ABSTRACT 
Title of Dissertation: STUDY OF CONDENSATION OF 
REFRIGERANTS IN MICRO-CHANNELS 
FOR DEVELOPMENT OF FUTURE 
COMPACT MICRO-CHANNEL 
CONDENSERS  
  
 Sourav Chowdhury, Doctor of Philosophy, 2008 
  
Directed By: Professor Michael Ohadi, Department of 
Mechanical Engineering 
 
Mini- and micro- channel technology has gained considerable ground in the recent years 
in industry and is favored due to its several advantages stemming from its high surface to 
volume ratio and high values of proof pressure it can withstand.  Micro-channel 
technology has paved the way to development of highly compact heat exchangers with 
low cost and mass penalties. In the present work, the issues related to the sizing of 
compact micro-channel condensers have been explored. The considered designs 
encompass both the conventional and MEMS fabrication techniques. In case of MEMS-
fabricated micro-channel condenser, wet etching of the micro-channel structures, 
followed by bonding of two such wafers with silicon nitride layers at the interface was 
attempted. It was concluded that the silicon nitride bonding requires great care in terms of 
high degree of surface flatness and absence of roughness and also high degree of surface 
purity and thus cannot be recommended for mass fabrication. Following this investigation, 
a carefully prepared experimental setup and test micro-channel with hydraulic diameter 
700 ?m and aspect ratio 7:1 was fabricated and overall heat transfer and pressure drop 
aspects of two condensing refrigerants, R134a and R245fa were studied at a variety of 
test conditions. To the best of author?s knowledge, so far no data has been reported in the 
 iii 
literature on condensation in such high aspect ratio micro-channels.  Most of the 
published experimental works on condensation of refrigerants are concerning 
conventional hydraulic diameter channels (> 3mm) and only recently some experimental 
data has been reported in the sub-millimeter scale channels for which the surface tension 
and viscosity effects play a dominant role and the effect of gravity is diminished. It is 
found that both experimental data and empirically-derived correlations tend to under-
predict the present data by an average of 25%. The reason for this deviation could be 
because a high aspect ratio channel tends to collect the condensate in the corners of its 
cross-section leaving only a thin liquid film on the flat side surfaces for better heat 
transfer than in circular or low aspect ratio channels. 
STUDY OF CONDENSATION OF REFRIGERANTS IN A MICRO-CHANNEL FOR 
DEVELOPMENT OF FUTURE COMPACT MICRO-CHANNEL CONDENSERS 
 
 
 
By 
Sourav Chowdhury 
Dissertation submitted to the Faculty of the Graduate School of the 
University of Maryland, College Park in partial fulfillment 
of the requirements for the degree of 
Doctor of Philosophy 
2008 
 
 
 
 
 
 
 
 
Advisory Committee: 
Professor Michael Ohadi, Chair/Advisor 
Professor Marino di Marzo 
Professor Jungho Kim 
Assistant Professor Bao Yang 
Professor Gary A. Pertmer 
 
 
 
 
 
 
 
 
?Copyright by 
Sourav Chowdhury 
2008 
 
 
 
 
 
 
 
 
 
 
 
 ii 
 
 
 
 
 
 
 
 
 
Dedication 
To my parents for whose unconditional love and dedication I am forever gratefully 
indebted 
 
 
 
 
 
 
 
 
 
 
 
 iii 
Acknowledgements 
First and foremost I would like to thank my advisor Dr. Michael Ohadi, whom I 
am indebted to for his encouragement, counsel and support throughout the course of my 
study and research. 
I also extend my appreciation to the other members of my advisory committee 
who provided support and guidance, including Dr. Marino di Marzo, Dr. Jungho Kim, Dr. 
Bao Yang, and  Dr. Gary A. Pertmer. 
During the course of my research, Dr. Serguei Dessiatoun and Dr. Amir Shoustari 
have afforded me invaluable their support and guidance. I am grateful for their generous 
help and counsel. Also, I had the privilege of working with and being advised by several 
fellow scientists, including  Dr. Francis Franca and Dr. John Lawler, to whom I wish to 
express my sincere gratitude. I would also like to thank Dr. Reza Ghodssi, for his 
suggestions and interaction and providing me with the opportunity of productive 
interaction with his team and access to micro-fabrication facilities. 
I am grateful to my many friends and colleagues at the Smart and Small Thermal 
Systems Laboratory for providing a stimulating and constructive environment to learn 
and grow. I am especially thankful to Dr. Vytenis Benetis, Dr. Parisa Foroughi, Dr. 
Guohua Kuang, Dr. Saeed Moghaddam, Arman Molki, Dr. Mihai Catalin Rada, Dr. 
Valentin Tudor, Dr. Jianlin Wu, Ebrahim Al-Hajri, Edvin Cetegen and Mohamed Al-
Shehhi. I would also like to thank my friends in MEMS Sensors and Actuators Lab who 
have assisted me in some aspects of micro-fabrication. 
Finally, I wish to thank ASHRAE Research Grant Committee, the AHX/EHD 
Consortium and ATEC Inc. for their invaluable financial support of this project. 
 iv 
 
TABLE OF CONTENTS 
 
Nomenclature?????????????????????????????..ix 
List of tables??????????????????????????????.x 
List of figures?????????????????????????????..xi 
 
CHAPTER 1: INTRODUCTION??????????????????????1 
1.1 Background??????????????????????????1 
1.2 Research Objectives And Scope??????????????????2 
1.3 Research Approach  and Presentation of Work.????????????2 
1.4 Summary???????????????????????????3 
 
CHAPTER 2: LITERATURE REVIEW OF CONDENSING FLOWS IN SMALL 
HYDRAULIC DIAMETER CHANNELS????????????..4 
2.1 Introduction??????????????????????????4 
2.2 Two-phase Flow Regimes????????????????.????.5 
  2.2.1 Void Fraction???????????????.??????.7 
                        2.2.2 Flow Regime Analysis?????????????????..10 
                        2.2.3 Mini- and Micro-channel Heat Transfer and Pressure Drop 
Studies???????????????????????...13 
                        2.3.4 Analytical Models????????..??????????...17 
 2.3 Summary??????????????????????????..19 
 
 
 v 
      CHAPTER 3: OBSERVATIONS FROM PRELIMINARY FLOW VISUALIZATION & 
NUMERICAL STUDY???????????????????.20 
 3.1 Introduction?????????????????????????.20 
 3.2 Description of Experimental Setups for Visualization????????....20 
   3.2.1 Micro-channel Array Visualization????????..?20 
   3.2.2 Single Micro-channel Visualization?????????.23 
      3.3 Discussion on Observations in the Present Preliminary Study??????25 
 3.4 CFD study of a bubble motion by Volume of Fluid (VOF) Approach??...26 
  3.4.1 Problem Formulation???????????????26 
  3.4.2 Results and Discussion??????????????.29 
 3.5 Summary??????????????????????????..33 
       
CHAPTER 4: DESIGN AND FABRICATION OF A MICRO-CHANNEL 
CONDENSER DEVICE IN SILICON WAFER USING MEMS 
TECHNIQUES???..34   
   4.1 Introduction??????????????????????????34 
4.2 Micro-channel Condenser Design?????????????????..34 
4.3 Fabrication Methodology and Results???????????????....36 
      4.3.1 Process Steps??????????????????.36 
4.3.2 Wafer-wafer Fusion Bonding Procedure?????????......39 
4.3.3 Results???????????????????????...40 
     4.4 Packaging???????????????????????43 
4.5 Conclusion & Summary??????????????????????44 
 vi 
CHAPTER 5: DESIGN AND FABRICATION OF THE TEST SECTION MICRO-
CHANNEL????????????????????????45  
5.1 Introduction?????????????????????????.45 
5.2 Design of Micro-channel Prototype????????????????47 
5.3 Fabrication Methodology????????????????????51 
5.4 Leakage Testing and Validation of Channel Geometry????????..56 
5.5 Summary??????????????????????????..58 
 
CHAPTER 6: EXPERIMENTAL APPARATUS AND 
INSTRUMENTATION??????????????????....59 
6.1 Introduction?????????????????????????.59 
6.2 Description of Apparatus????????????????????60 
   6.2.1 Description of Refrigerant Flow Loop????????.60 
   6.2.2 Description of Coolant (Water) Flow Loop??????.77 
6.3 Instruments and Measurement System???????????????79 
  6.3.1 Measurement of Temperatures???????????.80 
  6.3.2 Measurement of Pressures?????????????81 
  6.3.3 Measurement of Flows??????????????..81 
6.4 Instrument Calibration?????????????????????.82 
6.4.1 Calibration of Thermocouples???????????..83 
            6.4.2 Calibration of Differential Pressure Transducer????...85 
            6.4.3 Calibration of Absolute Pressure Transducer?????...85 
            6.4.4 Calibration of Water Flow Meters??????????87 
 vii 
            6.4.5 Calibration of Refrigerant Pump Flow Meter?????...89 
  6.4.6 Tuning of PID Controllers?????????????89 
  6.4.7 Summary of Calibration Results??????????...91 
6.5 Selection of Working Fluids???????????????????92 
6.6 Summary??????????????????????????..92 
 
CHAPTER 7: EXPERIMENTAL RESULTS????????????????...96 
7.1 Introduction?????????????????????????.96 
7.2 Experimental Procedure????????????????????..96 
   7.2.1 Definition of Thermodynamic State of Condensation??.99 
7.3 Experimental Domain?????????????????????100 
7.4 Energy Balance of Experimental Apparatus????????????..101 
7.5 Data Reduction Procedure and Uncertainty Analysis?????????103 
   7.5.1 Scheme for Finding Average Heat Transfer Coefficient?103 
   7.5.2 Data Uncertainty Analysis????????????..106 
7.6 Experimental Results with R134a as Working Fluid............??????.108 
           7.6.1 Effect of Variation in Inlet Quality??.???????.108 
      7.6.2 Effect of Variation in Saturation Temperature?????113 
      7.6.3 Effect of Change in Inlet Superheat?????????.115 
     7.6.4 CFD-based Approximate Inverse Local Heat Transfer 
Study...................................................................................118 
7.7 Experimental Results with R245fa????????????????132 
 viii 
7.8 Comparison with Correlations for Local Heat Transfer 
Coefficient???????????????????.????...?.136 
7.9 Comparison with Correlations for Average Heat Transfer 
Coefficient???????????????????.????...?.139 
7.10 Summary?????????????????????????..145 
 
CHAPTER 8: CONCLUSIONS & RECOMMENDATIONS FOR FUTURE 
WORK?????????????????????????.146 
8.1 Introduction?????????????????????????146 
8.2 Overview of Conclusions from Present Work???????????...148 
   8.2.1 Significant Findings from the Present Work?????..149 
8.3 Recommendations for Future Work???????????????...152 
 8.3.1 Recommendations for Experimental Analysis?????152 
 8.3.2 Recommendations for Analytical & Numerical Studies?.154 
8.4 Summary??????????????????????????158 
 
REFERENCES?????????????..???????????..???159 
 
 
 
 
 
 
 ix 
NOMENCLATURE 
 
 
A Constant 
 
Ac Cross-sectional Area (m2) 
ASurface Surface Area (m2) 
ar Aspect Ratio 
a Width (m) 
b Height (m) 
c, C Constant 
Co Confinement Number 
D Diameter (m) 
Dh Hydraulic Diameter (m) 
EB Energy Balance Ratio 
f Friction Factor 
Fr, Fr* Froude Number 
G Mass Flux (kg/m2s) 
g gravity(m/s2) 
Ga Galileo Number 
h Heat Transfer Coefficient 
(W/m2K) 
h  Average Heat Transfer 
Coefficient (W/m2K) 
i Specific Enthalpy (J/kg) 
jg, jg* Dimensionless Gas Velocity 
j Index Variable 
k Thermal Conductivity (W/m2K) 
L Channel Length (m) 
m&  Mass Flow Rate (kg/s) 
n Index Variable 
Nu  Average Nusselt Number 
p, q, r Constant 
P Perimeter (m) 
Pw Wetted Perimeter (m) 
P Pressure (Pa or kPa) 
q, Q Heat Transfer Rate (W) 
q?? Heat Flux (W/m2) 
q? Heat Transfer Rate Per Unit Length 
(W/m) 
R Thermal Resistance (K/W) 
Re Reynolds Number 
Sr Slip Ratio 
T Temperature (C) 
t Thickness (m) 
U Average Velocity (m/s) 
U Uncertainty 
V Velocity (m/s) 
 x 
V&  Volumetric Flow Rate 
(liter/min) 
We  Weber Number 
X Matrinelli Parameter 
x Vapor Mass Fraction, Quality  
X, Y Coordinate Axes (m) 
z Axial Coordinate (m) 
 
GREEK SYMBOLS 
? Void Fraction 
? Volumetric Void Fraction 
? Height (m) 
? Film Thickness (m) 
? Difference 
? Surface Roughness (m) 
? Dynamic Viscosity (kg/m.s) 
? Kinematic Viscosity (m2/s) 
? Two-Phase Multiplier 
? Surface Tension (N/m) 
? Shear Stress (N/m2) 
 
 
 
SUBSCRIPTS 
c capillary 
corr from correlation 
Cu Copper 
evap evaporator 
f liquid 
g gas/vapor 
GS superficial vapor 
h hydraulic 
h height 
i interfacial 
in inlet 
out outlet 
l liquid 
ln logarithmic 
 
LS superficial liquid 
ref refrigerant 
sat saturation 
th thermodynamic 
tt turbulent-turbulent 
v vapor 
w wall
 xi 
LIST OF TABLES 
Table 2.1: Constants for various void fraction models of the form proposed by 
Butterworth??????????????????????????8 
Table 2.2: Summary of Important Condensing Flow Regime Analysis???????10 
Table 2.3: Summary of experimental condensation literatures in horizontal channels with 
Dh < 3 mm??????????????????????????14 
Table 3.1: Summary of the preliminary CFD investigation on adiabatic bubble 
hydrodynamics???????????????????????.....30 
Table 4.1: Wafer Cleaning Procedure (RCA Cleaning)?????????????40 
 
Table 6.1: Summary of Measurement & Control System for Experimental Apparatus?91 
Table 6.2: Comparison of thermo-physical-environmental properties of some common 
refrigerants?????????????????????????..94 
Table 6.3: Comparison of thermo-physical properties of the two working fluids used in 
the present study???????????????????????.95 
Table 7.1: Domain of experiments with R134a and R245fa as working fluids???..101 
Table 7.2: Uncertainty Analysis on Average Heat Transfer Coefficient (R134a)??..107 
Table 7.3: Variation of relevant properties with saturation temperature of R134a??114 
Table 7.4: A sample of three tests with R134a as working fluid chosen for CFD-based 
analysis??????????????????????????..120 
Table 7.5: Variation of relevant properties with saturation temperature of R245fa?...136 
 
 
 
 
 xii 
LIST OF FIGURES 
Figure 2.1: Variation of void fraction with quality for different commonly used 
models???????????????????????????.9 
Figure 3.1: Schematic of the test loop for testing the EDM-machined Micro-
condenser??????????????????????????21 
Figure 3.2: (Top) Schematic of the test prototype (Bottom) Photograph of the assembly 
from above?????????????????????????..21 
Figure 3.3: Schematic of the visualization Test Setup, constructed similar to that used by 
Begg et al??????????????????????????23 
Figure 3.4 Photograph of the thermo-siphon visualization test setup???????...25 
Figure 3.5: Schematic describing the flow domain for simulation of a Taylor Bubble?28 
Figure 3.6: Velocity vector map inside and around a bubble (G=110.2 kg/m2s and Bubble 
volume=0.6786 mm3)?????????????????????.32 
Figure 4.1: Cross-section details of silicon micro-channel condenser???????.35 
Figure 4.2: Geometry of the cross-section of channels by (a) DRIE and (b) Anisotropic 
chemical etching??????????????????????.....36 
Figure 4.3 Photograph of the mask pattern used for photolithography???????37 
Figure 4.4: Process flow for fabrication of Silicon micro-channel condenser????.38 
Figure 4.5: (Left) Micro-channel pattern etched on 4 inch diameter <100> Silicon wafer 
with pre-deposited 0.1 ?m Silicon Oxide and 0.3 ?m Silicon Nitride (Right) 
Close-up of the V-groove feature????????????????...38 
Figure 4.6: Photograph showing result of fusion bond and dicing to create the first 
prototype micro-condenser???????????????????.41 
 xiii 
Figure 4.7 (Left) An assembled set-up for inspection of wafer bond quality consisting of 
an IR-Viewer and a CCD camera for digital capturing. (Right-above) View of 
sample under microscope (Right-below) Void fringes exhibited around the 
periphery of void regions???????????????????....42 
Figure 4.8 Photograph showing the result of wafer bonding for the simplified 
design???????????????????????????..43 
Figure 4.9 (Left) CAD Model showing the package configuration (Right) Photograph of 
the package?????????????????????????.43 
Figure 5.1: An ANSYS model for structural analysis of the micro-channel?????50 
Figure 5.2: Exploded view of CAD model of the micro-channel test section. Inset: 
Detailed view of channel end??????????????????..51 
Figure 5.3: Left ? Schematic of Copper Electro-deposition Setup. Right ? Result of first 
trial with wax mandrel?????????????????????52 
Figure 5.4: Six-step process in machining and closing a channel?????????55 
Figure 5.5: Top ? Micro-channel structure with integrated flow and pressure ports before 
deposition. Bottom Left ? Electro-deposited cover. Bottom Right ? 
Undesirable but removable nodular growth of Copper on the painted 
surfaces???????????????????????????56 
Figure 5.6: Graph of comparison of CFD simulation and experimental measurements of 
pressure drop across the micro-channel length????????????57 
Figure 6.1: Schematic of the experimental setup???????????????...61 
Figure 6.2: Tube evaporator showing labels of various parts described in text????63 
 xiv 
Figure 6.3:  Photograph of the two heat-insulating double-wall Dewar cylinders used in 
the experimental apparatus.???????????????????64 
Figure 6.4: Diagram depicting the thermocouple junctions formed by constantan wires 
mating with the copper micro-channel structure??????????....65 
Figure 6.5: Part of the refrigerant loop indicating the evaporator and condenser test 
section encased in double-wall glass Dewar cylinders?????????65 
Figure 6.6: Left ? The bottom half of the copper elbow bridging the evaporator and 
condenser Dewar segments. Right ? The top half of the elbow placed to 
complete the active heat shield??????????????????67 
Figure 6.7: Schematic illustrating the various parts of the Reservoir System???..?.73 
Figure 6.8: Photograph of the refrigerant loop assembly????????????..77 
Figure 6.9: Y-connection indicating the location of RTDs for accurate coolant 
temperature measurement at the inlet to the condenser. Similar arrangement 
was provided at the coolant outlet????????????????...81 
Figure 6.10: Calibration chart for the Thin-film Platinum RTDs that measured the coolant 
inlet and outlet temperatures???????????????????84 
Figure 6.11: Calibration chart for Condenser wall thermocouples?????????85 
Figure 6.12: Calibration results for Validyne? Differential Pressure Transducer using U-
tube manometer???????????????????????...86 
Figure 6.13: Results for the corroboration of the linearity and accuracy of factory 
calibrated Setra? Absolute Pressure Transducer???????????87 
Figure 6.14: Low Flow Water Flow Meter Calibration Chart??????????...88 
Figure 6.15: High Flow Water Flow Meter Calibration Chart??????????..88 
 xv 
Figure 6.16: Flow response versus monitor readings of the Ismatec Micropump internal 
flow meter??????????????????????????89 
Figure 6.17: Effects of different control options on the absolute pressure??????90 
Figure 7.1: Schematic depicting the parameters used for calculating percentage energy 
balance??????????????????????????...102 
 
Figure 7.2: Effect of change in inlet quality on (a) average heat transfer coefficient (b) 
total channel pressure drop (R134a)???????????????..110 
Figure 7.3: Wall temperature profile at different inlet qualities at refrigerant mass fluxes 
(a) 150 (b) 200   (c) 300 kg/m2.s respectively (R134a)????????.112 
Figure 7.4: Effect of change in saturation temperature (a) average heat transfer coefficient 
(b) total channel pressure drop (R134a)??????????????114 
Figure 7.5: Effect of change in inlet superheat on (a) average heat transfer coefficient (b) 
overall channel pressure drop. (R134a)?????????????.....117 
Figure 7.6: CFD-model to study the combination of convective heat transfer in the 
coolant jacket and heat conduction in the channel wall????????.118 
Figure 7.7: Schematic indicating the surfaces and points used in the post-processing 
analysis of CFD results????????????????????..121 
Figure 7.8: (a) Trial heat flux profiles along the channel as inputs to the FLUENT? 
model (b) Corresponding temperature profile as output to 
simulation?????????????????????????..122 
Figure 7.9: Contour Plots of (a) Temperature (in Centigrade); and (b) velocity (in m/s) of 
the coolant (shown with two orthogonal planes) and the outer channel along 
the length of the test section??????????????????..123 
 xvi 
Figure 7.10: Path-line trace for the coolant flow colored by (a) temperature (b) particle ID 
indicting a twisted trajectory of the main flow of the coolant in the coolant 
jacket???????????????????????????..125 
Figure 7.11: (a) Temperature contour plot (in Section A) and (b) temperature plot on 
Crosshairs A ? A? and B ? B???????????????????.126 
Figure 7.12: Axial Profile of the heat transfer coefficient (negative value indicating heat 
rejection) on the outer surface of the channel wetted by the coolant (Position z 
= 0 m is close to the coolant exit end)?????????????..?127 
Figure 7.13: Comparison of computed heat transfer coefficient with most relevant 
experimental data from literature. (R134a)???????????.?..128 
Figure 7.14: Comparison of average heat transfer coefficient values obtained by different 
wall averaging techniques. (R134a)???????????????..128 
Figure 7.15: Effect of changing mass flux on (a) the average heat transfer coefficient and 
(b) channel pressure drop. (R245fa)???????????????..132 
Figure 7.16: Effect of changing inlet superheat on (a) average heat transfer coefficient 
and (b) channel pressure drop. (R245fa)????????????.?..134 
Figure 7.17: Effect of change in saturation temperature (a) average heat transfer 
coefficient and (b) channel pressure drop. (R245fa)?????????.135 
Figure 7.18: Comparison of correlation and computed local Nusselt Numbers (at specific 
qualities) at different mass fluxes????????????????..138 
Figure 7.19: Schematic explaining the situation for deriving the average Nusselt number 
from a given heat transfer correlation???????????????142 
 xvii 
Figure 7.20: Comparison of average Nusselt numbers from correlations and from 
experimental data for R134a??????????????????.143 
Figure 8.1: A semi-exploded sketch of a proposed improved test section?????..155 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 1 
Chapter 1 
INTRODUCTION 
 
1.1 BACKGROUND 
 
Micro-channel technology has gained considerable ground in recent years in 
industry and is favored for its capability to yield low cost, compact heat and mass transfer 
devices due to high surface to volume ratio, among other advantages. Considerable 
volume of literature exists on micro-channel cooling techniques ? by liquid phase and 
phase changing fluids ? applied to cooling of electronics and other high heat flux devices 
for a wide range of both military and commercial applications. More recently, the interest 
in overall compactness of thermal cooling loops has spurred interest in studying 
condensing refrigerants in micro-channels to complement the micro-channel evaporator 
technology. Due to high heat transfer coefficients it offers and high values of proof 
pressures it can withstand, micro-channel technology also finds strong interest in two-
phase flows of high pressure refrigerants used automotive & stationary heating, 
ventilating, and air conditioning/refrigeration (HVAC-R) industry.  
With the significant demand in this field, the art of combining and utilizing the 
manifold benefits of the micro-channel technology is further ahead that the science of 
obtaining comprehensive understanding of fluidic heat and mass transfers in these 
channels. This is more so for condensing flows in micro-channels which is of recent 
interest as evidenced by a number of publications on experimental and theoretical works 
within the present decade. Several heat transfer and pressure drop correlations have been 
proposed for various micro- and mini- channel sizes and with a variety of refrigerants, 
mostly pressurized ones.  Most of these correlations have purely experimental basis and 
 2 
some were created with guidance from analytical and flow visualization understandings. 
These published experimental data and proposed correlations form a valuable basis for 
the design and development of compact micro-channel evaporators and condensers.   
In the present work, the feasibility of fabrication of a compact micro-channel 
condenser with interconnects, using conventional MEMS techniques, have been explored 
to understand its issues and limitations. Following this, an experimental investigation of 
condensation of refrigerants in a high aspect ratio micro-channel with two pressurized 
refrigerants, R134a and R245fa have been carried out to understand the applicability of 
the available correlations. The study involved parametric variations of the flow aspects 
and calculated/measured the overall heat transfer coefficient and pressure drop. This 
knowledge is essential for the sizing of the condenser in the design phase. Thus, the 
present work adds to the growing body of experimental literature on condensation in 
mini- and micro-channels. The present work is a part of a longer undertaking for a 
comprehensive investigation and assessment of the physics of such flows in a single 
micro-channel. 
 
1.2 RESEARCH OBJECTIVES & SCOPE 
 
Following are the broad objectives of the present work: 
 
 
1) To explore the issues in fabrication of a condenser structure in silicon for the use in 
electronics cooling applications. 
2) To generate experimental data of heat transfer and pressure drop for a single high 
aspect ratio micro-channel and compare those with reported correlations from the 
literature to understand their applicability in relation to condenser sizing and 
designing. 
 3 
1.3 RESEARCH APPROACH AND PRESENTATION OF WORK 
First, a thorough literature survey and review was conducted and is presented in 
Chapter 2. Next, to fulfill the objectives of research, preliminary experiments were 
carried out in a compact condenser made by Electro-discharge machining (EDM) on a 
brass plate and in a thermo-siphon test setup with a single transparent glass capillary tube, 
which are reported in Chapter 3. Then, in Chapter 4, the feasibility of micro-fabrication 
process as a means for making micro-channels in a silicon substrate is explored. In 
Chapters 5 through 7, an experimental investigation is reported with two pressurized 
refrigerants, R134a and R245fa, in a specially constructed horizontal single-channel 
micro-condenser of hydraulic diameter 0.7 mm and aspect ration 7:1. A compact 
experimental setup was constructed and instrumented with care to lower the heat 
loss/gain effects which affected the uncertainties in data. Overall heat transfer coefficient 
and pressure drop data were then collected and comparison of this data against various 
correlations meant both for conventional scale tubes and micro-channels were made. 
CFD-based approximate numerical inverse heat transfer modeling was adopted with 
manual iterations to investigate the local heat transfer effects in the channel for selected 
data points. Finally, in Chapter 8, several conclusions are presented and the future efforts 
needed to bring a closure to the present work are indicated. 
 
1.4 SUMMARY 
 In the present chapter, the background and need for the present research is stated 
and the research scope is broadly outlined. The research approach encompassing the 
objectives and issues of interest that arose during the research is briefly described. 
 
 
4 
Chapter 2 
 
LITERATURE REVIEW OF CONDENSING FLOWS IN SMALL HYDRAULIC 
DIAMETER CHANNELS 
 
 
2.1 INTRODUCTION 
 
In the past decade significant attention has been spent in heat transfer and fluid 
flow enhancement of the process for heat removal from small surface areas under high 
heat flux conditions such as high heat flux electronic chips and high current density 
electrical equipment. Although compared to single phase and boiling flows the 
phenomenon of condensation is less studied, there is a sizable and growing volume of 
literature available in two-phase condensing flows, mostly in conventional tubes. In a 
thermal management loop, the heat is dissipated from the target surface to the liquid. If 
the fluid changes phase then it has to be condensed by air or another cooling medium. 
Often the cooling is done with air. In such a case, since the air-side thermal resistance is 
significantly high, even when extended fin or louver-enhanced surface area is provided to 
the air-side, most recent research efforts have been spent in reducing the air-side thermal 
resistance through various enhancement techniques. Meanwhile, a more effective heat 
transfer on the refrigerant side means less volume/weight for the whole system. With the 
advent of miniaturization of components due to various reasons, chief amongst which is 
the cost, volume, weight and portability, compact condenser designs involving micro-
channels and micro-port flat tubes have been sought after by the refrigeration, air-
conditioning and automotive industries. This need has spurred studies in internal flow 
condensation in mini- and micro-channels. Still, at the present moment, the body of 
 
 
5 
experimental and analytical literature on condensing flows in small diameter channels is 
limited. 
Detailed and up-to-date surveys of literature condensation in general were 
reported by Garimella et al. [63] and Ghiaasiaan [64]. However, in the present chapter, a 
brief review of relevant literature related only to flow condensation in small diameter 
channels will be discussed. Since heat transfer and hydrodynamics of two-phase flow are 
strongly linked, predictive models on condensation involve the understanding of the 
complex flow regimes that exist in a micro-channel. Hence a brief review of the literature 
on flow regime studies in adiabatic and condensing two-phase flows is relevant and will 
be made.  
 
2.2 TWO-PHASE FLOW REGIMES 
Before the details on flow regimes in mini- and micro-channels are discussed, the 
issue of what should be regarded as a mini- or micro- channel is needed to be understood. 
This topic has been discussed by different researchers and there is a general agreement 
that when the channel size starts to influence the flow pattern, heat transfer and pressure 
loss gradients in a way not found in conventionally studied tubes/channels, it needs a 
different treatment. Thus, the broad term ?micro-channel? refers to those channel sizes for 
which the relative importance of gravity, interfacial shear, surface tension  start to 
become different than that for ?macro-channels? or certain mini channels. At smaller 
channel sizes, the influence of gravity is less (but not necessarily negligible) and the 
surface tension and viscous forces emerge as dominant players in determining the flow 
pattern. This is a broad distinction between micro- and macro-channels. A more nuanced 
 
 
6 
categorization can be made by making the distinctions between mini-channels (relatively 
larger sized channels, typically 1 to 2 mm in hydraulic diameter, but still distinctly 
different than the ?macro-? category), micro-channels (smaller in diameter than mini-
channels; less than 1 mm hydraulic diameter) and the extreme of the range, sub-micron 
channels (where the continuum regime is no longer applicable). Serizawa et al. [1] 
conducted steam-water and air-water flow visualization studies in a micro-channel with 
hydraulic diameter 50 ?m and recommended that the capillary length scale, given in 
Equation 2.1 below should be greater than the hydraulic diameter of the channel for it to 
belong to micro-channel category. 
                                                 )(
gl
c gL ??
?
?=                                                       (2.1) 
This parameter captures the competition between the surface tension and the buoyancy 
forces. In air-water studies, where the surface tension is strong, the micro-channel 
diameter limit will be higher than that for the refrigerant like R-134a. There is yet another 
parameter similar to the above and is termed as Confinement Number, Co, which 
essentially is the ratio of the capillary length scale to the hydraulic diameter (Co=Lc/Dh).  
Kew and Cornwell [2] had recommended that if Co > 0.5 then the flow is truly influenced 
by the hydraulic diameter. This can be understood for a channel in which, in order to 
keep the forces in balance, it is necessary that a single bubble, not multiple bubbles, 
occupies the entire cross-section of the channel at any instant. In practice, a survey of the 
mini- and micro-channel literature indicates that such strong demarcation in 
categorization is not used by researchers and the term micro-channel is sometimes, 
incorrectly, included to mean channels having sizes slightly larger than 1 mm hydraulic 
diameter. Specifically, in case of high aspect ratio rectangular channels, one of the 
 
 
7 
dimensions may be of several millimeters long while the other dimension may be about a 
few hundred microns. In such case, there is gravity effect if the longer side of the channel 
cross-section is aligned with the gravity vector and the effect is less significant when the 
other side is so. 
Unfortunately, because of its relative simplicity in technique, most of the flow 
regime studies for two-phase flows have been conducted with adiabatic flows in which 
the gaseous component is typically chosen as air or nitrogen and the liquid component as 
water or oil. There are parallels between such flows and condensing two-phase flows. 
However, the surface tension and gravity effects are significantly different. For example, 
the ratio of densities of air to that of water is about 1:1000 where as for a refrigerant fluid 
such as R134a the vapor to liquid density ration is in the range of 1:200. Similarly, the 
surface tension effects (and contact angles with respect to the channel material) could be 
significantly different. Also, the effect of momentum change in condensation, when gas 
molecules enters the liquid encountering reduction in velocity, is absent in adiabatic 
studies. Hence in the present discussion, the condensing flows in relatively small 
diameter conventional channels and micro-channels will be discussed. 
 
2.2.1 VOID FRACTION 
 Void fraction models express relationships of the void-fraction (?g  or simply ?) 
to the flow quality (x). Void fraction is a parameter that describes the mixed nature of 
two-phase flow and can be described as the time-averaged fraction of the cross-section of 
the channel occupied by the gas phase at a given location. A good model for void fraction 
is a useful component in the analysis for a one-dimensional flow in which one 
 
 
8 
momentum equation can be solved for fluid for a given void fraction. In condensing 
flows, in general, the void fraction decreases along the length of the channel because of 
the decrease in the quality. However, in order to obtain a physically appropriate model 
for the void fraction, the actual physics of the flow should be taken into account. In the 
different correlations proposed so far for condensing flows, the void fraction calculation 
relates to the slip between the two phases.  
Butterworth [3] proposed the general form of the more widely used void fraction 
models: 
                                
1
11
?
?
?
?
?
?
?
?
?
???
?
???
?
???
?
???
??
?
??
?
? ?+=
r
g
l
q
l
g
p
x
xA
?
?
?
??
                             (2.2) 
The constants in this relationship differ from model to model and are summarized for 
some of the commonly used models in the Table 2.1. 
Table 2.1: Constants for various void fraction models of the form proposed by Butterworth 
 
Correlation A P q r 
Homogenous Flow 1 1 1 0 
Lockhart-Martinelli (1949) 0.28 0.64 0.36 0.07 
Baroczy (1963) 1 0.74 0.65 0.13 
Zivi (1964) 1 1 .67 0 
Thom (1967) 1 1 0.89 .18 
 
 
 
 
 
 
 
9 
  
 
 
 
 
 
 
 
 
 
 
 
 
Figure 2.1: Variation of void fraction with quality for different commonly used models 
 
 
Several other researches have also used the model given by Smith [4]: 
1
14.01
14.0
6.04.011
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
??
?
?
?
?
?
??
?
?
?
?
?
?????? ?+
?????? ?+
+?????? ???
?
?
???
?+=
x
x
x
x
x
x g
l
l
g ?
?
?
??       (2.3) 
Smith stated that this model is useful for different flow regimes, mass fluxes, qualities etc. 
parameters. 
A related parameter of interest similar to void fraction is the volume fraction ?, 
which is the time-averaged fraction of gas/vapor volume at a specific location. The 
homogeneous void fraction,  ?Homogeneous is same as ?. The slip ratio, Sr, which is the 
Comparison of Void Fraction Models Commonly Used 
(R134a at T_sat=50C)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Vapor Quality, x
Vo
id
 F
ra
cti
on
, a
lp
ha
Homogeneous
Lockhart-Martinelli
Baroczy
Zivi
Thom
 
 
10 
ratio of the average velocities of the gas core and the liquid, is related to the void fraction 
(Premoli [5]) by: 
        ( )
lgr xSx
x
??? /)1( ?+=        (2.4) 
As mentioned before, majority of the work done in experimental studies on void 
fraction have been performed in air-water systems and therefore not readily applicable to 
condensation situation. However, mention should be made for some of the novel 
techniques that have been made use of in these studies. For small diameter channels 
(between 1 and 5 mm), Kariyasaki [6] measured void fractions and proposed empirically 
derived models of the measured void fraction as a function of  ?. Mishima, Hibiki and co-
researchers [7, 8] made several publications in which they used a novel neutron 
radiography technique to measure the void fraction profile in 1 to 4 mm diameter 
channels and found that their data is in good agreement with that of Kariyasaki et al. 
Triplett et al. [9, 10] had used image analysis technique, whereas Rezkallah and co-
workers [11] had used non-intrusive capacitance-type probe to measure void fraction in 
micro-gravity environment. Kawahara et al. [12] also used image analysis technique to 
measure void fraction of nitrogen and pure water or water-ethanol mixtures in millimeter 
and sub-millimeter channels and concluded that the millimeter channels follow 
homogeneous flow model. They have used different concentrations of ethanol to study 
the effect of fluid properties on void fraction but concluded that such effect was not 
significant. Thus, a variety of techniques and understanding have been developed in void 
fraction studies. However, as mentioned before, more research related to actual 
condensing flows is needed.  
 
 
 
11 
2.2.2 FLOW REGIME ANALYSIS 
Several researchers in the past have studied condensation in macro-sized tubes 
and have attempted to give the transition criterion from one flow regime to the next 
through various flow parameters. Table 2.2 depicts some of these studies on condensation 
flow regimes and the choice of the transition parameter(s) used by the various researchers. 
Table 2.2: Summary of important Condensing Flow Regime Analysis 
 
Researcher Parameter(s)  Relationship form Remarks 
Traviss-Rohsenow [13] 
Film Reynolds No. Rel=G(1-x)D/?l 
Froude No. Frl=V2/g? 
Galileo No. Gal=gD3/?l 
Martinelli Parameter, Xtt 
Rel=f (Frl,Gal,Xtt) 
Frl=45 marks the boundary 
between annular and stratified 
wavy, slug and plug flows. 
Breber et al [14] Wallis? jg*=Gx/(Dg?g(?l-?g))
0.5 
Martinelli Parameter, X 
Titel-Dukler type flow 
regime map with axes jg* 
and X 
Different combinations of jg* 
and X gives annular, 
Wavy/Stratified, Slug and 
Bubbly Flows. 
Sardesai [15] 
Modified Froude Fr*= Fr.(?v/Gx). UGS 
Martinelli Parameter, X 
Chisholm two-phase multiplier, ?g 
?g2.F=K, constant 
K=1.75 is lower limit of 
annular flow, K=1, start of 
gravity flow 
Soliman [16] 
Film Reynolds No. Rel=G(1-x)D/?l 
Martinelli Parameter, Xtt 
Film Thickness Parameter, ?+ 
Azer two-phase multiplier, ?g 
Galileo No. Gal=gD3/?l 
Rel=f (Frl,Gal,( ?g /Xtt)) 
Various data sources used. 
Annular to wavy/intermittent 
flow transition at Fr=7. 
Soliman [17] 
Weber No. We 
Azer two-phase multiplier, ?g 
Superficial gas & liquid Reynolds No. 
ReGS & ReLS 
Martinelli Parameter, Xtt 
We=f(ReGS, ?g, Xtt, etc..) 
at different ReLS 
Various data sources used. 
We<20 : always annular 
We>20: always mist 
Tandon et al. [18] Wallis? jg*=Gx/(Dg?v(?l-?v))
0.5 
Smith?s void fraction, ? 
Similar to Breber et al. 
with axes jg* & (1-?)/? 
Different combinations of jg* 
& (1-?)/? gives spray, 
annular, wavy, slug, plug 
flows. 
Wang et al. [19] Mass Flux, G Quality, x Flow Map with axes G & x 
50<G<700 kg/m2.s studied at 
different saturation 
temperatures and for different 
refrigerants. Flow transitions 
criteria given in the flow map. 
Dobson and Chato [20] 
Modified Superficial Vapor Velocity, 
jg=jg,Mandhane.( ?g/?air) 
Superficial Liquid Velocity, jl 
Mandhane (1974) Map 
was created with air-water 
studies. Hence the jg 
needed correction. 
Exhaustive condensation 
studies in 3 to 7 mm for 
different refrigerants and 
blends, and wide mass flux 
range 25<G<750 kg/m2.s 
Detailed description of flow 
regime changes for different G 
and x. Found very good 
agreement with corrected 
Mandhane Flow Map. 
Coleman and Garimella 
[21, 22] 
Mass Flux, G 
Quality, x 
Flow Map with axes G & 
x 
Very detailed study of 
condensation of R134a in 
round, square and rectangular 
tubes at different qualities 
with 1<Dh<4.9 mm and 
150<G<750 kg/m2.s. Effect of 
Hydraulic diameter and 
channel shape on flow regime 
 
 
12 
was notable. 
 
 
Amongst the different flow regime analyses tabulated here, the works of Dobson and 
Chato [20] and Coleman and Garimella [22] deserve special mention due to thoroughness 
of their studies. Especially, the latter work, which encompasses various tube shapes and 
tube hydraulic diameters and which includes a wide range of test conditions, is significant 
and creates a strong reference for further studies. Some of the salient observations noted 
by them are given below: 
? Four major flow regimes, annular, wavy, intermittent and dispersed were 
identified. Within each major regime, different patterns/mechanisms were 
discernible. For example, in the annular regime, mist, annular ring, wavy ring, 
wave packet and annular film patterns were visible. These various patterns are 
evidence of the balancing influences of gravity and interfacial shear.   
? Apart from overall flow regime maps for a particular tube at different mass flux 
and quality values, flow regime maps with G and x as axes have been erected to 
specifically show the effects of tube hydraulic diameter for square tubes 
(separately for intermittent and annular flow regimes) and tube shape effect at 
similar hydraulic diameters (for flow regime transitions).   
? For overall flow pattern map for a particular hydraulic diameter, the regions for 
major regimes and within each regime, the regions of different flow patterns were 
identified.  
? The flow map for studying effect of hydraulic diameter on square tubes indicates 
that with decreasing Dh, the region of intermittent regime is reduced. This effect is 
 
 
13 
significant at lower mass fluxes: 150<G<400 kg/m2.s. The effect was attributed to 
the fact that at smaller hydraulic diameters, the surface tension effect is more 
significant because it is easier to collect and retain liquid condensate in the 
corners. Thus the transition from intermittent to wavy/annular region occurs at 
higher qualities. Also, at larger hydraulic diameters, the annular flow tends to be 
wavy flow with discrete waves visible in the bottom condensate pool; the top 
condensate layer was relatively thinner than the bottom pool. 
? As hydraulic diameter decreases the wavy flow regime is increasingly replaced by 
annular regime since the surface tension tends to stabilize the flow. 
? In case of flow map for effect of tube shape, it was noted that the square and 
rectangular tubes tend to hold condensate at the corners and hence at a given mass 
flux, the transition from intermittent to wavy flow regimes was found to occur in 
their cases at qualities lower than that for round tubes. In this respect the behavior 
of rectangular tubes was found to be in between square and round tubes. 
? In conclusion, the effect of the tube hydraulic diameter was found to be stronger 
than that of the tube shape.  
In summary, although a wealth of investigations have been generated in flow regime 
mapping, only a few available literature, mostly involving diameters > 3 mm have been 
available that involves condensing flows. The flow maps created by Coleman-Garimella 
[22] provide a starting reference and guidance for future studies involving higher aspect 
ratio micro- and mini-channels.    
 
 
2.2.3 MINI- AND MICRO-CHANNEL HEAT TRANSFER AND PRESSURE DROP 
STUDIES  
 
 
14 
  Table 2.3 summarizes the available experimental literature that is relevant to the 
present work. As seen in this table, there is only a limited sub-set of work that concern 
channels with hydraulic diameter < 1 mm and some of them involve enhanced surface 
features such as micro-fins. A brief discussion on some of the more relevant publications 
amongst these will be given here.  
Yang and Webb [23, 24] employed a modified Wilson-plot technique to obtain data 
for plain and micro-finned tubes with a somewhat high mass flux range relative to the 
refrigeration and air-conditioning applications typically found. Their data indicate 
enhancement of the heat transfer coefficient with micro-fin features although the 
enhancement effects decreased for the upper ranges of mass fluxes. The heat transfer 
coefficient also showed a weak dependence on the heat flux and the authors attributed 
this to increased contribution of the momentum effect at higher heat fluxes.  
Table 2.3* Summary of experimental condensation literatures in horizontal channels with Dh < 3 mm 
 
Researcher 
Channel 
Dimensions & 
Features 
Dh (mm) Fluid Tsat (C) G (kg/m2.s) 
Yang and 
Webb [23, 24] 
Smooth, 
Micro-fin, 
Horizontal, 
MP 
2.64, 1.56 R-12 65 400-1400 
Zhang [25] 
Smooth, 
Horizontal, 
SP-C, MP-C, 
MP-R 
6.25, 3.25, 2.13, 1.45, 
0.96, 1.33 
R134a, R22, 
R404a 40, 65 200-1000 
Webb and 
Ermis [26] 
Smooth,  
Micro-fin, 
Horizontal, 
MP, R 
0.611,1.564, 
0.44,1.33 R134a 65 300-1000 
Garimella and 
Bandhauer 
[27] 
Smooth, 
Horizontal, 
MP-S. 
0.76 R134a 51.7 150-750 
Wilson et al. 
[28] 
Micro-fin, 
Horizontal, 
Flattened 
1.84, 4.4, 6.37, 7.79 R134a, R410a 35 75-400 
 
 
15 
Koyama et al. 
[29] 
Smooth, 
Horizontal, 
MP, R 
1.11, 0.807 R134a 60 100-700 
Wang [30] 
Smooth, 
Horizontal, 
MP-R 
1.46 R134a 61-66.5 79-760 
Baird et al. 
[31] 
Smooth, 
Horizontal, 
SP-C 
0.92, 1.95 R123, R11 20-72 70-600 
Cavallini et al. 
[32, 33] 
Smooth, 
Horizontal, 
MP-R 
1.4 
R134a, 
R236ea, 
R410A 
40 200-1400 
Kim et al. [34] 
Smooth, 
Horizontal, 
SP-C 
0.691 R134a 40 100-600 
Bandhauer et 
al. [35] 
Smooth, 
Horizontal, 
MP 
0.506, 0.761, 1.52 
mm R134a - 150-750 
*Note: Based in part on Cavallini [33]; MP- Multi-port, SP-Single port, R-Round, C-Circular 
The main mechanism for micro-fin enhancement, which was notable at low mass 
fluxes and higher qualities, was believed to be due to drainage of the condensate film 
from the tips of the fins towards their bases. At higher qualities, the fins are submerged in 
condensate and therefore were not able to provide enhancement.  
Koyama et al. [29] conducted a study on condensation of R134a in two 865 mm 
long extruded multi-port tubes, one with 19 ports (Dh =0.807 mm) and the other with 8 
ports (Dh =1.114 mm). The effective length is 600 mm for each tube with the balance 
length used to provide access to the pressure ports. Local heat transfer coefficients were 
measured using heat flux sensors for 75 mm length sections. The authors acknowledged 
that it is very difficult to accurately measure the local heat flux using Wilson plot method 
or by temperature rise in the coolant especially for small channels such as these. Limited 
data identifying individual test points were provided; however, the data has been 
presented in form of comparison with heat transfer and pressure drop correlations. The 
Nusselt number correlation used by them is of the form of a root-sum-square of the 
 
 
16 
annular and stratified Nusselt numbers obtained from Haraguchi et al. [36]. In addition to 
this modification, they had adopted the formula for two-phase multiplier proposed by 
Mishima-Hibiki [37], because the correlations proposed by these authors were best able 
to predict their data. Even then, their Nusselt number comparison graph showed 
significant over-prediction of the data; interestingly, the over-prediction was found to be 
higher for the tube of diameter 1.114 mm rather than for the 0.807 mm. Moser [38] 
correlation was found to be a better predictor for the smaller diameter tube than the larger 
one although for lower mass fluxes (<300 kg/m2.s) the correlation gave significant under-
prediction.  
Cavallini et al. [32, 33] conducted measurements of heat transfer coefficients and 
pressure drop in a multi-port flat tube with 1.4 mm port hydraulic diameter using R134a 
and R410A. The test section was divided into three segments to provide quasi-local 
conditions for their measurements. They converted the saturation temperature changes 
into pressure drop values and found reasonably good agreement with correlations by 
Friedel [39], Zhang and Webb [25], Mishima-Hibiki [37] and others. However, they 
found that their heat transfer coefficient data was under-predicted, especially at higher 
heat fluxes, by several available correlations, most notably by those proposed by Moser 
et al. [38], Zhang and Webb [25], Cavallini  et al. [40] and Koyama et al. [29]. The 
reason proposed for this was that at higher mass fluxes, a tendency of mist type flow was 
possible. Also, most of the available correlations come from larger diameter tubes with 
annular flow features.  
Garimella and Bandhauer [35] implemented an interesting thermal amplification 
technique to mitigate the issues of low heat duties due to low mass fluxes and high heat 
 
 
17 
transfer coefficients.  The local condensation heat transfer rates were measured within 
?10%. The heat transfer coefficient values reported by these authors were for 0.506, 
0.761 and 1.520 mm circular tubes with R134a as the working fluid. However, heat 
transfer coefficients beyond 85% quality have not been. In general, the data indicate an 
approximately linear trend between heat transfer coefficient and local quality, within the 
reported quality range. For the lower mass flow rate cases such as 150 and 300 kg/m2.s, 
the proper distinction was not possible to be made between heat transfer coefficient of the 
higher and lower mass fluxes, since the difference lie within the data uncertainty.  
Baird et al [31] implemented thermo-electric coolers for the purpose of local heat 
flux measurement in their study with 0.92 and 1.95 mm circular channels at different heat 
and mass fluxes for R123 and R11. Their data, unlike that of Yang and Webb [23], 
indicate a rather strong influence of heat flux and a weaker influence of the saturation 
pressure on heat transfer coefficient. Very little influence of the tube diameter was 
observed in their operating range. 
 
2.2.4 ANALYTICAL MODELS 
So far very few studies have been observed which concern the modeling of the 
two-phase condensing flows. Notable among them are the studies of Begg et al. [41] and 
by Wang and Rose [42, 43]. Begg et al. [41] have studied the physical and two-
dimensional mathematical model of annular film condensation in mini circular tubes with 
the vapor and liquid flows coupled through interfacial heat flux, shear stress, and pressure 
jump conditions due to the surface tension effects. In their model an equilibrium stable 
annular flow is assumed to occur with saturated vapor flow at the inlet and liquid 
 
 
18 
completely filling the tube at some distance from the tube?s inlet end. In their 
mathematical model, they have adopted the Munoz-Cobo et al.[44] interfacial shear stress 
formula meant for annular film condensation in vertical tubes with non-condensable 
gases, and friction factor as proposed by Wallis (1969). To validate their model, they 
created a thermo-siphon loop made with Pyrex? which circulated water vapor into test 
section. The test section consisted of three kinds of tubes with diameters 3.4, 2.3 and 1.6 
mm with a thermally insulated flow development region prior to the condensing section. 
Quasi-static annular flow pattern was observed and the pattern was found to be steady 
except when non-condensable gases were believed to be present within the test setup. 
One interesting observation in the predicted film thickness profile was that at the end of 
the annular flow where the liquid film collapses to form fully liquid flow, the film 
thickness suddenly decreases before increasing again. Also, in the flow visualization 
studies, the film thickness seems to increase (for smaller diameter tubes) before 
decreasing and increasing again (to converge into the liquid pool). It was not clear if non-
condensable elements in the vapor were responsible for this phenomenon.  
Wang and Rose [42, 43] have studied a truly three dimensional model of annular 
condensation in circular, square, rectangular and triangular channels at millimeter scale. 
In their model they considered effects of gravity, surface tension, and shear. For non-
circular cross-section tubes, they employed cylindrical coordinate system for the 
condensate film in the corner regions and Cartesian coordinates for flat edges. The 
condensation momentum loss effect (suction) was modeled after the approach proposed 
by Mickley et al. (1954) (as described by Kays, Crawford and Weigand [45]) and the 
interfacial friction factor (zero transpiration) was taken of the form proposed by Churchill 
 
 
19 
[46]. The effect of the film curvature along the axial direction was neglected. This is an 
issue where the condensate film tends to converge giving an error in the prediction for 
total condensation length. Due to competing effects of surface tension and gravity, the 
film thickness was not found to be uniform at different points on the perimeter at a given 
location. This would cause a non uniform heat transfer coefficient at different locations of 
the perimeter. In their graphical comparison of average heat transfer coefficient profiles 
along the axis for different geometries, significant differences were observed for 
rectangular and square geometries against circular geometries. In fact, near the vapor 
entrance, the rectangular geometry indicates 50% better heat transfer coefficient at the 
flow conditions they considered. This is a promising result. 
 
2.3 SUMMARY  
 In this chapter, a brief summary of available literature on study of various aspects 
of condensation in mini- and micro-channels has been presented. The relevant literature, 
broadly classified into void fraction related studies, flow regime analysis and flow 
condensation heat transfer and pressure drop studies has been summarized. In the end, a 
brief summary of few available analytical studies have been made. 
 
 
20 
 
Chapter 3 
 
OBSERVATIONS FROM PRELIMINARY FLOW VISUALIZATION & 
NUMERICAL STUDY 
 
3.1 INTRODUCTION 
 
In the present chapter, preliminary flow visualization and preliminary numerical 
(CFD) study of bubble flow hydrodynamics will be presented. The visualization studies 
were designed to obtain preliminary information about the flow process in a micro-
channel, such as prevailing flow regimes, issues in test setup preparation etc. and were 
not meant for rigorous investigation of the condensation phenomenon with care for 
accuracy. In the same light, the preliminary numerical study of bubble behavior has been 
made to understand the flow simulation process and its potential as well as its limitations.  
 
3.2 DESCRIPTION OF EXPERIMENTAL SETUPS AND VISUALIZATION 
 
3.2.1 MICRO-CHANNEL ARRAY VISUALIZATION 
 
Two experimental setups were used in the preliminary study of condensation. In 
the first setup, shown in Figure 3.1, a micro-channel test prototype (Figure 3.2) was 
installed. In preparing the test prototype, the wire-EDM (Electro Discharge Machining) 
process was employed to machine sixty-eight micro-channels in a brass block with 
overall dimensions of 30mm ? 30mm. Each channel was 250 ?m wide and 400 ?m deep 
(hydraulic diameter 0.31mm) with a wall of 175 ?m between adjacent channels. The 
micro-channels were covered from the top using a Plexi-glass? plate sealed along the 
sides with rubber gasket. Visualization of the condensation of vapor occurring in these 
micro-channels was achieved through this Plexi-glass? cover. The condenser was water-
cooled from below as shown in Fig. 3.2. The micro-evaporator has similar geometry but 
 
 
21 
had twenty channels and an overall channel array dimension of 10mm ? 10mm. Heat was 
supplied to the evaporator by means of a rod-shaped electrical heater embedded in the 
brass block underneath the channel structure. 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Figure 3.1: Schematic of the test loop for testing the EDM-machined Micro-condenser 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Figure 3.2: (Top) Schematic of the test prototype (Bottom) Photograph of the assembly from above 
Reservoir
Pressure 
Transducer
Condenser Test 
Section
Evaporator Section
Pre-heater
Filter
Calibrated
Pump
   Plexi-glass Cover 
HFE-7100 vapor HFE-7100 liquid 
Rubber Sealing 
Wire-EDM-created Channels  
400 (h) X 250 (w) X 175 (wall) ?m 
from Brass Alloy 360  
( k = 115 W/m.K) 
 
Cold Water  in Cavity 
From Refrigerated Bath/Circulator 
 
Inlet 
Outlet 
 
 
22 
 
HFE-7100 was used as the working fluid since there is an active interest to use 
this fluid for electronics cooling applications. Also, HFE-7100 is liquid at room 
conditions and was useful in this present case since the test setup was not designed for 
pressurized refrigerants.  
Saturated vapor inlet condition was not applied to the condenser due to the 
concern that the evaporator might experience ?dry out? condition in the high heat duty 
process and the evaporator heater might get burnt. Only medium to low condenser inlet 
qualities were used for sample tests. The EDM-machined micro-condenser, in one such 
test, exhibited heat removal capacity of 26 Watts at a flow rate of 134 kg/m2.s at a 
saturation temperature close to the room condition (25C).  The pressure drop across the 
test channel at this flow rate was 800 Pa. A heat transfer coefficient of 2900 W/m2-K was 
achieved at an inlet vapor quality of 0.24.  Further, it was observed that at higher mass 
fluxes, the bubbles in the two-phase flow entering the condenser inlet, tends to 
concentrate in the few channels near the middle of the channel array due to flow inertia. 
Also, the distribution of the entrant bubbles in the inlet header section was significantly 
influenced by the tilt angle of the test prototype with respect to the gravity vector. Thus, 
although the channels were sufficiently small for gravity to have an effect on the flow 
after it entered the channel, the effect occurred in the header section itself. (With this 
knowledge, the interconnect design described in Appendix A.4 was provided with web 
features to direct the two-phase flow evenly across the array of the channels.). Upon 
entering the channel, the two phase flow took the form of plug flow pattern, with a 
typical vapor bubble between two liquid plugs diminished in volume (due to vapor 
condensation) as the flow proceeded along a channel towards the exit. Small flow 
 
 
23 
oscillations were observed for all flow conditions tested. These oscillations could be 
inherent to the flow process. 
3.2.2 SINGLE MICRO-CHANNEL VISUALIZATION 
A second test setup was based on the thermo-siphon concept, similar in 
configuration to that used by Begg et al. [41] for the validation of their mathematical 
model. A schematic of this test setup is shown in Figure 3.3 and its actual photograph is 
presented in Figure 3.4. The reason this setup was chosen for a preliminary study was 
primarily because of its simplicity, and also in order to confirm and better understand the 
authors? visual observations.  
 
 
 
 
 
 
 
 
 
 
 
Figure 3.3: Schematic of the visualization Test Setup, constructed similar to that used by Begg et al [41]. 
 
The main apparatus in this setup was the Pyrex? U-tube with a tube-in-tube capillary 
condenser connected to its either end by means of flexible rubber hosing. The tube-in-
30 cm
? ?
Pyrex Glass 
Tube
Heater Wrap
Vacuum Stopcock
Rubber 
Hosing
3/8 ? Rubber Stopper
? ?
1 ?
12 cm
2 mm O.D., 1 mm 
I.D.
12 cm
 
 
24 
tube condenser consisted of a cylindrical Plexi-glass water jacket with inlet and outlet 
ports for coolant water flow. The coolant flow was countercurrent to the flow of 
condensing refrigerant in the capillary. The capillary was also of Pyrex? construction 
with inner diameter 1 mm. The effect length of the condenser was 12 cm. To one of the 
arms of the U-tube apparatus a heater wrap was affixed with heat conductive adhesive. 
The other arm of the U-tube was provided with a port equipped with a vacuum valve for 
charging and discharging of the working fluid. As in the previous case, the fluid chosen 
for the study was HFE-7100. To operate the test setup, first the coolant flow was started 
and then slowly the voltage input to the heater was increased to allow heat to dissipate 
into the liquid. The heat input was kept sufficiently low to allow evaporation from the 
surface of the liquid pool in the U-tube arm but to discourage violent boiling activity 
which might destabilize the condensation process in the capillary by introducing pressure 
fluctuations along the inlet path.   During operation of the loop, the pressure drop across 
the flow channel could be estimated from the difference of the levels of liquid in each 
arm of the U-tube and calculating the gravitational head there-from.  
 A test was conducted with heater power input of 4.95 Watts. It was observed that 
the condensation process, once established and stabilized, had an annular flow pattern 
with a net condensation length of 37 mm. However, the end of the vapor core where the 
condensate liquid film converges and marks the beginning of liquid-only flow was 
oscillating continually back and forth along the tube axis and small vapor bubbles were 
getting injected into the liquid flow. This unstable behavior of the oscillating liquid vapor 
meniscus was observed even at lower heat inputs, but with a lesser amplitude. In recent 
literature, some references to this kind of flow have been made. Chen and Cheng [47] had 
 
 
25 
observed this behavior in their study of condensation of pure steam in an array of micro-
channels with hydraulic diameter 82.8 ?m. In their study, the annular flow typically ends 
with ?injection flow? when, intermittently, parcels of vapor mass at the head of the 
annular core, get separated from the core, and get injected into the liquid condensate due 
to oscillation of the meniscus.  
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Figure 3.4 Photograph of the thermo-siphon visualization test setup 
 
 
3.3 DISCUSSION ON OBSERVATIONS IN THE PRESENT PRELIMINARY STUDY 
 
Three issues were apparent from the observations from these two test setups. First, 
it would be difficult to detect and control the uniformity of flow distribution if an array of 
parallel micro-channels were used for more rigorous study. Secondly, even if a single 
micro-channel was used, there remained the possibility flow-induced oscillations. Thirdly, 
 
Tube-in-tube capillary 
condenser with 
counter current 
coolant water flow in 
the jacket 
Enlarged 
Section for 
wrapping foil 
heater 
 
 
26 
thermo-siphon mechanism would not be a sufficient driver for the flow in the loop, and 
an additional pump would be needed, if higher mass-flux cases were to be investigated.  
 
3.4 CFD STUDY OF A BUBBLE MOTION BY VOLUME OF FLUID (VOF) 
APPROACH 
 Under low and moderate mass flux conditions and low local qualities,  where 
liquid and gas flows would both be laminar, an intermittent flow regime had been found 
to operate in small hydraulic diameter micro-channels by Coleman and Garimella [22]. 
This intermittent regime flow is of the pattern of discrete plug and slug flows, such that 
elongated capsule like bubbles, separated from each other by pure liquid slugs and from 
the channel walls by a thin liquid film were found to exist.  These elongated bubbles are 
often termed as Taylor bubbles, and have been subject to scrutiny by different authors 
such as Nijhuis et al. [48], Taha and Cui [49, 50, 51], van Baten and Krishna [52], Zheng 
et al. [53] and others, who employed CFD analysis to understand the hydrodynamics of 
these bubble motions under adiabatic conditions.  
Here, a brief description of the findings from a similar effort in characterizing 
different bubbles in motion in a circular channel with hydraulic diameter 1 mm and 
R134a at saturation temperature of 50C is provided. In this work, an adiabatic condition 
had been studied due to the lack of a comprehensive model for interfacial mass transfer 
that could be implemented in this numerical procedure. 
 
3.4.1 PROBLEM FORMULATION 
  A Volume of Fluid (VOF) approach to solution of the flow field was adopted. 
The CFD software FLUENT (version 6.3.26) was used to model the problem. VOF 
 
 
27 
model is able to track the motion of two immiscible fluids in contact. It solves a single set 
of momentum equation and the interface between the fluids is tracked based on the 
volume fraction of each fluid in each computational cell, throughout the computation 
domain. Thus in case of a gas-liquid flow, a cell having ?g =0 is completely in the liquid 
phase, where as those cells having ?g =1 is in the gas phase. Those cells with 0<?g<1 are 
in the interfacial zone between two the fluids. The details of the CFD model can be 
obtained from Fluent? User?s Guide [54].  
A 1 mm diameter circular tube was chosen for the study so that a 2D axially 
symmetric domain, 0.5 x 10 mm, could be simulated with a high mesh density, taking 
into account the limitation of available computing power. Figure 3.5 shows the details of 
the computational domain. The mesh density chosen for this study was 50 (X) x 500 (Y). 
In order to enable the simplification of study by having the three dimensional problem 
solved with 2D axially symmetric flow, gravity was not chosen to participate in the 
problem. 
The domain was initialized to a fully liquid state except for a region around the 
axis having specification of full vapor. The initial shape of this vapor region that was 
chosen was similar to the axial cross-section of a cylinder with hemispherical caps to aid 
rapid convergence. Any other shape (encompassing same vapor cross-sectional area) 
would eventually lead to similar steady state shape, but would require more computation 
time. 
 
 
 
 
 
28 
 
 
 
 
 
 
 
 
 
 
 
 
Figure 3.5: Schematic describing the flow domain for simulation of a Taylor Bubble 
 
Since the inlet liquid mass flow rate forces the bubble to rise in the Y-direction, in 
order to capture the evolution of the bubble, the wall was given a fictitious velocity, equal 
in magnitude and direction as that of the bubble, so that the bubble would appear 
stationary with respect to the wall. The steady state bubble velocity was unknown and 
hence the wall velocity, initially set to an arbitrary guess value, was continually updated 
in small increments by observing whether the bubble was moving relative to the wall as 
iterations through flow time were progressed. Courant number, which determines the 
solution time step, was set to 0.25 for unsteady iterative solution. 
 
 
Y   
X   
Axis of 
Symmetry   
Wall  moving   
upward  with 
velocity U bubble  
  
  
Domain outlet   
(O utflow)   
Bubble m oving 
up ward with 
velocity U bubble   
Domain Inlet 
(Liquid inlet velocity: Vin) 
 
 
29 
3.4.2 RESULTS AND DISCUSSION 
 The results of this preliminary CFD model are presented in Table 3.1. As seen in 
the table, two aspects of bubble motion were studied, namely,  
(a) effect of mass flux, and  
(b) effect of bubble volume 
 on the bubble shape and the flow pattern in the liquid plug upstream and downstream of 
the bubble. Four different model conditions were chosen for this brief study. For the first 
two conditions, the bubble volume was kept constant and only the inlet mass flux, G was 
changed between low and high values (11 and 110 kg/m2s). For the third and fourth 
conditions, the same flow state was studied with smaller bubble volume. 
 For all diagrams, the flow was from the bottom towards the top, as depicted in 
Figure 3.5. For the larger bubble volume, the influence of refrigerant mass flux on bubble 
shape and liquid flow pattern was clearly visible. In case of lower mass flux case, the 
bubble cap was more rounded than that in higher mass flux case. Also, the film thickness 
increased for larger mass flux case. For lower mass flux, there was a strong toroidal 
vortex structure ahead of the bubble with a similar but weaker and elongated toroidal 
structure behind it. As mass flux increased, this latter vortex shrunk in dimension where 
as the one ahead of the bubble was found elongated to span the entire remaining length of 
the tube. In case of the smaller bubble volume, both the shape and the flow structures 
ahead and behind the bubble were almost identical, with little or no effect of the mass 
flux, at least for the level of mass fluxes investigated here. Also, the bubble velocity, and 
hence the Capillary number, seems to be unaffected by refrigerant mass flux. This 
conclusion should to be verified for other bubble volume and mass flux cases. 
 
 
30 
 
Table 3.1 Summary of the preliminary CFD investigation on adiabatic bubble hydrodynamics 
Ste
ad
y S
tat
e B
ub
ble
 Sh
ap
e 
    
Co
nto
ur 
of 
Str
eam
 Fu
nc
tio
n i
n t
he
 Fl
ow
 D
om
ain
 
    
Mass Flux, G 
(kg/m2.s) 11.02 110.20 11.02 110.20 
Bubble Velocity, 
Ububble (m/s) 0.17 1.6 0.17 1.6 
Capillary Number, 
Ca= ?lUbubble/? 4.97 ? 10
-3  4.68 ? 10-2 4.97 ? 10-3  4.68 ? 10-2 
Bubble Volume 
(mm3) 0.6786 0.6786 0.3958 0.3958 
 
 
31 
It must be borne in mind that only single bubble motions had been studied here. In 
reality, in case of plug/slug flows, a train of bubbles are present. In such a case, the flow 
structure ahead and behind a particular bubble at a particular mass flux would have strong 
similarity due to periodicity of the problem. 
With regards to the phenomenon of condensation, the interesting aspect of the 
present study was the insight into the vortical structures ahead and behind a moving 
bubble. It may be easily understood that due to this motion, the liquid particles were 
continually brought from the liquid layers adjacent to the channel wall into the main flow 
stream. This would cause an enhancement in the wall heat transfer. Taha and Cui [51] 
also cited the earlier findings that the slug flow structures enhanced the radial mass 
transfers in reactors with catalytically active walls and in case of blood flow, the vortical 
motion of the blood plasma ahead and behind a red blood corpuscle augmented the 
distribution of nutrients. In Figure 3.6, the velocity vector map inside and around a larger 
bubble in motion for the higher mass flux case is shown. The colors of the vectors reflect 
the phase of the fluid. It can be seen that there was a vortex structure present inside the 
bubble as well, which might contribute to the heat transfer enhancement, especially if the 
bubble contained superheated vapor inside. Due to the vortical structures present both 
inside and ahead/behind a bubble, the modeling of heat and mass transfer through the 
bubble cap and tail and through the cylindrical side surfaces need to be modeled 
differently. The heat and mass transfer through the side surface may be expected to be of 
stronger influence since the liquid film around the bubble is of minimum thickness here 
and also the liquid motion is fastest in this region. However, as visible from the shape of 
the bubbles for smaller bubble volumes, the cylindrical bubble surface is not present.  
 
 
32 
 
 
 
 
 
 
 
 
 
 
 
Figure 3.6: Velocity vector map inside and around a bubble (G=110.2 kg/m2.s and Bubble volume=0.6786 
mm3) 
 
 In closing this preliminary study, it should be mentioned that in the present case 
the flow studied is adiabatic with zero mass exchange between the two phases. In case of 
actual condensation phenomenon, there is a net heat and mass transfer from the vapor to 
the liquid phases with a corresponding change in momentum. This issue has not been 
addressed here. A detailed note is provided in Chapter 8 concerning the lack of a 
confident understanding/model of the interfacial heat and mass transfer. If such a model 
is established, the present study can be made to include those effects to better reflect the 
real situation.  
 
 
 
 
 
33 
3.4 SUMMARY 
 In the present chapter, three preliminary studies have been presented ? two 
experimental and one numerical CFD-based ? with a view to understand and gain 
confidence in the condensation phenomenon in micro-channels as applicable to the 
functioning of a compact multi-channel condenser. In the experimental studies, the issues 
of flow maldistribution and flow oscillations were highlighted, and in the adiabatic 
numerical study of bubble hydrodynamics, the vortical flow structures inside the vapor 
and liquid regions were identified as factors that may be influencing the heat and mass 
transfer. The study also clearly suggests the need for more detailed flow visualization 
investigation that can realistically represent the actual situation.         
 
 
 
34 
 
Chapter 4 
DESIGN AND FABRICATION OF A MICRO-CHANNEL CONDENSER IN A 
SILICON WAFER USING MEMS TECHNIQUES 
4.1 INTRODUCTION 
 In this chapter, a brief account of activities in relation to design and fabrication of 
a micro-channel condenser in a silicon wafer has been provided. The objective was to 
create a compact two-phase forced convective cooling loop with components made with 
the same material as that of a microchip, i.e. silicon. Silicon micro-channel evaporators 
for heavy duty electronic chips and such other high heat flux application have been 
widely reported in literature. Compact micro-pumps, driven variously by electric current 
and suitable for the above application were reported as well. Therefore, to bring a closure 
to a complete compact thermal management loop, the feasibility study of a micro-
condenser made from a silicon substrate was felt to be needed. If such a condenser was 
found possible to conveniently fabricate, then the condenser, evaporator and micro-pump 
can all be fabricated on one silicon substrate within a small footprint, thereby eliminating 
the necessity of separate interconnections between these components. 
 
4.2 MICRO-CHANNEL CONDENSER DESIGN 
The aim of this design was to etch micro-channel structures onto silicon wafers by 
bulk micromachining techniques. As shown in Figure 4.1, the design consists of wafer-
wafer bonded micro-channel structure supported at ends but fluidic interconnects. Two 
possible varieties are shown in this sketch ? the first one having one wafer bearing the 
channel feature and capped with a blank wafer, and the second one having the channel 
 
 
35 
structure on both wafers and are bonded after the channels are aligned with each other 
and contacted. Clearly, the level of complexity in creating the second variety was more 
since it involved  alignment issues that needed to be dealt with by means of an some 
intra-red camera technique such that the channel structures would be ?visible? through the 
back side. As a first attempt, the design having a blank wafer cap was studied.  
 
 
 
 
 
 
 
 
 
 
Figure 4.1: Cross-section details of silicon micro-channel condenser 
 
 For development purposes, bulk chemical etching of silicon is selected for its 
flexibility and cost issues. Initially a cavity width of 500 ?m, depth of 350 ?m and wall 
width of 100 ?m was chosen to maximize heat transfer area. However, it was found that 
due to fabrication and handling limitations, the resultant wafers were damaged. A revised 
design with reduced cavity width of 330 ?m, depth of 200 ?m and wall width of 250 ?m 
were selected as the final design parameters. The channels are 64 mm long and there are 
68 channels in the array with an overall width of 40 mm. 
Flat region for bonding 40 mm 
Blank Wafer 
330 ?m  
50 ?m 
200 ?m 
 250 ?m 
 
 
36 
 
Channels or grooves can be etched in silicon wafers by either deep reactive ion 
etching (DRIE) or anisotropic wet etching techniques.  High aspect ratios channels can be 
fabricated more conveniently by DRIE than by wet etching, since DRIE produces vertical 
channel walls, whereas wet etching generates angled sidewalls, as illustrated in Figure 
4.2. 
 
 
a 
b 
w
a 
wm = b + 2*a*cot(54.74?) 
Wm = mask opening 
<100> wafer 
b 
 
 (a) (b) 
Figure 4.2: Geometry of the cross-section of channels by (a) DRIE and (b) anisotropic chemical etching 
 
4.3 FABRICATION METHODOLOGY & RESULTS 
4.3.1 PROCESS STEPS 
 
Based on the above design, a set of standard 4 inch diameter and 525 ?m <100> 
polished silicon wafers with pre-deposited 0.1 ?m silicon oxide (SiO2) and 0.3 ?m 
LPCVD silicon nitride (SixNy) were purchased.  Photolithography was performed using a 
standard 5 inch-square chrome-on-glass mask, shown in Figure 4.3.  The pattern was 
developed and cured. Next Reactive Ion Etching (RIE) was performed to produce a 
channel pattern on the silicon oxide ? silicon nitride layer.  Subsequently, the wafer was 
 
 
37 
immersed in an aqueous potassium hydroxide (KOH) solution, a standard silicon etchant, 
in a constant temperature etch bath.  The etching medium was an aqueous solution of 
44% KOH (by weight) at 80 C.  The measured etch rate was 1.4 ?m/min and etch was 
performed for 160 minutes. Towards the end of the etch process, the etch rate was found 
to reduce. Figure 4.4 delineates the various process steps that were carried out in the 
creation of a wafer with the desired micro-channel pattern. The result of etching is shown 
in Figure 4.5.  
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Figure 4.3 Photograph of the mask pattern used for photolithography 
 
 
 
 
 
 
 
 
Bond region 
Escape route for gas 
trapped during bonding 
Alignment 
mark 
4 inch diameter wafer profile 
 
 
38 
 
 
 
 
Figure 4.4: Process flow for fabrication of Silicon micro-channel condenser 
 
 
 
 
 
 
 
 
 
 
 
 
Figure 4.5: (Left) Micro-channel pattern etched on 4 inch diameter <100> Silicon wafer with pre-deposited 
0.1 ?m Silicon Oxide and 0.3 ?m Silicon Nitride (Right) Close-up of the V-groove feature 
 
 
 
 
Si - 500 ?m 
SiO2 (thermal) ? 0.1 ?m 
Photoresist (Shipley? 1813) ? 1.3 
?m 
Cr (E-beam deposited)- 200 ? 
Au (E-beam deposited) ? 8000 ? 
 
SixNy (LPCVD) ? 0.3 ?m 
Blank 
E-beam 
Deposition Cr/Au 
KOH Etch 
Removal of 
Cr/Au 
Aligned 
Wafer Bonding 
RIE 
(CF4: O2) 
Aligned 
Photolithography 
& RIE 
Photolithography 
 
 
39 
4.3.2 WAFER-WAFER FUSION BONDING PROCEDURE 
 
Wafer bonding is more commonly applied to Si-Si or SiO2-SiO2 layers. However, 
recent applications necessitate bonding between SixNy - SixNy . This is because silicon 
nitride, due to its relatively high insensitivity to etching solution can be used as a masking 
layer (with appropriate patterned openings per design). This eliminates the need for 
additional mask layer fabrication steps (such as inert gold layer deposition). However, 
bond strength for the bond between two silicon nitride layers been found to be 
significantly less than that between bare silicon layers or between silicon oxide to silicon 
oxide layers. Such a bonding process has been sparingly mentioned in literature (Sanchez 
et al. [59], Bower et al. [60]). Careful preparation of samples has been reported to result 
in reasonable bond strength in case of Nitride layers. A suggested mechanism for bond 
formation is formation of ?Si-N-Si? covalent bonds across contacting surfaces. Low 
surface roughness is a significant criterion for good bond strength. Direct bonding at 
room temperature is possible though it has been seen to produce lower yield strengths. A 
subsequent step of annealing has been shown to improve bond strength significantly. 
 The main steps in fusion bonding of wafers were:  
(a) chemical cleaning of the wafers to obtain very high surface purity,  
(b) hydration of wafer surfaces, 
(c) contact bonding of the wafers, also called pre-bonding, 
(d) baking of the wafers in a furnace at a high temperature and pressure.  
Cleaning of the wafer surfaces prior to bonding was a very crucial step and 
significantly influenced the bond quality. Because of the high degree of flatness of the 
wafers, any small particle or impurity at the interface might cause weak bond formation 
 
 
40 
or voids. Therefore, a pre-cleaning step was necessary. This step involved the established 
chemical treatment process known as RCA cleaning. The process involved three steps 
(and three wafers baths) as shown in Table 4.1. 
Table 4.1: Wafer Cleaning Procedure (RCA Cleaning) 
 
4.3.3 RESULTS 
Figure 4.6 shows the first result of fusion bond of wafers as per original design 
shown in Figure 4.1 (right). The bonding was performed, for demonstration purpose, at a 
professional foundry facility belonging to an outside vendor, due to the need for wafer 
aligment. The subsequent samples were tested in-house as per simplified design i.e. 
Figure 4.1 (left). This simplified design eliminated the necessity of wafer alignment at 
pre-bond stage thereby rendering the complexity, although it restricted the design to 
channels of smaller hydraulic diameter. 
 
 
 
 
 
Step Step Name  Procedure Details Procedure/Steps 
1 SC-1 5:1:1 of H2O:H2O2:NH4OH @ 70C for 15 mins. 
1. Heat Bath; add NH4OH; add 
H2O2 into hot bath. 
2. Add more Peroxide than 
necessary as it tends to be 
boiled off 
2 HF Dip 50:1 of H2O:HF for 2 mins. 
This removes thin layer of SiO2 
metallic contaminants tend to 
accumulate in step 1. 
3 SC-2 6:1:1 of H2O:H2O2:HCl @ 70C for 15 mins. 1. Add HCl; Heat Bath; add H
2O2 into hot bath. 
 
 
41 
 
 
 
 
 
 
Figure 4.6: Photograph showing result of fusion bond and dicing to create the first prototype micro-
condenser 
 
During the in-house process development, the pre-bonded wafers were inspected 
using an IR-camera-enabled wafer bond inspection system especially designed for the 
present work. The substrate (silicon) and the nitride and oxide layers permit infra-red 
radiation to pass through them. If the infra-red rays encountered voids, multiple 
reflections that occured between the disjoined surfaces, caused visible interference 
fringes to be formed. These fringes could be observed by means of the IR-camera as 
shown in Figure 4.7.  
 
 
 
 
 
 
 
 
 
 
Fusion 
bond 
Microchannels 
 
 
42 
 
 
 
 
 
 
 
 
 
 
 
Figure 4.7 (Left) An assembled set-up for inspection of wafer bond quality consisting of an IR-Viewer 
and a CCD camera for digital capturing. (Right-above) View of sample under microscope (Right-
below) Void fringes exhibited around the periphery of void regions 
 
 
Some samples showed evidence of large void regions, as indicated by the 
appearance of void fringes in the IR images.  Some samples only exhibited small voids at 
the extremeties of the wafers, so these samples were chosen for subsequent annealing 
step to complete the bonding.  Samples were annealed in a furnace at 1000 C for two 
hours with a clamping force of about 10 kN. Figure 4.8 shows a sample result of wafer 
bonding for simplified design produced at the in-house facility. 
 
 
 
 
 
Void Fringes 
Stand 
CCD Camera 
 Relay Lens IR Viewer 
Extension Tube 
4? Wafer 
Holder 
IR-Source 
Computer 
Lab-jack 
 
 
43 
 
 
 
 
 
 
 
Figure 4.8 Photograph showing the result of wafer bonding for the simplified design 
 
4.4 PACKAGING  
Figure 4.9 depicts the packaged micro-channel condenser. Two channeled copper 
heat sinks were fabricated to encase the micro-condenser channel structure from the two 
faces. A thermally-conductive paste was used as an interface layer between the silicon 
condenser and the two heat sinks.  Plastic interconnects were specially designed and 
fabricated by a rapid prototyping process for distributing the fluid to the micro-channels. 
These interconnects were sealed to the micro-condenser with a layer of epoxy glue. 
 
Figure 4.9 (Left) CAD Model showing the package configuration (Right) Photograph of the package. 
 
Channel Opening 
 
 
 
 
Copper Heat Sink
Microchannel Condenser
Copper Heat Sink 
Microchannel 
Condenser 
 
 
44 
4.5 CONCLUSIONS AND SUMMARY 
In this chapter, a process for Silicon Nitride to Silicon Nitride surface bonding 
technique was explored to see the viability of this process for the fabrication of wafer 
stack containing embedded flow structures. It is concluded, that while this fusion bonding 
process is possible to implement in laboratory condition only under great care for surface 
purity and surface flatness, by no means, it is an appropriate bonding technique if 
robustness of the condenser for handling, packaging and testing is desired for. A more 
convenient and productive technique would be anodic bonding to Pyrex? wafer or 
bonding of the Silicon Oxide layers which is a proven process but would require 
additional masking layers such as chrome-gold. Due to the fragile quality that caused 
breakage of the samples at various stages of the development process, sufficient samples 
were not available to complete the study of the fluid flow aspects of these micro-
condensers. In retrospect, it is also desired that substrate wafers be of at least 1 mm thick 
each in order to implement 200 ?m channel depth without risk of breakage.  
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
45 
Chapter 5 
 
DESIGN AND FABRICATION OF THE TEST SECTION MICRO-
CHANNEL 
 
5.1 INTRODUCTION 
 
This section explains the design and fabrication of the micro-channel test section 
that was used to perform experimental tests to generate condensation data. As explained 
in the previous chapter, the objective of creating and testing a micro-channel condenser 
by the bonding of silicon wafers with embedded channel structures was not realized due 
to the lack of robustness of the designed structure. Some recommendations have been 
made in the design and the process steps which may lead to a successful product. 
However, that avenue was not pursued due to a lack of resources and time. Based on the 
preliminary experiments on EDM-machined metallic condenser test section, as described 
in Chapter 3, it was felt that the actual condensation phenomenon should be explored in 
the present work, in order to address the issue of sizing the intended micro-channel array 
condenser for a particular cooling application. However, as found during the literature 
study, at the time of this research work, only a few papers were available to describe 
condensation in a small channel such as desired for the silicon micro-condenser. Hence, a 
more basic study on condensation of common refrigerants to understand the applicability 
of the reported correlations to condensation in small hydraulic diameter channels was 
deemed as the necessary step. Since the present work was intended to be a part of a 
longer study program, pursuing the design alterations recommended for the fabrication of 
a micro-condenser was planned to be a future work. Meanwhile the exploration of the 
basic phenomenon was planned to be completed. For this reason, the remainder of the 
 
 
46 
present work is devoted to the study of condensation phenomenon in a small hydraulic 
diameter and high aspect ratio micro-channels. 
The objective of the design of the test section was to create a high aspect ratio 
micro-channel with hydraulic diameter Dh ? 750 ?m and aspect ratio: 4 ? ar ?7. The 
rationale for choosing a high aspect ratio was that such aspect ratios micro-channels have 
been shown to be effective in flow boiling scenarios and can be expected to perform 
similarly in condensation scenarios as well. Using a multi-port extruded aluminum micro-
channel tube, similar to the works presented in most literature, was considered during 
experimental planning. However, there are several issues related to commercially 
available multiport tubes. First, as experienced during the preliminary tests presented in 
Chapter 3, it is difficult to ensure uniform flow distribution in the different ports at the 
inlet header section. Second, it may not be possible to control the coolant flow outside a 
multi-port tube such that it uniformly affects all flow channels. Third, it would be 
difficult to identify the exact locations for inlet and outlet pressure measurements, as 
entrance losses would be there. Fourth, the fluid flows in the ports would experience 
different wall temperatures on different sides of their rectangular cross-sections, if high 
aspect ratio channel is used. Finally, it would be difficult to obtain such a tube with all 
ports having exact rectangular shapes ? typically the end ports of the tube are of different 
shapes and dimensions than that of the ports in the middle due to extrusion limitations. It 
must be mentioned here that the major benefit of using the multi-port channels, as 
opposed to a single channel, is that the total heat transfer value would be few times the 
value of heat transfer from a single channel under similar inlet mass fluxes and 
temperature and pressure conditions and hence the error generated due to heat losses (or 
 
 
47 
gains) and other parasitic effects, being independent of the total amount of heat transfer 
within the condenser, would possibly be reduced in the multi-port tube case. A single 
channel test with small error margins was one of the main concerns in the design of the 
experimental apparatus. Further discussions on the desired error limits, the design of the 
apparatus, and the control systems to keep error within those limits will be discussed in a 
subsequent chapter.  
 
5.2 DESIGN OF MICRO-CHANNEL PROTOTYPE 
 Based on the reasons presented in the previous section, it was determined that a 
single micro-channel test section would be prepared for the present study. In order to 
obtain a rectangular cross-section with a high aspect ratio (7 ? ar ?10) and having the 
hydraulic diameter Dh ? 750 ?m, a channel with rectangular cross-section of 400 ?m 
?2800 ?m was created. For this channel, the cross sectional area was Ac = 400.2800 ?m2 
= 1120000 ?m2 or 1.12 mm2 and wetted perimeter was be Pw= 3.2 mm. Therefore, the 
hydraulic diameter was Dh = 700 ?m and the aspect ratio was ar = 7.0. This satisfied the 
dimensional criteria.  
The next question was the optimal length of this test section. Typical automotive 
condensers employ multiport tubes with lengths ranging from 300 to 600 mm where as 
for compact heat exchangers for, say, electronics cooling applications, the length of tubes 
is necessarily much smaller, and can be up to an order of magnitude smaller than 
automotive condensers.  
There are other aspects to this issue. First, creating a long test section increases 
the surface area for heat losses; since a single channel test section was to be employed 
 
 
48 
where the lower margin of the heat transfer rate (for the expected range of test 
parameters) would be about 10 W only, the amount of heat loss impacts would the 
significantly affect the accuracy of data . Second, in the case where the channel length is 
short (viz. 50 mm), similar to the EDM-machined micro-channel discussed previously in 
relation to the preliminary tests, the surface area available for heat transfer on the 
refrigerant-side is limited, and hence, in order to accommodate the latent heat transfer 
from the refrigerant at mass fluxes of 50 to 300 kg/m2.s, the temperature difference 
between the wall and the refrigerant need to be significantly high in order to condense a 
vapor flow at quality 1.0 (or even a 15C superheat) at the inlet to a liquid at quality 0.0 at 
the outlet of the channel. In order to have such high temperature differences, the coolant 
needs to operate at temperatures close to or even below the normal freezing point of 
water (for certain test cases) and thus water could no longer be used as coolant. This 
would significantly reduce the possibility of using available instruments and chillers in 
the experimental setup. Furthermore, cold coolant piping has its own issue of heat losses 
at different ambient conditions and it would be difficult to control such losses in a 
laboratory environment. Thus, operations at reduced coolant temperatures are difficult to 
design and also to control. Another problem with using very short test section is that the 
entrance effects and flow patterns of both the coolant and the refrigerant can be 
significantly different than the representative cases if the data is to be used for 
comparison with that from literature.  The only advantage of a short test section is that 
the test section surface area (i.e. coolant jacket area exposed to ambient air) for heat 
transfer is minimized which could reduce the parasitic heat transfer.  
 
 
49 
Since the intent of the present experimental study is to aid in the design of 
compact condenser heat exchangers, a moderate figure of 200 mm was selected for the 
test channel length. This length is comparable to those used in some of the recent 
condensation studies (Garimella et al. [35], Begg et al. [41]) of similar micro-channel 
hydraulic diameters, and hence an additional advantage is that their data can be compared 
with the present work to check validity of the data from the present work.  
Entrance and exit portions of the channel were expanded into a cylindrical flow 
cross-section to accommodate inlet and outlet tubes of 1/8? O.D. These tubes were 
stainless steel and thin-walled (0.1 mm wall thickness) to minimize axial heat conduction 
losses (or gains) from the test section. 1/16? O.D. pressure ports were created 5 mm away 
from the entrance and exit ends of the channel. Therefore, the effective length of micro-
channel flow (from the perspective of the pressure drop measurement), measured from 
center of entrance pressure port to the center of exit pressure port, was 190 mm. However, 
the entire 200 mm length is available for heat transfer. 
Next the wall thickness of the channel had to be decided. On one hand, a thin 
walled tube, subject to possible corrosion, could leak or even burst at high refrigerant 
operating pressures (up to 20 bars). There is also a possibility that the test channel could 
bulge in the middle and cause intended rectangular cross-section to distort.  On the other 
hand, a thick wall selection could lead to a slightly higher temperature difference between 
the refrigerant and the coolant and could also be affected by axial conduction. A 
controlled axial conduction is not entirely without benefit, as it can assist in keeping wall 
temperature uniform along the length of the test section. An analysis performed in 
ANSYS indicated that a wall thickness of 2 mm, with for copper as the wall material, is 
 
 
50 
sufficient to prevent burst and also maintains the intended rectangular shape at elevated 
refrigerant pressures of up to 20 bars. Figure 5.1 indicates the model in ANSYS for 
structural analysis. At 2 mm wall thickness, the model indicates that the maximum 
deflection at the middle of the channel is 4 ?m which is 1% of the channel width. This 
value of deflection is acceptable. If we had a channel with a smaller aspect ratio (i.e. 
higher channel width), then the deflection might be expected to be even smaller. 
 
 
 
 
 
 
 
 
 
Figure 5.1: An ANSYS model for structural analysis of the micro-channel 
 
Based on these considerations, the final structure was a channel 200 mm x 0.4 
mm x 2.8 mm and was created in a bar of copper with dimensions 200 mm x 6.8 mm x 
4.4 mm with 2.0 mm wall thickness on all sides of the channel. An exploded CAD model 
of the micro-channel with design dimensions is shown in Figure 5.2. 
 
 
 
 
 
51 
 
 
 
 
 
 
 
 
 
 
Figure 5.2: Exploded view of CAD model of the micro-channel test section. Inset: Detailed view of channel 
end 
 
5.3 FABRICATION METHODOLOGY 
The goal of the micro-channel fabrication was to create a rectangular cross-
section with minimal distortion and to maintain that cross-section straight along the entire 
length of the channel.  
The first step in the strategy for fabrication was to determine whether it was 
appropriate to create a one-piece channel (as in extrusion) or a two-piece structure (an 
open groove with a lid) which would be subsequently joined. With a view to create a one 
piece structure, extrusion process was considered. However, creating a customized 
extrusion die is expensive. Alternatively, to be a process similar to the fabrication of 
jewelry, was considered. In this process a wax mandrel was produced and its surface was 
coated with a thin silver solution, which was electrically conductive. After this solution 
dried, the mandrel piece was immersed in an appropriate electrolytic bath and the surface 
400 ?m 
Micro-channel 
400 ?m x 2800 ?m 
  
. Pressure Port 
1/16? O.D. 
  
 Flow Port 
1/8? O.D. 
  
 
 
52 
was metalized by electro-deposition. The thickness of deposited metal could be 
controlled by current and time of deposition. Subsequently, the metalized mandrel was 
gently heated up to the melting point of wax (70 to 80C) and the molten wax was drained 
off. As a trial for creating a wax mandrel for the micro-channel, for understanding the 
challenges that could be associated with this process, a portable extrusion setup was 
assembled in the laboratory with a controllable motor and screw injector. A trial 
extrusion die was prepared to study the process controls. Several difficulties presented 
themselves in this operation. The wax mandrel did not retain the intended shape as wax 
was found to expand after it passes out from the die. Secondly, it was difficult to keep the 
200 mm length of the wax mandrel straight during handling as wax is very pliable. Still a 
trial electro-deposition of copper was performed in a bath of copper sulphate solution to 
observe the final outcome. The result at the end of this trial run is exhibited in Figure 5.3. 
 
 
 
 
Figure 5.3: Left ? Schematic of copper electro-deposition setup. Right ? Result of first trial with wax 
mandrel. 
 
Air 
Supply 
+ - 
Copper 
Anode 
Aqueous CuSO4 at room temperature 
Wax Strip 
with 
Conductive 
Silver Paint 
Copper Electroplating Setup 
 
 
53 
This channel, created wholly from electro-deposition, had a distorted internal 
geometry, and the external surface of the channel exhibited uneven nodular formations 
because the gas bubbles that formed due to the deposition chemical reaction were 
removed too slowly. A stream of air was introduced, as shown in Figure 5.3 to augment 
the bubble breakout process. However, it needed significant attention and adjustments to 
produce smoother external surface of the channel. Although the final outcome was not 
completely discouraging, it was clear that this process inherently possessed significant 
logistical challenges and control issues and therefore, was deemed not worth pursuing. 
Instead, a traditional machining approach would be adequate. 
Electro-discharge machining (EDM) was initially thought to be the appropriate 
machining technique in this situation. The available cutting wire tool had various wire 
gages ranging from 0.125 mm and higher. However, because the material choice was 
copper, and because the length of cut was 200 mm it was advised that a thicker wire tool 
be used to avoid wire breakage. Also, it would be difficult, yet possible, to maintain a 400 
?m channel width while avoiding pitting and other roughening features on the side wall. 
Another reason was that due to the wire tool shape, the channel end profile would be 
rounded instead of the desired right-angled corners. Although a rounded end would not 
significantly alter the channel cross-sectional area and hydraulic diameter values, it 
would still be a defect. Therefore, it was decided that a computer-controlled milling 
operation would be a simple and appropriate method of channel creation. Further, a 400 
?m thick slitting saw blade was available commercially and would be a suitable tool.  
 After the channel was created by the well-controlled slitting operation, it was 
inspected for defects and measurements were taken with electronic calipers with 
 
 
54 
precision of 1 ?m. The next step was to create a channel cover. Typically, such channels 
are covered by brazing of a metal plate. However, it was feared that by such a brazing 
operation the channel height was not guaranteed to stay at design dimensions. Also the 
heat applied during brazing could distort the channel shape. It was determined that a 
more benign approach would be desirable. 
 The process flow for creating an integrated channel cover is shown in Figure 5.4. 
Step 1 was the creation of a 4.8 mm x 4.4 mm x 200 mm oxygen-free 99.9% pure Copper 
blank. A channel 2.8 mm x 0.4 mm was created in this blank (Step 2) by the well-
controlled machining process described in the previous paragraph. All oil and other 
organic deposits were cleaned from the surface of the structure by immersing it in a jar of 
10% strength aqueous solution of sulphuric acid. Then structure was cleaned in water and 
dried, and then jeweler?s wax was pressed into this channel (Step 3) to create a protective 
layer to prevent electro-deposition inside the channel. Prior to Step 4, the inlet, outlet and 
pressure port tubing pieces were brazed to the channel. Step 4 was a crucial preparatory 
step involving several sub-steps. First, remnants of wax were carefully scraped off from 
the top of the channel with a sharp razor blade, leaving the wax filled inside the channel 
and flush at the surface. Then a wax-dissolving solution was applied on the exposed 
copper surface using a cotton swab to further dissolve the wax and any organic residues 
that may contaminate the surface preventing electro-deposition. The remaining three 
faces of the channel structure were coated with paint to protect them from deposition. 
Finally, a fine paint brush was used to apply a coating of silver solution on the exposed 
wax layer so that it was ready to be coated with copper. After the silver solution dried, 
the structure was placed inside a freshly prepared copper sulphate bath and air flow was 
 
 
55 
provided using a small air pump to continually remove gas bubbles formed on the 
cathode surface. The current flow, the temperature of the bath, and immersion time were 
recorded to ensure desired deposition thickness.  
 
 
Figure 5.4: Six-step process in machining and closing a channel. 
 
After desired thickness of deposition was achieved, the structure was removed 
from the bath and the residual solution was cleaned off the object (Step 5). It was 
observed that some undesired nodule growths began to form from the painted sides of the 
channel structure, as shown in Figure 5.5. This is because the paint layer, although not 
electrically conductive per se, acts as a weak dielectric layer with certain impurities 
causing low conduction at certain places. These nodular growths were weakly attached to 
the object and were broken away by finger pressure. The last step, Step 6, was to clean 
the wax present inside by passing boiling water through the micro-channel and melting 
4.4 mm 
4.8 mm 
1 2 3 
6 
Jeweler?s 
Wax 
4 Non-conductive Paint for Protection 5 
~ 6.8 mm 
Micro-layer of 
Conductive Silver 
Paint 
 
 
56 
the wax. As an added precaution, the wax-dissolving solution was also passed through 
the micro-channel to dissolve any residual wax left in the channel.  This brings the 
process of fabrication to a conclusion. 
 
Figure 5.5: Top ? Micro-channel structure with integrated flow and pressure ports before deposition. 
Bottom Left ? Electro-deposited cover. Bottom Right ? Undesirable but removable nodular growth of 
Copper on the painted surfaces. 
 
5.4 LEAKAGE TESTING AND VALIDATION OF CHANNEL GEOMETRY 
 After the channel was created, it was connected to a compressed air cylinder by 
Swagelok? fittings and was tested for leakage by holding the channel under water and 
subjecting the channel to an internal air pressure of up to 20 bars. The channel was kept 
submerged at this condition for several hours. However, no signature of bubble formation 
was observed, attesting to the integrity of the channel. 
Although great care was taken to create channel geometry close to the design of 
the micro-channel, it was felt that an additional step of validation would be useful. After 
the test channel was assembled in the test apparatus, as described in the next chapter, a 
test was run in which liquid R245fa refrigerant at a known temperature, T=23.4 C and 
Mechanically cut 
channel 
0.4 mm X 2.8 mm 
Removable 
Copper nodules 
 
 
57 
pressure, P=140.7 kPa was circulated in the loop and pressure drop measurements 
between the inlet and exit ends of the channel were recorded by using a calibrated 
pressure transducer. These experimentally obtained data were compared with pressure 
drop obtained from CFD simulation with the FLUENT? software program by creating an 
exact design model of the channel. This comparison is shown in the graph in Figure 5.6. 
This corroboration, along with the visual measurements of the channel dimensions after 
machining operation, provided confidence that the final form of the channel closely 
matched the design dimensions. 
 
Figure 5.6: Graph of comparison of CFD simulation and experimental measurements of pressure drop 
across the micro-channel length. 
 
 
 
 
Liquid Mass Flux vs. Pressure Drop
(R-245fa)
0
0.2
0.4
0.6
0.8
1
1.2
1.4
0 50 100 150 200 250 300
Liquid Mass Flux in Channel (kg/m2.s)
Pr
es
su
re
 D
ro
p (
kP
a)
Experimental
Numerical Model
Linear (Experimental)
R245fa at 23.4C
 
 
58 
5.5 SUMMARY 
This chapter presents the criteria and thought process that led to the design of a 
non-traditional fabrication method for creating a suitable micro-channel structure for the 
present experimental work. Data from simulation software tools and adoption of a trial-
and-error approach were used to fine tune and/or validate the design and fabrication 
process. The result was a carefully prepared micro-channel structure with integrated flow 
and pressure ports that was ready to be incorporated in the test apparatus, described in the 
next chapter. 
 
 
 
59 
Chapter 6 
 
EXPERIMENTAL APPARATUS AND INSTRUMENTATION 
 
 
6.1 INTRODUCTION 
 
 It was necessary for the design of the experimental apparatus was to 
accommodate condensation tests for two refrigerants ? one, a high pressure refrigerant 
and another, a refrigerant which is in liquid state at temperature and pressures close to 
room conditions. The selection and properties of the two refrigerants, one of which was 
determined to be R134a, are discussed in a later section of the present chapter. The setup 
also needed to operate with R134a at mass fluxes through the channel: from 50 to 300 
kg/m2.s, saturation pressures from 30 to 70 C, and inlet super-heats from 0 to 20 C. These 
ranges are typical for compact condenser applications. It was also important to contain 
the mass flux measurement error within ? 5% and energy balance, uncertainty in heat 
transfer coefficients and pressure drops for each data point within ? 15%. These limits, 
especially the uncertainty limit on the heat transfer coefficient, are critical since many 
results in available literature are reported as much as 30 %. It was mentioned in Chapter 5 
that because a single channel was employed in this investigation, the heat transfer was 
about 10 W on the lower side of the heat inputs and such a low value of heat power 
would be especially subject to error in measurements and uncertainty due to heat losses. 
Hence it was necessary in the design of the apparatus to counter such errors as far as 
possible. Also, some of the measurement and instrumentation problems revealed 
themselves only during the course of study and hence several modifications were made to 
improve the system after the first configuration was created. The final form of the 
apparatus is presented here with some developmental notes as necessary. 
 
 
60 
6.2 DESCRIPTION OF APPARATUS 
 As mentioned before, the objective of the design of experimental set-up was to 
enable safe operation of a high pressure refrigerant with measurement accuracies at the 
level required for this study. The set-up consisted of two loops ?the refrigerant loop and 
the coolant water loop, as shown in Figure 6.1. The parts of these loops are described in 
the following sections. 
 
6.2.1 DESCRIPTION OF REFRIGERANT FLOW LOOP 
 
 The refrigerant loop consisted mainly of a tube evaporator or vapor generator, a 
condenser test section, a visualization section down stream of the condenser, a sub-cooler, 
a fluid reservoir/gas-liquid separator, a ceramic particulate filter and finally a micro-
pump with a built-in flow measurement mechanism.  
 
6.2.1.1 Tube Evaporator  
 The purpose of the evaporator located upstream to the condenser test section was 
to provide adequate heat input to the refrigerant flow with which to generate saturated 
and superheated vapor conditions at the entrance to the condenser test section.
 
 
61 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Figure 6.1: Schematic of the experimental setup 
 
+ 
_ 
B 
C 
V 
 
_ 
 P I D  
 PID 
+ 
+ 
I 
I 
B 
C 
Ev
ap
ora
tor
 
 C o n d e n s e r  
Vis
ua
liz
ati
on
 
Se
cti
on
 
Su
b-c
oo
ler
 
Mi
cro
pu
mp
 
Di
ffe
ren
tia
l 
Pr
es
su
re 
Tr
an
sd
uc
er 
Sh
un
t  
Va
lve
 
Ab
so
lut
e 
Pr
es
su
re 
Tr
an
sd
uc
er 
Sa
fet
y  
Va
lve
 
Ca
pil
lar
y Fl
ow
 M
ete
rs 
LO
 Fl
ow
 
HI
 Fl
ow
 
Re
frig
era
ted
 
Ba
th/
Ci
rcu
lat
or 
DC
 P
ow
er 
Su
pp
ly 
A 
Re
se
rvo
ir 
Le
ve
l 
Tr
an
sp
are
nt 
Te
flo
n Tu
be
 
Co
pp
er 
Elb
ow
 
(H
ea
t S
hie
ld)
 
T w
i1
 
 T w
i2
 
 T w
o2
 
T w
o1
 
T C
on
d-I
N 
 
T C
on
d-
OU
T 
T E
va
p-I
N 
T R
es
v1
 
 
T R
es
v3
 
 
T R
es
v2
 
 
 
Me
ter
ing
  
Va
lve
 
A 
Di
sc
ha
rge
  
Va
lve
 
He
ate
r 
 H e a t e r  
Co
ola
nt 
IN
 
Co
ola
nt 
OU
T 
+ 
 
 
62 
 The evaporator consisted of a coil of heater wire housed in a 3/16 inch I.D. 
thick-walled Pyrex? glass tube (wall thickness 2.0 mm) and 80 mm length and fitted 
with thin-walled stainless steel inlet and outlet flow ports (wall thickness 0.1 mm thick). 
A transparent Pyrex tube was used as the housing for the heater element to enable visual 
observation of the flow at the exit of the evaporator. The thin-walled stainless steel tubing 
for the inlet and the outlet was used to minimize the axial conduction of heat along the 
port walls. The ports were connected to the Pyrex piece with high temperature epoxy 
resin with carbon fiber reinforcement to prevent cracking of the glass. A tightly wound 
coil of Nichrome wire (80% nickel, 20% chromium) that served as the electrical heating 
element was inserted into the Pyrex glass tube and the two ends of the heating element 
were brazed to the inlet and outlet ports, respectively. Thus the inlet and the outlet ports 
also acted as electrical leads. This helped to eliminate the need for a complicated fluid 
connector through which external power supply leads could get access to the heating 
element. This reduced complexity in plumbing which in turn reduces possibility of 
refrigerant leakage. 
 It was estimated that the maximum heat input requirement amongst 
different test conditions would be about 70 W. The net resistance of the heating element 
including the resistances at the braze contacts and the lead arrangement described 
previously was about 54 ?. This value was decided upon apriori because the D.C. power 
supply acquired for heat input to the evaporator was equipped with an ammeter with 
having range of 0 - 1.2 A. The net resistance value exhibited a variation within 10% for 
different flow and heat input conditions. This was expected as the Nichrome wire 
resistivity varies as wire temperature varies. However, the heat input was calculated 
 
 
63 
based on input voltage and current measurements and hence heater resistance change did 
not influence the data. 
 The evaporator test section was pressure-tested in a separate loop 
consisting of a reciprocating oil pump and absolute pressure gage. The evaporator was 
subjected to an internal oil pressure of 400 psia sustained for a period of few hours while 
it was kept in an oven with an internal temperature of 70 C. It was concluded that the 
high temperature epoxy joint did not degrade at the elevated temperature and pressure 
and hence was acceptable for operation in the loop. Figure 6.2 shows the evaporator 
configuration described here.  
 
 
 
 
 
 
 
 
 
 
Figure 6.2: Tube evaporator with the labels indicating the various parts described in text. 
 
Both the evaporator and the condenser test sections were placed in double-walled 
Pyrex Dewar cylinders as shown in Figure 6.3. The reason for using a Dewar was to 
minimized heat loss/gain to the system. A high vacuum (2 millibars) created between the 
Dewar walls eliminates the possibility of convective heat losses. Each Dewar shell was 
50.8 mm O.D. and 25.4 mm I.D. As an additional measure of safety, the tube evaporator 
was placed inside a ? inch I.D. transparent plexi-glass pipe of appropriate length before 
Heating Element 
Pyrex Tube Housing 
Outlet port brazed 
internally to the heater wire 
Carbon fiber reinforced high 
temperature epoxy joint 
Swagelok 
Connector 
 
 
64 
the assembly was placed inside the eEvaporator Dewar shell. This was to prevent any 
damage to the Dewar cylinders in case the evaporator happened to shatter under high 
pressure.  
 
 
 
 
 
 
 
 
 
 
 
Figure 6.3:  Photograph of the two heat-insulating double-walled Dewar cylinders used in the experimental 
apparatus. 
 
6.2.1.2 Micro-channel Condenser Test Section 
 
 The design and fabrication of the micro-channel condenser test structure was 
described in details in chapter 5. This subsection elaborates on the integration and 
assembly of the test section structure with the refrigerant loop.  
Five type-T thermocouples were embedded into the wall of the condenser test 
channel. The two halves of the type-T thermocouple are copper and constantan. Since 
99.9% pure oxygen-free copper was used as the material for the micro-channel structure, 
it was appropriate to use the channel wall itself to represent one half of the type-T 
Condenser Dewar 
Evaporator Dewar 
50.8 mm 
O.D. 25.4 mm I.D. 
200 mm 
152 mm 
 
 
65 
thermocouple. Hence, five AWG 24 Teflon-sheath insulated constantan wires were 
soldered onto the channel wall at equi-spaced location. The diagram in Figure 6.4 
indicates the electrical connections forming the five thermocouple junctions. 
It was found necessary to use thicker gage constantan wires as thinner wires are 
more fragile and are often broken, thus resulting in disconnection. However, if for each 
thermocouple, a set of two wires were used, there would have been too many wires to be 
routed out of the Dewar cylinder and this might prevent the coolant to be uniformly 
distributed across the flow cross-section. Thus, using the material of the channel wall as 
?the second arm? of the thermocouples was helpful in reducing complexity.  
 
 
Figure 6.4: Diagram depicting the thermocouple junctions formed by Constantan soldered to the copper 
micro-channel structure. 
 
As described in the above section, Dewar cylinders were used as heat-insulating 
casings for the evaporator and the condenser test section. In case of the condenser test 
a 
a/2 
a 
a 
a 
Tw1 
Tw2 
Tw4 
Tw1 
Tw5 
a 
Copper Wire, AWG 24 
Constantan Wire, AWG 24 
Solder Joint 
= 40 mm 
Micro-channel 
Condenser 
Wall Thermocouple Circuit 
a/2 
 
 
66 
section, the Dewar cylinder additionally acted as a coolant water jacket as depicted in 
Figure 6.5.  
 
 
 
 
 
 
` 
 
 
 
 
 
 
 
 
 
 
 
Figure 6.5: Part of the refrigerant loop indicating the evaporator and condenser test section 
encased in double-wall glass Dewar cylinders. 
 
It was critical to contain the heat loss from the metal interconnect between the 
evaporator and the condenser section. Traditional heat insulations such as glass wool 
fillings and rubber foams were not appropriate, as this portion of the refrigerant flow path 
was somewhat complex and it was difficult to apply such insulations to this portion of the 
setup. First, this inter-connect provided an access to the thermocouple measuring 
temperature of the refrigerant at the inlet of the condenser. Second, this region also 
contained connections for coolant outlet and connection to the pressure port. Due to these 
connections, it was necessary to check for any refrigerant leakages by sniff tests which 
Condenser 
Section 
Sub-cooler 
Heat Exchanger 
Evaporator 
Section 
Visualization 
Section 
Pressure taps 
Copper 
Lead Wire 
Five Constantan Wires 
Coolant Inlet 
Refrigerant Flow 
Coolant Flow 
 
 
67 
would be difficult with traditional insulation. An alternative approach was therefore 
adopted. As shown in Figure 6.6, the two glass Dewar Cylinders were bridged by a 2 inch 
diameter copper elbow connected that was split in half.  
  
Figure 6.6: Left ? The bottom half of the Copper elbow bridging the evaporator and condenser Dewar 
segments. Right ? The top half of the elbow placed to complete the active heat shield. 
 
On the outer surfaces of the two halves of the copper elbow, two Kapton? film 
heater patches were affixed and a thermocouple was soldered on to the surface of the 
elbow to monitor its temperature. Heat inputs to these two heater patches were controlled 
manually by a variac so that the reading of the above thermocouple was kept close to the 
temperature reading of the refrigerant inlet thermocouple. Thus the split copper acted as 
an active heat shield and eliminated formation of convection currents within its volume 
that would cause heat dissipation. Even then, some heat losses were expected due to 
conduction along the pressure port tube and the connection to the absolute pressure 
transducer. However, this arrangement provided the necessary flexibility of monitoring 
refrigerant leakages and also provided access to the inter-connect when needed. 
 
 
 
68 
6.2.1.3 Visualization Section 
 The visualization section downstream to the condenser allowed visual 
confirmation of the completion of the condensation process. Figure 6.5 includes an image 
of the visualization section and indicates its location within the refrigerant loop. 
There is a difference between thermodynamically complete condensation and 
visually complete condensation, as described in the next chapter. In the former, the 
average enthalpy of the refrigerant at the outlet to the condenser is equal to the enthalpy 
of a saturated liquid. However, even at this condition, small vapor bubbles exist within a 
sub-cooled liquid. The condensation and elimination of all vapor bubbles require a 
certain period of time for heat dissipation. In order to completely suppress all vapor 
bubbles immediately at the exit of the condenser, sufficient sub-cooling was necessary. 
The amount of sub-cooling depended on the operating mass flux, and was found to have a 
range between 2 and 4 C. However, this sub-cooling effect was also dependant on the 
outlet geometry of the condenser and hence the sub-cooling levels necessary for the 
present test section and test conditions might not be replicated in another condenser with 
similar channel geometry. Also, the tests run at the two different outlet conditions ? 
thermodynamically-complete and visually-complete ? produced results for heat transfer 
coefficient that differed by up to 5%. Since the amount of sub-cooling necessary to reach 
the visually-complete state depended on test condition and outlet geometry and also 
because accurate control of sub-cooling temperature was found difficult, it was 
determined that the thermodynamically-complete state was a better target. 
Therefore, the visualization test section served as a monitoring window to identify 
the above difference between two exit conditions. The fabrication and pressure testing of 
 
 
69 
this section was similar to that of the tube evaporator, which has been already been 
discussed. 
6.2.1.4 Sub-cooler Section 
 The purpose of the sub-cooler section was to further extract heat from the 
refrigerant flow coming out of the condenser in order to suppress all remaining vapor 
bubbles. The necessity to suppress the bubbles was to ensure that only liquid refrigerant 
was passed into the micro-pump. Also, if sufficient sub-cooling was provided, the 
possibility of cavitation inside the micro-pump was eliminated. 
 The sub-cooler heat exchanger was a simple tube-in-tube heat exchanger in which 
coolant water flowed in the outer tube and refrigerant in the inner tube. To increase the 
heat transfer coefficient on the coolant side of the heat exchanger, copper wire mesh was 
inserted in the cavity between the inner and the outer tubes. The presence of the wire 
mesh enabled flow mixing. It should be noted that the sub-cooler was designed to take a 
heat load of no more that 10 W. Figure 6.5 indicates its location in the refrigerant loop. 
 
6.2.1.5 Fluid Reservoir and Vapor-Liquid Separator 
 The fluid reservoir served several important functions in the refrigerant loop. First, 
as its name implies, it acted as a reservoir for holding excess liquid refrigerant in the loop. 
In any refrigeration system, the refrigerant loop cannot be allowed to be filled with liquid 
only since the loop pressure would experience large differences due to the variation in 
operating temperature. Hence, the reservoir was meant to provide a ?cushion volume? for 
extra refrigerant to flow into it, if the pressure level in the loop increased.  
 
 
70 
Second, the reservoir also acted as a separator of any non-condensable (or 
condensable) gas/vapor bubble that emerged from the sub-cooler heat exchanger before 
the flow entered into the micro-pump. Although great precautions were taken to secure 
the refrigerant loop against any possible air leaks during charging and operation of the 
system, it was possible that a miniscule amount of non-condensable gas might have 
existed inside the refrigerant loop. The reservoir provided an opportunity for any trace 
non-condensable gas to be segregated from the flow after a few trial test runs. For this 
function, it was necessary that the reservoir was to be kept as a higher elevation than the 
rest of the loop. 
 Third, the reservoir also acted as an essential component in the feed-back loop 
for control of absolute pressure in the system. In order to measure the saturation 
temperature of the refrigerant at the inlet to the condenser, an absolute pressure 
transducer was connected to the 4-way interconnect which also accepted the connection 
to the pressure port as shown in schematic in Figure 6.1.  
The reservoir cylinder, as shown in the photograph in Figure 6.7, was a 
Swagelok? double-ended 316 stainless steel sample cylinder with an internal volume of 
300 cc. The top end of the reservoir cylinder was fitted to a 4-way connector. A 350 psia 
spring-loaded safety valve was fitted to this connector. A thin-walled 1/8-inch stainless 
steel tube, which was blinded at one end, was inserted into the reservoir in such a way 
that the blinded end of the tubing reached the mid-section of the reservoir cylinder. Three 
calibrated thermocouples were inserted into the tubing so that their tips were successively 
placed 1 inch apart as shown in the schematic. Due to varying levels at of the refrigerant 
at different operating points, it was best not to locate the tips of the three thermocouples 
 
 
71 
at one place only. The average readings of the three thermocouples would be a better 
estimate of the saturation temperature of the mixture.  
A 3/8-inch Teflon tube was connected in parallel to the reservoir; its function was 
to enable monitoring the fluid level inside the reservoir during the refrigerant charging 
process.  It was desired that of the three thermocouples, the middle thermocouple was to 
be kept approximately close to the vapor liquid interface inside the reservoir for typical 
operating conditions, and the level monitoring tube served as a visual tool to ensure that. 
During the heat transfer experimentation, the liquid level inside the reservoir typically 
shifted by about 2 to 3 mm due to presence of vapor in the evaporator and condenser 
sections.  
The reservoir was provided with a heater element wrapped around it at about the 
location of the liquid level. Power input to this heater was supplied by a DC power 
supply. In order to control the operating absolute pressure at the entrance to the 
condenser, the voltage signal from the absolute pressure transducer was fed to a PID 
controller which controlled the heat input to the reservoir heater. At a steady state 
operation of the system, the PID controller altered between a ?switched on? and 
?switched off? states; the duration of the stay at each state was affected, primarily, by the 
ambient conditions.  
However, during the actual operation of the experimental apparatus, it was found 
that the controller was ineffective in dampening the fluctuation of pressure in the system. 
This was possible for two reasons. Firstly, there was a response delay between the instant 
the reservoir heater was ?switched on? and instant when pressure increase was noted by 
the absolute pressure transducer. This was because there were several steps in this control 
 
 
72 
operation: the increase in the surface temperature of the reservoir translated to increase in 
the saturation pressure of the two-phase mixture contained in the reservoir. This pressure 
increase altered the effective average pressure inside the condenser which needed some 
time to stabilize. Once stabilization was reached, the corresponding pressure signal 
directed the PID controller to switch off and the surface temperature of the reservoir 
gradually reduced, thus completing the feed-back loop. These several steps contributed to 
the response delay. Secondly, due to the boiling process inside the tube evaporator, the 
absolute pressure transducer experienced certain low amplitude fluctuation which 
perhaps caused a certain amount of difficulty/delay for the controller in its switching 
decision and may have resulted in pressure ?swings? about the control point.  
In order to minimize the response delay and to measure the system?s operating 
pressure with greater accuracy, the signal from one of three thermocouples (in the tube 
that was inserted into the reservoir) to measure the saturation pressure was used as the 
control signal for the PID controller. It was observed that this significantly improved the 
response delay and the overall stability of the system. 
 
 
 
 
 
73 
 
Figure 6.7: Schematic illustrating the various parts of the Reservoir System 
 
 
6.2.1.6 Micro-pump 
 An ISMATEC? MCP-Z Standard gear-based micro-pump was selected as the 
prime-mover for the working fluid. The pump was equipped with control electronics 
which monitored, under different menu-based choices, the rotational speed, dispensed 
volume or the number of rotations. The flow range of the pump was 0 to 180 ml/min and 
it could be calibrated. The pump gear head was reported by the manufacturer to consist of 
T1 
T3 
T2 
1 
2 
3 
Liquid 
Vapor Reservoir 
Liquid Level 
?Pressure-control? 
heater wrap 
316 Stainless Steel 
Reservoir Cylinder 
4-way Connector 
Thin-walled 
Steel Tube 
Liquid Level 
Monitoring 
Tube 
+ 
? 
Safety Valve 
(Set at 350 psia) 
 
 
74 
accurately machined parts and the product of swept volume per rotation to the rotational 
speed of the gear head constituted the flow rate delivered by the pump. Due to a unique 
?suction shoe? arrangement in the gear-head, the flow was unidirectional and the flow 
rate was independent of the temperature and pressure drop across the pump for the 
operating range. Also, the flow was observed to be nearly pulse-free. 
The micro-pump flow rate was calibrated using distilled water at room 
temperature. A fixed number of rotations were allowed and then the weight of the 
dispensed water was measured to generate a flow calibration curve. The calibration factor, 
thus generated, was incorporated in the control memory of the pump so that the digital 
readout reflected the calibrated flow rate.  
Since the micro-pump was sensitive to any particulate in the flow, as a precaution, 
a 2 ?m porous ceramic filter was inserted in the refrigerant loop upstream to the micro-
pump. 
 
6.2.1.7 Refrigerant Loop Controls and Measurements 
The purpose of this section is to list the various measurements that were made at 
different locations of the refrigerant loop. Also, various control loops were created to 
maintain constancy of pressure or temperature and these will be discussed in the present 
subsection as well. 
Several thermocouple-based temperature measurements were made at different 
locations in the refrigerant loop. Primarily Omega? type-T AWG32 thermocouples with 
Teflon wire insulations were used. The copper and constantan wires of each 
thermocouple were butt-joined by a spark discharge and melting process which 
 
 
75 
consistently created small ball shapes. Also, unlike solder joints, this process of creation 
of the thermocouple junction did not introduce a foreign metal into the joint. Therefore it 
was expected that the thermocouple junctions would exhibit temperature responses close 
to each other. 
Temperature measurements were made at the inlets to the evaporator and to the 
condenser and at the exit of the condenser. In each case, the thermocouple tips were 
inserted into the tubing up to the entrance or exit location. Temperature measurements 
were also made by the five thermocouple junctions that were created on the wall of the 
micro-channel as described in Section 6.2.1.2. However, these thermocouples, unlike the 
other thermocouples in the system were AWG24 gage. Three thermocouples that 
measured the liquid-vapor interface temperature (or the saturation temperature) inside the 
reservoir cylinder as described in Section 6.2.1.4. A thermocouple was implanted on the 
wall of the copper elbow bridging the two Dewar cylinders to measure the surface 
temperature of the elbow. Split-wire thermocouples that were inserted into the refrigerant 
flow path to measure the flow directly were encased in small pieces of 1/8 inch stainless 
steel tubing filled with epoxy. The epoxy was necessary to create a leak-free connection. 
The piece of tubing was then inserted into the flow path by typical Swagelok? 
connectors. 
There were two pressure measurements in the refrigerant loop. A Setra? absolute 
pressure transducer was used to measure the condenser inlet pressure. It was discussed in 
Section 6.2.1.4 that a better way of measuring this pressure was to determine the 
saturation temperature inside the reservoir and convert it into saturation pressure. Still the 
absolute pressure transducer served as a corroborating value. The other pressure 
 
 
76 
measurement was the measurement of the differential pressure by means of diaphragm-
type differential pressure transducer from Validyne Corp. The differential pressure 
transducer was connected to the micro-channel pressure ports by 1/16 inch stainless steel 
tubing. The use of a tube with such small cross-section helped dampen the rapid 
fluctuations in the pressure signal experienced due to active boiling process. The 
differential pressure transducer was protected from sudden pressure surges, possible 
during the charging process, by two quarter-turn plug shut-off valves at its either port and 
was shunted by another such valve, as shown in Figure 6.1. 
There were three instrumented control loops in the refrigerant system, of which 
two were mentioned before. The first one was the control of the copper elbow heat shield 
temperature which was maintained close to the temperature at the inlet to the condenser. 
It was observed that a small input power to the patch heaters was sufficient for a stable 
control of this temperature and hence this input was manually adjusted from time-to-time 
making a PID controller unnecessary. The second control loop was the control of the 
saturation temperature inside the reservoir. As mentioned before, a Watlow? 
microelectronic PID controller was used for this purpose.  A third active feed-back 
control loop was created to control the temperature of the sub-cooled liquid refrigerant at 
the inlet to the tube evaporator. It was observed that as ambient temperature or air draft 
across varied, this temperature exhibited a drift. Since the evaporator heat input was 
calculated based on this inlet temperature, it was necessary to stabilize it at a value above 
the ambient range. For that end, a small heater wire element was wound on the stainless 
steel tube and connected to a DC power supply; the heater power was governed by a 
second PID controller which received the evaporator inlet temperature as its input signal. 
 
 
77 
This completes the description of the various components of the refrigerant loop. 
A photograph of the loop is shown in Figure 6.8. In this image, the condenser section and 
the reservoir are shown with insulation covering. 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Figure 6.8: Photograph of the refrigerant loop assembly 
 
6.2.2 DESCRIPTION OF COOLANT (WATER) FLOW LOOP 
The coolant water flow loop consists of a refrigerated bath circulator, coolant 
flow measurement and metering system and temperature-control heater. 
 
 
 
Reservoir 
at 45o 
incline 
To Micropump 
Sub-cooler 
Absolute Pressure 
Transducer 
Differential Pressure 
Transducer 
Insulated 
Condenser Section 
Level Visualization 
Tube Water Flow Meter 
 
 
78 
6.2.2.1  Refrigerated Bath Circulator 
The 300 W NESLAB refrigerated bath/circulator was selected to maintain a 
coolant water flow at a desired temperature in the coolant loop. It consisted of an 
internal refrigeration system and a water circulating centrifugal pump. The temperature 
control system consisted of a micro-electronic circuit and a digital read-out. The 
bath/circulator was capable of controlling water temperature in the range of 5-95 C at 
an accuracy of ?0.03 C. To prevent fouling in the coolant loop and also inside the 
bath/circulator, distilled water was used as the working coolant. 
 
6.2.2.2 Coolant Flow Measurement and Metering System 
 . The coolant flow from the refrigerated bath/circulator was split into two streams 
? one flowing through the condenser section (named B in schematic in Figure 5.1) and 
the other flowing through the sub-cooler heat exchanger (named A). The flow 
measurement and metering system for the coolant stream through the condenser section 
consisted of two capillary flow meters connected in parallel and a metering valve down-
stream of the flow meters.  This metering valve, together with a control heater, was used 
to vary the coolant flow rate through the condenser and its temperature difference across 
the condenser. 
 To decrease the variation in the wall temperature of the condenser, as will be 
further described in the next chapter, it was desired to keep the temperature difference 
between the coolant inlet and coolant outlet of the condenser at 2 to 2.5C. It was also 
observed during the analysis of errors in the system that the uncertainty in heat transfer 
coefficient decreased with the increase in the above difference of coolant inlet and outlet 
 
 
79 
temperatures. A value less than 2 C, if chosen, would significantly increase the 
uncertainty. For each test condition at the inlet to the condenser, a value of the flow rate 
of coolant was arrived at by calculation. This value was set by the flow measurement 
system described above. After the required flow was thus set, the heat input to the control 
heater was varied gently using a variac until the desired steady state coolant temperature 
difference was achieved. 
There was no measurement system provided for the other coolant stream A that 
was flowing through the sub-cooler. This led to a situation in which the outlet 
temperature of the sub-cooled refrigerant varied. However, as described in Section 
6.2.1.6, the refrigerant temperature was stabilized using a feedback controlled heating 
section upstream to the evaporator. Hence the lack of instrumentation and control on the 
sub-cooler did not cause a difficulty. This concludes the description of the coolant loop. 
 
6.3    INSTRUMENTATION AND MEASUREMENT SYSTEM 
In this section a summary of the various instruments and measurement devices, 
belonging to the three categories of measurements and/or controls in the refrigerant and 
coolant loops ? temperature, pressure and flow ? will be provided. It should be noted here 
that all thermocouple and pressure data were recorded using an Inlet Pentium? computer 
and Agilent? Data Acquisitions System (DAS) containing two 20-channel multiplexer 
data cards that were vendor-calibrated. Data for the flow measurements were monitored 
during an experimental run and were manually recorded. 
 
 
 
 
80 
6.3.1 MEASUREMENTS OF TEMPERATURE 
 As referenced in the earlier sections and also indicated in the Figure 5(i), there 
were several locations in the two flow loops where accurate temperature measurements 
were necessary. AWG32 type-T thermocouples that were calibrated in the refrigerated 
bath/circulator using an accurate RTD reference device were used in the refrigerant loop. 
The measurements were made at the evaporator inlet (TEvap-IN), condenser inlet (TCond-IN) 
and condenser outlet (TCond-OUT). The system pressure was calculated based on the three 
thermocouples suspended inside the reservoir (TResv i, i = 1, 2, 3). The reservoir heater 
was controlled using a self-tuned Watlow PID controller. The heater upstream to the 
evaporator was similarly controlled using another PID controller. 
 Each of the coolant temperatures at the inlet and outlet of the condenser was 
measured using two AWG32 calibrated thermocouples (Twi1 and Twi2). It was observed 
that the energy balance and estimated average heat transfer coefficients, as discussed in 
Chapter 7, were heavily dependant on these measurements, and hence two thermocouples 
were used at each location. In order to improve the accuracy, the thermocouples were 
later replaced with four Thin-film platinum resistance temperature detectors (RTDs). 
Further a Y-junction arrangement, as shown in Figure 6.9, allowed proper positioning of 
the RTD sensors and a wire mesh placed in the flow before it reached the RTDs was also 
helpful in mixing the flow, so that the temperatures measured corresponded to the coolant 
bulk temperature. 
 
 
 
 
81 
 
 
 
 
 
 
 
 
 
Figure 6.9: Y-connection indicating the location of RTDs for accurate coolant temperature measurement at 
the inlet to the condenser. A similar arrangement was provided at the coolant outlet. 
 
 
6.3.2 MEASUREMENTS OF PRESSURE 
The two pressure measurement devices were the Setra? absolute pressure 
transducer with a range of 0-500 psi, and a Validyne DP-15 differential pressure 
transducer with a differential pressure rang of 0-14 kPa. As mentioned earlier, the system 
pressure was derived using the measurement of the saturation pressure inside the 
reservoir. The absolute pressure signal was used to corroborate this value.          
  
6.3.3 MEASUREMENTS OF FLOWS 
As mentioned earlier, the refrigerant flow was measured by the calibrated micro-
pump which was provided with micro-electronic control and digital readout. 
The flow measurement for the coolant stream B (ref. to Figure 6.1) was measured 
using two capillary flow meters placed in a parallel configuration in such a way that flow 
was directed through one of the meters for a given test condition using the shut-off valves, 
Transparent ?Y? 
Connector 
Wire Mesh for 
Flow mixing 
Thin-film Platinum RTD 
(2 x 2.3 x 1mm sensor dimension) 
Silicon Rubber Seal 
Coolant Water 
Flow 
Twi1 
Twi2 
 
 
82 
shown in Figure 6.1. It was found after a survey of commercially available and affordable 
rotameters that the expected accuracies of such meters were within ? 2 to 4% of full-scale. 
It was decided that it would be better to create two customized capillary-pressure-drop-
based flow meters for two flow ranges necessary. In these flow meters, the flow was 
measured by the pressure drop across the capillaries, of a selected lengths and internal 
diameters, measured by the heights of water in the two columns located at the upstream 
and down stream of the capillary. The difference in heights ( ?h) of the water levels in 
these two columns indicated the difference in static pressure (?p = ?wg?h) across the 
capillary. The correspondence between the static pressure difference, ?p and the flow rate 
passing through the capillary was found by calibration. The height difference was 
measured by a millimeter scale with a least count of 0.5 mm. Therefore, the accuracy of 
this measurement was 1.0 mm. An appropriate choice of capillary was made so that the 
measurement accuracy was kept at or below ?2% of reading. As mentioned earlier, two 
different flow meters were fabricated and calibrated to serve the different flow ranges (20 
to 130 ml/min and 140-550 ml/min) necessary for different refrigerant test conditions.  
 
6.4 INSTRUMENT CALIBRATION 
In this section, the procedure to calibrate various instruments will be discussed, 
including calibration of thermocouples and RTDs, pressure transducers, and flow meters. 
 
 
 
 
 
 
83 
6.4.1 CALIBRATION OF THERMOCOUPLES 
For the purpose of calibration, all thermocouple and RTDs wires were put in a 
bunch in such a way that the sensor tips of the RTDs and the thermocouple junctions 
were closely located. The calibration was performed between the ranges of 10 to 90 C in 
steps of 10 C in a water bath inside the refrigerated bath/circulator which maintained a set 
point water temperature within ? 0.03 C stability. An RTD probe with accuracy of ? 
0.003 C was inserted into the water bath with its tip in close proximity to those of the 
thermocouples and RTDs. Temperatures were observed by a DAS software, and upon 
reaching stability for a continuous period of 5 minutes the data were recorded at each 
calibration point. It was found that the RTD?s as well as the thermocouples exhibited 
linearity with a correlation coefficient greater than 0.9999. For the sake of brevity, only 
the calibration chart for the RTDs is shown in Figure 6.10. 
 The five thermocouple junctions that were created on the wall of the micro-
channel structure were calibrated in situ. After calibration of all other instruments and 
sensors and their assembly in the experimental apparatus, these five thermocouples 
junctions were calibrated. Coolant water was circulated in the coolant jacket while 
refrigerant was circulated inside the micro-channel. Refrigerant was prevented from 
boiling inside the micro-channel by increasing the system pressure sufficiently. The 
refrigerant temperature was controlled by power input to the heater upstream of the  
 
 
 
 
 
 
 
84 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Figure 6.10: Calibration chart for the Thin-film Platinum RTDs that measured the coolant inlet and outlet 
temperatures. 
 
evaporator. The inlet and outlet temperatures of the refrigerant and the coolant water for 
the condenser test section were kept equal and observations were made to detect any 
possible temperature drift over a sufficient period of time. When assured of stability, 
temperatures were recorded and translated into the calibration chart shown in Figure 6.11. 
 
RTD-Calibration Data
0.000
5.000
10.000
15.000
20.000
25.000
30.000
35.000
40.000
45.000
0.000 5.000 10.000 15.000 20.000 25.000 30.000 35.000 40.000 45.000
RTD measurement(C)
Pr
ob
e T
em
pe
ra
tu
re
 (C
)
RTD-1
RTD-2
RTD-3
RTD-4
Linear (RTD-1)
 
 
85 
 
Figure 6.11: Calibration chart for condenser wall thermocouples. 
 
 
6.4.2  CALIBRATION OF DIFFERENTIAL PRESSURE TRANSDUCER 
  
                  The Validyne? DP-15 bidirectional differential pressure transducer is a 
diaphragm-type transducer with a diaphragm having a range of 14 kPa. It was calibrated 
using a vertical graduated U-tube manometer-like setup. The U-tube manometer was 
created with ?? Tygon? tubing, with one arm of the U-tube connected to one of the ports 
of the differential pressure transducer. The other arm of the U-tube and the remaining 
port of the transducer were both kept open to atmosphere. Water was poured into the free 
arm of U-tube using a squeeze bottle. Care was taken so that the water column in either 
arm of the U-tube was devoid of bubbles or air locks which might alter the calibration 
results. The pressure differences, estimated from the difference in the heights of the water 
Calibration of Micro-channel 
Wall Thermocouple Junctions
Tw 1:   y = 1.031x - 0.9806
R2 = 0.9999
Tw 2:   y = 1.021x - 0.7872
R2 = 1
Tw 5:   = 1.0151x - 0.6435
R2 = 1
Tw 4:  y = 1.0112x - 0.5082
R2 = 1
Tw 3:   = 1.0106x - 0.4742
R2 = 1
20
22
24
26
28
30
32
34
36
20 22 24 26 28 30 32 34 36
 Wall Thermocouple Temperature (C)
Av
er
ag
e W
ate
r T
em
pe
ra
tu
re
 (C
)
Tw1
Tw2
Tw3
Tw4
Tw5
Linear (Tw1)
Linear (Tw2)
Linear (Tw5)
Linear (Tw4)
Linear (Tw3)
Linear (Tw3)
 
 
86 
columns in the two arms of the U-tube, were thus exerted on one port of the differential 
pressure transducer and its voltage outputs were recorded. The transducer connection was 
reversed and the procedure was repeated. Figure 6.12 shows the calibration results from 
the above procedure.  
 
 
Figure 6.12: Calibration results for Validyne? differential pressure transducer using U-tube manometer. 
 
 
6.4.3 CALIBRATION OF THE ABSOLUTE PRESSURE TRANSDUCER 
        The Setra? absolute pressure transducer with a range of 0 ? 550 psia was purchased 
with a factory calibration of ?0.11% full scale accuracy. The transducer was measured 
against another accurate reference pressure transducer (pre-calibrated by dead weight 
method) by pressurizing both transducers, kept in parallel configuration, using 
Calibration of DP Transducer: Validyne DP-15-32 (0 - 14 kPa)
y = 1.3959x
R2 = 0.9999
-15
-12.5
-10
-7.5
-5
-2.5
0
2.5
5
7.5
10
12.5
15
-11 -8.5 -6 -3.5 -1 1.5 4 6.5 9
Signal Voltage (V)
Pr
es
su
re
 D
iff
er
en
ce
 (k
Pa
)
Experimental Data
Linear (Experimental Data)
 
 
87 
compressed air.  The pressure transducer was calibrated in the operating pressure range of 
0 ? 300 psia. The results of the two calibration runs, as shown in Figure 6.13, were 
deemed repeatable, linear and acceptable within the factory accuracy.  
Calibration of Setra 0 - 550 psia Absolute Pressure Transducer
y = 0.01x
R22 = 0.9999
0
50
100
150
200
250
300
350
0 1 2 3 4
Setra Transducer Voltage Output (V)
Re
fer
en
ce
 A
bs
olu
te 
Pr
es
su
re
 Tr
an
sd
uc
er
 R
ea
din
g 
(p
sia
)
Calibration Run #1
Calibration #2
Linear (Calibration Run #1)
 
Figure 6.13: Results for the corroboration of the linearity and accuracy of factory calibrated Setra? 
absolute pressure transducer. 
 
 
 
 
 
 
6.4.4 CALIBRATION OF FLOW METERS 
 The low-flow and high-flow coolant water flow meters were calibrated by 
measuring the water flow rate passing through their capillary passages and recording the 
difference in the heights of the water column in each arm of the flow meter.  Figures 6.14 
and 6.15 indicate the calibration charts for the capillary flow meters. The calibration 
 
 
88 
process was repeated a few times and the graphs indicate the cumulative calibration data 
used to derive the calibration correlation. 
 
 
 
 
 
 
 
 
 
 
Figure 6.14: Low flow water flow meter calibration chart 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Figure 6.15: High flow water flow meter calibration chart 
Low-flow Rate Meter Correlation
y = -4.504404E-07x6 + 4.644879E-05x5 - 2.155876E-03x4 + 5.766924E-
02x3 - 9.581607E-01x2 + 1.251246E+01x
R2 = 9.996162E-01
0
20
40
60
80
100
120
140
0 5 10 15 20 25 30
Height Difference (cm)
Wa
ter
 Fl
ow
 R
ate
 (m
l/m
in)
Series1
Poly. (Series1)
y = -0.00000001x6 + 0.00000294x5 - 0.00035624x4 + 0.02168786x3 - 
0.72462887x2 + 18.27399036x + 54.08967103
R2 = 0.99989714
0
100
200
300
400
500
600
0 20 40 60 80 100
 Height Difference (cm)
W
ate
r F
low
 R
ate
 (m
l/m
in)
Series1
Poly. (Series1)
High-flow Flow Meter Correlation
 
 
89 
6.4.5     CALIBRATION OF MICRO-PUMP FLOW METER 
                    The Ismatec ? micropump internal flow measurement system was calibrated 
by timed dispensation of water flowing through the pump at different monitor readings. 
Figure 6.16 indicates the linear response of the flow meter. 
Calibration Curve for ISMATEC Micro Pump
y = 0.9922x - 0.0276
R2 = 0.9999
y = x
0
5
10
15
20
25
0 5 10 15 20 25
Monitor Readings (ml/min)
Me
as
ur
ed
 Fl
ow
 R
at
e (
ml
/m
in)
Experimental Data
Linear (Y=X)
 
Figure 6.16: Flow response versus monitor readings of the Ismatec micropump internal flow meter. 
 
6.4.6    TUNING OF PID CONTROLLERS 
               A PID (Partial-Integral-Derivative) controller is a feedback mechanism which 
attempts to correct the error between the measured process variable and the desired set-
point by measuring the error value and outputting a corrective signal to remedy that error. 
The PID control algorithm involves three separate parameters ? the proportional 
parameter determines the reaction to the most current error, the Integral parameters 
 
 
90 
determines the reaction based on the sum of the recent errors, and finally the derivative 
parameter determines the reaction to the rate at which the error value is shifting. The 
weighted sum of these values is used to control the process via a control element, which 
in the present cases were heaters operated at a constant DC voltage. The PID controllers 
have a primitive option of acting as a control switch (heater circuit simply turned ?ON?/ 
?OFF? depending on the sign of error). The success of such a control strategy depends 
largely on the proportion of the corrective action to that of the measured fluctuations and 
also on the delay in reaction. However the utility of the PID controllers stems from its 
micro-electronic logic circuit with auto-tuning features. Although on certain occasions, 
manual tuning of the PID parameters was found convenient, in most cases the auto-tuning 
feature was used to set the most optimum parameters. Figure 6.17 indicates the difference 
between the control effects from manual adjustments and auto-tuning of the reservoir PID 
controller on the system pressure. To enable visibility on one graph, different desired set-
points (16 and 25 psi) were chosen for the two cases. 
 
 
 
 
 
 
 
 
 
Figure 6.17: Effects of different control options on the absolute pressure of the system. 
PID Controller Testing Data
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0 1000 2000 3000 4000 5000 6000 7000
Time (sec)
Pr
es
su
re
 (x
10
0 p
si)
Data w /o Tuning
Data w ith Tuning
10 per. Mov. Avg. (Data w ith
Tuning)
10 per. Mov. Avg. (Data w /o
Tuning)
 
 
91 
6.4.7 SUMMARY OF CALIBRATION RESULTS 
                   Sufficient care was adopted in the calibration of various instruments and 
meters present in the system. The calibration runs were repeated several times in order to 
verify the stability of the calibration process and the repeatability of the results. Final sets 
of calibration data were used in generating correlations that were presented in this chapter. 
Table 6.1 below indicates a summary of various instruments, their operating ranges and 
other specifications and error limits obtained from their calibration results. 
Table 6.1: Summary of Measurement and Control System for Experimental Apparatus 
 
 
  
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Stability: ?0.03 C Neslab 
Bath/Circulator 
0 ? 300 W Cooling for 
Refrigerant loop 
?0.1% of Span Watlow? Series 
SD  
PID Controller 
0 ? 5 V input Pressure 
Stabilization 
?0.1 C   
&  
?0.08 C 
AWG 24 & 32     
T-type 
Thermocouples and 
Platinum RTDs 
0 ? 100 C Temperature 
Measurement 
?0.25% FS Validyne DP15-32 
Differential 
Pressure 
Transducer 
0 ? 14 kPa 
(0 ? 2 psi) 
Differential 
Pressure 
Transducer 
?0.11% of FS 
(FS: 0 ? 500 psi) 
Setra 205 
Transducer 
40 ? 320 psi Absolute Pressure 
Transducer 
< ?1% of FS Calibrated 
Capillary Flow 
Meter 
0 ? 550 ml/min Water Flow 
Measurement 
?1% of rate ISMATEC?  
Micro-pump Flow 
Meter 
0.06 ? 0.6 gm/s Refrigerant Flow 
Rate 
Accuracy Instrument Operating Range Measurement 
/Operation 
 
 
92 
6.5  SELECTION OF WORKING FLUIDS 
 
Refrigerant R134a was selected as the first of the two working fluids for the 
present study. R134a is a high pressure refrigerant commonly used in the refrigeration 
and air-conditioning markets, including automotive HVACs. Further, the literature survey 
indicated that some relevant data with R134a were available from recent publications that 
might be useful for comparison with the data obtained in the present study.  
As mentioned in Section 6.1, it was intended that test data be generated in the 
range of 30 to 70 C for R134a.Therefore, with a view to use R134a in the system, the test 
loop was prepared and its parts were pressure-tested up to 400 psig pressure. However, it 
was decided that the second refrigerant to be chosen should have applications at lower 
saturation pressures such as in electronics cooling. It was also important that this 
refrigerant should have low environmental impact so that such a refrigerant would be 
acceptable within environmental regulations anticipated in near future. Table 6.2 contains 
a comparison of the relevant thermo-physical-environmental properties of a selection of 
currently available refrigerants with low or zero Ozone Depletion Potential (ODP). In 
Table 6.2, the properties of water are included for reference. Although water is regarded 
as one of the best coolants available in nature, due to corrosion and other issues, water is 
generally not accepted as a choice of refrigerant.  
Initially, it was thought that HFE-7100 would be a relevant choice for the 
second refrigerant, because it is liquid at room temperature and hence useful as a coolant 
in electronics cooling and other cooling applications that have preference to no using 
high pressure refrigerants. Furthermore, HFE-7100 has been categorized as a replacement 
fluids for several phased-out refrigerants and also has comparable or better heat transfer 
 
 
93 
properties, compared to other liquids of its category. Additionally, HFE-7100 is a low-
GWP (Global Warming Potential) substance and is already accepted as a dielectric fluid 
now commonly used for miniature liquid cooling loops e.g. those for power electronics. 
However, during trial runs in the apparatus, it was observed that there were several 
drawbacks for using this refrigerant in the present setup. First, the refrigerant was 
reported to have significant affinity for absorbing and retaining moisture from the 
atmosphere. Since the saturation pressure of the refrigerant at 30 C is below normal 
atmospheric pressure, a portion of the desired test conditions would require the setup to 
be run under partial vacuum conditions for which the apparatus would need additional 
instrumentation and arrangements. First, this increased the possibility of leakage of 
atmospheric air into the system which would be difficult to detect. Second, this fluid has 
high viscosity and consequently would exhibit high pressure drop in the system for the 
flow ranges desired. Since its saturation temperature is very sensitive to pressure, it was 
estimated that there would be significant variation of saturation temperature (as much as 
10 to 15 C across the channel) for this fluid when high mass flux tests were to be 
conducted. In such a situation, it would be difficult to obtain the desired operating 
saturation conditions with this fluid. Also, it affected the calculation of heat transfer as 
wall axial conduction would dominate due to a wide variation of saturation temperature.  
For these reasons, HFE-7100 was rejected as a choice for the working fluid.. 
Eventually, a new low-to-intermediate pressure refrigerant, R245fa was selected for study. 
Its ODP is zero and GWP is less than that of R134a and its operating saturation pressures 
were lower than R134a. 
 
 
 
94 
Table 6.2: Comparison of thermo-physical-environmental properties of some common refrigerants 
 HFC-134a HFC-125 HFC-143a HFE-7000 HFE-7100 Water* 
Saturation  
Temperature 
(oC) 
30 70 30 60 30 70 25 34 30 61 100 
Saturation 
Pressure     
(bar) 
7.71 21.18 15.68 31.71 14.35 28.77 0.695 0.969 0.3415 1.045 1.013 
Liquid 
Density         
(kg/m3) 
1187 996.3 1159 870.3 908.4 728.9 1414 1388 1455 1369 958.4 
Liquid 
Absolute 
Viscosity      
(cP) 
0.01827 0.01065 0.01259 0.00636 0.01162 0.005391 0.04251 0.03798 0.06556 0.06 0.02819 
Liquid 
Surface 
Tension 
(Dynes/cm) 
7.42 7.148 3.261 3.261 4.068 4.068 12.4 12.4 13.6 136 5.891 
Liquid 
Specific Heat 
(kJ/kg.K) 
1.446 1.804 1.462 3.247 1.722 2.712 1.088 1.388 1.134 1.158 4.217 
Heat of 
Vaporization 
(kJ/kg) 
173.1 124.3 104.6 51.78 152.5 94.01 135.3 132.7 120.9 111.2 2257 
Liquid 
Thermal 
Conductivity 
(W/m.K) 
0.081 0.059 0.058 0.043 0.068 0.05679 0.07426 0.0725 0.06785 0.063 0.6651 
GWP **          
(100 year 
ITH) 
1600 3800 5400 350 280 N/A 
Atmospheric 
Lifetime   
(years) 
13.6 32.6 53.5 4.9 4.1 N/A 
* For comparison purposes only 
** Global Warming Potential ? 100 year Integrated Time Horizon. 
 
 
 
 
95 
Further, its thermal properties are comparable or better that R134a. These reasons make 
R245fa a competitive choice as a future replacement of R134a. Table 6.3 provides a 
comparison of the thermo-physical properties of R245fa with R-134a. 
Table 6.3: Comparison of thermo-physical properties of the two working fluids used in the present study 
Density 
(kg/m3) 
Enthalpy 
(kJ/kg) 
Viscosity 
(?Pa.s) 
 
Tsat 
( oC ) 
Psat 
( kPa) 
liquid liquid vapor liquid vapor 
Surface 
Tension 
(mN/m) 
30 770 1187.5 241.7 414.8 185.8 12.04 7.42 R-134a 
70 2116 996.2 304.3 428.7 106.4 14.65 2.61 
30 179 1325.1 239.6 427.5 376.4 10.51 13.41 R-245fa 
70 610 1204.7 295.7 456.6 226.8 12.0 8.35 
 
 
6.6  SUMMARY 
 
 The present chapter describes the details of design and fabrication of the 
experimental apparatus and the calibration of various instruments. Various design 
concepts were presented and logical choices were made in creating the final form of the 
experimental setup. The chief criterion in the design was to contain the error in energy 
balance and heat transfer coefficient estimates in the data. To that end, several 
modifications were incrementally made to the design to improve accuracy and data 
repeatability. The various parts and instruments used in this apparatus are commonly 
available and hence this apparatus may be reproduced with ease. This chapter also 
contains a brief description of the thought process of the selection of the two working 
fluids used in this study. 
 
 
96 
Chapter 7 
 
EXPERIMENTAL RESULTS 
 
7.1 INTRODUCTION 
 
 In this chapter, the experimental procedure and data reduction process will be 
described, following which the data on heat transfer coefficient and pressure drop will be 
presented for the two fluids R134a and R245fa.  
  Since different researchers have adopted various experimental apparatus and 
investigative techniques, it is essential to understand both the process adopted in 
consistently executing the experiment in the experimental domain and also the data 
reduction technique that was used subsequently. A few data sets that have been found to 
be available in the literature contain data on investigation of condensing flows of R134a 
in micro-channels of similar size to that presented in this work. Although the 
experimental conditions of these data sets was not closely matched to those in this work, 
nevertheless, a comparison has been made with those data sets to assess the proximity of 
the results, since, as has been noted earlier, several investigations have found widely 
varying and even opposite results from similar test conditions. A comparison of the data 
with correlations based on data from conventional diameter channels and micro-channels 
will be made to show the applicability of available correlations. 
 
7.2 EXPERIMENTAL PROCEDURE 
 An experimental procedure had been adopted to consistently execute all test 
points in the experimental domain. The procedure involved controlling and/or monitoring 
four experimental parameters/inputs during a test: 
 
 
97 
a) Refrigerant Flow rate (INPUT) 
b) Refrigerant-side heat input (to the evaporator) (INPUT) 
c) Water flow rate (INPUT) 
d) Water inlet temperature (CONTROL) 
The steps of the experimental procedure are presented below: 
1. Determine desired inlet condition to the condenser (i.e., saturation temperature, 
vapor quality, degree of superheat) and desired refrigerant mass flux.  
2. Switch on the refrigerated bath/circulator and provide coolant flows in the 
condenser and the sub-cooler. 
3. Provide input to the PID controller for the reservoir heater to maintain a 
steady system pressure equal to the desired saturation pressure in the system. 
Provide input to the micro-pump to create a steady refrigerant flow in the loop 
to maintain the desired refrigerant mass flux. Provide input to the PID 
controller that controls the input to the heater on the refrigerant line located 
upstream of the evaporator so that a steady sub-cooled liquid temperature is 
established and maintained at the evaporator inlet through out the execution of 
the test point. Then provide input to the evaporator heater by controlling the 
power level to the desired value to maintain the refrigerant thermal condition 
desired at the inlet of the condenser. Also provide heat input to the copper 
elbow heat shield to keep its temperature close to the desired temperature of 
 
 
98 
the refrigerant at the condenser inlet. Thus the average temperature of the 
copper heat shield is closely maintained at the refrigerant saturation 
temperature, and hence reducing the convection losses from the refrigerant 
line in this region. 
4. Observe prevailing refrigerant saturation temperature (Tsat) and sub-cooled 
evaporator inlet temperature (Tevap,in) and review these inputs to maintain the 
desired condenser inlet condition. 
5. Calculate the amount of heat that would have to be extracted from the 
refrigerant flow in the condenser in order to achieve full thermodynamic 
condition i.e. xth = 0) at the condenser outlet. Also calculate the corresponding 
flow rate of coolant water required if the difference between the coolant inlet 
and coolant outlet temperatures is kept at 2 C (in some cases 2.5 C due to a 
limitation of the apparatus for water mass flux).  
6. Determine whether the range of the low-flow or high-flow meter is suitable 
for the above coolant flow rate. Apply this coolant water flow rate with the 
appropriate choice of water flow meter and monitor the steadiness of the flow 
continually. The only remaining parameter to be controlled is coolant water 
inlet temperature. 
7. Increase or decrease the set-point temperature of the refrigerated 
bath/circulator to a desired level. Then control the coolant heater power input 
to get a steady state temperature difference of 2 C between the inlet and outlet 
of the coolant stream in the condenser. 
 
 
99 
8. Monitor all inputs and outputs until steady state is reached. Capture the data 
for 10 minutes and transfer it to a computer for analysis while the system 
remains at steady state. While the system is in operation, calculate the inlet 
and outlet thermodynamic states of the refrigerant. If the calculated values are 
sufficiently close to the desired conditions, the experiment is concluded. 
Otherwise perform adjustments to the inputs and capture a new set of data. 
Repeat this process until a desired condition is reached. 
7.2.1 DEFINITION OF THERMODYNAMIC STATE OF CONDENSATION  
Two definitions for completion of condensation at the condenser outlet can be adopted: 
? Thermodynamic state of condensation ? The calculated vapor quality at the 
condenser outlet is 0%. That is, the calculated enthalpy at the condenser outlet is 
equal to enthalpy of the saturated liquid. In this case, at the condenser outlet, 
small vapor bubbles will exit the condenser channel along with sub-cooled liquid, 
and, if an adiabatic flow section is provided after the condenser, these bubbles 
will eventually collapse to form complete liquid, without further heat extraction. 
? Visual condensation ? The fluid refrigerant at the outlet of the condenser channel 
is liquid and completely devoid of bubbles, which is verified visually in the sight 
glass. However, in this case, the refrigerant at the outlet of the condenser channel 
is sub-cooled, and the calculated value of outlet quality is negative. 
From preliminary experiments, we have seen that, irrespective of the choice of 
refrigerant mass flux used for the test, the heat transfer coefficient, and to some extent the 
 
 
100 
channel pressure drop, are measurably different for the above two cases (about 20-30% 
change to average heat transfer coefficient value). The reason is that in the case of visual 
condensation, the end portion of the condenser is used for sufficient sub-cooling (beyond 
thermodynamic saturation state) so as to condense the bubbles completely before they 
leave the channel. Thus this end portion of the channel contains smaller vapor bubbles 
and larger slugs of liquid, and the rate of heat transfer (also pressure drop) is 
predominantly from single-phase liquid which is less than that for two-phase flow. Also, 
by decreasing the coolant temperature overall, the condensation rate is enhanced along 
the entire length of the channel, which increases the length of the condenser channel 
containing low quality liquid flow (x < ~ 0.1). 
Hence, it is crucial to consistently adopt one of the above definitions for completion 
of condensation. Because the rates of outlet sub-cooling and the collapse of remaining 
vapor bubbles depend largely on outlet header design, the definition of visual 
condensation will not be an objective one. Therefore, any data collected, based on visual 
condensation will not be helpful for future design requirements. Hence, we adopt the 
definition of thermodynamic state of condensation to be appropriate for our investigation.   
7.3 EXPERIMENTAL DOMAIN 
 The experiments conducted with the two refrigerants R134a and R245fa as 
working fluids are summarized in the Table 7.1. Results are given for tests with R134a 
for condenser inlet conditions at several different qualities, two different mass fluxes and 
at different inlet superheats. However, for the refrigerant R245fa only the tests with fully 
 
 
101 
saturated or superheated vapor inlet conditions were conducted, since no available data or 
correlations have  been reported for this relatively new refrigerant. 
Table 7.1: Domain of experiments with R134a and R245fa as working fluids 
 
 R134a R245fa 
Refrigerant Mass Flux 
(kg/m2.s) 150, 200, 300 50, 100, 150, 200, 250 
Refrigerant Saturation Temperature 
(oC) 30, 40, 50 30, 40, 50, 60, 70 
xinlet = 0.2 ? 1.0 Refrigerant Condition at Inlet 
Superheat = 0 ? 15oC Superheat = 0 ? 20
oC 
 
7.4 ENERGY BALANCE OF EXPERIMENTAL APPARATUS 
 The system energy-balance is expressed as a percentage of the heat input to the 
evaporator, as shown in the Equations 7.1 and 7.2, where Qparasitic-loss represents the net 
value of heat losses and/or heat gains to the system shown in Fig. 7.1. 
( ) ( ) lossparasiticoutwaterinwaterwaterrefrefrefin QiimiimQ ?+?=?+ ,,31 &&                                          (7.1)                           
100% ?= ?
in
lossparasitic
Q
QEB                                                                                                (7.2) 
The error in the system energy-balance can have different values under different 
conditions of input parameters as described in the experimental domain. For example, at 
low heat input (xinlet = 0.2) or low water inlet temperature (significantly below room 
temperature), the energy balance error can be significant. In order to determine the range 
and trend of variation of the error in energy balance, a series of experiments was 
conducted varying one of the input parameters and keeping the others fixed. 
 
 
102 
Several other experiments were also conducted close to settings required for data points 
(ref. Table 7.1). However, the percentage error in energy balance was found to vary, 
however, consistently within ? 6%. The absolute value of the percentage energy balance 
was lower when the heat input requirements were higher. From these experimental tests, 
we are confident that error in energy balance is between 0 to ?6%. 
 
Figure 7.1: Schematic depicting the parameters used for calculating percentage energy balance. 
It may be mentioned here that amongst the various parameters measured for energy 
balance analysis, the parameters which have greatest influence on the energy-balance are 
the water-side temperatures. Hence, as described in the previous chapter, the water-side 
temperatures were measured with two Platinum RTD's at each of the inlet and outlet 
locations to minimize the energy-balance and heat transfer data analysis. 
 
T1 
T2 T3 
P 
Evaporator 
Condenser 
Water-side 
Tw,in Tw,out 
Qin 
Refrigerant-side 
 
 
103 
7.5 DATA REDUCTION PROCEDURE AND DATA UNCERTAINTY ANALYSIS 
7.5.1 SCHEME FOR FINDING AVERAGE HEAT TRANSFER COEFFICIENT 
 In this sub-section, the steps for finding average heat transfer coefficient are 
described. The first task in data reduction is to identify the average inner wall 
temperature and saturation temperature of the refrigerant in the condenser. As described 
previously in the description of the micro-channel condenser test prototype, there are five 
thermocouples embedded in the outer wall of the channel, and the wall thickness of 
channel is 0.002 m. Therefore, the average outer wall temperature is given by 
5
5
1
,
,
?
=
? =
j
jwall
averagewallouter
T
T             (7.3) 
where Tj, j=1 to 5 represent the temperature readings of the wall thermocouples. 
The inner wall temperature must be a value higher than Touter-wall,average due to an outward 
average heat flux given by  
( ) t
Tk
Ltba
Qq averagewall
Cu
water
fluxheatwall
?
??
??
++=?? 22                     (7.4) 
where a = 0.0004 m and b=0.0028 m are the dimensions of the rectangular cross-section 
of the channel, t = 0.002 m is the channel wall thickness, L=0.2 mm is the length of 
channel and ?Twall-average is the approximate estimate of the temperature difference 
between the outer wall and the inner wall. Therefore, the inner wall temperature is given 
by 
 
 
104 
Cu
fluxheatwall
averagewallouteraveragewallinner k
qtTT ??
??
??+= .
,,          (7.5) 
It must be noted here that typically ?Twall-average < 0.2 C, and hence warrants the 
approximation in Equation 7.4. The average saturation temperature of the refrigerant is 
the average of the saturation temperatures at the inlet and the outlet of the channel, the 
difference of which is a function of the pressure drop in the channel and the nature and 
operating conditions of the fluid. The outlet saturation temperature is taken as the average 
of the temperature readings of the three thermocouples in the reservoir, as described 
previously, and is stated as  
3
3
1
,?
=
? =
j
jreservoir
saturationoutlet
T
T                (7.6) 
where Treservoir, j, j=1,2,3 are the readings of the three thermocouples in the reservoir. The 
average saturation temperature is therefore the Toutlet-saturation plus a correction term 
(?Tsaturation-correction) indicating the influence of the pressure drop. This correction term is 
found to vary linearly with pressure drop, since the pressure drop is less than 15 kPa and 
can be considered insignificant. 
correctionsaturationsaturtionoutletsaturationaverage TTT ??? ?+=                                                             (7.7a) 
where  
channelPchannelPinletPsaturationPinletPsaturationcorrectionsaturation PcTTT ???=? ??==? .)(5.0     (7.7b) 
 
 
105 
where ?Pchannel is the pressure drop in kilo-Pascal units and ?c? (in C/kPa units) represents 
a constant whose value must be identified at each saturation temperature level. Again, 
this approximation is warranted as there is only a small change in saturation temperature 
due to channel pressure drop. Next, the average heat transfer coefficient is based on the 
water-side heat transfer and is given as  
( )averagewallinnersaturationaveragesurface wateraverage TTA
Qh
??? ?
=          (7.8) 
where 
???
?
???
? +?+=
22
2,,1,,2,,1,, inwaterinwateroutwateroutwater
waterwater
iiiimQ &                                                (7.9) 
and,  
LbaAsurface ).(2 +=            (7.10) 
Here iwater,in,j and iwater, out, j , j=1,2 represent the water inlet and outlet enthalpy readings 
from the RTD's. It may be noted here that the water-side heat transfer, Qwater is used for 
calculation of average heat transfer coefficient since the electrical heat input is typically a 
higher value as it includes the Qparasitic-loss. Enthalpies of water are calculated at normal 
atmospheric pressure neglecting pressure drop in the water jacket. The mass flow rate of 
water is measured as a function of ?h (in mm), the difference in heights of the two water 
columns in the water flow meters and is given by the following correlation equations: 
 
 
 
106 
HIGH FLOW (i.e. min/120mlVwater >& ): 
54.0897  18.2739  107.2463 -
 102.169 103.5624 - 102.94  101- 
h
2
h
1-
3
h
-24
h
-45
h
-66
h
-8
++?
?+??+?=?
??
????highwaterV&     (7.11)     
LOW FLOW (i.e. min/120mlVwater <& ): 
 12.51246  109.5816 - 105.7669 
102.15588 - 104.6449  104.5044- 
h
2
h
1-3
h
2-
4
h
-35
h
-56
h
-7
???
???
+??+
??+?=?lowwaterV&      (7.12) 
7.5.2  DATA UNCERTAINTY ANALYSIS 
For the analysis of uncertainty, the usual method of root sum square (RSS) of weighted 
individual parametric uncertainty is adopted. The analysis has been conducted in the 
Engineering Equation Solver (EES?) software which uses this method as per guidelines 
given in the Guidelines for Evaluating and Expressing the Uncertainty of NIST 
Measurement Results, National Institute of Standards and Technology Technical Note 
1297, 1994. 
If Y=f(Xj), j=1 to n, then the uncertainty of the calculated value of Y may be expressed in 
terms of the uncertainties in Xj,s as 
?
= ?
?
?
?
???
?
?
?= n
j
jX
j
Y UX
YU
1
2
,
2
          (7.13) 
where U? represents the uncertainty of the variable ? =Y, Xj, j=1 to n. 
 
 
107 
As with the energy balance analysis, the data uncertainty of the computed value of 
average heat transfer coefficient, taking into account uncertainties of individual 
parameters, varies from one data point to another. Table 7.2 lists the uncertainties of 
individual parameters and their impact on the overall uncertainty value. The overall 
uncertainty values of different test points are different but lie within ?10% band. 
Accordingly, the test data are reported with an uncertainty of ?10% for all data points 
Table 7.2 Uncertainty Analysis of Average Heat Transfer Coefficient (R134a) 
 
Parameter Uncertainty, UX,j 
Partial Derivative,   ??
?
?
???
?
?
?
jX
Y  % of Total uncertainty 
a ?0.00004 (m) -1.27?10-6 4.72 
b ?0.00005 (m) -1.27?10-6 6.68 
L v0.001 (m) -21561 0.76 
t ?0.0002 (m) 18776 0.03 
kCu ?5 (W/m2.K) -0.0963 0.00 
?Pchannel ?40 Pa -14.86  
?left ?0.0005 (m) 86.38 3.39 
?right ?0.0005 (m) -86.38 3.39 
Patmosphere ?5 kPa -2.57?10-7 0.00 
Twall, j, j =1-5 ?0.1 oC 133 0.33 
Treservoir, j, j =1-3 ?0.1 oC -221.7 0.89 
Twater,in, j, j=1,2 ?0.08 oC -1284 19.17 
Twater, out, j, j=1,2 ?0.08 oC -1285 19.20 
Total: 100 
Computed Uncertainty for this test point haverage:  ? 5.77 % 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
108 
7.6  EXPERIMENTAL RESULTS WITH R134a AS WORKING FLUID 
The experimental data with R134a as the working fluid will be discussed in this sub- 
section. Specifically, different combinations of four effects that have been studied with 
this fluid are the effect of change in: 
1. inlet quality (0.2 to 1.0) 
2. refrigerant mass flux (150, 200 and 300 kg/m2s) 
3. refrigerant saturation temperature (30 C, 40 C) 
4. inlet superheat (0 to 15 C) 
The average heat transfer coefficient and total channel pressure drop will be discussed 
along with relevant comparisons of the data with literature or available correlations.  
7.6.1 EFFECT OF VARIATION IN INLET QUALITY 
Figures 7.2 (a) and (b) show the effect of change in inlet quality at three different 
mass fluxes on the average heat transfer coefficient and overall channel pressure drop, 
respectively. A cursory inspection of the data reveals that the average heat transfer 
coefficient values exhibit a fair degree of repeatability within uncertainty margins, 
whereas the pressure drop values exhibit a higher repeatability. It may be mentioned here 
that the pressure drop values are averages of instantaneous readings from the differential 
pressure transducer over the period of data recording. In general, for all inlet conditions, 
especially for the conditions with inlet quality, xin < 1.0, a high but stable fluctuation in 
instantaneous pressure drop readings was noticed, with amplitudes reaching ~ 2 kPa for 
 
 
109 
higher flow cases. This was possible due to the influence of in-tube boiling in the 
evaporator section upstream of the condenser channel. The scale in pressure fluctuations 
was relatively less for cases with inlet quality of 1.0 or inlet superheat since single-phase 
vapor may transmit less pressure fluctuation to the pressure port in the condenser channel. 
The heat transfer coefficients for different mass fluxes tends to collapse at inlet 
quality <0.1, since the flows with very low inlet qualities are dominantly forced 
convective single-phase (liquid) flows. At higher inlet qualities, and especially at and 
near the channel inlet, the slug-bubble flow pattern is expected, for which the dominant 
mechanism of heat transfer is the forced-convective effects in the thin liquid films 
separating the vapor bubbles from the channel wall. Even then a major portion of the 
channel will still be dominated by liquid slugs between bubbles. The pressure drop data is 
much more repeatable and coherent. At each quality, the channel pressure drop tends to 
form a parabolic rise with mass flux. For a given mass flux, especially for 300 kg/m2s, 
the pressure drop curve indicates a steep rise for initial qualities of 0 ? xinlet ? 0.5. 
However, at higher qualities the curves tend to flatten. This may indicate that there is a 
significant change in the flow pattern for low to medium qualities, but at high qualities 
(xinlet ? ~0.8) the flow pattern is essentially stable. 
 
 
 
 
 
 
110 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
(a) 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
(b) 
Figure 7.2: Effect of change in inlet quality on (a) average heat transfer coefficient (b) total channel 
pressure drop (R134a) 
Effect of Change in  Inlet Quality on Heat Transfer Coefficient  
(R134a, T_saturation = 30.2 C)
2000
2500
3000
3500
4000
4500
5000
5500
6000
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Inlet Quality
h_
av
er
ag
e (
W
/m
2.K
)
HTC at 150 kg/m2.s
HTC at 200 kg/m2.s
HTC at 300 kg/m2.s
Effect of Change in Inlet Quality on Pressure Drop
(R134a, T_sataturation = 30.2 C)
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Inlet Quality
Ch
an
ne
l P
re
ss
ur
e D
ro
p (
kP
a)
DP at 150 kg/m2.s
DP at 200 kg/m2.s
DP at 300 kg/m2.s
 
 
111 
Figures 7.3 (a) ? (c) show the different wall temperature profiles at different inlet 
conditions and mass fluxes. Insights into the local heat transfer coefficient along the 
channel would require the heat flux profile of the channel together with the wall 
temperature profiles. The heat flux profile is briefly discussed in the next sub-section 
employing computational techniques.  
 
Wall Temperature Profile at Different Inlet Qualities 
(R134a, Mass Flux = 150 kg/m2.s)
20
22
24
26
28
30
32
0 50 100 150 200
Thermocouple Location from Entrance (mm)
Te
m
pe
ra
tu
re
 (C
)
x_in = 0.2
x_in = 0.5
x_in = 0.8
T_sat
 
(a)  
Wall Temperature Profile at Different Inlet Qualities 
(R134a, Mass Flux = 200 kg/m2.s)
20
22
24
26
28
30
32
0 50 100 150 200
Thermocouple Location from Entrance (mm)
Te
m
pe
ra
tu
re
 (C
)
x_in = 0.2
x_in = 0.5
x_in = 0.8
T_sat
 
(b) 
 
 
112 
Wall Temperature Profile at Different Inlet Qualities 
(R134a, Mass Flux = 300 kg/m2.s)
20
22
24
26
28
30
32
0 50 100 150 200
Thermocouple Location from Entrance (mm)
Te
m
pe
ra
tu
re
 (C
)
x_in = 0.2
x_in = 0.5
T_sat
 
(c) 
 
Figure 7.3: Wall temperature profile at different inlet qualities at refrigerant mass fluxes (a) 150 (b) 200   
(c) 300 kg/m2.s respectively (R134a) 
 
 
However, here some observations can be made about the temperature profiles. 
The coolant water flowed in a counter-current direction with respect to the refrigerant, 
with a net rise of 2 C across the test section which is equivalent to a temperature gradient 
of 10 C/m, approximately. In the three different mass flow rates that are presented here, 
for the medium to high inlet qualities (xin > ~0.5), the value of the overall wall 
temperature gradient exceeds the coolant temperature gradient (the difference between 
the readings of the first and fifth thermocouples is more than 2 C). Therefore, if the 
coolant were to flow co-currently with the refrigerant, the overall wall temperature 
gradient could have been reduced. As stated before, one goal was to minimize the wall 
temperature gradient, since the change in wall temperature can influence the liquid 
properties of the refrigerant and hence may unduly influence, albeit only slightly, the 
 
 
113 
condensation phenomenon in the channel. As stated in Chapter 2, some researchers have 
found weak influence of heat flux on condensation heat transfer coefficients. 
 
7.6.2 EFFECT OF VARIATION IN SATURATION TEMPERATURE 
As shown in Figures 7.4 (a) and (b), a relatively smaller sample of test results 
with two different saturation temperatures were available in the present study to 
understand this effect for R134a. Across the range of inlet qualities, the effect of this 
variation on pressure drop is more prominent than on the overall heat transfer coefficient, 
the latter being observable only at inlet qualities higher than 0.5.  
 
 
(a) 
 
 
Effect of Saturation Temperature on Heat Transfer Coefficient
(R134a, Mass Flux = 200 kg/m2.s)
2000
3000
4000
5000
6000
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Inlet Quality
h_
av
er
ag
e (
W
/m
2.K
)
HTC at T_sat = 30 C
HTC at T_sat = 40 C
 
 
114 
 
 
 
 
 
 
 
 
(b) 
Figure 7.4: Effect of change in saturation temperature (a) average heat transfer coefficient (b) total channel 
pressure drop (R134a) 
 
The decrease in heat transfer coefficient and increase in overall pressure drop due to 
increase in saturation temperature is primarily due to the changes in the Prandtl Number 
and kinematic viscosity of the refrigerant as shown in Table 7.3.  
Table 7.3: Variation of relevant properties with saturation temperature of R134a 
 
Tsat(oC) Psat(kPa) ?liq(kg/m.s) ?vap(kg/m.s) ?liq(kg/m3) ?vap(kg/m3) Prandtlliq 
20 572.1 0.000207 1.18E-05 1225 27.8 3.392 
30 770.6 0.000183 1.22E-05 1187 37.56 3.271 
40 1017 0.000162 1.27E-05 1148 50.12 3.192 
50 1319 0.000142 1.32E-05 1102 66.32 3.151 
 
Effect of Saturation Temperature on Pressure Drop 
(R134a, Mass Flux = 200 kg/m2.s)
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Inlet Quality
Ch
an
ne
l P
re
ss
ur
e D
ro
p (
kP
a)
DP at T_sat = 30 C
DP at T_sat = 40 C
 
 
115 
7.6.3 EFFECT OF CHANGE IN INLET SUPERHEAT 
The effect of inlet superheat is shown in Figures 7.5 (a) and (b). The level of 
superheat was restricted to below 20 C because it was not advisable to conduct tests with 
superheat levels beyond 20 C due to the possibility of heater burnout caused by dry-out in 
the evaporator exit region.  
Although sufficient resolution of the wall temperature near the channel entrance is 
not available, we may suppose that just at the downstream of the entrance, there was a 
region of single phase vapor cooling zone. In this pure vapor zone, which was 
experiencing the highest fluid velocity in the entire loop, a thermal boundary layer was 
established such that the sub-layer of vapor within the thermal boundary layer adjacent to 
the wall had a temperature sufficiently close to the saturation temperature to precipitate 
condensation. This was the condensation incipience zone. Subsequent to the start of 
condensing zone, which marked the beginning of annular flow with a thin liquid layer 
separating the wall from the vapor, the vapor was more likely to be in a superheated state 
and would possibly continue to retain superheat up to a certain length downstream before 
reaching the equilibrium saturation temperature. Downstream of the condensation 
incipience zone the vapor velocity was significantly high, and the thin annulus of liquid 
driven by this high vapor shear may experience a high degree of turbulence and therefore 
exhibit high convective heat transfer. Since the wall heat transfer coefficient may be 
expected to be significantly high just downstream of the condensation incipience zone, 
the average temperature of the thin liquid layer of the annulus might be expected to be 
close to the wall temperature or at-least slightly sub-cooled. The direction of thermal 
energy transport was from the vapor core to the liquid layer. However, the mass transfer 
 
 
116 
might not necessarily be in the same direction since some evaporation or re-entrainment 
of sub-cooled liquid molecules into the vapor stream might take place thereby further 
driving down the superheat in the vapor core. This is a matter of conjecture in the present 
context, and  although it is not within the scope of this study to understand where and by 
what mechanism the vapor core temperature would reach the local saturation temperature 
of the fluid, this is an interesting problem and needs microscopic observations to confirm 
the ?near entrance? phenomena. The reason the ?near entrance? phenomenon is of concern 
is because it is influenced by the wall temperature which is subject to variation depending 
on the choice of the coolant, the local coolant temperature and the mechanism of coolant 
flow. As mentioned in 7.6.1, the direction of the coolant flow (with respect to the 
refrigerant flow direction) and the coolant temperature may influence the local wall 
temperatures and thus alter the condensation process and flow patterns in the interior of 
the channel. 
The first wall thermocouple, located 20 mm downstream to the channel entrance, 
exhibited a noticeable although small increase in temperature with an increase in 
superheat. The readings of other wall thermocouples were indistinguishable from their 
values for saturated vapor inlet condition, as were the pressure drops at different 
superheats. This indicated that the region of interest was near the entrance and that the 
flow downstream was not noticeably affected by the inlet superheat. Although limited to 
the near entrance zone, the significantly high heat transfer coefficient at this location 
noticeably influenced the overall average heat transfer coefficient value for superheats 
greater than or equal to 10 C. 
 
 
 
117 
Effect of Inlet Superheat on Heat Transfer Coefficient
(R134a, T_saturation = 50.10 C)
3500
3750
4000
4250
4500
4750
5000
5250
5500
0 50 100 150 200 250 300 350
Mass Flux (kg/m2.s)
h_
av
er
ag
e (
W
/m
2.K
)
HTC at Inlet SH=0 C
HTC at Inlet SH =10 C
HTC at Inlet SH=5 C
HTC at Inlet SH=15 C
 
(a) 
Effect of Inlet Superheat on Pressure Drop
(R134a, T_saturation = 50.10 C)
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
0 50 100 150 200 250 300 350
Mass Flux (kg/m2.s)
Ch
an
ne
l P
re
ss
ur
e D
ro
p (
kP
a)
DP at Inlet SH=0 C
DP at Inlet SH=10 C
DP at Inlet SH=5 C
DP at Inlet SH=15 C
 
(b) 
Figure 7.5: Effect of change in inlet superheat on (a) average heat transfer coefficient (b) overall channel 
pressure drop. (R134a) 
 
 
118 
7.6.4. CFD-BASED APPROXIMATE INVERSE HEAT TRANSFER STUDY 
 A purely 2-D analytical approach was initially attempted to obtain information 
about the local heat flux profile on the inside walls of the micro-channel condenser. This 
approach involved inverse heat transfer analysis with the wall temperature profile as one 
of the source functions. However, this approach required an assumption for the outer wall 
(i.e. water-side) heat flux or outer wall heat transfer coefficient profile along the channel 
length. Moreover, the flow in the coolant jacket was not radially symmetric and varied 
along the channel length which would lead to error. Instead, a CFD-based approach, 
involving manual iterations was adopted to estimate the desired local heat flux profile for 
a limited data set. This approach, though laborious and still an approximate estimate, 
reveals a better understanding of the coolant flow and heat transfer. 
Figure 7.6 shows a sketch of a conjugate heat transfer model used for the CFD-
analysis. It consisted of the coolant jacket enclosed around the condenser micro-channel. 
The coolant entrance and exit passages were modeled closely to the actual prototype.  
 
 
 
 
 
Figure 7.6: CFD-model to study the combination of convective heat transfer in the coolant jacket and heat 
conduction in the channel wall. 
Coolant Jacket 
Micro-channel 
Condenser 
Refrigerant Flow 
Coolant Flow 
 
 
119 
 
The various surfaces and points used for monitoring and measurements during the post-
processing of the CFD results are shown in Figure 7.7. In the CFD model, several 
temperature monitoring points were chosen underneath the outer surface of the channel. 
Ta through Te, respectively, corresponds to the locations of the wall thermocouples Tw1 
through Tw5 that are embedded in the wall as described in an earlier chapter. Tbeginning and 
Tend were two fictitious temperature-monitoring points close to the refrigerant entrance 
and refrigerant exit ends of the channel to complete the temperature profile on the 
channel surface. Although there are no actual thermocouples at the locations marked by 
Tbeginning & Tend , the purpose of monitoring the temperatures at these locations is to 
ensure that the values from the model are realistic. 
 
 
 
 
 
 
 
 
 
 
 
` 
 
 
 
Figure 7.7: Schematic indicating the surfaces and points used in the post-processing analysis of CFD results. 
 
Section A 
 
Section B 
 
Section C 
 
Section D 
 
Section E 
 
Tbeginning 
Ta 
Tb 
Tc 
Td 
Te 
Tend 
Inlet Cross-section 
Outlet Cross-section 
Temperature Monitoring  
Point 
50  ?m 
Location of Temperature 
monitoring point inside the 
channel wall 
 
 
120 
For the purpose of the CFD-based inverse heat transfer study, three relevant experimental 
test points were chosen, as shown in Table 7.4. These R134a test points represent the 
effect of variation in refrigerant mass flux on local heat transfer coefficients in the 
channel and the results may be compared with data from literature. 
Table 7.4: A sample of three tests with R134a as working fluid chosen for CFD-based analysis 
 
 Experimental Inputs 
Experimental Results used as Inputs to 
Computation Model 
Simulation 
ID # 
Refrigerant Mass 
Flux (kg/m2.s) Refrigerant Tsat  (oC) 
& Inlet Quality (xin) 
Water 
Mass Flow 
Rate (kg/s) 
Water Inlet 
Temperature 
(oC) 
Qwater (W) 
1 100 50.1 , 1.0 0.0020121 42.12 18.47 
2 200 50.1 , 1.0 0.00391773 36.45 34.08 
3 300 50.1 , 1.0 0.00485172 30.19 50.68 
 
 
A CAD-model, as shown in Figure 7.6, meshed with tetrahedral and hexahedral cells, 
totaling approximately 3.8 million, was created with the aid of Gambit ? software and 
the analysis was conducted with the CFD software tool, FLUENT ?. The computation 
involved the conductive heat transfer (in the channel walls) and convective heat transfer 
(caused by the coolant flowing in the coolant jacket). The wall of the coolant jacket and 
the walls of the coolant inlet and outlet ports were assigned adiabatic boundary condition. 
It has been mentioned in an earlier chapter that the inlet & outlet ports of the condenser 
channel were made with thin-walled stainless tubing to prevent heat leakages into or from 
the test prototype by wall conduction. Based on this, the faces of the channel at the inlet 
and outlet were also assigned adiabatic conditions, whereas the surfaces of the channel 
interior to the coolant jacket (i.e. the outer faces of the channel wetted by the coolant) 
were exposed to convective heat transfer. The inlet to the coolant jacket was assigned a 
mass flow boundary condition with a mass flow rate value chosen from the Table 7.4 as 
per the simulation ID. The outlet to the coolant jacket was assigned a pressure boundary 
 
 
121 
condition and the coolant-jacket inlet pressure is automatically adjusted during 
computational iterations to a level that allows the boundary condition coolant mass flow 
value to reach at the inlet of the jacket. 
With the given boundary condition assignments in place, an assumed polynomial 
profile of refrigerant wall heat flux, as given in Equation 7.14a under condition given in 
Equation 7.14b, was assigned as a boundary condition on the inner wall of the micro-
channel through a User Defined Function (UDF). 
( ) 44332210 zCzCzCzCCzfq ++++==?   (W/m)    (7.14a) 
( ) waterQdzzfdzq ?=? ?? 2.0
0
2.0
0
                              (7.14b) 
Thus, the coefficients of the polynomial function, Cj, where j = 0 to 4, may vary, but the 
total heat rejected by the refrigerant will remain approximately equal to the heat absorbed 
by the coolant. A slight difference (within a few percent) between the total heat 
calculated from the imposed heat flux profile and the Qwater value is allowable due to 
expected inaccuracies in the measurements on the coolant side. Several trials with 
different heat flux profiles were used for the simulation and the results of the temperature 
values of the monitoring points (Tbeginning, Ta , Tb ,Tc , Td , Te , Tend ) for each were 
compared with the actual temperatures at those locations. Modifications to the trial 
profile were made until the output temperature profile closely matched most of the 
experimentally measured values within ? 0.5 C and in some cases within ? 1 C. Figure 
7.8 (a) and (b) show the trial heat flux profiles and the corresponding results of the trial 
runs for Simulation ID#2 (ref. Table 7.4). The value of Tbeginning may be expected to be 
within the range: Tsat ? Tbeginning ? Tw1, whereas the value of Tend may be expected to be 
 
 
122 
close to the value of Tw5 since the quality of refrigerant near Tw5  and Tend  are close to or 
less than 0.1. Hence the heat transfer gradient was much less compared to the beginning 
or middle portion of the channel.  
 
 
 
 
 
 
  (a) 
 
 
 
 
 
 
 
 
(b) 
 
Figure 7.8: (a) Trial heat flux profiles along the channel as inputs to the FLUENT? model  (b) 
Corresponding temperature profile as output to simulation. 
 
Trial Wall Heat Flux Functions:q'=f(z) 
(R134a, Mass Flux = 200 kg/m2s)
120
130
140
150
160
170
180
190
200
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
z (m)
q' 
(W
/m
)
q'=f(z) (Trial_1)
q'=f(z) (Trial_2)
q'=f(z) (Trial_3)
q'=f(z) (Trial_4)
 Wall Temperatures for Different Trials
(R134a, Mass Flux = 200 kg/m2s)
312
314
316
318
320
322
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
z (m)
Te
mp
er
at
ur
e (
K)
Experimental Valuesl
Result of q'=f(z) (Trial_1)
Result of q'=f(z) (Trial_2)
Result of q'=f(z) (Trial_3)
Result of q'=f(z) (Trial_4)
 
 
123 
Clearly, this numerical simulation technique for solving the inverse heat transfer 
problem is neither the optimum nor the most precise way to obtain the local heat transfer 
results. However, the initial scope of this study was to derive average results only and by 
means of this approximate technique we may have some measure of confidence about the 
accuracy of the heat balance and results of the heat transfer coefficient. Also, this analysis 
allows us to compare the results of this study with those that have been reported in the 
literature, which will be discussed later in this chapter.  
 
 
 
 
 
 
 
 
 
 
Figure 7.9: Contour Plots of (a) Temperature (in Centigrade); and (b) velocity (in m/s) of the coolant 
(shown with two orthogonal planes) and the outer channel along the length of the test section 
 
 
(a)  
(b)  
 
 
124 
Figures 7.9(a) and (b) show the temperature and coolant velocity contour plots 
along the entire length of the test section. A careful inspection of Figure 7.9(b) reveals 
that the highest temperature on the channel surface was not at the refrigerant entrance end 
but somewhat downstream. This is because the refrigerant entrance end of the channel 
fell in the ?shadow? of the coolant flow. As shown in Figure 7.10(a), the coolant path-
lines (started from the coolant inlet), indicate that the main flow path of the coolant, 
crossed the channel axis near the temperature monitoring point Ta and approached the 
coolant exit, nearly missing the refrigerant entrance end of the channel. This caused 
channel surface in the neighborhood of the point Ta to experience higher heat transfer rate 
and hence exhibited a slightly higher temperature than the surface region near the point 
Tbeginning. In fact, as shown by Figure 7.10(b), another view of the coolant path-lines 
indicate that the coolant flow twisted around the channel near the middle and end portion 
of the coolant path, thereby locally enhancing the coolant heat transfer coefficient. It may 
be recalled here that there was a heat flux redistribution effect due to axial and transverse 
conduction by the copper wall of the channel; this effect was more dominant in the region 
of higher heat flux gradient which was occurring near the refrigerant entrance end of the 
channel. The radial conduction effect is captured by the approximation given in Equation 
(7.5). Figure 7.11 (a) ? (b) shows the temperature profile in orthogonal directions across 
the Section A. Slight asymmetry in the profiles indicate the tortuous nature of the coolant 
in the jacket. The transverse conduction within the channel wall indicates a difference of 
less than 0.2C between the outer and inner wall temperatures. 
 
 
 
 
125 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Figure 7.10: Path-line trace for the coolant flow colored by (a) temperature (b) particle ID indicting a 
twisted trajectory of the main flow of the coolant in the coolant jacket. 
 
 
 
Tbeginning 
Ta 
(a) 
(b) 
Section A 
 
 
126 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
(b) 
 
Figure 7.11: (a) Temperature contour plot (in Section A) and (b) temperature plot on Crosshairs A ? A? and 
B ? B? 
 
It may be mentioned here in the passing that the axial profile of wall temperature 
in the present study has similarity in trend to that obtained by Koyama et al. [29]. The 
Temperature Plot on Crosshairs in Section A 
37
38
39
40
41
42
43
44
45
46
47
48
-0.015 -0.01 -0.005 0 0.005 0.01 0.015
Distance from Origin (m)
Te
m
pe
ra
tu
re
 (K
)
Temperature Profile on A - A'
Temperature Profile on B - B '
 
 
127 
difference between their test section and that in the present study is in the length of the 
condenser channel. Whereas they employed a channel of effective heat transfer length of 
0.6 m, the present study has used 0.2 m. The trend in their wall temperature also indicates 
the low gradients near the inlet and the outlet of the channel. 
Figure 7.12: Axial profile of the heat transfer coefficient (negative value indicating heat rejection) on the 
outer surface of the channel wetted by the coolant (Position z = 0 m is close to the coolant exit end) 
 
Figure 7.12 depicts the axial profile of the heat transfer coefficient on the coolant-
side. For the zone with z: 0? z? 0.1m, the average heat transfer coefficient is lower that 
that of the other half of the channel. This, combined with the fact that the refrigerant side 
heat flux as well as the coolant temperature are higher for this zone, explains the change 
in gradient for the wall temperature profile, which was discussed in Section 7.6.1. 
 
 
 
 
128 
 
Figure 7.13: Comparison of computer heat transfer coefficient with most relevant experimental data from 
literature. (R134a) 
Comparison of Different Average Heat Transfer Coefficients
(R134a, T_sat=50.1 C, x_in=1.0)
0
1000
2000
3000
4000
5000
6000
100 200 300
Refrigerant Mass Flux (kg/m2.s)
He
at 
Tr
an
sfe
r C
oe
ffc
ien
t 
(W
/m
2.K
)
Based on Computation
Based on Arithmetic Wall Average Temperature
Based on Log Mean Wall Average Temperature
 
 
Figure 7.14: Comparison of average heat transfer coefficients from CFD-based technique with values 
obtained by different wall averaging techniques. (R134a) 
 
Comparison of Computed Heat Transfer Coefficient vs. Local Quality with Literature Data
(R134a, Microchannel Dh < 1.0 mm)
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
10000
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Local Quality
He
at 
Tr
an
sf
er
 C
oe
ffic
ien
t (W
/m
^2
.K
)
Present, Dh=0.7 mm, AR=7, G=100 kg/m2.s, T_sat=50.1 C
Present, Dh=0.7 mm, AR=7, G=200 kg/m2.s, T_sat=50.1 C
Present, Dh=0.7 mm, AR=7, G=300 kg/m2.s, T_sat=50.1 C
Garimella et al, Dh=0.761 mm, Circular, G=150 kg/m2.s, T_sat=51.7 C
Garimella et al, Dh=0.761 mm, Circular, G=300 kg/m2.s, T_sat=51.7 C
Garimella et al, Dh=0.761 mm, Circular, G=450 kg/m2.s, T_sat=51.7 C
Koyama et al, Dh=0.807 mm, AR=2, G=273 kg/m2.s, T_sat=60.4 C
 
 
129 
From the different Trial runs, we obtained the most likely heat flux profile, say q? 
= f1(z) [W/m] on the inner surface of the micro-channel which causes temperature values 
at the monitoring points in the wall of the model channel to be approximately close to 
within ?0.5 C for most of the points. From this heat flux profile, and the actual wall 
temperature profile (from experimental data), Tw = f2(z), we can easily find the axial 
profiles of refrigerant quality and refrigerant heat transfer coefficient easily. Figure 7.13 
shows a comparison of the computed heat transfer coefficient with the quality. In the 
same figure, these computed curves are compared with data obtained from literature 
published by Garimella [35] and Koyama [29].  
In case of experiments performed by Garimella et al. [35], extruded micro-
channel tubes with each port having a circular cross-section and hydraulic diameter of 
0.761 mm were used and inlet and outlet qualities were maintained close to each other 
and the heat transfer coefficient data was presented against the average quality in the 
channel. The data was obtained at saturation temperature of 51.7 C, which was close to 
the present test condition. The test methodology of Koyama et al. [29] is more similar to 
the present study. They studied the condensation phenomenon of R134a in extruded tubes, 
with effective length of 600 mm and having rectangular ports of hydraulic diameters 
0.807 mm (aspect ratio 2:1), at saturation pressure of 1.7 MPa (Tsat = 60.4 C). In their 
study, the flow of the coolant flow was in opposite direction to that of the refrigerant flow 
and local heat flux and local wall temperatures were measured along the tube.  
Webb et al. [26] have published experimental data for condensation of R134a for 
micro-finned extruded tubes with hydraulic diameter 0.611 and 0.44 mm. However, the 
 
 
130 
local heat transfer coefficient data in their publication is limited only to mass fluxes of 
600 and 1000 kg/m2.s and only at saturation temperatures of 65 C. Accounting for the 
difference of the saturation temperatures in their work from that of Garimella et al [35], 
and the fact that their micro-channel ports have smaller hydraulic diameters and 
contained micro-fin features,  the heat transfer coefficient values obtained by Webb et al. 
[26] are lower than the data from Garimella et al. [35].  
The heat transfer results of a 0.7 mm hydraulic diameter with high aspect ratio of 
7.0 may be expected to be higher than data presented in these two studies with similar 
hydraulic diameters but lower aspect ratios or circular cross-sections, and that is observed 
in the comparison in Figure 7.13. This is supported by the analytical works of Wang and 
Rose [42, 43]. However, the general trend in the data is similar and lies within ?10% 
band as indicated on the computed curves. In case of high aspect ratio such as in the 
present study, the upper and lower corners of the channel cross-section tend to collect the 
liquid and thereby tend to stretch thin the liquid film along the longer sides of the channel. 
This increases the heat transfer coefficient in case of high aspect ratio micro-channels and 
this phenomenon is also observed in case of forced convective boiling studies as well. 
However, the effectiveness of the high aspect ratio is only available at millimeter or sub-
millimeter scale channels since for larger channels the effect of surface tension to draw 
away the liquid from the flat wall of the channel to the corner regions is not effective due 
to larger film radius of curvature at the corners. For high aspect ratio micro-channels, the 
proportion of the channel length with annular and wavy annular flow pattern may also be 
expected to be relatively smaller than in the case of the channels studied by Garimella et 
al. [35] and referred here. Due to the close proximity of the liquid films on the sides of 
 
 
131 
the channel, the liquid surface instabilities or waves on the films, that are excited due to 
the difference in the velocities of the vapor in the core and liquid in the film, may lead to 
?pinching? or contact of the side-wall films (Teng et al[55]).  This bridging vapor leads to 
formation of slugs and may cause premature formation of large bubbles of vapor each 
followed by a smaller sized slug of liquid. This is referred to as the plug/slug flow by 
Coleman and Garimella [22] but it is suspected, in case of high aspect ratio channels, to 
occur at a quality higher than they expected for their lower aspect ratio channels. The 
process of formation of the slugs needs to be studied further by microscopic observation 
of the film thicknesses on the side walls.  
The graphical comparison with the available data in literature in Figure 7.13 also 
reveals that the computed heat transfer coefficients differ with the trend of the data at the 
low qualities. As mentioned earlier in this chapter, the accuracy in the wall temperature 
profile segments close to the inlet and the outlet of the channel is wanting and this may 
result in error at very low and very high qualities. 
The definition of average heat transfer coefficient (ref. Equation 7.8) used in the 
present study is based on the arithmetic average of the wall temperatures with the added 
correction of the transverse wall conduction, as given in Equation 7.5. Cavillini et al [56] 
used a logarithmic mean temperature difference between the wall and the refrigerant at 
the inlet and outlet, as given by  
???
?
???
?
?
?
?=?
inwallsat
outwallsat
outwallinwall
TT
TT
TTT
,
,
,,
ln
ln
)(             (7.15) 
where Twall, in  and Twall, out are the temperatures of the wall thermocouples embedded in 
 
 
132 
the wall near the inlet and outlet of the refrigerant flow and Tsat  is the average saturation 
temperature. However, in their case the inlet and outlet qualities did not vary as widely as 
in the present study. This definition of temperature difference, used subsequently in 
computing the average heat transfer coefficient of the flow at average quality, is 
acceptable. A comparison of the three techniques for measuring the average heat transfer 
coefficient is shown in Figure 7.14. The computed average heat transfer coefficient in 
this graph is based on the computed heat transfer coefficient profile given in Figure 7.13 
by the following equation: 
?=
L
computedaverage dzzhLh
0
, )(
1           (7.16) 
Clearly, due to wide variation of the quality and hence the wall temperature, the 
logarithmic mean temperature definition is not applicable for the present study. Even the 
arithmetic mean temperature, though it reflects the trend, does not follow the computed 
average value at lower mass fluxes.  
7.7. EXPERIMENTAL RESULTS WITH R245fa AS WORKING FLUID 
Shown in Figure 7.15 ? 7.17, the trends in data for R245fa are similar to that 
presented for R134a. All data for this refrigerant has been taken for conditions with either 
saturated or superheated vapor at the inlet. A set of relevant property values is presented 
in Table 7.5. Comparing this with R134a properties given in Table 7.3, it is readily seen 
that the Prandtl number for R245fa is twice that of R134a at 30 oC, which is the main 
reason why the heat transfer coefficient is nearly 50% higher at 200 kg/m2.s and inlet 
superheat = 0 oC. However, this comes with a three times higher pressure drop due to 
 
 
133 
higher liquid dynamic viscosity (nearly twice that of liquid dynamic viscosity of R134a at 
30 C). 
 
 
 
 
 
 
 
 
(a) 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
(b) 
 
Figure 7.15: Effect of changing mass flux on (a) the average heat transfer coefficient and (b) channel 
pressure drop. (R245fa) 
Effect of Mass Flux on Heat Transfer Coefficient 
(R245fa, T_sat= 50 C)
2000
3000
4000
5000
6000
7000
8000
9000
0 50 100 150 200 250 300
Refrigerant Mass Flux (kg/m2.s)
h_
av
er
ag
e (
W
/m
2.K
)
HTC at Inlet SH = 0 C
HTC at Inlet SH = 10 C
Effect of Mass Flux on Pressure Drop 
(R245fa, T_sat = 50 C)
0
4
8
12
16
20
0 50 100 150 200 250 300
Refrigerant Mass Flux (kg/m2.S)
Ch
an
ne
l P
re
ss
ur
e D
ro
p (
kP
a)
DP at Inlet SH = 0 C
DP at Inlet SH = 10 C
 
 
134 
 
 
 
 
 
 
 
 
 
(a) 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
(b) 
 
Figure 7.16: Effect of changing inlet superheat on (a) average heat transfer coefficient and (b) channel 
pressure drop. (R245fa) 
 
Effect of Inlet Superheat on Heat Transfer Coefficient 
(R245fa, T_sat = 50 C, Mass Flux = 200 kg/m2.s)
4000
5000
6000
7000
8000
9000
0 5 10 15 20
Inlet Superheat (C)
h_
av
er
ag
e (
W
/m
2.K
)
Effect of Inlet Superheat on Pressure Drop
(R245fa, T_sat = 50 C, Mass Flux = 200 kg/m2.s)
4
6
8
10
12
14
0 5 10 15 20
Inlet Superheat (C)
Ch
an
ne
l P
re
ss
ur
e D
ro
p 
(kP
a)
 
 
135 
(a) 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
(b) 
 
Figure 7.17: Effect of change in saturation temperature (a) average heat transfer coefficient and (b) channel 
pressure drop. (R245fa) 
 
 
Effect of Saturation Temperature on Heat Transfer 
Coefficient (R245fa, Mass Flux = 200 kg/m2.s)
4000
5000
6000
7000
8000
9000
20 30 40 50 60 70 80
T_sat (C)
h_
av
er
ag
e (
W
/m
2.K
)
HTC at Inlet SH = 0 C
HTC at Inlet SH = 10 C
Effect of Saturation Temperature on Pressure Drop 
(R245fa, Mass Flux = 200 kg/m2.s)
0
4
8
12
16
20
20 30 40 50 60 70 80
T_sat (C)
Ch
an
ne
l P
re
ss
ur
e D
ro
p 
(kP
a)
DP at Inlet SH = 0 C
DP at Inlet SH = 10 C
 
 
136 
 
Table 7.5: Variation of relevant properties with saturation temperature of R245fa 
 
Tsat(oC) Psat(kPa) ?liq(kg/m.s) ?vap(kg/m.s) ?liq(kg/m3) ?vap(kg/m3) Prandtlliq 
20 123.7 0.0004356 0.0000104 1352 7.121 7.011 
30 178.9 0.0003791 0.00001079 1325 10.11 6.514 
40 251.6 0.0003307 0.00001118 1297 14.01 6.059 
50 345.2 0.0002893 0.00001159 1268 19.03 5.648 
60 463.4 0.0002541 0.00001202 1237 25.41 5.282 
70 609.8 0.0002242 0.00001247 1205 33.43 4.961 
 
7.8 COMPARISON WITH CORRELATIONS FOR LOCAL HEAT TRANSFER 
COEFFICIENT 
The three graphs in Figure 7.18 present the comparison of Nusselt numbers at 
specific vapor qualities against the Nusselt numbers based on the computed data (ref. 
Figure 7.13). The outstanding feature of these graphs is that for almost all correlations, 
the correlation predictions for qualities: x<0.1 and x>0.9 suffer significant deviations 
from the present computed data. As mentioned previously, the computed data is only 
accurate in as much as the temperature profile near the inlet and the outlet portions of the 
channel are correctly calculated from the wall thermocouple temperature profile. Hence, 
there is some ambiguity in this regard for these qualities since thermocouples were not 
placed close to the inlet and outlet to confirm the temperature profile in these regions. 
However, it should be pointed out that the heat transfer coefficient for x>0.9 is expected 
to be higher than that from the computation (Garimella [35]) and hence the two 
 
 
137 
correlations that seems to correctly capture this trend are the Cavallini[40] and 
Bandhauer-Garimella[35] correlations. The Koyama [29] correlation, which is a modified 
version of the Mishima-Hibiki [37] correlation, seems to under-predict most of the local 
data. It will be discussed in the next section that Koyama et al. [29] have found 
significant under-prediction of their data for mass fluxes less than 300 kg/m2.s with the 
Moser [38] correlation. Cavallini [40] correlation seems to predict the data best, 
especially for lower mass fluxes and qualities: x>0.3. Since the uncertainty in the 
computed local data is not clearly understood, further judgment regarding the 
appropriateness of each correlation will not be made.  
 
 Comparison of Local Nusselt Number from Correlations to Computed 
Local Nusself Number (G: 100 kg/m2.s, R134a)
0.00
0.25
0.50
0.75
1.00
1.25
1.50
1.75
2.00
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Computed Local Quality
Nu
_c
or
r/N
u_
co
mp
ut
ed
Dobson-Chato
Moser
Soliman (1968)
Traviss
Boyko-Kruzhilin (1967)
Shah (1979)
Cavallini
Koyama
Bandhauer-Garimella
 
 
138 
 
 
 
 
 
 
 
 
 
Figure 7.18: Comparison of correlation and computed local Nusselt Numbers (at specific qualities) at 
different mass fluxes. 
Comparison of Local Nusselt Number from Correlations to Computed 
Local Nusself Number (G: 300 kg/m2.s, R134a)
0.00
0.25
0.50
0.75
1.00
1.25
1.50
1.75
2.00
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Computed Local Quality
Nu
_c
or
r/N
u_
co
mp
ut
ed
Dobson-Chato
Moser
Soliman (1968)
Traviss
Boyko-Kruzhilin (1967)
Shah (1979)
Cavallini
Koyama
Bandhauer-Garimella
 Comparison of Local Nusselt Number from Correlations to 
Computed Local Nusself Number (G: 200 kg/m2.s, R134a)
0.00
0.25
0.50
0.75
1.00
1.25
1.50
1.75
2.00
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Computed Local Quality
Nu
_c
or
r/N
u_
co
mp
ut
ed
Dobson-Chato
Moser
Soliman (1968)
Traviss
Boyko-Kruzhilin (1967)
Shah (1979)
Cavallini
Koyama
Bandhauer-Garimella
 
 
139 
7.9 COMPARISON WITH CORRELATIONS FOR AVERAGE CHANNEL NUSSELT 
NUMBER 
7.9.1 Scheme for deriving the average Nusselt number for a given correlation 
In order to compare with the average heat transfer coefficient data, presented in 
this chapter, with that from the literature, a scheme is used to find the average Nu from a 
given heat transfer correlation. In finding the experimental value of the average heat 
transfer coefficient (and hence average Nusselt number) in the present study, the 
refrigerant condition is set at the channel inlet and the coolant temperature is varied to 
reach a saturated liquid state at the channel outlet.  Hence, the wall temperature is the 
variable to be adjusted.  
In order to obtain the average Nusselt number from a given correlation, we can 
assume a similar hypothetical situation in which the desired refrigerant condition is set at 
the channel inlet and the wall temperature is iteratively varied so that when it reaches a 
particular value, Twall, avg, the total heat transfer is such that the desired refrigerant 
condition (i.e. saturated liquid state) is reached at the channel outlet.  Because the heat 
transfer coefficient profile of the refrigerant along the channel length is dictated by the 
correlation, the obtained Twall, avg may be different that the actual experimental value of 
the average wall temperature.  
The situation is illustrated schematically in Figure 7.18. In this schematic, the 
channel is divided into (n+1) equal segments with length of each segment ?z. Since the 
inlet refrigerant condition is known (quality x0: 0 < x0  ? 1.0 & mass flow rate, refm& ), the 
 
 
140 
heat transfer coefficient at z = z0 can be computed from the given correlation. This heat 
transfer coefficient value, h0 may be assumed to be valid for the channel length up to z = 
z1, if ?z is sufficiently small. Therefore, for a choice of Twall, avg, the amount of heat 
absorbed by the channel wall from z = z0 to z = z1 can be computed from  
( )avgwallavgsat TTzPhq ,,00 ??=                                                                             (7.16)  
where,  P = channel cross-section perimeter = 2 (a + b), with a = 0.0004 m and b = 
0.0028 m.   
The quality x1 of the refrigerant at z = z1 can thus be calculated: 
fgref im
qxx
&
0
01 ?=            (7.17) 
where, ifg  represents the difference of specific saturation enthalpies of vapor and liquid 
states at saturation temperature equal to Tsat, avg. The next step is to determine the value of 
heat transfer coefficient at z = z1 corresponding to quality x1 and repeat the procedure 
described above. The iterative process is terminated at zj, j = n+1. It may be noted here 
that for this iteration, the initial choice for Twall, average is arbitrary and after completion of 
the run, the condition that is to be checked is given by  
001.0>?total
actual
total
actual
total
iteration
Q
QQ           (7.18) 
 
 
141 
where, totaliterationQ  represents the total heat absorbed by the wall from z = z0 to z = zn+1 and is 
given by ?
=
n
oj
jq , where qj  is the heat absorbed from the refrigerant at the (j-1)-th segment, 
between z = zj-1 and z = zj; and, totalactualQ  represents the actual heat absorbed by the wall so 
that the quality at the outlet of the channel reaches the desired value of 0.0. Therefore, 
total
actualQ  is given by  
fgref
total
actual ixmQ 0&=               (7.19) 
If the condition given in Equation 7.18 is met, the next choice of Twall, avg is implemented 
( by changing Twall, avg by a small amount ?Twall, avg) and the iterative process is continued. 
Thus, there will be some value of Twall, avg = avgwallT ,?  for which the condition in Equation 
7.18 will not be met, that is to say the computed quality at the outlet will be 
approximately equal to the desired value (i.e. xn+1 ? 0.0). For this value of the wall 
average temperature, the average channel Nusselt number may be calculated by  
     
              (7.20) 
It may be readily observed that the average heat transfer coefficient derived above, may 
also be represented as: 
???
?
???
?
+=???
?
???
? ?
?+= ?? ==
n
j
j
n
j
jcorr hnzPhznPh
00 )1(
1
)1(
1                   (7.21) 
???
?
???
?
??+= )?()1.( ,, avgwallavgsat
total
actual
l
hcorr
TTznP
Q
k
DNu
 
 
142 
The concept in Equation 7.21 is schematically represented in Figure 7.18 which indicates 
that the total thermal resistance between the wall and the fluid is actually a parallel array 
of the wall thermal resistances from the individual segments of the channel, such that     
( )zPhR jj ?= /1 , j = 0 to n. 
 
 
 
 
 
 
Figure 7.19: Schematic explaining the situation for deriving the average Nusselt number from a given heat 
transfer correlation. 
 
 
 
 
 
 
 
z1 z2 z3 zn 
?z 
zn-1 
Twall, avg 
Tsat, avg 
Twall, avg 
Tsat, avg 
R0  R1  R2  Rn-1  Rn  
 
 
143 
7.9.2 Comparison of average Nusselt number from correlation with that from data 
Figure 7.20 presents the comparison of average Nusselt numbers computed from 
four relevant correlations with Nusselt numbers obtained from experimental average heat 
transfer data from the present work. It is evident that all these correlations that have been 
recommended in the literature under-predict the present data by an average of 25%. 
Specifically, among the four, the Dobson-Chato [20] correlation is a better predictor of 
the average heat transfer data. We must keep in mind that most of these correlations are 
built on data-bases for larger hydraulic diameter tubes (>3 mm).  
 
Figure 7.20: Comparison of average Nusselt numbers from correlations and from experimental data for 
R134a 
Comparison of Average Nusselt Number from Correlation with 
Experimental Nusselt Number (R134a)
0
5
10
15
20
25
30
35
40
45
50
0 5 10 15 20 25 30 35 40 45 50
Experimental Nu_average
Co
rre
lat
io
n 
Nu
_a
ve
ra
ge
Dobson-Chato
Koyama
Cavallini
Moser
-25%
 
 
144 
In the present case, the high aspect ratio micro-channel is considered. As per 
boiling heat transfer literature, the higher aspect ratio of the channel is responsible for 
augmented heat transfer. However, the conclusion that higher aspect ratio is indeed 
responsible for higher heat transfer coefficients in case of condensation should be 
reserved only after further data is available with different channel aspect ratios at 
equivalent hydraulic diameter, preferably from a study that employs the same 
experimental setup (so as to have same entrance/exit characteristics, data collection 
process, and subject of similar data uncertainties). 
Another issue regarding the deviation of the data from the different correlations 
could be due to the low mass fluxes that are studied in the present work and among them 
especially those cases in which the inlet quality is less than 0.5. As shown by Koyama et 
al. [29] in their data comparison, for micro-channels with aspect ratio 2:1 and R134a as 
fluid, the lower mass flux data (100 to 300 kg/m2.s) in general deviates from the Moser 
[38] correlation predictions by -30% to -70%. In the present case, the deviation with 
Moser correlation has been found to be beyond -50%. Cavallini et al. had similar 
conclusions (under-prediction) when they applied their conventional scale correlation to 
their data from 1.4 mm diameter channel. 
The present analysis of data and comparison of them with some of the 
correlations indicate the possibility that there is an effect of the high aspect ratio of the 
present channel which has resulted in higher heat transfer coefficients than that from a 
lower aspect ratio or circular cross-section channels. This aspect ratio effect should be 
represented as a correction factor incorporated in the correlations. As an extension of the 
present work, tests on a few channels with smaller aspect ratios should be carried out to 
 
 
145 
directly prove this effect. That data would enable the formulation of the correction factor 
for the aspect ratio. 
Lastly, it may be mentioned here that the average data for the refrigerant R245fa 
have not been compared with the correlations mainly due to the fact that the inlet 
qualities studied here for this refrigerant are mostly at or above 1.0 (i.e. inlet saturation or 
superheat condition). Most correlations fail to make predictions at x=1.0. 
7.10 SUMMARY 
In the present chapter, a detailed description of the test procedure and the 
average/overall heat transfer coefficient and pressure drop data has been provided for the 
present work with R134a and R245fa as the working fluids. Selected data from the R134a 
group have been taken for an approximate inverse heat transfer method using CFD 
software. This method generated local heat transfer coefficient data which have been 
compared with most relevant data available literature. Subsequently, this computed data 
and the average data on heat transfer for R134a have been compared against the available 
and most relevant correlations. It is acknowledged that the computed data is dependant on 
the accuracy of the axial temperature profile on the condenser channel wall near the inlet 
and the outlet. The data indicate that the most of the relevant correlations tend to under-
predict the average Nusselt number data by about 25%. 
 
 
  
  
 
 
146 
Chapter 8 
 
CONCLUSIONS AND RECOMMENDATIONS FOR FUTURE WORK 
 
 
8.1 INTRODUCTION 
 
In this last chapter a brief overview and major conclusions from the present work 
is provided. Additionally, since the present work   represents a research area of continued 
interest and with broad applications, a set of recommendations for future activities, both 
on the experimental as well as on the analytical fronts is made to provide a closure.  
 
8.2. OVERVIEW AND CONCLUSIONS FROM PRESENT WORK 
 
 
The present work mainly encompasses two areas: 
1. fabrication of a micro-channel condenser employing the traditional batch fabrication 
techniques applied to MEMS, and 
2.  the experimental heat transfer assessment of the condensation phenomenon in a 
single horizontal micro-channel with two refrigerants, R134a and R245fa as working 
fluids, under different test conditions.  
In the first area of exploratory research, the feasibility of a design involving bonding 
of two silicon wafers with interfacing silicon nitride layers and embedded flow channel 
structures was explorer. Silicon nitride layer bonding methodology has been sparingly 
reported in open literature and in the knowledge of the present author, this is perhaps the 
first time such a nitride bond has been explored as a possible technique for creating flow 
structures. If this process was found successful and convenient, it would reduce the effort 
in making such structures in silicon wafer by eliminating the requirement of chromium-
gold layers typically used as an etch mask for chemical etching of silicon. It was 
 
 
147 
concluded that the silicon nitride bonding was possible, as shown by some samples 
successfully bonded; however, the initial thickness of the silicon wafer need to be at least 
1 mm in order to etch and bond flow structures as in the present case. Otherwise, the 
possibility of breakage was found to be significant, and the handling of the samples to be 
quite difficult. As a precursor to creating the silicon micro-condenser, a similar micro-
channel structure conveniently machined in a brass piece was packaged and preliminary 
experiments were generated to indicate the functional design issues related to the micro-
condenser concept. 
In the second phase of research, the issue of sizing of condenser with the help of 
available correlations was explored. Since it was found from the literature survey that 
only few publications are available which concern the study of refrigerants condensing in 
sub-millimeter micro-channels, it was deemed necessary to first explore the condensation 
phenomenon of pressurized refrigerants, like R134a, in a compact condenser channel 
with high aspect ratio and sub-millimeter hydraulic diameter. Careful effort was made to 
design and fabricate this compact test setup to accommodate the required test conditions 
within certain limitations of time and cost. This compact setup provided data with 
predetermined accuracy and repeatability such that the associated over-all uncertainty for 
the heat transfer coefficient, was to be at or less than ?10%. The test setup, with 
modifications from the knowledge from the present work (as suggested later in this 
chapter),  may also enable future experimental tests on micro-channels for just about any 
orientation with respect to the horizontal, including completely vertical or inclined  
positions. The typical micro-channels experimental work reported in the literature 
represents an array of channels with a manifold, rather than a single channel as in the 
 
 
148 
present study.  With multiple channels and manifolding, the results are subjected to 
uncertainties such as channel internal geometry non-uniformity and flow mal-distribution 
in the channels.   
The test data for heat transfer coefficient have been compared with the values 
generated from the empirical correlations reported in the experimental literature. Also, 
for selected test points, the local heat transfer coefficient values were extracted through a 
CFD-based approximate inverse heat transfer method and were found to be close to the 
values reported in recent literature. The newer models which take into account the natural 
effects of surface tension and viscosity in such small channels on the prevailing flow 
pattern appear to agree better with the present average heat transfer coefficients data than 
earlier correlations conventional scale tube correlations which did not take into account 
the small channel effects. This emphasizes the learning that future correlation 
development should include flow regime identification. Within the present body of work 
only preliminary visualization observations were conducted, giving insights to the 
different flow patterns that may exist at some of the test conditions. CFD-based approach 
on understanding the motion of vapor bubbles in micro-channels  were also  explored  
and its applicability was demonstrated to serve as a potential tool to understand the heat 
and mass transfer behaviors at the vapor-liquid interface which ordinarily cannot be 
performed without an intrusive technique. However, in order to implement this technique, 
the interfacial phenomenon needs to be properly modeled with experimental verification. 
One of the important drawbacks in the present setup which needs to be eliminated in 
future research in this area is the resolution of the wall temperature and the wall heat flux 
 
 
149 
near the entrance and exit of the channel. A modification to that effect is suggested in 
more details in the recommendations for future work that follows next. 
 
Below, the specific findings from the present work are enumerated to help 
understand their significance and applicability in the area of design and fabrication of 
compact condensers for two-phase cooling applications. 
1. The research into the feasibility of employing the fusion bonding of two silicon 
wafers (with embedded flow structures) at their silicon nitride layer interfaces 
indicates that while this bonding technique is possible, it requires a very smooth 
and flat interface and also it requires significant effort in surface preparation to 
maintain surface cleanliness necessary to prevent the formation of voids or un-
bonded regions in the interface. This poses as a challenge in a routine operation or 
in mass fabrication methods. 
2. In order to have a robust embedded flow structure that can withstand the handling 
and packaging requirements, the channel structure must be created in silicon 
wafers thicker than 1 mm at least; otherwise possibility of breakage of the fragile 
channel structure is present. 
3. The research in the understanding of the flow conditions and in using the present 
correlations for sizing of the condenser indicates that for high aspect ratio micro-
channels, data from which is not reported previously in literature, the present 
correlations under-predict the overall heat transfer coefficients by about 25% on 
an average. An averaging technique has been explained to carry out this 
comparison. The reason for this under-prediction could stem from the fact that in 
 
 
150 
a high aspect ratio channel, such as used in the present work, the corners of the 
channel more effectively draw the condensate film towards them due to surface 
tension and maintain a thin film on the sides of the channel that promotes higher 
heat transfer rates. A recent analytical work by Wang and Rose [43] which 
explored only the annular condensation phenomenon alluded to this effect. Also, 
recent flow visualization data from Coleman and Garimella [22] that had explored 
the effect of channel shape on flow regimes also implied such an effect. Since 
most correlations are based on either circular or low aspect ratio rectangular 
cross-section channels, their under-prediction in the present case is possible. The 
experimental data at this size of channel is very limited. The comparison of these 
available data with the present data (derived through CFD-based inverse heat 
transfer technique), though not at exactly similar test conditions, indicates similar 
trends, which is encouraging to note. 
4. An averaging technique for the comparison of the overall pressure drop data with 
that from the limited data set in the literature could not be carried out since it 
required the knowledge of surface heat flux profile along of the channel axis. It is 
explained why the measurements in the present case does not enable the 
derivation of this profile. However, it should be noted that the pressure drop 
levels were of the order of magnitude similar to that in the relevant literature. 
5. The refrigerant R245fa which has significantly low GWP and ODP is a possible 
alternative to some of the present generation refrigerants. This being a new 
refrigerant, the data with this refrigerant is not available, as far as is known. In the 
present work, limited tests were carried out to understand the condensation heat 
 
 
151 
transfer potential of this refrigerant and the data so far is encouraging as it 
indicates that with this refrigerant, higher levels of heat fluxes can be achieved 
than with R134a. This helps in further miniaturization of the condenser footprint. 
6. The effects of saturation temperature changes and superheat conditions at the inlet 
to condenser have been presented in addition to the typically studied effect of 
mass flux variation. The effect of superheat change on the heat transfer coefficient 
is understandable, but it should be noted that the superheat condition does not 
change the overall channel pressure drop. The effect of change in saturation 
temperature is also understandable since the saturation temperature alters the 
viscous and surface tension effects and also the Prandtl number of the fluid. The 
variation of the heat transfer and pressure drop as reported in the present work 
will be helpful in the sizing of the condensers if precise knowledge of the 
refrigerant conditions is not known.  
7. Lastly, an outcome of the literature survey and research during preliminary study 
on adiabatic bubble hydrodynamics in a millimeter scale channel is the finding 
that a convincing interfacial mass and heat transfer phenomenon model is not 
available. If such a model is made available, with parameters defined for each 
fluid, the bubble motion model can be expanded to complement visualization 
work to understand the collapse of bubbles at different flow and wall temperature 
conditions. Also, the vortical structures in the liquid plugs ahead and behind a 
bubble (in a bubble train), alluded to in Chapter 3, can increase the heat transfer 
coefficient in the liquid plug and can be modeled easily if heat transfer is included 
in the model. 
 
 
152 
 
8.3 RECOMMENDATIONS FOR FUTURE WORK 
 
 
 Recommendations for future work are divided into two main categories, 
representing the experimental and numerical work conducted in the present work.   
 
8.3.1 RECOMMENDATIONS FOR EXPERIMENTAL ANALYSIS 
 
 
In the present experimental analysis, the test setup and test prototype enables 
measurement of only overall heat transfer coefficient and pressure drop to be obtained. In 
a practical situation, for the sizing of a compact condenser, such an approach is 
acceptable. However, as mentioned in the previous chapter, in order to compare the 
obtained data with available correlations, either an averaging technique needs to be 
applied or an inverse heat transfer technique needs to be devised to obtain local 
information for a direct comparison. A more involved approach would be to create a test 
prototype channel with possibility of direct local measurements of heat flux and wall 
temperatures, both on the refrigerant-side and the coolant-side. As was the experience in 
designing the current setup, such a test section having instrumentation for full-scale local 
measurements becomes too complicated to design and maintain. Also, the possibility of 
parasitic heat transfers (undesired losses/gains) encourages the design to be as un-
intrusive and uncluttered as possible. Several experimentalists have used a pre- and post- 
condenser heat exchangers upstream and downstream to their test sections (viz. Koyama 
et al. [29]) so that the inlet and exit qualities for the test sections were maintained close to 
the desired value. (The difference between the inlet and outlet qualities is allowed to be 
only a few percent points, so that the heat and pressure drop data can be ascribed to the 
 
 
153 
average quality in the test section.) In such a case, the amount of heat transfer from the 
refrigerant-side is small and cannot be directly measured with high degree of accuracy 
due to overwhelming parasitic losses and instrument limitations. Garimella et al [35] has 
employed a thermal magnification technique which involved a second coolant loop to 
magnify the refrigerant-to-coolant heat transfer while maintaining the desired accuracy. 
However, in their case, an array of channels were employed instead not a single channel 
which, while assisting the data certainty, leaves behind the question of possible mal-
distribution of flow in the channels. In any case, such a technique is desirable if the extra 
thermal loop and instrumentation for that loop can be afforded. However, whether or not 
the entrance and exit conditions to the test section (due to transition from pre- and post-
condensers) are altered is also another important issue that needs to be considered.  
Also, in the present test setup, the value of the coolant jacket flow diameter was 
so chosen as to enable wall thermocouple wires to be routed out of the jacket without 
significantly influencing the axial coolant flow. However, as evidenced in the results 
from CFD-modeling of this flow, which did not even consider the effect of the 
thermocouple wires, due to asymmetric inlet and outlet locations the flow tended to have 
a swirl and also the coolant-side heat transfer coefficient was not same at different axial 
locations of the channel outer surface. Due to this, a simple Wilson Plot technique could 
not be adopted to get the local information. Figure 8.1 below shows a sketch of an 
alternative design that may better serve the purposes of deriving local data. The three tier 
structure, with the refrigerant condenser channel sandwiched between two coolant tubes, 
will enable a predictable coolant flow pattern. If thermocouples (not shown) were 
inserted into the coolant flow and into the condenser channel wall at several appropriate 
 
 
154 
locations along the axis, the coolant-side local heat flux, coolant temperature and wall 
temperatures for both condenser and coolant tubes can be obtained. Since the 
condenser/coolant-tube wall temperature axial profile is obtained, the wall conduction 
can be computed within reasonable accuracy. With these measurements, it will be 
possible to obtain the refrigerant-side local heat flux axial profile and hence local heat 
transfer coefficients. If pressure ports with access into the refrigerant path are provided 
on the condenser wall at several places along the axis, local pressure measurements can 
be obtained as well. However, too many pressure ports can interfere with the internal 
flow dynamics and hence would not be advisable. 
 
8.3.2 RECOMMENDATIONS FOR ANALYTICAL & NUMERICAL STUDIES 
 
 
A few researchers have attempted to make simplified analytical models for the 
condensate films in small channels dominated by surface tension and vapor-liquid 
interfacial shear forces and to a less extent by gravitational acceleration. 
 
 
155 
 
 
Figure 8.1: A semi-exploded sketch of a proposed improved test section 
 
 
Notable among these attempts are those of Wang and Rose [42, 43] and Begg, 
Khrustalev and Faghri [41]. Bandhauer, Agarwal and Garimella [35] had attempted to 
create a semi-empirical model of the condensate film interfacial temperature based on an 
earlier work by Traviss et al. [13] and from there, they had attempted to calculate the 
local heat transfer coefficients. All of the above models depend heavily on much earlier 
works on the interfacial heat transfer, mass transfer and shear models based mostly on 
data from conventional scale tubes. For example, Wang and Rose?s work has adopted the 
approach suggested by Mickley et al which was later used by Kays, Crawford and 
Weigand [45], and Cavallini [57]. The interfacial shear stress is given by: 
25.0
vvi Uf?? =                       (8.1) 
where, the interfacial friction factor is denoted by f and the average vapor velocity by Uv. 
 
Copper Micro-channel 
Condenser 
Plexiglass ? Protections 
Epoxy Filler 
Solder Contact and Filler 
Coolant Flow 
Coolant Tube 
Refrigerant Flow 
A 
A? 
Section A-A? 
 
 
156 
The expression used for the interfacial friction factor is related to the condensation mass 
flux, m& ??  and the zero transpiration (i.e. no net mass exchange) value of the interfacial 
friction factor, f0,v by: 
12exp
2
,0
,0
???
?
?
???
? ???
???
?
???
? ???
=
vvv
vvv
Uf
m
Uf
m
f
?
?
&
&
            (8.2) 
The condensate mass flux, m& ??  is used as a variable in their set of equations while the 
zero transpiration friction factor, f0,v is adopted from Churchill [46]: 
12/1
5.1
16169.0
12
,0
Re
37530
71.3Re
9.6log6569.5
1
Re
82
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
??
?
?
?
??
?
?
?
??
?
??
?+
??
???
??
??? +
???
?
???
?
+??
?
?
???
?=
vhvv
v
v
D
f
?
                (8.3) 
where, ? is the surface roughness, Dhv is the vapor hydraulic diameter and Rev is the local 
vapor Reynolds Number, given by vhvvvv DU ?? /Re = . Here Dhv is another variable used 
in their equation set. 
Begg et al [41] have used a similar expression relating to the condensation friction 
factor, f to the zero transpiration friction factor, f0,v. However, in their work, the 
expression for the zero-transpiration friction factor is adopted from Bowman and 
Hitchcock [58] which is different from the expression in Equation 8.3. Bandhauer et al 
[35] has adopted a different way of relating the interfacial friction factor to the flow 
parameters and then using that to find the interfacial stress. However, in their approach, 
the interfacial suction effect due to condensation is indirectly related to the interfacial 
 
 
157 
friction factor through the local liquid Reynolds Number 
l
l
xGD
?? )1(
)1(Re
+
?= , where G is 
the total refrigerant mass flux, D is the hydraulic diameter of the channel, x is the local 
vapor quality and l? is the liquid viscosity. ? represents the void fraction given by 
Baroczy [61].  
 In parallel, similar work has been performed in describing the interfacial activity 
in case of cavitation of liquid into bubbles and condensation of the vapor back into liquid, 
most notably by Senocak and Shyy [62] who have compared experimental results with 
three recent models and proposed their own. However, their model for interfacial mass 
flux depends on defining a characteristic length scale and free stream velocity of the fluid 
which poses difficulty in application in case of condensation in micro-channels.  
In short, it is apparent that the approaches in modeling the interfacial phenomenon 
is varied and tend to depend on earlier works based on conventional scale tubes or 
channels. The preliminary investigation in CFD modeling of a bubble motion, as 
presented in this work, and similar work carried out by Taha and Cui [49, 50, 51], van 
Baten and Krishna [52], Zheng [53] and others indicate significant interest in capturing 
the actual physical phenomena and incorporating it in a numerical model. Any such 
attempt will be benefited if a convincing model of interfacial dynamics for condensing 
flows in micro-channels may be developed with experimentally-derived parameters (such 
as accommodation coefficient etc.) for different fluids. Clearly, this presents itself as an 
opportunity for further research in modeling and experimental corroboration. 
 
 
 
 
 
158 
8.4 SUMMARY 
 
 In this chapter a brief overview and conclusions from present work was provided, 
with specific findings enumerated. It was followed by recommendations for future work 
for certain new dimensions in the experimental measurements, as well as numerical 
modeling on condensation in micro channels. Of particular interest are local measurement 
of the heat transfer and pressure drop data; also, the need for a comprehensive and more 
convincing interfacial dynamics model were emphasized to facilitate more realistic 
modeling of the actual phenomenon of condensation in micro channels. 
 
 
 
 
 
159 
 
REFERENCES 
 
[1]  Sherizawa, A., Feng, Z., and Kawara, Z., ?Two-phase flow in micro-channels,? 
Experimental Thermal and Fluid Sciences, volume 26, issue 6-7, pages 703-714, 
2002. 
[2]  Kew, P. A., and Cornwell, K., ?Correlations for the prediction of boiling heat 
transfer in small-diameter micro-channels,? Applied Thermal Engineering, volume 
17, issue 8-10, pages 705-715, 1997. 
[3]  Butterworth, D., ?A comparison of some void-fraction relationships for co-current 
gas liquid flow,? International Journal of Multiphase Flow, volume 1, issue 6, 845-
850, 1975. 
[4]  Smith, S. L., ?Void-fractions in two-phase flow: a correlation based upon an equal 
velocity head model,? Proceedings of the Institution of Mechanical Engineers, 
Thermodynamics and Fluid Mechanics Group, pages 647-657, 1969. 
[5]  Premoli, A., Francesco, D., and Prina, A., ?A dimensionless correlation for 
determining the density of two-phase mixtures,? La Termotecnica, volume 25, 
pages 17-26, 1971. 
[6]  Kariyasaki, A., Fukano, T., Ousaka, A., and Kagawa, M., ?Characteristics of time-
varying void fraction in isothermal air-water co-current flow in horizontal capillary 
tube,? Transactions of JSME, volume 57(B), pages 4036-4043, 1991. 
 
 
160 
[7]  Mishima, K., and Hibiki, T., ?Visualization and measurement of two-phase flow by 
using neutron radiography,? Nuclear Engineering and Design, volume 175, issue 1-
2, pages 25-35, 1997. 
[8]  Mishima, K., and Hibiki, T., ?Development of high-frame-rate neutron radiography 
and quantitative measurement method for multi-phase flow research,? Nuclear 
Engineering and Design, volume 184, issue 2-3, pages 183-201, 1998. 
[9]  Triplett, K. A., Ghiaasiaan, S. M., Abdel-Khalik, S. I., LeMouel, A., and McCord, 
B. N., ?Gas-liquid two-phase flow in micro-channels: part II: void-fraction and 
pressure drop,? International Journal of Multiphase Flow, volume 25, issue 3, pages 
395-410, 1999a. 
[10]  Triplett, K. A., Ghiaasiaan, S. M., Abdel-Khalik, S. I., and Sadowski, D. L., ?Gas-
liquid two-phase flow in micro-channels: part I: two-phase flow patterns,? 
International Journal of Multiphase Flow, volume 25, issue 3, pages 377-394, 
1999b. 
[11]  Lowe, D. C., and Rezkallah, K. S., ?Flow regime identification in microgravity two-
phase flows using void fraction signals,? International Journal of Multiphase Flow, 
volume 25, issue 3, pages 433-457, 1999. 
[12]  Kawahara, A., Sadatomi, M., Okayama, K., Kawaji, M., and Chung, P. M.-Y., 
?Effect of channel diameter and liquid properties on void fraction in adiabatic two-
phase flow through micro-channels,? Heat Transfer Engineering, volume 26, issue 3, 
pages 13-19, 2005. 
 
 
161 
[13]  Traviss, D. P., Rohsenow, W. M., and Baron, A. B., ?Flow regimes in horizontal 
two-phase flow with condensation,? ASHRAE Transactions, volume 79, issue 2, 
pages 31-39, 1973. 
[14]  Breber, G., Palen J. W., and Taborek, J., ?Prediction of horizontal tube-side 
condensation of pure components using flow regime criteria,? Journal of Heat 
Transfer, Transactions of ASME, volume 102, issue 3, pages 471-476, 1980. 
[15]  Sardesai, R. G., Owen, R. G., Pulling, D. J., ?Flow regimes for condensation of a 
vapor inside a horizontal tube,? Chemical Engineering and Science, volume 36, 
issue 7, pages 1173-1180, 1981. 
[16]  Soliman, H. M., ?On the annular-to-wavy flow pattern transition during 
condensation inside horizontal tubes,? Canadian Journal of Chemical Engineering, 
volume 60, issue 4, pages 475-481, 1982. 
[17]  Soliman, H. M., ?Mist-annular transition during condensation and its influence on 
the heat transfer mechanism,? International Journal of Multiphase Flow, volume 12, 
issue 2, pages 277-288, 1986. 
[18]  Tandon, T. N., Varma, H. K., and Gupta, C. P., ?New flow regimes map for 
condensation inside horizontal tubes,? Journal of Heat Transfer, volume 104, issue 
4, pages 763-768, 1982. 
[19]  Wang, C. C., Chiang, C. S., and Lu, D. C., ?Visual observation of two-phase flow 
pattern for R22, R134a, and R407C in a 6.5 mm smooth tube,? Experimental 
Thermal and Fluid Sciences, volume 15, issue 4, pages 395-405, 1997. 
 
 
162 
[20]  Dobson, M. K., and Chato, J. C., ?Condensation in smooth horizontal tubes,? 
Journal of Heat Transfer, Transactions of ASME, volume 120, issue 1, pages 193-
213, 1998. 
[21]  Coleman, J. W., and Garimella, S., ?Characterization of two-phase flow patterns in 
small diameter round and rectangular tubes,? International Journal of Heat and 
Mass Transfer, volume 42, issue 15, pages 2869-2881, 1999. 
[22] Coleman, J. W., and Garimella, S., ?Two-phase flow regimes in round, square, and 
rectangular tubes during condensation of refrigerant R134a,? International Journal 
of Refrigeration, volume 26, issue 1, pages 117-128, 2003. 
[23]  Yang, C. Y., and Webb, R. L., ?Condensation of R-12 in small hydraulic diameter 
extruded aluminum tubes with and without micro-fins,? International Journal of 
Heat and Mass Transfer, volume 39, issue 4, pages 791-800, 1996a. 
[24]  Yang, C. Y., and Webb, R. L., ?Friction pressure drop of R-12 in small hydraulic 
diameter extruded aluminum tubes with and without micro-fins,? International 
Journal of Heat and Mass Transfer, volume 39, issue 4, pages 801-809, 1996b. 
[25] Zhang, M., and Webb, R. L., ?Correlation of two-phase friction for refrigerants in 
small-diameter tubes,? Experimental Thermal and Fluid Sciences, volume 25, issue 
3-4, pages 131-139, 2001. 
[26]  Webb, R. L., and Ermis, K., ?Effect of hydraulic diameter on the condensation of 
R134a in Flat Extruded Aluminum tubes,? Journal of Enhanced Heat Transfer, 
volume 8, issue 2, pages 77-90, 2001. 
 
 
163 
[27]  Garimella, S., and Bandhauer, T. M., ?Measurement of condensation heat transfer 
coefficients in micro-channel tubes,? Proceedings of ASME International 
Mechanical Engineering Congress and Exposition, N.Y., pages 243-249, 2001. 
[28] Wilson, M. J., Newell, T. A., Chato, J. C. and Infante Ferreira, C. A., ?Refrigerant 
charge, pressure drop and condensation heat transfer in flattened tubes,? 
International Journal of Refrigeration, volume 26, issue 4, pages 442-451, 2003. 
[29]  Koyama, S., Kuwahara, K., Nakashita, K., and Yamamoto, K., ?An experimental 
study of condensation of refrigerant R134a in a multi-port extruded tube,? 
International Journal of Refrigeration, volume 26, issue 4, 425-432, 2003. 
[30]  Wang, W. W.-W., Radcliff, T. D., and Christensen, R. N., ?A condensation heat 
transfer correlation for milli-meter scale tubing with flow regime trasition,? 
Experimental Thermal and Fluid Sciences, volume 26, issue 5, pages 473-485, 2002. 
[31]  Baird, J. R., Fletcher, D. F., and Haynes, B. S., ?Local Condensation Heat Transfer 
Rates in Fine Passages,? International Journal of Heat and Mass Transfer, volume 
46, pages 4453-4466, 2003. 
[32]  Cavallini, A., Del Col, D., Doretti, L., Matkovic, M., Rosetto, L., and Zilio, C., 
Condensation heat transfer and pressure gradient inside multiport minichannels,? 
Heat Transfer Engineering, volume 26, issue 3, pages 45-55, 2005. 
[33]  Cavallini, A., Del Col, D., Doretti, L., Matkovic, M., Rossetto, L., and Zilio, C., 
?Update on Condensation Heat Transfer and Pressure Drop inside Minichannels,? 
Heat Transfer Engineering, volume 27, no. 4, pages 74-87, 2006. 
[34]  Kim, M. H., Shin, J. S., Huh, C., Kim, T. J., and Seo, K. W., ?A Study of 
Condensation Heat Transfer in a Single Mini-Tube and Review of Korean Micro- 
 
 
164 
and Mini-Channel Studies,?  Proceedings of the First International Conference on 
Micro-channels and Mini-channels, Rochester, N.Y., pages 47-58, 2003. 
[35]  Bandhauer, T. M., Agarwal, A., and Garimella, S., ?Measurement and modeling of 
condensation heat transfer, coefficients in circular micro-channels, Journal of Heat 
Transfer, volume 128, pages 1050-1058, 2006. 
[36]  Haraguchi, H., Koyama, S., Fujii, T., ?Condensation of refrigerants HCFC22, 
HFC134a and HCFC123 in a horizontal smooth tube: proposal of empirical 
expressions for the local heat transfer coefficient,? Transactions of  JSME(B), 
volume 60(574), pages 245?252, 1994. 
[37]  Mishima, K., and Hibiki, T., ?Effect of inner diameter on some characteristics of 
air-water two-phase flows in capillary tubes,? Transactions of JSME(B), volume 61, 
pages 99?106, 1995. 
[38]  Moser, K. W., Webb, R. L., and Na, B., ?A new equivalent Reynolds number 
models for condensation in smooth tubes,? Journal of Heat Transfer, Transactions 
of ASME, volume 120, issue 2, pages 410-417, 1998. 
[39]   Friedel, L., ?Pressure drop during gas/vapor-liquid flow in pipes,? International 
Journal of Chemical Engineering, volume 20, issue 3, pages 352-367, 1980. 
[40]   Cavallini, A., Censi, G., Del Col, D., Doretti, L., Longo, G. A., and Rossetto, L., 
?Condensation of halogenated refrigerants inside smooth tubes,? HVAC and R 
Research, volume 8, issue 4, pages 429-451, 2002. 
[41]  Begg, E., Khrustalev, D., and Faghri, A., ?Complete condensation of forced-
convection two-phase flow in a miniature tube,? Journal of Heat Transfer, 
Transactions of ASME, volume 121, pages 904-905, 1999. 
 
 
165 
[42]  Wang, H. S., and Rose, R. W., ?A theory of film condensation in horizontal non-
circular section microchannels,? Journal of Heat Transfer, Transactions of ASME, 
volume 127, pages, 1096-1105, 2005. 
[43]  Wang, H. S., and Rose, R. W., ?Film condensation in horizontal microchannels: 
effect of channel shape,? International Journal of Thermal Sciences, volume 45, 
pages 1205-1212, 2006. 
[44]  Munoz-Cobo, J. L., Herranz, L., Sancho, J., Tkachenko, I., and Verdu, G., 
?Turbulent vapor condensation with non-condensable gases in vertical tubes,? 
International Journal of Mass Transfer, volume 39, no. 15, pages 3249-3260. 
[45]  Kays, W. M., Crawford, M. E. and Weigand, B., Convective Heat and Mass 
Transfer, 4th edition, McGraw-Hill, N. Y., pages 212?215, 2005. 
[46]  Churchill, S. W., ?Friction-factor equation spans all fluid-flow regimes,? Chemical 
Engineering Progress, volume 84, pages 91?92, 1977. 
[47]  Chen, Y., and Cheng, P., ?Condensation of steam in silicon microchannels,? 
International Communications of Heat and Mass Transfer, volume 32, pages 175-
183, 2005. 
[48]  Nijhuis, T. A., Kreutzer, M. T., Romijn, A. C. J., Kapteijn, F., and Moulijn, J. A., 
?Monolithic catalysts as efficient three phase reactors,? Chemical Engineering 
Science, volume 56, pages 823-829.  
[49]  Taha, T., and Cui, Z. F., ?Hydrodynamics of slug flow inside capillaries,? Chemical 
Engineering Science, volume 59, pages 1181-1190, 2004. 
[50]  Taha, T., and Cui, Z. F., ?CFD modeling of slug flow in vertical tubes,? Chemical 
Engineering Science, volume 61, pages 676-687, 2006. 
 
 
166 
[51]  Taha, T., and Cui, Z. F., ?CFD modeling of slug flow inside square capillaries,? 
Chemical Engineering Science, volume 61, pages 665-675, 2006. 
[52]  van Baten, J. M., Krishna, R., ?CFD simulations of mass transfer from Taylor 
bubbles rising in circular capillaries,? Chemical Engineering Science, volume 59, 
pages 2535-2545, 2004. 
[53]  Zheng, D., He, X., and Che, D., ?CFD simulations of hydrodynamic characteristics 
in a gas-liquid vertical upward slug flow,? International Journal of Heat and Mass 
Transfer, volume 50, pages 4151-4165, 2007. 
[54]  FLUENT ver. 6.3 User?s Guide, ANSYS, Inc., PA, U.S.A. 
[55]  Teng, H., Cheng, P., and Zhao, T. S., ?Instability of condensate film and capillary 
blocking in small-diameter-thermosyphon condensers,? International Journal of 
Heat and Mass Transfer, volume 42, pages 3071-3083, 1999. 
[56]  Cavallini, A., Censi, G., Del Col, D., Doretti, L., Longo, G. A., and Rossetto, L., 
?Experimental investigation on condensation heat transfer and pressure drop of new 
HFC refrigerants (R134a, R125, R32, R410A, R236ea) in a horizontal smooth 
tube,? International Journal of Refrigeration, volume 24, pages 73-87, 2001. 
[57] Cavallini, A., Del Col, D., Doretti, L., Longo, G. A., and Rossetto, L., ?A New 
Computational Procedure for Heat Transfer and Pressure Drop during Refrigerant 
Condensation inside Enhanced Tubes,? Journal of Enhanced Heat Transfer, volume 
6, pages 441?456, 1999. 
[58]  Bowman, W. J., and Hitchcock, J. E., ?Transient, Compressible Heat Pipe Vapor 
Dynamics,? Proceedings of ASME National Heat Transfer Conference, Houston, 
TX,  volume 1, pages 329-338. 1988. 
 
 
167 
[59]  Sanchez, S., Gui, C., Elwenspoek, M.,?Spontaneous Direct Bonding of Thick 
Silicon Nitride?, Journal of Micromechanics and Microengineering, volume 7, 
pages 111-113, 1997. 
[60] Bower, R. W., Ismail, M. S., Roberds, B. E., ?Low Temperature Si3N4 Direct 
Bonding?, Applied Physics Letters, volume 62, no. 26, pages 3485-3487, 1993. 
[61] Baroczy, C. J., ?Correlation of Liquid Fraction in Two-Phase Flow with 
Applications to Liquid Metals,? Chemical Engineering Prog., Symposium Ser., 
volume 61(57), pages 179?191, 1965. 
[62]  Senocak, I., Shyy, W., ?Interfacial Dynamics-based Modeling of Turbulent 
Cavitating Flows, Part-1: Model Development and Steady-state Computations,? 
International Journal for Numerical Methods in Fluids, volume 44, pages 975-995, 
2004. 
[63]  Kandlikar, S. G., Garimella, S., Li, D., Colin, S., King, M. R., ?Heat Transfer and 
Fluid Flow in Minichannels and Microchannels,? Elsevier (Singapore) Pte Ltd., 
2006. 
[64] Ghiaasiaan S. M., ?Two-phase Flow, Boiling and Condensation: In Conventional 
and Miniature Systems,? Cambridge University Press, 2008.