Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

W.

Escher
IBM Research GmbH,
Zurich Research Laboratory,
8803 Rüschlikon, Switzerland;
Department of Mechanical and Process
Engineering,
Laboratory of Thermodynamics in Emerging
Technologies,
ETH Zurich,
8092 Zurich, Switzerland

T. Brunschwiler
IBM Research GmbH,
Zurich Research Laboratory,
On the Cooling of Electronics
8803 Rüschlikon, Switzerland
With Nanofluids
N. Shalkevich
Laboratoire de chimie physique des surfaces, Nanofluids have been proposed to improve the performance of microchannel heat sinks.
Institut de Physique, In this paper, we present a systematic characterization of aqueous silica nanoparticle
Universite de Neuchâtel, suspensions with concentrations up to 31 vol %. We determined the particle morphology
Rue Emile-Argand 11, by transmission electron microscope imaging and its dispersion status by dynamic light
2009-Neuchatel, Switzerland scattering measurements. The thermophysical properties of the fluids, namely, their spe-
cific heat, density, thermal conductivity, and dynamic viscosity were experimentally mea-
A. Shalkevich sured. We fabricated microchannel heat sinks with three different channel widths and
Adolphe Merkle Institute, characterized their thermal performance as a function of volumetric flow rate for silica
Université de Fribourg, nanofluids at concentrations by volume of 0%, 5%, 16%, and 31%. The Nusselt number
P.O. Box 209 11, was extracted from the experimental results and compared with the theoretical predic-
CH-1723 Marly 1, Switzerland tions considering the change of fluids bulk properties. We demonstrated a deviation of
less than 10% between the experiments and the predictions. Hence, standard correlations
T. Burgi can be used to estimate the convective heat transfer of nanofluids. In addition, we applied
Laboratoire de chimie physique des surfaces, a one-dimensional model of the heat sink, validated by the experiments. We predicted the
Institut de Physique, potential of nanofluids to increase the performance of microchannel heat sinks. To this
Universite de Neuchâtel, end, we varied the individual thermophysical properties of the coolant and studied their
Rue Emile-Argand 11, impact on the heat sink performance. We demonstrated that the relative thermal conduc-
2009-Neuchatel, Switzerland; tivity enhancement must be larger than the relative viscosity increase in order to gain a
Physikalisch-Chemisches Institut, sizeable performance benefit. Furthermore, we showed that it would be preferable to
Ruprecht-Karls-Universitat Heidelberg, increase the volumetric heat capacity of the fluid instead of increasing its thermal
Im Neuenheimer Feld 253, conductivity. 关DOI: 10.1115/1.4003283兴
69120 Heidelberg, Germany
Keywords: nanofluid, nanoparticle, suspension, thermal conductivity, convective heat
B. Michel transfer, electronics cooling, experiment
IBM Research GmbH,
Zurich Research Laboratory,
8803 Rüschlikon, Switzerland

D. Poulikakos1
Department of Mechanical and Process
Engineering,
Laboratory of Thermodynamics in Emerging
Technologies,
ETH Zurich,
8092 Zurich, Switzerland
e-mail: dimos.poulikakos@ethz.ch

1 Introduction generating electronic component 关1–5兴. Commonly, water is sug-


gested to be used as a single-phase coolant in combination with
High heat flux removal is a major challenge in the design of
microchannel heat sinks for cooling electronics, as it possesses the
future electronic devices. The trend to address these high heat
most adequate thermal and hydrodynamic transport properties in
fluxes is to introduce microchannel arrays directly in the heat
the required range of operating temperatures. However, the ther-
mal conductivity of water is two to three orders of magnitude
1
Corresponding author. lower than of most metals and metal oxides. Therefore, an inno-
Contributed by the Heat Transfer Division of ASME for publication in the JOUR- vative way to elevate the thermal conductivity of fluids may be
NAL OF HEAT TRANSFER. Manuscript received August 31, 2009; final manuscript re-
ceived December 2, 2010; published online February 4, 2011. Assoc. Editor: Roger the addition of nanometer-sized metal or metal oxide particles into
Schmidt. a base-fluid, most suitably water. The use of nanometer-sized par-

Journal of Heat Transfer Copyright © 2011 by ASME MAY 2011, Vol. 133 / 051401-1

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 08/18/2013 Terms of Use: http://asme.org/terms


ticles with large specific surface area would provide higher stabi- 5%, 16%, and 31%. In a second step, we experimentally studied
lization against particle sedimentation and prevent clogging of the the heat transfer characteristic of the nanofluids and their potential
heat sink. Various types of nanoparticles such as metallic 关6–10兴, to enhance the performance of a microchannel heat sink.
nonmetallic 关11–13兴, with different shapes and sizes, can be sus- To achieve this goal, we fabricated three different test chips
pended in different fluids forming so-called nanofluids. with varying channel widths and determined the total thermal re-
The major interest of the research on nanofluids has focused on sistance and the pressure loss of the heat sinks in combination
the enhanced thermal conductivity of the colloids under stationary with the nanofluids as a function of volumetric flow rate. We
condition. There has been a broad range of experimental investi- compared the measured results to theoretical predictions and dem-
gations using common measurement techniques such as the onstrated that if the thermophysical properties of the fluids are
transient-hot-wire 共THW兲 method, the 3-␻ method, and the par- known; standard correlations can be applied to predict the convec-
allel plate method. However, there is a large scatter in the data of tive heat transfer of nanofluids even at particle concentration up to
thermal conductivity enhancements. To give just one example, 31 vol %. To draw a relevant comparison between the heat sink
Choi et al. 关14兴 reported a 160% thermal conductivity increase for performance with water and that with a nanofluid, the heat sink
1 vol % of multiwalled carbon nanotubes 共MWCNTs兲 in silicone performance has to be compared at the design optimum for each
oil, whereas Xie et al. 关15兴 measured only a 20% enhancement of combination of fluid properties. Therefore, we introduced a one-
thermal conductivity for a similar fluid. To resolve the inconsis- dimensional model validated by the experiments. The model is
tencies, an international nanofluid properties benchmark exercise capable to predict the total thermal resistance and the pressure loss
共INPBE兲 with more than 30 groups participating was triggered as a function of operating parameters and fluid properties. We
关16兴. The objective of the exercise was to generate a reliable da- applied this model to determine the optimum design of a micro-
tabase for representative nanofluids and to give more insight if channel heat sink depending of the coolant thermophysical prop-
there are any systematic discrepancies due to the applied disper- erties and studied the sensitivity of the heat sink performance at
sion and measurement method. In the scope of the benchmark the design optimum on the thermophysical properties of the cool-
study, a large variety of nanofluids was tested, comprising aque- ant. Finally, we discuss which combination of fluid properties in
ous and nonaqueous, metallic and metallic oxide particles, and terms of thermal conductivity and dynamic viscosity is required to
spherical and elongated particles from low to high particle con- enhance the performance of a microchannel heat sink.
centrations. Various experimental approaches were applied for the
thermal conductivity measurements; however, no significant sys- 2 Experimental Setup
tematic differences due to the measurement methods were ob- Figure 1 shows a schematic of the microchannel heat sink used
served. The thermal conductivity of the tested nanofluids in- in the present investigation. A test-vehicle consists of a silicon die
creased with particle loading, particle aspect ratios, and and a glass chip being 16⫻ 16 mm2 in size. The silicon die con-
decreasing base-fluid thermal conductivity. Nevertheless, all ex- tains a 10⫻ 10 mm2 array of parallel microchannels. The micro-
perimental data could be well explained by the classical effective channel array was fabricated by standard photolithography and
medium theory for well-dispersed particles. deep reactive ion etching 共DRIE兲 into the 525 ␮m thick silicon
Forced convective heat transfer is not only influenced by the chip. In the same process, step pressure taps at 0 mm and 10 mm
thermal conductivity of the coolant. Other transport properties were integrated to resolve the pressure drop across the microchan-
such as the fluid density, specific heat capacity, and dynamic vis- nel array. In a second DRIE sequence, lateral fluid ports with an
cosity have an impact on the performance of the cooling solution opening of 1 ⫻ 10 mm2 were implemented into the silicon chip
and thus are being investigated to greater or lesser extent 关17–19兴. for fluid supply and return. We integrated a resistive heater on the
In the following, we consider laminar forced convection. The center back side of the silicon chip covered with a 100 nm thick
small dimensions in microchannel heat sinks make turbulent flows silicon oxide layer for electrical insulation. The heater covered an
impractical as they would result in large pressure losses across the area of 10⫻ 10 mm2 and was realized by depositing a 300 nm
heat sink. Several experimental studies on nanofluid laminar thick NiCr 80/20 metal layer with a low temperature resistance
forced convective heat transfer have been reported in literature coefficient of ⫾80 ppm for optimal heat flux uniformity struc-
关20–23兴. The issue whether the nanoparticles themselves affect tured by a lift-off resist providing a total electrical resistance of
laminar forced convective heat transfer apart from the change of 3.1 ⍀. At the center line of the heater orthogonal to the flow
the fluid bulk properties is still under debate since there is a lack direction, the heater was separated into two sections by a resistive
of extensive studies on this topic. temperature device 共RTD兲. A 300 nm thick gold metallization was
Rea et al. 关23兴 investigated the laminar forced convective heat used as contact pads and the formation of the RTD, providing a
transfer and pressure loss within a vertical heated tube for temperature resistance coefficient of 2300 ppm/K. The heat trans-
alumina-water and zirconia water nanofluids at maximum particle fer structure, the fluid ports and the pressure taps were covered by
volume fractions of 0.06 and 0.013, respectively. They experimen- a 500 ␮m thick glass 共Pyrex, thermal expansion coefficient close
tally determined the change of thermal conductivity and viscosity to silicon兲 wafer to seal the structure and to allow optical access
as a function of particle loading and used standard correlations for from the top. The glass wafer was spin coated with a 4 ␮m to
the effective fluid density and specific heat capacity. However, if 5 ␮m thick polyimide layer. We achieved a leak-proof bond be-
the temperature dependent thermophysical material properties of tween the silicon and the glass wafer by applying a uniform pres-
the nanofluid are taken into account, both the enhanced convective sure of 7 bars in a membrane oven at 320° C and 1 mbar atmo-
heat transfer coefficient and the increased pressure loss are in sphere. The package was cut by a dicing saw into individual chips.
good agreement with the traditional model predictions for laminar The individual test-vehicles were connected to a test-section
flow. providing interfaces for the fluid supply and return and the pres-
In this paper, we extend the study of convective heat transfer sure taps and electrical connections in form of spring loaded
characteristics of nanofluids to highly concentrated fluids and probs. The pressure drop across the microchannel array was mea-
present an experimental and theoretical evaluation of their poten- sured by a differential pressure transducer with a maximum pres-
tial for electronics cooling. Therefore, a thorough characterization sure difference of 2 bars 共error⬍ 2 mbars兲. We used T-type ther-
of a SiO2-water nanofluid was conducted. We characterized the mocouples with a specified error of ⫾0.1 K to measure the fluid
particle dispersion state by transmission electron microscopy inlet and outlet temperatures. A spatial temperature distribution of
共TEM兲 imaging and dynamic light scattering 共DLS兲 measure- the heater backside was obtained by observing the heater by an
ments. All thermophysical properties of the fluid as density, spe- infrared 共IR兲 camera with a noise-equivalent temperature differ-
cific heat, thermal conductivity, and its dynamic viscosity were ence of 25 mK. The test-section was integrated into a closed fluid
experimentally measured at particle concentrations by volume of loop as depicted in Fig. 1共c兲. The flow was driven by a magnetic

051401-2 / Vol. 133, MAY 2011 Transactions of the ASME

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 08/18/2013 Terms of Use: http://asme.org/terms


a c
transparent glass cover IR
Test Camera
Coriolis Section
RTDs, UHeater, IHeater
Flowmeter
Power Supply

Inlet Heater Outlet Tin ∆P Tout Digital


Multimeter
b Particle Filter
pore size: 7 µm Controll
Fluid
Reservoir and
Lch Data Acquisition
Gear Pump
Heat Exchanger

hch ww wch

hbase Chiller

Fig. 1 Schematic of experimental setup, „a… side view of a test-vehicle, „b… isometric view of a section of a parallel
microchannel array, and „c… closed fluid loop

coupled gear pump. We incorporated a particle filter with a pore 共Theater,max − Tf,in兲
size of 7 ␮m to prevent the microchannel array from contamina- ⬙ =
Rtot 共2兲

q̇heater
tion with larger impurities. The flow rate and the fluid density
were measured by a coriolis flow meter. We monitored the fluid We theoretically described the total thermal resistance by an
density during all runs to assure that there was no change in the equivalent resistance network 关26兴. We considered a conduction
composition of the fluids. The heat was removed by a 1.8 kW thermal resistance, R⬙cnd, from the heater through the silicon base
chiller guaranteeing constant fluid inlet temperature of defined as
20⫾ 1.5° C. Other uncertainties are listed in Table 1, where the
uncertainties of the secondary variables were derived by applying hbase
R⬙cnd = 共3兲
standard uncertainty propagation theory. A more detailed descrip- ks
tion of the fabrication process, the test-section, and the fluid loop
can be found in Ref. 关24兴. where hbase is the thickness of the silicon base 共cf. Figure 1共b兲兲
and ks is the thermal conductivity of the silicon at 20° C. The
effective convection resistance was determined by assuming a par-
3 Theoretical Model allel heat flux from the channel base and from the channel wall
We estimated the hydrodynamic and thermal performance of into the fluid, which reads
the parallel microchannel heat sink by a one-dimensional math- wch + ww
ematical model. We assumed a laminar and incompressible fluid ⬙
Rcnv,eff = 共4兲
flow with constant heat capacity, cp,f, dynamic viscosity, ␮f, den- wch␣cnv + 2hch␣cnv␥w
sity, ␳f, and thermal conductivity. The pressure drop across the where wch and hch are the channel width and height, respectively,
heat sink was estimated by assuming laminar flow through a rect- ww is the channel wall thickness, and ␥w is the fin efficiency that
angular channel and reads can be found in Ref. 关1兴. The convective heat transfer coefficient,
2f app␳fū2Lch acnv, is defined by the Nusselt number, Nu, the thermal conduc-
⌬p = 共1兲 tivity of the fluid, kf, and the characteristic channel width, dh
dh
where p is the pressure, dh is the characteristic diameter of the Nu␭f
␣cnv = 共5兲
channel, ū is the mean velocity, and f app is the apparent Fanning dh
friction factor. We used a correlation proposed by Muzychka and
To determine the Nusselt number, we assumed simultaneously
Yovanovich 关25兴, considering hydrodynamically developing lami-
developing laminar flow with a uniform wall heat flux for rectan-
nar flow in rectangular ducts. The total thermal resistance of the
gular channels 关27兴. Furthermore, we account for the limited heat
heat sink was defined by the temperature difference between the
capacity of the fluid by introducing a bulk thermal resistance,
maximum heater temperature, Theater,max, and the fluid inlet tem-
⬙ , with
Rbulk
⬙ , as
perature, Tf,in, normalized by the heat flux density, q̇heater
Lch共wch + ww兲
⬙ =
Rbulk 共6兲
cp,f ␳ f V̇ch
Table 1 Experimental uncertainties
where V̇ch is the volumetric flow rate through one channel. Con-
T 共K兲 ⫾0.1
sequently, the total thermal resistance is defined as
T IR camera 共K兲 ⫾0.3
p 共bar兲 ⫾0.002 ⬙ = R⬙cnd + R⬙cnv + Rbulk
Rtot ⬙ 共7兲
V̇ ⫾1%
Rtot ⫾3.6% Recapitulating, the hydrodynamic and thermal performance of the
Rbulk ⫾3.7% heat sink results from the interaction of the heat sinks geometry
Rcnd ⫾0.5% and its thermal conductivity, the operation parameters and the
thermophysical properties of the coolant.

Journal of Heat Transfer MAY 2011, Vol. 133 / 051401-3

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 08/18/2013 Terms of Use: http://asme.org/terms


a 1.4 b 1
φvol = 5% φvol = 5%
φvol = 16% 0.9
1.2 φvol = 16%
φvol = 31% 0.8 φvol = 31%
1 0.7

relative weight / [−]


50 nm
0.6
0.8

g(1) / [−]
0.5
0.6
0.4

0.4 0.3

0.2
0.2
0.1

0 0
−5 0 1 10 100 1000 10000 100000
10 10
lag time / [s] hydrodynamic particle radius / [nm]

Fig. 2 „a… Measured intensity correlation functions for silica suspensions at different particle con-
centrations as a function of decay time with incorporated TEM image of silica nanoparticles and „b…
particle size distributions of silica suspensions at different particle concentrations

4 Nanofluid Properties TIN data were converted into intensity weighted distributions of
hydrodynamic radii, as shown in Fig. 2共b兲. The suspension at the
The nanofluid we tested in the experiments consisted of SiO2
lowest concentration demonstrated a monomodal particle size dis-
nanoparticles dispersed in water. The fluid was purchased from
tribution 共PSD兲 with mean hydrodynamic radius of around 15 nm,
Sigma-Aldrich® with a vendor specified average particle size of
which is coherent with TEM image. Increasing the silica concen-
22 nm, which matches our measurements obtained by transmis-
tration up to 16 vol % mainly shifts the PSD to the lower values
sion electron microscopy 共cf. Figure 2共a兲兲. The suspension is elec-
due to strong repulsive interaction between negatively charged
trostatically stabilized, where the particle surface is covered by
silica particles. We observe the same effect even more pronounced
deprotonated silanol groups with sodium as counterions. The ini-
at the maximum concentration of 31 vol %, where the mean hy-
tial concentration of the fluid was 50 wt %. To achieve lower
drodynamic radius dropped down below 10 nm. At the same time,
concentrations, the fluid was diluted by MilliQ water to obtain
we observed a second peak from large aggregates. Those aggre-
concentrations of 10 wt % and 30 wt %. The particle concentra-
gates remain stable even after ultrasonication and disappear only
tion was converted in fraction by volume, ␾vol, by under dilution.
␾wt␳bf As shown in Sec. 3, the hydrodynamic and thermal perfor-
␾vol = 共8兲 mance of the heat sinks depends on the thermophysical properties
␾wt␳bf + 共1 − ␾wt兲␳p
of the fluid, namely, its bulk thermal conductivity, density, heat
where ␾wt is the particle fraction by weight and ␳bf and ␳p are the capacity, and dynamic viscosity. Hence, to compare the theoreti-
bulk density of the base-fluid and the silica particle, respectively. cally estimated performance with the experimental one, we need
The nanoparticles bulk density was assumed to be identical to the to determine the thermophysical properties of the silica suspen-
bulk density of thermal silica at 20° C and was taken to be sions.
2.2 g / cm3. The thermal conductivity of the fluids was measured with a
We performed DLS measurements in order to evaluate the dis- transient hot wire apparatus with an estimated error of less than
persion state of the fluids. The measurements were conducted with ⫾2.5%. A detailed description of the apparatus and the demon-
an ALV-5000 spectrophotometer equipped with an argon-ion laser stration of its accuracy can be found in Ref. 关6兴. Figure 3 shows
共Coherent Innova 308, ␭ = 632.8 nm兲, a digital autocorrelator the relative thermal conductivity of the suspensions as a function
共ALV兲, and a variable angle detection system. Measurements were of particle volume fraction at 20° C and 40° C. The relative ther-
made at a fixed scattering angle of 90 deg and a temperature of mal conductivity showed a linear dependence on the particle load-
25.0⫾ 0.1° C. The individual correlation functions of the intensity ing and is not influenced by temperature. We compared the ex-
data were analyzed using a second-order cumulate fit. Figure 2共a兲 perimentally measured thermal conductivity enhancement with
shows the measured intensity correlation function for the silica the expected enhancement from Maxwell’s effective medium
suspensions at different concentrations as a function of the decay theory for perfectly dispersed spherical particles given by
time. For short decay times, the correlation is high, as Brownian
motion is too slow to change the particle constellation and there- kf 3共kp/kbf − 1兲␾vol
with the signal intensity due to a change of interparticle interfer- =1+ 共9兲
kbf 共kp/kbf + 2兲 − 共kp/kbf − 1兲␾vol
ence. As the decay time becomes longer, the distance between
particles changes and therewith the signal intensity. Consequently, As seen in Fig. 3, there is no significant deviation between the
at a characteristic decay time, the correlation exponentially decays experiment and the predicted value. The experimentally measured
to zero. We observed almost the same behavior of the samples at conductivity is actually slightly smaller than the expected theoret-
5 vol % and 16 vol %, where we see correlation functions with ical value, which could be attributed to an additional interfacial
relatively monodisperse monomodal decay. This indicates classi- thermal resistance, not considered in the theory.
cal Brownian motion of the particles and has only traces of large Specific heat capacity measurements of the silica nanofluids
aggregates. The most concentrated suspension exhibits two decays were carried out using a heat flux type differential scanning calo-
due to the presence of aggregates at this particle concentration. In rimeter 共DSC兲 共DSC Phoenix 204, Netzsch兲. For specific heat
addition, we analyzed the DLS data with the CONTIN method measurements, we conducted three different runs. First, two
关28兴 to obtain the inverse Laplace transformation of the autocor- empty aluminum crucibles with covering lid were placed in the
relation function and the distribution of decay times. The CON- sample, and the reference cells and the heat flow versus tempera-

051401-4 / Vol. 133, MAY 2011 Transactions of the ASME

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 08/18/2013 Terms of Use: http://asme.org/terms


1.5 10
6

k, T = 20 °C 0.05 φ = 5%, T = 20 °C
vol
k, T = 40 °C 4 φvol = 5%, T = 40 °C
10 0.04
cp

µ / [Pas]
1.3 φvol = 16%, T = 20 °C
ρ 0.03 φvol = 16%, T = 40 °C
2
10 φ = 31%, T = 20 °C
0.02 vol

µ / [Pas]
0 3000 6000 φ = 31%, T = 40 °C
1.1 shear rate / [s −1 ]
vol
x /x / [−] 10
0
bf

−2
f

0.9 10

−4
10
−3 −2 −1 0 1 2 3 4
0.7 10 10 10 10 10 10 10 10
shear rate / [s −1 ]

0.5 Fig. 4 Dynamic viscosity as a function of shear rate for differ-


0 10 20 30 40 ent particle concentrations at 20° C and 40° C
φ / [%]
vol

Fig. 3 Relative change of thermal properties, namely, thermal a function of the shear rate. The samples with 5 vol % and
conductivity, density, and specific heat as a function of particle 16 vol % exhibit the typical behavior of a Newtonian fluid with a
concentration; dots indicate experimental data, and lines indi- shear rate independent viscosity. Due to the low viscosity of these
cate theoretical predictions two samples, we were unable to detect significant shear stress at
very low shear rates as seen by the wide scattered data at low
shear rates. The concentrated suspension with 31 vol % of silica
ture curve was measured in a range from 10° C to 80° C with a reveals a shear thinning effect at shear rate below 0.01 s−1. The
constant heating rate of 10 K/min. In a second run, the heat flow effect becomes much more prominent at 40° C. At shear rates
versus temperature dependency of a standard reference material below 100 s−1, the viscosity increases by more than 7 orders of
was measured, in this case a 75 mg sapphire sample, with a magnitude. The viscosity at the high shear rate limit of 5740 s−1
known heat capacity in the temperature range of interest. The third was determined and summarized in Table 2.
run was conducted with the actual sample. During all three runs,
the reference cell was kept empty and identical conditions were 5 Experimental Results and their Discussion
maintained. The specific heat of the sample can be determined by We fabricated test-vehicles of three different designs with vary-
a comparison of the heat flow rates into the sample and the refer- ing channel width of 50 ␮m, 100 ␮m, and 200 ␮m. The exact
ence material and is given as dimensions of the test-vehicles are specified in Table 3. We deter-
mined the pressure loss across each test-vehicle and its total ther-
mref q̇sam共T兲 − q̇bl共T兲
cp,sam共T兲 = cp,ref共T兲 共10兲 mal resistance as a function of particle loading and volumetric
msam q̇ref共T兲 − q̇bl共T兲 flow rate. The measurements were conducted at a constant tem-
Figure 3 shows the experimentally determined relative specific perature difference between maximum heater temperature and
heat as a function of the silica nanoparticle concentration at 20° C. fluid inlet temperature of 20 K controlled by the heater input
The specific heat of the fluid drops with increasing particle load- power.
ing. We compared the data with an analytical formulation for the 5.1 Hydrodynamic Characteristics. Figure 5共a兲 depicts the
effective specific heat, assuming thermal equilibrium between the pressure drop across the three test-vehicles operated by water as a
particles and the surrounding fluid function of the volumetric flow rate. If we apply a certain power
共␳cp兲p␾vol + 共1 − ␾vol兲共␳cp兲bf density, the fluid temperature increases along the channel and
cp,f = = ␾wtcp,p + 共1 − ␾wt兲cp,bf therewith the dynamic viscosity of the fluid changes. The maxi-
␳f mum temperature difference between fluid inlet and fluid outlet
共11兲 was 15 K at low flow rates for the test-vehicle, providing the
As seen in Fig. 3, the analytical formulation is in close agreement
with the experimental results for the lower concentrations. For a Table 2 Dynamic viscosities at a high shear rate limit
particle concentration of 31 vol %, the deviation increases. How- „5750 s−1… as a function of particle concentration and
ever, the deviation is still smaller than 10%. temperature
During the convective heat transfer measurements, we obtained
the fluid density in situ by means of a coriolis flow meter 共Opti- ␮ ␮
mas 3000, Krohne兲 with an error of less than ⫾2 kg/ m3. Figure 3 共mPa s兲 共mPa s兲
shows the fluid density as a function of the particle volume frac- ␾vol
共%兲 20° C 40° C
tion. In addition, we plot the definition of the expected density of
the nanofluids defined as 5 1.62 1.23
␳f = ␳p␾vol + 共1 − ␾vol兲␳bf 共12兲 16 3 2.19
31 26.7 20.09
As expected from Eq. 共12兲, we observed a linear relationship be-
tween the nanofluids density and the particle volume fraction.
The rheological behavior of the dispersions was studied with a Table 3 Dimensions of test-vehicles
MCR 300 rheometer 共Paar Physica兲 in a temperature-controlled
narrow-gap Mooney–Ewart geometry. The Mooney–Ewart geom- wch ww hch hb
共␮m兲 共␮m兲 共␮m兲 共␮m兲
etry consists of a concentric cylinder system with the inner cylin-
der having a cone end. The cone end is used to account for end 204 196 185 340
effects and is designed to ensure that the shear rate is homogenous 103 97 187 338
throughout the geometry. All tests were conducted at 20° C and 53 50 193 330
40° C. Figure 4 shows the dynamic viscosity of the nanofluids as

Journal of Heat Transfer MAY 2011, Vol. 133 / 051401-5

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 08/18/2013 Terms of Use: http://asme.org/terms


a 2 b 30
µ from fit
eff
1.8 wch = 50 µm
wch = 100 µm µ from rheology, T = 20 °C
25
1.6 µ from rheology, T = 40 °C

1.4
20
1.2

∆p / [bar]

µ / [mPas]
1 15
0.8
10
0.6 wch = 200 µm
0.4
5
0.2

0 0
0 0.1 0.2 0.3 0.4 0.5 0 10 20 30 40
volumetric flow rate / [l/min] φ / [%]
vol

Fig. 5 „a… Experimentally determined pressure drop „marker… across the three different test-vehicles
operated with water. Lines indicate the theoretical expected pressure drop „cf. Eq. „1……, „b… compari-
son of average effective viscosity, fitted to the pressure drop volumetric flow rate data, to measured
viscosity by rheology at a shear rate of 5740 s−1 and T = 20° C and T = 40° C as a function of particle
fraction.

highest convective heat transfer coefficient 共wch = 50 ␮m兲. As bution obtained by the infrared camera and the actual heat flux
demonstrated, the viscosity changes with fluid temperature 共cf. density from the enthalpy increase of the coolant, which was in
Fig. 4兲. To consider this effect, we introduced an effective con- good agreement with the dissipated electrical power, and is de-
stant viscosity fitted to the experimentally determined pressure fined as
drop volumetric flow rate data according to Eq. 共1兲. The results for
each fluid and geometry are summarized in Table 4. The deviation
V̇cp,f␳f共Tf,out − Tf,in兲
of the effective viscosity and the literature value for water at 20° C ⬙ =
qheater 共13兲
is less than 6%. The effective viscosity data of the nanofluids and L2ch
the results of rheology measurements, summarized in Table 2, are
in close agreement for the two lower concentrated fluids. The Figure 6共a兲 shows the total thermal resistance of the three test-
effective viscosity of the high concentrated fluid is by a factor of vehicles operated with silica nanofluids at different particle con-
2–2.6 共depending on the effective mean temperature of the fluid兲 centrations as a function of volumetric flow rate. The markers
lower than expected from rheology 共cf. Fig. 5共b兲兲. As seen from indicate the experimental results, and the lines indicate the theo-
Fig. 4, the fluid with a particle loading of 31 vol % shows a retical predictions considering the experimentally determined fluid
strong shear thinning behavior at lower shear rates. At higher bulk properties. The test-vehicle with a channel width of wch
shear rates, the shear thinning behavior is less pronounced but is = 200 ␮m was operated in combination with all three nanofluids,
still present. Unfortunately, due to apparatus limitations, we were whereas the test-vehicles with narrower channels could only be
only able to conduct rheology measurements at shear rates up to operated with the more diluted ones, as the higher concentrated
5740 s−1. However, in the experiments, we have very high shear fluids resulted in high pressure drops, which were not practical for
rates 共higher than 1 ⫻ 104 s−1兲 at the channel wall, decreasing the present test setup. In all experiments, the total thermal resis-
toward the channel core and approaching zero at the centerline. tance shows a power law decay as a function of volumetric flow
Hence, if shear thinning effects are present, viscosity changes rate. This decay is on one hand caused by a reduction of the bulk
along the channel cross section rendering the difficulty to theo- thermal resistance. On the other hand, the convective thermal re-
retically determine an effective viscosity from rheology data. Con- sistance is reduced as well. In the present experiments, the con-
sequently, we took the effectively measured viscosity, summarized vective heat transfer coefficient is influenced by entrance effects.
in Table 4, as an input parameter for the following analysis 共this Especially for the test-vehicles with wch = 200 ␮m and wch
significantly affects the results for the 31 vol % suspension due to = 100 ␮m, it is a function of the Reynolds number, thus of the
shear thinning effect, cf. Fig. 5共b兲兲. volumetric flow rate. In addition, we can observe a decrease of the
total thermal resistance with decreasing channel width due to a
5.2 Convective Heat Transfer. The total thermal resistance reduction of the convective thermal resistance, as expected from
of a test-vehicle was defined by Eq. 共2兲, where the maximum Eq. 共5兲.
heater temperature was taken from the spatial temperature distri- The particle concentration, on the other hand, has no significant
impact on the total thermal resistance of the test-vehicles. We
would expect from Eq. 共5兲 an increased convective heat transfer
Table 4 Effective dynamic viscosities, fitted from pressure for higher concentrated nanofluids due to an increased thermal
drop volumetric flow rate data, for the three test-vehicles and conductivity. This statement only holds for constant 共fully devel-
different nanoparticle concentrations oped兲 Nusselt numbers. However, as mentioned before, the con-
vective heat transfer coefficient is influenced by entrance effects
wch ␾vol = 0% ␾vol = 5% ␾vol = 16% ␾vol = 31% and is a function of the Reynolds number. Since for a constant
共␮m兲 ␮/共mPa s兲 ␮/共mPa s兲 ␮/共mPa s兲 ␮/共mPa s兲
volumetric flow rate the Reynolds number changes with particle
50 0.95 1.27 2.2 – loading, owing to the increase of the nanofluid viscosity, the Nus-
100 1.06 1.21 2.23 9.9 selt number varies with particle concentration. To experimentally
200 1.03 1.16 2.5 10.6 determine the Nusselt number of the system, we computed an
effective convective resistance, defined as the total thermal resis-

051401-6 / Vol. 133, MAY 2011 Transactions of the ASME

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 08/18/2013 Terms of Use: http://asme.org/terms


a 1 b 10 φ vol = 0%
φ vol = 0% wch = 200 µm
wch = 200 µm φ vol = 5%
0.8 φ vol = 5%

R’’ / [cm 2 K/W]


8 φ vol = 16%
φ vol = 16%
0.6

Nu / [−]
φ vol = 31%
φ vol = 31%
0.4 6

tot
0.2
4
R’’bulk,H 2O
0
1 10
φ vol = 0%
wch = 100 µm φ vol = 0% wch = 100 µm
0.8 φ vol = 5%
φ vol = 5%
R’’tot / [cm K/W]

8 φ vol = 16%
0.6 φ vol = 16%

Nu / [−]
2

0.4 6

0.2 4
R’’bulk,H 2O
0
1 10
φ vol = 0% wch = 50 µm
wch = 50 µm φ vol = 0%
0.8 φ vol = 5%
φ vol = 5%
R’’tot / [cm K/W]

8
0.6

Nu / [−]
2

6
0.4

0.2 4
R’’bulk,H 2O
0
0 0.1 0.2 0.3 0.4 0.5 0 500 1000 1500 2000
volumetric flow rate / [l/min] Re / [−]

Fig. 6 „a… Total thermal resistance and „b… Nusselt number as a function of volumetric flow rate for
the three different test-vehicles and different particle concentrations „markers indicate experimental
results and lines the theoretical predictions…

tance minus the theoretical conduction resistance of the silicon number is only a function of the geometry and not of the fluid
base 共cf. Eq. 共3兲兲 and the bulk resistance of the fluid 共cf. Eq. 共6兲兲 properties and the flow characteristics.
In engineering applications such as cooling of electronics, we
R⬙cnv = Rtot
⬙ − R⬙cnd − Rbulk
⬙ 共14兲 are particularly interested in the total thermal resistance as a func-
where we used the experimentally measured bulk properties of the tion of the volumetric flow rate of the coolant or the invested
nanofluids as input parameters. The effective convective resis- pumping power. For a constant volumetric flow rate, the Reynolds
tance is theoretically described by Eq. 共4兲. Hence, we can itera- number decreases with particle loading due to an increased vis-
tively determine the convective heat transfer coefficient of the cosity, whereas the Prandtl number increases. The dimensionless
problem and use it as an input parameter in Eq. 共5兲. In combina- thermal entrance length is a function of those two dimensionless
tion with the dimensions of the test-vehicles and the experimen- numbers, namely, the Reynolds and Prandtl numbers, and is given
tally determined thermal conductivity of the fluid, we can deter- by
mine experimentally the Nusselt number.
Figure 6共b兲 shows the experimental Nusselt number as a func- x+
z+ = ⬇ 0.05 共16兲
tion of the Reynolds number for the three test-vehicles operated at RePrdh
different particle concentrations. For the device with a channel
width of wch = 200 ␮m, we observed for all four fluids an increase where x+ is the thermal entrance length in meters. Commonly, the
in Nusselt number with Reynolds number. At higher Reynolds entrance length is associated with the distance at which the devia-
numbers, the channel entrance length is increased and therewith tion between the local heat transfer coefficient and the asymptotic
the development of the hydrodynamic and thermal boundary layer value, the heat transfers coefficient for fully developed flow is less
is suppressed, leading to an enhancement of convective heat trans- than 5%. In case of simultaneous developing flow with a constant
fer. For a constant Reynolds number, we recorded an increase in heat flux boundary condition, the dimensional entrance length is
Nusselt number with increased particle concentration. We know equal to 0.05. For example, the entrance lengths in the present
that for a constant geometry, the entrance length is not only a experiments with water vary from the lowest Reynolds numbers
function of the Reynolds number but also of the Prandtl number, to the high once among 2.8–12 cm, 1.5–5.44 cm, and 0.7–1.4 cm
defined as for the three test-vehicles with wch = 200 ␮m, wch = 100 ␮m, and
wch = 50 ␮m, respectively. The ratio of the thermal entrance
␮fcp,f lengths for a nanofluid to its base-fluid in case of a constant volu-
Pr = 共15兲 metric flow rate can be defined as
kf
Using the experimentally determined fluid properties, we can xf+ ␳ f cp,f kbf
show that the Prandtl number increases with increasing particle = 共17兲
+
xbf ␳bfcp,bf kf
concentration. From theory, we would expect elongated entrance
lengths and thus increased Nusselt numbers at higher Prandtl Since both terms, namely, the ratio of the volumetric heat capacity
numbers, this being consistent with our experimental results. We and the thermal conductivity of the fluids, are smaller than unity
observed a similar behavior for the device with a channel width of and decrease with increased particle concentration, the entrance
wch = 100 ␮m. If the channel width is further reduced to wch length is reduced with the particle loading causing a reduction of
= 50 ␮m, entrance effects become negligible. Thus, the Nusselt the Nusselt number. Thus, to get back to Fig. 6共a兲, we do not

Journal of Heat Transfer MAY 2011, Vol. 133 / 051401-7

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 08/18/2013 Terms of Use: http://asme.org/terms


10 exercise 关16兴 demonstrated that the thermal conductivity enhance-
wch = 200 µm +10% ment increases with increasing particle aspect ratio and thermal
9 wch = 100 µm conductivity ratio of the particle and the base-fluid. Hence, there
wch = 50 µm might be better performing nanofluids than the present silica
8 nanofluids. To study the general potential of nanofluids to enhance

measured Nu / [−]
-10% the performance of microchannel heat sinks, we applied the ex-
7 perimentally validated one-dimensional model presented in Sec. 3
to determine the heat sink performance as a function of the fluid
6 thermophysical properties.
We demonstrated in a previous study that there exists an opti-
5 mum channel width for a constant filling factor of 0.5 and con-
stant channel height maximizing the COP of a microchannel heat
4 sink 关1兴. For a constant pumping power, the optimum channel
width, wch,opt, is defined by
3
3 4 5 6 7 8 9 10 ⬙
⳵ Rbulk ⳵ R⬙cnv
calculated Nu / [−] =− 共20兲
⳵ wch ⳵ wch
Fig. 7 Comparison of measured and predicted Nusselt num- This means that if the channel width is reduced further than the
ber of all experimentally obtained data points „also shown in optimum width, the increase of bulk thermal resistance due to an
Fig. 6… increased hydrodynamic resistance 共reduction of the volumetric
flow rate兲 cannot be compensated by a reduction of convective
thermal resistance. Since the bulk and the convective thermal re-
observed a decrease of total thermal resistance for increased par- sistance are a function of the thermophysical properties of the
ticle concentrations, as the rise of thermal conductivity with in- coolant, the optimum channel width changes with varying fluid
creasing particle concentration is partly compensated by a reduc- properties. Thus, for a meaningful evaluation of the impact of the
tion of entrance effects due to an increased fluid viscosity. In individual fluid properties, the heat sink performance has to be
addition, the volumetric heat capacity of the coolant is lowered compared at the respective design optimum.
with increased particle loading enlarging the bulk thermal resis- We studied the impact of a relative change of the thermophysi-
tance of the fluid. cal fluid properties on the heat sink performance for a constant
In Figs. 6共a兲 and 6共b兲, we present the theoretically expected pumping power of 1 W corresponding to the higher pumping
data and the experimental result. For the sake of clarity, we com- power range of the experiments. If water is used as the coolant,
pare in Fig. 7 the calculated to the experimentally determined the optimum channel width is estimated to be wch,opt = 35.7 ␮m,
Nusselt number for all experimental data points. All measured resulting in a total thermal resistance of 0.121 cm2 K / W for a
Nusselt numbers are within ⫾10% of predictions. Consequently, volumetric flow rate of 0.215 l/min and a pressure drop of 2.78
we could demonstrate that there is no additional impact of nano- bars. Figure 8 shows the relative change of the heat sink total
particles but the change of thermophysical properties on the con- thermal resistance, the optimum channel width, and the volumet-
vective heat transfer. Thus, standard correlations are adequate to ric flow rate for a relative change of the thermal conductivity,
predict the laminar convective heat transfer of suspensions with dynamic viscosity, specific heat, and density compared with water.
spherical nanoparticles, even up to high particle concentrations of If we consider a heat sink optimized for water and assume that the
31 vol %. thermal conductivity of the coolant is increased by 50% for ex-
ample, the convective thermal resistance will be reduced 共cf. Eqs.
6 Theoretical Evaluation of the Nanofluid Potential 共4兲 and 共5兲兲. Thus, the bulk thermal resistance becomes dominant.
Effectiveness for Electronics Cooling It would be preferential to increase the channel width in order to
In Sec. 5.2, we concentrated on the convective heat transfer of reduce the hydrodynamic resistance and thus increase the volu-
nanofluids and neglected to discuss the increased viscous pressure metric flow rate 共cf. Fig. 8共a兲兲. Consequently, the bulk thermal
drop for nanofluids and its impact on the performance of micro- resistance would be reduced at the expense of an increased con-
channel heat sinks. The characteristic number for a heat sink is its vective resistance. At the new optimum channel width, the total
coefficient of performance, COP, defined as the ratio of the dissi- thermal resistance would be reduced by about 10%.
pated heat to the invested pumping power A 50% increase in dynamic viscosity entails a shift of the op-
timum channel width to wider channels to reduce the hydrody-
q̇ namic resistance but, on the other hand, the convective thermal
COP = 共18兲 resistance is increased. The increased hydrodynamic resistance
P
due to higher viscosities is not fully compensated by the enlarge-
where P is the idealized pumping power ment of the channel width. Thus, the bulk thermal resistance is
increased as well, resulting in a 12% larger total thermal resis-
P = ⌬pV̇ 共19兲 tance 共cf. Fig. 8共b兲兲.
The dissipated heat scales with the total thermal resistance of the Increasing the specific heat capacity of the fluid or the fluid
system, whereas for a constant volumetric flow rate and geometry, density has a similar effect on the heat sinks performance. If one
the pressure drop across the heat sink; thus, the invested pumping of these properties is increased by 50%, the effective thermal mass
power are proportional to the dynamic viscosity of the coolant. As of the coolant is enlarged, causing a reduction of the bulk thermal
depicted in Fig. 6共a兲, none of the nanofluids provided a region resistance. Hence, the optimum is shifted to narrower channels
with a lower thermal resistance than water, irrespective of the and the convective thermal resistance is reduced, leading to a 20%
channel width of the heat sink. Since all silica nanofluids showed reduction of the heat sink total thermal resistance.
an increased viscosity compared with water, there would be no To evaluate the potential of nanofluids to increase the perfor-
benefit of using one of these nanofluids in combination with the mance of microchannel heat sinks, we considered a best case sce-
present heat sink designs. nario. Therefore, we neglected the reduction of volumetric heat
There are several investigations reporting higher thermal con- capacity of the fluid. We determined the relative COP of the heat
ductivity enhancements at lower particle concentrations than for sink operated by a nanofluid with a random combination of rela-
the present silica nanofluids. Also, the international benchmark tive thermal conductivities and viscosities compared with water

051401-8 / Vol. 133, MAY 2011 Transactions of the ASME

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 08/18/2013 Terms of Use: http://asme.org/terms


a 1.6 b 1.6
R’’tot R’’
tot
1.4 wopt 1.4 wopt
V V

x i/xw / [−]

x i/xw / [−]
1.2 1.2

1 1

0.8 0.8

0.6 0.6
0.5 0.75 1 1.25 1.5 0.5 0.75 1 1.25 1.5
k nf/kw [−] µnf/µw [−]
c 1.6 d 1.6
R’’tot R’’
tot
1.4 wopt 1.4 w
opt
V V

x i/xw / [−]
x /x / [−]

1.2 1.2
w

1 1
i

0.8 0.8

0.6 0.6
0.5 0.75 1 1.25 1.5 0.5 0.75 1 1.25 1.5
c p,nf/cp,w [−] ρnf/ρw [−]

Fig. 8 Impact of relative change of „a… thermal conductivity, „b… dynamic viscosity, „c… specific heat,
and „d… density on the total thermal resistance of the heat sink, the optimum channel width and the
volumetric flow rate for a constant pumping power of 1 W

共cf. Fig. 9兲. A constant pumping power of 1 W was applied, and TEM imaging, and the dispersion status of the suspensions was
the COP at the respective design optimum of the heat sink was studied by DLS measurements. We experimentally determined the
determined and normalized by the COP for water. Regions with thermophysical properties of these nanofluids and demonstrated
relative COPs larger than unity indicate combinations of relative that the density and specific heat capacity of the tested nanofluids
thermal conductivities and dynamic viscosities providing an en- can be well predicted by analytical correlations. We could not
hanced performance of the heat sink. As seen from Fig. 9, a ratio observe any abnormal thermal conductivity enhancement, which
between the thermal conductivity enhancement and the viscosity significantly deviates from the effective medium theory.
increase larger than unity is required to gain a significant benefit
To investigate the convective heat transfer potential of these
from using a nanofluid. However, the average value for the ratio
of the relative viscosity to the relative thermal conductivity as a nanofluids, we fabricated test-vehicles of microchannel heat sinks
function of particle concentration is about two 关18兴. To our knowl- with three different channel widths. We measured the total thermal
edge, there are no investigations reporting values of this ratio resistance as a function of volumetric flow rate and extracted the
lower than unity. Nusselt number for the individual parameters. The experimentally
determined Nusselt number was compared with the analytical pre-
7 Conclusions dictions, considering the change of fluid bulk properties. We could
demonstrate a deviation between the experiment and the predic-
In this paper, we evaluated the potential of nanofluids for cool-
tions of less than ⫾10%, indicating that there is no significant
ing of electronics by means of a characterization of aqueous sus-
pensions of silica nanoparticles at concentrations by volume of additional effect from the particles themselves on the convective
5%, 16%, and 31%. The particle morphology was investigated by heat transfer. Hence, standard correlations can be applied to ad-
equately determine the convective heat transfer of nanofluids if
the thermophysical properties of the fluid are known. Further-
more, none of the tested fluids could reduce the total thermal
1.7
resistance for a constant volumetric flow rate. The performance of
2
1.1

1.0

1.6
the heat sink even deteriorated, mainly due to a strong increase in
8
1.0

dynamic viscosity with particle loading.


06

1.5 1 Additionally, we applied a one-dimensional model validated by


1.

=
04

t the experiments to study the potential of nanofluids to enhance the


1.

P op
k f/kbf / [−]

1.4 CO performance of microchannel heat sinks. To this end, we con-


ducted a sensitivity study of the individual thermophysical prop-
1.3 erties of the coolant on the performance of the heat sink at its
respective design optimum. We actually demonstrated that the
1.2
02
1. 8
thermal conductivity has the lowest impact on the performance of
1.1 0.9 0.9
6
4
a microchannel heat sink and just compensates a one-to-one in-
0.9 2
1 0.9 crease of dynamic viscosity. This finding is in contrast to Prasher
1 et al. 关18兴, who suggested that a ratio of relative viscosity to
1 1.1 1.2 1.3 1.4 1.5 1.6 1.7
relative thermal conductivity increase of up to four would be tol-
µ /µ / [−]
f bf erable to still result in gain from nanofluids. However, in Ref.
关18兴, only the convective thermal resistance was taken into ac-
Fig. 9 Contour plot of COP at the respective design optimum
for different combinations of relative changes in thermal con- count, neglecting the fact that the convective resistance is only
ductivity and dynamic viscosity for a constant pumping power one part of the overall chain of affected resistances.
of 1 W An enlargement of the effective thermal mass, which can be

Journal of Heat Transfer MAY 2011, Vol. 133 / 051401-9

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 08/18/2013 Terms of Use: http://asme.org/terms


either achieved by increasing the heat capacity of the fluid or its 关2兴 Tuckerman, D. B., and Pease, R. F. W., 1981, “Implications of High-
Performance Heat Sinking for Electron Devices,” IEEE Trans. Electron De-
density, seems to be more promising to enhance the heat sink
vices, 28共10兲, pp. 1230–1231.
performance. 关3兴 Brunschwiler, T., Rothuizen, H., Fabbri, M., Kloter, U., and Michel, B., 2006,
“Direct Liquid Jet Impingement Cooling With Micronsized Nozzle Array and
Acknowledgment Distributed Return Architecture,” The 10th Intersociety Conference on Ther-
mal and Thermomechanical Phenomena in Electronics Systems, San Diego,
This work was supported by KTI/CTI under Project No. P. CA.
8074.1 NMPP-N. We acknowledge Reto Waelchli, Wulf Glatz, 关4兴 Escher, W., Michel, B., and Poulikakos, D., 2009, “Efficiency of Optimized
Stephan Paredes, Javier Goicochea, Ingmar Meijer, Ute Drechsler, Bifurcating Tree-Like and Parallel Microchannel Networks in the Cooling of
and Martin Witzig for their technical contributions and discus- Electronics,” Int. J. Heat Mass Transfer, 52共5–6兲, pp. 1421–1430.
sions and Walter Riess for the continuous support. 关5兴 Escher, W., Brunschwiler, T., Michel, B., and Poulikakos, D., 2009, “Experi-
mental Investigation of an Ultra-Thin Manifold Micro-Channel Heat Sink for
Liquid-Cooled Chips,” ASME Trans. J. Heat Transfer, 132共8兲, p. 081402.
Nomenclature 关6兴 Shalkevich, N., Escher, W., Buergi, T., Michel, B., Si-Ahmed, L., and Poul-
ikakos, D., 2010, “A Study on the Thermal Conductivity of Gold Nanoparticle
cp ⫽ specific heat capacity 共J kg−1 K−1兲 Colloids,” Langmuir, 26共2兲, pp. 663–670.
COP ⫽ coefficient of performance 关7兴 Putnam, S. A., Cahill, D. G., Braun, P. V., Ge, Z. B., and Shimmin, R. G.,
dh ⫽ characteristic diameter 共m兲 2006, “Thermal Conductivity of Nanoparticle Suspensions,” J. Appl. Phys.,
f ⫽ friction factor 99共8兲, p. 084308.
关8兴 Patel, H. E., Das, S. K., Sundararajan, T., Sreekumaran Nair, A., George, B.,
h ⫽ height 共m兲 and Pradeep, T., 2003, “Thermal Conductivities of Naked and Monolayer Pro-
k ⫽ thermal conductivity 共W m−1 K−1兲 tected Metal Nanoparticle Based Nanofluids: Manifestation of Anomalous En-
L ⫽ length 共m兲 hancement and Chemical Effects,” Appl. Phys. Lett., 83共14兲, pp. 2931–2933.
p ⫽ pressure 共Pa兲 关9兴 Liu, M. S., Lin, M. C. C., Tsai, C. Y., and Wang, C. C., 2006, “Enhancement
of Thermal Conductivity With Cu for Nanofluids Using Chemical Reduction
P ⫽ idealized pumping power 共W兲 Method,” Int. J. Heat Mass Transfer, 49共17–18兲, pp. 3028–3033.
Pr ⫽ Prandtl number 关10兴 Hong, T.-K., Yang, H.-S., and Choi, C. J., 2005, “Study of the Enhanced
q̇ ⫽ heat flux 共W兲 Thermal Conductivity of Fe Nanofluids,” J. Appl. Phys., 97共6兲, p. 064311.
R⬙ ⫽ specific thermal resistance 共cm2 K W−1兲 关11兴 Li, C. H., and Peterson, G. P., 2006, “Experimental Investigation of Tempera-
ture and Volume Fraction Variations on the Effective Thermal Conductivity of
Re ⫽ Reynolds number Nanoparticle Suspensions 共Nanofluids兲,” J. Appl. Phys., 99共8兲, p. 084314.
T ⫽ temperature 共K兲 关12兴 Lee, D., 2007, “Thermophysical Properties of Interfacial Layer in Nanofluids,”
u ⫽ velocity 共m s−1兲 Langmuir, 23共11兲, pp. 6011–6018.
关13兴 Zhu, H. T., Zhang, C. Y., Tang, Y. M., and Wang, J. X., 2007, “Novel Synthe-
V̇ ⫽ volumetric flow rate 共m3 s−1兲 sis and Thermal Conductivity of CuO Nanofluid,” J. Phys. Chem. C, 111共4兲,
w ⫽ width 共m兲 pp. 1646–1650.
x+ ⫽ thermal entrance length 共m兲 关14兴 Choi, S. U. S., Zhang, Z. G., Yu, W., Lockwood, F. E., and Grulke, E. A.,
2001, “Anomalous Thermal Conductivity Enhancement in Nanotube Suspen-
z+ ⫽ dimensionless thermal entrance length
sions,” Appl. Phys. Lett., 79共14兲, pp. 2252–2254.
Greek Letters 关15兴 Xie, H. Q., Lee, H., Youn, W., and Choi, M., 2003, “Nanofluids Containing
Multiwalled Carbon Nanotubes and Their Enhanced Thermal Conductivities,”
␣ ⫽ heat transfer coefficient 共W m−2 K−1兲 J. Appl. Phys., 94共8兲, pp. 4967–4971.
␥ ⫽ fin efficiency 关16兴 Buongiorno, J., Venerus, D. C., Prabhat, N., Mckrell, T., Townsend, J., Chris-
␾ ⫽ particle fraction tianson, R., Tolmachev, Y. V., Keblinski, P., Hu, L., Alvarado, J. L., Bang, I.
␮ ⫽ dynamic viscosity 共kg m−1 s−1兲 C., Bishnoi, S. W., Bonetti, M., Botz, F., Cecere, A., Chang, Y., Chen, G.,
Chen, H., Chung, S. J., Chyu, M. K., Das, S. K., Di Paola, R., Ding, Y.,
␳ ⫽ density 共kg m−3兲 Dubois, F., Dzido, G., Eapen, J., Escher, W., Funfschilling, D., Galand, Q., Ga,
J., Gharagozloo, P. E., Goodson, K. E., Gutierrez, J. G., Hong, H., Horton, M.
Superscripts and Subscripts S. H. K., Iorio, C. S., Pil Jang, S., Jarzebski, A. B., Jiang, Y., Jin, L., Kabelac,
i ⫽ index S., Kamath, A., Kedzierski, M. A., Geok Kieng, L., Kim, C., Kim, J., Kim, S.,
app ⫽ apparent Hyun Lee, S., Choong Leong, K., Manna, I., Michel, B., Ni, R., Patel, H. E.,
avg ⫽ average Philip, J., Poulikakos, D., Reynaud, C., Savino, R., Singh, P. K., Song, P.,
Sundararajan, T., Timofeeva, E., Tritcak, T., Turanov, A. N., Van Vaerenbergh,
base ⫽ substrate between heat transfer channels and
S., Wen, D., Witharana, S., Yang, C., Yeh, W., Zhao, X., and Zhou, S., 2009,
heat source “A Benchmark Study on the Thermal Conductivity of Nanofluids,” J. Appl.
bf ⫽ base-fluid Phys., 106共9兲, p. 094312.
bl ⫽ baseline 关17兴 Zhou, S.-Q., and Ni, R., 2008, “Measurement of the Specific Heat Capacity of
ch ⫽ channel Water-Based Al2O3 Nanofluid,” Appl. Phys. Lett., 92共9兲, p. 093123.
cnd ⫽ conduction 关18兴 Prasher, R., Song, D., Wang, J., and Phelan, P., 2006, “Measurements of Nano-
fluid Viscosity and Its Implications for Thermal Applications,” Appl. Phys.
cnv ⫽ convection Lett., 89共13兲, p. 133108.
f ⫽ fluid 关19兴 Kulkarni, D. P., Das, D. K., and Chukwu, G. A., 2006, “Temperature Depen-
in ⫽ inlet dent Rheological Property of Copper Oxide Nanoparticles Suspension 共Nano-
max ⫽ maximum fluid兲,” J. Nanosci. Nanotechnol., 6共4兲, pp. 1150–1154.
opt ⫽ optimized 关20兴 Wen, D. S., and Ding, Y. L., 2004, “Experimental Investigation Into Convec-
tive Heat Transfer of Nanofluids at the Entrance Region Under Laminar Flow
out ⫽ outlet Conditions,” Int. J. Heat Mass Transfer, 47共24兲, pp. 5181–5188.
p ⫽ particle 关21兴 Zeinali Heris, S., Nasr Esfahany, M., and Etemad, S. Gh., 2007, “Experimental
ref ⫽ reference Investigation of Convective Heat Transfer of Al2O3/Water Nanofluid in Circu-
s ⫽ solid lar Tube,” Int. J. Heat Fluid Flow, 28共2兲, pp. 203–210.
sam ⫽ sample 关22兴 Chen, H. S., Wei, Y., He, Y. R., Ding, W., Zhang, L. L., Tan, C. Q., Lapkin, A.
tot ⫽ total A., and Bavykin, D. V., 2007, “Heat Transfer and Flow Behaviour of Aqueous
Suspensions of Titanate Nanotubes 共Nanofluids兲,” Proceedings of the UK-
vol ⫽ by volume China Particle Technology Forum, Leeds, England, pp. 63–72.
w ⫽ wall 关23兴 Rea, U., Mckrell, T., Hu, L. W., and Buongiorno, J., 2009, “Laminar Convec-
wt ⫽ by weight tive Heat Transfer and Viscous Pressure Loss of Alumina-Water and Zirconia-
Water Nanofluids,” Int. J. Heat Mass Transfer, 52共7–8兲, pp. 2042–2048.
关24兴 Brunschwiler, T., Michel, B., Rothuizen, H., Kloter, U., Wunderle, B., Opper-
References mann, H., and Reichl, H., 2008, “Forced Convective Interlayer Cooling in
关1兴 Escher, W., Michel, B., and Poulikakos, D., 2010, “A Novel High Perfor- Vertically Integrated Packages,” Proceedings of the 11th Intersociety Confer-
mance, Ultra Thin Heat Sink for Electronics,” Int. J. Heat Fluid Flow, 31共4兲, ence on Thermal and Thermomechanical Phenomena in Electronic Systems,
pp. 586–598. Orlando, FL, pp. 1114–1125.

051401-10 / Vol. 133, MAY 2011 Transactions of the ASME

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 08/18/2013 Terms of Use: http://asme.org/terms


关25兴 Muzychka, Y. S., and Yovanovich, M. M., 1998, “Modeling Friction Factors in Heat Transfer in the Combined Entry Region of Non-Circular Ducts,” ASME
Non-Circular Ducts for Developing Laminar Flow,” AIAA Paper No. 98-2492 Trans. J. Heat Transfer, 126共1兲, pp. 54–61.
Albuquerque, NM. 关28兴 Provencher, S. W., 1982, “Contin—A General-Purpose Constrained Regular-
关26兴 Incropera, F. P., 1999, Liquid Cooling of Electronic Devices by Single-Phase ization Program for Inverting Noisy Linear Algebraic and Integral-Equations,”
Convection, Wiley, New York.
Comput. Phys. Commun., 27共3兲, pp. 229–242.
关27兴 Muzychka, Y. S., and Yovanovich, M. M., 2004, “Laminar Forced Convection

Journal of Heat Transfer MAY 2011, Vol. 133 / 051401-11

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 08/18/2013 Terms of Use: http://asme.org/terms

You might also like