the effect of the structure to the thermal stability
of proteins. This is the first time that A MEMS based DSC is used to study the effect of the structure
to the thermal stability of the complex antibody. The micro-DSC had the potential to characterize
the thermal stability of protein with significantly higher throughput by enabling arrays operation
and with lower sample consumption, which could mean a great reduction in cost and time for the
drug formulation in the pharmaceutical industry.
6.2 Future Work
6.2.1 MEMS based Isothermal Titration Calorimeter (ITC) Development
ITC is an important tool for direct thermodynamic characterization of biomolecular reactions
which allows direct determination of the binding affinity, stoichiometry, and entropy and enthalpy
141
of the binding reaction in solution, without the need to use labels [187]. It works by directly
measuring the heat that is either released or absorbed during a biomolecular reaction between two
mixed samples. It consists of two cells which are enclosed in an adiabatic jacket. The compounds
to be studied are placed in the sample cell, while the other cell, the reference cell, is used as a
control and contains the buffer in which the sample is dissolved [188]. During the measurement
process, sample and reference solutions are introduced to their respective measurement chambers,
and the difference in the thermal power from the reaction is measured. ITC is known to be the only
technique that can simultaneously determine all reaction associated binding parameters in a single
experiment, and is label-free and solution-based, requiring no molecular labeling or surface
immobilization. While widely used in basic biochemical studies, as well as practical applications
such as drug discovery and biotherapeutics development.
ITC was first described as a method for the simultaneous determination of binding affinity (𝐾𝑎)
and enthalpy change (∆𝐻) about 50 years ago by Christensen and Izatt [189] [190]. The ITC was
originally applied to a variety of weak acid–base equilibria and to metal ion complexation reactions
[191] [192]. These systems could be studied with the calorimetric instrumentation available at the
time which was limited to the determination of equilibrium constant (𝐾𝑒𝑞), values less than about
104 − 105𝑀−1 [193]. However, the sensitivity is low that the determination of larger association
constants requires more dilute solutions. Beaudette and Langerman published one of the first
calorimetric binding studies of a biological sy/stem using a small sample consumption (0.3µmol)
isothermal titration calorimeter [194]. While in 1979, Langerman and Biltonen published a paper
about microcalorimeters for biological chemistry. In the paper, they discussed about available
instrumentation, applications, experimental design, and data analysis and interpretation [195]. This
was the beginning of the application of titration calorimetry in study biological reaction study. The
142
first commercial titration calorimeter specifically designed for the study of biological systems was
available from MicroCal 10 years later [23]. ITC is now widely used to characterize the
thermodynamics of biopolymer binding interactions [196] due to theimprovements in ITC
instrumentation and integration of data analysis software. Modern ITC instruments make it
possible to measure heat exchange as small as 0.4 mJ [197], allowing the determination of binding
constants as high as 109𝑀−1. Improvements in modern micro-calorimeters such higher sensitivity,
faster rthe esponse makes it also be able to directly characterize the kinetic parameters for an
enzymes [198]. In recent years, a lot of papers published about the data analysis and new designs
of ITC [199] [77] [200] [201].
Conventional ITC instruments are limited by high cost, complicated design and construction, low
throughput, and slow response time [202] [203]. Microelectromechanical systems (MEMS)
technology can be used to significantly reduce the size of the ITC. However, it is hard to achieve
proper control of reaction conditions of biomolecular reaction systems. The previous flow-through
based MEMS calorimetric devices have limited sensitivity due to poor thermal isolation [204].
Other designs such as the open-type MEMS based DSC, it suffers the variation of reaction volumes
and the evaporation induced noise [205].
Inspired by the MEMS based DSC, we propose a novel MEMS based ITC device design (Fig. 6-
1). The device consists of two microfluidic chambers on a freestanding polyimide diaphragm and
surrounded by air cavities for thermal isolation. The chambers are each equipped with a thin-film
gold resistive heater and temperature sensitive thermistor. The two thermistors together with the
two external decade resistor boxes form a Wheatstone bridge to detect the temperature difference
between the two chambers. Compared to silicon, polyimide is much cheaper, and its thermal
143
conductivity is much lower. The fabrication process, increase the fabrication yield, reduce the
device cost, and improve the device reliability. During the measurement process, the flow rate
ratio is changed to study the heat release for different molar ratios of the two mixed samples.
Design and principle
The proposed MEMS based ITC for biomolecular reaction study consists of the PDMS/Flexdyne
double-layer microfluidic device, the micro heaters, thermistors, and the polyimide layers). There
are two identical Flexdyne based microfluidic chambers with an identical volume of 0.63µL,
which hold the sample and reference materials for calorimetric measurements. These chambers
referred to as the sample and reference chambers, respectively, are each connected to the inlets
through the highly efficient mixer and outlet ports. The polyimide diaphragm, along with air
cavities surrounding the chambers, provides high thermal isolation that enables sensitive
calorimetric measurements. The MEMS based ITC device is composed of multiple material layers.
From the bottom to top, there is a polyimide substrate layer, a layer of thin film VOx based
thermistor embedded located underneath the centers of the calorimetric chambers to measure the
temperature difference between the sample and reference solutions, a polyimide-based dielectric
layer, a layer of thin-film resistive micro heater, under each chamber center and a polyimide-based
layer for package. The micro heater provides heating to the calorimetric chambers to generate a
constant differential power for calorimetric calibration. VOx based thermistors are chosen for
differential temperature sensing due to their high TCR and ease of fabrication. Polymers are used
to construct the calorimetric chambers to facilitate thermal isolation. In particular, polyimide is
chosen as the diaphragm material due to its excellent mechanical stiffness (Young’s modulus: 2.5
GPa), thermal stability (glass transition temperature: 285 °C) and low thermal conductivity
144
(0.12W/mK), while PDMS and Flexdym are used to fabricate the calorimetric chambers for its
ease of fabrication and packaging, low thermal conductivity.
Figure 6-1. (a) The microfluidic device with a mixer. (b) The detailed structure of the micro mixer.
(c) The temperature sensor on the polyimide substrate.
This device integrates a microfluidic reactor that features efficient passive chaotic mixing as well
as thermally isolated, well-defined reaction-reference chambers on a robust polymer-based
thermoelectric sensing chip. The MEMS based ITC allows biomolecules to mix thoroughly, react
without fluid evaporation, real-time monitoring and controlling of the device temperature as well
as sensitive differential temperature sensing. In an ITC experiment for the characterization of a
ligand-receptor system, the ligand with known concentration and volume is titrated or added into
a known concentration of a receptor solution (Fig. 6-2).
Sample 1
Sample 2
buffer
(a)
(b)
(c)
145
Figure 6-2. A simplified model of a typical receptor/ligand-binding interaction. The ligand in this
representation is shown to geometrically match the binding site on the receptor to indicate a
specific binding interaction [206].
Figure 6-3. (a) The recording of the heat release for multiple titrations with the different molar
ratio. (b) The integration of all the peaks in Fig. 6-3a.
Before entering their respective chamber, a series of liquid segments of different concentration of
ligand in a constant volume at a controlled temperature are allowed to react with the same volume
of the receptor and a pure buffer, respectively. The differential power generated (i.e., the thermal
power difference between the reaction and reference chambers), can be represented as s rP P P
(a) (b)
146
, where Ps and Pr are the thermal power in the reaction and reference chambers, respectively. This
differential power can be determined by:
∆𝑃 =∆𝑈
𝑆 (6 − 1)
ΔU is the output of the thermopile and S is the power sensitivity of the ITC. The total heat released
or absorbed in the calorimetric cell on each injection of reagent is determined by integrating the
differential power generated over time (Eq.6-2). The molar heat capacity change can be derived
from Eq. Three where [𝐿] is the concentration of the ligand, Vs .is the volume of the chamber. The
equilibrium constant is determined by eq.6-4. The Gibbs free energy and the entropy change are
then determined by Eq. 6-5 and Eq. 6-6 respectively.
∆𝑄 = ∫∆𝑃𝑑𝑡 (6 − 2)
.∆𝑄 = ∆𝐻 × [𝐿] × 𝑉𝑠 (6 − 3)
𝐾𝑒𝑞 =[𝐶𝑜𝑚𝑝𝑙𝑒𝑥]
[𝑅𝑒𝑐𝑒𝑝𝑡𝑜𝑟][𝐿𝑖𝑔𝑎𝑛𝑑] (6 − 4)
∆𝐺 = −𝑅𝑇𝑙𝑛(𝐾𝑒𝑞) (6 − 5)
∆𝑆 =∆𝐻 − ∆𝐺
𝑇 (6 − 6)
6.2.2 MEMS-Based DSC Sensor Array Development.
The demand for fast, reliable and continuous measurements of chemical species drug development,
biotechnology, and environmental sciences has evolved the need for small, easy to handle and
inexpensive analytic sensors. In medical application and metabolism monitoring of cell cultures,
it is highly favored to reduce the sample consumption. One way to reduce sample consumption is
147
the miniaturization of the device. It is also of great interest to develop sensor array to achieve
parallel tests to increase the throughput. This is of vital importance for the drug screening in new
drug design. Higher throughput thermodynamic measurements are providing value in structure-
based drug discovery, in particular throughout the fragment-based drug discovery process where
enthalpy can be used to detect and characterize ligand binding. On the other hand, the integration
of small dimensioned biosensors causes well-known problems like the chemical cross-talk
between adjacent biosensors [207].
For the proposed MEMS based DSC sensor array, the main cross-talk between the two sensor units
is the heat conduction between the adjacent sensor unit. In Chapter 4.2, the temperature distribution
on the polyimide thin film is analyzed to study the heat conduction between the micro heaters. The
simulation results show the heat conduction between the two heaters can affect the temperature in
the heating area. As the distalnce between the heaters becomes smaller, the heat conduction
increases which leads to a larger peak temperature difference. In the real protein samples tests, if
the distance is too small, the heat release during the protein denaturation process can not only cause
a temperature shift in the sample cell but also generate a smaller temperature shift in the reference
cell. While the released heat is proportional to the temperature difference between the sample and
the reference cell, the temperature shift in the reference area caused by heat conduction will put
additional error to the DSC measurements. It is therefore important to choose a proper distance
between the sample and reference cells so that the heat conduction between them can be negligible.
However, it is desirable to minimal the distance to achieve a compact design of the MEMS DSC
arrays. The optimized distance between the sample and reference cell was 4mm.
The proposed calorimeter array is shown in Fig. 6-4.
148
Figure 6-4. (a) The schematic diagram for the MEMS based calorimeter array. (b) The single unit
of the calorimeter array.
(a) (b)
149
Reference
[1] E. S. Watson, J. Justin, N. Brenner, and M. J. Oneill, "Differential Scanning Calorimeter
for Quantitative Differential Thermal Analysis," (in English), Analytical Chemistry, vol.
36, no. 7, pp. 1233-&, 1964.
[2] M. T. DeCrosta, J. B. Schwartz, R. J. Wigent, and K. Marshall, "Thermodynamic analysis
of compact formation; compaction, unloading, and ejection. I. Design and development of
a compaction calorimeter and mechanical and thermal energy determinations of powder
compaction," Int J Pharm, vol. 198, no. 1, pp. 113-34, Mar 30 2000.
[3] V. Plotnikov et al., "An autosampling differential scanning calorimeter instrument for
studying molecular interactions," Assay Drug Dev Technol, vol. 1, no. 1 Pt 1, pp. 83-90,
Nov 2002.
[4] M. Sun, G. W. Reed, and J. O. Hill, "Modification of a whole room indirect calorimeter
for measurement of rapid changes in energy expenditure," J Appl Physiol (1985), vol. 76,
no. 6, pp. 2686-91, Jun 1994.
[5] R. P. Layton, W. H. Mints, Jr., J. F. Annis, M. J. Rack, and P. Webb, "Calorimetry with
heat flux transducers: comparison with a suit calorimeter," J Appl Physiol Respir Environ
Exerc Physiol, vol. 54, no. 5, pp. 1361-7, May 1983.
[6] S. Y. Wang et al., "A power compensated differential scanning calorimeter for protein
stability characterization," (in English), Sensors and Actuators B-Chemical, vol. 256, pp.
946-952, Mar 2018.
[7] E. Turi, Thermal characterization of polymeric materials. Elsevier, 2012.
[8] H. Huth, A. A. Minakov, and C. Schick, "Differential AC-chip calorimeter for glass
transition measurements in ultrathin films," (in English), Journal of Polymer Science Part
B-Polymer Physics, vol. 44, no. 20, pp. 2996-3005, Oct 15 2006.
[9] H. Huth, A. A. Minakov, A. Serghei, F. Kremer, and C. Schick, "Differential AC-chip
calorimeter for glass transition measurements in ultra thin polymeric films," (in English),
European Physical Journal-Special Topics, vol. 141, pp. 153-160, Feb 2007.
[10] J. B. Chaires, "Calorimetry and thermodynamics in drug design," (in English), Annual
Review of Biophysics, vol. 37, pp. 135-151, 2008.
[11] G. Bruylants, J. Wouters, and C. Michaux, "Differential scanning calorimetry in life
science: Thermodynamics, stability, molecular recognition and application in drug design,"
(in English), Current Medicinal Chemistry, vol. 12, no. 17, pp. 2011-2020, 2005.
[12] J. Holm, L. N. Babol, N. Markova, A. J. Lawaetz, and S. I. Hansen, "The interrelationship
between ligand binding and thermal unfolding of the folate binding protein. The role of
self-association and pH," (in English), Biochimica Et Biophysica Acta-Proteins and
Proteomics, vol. 1844, no. 3, pp. 512-519, Mar 2014.
[13] C. E. Chang, W. L. You, M. Raymundo, and Z. Y. Tang, "Modeling ligand-protein binding
kinetics: the continuous and dynamical processes in ligand association/dissociation," (in
English), Abstracts of Papers of the American Chemical Society, vol. 251, Mar 13 2016.
[14] M. S. Kang, C. Roberts, Y. H. Cheng, and C. E. A. Chang, "Modeling of ligand-protein
association processes and crowding effect using multi-scale simulation methods," (in
English), Abstracts of Papers of the American Chemical Society, vol. 241, Mar 27 2011.
[15] B. G. Archer and H. Krakauer, "Thermodynamics of Antibody-Antigen Reactions .1.
Binding of Simple Haptens to 2 Classes of Antibodies Fractionated According to Affinity,"
(in English), Biochemistry, vol. 16, no. 4, pp. 615-617, 1977.
150
[16] M. B. Keown, A. J. Henry, R. Ghirlando, B. J. Sutton, and H. J. Gould, "Thermodynamics
of the interaction of human immunoglobulin E with its high-affinity receptor Fc epsilon
RI," (in English), Biochemistry, vol. 37, no. 25, pp. 8863-8869, Jun 23 1998.
[17] W. M. Aumiller and C. D. Keating, "Experimental models for dynamic
compartmentalization of biomolecules in liquid organelles: Reversible formation and
partitioning in aqueous biphasic systems," (in English), Advances in Colloid and Interface
Science, vol. 239, pp. 75-87, Jan 2017.
[18] T. Georgelin, M. Jaber, H. Bazzi, and J. F. Lambert, "Formation of Activated Biomolecules
by Condensation on Mineral Surfaces - A Comparison of Peptide Bond Formation and
Phosphate Condensation," (in English), Origins of Life and Evolution of Biospheres, vol.
43, no. 4-5, pp. 429-443, Oct 2013.
[19] A. R. Saksena, R. Saxena, R. Srivastava, and A. Gupta, "Thermodynamics of molecular
association of biomolecules with iodine : A study of molecular complex formation," (in
English), Journal of the Indian Chemical Society, vol. 87, no. 5, pp. 583-587, May 2010.
[20] G. Cannazza et al., "Development of an in vitro liquid chromatography-mass spectrometry
method to evaluate stereo and chemical stability of new drug candidates employing
immobilized artificial membrane column," (in English), Journal of Chromatography A, vol.
1363, pp. 216-225, Oct 10 2014.
[21] P. Baranczewski et al., "Introduction to in vitro estimation of metabolic stability and drug
interactions of new chemical entities in drug discovery and development," (in English),
Pharmacological Reports, vol. 58, no. 4, pp. 453-472, Jul-Aug 2006.
[22] L. M. Ahmad, B. Towe, A. Wolf, F. Mertens, and J. Lerchner, "Binding event measurement
using a chip calorimeter coupled to magnetic beads," (in English), Sensors and Actuators
B-Chemical, vol. 145, no. 1, pp. 239-245, Mar 4 2010.
[23] T. Wiseman, S. Williston, J. F. Brandts, and L. N. Lin, "Rapid Measurement of Binding
Constants and Heats of Binding Using a New Titration Calorimeter," (in English),
Analytical Biochemistry, vol. 179, no. 1, pp. 131-137, May 15 1989.
[24] N. Markova, R. O'Brien, and M. Arsenault, "Performance and quality. making
microcalorimetry simple with microcal peaq-itc," (in English), Protein Science, vol. 24, pp.
311-312, Oct 2015.
[25] B. Saad, A. Harris, P. Venugopal, and B. Milosavljevic, "TA, DSC and photophysical study
of phase transitions in glycerol-water mixture," (in English), Abstracts of Papers of the
American Chemical Society, vol. 251, Mar 13 2016.
[26] W. L. Goh et al., "Molecular Rotors As Conditionally Fluorescent Labels for Rapid
Detection of Biomolecular Interactions," (in English), Journal of the American Chemical
Society, vol. 136, no. 17, pp. 6159-6162, Apr 30 2014.
[27] M. Jagle, "A fast and highly sensitive nano-calorimeter," (in German), Technisches Messen,
vol. 70, no. 12, pp. 557-560, Dec 2003.
[28] E. Zhuravlev and C. Schick, "Fast scanning power compensated differential scanning nano-
calorimeter: 1. The device," (in English), Thermochimica Acta, vol. 505, no. 1-2, pp. 1-13,
Jun 10 2010.
[29] J. Koh, W. Lee, and J. H. Shin, "High-sensitivity chip calorimeter platform for sub-nano
watt thermal measurement," (in English), Sensors and Actuators a-Physical, vol. 241, pp.
60-65, Apr 15 2016.
151
[30] E. Shoifet, Y. Z. Chua, H. Huth, and C. Schick, "High frequency alternating current chip
nano calorimeter with laser heating," (in English), Review of Scientific Instruments, vol.
84, no. 7, Jul 2013.
[31] O. Nakabeppu and K. Deno, "Nano-DTA and nano-DSC with cantilever-type calorimeter,"
(in English), Thermochimica Acta, vol. 637, pp. 1-10, Aug 10 2016.
[32] L. O. Tremblay and A. Herscovics, "Cloning and expression of a specific human alpha 1,2-
mannosidase that trims Man(9)GlcNAc(2) to Man(8)GlcNAc(2) isomer B during N-glycan
biosynthesis," (in English), Glycobiology, vol. 9, no. 10, pp. 1073-1078, Oct 1999.
[33] N. E. Hager, "High-Speed Thermal-Analysis with Thin-Foil Calorimeter," (in English),
Review of Scientific Instruments, vol. 43, no. 8, pp. 1116-&, 1972.
[34] B. Frochte, Y. Khan, and E. Kneller, "A Simple High-Speed Calorimeter," (in English),
Review of Scientific Instruments, vol. 61, no. 7, pp. 1954-1957, Jul 1990.
[35] M. P. Bruijn, M. L. Ridder, L. Gottardi, H. F. C. Hoevers, P. A. J. de Korte, and J. van der
Kuur, "Advances in fabrication of TES mu-calorimeter arrays and associated filter
structures for AC-Biased read out," (in English), Journal of Low Temperature Physics, vol.
151, no. 1-2, pp. 500-505, Apr 2008.
[36] T. Saab et al., "Characterization and modeling of transition edge sensors for high resolution
X-ray calorimeter arrays," (in English), Nuclear Instruments & Methods in Physics
Research Section a-Accelerators Spectrometers Detectors and Associated Equipment, vol.
520, no. 1-3, pp. 281-284, Mar 11 2004.
[37] S. Kempf et al., "Design, fabrication and characterization of a 64 pixel metallic magnetic
calorimeter array with integrated, on-chip microwave SQUID multiplexer," (in English),
Superconductor Science & Technology, vol. 30, no. 6, Jun 2017.
[38] M. Arenton et al., "Development and Performance of a Calibration System for a Large
Calorimeter Array," (in English), Ieee Transactions on Nuclear Science, vol. 30, no. 1, pp.
193-197, 1983.
[39] A. A. Antoniou and G. G. Litvan, "Calorimeter for Simultaneous Measurement of Thermal
Properties and Dimensional Changes," (in English), Review of Scientific Instruments, vol.
38, no. 11, pp. 1641-&, 1967.
[40] G. Privalov, V. Kavina, E. Freire, and P. L. Privalov, "Precise Scanning Calorimeter for
Studying Thermal-Properties of Biological Macromolecules in Dilute-Solution," (in
English), Analytical Biochemistry, vol. 232, no. 1, pp. 79-85, Nov 20 1995.
[41] P. D. Ross and M. V. Rekharsky, "Thermodynamics of hydrogen bond and hydrophobic
interactions in cyclodextrin complexes," (in English), Biophysical Journal, vol. 71, no. 4,
pp. 2144-2154, Oct 1996.
[42] E. A. Mills and S. S. Plotkin, "Protein Transfer Free Energy Obeys Entropy-Enthalpy
Compensation," (in English), Journal of Physical Chemistry B, vol. 119, no. 44, pp. 14130-
14144, Nov 5 2015.
[43] L. F. Whiting, M. S. Labean, and S. S. Eadie, "Evaluation of a Capillary-Tube Sample
Container for Differential Scanning Calorimetry," (in English), Thermochimica Acta, vol.
136, pp. 231-245, Dec 1 1988.
[44] S. Maggiolino, N. Scuor, R. L. Mahajan, and O. Sbaizero, "MEMS_DSC: a new device for
microcalorimetric analysis in the biological field," (in English), Microsystem
Technologies-Micro-and Nanosystems-Information Storage and Processing Systems, vol.
16, no. 6, pp. 967-971, Jun 2010.
152
[45] G. Vanden Poel, D. Istrate, A. Magon, and V. Mathot, "Performance and calibration of the
Flash DSC 1, a new, MEMS-based fast scanning calorimeter," (in English), Journal of
Thermal Analysis and Calorimetry, vol. 110, no. 3, pp. 1533-1546, Dec 2012.
[46] J. C. Ingraham and G. Miller, "Infrared Calorimeter for Time-Resolved Plasma Energy
Flux Measurement," (in English), Review of Scientific Instruments, vol. 54, no. 6, pp. 673-
676, 1983.
[47] C. Hany, C. Pradere, J. Toutain, and J. C. Batsale, "A millifluidic calorimeter with infrared
thermography for the measurement of chemical reaction enthalpy and kinetics," (in
English), Qirt Journal, vol. 5, no. 2, pp. 211-229, Jul-Dec 2008.
[48] I. Hatta, "Characteristic and applicability of a temperature-modulated differential scanning
calorimeter for steady state measurement of heat capacity," (in English), Thermochimica
Acta, vol. 352, pp. 279-283, Jul 3 2000.
[49] Y. Saruyama and G. Van Assche, "Mathematical modeling of the thermal system of
modulated temperature differential scanning calorimeter," (in English), Thermochimica
Acta, vol. 391, no. 1-2, pp. 87-95, Aug 12 2002.
[50] M. G. R. T. Decooker and P. J. Vandenberg, "Direct Synthesis of Methylchlorosilanes -
Kinetic Measurements with a Gas-Flow Calorimeter," (in English), Recueil Des Travaux
Chimiques Des Pays-Bas-Journal of the Royal Netherlands Chemical Society, vol. 94, no.
8, pp. 192-195, 1975.
[51] C. J. Wormald, "A new gas phase flow mixing calorimeter: Test measurements on
(nitrogen plus cyclohexane)," (in English), Journal of Chemical Thermodynamics, vol. 29,
no. 6, pp. 701-714, Jun 1997.
[52] V. V. Babintsev et al., "A hybrid gas calorimeter summing signals from parallel-plate
ionization chambers," (in English), Instruments and Experimental Techniques, vol. 39, no.
4, pp. 487-491, Jul-Aug 1996.
[53] N. Tamai, Y. Nambu, M. Tsutsui, M. Goto, H. Matsuki, and S. Kaneshina, "Application of
pressure perturbation calorimetry to the aqueous system of phospholipid vesicle
dispersion," (in English), High Pressure Research, vol. 30, no. 4, pp. 490-498, 2010.
[54] N. Tamai, Y. Nambu, S. Tanaka, M. Goto, H. Matsuki, and S. Kaneshina, "Volumetric
characterization of ester- and ether-linked lipid bilayers by pressure perturbation
calorimetry and densitometry," (in English), Colloids and Surfaces B-Biointerfaces, vol.
92, pp. 232-239, Apr 1 2012.
[55] P. P. Pandharipande and G. I. Makhatadze, "Applications of pressure perturbation
calorimetry to study factors contributing to the volume changes upon protein unfolding,"
(in English), Biochimica Et Biophysica Acta-General Subjects, vol. 1860, no. 5, pp. 1036-
1042, May 2016.
[56] F. Haloua et al., "Thermal behaviour modelling of a reference calorimeter for natural gas,"
(in English), International Journal of Thermal Sciences, vol. 55, pp. 40-47, May 2012.
[57] I. S. Williams, R. Street, and E. S. R. Gopal, "Automated High Precision Calorimeter for
the Temperature-Range 200k-400k .1. Design and Performance," (in English), Pramana,
vol. 11, no. 5, pp. 519-527, 1978.
[58] R. C. Wilhoit, "Recent Developments in Calorimetry .2. Some Associated
Measurements .3. Some Specific Types of Calorimeters," (in English), Journal of
Chemical Education, vol. 44, no. 9, pp. A685-&, 1967.
153
[59] J. W. Judy, "Microelectromechanical systems (MEMS): fabrication, design and
applications," (in English), Smart Materials and Structures, vol. 10, no. 6, pp. 1115-1134,
Dec 2001.
[60] M. Esashi, "MEMS Devices: Progress Towards Practical Applications," (in English), Ieej
Transactions on Electrical and Electronic Engineering, vol. 4, no. 1, pp. 54-55, Feb 2009.
[61] D. M. Mattox, "Physical vapor deposition: The technical process with a relatively short
past & a long, exciting future," (in English), Plating and Surface Finishing, vol. 83, no. 11,
pp. 32-35, Nov 1996.
[62] H. O. Pierson, E. Randich, and D. M. Mattox, "Chemical Vapor-Deposition of Tib2 on
Graphite," (in English), Journal of the Less-Common Metals, vol. 67, no. 2, pp. 381-388,
1979.
[63] A. C. King et al., "The progress made using the Combustion Chemical Vapor Deposition
(CCVD) technique to fabricate YBa2Cu3O7-x coated conductors," (in English), Ieee
Transactions on Applied Superconductivity, vol. 13, no. 2, pp. 2643-2645, Jun 2003.
[64] G. M. Rebeiz, RF MEMS: theory, design, and technology. John Wiley & Sons, 2004.
[65] D. M. Mattox, "Plasma technology: Plasma chemistry, etching & deposition," (in English),
Plating and Surface Finishing, vol. 88, no. 5, pp. 106-108, May 2001.
[66] S. F. Yu, S. Y. Wang, M. Lu, and L. Zuo, "Review of MEMS differential scanning
calorimetry for biomolecular study," (in English), Frontiers of Mechanical Engineering,
vol. 12, no. 4, pp. 526-538, Dec 2017.
[67] S. Youssef, J. Podlecki, R. Al Asmar, B. Sorli, O. Cyril, and A. Foucaran, "MEMS
Scanning Calorimeter With Serpentine-Shaped Platinum Resistors for Characterizations of
Microsamples," (in English), Journal of Microelectromechanical Systems, vol. 18, no. 2,
pp. 414-423, Apr 2009.
[68] L. H. Allen and S. L. Lai, "MEMS-based scanning calorimeter for thermodynamic
properties of nanostructures," (in English), Microscale Thermophysical Engineering, vol.
2, no. 1, pp. 11-19, Jan-Mar 1998.
[69] T. J. Plegue, K. M. Kovach, A. J. Thompson, and J. A. Potkay, "Stability of Polyethylene
Glycol and Zwitterionic Surface Modifications in PDMS Microfluidic Flow Chambers,"
(in English), Langmuir, vol. 34, no. 1, pp. 492-502, Jan 9 2018.
[70] Y. F. Ju et al., "Fabrication of large-volume microfluidic chamber embedded in glass using
three-dimensional femtosecond laser micromachining," (in English), Microfluidics and
Nanofluidics, vol. 11, no. 1, pp. 111-117, Jul 2011.
[71] R. An, Y. Li, Y. Dou, D. Liu, H. Yang, and Q. Gong, "Water-assisted drilling of
microfluidic chambers inside silica glass with femtosecond laser pulses," (in English),
Applied Physics a-Materials Science & Processing, vol. 83, no. 1, pp. 27-29, Apr 2006.
[72] M. Alizadehgiashi, A. Gevorkian, M. Tebbe, M. Seo, E. Prince, and E. Kumacheva, "3D-
Printed Microfluidic Devices for Materials Science," (in English), Advanced Materials
Technologies, vol. 3, no. 7, Jul 2018.
[73] Y. Hong et al., "3D Printed Microfluidic Device with Microporous Mn2O3-Modified
Screen Printed Electrode for Real-Time Determination of Heavy Metal Ions," (in English),
Acs Applied Materials & Interfaces, vol. 8, no. 48, pp. 32940-32947, Dec 7 2016.
[74] S. Bibi et al., "A preliminary investigation to group disparate batches of licit and illicit
diazepam tablets using differential scanning calorimetry," (in English), Analytical Methods,
vol. 7, no. 20, pp. 8597-8604, 2015.
[75] A. San-Miguel and H. Lu, "Microfluidics as a tool for C. elegans research," 2005.
154
[76] E. A. Johannessen, J. M. R. Weaver, L. Bourova, P. Svoboda, P. H. Cobbold, and J. M.
Cooper, "Micromachined nanocalorimetric sensor for ultra-low-volume cell-based
assays," (in English), Analytical Chemistry, vol. 74, no. 9, pp. 2190-2197, May 1 2002.
[77] S. Keller, C. Vargas, H. Y. Zhao, G. Piszczek, C. A. Brautigam, and P. Schuck, "High-
Precision Isothermal Titration Calorimetry with Automated Peak-Shape Analysis," (in
English), Analytical Chemistry, vol. 84, no. 11, pp. 5066-5073, Jun 5 2012.
[78] A. Gourishankar, S. Shukla, K. N. Ganesh, and M. Sastry, "Isothermal titration calorimetry
studies on the binding of DNA bases and PNA base monomers to gold nanoparticles," (in
English), Journal of the American Chemical Society, vol. 126, no. 41, pp. 13186-13187,
Oct 20 2004.
[79] E. A. Johannessen, J. M. R. Weaver, P. H. Cobbold, and J. M. Cooper, "A Suspended
Membrane Nanocalorimeter for Ultralow Volume Bioanalysis," (in English), Ieee
Transactions on Nanobioscience, vol. 1, no. 1, pp. 29-36, Mar 2002.
[80] E. A. Olson et al., "Scanning calorimeter for nanoliter-scale liquid samples," (in English),
Applied Physics Letters, vol. 77, no. 17, pp. 2671-2673, Oct 23 2000.
[81] P. M. Sarro, A. W. Vanherwaarden, and W. Vandervlist, "A Silicon Silicon-Nitride
Membrane Fabrication Process for Smart Thermal Sensors," (in English), Sensors and
Actuators a-Physical, vol. 42, no. 1-3, pp. 666-671, Apr 15 1994.
[82] L. Zuo, X. M. Chen, S. F. Yu, and M. Lu, "Design and fabrication of a differential scanning
nanocalorimeter," (in English), Journal of Micromechanics and Microengineering, vol. 27,
no. 2, Feb 2017.
[83] J. W. Liu, Q. A. Huang, J. T. Shang, J. Song, and J. Y. Tang, "Micromachining of
Pyrex7740 Glass and Their Applications to Wafer-Level Hermetic Packaging of Mems
Devices," (in English), Mems 2010: 23rd Ieee International Conference on Micro Electro
Mechanical Systems, Technical Digest, pp. 496-499, 2010.
[84] L. Wang, L. Lei, X. Ni, J. Shi, and Y. Chen, "Patterning bio-molecules for cell attachment
at single cell levels in PDMS microfluidic chips," Microelectronic Engineering, vol. 86,
no. 4-6, pp. 1462-1464, 2009.
[85] A. Blake, T. Pearce, N. Rao, S. Johnson, and J. Williams, "Multilayer PDMS microfluidic
chamber for controlling brain slice microenvironment," Lab on a Chip, vol. 7, no. 7, pp.
842-849, 2007.
[86] I. D. Johnston, D. K. McCluskey, C. K. L. Tan, and M. C. Tracey, "Mechanical
characterization of bulk Sylgard 184 for microfluidics and microengineering," (in English),
Journal of Micromechanics and Microengineering, vol. 24, no. 3, Mar 2014.
[87] J. Lerchner et al., "A new micro-fluid chip calorimeter for biochemical applications," (in
English), Thermochimica Acta, vol. 445, no. 2, pp. 144-150, Jun 15 2006.
[88] V. Baier et al., "Highly sensitive thermopile heat power sensor for micro-fluid calorimetry
of biochemical processes," (in English), Sensors and Actuators a-Physical, vol. 123-24, pp.
354-359, Sep 23 2005.
[89] Y. Jia, B. Wang, Z. X. Zhang, and Q. Lin, "A polymer-based MEMS differential scanning
calorimeter," (in English), Sensors and Actuators a-Physical, vol. 231, pp. 1-7, Jul 15 2015.
[90] B. Wang, J. J. Lai, H. Li, H. M. Hu, and S. H. Chen, "Nanostructured vanadium oxide thin
film with high TCR at room temperature for microbolometer," (in English), Infrared
Physics & Technology, vol. 57, pp. 8-13, Mar 2013.
155
[91] Y. H. Han, K. T. Kim, H. J. Shin, S. Moon, and I. H. Choi, "Enhanced characteristics of an
uncooled microbolometer using vanadium-tungsten oxide as a thermometric material," (in
English), Applied Physics Letters, vol. 86, no. 25, Jun 20 2005.
[92] B. Wang and Q. Lin, "Temperature-modulated differential scanning calorimetry in a
MEMS device," (in English), Sensors and Actuators B-Chemical, vol. 180, pp. 60-65, Apr
2013.
[93] J. L. Ford and T. E. Mann, "Fast-Scan DSC and its role in pharmaceutical physical form
characterisation and selection," (in English), Advanced Drug Delivery Reviews, vol. 64, no.
5, pp. 422-430, Apr 2012.
[94] J. A. Dimasi, "New drug development in the United States from 1963 to 1999," Clin
Pharmacol Ther, vol. 69, no. 5, pp. 286-96, May 2001.
[95] C. H. Spink, "Differential scanning calorimetry," Methods Cell Biol, vol. 84, pp. 115-41,
2008.
[96] J. C. Yin, S. F. Yu, S. Y. Wang, M. Lu, and L. Zuo, "Design and Fabrication of Flexible
Differential Scanning Nanocalorimeter," (in English), Proceedings of the Asme
International Design Engineering Technical Conferences and Computers and Information
in Engineering Conference, 2014, Vol 4, 2014.
[97] Y. W. Mo, Y. Okawa, M. Tajima, T. Nakai, N. Yoshiike, and K. Natukawa, "Micro-
machined gas sensor array based on metal film micro-heater," (in English), Sensors and
Actuators B-Chemical, vol. 79, no. 2-3, pp. 175-181, Oct 15 2001.
[98] A. W. Vanherwaarden, P. M. Sarro, J. W. Gardner, and P. Bataillard, "Liquid and Gas
Micro-Calorimeters for (Bio)Chemical Measurements," (in English), Sensors and
Actuators a-Physical, vol. 43, no. 1-3, pp. 24-30, May 1994.
[99] P. Furjes, C. Ducso, M. Adam, J. Zettner, and I. Barsony, "Thermal characterisation of
micro-hotplates used in sensor structures," (in English), Superlattices and Microstructures,
vol. 35, no. 3-6, pp. 455-464, Mar-Jun 2004.
[100] S. Y. Wu, Q. Lin, Y. Yuen, and Y. C. Tai, "MEMS flow sensors for nano-fluidic
applications," (in English), Sensors and Actuators a-Physical, vol. 89, no. 1-2, pp. 152-
158, Mar 20 2001.
[101] S. Noh, J. Seo, and E. Lee, "The Fabrication by using Surface MEMS of 3C-SiC Micro-
heaters and RTD Sensors and their Resultant Properties," (in English), Transactions on
Electrical and Electronic Materials, vol. 10, no. 4, pp. 131-134, Aug 2009.
[102] J. F. Creemer et al., "Microhotplates with TiN heaters," (in English), Sensors and Actuators
a-Physical, vol. 148, no. 2, pp. 416-421, Dec 3 2008.
[103] D. Briand et al., "Micro-hotplates on polyimide for sensors and actuators," (in English),
Sensors and Actuators a-Physical, vol. 132, no. 1, pp. 317-324, Nov 8 2006.
[104] D. L. Huber, R. P. Manginell, M. A. Samara, B. I. Kim, and B. C. Bunker, "Programmed
adsorption and release of proteins in a microfluidic device," (in English), Science, vol. 301,
no. 5631, pp. 352-354, Jul 18 2003.
[105] U. Sokmen et al., "Capabilities of ICP-RIE cryogenic dry etching of silicon: review of
exemplary microstructures," (in English), Journal of Micromechanics and
Microengineering, vol. 19, no. 10, Oct 2009.
[106] D. Teyssieux, D. Briand, J. Charnay, N. F. de Rooij, and B. Cretin, "Dynamic and static
thermal study of micromachined heaters: the advantages of visible and near-infrared
thermography compared to classical methods," (in English), Journal of Micromechanics
and Microengineering, vol. 18, no. 6, Jun 2008.
156
[107] J. Zhang et al., "Fabrication of highly oriented reduced graphene oxide microbelts array
for massive production of sensitive ammonia gas sensors," (in English), Journal of
Micromechanics and Microengineering, vol. 23, no. 9, Sep 2013.
[108] J. Puigcorbe et al., "Thermal and mechanical analysis of micromachined gas sensors," (in
English), Journal of Micromechanics and Microengineering, vol. 13, no. 5, pp. 548-556,
Sep 2003.
[109] A. Zulfiqar, A. Pfreundt, W. E. Svendsen, and M. Dimaki, "Fabrication of polyimide based
microfluidic channels for biosensor devices," (in English), Journal of Micromechanics and
Microengineering, vol. 25, no. 3, Mar 2015.
[110] S. Metz, R. Holzer, and P. Renaud, "Polyimide-based microfluidic devices," (in English),
Lab on a Chip, vol. 1, no. 1, pp. 29-34, 2001.
[111] M. Aslam, C. Gregory, and J. V. Hatfield, "Polyimide membrane for micro-heated gas
sensor array," (in English), Sensors and Actuators B-Chemical, vol. 103, no. 1-2, pp. 153-
157, Sep 29 2004.
[112] S. Langereis et al., "A Temperature-Sensitive Liposomal H-1 CEST and F-19 Contrast
Agent for MR Image-Guided Drug Delivery," (in English), Journal of the American
Chemical Society, vol. 131, no. 4, pp. 1380-+, Feb 4 2009.
[113] E. Cozzani, C. Summonte, L. Belsito, G. C. Cardinali, and A. Roncaglia, "Design study of
micromachined thermal emitters for NDIR gas sensing in the 9-12 mu m wavelength
range," (in English), 2007 Ieee Sensors, Vols 1-3, pp. 181-184, 2007.
[114] S. Y. Wang et al., "Micro-differential scanning calorimeter for liquid biological samples,"
(in English), Review of Scientific Instruments, vol. 87, no. 10, Oct 2016.
[115] T. Simon, N. Barsan, M. Bauer, and U. Weimar, "Micromachined metal oxide gas sensors:
opportunities to improve sensor performance," (in English), Sensors and Actuators B-
Chemical, vol. 73, no. 1, pp. 1-26, Feb 25 2001.
[116] S. L. Chen, H. K. Ma, and D. Y. Chen, "Radiation Blockage by the Interaction of Thermal-
Radiation with Conduction and Convection in the Combustion of Condensed Fuels," (in
English), International Communications in Heat and Mass Transfer, vol. 20, no. 1, pp.
145-157, Jan-Feb 1993.
[117] G. K. Batchelor, "Heat Convection and Buoyancy Effects in Fluids," (in English),
Quarterly Journal of the Royal Meteorological Society, vol. 80, no. 345, pp. 339-+, 1954.
[118] D. V. Widder, The heat equation. Academic Press, 1976.
[119] H. B. Awbi, "Calculation of convective heat transfer coefficients of room surfaces for
natural convection," (in English), Energy and Buildings, vol. 28, no. 2, pp. 219-227, Oct
1998.
[120] V. V. Gavrilov and I. K. Fasakhov, "Specific Features of Thermocouple Calorimeter
Application for Measurements of Pulsed X-ray Emission from Plasma," (in English),
Plasma Physics Reports, vol. 38, no. 1, pp. 52-56, Jan 2012.
[121] J. Soeda, Y. Ikeda, and T. Shiro, "Solution-processing of flexible thin-film negative-
temperature-coefficient silicon thermistors using silicon nanoparticles," (in English),
Japanese Journal of Applied Physics, vol. 56, no. 7, Jul 2017.
[122] A. Kosarev, M. Moreno, A. Torres, and C. Zuniga, "IR sensors based on silicon-
germanium-boron alloys deposited by plasma: Fabrication and characterization," (in
English), Journal of Non-Crystalline Solids, vol. 354, no. 19-25, pp. 2561-2564, May 1
2008.
157
[123] M. Malmivirta et al., "The Angular Dependence of the Critical Current of BaCeO3 Doped
YBa2Cu3O6+x Thin Films," (in English), Ieee Transactions on Applied Superconductivity,
vol. 25, no. 3, Jun 2015.
[124] T. Inoue, T. Tanaka, M. Konagai, and K. Takahashi, "Electronic and Optical-Properties of
P-Type Amorphous-Silicon and Wide Band-Gap Amorphous-Silicon Carbide Films
Prepared by Photochemical Vapor-Deposition," (in English), Applied Physics Letters, vol.
44, no. 9, pp. 871-873, 1984.
[125] S. Y. Wang, S. F. Yu, M. Lu, M. Z. Liu, and L. Zuo, "Atomic Layer-Deposited Titanium-
Doped Vanadium Oxide Thin Films and Their Thermistor Applications," (in English),
Journal of Electronic Materials, vol. 46, no. 4, pp. 2153-2157, Apr 2017.
[126] C. X. Luo et al., "PDMS microfludic device for optical detection of protein immunoassay
using gold nanoparticles," (in English), Lab on a Chip, vol. 5, no. 7, pp. 726-729, 2005.
[127] L. C. Xiong, P. Chen, and Q. S. Zhou, "Adhesion promotion between PDMS and glass by
oxygen plasma pre-treatment," (in English), Journal of Adhesion Science and Technology,
vol. 28, no. 11, pp. 1046-1054, Jun 3 2014.
[128] J. A. Jofre-Reche, J. Pulpytel, F. Arefi-Khonsari, and J. M. Martin-Martinez, "Increased
adhesion of polydimethylsiloxane (PDMS) to acrylic adhesive tape for medical use by
surface treatment with an atmospheric pressure rotating plasma jet," (in English), Journal
of Physics D-Applied Physics, vol. 49, no. 33, Aug 24 2016.
[129] Y. Cao, J. Bontrager-Singer, and L. Zhu, "A 3D microfluidic device fabrication method
using thermopress bonding with multiple layers of polystyrene film," (in English), Journal
of Micromechanics and Microengineering, vol. 25, no. 6, Jun 2015.
[130] Z. L. Hu and X. Y. Chen, "Fabrication of Polyethylene Terephthalate Microfluidic Chip
Using CO2 Laser System," (in English), International Polymer Processing, vol. 33, no. 1,
pp. 106-109, Mar 2018.
[131] H. J. Chiang, S. L. Yeh, C. C. Peng, W. H. Liao, and Y. C. Tung, "Polydimethylsiloxane-
polycarbonate Microfluidic Devices for Cell Migration Studies Under Perpendicular
Chemical and Oxygen Gradients," (in English), Jove-Journal of Visualized Experiments,
no. 120, Feb 2017.
[132] A. M. D. Wan, D. Devadas, and E. W. K. Young, "Recycled polymethylmethacrylate
(PMMA) microfluidic devices," (in English), Sensors and Actuators B-Chemical, vol. 253,
pp. 738-744, Dec 2017.
[133] X. Wang, L. Y. Zhang, and G. Chen, "Hot embossing and thermal bonding of poly(methyl
methacrylate) microfluidic chips using positive temperature coefficient ceramic heater,"
(in English), Analytical and Bioanalytical Chemistry, vol. 401, no. 8, pp. 2657-2665, Nov
2011.
[134] Y. I. Jung, H. G. Kim, I. H. Kim, J. Y. Park, and W. J. Kim, "Microstructures of laser
bonded SiC ceramics with Zr interlayers," (in English), Journal of Nuclear Materials, vol.
455, no. 1-3, pp. 586-590, Dec 2014.
[135] M. A. Moyano and J. M. Martin-Martinez, "Surface treatment with UV-ozone to improve
adhesion of vulcanized rubber formulated with an excess of processing oil," (in English),
International Journal of Adhesion and Adhesives, vol. 55, pp. 106-113, Dec 2014.
[136] W. Xia, L. Li, K. F. Deng, S. Li, W. G. Su, and W. Zhang, "Benzo-cyclo-butene bonding
process with 'stamp' printing for wafer level package," (in English), Micro & Nano Letters,
vol. 9, no. 5, pp. 363-366, May 2014.
158
[137] X. Li, N. Q. Wu, Y. Rojanasakul, and Y. X. Liu, "Selective stamp bonding of PDMS
microfluidic devices to polymer substrates for biological applications," (in English),
Sensors and Actuators a-Physical, vol. 193, pp. 186-192, Apr 15 2013.
[138] J. Lachaux et al., "Thermoplastic elastomer with advanced hydrophilization and bonding
performances for rapid (30 s) and easy molding of microfluidic devices," (in English), Lab
on a Chip, vol. 17, no. 15, pp. 2581-2594, Aug 7 2017.
[139] J. Y. Andersson, P. Ericsson, H. H. Radamson, S. G. E. Wissmar, and M. Kolandouz,
"SiGe/Si quantum structures as a thermistor material for low cost IR microbolometer focal
plane arrays," (in English), Solid-State Electronics, vol. 60, no. 1, pp. 100-104, Jun 2011.
[140] V. F. Mitin et al., "Nanocrystalline SiC film thermistors for cryogenic applications," (in
English), Review of Scientific Instruments, vol. 89, no. 2, Feb 2018.
[141] A. I. Gubin, N. T. Cherpak, A. A. Lavrinovich, and K. V. Oganisian, "Temperature
dependence of the microwave conductivity of a YBCuO film in the normal state
Temperature dependence of the microwave conductivity of a YBCuO film in the normal
state," (in English), Low Temperature Physics, vol. 33, no. 10, pp. 818-820, Oct 2007.
[142] F. Goodenough, "Monolithic silicon temperature sensors challenge thermistors, RTDs, and
thermocouples," (in English), Electronic Design, vol. 45, no. 22, pp. 54-&, Oct 13 1997.
[143] J. Park, K. Ahn, S. Yu, J. An, T. H. Lee, and M. G. Kim, "Oxidation state control of
solution-processed vanadium oxide thin-films and resistive switching of VO2 thin-film in
a metastable state," (in English), Thin Solid Films, vol. 648, pp. 69-75, Feb 28 2018.
[144] D. L. Nguyen, Y. B. Taarit, and J. M. M. Millet, "Determination of the oxidation state of
antimony, iron and vanadium in mixed vanadium and iron antimonate oxide catalysts," (in
English), Catalysis Letters, vol. 90, no. 1-2, pp. 65-70, Sep 2003.
[145] S. Lee, T. L. Meyer, S. Park, T. Egami, and H. N. Lee, "Growth control of the oxidation
state in vanadium oxide thin films," (in English), Applied Physics Letters, vol. 105, no. 22,
Dec 1 2014.
[146] H. C. Wang, X. J. Yi, and S. H. Chen, "Low temperature fabrication of vanadium oxide
films for uncooled bolometric detectors," (in English), Infrared Physics & Technology, vol.
47, no. 3, pp. 273-277, Jan 2006.
[147] R. T. R. Kumar, B. Karunagaran, D. Mangalaraj, S. K. Narayandass, P. Manoravi, and M.
Joseph, "Properties of pulsed laser deposited vanadium oxide thin film thermistor," (in
English), Materials Science in Semiconductor Processing, vol. 6, no. 5-6, pp. 375-377,
Oct-Dec 2003.
[148] H. C. Wang, X. J. Yi, S. H. Chen, S. W. He, and X. C. Fu, "Reactive ion beams sputtering
of vanadium oxides films for uncooled microbolometer," (in English), International
Journal of Infrared and Millimeter Waves, vol. 26, no. 3, pp. 421-431, Mar 2005.
[149] Y. H. Han et al., "Fabrication of a surface micromachined uncooled microbolometer based
on the V2O5/V/V2O5 sandwich structure," (in English), Journal of the Korean Physical
Society, vol. 45, no. 6, pp. 1655-1658, Dec 2004.
[150] Y. H. Han, I. H. Choi, S. Moon, and C. S. Son, "Characteristics of vanadium-tungsten-
oxide bolometric thin films for uncooled IR detectors," (in English), Journal of the Korean
Physical Society, vol. 45, pp. S902-S905, Dec 2004.
[151] H. J. Pauwels, "Conductivity and Mobility Measurements in Space Charge Region with
Method of Van Der Pauw," (in English), Solid-State Electronics, vol. 14, no. 12, pp. 1327-
&, 1971.
159
[152] A. Kraft, "Advanced materials," (in English), Chemistry & Industry, vol. 76, no. 12, pp.
53-54, Dec 2012.
[153] K. P. Wheeler, "Reviews of modern physics," (in English), Online Information Review, vol.
24, no. 2, pp. 171-173, 2000.
[154] F. J. Morin, "Oxides Which Show a Metal-to-Insulator Transition at the Neel
Temperature," (in English), Physical Review Letters, vol. 3, no. 1, pp. 34-36, 1959.
[155] J. B. Goodenough, "The two components of the crystallographic transition in VO2,"
Journal of Solid State Chemistry, vol. 3, no. 4, pp. 490-500, 1971.
[156] A. Velichko, A. Pergament, V. Putrolaynen, O. Berezina, and G. Stefanovich, "Effect of
memory electrical switching in metal/vanadium oxide/silicon structures with VO2 films
obtained by the sol-gel method," (in English), Materials Science in Semiconductor
Processing, vol. 29, pp. 315-320, Jan 2015.
[157] Y. H. Jung and S. Chung, "Fabrication of vanadium oxide thin films from colloidal
vanadium oxide (VOx) nanocrystals for smart window applications," (in English),
Abstracts of Papers of the American Chemical Society, vol. 253, Apr 2 2017.
[158] W. Cheng, G. B. Zeng, and M. Niederberger, "Design of vanadium oxide core-shell
nanoplatelets for lithium ion storage," (in English), Journal of Materials Chemistry A, vol.
3, no. 6, pp. 2861-2868, 2015.
[159] C. D. Reintsema, E. N. Grossman, and J. A. Koch, "Improved VO 2 microbolometers for
infrared imaging: operation on the semiconducting-metallic phase transition with negative
electrothermal feedback," in Infrared Technology and Applications XXV, 1999, vol. 3698,
pp. 190-201: International Society for Optics and Photonics.
[160] V. Y. Zerov, Y. V. Kulikov, V. N. Leonov, V. G. Malyarov, I. A. Khrebtov, and I. I.
Shaganov, "Features of the operation of a bolometer based on a vanadium dioxide film in
a temperature interval that includes a phase transition," (in English), Journal of Optical
Technology, vol. 66, no. 5, pp. 387-390, May 1999.
[161] M. J. Miller and J. L. Wang, "Influence of grain size on transition temperature of
thermochromic VO2," (in English), Journal of Applied Physics, vol. 117, no. 3, Jan 21
2015.
[162] J. Jian et al., "Roles of grain boundaries on the semiconductor to metal phase transition of
VO2 thin films," (in English), Applied Physics Letters, vol. 107, no. 10, Sep 7 2015.
[163] T. H. Yang, S. Nori, H. H. Zhou, and J. Narayan, "Defect-mediated room temperature
ferromagnetism in vanadium dioxide thin films," (in English), Applied Physics Letters, vol.
95, no. 10, Sep 7 2009.
[164] K. Nagashima, T. Yanagida, H. Tanaka, and T. Kawai, "Stress relaxation effect on
transport properties of strained vanadium dioxide epitaxial thin films," (in English),
Physical Review B, vol. 74, no. 17, Nov 2006.
[165] G. Y. Song, C. Oh, S. Sinha, J. Son, and J. Heo, "Facile Phase Control of Multivalent
Vanadium Oxide Thin Films (V2O5 and VO2) by Atomic Layer Deposition and
Postdeposition Annealing," (in English), Acs Applied Materials & Interfaces, vol. 9, no.
28, pp. 23909-23917, Jul 19 2017.
[166] M. B. Sahana, G. N. Subbanna, and S. A. Shivashankar, "Phase transformation and
semiconductor-metal transition in thin films of VO2 deposited by low-pressure
metalorganic chemical vapor deposition," (in English), Journal of Applied Physics, vol. 92,
no. 11, pp. 6495-6504, Dec 1 2002.
160
[167] B. N. Masina et al., "Phase-selective vanadium dioxide (VO2) nanostructured thin films
by pulsed laser deposition," (in English), Journal of Applied Physics, vol. 118, no. 16, Oct
28 2015.
[168] P. Dagur, A. U. Mane, and S. A. Shivashankar, "Thin films of VO2 on glass by atomic
layer deposition: microstructure and electrical properties," (in English), Journal of Crystal
Growth, vol. 275, no. 1-2, pp. E1223-E1228, Feb 15 2005.
[169] C. X. Zhang, W. Cao, A. V. Adedeji, and H. E. Elsayed-Ali, "Preparation and properties
of VO2 thin films by a novel sol-gel process," (in English), Journal of Sol-Gel Science and
Technology, vol. 69, no. 2, pp. 320-324, Feb 2014.
[170] K. Y. Tsai, T. S. Chin, and H. P. D. Shieh, "Properties of VO2 films sputter-deposited from
V2O5 target," (in English), Japanese Journal of Applied Physics Part 1-Regular Papers
Short Notes & Review Papers, vol. 42, no. 7a, pp. 4480-4483, Jul 2003.
[171] W. Z. Dai, H. F. Yu, and R. Nassar, "A fourth-order compact finite-difference scheme for
solving a 1-D Pennes' bioheat transfer equation in a triple-layered skin structure," (in
English), Numerical Heat Transfer Part B-Fundamentals, vol. 46, no. 5, pp. 447-461, Nov
2004.
[172] W. H. Dai and R. Nassar, "A compact finite-difference scheme for solving a one-
dimensional heat transport equation at the microscale," (in English), Journal of
Computational and Applied Mathematics, vol. 132, no. 2, pp. 431-441, Jul 15 2001.
[173] S. F. Yu, S. Y. Wang, M. Lu, and L. Zuo, "A novel polyimide based micro heater with high
temperature uniformity," (in English), Sensors and Actuators a-Physical, vol. 257, pp. 58-
64, Apr 15 2017.
[174] B. Ibarra-Molero, A. N. Naganathan, J. M. Sanchez-Ruiz, and V. Muñoz, "Modern analysis
of protein folding by differential scanning calorimetry," in Methods in enzymology, vol.
567: Elsevier, 2016, pp. 281-318.
[175] O. L. Mayorga and E. Freire, "Dynamic analysis of differential scanning calorimetry data,"
Biophysical chemistry, vol. 27, no. 1, pp. 87-96, 1987.
[176] M. V. Kok, "Thermal behavior and kinetics of crude oils at low heating rates by differential
scanning calorimeter," (in English), Fuel Processing Technology, vol. 96, pp. 123-127, Apr
2012.
[177] J. M. Sanchezruiz, "Theoretical-Analysis of Lumry-Eyring Models in Differential
Scanning Calorimetry," (in English), Biophysical Journal, vol. 61, no. 4, pp. 921-935, Apr
1992.
[178] A. E. Lyubarev and B. I. Kurganov, "Modeling of irreversible thermal protein denaturation
at varying temperature. II. The complete kinetic model of Lumry and Eyring," (in English),
Biochemistry-Moscow, vol. 64, no. 7, pp. 832-838, Jul 1999.
[179] D. M. Ecker, S. D. Jones, and H. L. Levine, "The therapeutic monoclonal antibody market,"
(in English), Mabs, vol. 7, no. 1, pp. 9-14, Jan-Feb 2015.
[180] O. Leavy, "Therapeutic antibodies: past, present and future FOREWORD," (in English),
Nature Reviews Immunology, vol. 10, no. 5, pp. 297-297, May 2010.
[181] E. Garber and S. J. Demarest, "A broad range of Fab stabilities within a host of therapeutic
IgGs," (in English), Biochemical and Biophysical Research Communications, vol. 355, no.
3, pp. 751-757, Apr 13 2007.
[182] R. M. Ionescu, J. Vlasak, C. Price, and M. Kirchmeier, "Contribution of variable domains
to the stability of humanized IgG1 monoclonal antibodies," (in English), Journal of
Pharmaceutical Sciences, vol. 97, no. 4, pp. 1414-1426, Apr 2008.
161
[183] R. Bos and W. Nieuwenhuizen, "The Potential Improvement of Thrombolytic Therapy by
Targeting with Bispecific Monoclonal-Antibodies - Why They Are Used and How They
Are Made," (in English), Biotherapy, vol. 5, no. 3, pp. 187-199, Oct 1992.
[184] S. Ewert, A. Honegger, and A. Pluckthun, "Stability improvement of antibodies for
extracellular and intracellular applications: CDR grafting to stable frameworks and
structure-based framework engineering," (in English), Methods, vol. 34, no. 2, pp. 184-199,
Oct 2004.
[185] W. Wang, S. Singh, D. L. Zeng, K. King, and S. Nema, "Antibody structure, instability,
and formulation," (in English), Journal of Pharmaceutical Sciences, vol. 96, no. 1, pp. 1-
26, Jan 2007.
[186] J. Wen, K. Arthur, L. Chemmalil, S. Muzammil, J. Gabrielson, and Y. J. Jiang,
"Applications of differential scanning calorimetry for thermal stability analysis of proteins:
Qualification of DSC," (in English), Journal of Pharmaceutical Sciences, vol. 101, no. 3,
pp. 955-964, Mar 2012.
[187] M. M. Pierce, C. S. Raman, and B. T. Nall, "Isothermal titration calorimetry of protein-
protein interactions," (in English), Methods-a Companion to Methods in Enzymology, vol.
19, no. 2, pp. 213-221, Oct 1999.
[188] M. W. Freyer and E. A. Lewis, "Isothermal titration calorimetry: Experimental design, data
analysis, and probing Macromolecule/Ligand binding and kinetic interactions," (in
English), Biophysical Tools for Biologists: Vol 1 in Vitro Techniques, vol. 84, pp. 79-113,
2008.
[189] J. J. Christensen, R. M. Izatt, L. D. Hansen, and Partridg.Fa, "Entropy Titration . A
Calorimetric Method for Determination of Deltag Deltah and Deltas from a Single
Thermometric Titration," (in English), Journal of Physical Chemistry, vol. 70, no. 6, pp.
2003-+, 1966.
[190] Christen.Jj, R. M. Izatt, and L. D. Hansen, "New Precision Thermometric Titration
Calorimeter," (in English), Review of Scientific Instruments, vol. 36, no. 6, pp. 779-&, 1965.
[191] J. J. Christensen, H. D. Johnston, and R. M. Izatt, "An Isothermal Titration Calorimeter,"
(in English), Review of Scientific Instruments, vol. 39, no. 9, pp. 1356-+, 1968.
[192] D. J. Eatough, "Calorimetric Determination of Equilibrium Constants for Very Stable
Metal-Ligand Complexes," (in English), Analytical Chemistry, vol. 42, no. 6, pp. 635-&,
1970.
[193] D. J. Eatough, V. F. White, L. D. Hansen, N. L. Eatough, and E. C. Ellis, "Hydration of
Nitric-Acid and Its Collection in the Atmosphere by Diffusion Denuders," (in English),
Analytical Chemistry, vol. 57, no. 3, pp. 743-748, 1985.
[194] N. V. Beaudette and N. Langerman, "Improved Method for Obtaining Thermal Titration
Curves Using Micromolar Quantities of Protein," (in English), Analytical Biochemistry,
vol. 90, no. 2, pp. 693-704, 1978.
[195] E. Freire and R. L. Biltonen, "Thermodynamics of Protein-Lipid Interactions," (in English),
Biophysical Journal, vol. 25, no. 2, pp. A284-A284, 1979.
[196] E. Freire, O. L. Mayorga, and M. Straume, "Isothermal Titration Calorimetry," (in English),
Analytical Chemistry, vol. 62, no. 18, pp. A950-A959, Sep 15 1990.
[197] C. A. Brautigam, H. Y. Zhao, C. Vargas, S. Keller, and P. Schuck, "Integration and global
analysis of isothermal titration calorimetry data for studying macromolecular interactions,"
(in English), Nature Protocols, vol. 11, no. 5, pp. 882-894, May 2016.
162
[198] M. J. Todd and J. Gomez, "Enzyme kinetics determined using calorimetry: A general assay
for enzyme activity?," (in English), Analytical Biochemistry, vol. 296, no. 2, pp. 179-187,
Sep 15 2001.
[199] L. S. Roselin, M. S. Lin, P. H. Lin, Y. Chang, and W. Y. Chen, "Recent trends and some
applications of isothermal titration calorimetry in biotechnology," (in English),
Biotechnology Journal, vol. 5, no. 1, pp. 85-98, Jan 2010.
[200] K. Rajarathnam and J. Rosgen, "Isothermal titration calorimetry of membrane proteins -
Progress and challenges," (in English), Biochimica Et Biophysica Acta-Biomembranes, vol.
1838, no. 1, pp. 69-77, Jan 2014.
[201] S. J. Darby, L. Platts, M. S. Daniel, A. J. Cowieson, and R. J. Falconer, "An isothermal
titration calorimetry study of phytate binding to lysozyme," (in English), Journal of
Thermal Analysis and Calorimetry, vol. 127, no. 2, pp. 1201-1208, Feb 2017.
[202] I. Jelesarov and H. R. Bosshard, "Isothermal titration calorimetry and differential scanning
calorimetry as complementary tools to investigate the energetics of biomolecular
recognition," (in English), Journal of Molecular Recognition, vol. 12, no. 1, pp. 3-18, Jan-
Feb 1999.
[203] K. Bouchemal, "New challenges for pharmaceutical formulations and drug delivery
systems characterization using isothermal titration calorimetry," (in English), Drug
Discovery Today, vol. 13, no. 21-22, pp. 960-972, Nov 2008.
[204] L. Wang, B. Wang, and Q. Lin, "Demonstration of MEMS-based differential scanning
calorimetry for determining thermodynamic properties of biomolecules," (in English),
Sensors and Actuators B-Chemical, vol. 134, no. 2, pp. 953-958, Sep 25 2008.
[205] F. E. Torres et al., "Enthalpy arrays," (in English), Proceedings of the National Academy
of Sciences of the United States of America, vol. 101, no. 26, pp. 9517-9522, Jun 29 2004.
[206] M. W. Freyer and E. A. Lewis, "Isothermal titration calorimetry: experimental design, data
analysis, and probing macromolecule/ligand binding and kinetic interactions," Methods in
cell biology, vol. 84, pp. 79-113, 2008.
[207] M. Suzuki and H. Akaguma, "Chemical cross-talk in flow-type integrated enzyme
sensors," (in English), Sensors and Actuators B-Chemical, vol. 64, no. 1-3, pp. 136-141,
Jun 10 2000.