INTERESSANT!!!!!4

THERMAL CONDUCTIVITY OF METAL OXIDE NANOFLUIDS
A Thesis
Presented to
The Academic Faculty
by
Michael Peter Beck
In Partial Fulfillment
of the Requirements for the Degree
Doctor of Philosophy in the
School of Chemical and Biomolecular Engineering
Georgia Institute of Technology

December 2008
THERMAL CONDUCTIVITY OF METAL OXIDE NANOFLUIDS
Approved by:
Dr. Amyn S. Teja, Advisor Dr. Sankar Nair

School of Chemical and Biomolecular School of Chemical and Biomolecular
Engineering Engineering
Georgia Institute of Technology Georgia Institute of Technology
Dr. Carson Meredith Dr. Ganesh Skandan

School of Chemical and Biomolecular CEO
Engineering NEI Corporation
Dr. Said I. Abdel-Khalik

School of Mechanical Engineering
Date Approved: August 12, 2008

Dedicated to the memory of
my mother
Linda S. Beck
(August 24, 1947 – October 25, 2000)
ACKNOWLEDGEMENTS
I would like to thank all of the people that have helped me throughout my studies
here at Georgia Tech. First, I would like to acknowledge my advisor Dr. Amyn Teja for
his support and guidance in my research. I would also like to acknowledge my
committee members, Dr. Nair, Dr. Meredith, Dr. Abdel-Khalik, and Dr. Skandan for their
help and advice over the years.
I would also like to thank the past and present members of the Teja research
group. Dr. Sun, Kerry, Ibrahim, Shutaro, Izumi, Chunbao, Nan, James, Anu, Pei Yoong,
Angel, Yanhui, Pramod, and Nelson have all been very helpful and have been great
friends. Dr. Sun taught me how to perform my experiments and was instrumental in the
initial stages of my research. I am grateful to Chunbao, Pei Yoong, Yanhui, and Pramod
who spent their time helping me through various stages of my work.
I would also like to thank Chris Gill of the Jones group for his help with the TEM
imaging of my samples and Jason Ward of the Koros group for his help with surface area
measurements. I appreciate the efforts of Brad and Jeff in the ChBE machine shop and
the technicians in the Mechanical Engineering electronics shop for their help during the
construction of my measurement apparatus.
I would also like to acknowledge my family for all of the encouragement and
support they have provided throughout my education. Most importantly, I would like to
thank my wife, Nan. Her love and support have helped me through many obstacles and
have allowed me to accomplish my goals. I would not be where I am today without her.
iv
TABLE OF CONTENTS
Page
ACKNOWLEDGEMENTS iv
LIST OF TABLES x
LIST OF FIGURES xi
LIST OF SYMBOLS xvii
SUMMARY xx
CHAPTER 1
INTRODUCTION 1
CHAPTER 2
BACKGROUND - EXPERIMENTAL 6
2.1 Thermal Conductivity of Liquids 6
2.1.1 Temperature and Pressure Dependence 8
2.2 Thermal Conductivity of Solids 12
2.2.1 Lattice Thermal Conductivity 13
2.2.2 Temperature Dependence 15
2.2.3 Thermal Conductivity of Solids as a Function of Size and Dimension 17
2.3 Thermal Conductivity of Heterogeneous Systems 20
2.3.1 Thermal Conductivity of Microparticle Dispersions 20
2.3.2 Thermal Conductivity of Nanofluids 21
2.3.2.1 Effect of Volume Fraction 25
2.3.2.2 Effect of Particle Size 29
2.3.2.3 Effect of Temperature 31
v
2.3.2.4 Effect of Ratio of Particle to Fluid Thermal Conductivity 33
2.3.2.5 Effect of Particle Surface Charge 33
2.3.2.6 Effect of Particle Arrangement in Suspension 34
CHAPTER 3
TRANSIENT HOT-WIRE METHOD 36
3.1 Basic Measurement Technique 38
3.2 Background of Transient Hot-wire Method 39
3.2.1 Basic Transient Hot-wire Method 39
3.2.2 Insulated Wire Method 40
3.2.3 Liquid Metal Wire Method 40
3.3 Construction of Transient Hot-wire Instrument 41
3.3.1 Instrumentation 44
3.3.1.1 Hot Wire Cell 44
3.3.1.2 Instrumentation of Bleazard and Teja 44
3.3.1.3 Updated Instrumentation 44
3.3.2 Data Acquisition and Analysis 45
3.4 Procedure, Analysis and Corrections 48
3.4.1 Procedure 48
3.4.2 Analysis 48
3.4.3 Corrections and Calibration 50
3.4.3.1 Insulating Layer and Three Dimensional Aspect of Wire 50
3.4.3.2 Finite Extent of the Liquid 51
3.4.3.3 Heat Transfer by Radiation 51
vi
3.4.3.4 Calibration 52
3.5 Calibration of Updated Instrument 53
3.6 Validation of Updated Instrument 60
CHAPTER 4
THERMAL CONDUCTIVITY MEASUREMENTS AND CHARACTERIZATION

OF SOLID / LIQUID DISPERSIONS 63
4.1 Materials and Methods 63
4.1.1 Materials 63
4.1.2 Sample Preparation 65
4.2 Nanofluid Characterization 67
4.2.1 Primary Particle Size 67
4.2.2 Secondary Particle Size 68
4.3 Thermal Conductivity Measurements 73
4.3.1 Temperature Studies 73
4.3.2 Particle Size Studies 86
4.4 Thermal Conductivity Behavior 95
CHAPTER 5
BACKGROUND - THEORY 96
5.1 Type I Thermal Conductivity Models 97
5.2 Type II Thermal Conductivity Models 100
5.3 Theoretical Bounds for Thermal Conductivity of Heterogeneous Materials 101
5.4 Accuracy of Thermal Conductivity Models 101
5.5 Geometric Mean 102
5.6 Nanofluid Thermal Conductivity Models 103
vii
5.6.1 Brownian Motion 104
5.6.2 Ordered Liquid Molecules 107

5.6.3 Nanoparticle Clustering 108
5.6.4 Interfacial Thermal Resistance 109
5.7 Summary of Thermal Conductivity Models for Nanofluids 110
CHAPTER 6
THERMAL CONDUCTIVITY MODELING OF SOLID / LIQUID DISPERSIONS

112
6.1 Evaluation of Thermal Conductivity Models 112
6.1.1 Particle Volume Fraction 112
6.1.2 Temperature 115
6.1.3 Thermal Conductivity of Individual Phases 119
6.1.4 Particle Diameter 123
6.1.5 Geometric Mean 127
6.2 Size Dependent Particle Thermal Conductivity 131
6.3 Prediction of Thermal Conductivity of Nanofluids 134
6.4 Particle Size Polydispersity 143
CHAPTER 7
CONCLUSIONS AND RECOMMENDATIONS 147
7.1 Conclusions 147
7.2 Future Work 149
APPENDIX A
CALIBRATION OF THERMOCOUPLE 151
APPENDIX B
viii
DATA ACQUISITION AND ANALYSIS PROGRAMS 153
B.1 LabVIEW Code 153
B.2 MATLAB Code 163
APPENDIX C
IMAGING OF NANOPARTICLES 169
REFERENCES 173
VITA 189
ix
LIST OF TABLES
Page
Table 2.1 Thermal conductivity of some solids and liquids at 25 ºC ............................. 7
Table 2.2 List of references containing thermal conductivity data of nanofluids......... 23
Table 3.1 Validation of transient hot wire apparatus with ethylene glycol .................. 61
Table 4.1. Sources of Alumina and Ceria powders and their properties provided by the
manufacturer ................................................................................................. 64
Table 4.2 Nanofluids provided by NEI Corporation (Somerset, NJ)............................ 64
Table 4.3 Mean primary and secondary particle sizes (nm) of powders dispersed in
water from specific surface area measurements, transmission electron
microscopy, and dynamic light scattering .................................................... 72
Table 4.4 Thermal conductivity of ethylene glycol nanofluids containing Al2O3 ........ 75
Table 4.5 Thermal conductivity of aqueous nanofluids containing Al2O3 ................... 75
Table 4.6 Thermal conductivity of aqueous nanofluids containing Al2O3 from NEI ... 76
Table 4.7 Thermal conductivity of water + ethylene glycol nanofluids containing Al2O3
(50 nm) from NEI ......................................................................................... 77
(10 nm) from NEI ......................................................................................... 78
Table 4.9 Thermal conductivity of water + ethylene glycol nanofluids containing TiO2
from NEI ...........................................................................................................
....................................................................................................................... 79
Table 4.10 Thermal conductivity of aqueous nanofluids containing Al2O3 ................... 88
Table 4.11 Thermal conductivity of nanofluids containing Al2O3 in ethylene glycol .... 89
Table 4.12 Thermal conductivity of aqueous nanofluids containing CeO2 .................... 89
Table 6.1 Deviations between thermal conductivity model predictions and

experimental values available in the literature............................................ 141
Table A.1 Calibration of type E thermocouple ............................................................ 151
x
LIST OF FIGURES
Page
Figure 2.1 Reduced thermal conductivity of diatomic fluids as a function of reduced

temperature and pressure adapted from Bird et al. [1]. The solid line
represents a saturated fluid. The reduced values are defined as the absolute
values normalized by the critical values. ...................................................... 10
Figure 2.2 Reduced thermal conductivity of water as a function of reduced temperature

and pressure [23]. The solid line represents saturated water. The reduced
values are defined as the absolute values normalized by the critical values. 11
Figure 2.3 Illustration of a phonon wave [36] ................................................................ 14
Figure 2.4 Thermal conductivity of alumina from various studies. Adapted from Bansal
& Zhu [37]. ................................................................................................... 16
Figure 2.5 Out-of-plane thermal conductivity at room temperature for silicon films.
Adapted from Liu & Asheghi [40]. A linear fit is provided as a visual aid. 19
Figure 2.6 Thermal conductivity of aqueous nanofluids containing metal oxides as

measured by Eastman et al. [60]. The dashed lines represent linear fits of the
data. ............................................................................................................... 26
Figure 2.7 Thermal conductivity enhancement of nanofluids containing TiO2 in water

for both spherical and rod shaped nanoparticles as measured by Murshed et
al. [66]. The lines represent fits of each linear regime................................. 27
Figure 2.8 Thermal conductivity enhancement for dilute metal nanofluids. The circles
(●) represent the measurements by Eastman et al. [74] of Cu (10 nm) in
ethylene glycol + 1 % (v/v) thioglycolic acid. The squares (■) represent the
measurements by Jana et al. [73] of Cu (35 – 50 nm) in water + laurate salt.
The triangles (▲) represent measurements by Putnam et al. [78] of
alkanethiolate – stabilized Au (2 – 4 nm) in toluene. ................................... 28
Figure 2.9 Thermal conductivity enhancement of nanofluids consisting of alumina

particles in pump oil and ethylene glycol as a function of particle size as
measured by Xie et al. [12] ........................................................................... 30
Figure 2.10 Thermal conductivity enhancement nanofluids as a function of temperature.

The circles and squares represent measurements by Das et al. [8] of aqueous
nanofluids containing alumina. The triangles represent measurements by
Yang and Han [10] of nanofluids consisting of perfluorohexane and Bi2Te3
nanorods. ....................................................................................................... 32
xi
Figure 3.1 Liquid metal transient hot-wire cell. Adapted from Bleazard and Teja [104] .
................................................................................................................... 42
Figure 3.2 Electrical diagram of transient hot-wire apparatus. ...................................... 43
Figure 3.4 Multiple calibrations of the effective length of the hot wire with dimethyl
phthalate (DMP) at 73 ºC as a function of the time period. .......................... 56
Figure 3.5 The mean effective length as a function of the time period determined from
each measurement of water and dimethyl phthalate at all temperatures ...... 57
Figure 3.6 The standard deviation of the effective wire length as a function of the time
period analyzed. ............................................................................................ 58
Figure 3.7 Residuals of the effective wire length for water and dimethyl phthalate at
various temperatures ..................................................................................... 59
Figure 3.8 Thermal conductivity enhancement of aqueous alumina nanofluids. The

solid circles (●) represent measurements from this study (47 nm diameter
alumina), and the others represent data from the literature (38 – 50 nm
diameter alumina) [8, 13, 50, 52, 57, 58]. ..................................................... 62
Figure 4.1 Aqueous nanofluid containing 3 % (v/v) alumina particles.......................... 66
Figure 4.2 Transmission electron microscopy image of alumina particles with a nominal
size of 47 nm from Nanophase Technologies (Romeoville, IL). The average
diameter of these particles is 77 nm and 46 nm as determined by TEM and
BET, respectively. Magnification = 100,000 ............................................. 70
size of 50 nm from Electron Microscopy Sciences (Hatfield, PA). The
average diameter of these particles is 20 nm and 16 nm as determined by
TEM and BET, respectively Magnification = 100,000................................ 71
Figure 4.4 Thermal conductivity of ethylene glycol and nanofluids consisting of

alumina nanoparticles (diameter = 12 nm) in ethylene glycol. Each data set
represents a different volume fraction of alumina (calculated at room
temperature). The curve represents the thermal conductivity of ethylene
glycol............................................................................................................. 80
Figure 4.5 Thermal conductivity of water and aqueous nanofluids containing alumina
nanoparticles (diameter = 12 nm). Each data set represents a different
volume fraction of alumina (calculated at room temperature). The curve
represents the thermal conductivity of water. Data of Das et al. [8] is
presented for comparison (diameter = 38 nm). ............................................. 81
xii
nanoparticles (diameter = 50 nm) with a dispersant. Each data set represents
a different volume fraction of alumina (calculated at room temperature). The
curve represents the thermal conductivity of water. ..................................... 82
Figure 4.7 Thermal conductivity of a 50 % (w/w) ethylene glycol and water mixture and
nanofluids consisting of this mixture and alumina nanoparticles (diameter =
50 nm) with a dispersant. Each data set represents a different volume
fraction of alumina (calculated at room temperature). The curve represents
the thermal conductivity of the ethylene glycol and water mixture.............. 83
the thermal conductivity of the ethylene glycol and water mixture.............. 84
nanofluids consisting of this mixture and titania nanoparticles (diameter = 2
nm) with a dispersant. Each data set represents a different volume fraction
of alumina (calculated at room temperature). The curve represents the
thermal conductivity of the ethylene glycol and water mixture. .................. 85
Figure 4.10 Thermal conductivity versus volume fraction for aqueous nanofluids
containing alumina at room temperature. The lines represent linear fits. .... 90
containing ceria at room temperature. The lines represent linear fits. ......... 91
Figure 4.12 Thermal conductivity of aqueous nanofluids containing 2 - 4 % (v/v)

alumina at room temperature. Empirical curve fits are provided to aid in
visual detection of trends in the data. Data from Chon et al. [50] and
Timofeeva et al. [58] is presented for comparison. ...................................... 92
Figure 4.13 Thermal conductivity of nanofluids consisting of 2 - 4 % (v/v) alumina in

ethylene glycol at room temperature............................................................. 93
Figure 4.14 Thermal conductivity enhancement of nanofluids containing 3 % (v/v)

alumina as a function of the ratio of solid to liquid thermal conductivity at
room temperature. ......................................................................................... 94
Figure 5.1 Schematic of method by Prasher et al. [150] to determine the thermal
conductivity of an aggregate which includes the backbone of nanoparticles,
the dead end nanoparticles and the fluid surrounding the particles. ........... 109
xiii
Figure 6.1 Thermal conductivity measurements and predictions for an aqueous
nanofluid containing alumina (diameter = 72 nm) from several models as a
function of volume fraction (φ). The error bars represent the estimated
measurement error. ..................................................................................... 114
Figure 6.2 Thermal conductivity enhancement predictions for an aqueous nanofluid

containing 5 % (v/v) alumina (40 nm) from several models as a function of
temperature. ................................................................................................ 117

ethylene glycol and alumina nanoparticles (diameter = 12 nm). Each data set
temperature). The dashed lines represent the Maxwell equation for each
concentration. The error bars represent estimated error for the
measurements. ............................................................................................. 118
Figure 6.4 The thermal conductivity of 5 % (v/v) alumina in pump oil, ethylene glycol,
glycerol, and water from Xie et al. [59]. The dashed lines represents the
volume fraction weighted geometric mean or the Maxwell equation......... 121
Figure 6.5 The Hashin and Shtrikman bounds for the thermal conductivity of a
heterogeneous material as a function of the ratio of the individual phase
thermal conductivies (α). The lower bound is equivalent to the Maxwell
model. The volume fraction in these calculations is 5 %. ......................... 122

containing 5 % (v/v) alumina at room temperature from several models as a
function of particle diameter. ...................................................................... 125
Figure 6.7 Thermal conductivity enhancement of aqueous nanofluids containing 2 - 4 %

(v/v) alumina at room temperature. Empirical curve fits are provided to aid
in visual detection of trends in the data. ..................................................... 126
Figure 6.8 Thermal conductivity enhancement of aqueous nanofluids containing 4 %

(v/v) alumina at room temperature. The Nan et al. [155] model and the
Maxwell model [118] are provided for comparison. The thermal
conductivity of bulk alumina (40 W m-1 K-1) was used in the predictions. 129

in visual detection of trends in the data. The dashed lines represent the
volume fraction weighted geometric mean at each concentration using the
bulk alumina thermal conductivity (40 W m-1 K-1). ................................... 130
xiv
Figure 6.10 Thermal conductivity enhancement of aqueous nanofluids containing
alumina at room temperature for various mean particle diameters. The lines
represent least squares fits of the volume fraction weighted geometric mean
by adjusting the thermal conductivity of the particle (k2)........................... 132
Figure 6.11 Estimated thermal conductivity of alumina particles from the volume
fraction weighted geometric mean. The curve represents a semi-empirical fit
(eq. 6.5). ...................................................................................................... 133
Figure 6.12 Thermal conductivity enhancement of nanofluids consisting of 2 and 3 %

(v/v) alumina in ethylene glycol at room temperature. The circles and
squares represent experimental data at 2 % and 3 % respectively. The
dashed lines represent predictions using Equation 6.5 with the volume
fraction weighted geometric mean. ............................................................. 136

temperature). The dashed lines represent predictions using Equation 6.5 with
the volume fraction weighted geometric mean. .......................................... 137
Figure 6.14 Thermal conductivity of alumina nanofluids as a function of αbulk from this
work and the work of Xie et al. [59]. The thermal conductivity of the bulk
solid was used to determine αbulk. The dashed lines represent predictions
using Equation 6.5 with the volume fraction weighted geometric mean. ... 138
Figure 6.15 Thermal conductivity enhancement of nanofluids consisting of alumina in

ethylene glycol at room temperature from Xie et al. [12]. The dashed lines
represent predictions using Equation 6.5 with the volume fraction weighted
geometric mean. .......................................................................................... 139

alumina (diameter = 35 nm) from Eastman et al. [60]. The dashed line
represents a prediction using Equation 6.5 with the volume fraction weighted
geometric mean. .......................................................................................... 140
size of 300 nm from Electron Microscopy Sciences. The average diameter
of these particles is 99 nm and 71 nm as determined by TEM and BET,
respectively. Magnification = 20,000 ........................................................ 145
size of 47 nm from Nanophase Technologies The average diameter of these
particles is 77 nm and 46 nm as determined by TEM and BET, respectively.
Magnification = 30,000............................................................................... 146
xv
Figure A.1 Calibration of type E thermocouple ............................................................ 152
Figure C.1 Transmission electron microscopy image of alumina particles with a nominal
size of 11 nm from Nanostructure and Amorphous Materials. The average
BET, respectively Magnification = 10,000 ................................................ 171
respectively. Magnification = 15,000 ........................................................ 172
xvi
LIST OF SYMBOLS
a radius of dispersed spheres
A heat transfer area
A1 adjustable parameter in Eq. 6.4
b inside diameter of cell
b radius of spherical volume (Eq. 5.8)
B radiation parameter in Eq. 3.12
c cell constant in Eq. 3.1
C1 proportionality constant in Eq. 5.20
d diameter of particle
df diameter of fluid molecule
D diffusivity
D pipe diameter (Eq. 1.2)
f empirical factor in Eq. 5.16
gv roots of J0
G temperature gradient
h heat transfer coefficient
h thickness of ordered liquid molecular layer (Eq. 5.27)
H overall heat transfer coefficient
I electric current
J0 zero-order Bessel function of the first kind
k thermal conductivity
k1 thermal conductivity of liquid
k2 thermal conductivity of solid
xvii
kB Boltzmann constant
kbulk thermal conductivity of bulk solid
keff effective thermal conductivity of dispersion
klayer thermal conductivity of ordered liquid molecules
L wire length
L pipe length (Eq. 1.2)
m mass
n empirical shape factor in Eq. 5.10
nS number of spheres
q heat dissipated per unit length
P1 one dimensional porosity
Q heat flow
r radial distance from heat source
ri inner radius of capillary
ro outer radius of capillary
R1, R2 electric resistance of fixed resistors
RB interfacial thermal resistance (Kapitza resistance)
RD electric resistance of decade resistor
RW electric resistance of wire
t time
T temperature
Tid ideal temperature of wire
V electric potential across Wheatstone bridge
V electric potential (Eq. 3.2)
V specific volume (Eq. 2.1)
xviii
VS source electric potential
Y0 zero-order Bessel function of the second kind
Greek Symbols
α ratio of solid to liquid thermal conductivity
α thermal diffusivity (Chapter 3)
γ exponent of Euler’s constant
ε constant related to Kapitza resistance in Eq. 5.20
κ temperature coefficient of mercury resistivity
μ viscosity
μb viscosity of fluid at bulk temperature
μ0 viscosity of fluid at wall temperature
ρ density
σ standard deviation of P1
τ relaxation time constant
φ particle volume fraction
ψ sphericity
Subscripts
G borosilicate glass
Hg mercury
xix
SUMMARY
The focus of this work is on nanofluids containing metal oxides. The specific
goals were the determination of the effects of temperature and particle size on the thermal
conductivity of nanofluids, the elucidation of the mechanism of conduction, and the
development of a predictive thermal conductivity model.
Thermal conductivity measurements were performed with a liquid metal transient
hot wire apparatus, which made possible measurements of electrically conducting fluids
at elevated temperature. The nanofluid samples consisted of either alumina, titania, or
ceria particles dispersed in deionized water, ethylene glycol, or a mixture of the two.
Thermal conductivity measurements were performed at room temperature and at other
temperatures up to 422 K that is over the largest temperature range that has been hitherto
reported in the literature. Additionally, measurements were performed on nanofluids
containing seven sizes of alumina particles with average diameters ranging from 8 to 282
nm, which is the largest number of sizes than previously considered.
The results of the thermal conductivity measurements at elevated temperatures
revealed that the thermal conductivity relationship with temperature mimics the
temperature relationship for the base liquid. The results of the thermal conductivity
measurements of nanofluids containing different sizes of particles revealed that the
addition of the smaller nanoparticles yielded lower thermal conductivity. The thermal
conductivity measurements of dispersions containing larger particles (> 50 nm) yielded
values that were well represented by predictions from the volume fraction – weighted
geometric mean, which display agreement with measurements than predictions from
xx
theoretical models for solid – liquid dispersions. These results also suggest that the
thermal conductivity of the solid nanoparticles is less than the thermal conductivity of the
bulk solid due to phonon scattering at the solid – liquid interface.
A predictive model has been developed for the thermal conductivity of nanofluids
containing alumina. The model incorporates the particle size dependence of the thermal
conductivity of solids in the volume fraction – weighted geometric mean. The model was
fit to the experimental data for aqueous nanofluids containing alumina by adjusting a
single parameter. The resulting model was capable of predicting the thermal conductivity
of nanofluids containing ethylene glycol and alumina at various temperatures and
containing various particle sizes from this work (within 2.3 %) and from the literature
(within 5.0 %). Lastly, the model was used to evaluate the consistency of published data
on the thermal conductivity of nanofluids.
xxi
CHAPTER 1
INTRODUCTION
The advent of high heat flow processes has created significant demand for new
technologies to enhance heat transfer. For example, microprocessors have continually
become smaller and more powerful, and as a result heat flow demands have steadily
increased over time leading to new challenges in thermal management. Furthermore,
there is increasing interest in improving the efficiency of existing heat transfer processes.
An example is in automotive systems where improved heat transfer could lead to smaller
heat exchangers for cooling resulting in reduced weight of the vehicle.
Many methods are available to improve heat transfer in processes. The flow of
heat in a process can be calculated using,
Q = hAΔT (1.1)
where Q is the heat flow, h is the heat transfer coefficient, A is the heat transfer area, and
ΔT is the temperature difference that results in heat flow [1]. We can see from this
equation that increased heat transfer can be achieved by:
(i) increasing ΔT
(ii) increasing A
(iii) increasing h
A greater temperature difference ΔT can lead to increased the heat flow, but ΔT is
often limited by process or materials constraints. For example, the maximum
temperature in a nuclear reactor must be kept below a certain value to avoid runaway
reactions and meltdown. Therefore, increased ΔT can only be achieved by decreasing the
1
the temperature of the coolant. However, this would reduce the rate of the nuclear
reaction and decrease the efficiency of the process.
Maximizing the heat transfer area A is a common strategy to improve heat
transfer, and many heat exchangers such as radiators and plate-and-frame heat
exchangers are designed to maximize the heat transfer area [2]. However, this strategy
cannot be employed in microprocessors and microelectromechanical systems (MEMS)
because the area can not be increased. In aerospace and automotive systems, increasing
the heat transfer area can only be achieved by increasing the size of the heat exchanger
which can lead to unwanted increases in weight.
Heat transfer improvements can also be achieved by increasing the heat transfer
coefficient h either by using more efficient heat transfer methods, or by improving the
transport properties of the heat transfer material. For example, heat transfer systems
which employ forced convection of a gas exhibit a greater heat transfer coefficient than
systems which employ free convection of a gas. Alternatively, the heat transfer
coefficient can be increased by enhancing the properties of the coolant for a given
method of heat transfer. Additives are often added to liquid coolants to improve specific
properties. For example, glycols are added to water to depress its freezing point and to
increase its boiling point. The heat transfer coefficient can be improved via the addition
of solid particles to the liquid coolant. In the case of nanosized particles, the resulting
dispersion is known as a nanofluid [3]. Xuan and Li [4] studied the forced convective
heat transfer of an aqueous nanofluid containing 2 % (v/v) copper nanoparticles in a
horizontal tube and observed a 39 % increase in heat transfer coefficient over that of the
pure coolant. Similarly, Wen and Ding [5] performed a pool boiling heat transfer
2
experiment with an aqueous nanofluid containing 1.25 % (v/v) alumina nanoparticles on
a polished stainless steel surface and observed a 40 % enhancement of the heat transfer
coefficient compared to that of pure water. Based on these results, the addition of
nanoparticles to a coolant appears to significantly enhance the heat transfer coefficient.
The relationship between the heat transfer coefficient and transport and
thermodynamic properties of the coolant can be ascertained from heat transfer
correlations. For example, in laminar flow through a pipe, the heat transfer coefficient is
obtained from the Nusselt number Nu via the Sieder and Tate correlation [1],
1 0.14
hD ⎛ D⎞ 3⎛μ ⎞
Nu = = 1.86⎜ Re Pr ⎟ ⎜⎜ b ⎟⎟ (1.2)
k ⎝ L ⎠ ⎝ μ0 ⎠
where D is the diameter of the pipe, k is the thermal conductivity of the fluid, Re is the
Reynolds number, Pr is the Prandtl number, L is the length of the pipe, and μb and μ0 are
the viscosities of the fluid at the bulk temperature and the wall temperature, respectively.
Equation 1.2 shows the relationship between the heat transfer coefficient and the thermal
conductivity of the coolant. Increases in thermal conductivity therefore relate directly to
increases in the heat transfer coefficient. It is not surprising therefore that nanofluids
exhibit enhanced heat transfer since they also exhibit enhanced thermal conductivity [6].
The thermal conductivity of a liquid coolant can be increased with the addition of
a more thermally conductive liquid or with the addition of solids, which are inherently
more thermally conductive than liquids. Whereas a highly thermally conductive liquid,
such as water, has a thermal conductivity that is 4 – 6 times greater than that of a less
conductive liquid, solid thermal conductivity can be as much as three orders of magnitude
3
greater than liquid thermal conductivity. Thus, much research has been dedicated to
nanofluids containing solid nanoparticles [6].
The thermal conductivities of nanofluids containing solid particles are generally
greater than values predicted by theories of transport in heterogeneous materials.
Existing models predict no more than 6 % enhancement of the thermal conductivity of
poly (α-olefin) oil containing dispersed 1 % carbon nanotubes. However, Choi et al. [7]
observed that the thermal conductivity of this nanofluid was 150 % greater than that of
the oil alone. Thermal conductivities for a variety of nanofluids are discussed in chapter
2, and it can be seen that many of these nanofluids exhibit significant thermal
conductivity enhancement. However, most of the studies are limited to nanofluids
containing a single size of nanoparticles at room temperature [6]. The few studies
focusing on the effect of temperature on the thermal conductivity of nanofluids have
produced seemingly conflicting results. Some have observed increasing thermal
conductivity with increasing temperature [8, 9], while others have reported the opposite
[10, 11]. Similar discrepancies have arisen as to the effect of particle size on the thermal
conductivity of nanofluids [12, 13]. Based on these data, several theories have been
proposed to explain the anomalous thermal conductivity behavior. The most prevalent
theories involve the Brownian motion of particles to create a microconvective effect, or
the ordering of liquid molecules at the solid interface to enhance conduction through
those molecules, or the clustering of nanoparticles to form pathways of lower thermal
resistance [14].
Semi-empirical thermal conductivity models have been developed based on these
mechanisms as detailed in Chapter 3. These models are effective in fitting some of the
4
thermal conductivity data from the literature. However, they are ineffective in predicting
the general behavior of the thermal conductivity of nanofluids. Thus, the mechanisms for
thermal conduction in nanofluids remain to be resolved. Furthermore, it remains unclear
how parameters such as temperature and particle size affect the thermal conductivity of
nanofluids.
This work focuses on the effects of various parameters on the thermal
conductivity of nanofluids and the development of a new model for predictive purposes.
A systematic experimental study has been performed in which thermal conductivity has
been measured for nanofluids containing metal oxides over a wider temperature range
than previously studied, and for more particle sizes than previously considered. The
measurements have been performed using a liquid metal transient hot wire method
described in Chapter 3. The resulting data are presented in Chapter 4 and are used to
resolve discrepancies in the temperature dependence of the thermal conductivity and
particle size dependence of the thermal conductivity.
The data are also used to rigorously test the prediction of published models for the
thermal conductivity of nanofluids and thereby identify (or eliminate) mechanisms for
heat transport in these fluids. Based on these tests, a new model for thermal transport in
nanofluids is proposed and is described in Chapter 6. The predictive capabilities of the
model are outlined, so that new nanofluids and/or new thermal management options can
be explained for different applications that require enhanced heat transfer fluids.
5
CHAPTER 2
BACKGROUND - EXPERIMENTAL
Nanofluids can be considered to be composite materials consisting of a solid
discrete phase and a liquid continuous phase, and the behavior of their thermal
conductivity can best be understood by considering the thermal conductivity of the
individual phases. This chapter provides a review of thermal conductivity data for
liquids, solids, and solid – liquid dispersions.
2.1 Thermal Conductivity of Liquids
The thermal conductivity of liquids is generally greater than that of gases but
much less than that of solids. Values range from 0.06 W m-1 K-1 for fluorocarbons (FC-
72 from 3M) to 0.6 W m-1 K-1 for water. Associating liquids such as water and ethylene
glycol exhibit greater thermal conductivity than nonpolar liquids such as hexane. Table
2.1 displays the thermal conductivity of several solids and liquids that have been used in
nanofluid investigations.
Heat conduction occurs in liquids through the interactions between vibrating
molecules in a temperature gradient. As liquid molecules vibrate around their
equilibrium positions, their force fields overlap with their nearest neighbors. The
molecules at a greater temperature vibrate at a higher frequency and transfer heat through
these vibrationss to the molecules vibrating at a lower frequency due to their lower
temperature [15]. Liquids such as water, ammonia, or glycols have greater thermal
conductivity than non-polar liquids such as alkanes because of the stronger
intermolecular force between polar molecules [16].
6
Table 2.1 Thermal conductivity of some solids and liquids at 25 ºC
Material Thermal Conductivity (W m-1 K-1)

Carbon Nanotubes 2000 [7]
Diamond 900 – 2320 [17]
SiC 490 [2]
Silver 424 [2]
Copper 398 [2]
Gold 315 [2]
Aluminum 273 [2]
Graphite 119 – 165 [7]
Iron 80 [2]
Cupric Oxide 77 [18]
Alumina 40 [19]
Zinc Oxide 29 [13]
Carbon Nanofibers 13 [20]
Titania 8.4 [13]
Iron (II, III) Oxide 7.0 [21]
Bismuth Telluride 5.0 [22]
Carbon (amorphous) 1.59 [17]
Silica 1.34 [2]
C60 – C70 (Fullerenes) 0.4 [18]
Water 0.608 [23]
Glycerol 0.285 [2]
Ethylene Glycol 0.257 [2]
Ethanol 0.172 [2]
Decene 0.14 [24]
Toluene 0.133 [2]
Perfluorotriethylamine 0.13 [25]
Hexane 0.126 [26]
Poly (α-olefin) oil 0.117 [27]
Perfluorohexane (FC-72) 0.057 [10]
7
2.1.1 Temperature and Pressure Dependence
The thermal conductivity of liquids generally exhibits a linearly decreasing
relationship with increasing temperature due to thermal expansion. As the density
decreases, the thermal conductivity also decreases. This is demonstrated by the following
relationship between the temperature coefficient of thermal conductivity and the
coefficient of thermal expansion,
1 ⎛ ∂k ⎞ 1 ⎛ ∂V ⎞ 1 ⎛ ∂ρ ⎞
⎜ ⎟∝− ⎜ ⎟= ⎜ ⎟ (2.1)
k ⎝ ∂T ⎠ V ⎝ ∂T ⎠ ρ ⎝ ∂T ⎠
Thus, the relative change of thermal conductivity with temperature is directly
proportional to the relative change of density [28]. The temperature dependence is
relatively weak when compared to that for the viscosity. For example, the thermal
conductivity of hexane decreases 22.6 % when heated from 235 K to 335 K at
atmospheric pressure, but the viscosity decreases 68.7 % for the same temperature change
[26]. The difference arises from kinetic theory, which yields an exponential decreasing
of viscosity with increasing temperature. The pressure has an even weaker effect on the
thermal conductivity of liquids. The thermal conductivity of hexane increases just 4.8 %
when pressurized from 1 to 101 bar at 298 K [26]. The pressure dependence also arises
from the relationship between thermal conductivity and density. These temperature and
pressure trends are illustrated in Figure 2.1, which displays a corresponding – states plot
for the reduced thermal conductivity of liquids. Note the relationship between the
thermal conductivity and temperature is linear except near the critical point. The plot
was developed from the thermal conductivity of several monatomic liquids, and is often
used to estimate the thermal conductivity of many polyatomic liquids as well [1].
8
However, this plot is insufficient when estimating the thermal conductivity of polar or
associating liquids such as water and ethylene glycol. For these liquids, there may exist a
maximum on the thermal conductivity versus temperature curve, as is the case for water
and glycols [23, 29]. This phenomenon for water and other associating liquids is due to
changes in the local structure of the hydrogen bonding network with changes in
temperature. At lower temperatures, a portion of the energy being transferred becomes
stored in the hydrogen bonds as they form a network. With increasing temperature, less
energy is captured by the structural changes of the hydrogen bonding network leading to
increased thermal conductivity [30]. This phenomenon competes with the typical
relationship between thermal conductivity and temperature, where the thermal
conductivity continually decreases with increasing temperature due to thermal expansion.
Figure 2.2 displays the effect of temperature and pressure on the thermal conductivity
water. The maximum thermal conductivity occurs at approximately 404 K for saturated
water.
9
10
c
Reduced Thermal Conductivity, k = k / k
r
Sat'd
P =0
r
P = 0.1
r
1 P =1
r
P =5
r
P = 10
r
0.1
0.2 0.4 0.6 0.8 1 3
Reduced Temperature, T = T / T
r c
Figure 2.1 Reduced thermal conductivity of diatomic fluids as a function of reduced

temperature and pressure adapted from Bird et al. [1]. The solid line
represents a saturated fluid. The reduced values are defined as the absolute
values normalized by the critical values.
10
c
Reduced Thermal Conductivity, k = k / k
r
1
Sat'd
P = 0.0046
r
P = 0.92
r
P = 2.3
r
0.1
0.4 0.5 0.6 0.7 0.8 0.9 1 2

Reduced Temperature, T r = T / Tc
Figure 2.2 Reduced thermal conductivity of water as a function of reduced temperature

and pressure [23]. The solid line represents saturated water. The reduced
values are defined as the absolute values normalized by the critical values.
11
2.2 Thermal Conductivity of Solids
Heat can be conducted through solids by several different excitations. In metals,
the primary heat carriers are free electrons. In insulators and in some semiconductors,
lattice waves (phonons) are primarily responsible for heat transfer. Thermal conduction
by electrons is more effective than conduction by phonons, which is demonstrated by the
orders of magnitude difference between the thermal conductivity of metals and insulators
in Table 2.1. As a rule of thumb materials that exhibit greater electrical conductivity also
exhibit greater thermal conductivity. Thus, copper is a better conductor of electricity
than aluminum and also exhibits a greater thermal conductivity. There are some notable
exceptions to this rule such as diamond, which is an excellent electrical insulator (1017 Ω
m-1 [31]), but has a thermal conductivity of 900 W m-1 K-1 [32].
The thermal conductivity is greatly dependent on the structure of the solid, even
when the solids have identical chemical formulas. For example, amorphous carbon has a
thermal conductivity of approximately 1.6 W m-1 K-1 [17], but diamond and carbon
nanotubes can exhibit thermal conductivities as high as 900 W m-1 K-1 [32] and 2000 W
m-1 K-1 [7], respectively. Crystalline solids typically conduct heat more readily than
amorphous solids and therefore their thermal conductivities are higher than those of
amorphous solids. In crystalline solids, the phase, crystallite size, and impurities affect
the thermal conductivity. In amorphous solids, the degree of molecular order is the
dominant variable in heat conduction [1] This work focuses on insulators, so the
following discussion is limited to conduction by phonons, which gives rise to lattice
thermal conductivity [33].
12
2.2.1 Lattice Thermal Conductivity
Phonons are defined as quantized modes of vibration in a solid crystalline lattice,
which arise from the vibrations of atoms within the lattice. Due to the proximity of
atoms, their vibrations are strongly coupled with those of neighboring atoms. Since
chemical bonds between atoms are generally not rigid and are similar to springs, the
displacement of an atom from its equilibrium position also affects neighboring atoms.
The oscillating motion of these coupled atoms is analogous to acoustic waves moving
through the lattice. In the presence of a temperature gradient, energy is propagated
through the lattice by these phonon waves [34], as illustrated in Figure 2.3. If the atoms
oscillated harmonically, the velocity of phonon waves would be the speed of sound in a
crystal. However, anharmonicity is caused by higher order interactions among atoms,
known as phonon scattering, which lead to a change in direction of the phonon wave
[35]. Phonon scattering can be divided into elastic phonon scattering, where phonon
momentum is conserved, and inelastic scattering, where it is not. Inelastic scattering
creates resistance to thermal transport and lowers the thermal conductivity. Scattering
can result from collisions of phonons with each other (Umklapp scattering) or defects in
the crystal structure such as impurities and grain boundaries.
13
Figure 2.3 Illustration of a phonon wave [36]
14
2.2.2 Temperature Dependence
The thermal conductivities generally increase for nonmetals and decrease for
metals as temperature increases [1]. An exception to this rule is alumina, which despite
being a nonmetal, exhibits a decreasing thermal conductivity with increasing temperature
(Figure 2.4). The thermal conductivity of alumina decreases 25 % when heated from 300
to 400 K. The magnitude of this reduction is similar to that generally exhibited by liquids
such as hexane. The thermal conductivity of hexane decreases 22.6 % when it is heated
from 235 to 335 K.
The decrease in thermal conductivity with increasing temperature for alumina and
other crystalline solids arises from the temperature dependence of Umklapp scattering,
which is the dominant source of heat transfer resistance at ambient and higher
temperatures. The mean free path for phonons (l) is limited by Umklapp scattering and
exhibits the following relationship:
l ∝ T −n (2.2)
where n is typically greater than 1.
15
40
Touloukian
Thermal Conductivity (W m K )
35
-1
Munro
Bansal & Zhu
-1
30 Youngblood
Barea et al.
25 Santos & Taylor
20
15
10
0
200 400 600 800 1000 1200 1400 1600 1800
Temperature (K)
Figure 2.4 Thermal conductivity of alumina from various studies. Adapted from Bansal
& Zhu [37].
16
2.2.3 Thermal Conductivity of Solids as a Function of Size and Dimension
Several recent studies have focused on thermal conduction in nanoscale thin films
due to their use in microelectronics [38-40]. Thermal conductivities of submicron films
were measured and shown to decrease as the thickness of the film decreased. For
example, Liu et al. [40] found that for a 20 nm thick silicon film, the out-of-plane thermal
conductivity was nearly an order of magnitude less than the bulk value. Figure 2.5
displays their data along with those of others for the out-of-plane thermal conductivity of
thin silicon films. They suggested that phonon scattering at the interface of the solid
becomes the dominant source of thermal resistance in solid nanomaterials because of
their large specific surface area. A less substantial decrease of the in-plane thermal
conductivity (~10 % at 300 K) was observed by Yu et al. [41] in a superlattice with a 70
nm periodic structure.
Phonon-interface scattering is not as well understood as other phonon scattering
processes (such as boundary scattering or phonon-phonon scattering), and it is seldom
incorporated into predictive methods for the thermal conductivity of solids. This could
be the reason why most methods are unable to predict the reduced thermal conductivity
of nanostructured materials [42]. Ziambaras and Hyldgaard [43] examined the thermal
conductivity of nanoscale films and wires using the Boltzmann transport equation and
including the effect of phonon-interface scattering. Their results indicate that the axial
thermal conductivity of a wire is less than the in-plane thermal conductivity of a film of
the same thickness. They suggested this effect is caused by confinement of the phonon
wave since the thickness of the nanomaterial is similar to the phonon mean free path,
similar to Knudsen diffusion. Thus, nanowires, which are confined in two dimensions,
17
should exhibit a lower thermal conductivity than nanofilms which are only confined in
one dimension. Li et al. [44] demonstrated this effect when they measured axial heat
conduction in silicon wires as small as 22 nm in diameter, and found the thermal
conductivity to be more than two orders of magnitude less than the bulk value. The axial
thermal conductivity of a 22 nm diameter Si nanowire is approximately 6 W m-1 K-1,
while the out-of-plane thermal conductivity for a 20 nm thick Si film is 22 W m-1 K-1
[40].
Nanoparticles should exhibit an even lower thermal conductivity than nanowires
or nanofilms because they are confined in three dimensions. Fang et al. [45] came to the
same conclusion using their molecular dynamic simulations to estimate the thermal
conductivity of silicon nanoparticles. They found that below 8 nm, silicon nanoparticles
exhibited a two order of magnitude decrease in thermal conductivity (~ 2 W m-1 K-1)
compared to the bulk material (237 W m-1 K-1). This decreased thermal conductivity of
nanoparticles should be considered in the study of the thermal conductivity of nanofluids.
18
150
-1
-1
100
50
Liu & Asheghi

Ju & Goodson
Asheghi et al.
0 4
10 100 1000 10
Film Thickness (nm)
Figure 2.5 Out-of-plane thermal conductivity at room temperature for silicon films.
Adapted from Liu & Asheghi [40]. A linear fit is provided as a visual aid.
19
2.3 Thermal Conductivity of Heterogeneous Systems
Thermal transport in heterogeneous systems has been studied extensively due to
their wide range of industrial applications. Some examples include the modeling of
transport in packed beds and the freezing and thawing of food products. Nanofluids
represent a subset of heterogeneous materials. Consequently, a comparison of the
thermal conductivity behavior of nanofluids and of other heterogeneous materials should
provide insight into the effect of decreasing particle size on the conduction mechanism.
2.3.1 Thermal Conductivity of Microparticle Dispersions
Shin and Lee [46] measured the thermal conductivity of polyethylene and
polypropylene particles suspended in mixtures of silicon oil and kerosene. The particle
diameters ranged from 25 to 300 μm. At 10 % (v/v) particles, all suspensions exhibited a
13 % thermal conductivity enhancement. Thermal conductivity enhancement is defined
as the relative difference in thermal conductivity of the dispersion (keff) and the pure
liquid (k1), (k eff − k1 ) k1 . Shin and Lee observed a linear relationship between the
thermal conductivity and volume fraction of particles, but they did not observe any
dependence on the particle size. Bjorneklett et al. [47] reported measurements of the
thermal conductivity of epoxy containing silver particles for various volume fractions of
silver. The epoxy containing 6.25 % (v/v) silver particles exhibited a 160 % thermal
conductivity enhancement due to the large thermal conductivity difference between the
silver and epoxy. The thermal conductivity varied linearly with volume fraction up to
6.25 % volume fraction. At volume fractions greater than 6.25 %, the slope increased
20
until the thermal conductivity of epoxy containing 23.1 % (v/v) silver particles was 790
% greater than the pure epoxy.
Turian et al. [48] measured the thermal conductivity of a number of fluids
containing as much as 50 % (v/v) suspended coal, glass, gypsum, and silica particles.
The ratios of the thermal conductivities of these solids to liquids (k2 / k1) were relatively
small (< 14). For example, silica has a thermal conductivity of approximately 1.3 W m-1
K-1 while for water it is 0.6 W m-1 K-1 giving a thermal conductivity ratio of 2.2.
Consequently, the thermal conductivity enhancement for such suspensions was found to
be small, 18 % enhancement in the case of an aqueous dispersion containing 12 % (v/v)
silica particles. Turian et al. also observed a linear relationship between thermal
conductivity and volume fraction at dilute concentrations, although beyond 10 % (v/v) of
particles, the slope increased as the concentration increased.
2.3.2 Thermal Conductivity of Nanofluids
The thermal conductivity of nanofluids has drawn increasing attention since Choi
[3] first postulated that heat transfer could be improved through the addition of metallic
nanoparticles to the heat transfer fluid. He addressed the limitation in thermal
conductivity of typical heat transfer fluids and suggested the addition of more conductive
solid particles would enhance the fluid thermal conductivity beyond that suggested by
conventional models. The advantages of using nanoparticles are that they are more easily
suspended in the fluid, they may be used in microchannels, and the small size causes less
wear to machinery. However, aggregation of particles must be minimized in order to
benefit from these effects of small particle size.
21
Eastman et al. [49] measured the thermal conductivity of nanofluids, and found
that an aqueous nanofluid containing 5 % (v/v) CuO nanoparticles exhibited a thermal
conductivity 60 % greater than that of water. Additionally, they reported a 40 % greater
thermal conductivity compared to water for an aqueous nanofluid containing 5 % volume
fraction of Al2O3 nanoparticles [49]. Subsequently, many authors have presented data for
a wide variety of nanofluids. A list of nanofluid systems for which the thermal
conductivity has been measured is presented in Table 2.2. Some of these nanofluids
exhibit thermal conductivities that are in good agreement with the conventional models
described in Chapter 5 (within 10 %), while other nanofluids exhibit anomalous thermal
conductivities which are greater than predicted (> 25% deviation). The reasons for these
discrepancies remain unknown, but they may arise from experimental error in the
measurement technique or differences in sample preparation.
22
Table 2.2 List of references containing thermal conductivity data of nanofluids
Nanoparticle k2 / k1 Fluid
Al2O3 66 Water [8, 9, 11-13, 49-60]
156 Ethylene Glycol [12, 13, 53, 57-59]
140 Glycerol [59]
342 Oil [12, 53, 59, 61]
CuO 127 Water [8, 9, 11, 18, 49, 51, 53, 57, 60, 62, 63]
300 Ethylene Glycol [18, 53, 57, 62, 64, 65]
TiO2 14 Water [11, 13, 51, 56, 66, 67]
33 Ethylene Glycol [13, 68]
Fe3O4 11.5 Water [69]
ZrO2 Water [51]
WO3 Ethylene Glycol [56]
ZnO 48 Water [13]
113 Ethylene Glycol [13]
SiO2 2.2 Water [18, 32, 62, 67]
5.2 Ethylene Glycol [32, 67]
7.8 Ethanol [67]
SiC 806 Water [70]
1910 Ethylene Glycol [70]
Cu 655 Water [71-73]
1550 Ethylene Glycol [60, 74, 75]
Water + Ethylene Glycol [76]
3400 Oil [49, 71]
3060 Perfluorotriethylamine [25]
Ag 697 Water [32, 77]
Water + Ammonia [77]
Water + Ethylene Glycol [77]
Au 518 Water [61, 73]
1830 Ethanol [78]
2370 Toluene [11, 54, 78]
Fe 132 Water [68]
311 Ethylene Glycol [55, 56, 79, 80]
AlxCuy Water [81, 82]
Ethylene Glycol [81, 82]
AgxCuy Oil [83]
AgxAly Water [82]
Ethylene Glycol [82]
Bi2Te3 42.7 Oil [10]
87.7 Perfluorohexane [10]
23
Table 2.2 (continued)
Carbon Nanotubes 3290 Water [11, 18, 24, 62, 73, 75, 84-89]
7780 Ethylene Glycol [24, 62, 75, 90, 91]
Antifreeze [87]
17100 Oil [7, 18, 27, 75, 90, 92, 93]
14300 Decene [24]
Carbon Nanofibers 21.4 Water [54]
111 Oil [27]
C60-C70 (Fullerenes) 0.66 Water [18]
3.01 Toluene [78]
3.42 Oil [18]
Graphite 196 Water [94]
1020 Oil [27]
Diamond 3500 Ethylene Glycol [32]
The following examples are some of the more notable thermal conductivity results for
nanofluids. Choi et al. [7] reported a 150 % thermal conductivity enhancement of poly
(α-olefin) oil with the addition of multiwalled carbon nanotubes (MWCNT) at 1 %
volume fraction. Similarly, Yang et al. [93] reported a 200 % thermal conductivity
enhancement for poly (α-olefin) oil containing 0.35 % (v/v) MWCNT. It is important to
note that this thermal conductivity enhancement was accompanied by a three order of
magnitude increase in viscosity. Eastman et al. [74] found a 40 % thermal conductivity
enhancement for ethylene glycol with 0.3 % (v/v) copper nanoparticles (10 nm diameter),
although the authors added about 1 % (v/v) thioglycolic acid to aid in the dispersion of
the nanoparticles. The addition of this dispersant yielded a greater thermal conductivity
than the same concentration of nanoparticles in the ethylene glycol without the
dispersant. Jana et al. [73] measured the thermal conductivity of a similar copper
containing nanofluid, except the base fluid was water and laurate salt was used as a
dispersant. They observed a 70 % thermal conductivity enhancement for 0.3 % (v/v) Cu
24
nanoparticles in water. Kang et al. [32] reported a 75 % thermal conductivity
enhancement for ethylene glycol with 1.2 % (v/v) diamond nanoparticles between 30 and
50 nm in diameter. Despite these remarkable results, some researchers have measured
the thermal conductivity of nanofluids and have found no anomalous results. Also, those
results can often be predicted by conventional thermal conductivity models [11, 51, 54,
58]. It is therefore unclear why certain nanofluids exhibit anomalous thermal
conductivity enhancement while other nanofluids do not.
2.3.2.1 Effect of Volume Fraction
Most of the nanofluid thermal conductivity data in the literature exhibit a linear
relationship with the volume fraction of particles as shown in Figure 2.6. However, some
exceptions have shown a non-linear relationship especially at low volume fraction (< 1
%) [66, 69]. In these studies, the slope of the thermal conductivity versus volume
fraction can be divided into two linear regimes. At low concentrations, the slope was
greater than at high concentrations. This typically occurred around 1 % (v/v) as seen in
Figure 2.7.
Some researchers have observed anomalous thermal conductivity enhancement
for dilute suspensions (< 1 % by volume) of metallic nanoparticles [72-74, 83]. In
addition to the aforementioned results of Eastman et al. [74] and Jana et al. [73], Ceylan
et al. [83] measured a thermal conductivity enhancement of 33 % for 0.006 % (v/v) Ag –
Cu alloy nanoparticles in pump oil. Yet others have reported no anomalous enhancement
for similar systems [54, 78]. Limited data are available for these ultra dilute metal
nanofluids, but there does not seem to be any explanations for the large differences in
results (Figure 2.8).
25
0.76
Al O in H O
-1
0.74 2 3 2
CuO in H O
-1
2
0.72
0.7
0.68
0.66
0.64
0.62
0 1 2 3 4 5 6
φ (%)
Figure 2.6 Thermal conductivity of aqueous nanofluids containing metal oxides as

measured by Eastman et al. [60]. The dashed lines represent linear fits of the
data.
26
35
30
25
(keff - k1) / k1 (%)
20
15
10
Spheres
5 Cylinders
0
0 1 2 3 4 5 6
φ (%)
Figure 2.7 Thermal conductivity enhancement of nanofluids containing TiO2 in water

for both spherical and rod shaped nanoparticles as measured by Murshed et
al. [66]. The lines represent fits of each linear regime.
27
80
Eastman et al.
60 Jana et al.
Putnam et al.
- k ) / k (%)
40
1
1
20
eff
(k
-20
0 0.2 0.4 0.6 0.8 1
φ (%)
Figure 2.8 Thermal conductivity enhancement for dilute metal nanofluids. The circles
(●) represent the measurements by Eastman et al. [74] of Cu (10 nm) in
ethylene glycol + 1 % (v/v) thioglycolic acid. The squares (■) represent the
measurements by Jana et al. [73] of Cu (35 – 50 nm) in water + laurate salt.
The triangles (▲) represent measurements by Putnam et al. [78] of
alkanethiolate – stabilized Au (2 – 4 nm) in toluene.
28
2.3.2.2 Effect of Particle Size
Xie et al [12] measured the thermal conductivity of nanofluids containing
different sizes of alumina nanoparticles with diameters between 12 nm (124 m2 g-1) and
304 nm (5 m2 g-1). With the exception of the largest particles, the thermal conductivity
decreased as particle size decreased as seen in Figure 2.9. They concluded that there is
an optimal particle size which yields the greatest thermal conductivity enhancement.
However, Kim et al. [13] measured the thermal conductivity of nanofluids containing
different sizes of Al2O3, TiO2, and ZnO in water and in ethylene glycol. They observed
greater thermal conductivity for nanofluids containing the smaller nanoparticles. For
nanofluids containing 3 % (v/v) TiO2 in ethylene glycol, the thermal conductivity
enhancement for the 10 nm sample (16 %) was approximately double the enhancement
for the 70 nm sample. Li and Peterson [52] also observed up to 8 % greater thermal
conductivity enhancement for aqueous nanofluids containing 36 nm Al2O3 particles
compared to nanofluids containing 47 nm Al2O3 particles.
29
40
35
(k - k ) / k (%)
30
f
f
25
20 Pump Oil
Ethylene Glycol
15
0 50 100 150 200 250 300 350
Average Particle Diameter (nm)
Figure 2.9 Thermal conductivity enhancement of nanofluids consisting of alumina

particles in pump oil and ethylene glycol as a function of particle size as
measured by Xie et al. [12]
30
2.3.2.3 Effect of Temperature
Das et al [8] measured the thermal conductivity of aqueous nanofluids containing
Al2O3 and CuO at temperatures between 20 and 50 ºC. They observed that the thermal
conductivity increased as the temperature increased and speculated that this behavior is
typical of nanofluids over greater temperature ranges as well. However, they did not
measure the thermal conductivity of aqueous nanofluids at temperatures greater than 130
ºC where the thermal conductivity of saturated water attains a maximum. Over the
limited temperature range considered in their study, a gradual curve could appear linear.
Thus, more comprehensive data is required before concluding whether the thermal
conductivity exhibits a linear relationship with temperature. In contrast, Yang and Han
[10] reported thermal conductivity measurements of perfluorohexane (FC-72) containing
semiconductor (Bi2Te3) nanorods. They observed a decrease in the effective thermal
conductivity as the temperature increased from 5 to 50 ºC. This trend is similar to the
relationship between the thermal conductivity of non-polar liquids and thermal
conductivity [1]. Figure 2.10 illustrates the contrast among the observed temperature
dependence of the thermal conductivity of nanofluids.
31
25
1 % Das et al.
20 4 % Das et al.
Yang and Han
- k ) / k (%)
15
1
1
10
eff
(k
0
270 280 290 300 310 320 330
Temperature (K)
Figure 2.10 Thermal conductivity enhancement nanofluids as a function of temperature.

The circles and squares represent measurements by Das et al. [8] of aqueous
nanofluids containing alumina. The triangles represent measurements by
Yang and Han [10] of nanofluids consisting of perfluorohexane and Bi2Te3
nanorods.
32
2.3.2.4 Effect of Ratio of Particle to Fluid Thermal Conductivity
Xie et al. [59] also examined the effect of the thermal conductivity of the base
fluid using measurements of the thermal conductivity of nanofluids containing 60 nm
diameter (25 m2 g-1) alumina nanoparticles dispersed in different base fluids. The ratio of
particle to fluid thermal conductivity, α, ranged from 75 to 326. They observed a greater
relative thermal conductivity enhancement for nanofluids with a greater α. The
enhancement for the least conductive nanofluid (alumina in pump oil, α = 326) was
nearly 40 %, while the enhancement for the most conductive nanofluid (alumina in water,
α = 75) was about 23 %.
Several studies have explored the thermal conductivity of nanofluids with the
same base fluid but with different nanoparticles [8, 9, 49, 53, 57, 62]. The thermal
conductivity enhancement was greater for systems with greater α.
Similarly, some of the largest thermal conductivity enhancements observed in
nanofluids have contained highly thermally conductive particles (copper [74], carbon
nanotubes [7], diamonds [32]) in the same base fluids as nanofluids containing less
conductive particles, which exhibited much lower thermal conductivity enhancements
[59].
2.3.2.5 Effect of Particle Surface Charge
Lee et al. [63] explored the effect of the charge at the particle surface by varying
the pH of the water before dispersing the nanoparticles. They observed a greater thermal
conductivity enhancement at the acidic and basic pH range and a lower enhancement for
33
nanofluids with neutral pH values and concluded that greater surface charges increase the
stability of the dispersion leading to an increased thermal conductivity.
2.3.2.6 Effect of Particle Arrangement in Suspension
Wright et al. [88] studied the thermal conductivity of dilute nanofluids containing
0.01 – 0.02 % Ni coated single wall carbon nanotubes (SWCNTs) within a magnetic
field. They observed greater thermal conductivity enhancement when the magnetic field
was applied, which suggested that the SWCNTs aligned to form chains creating greater
conductive paths within the nanofluid. Hong et al. [95] and Wensel et al. [89] observed a
similar phenomenon for dilute nanofluids containing both Fe2O3 and SWCNTs.
However, after the nanofluids remain in the magnetic field for a certain amount of time
the thermal conductivity decreases due to particle settling caused by greater particle
agglomeration.
In this review, several discrepancies among the thermal conductivity data of
nanofluids have been highlighted. Anomalous thermal conductivity enhancement ( up to
200 %) has been observed in certain nanofluids, while not in others. The effect of
smaller particle size has been shown to enhance conduction in some nanofluids and
hinder conduction in others. Similarly, the effect of higher temperature has been shown
to enhance conduction in some nanofluids, but it has also been shown to mimic the
thermal conductivity behavior of the base fluid. Alternatively, there appears to be
agreement on the relationship of the thermal conductivity of nanofluids with particle
volume fraction φ and with the ratio of solid to liquid thermal conductivity α. The
thermal conductivity linearly increases with particle volume fraction, and thermal
conductivity enhancement is greater for nanofluids with greater α.
34
The reliability of thermal conductivity values of nanofluids in the literature is
difficult to determine as evident by the aforementioned discrepancies in the observed
behavior of the thermal conductivity with temperature and particle size. However,
thermal conductivity measurements of commonly studied systems such as alumina in
water have been reproduced within 2 % at room temperature [8, 9, 13, 50, 57, 58]. The
lack of reliable and reproducible data at different temperatures and particle sizes make
the development of predictive models difficult.
35
CHAPTER 3
TRANSIENT HOT-WIRE METHOD
In order to measure the thermal conductivity of a fluid, the measurement
technique must be capable of isolating contributions due to heat conduction from those
resulting from convection and radiation. Steady-state or transient methods can be used to
achieve this. Steady-state methods generally confine the fluid between two surfaces,
maintained at different temperatures. Two of the common geometries employed in these
measurements are coaxial cylinders and parallel plates [96]. In the first method, the fluid
occupies a thin annular gap between the two vertically oriented cylinders. By uniformly
heating the inner cylinder while maintaining the outer cylinder at a constant lower
temperature, a constant axial thermal gradient can be achieved. In the second method,
the fluid occupies a thin gap between two horizontally aligned plates. The upper plate is
maintained at a higher temperature so that the adjacent fluid is hotter and therefore less
dense. This helps to eliminate free convection. Both methods are versatile and can be
used to measure thermal conductivity over a wide range of temperature and pressure.
Both methods are suitable for nonpolar as well as electrically conducting fluids. The
basic equation for the thermal conductivity in these methods is rather straightforward,
Q
k =c (3.1)
ΔT
where Q is the heat flow by conduction, ΔT is the temperature difference between the two
surfaces, and c is the cell constant for the specific cell geometry. The difficulty with
these methods is in the careful design and construction of the measurement cell.
36
Convection is minimized by keeping the coaxial cylinders perfectly aligned on a vertical
axis, or keeping the parallel plates perfectly horizontal. Improper placement of one
surface in relation to the other can yield a significant error due to convection. In addition,
a correction must be applied to account for radiation and other heat losses. Heat is
generated by resistive heating, and the heat flow Q due to conduction is obtained using,
Q = VI − Q S − Q R − QC (3.2)
where V is the electric potential, I is the electric current, QS is the heat flow to the
surroundings, QR is the heat flow by radiation, and QC is the heat flow by convection.
The heat flow to the surroundings must be determined experimentally by calibrating the
cell. The radiative heat flow can be found analytically for fluids which are entirely
transparent. However, no analytical solution is available for partially transparent fluids.
In this case, the heat transferred by radiation is found by measuring the apparent thermal
conductivity while varying the thickness of the fluid gap between the solid surfaces. Due
to this difficulty, the radiative heat transfer is generally minimized by designing the cell
with a thin gap (0.2 – 0.3 mm) between the surfaces and using solid materials with low
emissivity and highly polished surfaces. The convective heat transfer in these methods is
often negligible when the cell is designed properly, but a correction can be applied by
using the Rayleigh number with an empirical formula developed by Le Neindre and
Tufeu as listed in Wakeham et al. [96]. When all sources of error are accounted for in a
careful design, steady-state thermal conductivity measurements can be made with
accuracies in the 1 – 3 % range.
Transient methods generally require that the fluid only be heated for several
seconds to take advantage of the longer timescale for the onset of convection compared to
37
the propagation of thermal energy by conduction. These methods utilize a single surface
to heat the fluid and to measure the temperature response to find the thermal conductivity
of the surrounding medium. Transient methods eliminate the need for precise
temperature control needed in steady-state methods. They employ a heated cylinder,
plate, or wire suspended in a liquid to measure th thermal conductivity. The transient hot
wire method is particularly simple to use because the wire can be modeled as a line. This
simplicity also minimizes sources of heat loss compared to the steady-state systems. A
significant advantage of the transient hot wire method is the ability to analytically verify
the elimination of convective and radiative contributions to heat transfer. Free
convection would appear in the data analysis as a deviation from the transient conduction
model. Additionally, the radiative contribution to heat transfer is calculated in the
analysis as detailed in Section 3.4.3.3. These properties of the transient hot wire method
have made it the preferred method for thermal conductivity measurement, as
demonstrated by the use of this method in many recent nanofluid studies [7, 11-13, 18,
24, 25, 32, 49-51, 54-59, 62-66, 68-72, 74-77, 79, 80, 84-86, 90, 93, 94, 97].
3.1 Basic Measurement Technique
The transient hot-wire technique utilizes a thin metal wire immersed vertically in the
fluid of interest. This wire forms one resistor of a Wheatstone bridge with other resistors
of known resistances. When a constant voltage is applied to the initially balanced bridge,
heat is dissipated along the wire, and causes a temperature change in the wire. This
temperature change and heating period must be small enough that convection does not
occur during the measurement, and power is therefore supplied to the bridge for only a
few seconds. The small temperature change in the wire induces a small change in
38
electrical resistance of the wire, which can then be used to determine the temperature
change. Thus, the wire acts both as the heat source and a thermometer. The temperature
versus time relationship is related to the thermal conductivity of the fluid via a specific
solution to Fourier’s law for an infinite line heat source in an infinite medium,
q ⎛ 4αt ⎞
ΔT = ln⎜⎜ 2 ⎟⎟ (3.3)
4πk ⎝ r γ ⎠
where ΔT is the temperature change of the wire, q is the heat dissipated per unit length, r
is the distance from the heat source, γ is the exponent of Euler’s constant, and k and α are
the thermal conductivity and thermal diffusivity of the fluid, respectively [98].
Corrections must be applied for the finite extent of the real system, and are addressed in
section 3.3.3.
3.2 Background of Transient Hot-wire Method
3.2.1 Basic Transient Hot-wire Method
The transient hot-wire method had been developed over many years, but did not
become the predominant method of thermal conductivity measurement for fluids until the
work of de Groot et al. [99], who showed that this method could be used to make
absolute measurements of the thermal conductivity of gases up to 800 °C and 400 atm
with an estimated accuracy and precision of 0.2 %. The accuracy was due to careful
design of the system to minimize differences between the real and the model system.
Each assumption that was made in the derivation of Equation 3.3 was analyzed to
determine its significance [100], and corrections were applied to account for any
deviations, particularly the change in physical properties due to the temperature increase
and the finite extent of the wire and the fluid.
39
3.2.2 Insulated Wire Method
A transient hot-wire device with a bare metal wire can be used to measure the
thermal conductivity of fluids which are not electrically conductive, such as gases and
most organic liquids. In the case of electrically conducting liquids, current leakage
would occur from the wire into the fluid, which would introduce significant error in the
measurement of heat dissipated in the wire. Nagasaka and Nagashima [101] extended the
method to measure electrolytes by adding a polyester insulating layer to the platinum
wire to prevent electrical leakage. This layer of insulation added more complexity to the
method, and created another correction during data analysis. However, they were able to
measure the thermal conductivity of aqueous NaCl solutions between 0 and 45 °C with
an estimated accuracy of 0.5 %.
3.2.3 Liquid Metal Wire Method
Hoshi et al. [102] further extended the insulated transient hot-wire method by
using a borosilicate glass capillary as the insulating layer with a mercury wire. This
method allows measurement over a greater temperature range than a coated wire, which
is prone to cracking at high temperatures because of differences in thermal expansion of
the two materials. They measured the thermal conductivity of molten salts at
temperatures up to 300 °C with an estimated accuracy of 3.5 %.
The method developed by Hoshi et al. has been used extensively for the
measurement of the thermal conductivity of electrically conducting liquids by DiGuilio
and Teja who measured the thermal conductivity of aqueous salt solutions [103], and by
Bleazard and Teja who measured the thermal conductivity of aqueous LiBr solutions and
propionic acid and water mixtures [104], alkanediols [105], and acetic acid and water
40
mixtures [106]. More recently, Sun and Teja measured the thermal conductivity of
various glycol and water mixtures [107, 108] and benzoic acid and water mixtures [109]
using this method. The estimated accuracy of these measurements is approximately 2 %.
Two methods have been used to construct the capillary. The first method
employed a high temperature rubber cement to seal the ends of a fine pyrex capillary to
the ends of larger pyrex tubes to form a U shape shown in Figure 3.1 [110]. This method
limits the temperature range of the thermal conductivity measurements because the
capillary can break due to differences in the thermal expansion of the two materials. The
second method consists of a single pyrex tube that is heated and stretched to form the
capillary [111]. No sealant is needed, but the disadvantage is that the capillary created by
this method has a non-uniform cross sectional area. However, this method allows
measurements at higher temperatures than any of the previously mentioned methods.
Due to the versatility and demonstrated effectiveness of the second method, it was chosen
for measurement of the thermal conductivity of nanofluids in this work.
3.3 Construction of Transient Hot-wire Instrument
Two transient hot-wire instruments were used in this research. The first was
designed and constructed by Bleazard and Teja [104], and consisted of a transient hot-
wire cell and the electric apparatus illustrated in Figures 3.1 and 3.2, respectively. This
instrument is described in detail by Bleazard and Teja [104] and was used to make the
thermal conductivity measurements in Section 5.3.1. An updated apparatus which
utilizes greater computing power, provides faster analysis, and is more user-friendly, was
used to make the thermal conductivity measurements presented in Section 5.3.2.
41
Figure 3.1 Liquid metal transient hot-wire cell. Adapted from Bleazard and Teja [104]
42
Figure 3.2 Electrical diagram of transient hot-wire apparatus.
43
3.3.1 Instrumentation
3.3.1.1 Hot Wire Cell
Both the original and updated instrument employed the same transient hot-wire
cell. The cell consisted of a capillary formed from a standard wall borosilicate glass tube
(4 mm OD, 2 mm ID) which was fixed in a ceramic support with ceramic epoxy
(Thermeez Hi-Seal, Cotronics Corp.). The capillary was examined with a microscope to
determine the dimensions. The average inner diameter was approximately 100 microns
and the average outer diameter was approximately 250 microns. Mercury was drawn
through the capillary with a vacuum pump, and tungsten wires were used to connect the
mercury to the rest of the Wheatstone bridge. The mercury filled capillary was placed in
a glass test tube containing the fluid to be studied, and the test tube was sealed and placed
in a pressure vessel in a Techne fluidized sand bath (Model SBL-2D). The temperature
of the fluid was determined with a type E thermocouple (see Appendix A for calibration).
3.3.1.2 Instrumentation of Bleazard and Teja
The electrical circuit is shown in Figure 3.2. The Wheatstone bridge consisted of
the hot wire, two 100 ± 0.01 Ω precision resistors, and a 4-dial decade resistance box
(General Radio 1433A) adjustable by ± 0.01 Ω with a maximum of 111.1 Ω. The bridge
was powered by a Hewlett-Packard power supply (model 6213A), and a relay
(Magnecraft W172DIP-1) was placed in series with the power supply to close the circuit
when activated. An IBM PC XT and a Strawberry Tree analog to digital converter card
(ACPC-16) was used to control the relay and for data acquisition.
3.3.1.3 Updated Instrumentation
44
All of the aforementioned components were replaced in the updated instrument.
The new Wheatstone bridge circuit consisted of two 100 ± 0.005 % Ω precision
wirewound resistors (RCD Components, Inc., Manchester, NH, Model 150B-1000-VB)
and a 4-dial decade resistance box (IET Labs, Inc., Westbury, NY, HARS-X-4-0.01)
adjustable by ± 0.01 Ω with a maximum of 111.1 Ω. The bridge was powered by an DC
supply (Agilent, Santa Clara, CA, Model E3610A). A generic 100 μF capacitor was
placed in parallel with the power supply for signal conditioning. A solid state relay
(Omron Electronic Components, Schaumburg, IL, G3VM-61B1) was placed in series
with the power supply, which closed the circuit when activated. A Dell computer
containing an analog to digital (A/D) converter (IOtech, Cleveland, OH,
ADAC/5502MF) was used to control the relay and for data acquisition and analysis.
There is no significant difference between the resistors in the updated instrument
compared to the original instrument. However, the Agilent power supply is an upgrade
from the original power supply because it produces a more reliable constant voltage than
the original. Additionally, there was no capacitor for signal conditioning in the original
instrument. The greatest difference between the original hardware and the updated
hardware is the improved computing power of the Dell computer with the IOtech A/D.
The original system recorded data at approximately 60 Hz for each channel, while the
updated system records data at 1000 Hz for each channel.
3.3.2 Data Acquisition and Analysis
The apparatus designed by Bleazard and Teja employed a BASIC program to
control the relay and for data acquisition and a separate Fortran program to analyze the
data and obtain a thermal conductivity value.
45
In the updated apparatus, the graphical programming software LabVIEW (version
7.1) was chosen for the data acquisition and analysis and to control the relay. The
LabVIEW program was designed to automatically collect the data, import data into
MATLAB for data analysis, and then provide graphical analysis and a thermal
conductivity value. The results from a typical thermal conductivity measurement are
shown in Figure 3.3, and the actual LabVIEW and MATLAB code can be found in
Appendix B. The graphical analysis includes plots of the measured data (voltage offset
of the bridge and source voltage) and two plots of ΔT versus ln(t), one over the entire
time period of the measurement, and the other a snapshot of the linear time period used to
calculate thermal conductivity. The selection of this linear portion is described in Section
3.5. In the second plot, the experimental temperature changes are compared with
predicted values for the wire. Graphical analysis provides the user with the ability to
identify any errors that may arise during the measurement. Such errors include
disturbances in the voltage source or the onset of convection, which would appear as a
noticeable decrease in the slope on the ΔT versus ln(t) plot at later times.
46
47
Figure 3.3 User interface in LabVIEW for transient hot wire device
3.4 Procedure, Analysis and Corrections
The following is a detailed description of the procedure and analysis used in this
research.
3.4.1 Procedure
With the mercury filled capillary immersed in the fluid, the bridge is manually
balanced by adjusting the decade resistance to a value which minimizes the absolute
voltage offset. Then, the temperature of the fluid is measured using a thermocouple.
This is followed by activation of a relay to close the circuit and allow a constant voltage
to be supplied to the bridge. The current flowing through the wire gives rise to heat
dissipation, which leads to an increase in the temperature of the wire. After several
seconds, the relay is deactivated to open the circuit and stop current flow. The voltage
across the bridge is measured at regular intervals while current flows through the circuit,
and the voltage versus time data are analyzed as described below to determine the
thermal conductivity.
3.4.2 Analysis
The thermal conductivity of the liquid is found from Fourier’s law for an infinite
line heat source in an infinite medium (Eq. 3.3). Equation 3.3 can also be written,
⎛ 4α ⎞
ln(t ) +
q q
ΔT = ln⎜⎜ 2 ⎟⎟ (3.4)
4πk 4πk ⎝ r γ ⎠
Thus, if the physical properties of the fluid are effectively constant during the
measurement, the relationship between ΔT and ln (t) has the following slope,
d (ΔT ) q
= (3.5)
d (ln t ) 4πk
48
Since the heat dissipated in the wire is effectively constant, the slope is constant and the
relationship between ΔT and ln (t) is linear. Thus, the thermal conductivity can be found
from the slope and the heat dissipated (Eq. 3.5).
The temperature change of the wire is calculated from the voltage offset data
using the following equation to find the resistance of the wire RW,
VS RD (R1 + R2 )
RW (t ) = − RD (3.6)
VS R1 + V (t )(R1 + R2 )
where RD is the resistance of the decade box, R1 and R2 are the resistances of the
corresponding fixed resistors in Figure 3.2, VS is the voltage supplied to the bridge, and V
is the voltage across the bridge. Then, the resistance of the wire was converted to the
temperature change of the wire by,
RW (t ) − RW (0 )
ΔT (t ) = (3.7)
κRW (0 )
where κ is the temperature coefficient of the electrical resistivity of mercury. Lastly, the
average heat dissipated in the wire per unit length, q, is calculated from Joule’s first law,
I 2 RWav VS2 RWav

q= = (3.8)
L (RD + RWav )2 L
where I is the current flow through the wire, L is the length of the wire, and RWav is the
average resistance of the wire. The heat dissipated per unit length and the slope of the
linear relationship between ΔT and ln (t) are all that are needed to determine the thermal
conductivity using equation 3.5 if all of the assumptions are assumed to be correct.
However, some assumptions create significant error in the thermal conductivity
measurement. Some of these errors can be corrected theoretically, and others can only be
corrected through calibration.
49
3.4.3 Corrections and Calibration
The thermal conductivity is calculated from the temperature versus time response
of an ideal system using equation 3.5. The ideal temperature change ΔTid is obtained by
adding theoretical corrections to the measured temperature change of the wire,
ΔTid (t ) = ΔT (t ) + ∑ δTi (t ) (3.9)

i
where δTi are corrections to the temperature change.
3.4.3.1 Insulating Layer and Three Dimensional Aspect of Wire
The analysis in the previous section is based on a one dimensional linear heat
source immersed in a fluid. The radius of the wire, and the radius and physical properties
of the pyrex capillary were not considered in the model. Nagasaka and Nagashima [101]
derived the following expression to account for the temperature correction caused by the
insulating pyrex layer and for the three dimensional aspect of the wire:
− q ⎧⎪ ⎛ ri ⎞ 2k ⎛ ro ⎞ 1⎡ ⎛ 4αt ⎞⎤ ⎫⎪
δT1 (t ) =
k
⎨2 ln⎜ ⎟⎟ + ln⎜⎜ ⎟⎟ + + ⎢C1 + C 2 + C 3 ln⎜⎜ 2 ⎟⎟⎥ ⎬ (3.10)
4πk ⎪⎩ ⎜⎝ ro ⎠ k G ⎝ ri ⎠ 2k Hg t ⎢⎣ ⎝ ro γ ⎠⎥⎦ ⎪⎭
ri 2 ⎡⎛ k − k G ⎞⎛ 1 1 ⎞⎟ 4 2 ⎤
C1 = ⎢⎜⎜ ⎟⎜
⎟⎜ α
−
⎟
+ − ⎥
8 ⎢⎣⎝ k Hg ⎠⎝ Hg α G ⎠ α G α Hg ⎥⎦
ro2 ⎛1 1 ⎞ ri 2 ⎛⎜ k G k Hg ⎞ ⎛ ro
⎟ ln⎜
⎞
C2 = ⎜⎜ − ⎟⎟ + − ⎟⎟
⎜ ⎟ ⎜r
2 ⎝ α α G ⎠ k G ⎝ α G α Hg ⎠ ⎝ i ⎠
ri 2 ⎛⎜ k G k Hg ⎞ ro2 ⎛ k k G ⎞
⎟+
C3 = − ⎜ − ⎟
2k ⎜⎝ α G α Hg ⎟ 2k ⎜ α α ⎟
⎠ ⎝ G ⎠
where ri and ro are the inner and outer radii of the capillary, respectively, t is time, k is the
thermal conductivity, α is the thermal diffusivity, and γ is the exponential of Euler’s
50
constant. The subscripts Hg and G refer to properties of mercury and borosilicate glass,
respectively, while properties with no subscript refer to the fluid. The magnitude of this
correction is strongly dependent on the geometry of the capillary and can cause as much
as 10 % error in the thermal conductivity measurement if omitted.
3.4.3.2 Finite Extent of the Liquid
Equation 3.3 is based on an infinite volume of liquid to act as a heat sink. The
following expression was derived by Healy et al. [100] to calculate the magnitude of this
error:
⎡ ⎛ 4αt ⎞ ∞ ⎛ − gν2αt ⎞ ⎤
δT2 (t ) = ⎟⎟[πY0 ( gν )]2 ⎥
q
⎢ ⎜ 2 ⎟ ∑
ln ⎜ ⎟ + exp⎜⎜ (3.11)
4πk ⎢⎣ ⎝ b γ ⎠ ν =1 ⎝ b
2
⎠ ⎥⎦
where b is the inside diameter of the cell, Y0 is the zero-order Bessel function of the
second kind, and gν are the roots of J0, the zero-order Bessel function of the first kind.
The magnitude of this correction was calculated during each measurement to ensure the
assumption caused negligible error. In the present work, this correction was insignificant
( < 0.1 %).
3.4.3.3 Heat Transfer by Radiation
Nieto de Castro et al. [112] developed the following equation to determine the
amount of heat transfer by radiation during the measurement:
− qB ⎡ ri 2 ⎛ 4αt ⎞ ri 2 ⎤
δT3 (t ) = ⎢ ln⎜⎜ 2 ⎟⎟ + − t⎥ (3.12)
4πk ⎢⎣ 4α ⎝ ri γ ⎠ 4α ⎥⎦
where B is a radiation parameter. This parameter was determined from the following
equation after the first two temperature corrections are made:
51
q ⎛ Br 2 ⎞ ⎛ 4αt ⎞ Bqri 2
ΔTid (t ) =
Bqt
⎜1 + i ⎟⎟ ln⎜⎜ 2 ⎟⎟ + − (3.13)
⎜
4πk ⎝ 4α ⎠ ⎝ ri γ ⎠ 16παk 4πk
This correction was also insignificant for data measuremed in this study ( < 0.1 %).
3.4.3.4 Calibration
There are two sources of error that can not be accounted for theoretically. The
first is the non-uniform diameter of the capillary. The capillary was produced by heating
and stretching the borosilicate glass, which creates a capillary of varying thickness along
its length. In the preceding analysis, the average inner and outer diameters of the wire
were used in the calculations. Those values were approximated by examining the
capillary under a microscope. However, the use of an average wire diameter did not
eliminate this source of error. During Joule heating, a wire of varying diameter is
analogous to a series of resistors with varying resistance, which causes different levels of
heat dissipation along the length of the wire.
Secondly, axial heat conduction was not considered in any of the previous
calculations. The analysis was based on all heat conducting radially through the wire,
glass and fluid, since the model is of an infinite wire with no ends. The real system was
designed to minimize end effects, but they are not eliminated. Despite far less heat
transfer area, a small amount of heat is conducted axially in the mercury wire because
there is much less resistance to heat transfer through the mercury at the end of the wire
than through the borosilicate glass. De Groot et al. [113] solved this problem for the bare
platinum wire by constructing an instrument from two wires of the same diameter but
different lengths. DiGuilio and Teja [110] extended this method to the liquid metal
transient hot wire method. The end effects would be similar in both of these wires and
52
would effectively offset each other. The system would then be modeled as a single wire
with a length equal to the difference between the two wires. The differential method
requires that both wires be the same diameter, and that the insulation for each wire have
the same thickness. The differential method was not used in this work because of the
difficulty with producing capillaries of different lengths but equal diameters.
There is no analytical correction available for either of these sources of error.
However, calibration of the apparatus was used to compensate for these errors. By
measuring a reference fluid with a well known thermal conductivity, an effective wire
length was calculated. The effective wire length determined during the calibration was
then used for subsequent measurements and provides accurate results for various liquids,
temperatures, and pressures, since these errors are primarily dependent on the geometry
of the wire. To validate this approach, the thermal conductivity of another reference fluid
was measured as a function of temperature. Bleazard and Teja [104] used this approach
to show that overall the error was < 2 % using this method.
3.5 Calibration of Updated Instrument
The effective length of the wire was adjusted to calibrate the updated apparatus.
Two other parameters were considered in the calibration experiment. First, the voltage
supplied to the bridge had to be selected since this directly determines the temperature
rise in the wire. Since the voltage supplied to the bridge was constant, a higher voltage
leads to a greater temperature rise in the wire. If the voltage is too low, the noise in the
data becomes more significant and leads to lower resolution in the experimental
measurement. However, if the voltage is too large, convection may occur during the
measurement due to a larger temperature rise in the wire. A precise optimization was
53
unnecessary for this parameter. However, it is most important to operate within a range
that avoids both of these problems for the variety of fluids to be studied. For this
research, a value of 4 V was chosen based on several trials.
The second parameter is the time period corresponding to the linear portion of the
ΔT vs ln(t) relationship, which was determined through calibration. Time periods of 1,
1.5, 2, and 2.5 seconds were considered. At 1 second, the variation between
measurements was greater than at larger time periods. However, as the time period
increased, there was a greater change in temperature and therefore in the physical
properties of the fluid, which caused an increasingly larger change in the slope. For this
reason, 1.5 seconds was selected as the time period for the experiment. The specific time
corresponding to the beginning of the linear portion was then selected. This value along
with the effective wire length was determined during the calibration with reference fluids
water [23] and dimethyl phthalate [114]. Calibration data are presented in Figure 3.4 for
dimethyl phthalate at 73 ºC. The same data were collected for this fluid at 24, 117, and
145 ºC and for water at 24, 59, and 84 ºC. At least five measurements were performed
for each set of conditions. For water above 84 ºC, convection occurred and the data were
not considered in the calibration. The individual measurements were averaged to obtain a
single curve corresponding to a specific fluid at one temperature. From these seven
curves, the average and standard deviation were found for each time period and displayed
in Figures 3.5 and 3.6, respectively. The effective length and time period were selected
that produced the minimum standard deviation. The standard deviation reached a
minimum value of 0.055 cm for the time period of 0.88 to 2.38 seconds with a
54
corresponding effective length of 9.9 cm. A plot of residuals of the effective mean is
presented in Figure 3.7.
55
10.6
10.4
Effective Length (cm)
10.2
10
9.8 73A
73B
9.6 73C
73D
9.4 73E
73F
9.2
0 500 1000 1500 2000 2500 3000 3500 4000
Initial Time of Linear Section (ms)
Figure 3.4 Multiple calibrations of the effective length of the hot wire with dimethyl
phthalate (DMP) at 73 ºC as a function of the time period.
56
10.6
10.4
Effective Length (cm)
10.2
10
9.8
9.6
9.4
500 1000 1500 2000 2500 3000 3500 4000
Figure 3.5 The mean effective length as a function of the time period determined from
each measurement of water and dimethyl phthalate at all temperatures
57
0.35
0.3
Standard Deviation (cm)
0.25
0.2
0.15
0.1
0.05
500 1000 1500 2000 2500 3000 3500 4000
Figure 3.6 The standard deviation of the effective wire length as a function of the time
period analyzed.
58
1.5
0.5
Residuals (%)
-0.5
DMP
-1
HO
2
-1.5
280 300 320 340 360 380 400 420
Temperature (K)
Figure 3.7 Residuals of the effective wire length for water and dimethyl phthalate at
various temperatures
59
3.6 Validation of Updated Instrument
Ethylene glycol was used to validate the transient hot wire apparatus because it is
one of the base fluids considered in this study. Table 3.1 contains a comparison of the
measured thermal conductivity and reference values [115] at three temperatures. Each
reported measured value is the average of at least 5 measurements. The error for the
reference data is reported to be less than 5 %, and measured values from this work are
within 2 % of the reported values.
A heterogeneous system was also selected to validate the instrument for nanofluid
samples. An aqueous nanofluid containing alumina was selected due to the large amount
of data available for this system in the literature. Six thermal conductivity data sets [8,
13, 50, 52, 57, 58] are available for this system containing 0 – 5 % (v/v) alumina
nanoparticles with average particle diameters between 38 – 50 nm at room temperature.
Moreover, these six sets of data are in general agreement with each other within 1.5 %.
The measurements in this study for the aqueous nanofluid sample containing alumina (46
nm diameter) compare favorably with the results from the literature as illustrated in
Figure 3.8. The measured values in this study were within 1.1 % of a linear fit to the
literature values, which is well within the 2 % reported accuracy of this method.
60
Table 3.1 Validation of transient hot wire apparatus with ethylene glycol
Temperature (K) 298.0 323.9 352.4

kexp (W m-1 K-1) 0.257 0.260 0.264
kref (W m-1 K-1) [115] 0.254 0.257 0.258
% error 1.36 % 1.02 % 1.85 %
61
12
10
(keff - k1) / k1 (%)
6
Lee et al.
Chon et al.
4 Das et al.
Kim et al.
2 Timofeeva et al.
Li & Peterson
This work
0
0 1 2 3 4 5 6
φ (%)
Figure 3.8 Thermal conductivity enhancement of aqueous alumina nanofluids. The

solid circles (●) represent measurements from this study (47 nm diameter
alumina), and the others represent data from the literature (38 – 50 nm
diameter alumina) [8, 13, 50, 52, 57, 58].
62
CHAPTER 4
THERMAL CONDUCTIVITY MEASUREMENTS AND
CHARACTERIZATION OF SOLID / LIQUID DISPERSIONS
The focus of this work is on the thermal conductivity of dispersions containing
metal oxides in electrically conducting liquids. This chapter describes the systems that
were studied and the results of the thermal conductivity measurements as well as methods
for characterization of the nanofluids.
4.1 Materials and Methods
4.1.1 Materials
Nanofluids were prepared using alumina, ceria, or titania particles in deionized
water, ethylene glycol, or a mixture of the two. The resistivity of the pure deionized
water was greater than 18 MΩ·cm, and the ethylene glycol was reagent grade (99.0 %)
from VWR International (West Chester, PA). The particles were obtained from different
vendors as dry powders. Sources are listed in Table 4.1 together with the crystal phase
and average particle size reported by the manufacturer.
Commercial nanofluids were supplied by NEI Corporation (Somerset, NJ) and are
listed in Table 4.2. Each of these nanofluids contained a proprietary dispersant.
63
Table 4.1 Sources of Alumina and Ceria powders and their properties provided by the
manufacturer
Manufacturer (Product Name) Crystal Phase & Average Particle

Material Diameter (nm)
Electron Microscopy Sciences γ - Al2O3 50
Electron Microscopy Sciences α - Al2O3 300
Electron Microscopy Sciences α - Al2O3 1000
Nanophase (NanoTek) δ/γ - Al2O3 47
Nanostructured and Amorphous Materials γ - Al2O3 11
Nanostructured and Amorphous Materials γ - Al2O3 20
Nanostructured and Amorphous Materials α-Al2O3 150
Nanostructured and Amorphous Materials CeO2 15 - 30
Nanostructured and Amorphous Materials CeO2 70 - 100
Electron Microscopy Sciences (Hatfield, PA)

Nanophase (Romeoville, IL)
Nanostructured and Amorphous Materials (Los Alamos, NM)
Table 4.2 Nanofluids provided by NEI Corporation (Somerset, NJ)
Particle Average Particle Base Fluid Volume

Material Diameter (nm) Fractions
γ - Al2O3 50 nm Water 1.4, 2.9, 4.6 %
γ - Al2O3 50 nm 50 % (w/w) Water + Ethylene Glycol 0.5, 1, 2, 3, 4 %
γ - Al2O3 10 nm 50 % (w/w) Water + Ethylene Glycol 0.5, 1, 2, 3 %
TiO2 2 nm 50 % (w/w) Water + Ethylene Glycol 2, 4, 6, 8.5 %
64
4.1.2 Sample Preparation
Nanofluid samples were prepared by dispersing pre-weighed quantities of dry
particles in either deionized water or ethylene glycol. The pH of each aqueous mixture
was measured and, if necessary, was adjusted to 4.0 ± 0.2 by the addition of HCl to
promote dispersion. The mixtures were then subjected to ultrasonic mixing (Sonics &
Materials, Inc. Vibra-Cell VCX 750) for several minutes to break up any particle
aggregates. The acidic pH is much less than the isoelectric point of these particles (7-9
for alumina [116] and 6.7-8.7 for ceria [117]), thus ensuring a positive surface charge on
the particles. The surface charge enhanced repulsion between the particles, which
resulted in uniform dispersions for the duration of the experiments. An image of an
aqueous nanofluids containing alumina is displayed in Figure 4.1
65
Figure 4.1 Aqueous nanofluid containing 3 % (v/v) alumina particles
66
4.2 Nanofluid Characterization
4.2.1 Primary Particle Size
Average particle sizes provided by the manufacturer are typically determined by
surface area measurements of the dry powder. However, different lots of particles may
have different sizes, and this measurement provides no information regarding the size
distribution of particles. Thus, transmission electron microscopy (TEM) was employed
to obtain images of each type of alumina particle, and surface area measurements were
performed to determine the average particle size.
TEM images were obtained using a Hitachi (Pleasanton, CA) HF-2000
transmission electron microscope. TEM samples were prepared by diluting a small
amount of a nanofluid sample (approximately 10:1) and placing a few drops on the TEM
grid and allowing it to dry for 24 hours.
The Brunauer – Emmett – Teller (BET) surface areas were obtained from nitrogen
adsorption data measured by an Accelerated Surface Area and Porosity (ASAP 2020)
system from Micromeritics (Norcross, GA). The samples were degassed before the
measurements by ramping the temperature up to 300 ºC at a rate of 10 ºC / min and held
at that temperature for 100 minutes. Afterwards, free space in the sample tube was
measured using the analysis port of the ASAP 2000. The sample was then degassed by
heating to 300 ºC under vacuum for at least 4 hours. The BET surface areas were
calculated from P / P0 data between 0.05 and 0.30, where P and P0 are the equilibrium
and saturation pressures of nitrogen, respectively.
The average particle diameter was determined from the specific surface area and
an estimate of the shape of the particles obtained from the TEM images. All of the
67
particles were either spherical or ellipsoidal (sphericity > 0.9). Estimates of the average
and standard deviation of the particle diameter were obtained by measuring individual
particles in the TEM images. The images display aggregates consisting of many
particles, which make it difficult to differentiate individual particles. Therefore, only ten
particles were considered from each image. Examples of these images are displayed in
figures 4.2 and 4.3. The remaining images are shown in Appendix C. Table 4.3 displays
the average particle diameters obtained from both the surface area measurements and the
TEM images. The large differences between a few of the particle samples can be
attributed to the polydispersity of those samples (Figure 4.2). The surface area
measurements provide a more accurate mean than the TEM images due to the extremely
small sample size considered in the analysis of TEM images. Therefore, only the
diameters obtained from the surface area measurements are considered in the analysis of
thermal conductivity results. Only two samples of ceria were used in this study, thus
surface area measurements were sufficient without TEM imaging to demonstrate the
difference in size. The average particle size of the ceria was calculated assuming
spherical particles.
4.2.2 Secondary Particle Size
Particles in each sample were aggregated to some degree, which can be seen in
the TEM images. The aggregate size is also referred to as the secondary particle size and
is important in determining the stability of the dispersion. It is this size which determines
the Brownian velocity of particles in a liquid. An average hydrodynamic radius can be
calculated by measuring the velocity of the aggregates with dynamic light scattering. The
measurement is only valid for transparent or nearly transparent samples, so the technique
68
is not appropriate for the concentrations used in this work (0.5 – 5.0 %). Aqueous
samples were prepared as stated previously, and then a portion was diluted until the
liquid was transparent. The level of dilution was adjusted so that a similar average count
rate was achieved in each measurement. The measurements were performed with a
90Plus particle size analyzer from Brookhaven Instruments Corporation (Holtsville, NY).
The samples were measured at room temperature. Two runs were performed on each
sample, and each run lasted 2 minutes at a measurement rate of 1 per second. The results
are displayed in Table 4.3 and are presented as hydrodynamic diameters for consistency.
69
size of 47 nm from Nanophase Technologies (Romeoville, IL). The average
BET, respectively. Magnification = 100,000
70
size of 50 nm from Electron Microscopy Sciences (Hatfield, PA). The
average diameter of these particles is 20 nm and 16 nm as determined by
TEM and BET, respectively Magnification = 100,000
71
Table 4.3 Mean primary and secondary particle sizes (nm) of powders dispersed in water from specific surface area measurements,
transmission electron microscopy, and dynamic light scattering
Material Particle Diameter Particle Tranmission Electron Microscopy Hydrodynamic

from Diameter from Diameter
Diameter Standard Coefficient of
Manufacturer BET
Deviation Variation
Al2O3 11 8 6 2 33 % 320
Al2O3 20 12 10 2 20 % 265
Al2O3 47 46 77 110 140 % 205
Al2O3 50 16 20 5 25 % 520
Al2O3 150 245 180 38 21 % 790
Al2O3 300 71 99 95 96 % 492
Al2O3 1000 282 290 51 18 % 522
72
CeO2 15 – 30 12 NA NA NA 280
CeO2 70 – 100 74 NA NA NA 550
4.3 Thermal Conductivity Measurements
4.3.1 Temperature Studies
The thermal conductivity of nanofluids of specific mass fractions is presented in
Tables 4.4 - 4.9 and Figures 4.4 – 4.9. Each data point represents the average value of
five thermal conductivity measurements. Standard deviations for each series of
measurements are also provided. The measurements were made at temperatures ranging
from 294 K to 422 K using the original transient hot wire instrument constructed by
Bleazard and Teja [104], after the sample was allowed to come to thermal equilibrium at
each temperature for no less than 30 minutes. The samples consisted of 12 nm alumina
nanoparticles dispersed in water and in ethylene glycol, and the nanofluids provided by
NEI listed in Table 4.2. Note that for each mass fraction, the volume fraction decreases
with temperature since the density of the liquid phase decreases with temperature. Thus,
the volume fractions listed in the figures refer to the volume fraction at room
temperature.
The nanofluids containing ethylene glycol (Figure 4.4) and the ethylene glycol
and water mixture (Figures 4.7 and 4.8) clearly display curvature in the thermal
conductivity – temperature relationship. Furthermore, the maximum thermal
conductivity occurs at approximately the same temperature as the base liquid. This
relationship may be valid in aqueous nanofluids (Figures 4.5 and 4.6), but this cannot be
stated unambiguously since the thermal conductivity maximum for water occurs at a
temperature (~ 403 K) near the maximum temperature measured in this work (~ 422 K).
At temperatures greater than 422 K, convection was evident during the measurement,
which invalidated those measurements. Figure 4.5 displays a comparison of the thermal
73
conductivities for aqueous data from this work and from Das et al. [8]. Their data lies at
the lower temperature range of this work (< 320 K), but the thermal conductivity data
displays a stronger temperature dependence than observed in this work. This work and
the data from Das et al. appear to exhibit a linear relationship between thermal
conductivity and temperature below 320 K.
The nanofluids containing TiO2 nanoparticles exhibited a lower thermal
conductivity than the ethylene glycol and water mixtures (Figure 4.9). Furthermore, the
thermal conductivity of the nanofluids decreased as the volume fraction increased. This
suppression of the thermal conductivity with the addition of nanoparticles to a liquid has
not been observed previously in the literature.
74
Table 4.4 Thermal conductivity of ethylene glycol nanofluids containing Al2O3
Mass fraction T (K) Volume fraction Mean k Standard

of Al2O3 of Al2O3 W m-1 K-1 Deviation
3.26 % 302.0 1.00 % 0.258 0.004
3.26 % 323.4 0.98 % 0.259 0.004
3.26 % 347.3 0.97 % 0.262 0.001
3.26 % 372.2 0.95 % 0.267 0.003
3.26 % 392.4 0.94 % 0.264 0.002
3.26 % 411.1 0.92 % 0.260 0.002
9.34 % 296.3 3.00 % 0.276 0.003
9.34 % 323.6 2.95 % 0.282 0.002
9.34 % 349.0 2.90 % 0.284 0.005
9.34 % 373.3 2.85 % 0.285 0.005
9.34 % 392.1 2.81 % 0.287 0.003
9.34 % 409.6 2.78 % 0.280 0.007
12.2 % 304.0 3.99 % 0.290 0.005
12.2 % 323.7 3.94 % 0.291 0.005
12.2 % 348.5 3.87 % 0.294 0.003
12.2 % 373.3 3.81 % 0.293 0.006
12.2 % 391.0 3.76 % 0.288 0.007
12.2 % 409.0 3.71 % 0.285 0.007
Table 4.5 Thermal conductivity of aqueous nanofluids containing Al2O3

3.61 % 297.2 1.00 % 0.625 .014
3.61 % 321.6 0.99 % 0.641 .017
3.61 % 345.5 0.98 % 0.653 .011
3.61 % 371.1 0.96 % 0.670 .007
3.61 % 396.9 0.94 % 0.678 .005
10.3 % 298.7 3.00 % 0.636 .014
10.3 % 322.3 2.97 % 0.646 .009
10.3 % 346.9 2.94 % 0.674 .010
10.3 % 371.4 2.89 % 0.680 .006
10.3 % 391 2.85 % 0.696 .013
10.3 % 410.9 2.80 % 0.687 .010
16.3 % 302.3 4.99 % 0.650 .011
16.3 % 324.4 4.95 % 0.671 .005
16.3 % 348.3 4.89 % 0.685 .006
16.3 % 368.6 4.82 % 0.712 .006
16.3 % 391.9 4.74 % 0.720 .011
75
Table 4.6 Thermal conductivity of aqueous nanofluids containing Al2O3 from NEI

5.00 % 294.4 1.40 % 0.621 0.002
5.00 % 322.2 1.39 % 0.632 0.007
5.00 % 346.6 1.37 % 0.651 0.005
5.00 % 373.5 1.34 % 0.677 0.002
5.00 % 398.0 1.32 % 0.693 0.009
5.00 % 421.5 1.29 % 0.701 0.005
10.0 % 294.9 2.91 % 0.635 0.004
10.0 % 326.0 2.88 % 0.662 0.003
10.0 % 349.5 2.84 % 0.680 0.007
10.0 % 373.0 2.80 % 0.697 0.005
10.0 % 396.1 2.75 % 0.701 0.006
15.0 % 294.9 4.54 % 0.645 0.005
15.0 % 322.2 4.50 % 0.675 0.008
15.0 % 348.7 4.44 % 0.694 0.003
15.0 % 372.7 4.37 % 0.703 0.008
15.0 % 397.3 4.29 % 0.728 0.007
76
(50 nm) from NEI

1.71 % 294.7 0.50 % 0.403 0.002
1.71 % 322.0 0.49 % 0.422 0.003
1.71 % 349.6 0.48 % 0.430 0.004
1.71 % 373.1 0.47 % 0.440 0.002
1.71 % 397.3 0.46 % 0.446 0.003
1.71 % 418.6 0.45 % 0.431 0.005
3.37 % 295.4 1.00 % 0.411 0.003
3.37 % 322.2 0.99 % 0.425 0.003
3.37 % 348.4 0.97 % 0.435 0.006
3.37 % 372.6 0.95 % 0.442 0.003
3.37 % 397.1 0.93 % 0.453 0.003
3.37 % 419.6 0.91 % 0.440 0.009
6.59 % 294.6 2.00 % 0.424 0.004
6.59 % 326.1 1.97 % 0.454 0.004
6.59 % 348.0 1.94 % 0.461 0.005
6.59 % 373.5 1.90 % 0.469 0.005
6.59 % 397.2 1.86 % 0.473 0.004
6.59 % 418.1 1.82 % 0.474 0.008
9.65 % 295.1 3.00 % 0.432 0.004
9.65 % 322.8 2.96 % 0.446 0.004
9.65 % 347.4 2.91 % 0.457 0.002
9.65 % 374.4 2.85 % 0.464 0.003
9.65 % 397.5 2.79 % 0.467 0.003
9.65 % 418.6 2.73 % 0.468 0.002
12.6 % 295.3 4.00 % 0.440 0.002
12.6 % 326.5 3.93 % 0.460 0.009
12.6 % 349.5 3.87 % 0.471 0.005
12.6 % 372.7 3.81 % 0.482 0.004
12.6 % 398.1 3.72 % 0.481 0.002
12.6 % 420.1 3.64 % 0.484 0.004
77
(10 nm) from NEI

1.71 % 296.4 0.50 % 0.403 0.002
1.71 % 324.1 0.49 % 0.414 0.004
1.71 % 348.4 0.48 % 0.429 0.003
1.71 % 372.2 0.48 % 0.435 0.005
1.71 % 397.8 0.46 % 0.441 0.004
1.71 % 422.3 0.45 % 0.438 0.007
3.37 % 296.6 1.00 % 0.413 0.003
3.37 % 324.8 0.98 % 0.431 0.004
3.37 % 348.0 0.97 % 0.438 0.003
3.37 % 373.1 0.95 % 0.447 0.004
3.37 % 397.6 0.93 % 0.457 0.005
3.37 % 421.3 0.91 % 0.449 0.004
6.59 % 296.7 2.00 % 0.423 0.002
6.59 % 315.2 1.98 % 0.438 0.002
6.59 % 333.2 1.96 % 0.451 0.004
6.59 % 353.6 1.93 % 0.458 0.007
6.59 % 372.5 1.90 % 0.464 0.004
9.66 % 296.0 3.00 % 0.429 0.001
9.66 % 313.6 2.97 % 0.445 0.005
9.66 % 333.8 2.94 % 0.456 0.001
9.66 % 353.6 2.90 % 0.468 0.004
78
Table 4.9 Thermal conductivity of water + ethylene glycol nanofluids containing TiO2
from NEI

ofTiO2 of TiO2 W m-1 K-1 Deviation
6.59 % 297.3 2.00 % 0.392 0.004
6.59 % 321.3 1.97 % 0.397 0.004
6.59 % 348.0 1.94 % 0.414 0.006
6.59 % 372.9 1.90 % 0.422 0.006
6.59 % 397.0 1.86 % 0.424 0.003
6.59 % 421.2 1.82 % 0.420 0.008
12.6 % 297.6 4.00 % 0.384 0.002
12.6 % 324.4 3.94 % 0.398 0.003
12.6 % 348.3 3.88 % 0.402 0.003
12.6 % 373.4 3.81 % 0.413 0.003
12.6 % 397.4 3.73 % 0.414 0.002
12.6 % 419.0 3.65 % 0.417 0.002
18.1 % 296.8 6.00 % 0.383 0.002
18.1 % 321.9 5.92 % 0.389 0.003
18.1 % 348.5 5.82 % 0.402 0.003
18.1 % 373.7 5.71 % 0.414 0.006
18.1 % 397.4 5.60 % 0.417 0.005
18.1 % 419.0 5.48 % 0.416 0.002
24.3 % 296.8 8.50 % 0.362 0.003
24.3 % 323.7 8.38 % 0.375 0.001
24.3 % 347.9 8.26 % 0.375 0.002
24.3 % 372.7 8.11 % 0.383 0.001
24.3 % 397.3 7.94 % 0.384 0.003
24.3 % 418.6 7.78 % 0.389 0.003
79
0.3
-1
0.29
-1
0.28
EG
1%
0.27 3%
4%
0.26
0.25
280 300 320 340 360 380 400 420 440
Temperature (K)

alumina nanoparticles (diameter = 12 nm) in ethylene glycol. Each data set
temperature). The curve represents the thermal conductivity of ethylene
glycol.
80
0.8
-1
0.75
-1
0.7
0.65
Water
1%
3%
0.6 5%
1% [7]
4% [7]
0.55
280 300 320 340 360 380 400 420 440
Temperature (K)
nanoparticles (diameter = 12 nm). Each data set represents a different
volume fraction of alumina (calculated at room temperature). The curve
represents the thermal conductivity of water. Data of Das et al. [8] is
presented for comparison (diameter = 38 nm).
81
0.74
0.72
-1
-1
0.7
0.68
0.66
0.64
0.62 H2O
1.4%
0.6 2.9%
4.6%
0.58
280 300 320 340 360 380 400 420 440
Temperature (K)
nanoparticles (diameter = 50 nm) with a dispersant. Each data set represents
a different volume fraction of alumina (calculated at room temperature). The
curve represents the thermal conductivity of water.
82
0.5
-1
0.48
-1
EG /
0.46
HO
2
0.5%
0.44 1%
2%
3%
4%
0.42
0.4
280 300 320 340 360 380 400 420 440
Temperature (K)
the thermal conductivity of the ethylene glycol and water mixture.
83
0.5
-1
0.48
-1
EG /
0.46
HO
2
0.5%
0.44 1%
2%
3%
0.42
0.4
280 300 320 340 360 380 400 420 440
Temperature (K)
the thermal conductivity of the ethylene glycol and water mixture.
84
0.43
-1
0.42
-1
0.41
0.4 EG /
HO
2
0.39 2%
4%
0.38 6%
8.5%
0.37
0.36
280 300 320 340 360 380 400 420 440
Temperature (K)
nanofluids consisting of this mixture and titania nanoparticles (diameter = 2
nm) with a dispersant. Each data set represents a different volume fraction
of alumina (calculated at room temperature). The curve represents the
thermal conductivity of the ethylene glycol and water mixture.
85
4.3.2 Particle Size Studies
The thermal conductivity of nanofluids measured at specific mass fractions and
particle sizes is presented in Tables 4.10 - 4.12 and Figures 4.10 – 4.13. Each data point
represents the average value of at least five measurements. The standard deviations for
each series of measurements are also provided. All measurements were made at room
temperature with the new transient hot wire instrument described in section 3.3.1. The
samples consisted of alumina or ceria nanoparticles dispersed in water and alumina in
ethylene glycol. The error bars in the figures represent the estimated experimental error.
Figure 4.10 and 4.11 display a linear particle volume fraction dependence for
aqueous alumina and ceria nanofluids, respectively. The same trend has been observed in
many studies in the literature as discussed in section 2.3.2.1. Note that the slope
increases as the particle size increases. Figure 4.12 displays the thermal conductivities
for aqueous alumina nanofluids as a function of particle diameter. The thermal
conductivity generally decreases with decreasing particle size below a certain particle
diameter (~ 50 nm). The thermal conductivity of nanofluids containing larger particles
appears nearly constant with particle diameter. Additionally, the thermal conductivity
decreased as the particle size decreased for the nanofluids consisting of alumina in
ethylene glycol (Figure 4.13).
Figure 4.12 displays a comparison of thermal conductivities for aqueous
nanofluids containing alumina from this work and from the work of Chon et al. [50] and
Timofeeva et al. [58]. Chon et al observed an increase in the thermal conductivity of the
nanofluids as the particle diameter decreased. Alternatively, the data from Timofeeva et
al. is similar to data from this work, where the addition of the largest particles yielded the
86
greatest thermal conductivity of the nanofluids. In each study, only three different sizes
of nanoparticles were considered, but they demonstrate the discrepancy in nanofluid
thermal conductivity data in the literature.
The effect of the ratio of solid to liquid thermal conductivities α has not been
rigorously studied in this work. However, the thermal conductivity enhancement of
nanofluids containing ethylene glycol and alumina was greater than that of aqueous
nanofluids containing the same size of alumina particles. The ethylene glycol has a lower
thermal conductivity than water, so α is higher for these nanofluids containing ethylene
glycol. Figure 4.14 displays this difference in thermal conductivity enhancement.
87
Table 4.10 Thermal conductivity of aqueous nanofluids containing Al2O3
Average Particle T (K) Volume fraction Mean k Standard

Diameter (nm) of Al2O3 W m-1 K-1 Deviation
8 299.1 1.93 % 0.620 0.017
8 297.4 2.99 % 0.621 0.008
8 297.7 3.99 % 0.623 0.016
12 297.4 2.00 % 0.622 0.006
12 297.4 3.00 % 0.630 0.009
12 297.8 4.00 % 0.640 0.005
16 296.8 2.00 % 0.635 0.003
16 298.0 3.00 % 0.642 0.015
16 298.3 3.98 % 0.662 0.015
46 298.0 2.00 % 0.637 0.017
46 298.3 2.99 % 0.644 0.018
46 297.8 3.99 % 0.660 0.011
71 299.9 0.93 % 0.633 0.009
71 296.8 1.86 % 0.657 0.010
71 297.7 3.00 % 0.696 0.017
71 296.6 3.99 % 0.712 0.016
245 300.5 1.86 % 0.656 0.013
245 298.5 3.00 % 0.678 0.012
245 299.1 4.00 % 0.709 0.013
282 297.7 1.00 % 0.646 0.016
282 297.4 2.00 % 0.669 0.011
282 298.0 3.00 % 0.693 0.008
282 298.7 4.00 % 0.719 0.020
88
Table 4.11 Thermal conductivity of nanofluids containing Al2O3 in ethylene glycol

Diameter (nm) of Al2O3 W m-1 K-1 Deviation
12 298.6 2.00 % 0.272 0.001
12 298.0 3.00 % 0.276 0.002
16 298.6 2.00 % 0.273 0.001
16 297.9 3.00 % 0.281 0.002
245 298.6 2.00 % 0.280 0.002
245 298.3 2.99 % 0.294 0.003
282 297.9 2.00 % 0.283 0.002
282 298.0 3.01 % 0.299 0.002
Table 4.12 Thermal conductivity of aqueous nanofluids containing CeO2

Diameter (nm) of CeO2 W m-1 K-1 Deviation
12 297.7 2.00 % 0.630 0.006
12 297.9 3.00 % 0.649 0.010
12 298.0 4.00 % 0.671 0.013
74 297.8 2.00 % 0.659 0.010
74 298.7 3.00 % 0.693 0.016
74 298.1 4.00 % 0.730 0.020
89
0.74
-1
0.72
-1
0.7
0.68 12 nm
46 nm
245 nm
0.66
0.64
0.62
1.5 2 2.5 3 3.5 4 4.5
φ (%)
containing alumina at room temperature. The lines represent linear fits.
90
0.75
-1
-1
0.7
12 nm
74 nm
0.65
0.6
1.5 2 2.5 3 3.5 4 4.5
φ (%)
containing ceria at room temperature. The lines represent linear fits.
91
0.75
-1
-1
0.7
2%
3%
0.65 4%
1% [30]
2.5% [38]
0.6
0 50 100 150 200 250 300
Figure 4.12 Thermal conductivity of aqueous nanofluids containing 2 - 4 % (v/v)

alumina at room temperature. Empirical curve fits are provided to aid in
visual detection of trends in the data. Data from Chon et al. [50] and
Timofeeva et al. [58] is presented for comparison.
92
0.31
-1
0.3
-1
0.29
2%
3%
0.28
0.27
0.26
0 50 100 150 200 250 300
Figure 4.13 Thermal conductivity of nanofluids consisting of 2 - 4 % (v/v) alumina in

ethylene glycol at room temperature.
93
20
15
(keff - k1) / k1 (%)
12 nm
10 16 nm
245 nm
282 nm
0
60 80 100 120 140 160
α
Figure 4.14 Thermal conductivity enhancement of nanofluids containing 3 % (v/v)

alumina as a function of the ratio of solid to liquid thermal conductivity at
room temperature.
94
4.4 Thermal Conductivity Behavior
The thermal conductivity of nanofluids displayed a temperature dependence
similar to the base fluid in this work. This trend was observed in nanofluids consisting of
alumina and water, alumina and ethylene glycol, and alumina and a mixture of ethylene
glycol and water. A maximum was observed in the thermal conductivity versus
temperature curve for nanofluids at approximately the same temperature as the maximum
for the base fluid.
The thermal conductivity of nanofluids appears to decrease with decreasing
particle size below a certain diameter (< 50 nm), but seems constant above that size. This
trend was observed for aqueous alumina nanofluids. Nanofluids containing alumina and
ethylene glycol and aqueous ceria nanofluids also exhibited a decreasing thermal
conductivity with decreasing particle diameter.
The thermal conductivities of nanofluids in this work exhibited particle volume
fraction and α dependences in agreement with the literature. The thermal conductivity
linearly increases with increasing particle volume fraction, and the thermal conductivity
enhancement increases with increasing α.
95
CHAPTER 5
BACKGROUND - THEORY
The transport properties of heterogeneous systems have been the subject of many
investigation beginning with Maxwell [118] who derived an expression for the effective
conductivity of a medium containing a dilute suspension of non-interacting spheres as
follows:
k eff 3(α − 1)φ

= 1+ (5.1)
k1 (α + 2) − (α − 1)φ
where φ is the volume fraction of spheres and,
α = k2 k (5.2)
1
with k1 and k2 being the conductivities of the continuous and discrete phases,
respectively. This equation was derived for the electrical conductivity, but it applies to
other transport properties such as thermal conductivity. Many studies have focused on
extensions of this model to account for the interaction of particles in more concentrated
suspensions, particle shape, and arrangements of particles. The more commonly used
models are discussed in this section. A more detailed discussion of thermal conductivity
models for heterogeneous materials can be found elsewhere [119-121].
Tsotsas and Martin [120] classified thermal conductivity models for
heterogeneous systems into three categories, solutions (analytical or numerical) of the
Laplace equation (type I), empirical models consisting of combinations of resistances in
series and in parallel (type II), and solutions of a unit cell with either parallel heat flux
96
vectors or parallel isotherms (type III). This discussion will be limited to the first two of
these categories since they are more commonly used.
5.1 Type I Thermal Conductivity Models
The Maxwell equation (Eq. 5.1) was the first model for transport properties of a
heterogeneous system derived by solving the Laplace equation,
∇ 2T = 0 (5.3)
He considered a dilute dispersion of nS spheres with radius a and thermal conductivity k2
in a continuous medium with thermal conductivity k1.and a constant temperature
gradient, G, far from the spheres. The Laplace equation can then be solved in spherical
coordinates for the temperature inside and outside the spheres with the following
boundary conditions:
T (r , θ )r →∞ = Gr cos θ
T2 (r , θ ) r = a = T1 (r ,θ ) r = a (5.4)
∂T2 ∂T1
k2 = k1
∂r r =a ∂r r =a
The last boundary condition is based on the assumption that each sphere is surrounded by
the base fluid and does not interact with other spheres. Each sphere is considered
separately, and the total effect is obtained by multiplying by the number of spheres. The
resulting solutions for the temperature within and outside the sphere is as follows,
⎛ 3k1 ⎞
T2 (r , θ ) − T0 = ⎜⎜ ⎟⎟Gr cos θ (5.5)
⎝ 2 k1 + k 2 ⎠
⎛ k − k1 n S a 3 ⎞
T1 (r , θ ) − T0 = ⎜⎜1 − 2 3 ⎟
⎟Gr cos θ (5.6)
⎝ 2 k1 + k 2 r ⎠
97
where T0 is the temperature at the center of a sphere. If this dispersion with an effective
thermal conductivity keff is contained within a single sphere of radius b and surrounded by
the same base fluid, equation 5.3 can be solved with the following boundary conditions:
T (r , θ )r →∞ = Gr cos θ
Tm (r , θ ) r =b = T1 (r ,θ ) r =b (5.7)
∂Tm ∂T1
k eff = k1
∂r r =b ∂r r =b
The resulting solution is,
⎛ k eff − k1 b 3 ⎞
T1 (r ,θ ) − T0 = ⎜1 − ⎟Gr cos θ (5.8)
⎜ 2k + k r 3 ⎟
⎝ 1 eff ⎠
By introducing the volume fraction of spheres,
3
φ = nS a (5.9)
b3
equations 5.6 and 5.8 can be combined to yield equation 5.1.
Many models have been proposed to overcome some of the limitations of the
Maxwell model. For example, Lord Rayleigh [122] derived a series of linear equations
which provide a thermal conductivity for concentrated systems of square arrays of
cylinders or cubic arrays of spheres. Hamilton and Crosser [123] extended the Maxwell
equation to include an empirical shape factor, n,
k eff n(α − 1)φ

=1+ (5.10)
k1 (α + n − 1) − (α − 1)φ
For systems where α < 100, n is equal to 3, which is equivalent to the Maxwell equation.
For systems where α >100, n is defined by the following equation,
n= 3 (5.11)
ψ
98
where ψ is the sphericity of the particles. This model provided predictions within 2 % of
measured thermal conductivity values for mixtures of rubber with aluminum particles of
different shapes (0.5 < ψ < 1.0). Additionally, Jeffrey [124] extended the first order
Maxwell equation to include a second order term by including the effect of interactions
between pairs of spheres to give:
k eff ⎛ 3β 3 9β 3 α + 2 3β 4 ⎞
= 1 + 3βφ + φ 2 ⎜⎜ 3β 2 + + + 6 + ...⎟⎟ (5.12)
k1 ⎝ 4 16 2α + 3 2 ⎠
where,
α −1
β= (5.13)
α +2
The terms inside the parentheses are a convergent infinite series. This equation was also
developed for a dilute suspension. Thus, for small values of φ, the equation is equivalent
to the Maxwell equation if the second order term is neglected.
Progelhof et al. [121] described an equation developed by Bruggeman [125] for
the thermal conductivity of dilute suspensions of spheres. In contrast to the Maxwell
equation, the effective medium approximation was used in the derivation to give:
1
⎛ k 2 − k eff ⎞⎛ k1 ⎞ 3
1 − φ = ⎜⎜ ⎟⎜ ⎟ (5.14)
⎟⎜ k ⎟
⎝ k 2 − k1 ⎠⎝ eff ⎠
With the effective medium approximation, each sphere is considered to be suspended in a
medium with a homogenous conductivity of keff. Landauer [126] also employed this
approximation in his derivation for the thermal conductivity of a suspension of spheres,
k1 − k eff k 2 − k eff
(1 − φ ) +φ =0 (5.15)
k1 + 2k eff k 2 + 2k eff
99
Landauer observed that predictions of this equation compared favorably with
experimental electrical conductivity data for binary metallic mixtures over the entire
range of concentrations [126]. The derivation did not require the assumption of a dilute
suspension of spheres. However, the equation was developed under the assumption that a
single particle is surrounded by a random arrangement of volumes containing either the
medium or a particle. As a result, this model does not provide accurate predictions of the
thermal conductivity of dispersions containing aggregated particles [48].
5.2 Type II Thermal Conductivity Models
The second type of thermal conductivity model for heterogeneous materials
considers a suspension as an array of elements with specific resistances. This system is
then modeled as a combination of resistances in series and in parallel. The simplest of
these models was described by Keey [127] and developed by Krischer [128],
−1
⎡ 1− f ⎛ 1 − φ 2 φ 2 ⎞⎤
k eff =⎢ + f ⎜⎜ + ⎟⎟⎥ (5.16)
⎣ (1 − φ 2 )k 1 + φ k
2 2 ⎝ 1 k k 2 ⎠⎦
where f is an empirical factor equivalent to the relative proportion of parallel resistances
in a rectangular array of elements. No assumption is made regarding the spacing of the
discrete phase. Thus, the model is capable of incorporating the irregular arrangement of
particles. However, experimental data is required to estimate the empirical factor. A
useful aspect of this model is that it provides upper (f = 0) and lower (f = 1) bounds for
the thermal conductivity of heterogeneous materials. If f = 0, all of the particles are
arranged in series which creates a pathway of high thermal conductivity and maximizes
the effective thermal conductivity of the dispersion. If f = 1, all of the particles are
arranged in parallel and therefore minimizes the effective thermal conductivity of the
100
dispersion. Tsao [129] developed a model similar to the Krischer model which includes
the effect of different geometries of the discrete phase. His model gives:
−1
⎡ ⎤
⎢ 1 ⎥
⎢ dP1 ⎥
k eff = ∫ (5.17)
⎢0 1 ⎛ P − φ ⎞ ⎥
⎢ k1 + (k 2 − k1 )∫P1
1
e −1 2 ⎜ 1 ⎟dP1 ⎥
⎣ σ 2π ⎝ σ ⎠ ⎦
where P1 and σ are parameters that must be determined experimentally. Tsao defined P1
as the one dimensional porosity which is the fraction of the discrete phase on any line
drawn through a cubic volume of the heterogeneous material, and σ is the standard
deviation of P1.
5.3 Theoretical Bounds for Thermal Conductivity of Heterogeneous Materials
It may be deduced from the above discussion that the thermal conductivity of
heterogeneous materials depends on the spatial distribution of each phase [130]. Since
this information is typically unavailable, theoretical bounds for the thermal conductivity
are often considered. As mentioned earlier, Krischer’s model gives upper and lower
bounds when f = 0 and f = 1. Hashin and Shtrikman [130] derived more restrictive
bounds using variational theorems to give:
3φ (α − 1) k eff ⎡ 3(α − 1)(1 − φ ) ⎤

1+ ≤ ≤ α ⎢1 − ⎥ (5.18)
α + 2 − φ (α − 1) k1 ⎣ 3α − φ (α − 1) ⎦
Note that the lower bound is equivalent to the Maxwell equation. Thus, an idealized
suspension with homogeneously dispersed spheres yields the lowest thermal conductivity
of any spatial arrangement.
5.4 Accuracy of Thermal Conductivity Models
101
Turian et al. [48] compared many theoretical models for the thermal conductivity
of heterogeneous materials with available data on solid – liquid dispersions (0.16 < α <
1507). They demonstrated that when the difference between the thermal conductivities
of each phase is small (0.4 < α < 2.4), many of the thermal conductivity models
(Maxwell [118], Jeffrey [124], Bruggeman [125]) agree within 2 % with experimental
data for dilute suspensions (φ < 0.15). For these values of α, the Jeffrey and the
Bruggeman equations were accurate within 2 % for more concentrated suspensions as
well. However, as α increased, agreement with experimental data was less satisfactory.
For each dispersion considered, the Jeffrey model was more accurate than the Maxwell
equation, and the Bruggeman model was more accurate than both the Maxwell and
Bruggemann models.
5.5 Geometric Mean
Turian et al. [48] found that the volume fraction – weighted geometric mean of
the thermal conductivities of the individual phases,
k eff = k11−φ k 2φ (5.19)
provided as good or a better prediction of the thermal conductivity than any of the
theoretical models when α > 3.5. This empirical equation seems to better reflect the
thermal conductivity of real dispersions containing interacting particles with irregular
spatial arrangement as opposed to the ideal homogeneously dispersed systems considered
by Maxwell. Turian et al. investigated the volume – fraction weighted arithmetic,
geometric, and harmonic means in addition to theoretical thermal conductivity models.
The arithmetic mean and the harmonic mean are equivalent to the upper and lower
102
bounds provided by the Krischer model, and the geometric mean falls between these
bounds. Moreover, the geometric mean is within the more restrictive Hashin – Shtrikman
bounds for α > 5. In their comparison of various model predictions with experimental
data, the average deviation for 3.5 < α < 70 was 14.3 % with the Maxwell equation and
5.7 % with equation 5.19, and for 70 < α < 200 the average deviation was 26.3 % with
the Maxwell equation and 9.9 % with equation 5.19. Overall, the volume fraction –
weighted geometric mean provided the best prediction of experimentally measured
thermal conductivities of solid – liquid dispersions.
5.6 Nanofluid Thermal Conductivity Models
Keblinski et al. [14] considered several size dependent contributions to the
thermal conductivity of nanofluids and suggested that the enhancement in thermal
conductivity is due to Brownian motion of the particles, or to ordered liquid molecules at
the liquid / particle interface, or nanoparticle clustering. The Brownian motion of the
nanoparticles results in collisions between particles that can cause heat to transfer directly
from particle to particle. However, a simple kinetic analysis showed that mechanism
could be discounted. They then assumed that the liquid molecules at the surface of the
particles are more conductive because they exhibit higher order than bulk liquid
molecules. As there is greater surface area associated with smaller particles, a greater
fraction of the liquid molecules is ordered and contributes to enhanced conduction.
Keblinski et al. also considered aggregation or clustering of suspended particles and
showed that it is possible for these clusters to create paths of lower thermal resistance.
The higher specific surface area of nanoparticles promotes a greater degree of
aggregation than with a suspension of larger particles. This last theory is very similar to
103
that of Krischer [128], who developed an empirical model to describe the irregular
arrangement of suspended particles. Most nanofluid thermal conductivity models were
developed based on one or more of these mechanisms.
5.6.1 Brownian Motion
Many models attribute the enhanced thermal conductivity of nanofluids to the
effect of Brownian motion. Instead of heat transfer between individual particles, the
hypothesis is that interaction between the dynamic particles and fluid molecules enhances
conduction. As the particles move randomly through the fluid, they carry fluid molecules
with them and create a local convective effect at the microscale level, thus enhancing
thermal conduction. As an example, the Jang and Choi [131] model is based on a linear
combination of contributions from the liquid, the suspended particles, and the Brownian
motion of the particles to give:
k eff , m df
= (1 − φ ) + εαφ + C1 Re 2 Pr φ (5.20)
k1 d
where ε is a constant related to the Kapitza resistance (see Section 5.6.4), C1 is a
proportionality constant, df is the diameter of a fluid molecule, and Re and Pr are the
Reynolds the Prandtl numbers of the fluid, respectively. The Reynolds number, Re, is
defined by,
ρk B T
Re = (5.21)
3πμ 2 l f
where kB is the Boltzmann constant, lf is the mean free path of a fluid molecule, and ρ and
μ are the density and viscosity of the fluid, respectively. Their model reflects a strong
temperature dependence due to Brownian motion and a simple inverse relationship with
104
the particle diameter. As the particle diameter increases, the third term becomes
negligible and the model reduces to a linear combination of the thermal conductivity of
the fluid and the contribution from the particle. However, it has been demonstrated
repeatedly that a linear combination of the individual thermal conductivity contributions
is a poor predictor of the effective thermal conductivity in heterogeneous systems [48,
120]. Based on the Jang and Choi model, Chon et al. [50] employed the Buckingham-Pi
theorem to develop the following empirical correlation,
0.3690
k eff , m ⎛df ⎞
= 1 + 64.7φ 0.7460
⎜ ⎟ α 0.7476 Pr 0.9955 Re 1.2321 (5.22)
⎜ d ⎟
k1 ⎝ ⎠
where the Reynolds and Prandtl numbers are the same as in the Jang and Choi model.
The equation was fit to their measurements of aqueous nanofluids containing three sizes
of alumina particles. However, their correlation is of limited use, since it is based on
measurements over a limited temperature range (20 – 70 ºC) and it was fit to thermal
conductivity data for a single nanoparticle material in a single base fluid. Chon et al. did
not demonstrate any ability of their model to predict the thermal conductivity of other
nanofluids. Other models are available that are fitted to similarly limited nanofluid data
and include no consideration for the more conventional thermal conductivity models
[132-134]. However, some researchers have used conventional heterogeneous thermal
conductivity models as a starting point and extended these to include a particle size
dependence based on Brownian motion [135-139]. For instance, Xuan et al. [138]
developed an extension of the Maxwell equation to include the microconvective effect of
the dynamic particles and the heat transfer between the particles and fluid to give:
k eff , m α + 2 + 2φ (α − 1) 18φHAk B T
= + 2 6 τ (5.23)
k1 α + 2 − φ (α − 1) π ρd k1
105
where H is the overall heat transfer coefficient between the particle and the fluid, A is the
corresponding heat transfer area, and τ is the comprehensive relaxation time constant.
The heat transfer area should be proportional to the square of the diameter, thus the
effective thermal conductivity is proportional to the inverse of the particle diameter to the
fourth power. Such a strong particle size dependence has yet to be demonstrated
experimentally. Additionally, the equation reduces to the Maxwell equation with
increasing particle size. As discussed previously, thermal conductivity enhancements
greater than those predicted by the Maxwell equation have been reported for nanofluids
containing relatively large nanoparticles (d > 30 nm) [59]. It is therefore obvious that
models that reduce to the Maxwell equation at large nanoparticle sizes will not be able to
represent published data.
The Brownian motion models share a common attribute that distinguishes them
from other models. Each model exhibits a monotonically increasing relationship between
the thermal conductivity and temperature due to the temperature dependence of the
velocity of the particles. This behavior has been observed over a limited temperature
range for nanofluids between 10 ºC and 70 ºC [8, 9, 50, 52, 58]. However, all of these
studies have focused on nanofluids consisting of polar liquids, water and ethylene glycol.
The thermal conductivities of these liquids increase with temperature unlike most other
liquids. In the one study [10] which considered nanoparticles in a liquid with decreasing
thermal conductivity, perfluorohexane, the thermal conductivity of the nanofluid
decreased with increasing temperature.
106
5.6.2 Ordered Liquid Molecules
A layer of ordered liquid molecules adjacent to the surface of the solid particles
has been credited with the enhanced thermal conductivity observed in nanofluids.
Molecules in this layer exhibit greater order than bulk liquid molecules, and with greater
order, they should exhibit a greater thermal conductivity than that of the bulk liquid. Yu
and Choi [140] developed the following extension of the Maxwell equation which
incorporates a layer of ordered liquid molecules into an effective particle,
k pe + 2k1 + 2(1 + β l ) (k pe − k1 )
3
k eff , m
= , (5.24)
k pe + 2k1 − (1 + β l ) (k pe − k1 )
3
k1
where the thermal conductivity of the effective particle is defined as,
k pe =
[2(1 − γ ) + (1 + β ) (1 + 2γ )]γ k
l
3
, (5.25)
− (1 − γ ) + (1 + β l ) (1 + 2γ )
3 2
k layer
γ= k2
, (5.26)
βl = h d , (5.27)
where klayer is the thermal conductivity of the ordered liquid layer of molecules and h is
the thickness of that layer. Others [141-146] have developed similar models using the
same method of defining an effective particle which includes the solid particle and the
surrounding ordered liquid layer. These models are capable of fitting some of the
experimental data available in the literature by adjusting the thickness and thermal
conductivity of the ordered liquid layer. However, Evans et al. [147] used molecular
dynamics simulations to estimate the thermal conductivity of water molecules with
greater order and found that the thermal conductivity was greater by an order of three
when the water had crystalline order. Thus, the thermal conductivity of the ordered water
107
layer should be less than or equal to three times the thermal conductivity of bulk water.
By using this value in any of the aforementioned models, the thermal conductivity
enhancement is equivalent to the Maxwell equation except at very small values of
particle diameter (< 5 nm). Based on this data, nanofluids would be more thermally
conductive than other dispersions only if the diameter of the nanoparticles is less than 5
nm.
An inverse relationship between thermal conductivity and particle size is a
common attribute among both the Brownian motion and ordered liquid thermal
conductivity models. In certain models, this relationship is stronger (Xuan et al. [138])
than others (Chon et al. [50]). However, there is little to no experimental evidence to
support this relationship. Some studies have examined the effect of particle size on
thermal conductivity, and have found slightly greater enhancement [50, 52] or lower
enhancement [12, 58] with decreasing particle size.
5.6.3 Nanoparticle Clustering
The difference between models developed to account for nanoparticle clustering
[148-150] and models which reflect the spatial arrangement of particles is that the
nanoparticle clustering models include a particle size dependence. The smaller particle
size creates greater attraction between the individual particles, which leads to a greater
extent of aggregation. Prasher et al. [150] extended the Maxwell equation for use with a
dispersion of particle aggregates. Each aggregate consists of nanoparticles, a fraction of
which form a conductive pathway, or backbone, while the remaining nanoparticles are
considered dead ends. They calculated the thermal conductivity of the fluid and dead
ends inside the aggregate (knc) and then found the thermal conductivity of the effective
108
aggregate (ka) by including the backbone. This concept is illustrated in Figure 5.1. This
model is quite difficult to use as it requires knowledge of the aggregate size as well as the
proportion of nanoparticles forming conductive pathways through the aggregate.
The hypothesis that particle clustering enhances conduction in suspensions is well
supported by numerical simulations and molecular dynamics studies [151-153].
However, those studies did not find a significant particle size dependence.
Figure 5.1 Schematic of method by Prasher et al. [150] to determine the thermal
conductivity of an aggregate which includes the backbone of nanoparticles,
the dead end nanoparticles and the fluid surrounding the particles.
5.6.4 Interfacial Thermal Resistance
In contrast to the previously discussed phenomena, thermal resistance at the solid-
liquid interface should lead to a reduction in the thermal conductivity of the dispersion.
The interfacial thermal resistance is defined as a temperature discontinuity at the interface
and is sometimes referred to as the Kapitza resistance [154]. This empirical property
109
incorporates phonon scattering that occurs at the interface as well as any other
phenomena that create resistance to heat transport across the interface such as poor
contact between the substances. Nan et al. [155] presented several models which
incorporate the shape of the particle and the interfacial thermal resistance into the
Maxwell equation. For spheres the model is as follows,
k eff α (1 + 2 χ ) + 2 + 2φ [α (1 − χ ) − 1]
= , (5.28)
k1 α (1 + 2 χ ) + 2 − φ [α (1 − χ ) − 1]
2 R B k1
χ= , (5.29)
d
where RB is the interfacial thermal resistance. When χ << 1, equation 5.28 reduces to the
Maxwell equation. As χ increases due to smaller diameter or greater resistance, the
effective thermal conductivity of the nanofluid decreases. The work of Prasher et al.
[136] extends this model to include the effect of Brownian motion. The interfacial
thermal resistance has been measured for a few systems, carbon nanotubes with sodium
dodecyl sulfate (SDS) in D2O [156], citrate-stabilized Pt nanoparticles in water and thiol-
stabilized AuPd nanoparticles in toluene [157]. The thermal conductivity of these
systems can be estimated by using these values of the interfacial thermal resistance in
conjunction with equation 5.28. For example, based on the value for Pt in water, an
aqueous nanofluid containing 25 nm diameter Pt particles would enhance the thermal
conductivity half as much as a nanofluid containing 100 nm diameter Pt particles.
5.7 Summary of Thermal Conductivity Models for Nanofluids
Numerous thermal conductivity models have been developed for heterogeneous
systems and specifically for nanofluids. Theoretical models such as those by Maxwell
110
[118] and Bruggeman [125] were derived by assuming a homogeneous or random
arrangement of particles. However, these assumptions are not valid for dispersions
containing aggregates. Empirical models [48, 128] have been successfully employed to
account for the spatial arrangement of particles. More recently, particle size has been
incorporated into many models in an attempt to describe the thermal conductivity of
nanofluids. Several mechanisms have been described in the previous section that may
affect the thermal conductivity of nanofluids, including Brownian motion of the particles,
ordered liquid molecules at the solid / liquid interface, nanoparticle clustering, and
interfacial thermal resistance. However, there is no consensus as to which mechanism
has the dominant effect on the thermal conductivity.
111
CHAPTER 6
THERMAL CONDUCTIVITY MODELING OF SOLID / LIQUID
DISPERSIONS
There have been many attempts to model the thermal conductivity of nanofluids
and to elucidate the mechanisms for conduction in these fluids [50, 131, 136, 140, 150].
In this chapter, some of the models and mechanisms discussed in chapter 2 are evaluated
using the data presented in Chapter 4.
6.1 Evaluation of Thermal Conductivity Models
A number of thermal conductivity models have been discussed in Chapter 5.
These models have been evaluated in terms of their predictions and through a comparison
trends in the thermal conductivity data presented in Chapter 5. This assessment has been
performed through an examination of the relationships between the thermal conductivity
of nanofluids and volume fraction (φ), temperature (T), the ratio of the pure phase
thermal conductivities (α), and the particle diameter (d).
6.1.1 Particle Volume Fraction
The thermal conductivity of nanofluids in this work generally exhibited a linear
relationship with the volume fraction between 1 and 5 % as displayed in Figure 6.1. The
linear trend observed in this work has also been commonly observed in the literature [8,
12, 49, 57, 60]. Some exceptions have been reported in the case of particle volume
fractions less than 1 % (v/v) [66, 69], but such dilute nanofluids were not considered in
this work. Thus, the nonlinear relationship observed by Murshed et al. [66] displayed in
112
Figure 2.6 was not observed in this work. Throughout the range of particle volume
fractions considered in this work (1 – 5 %), all of the thermal conductivity models
discussed in chapter 5 exhibit a linear relationship with respect to particle volume
fraction (Figures 4.10 and 4.11). Figure 6.1 includes a comparison of the predicted
thermal conductivity versus volume fraction relationship for various models. At particle
volume fractions between 1 and 5 %, the models by Bruggemann, Landauer, and Jeffrey
predict thermal conductivity values within 2 % of the Maxwell equation. Thus, the
Maxwell model effectively predicts the same trends as these models and will be the only
of these models discussed in this chapter.
113
0.75
-1
-1
0.7
d = 71 nm
Maxwell
Nan et al.
Yu & Choi
Jang & Choi
Chon et al.
Prasher et al.
0.65 Geometric Mean
0.6
0 1 2 3 4 5 6
φ (%)
Figure 6.1 Thermal conductivity measurements and predictions for an aqueous

nanofluid containing alumina (diameter = 72 nm) from several models as a
function of volume fraction (φ). The error bars represent the estimated
measurement error.
114
6.1.2 Temperature
The thermal conductivity models based on Brownian motion exhibit a strong
temperature dependence while the temperature dependence of other thermal conductivity
models arises from the inherent temperature dependence of the individual phases. Figure
6.2 illustrates the difference in the predicted thermal conductivity enhancement versus
temperature relationship for several models. Note that the Brownian motion models by
Jang and Choi [131], Chon et al. [50], and Prasher et al. [136] display a monotonically
increasing thermal conductivity as the temperature rises, but the other models exhibit
thermal conductivity enhancements which are relatively constant with temperature. This
temperature dependence due to the Brownian motion of the nanoparticles can be seen in
the Stokes – Einstein relation,
k BT
D= (6.1)
3πμd
where D is the diffusivity of the particle, kB is the Boltzmann constant, T is the
temperature, μ is the viscosity, and d is the radius of the particle. The diffusivity contains
both direct and indirect temperature dependence because of the viscosity, and the
temperature dependence is greater than first order. In many of the Brownian motion
models, the thermal conductivity is directly proportional to the diffusivity or velocity of
the suspended particles. Consequently, the relationship between the thermal conductivity
of a nanofluid and temperature greater than first order. However, the experimental
results clearly demonstrate curvature in the thermal conductivity versus temperature
behavior for each nanofluid, as illustrated in Figures 5.4, 5.7 – 5.8. Each nanofluid
exhibits a maximum thermal conductivity at approximately the same temperature as the
115
maximum observed for the pure fluid. Figure 6.3 displays the thermal conductivity of the
nanofluids consisting of alumina in ethylene glycol
These results suggest that the thermal conductivity versus temperature behavior of
the nanofluid closely follows the behavior of the base fluid. Due to their strong
temperature dependence, the Brownian motion based models are incapable of fitting
these experimental results. A more appropriate thermal conductivity model would
contain no direct temperature dependence, but would be approximately proportional to
the thermal conductivity of the base fluid, similar to the Maxwell model. A comparison
between the Maxwell model and the experimental data in Figure 6.3 illustrates the
similarity between the experimental and theoretical relationship between the thermal
conductivity of nanofluids and temperature.
116
200
150
- k ) / k (%)
Maxwell
1
Nan et al.
100 Yu and Choi
Jang and Choi
1
Chon et al
Prasher et al
eff
(k
50
0
280 300 320 340 360 380 400
Temperature (K)

containing 5 % (v/v) alumina (40 nm) from several models as a function of
temperature.
117
0.3
-1
0.29
-1
0.28
EG
1%
0.27 3%
4%
0.26
0.25
280 300 320 340 360 380 400 420 440
Temperature (K)

temperature). The dashed lines represent the Maxwell equation for each
concentration. The error bars represent estimated error for the
measurements.
118
6.1.3 Thermal Conductivity of Individual Phases
As the ratio of solid to liquid thermal conductivity, α, increases, the thermal
conductivity predictions from the Maxwell equation approach a limiting value, which is
only dependent on the particle volume fraction and the thermal conductivity of the liquid.
The following equation is obtained from the Maxwell equation when α >> 1.
⎛ 3φ ⎞
k eff = k1 ⎜⎜1 + ⎟⎟ (6.2)
⎝ 1−φ ⎠
This shows that the thermal conductivity of a nanofluid does not depend on the solid
thermal conductivity when α >> 1 or k2 >> k1. (Note that the thermal conductivity
enhancement is also independent of the liquid thermal conductivity.) However, this
limiting value given by eq. 6.2 is not observed experimentally. Xie et al. [59] measured
the thermal conductivity of alumina nanoparticles in various base fluids and observed that
the relative thermal conductivity enhancement of the nanofluid was greater for less
conductive fluids (greater α). In this work, nanofluids consisting of ethylene glycol and
alumina exhibited greater thermal conductivity enhancement than the aqueous nanofluids
containing the same concentration of the same size of alumina particles.
Turian et al. [48] observed this deficiency in the Maxwell model and similar
models [124-126]. They found that deviations between thermal conductivity predictions
of the Maxwell equation and experimental data increased as α increased, and that the
volume fraction – weighted geometric mean provided more accurate predictions of the
thermal conductivity of solid – liquid dispersions when α is high (α > 70),
k eff
=αφ (6.3)
k1
119
A comparison of thermal conductivity predictions from the geometric mean and the
Maxwell equation are compared with the data from Xie et al. [59] in Figure 6.4. The
geometric mean exhibits a relationship between the thermal conductivity and α that is
similar to the empirical trend. Furthermore, the geometric mean predictions are within 6
% (within 1 % for α ≤ 150) of the experimental data while the predictions from the
Maxwell equation deviate from the experimental data by as much as 23 %.
The geometric mean is compared with the Hashin and Shtrikman (HS) bounds in
Figure 6.5. The HS bounds represent the limits of thermal conductivity throughout the
range of spatial arrangement of particles in suspension. Note that the lower HS bound is
the same as the Maxwell equation. Thus, the lower HS bound represents the thermal
conductivity of a dispersion containing homogeneously dispersed particles with no
particle interactions. Consequently, aggregation of suspended particles promotes
conduction and yields a greater thermal conductivity than that predicted by Maxwell.
The upper HS bound represents the arrangement of particles that maximizes the thermal
conductivity of a dispersion. Also the thermal conductivity of the upper HS bound
increases monotonically with increasing α, and the lower HS bound reaches a limiting
value. Thus, the thermal conductivity of dispersions is a function of α due to aggregation
of the particles and is best represented by the volume fraction – weighted geometric
mean.
120
40
Xie et al. 2002
35 Geometric Mean
Maxwell
(keff - k1) / k1 (%)
30
25
20
15
10
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
-1 -1
k1 (W m K )
Figure 6.4 The thermal conductivity of 5 % (v/v) alumina in pump oil, ethylene glycol,
glycerol, and water from Xie et al. [59]. The dashed lines represents the
volume fraction weighted geometric mean or the Maxwell equation.
121
100
Maxwel / Lower HS bound
Geometric Mean
80 Upper HS bound
(keff - k1) / k1 (%)
60
40
20
0
0 5 10 15 20 25 30
α
Figure 6.5 The Hashin and Shtrikman bounds for the thermal conductivity of a
heterogeneous material as a function of the ratio of the individual phase
thermal conductivies (α). The lower bound is equivalent to the Maxwell
model. The volume fraction in these calculations is 5 %.
122
6.1.4 Particle Diameter
The models for the thermal conductivity of nanofluids exhibit an inverse
relationship between thermal conductivity and particle diameter. In the Brownian motion
based models, the size dependence typically arises from the Brownian velocity. In the
ordered liquid layer models, the particle diameter is related to the total surface area,
which determines the volume of ordered liquid molecules. Figure 6.6 displays a
comparison of thermal conductivity predictions from various models and their
relationship with particle diameter. These types of models yield thermal conductivity
predictions that increase as the particle size decreases below a threshold. For example,
the predictions of the Yu and Choi [140] model are weakly dependent on particle size
above a diameter of 20 nm. However, below 20 nm, the thermal conductivity exhibits a
strong particle size dependence. This inverse relationship between the thermal
conductivity and the particle size has not been validated experimentally.
The experimental data in this work (Tables 4.10 – 4.12) exhibits a relationship
between the thermal conductivity of the dispersion and the particle size similar to that
exhibited by the Nan et al. [155] model. As illustrated in Figure 6.7, the thermal
conductivity of the dispersion decreases with decreasing particle size. A decrease in
thermal conductivity enhancement was also observed for decreasing particle size in
aqueous ceria nanofluids (Figure 4.11). The thermal conductivity models for nanofluids
are incapable of fitting these results due to their inverse dependence between thermal
conductivity and particle size. The model by Nan et al. is the only available model that
exhibits decreasing thermal conductivity with decreasing particle size. They attributed
this trend to thermal resistance at the interface of the solid nanoparticles and the liquid. .
123
The interfacial thermal resistance is defined as a temperature discontinuity at the
interface. Thus, it is an empirical phenomenon and can be caused by various mechanisms
such as phonon dynamics and poor contact between the individual phases. In nanofluids,
the thermal resistance is most likely phonon boundary scattering occurring at the
interface of the solid and liquid. A number of studies [38, 40, 44] have experimentally
demonstrated lower thermal conductivity of silicon nanomaterials attributed to phonon
boundary scattering. Also, the thermal conductivity of solids is lower when the solid
dimensions are confined. For example, the thermal conductivity of nanowires is less than
that for nanofilms when the diameter of the wire is comparable to the thickness of the
film as a result of confinement in two dimensions instead of one [43]. Consequently,
nanoparticles should exhibit lower thermal conductivity than either wires or films due to
confinement in three dimensions. The thermal conductivity of nanofluids consisting of
titania (d = 2 nm) in a mixture of ethylene glycol and water exhibited thermal
conductivity less than the mixture without particles (Figure 5.9). Furthermore, the
thermal conductivity decreased with increasing particle concentration. These results
suggest that the thermal conductivity of the titania is less than the thermal conductivity of
the fluid. Thus, the thermal conductivity of solid nanoparticles should be represented by
a model which is a function of the particle size rather than incorporating the particle size
dependence into the model for the dispersion.
124
100
Maxwell
80 Nan et al 1997
Yu and Choi 2003
Jang and Choi 2004
(keff - k1) / k1 (%)
60 Chon et al 2005
Prasher et al 2005
40
20
-20
0 20 40 60 80 100
Particle Diameter (nm)

containing 5 % (v/v) alumina at room temperature from several models as a
function of particle diameter.
125
20
15
(keff - k1) / k1 (%)
10 2%
3%
4%
5
0
0 50 100 150 200 250 300
Mean Particle Diameter (nm)

in visual detection of trends in the data.
126
6.1.5 Geometric Mean
All of the published models for the thermal conductivity of nanofluids represent
the enhancement versus volume fraction trend well, as shown in Figure 6.1 and discussed
in section 6.1.1. However, the relationship of the enhancement with temperature, α, and
particle size is not represented adequately by any of the models as discussed in the
previous three sections (6.1.2 – 6.1.4). Based on this analysis in the previous sections,
only two models are capable of predicting three of these four observed trends, the model
by Nan et al. and the volume fraction – weighted geometric mean.
The model by Nan et al. [155] displays trends that reflect the behavior of the
thermal conductivity as a function of particle volume fraction, temperature, and particle
diameter. However, the model reduces to the Maxwell model at large particle size and
includes the same deficiency in reflecting the relationship between the thermal
conductivity of the dispersion and the thermal conductivities of the individual phases.
Figure 6.8 displays a comparison of predictions from the Nan et al. model compared to
the Maxwell model and experimental data. The Nan et al. model is incapable of
predicting the thermal conductivities of these nanofluids containing larger particles (d >
50 nm).
The volume fraction – weighted geometric mean reflects the empirical trends
between the thermal conductivity and the particle volume fraction, temperature, and the
thermal conductivities of the individual phases. Figure 6.9 displays experimental data
and the corresponding predictions from the geometric mean. The geometric mean yields
predictions that are within 2 % of the experimental values for the dispersions containing
larger particles (d > 50 nm), but it contains no direct particle size dependence. However,
127
the thermal conductivity of the solid is dependent on the particle size in the nanoscale
range (section 3.2.2). Therefore, a particle size dependence can be incorporated into the
volume fraction – weighted geometric mean through the thermal conductivity of the solid
particles.
128
20
15
(keff - k1) / k1 (%)
10
exp
0 Nan et al.
Maxwell
-5
0 50 100 150 200 250 300
Figure 6.8 Thermal conductivity enhancement of aqueous nanofluids containing 4 %

(v/v) alumina at room temperature. The Nan et al. [155] model and the
Maxwell model [118] are provided for comparison. The thermal
conductivity of bulk alumina (40 W m-1 K-1) was used in the predictions.
129
20
15
(k - k ) / k (%)
f
10
f
5
2%
3%
4%
0
0 50 100 150 200 250 300

in visual detection of trends in the data. The dashed lines represent the
volume fraction weighted geometric mean at each concentration using the
bulk alumina thermal conductivity (40 W m-1 K-1).
130
6.2 Size Dependent Particle Thermal Conductivity
Molecular dynamics studies by Ziambaras and Hyldgaard [43] and Fang et al.
[45] predict that the thermal conductivity of nanomaterials increases linearly with particle
size at small sizes and eventually attains the bulk thermal conductivity at larger particle
sizes. It is therefore proposed here that the thermal conductivity of the particle be
represented by the following equation,
(
k 2 (d ) = k bulk 1 − e − A1d ) (6.4)
where kbulk is the thermal conductivity of the bulk material (40 W m-1 K-1 for alumina
[19]) and A is an empirical parameter. To estimate this parameter, the thermal
conductivity of each size of alumina particles was estimated by performing a least
squares fit of the geometric mean to the experimental data of the aqueous nanofluids
(Table 5.10) as displayed in Figure 6.10. The semi-empirical fit and the estimated
thermal conductivities of alumina particles versus average particle diameter are displayed
in Figure 6.11. The error bars represent the experimental error in the thermal
conductivity measurement of the fluid. The following equation was found with the least
squares method to find the fitted parameter,
⎛
k 2 (d ) = ⎜ 40
W ⎞
⎝ m⋅K ⎠
⎟1− e (
− (0.0126 nm −1 )d
). (6.5)
131
18
16
14
(keff - k1) / k1 (%)
12
10
12 nm
8 46 nm
245 nm
6
2
1.5 2 2.5 3 3.5 4 4.5
φ (%)

alumina at room temperature for various mean particle diameters. The lines
represent least squares fits of the volume fraction weighted geometric mean
by adjusting the thermal conductivity of the particle (k2).
132
80
-1
60
-1
40
20
-20
0 50 100 150 200 250 300
Figure 6.11 Estimated thermal conductivity of alumina particles from the volume
fraction weighted geometric mean. The curve represents a semi-empirical fit
(eq. 6.5).
133
6.3 Prediction of Thermal Conductivity of Nanofluids
Equation 6.5 and the geometric mean were used to predict the thermal
conductivity of alumina in ethylene glycol as a function of temperature and particle size.
Figures 6.12 and 6.13 display comparisons of these predictions of thermal conductivity
enhancement with the experimental data from this study. The predictions are all within
1.5 % of the experimental data in Figure 6.12 and within 2.3 % in Figure 6.13. Figure
6.14 displays a comparison between the predictions of the model and experimental data
as a function of α from this work and that of Xie et al. [59]. The predictions were within
4 % of the data of Xie et al. and within 2 % of the data from this work. Additionally,
these equations provided predictions within 4 % of the experimental data from Xie et al.
[12] (Figure 6.15) and within 1 % of Eastman et al. [60] (Figure 6.16). Furthermore,
these equations provided predictions within 4.8, 4.5, 2.2, and 1.8 % of the measured
values of Lee et al. [57], Wang et al. [53], Kim et al. [13], and Yoo et al. [55],
respectively. However, some experimental thermal conductivity data deviates from
predictions of this model due to a temperature dependence which conflicts with the
results of this work [8, 9, 50, 52]. The reasons for other discrepancies between the
measured thermal conductivity and these predictions are unknown [49, 51, 58].
The dearth of thermal conductivity data for nanofluids containing different
particle sizes makes it difficult to apply this model to other systems. Thermal
conductivity data is available for CuO nanofluids containing particles between 23 and 36
nm [11, 18, 53, 57, 60, 65], but this size range is too narrow to obtain a reliable
correlation between thermal conductivity and particle size. However, the model is
capable of fitting data available in the literature for CuO nanofluids because the measured
134
values are less than the maximum value of the model given by the geometric mean with
the thermal conductivity of the bulk solid. Alternatively, the model is incapable of
predicting the measured thermal conductivity values for nanofluids containing more
conductive particles such as Cu [25, 49, 60, 71-74], diamond [32], Ag [32, 77], Au [73]
or alloys [81-83]. Specifically, the maximum predicted thermal conductivity
enhancement for a nanofluid consisting of 0.3 % (v/v) Cu in ethylene glycol is 2.1 %,
whereas the measured value was 40 % [74]. In addition, the maximum predicted thermal
conductivity enhancement for a nanofluid consisting of 1.3 % (v/v) diamond particles in
ethylene glycol is 11.2 %, and the measured value was 75 % [32]. Further comments
about the available thermal conductivity studies in relation to this new model are
provided in Table 6.1. Studies of nanofluids containing particles which were not
spherical or nearly spherical were not considered, since the effect of particle shape has
not been incorporated into the model [10, 24, 27, 75, 84-93].
Thus, the volume fraction – weighted geometric mean with a single parameter
semi-empirical equation for the solid thermal conductivity is capable of predicting the
thermal conductivity of nanofluids containing alumina, but it unproven with other
systems. Furthermore, the model is not appropriate for predicting the thermal
conductivity of nanofluids containing highly thermally conductive particles such as
copper, silver, or diamond.
135
18
16
14
(k - k ) / k (%)
f
12
f
10
6
0 50 100 150 200 250 300
Figure 6.12 Thermal conductivity enhancement of nanofluids consisting of 2 and 3 %

(v/v) alumina in ethylene glycol at room temperature. The circles and
squares represent experimental data at 2 % and 3 % respectively. The
dashed lines represent predictions using Equation 6.5 with the volume
fraction weighted geometric mean.
136
0.3
-1
0.29
-1
0.28
EG
1%
0.27
3%
4%
0.26
0.25
280 300 320 340 360 380 400 420 440
Temperature (K)

temperature). The dashed lines represent predictions using Equation 6.5 with
the volume fraction weighted geometric mean.
137
30
25
(k - k ) / k (%)
20
f
5 %, 60 nm
Xie et al.
f
15 3 %, 16 nm
3 %, 282 nm
10
5
60 80 100 120 140 160
αbulk
Figure 6.14 Thermal conductivity of alumina nanofluids as a function of αbulk from this
work and the work of Xie et al. [59]. The thermal conductivity of the bulk
solid was used to determine αbulk. The dashed lines represent predictions
using Equation 6.5 with the volume fraction weighted geometric mean.
138
30
25
(k - k ) / k (%)
20
f
1.8%
3.5%
f
15 5%
10
5
0 50 100 150 200 250 300 350
Figure 6.15 Thermal conductivity enhancement of nanofluids consisting of alumina in

ethylene glycol at room temperature from Xie et al. [12]. The dashed lines
represent predictions using Equation 6.5 with the volume fraction weighted
geometric mean.
139
18
16
14
(keff - k1) / k1 (%)
12
10
2
0 1 2 3 4 5 6
φ (%)

alumina (diameter = 35 nm) from Eastman et al. [60]. The dashed line
represents a prediction using Equation 6.5 with the volume fraction weighted
geometric mean.
140
Table 6.1 Deviations between thermal conductivity model predictions and experimental
values available in the literature
Authors and Nanofluids Maximum Comments

Reference deviation of
data from
model*
Eastman et al. Al2O3 / Water 10 % Data has not been
1997 [49] CuO / Water 24 % reproduced by others
Eastman et al. Al2O3 / Water 1%
1999 [60] CuO / Water 5%
Cu / Ethylene Glycol Maximum > geometric mean
Lee et al. 1999 Al2O3 / Water 5%
[57] Al2O3 / Ethylene Glycol 4%
CuO / Water 5.4 %
CuO / Ethylene Glycol 2%
Wang et al. 1999 Al2O3 / Water 2.3 %
[53] Al2O3 / Ethylene Glycol 7.5 %
Al2O3 / Engine Oil 8%
Al2O3 / Pump Fluid 16 %
CuO / Water 13 %
CuO / Ethylene Glycol 23 %
Xie et al. 2002 [12, Al2O3 / Water 4.4 %
59] Al2O3 / Ethylene Glycol 4%
Al2O3 / Pump Oil 7.3 %
Xie et al. 2002 SiC / Water 22 %
[70] SiC / Ethylene Glycol 28 %
Das et al. 2003 [8] Al2O3 / Water Temperature dependence of data is
CuO / Water apparently erroneous
Chon et al. 2005 Al2O3 / Water Temperature dependence of data is
[50] apparently erroneous
Kwak and Kim CuO / Ethylene Glycol 1.2 %
2005 [64]
Murshed et al. TiO2 / Water 20 % > maximum model
2005 predictions
Hwang et al. 2006, CuO / Water 1.3 %
2007 [18, 62] CuO / Ethylene Glycol 4.1 %
SiO2 / Water 4.3 %
CNT / Water NA not spherical
CNT / Mineral Oil NA not spherical
Lee et al. 2006 CuO / Water 10 %
[63]
Li and Peterson Al2O3 / Water Temperature and particle size
2006, 2007 [9] CuO / Water dependence are apparently erroneous
Liu et al. 2006 [65] CuO / Ethylene Glycol 2.5 %
141
Table 6.1 (continued)
Zhang et al. 2006, Al2O3 / Water 7% when φ < 6 %

2007 [11, 51, 54] 24 % when φ > 6 %
CuO / Water 4%
TiO2 / Water 2%
CNT / Water NA not spherical
Au / Toluene NA no enhancement
(φ = 0.003 %)
Kim et al. 2007 Al2O3 / Water 2%
[13] Al2O3 / Ethylene Glycol 2.2 %
ZnO / Water Particle size dependence of data are
ZnO / Ethylene Glycol apparently erroneous
TiO2 / Water
TiO2 / Ethylene Glycol
Yoo et al. 2007 Al2O3 / Water 2%
[55, 56] TiO2 / Water 12 %
WO3 / Ethylene Glycol NA
Fe / Ethylene Glycol 14 %
Timofeeva et al. Al2O3 / Water Particle size dependence of data are
2007 [58] Al2O3 / Ethylene Glycol apparently erroneous
Wang et al. 2007 TiO2 / Water 6.1 %
[67] SiO2 / Water 3.3 %
SiO2 / Ethylene Glycol 3.4 %
SiO2 / Ethanol 3.7 %
* Maximum deviation of data from model is defined as (k measured − k predicted ) k predicted
( )
Model predictions were obtained using k 2 (d ) = k bulk 1 − e − (0.0126 nm )d .
−1
Note that the exponent is the same as that for alumina.
142
6.4 Particle Size Polydispersity
The relationship between the thermal conductivity of the solid and particle size
raises a concern about the use of mean particle size to characterize the samples. Some of
the particle samples in this study have more polydisperse distributions than others.
Specifically, the alumina nanoparticles with an average size of 46 nm and 71 nm have a
much greater standard deviation than the other samples as observed in the TEM images
(see Table 4.3). The range of particle diameters observed in TEM images was 14 – 260
nm for the 71 nm average diameter particles (Figure 6.17). Similarly, the range was 8 –
370 nm for the 46 nm average diameter particles (Figures 4.2 and 6.18). When studying
the thermal conductivity of the solid particles, the primary size distribution would be a
more appropriate characterization than the average particle size. However, the size
distribution is quite difficult to measure in aggregated particles. The use of average
particle size to characterize polydisperse particle samples is especially poor when the
thermal conductivity versus particle size relationship is not linear over the entire size
distribution. Thus, the polydisperse size distribution creates some uncertainty when
examining the relationship between size and thermal conductivity. Consequently, the
experimental data points representing nanofluids containing polydisperse particles (d =
46 nm and 71 nm) display the largest deviation from the model.
A possible method to incorporate the particle size distribution into the model
would be to treat each size of particles separately. For example, for a binodal distribution
with particle sizes dA and dB, the geometric mean would be written,
φA φB
k eff ⎛ k (d ) ⎞ ⎛ k 2 (d B ) ⎞
= ⎜⎜ 2 A ⎟⎟ ⎜⎜ ⎟⎟ (6.6)
k1 ⎝ k 1 ⎠ ⎝ k1 ⎠
143
where φA and φB are the volume fractions of each size of particle, and the sum of φA and
φB is the volume fraction of particles in the dispersion. Thus, for a distribution of n sizes
of particles, the model would be,
⎛ k (d ) ⎞
φi
n
∏i ⎜⎜ 2k i ⎟⎟ (6.7)
⎝ 1 ⎠
144
respectively. Magnification = 20,000
145
size of 47 nm from Nanophase Technologies The average diameter of these
particles is 77 nm and 46 nm as determined by TEM and BET, respectively.
Magnification = 30,000
146
CHAPTER 7
CONCLUSIONS AND RECOMMENDATIONS
7.1 Conclusions
The thermal conductivities of dispersions containing alumina in water, alumina in
ethylene glycol, alumina in ethylene glycol + water, and ceria in water were measured
with the transient hot wire apparatus at temperatures ranging from 296 K to 422 K and
average particle diameters ranging from 8 nm to 282 nm. To date, these measurements
cover the largest number of particle sizes in a single study and the widest temperature
range considered for any nanofluid.
The effects of temperature and particle size on the thermal conductivity of these
nanofluids have been discovered. The temperature dependence arises mostly from the
temperature dependence of the thermal conductivity of the liquid. Thus, a maximum was
observed in the thermal conductivity versus temperature relationship of nanofluids with
water and ethylene glycol as the base fluids. This maximum occurred at approximately
the same temperature as the maximum for the base liquid. The thermal conductivity of
the nanofluids with smaller particles (d < 50 nm) exhibited a lower thermal conductivity
than the dispersions containing larger particles.
A semi-empirical predictive model containing a single adjustable parameter was
developed based on the experimental thermal conductivity data from aqueous nanofluids
containing alumina. The model consists of the volume fraction – weighted geometric
mean with a size dependent solid thermal conductivity. Turian et al. [48] demonstrated
that the geometric mean provides the most accurate predictions compared with other
147
models for a wide range of solid-liquid systems. The size dependence of solid thermal
conductivity was demonstrated previously for ultrathin films [40] and nanowires [44]. In
the present work, the effect of this phenomenon was observed in measurements of
nanofluids containing different sizes of nanoparticles. Nanofluids containing smaller
nanoparticles (< 50 nm) exhibited an increasing thermal conductivity with increasing
particle diameter, and as the particle size increased the thermal conductivity reached a
plateau. This plateau is equivalent to the thermal conductivity predictions from the
volume fraction – weighted geometric mean using the bulk thermal conductivity of
alumina. Thus, the thermal conductivities of each size of alumina particles was estimated
from the nanofluid thermal conductivity data by using the geometric mean, and a semi-
empirical equation with a single adjustable parameter was fit to this data. The new
predictive model yielded values that were within 5 % for several sets of experimental
thermal conductivity data for nanofluids containing alumina.
The model presented here is the only available model that reflects observed
relationships between the thermal conductivity of nanofluids and the volume fraction (φ),
temperature (T), particle diameter (d), and the ratio of the individual phase thermal
conductivies (α). Specifically, the model is approximately linear at low volume fractions
(φ < 5 %), which has been demonstrated throughout the literature and in these
experiments. Additionally, this model exhibits a temperature dependence similar to the
base fluid. This behavior was demonstrated in the measurements of the nanofluids
containing ethylene glycol and alumina, where both the nanofluids and pure ethylene
glycol exhibited maximum thermal conductivity at approximately the same temperature.
Additionally, the decrease in thermal conductivity of the nanoparticles with particle size
148
predicted by Fang et al. [45] was observed in the thermal conductivity measurements for
nanofluids containing alumina and for those containing ceria. Lastly, Xie et al. [59] and
Turian et al. [48] observed increasing thermal conductivity enhancement as α increased,
and this relationship was also evident in this work for the nanofluids containing the same
size of particles, but different base fluids (water and ethylene glycol).
7.2 Future Work
Future work should focus on applying the model presented here to other systems.
The semi-empirical equation for the particle thermal conductivity was developed
specifically for alumina. The available thermal conductivity data for other nanofluid
systems are not as comprehensive as for nanofluids containing alumina, and the
relationship between the thermal conductivity of nanofluids and the particle size must be
determined for those systems. With this information, one could estimate the adjustable
parameter in the particle thermal conductivity equation and try to find a relationship
between that parameter and certain solid properties, such as the phonon mean free path.
For metal oxides and other ceramics, the difficulty in these experiments would be in
obtaining particles of various sizes. However, certain nanofluids have exhibited thermal
conductivity enhancement beyond the predictions of this model. These nanofluids often
consist of highly thermally conductive solid particles such as diamond [32], carbon
nanotubes [7], and metals [74]. There are no studies that focus on the effects of
parameters such as temperature and particle size for these systems. A further difficulty is
that these systems often require a dispersant to create a stable nanofluid, which makes
these ternary systems. Thus, the model would have to be extended to include the effect
of the dispersant.
149
Other work could focus on the spatial arrangement of the particles in the
nanofluids. The volume fraction weighted geometric mean represents a spatial
arrangement where clusters of particles form small networks of low thermal resistance
whereas the Maxwell equation is a theoretical solution for a dispersion where there is no
particle interaction [48]. Such an experiment would require fine control over the degree
of aggregation of the suspended particles, possibly by adjusting pH and/or the ionic
strength. Aggregation appears to enhance conduction in these dispersions to some extent,
but extensive aggregation would lead to particle settling. Thus, there may be an optimal
degree of aggregation that yields the greatest thermal conductivity enhancement. This
experiment may provide greater insight into the specific effect aggregation has on the
thermal conductivity of the dispersion.
150
APPENDIX A
CALIBRATION OF THERMOCOUPLE
A type E thermocouple was used to measure the initial temperature of the fluid
before each thermal conductivity measurement. A second junction was held at 0 C in a
ice bath during measurements. To calibrate the thermocouple between 25 and 50 ºC, the
thermocouple and a RTD probe (ASL Inc. B463712) were placed in a recirculating heater
(Thermo NESLAB RTE 740). At each temperature, the voltage of the thermocouple was
recorded from a Fluke 8840A multimeter, and the temperature was recorded from an
Omega DP251 precision digital thermometer. Above 50 ºC, the thermocouple was
calibrated with the same RTD probe in an Omega hot point® dry block probe calibrator.
Table A.1 Calibration of type E thermocouple
Temperature (ºC) Voltage (mV)

24.99 1.483
30.00 1.792
34.95 2.097
39.93 2.407
44.94 2.722
49.95 3.041
70.06 4.260
90.16 5.56
110.23 6.898
130.12 8.2215
151
140
T = 2.2861 + 15.656V R= 0.99987

120
Temperature ( C)
100
o
80
60
40
20
1 2 3 4 5 6 7 8 9
V (mV)
Figure A.1 Calibration of type E thermocouple
152
APPENDIX B
DATA ACQUISITION AND ANALYSIS PROGRAMS
B.1 LabVIEW Code
Figure B.1 LabVIEW Code: Case 0 of Nested Sequence Structure
153
154
155
156
157
158
159
160
Figure B.8 LabVIEW Code: Case 0 of Sequence Structure
161
Figure B.9 LabVIEW Code: Case 1 of Sequence Structure
162
B.2 MATLAB Code
format long e
%Wire Geometric Parameters

L = 9.9; %Effective Length of capillary (cm)
ri = 50*10^-6; %Effective Inner Radius of capillary (meters)
ro = 0.240*10^-3; %Effective Outer Radius of capillary (meters)
final = Nsample*period;
time0 = linspace(period, final, Nsample)';
%Resistivity of mercury (ohms*cm)

resistivity = 78.69943*10^-6 + 2.9613798*10^-8*T + 9.771140*10^-11*T^2;
%Temperature coefficient of resistivity for mercury (1/K)

res = (2.9613798*10^-8 + 1.954228*10^-10*T) / resistivity;
%Transport properties of fluid
if (fluid == 1) %Dimethyl Phthalate

kW = -2.393E-07 * T^2 + 0.000026236 * T + 0.16083
rhoW = 396.30 * (1 + 3.71131*(1 - T/765)^(1/3) - 5.02261*(1 - T/765)^(2/3) +
4.03136*(1 - T/765));
CpW = 8.314/0.19419 * (32.6009-0.604958*10^-2*T+0.736242*10^-4*T^2);
elseif (fluid == 2) %Ethylene Glycol
kW = -4.8687E-07*T^2 + 3.7311E-04*T + 1.8625E-01
rhoW = 62.068*(1.3151 / 0.25125^(1+ (1 - T/719.7)^0.2187)); %from Table
2-30 of Perry's 7th ed.
CpW = 6.4884E-04*T^2 + 3.8644E+00*T + 1.0450E+03;
elseif (fluid == 3) %Toluene
kW = 0.1311*(1.68182-0.682022*T/298.15)
rhoW = 92.141*(0.8488 / 0.26655^(1+(1-T/591.8)^0.2878)); %from Table
CpW = 0.010606*T^2 - 3.5303*T + 1816.9; %from fit of data taken from
NIST Chemistry WebBook
elseif (fluid == 4) %Methanol
kW = 9.790E-07*T^2 - 9.309E-04*T + 3.909E-01 %from fit from Table
rhoW = 32.042*(2.288 / 0.2685^(1+ (1 - T/512.64)^0.2453)); %from Table 2-30 of
Perry's 7th ed.
CpW = (1.0580e5 - 3.6223e2*T+.93790*T^2)/32.042; %from Table 2-196 of
Perry's 7th ed.
elseif (fluid == 5) %Hexane
kW = -2.972E-04*T + 0.2143 %from fit of data taken from NIST
Chemistry WebBook
163
rhoW = 86.177 * 0.70824 / 0.26411^(1 + (1 - T/507.6)^0.27537); %from Table
CpW = (434.6*T + 63741) / 86.18; %from fit of data taken from NIST
Chemistry WebBook
elseif (fluid == 6) %PDMS
kW = 0.155 %from Manufacturer (3M)
rhoW = 960; %from Manufacturer (3M)
CpW = 1460; %from Manufacturer (3M)
else %Water
kW = 1.8822E-08*T^3 - 2.7872E-05*T^2 + 1.3274E-02*T - 1.3710 % from fit of
data from IAPWS between 275 and 420 K
rhoW = -0.0025803*T^2 + 1.2333*T + 857.81; % from fit of data from
IAPWS between 275 and 420 K
CpW = 0.013564*T^2 - 8.7637*T + 5594.7; % from fit of data from
IAPWS between 275 and 420 K
end
rhof = rhoW;
%Transport Properties of Mercury

CpHg = 152.039 - 0.0598907*T + 5.34676*10^-5*T^2; %Heat Capacity for 273 K <
T < 473 K (CRC Handbook of Chemistry and Physics, 71st ed., 1990-1991, p. 6-102)
rhoHg = 14291 - 2.6325*T +3.0946*10^-4*T^2; %Density for 273 K < T < 473 K
(CRC Handbook of Chemistry and Physics, 71st ed., 1990-1991, p. 15-18)
kHg = 1.1097794 + 3.064102*10^-2*T - 2.195573*10^-5*T^2; %Thermal
Conductivity from Bleazard Thesis
alphaHg = kHg / (CpHg * rhoHg); %Thermal Diffusivity at T
%Transport Properties of Pyrex Insulating Layer

CpIns = -0.001927*T^2 + 2.98*T + 14.72; %Heat Capacity (J /kg /K) from
Thermophysical Properties of Matter, Vol. 5, 1970
rhoIns = 2230; %Density (kg /m3) from Corning website (Pyrex brand 7740
Borosilicate type 1 glass)
kIns = 0.81039 + 1.0914*10^-3*T; %Thermal Conductivity from Bleazard
Thesis
alphaIns = kIns / (CpIns * rhoIns); %Thermal Diffusivity at T
%Transport Properties of Gamma - Alumina

CpAl = (108.683 + 37.2263*(T/1000) - 14.2065*(T/1000)^2 + 2.193601*(T/1000)^3 -
3.20988*(1000/T)^2) / 0.10196; %Heat Capacity from NIST (J /kg /K)
rhoAl = 3700; %from Manufacturer
%Transport Properties of Nanofluid

CpW = w*CpAl + (1-w)*CpW;
rhoW = 1/(w/rhoAl + (1-w)/rhoW);
164
eEuler = exp(0.577215665); %exponent of Euler's constant
R1 = 100; %Resistance of fixed resistor 1 (ohms)

R2 = 100; %Resistance of fixed resistor 2 (ohms)
X = R2 / (R1 + R2);
Y = length(find(E0 > 0.05));
Z = Nsample + 1 - Y;
e = e0(Z:Nsample);
E = E0(Z:Nsample);
Eavg = mean(E(2:Y));
time = time0(1:Y);
Rwire = (Rdecade)*(e + Eavg*X) ./ (Eavg - Eavg*X - e); %Resistance of
capillary (ohms)
aa = polyfit(time(2:4)', Rwire(2:4), 1);
Rinitial = aa(2); %Initial Wire Resistance
DTw = (Rwire - Rinitial) ./ (res*Rinitial); %Temperature change of wire (K)
T1 = T + DTw;
resistivity1 = 78.69943*10^-6 + 2.9613798*10^-8*T1 + 9.771140*10^-11*T1.^2;
res1 = (2.9613798*10^-8 + 1.954228*10^-10*T1) ./ resistivity1;
DTw = (Rwire - Rinitial) ./ (res1*Rinitial);
start =880;
stop = start + 1500;
Emean = mean(E(start:stop));
Rmean = mean(Rwire(start:stop));
timelin = time(start:stop)';
lntime = log(timelin); %Linear part of curve
DTwlin = DTw(start:stop);
fit0 = polyfit(lntime, DTwlin, 1); %Linear fit of Temp vs ln Time
slope = fit0(1);
DTfit = slope * lntime + fit0(2);
qmean = Emean^2*Rmean / ( L * (Rdecade + Rmean)^2)*100; %Power

dissipated (W /m)
kW = qmean / (4 * pi * slope)
kW0 = kW;
gv = [2.40482, 5.52008, 8.65373, 11.79153, 14.93092, 18.07106, 21.21164, 24.35247,

27.49348, 30.63461]';
index = linspace(11, 100, 90)';
CN = index * pi - pi/4;
165
gv(11:100) = CN + 1 ./ (8*CN) - 31 ./ (394*CN.^3) + 3779 ./ (15360*CN.^5); %
roots of the zero-order Bessel function of the first kind
Yo = bessely(0,gv); %zero-order Bessel function of the second kind
b = 0.0219;
gv2 = gv.^2;
[dim2, dim] = size(timelin);
p = 1;
count = 0;
while p > 0.01
count = count +1;
alphaW = kW / (CpW * rhoW); %Thermal Diffusivity at T
%Finite physical properties of the wire and correction for insulating layer
C1 = ri^2 / 8 * ((kW - kIns) / kHg * (1/alphaHg - 1/alphaIns) + 4 / alphaIns - 2 /
alphaHg);
C2 = ro^2 / 2 * (1/alphaW - 1/alphaIns) + ri^2 / kIns * (kIns / alphaIns - kHg /
alphaHg) * log(ro / ri);
C3 = ri^2 / (2 * kW) * (kIns / alphaIns - kHg / alphaHg) + ro^2 / (2 * kW) * (kW /
alphaW - kIns / alphaIns);
dT1 = -qmean / (4 * pi * kW) * (2 * log(ri / ro) + 2 * kW / kIns * log(ro / ri) + kW / (2

* kHg) + ( C1 + C2 + C3 * log(4 * alphaW * timelin / (ro^2 * eEuler))) ./timelin);
AA = exp((-alphaW/b^2)*gv2*timelin);
BB = (pi*Yo*ones(1,dim)).^2;
dT2 = qmean / (4 * pi *kW) * (log (4 * alphaW * timelin / (b^2 * eEuler)) + sum (AA
.* BB));
DTw12 = DTwlin + dT1 + dT2;

fit2 = polyfit(lntime, DTw12, 1); %Linear fit of Corrected Temp vs ln Time
M2 = fit2(1);
B2 = fit2(2);
ebm = exp(B2 / M2);
B = fminsearch(@(B) LLSQ(B, DTw12, timelin, M2, ebm, eEuler), -0.1);
dT3 = - M2 * B * (log (ebm * timelin) / (ebm * eEuler) + 1 / (ebm * eEuler) - timelin);
DTid = DTw12 + dT3;

fitf = polyfit(lntime, DTid, 1);
Mf = fitf(1);
bf = fitf(2);
kWnew = qmean / (4 * pi * Mf);
166
p = abs(kWnew - kW)/kW*100;
kW = kWnew
end
alphaW = kW / (CpW * rhoW); %Thermal Diffusivity at T
DTw1 = DTwlin + dT1;

fit1 = polyfit(lntime, DTw1, 1);
M1 = fit1(1);
kW1 = qmean / (4 * pi * M1);
error1 = (kW1 - kW0) / kW1 * 100;
kW2 = qmean / (4 * pi * M2);

error2 = (kW2 - kW1) / kW2 * 100;
error3 = (kW - kW2) / kW * 100;
phi = w*rhof/(w*rhof+(1-w)*rhoAl)*100
P1 = ri^2/4 * (1/alphaIns - 1/(2*alphaHg)) + ro^2/4 * (2 / alphaW - 1 / alphaIns);

P2 = ri^2/(2*kIns) * (kIns / alphaIns - kHg / alphaHg) * log(ro/ri);
dTf = qmean / (4*pi*kW) * (log(4*alphaW*timelin/(ro^2*eEuler)) + (P1 + P2 + C3 *
log(4*alphaW*timelin/(ro^2*eEuler))) ./timelin);
Tf = T + mean(dTf);
if (fluid == 1) %Dimethyl Phthalate

kWf = -2.393E-07 * Tf^2 + 0.000026236 * Tf + 0.16083;
elseif (fluid == 2) %Ethylene Glycol
kWf = -4.8687E-07*Tf^2 + 3.7311E-04*Tf + 1.8625E-01;
elseif (fluid == 3) %Toluene
kWf = 0.1311*(1.68182-0.682022*Tf/298.15);
elseif (fluid == 4) %Methanol
kWf = 9.790E-07*Tf^2 - 9.309E-04*Tf + 3.909E-01; %from fit from
Table 2-370 of Perry's 7th ed.
elseif (fluid == 5) %Hexane
kWf = -2.972E-04*Tf + 0.2143; %from fit of data taken from
NIST Chemistry WebBook
elseif (fluid == 6) %PDMS
kWf = 0.155;
else %Water
kWf = 1.8822E-08*Tf^3 - 2.7872E-05*Tf^2 + 1.3274E-02*Tf - 1.3710 % from fit of
data from IAPWS between 275 and 420 K
end
props = [kW rhoW CpW];
167
DTwcalc = qmean / (4 * pi * kW) * (log (4 * alphaW * timelin / (ro^2 * eEuler)) + 2 *
kW / kIns * log (ro / ri) + kW / (2 * kHg) + ( C1 + C2 + C3 * log(4 * alphaW * timelin /
(ro^2 * eEuler))) ./timelin);
168
APPENDIX C
IMAGING OF NANOPARTICLES
169
170
BET, respectively Magnification = 10,000
171
respectively. Magnification = 15,000
172
REFERENCES
1. Bird, R.B., W.E. Stewart, and E.N. Lightfoot, Transport Phenomena. 2nd ed.
2002, New York: John Wiley & Sons, Inc. 895.
2. Perry, R.H. and D.W. Green, eds. Perry's Chemical Engineers' Handbook. 7th ed.
1997, McGraw-Hill, Inc.: New York.
3. Choi, S.U.S., Enhancing thermal conductivity of fluids with nanoparticles. FED,

1995. 231(Developments and Applications of Non-Newtonian Flows): p. 99-105.
4. Xuan, Y.M. and Q. Li, Investigation on convective heat transfer and flow features
of nanofluids. Journal of Heat Transfer-Transactions of the Asme, 2003. 125(1):
p. 151-155.
5. Wen, D.S. and Y.L. Ding, Experimental investigation into the pool boiling heat
transfer of aqueous based gamma-alumina nanofluids. Journal of Nanoparticle
Research, 2005. 7(2): p. 265-274.
6. Wang, X.Q. and A.S. Mujumdar, Heat transfer characteristics of nanofluids: a

review. International Journal of Thermal Sciences, 2007. 46(1): p. 1-19.
7. Choi, S.U.S., Z.G. Zhang, W. Yu, F.E. Lockwood, and E.A. Grulke, Anomalous
thermal conductivity enhancement in nanotube suspensions. Applied Physics
Letters, 2001. 79(14): p. 2252-2254.
8. Das, S.K., N. Putra, P. Thiesen, and W. Roetzel, Temperature dependence of

thermal conductivity enhancement for nanofluids. Journal of Heat Transfer-
Transactions of the Asme, 2003. 125(4): p. 567-574.
9. Li, C.H. and G.P. Peterson, Experimental investigation of temperature and

volume fraction variations on the effective thermal conductivity of nanoparticle
suspensions (nanofluids). Journal of Applied Physics, 2006. 99(8).
10. Yang, B. and Z.H. Han, Temperature-dependent thermal conductivity of nanorod-

based nanofluids. Applied Physics Letters, 2006. 89(8).
173
11. Zhang, X., H. Gu, and M. Fujii, Effective thermal conductivity and thermal
diffusivity of nanofluids containing spherical and cylindrical nanoparticles.
Experimental Thermal and Fluid Science, 2007. 31(6): p. 593-599.
12. Xie, H.Q., J.C. Wang, T.G. Xi, Y. Liu, F. Ai, and Q.R. Wu, Thermal conductivity
enhancement of suspensions containing nanosized alumina particles. Journal of
Applied Physics, 2002. 91(7): p. 4568-4572.
13. Kim, S.H., S.R. Choi, and D. Kim, Thermal conductivity of metal-oxide
nanofluids: Particle size dependence and effect of laser irradiation. Journal of
Heat Transfer-Transactions of the ASME, 2007. 129(3): p. 298-307.
14. Keblinski, P., S.R. Phillpot, S.U.S. Choi, and J.A. Eastman, Mechanisms of heat
flow in suspensions of nano-sized particles (nanofluids). International Journal of
Heat and Mass Transfer, 2002. 45(4): p. 855-863.
15. Tsederberg, N.V., Thermal Conductivity of Gases and Liquids. 1965, Cambridge,
Massachusetts: THE M.I.T. Press. 246.
16. Poling, B.E., J.M. Prausnitz, and J.P. O'Connell, The Properties of Gases and
Liquids. 5th ed. 2001, Boston: The McGraw-Hill Companies.
17. Dean, J.A., Lange's Handbook of Chemistry. 14th ed. 1992, New York: McGraw-
Hill, Inc.
18. Hwang, Y., J.K. Lee, C.H. Lee, Y.M. Jung, S.I. Cheong, C.G. Lee, B.C. Ku, and
S.P. Jang, Stability and thermal conductivity characteristics of nanofluids.
Thermochimica Acta, 2007. 455(1-2): p. 70-74.
19. Shackelford, J.F. and W. Alexander, eds. CRC Materials Science and Engineering
Handbook. 3rd ed. 2001, CRC Press: Boca Raton, FL. 1949.
20. Yu, C.H., S. Saha, J.H. Zhou, L. Shi, A.M. Cassell, B.A. Cruden, Q. Ngo, and J.
Li, Thermal contact resistance and thermal conductivity of a carbon nanofiber.
Journal of Heat Transfer-Transactions of the Asme, 2006. 128(3): p. 234-239.
21. Slack, G.A., THERMAL CONDUCTIVITY OF MGO, AL2O3, MGAL2O4, AND

FE3O4 CRYSTALS FROM 3 DEGREES TO 300 DEGREES K. Physical Review,
1962. 126(2): p. 427-&.
174
22. Zhitinskaya, M.K., S.A. Nemov, T.E. Svechnikova, L.N. Luk'yanova, P.P.
Konstantinov, and V.A. Kutasov, Thermal conductivity of Bi2Te3 : Sn and the
effect of codoping by Pb and I atoms. Physics of the Solid State, 2003. 45(7): p.
1251-1253.
23. Meyer, C.A., ed. ASME Steam Tables: Thermodynamic and Transport Properties
of Steam. 6th ed. 1993, American Society of Mechanical Engineers: New York.
436.
24. Xie, H.Q., H. Lee, W. Youn, and M. Choi, Nanofluids containing multiwalled
carbon nanotubes and their enhanced thermal conductivities. Journal of Applied
Physics, 2003. 94(8): p. 4967-4971.
25. Li, Q. and Y.M. Xuan, Enhanced heat transfer behaviors of new heat carrier for
spacecraft thermal management. Journal of Spacecraft and Rockets, 2006. 43(3):
p. 687-690.
26. NIST Reference Fluid Thermodynamic and Transport Properties Database

version 7. 2005.
27. Shaikh, S., K. Lafdi, and R. Ponnappan, Thermal conductivity improvement in

carbon nanoparticle doped PAO oil: An experimental study. Journal of Applied
Physics, 2007. 101(6): p. 7.
28. McLaughlin, E., THERMAL CONDUCTIVITY OF LIQUIDS + DENSE GASES.

Chemical Reviews, 1964. 64(4): p. 389-428.
29. Diguilio, R. and A.S. Teja, THERMAL-CONDUCTIVITY OF POLY(ETHYLENE

GLYCOLS) AND THEIR BINARY-MIXTURES. Journal of Chemical and
Engineering Data, 1990. 35(2): p. 117-121.
30. Doye, J.P.K. and D.J. Wales, Polytetrahedral Clusters. Physical Review Letters,
2001. 86(25): p. 5719 - 5722.
31. Kutz, M., ed. Handbook of Materials Selection. 2002, John Wiley and Sons: New
York. 1497.
175
32. Kang, H.U., S.H. Kim, and J.M. Oh, Estimation of thermal conductivity of
nanofluid using experimental effective particle volume. Experimental Heat
Transfer, 2006. 19(3): p. 181-191.
33. Tritt, T., ed. Thermal conductivity: theory, properties, and applications. 2004,
Kluwer Academic/Plenum Publishers: New York. 290.
34. Tien, C.-L., A. Majumdar, and F.M. Gerner, eds. Microscale Energy Transport.
1998, Taylor & Francis: Washington D.C. 395.
35. Tien, C.-L., ed. Annual Review of Heat Transfer. Vol. 7. 1996, Begell House, Inc:
New York. 444.
36. Marquardt, F. Sound wave traveling through crystal lattice. 2007 [cited 2008];
Available from: http://en.wikipedia.org/wiki/Image:Lattice_wave.svg.
37. Bansal, N.P. and D. Zhu, Thermal Conductivity of Alumina-Toughened Ziconia

Composites. 2003, NASA: Cleveland. p. 15.
38. Ju, Y.S., Phonon heat transport in silicon nanostructures. Applied Physics
Letters, 2005. 87(15): p. 3.
39. Behkam, B., Y.Z. Yang, and M. Asheghi, Thermal property measurement of thin
aluminum oxide layers for giant magnetoresistive (GMR) head applications.
International Journal of Heat and Mass Transfer, 2005. 48(10): p. 2023-2031.
40. Liu, W. and M. Asheghi, Phonon-boundary scattering in ultrathin single-crystal

silicon layers. Applied Physics Letters, 2004. 84(19): p. 3819-3821.
41. Yu, X.Y., G. Chen, A. Verma, and J.S. Smith, TEMPERATURE-DEPENDENCE

OF THERMOPHYSICAL PROPERTIES OF GAAS/ALAS PERIODIC
STRUCTURE. Applied Physics Letters, 1995. 67(24): p. 3554-3556.
42. Cahill, D.G., W.K. Ford, K.E. Goodson, G.D. Mahan, A. Majumdar, H.J. Maris,
R. Merlin, and S.R. Phillpot, Nanoscale thermal transport. Journal of Applied
Physics, 2003. 93(2): p. 793-818.
176
43. Ziambaras, E. and P. Hyldgaard, Phonon Knudsen flow in nanostructured
semiconductor systems. Journal of Applied Physics, 2006. 99(5): p. 054303.
44. Li, D.Y., Y.Y. Wu, P. Kim, L. Shi, P.D. Yang, and A. Majumdar, Thermal
conductivity of individual silicon nanowires. Applied Physics Letters, 2003.
83(14): p. 2934-2936.
45. Fang, K.C., C.I. Weng, and S.P. Ju, An investigation into the structural features
and thermal conductivity of silicon nanoparticles using molecular dynamics
simulations. Nanotechnology, 2006. 17(15): p. 3909-3914.
46. Shin, S. and S.H. Lee, Thermal conductivity of suspensions in shear flow fields.
47. Bjorneklett, A., L. Halbo, and H. Kristiansen, THERMAL-CONDUCTIVITY OF

EPOXY ADHESIVES FILLED WITH SILVER PARTICLES. International Journal
of Adhesion and Adhesives, 1992. 12(2): p. 99-104.
48. Turian, R.M., D.J. Sung, and F.L. Hsu, Thermal conductivity of granular coals,
coal-water mixtures and multi-solid/liquid suspensions. Fuel, 1991. 70(10): p.
1157-1172.
49. Eastman, J.A., U.S. Choi, S. Li, L.J. Thompson, and S. Lee, Enhanced thermal
conductivity through the development of nanofluids. Materials Research Society
Symposium Proceedings, 1997. 457(Nanophase and Nanocomposite Materials II):
p. 3-11.
50. Chon, C.H., K.D. Kihm, S.P. Lee, and S.U.S. Choi, Empirical correlation finding
the role of temperature and particle size for nanofluid (Al2O3) thermal
conductivity enhancement. Applied Physics Letters, 2005. 87(15).
51. Zhang, X., H. Gu, and M. Fujii, Experimental study on the effective thermal
conductivity and thermal diffusivity of nanofluids. International Journal of
Thermophysics, 2006. 27(2): p. 569-580.
52. Li, C.H. and G.P. Peterson, The effect of particle size on the effective thermal
conductivity of Al2O3-water nanofluids. Journal of Applied Physics, 2007.
101(4): p. 044312.
177
53. Wang, X.W., X.F. Xu, and S.U.S. Choi, Thermal conductivity of nanoparticle-
fluid mixture. Journal of Thermophysics and Heat Transfer, 1999. 13(4): p. 474-
480.
54. Zhang, X., H. Gu, and M. Fujii, Effective thermal conductivity and thermal
diffusivity of nanofluids containing spherical and cylindrical nanoparticles.
Journal of Applied Physics, 2006. 100(4): p. 044325.
55. Yoo, D.H., K.S. Hong, T.E. Hong, J.A. Eastman, and H.S. Yang, Thermal
conductivity of Al2O3/water nanofluids. Journal of the Korean Physical Society,
2007. 51: p. S84-S87.
56. Yoo, D.H., K.S. Hong, and H.S. Yang, Study of thermal conductivity of
nanofluids for the application of heat transfer fluids. Thermochimica Acta, 2007.
455(1-2): p. 66-69.
57. Lee, S., S.U.S. Choi, S. Li, and J.A. Eastman, Measuring thermal conductivity of
fluids containing oxide nanoparticles. Journal of Heat Transfer-Transactions of
the Asme, 1999. 121(2): p. 280-289.
58. Timofeeva, E.V., A.N. Gavrilov, J.M. McCloskey, Y.V. Tolmachev, S. Sprunt,
L.M. Lopatina, and J.V. Selinger, Thermal conductivity and particle
agglomeration in alumina nanofluids: Experiment and theory. Physical Review E,
2007. 76(6): p. 16.
59. Xie, H.Q., J.C. Wang, T.G. Xi, Y. Liu, and F. Ai, Dependence of the thermal
conductivity of nanoparticle-fluid mixture on the base fluid. Journal of Materials
Science Letters, 2002. 21(19): p. 1469-1471.
60. Eastman, J.A., U.S. Choi, G. Soyez, L.J. Thompson, and R.J. DiMelfi, Novel
thermal properties of nanostructured materials. Materials Science Forum, 1999.
312-314(Metastable, Mechanically Alloyed and Nanocrystalline Materials): p.
629-634.
61. Venerus, D.C., M.S. Kabadi, S. Lee, and V. Perez-Luna, Study of thermal
transport in nanoparticle suspensions using forced Rayleigh scattering. Journal of
Applied Physics, 2006. 100(9).
178
62. Hwang, Y.J., Y.C. Ahn, H.S. Shin, C.G. Lee, G.T. Kim, H.S. Park, and J.K. Lee,
Investigation on characteristics of thermal conductivity enhancement of
nanofluids. Current Applied Physics, 2006. 6(6): p. 1068-1071.
63. Lee, D., J.W. Kim, and B.G. Kim, A new parameter to control heat transport in
nanofluids: Surface charge state of the particle in suspension. Journal of Physical
Chemistry B, 2006. 110(9): p. 4323-4328.
64. Kwak, K. and C. Kim, Viscosity and thermal conductivity of copper oxide
nanofluid dispersed in ethylene glycol. Korea-Australia Rheology Journal, 2005.
17(2): p. 35-40.
65. Liu, M.S., M.C.C. Lin, I.T. Huang, and C.C. Wang, Enhancement of thermal
conductivity with CuO for nanofluids. Chemical Engineering & Technology,
2006. 29(1): p. 72-77.
66. Murshed, S.M.S., K.C. Leong, and C. Yang, Enhanced thermal conductivity of
TiO2 - water based nanofluids. International Journal of Thermal Sciences, 2005.
44(4): p. 367-373.
67. Wang, Z.L., D.W. Tang, S. Liu, X.H. Zheng, and N. Araki, Thermal-conductivity
and thermal-diffusivity measurements of nanofluids by 3 omega method and
mechanism analysis of heat transport. International Journal of Thermophysics,
2007. 28(4): p. 1255-1268.
68. Hong, T.K. and H.S. Yang, Nanoparticle-dispersion-dependent thermal

conductivity in nanofluids. Journal of the Korean Physical Society, 2005. 47: p.
S321-S324.
69. Zhu, H.T., C.Y. Zhang, S.Q. Liu, Y.M. Tang, and Y.S. Yin, Effects of
nanoparticle clustering and alignment on thermal conductivities of Fe3O4
aqueous nanofluids. Applied Physics Letters, 2006. 89(2).
70. Xie, H.Q., J.C. Wang, T.G. Xi, Y. Liu, and F. Ai, Thermal conductivity of
suspension containing SiC particles. Journal of Materials Science Letters, 2002.
21(3): p. 193-195.
71. Xuan, Y.M. and Q. Li, Heat transfer enhancement of nanofluids. International
Journal of Heat and Fluid Flow, 2000. 21(1): p. 58-64.
179
72. Liu, M.S., M.C.C. Lin, C.Y. Tsai, and C.C. Wang, Enhancement of thermal
conductivity with Cu for nanofluids using chemical reduction method.
International Journal of Heat and Mass Transfer, 2006. 49(17-18): p. 3028-3033.
73. Jana, S., A. Salehi-Khojin, and W.H. Zhong, Enhancement of fluid thermal
conductivity by the addition of single and hybrid nano-additives. Thermochimica
Acta, 2007. 462(1-2): p. 45-55.
74. Eastman, J.A., S.U.S. Choi, S. Li, W. Yu, and L.J. Thompson, Anomalously
increased effective thermal conductivities of ethylene glycol-based nanofluids
containing copper nanoparticles. Applied Physics Letters, 2001. 78(6): p. 718-
720.
75. Assael, M.J., I.N. Metaxa, K. Kakosimos, and D. Constantinou, Thermal

conductivity of nanofluids - Experimental and theoretical. International Journal of
Thermophysics, 2006. 27(4): p. 999-1017.
76. Jwo, C.S. and T.P. Teng, Experimental study on thermal properties of brines
containing nanoparticles. Reviews on Advanced Materials Science, 2005. 10(1):
p. 79-83.
77. Lee, C.H., S.W. Kang, and S.H. Kim, Effects of nano-sized ag particles on heat
transfer of nanofluids. Journal of Industrial and Engineering Chemistry, 2005.
11(1): p. 152-158.
78. Putnam, S.A., D.G. Cahill, P.V. Braun, Z.B. Ge, and R.G. Shimmin, Thermal
conductivity of nanoparticle suspensions. Journal of Applied Physics, 2006.
99(8).
79. Hong, T.K., H.S. Yang, and C.J. Choi, Study of the enhanced thermal conductivity
of Fe nanofluids. Journal of Applied Physics, 2005. 97(6).
80. Hong, K.S., T.K. Hong, and H.S. Yang, Thermal conductivity of Fe nanofluids
depending on the cluster size of nanoparticles. Applied Physics Letters, 2006.
88(3).
81. Manna, I., M. Chopkar, and P.K. Das, Nanofluid - A new concept in heat transfer
and thermal management. Transactions of the Indian Institute of Metals, 2005.
58(6): p. 1045-1055.
180
82. Chopkar, M., S. Kumar, D.R. Bhandari, P.K. Das, and I. Manna, Development
and characterization of Al2Cu and Ag2Al nanoparticle dispersed water and
ethylene glycol based nanofluid. Materials Science and Engineering B-Solid State
Materials for Advanced Technology, 2007. 139(2-3): p. 141-148.
83. Ceylan, A., K. Jastrzembski, and S.I. Shah, Enhanced solubility Ag-Cu
nanoparticles and their thermal transport properties. Metallurgical and Materials
Transactions a-Physical Metallurgy and Materials Science, 2006. 37A(7): p.
2033-2038.
84. Wen, D.S. and Y.L. Ding, Effective thermal conductivity of aqueous suspensions
of carbon nanotubes (carbon nanotubes nanofluids). Journal of Thermophysics
and Heat Transfer, 2004. 18(4): p. 481-485.
85. Assael, M.J., C.F. Chen, I. Metaxa, and W.A. Wakeham, Thermal conductivity of
suspensions of carbon nanotubes in water. International Journal of
Thermophysics, 2004. 25(4): p. 971-985.
86. Assael, M.J., I.N. Metaxa, J. Arvanitidis, D. Christofilos, and C. Lioutas, Thermal
conductivity enhancement in aqueous suspensions of carbon multi-walled and
double-walled nanotubes in the presence of two different dispersants.
International Journal of Thermophysics, 2005. 26(3): p. 647-664.
87. Hong, H.P., J. Wensel, F. Liang, W.E. Billups, and W. Roy, Heat transfer
nanofluids based on carbon nanotubes. Journal of Thermophysics and Heat
Transfer, 2007. 21(1): p. 234-236.
88. Wright, B., D. Thomas, H. Hong, L. Groven, J. Puszynski, E. Duke, X. Ye, and S.
Jin, Magnetic field enhanced thermal conductivity in heat transfer nanofluids
containing Ni coated single wall carbon nanotubes. Applied Physics Letters,
2007. 91(17): p. 3.
89. Wensel, J., B. Wright, D. Thomas, W. Douglas, B. Mannhalter, W. Cross, H.P.

Hong, J. Kellar, P. Smith, and W. Roy, Enhanced thermal conductivity by
aggregation in heat transfer nanofluids containing metal oxide nanoparticles and
carbon nanotubes. Applied Physics Letters, 2008. 92(2): p. 3.
90. Liu, M.S., M.C.C. Lin, I.T. Huang, and C.C. Wang, Enhancement of thermal
conductivity with carbon nanotube for nanofluids. International Communications
in Heat and Mass Transfer, 2005. 32(9): p. 1202-1210.
181
91. Casquillas, G.V., M. Le Berre, C. Peroz, Y. Chen, and J.J. Greffet, Microlitre hot
strip devices for thermal characterization of nanofluids. Microelectronic
Engineering, 2007. 84(5-8): p. 1194-1197.
92. Marquis, F.D.S. and L.P.F. Chibante, Improving the heat transfer of nanofluids
and nanolubricants with carbon nanotubes. JOM, 2005. 57(12): p. 32-43.
93. Yang, Y., E.A. Grulke, Z.G. Zhang, and G.F. Wu, Thermal and rheological
properties of carbon nanotube-in-oil dispersions. Journal of Applied Physics,
2006. 99(11).
94. Zhu, H.T., C.Y. Zhang, Y.M. Tang, J.X. Wang, B. Ren, and Y.S. Yin,
Preparation and thermal conductivity of suspensions of graphite nanoparticles.
Carbon, 2007. 45(1): p. 226-228.
95. Hong, H.P., B. Wright, J. Wensel, S.H. Jin, X.R. Ye, and W. Roy, Enhanced
thermal conductivity by the magnetic field in heat transfer nanofluids containing
carbon nanotube. Synthetic Metals, 2007. 157(10-12): p. 437-440.
96. Wakeham, W.A., A. Nagashima, and J.V. Sengers, eds. Measurement of the
Transport Properties of Fluids. Experimental Thermodynamics. 1991, Blackwell
Scientific Publications: Oxford. 459.
97. Murshed, S.M.S., K.C. Leong, and C. Yang, Determination of the effective
thermal diffusivity of nanofluids by the double hot-wire technique. Journal of
Physics D-Applied Physics, 2006. 39(24): p. 5316-5322.
98. Carslaw, H.S. and J.C. Jaeger, Conduction of Heat in Solids. 2nd ed. 1959,
Oxford: Clarendon Press. 510.
99. Degroot, J.J., J. Kestin, and Sookiazi.H, INSTRUMENT TO MEASURE

THERMAL-CONDUCTIVITY OF GASES. Physica, 1974. 75(3): p. 454-482.
100. Healy, J.J., J.J. de Groot, and J. Kestin, THEORY OF TRANSIENT HOT-WIRE
METHOD FOR MEASURING THERMAL-CONDUCTIVITY. Physica B & C,
1976. 82(2): p. 392-408.
101. Nagasaka, Y. and A. Nagashima, ABSOLUTE MEASUREMENT OF THE

THERMAL-CONDUCTIVITY OF ELECTRICALLY CONDUCTING LIQUIDS BY
182
THE TRANSIENT HOT-WIRE METHOD. Journal of Physics E-Scientific
Instruments, 1981. 14(12): p. 1435-1440.
102. Hoshi, M., T. Omotani, and A. Nagashima, TRANSIENT METHOD TO

MEASURE THE THERMAL-CONDUCTIVITY OF HIGH-TEMPERATURE
MELTS USING A LIQUID-METAL PROBE. Review of Scientific Instruments,
1981. 52(5): p. 755-758.
103. Diguilio, R.M. and A.S. Teja, THERMAL-CONDUCTIVITY OF AQUEOUS

SALT-SOLUTIONS AT HIGH-TEMPERATURES AND HIGH-
CONCENTRATIONS. Industrial & Engineering Chemistry Research, 1992. 31(4):
p. 1081-1085.
104. Bleazard, J.G. and A.S. Teja, Thermal-Conductivity of Electrically Conducting

Liquids by the Transient Hot-Wire Method. Journal of Chemical and Engineering
Data, 1995. 40(4): p. 732-737.
105. Bleazard, J.G., T.F. Sun, R.D. Johnson, R.M. DiGuilio, and A.S. Teja, The
transport properties of seven alkanediols. Fluid Phase Equilibria, 1996. 117(1-2):
p. 386-393.
106. Bleazard, J.G., T.F. Sun, and A.S. Teja, The thermal conductivity and viscosity of
acetic acid-water mixtures. International Journal of Thermophysics, 1996. 17(1):
p. 111-125.
107. Sun, T.F. and A.S. Teja, Density, viscosity and thermal conductivity of aqueous
solutions of propylene glycol, dipropylene glycol, and tripropylene glycol between
290 K and 460 K. Journal of Chemical and Engineering Data, 2004. 49(5): p.
1311-1317.
108. Sun, T.F. and A.S. Teja, Density, viscosity, and thermal conductivity of aqueous
ethylene, diethylene, and triethylene glycol mixtures between 290 K and 450 K.
Journal of Chemical and Engineering Data, 2003. 48(1): p. 198-202.
109. Sun, T. and A.S. Teja, Density, viscosity, and thermal conductivity of aqueous
benzoic acid mixtures between 375 K and 465 K. Journal of Chemical and
Engineering Data, 2004. 49(6): p. 1843-1846.
183
110. Diguilio, R.M. and A.S. Teja, THE THERMAL-CONDUCTIVITY OF THE
MOLTEN NANO3-KNO3 EUTECTIC BETWEEN 525-K AND 590-K.
111. Omotani, T., Y. Nagasaka, and A. Nagashima, Measurement of the thermal

conductivity of KNO3-NaNO3 mixtures with liquid metal in a capillary probe.
112. Nieto de Castro, C.A., S.F.Y. Li, C. Maitland, and W.A. Wakeham, Thermal
conductivity of toluene in the temperature range 35-90 oC at pressures up to 600
MPa. International Journal of Thermophysics, 1983. 4(4): p. 311-327.
113. de Groot, J.J., J. Kestin, and Sookiazi.H, INSTRUMENT TO MEASURE

THERMAL-CONDUCTIVITY OF GASES. Physica, 1974. 75(3): p. 454-482.
114. Marsh, K.N., ed. Recommended Reference Materials for the Realization of
Physicochemical Properties. 1987, Blackwell Scientific Publications: Boston.
500.
115. DIPPR, DIPPR Project 801. 2007, Design Institute for Physical Properties /
AIChE.
116. Drelich, J., J.S. Laskowski, and K.L. Mittal, eds. Apparent and Microscopic
Contact Angles. 2000, Utrecht: Boston. 524.
117. Kosmulski, M., Chemical Properties and Material Surfaces. 2001, New York:
Marcel Dekker. 753.
118. Maxwell, J.C., A Treatise on Electricity and Magnetism. 3rd ed. Vol. II. 1892,
London: Oxford University Press.
119. Carson, J.K., Review of effective thermal conductivity models for foods.
International Journal of Refrigeration-Revue Internationale Du Froid, 2006. 29(6):
p. 958-967.
120. Tsotsas, E. and H. Martin, THERMAL-CONDUCTIVITY OF PACKED-BEDS - A

REVIEW. Chemical Engineering and Processing, 1987. 22(1): p. 19-37.
184
121. Progelhof, R.C., J.L. Throne, and R.R. Ruetsch, METHODS FOR PREDICTING
THERMAL-CONDUCTIVITY OF COMPOSITE SYSTEMS. Polymer Engineering
and Science, 1976. 16(9): p. 615-625.
122. Rayleigh, L., On the Influence of Obstacles arranged in Rectangular Order upon
the Properties of a Medium. Philosophical Magazine and Journal of Science,
1892. 34: p. 481-502.
123. Hamilton, R.L. and O.K. Crosser, Thermal Conductivity of Heterogeneous 2-

Component Systems. Industrial & Engineering Chemistry Fundamentals, 1962.
1(3): p. 187.
124. Jeffrey, D.J., CONDUCTION THROUGH A RANDOM SUSPENSION OF

SPHERES. Proceedings of the Royal Society of London Series a-Mathematical
Physical and Engineering Sciences, 1973. 335(1602): p. 355-367.
125. Bruggeman, D.A.G., Calculation of various physics constants in heterogenous

substances I Dielectricity constants and conductivity of mixed bodies from
isotropic substances. Annalen Der Physik, 1935. 24(7): p. 636-664.
126. Landauer, R., THE ELECTRICAL RESISTANCE OF BINARY METALLIC

MIXTURES. Journal of Applied Physics, 1952. 23(7): p. 779-784.
127. Keey, R.B., Drying of Loose and Particulate Materials. 1992, New York:
Hemisphere Publishing Corporation. 504.
128. Krischer, O., Die wissenschaftlichen Grundlagen der Trocknungstechnik (The

Scientific Fundamentals of Drying Technology). 2nd ed. 1963, Berlin: Springer-
Verlag.
129. Tsao, G.T.N., THERMAL CONDUCTIVITY OF 2-PHASE MATERIALS.

Industrial and Engineering Chemistry, 1961. 53(5): p. 395-397.
130. Hashin, Z. and S. Shtrikman, A Variational Approach to the Theory of the

Effective Magnetic Permeability of Multiphase Materials. Journal of Applied
Physics, 1962. 33(10): p. 3125.
131. Jang, S.P. and S.U.S. Choi, Role of Brownian motion in the enhanced thermal
conductivity of nanofluids. Applied Physics Letters, 2004. 84(21): p. 4316-4318.
185
132. Kumar, D.H., H.E. Patel, V.R.R. Kumar, T. Sundararajan, T. Pradeep, and S.K.
Das, Model for heat conduction in nanofluids. Physical Review Letters, 2004.
93(14): p. 144301.
133. Patel, H.E., T. Sundararajan, T. Pradeep, A. Dasgupta, N. Dasgupta, and S.K.

Das, A micro-convection model for thermal conductivity of nanofluids. Pramana-
Journal of Physics, 2005. 65(5): p. 863-869.
134. Patel, H.E., T. Sundararajan, and S.K. Das, A cell model approach for thermal
conductivity of nanofluids. Journal of Nanoparticle Research, 2008. 10(1): p. 87-
97.
135. Koo, J. and C. Kleinstreuer, A new thermal conductivity model for nanofluids.
Journal of Nanoparticle Research, 2004. 6(6): p. 577-588.
136. Prasher, R., P. Bhattacharya, and P.E. Phelan, Thermal conductivity of nanoscale
colloidal solutions (nanofluids). Physical Review Letters, 2005. 94(2): p. 025901.
137. Ren, Y., H. Xie, and A. Cai, Effective thermal conductivity of nanofluids
containing spherical nanoparticles. Journal of Physics D-Applied Physics, 2005.
38(21): p. 3958-3961.
138. Xuan, Y.M., Q. Li, X. Zhang, and M. Fujii, Stochastic thermal transport of
nanoparticle suspensions. Journal of Applied Physics, 2006. 100(4).
139. Prakash, M. and E.P. Giannelis, Mechanism of heat transport in nanofluids.

Journal of Computer-Aided Materials Design, 2007. 14(1): p. 109-117.
140. Yu, W. and S.U.S. Choi, The role of interfacial layers in the enhanced thermal
conductivity of nanofluids: A renovated Maxwell model. Journal of Nanoparticle
Research, 2003. 5(1-2): p. 167-171.
141. Yu, W. and S.U.S. Choi, The role of interfacial layers in the enhanced thermal
conductivity of nanofluids: A renovated Hamilton-Crosser model. Journal of
Nanoparticle Research, 2004. 6(4): p. 355-361.
142. Xie, H.Q., M. Fujii, and X. Zhang, Effect of interfacial nanolayer on the effective
thermal conductivity of nanoparticle-fluid mixture. International Journal of Heat
and Mass Transfer, 2005. 48(14): p. 2926-2932.
186
143. Xue, Q.Z., Model for thermal conductivity of carbon nanotube-based composites.
Physica B-Condensed Matter, 2005. 368(1-4): p. 302-307.
144. Leong, K.C., C. Yang, and S.M.S. Murshed, A model for the thermal conductivity
of nanofluids - the effect of interfacial layer. Journal of Nanoparticle Research,
2006. 8(2): p. 245-254.
145. Feng, Y.J., B.M. Yu, P. Xu, and M.Q. Zou, The effective thermal conductivity of
nanofluids based on the nanolayer and the aggregation of nanoparticles. Journal
of Physics D-Applied Physics, 2007. 40(10): p. 3164-3171.
146. Lee, D., Thermophysical properties of interfacial layer in nanofluids. Langmuir,

2007. 23(11): p. 6011-6018.
147. Evans, W., J. Fish, and P. Keblinski, Thermal conductivity of ordered molecular
water. Journal of Chemical Physics, 2007. 126(15).
148. Wang, B.X., L.P. Zhou, and X.F. Peng, A fractal model for predicting the
effective thermal conductivity of liquid with suspension of nanoparticles.
149. Xuan, Y.M., Q. Li, and W.F. Hu, Aggregation structure and thermal conductivity
of nanofluids. Aiche Journal, 2003. 49(4): p. 1038-1043.
150. Prasher, R., W. Evans, P. Meakin, J. Fish, P. Phelan, and P. Keblinski, Effect of
aggregation on thermal conduction in colloidal nanofluids. Applied Physics
Letters, 2006. 89(14): p. 143119.
151. Kumar, S. and J.Y. Murthy, A numerical technique for computing effective
thermal conductivity of fluid-particle mixtures. Numerical Heat Transfer Part B-
Fundamentals, 2005. 47(6): p. 555-572.
152. Gao, L. and X.F. Zhou, Differential effective medium theory for thermal
conductivity in nanofluids. Physics Letters A, 2006. 348(3-6): p. 355-360.
153. Eapen, J., J. Li, and S. Yip, Beyond the Maxwell limit: Thermal conduction in
nanofluids with percolating fluid structures. Physical Review E, 2007. 76(6): p. 4.
187
154. Swartz, E.T. and R.O. Pohl, Thermal boundary resistance. Reviews of Modern
Physics, 1989. 61(3): p. 605-668.
155. Nan, C.W., R. Birringer, D.R. Clarke, and H. Gleiter, Effective thermal
conductivity of particulate composites with interfacial thermal resistance. Journal
of Applied Physics, 1997. 81(10): p. 6692-6699.
156. Huxtable, S.T., D.G. Cahill, S. Shenogin, L.P. Xue, R. Ozisik, P. Barone, M.
Usrey, M.S. Strano, G. Siddons, M. Shim, and P. Keblinski, Interfacial heat flow
in carbon nanotube suspensions. Nature Materials, 2003. 2(11): p. 731-734.
157. Wilson, O.M., X.Y. Hu, D.G. Cahill, and P.V. Braun, Colloidal metal particles as
probes of nanoscale thermal transport in fluids. Physical Review B, 2002. 66(22):
p. 6.
188
VITA
MICHAEL P. BECK
Michael P. Beck was born in Evanston, Illinois on October 27, 1975 to Donald P.
Beck and Linda S. Beck. He grew up with three brothers and attended Beloit Catholic
High School in Beloit, Wisconsin. He received his Bachelor of Science degrees in
Chemical Engineering and in Biochemistry from the University of Wisconsin in Madison
in 1999. From 1999 until 2002, he worked for Abbott Laboratories in North Chicago,
Illinois in the Corporate Engineering Division and the Abbott Diagnostics Division
before coming to Georgia Tech in 2002 to pursue a doctorate in chemical engineering.
On June 4, 2005, he married Kanrakot (Nan) Thamanavat in Atlanta, GA. Following his
graduation from Georgia Tech, he will work at Nalco Company in Naperville, IL as a
senior engineer.
189

INTERESSANT!!!!!4

Uploaded by

Copyright:

Available Formats

INTERESSANT!!!!!4

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

INTERESSANT!!!!!4

Uploaded by

Copyright:

Available Formats

THERMAL CONDUCTIVITY OF METAL OXIDE NANOFLUIDS

Michael Peter Beck

Georgia Institute of Technology

Dr. Amyn S. Teja, Advisor Dr. Sankar Nair

Dr. Carson Meredith Dr. Ganesh Skandan

Dr. Said I. Abdel-Khalik

Date Approved: August 12, 2008

his support and guidance in my research. I would also like to acknowledge my

help and advice over the years.

who spent their time helping me through various stages of my work.

construction of my measurement apparatus.

LIST OF SYMBOLS xvii

2.1 Thermal Conductivity of Liquids 6

2.1.1 Temperature and Pressure Dependence 8

2.2 Thermal Conductivity of Solids 12

2.2.1 Lattice Thermal Conductivity 13

2.2.2 Temperature Dependence 15

2.2.3 Thermal Conductivity of Solids as a Function of Size and Dimension 17

2.3 Thermal Conductivity of Heterogeneous Systems 20

2.3.1 Thermal Conductivity of Microparticle Dispersions 20

2.3.2 Thermal Conductivity of Nanofluids 21

2.3.2.1 Effect of Volume Fraction 25

2.3.2.2 Effect of Particle Size 29

2.3.2.3 Effect of Temperature 31

2.3.2.5 Effect of Particle Surface Charge 33

2.3.2.6 Effect of Particle Arrangement in Suspension 34

TRANSIENT HOT-WIRE METHOD 36

3.1 Basic Measurement Technique 38

3.2 Background of Transient Hot-wire Method 39

3.2.1 Basic Transient Hot-wire Method 39

3.2.2 Insulated Wire Method 40

3.2.3 Liquid Metal Wire Method 40

3.3 Construction of Transient Hot-wire Instrument 41

3.3.1.1 Hot Wire Cell 44

3.3.1.2 Instrumentation of Bleazard and Teja 44

3.3.1.3 Updated Instrumentation 44

3.3.2 Data Acquisition and Analysis 45

3.4 Procedure, Analysis and Corrections 48

3.4.3 Corrections and Calibration 50

3.4.3.1 Insulating Layer and Three Dimensional Aspect of Wire 50

3.4.3.2 Finite Extent of the Liquid 51

3.4.3.3 Heat Transfer by Radiation 51

3.6 Validation of Updated Instrument 60

THERMAL CONDUCTIVITY MEASUREMENTS AND CHARACTERIZATION

4.1 Materials and Methods 63

4.1.2 Sample Preparation 65

4.2 Nanofluid Characterization 67

4.2.1 Primary Particle Size 67

4.2.2 Secondary Particle Size 68

4.3 Thermal Conductivity Measurements 73

4.3.1 Temperature Studies 73

4.3.2 Particle Size Studies 86

4.4 Thermal Conductivity Behavior 95

5.1 Type I Thermal Conductivity Models 97

5.2 Type II Thermal Conductivity Models 100

5.3 Theoretical Bounds for Thermal Conductivity of Heterogeneous Materials 101

5.4 Accuracy of Thermal Conductivity Models 101

5.5 Geometric Mean 102

5.6 Nanofluid Thermal Conductivity Models 103